diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,117016 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.844406584476857,
+  "global_step": 19500,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.886002886002886e-07,
+      "loss": 7.6125,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.772005772005772e-07,
+      "loss": 8.8271,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.658008658008658e-07,
+      "loss": 10.8678,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1544011544011545e-06,
+      "loss": 10.3347,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4430014430014432e-06,
+      "loss": 10.3507,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7316017316017317e-06,
+      "loss": 9.5616,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0202020202020206e-06,
+      "loss": 10.9455,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.308802308802309e-06,
+      "loss": 10.556,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5974025974025976e-06,
+      "loss": 9.8022,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8860028860028863e-06,
+      "loss": 10.7835,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1746031746031746e-06,
+      "loss": 10.9482,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4632034632034634e-06,
+      "loss": 9.3454,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.751803751803752e-06,
+      "loss": 10.3258,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.040404040404041e-06,
+      "loss": 9.7472,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.329004329004329e-06,
+      "loss": 9.3532,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.617604617604618e-06,
+      "loss": 7.7292,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.906204906204907e-06,
+      "loss": 7.9826,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.194805194805195e-06,
+      "loss": 7.8704,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.4834054834054835e-06,
+      "loss": 8.6594,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.772005772005773e-06,
+      "loss": 9.5604,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.060606060606061e-06,
+      "loss": 9.0706,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.349206349206349e-06,
+      "loss": 8.1273,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.637806637806638e-06,
+      "loss": 7.0958,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.926406926406927e-06,
+      "loss": 8.3516,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.215007215007215e-06,
+      "loss": 8.1129,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.503607503607504e-06,
+      "loss": 6.7493,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.792207792207792e-06,
+      "loss": 7.6819,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.080808080808082e-06,
+      "loss": 7.0955,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.36940836940837e-06,
+      "loss": 7.9264,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.658008658008657e-06,
+      "loss": 6.3341,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.946608946608948e-06,
+      "loss": 7.2887,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.235209235209236e-06,
+      "loss": 6.8606,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.523809523809523e-06,
+      "loss": 5.163,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.812409812409814e-06,
+      "loss": 5.3988,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0101010101010101e-05,
+      "loss": 4.5882,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.038961038961039e-05,
+      "loss": 4.2335,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.067821067821068e-05,
+      "loss": 3.3154,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0966810966810967e-05,
+      "loss": 3.1046,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1255411255411256e-05,
+      "loss": 2.8186,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1544011544011545e-05,
+      "loss": 3.0244,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1832611832611833e-05,
+      "loss": 2.8757,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2121212121212122e-05,
+      "loss": 2.671,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.240981240981241e-05,
+      "loss": 2.651,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2698412698412699e-05,
+      "loss": 2.7196,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2987012987012986e-05,
+      "loss": 2.2879,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3275613275613277e-05,
+      "loss": 2.5259,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3564213564213566e-05,
+      "loss": 2.6204,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3852813852813853e-05,
+      "loss": 2.3755,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4141414141414141e-05,
+      "loss": 1.9201,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.443001443001443e-05,
+      "loss": 2.0139,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.471861471861472e-05,
+      "loss": 1.9907,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5007215007215008e-05,
+      "loss": 1.8586,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5295815295815297e-05,
+      "loss": 1.6858,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5584415584415583e-05,
+      "loss": 1.5651,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5873015873015872e-05,
+      "loss": 1.3707,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6161616161616165e-05,
+      "loss": 1.3935,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.645021645021645e-05,
+      "loss": 1.4545,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.673881673881674e-05,
+      "loss": 1.3245,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.702741702741703e-05,
+      "loss": 1.625,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7316017316017315e-05,
+      "loss": 1.3105,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7604617604617604e-05,
+      "loss": 1.1881,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7893217893217896e-05,
+      "loss": 1.4875,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8181818181818182e-05,
+      "loss": 1.3892,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.847041847041847e-05,
+      "loss": 1.4845,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.875901875901876e-05,
+      "loss": 1.5336,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9047619047619046e-05,
+      "loss": 1.3862,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.933621933621934e-05,
+      "loss": 1.2489,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9624819624819628e-05,
+      "loss": 1.215,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9913419913419914e-05,
+      "loss": 1.363,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0202020202020203e-05,
+      "loss": 1.3769,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0490620490620492e-05,
+      "loss": 1.0701,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.077922077922078e-05,
+      "loss": 1.3447,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.106782106782107e-05,
+      "loss": 1.4481,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.135642135642136e-05,
+      "loss": 1.5347,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1645021645021645e-05,
+      "loss": 1.5027,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1933621933621934e-05,
+      "loss": 1.154,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2222222222222223e-05,
+      "loss": 1.1805,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2510822510822512e-05,
+      "loss": 1.3845,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.27994227994228e-05,
+      "loss": 1.2483,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.308802308802309e-05,
+      "loss": 1.2181,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3376623376623376e-05,
+      "loss": 1.3535,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3665223665223666e-05,
+      "loss": 1.1389,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3953823953823955e-05,
+      "loss": 1.1088,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4242424242424244e-05,
+      "loss": 1.2891,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4531024531024533e-05,
+      "loss": 1.2511,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.481962481962482e-05,
+      "loss": 1.2369,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5108225108225108e-05,
+      "loss": 1.3828,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5396825396825397e-05,
+      "loss": 1.1722,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5685425685425686e-05,
+      "loss": 1.2283,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5974025974025972e-05,
+      "loss": 1.269,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6262626262626268e-05,
+      "loss": 1.3571,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6551226551226554e-05,
+      "loss": 1.3368,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6839826839826843e-05,
+      "loss": 1.1563,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7128427128427132e-05,
+      "loss": 1.4131,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7417027417027418e-05,
+      "loss": 1.1982,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7705627705627707e-05,
+      "loss": 1.0464,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7994227994227996e-05,
+      "loss": 1.3178,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8282828282828282e-05,
+      "loss": 1.2457,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.857142857142857e-05,
+      "loss": 1.3823,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.886002886002886e-05,
+      "loss": 1.2839,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9148629148629146e-05,
+      "loss": 1.3125,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.943722943722944e-05,
+      "loss": 1.3673,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.972582972582973e-05,
+      "loss": 1.3133,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0014430014430017e-05,
+      "loss": 1.345,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0303030303030306e-05,
+      "loss": 1.2872,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0591630591630595e-05,
+      "loss": 1.2523,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0880230880230884e-05,
+      "loss": 1.4092,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1168831168831166e-05,
+      "loss": 1.1815,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1457431457431456e-05,
+      "loss": 1.0842,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1746031746031745e-05,
+      "loss": 1.3207,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2034632034632034e-05,
+      "loss": 1.1827,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.232323232323233e-05,
+      "loss": 1.2096,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.261183261183262e-05,
+      "loss": 1.1902,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.29004329004329e-05,
+      "loss": 1.304,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.318903318903319e-05,
+      "loss": 0.9799,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.347763347763348e-05,
+      "loss": 1.1338,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.376623376623377e-05,
+      "loss": 1.4505,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.405483405483406e-05,
+      "loss": 1.2069,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.434343434343435e-05,
+      "loss": 1.1274,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.463203463203463e-05,
+      "loss": 1.0937,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.492063492063492e-05,
+      "loss": 1.1189,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.520923520923521e-05,
+      "loss": 1.2316,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5497835497835503e-05,
+      "loss": 1.2041,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.578643578643579e-05,
+      "loss": 1.3155,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6075036075036075e-05,
+      "loss": 1.1561,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6363636363636364e-05,
+      "loss": 1.2137,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.665223665223665e-05,
+      "loss": 1.2759,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.694083694083694e-05,
+      "loss": 1.2265,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.722943722943723e-05,
+      "loss": 1.229,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.751803751803752e-05,
+      "loss": 1.1165,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.780663780663781e-05,
+      "loss": 1.4235,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.809523809523809e-05,
+      "loss": 1.3249,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.838383838383838e-05,
+      "loss": 1.1531,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.867243867243868e-05,
+      "loss": 1.2747,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8961038961038966e-05,
+      "loss": 1.1682,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.9249639249639256e-05,
+      "loss": 1.2233,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.953823953823954e-05,
+      "loss": 1.2441,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.982683982683983e-05,
+      "loss": 1.1946,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0115440115440116e-05,
+      "loss": 1.3605,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0404040404040405e-05,
+      "loss": 1.1423,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0692640692640695e-05,
+      "loss": 1.2404,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.0981240981240984e-05,
+      "loss": 1.3583,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.126984126984127e-05,
+      "loss": 1.1826,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.155844155844156e-05,
+      "loss": 1.2441,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.184704184704185e-05,
+      "loss": 1.4142,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.213564213564214e-05,
+      "loss": 1.1478,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.242424242424243e-05,
+      "loss": 1.0778,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.271284271284272e-05,
+      "loss": 1.1882,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3001443001443e-05,
+      "loss": 1.2851,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.329004329004329e-05,
+      "loss": 1.1761,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.357864357864358e-05,
+      "loss": 1.0658,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.386724386724387e-05,
+      "loss": 1.1697,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.415584415584416e-05,
+      "loss": 0.874,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4444444444444447e-05,
+      "loss": 1.2423,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4733044733044736e-05,
+      "loss": 1.1098,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5021645021645025e-05,
+      "loss": 1.2382,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.5310245310245314e-05,
+      "loss": 1.1303,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.55988455988456e-05,
+      "loss": 1.168,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.588744588744589e-05,
+      "loss": 1.0233,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.617604617604618e-05,
+      "loss": 1.1518,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6464646464646464e-05,
+      "loss": 1.0742,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.675324675324675e-05,
+      "loss": 1.19,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.704184704184704e-05,
+      "loss": 1.1227,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.733044733044733e-05,
+      "loss": 1.0193,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.761904761904762e-05,
+      "loss": 1.3974,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.790764790764791e-05,
+      "loss": 0.8994,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.81962481962482e-05,
+      "loss": 1.2004,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.848484848484849e-05,
+      "loss": 1.2959,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.877344877344878e-05,
+      "loss": 1.0457,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9062049062049066e-05,
+      "loss": 1.2127,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9350649350649355e-05,
+      "loss": 1.3286,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.963924963924964e-05,
+      "loss": 1.1525,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.992784992784993e-05,
+      "loss": 1.2169,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0216450216450216e-05,
+      "loss": 1.3666,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.050505050505051e-05,
+      "loss": 1.0681,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0793650793650794e-05,
+      "loss": 1.1922,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.108225108225109e-05,
+      "loss": 1.1456,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.137085137085137e-05,
+      "loss": 1.1649,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.165945165945166e-05,
+      "loss": 1.2907,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1948051948051944e-05,
+      "loss": 1.0958,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.223665223665224e-05,
+      "loss": 1.2116,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2525252525252536e-05,
+      "loss": 1.3743,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.281385281385282e-05,
+      "loss": 1.1749,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.310245310245311e-05,
+      "loss": 1.1179,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.339105339105339e-05,
+      "loss": 1.2104,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.3679653679653686e-05,
+      "loss": 1.0638,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.396825396825397e-05,
+      "loss": 1.0118,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4256854256854264e-05,
+      "loss": 1.319,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4545454545454546e-05,
+      "loss": 1.1117,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.4834054834054835e-05,
+      "loss": 1.2413,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.512265512265512e-05,
+      "loss": 1.1447,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.5411255411255414e-05,
+      "loss": 1.0563,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.569985569985571e-05,
+      "loss": 1.1853,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.598845598845599e-05,
+      "loss": 1.1164,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.627705627705628e-05,
+      "loss": 1.2679,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.6565656565656563e-05,
+      "loss": 1.2541,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.685425685425686e-05,
+      "loss": 1.1147,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.714285714285714e-05,
+      "loss": 1.3745,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.743145743145744e-05,
+      "loss": 1.1545,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.772005772005772e-05,
+      "loss": 1.4559,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.800865800865801e-05,
+      "loss": 1.0335,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.829725829725829e-05,
+      "loss": 1.211,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.858585858585859e-05,
+      "loss": 1.1932,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.887445887445888e-05,
+      "loss": 1.1555,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.9163059163059166e-05,
+      "loss": 1.0703,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.945165945165946e-05,
+      "loss": 1.0208,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.9740259740259744e-05,
+      "loss": 1.1462,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.002886002886003e-05,
+      "loss": 0.9538,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.0317460317460316e-05,
+      "loss": 1.2064,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.060606060606061e-05,
+      "loss": 1.1433,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.0894660894660894e-05,
+      "loss": 1.1876,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.118326118326119e-05,
+      "loss": 1.0998,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.147186147186147e-05,
+      "loss": 1.4294,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.176046176046177e-05,
+      "loss": 1.446,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.204906204906206e-05,
+      "loss": 1.289,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.233766233766233e-05,
+      "loss": 1.1204,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.262626262626264e-05,
+      "loss": 0.9687,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.291486291486291e-05,
+      "loss": 1.0995,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.320346320346321e-05,
+      "loss": 1.2036,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.349206349206349e-05,
+      "loss": 1.3539,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.378066378066379e-05,
+      "loss": 1.2166,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.406926406926407e-05,
+      "loss": 1.0041,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.435786435786436e-05,
+      "loss": 1.1425,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.464646464646466e-05,
+      "loss": 1.2094,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.493506493506494e-05,
+      "loss": 1.1742,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.522366522366524e-05,
+      "loss": 1.0771,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.551226551226551e-05,
+      "loss": 1.3887,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.58008658008658e-05,
+      "loss": 1.2529,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.608946608946609e-05,
+      "loss": 1.0665,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.637806637806638e-05,
+      "loss": 1.0781,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 1.1535,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.695526695526696e-05,
+      "loss": 1.0534,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.724386724386725e-05,
+      "loss": 1.0247,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.753246753246754e-05,
+      "loss": 1.2899,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.782106782106783e-05,
+      "loss": 1.3335,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.810966810966812e-05,
+      "loss": 1.1695,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.83982683982684e-05,
+      "loss": 1.0978,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.86868686868687e-05,
+      "loss": 0.9109,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.897546897546898e-05,
+      "loss": 0.9093,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.926406926406926e-05,
+      "loss": 1.2479,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.955266955266956e-05,
+      "loss": 1.1414,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.984126984126984e-05,
+      "loss": 1.2694,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.012987012987014e-05,
+      "loss": 1.2616,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.041847041847042e-05,
+      "loss": 1.0606,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.07070707070707e-05,
+      "loss": 1.1835,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.099567099567101e-05,
+      "loss": 1.1282,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.128427128427128e-05,
+      "loss": 1.3455,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.157287157287159e-05,
+      "loss": 1.1219,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.186147186147186e-05,
+      "loss": 1.1324,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.215007215007215e-05,
+      "loss": 1.2777,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.243867243867244e-05,
+      "loss": 1.0446,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.272727272727273e-05,
+      "loss": 1.2375,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.301587301587302e-05,
+      "loss": 1.2527,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.33044733044733e-05,
+      "loss": 1.1987,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.35930735930736e-05,
+      "loss": 0.9354,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.388167388167388e-05,
+      "loss": 1.0634,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.417027417027417e-05,
+      "loss": 1.0534,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.445887445887446e-05,
+      "loss": 1.1484,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.474747474747475e-05,
+      "loss": 1.216,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.503607503607504e-05,
+      "loss": 1.0601,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.532467532467533e-05,
+      "loss": 1.1583,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.561327561327562e-05,
+      "loss": 1.3078,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.590187590187591e-05,
+      "loss": 1.1966,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.619047619047618e-05,
+      "loss": 1.2468,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.647907647907649e-05,
+      "loss": 1.1004,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.676767676767676e-05,
+      "loss": 1.2992,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.705627705627707e-05,
+      "loss": 0.885,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.734487734487735e-05,
+      "loss": 1.1407,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.763347763347763e-05,
+      "loss": 1.0941,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.792207792207793e-05,
+      "loss": 1.2176,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.821067821067821e-05,
+      "loss": 1.4615,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.849927849927851e-05,
+      "loss": 1.0972,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.878787878787879e-05,
+      "loss": 1.2857,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.907647907647908e-05,
+      "loss": 1.0825,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.936507936507937e-05,
+      "loss": 1.0555,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.965367965367965e-05,
+      "loss": 1.2898,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.994227994227994e-05,
+      "loss": 1.0303,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.023088023088023e-05,
+      "loss": 0.9915,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.051948051948052e-05,
+      "loss": 1.1086,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.080808080808081e-05,
+      "loss": 0.8798,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.10966810966811e-05,
+      "loss": 1.0243,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.138528138528139e-05,
+      "loss": 1.2426,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.167388167388168e-05,
+      "loss": 1.2824,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.196248196248197e-05,
+      "loss": 1.2335,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.225108225108226e-05,
+      "loss": 1.2083,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.253968253968255e-05,
+      "loss": 1.2004,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.282828282828283e-05,
+      "loss": 1.0998,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.311688311688312e-05,
+      "loss": 1.1425,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.340548340548341e-05,
+      "loss": 1.0367,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.36940836940837e-05,
+      "loss": 0.9224,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.398268398268399e-05,
+      "loss": 0.9904,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.427128427128428e-05,
+      "loss": 0.8774,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.455988455988456e-05,
+      "loss": 1.1096,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.484848484848486e-05,
+      "loss": 1.0099,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.513708513708513e-05,
+      "loss": 1.2278,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.542568542568544e-05,
+      "loss": 1.3363,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.571428571428571e-05,
+      "loss": 1.09,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.6002886002886e-05,
+      "loss": 1.1606,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.62914862914863e-05,
+      "loss": 1.256,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.658008658008658e-05,
+      "loss": 1.011,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.686868686868688e-05,
+      "loss": 1.165,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.715728715728716e-05,
+      "loss": 1.069,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.744588744588745e-05,
+      "loss": 1.108,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.773448773448774e-05,
+      "loss": 0.9554,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.802308802308803e-05,
+      "loss": 1.1837,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.831168831168831e-05,
+      "loss": 1.0544,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.86002886002886e-05,
+      "loss": 1.0408,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.888888888888889e-05,
+      "loss": 1.3589,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.917748917748918e-05,
+      "loss": 1.0303,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.946608946608947e-05,
+      "loss": 1.1877,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.975468975468976e-05,
+      "loss": 1.3239,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.004329004329005e-05,
+      "loss": 1.0641,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.033189033189034e-05,
+      "loss": 1.0169,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.062049062049063e-05,
+      "loss": 1.0796,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.090909090909092e-05,
+      "loss": 1.1633,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.11976911976912e-05,
+      "loss": 1.2328,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.148629148629148e-05,
+      "loss": 1.078,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.177489177489178e-05,
+      "loss": 1.1308,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.206349206349206e-05,
+      "loss": 1.3587,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.235209235209236e-05,
+      "loss": 1.0107,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.264069264069265e-05,
+      "loss": 0.9742,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.292929292929293e-05,
+      "loss": 1.122,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.321789321789323e-05,
+      "loss": 1.0089,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.35064935064935e-05,
+      "loss": 1.1734,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.379509379509381e-05,
+      "loss": 1.0029,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.408369408369408e-05,
+      "loss": 1.0406,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.437229437229437e-05,
+      "loss": 1.2416,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.466089466089466e-05,
+      "loss": 1.0181,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.494949494949495e-05,
+      "loss": 0.9773,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.523809523809524e-05,
+      "loss": 1.0277,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.552669552669553e-05,
+      "loss": 1.0678,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.581529581529582e-05,
+      "loss": 1.0942,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.610389610389611e-05,
+      "loss": 0.9977,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.63924963924964e-05,
+      "loss": 1.0147,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.668109668109669e-05,
+      "loss": 0.9568,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.696969696969698e-05,
+      "loss": 1.1509,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.725829725829726e-05,
+      "loss": 1.1969,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.754689754689755e-05,
+      "loss": 1.2036,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.783549783549783e-05,
+      "loss": 1.2046,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.812409812409813e-05,
+      "loss": 1.1206,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.841269841269841e-05,
+      "loss": 1.0471,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.870129870129871e-05,
+      "loss": 0.848,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.8989898989899e-05,
+      "loss": 1.1224,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.927849927849928e-05,
+      "loss": 1.1023,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.956709956709958e-05,
+      "loss": 0.9624,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.985569985569985e-05,
+      "loss": 0.9095,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010014430014430014,
+      "loss": 1.1904,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010043290043290043,
+      "loss": 0.9744,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010072150072150073,
+      "loss": 1.1864,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010101010101010102,
+      "loss": 1.1066,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001012987012987013,
+      "loss": 1.1247,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010158730158730159,
+      "loss": 1.1995,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010187590187590188,
+      "loss": 1.0529,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010216450216450218,
+      "loss": 1.0739,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010245310245310247,
+      "loss": 1.1949,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010274170274170274,
+      "loss": 1.1121,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010303030303030303,
+      "loss": 1.151,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010331890331890332,
+      "loss": 1.2003,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010360750360750363,
+      "loss": 0.9023,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010389610389610389,
+      "loss": 1.1397,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010418470418470419,
+      "loss": 1.1268,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010447330447330448,
+      "loss": 0.9943,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010476190476190477,
+      "loss": 1.1474,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010505050505050507,
+      "loss": 0.9632,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010533910533910533,
+      "loss": 0.996,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010562770562770564,
+      "loss": 1.1259,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010591630591630593,
+      "loss": 1.0806,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010620490620490621,
+      "loss": 0.8797,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010649350649350649,
+      "loss": 1.2003,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010678210678210678,
+      "loss": 1.1958,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010707070707070708,
+      "loss": 0.9832,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010735930735930737,
+      "loss": 1.2092,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010764790764790765,
+      "loss": 1.1859,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010793650793650794,
+      "loss": 1.1959,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010822510822510823,
+      "loss": 1.0971,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010851370851370853,
+      "loss": 0.9908,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010880230880230882,
+      "loss": 0.9946,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010909090909090909,
+      "loss": 1.1453,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010937950937950938,
+      "loss": 1.0916,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010966810966810967,
+      "loss": 1.0167,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00010995670995670997,
+      "loss": 1.1196,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011024531024531024,
+      "loss": 1.0849,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011053391053391054,
+      "loss": 1.3129,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011082251082251083,
+      "loss": 1.0774,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011111111111111112,
+      "loss": 1.1488,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011139971139971142,
+      "loss": 0.9528,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011168831168831168,
+      "loss": 1.1454,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011197691197691198,
+      "loss": 1.0262,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011226551226551227,
+      "loss": 1.011,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011255411255411256,
+      "loss": 1.1826,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011284271284271284,
+      "loss": 1.1539,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011313131313131313,
+      "loss": 1.0344,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011341991341991343,
+      "loss": 0.9586,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011370851370851372,
+      "loss": 1.1606,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011399711399711401,
+      "loss": 1.1175,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011428571428571428,
+      "loss": 0.9822,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011457431457431457,
+      "loss": 1.2219,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011486291486291488,
+      "loss": 1.1382,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011515151515151516,
+      "loss": 0.9928,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011544011544011544,
+      "loss": 1.1452,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011572871572871573,
+      "loss": 1.1609,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011601731601731602,
+      "loss": 1.1372,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011630591630591632,
+      "loss": 1.1721,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011659451659451658,
+      "loss": 1.1272,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011688311688311689,
+      "loss": 1.1348,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011717171717171717,
+      "loss": 1.2989,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011746031746031746,
+      "loss": 1.072,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011774891774891777,
+      "loss": 0.8568,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011803751803751804,
+      "loss": 1.1281,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011832611832611833,
+      "loss": 0.9275,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011861471861471862,
+      "loss": 1.0966,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011890331890331892,
+      "loss": 0.9316,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011919191919191919,
+      "loss": 1.1065,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011948051948051949,
+      "loss": 1.0671,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00011976911976911978,
+      "loss": 1.0384,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012005772005772007,
+      "loss": 0.9419,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012034632034632037,
+      "loss": 1.0876,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012063492063492063,
+      "loss": 0.9717,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012092352092352093,
+      "loss": 1.0366,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012121212121212122,
+      "loss": 0.9535,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012150072150072151,
+      "loss": 1.0807,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012178932178932179,
+      "loss": 1.1561,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012207792207792208,
+      "loss": 0.9338,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012236652236652238,
+      "loss": 1.2193,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012265512265512266,
+      "loss": 0.9173,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012294372294372293,
+      "loss": 0.9594,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012323232323232323,
+      "loss": 1.0657,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012352092352092354,
+      "loss": 0.9922,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001238095238095238,
+      "loss": 0.871,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012409812409812411,
+      "loss": 1.1817,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001243867243867244,
+      "loss": 0.9772,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012467532467532467,
+      "loss": 1.1538,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012496392496392497,
+      "loss": 0.9961,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012525252525252527,
+      "loss": 1.0304,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012554112554112555,
+      "loss": 1.1533,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012582972582972582,
+      "loss": 1.076,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012611832611832612,
+      "loss": 1.0718,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012640692640692643,
+      "loss": 0.9955,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001266955266955267,
+      "loss": 1.0748,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012698412698412698,
+      "loss": 1.0167,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012727272727272728,
+      "loss": 1.0155,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012756132756132758,
+      "loss": 1.0057,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012784992784992786,
+      "loss": 1.063,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012813852813852814,
+      "loss": 1.0901,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012842712842712844,
+      "loss": 1.282,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001287157287157287,
+      "loss": 1.0557,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012900432900432902,
+      "loss": 1.0898,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012929292929292932,
+      "loss": 1.077,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001295815295815296,
+      "loss": 1.1012,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00012987012987012987,
+      "loss": 1.1654,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013015873015873017,
+      "loss": 1.1662,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013044733044733048,
+      "loss": 1.2119,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013073593073593072,
+      "loss": 1.0823,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013102453102453103,
+      "loss": 1.0259,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013131313131313133,
+      "loss": 1.1746,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001316017316017316,
+      "loss": 1.0331,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013189033189033188,
+      "loss": 1.0464,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013217893217893218,
+      "loss": 0.9779,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013246753246753249,
+      "loss": 0.8909,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013275613275613276,
+      "loss": 1.0354,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013304473304473306,
+      "loss": 1.2457,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013333333333333334,
+      "loss": 1.0983,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013362193362193362,
+      "loss": 1.1041,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013391053391053392,
+      "loss": 0.9978,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013419913419913422,
+      "loss": 1.0723,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001344877344877345,
+      "loss": 1.0536,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013477633477633477,
+      "loss": 1.1692,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013506493506493507,
+      "loss": 1.1561,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013535353535353538,
+      "loss": 1.0377,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013564213564213565,
+      "loss": 1.1059,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013593073593073593,
+      "loss": 0.9347,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013621933621933623,
+      "loss": 1.0678,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001365079365079365,
+      "loss": 1.1543,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001367965367965368,
+      "loss": 1.1314,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013708513708513709,
+      "loss": 1.2496,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001373737373737374,
+      "loss": 0.9759,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013766233766233766,
+      "loss": 1.1213,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013795093795093797,
+      "loss": 1.1166,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013823953823953824,
+      "loss": 0.9664,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013852813852813852,
+      "loss": 0.9856,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013881673881673882,
+      "loss": 1.1434,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013910533910533912,
+      "loss": 1.1296,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001393939393939394,
+      "loss": 1.1432,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013968253968253967,
+      "loss": 1.1637,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00013997113997113998,
+      "loss": 1.1163,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014025974025974028,
+      "loss": 1.0949,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014054834054834055,
+      "loss": 0.9517,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014083694083694083,
+      "loss": 1.1795,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014112554112554113,
+      "loss": 1.0832,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001414141414141414,
+      "loss": 1.4059,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001417027417027417,
+      "loss": 1.0848,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014199134199134201,
+      "loss": 1.2182,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001422799422799423,
+      "loss": 0.8431,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014256854256854257,
+      "loss": 0.9374,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014285714285714287,
+      "loss": 1.1453,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014314574314574317,
+      "loss": 1.0252,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014343434343434342,
+      "loss": 1.1618,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014372294372294372,
+      "loss": 1.0155,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014401154401154402,
+      "loss": 0.965,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001443001443001443,
+      "loss": 1.1413,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014458874458874458,
+      "loss": 1.0931,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014487734487734488,
+      "loss": 1.0887,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014516594516594518,
+      "loss": 0.9833,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014545454545454546,
+      "loss": 1.2462,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014574314574314576,
+      "loss": 1.0699,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014603174603174603,
+      "loss": 1.1002,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014632034632034634,
+      "loss": 1.2184,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001466089466089466,
+      "loss": 1.1785,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014689754689754692,
+      "loss": 1.1364,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001471861471861472,
+      "loss": 1.0169,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014747474747474747,
+      "loss": 0.9605,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014776334776334777,
+      "loss": 1.0669,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014805194805194807,
+      "loss": 1.0517,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014834054834054835,
+      "loss": 0.9407,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014862914862914862,
+      "loss": 0.9913,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014891774891774893,
+      "loss": 1.0185,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014920634920634923,
+      "loss": 1.2083,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001494949494949495,
+      "loss": 1.0745,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00014978354978354978,
+      "loss": 0.9911,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015007215007215008,
+      "loss": 1.0851,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015036075036075036,
+      "loss": 1.0935,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015064935064935066,
+      "loss": 0.8888,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015093795093795096,
+      "loss": 0.9505,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015122655122655124,
+      "loss": 1.1142,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015151515151515152,
+      "loss": 1.0314,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015180375180375182,
+      "loss": 1.0097,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015209235209235212,
+      "loss": 1.1259,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015238095238095237,
+      "loss": 0.9561,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015266955266955267,
+      "loss": 0.9799,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015295815295815297,
+      "loss": 1.0485,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015324675324675325,
+      "loss": 0.9255,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015353535353535353,
+      "loss": 1.0618,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015382395382395383,
+      "loss": 0.9535,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015411255411255413,
+      "loss": 1.3033,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001544011544011544,
+      "loss": 1.1048,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001546897546897547,
+      "loss": 1.2068,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015497835497835498,
+      "loss": 1.0173,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015526695526695526,
+      "loss": 1.0024,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015555555555555556,
+      "loss": 1.2909,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015584415584415587,
+      "loss": 0.8689,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015613275613275614,
+      "loss": 1.2516,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015642135642135642,
+      "loss": 1.0803,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015670995670995672,
+      "loss": 1.212,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015699855699855702,
+      "loss": 0.851,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001572871572871573,
+      "loss": 0.9918,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015757575757575757,
+      "loss": 1.1107,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015786435786435788,
+      "loss": 1.2109,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015815295815295815,
+      "loss": 1.0863,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015844155844155845,
+      "loss": 1.0551,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015873015873015873,
+      "loss": 0.9644,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00015901875901875903,
+      "loss": 1.0375,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001593073593073593,
+      "loss": 1.0146,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001595959595959596,
+      "loss": 1.1276,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001598845598845599,
+      "loss": 1.2574,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016017316017316016,
+      "loss": 1.0466,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016046176046176046,
+      "loss": 1.204,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016075036075036077,
+      "loss": 1.0884,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016103896103896104,
+      "loss": 0.9306,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016132756132756132,
+      "loss": 1.1162,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016161616161616162,
+      "loss": 1.0071,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016190476190476192,
+      "loss": 1.1779,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001621933621933622,
+      "loss": 1.0954,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016248196248196248,
+      "loss": 1.1533,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016277056277056278,
+      "loss": 1.0735,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016305916305916305,
+      "loss": 0.8837,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016334776334776336,
+      "loss": 1.0408,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016363636363636366,
+      "loss": 1.2971,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016392496392496393,
+      "loss": 1.0771,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001642135642135642,
+      "loss": 1.013,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001645021645021645,
+      "loss": 0.9531,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016479076479076482,
+      "loss": 1.1147,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0001650793650793651,
+      "loss": 0.9577,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016536796536796537,
+      "loss": 1.22,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016565656565656567,
+      "loss": 1.0254,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016594516594516597,
+      "loss": 1.012,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016623376623376625,
+      "loss": 1.0438,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.00016652236652236652,
+      "loss": 1.2036,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016681096681096683,
+      "loss": 0.9732,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001670995670995671,
+      "loss": 1.0935,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001673881673881674,
+      "loss": 1.2093,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016767676767676768,
+      "loss": 0.9298,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016796536796536798,
+      "loss": 0.9751,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016825396825396826,
+      "loss": 1.1146,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016854256854256856,
+      "loss": 1.2354,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016883116883116884,
+      "loss": 1.175,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001691197691197691,
+      "loss": 1.0086,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016940836940836941,
+      "loss": 1.0601,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016969696969696972,
+      "loss": 1.0348,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00016998556998557,
+      "loss": 1.026,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017027417027417027,
+      "loss": 0.9565,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017056277056277057,
+      "loss": 0.8378,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017085137085137087,
+      "loss": 0.9822,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017113997113997115,
+      "loss": 0.8078,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017142857142857143,
+      "loss": 0.9858,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017171717171717173,
+      "loss": 1.188,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000172005772005772,
+      "loss": 0.9537,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001722943722943723,
+      "loss": 1.0658,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001725829725829726,
+      "loss": 1.159,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017287157287157288,
+      "loss": 1.0364,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017316017316017316,
+      "loss": 1.0191,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017344877344877346,
+      "loss": 1.1866,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017373737373737377,
+      "loss": 1.0317,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017402597402597401,
+      "loss": 1.0455,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017431457431457432,
+      "loss": 1.061,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017460317460317462,
+      "loss": 1.0293,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001748917748917749,
+      "loss": 0.9567,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017518037518037517,
+      "loss": 0.8392,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017546897546897547,
+      "loss": 1.0264,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017575757575757578,
+      "loss": 1.1184,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017604617604617605,
+      "loss": 0.9428,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017633477633477635,
+      "loss": 1.1975,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017662337662337663,
+      "loss": 0.996,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001769119769119769,
+      "loss": 1.3883,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001772005772005772,
+      "loss": 1.0234,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001774891774891775,
+      "loss": 1.0085,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017777777777777779,
+      "loss": 0.9718,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017806637806637806,
+      "loss": 1.109,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017835497835497836,
+      "loss": 1.0494,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017864357864357867,
+      "loss": 0.9692,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017893217893217894,
+      "loss": 1.1474,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017922077922077922,
+      "loss": 1.0897,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00017950937950937952,
+      "loss": 1.0868,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001797979797979798,
+      "loss": 1.0506,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001800865800865801,
+      "loss": 1.0207,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018037518037518038,
+      "loss": 0.9182,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018066378066378068,
+      "loss": 1.1089,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018095238095238095,
+      "loss": 1.0813,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018124098124098126,
+      "loss": 1.1292,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018152958152958153,
+      "loss": 0.971,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018181818181818183,
+      "loss": 1.1204,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001821067821067821,
+      "loss": 0.9599,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001823953823953824,
+      "loss": 1.083,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018268398268398272,
+      "loss": 1.0845,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018297258297258296,
+      "loss": 1.1854,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018326118326118327,
+      "loss": 1.2026,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018354978354978357,
+      "loss": 1.0325,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018383838383838384,
+      "loss": 1.0602,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018412698412698412,
+      "loss": 1.3358,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018441558441558442,
+      "loss": 0.9248,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018470418470418473,
+      "loss": 1.0665,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000184992784992785,
+      "loss": 1.1199,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001852813852813853,
+      "loss": 0.9906,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018556998556998558,
+      "loss": 0.8505,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018585858585858586,
+      "loss": 0.9731,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018614718614718616,
+      "loss": 0.9267,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018643578643578646,
+      "loss": 1.0345,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018672438672438674,
+      "loss": 1.1232,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000187012987012987,
+      "loss": 0.9959,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018730158730158731,
+      "loss": 1.2434,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018759018759018762,
+      "loss": 1.1891,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001878787878787879,
+      "loss": 0.9876,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018816738816738817,
+      "loss": 0.9558,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018845598845598847,
+      "loss": 1.0588,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018874458874458875,
+      "loss": 1.1437,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018903318903318905,
+      "loss": 1.1017,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018932178932178932,
+      "loss": 0.8702,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00018961038961038963,
+      "loss": 0.9612,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001898989898989899,
+      "loss": 1.15,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001901875901875902,
+      "loss": 1.0477,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019047619047619048,
+      "loss": 1.0949,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019076479076479076,
+      "loss": 0.885,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019105339105339106,
+      "loss": 0.9373,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019134199134199136,
+      "loss": 0.9945,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019163059163059164,
+      "loss": 1.0349,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019191919191919191,
+      "loss": 1.0155,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019220779220779222,
+      "loss": 1.0549,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019249639249639252,
+      "loss": 1.0691,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001927849927849928,
+      "loss": 1.0422,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019307359307359307,
+      "loss": 1.3045,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019336219336219337,
+      "loss": 1.2233,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019365079365079365,
+      "loss": 1.0247,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019393939393939395,
+      "loss": 1.1222,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019422799422799425,
+      "loss": 1.2032,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019451659451659453,
+      "loss": 0.8935,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001948051948051948,
+      "loss": 1.0427,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001950937950937951,
+      "loss": 1.052,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001953823953823954,
+      "loss": 1.0823,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019567099567099566,
+      "loss": 1.0705,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019595959595959596,
+      "loss": 0.9947,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019624819624819626,
+      "loss": 0.9685,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019653679653679654,
+      "loss": 0.9637,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019682539682539682,
+      "loss": 0.9622,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019711399711399712,
+      "loss": 1.1267,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019740259740259742,
+      "loss": 0.9728,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001976911976911977,
+      "loss": 0.8913,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000197979797979798,
+      "loss": 1.0994,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019826839826839827,
+      "loss": 1.286,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019855699855699855,
+      "loss": 1.022,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019884559884559885,
+      "loss": 1.073,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019913419913419916,
+      "loss": 1.0043,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019942279942279943,
+      "loss": 0.9639,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001997113997113997,
+      "loss": 1.0648,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0002,
+      "loss": 1.2054,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999999990165015,
+      "loss": 0.8927,
+      "step": 694
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999999606600596,
+      "loss": 1.1129,
+      "step": 695
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999999911485134,
+      "loss": 1.1867,
+      "step": 696
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.000199999984264024,
+      "loss": 0.8848,
+      "step": 697
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999999754125379,
+      "loss": 1.0921,
+      "step": 698
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999996459405518,
+      "loss": 1.1204,
+      "step": 699
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999995180857615,
+      "loss": 0.9946,
+      "step": 700
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999993705610097,
+      "loss": 1.0706,
+      "step": 701
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999992033663003,
+      "loss": 1.0618,
+      "step": 702
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999999016501636,
+      "loss": 1.0541,
+      "step": 703
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999988099670204,
+      "loss": 1.0323,
+      "step": 704
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999998583762458,
+      "loss": 1.0161,
+      "step": 705
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999983378879527,
+      "loss": 1.0622,
+      "step": 706
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999980723435097,
+      "loss": 0.9829,
+      "step": 707
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999977871291345,
+      "loss": 0.9631,
+      "step": 708
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999974822448318,
+      "loss": 0.9853,
+      "step": 709
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999971576906085,
+      "loss": 1.0013,
+      "step": 710
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999968134664706,
+      "loss": 1.0948,
+      "step": 711
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999964495724247,
+      "loss": 0.9227,
+      "step": 712
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999960660084783,
+      "loss": 1.0126,
+      "step": 713
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999956627746388,
+      "loss": 1.018,
+      "step": 714
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999952398709142,
+      "loss": 1.006,
+      "step": 715
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999947972973128,
+      "loss": 1.0092,
+      "step": 716
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999994335053843,
+      "loss": 0.8092,
+      "step": 717
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999993853140514,
+      "loss": 0.924,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999933515573362,
+      "loss": 1.1524,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999992830304318,
+      "loss": 0.9859,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999922893814705,
+      "loss": 1.2452,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999917287888045,
+      "loss": 1.0532,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999911485263304,
+      "loss": 1.2354,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999905485940602,
+      "loss": 1.017,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999899289920055,
+      "loss": 1.1283,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999989289720178,
+      "loss": 0.9692,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999886307785908,
+      "loss": 1.1646,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999987952167257,
+      "loss": 1.1227,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999872538861894,
+      "loss": 1.0253,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999865359354022,
+      "loss": 1.0175,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999857983149096,
+      "loss": 1.0863,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999850410247255,
+      "loss": 0.9601,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999842640648654,
+      "loss": 0.9725,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999834674353442,
+      "loss": 1.0669,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999826511361778,
+      "loss": 0.9745,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999818151673822,
+      "loss": 1.2486,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999980959528974,
+      "loss": 1.1224,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999800842209696,
+      "loss": 0.807,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999791892433866,
+      "loss": 1.1586,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999782745962422,
+      "loss": 1.1331,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999773402795549,
+      "loss": 1.057,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999763862933427,
+      "loss": 1.1814,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999754126376248,
+      "loss": 0.9859,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999744193124197,
+      "loss": 0.945,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999734063177473,
+      "loss": 1.1019,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999972373653628,
+      "loss": 0.9996,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999713213200812,
+      "loss": 1.2115,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999702493171283,
+      "loss": 1.192,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999691576447898,
+      "loss": 1.0587,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999680463030875,
+      "loss": 0.9847,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999669152920437,
+      "loss": 1.029,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999657646116797,
+      "loss": 0.9361,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999645942620192,
+      "loss": 1.1585,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999634042430843,
+      "loss": 0.9669,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999621945548986,
+      "loss": 0.9502,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999609651974863,
+      "loss": 1.0117,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999597161708712,
+      "loss": 1.068,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999584474750782,
+      "loss": 1.0196,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999571591101318,
+      "loss": 1.0681,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999558510760577,
+      "loss": 0.9002,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999954523372882,
+      "loss": 0.9628,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999531760006296,
+      "loss": 0.9826,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999518089593282,
+      "loss": 1.1348,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999950422249004,
+      "loss": 1.014,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999490158696846,
+      "loss": 1.0407,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999475898213977,
+      "loss": 1.1554,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999461441041713,
+      "loss": 1.1203,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999446787180336,
+      "loss": 1.1979,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999431936630136,
+      "loss": 1.0577,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999416889391404,
+      "loss": 1.1163,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999940164546444,
+      "loss": 1.004,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999938620484954,
+      "loss": 1.1912,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999370567547008,
+      "loss": 0.7786,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999354733557149,
+      "loss": 1.1817,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999338702880284,
+      "loss": 0.9681,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999322475516718,
+      "loss": 1.1833,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999306051466772,
+      "loss": 1.004,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999289430730775,
+      "loss": 0.8978,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999927261330905,
+      "loss": 1.0825,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999255599201926,
+      "loss": 1.3164,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999238388409744,
+      "loss": 1.0792,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999220980932832,
+      "loss": 1.2182,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999203376771542,
+      "loss": 0.8958,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999185575926215,
+      "loss": 0.8956,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999167578397205,
+      "loss": 1.2717,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999149384184863,
+      "loss": 0.9559,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999913099328955,
+      "loss": 1.2047,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999112405711623,
+      "loss": 0.8748,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999909362145145,
+      "loss": 1.1355,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999074640509402,
+      "loss": 1.1017,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999905546288585,
+      "loss": 1.0437,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999036088581173,
+      "loss": 1.042,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019999016517595753,
+      "loss": 1.3012,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999899674992997,
+      "loss": 1.0846,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998976785584217,
+      "loss": 0.8822,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998956624558886,
+      "loss": 1.0716,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998936266854377,
+      "loss": 0.9416,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999891571247108,
+      "loss": 0.9538,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998894961409413,
+      "loss": 1.0619,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998874013669774,
+      "loss": 1.0982,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998852869252578,
+      "loss": 1.1404,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998831528158243,
+      "loss": 1.0204,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998809990387186,
+      "loss": 0.9428,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001999878825593983,
+      "loss": 1.0378,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998766324816607,
+      "loss": 0.9863,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998744197017944,
+      "loss": 0.9295,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998721872544278,
+      "loss": 1.2478,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.00019998699351396048,
+      "loss": 0.8915,
+      "step": 808
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000199986766335737,
+      "loss": 0.9963,
+      "step": 809
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998653719077676,
+      "loss": 0.8287,
+      "step": 810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998630607908426,
+      "loss": 1.0206,
+      "step": 811
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998607300066409,
+      "loss": 1.0253,
+      "step": 812
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998583795552083,
+      "loss": 0.8368,
+      "step": 813
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998560094365906,
+      "loss": 0.9939,
+      "step": 814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999853619650835,
+      "loss": 1.0169,
+      "step": 815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999851210197988,
+      "loss": 1.2033,
+      "step": 816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998487810780975,
+      "loss": 1.1399,
+      "step": 817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998463322912108,
+      "loss": 0.9308,
+      "step": 818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998438638373762,
+      "loss": 1.0449,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998413757166424,
+      "loss": 1.0932,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998388679290583,
+      "loss": 1.2369,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998363404746732,
+      "loss": 1.1766,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998337933535367,
+      "loss": 1.3067,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998312265656992,
+      "loss": 0.9334,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998286401112108,
+      "loss": 0.8693,
+      "step": 825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998260339901226,
+      "loss": 1.0963,
+      "step": 826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999823408202486,
+      "loss": 1.0156,
+      "step": 827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998207627483524,
+      "loss": 0.9356,
+      "step": 828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998180976277735,
+      "loss": 1.1515,
+      "step": 829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998154128408026,
+      "loss": 1.0117,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999812708387492,
+      "loss": 1.2033,
+      "step": 831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998099842678948,
+      "loss": 1.0326,
+      "step": 832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999807240482065,
+      "loss": 1.1457,
+      "step": 833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999804477030056,
+      "loss": 1.0839,
+      "step": 834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019998016939119225,
+      "loss": 1.1693,
+      "step": 835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997988911277194,
+      "loss": 1.0061,
+      "step": 836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997960686775015,
+      "loss": 1.0605,
+      "step": 837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997932265613247,
+      "loss": 1.0886,
+      "step": 838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997903647792442,
+      "loss": 1.1261,
+      "step": 839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999787483331317,
+      "loss": 1.1477,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997845822175995,
+      "loss": 1.1341,
+      "step": 841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999781661438149,
+      "loss": 1.0776,
+      "step": 842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997787209930222,
+      "loss": 1.0535,
+      "step": 843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997757608822777,
+      "loss": 1.1267,
+      "step": 844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997727811059738,
+      "loss": 1.0193,
+      "step": 845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997697816641688,
+      "loss": 0.9869,
+      "step": 846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997667625569214,
+      "loss": 1.1366,
+      "step": 847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997637237842916,
+      "loss": 0.9413,
+      "step": 848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997606653463385,
+      "loss": 0.878,
+      "step": 849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999757587243123,
+      "loss": 1.1038,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999754489474705,
+      "loss": 0.9692,
+      "step": 851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997513720411457,
+      "loss": 1.1077,
+      "step": 852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997482349425066,
+      "loss": 0.9107,
+      "step": 853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997450781788492,
+      "loss": 0.9738,
+      "step": 854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997419017502358,
+      "loss": 1.01,
+      "step": 855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997387056567285,
+      "loss": 1.1947,
+      "step": 856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000199973548989839,
+      "loss": 1.0346,
+      "step": 857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997322544752848,
+      "loss": 1.0991,
+      "step": 858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999728999387475,
+      "loss": 1.1086,
+      "step": 859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997257246350255,
+      "loss": 1.1061,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997224302180006,
+      "loss": 1.25,
+      "step": 861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999719116136465,
+      "loss": 1.0826,
+      "step": 862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997157823904837,
+      "loss": 1.251,
+      "step": 863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999712428980123,
+      "loss": 1.0167,
+      "step": 864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997090559054477,
+      "loss": 1.0471,
+      "step": 865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999705663166525,
+      "loss": 0.9936,
+      "step": 866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019997022507634216,
+      "loss": 1.1372,
+      "step": 867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996988186962041,
+      "loss": 1.0813,
+      "step": 868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999695366964941,
+      "loss": 1.0252,
+      "step": 869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996918955696988,
+      "loss": 1.0516,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999688404510547,
+      "loss": 0.9617,
+      "step": 871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996848937875537,
+      "loss": 1.0082,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996813634007881,
+      "loss": 1.1772,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996778133503194,
+      "loss": 1.2548,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996742436362178,
+      "loss": 0.9254,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996706542585533,
+      "loss": 1.0579,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996670452173966,
+      "loss": 1.1432,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996634165128185,
+      "loss": 1.0323,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996597681448905,
+      "loss": 1.1426,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996561001136843,
+      "loss": 0.9539,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996524124192723,
+      "loss": 1.1456,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999648705061727,
+      "loss": 1.1858,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999644978041121,
+      "loss": 1.1813,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999641231357528,
+      "loss": 0.9676,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999637465011021,
+      "loss": 1.1472,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996336790016752,
+      "loss": 1.0214,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999629873329564,
+      "loss": 1.0384,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999626047994763,
+      "loss": 1.1153,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999622202997347,
+      "loss": 1.0888,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999618338337392,
+      "loss": 1.0467,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996144540149736,
+      "loss": 1.1105,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996105500301684,
+      "loss": 0.9785,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019996066263830531,
+      "loss": 1.3724,
+      "step": 893
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999602683073705,
+      "loss": 0.9793,
+      "step": 894
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995987201022018,
+      "loss": 0.9597,
+      "step": 895
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999594737468621,
+      "loss": 1.0249,
+      "step": 896
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995907351730416,
+      "loss": 1.0647,
+      "step": 897
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995867132155417,
+      "loss": 1.0198,
+      "step": 898
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995826715962008,
+      "loss": 0.9981,
+      "step": 899
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995786103150983,
+      "loss": 0.9785,
+      "step": 900
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995745293723138,
+      "loss": 1.1607,
+      "step": 901
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999570428767928,
+      "loss": 1.1257,
+      "step": 902
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995663085020212,
+      "loss": 1.0129,
+      "step": 903
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995621685746747,
+      "loss": 1.2347,
+      "step": 904
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995580089859701,
+      "loss": 1.2185,
+      "step": 905
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999553829735989,
+      "loss": 0.8736,
+      "step": 906
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995496308248135,
+      "loss": 1.0393,
+      "step": 907
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999545412252526,
+      "loss": 1.1708,
+      "step": 908
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.000199954117401921,
+      "loss": 0.9048,
+      "step": 909
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995369161249484,
+      "loss": 1.0997,
+      "step": 910
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995326385698254,
+      "loss": 1.0387,
+      "step": 911
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999528341353925,
+      "loss": 1.1196,
+      "step": 912
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995240244773314,
+      "loss": 1.0047,
+      "step": 913
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995196879401297,
+      "loss": 1.1511,
+      "step": 914
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995153317424054,
+      "loss": 1.2133,
+      "step": 915
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999510955884244,
+      "loss": 1.033,
+      "step": 916
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995065603657316,
+      "loss": 1.0304,
+      "step": 917
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019995021451869546,
+      "loss": 1.0034,
+      "step": 918
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994977103480002,
+      "loss": 1.0482,
+      "step": 919
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994932558489552,
+      "loss": 1.1258,
+      "step": 920
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994887816899074,
+      "loss": 1.1118,
+      "step": 921
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994842878709446,
+      "loss": 1.1281,
+      "step": 922
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994797743921555,
+      "loss": 1.0264,
+      "step": 923
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994752412536286,
+      "loss": 1.1218,
+      "step": 924
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994706884554533,
+      "loss": 1.1132,
+      "step": 925
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999466115997719,
+      "loss": 0.9288,
+      "step": 926
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999461523880516,
+      "loss": 0.861,
+      "step": 927
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999456912103934,
+      "loss": 0.9791,
+      "step": 928
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994522806680642,
+      "loss": 1.1233,
+      "step": 929
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999447629572998,
+      "loss": 1.0885,
+      "step": 930
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994429588188255,
+      "loss": 1.084,
+      "step": 931
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994382684056401,
+      "loss": 1.0745,
+      "step": 932
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994335583335335,
+      "loss": 1.3472,
+      "step": 933
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994288286025982,
+      "loss": 1.1308,
+      "step": 934
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994240792129274,
+      "loss": 1.1805,
+      "step": 935
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999419310164614,
+      "loss": 1.109,
+      "step": 936
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994145214577528,
+      "loss": 1.029,
+      "step": 937
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994097130924374,
+      "loss": 0.9729,
+      "step": 938
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994048850687627,
+      "loss": 1.1227,
+      "step": 939
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019994000373868226,
+      "loss": 0.9136,
+      "step": 940
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993951700467139,
+      "loss": 1.0139,
+      "step": 941
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993902830485317,
+      "loss": 1.3393,
+      "step": 942
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993853763923723,
+      "loss": 1.121,
+      "step": 943
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993804500783316,
+      "loss": 0.9489,
+      "step": 944
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999375504106507,
+      "loss": 0.8843,
+      "step": 945
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999370538476996,
+      "loss": 1.143,
+      "step": 946
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993655531898958,
+      "loss": 1.0442,
+      "step": 947
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999360548245305,
+      "loss": 1.0906,
+      "step": 948
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993555236433213,
+      "loss": 1.1008,
+      "step": 949
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993504793840443,
+      "loss": 1.1443,
+      "step": 950
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993454154675728,
+      "loss": 0.9903,
+      "step": 951
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993403318940064,
+      "loss": 1.0523,
+      "step": 952
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993352286634453,
+      "loss": 0.9615,
+      "step": 953
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993301057759895,
+      "loss": 1.1601,
+      "step": 954
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993249632317404,
+      "loss": 0.8684,
+      "step": 955
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993198010307987,
+      "loss": 1.068,
+      "step": 956
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019993146191732658,
+      "loss": 0.9099,
+      "step": 957
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999309417659244,
+      "loss": 1.0003,
+      "step": 958
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999304196488835,
+      "loss": 1.1213,
+      "step": 959
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992989556621424,
+      "loss": 0.9753,
+      "step": 960
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992936951792692,
+      "loss": 1.1764,
+      "step": 961
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992884150403178,
+      "loss": 0.8319,
+      "step": 962
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992831152453933,
+      "loss": 1.1431,
+      "step": 963
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999277795794599,
+      "loss": 0.9728,
+      "step": 964
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992724566880404,
+      "loss": 1.0253,
+      "step": 965
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999267097925822,
+      "loss": 1.1388,
+      "step": 966
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999261719508049,
+      "loss": 1.0784,
+      "step": 967
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992563214348276,
+      "loss": 1.173,
+      "step": 968
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992509037062642,
+      "loss": 0.8153,
+      "step": 969
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992454663224644,
+      "loss": 1.084,
+      "step": 970
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992400092835362,
+      "loss": 1.0255,
+      "step": 971
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992345325895865,
+      "loss": 0.9798,
+      "step": 972
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999229036240723,
+      "loss": 1.1103,
+      "step": 973
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992235202370538,
+      "loss": 1.1708,
+      "step": 974
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992179845786876,
+      "loss": 1.1105,
+      "step": 975
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999212429265733,
+      "loss": 1.161,
+      "step": 976
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992068542982997,
+      "loss": 0.9814,
+      "step": 977
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019992012596764967,
+      "loss": 1.3146,
+      "step": 978
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991956454004346,
+      "loss": 1.2225,
+      "step": 979
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991900114702235,
+      "loss": 1.0548,
+      "step": 980
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991843578859747,
+      "loss": 1.0187,
+      "step": 981
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991786846477989,
+      "loss": 1.0351,
+      "step": 982
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999172991755808,
+      "loss": 1.007,
+      "step": 983
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991672792101138,
+      "loss": 0.922,
+      "step": 984
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991615470108285,
+      "loss": 0.9502,
+      "step": 985
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991557951580652,
+      "loss": 1.1561,
+      "step": 986
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999150023651937,
+      "loss": 1.1801,
+      "step": 987
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991442324925572,
+      "loss": 0.9828,
+      "step": 988
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991384216800398,
+      "loss": 1.184,
+      "step": 989
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991325912144993,
+      "loss": 1.1175,
+      "step": 990
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991267410960503,
+      "loss": 0.9098,
+      "step": 991
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991208713248076,
+      "loss": 1.0777,
+      "step": 992
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999114981900887,
+      "loss": 1.1441,
+      "step": 993
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991090728244042,
+      "loss": 1.0169,
+      "step": 994
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019991031440954755,
+      "loss": 1.1868,
+      "step": 995
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990971957142172,
+      "loss": 1.0641,
+      "step": 996
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990912276807466,
+      "loss": 1.1754,
+      "step": 997
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990852399951812,
+      "loss": 0.9983,
+      "step": 998
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990792326576387,
+      "loss": 1.0229,
+      "step": 999
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990732056682372,
+      "loss": 0.9746,
+      "step": 1000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999067159027095,
+      "loss": 1.1109,
+      "step": 1001
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990610927343315,
+      "loss": 0.9648,
+      "step": 1002
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999055006790066,
+      "loss": 1.0449,
+      "step": 1003
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990489011944177,
+      "loss": 1.0548,
+      "step": 1004
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990427759475067,
+      "loss": 1.1725,
+      "step": 1005
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001999036631049454,
+      "loss": 0.817,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990304665003805,
+      "loss": 1.0725,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990242823004074,
+      "loss": 1.3174,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990180784496556,
+      "loss": 1.3331,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990118549482482,
+      "loss": 1.2112,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019990056117963069,
+      "loss": 1.1341,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989993489939547,
+      "loss": 1.0683,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989930665413147,
+      "loss": 1.0515,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989867644385108,
+      "loss": 1.0812,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989804426856665,
+      "loss": 1.0617,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989741012829066,
+      "loss": 1.1014,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989677402303555,
+      "loss": 1.2491,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989613595281384,
+      "loss": 1.0759,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998954959176381,
+      "loss": 1.1116,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998948539175209,
+      "loss": 1.0801,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998942099524749,
+      "loss": 1.0043,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998935640225127,
+      "loss": 1.0889,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989291612764704,
+      "loss": 0.9795,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989226626789067,
+      "loss": 1.0002,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989161444325642,
+      "loss": 0.9953,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989096065375702,
+      "loss": 1.4071,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019989030489940538,
+      "loss": 0.8976,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998896471802144,
+      "loss": 0.9479,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019988898749619702,
+      "loss": 1.095,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998883258473662,
+      "loss": 1.0742,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019988766223373494,
+      "loss": 1.2102,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998869966553163,
+      "loss": 1.0935,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019988632911212344,
+      "loss": 1.0718,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998856596041694,
+      "loss": 1.0757,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019988498813146736,
+      "loss": 1.0561,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019988431469403063,
+      "loss": 0.9842,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.0001998836392918723,
+      "loss": 1.0042,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019988296192500576,
+      "loss": 1.0444,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0.00019988228259344433,
+      "loss": 1.154,
+      "step": 1039
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019988160129720133,
+      "loss": 1.0303,
+      "step": 1040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019988091803629018,
+      "loss": 1.0995,
+      "step": 1041
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019988023281072433,
+      "loss": 1.164,
+      "step": 1042
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987954562051725,
+      "loss": 1.1527,
+      "step": 1043
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987885646568246,
+      "loss": 1.0216,
+      "step": 1044
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987816534623352,
+      "loss": 1.0693,
+      "step": 1045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987747226218397,
+      "loss": 0.9332,
+      "step": 1046
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987677721354754,
+      "loss": 1.1069,
+      "step": 1047
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987608020033783,
+      "loss": 1.1191,
+      "step": 1048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987538122256856,
+      "loss": 1.0522,
+      "step": 1049
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987468028025355,
+      "loss": 1.0805,
+      "step": 1050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987397737340645,
+      "loss": 0.9561,
+      "step": 1051
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987327250204118,
+      "loss": 1.0089,
+      "step": 1052
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987256566617162,
+      "loss": 1.1336,
+      "step": 1053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987185686581163,
+      "loss": 1.0315,
+      "step": 1054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019987114610097516,
+      "loss": 0.9382,
+      "step": 1055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998704333716762,
+      "loss": 1.1734,
+      "step": 1056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986971867792875,
+      "loss": 1.2827,
+      "step": 1057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998690020197469,
+      "loss": 1.0395,
+      "step": 1058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986828339714475,
+      "loss": 1.1241,
+      "step": 1059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998675628101364,
+      "loss": 1.1068,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986684025873598,
+      "loss": 1.1197,
+      "step": 1061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986611574295782,
+      "loss": 0.8394,
+      "step": 1062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998653892628161,
+      "loss": 0.8515,
+      "step": 1063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986466081832515,
+      "loss": 1.0698,
+      "step": 1064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986393040949922,
+      "loss": 1.1615,
+      "step": 1065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986319803635275,
+      "loss": 0.907,
+      "step": 1066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998624636989001,
+      "loss": 0.9806,
+      "step": 1067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998617273971558,
+      "loss": 1.1241,
+      "step": 1068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998609891311342,
+      "loss": 1.2036,
+      "step": 1069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019986024890084994,
+      "loss": 1.0935,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998595067063175,
+      "loss": 1.1263,
+      "step": 1071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985876254755152,
+      "loss": 0.8884,
+      "step": 1072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985801642456665,
+      "loss": 0.9934,
+      "step": 1073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985726833737753,
+      "loss": 1.0939,
+      "step": 1074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985651828599888,
+      "loss": 1.2312,
+      "step": 1075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985576627044548,
+      "loss": 1.1696,
+      "step": 1076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998550122907321,
+      "loss": 1.0833,
+      "step": 1077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998542563468736,
+      "loss": 0.872,
+      "step": 1078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985349843888478,
+      "loss": 1.2606,
+      "step": 1079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998527385667806,
+      "loss": 0.9036,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985197673057602,
+      "loss": 1.2751,
+      "step": 1081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985121293028603,
+      "loss": 1.2419,
+      "step": 1082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019985044716592558,
+      "loss": 1.1781,
+      "step": 1083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998496794375098,
+      "loss": 1.0242,
+      "step": 1084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984890974505381,
+      "loss": 1.1037,
+      "step": 1085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984813808857267,
+      "loss": 1.0772,
+      "step": 1086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984736446808162,
+      "loss": 0.9628,
+      "step": 1087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984658888359586,
+      "loss": 1.0623,
+      "step": 1088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984581133513065,
+      "loss": 0.9392,
+      "step": 1089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984503182270125,
+      "loss": 1.1415,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984425034632306,
+      "loss": 1.1902,
+      "step": 1091
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984346690601136,
+      "loss": 0.9895,
+      "step": 1092
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984268150178167,
+      "loss": 1.2027,
+      "step": 1093
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984189413364936,
+      "loss": 0.8283,
+      "step": 1094
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984110480162994,
+      "loss": 1.1377,
+      "step": 1095
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019984031350573896,
+      "loss": 1.0685,
+      "step": 1096
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998395202459919,
+      "loss": 1.1899,
+      "step": 1097
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019983872502240448,
+      "loss": 1.0753,
+      "step": 1098
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019983792783499226,
+      "loss": 0.9677,
+      "step": 1099
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000199837128683771,
+      "loss": 1.2352,
+      "step": 1100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998363275687563,
+      "loss": 1.202,
+      "step": 1101
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019983552448996398,
+      "loss": 1.1415,
+      "step": 1102
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998347194474099,
+      "loss": 1.0961,
+      "step": 1103
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998339124411098,
+      "loss": 1.1419,
+      "step": 1104
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998331034710796,
+      "loss": 1.1383,
+      "step": 1105
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019983229253733517,
+      "loss": 1.3991,
+      "step": 1106
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019983147963989253,
+      "loss": 1.0129,
+      "step": 1107
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019983066477876764,
+      "loss": 0.8857,
+      "step": 1108
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998298479539765,
+      "loss": 1.1629,
+      "step": 1109
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982902916553522,
+      "loss": 0.9737,
+      "step": 1110
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982820841345987,
+      "loss": 0.9808,
+      "step": 1111
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998273856977666,
+      "loss": 1.1355,
+      "step": 1112
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982656101847162,
+      "loss": 1.1035,
+      "step": 1113
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982573437559112,
+      "loss": 1.1428,
+      "step": 1114
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982490576914133,
+      "loss": 1.0884,
+      "step": 1115
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982407519913867,
+      "loss": 0.9565,
+      "step": 1116
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982324266559935,
+      "loss": 1.1208,
+      "step": 1117
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982240816853982,
+      "loss": 1.1449,
+      "step": 1118
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019982157170797643,
+      "loss": 1.108,
+      "step": 1119
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998207332839257,
+      "loss": 0.9198,
+      "step": 1120
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998198928964041,
+      "loss": 1.0119,
+      "step": 1121
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981905054542815,
+      "loss": 0.9456,
+      "step": 1122
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998182062310144,
+      "loss": 0.9116,
+      "step": 1123
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998173599531795,
+      "loss": 1.108,
+      "step": 1124
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981651171194006,
+      "loss": 0.9054,
+      "step": 1125
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981566150731282,
+      "loss": 1.2281,
+      "step": 1126
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981480933931443,
+      "loss": 0.9811,
+      "step": 1127
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981395520796168,
+      "loss": 1.1023,
+      "step": 1128
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981309911327143,
+      "loss": 1.0316,
+      "step": 1129
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981224105526042,
+      "loss": 1.0428,
+      "step": 1130
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998113810339456,
+      "loss": 1.1693,
+      "step": 1131
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019981051904934384,
+      "loss": 1.1609,
+      "step": 1132
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980965510147213,
+      "loss": 1.0303,
+      "step": 1133
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980878919034748,
+      "loss": 0.9143,
+      "step": 1134
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980792131598683,
+      "loss": 1.1902,
+      "step": 1135
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980705147840735,
+      "loss": 1.0441,
+      "step": 1136
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980617967762615,
+      "loss": 1.2065,
+      "step": 1137
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001998053059136603,
+      "loss": 1.0904,
+      "step": 1138
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000199804430186527,
+      "loss": 0.985,
+      "step": 1139
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980355249624355,
+      "loss": 0.8408,
+      "step": 1140
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980267284282717,
+      "loss": 0.956,
+      "step": 1141
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980179122629514,
+      "loss": 0.9583,
+      "step": 1142
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980090764666484,
+      "loss": 1.1046,
+      "step": 1143
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019980002210395363,
+      "loss": 1.0757,
+      "step": 1144
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979913459817895,
+      "loss": 1.0299,
+      "step": 1145
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979824512935822,
+      "loss": 0.9845,
+      "step": 1146
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979735369750893,
+      "loss": 1.0187,
+      "step": 1147
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979646030264867,
+      "loss": 0.9243,
+      "step": 1148
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.000199795564944795,
+      "loss": 1.0652,
+      "step": 1149
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997946676239655,
+      "loss": 1.1882,
+      "step": 1150
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979376834017778,
+      "loss": 1.0281,
+      "step": 1151
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979286709344965,
+      "loss": 1.0361,
+      "step": 1152
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979196388379876,
+      "loss": 1.1338,
+      "step": 1153
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019979105871124287,
+      "loss": 1.1081,
+      "step": 1154
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997901515757998,
+      "loss": 1.091,
+      "step": 1155
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978924247748738,
+      "loss": 0.9609,
+      "step": 1156
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978833141632354,
+      "loss": 0.9255,
+      "step": 1157
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978741839232615,
+      "loss": 1.0705,
+      "step": 1158
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997865034055132,
+      "loss": 1.0514,
+      "step": 1159
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978558645590265,
+      "loss": 1.1729,
+      "step": 1160
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978466754351262,
+      "loss": 1.1323,
+      "step": 1161
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978374666836107,
+      "loss": 1.056,
+      "step": 1162
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978282383046619,
+      "loss": 1.0469,
+      "step": 1163
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997818990298461,
+      "loss": 1.2245,
+      "step": 1164
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978097226651902,
+      "loss": 1.0421,
+      "step": 1165
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019978004354050314,
+      "loss": 0.9762,
+      "step": 1166
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977911285181676,
+      "loss": 0.955,
+      "step": 1167
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977818020047817,
+      "loss": 1.3432,
+      "step": 1168
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977724558650572,
+      "loss": 1.0477,
+      "step": 1169
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977630900991778,
+      "loss": 0.957,
+      "step": 1170
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977537047073286,
+      "loss": 1.1093,
+      "step": 1171
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977442996896926,
+      "loss": 1.0648,
+      "step": 1172
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997734875046456,
+      "loss": 1.0945,
+      "step": 1173
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977254307778042,
+      "loss": 1.2101,
+      "step": 1174
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019977159668839223,
+      "loss": 1.0743,
+      "step": 1175
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997706483364997,
+      "loss": 1.0911,
+      "step": 1176
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019976969802212144,
+      "loss": 0.9355,
+      "step": 1177
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019976874574527618,
+      "loss": 0.8967,
+      "step": 1178
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997677915059826,
+      "loss": 0.9634,
+      "step": 1179
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019976683530425958,
+      "loss": 0.995,
+      "step": 1180
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997658771401258,
+      "loss": 0.9818,
+      "step": 1181
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019976491701360018,
+      "loss": 1.0801,
+      "step": 1182
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997639549247016,
+      "loss": 0.8687,
+      "step": 1183
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019976299087344894,
+      "loss": 1.1587,
+      "step": 1184
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997620248598612,
+      "loss": 1.0027,
+      "step": 1185
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997610568839574,
+      "loss": 0.8936,
+      "step": 1186
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019976008694575653,
+      "loss": 1.1894,
+      "step": 1187
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019975911504527773,
+      "loss": 0.9609,
+      "step": 1188
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019975814118254007,
+      "loss": 1.0319,
+      "step": 1189
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997571653575627,
+      "loss": 0.9553,
+      "step": 1190
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019975618757036486,
+      "loss": 1.0917,
+      "step": 1191
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019975520782096573,
+      "loss": 1.0545,
+      "step": 1192
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019975422610938462,
+      "loss": 1.1741,
+      "step": 1193
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019975324243564081,
+      "loss": 1.1427,
+      "step": 1194
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997522567997537,
+      "loss": 0.9623,
+      "step": 1195
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997512692017426,
+      "loss": 1.0206,
+      "step": 1196
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019975027964162702,
+      "loss": 0.8896,
+      "step": 1197
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974928811942636,
+      "loss": 1.2582,
+      "step": 1198
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974829463516018,
+      "loss": 0.9834,
+      "step": 1199
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974729918884797,
+      "loss": 1.0382,
+      "step": 1200
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974630178050932,
+      "loss": 0.9319,
+      "step": 1201
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974530241016387,
+      "loss": 0.9493,
+      "step": 1202
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974430107783128,
+      "loss": 1.1111,
+      "step": 1203
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997432977835312,
+      "loss": 1.2217,
+      "step": 1204
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974229252728342,
+      "loss": 1.1163,
+      "step": 1205
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974128530910766,
+      "loss": 1.2649,
+      "step": 1206
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019974027612902382,
+      "loss": 1.1939,
+      "step": 1207
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019973926498705166,
+      "loss": 1.0879,
+      "step": 1208
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997382518832111,
+      "loss": 0.9939,
+      "step": 1209
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019973723681752204,
+      "loss": 0.9036,
+      "step": 1210
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019973621979000453,
+      "loss": 1.1234,
+      "step": 1211
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019973520080067853,
+      "loss": 1.0932,
+      "step": 1212
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019973417984956403,
+      "loss": 0.839,
+      "step": 1213
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997331569366812,
+      "loss": 1.0686,
+      "step": 1214
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997321320620501,
+      "loss": 1.1072,
+      "step": 1215
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019973110522569087,
+      "loss": 1.1214,
+      "step": 1216
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997300764276238,
+      "loss": 0.8993,
+      "step": 1217
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019972904566786903,
+      "loss": 1.0476,
+      "step": 1218
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997280129464469,
+      "loss": 1.1685,
+      "step": 1219
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019972697826337767,
+      "loss": 0.9204,
+      "step": 1220
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019972594161868178,
+      "loss": 1.1385,
+      "step": 1221
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019972490301237953,
+      "loss": 1.141,
+      "step": 1222
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019972386244449135,
+      "loss": 1.1074,
+      "step": 1223
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997228199150378,
+      "loss": 0.9383,
+      "step": 1224
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997217754240393,
+      "loss": 1.1815,
+      "step": 1225
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997207289715164,
+      "loss": 1.1733,
+      "step": 1226
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971968055748973,
+      "loss": 1.1865,
+      "step": 1227
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971863018197988,
+      "loss": 1.1563,
+      "step": 1228
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971757784500752,
+      "loss": 1.0737,
+      "step": 1229
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971652354659335,
+      "loss": 0.9439,
+      "step": 1230
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971546728675808,
+      "loss": 1.2904,
+      "step": 1231
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971440906552256,
+      "loss": 0.9589,
+      "step": 1232
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971334888290753,
+      "loss": 1.1378,
+      "step": 1233
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019971228673893387,
+      "loss": 1.0786,
+      "step": 1234
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997112226336225,
+      "loss": 1.0095,
+      "step": 1235
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997101565669943,
+      "loss": 1.0519,
+      "step": 1236
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970908853907026,
+      "loss": 0.9618,
+      "step": 1237
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970801854987143,
+      "loss": 1.1411,
+      "step": 1238
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001997069465994188,
+      "loss": 0.9455,
+      "step": 1239
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970587268773344,
+      "loss": 1.142,
+      "step": 1240
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970479681483654,
+      "loss": 0.9886,
+      "step": 1241
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970371898074923,
+      "loss": 1.2592,
+      "step": 1242
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970263918549274,
+      "loss": 1.026,
+      "step": 1243
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970155742908823,
+      "loss": 0.7936,
+      "step": 1244
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019970047371155704,
+      "loss": 1.0847,
+      "step": 1245
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019969938803292048,
+      "loss": 0.9299,
+      "step": 1246
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001996983003931999,
+      "loss": 1.057,
+      "step": 1247
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001996972107924167,
+      "loss": 1.1362,
+      "step": 1248
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001996961192305923,
+      "loss": 0.9014,
+      "step": 1249
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019969502570774822,
+      "loss": 1.0342,
+      "step": 1250
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019969393022390585,
+      "loss": 1.139,
+      "step": 1251
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019969283277908689,
+      "loss": 1.1385,
+      "step": 1252
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001996917333733128,
+      "loss": 1.1755,
+      "step": 1253
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001996906320066053,
+      "loss": 1.0614,
+      "step": 1254
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968952867898597,
+      "loss": 0.9961,
+      "step": 1255
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001996884233904766,
+      "loss": 0.822,
+      "step": 1256
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968731614109886,
+      "loss": 0.8999,
+      "step": 1257
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968620693087453,
+      "loss": 1.0307,
+      "step": 1258
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968509575982548,
+      "loss": 1.004,
+      "step": 1259
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968398262797354,
+      "loss": 1.2491,
+      "step": 1260
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968286753534062,
+      "loss": 1.0914,
+      "step": 1261
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968175048194864,
+      "loss": 0.9919,
+      "step": 1262
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019968063146781953,
+      "loss": 1.1235,
+      "step": 1263
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019967951049297536,
+      "loss": 1.2705,
+      "step": 1264
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019967838755743816,
+      "loss": 1.0583,
+      "step": 1265
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019967726266123003,
+      "loss": 1.0571,
+      "step": 1266
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0001996761358043731,
+      "loss": 1.0769,
+      "step": 1267
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019967500698688952,
+      "loss": 1.0346,
+      "step": 1268
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019967387620880146,
+      "loss": 1.0595,
+      "step": 1269
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00019967274347013124,
+      "loss": 0.9178,
+      "step": 1270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019967160877090107,
+      "loss": 1.1532,
+      "step": 1271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019967047211113333,
+      "loss": 1.2866,
+      "step": 1272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996693334908503,
+      "loss": 1.0138,
+      "step": 1273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019966819291007446,
+      "loss": 0.9124,
+      "step": 1274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996670503688282,
+      "loss": 0.9773,
+      "step": 1275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019966590586713398,
+      "loss": 0.9954,
+      "step": 1276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019966475940501435,
+      "loss": 1.008,
+      "step": 1277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019966361098249186,
+      "loss": 1.0841,
+      "step": 1278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019966246059958907,
+      "loss": 1.0485,
+      "step": 1279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996613082563286,
+      "loss": 1.0943,
+      "step": 1280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019966015395273318,
+      "loss": 1.0013,
+      "step": 1281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019965899768882545,
+      "loss": 1.1789,
+      "step": 1282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019965783946462818,
+      "loss": 0.9874,
+      "step": 1283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996566792801642,
+      "loss": 1.1735,
+      "step": 1284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996555171354562,
+      "loss": 0.8417,
+      "step": 1285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019965435303052718,
+      "loss": 1.0935,
+      "step": 1286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019965318696539994,
+      "loss": 0.9016,
+      "step": 1287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996520189400975,
+      "loss": 1.0523,
+      "step": 1288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019965084895464275,
+      "loss": 1.1538,
+      "step": 1289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019964967700905874,
+      "loss": 0.9428,
+      "step": 1290
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019964850310336853,
+      "loss": 1.0191,
+      "step": 1291
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019964732723759524,
+      "loss": 1.1568,
+      "step": 1292
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019964614941176195,
+      "loss": 0.995,
+      "step": 1293
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996449696258918,
+      "loss": 1.231,
+      "step": 1294
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996437878800081,
+      "loss": 1.1346,
+      "step": 1295
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199642604174134,
+      "loss": 1.0294,
+      "step": 1296
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019964141850829282,
+      "loss": 1.3396,
+      "step": 1297
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019964023088250786,
+      "loss": 1.1282,
+      "step": 1298
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019963904129680256,
+      "loss": 1.0707,
+      "step": 1299
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019963784975120022,
+      "loss": 1.0853,
+      "step": 1300
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019963665624572432,
+      "loss": 1.0403,
+      "step": 1301
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019963546078039833,
+      "loss": 1.0834,
+      "step": 1302
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996342633552458,
+      "loss": 1.0714,
+      "step": 1303
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019963306397029022,
+      "loss": 1.1036,
+      "step": 1304
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996318626255552,
+      "loss": 0.9654,
+      "step": 1305
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019963065932106442,
+      "loss": 1.0346,
+      "step": 1306
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996294540568415,
+      "loss": 1.2155,
+      "step": 1307
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019962824683291016,
+      "loss": 1.025,
+      "step": 1308
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019962703764929413,
+      "loss": 1.1001,
+      "step": 1309
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019962582650601723,
+      "loss": 1.1773,
+      "step": 1310
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019962461340310327,
+      "loss": 0.9627,
+      "step": 1311
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019962339834057608,
+      "loss": 0.9357,
+      "step": 1312
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996221813184596,
+      "loss": 0.9882,
+      "step": 1313
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019962096233677777,
+      "loss": 1.2171,
+      "step": 1314
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019961974139555454,
+      "loss": 0.8659,
+      "step": 1315
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996185184948139,
+      "loss": 1.0169,
+      "step": 1316
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019961729363458003,
+      "loss": 1.0857,
+      "step": 1317
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019961606681487685,
+      "loss": 1.1218,
+      "step": 1318
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019961483803572865,
+      "loss": 1.0003,
+      "step": 1319
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019961360729715946,
+      "loss": 1.0676,
+      "step": 1320
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996123745991936,
+      "loss": 1.0885,
+      "step": 1321
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019961113994185526,
+      "loss": 0.9626,
+      "step": 1322
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019960990332516874,
+      "loss": 1.0254,
+      "step": 1323
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019960866474915835,
+      "loss": 0.9999,
+      "step": 1324
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019960742421384846,
+      "loss": 1.3214,
+      "step": 1325
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996061817192635,
+      "loss": 1.0062,
+      "step": 1326
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019960493726542788,
+      "loss": 1.0235,
+      "step": 1327
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001996036908523661,
+      "loss": 0.947,
+      "step": 1328
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019960244248010264,
+      "loss": 1.1528,
+      "step": 1329
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019960119214866207,
+      "loss": 0.8263,
+      "step": 1330
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019959993985806901,
+      "loss": 1.011,
+      "step": 1331
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019959868560834806,
+      "loss": 1.0117,
+      "step": 1332
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019959742939952392,
+      "loss": 1.0277,
+      "step": 1333
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019959617123162127,
+      "loss": 1.1409,
+      "step": 1334
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019959491110466487,
+      "loss": 0.9694,
+      "step": 1335
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995936490186795,
+      "loss": 0.9409,
+      "step": 1336
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019959238497369003,
+      "loss": 1.073,
+      "step": 1337
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019959111896972128,
+      "loss": 1.0884,
+      "step": 1338
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958985100679814,
+      "loss": 0.9018,
+      "step": 1339
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995885810849456,
+      "loss": 1.1894,
+      "step": 1340
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958730920418858,
+      "loss": 0.9337,
+      "step": 1341
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958603536455213,
+      "loss": 1.0211,
+      "step": 1342
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958475956606133,
+      "loss": 1.0155,
+      "step": 1343
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019958348180874124,
+      "loss": 1.0774,
+      "step": 1344
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199582202092617,
+      "loss": 0.9777,
+      "step": 1345
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995809204177138,
+      "loss": 1.0602,
+      "step": 1346
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995796367840568,
+      "loss": 1.1182,
+      "step": 1347
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995783511916713,
+      "loss": 1.0374,
+      "step": 1348
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019957706364058257,
+      "loss": 1.1261,
+      "step": 1349
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019957577413081597,
+      "loss": 1.0019,
+      "step": 1350
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995744826623968,
+      "loss": 1.0157,
+      "step": 1351
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019957318923535052,
+      "loss": 1.0196,
+      "step": 1352
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995718938497025,
+      "loss": 1.3088,
+      "step": 1353
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995705965054783,
+      "loss": 1.3097,
+      "step": 1354
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019956929720270342,
+      "loss": 1.1619,
+      "step": 1355
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019956799594140337,
+      "loss": 1.0795,
+      "step": 1356
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019956669272160383,
+      "loss": 0.912,
+      "step": 1357
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019956538754333034,
+      "loss": 0.9614,
+      "step": 1358
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019956408040660865,
+      "loss": 1.0774,
+      "step": 1359
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995627713114644,
+      "loss": 0.973,
+      "step": 1360
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995614602579234,
+      "loss": 1.1799,
+      "step": 1361
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019956014724601138,
+      "loss": 1.1169,
+      "step": 1362
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019955883227575427,
+      "loss": 1.1096,
+      "step": 1363
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019955751534717783,
+      "loss": 1.1519,
+      "step": 1364
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019955619646030802,
+      "loss": 1.1592,
+      "step": 1365
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019955487561517074,
+      "loss": 1.0991,
+      "step": 1366
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199553552811792,
+      "loss": 1.1525,
+      "step": 1367
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019955222805019784,
+      "loss": 1.0921,
+      "step": 1368
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019955090133041427,
+      "loss": 1.0996,
+      "step": 1369
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019954957265246742,
+      "loss": 1.0807,
+      "step": 1370
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019954824201638344,
+      "loss": 1.0991,
+      "step": 1371
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019954690942218842,
+      "loss": 1.037,
+      "step": 1372
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019954557486990868,
+      "loss": 1.2911,
+      "step": 1373
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995442383595704,
+      "loss": 1.0392,
+      "step": 1374
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995428998911999,
+      "loss": 0.9148,
+      "step": 1375
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995415594648235,
+      "loss": 1.0364,
+      "step": 1376
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019954021708046754,
+      "loss": 0.9896,
+      "step": 1377
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995388727381585,
+      "loss": 1.1423,
+      "step": 1378
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019953752643792272,
+      "loss": 1.1103,
+      "step": 1379
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019953617817978673,
+      "loss": 1.064,
+      "step": 1380
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019953482796377712,
+      "loss": 1.1289,
+      "step": 1381
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019953347578992032,
+      "loss": 1.1299,
+      "step": 1382
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199532121658243,
+      "loss": 0.9457,
+      "step": 1383
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995307655687718,
+      "loss": 0.9698,
+      "step": 1384
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019952940752153338,
+      "loss": 1.2068,
+      "step": 1385
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019952804751655444,
+      "loss": 1.0911,
+      "step": 1386
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019952668555386178,
+      "loss": 0.9965,
+      "step": 1387
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019952532163348212,
+      "loss": 1.0255,
+      "step": 1388
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019952395575544233,
+      "loss": 1.0259,
+      "step": 1389
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019952258791976926,
+      "loss": 1.0752,
+      "step": 1390
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019952121812648983,
+      "loss": 1.058,
+      "step": 1391
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019951984637563098,
+      "loss": 0.901,
+      "step": 1392
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995184726672197,
+      "loss": 1.3004,
+      "step": 1393
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199517097001283,
+      "loss": 1.1426,
+      "step": 1394
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019951571937784792,
+      "loss": 1.0119,
+      "step": 1395
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995143397969416,
+      "loss": 0.9811,
+      "step": 1396
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019951295825859113,
+      "loss": 1.0617,
+      "step": 1397
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019951157476282374,
+      "loss": 1.3051,
+      "step": 1398
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001995101893096666,
+      "loss": 0.8295,
+      "step": 1399
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019950880189914695,
+      "loss": 1.0564,
+      "step": 1400
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019950741253129214,
+      "loss": 1.2663,
+      "step": 1401
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019950602120612945,
+      "loss": 1.0431,
+      "step": 1402
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019950462792368625,
+      "loss": 0.9666,
+      "step": 1403
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019950323268398999,
+      "loss": 1.192,
+      "step": 1404
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019950183548706805,
+      "loss": 1.0132,
+      "step": 1405
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019950043633294793,
+      "loss": 1.0153,
+      "step": 1406
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994990352216572,
+      "loss": 0.9716,
+      "step": 1407
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019949763215322338,
+      "loss": 1.0771,
+      "step": 1408
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019949622712767405,
+      "loss": 0.9546,
+      "step": 1409
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019949482014503688,
+      "loss": 1.0802,
+      "step": 1410
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019949341120533955,
+      "loss": 1.0196,
+      "step": 1411
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019949200030860975,
+      "loss": 1.0642,
+      "step": 1412
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019949058745487522,
+      "loss": 1.3107,
+      "step": 1413
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019948917264416382,
+      "loss": 1.2407,
+      "step": 1414
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994877558765033,
+      "loss": 1.0374,
+      "step": 1415
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019948633715192158,
+      "loss": 1.1204,
+      "step": 1416
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994849164704465,
+      "loss": 1.1865,
+      "step": 1417
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994834938321061,
+      "loss": 0.9601,
+      "step": 1418
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019948206923692832,
+      "loss": 0.9596,
+      "step": 1419
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019948064268494113,
+      "loss": 1.1706,
+      "step": 1420
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019947921417617267,
+      "loss": 1.0477,
+      "step": 1421
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019947778371065097,
+      "loss": 1.0028,
+      "step": 1422
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019947635128840423,
+      "loss": 1.1585,
+      "step": 1423
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994749169094606,
+      "loss": 1.323,
+      "step": 1424
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019947348057384827,
+      "loss": 0.9752,
+      "step": 1425
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019947204228159553,
+      "loss": 1.0841,
+      "step": 1426
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019947060203273065,
+      "loss": 0.9941,
+      "step": 1427
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019946915982728197,
+      "loss": 1.1073,
+      "step": 1428
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019946771566527785,
+      "loss": 1.0815,
+      "step": 1429
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994662695467467,
+      "loss": 1.0222,
+      "step": 1430
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019946482147171696,
+      "loss": 1.1548,
+      "step": 1431
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994633714402171,
+      "loss": 1.0683,
+      "step": 1432
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019946191945227572,
+      "loss": 1.0807,
+      "step": 1433
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019946046550792126,
+      "loss": 1.1221,
+      "step": 1434
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994590096071824,
+      "loss": 0.9279,
+      "step": 1435
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019945755175008776,
+      "loss": 1.0494,
+      "step": 1436
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199456091936666,
+      "loss": 1.1625,
+      "step": 1437
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994546301669459,
+      "loss": 1.1691,
+      "step": 1438
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994531664409561,
+      "loss": 1.1659,
+      "step": 1439
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994517007587255,
+      "loss": 1.167,
+      "step": 1440
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019945023312028287,
+      "loss": 1.4275,
+      "step": 1441
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019944876352565705,
+      "loss": 1.1271,
+      "step": 1442
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199447291974877,
+      "loss": 1.1877,
+      "step": 1443
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994458184679717,
+      "loss": 1.0009,
+      "step": 1444
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019944434300497006,
+      "loss": 1.2127,
+      "step": 1445
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019944286558590114,
+      "loss": 1.3325,
+      "step": 1446
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019944138621079397,
+      "loss": 1.0785,
+      "step": 1447
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019943990487967766,
+      "loss": 0.9721,
+      "step": 1448
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994384215925814,
+      "loss": 0.8762,
+      "step": 1449
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019943693634953428,
+      "loss": 1.0298,
+      "step": 1450
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019943544915056557,
+      "loss": 1.1801,
+      "step": 1451
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994339599957045,
+      "loss": 1.1463,
+      "step": 1452
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019943246888498041,
+      "loss": 1.2042,
+      "step": 1453
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019943097581842257,
+      "loss": 1.1979,
+      "step": 1454
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994294807960604,
+      "loss": 1.219,
+      "step": 1455
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019942798381792325,
+      "loss": 1.1586,
+      "step": 1456
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019942648488404058,
+      "loss": 1.0535,
+      "step": 1457
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019942498399444188,
+      "loss": 1.2647,
+      "step": 1458
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019942348114915674,
+      "loss": 1.2154,
+      "step": 1459
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019942197634821462,
+      "loss": 0.9522,
+      "step": 1460
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019942046959164515,
+      "loss": 1.1261,
+      "step": 1461
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199418960879478,
+      "loss": 0.8639,
+      "step": 1462
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019941745021174282,
+      "loss": 1.0101,
+      "step": 1463
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019941593758846936,
+      "loss": 0.9755,
+      "step": 1464
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019941442300968728,
+      "loss": 1.1115,
+      "step": 1465
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019941290647542648,
+      "loss": 1.0894,
+      "step": 1466
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019941138798571672,
+      "loss": 1.1762,
+      "step": 1467
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019940986754058793,
+      "loss": 1.0145,
+      "step": 1468
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019940834514006994,
+      "loss": 1.3063,
+      "step": 1469
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019940682078419276,
+      "loss": 0.9813,
+      "step": 1470
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019940529447298635,
+      "loss": 0.9601,
+      "step": 1471
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001994037662064807,
+      "loss": 1.0346,
+      "step": 1472
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019940223598470595,
+      "loss": 1.1643,
+      "step": 1473
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019940070380769213,
+      "loss": 1.1641,
+      "step": 1474
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001993991696754694,
+      "loss": 1.1571,
+      "step": 1475
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019939763358806794,
+      "loss": 0.9493,
+      "step": 1476
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199396095545518,
+      "loss": 1.0559,
+      "step": 1477
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019939455554784973,
+      "loss": 1.0224,
+      "step": 1478
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019939301359509352,
+      "loss": 1.0163,
+      "step": 1479
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019939146968727967,
+      "loss": 0.8265,
+      "step": 1480
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019938992382443856,
+      "loss": 1.1808,
+      "step": 1481
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019938837600660055,
+      "loss": 1.0342,
+      "step": 1482
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001993868262337961,
+      "loss": 1.1838,
+      "step": 1483
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019938527450605577,
+      "loss": 1.0842,
+      "step": 1484
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019938372082341,
+      "loss": 0.9719,
+      "step": 1485
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019938216518588936,
+      "loss": 1.1932,
+      "step": 1486
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019938060759352452,
+      "loss": 0.9669,
+      "step": 1487
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.000199379048046346,
+      "loss": 1.1413,
+      "step": 1488
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019937748654438455,
+      "loss": 1.1317,
+      "step": 1489
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019937592308767087,
+      "loss": 0.9919,
+      "step": 1490
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019937435767623576,
+      "loss": 1.1086,
+      "step": 1491
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019937279031010993,
+      "loss": 1.1683,
+      "step": 1492
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019937122098932428,
+      "loss": 1.0939,
+      "step": 1493
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019936964971390962,
+      "loss": 1.0751,
+      "step": 1494
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001993680764838969,
+      "loss": 1.0252,
+      "step": 1495
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019936650129931703,
+      "loss": 1.2316,
+      "step": 1496
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019936492416020103,
+      "loss": 1.0084,
+      "step": 1497
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0001993633450665799,
+      "loss": 1.1706,
+      "step": 1498
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019936176401848469,
+      "loss": 1.1262,
+      "step": 1499
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019936018101594652,
+      "loss": 1.2745,
+      "step": 1500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.00019935859605899655,
+      "loss": 0.8727,
+      "step": 1501
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019935700914766594,
+      "loss": 0.8789,
+      "step": 1502
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019935542028198583,
+      "loss": 0.9287,
+      "step": 1503
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993538294619876,
+      "loss": 1.1388,
+      "step": 1504
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019935223668770245,
+      "loss": 1.0665,
+      "step": 1505
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019935064195916177,
+      "loss": 1.0759,
+      "step": 1506
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019934904527639684,
+      "loss": 1.0021,
+      "step": 1507
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019934744663943917,
+      "loss": 1.0965,
+      "step": 1508
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019934584604832013,
+      "loss": 1.2173,
+      "step": 1509
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019934424350307123,
+      "loss": 1.3284,
+      "step": 1510
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199342639003724,
+      "loss": 1.1508,
+      "step": 1511
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019934103255031,
+      "loss": 0.995,
+      "step": 1512
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019933942414286082,
+      "loss": 1.2311,
+      "step": 1513
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993378137814081,
+      "loss": 1.1342,
+      "step": 1514
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993362014659835,
+      "loss": 0.9689,
+      "step": 1515
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019933458719661876,
+      "loss": 1.164,
+      "step": 1516
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019933297097334565,
+      "loss": 0.9936,
+      "step": 1517
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993313527961959,
+      "loss": 0.9319,
+      "step": 1518
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019932973266520142,
+      "loss": 1.0895,
+      "step": 1519
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019932811058039396,
+      "loss": 0.8248,
+      "step": 1520
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019932648654180554,
+      "loss": 1.09,
+      "step": 1521
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019932486054946807,
+      "loss": 1.1402,
+      "step": 1522
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019932323260341353,
+      "loss": 1.0178,
+      "step": 1523
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993216027036739,
+      "loss": 1.1977,
+      "step": 1524
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019931997085028132,
+      "loss": 1.1705,
+      "step": 1525
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993183370432678,
+      "loss": 1.1458,
+      "step": 1526
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019931670128266557,
+      "loss": 1.0333,
+      "step": 1527
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993150635685067,
+      "loss": 1.171,
+      "step": 1528
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993134239008235,
+      "loss": 1.3189,
+      "step": 1529
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993117822796482,
+      "loss": 0.958,
+      "step": 1530
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019931013870501304,
+      "loss": 0.8747,
+      "step": 1531
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001993084931769504,
+      "loss": 1.2225,
+      "step": 1532
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019930684569549264,
+      "loss": 0.9964,
+      "step": 1533
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019930519626067215,
+      "loss": 1.1457,
+      "step": 1534
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019930354487252138,
+      "loss": 1.0737,
+      "step": 1535
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019930189153107282,
+      "loss": 1.1682,
+      "step": 1536
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019930023623635897,
+      "loss": 1.155,
+      "step": 1537
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019929857898841243,
+      "loss": 1.001,
+      "step": 1538
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019929691978726576,
+      "loss": 1.1238,
+      "step": 1539
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019929525863295158,
+      "loss": 0.8596,
+      "step": 1540
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019929359552550265,
+      "loss": 1.1032,
+      "step": 1541
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992919304649516,
+      "loss": 1.0843,
+      "step": 1542
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019929026345133122,
+      "loss": 1.0024,
+      "step": 1543
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019928859448467426,
+      "loss": 1.0327,
+      "step": 1544
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992869235650136,
+      "loss": 1.112,
+      "step": 1545
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019928525069238208,
+      "loss": 0.9799,
+      "step": 1546
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019928357586681262,
+      "loss": 1.1031,
+      "step": 1547
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019928189908833814,
+      "loss": 0.8914,
+      "step": 1548
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019928022035699166,
+      "loss": 0.9882,
+      "step": 1549
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019927853967280617,
+      "loss": 0.9703,
+      "step": 1550
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019927685703581468,
+      "loss": 1.1153,
+      "step": 1551
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019927517244605042,
+      "loss": 0.9425,
+      "step": 1552
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019927348590354643,
+      "loss": 1.11,
+      "step": 1553
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992717974083359,
+      "loss": 1.0314,
+      "step": 1554
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019927010696045203,
+      "loss": 1.05,
+      "step": 1555
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019926841455992808,
+      "loss": 0.9589,
+      "step": 1556
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019926672020679736,
+      "loss": 0.9143,
+      "step": 1557
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992650239010932,
+      "loss": 1.0496,
+      "step": 1558
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019926332564284893,
+      "loss": 1.0154,
+      "step": 1559
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019926162543209799,
+      "loss": 1.1288,
+      "step": 1560
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019925992326887379,
+      "loss": 1.3342,
+      "step": 1561
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019925821915320983,
+      "loss": 1.0676,
+      "step": 1562
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019925651308513967,
+      "loss": 1.0572,
+      "step": 1563
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992548050646968,
+      "loss": 1.2167,
+      "step": 1564
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019925309509191482,
+      "loss": 1.1699,
+      "step": 1565
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019925138316682743,
+      "loss": 1.2471,
+      "step": 1566
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019924966928946828,
+      "loss": 1.0127,
+      "step": 1567
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199247953459871,
+      "loss": 1.1219,
+      "step": 1568
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019924623567806946,
+      "loss": 1.0511,
+      "step": 1569
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019924451594409737,
+      "loss": 1.1926,
+      "step": 1570
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992427942579886,
+      "loss": 1.0369,
+      "step": 1571
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019924107061977697,
+      "loss": 1.1145,
+      "step": 1572
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019923934502949644,
+      "loss": 1.0925,
+      "step": 1573
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019923761748718088,
+      "loss": 1.1197,
+      "step": 1574
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019923588799286438,
+      "loss": 1.0163,
+      "step": 1575
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019923415654658083,
+      "loss": 1.0014,
+      "step": 1576
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992324231483644,
+      "loss": 0.9393,
+      "step": 1577
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992306877982491,
+      "loss": 0.9336,
+      "step": 1578
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019922895049626912,
+      "loss": 1.1126,
+      "step": 1579
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992272112424586,
+      "loss": 1.0859,
+      "step": 1580
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019922547003685176,
+      "loss": 1.0326,
+      "step": 1581
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019922372687948287,
+      "loss": 1.1303,
+      "step": 1582
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992219817703862,
+      "loss": 0.8763,
+      "step": 1583
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019922023470959609,
+      "loss": 1.0203,
+      "step": 1584
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019921848569714686,
+      "loss": 1.1338,
+      "step": 1585
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199216734733073,
+      "loss": 1.0165,
+      "step": 1586
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019921498181740886,
+      "loss": 0.9315,
+      "step": 1587
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019921322695018893,
+      "loss": 0.8543,
+      "step": 1588
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992114701314478,
+      "loss": 1.1134,
+      "step": 1589
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019920971136121995,
+      "loss": 0.9273,
+      "step": 1590
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019920795063954003,
+      "loss": 1.0756,
+      "step": 1591
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019920618796644263,
+      "loss": 1.0801,
+      "step": 1592
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019920442334196248,
+      "loss": 0.98,
+      "step": 1593
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001992026567661342,
+      "loss": 1.1128,
+      "step": 1594
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019920088823899263,
+      "loss": 1.2339,
+      "step": 1595
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991991177605725,
+      "loss": 0.8721,
+      "step": 1596
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019919734533090867,
+      "loss": 1.0949,
+      "step": 1597
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199195570950036,
+      "loss": 0.9252,
+      "step": 1598
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019919379461798932,
+      "loss": 0.9623,
+      "step": 1599
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019919201633480368,
+      "loss": 1.0751,
+      "step": 1600
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019919023610051397,
+      "loss": 0.9657,
+      "step": 1601
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991884539151553,
+      "loss": 1.1277,
+      "step": 1602
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991866697787626,
+      "loss": 0.9146,
+      "step": 1603
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991848836913711,
+      "loss": 1.0979,
+      "step": 1604
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991830956530158,
+      "loss": 1.0156,
+      "step": 1605
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199181305663732,
+      "loss": 1.0966,
+      "step": 1606
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991795137235548,
+      "loss": 1.1788,
+      "step": 1607
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019917771983251953,
+      "loss": 1.1432,
+      "step": 1608
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991759239906614,
+      "loss": 1.2402,
+      "step": 1609
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019917412619801578,
+      "loss": 1.1636,
+      "step": 1610
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019917232645461804,
+      "loss": 0.9499,
+      "step": 1611
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019917052476050358,
+      "loss": 1.0341,
+      "step": 1612
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019916872111570784,
+      "loss": 0.8374,
+      "step": 1613
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019916691552026623,
+      "loss": 0.8894,
+      "step": 1614
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019916510797421435,
+      "loss": 0.9047,
+      "step": 1615
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019916329847758773,
+      "loss": 1.081,
+      "step": 1616
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019916148703042193,
+      "loss": 0.9071,
+      "step": 1617
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019915967363275263,
+      "loss": 1.0914,
+      "step": 1618
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019915785828461546,
+      "loss": 1.1016,
+      "step": 1619
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019915604098604618,
+      "loss": 0.9408,
+      "step": 1620
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019915422173708046,
+      "loss": 1.1723,
+      "step": 1621
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019915240053775416,
+      "loss": 1.0424,
+      "step": 1622
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019915057738810306,
+      "loss": 0.9422,
+      "step": 1623
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019914875228816304,
+      "loss": 1.0639,
+      "step": 1624
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019914692523796993,
+      "loss": 1.091,
+      "step": 1625
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991450962375598,
+      "loss": 1.1333,
+      "step": 1626
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019914326528696851,
+      "loss": 0.9132,
+      "step": 1627
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019914143238623216,
+      "loss": 1.2037,
+      "step": 1628
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019913959753538674,
+      "loss": 1.0351,
+      "step": 1629
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019913776073446834,
+      "loss": 1.0887,
+      "step": 1630
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019913592198351313,
+      "loss": 0.9374,
+      "step": 1631
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019913408128255727,
+      "loss": 1.0189,
+      "step": 1632
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019913223863163697,
+      "loss": 1.0963,
+      "step": 1633
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019913039403078844,
+      "loss": 1.078,
+      "step": 1634
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199128547480048,
+      "loss": 1.0201,
+      "step": 1635
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019912669897945195,
+      "loss": 1.0237,
+      "step": 1636
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019912484852903668,
+      "loss": 0.9666,
+      "step": 1637
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019912299612883852,
+      "loss": 0.9296,
+      "step": 1638
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019912114177889398,
+      "loss": 1.0113,
+      "step": 1639
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019911928547923952,
+      "loss": 1.035,
+      "step": 1640
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019911742722991162,
+      "loss": 0.9294,
+      "step": 1641
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019911556703094686,
+      "loss": 0.8983,
+      "step": 1642
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019911370488238182,
+      "loss": 1.0952,
+      "step": 1643
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019911184078425314,
+      "loss": 1.275,
+      "step": 1644
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001991099747365975,
+      "loss": 0.9445,
+      "step": 1645
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019910810673945154,
+      "loss": 1.2132,
+      "step": 1646
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019910623679285206,
+      "loss": 0.9869,
+      "step": 1647
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019910436489683586,
+      "loss": 1.0844,
+      "step": 1648
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019910249105143968,
+      "loss": 1.0587,
+      "step": 1649
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019910061525670045,
+      "loss": 1.1262,
+      "step": 1650
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019909873751265504,
+      "loss": 0.9004,
+      "step": 1651
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990968578193404,
+      "loss": 1.1638,
+      "step": 1652
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019909497617679348,
+      "loss": 0.9767,
+      "step": 1653
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019909309258505132,
+      "loss": 1.0415,
+      "step": 1654
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019909120704415093,
+      "loss": 1.0918,
+      "step": 1655
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019908931955412947,
+      "loss": 1.1927,
+      "step": 1656
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019908743011502397,
+      "loss": 1.1829,
+      "step": 1657
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019908553872687165,
+      "loss": 0.8071,
+      "step": 1658
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990836453897097,
+      "loss": 1.0962,
+      "step": 1659
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990817501035754,
+      "loss": 1.0393,
+      "step": 1660
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199079852868506,
+      "loss": 1.0379,
+      "step": 1661
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019907795368453882,
+      "loss": 1.0806,
+      "step": 1662
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990760525517112,
+      "loss": 0.8672,
+      "step": 1663
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019907414947006058,
+      "loss": 1.1502,
+      "step": 1664
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019907224443962434,
+      "loss": 1.0233,
+      "step": 1665
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019907033746044,
+      "loss": 1.0902,
+      "step": 1666
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199068428532545,
+      "loss": 1.1785,
+      "step": 1667
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199066517655977,
+      "loss": 1.2735,
+      "step": 1668
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019906460483077348,
+      "loss": 0.8215,
+      "step": 1669
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990626900569721,
+      "loss": 1.0358,
+      "step": 1670
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019906077333461055,
+      "loss": 1.115,
+      "step": 1671
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990588546637265,
+      "loss": 1.1867,
+      "step": 1672
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019905693404435773,
+      "loss": 1.1547,
+      "step": 1673
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019905501147654198,
+      "loss": 0.8653,
+      "step": 1674
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019905308696031707,
+      "loss": 1.1816,
+      "step": 1675
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019905116049572088,
+      "loss": 1.1369,
+      "step": 1676
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990492320827913,
+      "loss": 1.2858,
+      "step": 1677
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990473017215662,
+      "loss": 1.0674,
+      "step": 1678
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019904536941208365,
+      "loss": 1.0968,
+      "step": 1679
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019904343515438156,
+      "loss": 1.1373,
+      "step": 1680
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019904149894849806,
+      "loss": 1.0157,
+      "step": 1681
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990395607944712,
+      "loss": 0.967,
+      "step": 1682
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019903762069233908,
+      "loss": 0.8577,
+      "step": 1683
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990356786421399,
+      "loss": 1.1557,
+      "step": 1684
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019903373464391186,
+      "loss": 0.8342,
+      "step": 1685
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019903178869769316,
+      "loss": 1.0197,
+      "step": 1686
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019902984080352208,
+      "loss": 1.251,
+      "step": 1687
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.000199027890961437,
+      "loss": 0.8847,
+      "step": 1688
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019902593917147624,
+      "loss": 1.3314,
+      "step": 1689
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019902398543367812,
+      "loss": 0.9865,
+      "step": 1690
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990220297480812,
+      "loss": 1.2786,
+      "step": 1691
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990200721147238,
+      "loss": 1.1378,
+      "step": 1692
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019901811253364456,
+      "loss": 1.1137,
+      "step": 1693
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019901615100488199,
+      "loss": 1.0779,
+      "step": 1694
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019901418752847462,
+      "loss": 0.922,
+      "step": 1695
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019901222210446106,
+      "loss": 0.8978,
+      "step": 1696
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990102547328801,
+      "loss": 1.0207,
+      "step": 1697
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001990082854137703,
+      "loss": 1.0011,
+      "step": 1698
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019900631414717046,
+      "loss": 0.9732,
+      "step": 1699
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019900434093311935,
+      "loss": 1.1203,
+      "step": 1700
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019900236577165576,
+      "loss": 1.0633,
+      "step": 1701
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019900038866281858,
+      "loss": 1.0279,
+      "step": 1702
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019899840960664665,
+      "loss": 0.9637,
+      "step": 1703
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019899642860317897,
+      "loss": 0.9621,
+      "step": 1704
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019899444565245445,
+      "loss": 0.9086,
+      "step": 1705
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019899246075451207,
+      "loss": 1.0336,
+      "step": 1706
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019899047390939094,
+      "loss": 0.9459,
+      "step": 1707
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989884851171301,
+      "loss": 1.1832,
+      "step": 1708
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989864943777687,
+      "loss": 0.9817,
+      "step": 1709
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019898450169134586,
+      "loss": 0.9445,
+      "step": 1710
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989825070579008,
+      "loss": 1.1879,
+      "step": 1711
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019898051047747276,
+      "loss": 1.1224,
+      "step": 1712
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897851195010098,
+      "loss": 1.065,
+      "step": 1713
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897651147582482,
+      "loss": 1.1211,
+      "step": 1714
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897450905468356,
+      "loss": 0.889,
+      "step": 1715
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019897250468671666,
+      "loss": 1.166,
+      "step": 1716
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989704983719635,
+      "loss": 0.9479,
+      "step": 1717
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019896849011046358,
+      "loss": 0.9977,
+      "step": 1718
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019896647990225635,
+      "loss": 0.9968,
+      "step": 1719
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019896446774738142,
+      "loss": 0.998,
+      "step": 1720
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019896245364587828,
+      "loss": 1.227,
+      "step": 1721
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019896043759778665,
+      "loss": 1.0738,
+      "step": 1722
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019895841960314612,
+      "loss": 1.3038,
+      "step": 1723
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989563996619964,
+      "loss": 1.0493,
+      "step": 1724
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989543777743772,
+      "loss": 1.1893,
+      "step": 1725
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019895235394032831,
+      "loss": 1.0871,
+      "step": 1726
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989503281598896,
+      "loss": 1.0514,
+      "step": 1727
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989483004331008,
+      "loss": 0.9259,
+      "step": 1728
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019894627076000185,
+      "loss": 1.0825,
+      "step": 1729
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.0001989442391406327,
+      "loss": 1.0171,
+      "step": 1730
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 0.00019894220557503328,
+      "loss": 1.1273,
+      "step": 1731
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019894017006324356,
+      "loss": 1.0412,
+      "step": 1732
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019893813260530368,
+      "loss": 1.101,
+      "step": 1733
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001989360932012536,
+      "loss": 1.0339,
+      "step": 1734
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019893405185113352,
+      "loss": 0.8728,
+      "step": 1735
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019893200855498354,
+      "loss": 1.1526,
+      "step": 1736
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019892996331284391,
+      "loss": 1.1485,
+      "step": 1737
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001989279161247548,
+      "loss": 1.0305,
+      "step": 1738
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001989258669907565,
+      "loss": 1.214,
+      "step": 1739
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001989238159108893,
+      "loss": 1.2694,
+      "step": 1740
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001989217628851936,
+      "loss": 1.0061,
+      "step": 1741
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019891970791370973,
+      "loss": 1.0869,
+      "step": 1742
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001989176509964781,
+      "loss": 1.1221,
+      "step": 1743
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019891559213353922,
+      "loss": 0.9141,
+      "step": 1744
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019891353132493357,
+      "loss": 0.9025,
+      "step": 1745
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019891146857070166,
+      "loss": 1.0132,
+      "step": 1746
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019890940387088407,
+      "loss": 0.9787,
+      "step": 1747
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019890733722552147,
+      "loss": 1.1004,
+      "step": 1748
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019890526863465443,
+      "loss": 1.2964,
+      "step": 1749
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019890319809832372,
+      "loss": 1.1846,
+      "step": 1750
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019890112561657,
+      "loss": 0.9527,
+      "step": 1751
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000198899051189434,
+      "loss": 0.9426,
+      "step": 1752
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019889697481695668,
+      "loss": 0.9973,
+      "step": 1753
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988948964991787,
+      "loss": 1.0878,
+      "step": 1754
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019889281623614111,
+      "loss": 1.0287,
+      "step": 1755
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988907340278847,
+      "loss": 1.1516,
+      "step": 1756
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988886498744505,
+      "loss": 1.1957,
+      "step": 1757
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019888656377587942,
+      "loss": 0.8257,
+      "step": 1758
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019888447573221258,
+      "loss": 0.9537,
+      "step": 1759
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019888238574349102,
+      "loss": 1.0407,
+      "step": 1760
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019888029380975587,
+      "loss": 1.0402,
+      "step": 1761
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019887819993104824,
+      "loss": 1.1584,
+      "step": 1762
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019887610410740934,
+      "loss": 1.0097,
+      "step": 1763
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988740063388804,
+      "loss": 0.9108,
+      "step": 1764
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019887190662550266,
+      "loss": 1.0342,
+      "step": 1765
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019886980496731745,
+      "loss": 1.0406,
+      "step": 1766
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988677013643661,
+      "loss": 1.239,
+      "step": 1767
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019886559581668999,
+      "loss": 0.9017,
+      "step": 1768
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019886348832433049,
+      "loss": 1.3598,
+      "step": 1769
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988613788873291,
+      "loss": 1.0415,
+      "step": 1770
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019885926750572736,
+      "loss": 1.0671,
+      "step": 1771
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988571541795667,
+      "loss": 1.1986,
+      "step": 1772
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019885503890888876,
+      "loss": 1.0493,
+      "step": 1773
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019885292169373507,
+      "loss": 0.8547,
+      "step": 1774
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019885080253414738,
+      "loss": 0.9722,
+      "step": 1775
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019884868143016732,
+      "loss": 0.8647,
+      "step": 1776
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988465583818366,
+      "loss": 1.2996,
+      "step": 1777
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000198844433389197,
+      "loss": 1.1078,
+      "step": 1778
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019884230645229032,
+      "loss": 1.1285,
+      "step": 1779
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988401775711584,
+      "loss": 0.8843,
+      "step": 1780
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988380467458431,
+      "loss": 0.9705,
+      "step": 1781
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019883591397638632,
+      "loss": 1.0358,
+      "step": 1782
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019883377926283006,
+      "loss": 1.0161,
+      "step": 1783
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019883164260521623,
+      "loss": 0.9708,
+      "step": 1784
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019882950400358694,
+      "loss": 1.1599,
+      "step": 1785
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019882736345798422,
+      "loss": 1.1182,
+      "step": 1786
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988252209684502,
+      "loss": 1.0072,
+      "step": 1787
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019882307653502698,
+      "loss": 1.0393,
+      "step": 1788
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019882093015775678,
+      "loss": 1.092,
+      "step": 1789
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019881878183668176,
+      "loss": 1.0528,
+      "step": 1790
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019881663157184426,
+      "loss": 1.1713,
+      "step": 1791
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019881447936328652,
+      "loss": 1.1878,
+      "step": 1792
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019881232521105089,
+      "loss": 1.09,
+      "step": 1793
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019881016911517976,
+      "loss": 0.9088,
+      "step": 1794
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019880801107571548,
+      "loss": 1.1554,
+      "step": 1795
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019880585109270058,
+      "loss": 1.0406,
+      "step": 1796
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001988036891661775,
+      "loss": 1.062,
+      "step": 1797
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019880152529618873,
+      "loss": 1.0779,
+      "step": 1798
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987993594827769,
+      "loss": 0.9129,
+      "step": 1799
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019879719172598458,
+      "loss": 0.9514,
+      "step": 1800
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019879502202585446,
+      "loss": 1.1892,
+      "step": 1801
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019879285038242912,
+      "loss": 0.9981,
+      "step": 1802
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019879067679575134,
+      "loss": 1.1832,
+      "step": 1803
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987885012658639,
+      "loss": 1.0027,
+      "step": 1804
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987863237928095,
+      "loss": 0.823,
+      "step": 1805
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987841443766311,
+      "loss": 0.9415,
+      "step": 1806
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019878196301737142,
+      "loss": 1.1255,
+      "step": 1807
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987797797150735,
+      "loss": 1.2221,
+      "step": 1808
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987775944697802,
+      "loss": 1.0695,
+      "step": 1809
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019877540728153455,
+      "loss": 1.1859,
+      "step": 1810
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019877321815037956,
+      "loss": 0.9468,
+      "step": 1811
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019877102707635826,
+      "loss": 0.8774,
+      "step": 1812
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019876883405951377,
+      "loss": 1.0977,
+      "step": 1813
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019876663909988926,
+      "loss": 0.9691,
+      "step": 1814
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019876444219752787,
+      "loss": 1.0798,
+      "step": 1815
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987622433524728,
+      "loss": 0.9379,
+      "step": 1816
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019876004256476733,
+      "loss": 1.0277,
+      "step": 1817
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019875783983445473,
+      "loss": 1.2691,
+      "step": 1818
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019875563516157837,
+      "loss": 1.1509,
+      "step": 1819
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019875342854618153,
+      "loss": 1.3267,
+      "step": 1820
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987512199883077,
+      "loss": 1.3145,
+      "step": 1821
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019874900948800027,
+      "loss": 1.2042,
+      "step": 1822
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019874679704530278,
+      "loss": 1.1176,
+      "step": 1823
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019874458266025866,
+      "loss": 1.3184,
+      "step": 1824
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019874236633291156,
+      "loss": 0.8924,
+      "step": 1825
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019874014806330498,
+      "loss": 0.8903,
+      "step": 1826
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019873792785148262,
+      "loss": 1.0869,
+      "step": 1827
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019873570569748817,
+      "loss": 1.2095,
+      "step": 1828
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019873348160136524,
+      "loss": 1.0089,
+      "step": 1829
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987312555631577,
+      "loss": 1.128,
+      "step": 1830
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019872902758290926,
+      "loss": 1.0281,
+      "step": 1831
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019872679766066377,
+      "loss": 1.1305,
+      "step": 1832
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019872456579646505,
+      "loss": 1.2338,
+      "step": 1833
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019872233199035705,
+      "loss": 1.0999,
+      "step": 1834
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987200962423837,
+      "loss": 0.954,
+      "step": 1835
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000198717858552589,
+      "loss": 1.0464,
+      "step": 1836
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019871561892101692,
+      "loss": 0.8622,
+      "step": 1837
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987133773477115,
+      "loss": 1.1033,
+      "step": 1838
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987111338327169,
+      "loss": 1.1121,
+      "step": 1839
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001987088883760772,
+      "loss": 1.0422,
+      "step": 1840
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019870664097783655,
+      "loss": 0.9906,
+      "step": 1841
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019870439163803923,
+      "loss": 1.0709,
+      "step": 1842
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019870214035672942,
+      "loss": 1.1935,
+      "step": 1843
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986998871339514,
+      "loss": 1.1186,
+      "step": 1844
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019869763196974957,
+      "loss": 1.0569,
+      "step": 1845
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019869537486416822,
+      "loss": 1.1742,
+      "step": 1846
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019869311581725173,
+      "loss": 1.0567,
+      "step": 1847
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019869085482904458,
+      "loss": 0.9626,
+      "step": 1848
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019868859189959125,
+      "loss": 1.0569,
+      "step": 1849
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986863270289362,
+      "loss": 1.3231,
+      "step": 1850
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019868406021712404,
+      "loss": 1.0058,
+      "step": 1851
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019868179146419933,
+      "loss": 1.2699,
+      "step": 1852
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019867952077020666,
+      "loss": 1.0362,
+      "step": 1853
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019867724813519075,
+      "loss": 1.1666,
+      "step": 1854
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019867497355919626,
+      "loss": 0.9084,
+      "step": 1855
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019867269704226798,
+      "loss": 1.2844,
+      "step": 1856
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019867041858445065,
+      "loss": 1.1235,
+      "step": 1857
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986681381857891,
+      "loss": 1.2157,
+      "step": 1858
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986658558463282,
+      "loss": 1.059,
+      "step": 1859
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019866357156611284,
+      "loss": 1.0647,
+      "step": 1860
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986612853451879,
+      "loss": 1.0692,
+      "step": 1861
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986589971835984,
+      "loss": 1.0025,
+      "step": 1862
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986567070813894,
+      "loss": 1.1814,
+      "step": 1863
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019865441503860582,
+      "loss": 1.0371,
+      "step": 1864
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019865212105529283,
+      "loss": 1.0685,
+      "step": 1865
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019864982513149553,
+      "loss": 1.0909,
+      "step": 1866
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019864752726725908,
+      "loss": 0.9237,
+      "step": 1867
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019864522746262865,
+      "loss": 1.1004,
+      "step": 1868
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019864292571764955,
+      "loss": 1.0809,
+      "step": 1869
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019864062203236703,
+      "loss": 1.0088,
+      "step": 1870
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019863831640682636,
+      "loss": 1.0338,
+      "step": 1871
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986360088410729,
+      "loss": 1.1873,
+      "step": 1872
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019863369933515208,
+      "loss": 1.1099,
+      "step": 1873
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986313878891093,
+      "loss": 0.9715,
+      "step": 1874
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019862907450299004,
+      "loss": 1.0653,
+      "step": 1875
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019862675917683978,
+      "loss": 1.0449,
+      "step": 1876
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019862444191070408,
+      "loss": 1.1574,
+      "step": 1877
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019862212270462855,
+      "loss": 1.0446,
+      "step": 1878
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019861980155865875,
+      "loss": 1.0187,
+      "step": 1879
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019861747847284037,
+      "loss": 1.1864,
+      "step": 1880
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986151534472191,
+      "loss": 0.9349,
+      "step": 1881
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019861282648184068,
+      "loss": 0.8958,
+      "step": 1882
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019861049757675088,
+      "loss": 1.1287,
+      "step": 1883
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986081667319955,
+      "loss": 1.2318,
+      "step": 1884
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001986058339476204,
+      "loss": 0.9801,
+      "step": 1885
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019860349922367143,
+      "loss": 1.1774,
+      "step": 1886
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019860116256019457,
+      "loss": 0.9401,
+      "step": 1887
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019859882395723574,
+      "loss": 1.1101,
+      "step": 1888
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019859648341484097,
+      "loss": 1.1632,
+      "step": 1889
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019859414093305623,
+      "loss": 1.0494,
+      "step": 1890
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019859179651192773,
+      "loss": 0.8645,
+      "step": 1891
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019858945015150146,
+      "loss": 1.0589,
+      "step": 1892
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985871018518236,
+      "loss": 1.01,
+      "step": 1893
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019858475161294037,
+      "loss": 1.0877,
+      "step": 1894
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019858239943489802,
+      "loss": 0.9032,
+      "step": 1895
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019858004531774274,
+      "loss": 1.0114,
+      "step": 1896
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985776892615209,
+      "loss": 1.1078,
+      "step": 1897
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985753312662788,
+      "loss": 1.2297,
+      "step": 1898
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019857297133206285,
+      "loss": 1.1117,
+      "step": 1899
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019857060945891947,
+      "loss": 1.1018,
+      "step": 1900
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019856824564689515,
+      "loss": 1.2094,
+      "step": 1901
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019856587989603632,
+      "loss": 1.0067,
+      "step": 1902
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019856351220638954,
+      "loss": 1.0186,
+      "step": 1903
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019856114257800137,
+      "loss": 1.1971,
+      "step": 1904
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019855877101091846,
+      "loss": 1.2645,
+      "step": 1905
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019855639750518742,
+      "loss": 1.1142,
+      "step": 1906
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019855402206085497,
+      "loss": 0.9933,
+      "step": 1907
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985516446779678,
+      "loss": 1.0967,
+      "step": 1908
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019854926535657267,
+      "loss": 1.1182,
+      "step": 1909
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019854688409671646,
+      "loss": 0.9844,
+      "step": 1910
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985445008984459,
+      "loss": 0.8921,
+      "step": 1911
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019854211576180792,
+      "loss": 1.0728,
+      "step": 1912
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019853972868684943,
+      "loss": 1.0986,
+      "step": 1913
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019853733967361737,
+      "loss": 1.1106,
+      "step": 1914
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985349487221588,
+      "loss": 0.924,
+      "step": 1915
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019853255583252065,
+      "loss": 1.1899,
+      "step": 1916
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019853016100475007,
+      "loss": 1.2506,
+      "step": 1917
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985277642388941,
+      "loss": 1.041,
+      "step": 1918
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000198525365535,
+      "loss": 0.9784,
+      "step": 1919
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019852296489311478,
+      "loss": 1.0578,
+      "step": 1920
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019852056231328577,
+      "loss": 1.0444,
+      "step": 1921
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019851815779556021,
+      "loss": 0.934,
+      "step": 1922
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019851575133998542,
+      "loss": 1.1045,
+      "step": 1923
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019851334294660868,
+      "loss": 1.1505,
+      "step": 1924
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985109326154774,
+      "loss": 1.3403,
+      "step": 1925
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019850852034663898,
+      "loss": 0.9984,
+      "step": 1926
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019850610614014088,
+      "loss": 1.0577,
+      "step": 1927
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985036899960306,
+      "loss": 1.0344,
+      "step": 1928
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001985012719143556,
+      "loss": 1.0306,
+      "step": 1929
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019849885189516354,
+      "loss": 1.0013,
+      "step": 1930
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019849642993850192,
+      "loss": 1.2272,
+      "step": 1931
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019849400604441848,
+      "loss": 0.9713,
+      "step": 1932
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019849158021296081,
+      "loss": 1.0395,
+      "step": 1933
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984891524441767,
+      "loss": 0.8185,
+      "step": 1934
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019848672273811387,
+      "loss": 1.1686,
+      "step": 1935
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984842910948201,
+      "loss": 1.035,
+      "step": 1936
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984818575143432,
+      "loss": 1.2903,
+      "step": 1937
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019847942199673112,
+      "loss": 1.0884,
+      "step": 1938
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984769845420317,
+      "loss": 1.4385,
+      "step": 1939
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019847454515029288,
+      "loss": 1.2153,
+      "step": 1940
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019847210382156267,
+      "loss": 1.1371,
+      "step": 1941
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019846966055588912,
+      "loss": 1.0109,
+      "step": 1942
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019846721535332021,
+      "loss": 1.1352,
+      "step": 1943
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019846476821390411,
+      "loss": 0.8708,
+      "step": 1944
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984623191376889,
+      "loss": 0.9589,
+      "step": 1945
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984598681247228,
+      "loss": 1.0139,
+      "step": 1946
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.000198457415175054,
+      "loss": 1.0331,
+      "step": 1947
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019845496028873076,
+      "loss": 0.912,
+      "step": 1948
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019845250346580136,
+      "loss": 1.0983,
+      "step": 1949
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984500447063141,
+      "loss": 1.0241,
+      "step": 1950
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984475840103174,
+      "loss": 0.9787,
+      "step": 1951
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984451213778596,
+      "loss": 1.0254,
+      "step": 1952
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019844265680898918,
+      "loss": 1.0466,
+      "step": 1953
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984401903037546,
+      "loss": 1.3032,
+      "step": 1954
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984377218622044,
+      "loss": 1.0428,
+      "step": 1955
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019843525148438711,
+      "loss": 1.2397,
+      "step": 1956
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984327791703513,
+      "loss": 0.9571,
+      "step": 1957
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019843030492014565,
+      "loss": 1.0919,
+      "step": 1958
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0001984278287338188,
+      "loss": 1.0423,
+      "step": 1959
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019842535061141952,
+      "loss": 1.036,
+      "step": 1960
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019842287055299644,
+      "loss": 0.9761,
+      "step": 1961
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.00019842038855859842,
+      "loss": 1.1997,
+      "step": 1962
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019841790462827428,
+      "loss": 1.0433,
+      "step": 1963
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019841541876207283,
+      "loss": 1.0558,
+      "step": 1964
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019841293096004302,
+      "loss": 1.1226,
+      "step": 1965
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019841044122223376,
+      "loss": 1.0008,
+      "step": 1966
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019840794954869404,
+      "loss": 1.0352,
+      "step": 1967
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019840545593947283,
+      "loss": 1.1039,
+      "step": 1968
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019840296039461922,
+      "loss": 1.1229,
+      "step": 1969
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019840046291418227,
+      "loss": 1.1959,
+      "step": 1970
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019839796349821115,
+      "loss": 0.9444,
+      "step": 1971
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019839546214675497,
+      "loss": 1.0901,
+      "step": 1972
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019839295885986296,
+      "loss": 1.0764,
+      "step": 1973
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019839045363758437,
+      "loss": 1.0372,
+      "step": 1974
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019838794647996847,
+      "loss": 1.1593,
+      "step": 1975
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019838543738706453,
+      "loss": 1.2054,
+      "step": 1976
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019838292635892196,
+      "loss": 0.9478,
+      "step": 1977
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019838041339559012,
+      "loss": 1.0408,
+      "step": 1978
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019837789849711846,
+      "loss": 1.0112,
+      "step": 1979
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019837538166355648,
+      "loss": 0.8595,
+      "step": 1980
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019837286289495361,
+      "loss": 0.9292,
+      "step": 1981
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001983703421913594,
+      "loss": 1.0198,
+      "step": 1982
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019836781955282352,
+      "loss": 1.2069,
+      "step": 1983
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019836529497939553,
+      "loss": 1.1524,
+      "step": 1984
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019836276847112508,
+      "loss": 1.3069,
+      "step": 1985
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019836024002806187,
+      "loss": 1.0715,
+      "step": 1986
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019835770965025563,
+      "loss": 1.0063,
+      "step": 1987
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019835517733775615,
+      "loss": 0.9492,
+      "step": 1988
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019835264309061326,
+      "loss": 1.1269,
+      "step": 1989
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019835010690887677,
+      "loss": 1.0344,
+      "step": 1990
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001983475687925966,
+      "loss": 0.9196,
+      "step": 1991
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019834502874182261,
+      "loss": 0.9267,
+      "step": 1992
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019834248675660486,
+      "loss": 1.1602,
+      "step": 1993
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019833994283699328,
+      "loss": 1.0945,
+      "step": 1994
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001983373969830379,
+      "loss": 1.0923,
+      "step": 1995
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019833484919478886,
+      "loss": 1.135,
+      "step": 1996
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019833229947229623,
+      "loss": 1.0311,
+      "step": 1997
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019832974781561018,
+      "loss": 1.1006,
+      "step": 1998
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019832719422478088,
+      "loss": 0.9793,
+      "step": 1999
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019832463869985857,
+      "loss": 0.9289,
+      "step": 2000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019832208124089356,
+      "loss": 0.9596,
+      "step": 2001
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019831952184793606,
+      "loss": 1.0486,
+      "step": 2002
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001983169605210365,
+      "loss": 1.0556,
+      "step": 2003
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001983143972602452,
+      "loss": 1.1271,
+      "step": 2004
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019831183206561262,
+      "loss": 1.167,
+      "step": 2005
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019830926493718924,
+      "loss": 1.0497,
+      "step": 2006
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019830669587502547,
+      "loss": 1.2781,
+      "step": 2007
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019830412487917195,
+      "loss": 1.2034,
+      "step": 2008
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019830155194967917,
+      "loss": 1.067,
+      "step": 2009
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019829897708659777,
+      "loss": 1.1177,
+      "step": 2010
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019829640028997837,
+      "loss": 1.1669,
+      "step": 2011
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982938215598717,
+      "loss": 1.186,
+      "step": 2012
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019829124089632845,
+      "loss": 1.0867,
+      "step": 2013
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019828865829939942,
+      "loss": 1.0021,
+      "step": 2014
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019828607376913535,
+      "loss": 1.0716,
+      "step": 2015
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019828348730558713,
+      "loss": 1.1836,
+      "step": 2016
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019828089890880563,
+      "loss": 0.9279,
+      "step": 2017
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019827830857884173,
+      "loss": 1.3141,
+      "step": 2018
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982757163157464,
+      "loss": 0.9726,
+      "step": 2019
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019827312211957064,
+      "loss": 0.9906,
+      "step": 2020
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019827052599036545,
+      "loss": 1.0227,
+      "step": 2021
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019826792792818193,
+      "loss": 1.0794,
+      "step": 2022
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982653279330712,
+      "loss": 1.057,
+      "step": 2023
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982627260050843,
+      "loss": 1.0539,
+      "step": 2024
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019826012214427255,
+      "loss": 0.9416,
+      "step": 2025
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019825751635068706,
+      "loss": 1.005,
+      "step": 2026
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019825490862437914,
+      "loss": 1.0373,
+      "step": 2027
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019825229896540007,
+      "loss": 0.9685,
+      "step": 2028
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019824968737380117,
+      "loss": 1.1828,
+      "step": 2029
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019824707384963382,
+      "loss": 0.99,
+      "step": 2030
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019824445839294942,
+      "loss": 1.0481,
+      "step": 2031
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019824184100379947,
+      "loss": 0.9432,
+      "step": 2032
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019823922168223538,
+      "loss": 0.987,
+      "step": 2033
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019823660042830868,
+      "loss": 0.9007,
+      "step": 2034
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019823397724207098,
+      "loss": 1.0036,
+      "step": 2035
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982313521235738,
+      "loss": 1.0918,
+      "step": 2036
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982287250728689,
+      "loss": 1.1179,
+      "step": 2037
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019822609609000783,
+      "loss": 1.0986,
+      "step": 2038
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019822346517504237,
+      "loss": 0.8765,
+      "step": 2039
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019822083232802422,
+      "loss": 0.8671,
+      "step": 2040
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982181975490052,
+      "loss": 0.9974,
+      "step": 2041
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019821556083803716,
+      "loss": 1.0694,
+      "step": 2042
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019821292219517192,
+      "loss": 1.2145,
+      "step": 2043
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001982102816204614,
+      "loss": 1.314,
+      "step": 2044
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019820763911395758,
+      "loss": 1.1685,
+      "step": 2045
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019820499467571235,
+      "loss": 0.8908,
+      "step": 2046
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019820234830577777,
+      "loss": 0.9681,
+      "step": 2047
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019819970000420592,
+      "loss": 0.9904,
+      "step": 2048
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019819704977104888,
+      "loss": 1.0413,
+      "step": 2049
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019819439760635876,
+      "loss": 1.1813,
+      "step": 2050
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019819174351018776,
+      "loss": 1.1649,
+      "step": 2051
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019818908748258802,
+      "loss": 1.0616,
+      "step": 2052
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019818642952361187,
+      "loss": 1.1719,
+      "step": 2053
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981837696333115,
+      "loss": 0.9884,
+      "step": 2054
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019818110781173932,
+      "loss": 1.1105,
+      "step": 2055
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019817844405894763,
+      "loss": 1.1992,
+      "step": 2056
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019817577837498887,
+      "loss": 1.1159,
+      "step": 2057
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019817311075991543,
+      "loss": 1.0633,
+      "step": 2058
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981704412137798,
+      "loss": 1.0335,
+      "step": 2059
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981677697366345,
+      "loss": 1.1115,
+      "step": 2060
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019816509632853208,
+      "loss": 1.2951,
+      "step": 2061
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019816242098952508,
+      "loss": 0.9623,
+      "step": 2062
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019815974371966614,
+      "loss": 0.8879,
+      "step": 2063
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019815706451900797,
+      "loss": 1.1442,
+      "step": 2064
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019815438338760327,
+      "loss": 1.0096,
+      "step": 2065
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981517003255047,
+      "loss": 1.1603,
+      "step": 2066
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019814901533276512,
+      "loss": 1.1586,
+      "step": 2067
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019814632840943727,
+      "loss": 1.0081,
+      "step": 2068
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019814363955557404,
+      "loss": 1.1375,
+      "step": 2069
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019814094877122833,
+      "loss": 1.0986,
+      "step": 2070
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019813825605645305,
+      "loss": 1.0544,
+      "step": 2071
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019813556141130118,
+      "loss": 1.0309,
+      "step": 2072
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019813286483582574,
+      "loss": 1.049,
+      "step": 2073
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981301663300797,
+      "loss": 1.0513,
+      "step": 2074
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981274658941162,
+      "loss": 0.9532,
+      "step": 2075
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019812476352798835,
+      "loss": 1.0912,
+      "step": 2076
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981220592317493,
+      "loss": 1.2094,
+      "step": 2077
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019811935300545223,
+      "loss": 0.9742,
+      "step": 2078
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981166448491504,
+      "loss": 0.8724,
+      "step": 2079
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019811393476289706,
+      "loss": 1.0442,
+      "step": 2080
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019811122274674553,
+      "loss": 0.999,
+      "step": 2081
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019810850880074913,
+      "loss": 1.2401,
+      "step": 2082
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019810579292496122,
+      "loss": 1.0254,
+      "step": 2083
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019810307511943531,
+      "loss": 1.1612,
+      "step": 2084
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001981003553842248,
+      "loss": 1.068,
+      "step": 2085
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980976337193832,
+      "loss": 1.0909,
+      "step": 2086
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019809491012496403,
+      "loss": 0.9823,
+      "step": 2087
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019809218460102088,
+      "loss": 1.1167,
+      "step": 2088
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019808945714760737,
+      "loss": 1.161,
+      "step": 2089
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980867277647771,
+      "loss": 1.1879,
+      "step": 2090
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019808399645258382,
+      "loss": 1.0423,
+      "step": 2091
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980812632110812,
+      "loss": 1.17,
+      "step": 2092
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019807852804032305,
+      "loss": 1.1661,
+      "step": 2093
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019807579094036315,
+      "loss": 1.0796,
+      "step": 2094
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019807305191125532,
+      "loss": 1.2859,
+      "step": 2095
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019807031095305346,
+      "loss": 1.1839,
+      "step": 2096
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980675680658115,
+      "loss": 1.0398,
+      "step": 2097
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019806482324958335,
+      "loss": 1.0329,
+      "step": 2098
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019806207650442304,
+      "loss": 1.04,
+      "step": 2099
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019805932783038458,
+      "loss": 1.2167,
+      "step": 2100
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019805657722752202,
+      "loss": 1.2142,
+      "step": 2101
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980538246958895,
+      "loss": 1.2266,
+      "step": 2102
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980510702355411,
+      "loss": 1.0618,
+      "step": 2103
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019804831384653108,
+      "loss": 1.1327,
+      "step": 2104
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980455555289136,
+      "loss": 1.0662,
+      "step": 2105
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019804279528274295,
+      "loss": 1.0778,
+      "step": 2106
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019804003310807343,
+      "loss": 1.2449,
+      "step": 2107
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019803726900495937,
+      "loss": 1.4879,
+      "step": 2108
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019803450297345508,
+      "loss": 1.1791,
+      "step": 2109
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019803173501361504,
+      "loss": 0.9464,
+      "step": 2110
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019802896512549362,
+      "loss": 0.9106,
+      "step": 2111
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019802619330914537,
+      "loss": 1.3795,
+      "step": 2112
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980234195646248,
+      "loss": 1.0615,
+      "step": 2113
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980206438919865,
+      "loss": 1.1862,
+      "step": 2114
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019801786629128495,
+      "loss": 1.1073,
+      "step": 2115
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019801508676257492,
+      "loss": 0.9646,
+      "step": 2116
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019801230530591105,
+      "loss": 1.1123,
+      "step": 2117
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019800952192134796,
+      "loss": 1.1397,
+      "step": 2118
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001980067366089405,
+      "loss": 0.9754,
+      "step": 2119
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019800394936874343,
+      "loss": 0.9779,
+      "step": 2120
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019800116020081158,
+      "loss": 0.9338,
+      "step": 2121
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019799836910519978,
+      "loss": 1.0509,
+      "step": 2122
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019799557608196294,
+      "loss": 1.0384,
+      "step": 2123
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019799278113115604,
+      "loss": 1.1852,
+      "step": 2124
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019798998425283403,
+      "loss": 1.0788,
+      "step": 2125
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001979871854470519,
+      "loss": 1.1559,
+      "step": 2126
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019798438471386475,
+      "loss": 1.0544,
+      "step": 2127
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019798158205332764,
+      "loss": 1.3031,
+      "step": 2128
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001979787774654957,
+      "loss": 1.0745,
+      "step": 2129
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001979759709504241,
+      "loss": 0.9759,
+      "step": 2130
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019797316250816805,
+      "loss": 1.1145,
+      "step": 2131
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019797035213878276,
+      "loss": 1.0584,
+      "step": 2132
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019796753984232358,
+      "loss": 0.954,
+      "step": 2133
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019796472561884576,
+      "loss": 0.9828,
+      "step": 2134
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019796190946840464,
+      "loss": 1.0783,
+      "step": 2135
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001979590913910557,
+      "loss": 0.9864,
+      "step": 2136
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019795627138685434,
+      "loss": 1.2552,
+      "step": 2137
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019795344945585598,
+      "loss": 1.1633,
+      "step": 2138
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019795062559811618,
+      "loss": 1.0998,
+      "step": 2139
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019794779981369045,
+      "loss": 1.0464,
+      "step": 2140
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019794497210263442,
+      "loss": 1.0049,
+      "step": 2141
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019794214246500363,
+      "loss": 1.2151,
+      "step": 2142
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001979393109008538,
+      "loss": 0.9441,
+      "step": 2143
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019793647741024066,
+      "loss": 1.0328,
+      "step": 2144
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001979336419932199,
+      "loss": 1.1457,
+      "step": 2145
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019793080464984727,
+      "loss": 1.0711,
+      "step": 2146
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001979279653801786,
+      "loss": 0.8554,
+      "step": 2147
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019792512418426975,
+      "loss": 1.0007,
+      "step": 2148
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019792228106217658,
+      "loss": 0.8662,
+      "step": 2149
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019791943601395504,
+      "loss": 0.9922,
+      "step": 2150
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019791658903966112,
+      "loss": 1.0249,
+      "step": 2151
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019791374013935077,
+      "loss": 1.0047,
+      "step": 2152
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019791088931308003,
+      "loss": 1.0201,
+      "step": 2153
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019790803656090494,
+      "loss": 1.1499,
+      "step": 2154
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019790518188288173,
+      "loss": 1.0282,
+      "step": 2155
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019790232527906648,
+      "loss": 1.0135,
+      "step": 2156
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019789946674951536,
+      "loss": 0.8697,
+      "step": 2157
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019789660629428464,
+      "loss": 0.952,
+      "step": 2158
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019789374391343051,
+      "loss": 0.9079,
+      "step": 2159
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978908796070094,
+      "loss": 1.0718,
+      "step": 2160
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019788801337507753,
+      "loss": 1.1452,
+      "step": 2161
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019788514521769135,
+      "loss": 1.0608,
+      "step": 2162
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019788227513490723,
+      "loss": 1.1422,
+      "step": 2163
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978794031267817,
+      "loss": 0.9809,
+      "step": 2164
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019787652919337116,
+      "loss": 1.2145,
+      "step": 2165
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019787365333473217,
+      "loss": 0.9033,
+      "step": 2166
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019787077555092134,
+      "loss": 1.1081,
+      "step": 2167
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019786789584199524,
+      "loss": 1.1487,
+      "step": 2168
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019786501420801048,
+      "loss": 1.1738,
+      "step": 2169
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978621306490238,
+      "loss": 0.9728,
+      "step": 2170
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019785924516509194,
+      "loss": 1.0841,
+      "step": 2171
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019785635775627157,
+      "loss": 1.1417,
+      "step": 2172
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019785346842261957,
+      "loss": 1.0405,
+      "step": 2173
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978505771641927,
+      "loss": 0.8563,
+      "step": 2174
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019784768398104787,
+      "loss": 1.034,
+      "step": 2175
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019784478887324198,
+      "loss": 1.0274,
+      "step": 2176
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197841891840832,
+      "loss": 1.3921,
+      "step": 2177
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978389928838749,
+      "loss": 1.0447,
+      "step": 2178
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978360920024277,
+      "loss": 0.9808,
+      "step": 2179
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019783318919654744,
+      "loss": 0.9535,
+      "step": 2180
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019783028446629123,
+      "loss": 0.9022,
+      "step": 2181
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019782737781171624,
+      "loss": 1.0109,
+      "step": 2182
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019782446923287964,
+      "loss": 1.2634,
+      "step": 2183
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019782155872983856,
+      "loss": 1.0055,
+      "step": 2184
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019781864630265034,
+      "loss": 0.9284,
+      "step": 2185
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978157319513722,
+      "loss": 0.9476,
+      "step": 2186
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019781281567606156,
+      "loss": 1.3249,
+      "step": 2187
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.0001978098974767757,
+      "loss": 0.9886,
+      "step": 2188
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019780697735357204,
+      "loss": 1.2273,
+      "step": 2189
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.000197804055306508,
+      "loss": 1.0258,
+      "step": 2190
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019780113133564112,
+      "loss": 1.1839,
+      "step": 2191
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019779820544102886,
+      "loss": 1.0924,
+      "step": 2192
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 0.00019779527762272877,
+      "loss": 1.3315,
+      "step": 2193
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977923478807985,
+      "loss": 1.174,
+      "step": 2194
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977894162152956,
+      "loss": 1.3412,
+      "step": 2195
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019778648262627776,
+      "loss": 1.0035,
+      "step": 2196
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977835471138027,
+      "loss": 0.9509,
+      "step": 2197
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019778060967792817,
+      "loss": 1.0256,
+      "step": 2198
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019777767031871191,
+      "loss": 1.0289,
+      "step": 2199
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977747290362118,
+      "loss": 1.1402,
+      "step": 2200
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019777178583048563,
+      "loss": 1.0204,
+      "step": 2201
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019776884070159134,
+      "loss": 1.0051,
+      "step": 2202
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977658936495868,
+      "loss": 0.9597,
+      "step": 2203
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019776294467453007,
+      "loss": 1.0932,
+      "step": 2204
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977599937764791,
+      "loss": 1.0451,
+      "step": 2205
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977570409554919,
+      "loss": 0.9425,
+      "step": 2206
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019775408621162661,
+      "loss": 1.5815,
+      "step": 2207
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019775112954494131,
+      "loss": 1.0188,
+      "step": 2208
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977481709554942,
+      "loss": 0.9303,
+      "step": 2209
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019774521044334345,
+      "loss": 0.9897,
+      "step": 2210
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977422480085473,
+      "loss": 1.1759,
+      "step": 2211
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019773928365116402,
+      "loss": 0.9882,
+      "step": 2212
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019773631737125192,
+      "loss": 1.2119,
+      "step": 2213
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019773334916886934,
+      "loss": 1.1507,
+      "step": 2214
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019773037904407468,
+      "loss": 0.9317,
+      "step": 2215
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019772740699692634,
+      "loss": 1.0853,
+      "step": 2216
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019772443302748277,
+      "loss": 1.1214,
+      "step": 2217
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977214571358025,
+      "loss": 0.9724,
+      "step": 2218
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019771847932194405,
+      "loss": 0.983,
+      "step": 2219
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000197715499585966,
+      "loss": 1.134,
+      "step": 2220
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019771251792792698,
+      "loss": 1.1707,
+      "step": 2221
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001977095343478856,
+      "loss": 1.2299,
+      "step": 2222
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019770654884590055,
+      "loss": 1.037,
+      "step": 2223
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019770356142203056,
+      "loss": 1.2093,
+      "step": 2224
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019770057207633445,
+      "loss": 1.1625,
+      "step": 2225
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019769758080887093,
+      "loss": 1.063,
+      "step": 2226
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976945876196989,
+      "loss": 1.1327,
+      "step": 2227
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976915925088772,
+      "loss": 0.979,
+      "step": 2228
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019768859547646478,
+      "loss": 0.936,
+      "step": 2229
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019768559652252058,
+      "loss": 1.083,
+      "step": 2230
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019768259564710355,
+      "loss": 0.9632,
+      "step": 2231
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019767959285027276,
+      "loss": 0.9585,
+      "step": 2232
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019767658813208726,
+      "loss": 0.9428,
+      "step": 2233
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019767358149260614,
+      "loss": 1.0735,
+      "step": 2234
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976705729318886,
+      "loss": 1.197,
+      "step": 2235
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019766756244999373,
+      "loss": 0.8308,
+      "step": 2236
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019766455004698082,
+      "loss": 0.964,
+      "step": 2237
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019766153572290906,
+      "loss": 0.8757,
+      "step": 2238
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976585194778378,
+      "loss": 1.0986,
+      "step": 2239
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019765550131182636,
+      "loss": 1.1206,
+      "step": 2240
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019765248122493407,
+      "loss": 1.1456,
+      "step": 2241
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019764945921722037,
+      "loss": 1.0025,
+      "step": 2242
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976464352887447,
+      "loss": 1.0213,
+      "step": 2243
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019764340943956648,
+      "loss": 1.1029,
+      "step": 2244
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019764038166974535,
+      "loss": 1.193,
+      "step": 2245
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019763735197934077,
+      "loss": 1.1322,
+      "step": 2246
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019763432036841235,
+      "loss": 1.1788,
+      "step": 2247
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019763128683701972,
+      "loss": 1.0807,
+      "step": 2248
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019762825138522257,
+      "loss": 1.243,
+      "step": 2249
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019762521401308058,
+      "loss": 0.9817,
+      "step": 2250
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019762217472065355,
+      "loss": 1.1965,
+      "step": 2251
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976191335080012,
+      "loss": 0.939,
+      "step": 2252
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976160903751834,
+      "loss": 1.257,
+      "step": 2253
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019761304532225994,
+      "loss": 1.0701,
+      "step": 2254
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001976099983492908,
+      "loss": 1.3319,
+      "step": 2255
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019760694945633586,
+      "loss": 0.9951,
+      "step": 2256
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019760389864345514,
+      "loss": 1.1098,
+      "step": 2257
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019760084591070856,
+      "loss": 1.059,
+      "step": 2258
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019759779125815629,
+      "loss": 1.0577,
+      "step": 2259
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019759473468585829,
+      "loss": 1.0458,
+      "step": 2260
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019759167619387476,
+      "loss": 1.0467,
+      "step": 2261
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019758861578226581,
+      "loss": 1.1784,
+      "step": 2262
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019758555345109172,
+      "loss": 1.0449,
+      "step": 2263
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019758248920041265,
+      "loss": 1.1874,
+      "step": 2264
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019757942303028889,
+      "loss": 1.1002,
+      "step": 2265
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019757635494078077,
+      "loss": 1.1074,
+      "step": 2266
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019757328493194863,
+      "loss": 0.8134,
+      "step": 2267
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019757021300385286,
+      "loss": 1.0343,
+      "step": 2268
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001975671391565539,
+      "loss": 0.9047,
+      "step": 2269
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019756406339011215,
+      "loss": 1.1189,
+      "step": 2270
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019756098570458818,
+      "loss": 0.9135,
+      "step": 2271
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001975579061000425,
+      "loss": 1.1129,
+      "step": 2272
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001975548245765357,
+      "loss": 1.0215,
+      "step": 2273
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019755174113412834,
+      "loss": 0.9768,
+      "step": 2274
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019754865577288112,
+      "loss": 0.8197,
+      "step": 2275
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019754556849285474,
+      "loss": 1.0102,
+      "step": 2276
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019754247929410995,
+      "loss": 1.0386,
+      "step": 2277
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001975393881767074,
+      "loss": 1.0207,
+      "step": 2278
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000197536295140708,
+      "loss": 1.0389,
+      "step": 2279
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019753320018617255,
+      "loss": 1.3199,
+      "step": 2280
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019753010331316194,
+      "loss": 1.0943,
+      "step": 2281
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019752700452173704,
+      "loss": 0.8928,
+      "step": 2282
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019752390381195888,
+      "loss": 0.9187,
+      "step": 2283
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001975208011838884,
+      "loss": 1.0208,
+      "step": 2284
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019751769663758665,
+      "loss": 1.0045,
+      "step": 2285
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019751459017311468,
+      "loss": 0.9849,
+      "step": 2286
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001975114817905336,
+      "loss": 0.9834,
+      "step": 2287
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019750837148990453,
+      "loss": 0.99,
+      "step": 2288
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001975052592712887,
+      "loss": 1.0236,
+      "step": 2289
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019750214513474727,
+      "loss": 1.0816,
+      "step": 2290
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019749902908034156,
+      "loss": 1.0565,
+      "step": 2291
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019749591110813282,
+      "loss": 1.1727,
+      "step": 2292
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019749279121818235,
+      "loss": 0.8809,
+      "step": 2293
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019748966941055161,
+      "loss": 1.1237,
+      "step": 2294
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001974865456853019,
+      "loss": 1.0218,
+      "step": 2295
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019748342004249477,
+      "loss": 1.184,
+      "step": 2296
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001974802924821916,
+      "loss": 1.074,
+      "step": 2297
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019747716300445398,
+      "loss": 1.269,
+      "step": 2298
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019747403160934343,
+      "loss": 1.055,
+      "step": 2299
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019747089829692157,
+      "loss": 0.836,
+      "step": 2300
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019746776306724998,
+      "loss": 1.1213,
+      "step": 2301
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019746462592039042,
+      "loss": 1.054,
+      "step": 2302
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019746148685640451,
+      "loss": 1.2323,
+      "step": 2303
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019745834587535407,
+      "loss": 1.1774,
+      "step": 2304
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001974552029773008,
+      "loss": 1.1058,
+      "step": 2305
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019745205816230657,
+      "loss": 0.7023,
+      "step": 2306
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019744891143043324,
+      "loss": 1.0154,
+      "step": 2307
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019744576278174272,
+      "loss": 1.0628,
+      "step": 2308
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001974426122162969,
+      "loss": 1.2649,
+      "step": 2309
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019743945973415777,
+      "loss": 0.9417,
+      "step": 2310
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019743630533538737,
+      "loss": 1.0408,
+      "step": 2311
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001974331490200477,
+      "loss": 0.911,
+      "step": 2312
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019742999078820088,
+      "loss": 0.9335,
+      "step": 2313
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019742683063990902,
+      "loss": 1.2093,
+      "step": 2314
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019742366857523427,
+      "loss": 1.0096,
+      "step": 2315
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019742050459423883,
+      "loss": 1.1196,
+      "step": 2316
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019741733869698495,
+      "loss": 0.8507,
+      "step": 2317
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001974141708835349,
+      "loss": 1.1292,
+      "step": 2318
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019741100115395094,
+      "loss": 0.9387,
+      "step": 2319
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001974078295082955,
+      "loss": 1.226,
+      "step": 2320
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019740465594663092,
+      "loss": 1.0837,
+      "step": 2321
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019740148046901968,
+      "loss": 1.0555,
+      "step": 2322
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019739830307552417,
+      "loss": 1.2529,
+      "step": 2323
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019739512376620687,
+      "loss": 1.1218,
+      "step": 2324
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973919425411304,
+      "loss": 1.0291,
+      "step": 2325
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973887594003573,
+      "loss": 1.1634,
+      "step": 2326
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019738557434395015,
+      "loss": 1.2997,
+      "step": 2327
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019738238737197165,
+      "loss": 0.9859,
+      "step": 2328
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019737919848448449,
+      "loss": 0.928,
+      "step": 2329
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973760076815513,
+      "loss": 1.2222,
+      "step": 2330
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019737281496323496,
+      "loss": 1.0313,
+      "step": 2331
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019736962032959824,
+      "loss": 1.1963,
+      "step": 2332
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019736642378070392,
+      "loss": 1.1347,
+      "step": 2333
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019736322531661496,
+      "loss": 1.0104,
+      "step": 2334
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973600249373942,
+      "loss": 0.9605,
+      "step": 2335
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019735682264310464,
+      "loss": 1.0045,
+      "step": 2336
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019735361843380926,
+      "loss": 0.9742,
+      "step": 2337
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973504123095711,
+      "loss": 1.0305,
+      "step": 2338
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019734720427045315,
+      "loss": 0.9967,
+      "step": 2339
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973439943165186,
+      "loss": 0.9954,
+      "step": 2340
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019734078244783054,
+      "loss": 0.9802,
+      "step": 2341
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019733756866445215,
+      "loss": 0.9729,
+      "step": 2342
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973343529664467,
+      "loss": 1.0429,
+      "step": 2343
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019733113535387738,
+      "loss": 0.9793,
+      "step": 2344
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973279158268075,
+      "loss": 0.9078,
+      "step": 2345
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973246943853004,
+      "loss": 1.2503,
+      "step": 2346
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973214710294194,
+      "loss": 0.9542,
+      "step": 2347
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000197318245759228,
+      "loss": 1.2059,
+      "step": 2348
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019731501857478954,
+      "loss": 0.9737,
+      "step": 2349
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019731178947616753,
+      "loss": 1.0197,
+      "step": 2350
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001973085584634255,
+      "loss": 0.9636,
+      "step": 2351
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019730532553662703,
+      "loss": 1.2567,
+      "step": 2352
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019730209069583566,
+      "loss": 1.094,
+      "step": 2353
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019729885394111503,
+      "loss": 1.2116,
+      "step": 2354
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019729561527252884,
+      "loss": 1.1388,
+      "step": 2355
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972923746901407,
+      "loss": 1.0358,
+      "step": 2356
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972891321940145,
+      "loss": 0.9798,
+      "step": 2357
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972858877842139,
+      "loss": 0.9571,
+      "step": 2358
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972826414608028,
+      "loss": 1.1291,
+      "step": 2359
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000197279393223845,
+      "loss": 1.1813,
+      "step": 2360
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972761430734044,
+      "loss": 1.0485,
+      "step": 2361
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019727289100954492,
+      "loss": 0.9967,
+      "step": 2362
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019726963703233056,
+      "loss": 1.0842,
+      "step": 2363
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972663811418253,
+      "loss": 1.165,
+      "step": 2364
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019726312333809322,
+      "loss": 1.0238,
+      "step": 2365
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019725986362119838,
+      "loss": 1.2224,
+      "step": 2366
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019725660199120486,
+      "loss": 0.9206,
+      "step": 2367
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019725333844817686,
+      "loss": 1.2235,
+      "step": 2368
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972500729921786,
+      "loss": 1.1087,
+      "step": 2369
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019724680562327422,
+      "loss": 1.0814,
+      "step": 2370
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972435363415281,
+      "loss": 0.8649,
+      "step": 2371
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019724026514700445,
+      "loss": 0.9873,
+      "step": 2372
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019723699203976766,
+      "loss": 1.2097,
+      "step": 2373
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019723371701988214,
+      "loss": 0.9787,
+      "step": 2374
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019723044008741224,
+      "loss": 0.8791,
+      "step": 2375
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019722716124242248,
+      "loss": 1.1653,
+      "step": 2376
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972238804849773,
+      "loss": 1.0653,
+      "step": 2377
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019722059781514132,
+      "loss": 1.2423,
+      "step": 2378
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000197217313232979,
+      "loss": 1.1817,
+      "step": 2379
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019721402673855497,
+      "loss": 1.0172,
+      "step": 2380
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019721073833193397,
+      "loss": 1.0038,
+      "step": 2381
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019720744801318062,
+      "loss": 0.997,
+      "step": 2382
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001972041557823596,
+      "loss": 1.2454,
+      "step": 2383
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019720086163953568,
+      "loss": 1.0727,
+      "step": 2384
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019719756558477373,
+      "loss": 1.009,
+      "step": 2385
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019719426761813852,
+      "loss": 1.0034,
+      "step": 2386
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019719096773969494,
+      "loss": 0.9744,
+      "step": 2387
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001971876659495079,
+      "loss": 0.9665,
+      "step": 2388
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019718436224764233,
+      "loss": 1.033,
+      "step": 2389
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019718105663416323,
+      "loss": 1.0669,
+      "step": 2390
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001971777491091356,
+      "loss": 0.9334,
+      "step": 2391
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019717443967262455,
+      "loss": 0.9934,
+      "step": 2392
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001971711283246951,
+      "loss": 0.9902,
+      "step": 2393
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019716781506541245,
+      "loss": 1.2015,
+      "step": 2394
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019716449989484175,
+      "loss": 0.9153,
+      "step": 2395
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019716118281304818,
+      "loss": 0.8022,
+      "step": 2396
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019715786382009705,
+      "loss": 1.0205,
+      "step": 2397
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019715454291605356,
+      "loss": 1.0515,
+      "step": 2398
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019715122010098313,
+      "loss": 1.102,
+      "step": 2399
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019714789537495102,
+      "loss": 1.0059,
+      "step": 2400
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019714456873802272,
+      "loss": 1.0979,
+      "step": 2401
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019714124019026358,
+      "loss": 0.9497,
+      "step": 2402
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019713790973173914,
+      "loss": 0.9872,
+      "step": 2403
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019713457736251488,
+      "loss": 0.9468,
+      "step": 2404
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019713124308265635,
+      "loss": 1.0681,
+      "step": 2405
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019712790689222913,
+      "loss": 1.031,
+      "step": 2406
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019712456879129886,
+      "loss": 1.0019,
+      "step": 2407
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001971212287799312,
+      "loss": 1.0396,
+      "step": 2408
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001971178868581918,
+      "loss": 1.1993,
+      "step": 2409
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019711454302614648,
+      "loss": 1.0197,
+      "step": 2410
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019711119728386094,
+      "loss": 1.0132,
+      "step": 2411
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000197107849631401,
+      "loss": 1.1048,
+      "step": 2412
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019710450006883256,
+      "loss": 1.0249,
+      "step": 2413
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019710114859622144,
+      "loss": 1.0693,
+      "step": 2414
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001970977952136336,
+      "loss": 1.0299,
+      "step": 2415
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.000197094439921135,
+      "loss": 1.0908,
+      "step": 2416
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019709108271879163,
+      "loss": 1.212,
+      "step": 2417
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019708772360666957,
+      "loss": 1.0705,
+      "step": 2418
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019708436258483484,
+      "loss": 0.9453,
+      "step": 2419
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001970809996533535,
+      "loss": 1.1853,
+      "step": 2420
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019707763481229183,
+      "loss": 1.0792,
+      "step": 2421
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.0001970742680617159,
+      "loss": 1.0521,
+      "step": 2422
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019707089940169202,
+      "loss": 1.1159,
+      "step": 2423
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 0.00019706752883228644,
+      "loss": 1.0685,
+      "step": 2424
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001970641563535654,
+      "loss": 1.0608,
+      "step": 2425
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019706078196559526,
+      "loss": 0.9565,
+      "step": 2426
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001970574056684424,
+      "loss": 1.0377,
+      "step": 2427
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019705402746217327,
+      "loss": 1.1179,
+      "step": 2428
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019705064734685425,
+      "loss": 1.0377,
+      "step": 2429
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019704726532255188,
+      "loss": 0.954,
+      "step": 2430
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019704388138933268,
+      "loss": 0.9896,
+      "step": 2431
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019704049554726316,
+      "loss": 0.8132,
+      "step": 2432
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019703710779640999,
+      "loss": 1.194,
+      "step": 2433
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019703371813683978,
+      "loss": 1.1793,
+      "step": 2434
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019703032656861918,
+      "loss": 1.032,
+      "step": 2435
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019702693309181494,
+      "loss": 1.0092,
+      "step": 2436
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019702353770649378,
+      "loss": 1.1086,
+      "step": 2437
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019702014041272247,
+      "loss": 1.142,
+      "step": 2438
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001970167412105679,
+      "loss": 0.9467,
+      "step": 2439
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019701334010009688,
+      "loss": 1.2065,
+      "step": 2440
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001970099370813763,
+      "loss": 1.0731,
+      "step": 2441
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001970065321544732,
+      "loss": 1.0428,
+      "step": 2442
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019700312531945442,
+      "loss": 1.0464,
+      "step": 2443
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019699971657638702,
+      "loss": 1.0922,
+      "step": 2444
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019699630592533808,
+      "loss": 1.0502,
+      "step": 2445
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019699289336637467,
+      "loss": 1.241,
+      "step": 2446
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969894788995639,
+      "loss": 0.9801,
+      "step": 2447
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019698606252497295,
+      "loss": 0.7503,
+      "step": 2448
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019698264424266898,
+      "loss": 0.8994,
+      "step": 2449
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019697922405271932,
+      "loss": 1.27,
+      "step": 2450
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019697580195519114,
+      "loss": 1.2081,
+      "step": 2451
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019697237795015182,
+      "loss": 1.0972,
+      "step": 2452
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969689520376687,
+      "loss": 1.1647,
+      "step": 2453
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019696552421780914,
+      "loss": 1.0314,
+      "step": 2454
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969620944906406,
+      "loss": 0.9797,
+      "step": 2455
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969586628562305,
+      "loss": 1.1002,
+      "step": 2456
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019695522931464636,
+      "loss": 1.0469,
+      "step": 2457
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019695179386595575,
+      "loss": 1.1034,
+      "step": 2458
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969483565102262,
+      "loss": 1.354,
+      "step": 2459
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019694491724752533,
+      "loss": 0.9879,
+      "step": 2460
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019694147607792083,
+      "loss": 1.1096,
+      "step": 2461
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019693803300148034,
+      "loss": 1.0511,
+      "step": 2462
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969345880182716,
+      "loss": 1.1379,
+      "step": 2463
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019693114112836238,
+      "loss": 1.1002,
+      "step": 2464
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019692769233182046,
+      "loss": 0.9284,
+      "step": 2465
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019692424162871371,
+      "loss": 1.0833,
+      "step": 2466
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019692078901911,
+      "loss": 1.3046,
+      "step": 2467
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969173345030772,
+      "loss": 1.0511,
+      "step": 2468
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969138780806833,
+      "loss": 1.0045,
+      "step": 2469
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001969104197519963,
+      "loss": 0.9777,
+      "step": 2470
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019690695951708422,
+      "loss": 0.9322,
+      "step": 2471
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019690349737601508,
+      "loss": 1.2099,
+      "step": 2472
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000196900033328857,
+      "loss": 1.2142,
+      "step": 2473
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019689656737567813,
+      "loss": 1.2729,
+      "step": 2474
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019689309951654663,
+      "loss": 0.9006,
+      "step": 2475
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019688962975153077,
+      "loss": 1.1023,
+      "step": 2476
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019688615808069872,
+      "loss": 1.0558,
+      "step": 2477
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968826845041188,
+      "loss": 1.247,
+      "step": 2478
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019687920902185937,
+      "loss": 1.1423,
+      "step": 2479
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019687573163398872,
+      "loss": 1.1421,
+      "step": 2480
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019687225234057528,
+      "loss": 0.9068,
+      "step": 2481
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019686877114168753,
+      "loss": 1.0595,
+      "step": 2482
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968652880373939,
+      "loss": 1.0418,
+      "step": 2483
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968618030277629,
+      "loss": 1.2139,
+      "step": 2484
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968583161128631,
+      "loss": 1.0844,
+      "step": 2485
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968548272927631,
+      "loss": 1.0861,
+      "step": 2486
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019685133656753152,
+      "loss": 1.1114,
+      "step": 2487
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019684784393723698,
+      "loss": 1.1491,
+      "step": 2488
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968443494019482,
+      "loss": 1.0823,
+      "step": 2489
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019684085296173395,
+      "loss": 0.968,
+      "step": 2490
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019683735461666297,
+      "loss": 1.1498,
+      "step": 2491
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968338543668041,
+      "loss": 1.1358,
+      "step": 2492
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019683035221222618,
+      "loss": 1.0278,
+      "step": 2493
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019682684815299805,
+      "loss": 1.3081,
+      "step": 2494
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001968233421891887,
+      "loss": 0.9908,
+      "step": 2495
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019681983432086705,
+      "loss": 1.0425,
+      "step": 2496
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019681632454810215,
+      "loss": 0.9594,
+      "step": 2497
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000196812812870963,
+      "loss": 1.019,
+      "step": 2498
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019680929928951864,
+      "loss": 0.9191,
+      "step": 2499
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019680578380383828,
+      "loss": 1.1027,
+      "step": 2500
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019680226641399096,
+      "loss": 0.9566,
+      "step": 2501
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019679874712004593,
+      "loss": 1.0,
+      "step": 2502
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001967952259220724,
+      "loss": 1.1746,
+      "step": 2503
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019679170282013969,
+      "loss": 0.9972,
+      "step": 2504
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019678817781431697,
+      "loss": 1.0783,
+      "step": 2505
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019678465090467367,
+      "loss": 1.1109,
+      "step": 2506
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019678112209127916,
+      "loss": 1.045,
+      "step": 2507
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019677759137420284,
+      "loss": 0.9914,
+      "step": 2508
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019677405875351414,
+      "loss": 1.0786,
+      "step": 2509
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019677052422928254,
+      "loss": 1.009,
+      "step": 2510
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019676698780157764,
+      "loss": 1.1322,
+      "step": 2511
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019676344947046893,
+      "loss": 1.1005,
+      "step": 2512
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019675990923602598,
+      "loss": 1.2307,
+      "step": 2513
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019675636709831854,
+      "loss": 1.081,
+      "step": 2514
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019675282305741616,
+      "loss": 1.1752,
+      "step": 2515
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019674927711338863,
+      "loss": 0.9739,
+      "step": 2516
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019674572926630567,
+      "loss": 1.3269,
+      "step": 2517
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019674217951623707,
+      "loss": 1.0076,
+      "step": 2518
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019673862786325267,
+      "loss": 1.3337,
+      "step": 2519
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001967350743074223,
+      "loss": 1.2847,
+      "step": 2520
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019673151884881592,
+      "loss": 1.2361,
+      "step": 2521
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019672796148750336,
+      "loss": 1.1709,
+      "step": 2522
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019672440222355468,
+      "loss": 1.1486,
+      "step": 2523
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019672084105703985,
+      "loss": 1.0926,
+      "step": 2524
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019671727798802895,
+      "loss": 1.0342,
+      "step": 2525
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019671371301659206,
+      "loss": 1.0668,
+      "step": 2526
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019671014614279926,
+      "loss": 1.0313,
+      "step": 2527
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019670657736672075,
+      "loss": 0.8057,
+      "step": 2528
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019670300668842673,
+      "loss": 1.1533,
+      "step": 2529
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966994341079874,
+      "loss": 1.0223,
+      "step": 2530
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019669585962547306,
+      "loss": 0.9808,
+      "step": 2531
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019669228324095404,
+      "loss": 1.1793,
+      "step": 2532
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019668870495450066,
+      "loss": 0.995,
+      "step": 2533
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966851247661833,
+      "loss": 1.1607,
+      "step": 2534
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019668154267607238,
+      "loss": 1.0172,
+      "step": 2535
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966779586842384,
+      "loss": 1.2904,
+      "step": 2536
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966743727907518,
+      "loss": 1.0688,
+      "step": 2537
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019667078499568315,
+      "loss": 0.8764,
+      "step": 2538
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000196667195299103,
+      "loss": 1.0738,
+      "step": 2539
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000196663603701082,
+      "loss": 1.1514,
+      "step": 2540
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019666001020169073,
+      "loss": 1.2624,
+      "step": 2541
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019665641480099994,
+      "loss": 1.0537,
+      "step": 2542
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019665281749908033,
+      "loss": 1.1786,
+      "step": 2543
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019664921829600263,
+      "loss": 1.2599,
+      "step": 2544
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966456171918377,
+      "loss": 1.0437,
+      "step": 2545
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966420141866563,
+      "loss": 0.9871,
+      "step": 2546
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019663840928052935,
+      "loss": 1.025,
+      "step": 2547
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019663480247352773,
+      "loss": 0.8561,
+      "step": 2548
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019663119376572242,
+      "loss": 1.1023,
+      "step": 2549
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019662758315718435,
+      "loss": 0.8884,
+      "step": 2550
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019662397064798461,
+      "loss": 0.8908,
+      "step": 2551
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019662035623819416,
+      "loss": 0.9872,
+      "step": 2552
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966167399278842,
+      "loss": 1.0695,
+      "step": 2553
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019661312171712582,
+      "loss": 0.9552,
+      "step": 2554
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001966095016059902,
+      "loss": 0.884,
+      "step": 2555
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019660587959454852,
+      "loss": 0.9508,
+      "step": 2556
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019660225568287202,
+      "loss": 0.9901,
+      "step": 2557
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019659862987103204,
+      "loss": 1.0802,
+      "step": 2558
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019659500215909984,
+      "loss": 1.1725,
+      "step": 2559
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019659137254714678,
+      "loss": 1.0344,
+      "step": 2560
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001965877410352443,
+      "loss": 0.9382,
+      "step": 2561
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019658410762346385,
+      "loss": 0.9713,
+      "step": 2562
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001965804723118768,
+      "loss": 0.9521,
+      "step": 2563
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019657683510055472,
+      "loss": 1.1363,
+      "step": 2564
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019657319598956913,
+      "loss": 1.0661,
+      "step": 2565
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019656955497899164,
+      "loss": 0.8593,
+      "step": 2566
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019656591206889385,
+      "loss": 1.2275,
+      "step": 2567
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019656226725934744,
+      "loss": 1.1781,
+      "step": 2568
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019655862055042406,
+      "loss": 1.1247,
+      "step": 2569
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019655497194219547,
+      "loss": 1.1012,
+      "step": 2570
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019655132143473345,
+      "loss": 1.2077,
+      "step": 2571
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019654766902810978,
+      "loss": 1.0919,
+      "step": 2572
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001965440147223963,
+      "loss": 1.1527,
+      "step": 2573
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019654035851766492,
+      "loss": 1.0871,
+      "step": 2574
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019653670041398753,
+      "loss": 0.9259,
+      "step": 2575
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001965330404114361,
+      "loss": 1.0322,
+      "step": 2576
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001965293785100826,
+      "loss": 1.0831,
+      "step": 2577
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019652571470999908,
+      "loss": 0.9165,
+      "step": 2578
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001965220490112576,
+      "loss": 1.1864,
+      "step": 2579
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019651838141393027,
+      "loss": 1.1695,
+      "step": 2580
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019651471191808924,
+      "loss": 1.0237,
+      "step": 2581
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019651104052380666,
+      "loss": 0.8718,
+      "step": 2582
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019650736723115475,
+      "loss": 1.0188,
+      "step": 2583
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019650369204020577,
+      "loss": 1.1001,
+      "step": 2584
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019650001495103206,
+      "loss": 0.9962,
+      "step": 2585
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019649633596370588,
+      "loss": 1.1054,
+      "step": 2586
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019649265507829964,
+      "loss": 1.0702,
+      "step": 2587
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019648897229488567,
+      "loss": 1.0443,
+      "step": 2588
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019648528761353652,
+      "loss": 1.1545,
+      "step": 2589
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019648160103432458,
+      "loss": 0.9924,
+      "step": 2590
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019647791255732243,
+      "loss": 1.0991,
+      "step": 2591
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019647422218260256,
+      "loss": 0.9705,
+      "step": 2592
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001964705299102376,
+      "loss": 0.9789,
+      "step": 2593
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019646683574030017,
+      "loss": 1.0104,
+      "step": 2594
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001964631396728629,
+      "loss": 1.0539,
+      "step": 2595
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019645944170799856,
+      "loss": 1.0312,
+      "step": 2596
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019645574184577982,
+      "loss": 1.17,
+      "step": 2597
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001964520400862795,
+      "loss": 1.1055,
+      "step": 2598
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001964483364295704,
+      "loss": 1.2829,
+      "step": 2599
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019644463087572535,
+      "loss": 1.1771,
+      "step": 2600
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019644092342481728,
+      "loss": 0.9981,
+      "step": 2601
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001964372140769191,
+      "loss": 1.1765,
+      "step": 2602
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019643350283210374,
+      "loss": 1.0354,
+      "step": 2603
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019642978969044425,
+      "loss": 1.0864,
+      "step": 2604
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019642607465201364,
+      "loss": 1.1097,
+      "step": 2605
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019642235771688497,
+      "loss": 1.1456,
+      "step": 2606
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001964186388851314,
+      "loss": 0.9878,
+      "step": 2607
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019641491815682602,
+      "loss": 0.9891,
+      "step": 2608
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019641119553204207,
+      "loss": 1.1937,
+      "step": 2609
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019640747101085272,
+      "loss": 1.0053,
+      "step": 2610
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001964037445933313,
+      "loss": 0.9866,
+      "step": 2611
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019640001627955103,
+      "loss": 1.0687,
+      "step": 2612
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019639628606958533,
+      "loss": 1.0602,
+      "step": 2613
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001963925539635075,
+      "loss": 0.9208,
+      "step": 2614
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019638881996139103,
+      "loss": 1.123,
+      "step": 2615
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019638508406330925,
+      "loss": 1.1215,
+      "step": 2616
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019638134626933574,
+      "loss": 1.1705,
+      "step": 2617
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.000196377606579544,
+      "loss": 1.0996,
+      "step": 2618
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019637386499400759,
+      "loss": 0.976,
+      "step": 2619
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019637012151280009,
+      "loss": 1.0644,
+      "step": 2620
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001963663761359951,
+      "loss": 0.9566,
+      "step": 2621
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001963626288636664,
+      "loss": 0.889,
+      "step": 2622
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019635887969588763,
+      "loss": 1.1685,
+      "step": 2623
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019635512863273253,
+      "loss": 1.0517,
+      "step": 2624
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019635137567427488,
+      "loss": 1.0623,
+      "step": 2625
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019634762082058852,
+      "loss": 1.0511,
+      "step": 2626
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019634386407174734,
+      "loss": 1.1571,
+      "step": 2627
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019634010542782514,
+      "loss": 1.1225,
+      "step": 2628
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019633634488889595,
+      "loss": 1.3627,
+      "step": 2629
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001963325824550337,
+      "loss": 1.1343,
+      "step": 2630
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001963288181263124,
+      "loss": 1.139,
+      "step": 2631
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019632505190280607,
+      "loss": 1.2176,
+      "step": 2632
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019632128378458882,
+      "loss": 0.8947,
+      "step": 2633
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019631751377173476,
+      "loss": 1.0553,
+      "step": 2634
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019631374186431807,
+      "loss": 1.2592,
+      "step": 2635
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019630996806241287,
+      "loss": 0.9839,
+      "step": 2636
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001963061923660935,
+      "loss": 1.1952,
+      "step": 2637
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019630241477543414,
+      "loss": 0.9762,
+      "step": 2638
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019629863529050913,
+      "loss": 1.1017,
+      "step": 2639
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001962948539113928,
+      "loss": 1.012,
+      "step": 2640
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019629107063815954,
+      "loss": 1.0292,
+      "step": 2641
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001962872854708838,
+      "loss": 1.2882,
+      "step": 2642
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019628349840963995,
+      "loss": 1.0453,
+      "step": 2643
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019627970945450255,
+      "loss": 0.8566,
+      "step": 2644
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0001962759186055461,
+      "loss": 1.0917,
+      "step": 2645
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019627212586284518,
+      "loss": 0.9585,
+      "step": 2646
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019626833122647437,
+      "loss": 1.0324,
+      "step": 2647
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019626453469650837,
+      "loss": 1.0606,
+      "step": 2648
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019626073627302177,
+      "loss": 0.997,
+      "step": 2649
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019625693595608934,
+      "loss": 1.0091,
+      "step": 2650
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019625313374578584,
+      "loss": 1.0173,
+      "step": 2651
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019624932964218604,
+      "loss": 1.0357,
+      "step": 2652
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019624552364536473,
+      "loss": 0.9964,
+      "step": 2653
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019624171575539684,
+      "loss": 0.9558,
+      "step": 2654
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.00019623790597235726,
+      "loss": 0.923,
+      "step": 2655
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019623409429632088,
+      "loss": 1.0744,
+      "step": 2656
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019623028072736273,
+      "loss": 0.9308,
+      "step": 2657
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001962264652655578,
+      "loss": 1.2197,
+      "step": 2658
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001962226479109811,
+      "loss": 0.9273,
+      "step": 2659
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019621882866370779,
+      "loss": 1.1082,
+      "step": 2660
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019621500752381294,
+      "loss": 1.0939,
+      "step": 2661
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019621118449137176,
+      "loss": 1.1365,
+      "step": 2662
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019620735956645943,
+      "loss": 1.0356,
+      "step": 2663
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019620353274915116,
+      "loss": 1.1245,
+      "step": 2664
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019619970403952222,
+      "loss": 1.0056,
+      "step": 2665
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019619587343764795,
+      "loss": 1.1786,
+      "step": 2666
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019619204094360374,
+      "loss": 0.9238,
+      "step": 2667
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019618820655746487,
+      "loss": 1.1085,
+      "step": 2668
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019618437027930683,
+      "loss": 1.058,
+      "step": 2669
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019618053210920507,
+      "loss": 0.9241,
+      "step": 2670
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019617669204723508,
+      "loss": 0.9805,
+      "step": 2671
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019617285009347239,
+      "loss": 0.9134,
+      "step": 2672
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001961690062479926,
+      "loss": 0.9219,
+      "step": 2673
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001961651605108713,
+      "loss": 1.1063,
+      "step": 2674
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001961613128821841,
+      "loss": 1.0939,
+      "step": 2675
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019615746336200672,
+      "loss": 1.0103,
+      "step": 2676
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019615361195041492,
+      "loss": 1.0657,
+      "step": 2677
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019614975864748436,
+      "loss": 1.0109,
+      "step": 2678
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001961459034532909,
+      "loss": 0.8764,
+      "step": 2679
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019614204636791036,
+      "loss": 0.9314,
+      "step": 2680
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019613818739141862,
+      "loss": 1.0913,
+      "step": 2681
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019613432652389157,
+      "loss": 0.8528,
+      "step": 2682
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001961304637654051,
+      "loss": 0.8981,
+      "step": 2683
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019612659911603532,
+      "loss": 1.2202,
+      "step": 2684
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019612273257585815,
+      "loss": 1.1408,
+      "step": 2685
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019611886414494967,
+      "loss": 1.212,
+      "step": 2686
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019611499382338597,
+      "loss": 1.4314,
+      "step": 2687
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001961111216112432,
+      "loss": 0.9237,
+      "step": 2688
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019610724750859746,
+      "loss": 1.19,
+      "step": 2689
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000196103371515525,
+      "loss": 1.0189,
+      "step": 2690
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001960994936321021,
+      "loss": 1.3077,
+      "step": 2691
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019609561385840497,
+      "loss": 0.874,
+      "step": 2692
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019609173219450998,
+      "loss": 1.0136,
+      "step": 2693
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019608784864049344,
+      "loss": 0.9015,
+      "step": 2694
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019608396319643174,
+      "loss": 1.0386,
+      "step": 2695
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019608007586240133,
+      "loss": 1.0757,
+      "step": 2696
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019607618663847865,
+      "loss": 1.2818,
+      "step": 2697
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019607229552474022,
+      "loss": 0.9065,
+      "step": 2698
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019606840252126259,
+      "loss": 1.0295,
+      "step": 2699
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019606450762812232,
+      "loss": 1.1488,
+      "step": 2700
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000196060610845396,
+      "loss": 1.0031,
+      "step": 2701
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019605671217316032,
+      "loss": 1.0914,
+      "step": 2702
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019605281161149196,
+      "loss": 1.0544,
+      "step": 2703
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019604890916046763,
+      "loss": 1.1689,
+      "step": 2704
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019604500482016407,
+      "loss": 1.1093,
+      "step": 2705
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019604109859065811,
+      "loss": 1.0668,
+      "step": 2706
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019603719047202659,
+      "loss": 1.0914,
+      "step": 2707
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019603328046434635,
+      "loss": 1.1903,
+      "step": 2708
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001960293685676943,
+      "loss": 0.9751,
+      "step": 2709
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019602545478214744,
+      "loss": 1.178,
+      "step": 2710
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019602153910778272,
+      "loss": 1.1109,
+      "step": 2711
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019601762154467713,
+      "loss": 1.2275,
+      "step": 2712
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001960137020929078,
+      "loss": 1.0913,
+      "step": 2713
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019600978075255178,
+      "loss": 1.0702,
+      "step": 2714
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001960058575236862,
+      "loss": 1.2654,
+      "step": 2715
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001960019324063882,
+      "loss": 1.2328,
+      "step": 2716
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019599800540073505,
+      "loss": 0.8707,
+      "step": 2717
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019599407650680397,
+      "loss": 1.1954,
+      "step": 2718
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019599014572467224,
+      "loss": 0.9127,
+      "step": 2719
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019598621305441717,
+      "loss": 1.0498,
+      "step": 2720
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019598227849611615,
+      "loss": 1.0239,
+      "step": 2721
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019597834204984654,
+      "loss": 1.2201,
+      "step": 2722
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019597440371568574,
+      "loss": 0.9784,
+      "step": 2723
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019597046349371128,
+      "loss": 0.9896,
+      "step": 2724
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019596652138400064,
+      "loss": 1.0578,
+      "step": 2725
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019596257738663136,
+      "loss": 1.1683,
+      "step": 2726
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019595863150168098,
+      "loss": 1.1261,
+      "step": 2727
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001959546837292272,
+      "loss": 1.1437,
+      "step": 2728
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019595073406934762,
+      "loss": 1.0607,
+      "step": 2729
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019594678252211992,
+      "loss": 1.0985,
+      "step": 2730
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019594282908762183,
+      "loss": 1.1599,
+      "step": 2731
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019593887376593113,
+      "loss": 0.9978,
+      "step": 2732
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001959349165571256,
+      "loss": 1.1386,
+      "step": 2733
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001959309574612831,
+      "loss": 1.1325,
+      "step": 2734
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001959269964784815,
+      "loss": 1.111,
+      "step": 2735
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019592303360879877,
+      "loss": 1.1745,
+      "step": 2736
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019591906885231276,
+      "loss": 0.904,
+      "step": 2737
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019591510220910146,
+      "loss": 1.3533,
+      "step": 2738
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019591113367924297,
+      "loss": 0.9898,
+      "step": 2739
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019590716326281532,
+      "loss": 1.1243,
+      "step": 2740
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001959031909598966,
+      "loss": 1.0758,
+      "step": 2741
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958992167705649,
+      "loss": 1.2051,
+      "step": 2742
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958952406948985,
+      "loss": 1.0128,
+      "step": 2743
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958912627329755,
+      "loss": 1.0561,
+      "step": 2744
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019588728288487422,
+      "loss": 1.0491,
+      "step": 2745
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958833011506729,
+      "loss": 1.0901,
+      "step": 2746
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019587931753044986,
+      "loss": 1.0683,
+      "step": 2747
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019587533202428353,
+      "loss": 1.329,
+      "step": 2748
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958713446322522,
+      "loss": 1.055,
+      "step": 2749
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019586735535443437,
+      "loss": 1.0331,
+      "step": 2750
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958633641909085,
+      "loss": 0.8773,
+      "step": 2751
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019585937114175307,
+      "loss": 1.2642,
+      "step": 2752
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019585537620704665,
+      "loss": 0.9998,
+      "step": 2753
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958513793868678,
+      "loss": 1.3747,
+      "step": 2754
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019584738068129519,
+      "loss": 1.0521,
+      "step": 2755
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019584338009040738,
+      "loss": 1.0805,
+      "step": 2756
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019583937761428314,
+      "loss": 1.1656,
+      "step": 2757
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019583537325300118,
+      "loss": 1.2537,
+      "step": 2758
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958313670066402,
+      "loss": 1.0875,
+      "step": 2759
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958273588752791,
+      "loss": 1.1142,
+      "step": 2760
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958233488589967,
+      "loss": 0.8597,
+      "step": 2761
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019581933695787186,
+      "loss": 1.0276,
+      "step": 2762
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019581532317198343,
+      "loss": 1.1907,
+      "step": 2763
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019581130750141048,
+      "loss": 0.9774,
+      "step": 2764
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019580728994623195,
+      "loss": 1.006,
+      "step": 2765
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001958032705065268,
+      "loss": 0.8993,
+      "step": 2766
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019579924918237417,
+      "loss": 1.1144,
+      "step": 2767
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019579522597385315,
+      "loss": 1.1225,
+      "step": 2768
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019579120088104288,
+      "loss": 1.1469,
+      "step": 2769
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019578717390402248,
+      "loss": 1.0281,
+      "step": 2770
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019578314504287123,
+      "loss": 1.1099,
+      "step": 2771
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019577911429766832,
+      "loss": 0.8439,
+      "step": 2772
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019577508166849304,
+      "loss": 0.9314,
+      "step": 2773
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019577104715542477,
+      "loss": 1.0007,
+      "step": 2774
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019576701075854283,
+      "loss": 1.0913,
+      "step": 2775
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001957629724779266,
+      "loss": 1.0584,
+      "step": 2776
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019575893231365554,
+      "loss": 1.084,
+      "step": 2777
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019575489026580912,
+      "loss": 0.9625,
+      "step": 2778
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001957508463344668,
+      "loss": 1.0148,
+      "step": 2779
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019574680051970817,
+      "loss": 0.9537,
+      "step": 2780
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019574275282161283,
+      "loss": 0.9703,
+      "step": 2781
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019573870324026034,
+      "loss": 1.1408,
+      "step": 2782
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019573465177573037,
+      "loss": 1.1247,
+      "step": 2783
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019573059842810267,
+      "loss": 1.1258,
+      "step": 2784
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019572654319745692,
+      "loss": 1.0329,
+      "step": 2785
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019572248608387284,
+      "loss": 1.0787,
+      "step": 2786
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019571842708743033,
+      "loss": 1.1174,
+      "step": 2787
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019571436620820917,
+      "loss": 1.061,
+      "step": 2788
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019571030344628926,
+      "loss": 1.1274,
+      "step": 2789
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001957062388017505,
+      "loss": 0.962,
+      "step": 2790
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019570217227467287,
+      "loss": 1.1957,
+      "step": 2791
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956981038651363,
+      "loss": 0.8537,
+      "step": 2792
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956940335732209,
+      "loss": 1.1129,
+      "step": 2793
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019568996139900666,
+      "loss": 0.9583,
+      "step": 2794
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956858873425737,
+      "loss": 1.1912,
+      "step": 2795
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019568181140400216,
+      "loss": 0.8911,
+      "step": 2796
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019567773358337222,
+      "loss": 0.9594,
+      "step": 2797
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019567365388076412,
+      "loss": 1.1342,
+      "step": 2798
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000195669572296258,
+      "loss": 1.0917,
+      "step": 2799
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019566548882993424,
+      "loss": 1.0947,
+      "step": 2800
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019566140348187318,
+      "loss": 1.0371,
+      "step": 2801
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019565731625215512,
+      "loss": 1.0842,
+      "step": 2802
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019565322714086046,
+      "loss": 1.1293,
+      "step": 2803
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019564913614806965,
+      "loss": 1.0879,
+      "step": 2804
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019564504327386316,
+      "loss": 1.0482,
+      "step": 2805
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956409485183215,
+      "loss": 1.0119,
+      "step": 2806
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956368518815252,
+      "loss": 1.2269,
+      "step": 2807
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019563275336355483,
+      "loss": 1.0743,
+      "step": 2808
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019562865296449106,
+      "loss": 1.0739,
+      "step": 2809
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956245506844145,
+      "loss": 1.1254,
+      "step": 2810
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019562044652340584,
+      "loss": 1.0721,
+      "step": 2811
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019561634048154584,
+      "loss": 0.8997,
+      "step": 2812
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956122325589152,
+      "loss": 1.0579,
+      "step": 2813
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001956081227555948,
+      "loss": 1.1026,
+      "step": 2814
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019560401107166544,
+      "loss": 0.9841,
+      "step": 2815
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.000195599897507208,
+      "loss": 1.0557,
+      "step": 2816
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955957820623034,
+      "loss": 0.9992,
+      "step": 2817
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019559166473703262,
+      "loss": 1.1737,
+      "step": 2818
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955875455314766,
+      "loss": 0.9893,
+      "step": 2819
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019558342444571638,
+      "loss": 1.2942,
+      "step": 2820
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019557930147983302,
+      "loss": 1.0444,
+      "step": 2821
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019557517663390764,
+      "loss": 0.9778,
+      "step": 2822
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019557104990802133,
+      "loss": 1.0873,
+      "step": 2823
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955669213022553,
+      "loss": 1.0943,
+      "step": 2824
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019556279081669073,
+      "loss": 0.9881,
+      "step": 2825
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019555865845140893,
+      "loss": 1.0379,
+      "step": 2826
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955545242064911,
+      "loss": 1.0718,
+      "step": 2827
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019555038808201865,
+      "loss": 1.1442,
+      "step": 2828
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019554625007807283,
+      "loss": 1.0864,
+      "step": 2829
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955421101947351,
+      "loss": 0.9265,
+      "step": 2830
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955379684320869,
+      "loss": 1.2864,
+      "step": 2831
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955338247902097,
+      "loss": 1.1231,
+      "step": 2832
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019552967926918499,
+      "loss": 1.1858,
+      "step": 2833
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019552553186909428,
+      "loss": 1.1841,
+      "step": 2834
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019552138259001917,
+      "loss": 1.3581,
+      "step": 2835
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019551723143204132,
+      "loss": 1.0779,
+      "step": 2836
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955130783952423,
+      "loss": 0.8033,
+      "step": 2837
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955089234797039,
+      "loss": 1.1203,
+      "step": 2838
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019550476668550777,
+      "loss": 1.1221,
+      "step": 2839
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001955006080127357,
+      "loss": 1.238,
+      "step": 2840
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019549644746146952,
+      "loss": 1.0624,
+      "step": 2841
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019549228503179097,
+      "loss": 1.2284,
+      "step": 2842
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019548812072378205,
+      "loss": 1.213,
+      "step": 2843
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001954839545375246,
+      "loss": 1.0619,
+      "step": 2844
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019547978647310057,
+      "loss": 0.9207,
+      "step": 2845
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019547561653059198,
+      "loss": 1.0832,
+      "step": 2846
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019547144471008084,
+      "loss": 1.1249,
+      "step": 2847
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019546727101164916,
+      "loss": 1.1333,
+      "step": 2848
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001954630954353791,
+      "loss": 1.0112,
+      "step": 2849
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001954589179813528,
+      "loss": 0.9757,
+      "step": 2850
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019545473864965236,
+      "loss": 1.0634,
+      "step": 2851
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019545055744036005,
+      "loss": 1.1706,
+      "step": 2852
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019544637435355808,
+      "loss": 1.182,
+      "step": 2853
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019544218938932875,
+      "loss": 0.9381,
+      "step": 2854
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019543800254775436,
+      "loss": 1.1275,
+      "step": 2855
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019543381382891733,
+      "loss": 1.3267,
+      "step": 2856
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019542962323289994,
+      "loss": 1.1922,
+      "step": 2857
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019542543075978471,
+      "loss": 0.9996,
+      "step": 2858
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019542123640965407,
+      "loss": 0.9548,
+      "step": 2859
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019541704018259052,
+      "loss": 1.0819,
+      "step": 2860
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019541284207867662,
+      "loss": 1.1124,
+      "step": 2861
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019540864209799491,
+      "loss": 1.1053,
+      "step": 2862
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019540444024062804,
+      "loss": 1.1938,
+      "step": 2863
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019540023650665866,
+      "loss": 1.1454,
+      "step": 2864
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019539603089616944,
+      "loss": 1.0813,
+      "step": 2865
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001953918234092431,
+      "loss": 1.154,
+      "step": 2866
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019538761404596242,
+      "loss": 1.0341,
+      "step": 2867
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001953834028064102,
+      "loss": 1.2206,
+      "step": 2868
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001953791896906692,
+      "loss": 1.0213,
+      "step": 2869
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019537497469882242,
+      "loss": 1.1475,
+      "step": 2870
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019537075783095266,
+      "loss": 0.8716,
+      "step": 2871
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019536653908714294,
+      "loss": 1.0082,
+      "step": 2872
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019536231846747622,
+      "loss": 1.0896,
+      "step": 2873
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001953580959720355,
+      "loss": 1.4202,
+      "step": 2874
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019535387160090384,
+      "loss": 1.1199,
+      "step": 2875
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019534964535416433,
+      "loss": 1.1411,
+      "step": 2876
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001953454172319001,
+      "loss": 0.9853,
+      "step": 2877
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001953411872341944,
+      "loss": 1.0275,
+      "step": 2878
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.0001953369553611303,
+      "loss": 1.101,
+      "step": 2879
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019533272161279113,
+      "loss": 0.9833,
+      "step": 2880
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019532848598926013,
+      "loss": 1.3588,
+      "step": 2881
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019532424849062063,
+      "loss": 1.2307,
+      "step": 2882
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019532000911695596,
+      "loss": 0.954,
+      "step": 2883
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019531576786834952,
+      "loss": 1.0403,
+      "step": 2884
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019531152474488477,
+      "loss": 1.2148,
+      "step": 2885
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 0.00019530727974664513,
+      "loss": 1.3072,
+      "step": 2886
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001953030328737141,
+      "loss": 1.0381,
+      "step": 2887
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019529878412617526,
+      "loss": 0.9309,
+      "step": 2888
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019529453350411214,
+      "loss": 0.9982,
+      "step": 2889
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001952902810076083,
+      "loss": 1.0003,
+      "step": 2890
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019528602663674752,
+      "loss": 1.0001,
+      "step": 2891
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001952817703916134,
+      "loss": 1.3001,
+      "step": 2892
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019527751227228963,
+      "loss": 0.949,
+      "step": 2893
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019527325227886003,
+      "loss": 1.2746,
+      "step": 2894
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001952689904114084,
+      "loss": 1.0073,
+      "step": 2895
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001952647266700185,
+      "loss": 1.1405,
+      "step": 2896
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019526046105477424,
+      "loss": 1.0269,
+      "step": 2897
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019525619356575952,
+      "loss": 1.1018,
+      "step": 2898
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019525192420305834,
+      "loss": 1.2376,
+      "step": 2899
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019524765296675457,
+      "loss": 0.979,
+      "step": 2900
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001952433798569323,
+      "loss": 1.2353,
+      "step": 2901
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019523910487367555,
+      "loss": 1.0568,
+      "step": 2902
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001952348280170684,
+      "loss": 0.9935,
+      "step": 2903
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019523054928719503,
+      "loss": 0.9502,
+      "step": 2904
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019522626868413956,
+      "loss": 1.1691,
+      "step": 2905
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019522198620798619,
+      "loss": 1.0356,
+      "step": 2906
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019521770185881916,
+      "loss": 1.0828,
+      "step": 2907
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019521341563672273,
+      "loss": 1.1044,
+      "step": 2908
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019520912754178123,
+      "loss": 1.0432,
+      "step": 2909
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019520483757407906,
+      "loss": 1.0978,
+      "step": 2910
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001952005457337005,
+      "loss": 0.9827,
+      "step": 2911
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019519625202073,
+      "loss": 0.9566,
+      "step": 2912
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019519195643525206,
+      "loss": 1.0158,
+      "step": 2913
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019518765897735113,
+      "loss": 0.9566,
+      "step": 2914
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001951833596471118,
+      "loss": 1.2145,
+      "step": 2915
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019517905844461855,
+      "loss": 1.1215,
+      "step": 2916
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019517475536995604,
+      "loss": 1.1861,
+      "step": 2917
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019517045042320892,
+      "loss": 1.1729,
+      "step": 2918
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019516614360446183,
+      "loss": 1.1745,
+      "step": 2919
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001951618349137995,
+      "loss": 1.106,
+      "step": 2920
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001951575243513067,
+      "loss": 1.0994,
+      "step": 2921
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019515321191706822,
+      "loss": 1.0552,
+      "step": 2922
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019514889761116885,
+      "loss": 1.3274,
+      "step": 2923
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001951445814336935,
+      "loss": 1.1283,
+      "step": 2924
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000195140263384727,
+      "loss": 1.138,
+      "step": 2925
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019513594346435437,
+      "loss": 1.1624,
+      "step": 2926
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019513162167266053,
+      "loss": 1.2811,
+      "step": 2927
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001951272980097305,
+      "loss": 1.0852,
+      "step": 2928
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001951229724756493,
+      "loss": 1.1637,
+      "step": 2929
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019511864507050207,
+      "loss": 0.9411,
+      "step": 2930
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019511431579437386,
+      "loss": 0.9771,
+      "step": 2931
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001951099846473499,
+      "loss": 1.1544,
+      "step": 2932
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019510565162951537,
+      "loss": 1.1055,
+      "step": 2933
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019510131674095545,
+      "loss": 1.2633,
+      "step": 2934
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019509697998175546,
+      "loss": 1.229,
+      "step": 2935
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019509264135200068,
+      "loss": 1.2335,
+      "step": 2936
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950883008517764,
+      "loss": 1.1946,
+      "step": 2937
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950839584811681,
+      "loss": 1.0468,
+      "step": 2938
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019507961424026113,
+      "loss": 1.1363,
+      "step": 2939
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019507526812914097,
+      "loss": 1.1208,
+      "step": 2940
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950709201478931,
+      "loss": 1.0416,
+      "step": 2941
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000195066570296603,
+      "loss": 0.9807,
+      "step": 2942
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950622185753563,
+      "loss": 1.0028,
+      "step": 2943
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019505786498423856,
+      "loss": 1.0029,
+      "step": 2944
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950535095233354,
+      "loss": 1.0865,
+      "step": 2945
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019504915219273255,
+      "loss": 1.0868,
+      "step": 2946
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019504479299251566,
+      "loss": 1.2557,
+      "step": 2947
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950404319227705,
+      "loss": 1.1401,
+      "step": 2948
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019503606898358284,
+      "loss": 0.896,
+      "step": 2949
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019503170417503853,
+      "loss": 1.0834,
+      "step": 2950
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950273374972234,
+      "loss": 1.0833,
+      "step": 2951
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019502296895022335,
+      "loss": 1.0791,
+      "step": 2952
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001950185985341243,
+      "loss": 1.0866,
+      "step": 2953
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019501422624901223,
+      "loss": 1.0659,
+      "step": 2954
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019500985209497312,
+      "loss": 1.0694,
+      "step": 2955
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019500547607209304,
+      "loss": 0.9631,
+      "step": 2956
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019500109818045805,
+      "loss": 1.0045,
+      "step": 2957
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019499671842015425,
+      "loss": 1.0084,
+      "step": 2958
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001949923367912678,
+      "loss": 1.1091,
+      "step": 2959
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001949879532938849,
+      "loss": 1.135,
+      "step": 2960
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019498356792809176,
+      "loss": 1.2599,
+      "step": 2961
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019497918069397464,
+      "loss": 1.1226,
+      "step": 2962
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019497479159161988,
+      "loss": 0.988,
+      "step": 2963
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019497040062111374,
+      "loss": 1.0499,
+      "step": 2964
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019496600778254263,
+      "loss": 1.0266,
+      "step": 2965
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019496161307599295,
+      "loss": 1.0303,
+      "step": 2966
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019495721650155114,
+      "loss": 1.1993,
+      "step": 2967
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019495281805930367,
+      "loss": 1.2023,
+      "step": 2968
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001949484177493371,
+      "loss": 1.0945,
+      "step": 2969
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019494401557173795,
+      "loss": 1.0461,
+      "step": 2970
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001949396115265928,
+      "loss": 1.3133,
+      "step": 2971
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019493520561398832,
+      "loss": 1.109,
+      "step": 2972
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019493079783401113,
+      "loss": 1.2476,
+      "step": 2973
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019492638818674796,
+      "loss": 1.1883,
+      "step": 2974
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019492197667228553,
+      "loss": 1.205,
+      "step": 2975
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019491756329071062,
+      "loss": 1.1042,
+      "step": 2976
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019491314804211005,
+      "loss": 1.0712,
+      "step": 2977
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019490873092657064,
+      "loss": 1.0127,
+      "step": 2978
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019490431194417934,
+      "loss": 0.9033,
+      "step": 2979
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000194899891095023,
+      "loss": 1.0594,
+      "step": 2980
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001948954683791886,
+      "loss": 1.3817,
+      "step": 2981
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019489104379676313,
+      "loss": 0.9904,
+      "step": 2982
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019488661734783365,
+      "loss": 0.9207,
+      "step": 2983
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019488218903248718,
+      "loss": 1.1581,
+      "step": 2984
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019487775885081087,
+      "loss": 0.8845,
+      "step": 2985
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019487332680289185,
+      "loss": 0.8998,
+      "step": 2986
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019486889288881728,
+      "loss": 1.0635,
+      "step": 2987
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019486445710867442,
+      "loss": 0.9702,
+      "step": 2988
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019486001946255046,
+      "loss": 1.0577,
+      "step": 2989
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019485557995053274,
+      "loss": 1.1458,
+      "step": 2990
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019485113857270855,
+      "loss": 1.2567,
+      "step": 2991
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019484669532916527,
+      "loss": 0.8205,
+      "step": 2992
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001948422502199903,
+      "loss": 0.9792,
+      "step": 2993
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019483780324527103,
+      "loss": 1.1099,
+      "step": 2994
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000194833354405095,
+      "loss": 1.0669,
+      "step": 2995
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019482890369954968,
+      "loss": 1.2916,
+      "step": 2996
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019482445112872264,
+      "loss": 1.1088,
+      "step": 2997
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019481999669270143,
+      "loss": 1.0801,
+      "step": 2998
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001948155403915737,
+      "loss": 1.1581,
+      "step": 2999
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019481108222542707,
+      "loss": 1.0516,
+      "step": 3000
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001948066221943493,
+      "loss": 0.8512,
+      "step": 3001
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000194802160298428,
+      "loss": 1.0478,
+      "step": 3002
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019479769653775106,
+      "loss": 1.1806,
+      "step": 3003
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947932309124062,
+      "loss": 1.187,
+      "step": 3004
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019478876342248128,
+      "loss": 0.9339,
+      "step": 3005
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947842940680642,
+      "loss": 0.9426,
+      "step": 3006
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019477982284924289,
+      "loss": 1.0148,
+      "step": 3007
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019477534976610522,
+      "loss": 1.0075,
+      "step": 3008
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947708748187392,
+      "loss": 0.9864,
+      "step": 3009
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947663980072329,
+      "loss": 1.0351,
+      "step": 3010
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019476191933167435,
+      "loss": 1.0686,
+      "step": 3011
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947574387921516,
+      "loss": 1.1719,
+      "step": 3012
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947529563887529,
+      "loss": 1.1707,
+      "step": 3013
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947484721215663,
+      "loss": 1.0323,
+      "step": 3014
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019474398599068007,
+      "loss": 1.1984,
+      "step": 3015
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019473949799618244,
+      "loss": 1.0314,
+      "step": 3016
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019473500813816164,
+      "loss": 1.0264,
+      "step": 3017
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019473051641670606,
+      "loss": 0.9678,
+      "step": 3018
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.000194726022831904,
+      "loss": 1.0971,
+      "step": 3019
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001947215273838439,
+      "loss": 1.0936,
+      "step": 3020
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019471703007261412,
+      "loss": 1.2481,
+      "step": 3021
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019471253089830317,
+      "loss": 1.0278,
+      "step": 3022
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019470802986099953,
+      "loss": 1.1176,
+      "step": 3023
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019470352696079174,
+      "loss": 1.0467,
+      "step": 3024
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019469902219776836,
+      "loss": 1.0407,
+      "step": 3025
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019469451557201802,
+      "loss": 0.8734,
+      "step": 3026
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019469000708362936,
+      "loss": 0.8802,
+      "step": 3027
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019468549673269106,
+      "loss": 0.919,
+      "step": 3028
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019468098451929183,
+      "loss": 0.8782,
+      "step": 3029
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019467647044352042,
+      "loss": 1.1049,
+      "step": 3030
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019467195450546565,
+      "loss": 0.9373,
+      "step": 3031
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019466743670521633,
+      "loss": 1.0555,
+      "step": 3032
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019466291704286129,
+      "loss": 1.321,
+      "step": 3033
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001946583955184895,
+      "loss": 0.9342,
+      "step": 3034
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019465387213218985,
+      "loss": 1.0266,
+      "step": 3035
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019464934688405135,
+      "loss": 1.0651,
+      "step": 3036
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019464481977416297,
+      "loss": 1.065,
+      "step": 3037
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019464029080261378,
+      "loss": 1.0534,
+      "step": 3038
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019463575996949288,
+      "loss": 1.1478,
+      "step": 3039
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019463122727488935,
+      "loss": 1.2791,
+      "step": 3040
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001946266927188924,
+      "loss": 1.0165,
+      "step": 3041
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019462215630159117,
+      "loss": 1.1086,
+      "step": 3042
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019461761802307495,
+      "loss": 1.1067,
+      "step": 3043
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019461307788343297,
+      "loss": 1.2073,
+      "step": 3044
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019460853588275454,
+      "loss": 1.2325,
+      "step": 3045
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019460399202112901,
+      "loss": 1.2672,
+      "step": 3046
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019459944629864572,
+      "loss": 0.96,
+      "step": 3047
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019459489871539412,
+      "loss": 1.2141,
+      "step": 3048
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019459034927146369,
+      "loss": 1.0663,
+      "step": 3049
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019458579796694387,
+      "loss": 1.1186,
+      "step": 3050
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001945812448019242,
+      "loss": 1.0202,
+      "step": 3051
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001945766897764942,
+      "loss": 1.1958,
+      "step": 3052
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019457213289074355,
+      "loss": 1.1959,
+      "step": 3053
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001945675741447618,
+      "loss": 1.1615,
+      "step": 3054
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019456301353863872,
+      "loss": 0.8316,
+      "step": 3055
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001945584510724639,
+      "loss": 1.0075,
+      "step": 3056
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019455388674632714,
+      "loss": 0.9613,
+      "step": 3057
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019454932056031825,
+      "loss": 0.7823,
+      "step": 3058
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019454475251452698,
+      "loss": 1.0473,
+      "step": 3059
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019454018260904325,
+      "loss": 1.2051,
+      "step": 3060
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019453561084395687,
+      "loss": 0.9258,
+      "step": 3061
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019453103721935787,
+      "loss": 0.9424,
+      "step": 3062
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019452646173533618,
+      "loss": 0.969,
+      "step": 3063
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019452188439198173,
+      "loss": 1.1043,
+      "step": 3064
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001945173051893846,
+      "loss": 0.9289,
+      "step": 3065
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001945127241276349,
+      "loss": 0.9206,
+      "step": 3066
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019450814120682269,
+      "loss": 1.1829,
+      "step": 3067
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019450355642703813,
+      "loss": 1.1269,
+      "step": 3068
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019449896978837141,
+      "loss": 1.0206,
+      "step": 3069
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019449438129091274,
+      "loss": 1.1798,
+      "step": 3070
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019448979093475237,
+      "loss": 1.2057,
+      "step": 3071
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019448519871998062,
+      "loss": 0.9464,
+      "step": 3072
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019448060464668783,
+      "loss": 1.0845,
+      "step": 3073
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019447600871496428,
+      "loss": 1.2005,
+      "step": 3074
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019447141092490043,
+      "loss": 1.2378,
+      "step": 3075
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019446681127658678,
+      "loss": 1.1564,
+      "step": 3076
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001944622097701137,
+      "loss": 1.2422,
+      "step": 3077
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019445760640557174,
+      "loss": 1.0106,
+      "step": 3078
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019445300118305144,
+      "loss": 1.1823,
+      "step": 3079
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019444839410264343,
+      "loss": 1.015,
+      "step": 3080
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019444378516443828,
+      "loss": 1.122,
+      "step": 3081
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019443917436852664,
+      "loss": 1.1098,
+      "step": 3082
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019443456171499925,
+      "loss": 1.1479,
+      "step": 3083
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001944299472039468,
+      "loss": 1.272,
+      "step": 3084
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001944253308354601,
+      "loss": 1.2541,
+      "step": 3085
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019442071260962994,
+      "loss": 1.1837,
+      "step": 3086
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019441609252654712,
+      "loss": 1.1403,
+      "step": 3087
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019441147058630258,
+      "loss": 1.1029,
+      "step": 3088
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001944068467889872,
+      "loss": 1.1059,
+      "step": 3089
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001944022211346919,
+      "loss": 0.9008,
+      "step": 3090
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001943975936235077,
+      "loss": 1.0265,
+      "step": 3091
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019439296425552565,
+      "loss": 1.2577,
+      "step": 3092
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019438833303083678,
+      "loss": 1.2344,
+      "step": 3093
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019438369994953216,
+      "loss": 0.9878,
+      "step": 3094
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019437906501170295,
+      "loss": 0.9334,
+      "step": 3095
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019437442821744035,
+      "loss": 1.0671,
+      "step": 3096
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019436978956683548,
+      "loss": 0.9159,
+      "step": 3097
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001943651490599797,
+      "loss": 1.1768,
+      "step": 3098
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019436050669696417,
+      "loss": 0.991,
+      "step": 3099
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001943558624778803,
+      "loss": 0.9789,
+      "step": 3100
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019435121640281938,
+      "loss": 1.2673,
+      "step": 3101
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019434656847187282,
+      "loss": 1.0108,
+      "step": 3102
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019434191868513204,
+      "loss": 1.083,
+      "step": 3103
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019433726704268852,
+      "loss": 1.1675,
+      "step": 3104
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019433261354463371,
+      "loss": 1.2653,
+      "step": 3105
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001943279581910592,
+      "loss": 1.1183,
+      "step": 3106
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019432330098205657,
+      "loss": 1.1534,
+      "step": 3107
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019431864191771732,
+      "loss": 1.3292,
+      "step": 3108
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019431398099813323,
+      "loss": 1.0373,
+      "step": 3109
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001943093182233959,
+      "loss": 1.1789,
+      "step": 3110
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019430465359359707,
+      "loss": 1.1721,
+      "step": 3111
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019429998710882845,
+      "loss": 1.2572,
+      "step": 3112
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019429531876918193,
+      "loss": 0.946,
+      "step": 3113
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019429064857474923,
+      "loss": 0.9479,
+      "step": 3114
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.0001942859765256223,
+      "loss": 0.9794,
+      "step": 3115
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019428130262189294,
+      "loss": 1.2602,
+      "step": 3116
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 0.00019427662686365314,
+      "loss": 1.0606,
+      "step": 3117
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019427194925099493,
+      "loss": 1.1141,
+      "step": 3118
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001942672697840102,
+      "loss": 1.0307,
+      "step": 3119
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019426258846279111,
+      "loss": 1.2101,
+      "step": 3120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001942579052874297,
+      "loss": 1.0203,
+      "step": 3121
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000194253220258018,
+      "loss": 1.2277,
+      "step": 3122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001942485333746483,
+      "loss": 1.1096,
+      "step": 3123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019424384463741272,
+      "loss": 1.0088,
+      "step": 3124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001942391540464035,
+      "loss": 1.0742,
+      "step": 3125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019423446160171288,
+      "loss": 0.9768,
+      "step": 3126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001942297673034332,
+      "loss": 0.9507,
+      "step": 3127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019422507115165677,
+      "loss": 1.2336,
+      "step": 3128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019422037314647598,
+      "loss": 1.2204,
+      "step": 3129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019421567328798325,
+      "loss": 1.0244,
+      "step": 3130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000194210971576271,
+      "loss": 1.0463,
+      "step": 3131
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001942062680114317,
+      "loss": 0.9521,
+      "step": 3132
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019420156259355791,
+      "loss": 1.0373,
+      "step": 3133
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019419685532274215,
+      "loss": 0.9568,
+      "step": 3134
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019419214619907703,
+      "loss": 1.017,
+      "step": 3135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019418743522265519,
+      "loss": 1.1727,
+      "step": 3136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019418272239356927,
+      "loss": 1.0598,
+      "step": 3137
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019417800771191196,
+      "loss": 1.1586,
+      "step": 3138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019417329117777605,
+      "loss": 1.0617,
+      "step": 3139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019416857279125428,
+      "loss": 1.0549,
+      "step": 3140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019416385255243944,
+      "loss": 0.8782,
+      "step": 3141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019415913046142441,
+      "loss": 1.098,
+      "step": 3142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019415440651830208,
+      "loss": 1.0449,
+      "step": 3143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019414968072316535,
+      "loss": 0.9548,
+      "step": 3144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019414495307610717,
+      "loss": 1.1002,
+      "step": 3145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019414022357722055,
+      "loss": 0.9214,
+      "step": 3146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001941354922265985,
+      "loss": 0.9812,
+      "step": 3147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019413075902433412,
+      "loss": 1.0302,
+      "step": 3148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019412602397052048,
+      "loss": 0.9236,
+      "step": 3149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019412128706525073,
+      "loss": 1.1076,
+      "step": 3150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019411654830861803,
+      "loss": 0.9288,
+      "step": 3151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019411180770071562,
+      "loss": 1.3028,
+      "step": 3152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019410706524163672,
+      "loss": 0.9974,
+      "step": 3153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019410232093147463,
+      "loss": 1.044,
+      "step": 3154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019409757477032265,
+      "loss": 0.9997,
+      "step": 3155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019409282675827418,
+      "loss": 1.0053,
+      "step": 3156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019408807689542257,
+      "loss": 1.4024,
+      "step": 3157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019408332518186126,
+      "loss": 1.3905,
+      "step": 3158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019407857161768372,
+      "loss": 1.1424,
+      "step": 3159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019407381620298347,
+      "loss": 1.2491,
+      "step": 3160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019406905893785403,
+      "loss": 1.0955,
+      "step": 3161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019406429982238895,
+      "loss": 1.0234,
+      "step": 3162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001940595388566819,
+      "loss": 1.0331,
+      "step": 3163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001940547760408265,
+      "loss": 1.0565,
+      "step": 3164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019405001137491642,
+      "loss": 0.9681,
+      "step": 3165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001940452448590454,
+      "loss": 0.9899,
+      "step": 3166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001940404764933072,
+      "loss": 1.1157,
+      "step": 3167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001940357062777956,
+      "loss": 0.9985,
+      "step": 3168
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019403093421260442,
+      "loss": 1.1515,
+      "step": 3169
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019402616029782753,
+      "loss": 1.1174,
+      "step": 3170
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019402138453355886,
+      "loss": 1.1776,
+      "step": 3171
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001940166069198923,
+      "loss": 1.123,
+      "step": 3172
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001940118274569219,
+      "loss": 0.989,
+      "step": 3173
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019400704614474162,
+      "loss": 1.2552,
+      "step": 3174
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019400226298344552,
+      "loss": 1.0276,
+      "step": 3175
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001939974779731277,
+      "loss": 1.2489,
+      "step": 3176
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019399269111388222,
+      "loss": 1.381,
+      "step": 3177
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019398790240580333,
+      "loss": 1.1361,
+      "step": 3178
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019398311184898513,
+      "loss": 1.0054,
+      "step": 3179
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001939783194435219,
+      "loss": 1.0706,
+      "step": 3180
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019397352518950794,
+      "loss": 1.0324,
+      "step": 3181
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019396872908703747,
+      "loss": 1.1683,
+      "step": 3182
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019396393113620494,
+      "loss": 0.808,
+      "step": 3183
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019395913133710458,
+      "loss": 0.9972,
+      "step": 3184
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019395432968983092,
+      "loss": 1.1,
+      "step": 3185
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019394952619447838,
+      "loss": 1.0338,
+      "step": 3186
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001939447208511414,
+      "loss": 1.0298,
+      "step": 3187
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019393991365991453,
+      "loss": 1.1316,
+      "step": 3188
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019393510462089236,
+      "loss": 1.0261,
+      "step": 3189
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019393029373416946,
+      "loss": 0.962,
+      "step": 3190
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019392548099984045,
+      "loss": 1.1682,
+      "step": 3191
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193920666418,
+      "loss": 1.2414,
+      "step": 3192
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001939158499887428,
+      "loss": 1.0261,
+      "step": 3193
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001939110317121636,
+      "loss": 0.9623,
+      "step": 3194
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001939062115883572,
+      "loss": 1.0663,
+      "step": 3195
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001939013896174184,
+      "loss": 1.0183,
+      "step": 3196
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193896565799442,
+      "loss": 1.1063,
+      "step": 3197
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019389174013452297,
+      "loss": 1.1022,
+      "step": 3198
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019388691262275614,
+      "loss": 1.1675,
+      "step": 3199
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001938820832642365,
+      "loss": 1.0594,
+      "step": 3200
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001938772520590591,
+      "loss": 1.1844,
+      "step": 3201
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019387241900731893,
+      "loss": 1.0231,
+      "step": 3202
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193867584109111,
+      "loss": 1.1456,
+      "step": 3203
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019386274736453053,
+      "loss": 0.9682,
+      "step": 3204
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019385790877367255,
+      "loss": 1.0205,
+      "step": 3205
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019385306833663227,
+      "loss": 1.0585,
+      "step": 3206
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001938482260535049,
+      "loss": 1.2397,
+      "step": 3207
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019384338192438572,
+      "loss": 0.9706,
+      "step": 3208
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019383853594936995,
+      "loss": 0.9026,
+      "step": 3209
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193833688128553,
+      "loss": 0.9595,
+      "step": 3210
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019382883846203014,
+      "loss": 1.2439,
+      "step": 3211
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001938239869498968,
+      "loss": 0.7961,
+      "step": 3212
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019381913359224842,
+      "loss": 1.0543,
+      "step": 3213
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019381427838918044,
+      "loss": 1.2971,
+      "step": 3214
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019380942134078838,
+      "loss": 1.0526,
+      "step": 3215
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019380456244716778,
+      "loss": 1.1107,
+      "step": 3216
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001937997017084142,
+      "loss": 0.9433,
+      "step": 3217
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019379483912462326,
+      "loss": 1.179,
+      "step": 3218
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001937899746958906,
+      "loss": 1.0684,
+      "step": 3219
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001937851084223119,
+      "loss": 0.8775,
+      "step": 3220
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001937802403039829,
+      "loss": 0.9254,
+      "step": 3221
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019377537034099933,
+      "loss": 1.0969,
+      "step": 3222
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193770498533457,
+      "loss": 1.4992,
+      "step": 3223
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019376562488145174,
+      "loss": 1.2233,
+      "step": 3224
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001937607493850794,
+      "loss": 0.9677,
+      "step": 3225
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001937558720444359,
+      "loss": 1.2325,
+      "step": 3226
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019375099285961716,
+      "loss": 1.0338,
+      "step": 3227
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019374611183071912,
+      "loss": 1.1687,
+      "step": 3228
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001937412289578379,
+      "loss": 1.1672,
+      "step": 3229
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019373634424106942,
+      "loss": 1.0172,
+      "step": 3230
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019373145768050983,
+      "loss": 1.0704,
+      "step": 3231
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019372656927625523,
+      "loss": 1.2332,
+      "step": 3232
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019372167902840178,
+      "loss": 0.8643,
+      "step": 3233
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019371678693704566,
+      "loss": 1.0344,
+      "step": 3234
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019371189300228315,
+      "loss": 0.9428,
+      "step": 3235
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019370699722421044,
+      "loss": 1.2809,
+      "step": 3236
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019370209960292386,
+      "loss": 1.0118,
+      "step": 3237
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019369720013851977,
+      "loss": 1.1239,
+      "step": 3238
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001936922988310945,
+      "loss": 1.2706,
+      "step": 3239
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019368739568074447,
+      "loss": 1.1117,
+      "step": 3240
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019368249068756613,
+      "loss": 0.9411,
+      "step": 3241
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019367758385165602,
+      "loss": 0.9558,
+      "step": 3242
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019367267517311057,
+      "loss": 0.971,
+      "step": 3243
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019366776465202634,
+      "loss": 1.2657,
+      "step": 3244
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019366285228849996,
+      "loss": 0.8712,
+      "step": 3245
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019365793808262804,
+      "loss": 1.0614,
+      "step": 3246
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019365302203450723,
+      "loss": 1.194,
+      "step": 3247
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019364810414423427,
+      "loss": 0.9362,
+      "step": 3248
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019364318441190585,
+      "loss": 1.0819,
+      "step": 3249
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019363826283761877,
+      "loss": 1.2976,
+      "step": 3250
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019363333942146982,
+      "loss": 1.1219,
+      "step": 3251
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019362841416355585,
+      "loss": 1.2564,
+      "step": 3252
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019362348706397373,
+      "loss": 0.9078,
+      "step": 3253
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001936185581228204,
+      "loss": 1.3947,
+      "step": 3254
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019361362734019276,
+      "loss": 1.1055,
+      "step": 3255
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019360869471618788,
+      "loss": 1.2748,
+      "step": 3256
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001936037602509027,
+      "loss": 0.9861,
+      "step": 3257
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019359882394443434,
+      "loss": 0.9887,
+      "step": 3258
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019359388579687987,
+      "loss": 1.1846,
+      "step": 3259
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019358894580833643,
+      "loss": 1.3532,
+      "step": 3260
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001935840039789012,
+      "loss": 1.0369,
+      "step": 3261
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019357906030867137,
+      "loss": 0.9901,
+      "step": 3262
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019357411479774416,
+      "loss": 1.0986,
+      "step": 3263
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001935691674462169,
+      "loss": 1.0461,
+      "step": 3264
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019356421825418688,
+      "loss": 1.145,
+      "step": 3265
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019355926722175145,
+      "loss": 1.024,
+      "step": 3266
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193554314349008,
+      "loss": 0.9978,
+      "step": 3267
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019354935963605393,
+      "loss": 1.2791,
+      "step": 3268
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019354440308298675,
+      "loss": 1.0349,
+      "step": 3269
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019353944468990388,
+      "loss": 1.1013,
+      "step": 3270
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019353448445690293,
+      "loss": 1.2151,
+      "step": 3271
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019352952238408145,
+      "loss": 1.1366,
+      "step": 3272
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019352455847153696,
+      "loss": 0.9303,
+      "step": 3273
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019351959271936723,
+      "loss": 0.8739,
+      "step": 3274
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019351462512766983,
+      "loss": 1.0588,
+      "step": 3275
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019350965569654255,
+      "loss": 1.0969,
+      "step": 3276
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019350468442608305,
+      "loss": 1.2432,
+      "step": 3277
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019349971131638923,
+      "loss": 1.0735,
+      "step": 3278
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001934947363675588,
+      "loss": 1.1255,
+      "step": 3279
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001934897595796897,
+      "loss": 0.9664,
+      "step": 3280
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001934847809528798,
+      "loss": 1.1129,
+      "step": 3281
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019347980048722697,
+      "loss": 1.3137,
+      "step": 3282
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019347481818282925,
+      "loss": 1.1042,
+      "step": 3283
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001934698340397846,
+      "loss": 1.0401,
+      "step": 3284
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001934648480581911,
+      "loss": 0.9238,
+      "step": 3285
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001934598602381468,
+      "loss": 1.0563,
+      "step": 3286
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019345487057974978,
+      "loss": 0.9615,
+      "step": 3287
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019344987908309825,
+      "loss": 1.1663,
+      "step": 3288
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019344488574829033,
+      "loss": 1.0969,
+      "step": 3289
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019343989057542428,
+      "loss": 1.2487,
+      "step": 3290
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019343489356459835,
+      "loss": 1.0021,
+      "step": 3291
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019342989471591082,
+      "loss": 1.0743,
+      "step": 3292
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019342489402945998,
+      "loss": 1.1659,
+      "step": 3293
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019341989150534428,
+      "loss": 0.753,
+      "step": 3294
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019341488714366207,
+      "loss": 1.1816,
+      "step": 3295
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019340988094451175,
+      "loss": 1.0079,
+      "step": 3296
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019340487290799184,
+      "loss": 1.1285,
+      "step": 3297
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019339986303420088,
+      "loss": 0.9395,
+      "step": 3298
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019339485132323733,
+      "loss": 1.2377,
+      "step": 3299
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019338983777519983,
+      "loss": 0.9995,
+      "step": 3300
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193384822390187,
+      "loss": 1.064,
+      "step": 3301
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019337980516829744,
+      "loss": 1.2125,
+      "step": 3302
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019337478610962987,
+      "loss": 1.1654,
+      "step": 3303
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193369765214283,
+      "loss": 1.0767,
+      "step": 3304
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019336474248235564,
+      "loss": 1.1386,
+      "step": 3305
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019335971791394654,
+      "loss": 1.068,
+      "step": 3306
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019335469150915454,
+      "loss": 1.2616,
+      "step": 3307
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001933496632680785,
+      "loss": 0.9367,
+      "step": 3308
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019334463319081735,
+      "loss": 1.3378,
+      "step": 3309
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019333960127747,
+      "loss": 1.2015,
+      "step": 3310
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001933345675281355,
+      "loss": 0.7823,
+      "step": 3311
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001933295319429128,
+      "loss": 1.0613,
+      "step": 3312
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019332449452190095,
+      "loss": 1.1286,
+      "step": 3313
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019331945526519903,
+      "loss": 1.2904,
+      "step": 3314
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019331441417290622,
+      "loss": 1.0132,
+      "step": 3315
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001933093712451216,
+      "loss": 1.0304,
+      "step": 3316
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019330432648194444,
+      "loss": 1.3161,
+      "step": 3317
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932992798834739,
+      "loss": 0.9834,
+      "step": 3318
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932942314498093,
+      "loss": 1.0741,
+      "step": 3319
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932891811810499,
+      "loss": 1.0249,
+      "step": 3320
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932841290772951,
+      "loss": 1.0268,
+      "step": 3321
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019327907513864417,
+      "loss": 0.9494,
+      "step": 3322
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019327401936519665,
+      "loss": 1.1677,
+      "step": 3323
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932689617570519,
+      "loss": 1.1976,
+      "step": 3324
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019326390231430942,
+      "loss": 1.0778,
+      "step": 3325
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019325884103706874,
+      "loss": 0.9385,
+      "step": 3326
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932537779254294,
+      "loss": 0.9595,
+      "step": 3327
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019324871297949104,
+      "loss": 1.1459,
+      "step": 3328
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932436461993532,
+      "loss": 1.2155,
+      "step": 3329
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019323857758511562,
+      "loss": 1.054,
+      "step": 3330
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019323350713687797,
+      "loss": 1.0318,
+      "step": 3331
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019322843485473999,
+      "loss": 1.0692,
+      "step": 3332
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019322336073880142,
+      "loss": 1.0698,
+      "step": 3333
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019321828478916216,
+      "loss": 1.2024,
+      "step": 3334
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019321320700592192,
+      "loss": 1.0488,
+      "step": 3335
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001932081273891807,
+      "loss": 0.9837,
+      "step": 3336
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019320304593903833,
+      "loss": 0.888,
+      "step": 3337
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019319796265559482,
+      "loss": 1.1512,
+      "step": 3338
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019319287753895013,
+      "loss": 1.0278,
+      "step": 3339
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019318779058920428,
+      "loss": 1.0185,
+      "step": 3340
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019318270180645737,
+      "loss": 0.8386,
+      "step": 3341
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019317761119080944,
+      "loss": 1.0344,
+      "step": 3342
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019317251874236067,
+      "loss": 1.3591,
+      "step": 3343
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0001931674244612112,
+      "loss": 1.1399,
+      "step": 3344
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019316232834746123,
+      "loss": 1.1435,
+      "step": 3345
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193157230401211,
+      "loss": 0.8974,
+      "step": 3346
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.00019315213062256085,
+      "loss": 1.2323,
+      "step": 3347
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.000193147029011611,
+      "loss": 1.0872,
+      "step": 3348
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019314192556846185,
+      "loss": 1.2184,
+      "step": 3349
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019313682029321375,
+      "loss": 1.1262,
+      "step": 3350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019313171318596716,
+      "loss": 0.9231,
+      "step": 3351
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019312660424682253,
+      "loss": 1.0545,
+      "step": 3352
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019312149347588037,
+      "loss": 0.743,
+      "step": 3353
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019311638087324116,
+      "loss": 0.9555,
+      "step": 3354
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019311126643900548,
+      "loss": 0.9296,
+      "step": 3355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019310615017327393,
+      "loss": 0.9364,
+      "step": 3356
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019310103207614718,
+      "loss": 0.9483,
+      "step": 3357
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019309591214772587,
+      "loss": 0.911,
+      "step": 3358
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019309079038811068,
+      "loss": 0.9762,
+      "step": 3359
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019308566679740246,
+      "loss": 1.0378,
+      "step": 3360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019308054137570188,
+      "loss": 1.2083,
+      "step": 3361
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001930754141231098,
+      "loss": 1.1766,
+      "step": 3362
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019307028503972708,
+      "loss": 0.9725,
+      "step": 3363
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001930651541256546,
+      "loss": 0.9413,
+      "step": 3364
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019306002138099327,
+      "loss": 1.105,
+      "step": 3365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001930548868058441,
+      "loss": 1.1699,
+      "step": 3366
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019304975040030804,
+      "loss": 0.8798,
+      "step": 3367
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019304461216448614,
+      "loss": 1.2345,
+      "step": 3368
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019303947209847944,
+      "loss": 1.1389,
+      "step": 3369
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001930343302023891,
+      "loss": 1.1078,
+      "step": 3370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019302918647631624,
+      "loss": 0.9628,
+      "step": 3371
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000193024040920362,
+      "loss": 1.1631,
+      "step": 3372
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019301889353462762,
+      "loss": 1.0682,
+      "step": 3373
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019301374431921438,
+      "loss": 1.218,
+      "step": 3374
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019300859327422352,
+      "loss": 0.9965,
+      "step": 3375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019300344039975637,
+      "loss": 1.1497,
+      "step": 3376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019299828569591432,
+      "loss": 1.1021,
+      "step": 3377
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001929931291627987,
+      "loss": 1.2323,
+      "step": 3378
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019298797080051098,
+      "loss": 0.9602,
+      "step": 3379
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019298281060915265,
+      "loss": 0.9858,
+      "step": 3380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019297764858882514,
+      "loss": 0.9724,
+      "step": 3381
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019297248473963005,
+      "loss": 1.119,
+      "step": 3382
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019296731906166893,
+      "loss": 1.0302,
+      "step": 3383
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001929621515550434,
+      "loss": 0.9616,
+      "step": 3384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019295698221985508,
+      "loss": 1.0268,
+      "step": 3385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019295181105620567,
+      "loss": 0.8835,
+      "step": 3386
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019294663806419688,
+      "loss": 1.2599,
+      "step": 3387
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019294146324393046,
+      "loss": 0.9037,
+      "step": 3388
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001929362865955082,
+      "loss": 1.0634,
+      "step": 3389
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019293110811903193,
+      "loss": 1.0832,
+      "step": 3390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019292592781460353,
+      "loss": 0.9095,
+      "step": 3391
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019292074568232482,
+      "loss": 1.1898,
+      "step": 3392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019291556172229785,
+      "loss": 1.1723,
+      "step": 3393
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019291037593462446,
+      "loss": 0.9593,
+      "step": 3394
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019290518831940675,
+      "loss": 1.0763,
+      "step": 3395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019289999887674674,
+      "loss": 1.1767,
+      "step": 3396
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001928948076067465,
+      "loss": 1.1059,
+      "step": 3397
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001928896145095081,
+      "loss": 1.0373,
+      "step": 3398
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019288441958513375,
+      "loss": 1.1707,
+      "step": 3399
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001928792228337256,
+      "loss": 1.3437,
+      "step": 3400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019287402425538586,
+      "loss": 1.117,
+      "step": 3401
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019286882385021684,
+      "loss": 0.9623,
+      "step": 3402
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019286362161832082,
+      "loss": 1.1176,
+      "step": 3403
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019285841755980007,
+      "loss": 1.1632,
+      "step": 3404
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019285321167475698,
+      "loss": 1.1405,
+      "step": 3405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019284800396329396,
+      "loss": 1.1238,
+      "step": 3406
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019284279442551347,
+      "loss": 0.926,
+      "step": 3407
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019283758306151794,
+      "loss": 0.954,
+      "step": 3408
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019283236987140988,
+      "loss": 1.1913,
+      "step": 3409
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019282715485529185,
+      "loss": 1.103,
+      "step": 3410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019282193801326646,
+      "loss": 1.1798,
+      "step": 3411
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019281671934543622,
+      "loss": 1.0828,
+      "step": 3412
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001928114988519039,
+      "loss": 1.1542,
+      "step": 3413
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019280627653277214,
+      "loss": 0.9943,
+      "step": 3414
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019280105238814367,
+      "loss": 1.038,
+      "step": 3415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001927958264181212,
+      "loss": 1.1066,
+      "step": 3416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019279059862280758,
+      "loss": 1.0384,
+      "step": 3417
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019278536900230563,
+      "loss": 1.1679,
+      "step": 3418
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001927801375567182,
+      "loss": 1.0196,
+      "step": 3419
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001927749042861482,
+      "loss": 1.0261,
+      "step": 3420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019276966919069862,
+      "loss": 1.049,
+      "step": 3421
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019276443227047235,
+      "loss": 1.0267,
+      "step": 3422
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019275919352557241,
+      "loss": 0.894,
+      "step": 3423
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001927539529561019,
+      "loss": 1.0335,
+      "step": 3424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019274871056216387,
+      "loss": 1.1558,
+      "step": 3425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019274346634386147,
+      "loss": 1.2244,
+      "step": 3426
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001927382203012978,
+      "loss": 1.0601,
+      "step": 3427
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019273297243457607,
+      "loss": 0.9127,
+      "step": 3428
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001927277227437995,
+      "loss": 1.0497,
+      "step": 3429
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019272247122907138,
+      "loss": 1.167,
+      "step": 3430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000192717217890495,
+      "loss": 0.9464,
+      "step": 3431
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019271196272817366,
+      "loss": 1.0,
+      "step": 3432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019270670574221077,
+      "loss": 1.3865,
+      "step": 3433
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001927014469327097,
+      "loss": 1.0585,
+      "step": 3434
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001926961862997739,
+      "loss": 1.1276,
+      "step": 3435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019269092384350688,
+      "loss": 1.1293,
+      "step": 3436
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019268565956401208,
+      "loss": 0.9249,
+      "step": 3437
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019268039346139314,
+      "loss": 1.2208,
+      "step": 3438
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019267512553575357,
+      "loss": 1.0113,
+      "step": 3439
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019266985578719702,
+      "loss": 1.2372,
+      "step": 3440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019266458421582713,
+      "loss": 0.9451,
+      "step": 3441
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019265931082174765,
+      "loss": 0.867,
+      "step": 3442
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019265403560506222,
+      "loss": 1.2371,
+      "step": 3443
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019264875856587468,
+      "loss": 1.0976,
+      "step": 3444
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019264347970428876,
+      "loss": 1.248,
+      "step": 3445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019263819902040833,
+      "loss": 1.2222,
+      "step": 3446
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019263291651433728,
+      "loss": 1.2721,
+      "step": 3447
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001926276321861795,
+      "loss": 0.8618,
+      "step": 3448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019262234603603892,
+      "loss": 1.0678,
+      "step": 3449
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019261705806401954,
+      "loss": 0.9112,
+      "step": 3450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019261176827022533,
+      "loss": 1.0771,
+      "step": 3451
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001926064766547604,
+      "loss": 1.1733,
+      "step": 3452
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001926011832177288,
+      "loss": 1.0045,
+      "step": 3453
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019259588795923467,
+      "loss": 1.0126,
+      "step": 3454
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019259059087938215,
+      "loss": 1.3081,
+      "step": 3455
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019258529197827543,
+      "loss": 1.1695,
+      "step": 3456
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019257999125601877,
+      "loss": 1.0787,
+      "step": 3457
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001925746887127164,
+      "loss": 1.0303,
+      "step": 3458
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019256938434847264,
+      "loss": 0.9108,
+      "step": 3459
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019256407816339182,
+      "loss": 1.001,
+      "step": 3460
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019255877015757835,
+      "loss": 0.9985,
+      "step": 3461
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019255346033113658,
+      "loss": 1.0861,
+      "step": 3462
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019254814868417097,
+      "loss": 1.1675,
+      "step": 3463
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000192542835216786,
+      "loss": 1.1765,
+      "step": 3464
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019253751992908622,
+      "loss": 1.2458,
+      "step": 3465
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019253220282117615,
+      "loss": 0.9418,
+      "step": 3466
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019252688389316037,
+      "loss": 1.0547,
+      "step": 3467
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019252156314514352,
+      "loss": 0.9058,
+      "step": 3468
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019251624057723023,
+      "loss": 0.9839,
+      "step": 3469
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019251091618952527,
+      "loss": 1.207,
+      "step": 3470
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001925055899821333,
+      "loss": 1.0121,
+      "step": 3471
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019250026195515908,
+      "loss": 1.04,
+      "step": 3472
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019249493210870747,
+      "loss": 1.1598,
+      "step": 3473
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019248960044288327,
+      "loss": 1.0207,
+      "step": 3474
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019248426695779136,
+      "loss": 0.9956,
+      "step": 3475
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019247893165353665,
+      "loss": 1.0593,
+      "step": 3476
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019247359453022407,
+      "loss": 1.1533,
+      "step": 3477
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019246825558795862,
+      "loss": 1.0361,
+      "step": 3478
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019246291482684534,
+      "loss": 1.1742,
+      "step": 3479
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019245757224698923,
+      "loss": 0.8458,
+      "step": 3480
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019245222784849538,
+      "loss": 1.0892,
+      "step": 3481
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019244688163146898,
+      "loss": 1.0935,
+      "step": 3482
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019244153359601515,
+      "loss": 1.0084,
+      "step": 3483
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019243618374223907,
+      "loss": 0.9697,
+      "step": 3484
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019243083207024598,
+      "loss": 1.3763,
+      "step": 3485
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019242547858014115,
+      "loss": 1.2348,
+      "step": 3486
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001924201232720299,
+      "loss": 0.9639,
+      "step": 3487
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019241476614601752,
+      "loss": 1.2434,
+      "step": 3488
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019240940720220945,
+      "loss": 1.1331,
+      "step": 3489
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019240404644071108,
+      "loss": 1.2474,
+      "step": 3490
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019239868386162782,
+      "loss": 1.0179,
+      "step": 3491
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923933194650652,
+      "loss": 0.9327,
+      "step": 3492
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923879532511287,
+      "loss": 1.0723,
+      "step": 3493
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019238258521992387,
+      "loss": 1.161,
+      "step": 3494
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019237721537155634,
+      "loss": 0.9519,
+      "step": 3495
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923718437061317,
+      "loss": 1.0507,
+      "step": 3496
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019236647022375563,
+      "loss": 1.2682,
+      "step": 3497
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923610949245338,
+      "loss": 1.3523,
+      "step": 3498
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019235571780857198,
+      "loss": 0.9773,
+      "step": 3499
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923503388759759,
+      "loss": 1.2381,
+      "step": 3500
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923449581268514,
+      "loss": 0.9945,
+      "step": 3501
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019233957556130432,
+      "loss": 1.3687,
+      "step": 3502
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019233419117944048,
+      "loss": 0.9664,
+      "step": 3503
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019232880498136582,
+      "loss": 0.8872,
+      "step": 3504
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923234169671863,
+      "loss": 1.0232,
+      "step": 3505
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001923180271370079,
+      "loss": 1.1626,
+      "step": 3506
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019231263549093663,
+      "loss": 1.1492,
+      "step": 3507
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019230724202907854,
+      "loss": 1.0804,
+      "step": 3508
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019230184675153976,
+      "loss": 1.4227,
+      "step": 3509
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019229644965842633,
+      "loss": 1.1893,
+      "step": 3510
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001922910507498445,
+      "loss": 1.1122,
+      "step": 3511
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001922856500259004,
+      "loss": 1.1446,
+      "step": 3512
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001922802474867003,
+      "loss": 1.2888,
+      "step": 3513
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019227484313235046,
+      "loss": 1.0737,
+      "step": 3514
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019226943696295715,
+      "loss": 1.2192,
+      "step": 3515
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019226402897862678,
+      "loss": 0.9988,
+      "step": 3516
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019225861917946565,
+      "loss": 1.1875,
+      "step": 3517
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019225320756558023,
+      "loss": 1.0538,
+      "step": 3518
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019224779413707692,
+      "loss": 0.9988,
+      "step": 3519
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019224237889406225,
+      "loss": 1.1432,
+      "step": 3520
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019223696183664267,
+      "loss": 0.98,
+      "step": 3521
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019223154296492482,
+      "loss": 1.0505,
+      "step": 3522
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019222612227901524,
+      "loss": 1.0112,
+      "step": 3523
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019222069977902054,
+      "loss": 0.9899,
+      "step": 3524
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001922152754650474,
+      "loss": 1.2479,
+      "step": 3525
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019220984933720253,
+      "loss": 1.0494,
+      "step": 3526
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001922044213955926,
+      "loss": 0.9639,
+      "step": 3527
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019219899164032447,
+      "loss": 1.1788,
+      "step": 3528
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001921935600715049,
+      "loss": 1.1002,
+      "step": 3529
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019218812668924068,
+      "loss": 0.8854,
+      "step": 3530
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019218269149363874,
+      "loss": 1.1905,
+      "step": 3531
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.000192177254484806,
+      "loss": 1.0287,
+      "step": 3532
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001921718156628494,
+      "loss": 1.0934,
+      "step": 3533
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019216637502787588,
+      "loss": 1.1604,
+      "step": 3534
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001921609325799925,
+      "loss": 1.0708,
+      "step": 3535
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001921554883193063,
+      "loss": 0.9629,
+      "step": 3536
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019215004224592435,
+      "loss": 0.9944,
+      "step": 3537
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001921445943599538,
+      "loss": 1.0767,
+      "step": 3538
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019213914466150177,
+      "loss": 1.1802,
+      "step": 3539
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019213369315067554,
+      "loss": 1.0886,
+      "step": 3540
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019212823982758226,
+      "loss": 1.2348,
+      "step": 3541
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001921227846923292,
+      "loss": 0.8661,
+      "step": 3542
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019211732774502372,
+      "loss": 0.7645,
+      "step": 3543
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019211186898577311,
+      "loss": 1.0524,
+      "step": 3544
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019210640841468477,
+      "loss": 1.102,
+      "step": 3545
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019210094603186608,
+      "loss": 1.1881,
+      "step": 3546
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001920954818374245,
+      "loss": 1.1966,
+      "step": 3547
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019209001583146752,
+      "loss": 0.9306,
+      "step": 3548
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019208454801410266,
+      "loss": 0.9701,
+      "step": 3549
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001920790783854374,
+      "loss": 1.0218,
+      "step": 3550
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019207360694557942,
+      "loss": 1.1059,
+      "step": 3551
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019206813369463635,
+      "loss": 0.9373,
+      "step": 3552
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019206265863271578,
+      "loss": 1.087,
+      "step": 3553
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019205718175992542,
+      "loss": 1.1824,
+      "step": 3554
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019205170307637302,
+      "loss": 1.0135,
+      "step": 3555
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019204622258216635,
+      "loss": 1.1829,
+      "step": 3556
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001920407402774132,
+      "loss": 1.1553,
+      "step": 3557
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001920352561622214,
+      "loss": 1.1802,
+      "step": 3558
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001920297702366988,
+      "loss": 1.1288,
+      "step": 3559
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019202428250095338,
+      "loss": 1.1667,
+      "step": 3560
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019201879295509302,
+      "loss": 0.9108,
+      "step": 3561
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019201330159922573,
+      "loss": 1.0754,
+      "step": 3562
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001920078084334595,
+      "loss": 1.085,
+      "step": 3563
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001920023134579024,
+      "loss": 0.9696,
+      "step": 3564
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019199681667266252,
+      "loss": 0.9295,
+      "step": 3565
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019199131807784797,
+      "loss": 0.9716,
+      "step": 3566
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001919858176735669,
+      "loss": 1.1343,
+      "step": 3567
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001919803154599275,
+      "loss": 0.9887,
+      "step": 3568
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019197481143703803,
+      "loss": 1.1349,
+      "step": 3569
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019196930560500675,
+      "loss": 1.1201,
+      "step": 3570
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019196379796394193,
+      "loss": 1.1435,
+      "step": 3571
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019195828851395193,
+      "loss": 1.1272,
+      "step": 3572
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001919527772551451,
+      "loss": 0.9556,
+      "step": 3573
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019194726418762982,
+      "loss": 1.0518,
+      "step": 3574
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001919417493115146,
+      "loss": 1.0788,
+      "step": 3575
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.0001919362326269079,
+      "loss": 0.8673,
+      "step": 3576
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019193071413391822,
+      "loss": 1.0915,
+      "step": 3577
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019192519383265407,
+      "loss": 1.0601,
+      "step": 3578
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 0.00019191967172322409,
+      "loss": 1.0234,
+      "step": 3579
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001919141478057369,
+      "loss": 1.0636,
+      "step": 3580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001919086220803011,
+      "loss": 1.1564,
+      "step": 3581
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019190309454702545,
+      "loss": 1.0654,
+      "step": 3582
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019189756520601863,
+      "loss": 1.0644,
+      "step": 3583
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019189203405738943,
+      "loss": 0.931,
+      "step": 3584
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019188650110124662,
+      "loss": 1.0471,
+      "step": 3585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019188096633769905,
+      "loss": 1.2935,
+      "step": 3586
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001918754297668556,
+      "loss": 1.1156,
+      "step": 3587
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019186989138882514,
+      "loss": 1.0768,
+      "step": 3588
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001918643512037166,
+      "loss": 0.9582,
+      "step": 3589
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019185880921163904,
+      "loss": 0.8986,
+      "step": 3590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001918532654127014,
+      "loss": 1.0374,
+      "step": 3591
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019184771980701272,
+      "loss": 1.0434,
+      "step": 3592
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019184217239468212,
+      "loss": 1.1323,
+      "step": 3593
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001918366231758187,
+      "loss": 1.0464,
+      "step": 3594
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001918310721505316,
+      "loss": 0.9464,
+      "step": 3595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019182551931893005,
+      "loss": 1.11,
+      "step": 3596
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019181996468112323,
+      "loss": 0.9234,
+      "step": 3597
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001918144082372204,
+      "loss": 1.0109,
+      "step": 3598
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001918088499873309,
+      "loss": 0.9193,
+      "step": 3599
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000191803289931564,
+      "loss": 1.1888,
+      "step": 3600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001917977280700291,
+      "loss": 1.1052,
+      "step": 3601
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019179216440283562,
+      "loss": 0.9446,
+      "step": 3602
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000191786598930093,
+      "loss": 1.0071,
+      "step": 3603
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019178103165191064,
+      "loss": 0.8279,
+      "step": 3604
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019177546256839812,
+      "loss": 1.18,
+      "step": 3605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019176989167966495,
+      "loss": 1.0536,
+      "step": 3606
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019176431898582075,
+      "loss": 1.0678,
+      "step": 3607
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019175874448697507,
+      "loss": 1.0355,
+      "step": 3608
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019175316818323762,
+      "loss": 1.0472,
+      "step": 3609
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019174759007471804,
+      "loss": 0.9899,
+      "step": 3610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001917420101615261,
+      "loss": 0.9249,
+      "step": 3611
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019173642844377148,
+      "loss": 1.1306,
+      "step": 3612
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019173084492156407,
+      "loss": 1.2015,
+      "step": 3613
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019172525959501363,
+      "loss": 1.1156,
+      "step": 3614
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019171967246423005,
+      "loss": 1.0886,
+      "step": 3615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019171408352932322,
+      "loss": 1.1216,
+      "step": 3616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019170849279040306,
+      "loss": 0.9048,
+      "step": 3617
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019170290024757956,
+      "loss": 1.0172,
+      "step": 3618
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019169730590096271,
+      "loss": 0.9045,
+      "step": 3619
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019169170975066258,
+      "loss": 0.9991,
+      "step": 3620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019168611179678924,
+      "loss": 1.1543,
+      "step": 3621
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019168051203945273,
+      "loss": 1.1404,
+      "step": 3622
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019167491047876331,
+      "loss": 1.1526,
+      "step": 3623
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001916693071148311,
+      "loss": 0.8053,
+      "step": 3624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019166370194776634,
+      "loss": 1.1311,
+      "step": 3625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019165809497767922,
+      "loss": 0.9498,
+      "step": 3626
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019165248620468011,
+      "loss": 0.9231,
+      "step": 3627
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019164687562887931,
+      "loss": 1.1183,
+      "step": 3628
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019164126325038714,
+      "loss": 1.0141,
+      "step": 3629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019163564906931403,
+      "loss": 0.9308,
+      "step": 3630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019163003308577046,
+      "loss": 1.172,
+      "step": 3631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019162441529986678,
+      "loss": 1.113,
+      "step": 3632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019161879571171362,
+      "loss": 1.2332,
+      "step": 3633
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019161317432142138,
+      "loss": 0.9738,
+      "step": 3634
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019160755112910077,
+      "loss": 1.1249,
+      "step": 3635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001916019261348623,
+      "loss": 1.1227,
+      "step": 3636
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019159629933881666,
+      "loss": 1.1715,
+      "step": 3637
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019159067074107453,
+      "loss": 0.9508,
+      "step": 3638
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019158504034174657,
+      "loss": 1.0252,
+      "step": 3639
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001915794081409436,
+      "loss": 1.0726,
+      "step": 3640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019157377413877636,
+      "loss": 1.0266,
+      "step": 3641
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001915681383353557,
+      "loss": 1.3487,
+      "step": 3642
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001915625007307925,
+      "loss": 0.9312,
+      "step": 3643
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019155686132519753,
+      "loss": 0.9031,
+      "step": 3644
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019155122011868186,
+      "loss": 1.1078,
+      "step": 3645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001915455771113564,
+      "loss": 0.9904,
+      "step": 3646
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001915399323033321,
+      "loss": 1.1707,
+      "step": 3647
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019153428569472006,
+      "loss": 1.1884,
+      "step": 3648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001915286372856313,
+      "loss": 0.9744,
+      "step": 3649
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000191522987076177,
+      "loss": 1.1132,
+      "step": 3650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001915173350664682,
+      "loss": 0.8835,
+      "step": 3651
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019151168125661615,
+      "loss": 1.016,
+      "step": 3652
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019150602564673198,
+      "loss": 0.9704,
+      "step": 3653
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019150036823692703,
+      "loss": 1.1885,
+      "step": 3654
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019149470902731254,
+      "loss": 0.9082,
+      "step": 3655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019148904801799982,
+      "loss": 1.0277,
+      "step": 3656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001914833852091002,
+      "loss": 1.2307,
+      "step": 3657
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019147772060072512,
+      "loss": 1.0419,
+      "step": 3658
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019147205419298597,
+      "loss": 1.1602,
+      "step": 3659
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019146638598599425,
+      "loss": 1.2197,
+      "step": 3660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019146071597986138,
+      "loss": 1.1175,
+      "step": 3661
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019145504417469894,
+      "loss": 1.2447,
+      "step": 3662
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019144937057061844,
+      "loss": 0.9999,
+      "step": 3663
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019144369516773158,
+      "loss": 1.1574,
+      "step": 3664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001914380179661499,
+      "loss": 1.1827,
+      "step": 3665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019143233896598512,
+      "loss": 1.1756,
+      "step": 3666
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019142665816734893,
+      "loss": 1.0243,
+      "step": 3667
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019142097557035308,
+      "loss": 1.203,
+      "step": 3668
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019141529117510932,
+      "loss": 0.9385,
+      "step": 3669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001914096049817295,
+      "loss": 1.0146,
+      "step": 3670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019140391699032545,
+      "loss": 1.1674,
+      "step": 3671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019139822720100905,
+      "loss": 0.947,
+      "step": 3672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019139253561389221,
+      "loss": 1.1585,
+      "step": 3673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001913868422290869,
+      "loss": 1.1631,
+      "step": 3674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019138114704670512,
+      "loss": 0.8118,
+      "step": 3675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019137545006685884,
+      "loss": 1.0974,
+      "step": 3676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019136975128966017,
+      "loss": 0.8919,
+      "step": 3677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019136405071522119,
+      "loss": 1.0357,
+      "step": 3678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019135834834365403,
+      "loss": 1.0574,
+      "step": 3679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019135264417507086,
+      "loss": 1.1142,
+      "step": 3680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019134693820958387,
+      "loss": 1.118,
+      "step": 3681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019134123044730532,
+      "loss": 1.1805,
+      "step": 3682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019133552088834743,
+      "loss": 0.9393,
+      "step": 3683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019132980953282257,
+      "loss": 1.1124,
+      "step": 3684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019132409638084304,
+      "loss": 1.1625,
+      "step": 3685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019131838143252123,
+      "loss": 1.1639,
+      "step": 3686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019131266468796956,
+      "loss": 0.9899,
+      "step": 3687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019130694614730048,
+      "loss": 0.8618,
+      "step": 3688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019130122581062644,
+      "loss": 1.2562,
+      "step": 3689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019129550367806002,
+      "loss": 1.0356,
+      "step": 3690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001912897797497137,
+      "loss": 1.0979,
+      "step": 3691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001912840540257001,
+      "loss": 1.0937,
+      "step": 3692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019127832650613189,
+      "loss": 1.2577,
+      "step": 3693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001912725971911217,
+      "loss": 1.2239,
+      "step": 3694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001912668660807822,
+      "loss": 0.9505,
+      "step": 3695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019126113317522613,
+      "loss": 1.05,
+      "step": 3696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001912553984745663,
+      "loss": 1.0837,
+      "step": 3697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019124966197891542,
+      "loss": 0.9976,
+      "step": 3698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019124392368838642,
+      "loss": 0.8164,
+      "step": 3699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001912381836030921,
+      "loss": 1.0682,
+      "step": 3700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019123244172314545,
+      "loss": 1.1425,
+      "step": 3701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019122669804865934,
+      "loss": 1.1757,
+      "step": 3702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019122095257974677,
+      "loss": 0.9249,
+      "step": 3703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019121520531652076,
+      "loss": 1.1397,
+      "step": 3704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019120945625909438,
+      "loss": 1.2306,
+      "step": 3705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019120370540758066,
+      "loss": 1.0862,
+      "step": 3706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019119795276209276,
+      "loss": 1.1655,
+      "step": 3707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019119219832274383,
+      "loss": 1.08,
+      "step": 3708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019118644208964702,
+      "loss": 1.0601,
+      "step": 3709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001911806840629156,
+      "loss": 0.7653,
+      "step": 3710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001911749242426628,
+      "loss": 1.293,
+      "step": 3711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019116916262900197,
+      "loss": 1.0117,
+      "step": 3712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019116339922204637,
+      "loss": 0.9962,
+      "step": 3713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019115763402190945,
+      "loss": 1.0826,
+      "step": 3714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001911518670287045,
+      "loss": 0.9453,
+      "step": 3715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019114609824254507,
+      "loss": 1.1746,
+      "step": 3716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019114032766354453,
+      "loss": 1.1474,
+      "step": 3717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019113455529181647,
+      "loss": 1.0644,
+      "step": 3718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019112878112747435,
+      "loss": 1.1663,
+      "step": 3719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019112300517063187,
+      "loss": 1.098,
+      "step": 3720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019111722742140252,
+      "loss": 1.1332,
+      "step": 3721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019111144787990002,
+      "loss": 1.0787,
+      "step": 3722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000191105666546238,
+      "loss": 1.097,
+      "step": 3723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019109988342053025,
+      "loss": 1.0689,
+      "step": 3724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019109409850289044,
+      "loss": 1.2555,
+      "step": 3725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019108831179343242,
+      "loss": 0.9672,
+      "step": 3726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019108252329227,
+      "loss": 1.1434,
+      "step": 3727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019107673299951703,
+      "loss": 1.0462,
+      "step": 3728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019107094091528741,
+      "loss": 1.2906,
+      "step": 3729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001910651470396951,
+      "loss": 1.127,
+      "step": 3730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.000191059351372854,
+      "loss": 0.8651,
+      "step": 3731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019105355391487816,
+      "loss": 1.1355,
+      "step": 3732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019104775466588161,
+      "loss": 1.0138,
+      "step": 3733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001910419536259784,
+      "loss": 1.0916,
+      "step": 3734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019103615079528265,
+      "loss": 1.2103,
+      "step": 3735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019103034617390851,
+      "loss": 0.8793,
+      "step": 3736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019102453976197015,
+      "loss": 0.9545,
+      "step": 3737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019101873155958179,
+      "loss": 1.0524,
+      "step": 3738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019101292156685764,
+      "loss": 1.1548,
+      "step": 3739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019100710978391203,
+      "loss": 1.0344,
+      "step": 3740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019100129621085925,
+      "loss": 1.0257,
+      "step": 3741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019099548084781367,
+      "loss": 0.8489,
+      "step": 3742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019098966369488965,
+      "loss": 1.0096,
+      "step": 3743
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019098384475220166,
+      "loss": 1.1948,
+      "step": 3744
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001909780240198641,
+      "loss": 1.0115,
+      "step": 3745
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001909722014979915,
+      "loss": 1.0588,
+      "step": 3746
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019096637718669842,
+      "loss": 1.1833,
+      "step": 3747
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019096055108609936,
+      "loss": 1.1451,
+      "step": 3748
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019095472319630892,
+      "loss": 1.1566,
+      "step": 3749
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019094889351744181,
+      "loss": 1.2139,
+      "step": 3750
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001909430620496126,
+      "loss": 1.0509,
+      "step": 3751
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019093722879293609,
+      "loss": 0.9638,
+      "step": 3752
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019093139374752697,
+      "loss": 1.267,
+      "step": 3753
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019092555691349995,
+      "loss": 1.1168,
+      "step": 3754
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019091971829096998,
+      "loss": 0.9936,
+      "step": 3755
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019091387788005185,
+      "loss": 1.0552,
+      "step": 3756
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019090803568086036,
+      "loss": 1.02,
+      "step": 3757
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019090219169351052,
+      "loss": 1.0673,
+      "step": 3758
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019089634591811728,
+      "loss": 1.0526,
+      "step": 3759
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019089049835479556,
+      "loss": 1.0525,
+      "step": 3760
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001908846490036605,
+      "loss": 1.1515,
+      "step": 3761
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019087879786482698,
+      "loss": 1.0154,
+      "step": 3762
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001908729449384102,
+      "loss": 1.1082,
+      "step": 3763
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019086709022452533,
+      "loss": 1.1065,
+      "step": 3764
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019086123372328746,
+      "loss": 0.9566,
+      "step": 3765
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019085537543481177,
+      "loss": 1.0783,
+      "step": 3766
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019084951535921357,
+      "loss": 1.1013,
+      "step": 3767
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001908436534966081,
+      "loss": 0.8005,
+      "step": 3768
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001908377898471106,
+      "loss": 0.9088,
+      "step": 3769
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019083192441083648,
+      "loss": 1.2698,
+      "step": 3770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019082605718790107,
+      "loss": 1.0266,
+      "step": 3771
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019082018817841983,
+      "loss": 1.02,
+      "step": 3772
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019081431738250814,
+      "loss": 1.0688,
+      "step": 3773
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001908084448002815,
+      "loss": 1.2868,
+      "step": 3774
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019080257043185546,
+      "loss": 1.2056,
+      "step": 3775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019079669427734554,
+      "loss": 0.9385,
+      "step": 3776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019079081633686733,
+      "loss": 1.2566,
+      "step": 3777
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001907849366105364,
+      "loss": 1.1856,
+      "step": 3778
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019077905509846848,
+      "loss": 1.215,
+      "step": 3779
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019077317180077926,
+      "loss": 1.1202,
+      "step": 3780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001907672867175844,
+      "loss": 0.8339,
+      "step": 3781
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019076139984899965,
+      "loss": 1.0217,
+      "step": 3782
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001907555111951409,
+      "loss": 1.092,
+      "step": 3783
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001907496207561239,
+      "loss": 1.0364,
+      "step": 3784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019074372853206454,
+      "loss": 1.0347,
+      "step": 3785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019073783452307872,
+      "loss": 1.0112,
+      "step": 3786
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019073193872928238,
+      "loss": 1.1188,
+      "step": 3787
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001907260411507915,
+      "loss": 1.0233,
+      "step": 3788
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019072014178772202,
+      "loss": 0.966,
+      "step": 3789
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019071424064019004,
+      "loss": 0.9908,
+      "step": 3790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019070833770831166,
+      "loss": 1.0397,
+      "step": 3791
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019070243299220297,
+      "loss": 1.1309,
+      "step": 3792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019069652649198005,
+      "loss": 1.1044,
+      "step": 3793
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019069061820775916,
+      "loss": 1.1089,
+      "step": 3794
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019068470813965647,
+      "loss": 0.9981,
+      "step": 3795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019067879628778826,
+      "loss": 1.1585,
+      "step": 3796
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019067288265227082,
+      "loss": 1.1314,
+      "step": 3797
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019066696723322042,
+      "loss": 1.0074,
+      "step": 3798
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019066105003075348,
+      "loss": 1.2008,
+      "step": 3799
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019065513104498636,
+      "loss": 1.0866,
+      "step": 3800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019064921027603547,
+      "loss": 1.0531,
+      "step": 3801
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019064328772401732,
+      "loss": 0.9862,
+      "step": 3802
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019063736338904836,
+      "loss": 1.0526,
+      "step": 3803
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019063143727124512,
+      "loss": 0.9876,
+      "step": 3804
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019062550937072421,
+      "loss": 1.2359,
+      "step": 3805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0001906195796876022,
+      "loss": 1.089,
+      "step": 3806
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019061364822199573,
+      "loss": 0.9551,
+      "step": 3807
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019060771497402147,
+      "loss": 1.1679,
+      "step": 3808
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019060177994379617,
+      "loss": 1.2464,
+      "step": 3809
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00019059584313143647,
+      "loss": 1.1069,
+      "step": 3810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019058990453705926,
+      "loss": 1.0216,
+      "step": 3811
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001905839641607813,
+      "loss": 1.1221,
+      "step": 3812
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019057802200271942,
+      "loss": 1.1674,
+      "step": 3813
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019057207806299052,
+      "loss": 1.0326,
+      "step": 3814
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019056613234171151,
+      "loss": 1.0856,
+      "step": 3815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019056018483899938,
+      "loss": 1.2521,
+      "step": 3816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019055423555497107,
+      "loss": 0.8433,
+      "step": 3817
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019054828448974363,
+      "loss": 1.0907,
+      "step": 3818
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001905423316434341,
+      "loss": 0.8388,
+      "step": 3819
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019053637701615958,
+      "loss": 1.1668,
+      "step": 3820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001905304206080372,
+      "loss": 0.9996,
+      "step": 3821
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019052446241918413,
+      "loss": 0.8112,
+      "step": 3822
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019051850244971753,
+      "loss": 0.8559,
+      "step": 3823
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001905125406997547,
+      "loss": 0.9935,
+      "step": 3824
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019050657716941286,
+      "loss": 1.2031,
+      "step": 3825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001905006118588093,
+      "loss": 0.9789,
+      "step": 3826
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019049464476806138,
+      "loss": 1.184,
+      "step": 3827
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019048867589728646,
+      "loss": 1.1409,
+      "step": 3828
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019048270524660196,
+      "loss": 1.0648,
+      "step": 3829
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019047673281612532,
+      "loss": 1.04,
+      "step": 3830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019047075860597403,
+      "loss": 1.0721,
+      "step": 3831
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001904647826162656,
+      "loss": 1.0335,
+      "step": 3832
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019045880484711754,
+      "loss": 1.0348,
+      "step": 3833
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019045282529864745,
+      "loss": 1.0421,
+      "step": 3834
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019044684397097298,
+      "loss": 1.0072,
+      "step": 3835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019044086086421178,
+      "loss": 0.9617,
+      "step": 3836
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001904348759784815,
+      "loss": 0.8923,
+      "step": 3837
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019042888931389987,
+      "loss": 1.1104,
+      "step": 3838
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019042290087058463,
+      "loss": 1.2026,
+      "step": 3839
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019041691064865363,
+      "loss": 1.3371,
+      "step": 3840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019041091864822467,
+      "loss": 0.9262,
+      "step": 3841
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019040492486941562,
+      "loss": 1.0664,
+      "step": 3842
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019039892931234435,
+      "loss": 1.0855,
+      "step": 3843
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001903929319771288,
+      "loss": 0.9456,
+      "step": 3844
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019038693286388698,
+      "loss": 0.9395,
+      "step": 3845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001903809319727368,
+      "loss": 1.0367,
+      "step": 3846
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019037492930379642,
+      "loss": 1.1,
+      "step": 3847
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019036892485718382,
+      "loss": 1.1472,
+      "step": 3848
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019036291863301714,
+      "loss": 1.0607,
+      "step": 3849
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019035691063141452,
+      "loss": 1.1862,
+      "step": 3850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019035090085249413,
+      "loss": 1.0183,
+      "step": 3851
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019034488929637423,
+      "loss": 1.116,
+      "step": 3852
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019033887596317298,
+      "loss": 1.0889,
+      "step": 3853
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001903328608530087,
+      "loss": 1.1404,
+      "step": 3854
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019032684396599973,
+      "loss": 1.0582,
+      "step": 3855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019032082530226442,
+      "loss": 1.1564,
+      "step": 3856
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019031480486192114,
+      "loss": 1.023,
+      "step": 3857
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019030878264508826,
+      "loss": 1.141,
+      "step": 3858
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019030275865188436,
+      "loss": 1.1958,
+      "step": 3859
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019029673288242785,
+      "loss": 1.1649,
+      "step": 3860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019029070533683725,
+      "loss": 1.2674,
+      "step": 3861
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019028467601523114,
+      "loss": 1.0556,
+      "step": 3862
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019027864491772816,
+      "loss": 1.2323,
+      "step": 3863
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019027261204444687,
+      "loss": 0.9177,
+      "step": 3864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019026657739550594,
+      "loss": 1.0733,
+      "step": 3865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019026054097102412,
+      "loss": 1.2136,
+      "step": 3866
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019025450277112014,
+      "loss": 1.0752,
+      "step": 3867
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019024846279591274,
+      "loss": 1.0425,
+      "step": 3868
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019024242104552074,
+      "loss": 0.9462,
+      "step": 3869
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000190236377520063,
+      "loss": 0.998,
+      "step": 3870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019023033221965836,
+      "loss": 1.1545,
+      "step": 3871
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019022428514442577,
+      "loss": 1.2121,
+      "step": 3872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019021823629448414,
+      "loss": 1.1796,
+      "step": 3873
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019021218566995246,
+      "loss": 1.2843,
+      "step": 3874
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019020613327094975,
+      "loss": 1.1183,
+      "step": 3875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019020007909759508,
+      "loss": 1.0594,
+      "step": 3876
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019019402315000752,
+      "loss": 0.9878,
+      "step": 3877
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019018796542830617,
+      "loss": 1.1722,
+      "step": 3878
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001901819059326102,
+      "loss": 1.0243,
+      "step": 3879
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019017584466303884,
+      "loss": 1.2559,
+      "step": 3880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019016978161971124,
+      "loss": 1.0034,
+      "step": 3881
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019016371680274672,
+      "loss": 0.903,
+      "step": 3882
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019015765021226453,
+      "loss": 1.0971,
+      "step": 3883
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000190151581848384,
+      "loss": 1.0036,
+      "step": 3884
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019014551171122457,
+      "loss": 1.0868,
+      "step": 3885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019013943980090558,
+      "loss": 0.9927,
+      "step": 3886
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019013336611754645,
+      "loss": 1.1611,
+      "step": 3887
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019012729066126668,
+      "loss": 1.1215,
+      "step": 3888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019012121343218576,
+      "loss": 1.2652,
+      "step": 3889
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019011513443042323,
+      "loss": 1.0654,
+      "step": 3890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019010905365609868,
+      "loss": 0.9584,
+      "step": 3891
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001901029711093317,
+      "loss": 0.9866,
+      "step": 3892
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001900968867902419,
+      "loss": 0.9708,
+      "step": 3893
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019009080069894906,
+      "loss": 1.0668,
+      "step": 3894
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001900847128355728,
+      "loss": 0.8556,
+      "step": 3895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019007862320023292,
+      "loss": 0.9412,
+      "step": 3896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019007253179304914,
+      "loss": 1.1838,
+      "step": 3897
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019006643861414136,
+      "loss": 1.268,
+      "step": 3898
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019006034366362935,
+      "loss": 1.1786,
+      "step": 3899
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019005424694163304,
+      "loss": 1.1669,
+      "step": 3900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001900481484482724,
+      "loss": 1.0275,
+      "step": 3901
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019004204818366728,
+      "loss": 1.0649,
+      "step": 3902
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001900359461479378,
+      "loss": 1.1623,
+      "step": 3903
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019002984234120387,
+      "loss": 0.8938,
+      "step": 3904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001900237367635856,
+      "loss": 1.1686,
+      "step": 3905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001900176294152031,
+      "loss": 1.1663,
+      "step": 3906
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019001152029617647,
+      "loss": 1.0959,
+      "step": 3907
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00019000540940662593,
+      "loss": 1.0208,
+      "step": 3908
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001899992967466716,
+      "loss": 1.1445,
+      "step": 3909
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001899931823164338,
+      "loss": 0.9067,
+      "step": 3910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018998706611603273,
+      "loss": 1.2766,
+      "step": 3911
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018998094814558876,
+      "loss": 1.2301,
+      "step": 3912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018997482840522217,
+      "loss": 1.2578,
+      "step": 3913
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018996870689505334,
+      "loss": 1.1394,
+      "step": 3914
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018996258361520275,
+      "loss": 1.2377,
+      "step": 3915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018995645856579075,
+      "loss": 1.1853,
+      "step": 3916
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001899503317469379,
+      "loss": 1.0999,
+      "step": 3917
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018994420315876468,
+      "loss": 0.8589,
+      "step": 3918
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001899380728013916,
+      "loss": 1.0396,
+      "step": 3919
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018993194067493935,
+      "loss": 0.9446,
+      "step": 3920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018992580677952846,
+      "loss": 0.9503,
+      "step": 3921
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018991967111527957,
+      "loss": 1.328,
+      "step": 3922
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018991353368231343,
+      "loss": 1.0099,
+      "step": 3923
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018990739448075075,
+      "loss": 1.1207,
+      "step": 3924
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018990125351071226,
+      "loss": 0.9768,
+      "step": 3925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018989511077231876,
+      "loss": 1.0407,
+      "step": 3926
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018988896626569109,
+      "loss": 1.0126,
+      "step": 3927
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001898828199909501,
+      "loss": 1.0776,
+      "step": 3928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001898766719482167,
+      "loss": 0.9677,
+      "step": 3929
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018987052213761184,
+      "loss": 0.906,
+      "step": 3930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018986437055925644,
+      "loss": 1.0079,
+      "step": 3931
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018985821721327154,
+      "loss": 1.08,
+      "step": 3932
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018985206209977813,
+      "loss": 1.1877,
+      "step": 3933
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018984590521889735,
+      "loss": 1.0919,
+      "step": 3934
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018983974657075026,
+      "loss": 1.203,
+      "step": 3935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018983358615545798,
+      "loss": 1.0918,
+      "step": 3936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018982742397314172,
+      "loss": 1.1382,
+      "step": 3937
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018982126002392268,
+      "loss": 0.9889,
+      "step": 3938
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018981509430792212,
+      "loss": 0.8404,
+      "step": 3939
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018980892682526134,
+      "loss": 1.0887,
+      "step": 3940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018980275757606157,
+      "loss": 1.1264,
+      "step": 3941
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018979658656044423,
+      "loss": 1.0489,
+      "step": 3942
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018979041377853068,
+      "loss": 0.8589,
+      "step": 3943
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018978423923044234,
+      "loss": 1.0793,
+      "step": 3944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018977806291630068,
+      "loss": 0.8882,
+      "step": 3945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018977188483622714,
+      "loss": 1.1831,
+      "step": 3946
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001897657049903433,
+      "loss": 1.0331,
+      "step": 3947
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001897595233787707,
+      "loss": 1.0459,
+      "step": 3948
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018975334000163093,
+      "loss": 1.0079,
+      "step": 3949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018974715485904557,
+      "loss": 0.9865,
+      "step": 3950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018974096795113638,
+      "loss": 0.8624,
+      "step": 3951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018973477927802497,
+      "loss": 0.9748,
+      "step": 3952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018972858883983308,
+      "loss": 1.1399,
+      "step": 3953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001897223966366825,
+      "loss": 0.9699,
+      "step": 3954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018971620266869507,
+      "loss": 1.0396,
+      "step": 3955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018971000693599258,
+      "loss": 1.3158,
+      "step": 3956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018970380943869687,
+      "loss": 1.0162,
+      "step": 3957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018969761017692988,
+      "loss": 0.9083,
+      "step": 3958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018969140915081357,
+      "loss": 1.0899,
+      "step": 3959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001896852063604699,
+      "loss": 1.0459,
+      "step": 3960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018967900180602083,
+      "loss": 1.0183,
+      "step": 3961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018967279548758845,
+      "loss": 1.253,
+      "step": 3962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018966658740529483,
+      "loss": 0.9413,
+      "step": 3963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018966037755926212,
+      "loss": 1.0428,
+      "step": 3964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018965416594961238,
+      "loss": 1.0016,
+      "step": 3965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018964795257646786,
+      "loss": 1.1024,
+      "step": 3966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018964173743995076,
+      "loss": 1.2285,
+      "step": 3967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018963552054018334,
+      "loss": 0.9543,
+      "step": 3968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001896293018772879,
+      "loss": 0.9643,
+      "step": 3969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018962308145138666,
+      "loss": 1.022,
+      "step": 3970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018961685926260213,
+      "loss": 0.9881,
+      "step": 3971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001896106353110566,
+      "loss": 1.1173,
+      "step": 3972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018960440959687254,
+      "loss": 1.1595,
+      "step": 3973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018959818212017235,
+      "loss": 1.091,
+      "step": 3974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001895919528810786,
+      "loss": 1.0956,
+      "step": 3975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018958572187971376,
+      "loss": 0.9114,
+      "step": 3976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001895794891162004,
+      "loss": 1.2132,
+      "step": 3977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018957325459066117,
+      "loss": 1.2322,
+      "step": 3978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018956701830321864,
+      "loss": 1.0555,
+      "step": 3979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001895607802539955,
+      "loss": 1.1162,
+      "step": 3980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001895545404431145,
+      "loss": 0.9517,
+      "step": 3981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018954829887069829,
+      "loss": 1.3061,
+      "step": 3982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001895420555368697,
+      "loss": 1.0183,
+      "step": 3983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018953581044175153,
+      "loss": 0.9218,
+      "step": 3984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018952956358546662,
+      "loss": 1.119,
+      "step": 3985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001895233149681378,
+      "loss": 1.2016,
+      "step": 3986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018951706458988804,
+      "loss": 1.1441,
+      "step": 3987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001895108124508403,
+      "loss": 1.0382,
+      "step": 3988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018950455855111745,
+      "loss": 1.0239,
+      "step": 3989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018949830289084264,
+      "loss": 0.8367,
+      "step": 3990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001894920454701388,
+      "loss": 1.2166,
+      "step": 3991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001894857862891291,
+      "loss": 1.0929,
+      "step": 3992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018947952534793661,
+      "loss": 1.1454,
+      "step": 3993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001894732626466845,
+      "loss": 1.045,
+      "step": 3994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000189466998185496,
+      "loss": 1.0236,
+      "step": 3995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018946073196449424,
+      "loss": 1.074,
+      "step": 3996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001894544639838025,
+      "loss": 0.9061,
+      "step": 3997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018944819424354415,
+      "loss": 1.0202,
+      "step": 3998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018944192274384243,
+      "loss": 1.1124,
+      "step": 3999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018943564948482075,
+      "loss": 0.902,
+      "step": 4000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018942937446660247,
+      "loss": 1.1786,
+      "step": 4001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018942309768931105,
+      "loss": 1.0558,
+      "step": 4002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018941681915306992,
+      "loss": 1.1087,
+      "step": 4003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018941053885800262,
+      "loss": 1.1881,
+      "step": 4004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018940425680423264,
+      "loss": 1.2116,
+      "step": 4005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018939797299188355,
+      "loss": 1.0449,
+      "step": 4006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.000189391687421079,
+      "loss": 1.0672,
+      "step": 4007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001893854000919426,
+      "loss": 1.0346,
+      "step": 4008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018937911100459803,
+      "loss": 0.9655,
+      "step": 4009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018937282015916895,
+      "loss": 0.9943,
+      "step": 4010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018936652755577918,
+      "loss": 1.0204,
+      "step": 4011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018936023319455242,
+      "loss": 0.9647,
+      "step": 4012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018935393707561251,
+      "loss": 1.0355,
+      "step": 4013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018934763919908329,
+      "loss": 1.1533,
+      "step": 4014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018934133956508866,
+      "loss": 1.0799,
+      "step": 4015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018933503817375255,
+      "loss": 0.9968,
+      "step": 4016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018932873502519883,
+      "loss": 0.9558,
+      "step": 4017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018932243011955154,
+      "loss": 0.9569,
+      "step": 4018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018931612345693467,
+      "loss": 1.0949,
+      "step": 4019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018930981503747233,
+      "loss": 1.0044,
+      "step": 4020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018930350486128856,
+      "loss": 1.1316,
+      "step": 4021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018929719292850745,
+      "loss": 0.8828,
+      "step": 4022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018929087923925323,
+      "loss": 1.1642,
+      "step": 4023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018928456379365002,
+      "loss": 1.1574,
+      "step": 4024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001892782465918221,
+      "loss": 1.1733,
+      "step": 4025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001892719276338937,
+      "loss": 1.0985,
+      "step": 4026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001892656069199891,
+      "loss": 1.1777,
+      "step": 4027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018925928445023268,
+      "loss": 1.2142,
+      "step": 4028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018925296022474878,
+      "loss": 1.001,
+      "step": 4029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018924663424366172,
+      "loss": 0.8662,
+      "step": 4030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018924030650709604,
+      "loss": 1.2608,
+      "step": 4031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018923397701517617,
+      "loss": 1.1227,
+      "step": 4032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001892276457680266,
+      "loss": 1.2009,
+      "step": 4033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001892213127657719,
+      "loss": 0.9185,
+      "step": 4034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.0001892149780085366,
+      "loss": 0.7456,
+      "step": 4035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018920864149644527,
+      "loss": 1.1561,
+      "step": 4036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018920230322962264,
+      "loss": 1.1808,
+      "step": 4037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018919596320819335,
+      "loss": 1.1389,
+      "step": 4038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018918962143228208,
+      "loss": 0.9511,
+      "step": 4039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018918327790201357,
+      "loss": 1.1874,
+      "step": 4040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 0.00018917693261751264,
+      "loss": 1.1621,
+      "step": 4041
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018917058557890406,
+      "loss": 1.2055,
+      "step": 4042
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018916423678631272,
+      "loss": 1.0537,
+      "step": 4043
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018915788623986347,
+      "loss": 1.178,
+      "step": 4044
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001891515339396812,
+      "loss": 0.8601,
+      "step": 4045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018914517988589092,
+      "loss": 1.204,
+      "step": 4046
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001891388240786176,
+      "loss": 1.1103,
+      "step": 4047
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001891324665179862,
+      "loss": 1.0338,
+      "step": 4048
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018912610720412185,
+      "loss": 0.9816,
+      "step": 4049
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018911974613714959,
+      "loss": 1.035,
+      "step": 4050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018911338331719453,
+      "loss": 1.1378,
+      "step": 4051
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018910701874438188,
+      "loss": 1.0479,
+      "step": 4052
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001891006524188368,
+      "loss": 1.2548,
+      "step": 4053
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001890942843406845,
+      "loss": 1.1496,
+      "step": 4054
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001890879145100503,
+      "loss": 1.1248,
+      "step": 4055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018908154292705942,
+      "loss": 1.0427,
+      "step": 4056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001890751695918372,
+      "loss": 1.1103,
+      "step": 4057
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018906879450450906,
+      "loss": 1.2397,
+      "step": 4058
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018906241766520034,
+      "loss": 1.1942,
+      "step": 4059
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018905603907403651,
+      "loss": 1.1264,
+      "step": 4060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000189049658731143,
+      "loss": 0.998,
+      "step": 4061
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018904327663664535,
+      "loss": 1.0998,
+      "step": 4062
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018903689279066907,
+      "loss": 1.1841,
+      "step": 4063
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018903050719333976,
+      "loss": 1.1649,
+      "step": 4064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018902411984478298,
+      "loss": 0.9257,
+      "step": 4065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018901773074512441,
+      "loss": 1.0218,
+      "step": 4066
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018901133989448967,
+      "loss": 1.0361,
+      "step": 4067
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018900494729300453,
+      "loss": 1.0917,
+      "step": 4068
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018899855294079472,
+      "loss": 1.0693,
+      "step": 4069
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018899215683798595,
+      "loss": 1.0178,
+      "step": 4070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018898575898470413,
+      "loss": 1.1857,
+      "step": 4071
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018897935938107504,
+      "loss": 1.0769,
+      "step": 4072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018897295802722456,
+      "loss": 1.066,
+      "step": 4073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018896655492327863,
+      "loss": 0.965,
+      "step": 4074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018896015006936322,
+      "loss": 0.9608,
+      "step": 4075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018895374346560427,
+      "loss": 1.0146,
+      "step": 4076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018894733511212782,
+      "loss": 1.0685,
+      "step": 4077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001889409250090599,
+      "loss": 1.0183,
+      "step": 4078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001889345131565266,
+      "loss": 1.094,
+      "step": 4079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018892809955465406,
+      "loss": 1.067,
+      "step": 4080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018892168420356847,
+      "loss": 1.0216,
+      "step": 4081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018891526710339594,
+      "loss": 1.185,
+      "step": 4082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001889088482542627,
+      "loss": 1.0032,
+      "step": 4083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018890242765629508,
+      "loss": 1.0712,
+      "step": 4084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018889600530961934,
+      "loss": 1.0127,
+      "step": 4085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018888958121436177,
+      "loss": 1.2352,
+      "step": 4086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001888831553706488,
+      "loss": 1.0536,
+      "step": 4087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018887672777860676,
+      "loss": 0.8779,
+      "step": 4088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018887029843836213,
+      "loss": 1.2022,
+      "step": 4089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018886386735004131,
+      "loss": 1.1233,
+      "step": 4090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001888574345137709,
+      "loss": 1.2301,
+      "step": 4091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018885099992967734,
+      "loss": 1.0632,
+      "step": 4092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018884456359788724,
+      "loss": 1.0704,
+      "step": 4093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001888381255185272,
+      "loss": 1.0307,
+      "step": 4094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018883168569172387,
+      "loss": 1.1437,
+      "step": 4095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001888252441176039,
+      "loss": 1.1228,
+      "step": 4096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018881880079629396,
+      "loss": 1.1027,
+      "step": 4097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001888123557279209,
+      "loss": 1.2231,
+      "step": 4098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018880590891261136,
+      "loss": 1.0225,
+      "step": 4099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018879946035049226,
+      "loss": 0.8977,
+      "step": 4100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018879301004169036,
+      "loss": 1.2296,
+      "step": 4101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001887865579863326,
+      "loss": 0.953,
+      "step": 4102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018878010418454586,
+      "loss": 1.0463,
+      "step": 4103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001887736486364571,
+      "loss": 0.9642,
+      "step": 4104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018876719134219326,
+      "loss": 1.1072,
+      "step": 4105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001887607323018814,
+      "loss": 1.0066,
+      "step": 4106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001887542715156486,
+      "loss": 1.0502,
+      "step": 4107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018874780898362183,
+      "loss": 0.8808,
+      "step": 4108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018874134470592835,
+      "loss": 1.0967,
+      "step": 4109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018873487868269518,
+      "loss": 0.9847,
+      "step": 4110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001887284109140496,
+      "loss": 0.9488,
+      "step": 4111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018872194140011879,
+      "loss": 1.1602,
+      "step": 4112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018871547014103,
+      "loss": 1.138,
+      "step": 4113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018870899713691055,
+      "loss": 1.1108,
+      "step": 4114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018870252238788773,
+      "loss": 0.9521,
+      "step": 4115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001886960458940889,
+      "loss": 0.9179,
+      "step": 4116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001886895676556415,
+      "loss": 1.0909,
+      "step": 4117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018868308767267292,
+      "loss": 0.9711,
+      "step": 4118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018867660594531062,
+      "loss": 0.9783,
+      "step": 4119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018867012247368207,
+      "loss": 1.1721,
+      "step": 4120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018866363725791486,
+      "loss": 1.0198,
+      "step": 4121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018865715029813652,
+      "loss": 1.1705,
+      "step": 4122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018865066159447466,
+      "loss": 1.0258,
+      "step": 4123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018864417114705689,
+      "loss": 1.0206,
+      "step": 4124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018863767895601087,
+      "loss": 0.9801,
+      "step": 4125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018863118502146438,
+      "loss": 1.1249,
+      "step": 4126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018862468934354506,
+      "loss": 1.0913,
+      "step": 4127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001886181919223807,
+      "loss": 1.1032,
+      "step": 4128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018861169275809914,
+      "loss": 0.98,
+      "step": 4129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018860519185082822,
+      "loss": 0.9286,
+      "step": 4130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018859868920069575,
+      "loss": 1.0216,
+      "step": 4131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001885921848078297,
+      "loss": 1.0292,
+      "step": 4132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000188585678672358,
+      "loss": 1.113,
+      "step": 4133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001885791707944086,
+      "loss": 1.0074,
+      "step": 4134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018857266117410952,
+      "loss": 1.109,
+      "step": 4135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018856614981158884,
+      "loss": 0.8248,
+      "step": 4136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001885596367069746,
+      "loss": 1.1551,
+      "step": 4137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001885531218603949,
+      "loss": 0.8863,
+      "step": 4138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018854660527197793,
+      "loss": 1.0712,
+      "step": 4139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018854008694185184,
+      "loss": 1.2301,
+      "step": 4140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018853356687014485,
+      "loss": 0.9908,
+      "step": 4141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001885270450569852,
+      "loss": 1.0872,
+      "step": 4142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018852052150250122,
+      "loss": 1.1529,
+      "step": 4143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001885139962068212,
+      "loss": 1.0461,
+      "step": 4144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018850746917007345,
+      "loss": 0.9487,
+      "step": 4145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001885009403923864,
+      "loss": 1.1255,
+      "step": 4146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018849440987388848,
+      "loss": 1.0516,
+      "step": 4147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018848787761470813,
+      "loss": 1.0583,
+      "step": 4148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018848134361497385,
+      "loss": 1.0528,
+      "step": 4149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018847480787481416,
+      "loss": 1.2051,
+      "step": 4150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018846827039435758,
+      "loss": 1.0223,
+      "step": 4151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018846173117373276,
+      "loss": 1.1619,
+      "step": 4152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018845519021306828,
+      "loss": 0.9529,
+      "step": 4153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018844864751249284,
+      "loss": 0.9821,
+      "step": 4154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001884421030721351,
+      "loss": 1.1225,
+      "step": 4155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018843555689212379,
+      "loss": 0.8329,
+      "step": 4156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001884290089725877,
+      "loss": 1.1184,
+      "step": 4157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018842245931365562,
+      "loss": 1.1904,
+      "step": 4158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018841590791545638,
+      "loss": 1.0877,
+      "step": 4159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018840935477811886,
+      "loss": 0.8627,
+      "step": 4160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001884027999017719,
+      "loss": 1.1379,
+      "step": 4161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001883962432865445,
+      "loss": 0.8158,
+      "step": 4162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001883896849325656,
+      "loss": 1.1322,
+      "step": 4163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018838312483996421,
+      "loss": 1.0487,
+      "step": 4164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018837656300886937,
+      "loss": 1.2874,
+      "step": 4165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001883699994394101,
+      "loss": 1.2517,
+      "step": 4166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001883634341317156,
+      "loss": 1.2067,
+      "step": 4167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018835686708591496,
+      "loss": 1.2203,
+      "step": 4168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018835029830213733,
+      "loss": 1.3238,
+      "step": 4169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018834372778051196,
+      "loss": 1.045,
+      "step": 4170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018833715552116806,
+      "loss": 0.9322,
+      "step": 4171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018833058152423492,
+      "loss": 1.1129,
+      "step": 4172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018832400578984183,
+      "loss": 1.0321,
+      "step": 4173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018831742831811822,
+      "loss": 1.338,
+      "step": 4174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018831084910919335,
+      "loss": 0.9337,
+      "step": 4175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018830426816319665,
+      "loss": 0.9834,
+      "step": 4176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018829768548025766,
+      "loss": 1.1162,
+      "step": 4177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001882911010605058,
+      "loss": 1.0667,
+      "step": 4178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018828451490407057,
+      "loss": 1.2615,
+      "step": 4179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018827792701108154,
+      "loss": 1.0645,
+      "step": 4180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001882713373816683,
+      "loss": 0.923,
+      "step": 4181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018826474601596046,
+      "loss": 1.1128,
+      "step": 4182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018825815291408768,
+      "loss": 1.0885,
+      "step": 4183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018825155807617963,
+      "loss": 1.2963,
+      "step": 4184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018824496150236604,
+      "loss": 1.2232,
+      "step": 4185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018823836319277666,
+      "loss": 0.9967,
+      "step": 4186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001882317631475413,
+      "loss": 0.8612,
+      "step": 4187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018822516136678973,
+      "loss": 1.0453,
+      "step": 4188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018821855785065185,
+      "loss": 1.1458,
+      "step": 4189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018821195259925756,
+      "loss": 1.1762,
+      "step": 4190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018820534561273674,
+      "loss": 1.0843,
+      "step": 4191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018819873689121942,
+      "loss": 1.0761,
+      "step": 4192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001881921264348355,
+      "loss": 0.8495,
+      "step": 4193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018818551424371508,
+      "loss": 1.1009,
+      "step": 4194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001881789003179882,
+      "loss": 1.117,
+      "step": 4195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018817228465778496,
+      "loss": 1.1639,
+      "step": 4196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018816566726323548,
+      "loss": 1.2129,
+      "step": 4197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001881590481344699,
+      "loss": 1.077,
+      "step": 4198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001881524272716185,
+      "loss": 1.2331,
+      "step": 4199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018814580467481142,
+      "loss": 1.1828,
+      "step": 4200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018813918034417897,
+      "loss": 0.9361,
+      "step": 4201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018813255427985146,
+      "loss": 1.1248,
+      "step": 4202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001881259264819592,
+      "loss": 0.9883,
+      "step": 4203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018811929695063256,
+      "loss": 0.9261,
+      "step": 4204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018811266568600194,
+      "loss": 1.1125,
+      "step": 4205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018810603268819783,
+      "loss": 1.0907,
+      "step": 4206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018809939795735062,
+      "loss": 1.0946,
+      "step": 4207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018809276149359087,
+      "loss": 1.1455,
+      "step": 4208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001880861232970491,
+      "loss": 0.9566,
+      "step": 4209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018807948336785587,
+      "loss": 1.054,
+      "step": 4210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018807284170614184,
+      "loss": 1.1336,
+      "step": 4211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018806619831203758,
+      "loss": 1.1346,
+      "step": 4212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001880595531856738,
+      "loss": 1.0309,
+      "step": 4213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018805290632718121,
+      "loss": 1.1889,
+      "step": 4214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001880462577366906,
+      "loss": 1.0726,
+      "step": 4215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018803960741433265,
+      "loss": 1.1606,
+      "step": 4216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018803295536023824,
+      "loss": 1.1912,
+      "step": 4217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018802630157453817,
+      "loss": 1.0607,
+      "step": 4218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018801964605736339,
+      "loss": 0.7113,
+      "step": 4219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018801298880884474,
+      "loss": 1.1485,
+      "step": 4220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018800632982911322,
+      "loss": 1.0297,
+      "step": 4221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018799966911829975,
+      "loss": 0.9542,
+      "step": 4222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018799300667653542,
+      "loss": 1.3231,
+      "step": 4223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018798634250395124,
+      "loss": 0.9147,
+      "step": 4224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018797967660067833,
+      "loss": 1.1337,
+      "step": 4225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018797300896684774,
+      "loss": 0.9485,
+      "step": 4226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018796633960259067,
+      "loss": 1.0168,
+      "step": 4227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001879596685080383,
+      "loss": 0.9663,
+      "step": 4228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018795299568332184,
+      "loss": 1.0967,
+      "step": 4229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018794632112857258,
+      "loss": 0.9016,
+      "step": 4230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018793964484392175,
+      "loss": 1.1748,
+      "step": 4231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001879329668295007,
+      "loss": 1.1006,
+      "step": 4232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001879262870854408,
+      "loss": 0.9314,
+      "step": 4233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018791960561187346,
+      "loss": 0.9804,
+      "step": 4234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018791292240893004,
+      "loss": 1.0625,
+      "step": 4235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018790623747674202,
+      "loss": 0.9076,
+      "step": 4236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018789955081544092,
+      "loss": 1.1788,
+      "step": 4237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018789286242515828,
+      "loss": 0.8658,
+      "step": 4238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018788617230602558,
+      "loss": 1.1439,
+      "step": 4239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001878794804581745,
+      "loss": 0.9706,
+      "step": 4240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018787278688173662,
+      "loss": 0.9828,
+      "step": 4241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001878660915768436,
+      "loss": 1.2188,
+      "step": 4242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018785939454362717,
+      "loss": 0.9345,
+      "step": 4243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018785269578221905,
+      "loss": 1.0712,
+      "step": 4244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000187845995292751,
+      "loss": 0.9815,
+      "step": 4245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018783929307535479,
+      "loss": 1.1898,
+      "step": 4246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001878325891301623,
+      "loss": 0.9691,
+      "step": 4247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018782588345730538,
+      "loss": 0.9467,
+      "step": 4248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001878191760569159,
+      "loss": 1.1844,
+      "step": 4249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018781246692912585,
+      "loss": 1.0717,
+      "step": 4250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001878057560740672,
+      "loss": 1.0898,
+      "step": 4251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018779904349187186,
+      "loss": 1.1969,
+      "step": 4252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018779232918267195,
+      "loss": 1.0642,
+      "step": 4253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001877856131465995,
+      "loss": 1.136,
+      "step": 4254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018777889538378664,
+      "loss": 1.158,
+      "step": 4255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018777217589436548,
+      "loss": 0.956,
+      "step": 4256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018776545467846827,
+      "loss": 1.209,
+      "step": 4257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001877587317362271,
+      "loss": 0.8708,
+      "step": 4258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001877520070677743,
+      "loss": 1.1556,
+      "step": 4259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001877452806732421,
+      "loss": 1.113,
+      "step": 4260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018773855255276283,
+      "loss": 1.1111,
+      "step": 4261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001877318227064688,
+      "loss": 1.1206,
+      "step": 4262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018772509113449245,
+      "loss": 1.2553,
+      "step": 4263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001877183578369661,
+      "loss": 1.1996,
+      "step": 4264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018771162281402227,
+      "loss": 1.0005,
+      "step": 4265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.0001877048860657934,
+      "loss": 1.3631,
+      "step": 4266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.000187698147592412,
+      "loss": 1.1133,
+      "step": 4267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018769140739401062,
+      "loss": 0.8881,
+      "step": 4268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018768466547072188,
+      "loss": 1.0271,
+      "step": 4269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018767792182267833,
+      "loss": 1.0333,
+      "step": 4270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018767117645001265,
+      "loss": 1.1939,
+      "step": 4271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00018766442935285756,
+      "loss": 1.1519,
+      "step": 4272
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001876576805313457,
+      "loss": 0.9768,
+      "step": 4273
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018765092998560982,
+      "loss": 1.3687,
+      "step": 4274
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018764417771578279,
+      "loss": 0.8994,
+      "step": 4275
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018763742372199734,
+      "loss": 0.9468,
+      "step": 4276
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018763066800438636,
+      "loss": 1.0745,
+      "step": 4277
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018762391056308271,
+      "loss": 1.1126,
+      "step": 4278
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018761715139821935,
+      "loss": 1.2344,
+      "step": 4279
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018761039050992922,
+      "loss": 1.2234,
+      "step": 4280
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001876036278983453,
+      "loss": 1.001,
+      "step": 4281
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001875968635636006,
+      "loss": 0.7522,
+      "step": 4282
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018759009750582816,
+      "loss": 0.9353,
+      "step": 4283
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018758332972516106,
+      "loss": 0.9758,
+      "step": 4284
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018757656022173248,
+      "loss": 0.9676,
+      "step": 4285
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018756978899567558,
+      "loss": 0.9023,
+      "step": 4286
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018756301604712348,
+      "loss": 0.8583,
+      "step": 4287
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018755624137620943,
+      "loss": 1.1752,
+      "step": 4288
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018754946498306674,
+      "loss": 1.0492,
+      "step": 4289
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018754268686782863,
+      "loss": 1.2384,
+      "step": 4290
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018753590703062843,
+      "loss": 1.2499,
+      "step": 4291
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018752912547159958,
+      "loss": 1.1699,
+      "step": 4292
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018752234219087538,
+      "loss": 1.0724,
+      "step": 4293
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001875155571885893,
+      "loss": 0.8448,
+      "step": 4294
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001875087704648748,
+      "loss": 1.1139,
+      "step": 4295
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018750198201986535,
+      "loss": 0.9786,
+      "step": 4296
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001874951918536945,
+      "loss": 0.8946,
+      "step": 4297
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018748839996649583,
+      "loss": 1.1321,
+      "step": 4298
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018748160635840292,
+      "loss": 1.0368,
+      "step": 4299
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018747481102954935,
+      "loss": 1.2199,
+      "step": 4300
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001874680139800689,
+      "loss": 0.879,
+      "step": 4301
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018746121521009515,
+      "loss": 0.9211,
+      "step": 4302
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018745441471976188,
+      "loss": 0.9654,
+      "step": 4303
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018744761250920287,
+      "loss": 1.1019,
+      "step": 4304
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018744080857855191,
+      "loss": 1.1583,
+      "step": 4305
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001874340029279428,
+      "loss": 1.1036,
+      "step": 4306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018742719555750944,
+      "loss": 0.9886,
+      "step": 4307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018742038646738575,
+      "loss": 1.1264,
+      "step": 4308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001874135756577056,
+      "loss": 1.1611,
+      "step": 4309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018740676312860305,
+      "loss": 1.0032,
+      "step": 4310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018739994888021202,
+      "loss": 1.1726,
+      "step": 4311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018739313291266655,
+      "loss": 0.9668,
+      "step": 4312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018738631522610078,
+      "loss": 1.0186,
+      "step": 4313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018737949582064872,
+      "loss": 1.1867,
+      "step": 4314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018737267469644462,
+      "loss": 1.096,
+      "step": 4315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018736585185362255,
+      "loss": 1.099,
+      "step": 4316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018735902729231677,
+      "loss": 1.1649,
+      "step": 4317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001873522010126615,
+      "loss": 1.1427,
+      "step": 4318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000187345373014791,
+      "loss": 1.3555,
+      "step": 4319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001873385432988396,
+      "loss": 0.926,
+      "step": 4320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018733171186494164,
+      "loss": 1.0264,
+      "step": 4321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001873248787132315,
+      "loss": 1.1271,
+      "step": 4322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001873180438438435,
+      "loss": 1.1308,
+      "step": 4323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018731120725691222,
+      "loss": 1.2582,
+      "step": 4324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018730436895257206,
+      "loss": 1.0486,
+      "step": 4325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018729752893095755,
+      "loss": 1.0473,
+      "step": 4326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001872906871922032,
+      "loss": 1.1938,
+      "step": 4327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018728384373644364,
+      "loss": 1.0706,
+      "step": 4328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018727699856381345,
+      "loss": 1.0243,
+      "step": 4329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018727015167444724,
+      "loss": 1.198,
+      "step": 4330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018726330306847975,
+      "loss": 1.0996,
+      "step": 4331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018725645274604566,
+      "loss": 1.2692,
+      "step": 4332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018724960070727972,
+      "loss": 1.0572,
+      "step": 4333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001872427469523167,
+      "loss": 1.3746,
+      "step": 4334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018723589148129146,
+      "loss": 1.0482,
+      "step": 4335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018722903429433876,
+      "loss": 1.095,
+      "step": 4336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018722217539159357,
+      "loss": 1.1118,
+      "step": 4337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018721531477319074,
+      "loss": 1.0776,
+      "step": 4338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018720845243926527,
+      "loss": 1.3747,
+      "step": 4339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001872015883899521,
+      "loss": 1.2752,
+      "step": 4340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018719472262538624,
+      "loss": 1.0107,
+      "step": 4341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001871878551457028,
+      "loss": 0.9215,
+      "step": 4342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001871809859510368,
+      "loss": 0.8539,
+      "step": 4343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018717411504152337,
+      "loss": 1.0043,
+      "step": 4344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001871672424172977,
+      "loss": 1.0379,
+      "step": 4345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018716036807849492,
+      "loss": 1.0917,
+      "step": 4346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018715349202525028,
+      "loss": 0.8885,
+      "step": 4347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018714661425769903,
+      "loss": 1.234,
+      "step": 4348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018713973477597646,
+      "loss": 1.1176,
+      "step": 4349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018713285358021788,
+      "loss": 0.9392,
+      "step": 4350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018712597067055864,
+      "loss": 1.0128,
+      "step": 4351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001871190860471341,
+      "loss": 0.8943,
+      "step": 4352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018711219971007974,
+      "loss": 1.254,
+      "step": 4353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000187105311659531,
+      "loss": 1.1262,
+      "step": 4354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018709842189562333,
+      "loss": 1.0607,
+      "step": 4355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018709153041849224,
+      "loss": 1.0315,
+      "step": 4356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018708463722827336,
+      "loss": 1.3347,
+      "step": 4357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018707774232510222,
+      "loss": 1.1665,
+      "step": 4358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018707084570911446,
+      "loss": 1.0428,
+      "step": 4359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018706394738044575,
+      "loss": 1.2204,
+      "step": 4360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018705704733923176,
+      "loss": 0.9076,
+      "step": 4361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018705014558560823,
+      "loss": 1.2527,
+      "step": 4362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018704324211971088,
+      "loss": 1.0287,
+      "step": 4363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018703633694167553,
+      "loss": 1.2898,
+      "step": 4364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018702943005163803,
+      "loss": 1.007,
+      "step": 4365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018702252144973418,
+      "loss": 1.1093,
+      "step": 4366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018701561113609993,
+      "loss": 1.1313,
+      "step": 4367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018700869911087115,
+      "loss": 1.0887,
+      "step": 4368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018700178537418385,
+      "loss": 1.0287,
+      "step": 4369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186994869926174,
+      "loss": 1.2425,
+      "step": 4370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001869879527669776,
+      "loss": 1.1146,
+      "step": 4371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018698103389673076,
+      "loss": 1.1077,
+      "step": 4372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018697411331556956,
+      "loss": 1.0159,
+      "step": 4373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001869671910236301,
+      "loss": 1.0236,
+      "step": 4374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018696026702104857,
+      "loss": 1.0111,
+      "step": 4375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018695334130796115,
+      "loss": 1.1535,
+      "step": 4376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018694641388450407,
+      "loss": 1.0286,
+      "step": 4377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001869394847508136,
+      "loss": 1.1494,
+      "step": 4378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018693255390702605,
+      "loss": 1.0715,
+      "step": 4379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018692562135327773,
+      "loss": 1.0113,
+      "step": 4380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018691868708970496,
+      "loss": 1.1813,
+      "step": 4381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018691175111644422,
+      "loss": 1.1448,
+      "step": 4382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001869048134336319,
+      "loss": 1.1555,
+      "step": 4383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018689787404140446,
+      "loss": 1.1144,
+      "step": 4384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001868909329398984,
+      "loss": 1.1274,
+      "step": 4385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018688399012925023,
+      "loss": 1.1534,
+      "step": 4386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001868770456095966,
+      "loss": 1.0237,
+      "step": 4387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186870099381074,
+      "loss": 1.0991,
+      "step": 4388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018686315144381913,
+      "loss": 0.9993,
+      "step": 4389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018685620179796864,
+      "loss": 1.0832,
+      "step": 4390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001868492504436592,
+      "loss": 0.9663,
+      "step": 4391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001868422973810276,
+      "loss": 1.0396,
+      "step": 4392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018683534261021057,
+      "loss": 1.117,
+      "step": 4393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001868283861313449,
+      "loss": 1.0106,
+      "step": 4394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018682142794456742,
+      "loss": 1.0937,
+      "step": 4395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018681446805001505,
+      "loss": 1.0467,
+      "step": 4396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018680750644782467,
+      "loss": 1.1696,
+      "step": 4397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018680054313813314,
+      "loss": 1.0897,
+      "step": 4398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018679357812107755,
+      "loss": 1.0956,
+      "step": 4399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001867866113967948,
+      "loss": 1.0095,
+      "step": 4400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186779642965422,
+      "loss": 1.2963,
+      "step": 4401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018677267282709615,
+      "loss": 0.85,
+      "step": 4402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018676570098195443,
+      "loss": 1.1271,
+      "step": 4403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001867587274301339,
+      "loss": 1.0143,
+      "step": 4404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018675175217177175,
+      "loss": 0.9033,
+      "step": 4405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018674477520700525,
+      "loss": 1.0121,
+      "step": 4406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001867377965359715,
+      "loss": 0.9737,
+      "step": 4407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018673081615880793,
+      "loss": 1.1367,
+      "step": 4408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018672383407565175,
+      "loss": 1.1445,
+      "step": 4409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018671685028664032,
+      "loss": 1.1904,
+      "step": 4410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018670986479191098,
+      "loss": 1.0634,
+      "step": 4411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018670287759160116,
+      "loss": 1.0149,
+      "step": 4412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866958886858483,
+      "loss": 1.1635,
+      "step": 4413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866888980747899,
+      "loss": 1.1775,
+      "step": 4414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001866819057585634,
+      "loss": 1.1621,
+      "step": 4415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018667491173730638,
+      "loss": 1.1363,
+      "step": 4416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018666791601115642,
+      "loss": 1.132,
+      "step": 4417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018666091858025112,
+      "loss": 1.1266,
+      "step": 4418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018665391944472807,
+      "loss": 1.132,
+      "step": 4419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186646918604725,
+      "loss": 1.0274,
+      "step": 4420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018663991606037962,
+      "loss": 1.0016,
+      "step": 4421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018663291181182965,
+      "loss": 1.1806,
+      "step": 4422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018662590585921286,
+      "loss": 1.1864,
+      "step": 4423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186618898202667,
+      "loss": 1.3013,
+      "step": 4424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018661188884233003,
+      "loss": 1.1093,
+      "step": 4425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018660487777833975,
+      "loss": 0.9875,
+      "step": 4426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018659786501083408,
+      "loss": 1.1432,
+      "step": 4427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018659085053995098,
+      "loss": 0.9979,
+      "step": 4428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001865838343658284,
+      "loss": 1.0758,
+      "step": 4429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018657681648860433,
+      "loss": 1.0376,
+      "step": 4430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018656979690841684,
+      "loss": 1.1252,
+      "step": 4431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186562775625404,
+      "loss": 1.3548,
+      "step": 4432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018655575263970396,
+      "loss": 0.8801,
+      "step": 4433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018654872795145476,
+      "loss": 1.0878,
+      "step": 4434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001865417015607947,
+      "loss": 1.1053,
+      "step": 4435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018653467346786185,
+      "loss": 0.9774,
+      "step": 4436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018652764367279461,
+      "loss": 0.9661,
+      "step": 4437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018652061217573114,
+      "loss": 1.1579,
+      "step": 4438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018651357897680977,
+      "loss": 1.0556,
+      "step": 4439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018650654407616885,
+      "loss": 0.9261,
+      "step": 4440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001864995074739468,
+      "loss": 1.061,
+      "step": 4441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018649246917028196,
+      "loss": 0.9187,
+      "step": 4442
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001864854291653128,
+      "loss": 1.1585,
+      "step": 4443
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018647838745917782,
+      "loss": 1.2086,
+      "step": 4444
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001864713440520155,
+      "loss": 1.1857,
+      "step": 4445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018646429894396443,
+      "loss": 1.1354,
+      "step": 4446
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018645725213516312,
+      "loss": 1.0156,
+      "step": 4447
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018645020362575022,
+      "loss": 1.0342,
+      "step": 4448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018644315341586438,
+      "loss": 1.1323,
+      "step": 4449
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018643610150564422,
+      "loss": 1.1301,
+      "step": 4450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018642904789522855,
+      "loss": 1.2203,
+      "step": 4451
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018642199258475604,
+      "loss": 1.0308,
+      "step": 4452
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001864149355743655,
+      "loss": 1.2127,
+      "step": 4453
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001864078768641957,
+      "loss": 1.1265,
+      "step": 4454
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018640081645438552,
+      "loss": 1.1934,
+      "step": 4455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018639375434507382,
+      "loss": 1.0268,
+      "step": 4456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018638669053639957,
+      "loss": 0.9578,
+      "step": 4457
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001863796250285016,
+      "loss": 1.0445,
+      "step": 4458
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186372557821519,
+      "loss": 1.2145,
+      "step": 4459
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018636548891559076,
+      "loss": 1.1335,
+      "step": 4460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018635841831085586,
+      "loss": 0.9056,
+      "step": 4461
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018635134600745342,
+      "loss": 1.0392,
+      "step": 4462
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018634427200552261,
+      "loss": 1.2991,
+      "step": 4463
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018633719630520247,
+      "loss": 1.24,
+      "step": 4464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018633011890663222,
+      "loss": 1.2901,
+      "step": 4465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001863230398099511,
+      "loss": 1.0288,
+      "step": 4466
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018631595901529835,
+      "loss": 1.1917,
+      "step": 4467
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018630887652281323,
+      "loss": 1.1358,
+      "step": 4468
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018630179233263504,
+      "loss": 0.9819,
+      "step": 4469
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018629470644490313,
+      "loss": 1.0188,
+      "step": 4470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001862876188597569,
+      "loss": 1.13,
+      "step": 4471
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001862805295773358,
+      "loss": 1.1508,
+      "step": 4472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001862734385977792,
+      "loss": 1.1633,
+      "step": 4473
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001862663459212266,
+      "loss": 1.2615,
+      "step": 4474
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018625925154781752,
+      "loss": 1.1686,
+      "step": 4475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018625215547769152,
+      "loss": 1.293,
+      "step": 4476
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018624505771098817,
+      "loss": 1.1384,
+      "step": 4477
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018623795824784708,
+      "loss": 1.1091,
+      "step": 4478
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018623085708840788,
+      "loss": 1.2174,
+      "step": 4479
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018622375423281028,
+      "loss": 0.8095,
+      "step": 4480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018621664968119402,
+      "loss": 1.2456,
+      "step": 4481
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018620954343369874,
+      "loss": 1.2514,
+      "step": 4482
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018620243549046433,
+      "loss": 1.2317,
+      "step": 4483
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018619532585163057,
+      "loss": 0.9551,
+      "step": 4484
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018618821451733724,
+      "loss": 1.1529,
+      "step": 4485
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001861811014877243,
+      "loss": 0.9626,
+      "step": 4486
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018617398676293165,
+      "loss": 0.9665,
+      "step": 4487
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018616687034309922,
+      "loss": 1.2934,
+      "step": 4488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186159752228367,
+      "loss": 1.0794,
+      "step": 4489
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186152632418875,
+      "loss": 0.9197,
+      "step": 4490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018614551091476324,
+      "loss": 1.023,
+      "step": 4491
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018613838771617186,
+      "loss": 1.147,
+      "step": 4492
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018613126282324092,
+      "loss": 1.2518,
+      "step": 4493
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018612413623611058,
+      "loss": 1.0791,
+      "step": 4494
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.000186117007954921,
+      "loss": 0.8917,
+      "step": 4495
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018610987797981245,
+      "loss": 1.0894,
+      "step": 4496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018610274631092514,
+      "loss": 1.0059,
+      "step": 4497
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001860956129483993,
+      "loss": 1.1295,
+      "step": 4498
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018608847789237536,
+      "loss": 1.2261,
+      "step": 4499
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018608134114299357,
+      "loss": 1.1245,
+      "step": 4500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001860742027003944,
+      "loss": 0.9916,
+      "step": 4501
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00018606706256471813,
+      "loss": 0.9889,
+      "step": 4502
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0001860599207361053,
+      "loss": 1.0267,
+      "step": 4503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018605277721469638,
+      "loss": 1.1588,
+      "step": 4504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018604563200063187,
+      "loss": 1.1254,
+      "step": 4505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018603848509405233,
+      "loss": 1.0606,
+      "step": 4506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018603133649509833,
+      "loss": 0.9644,
+      "step": 4507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018602418620391044,
+      "loss": 1.0539,
+      "step": 4508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860170342206294,
+      "loss": 1.1795,
+      "step": 4509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018600988054539582,
+      "loss": 1.0201,
+      "step": 4510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001860027251783504,
+      "loss": 1.1088,
+      "step": 4511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018599556811963393,
+      "loss": 1.0073,
+      "step": 4512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001859884093693872,
+      "loss": 1.2002,
+      "step": 4513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000185981248927751,
+      "loss": 1.1387,
+      "step": 4514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018597408679486615,
+      "loss": 1.3755,
+      "step": 4515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018596692297087353,
+      "loss": 1.3348,
+      "step": 4516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001859597574559141,
+      "loss": 0.8605,
+      "step": 4517
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018595259025012877,
+      "loss": 0.9511,
+      "step": 4518
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018594542135365853,
+      "loss": 0.8952,
+      "step": 4519
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001859382507666444,
+      "loss": 0.898,
+      "step": 4520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018593107848922743,
+      "loss": 0.9657,
+      "step": 4521
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018592390452154862,
+      "loss": 1.0248,
+      "step": 4522
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001859167288637492,
+      "loss": 1.1533,
+      "step": 4523
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018590955151597026,
+      "loss": 1.0582,
+      "step": 4524
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018590237247835293,
+      "loss": 1.0263,
+      "step": 4525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018589519175103852,
+      "loss": 0.9182,
+      "step": 4526
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018588800933416822,
+      "loss": 1.1901,
+      "step": 4527
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858808252278833,
+      "loss": 1.1877,
+      "step": 4528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018587363943232504,
+      "loss": 1.1366,
+      "step": 4529
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018586645194763491,
+      "loss": 1.1342,
+      "step": 4530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018585926277395415,
+      "loss": 1.0977,
+      "step": 4531
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018585207191142424,
+      "loss": 0.8708,
+      "step": 4532
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018584487936018661,
+      "loss": 1.1656,
+      "step": 4533
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018583768512038277,
+      "loss": 1.0093,
+      "step": 4534
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018583048919215415,
+      "loss": 0.8246,
+      "step": 4535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018582329157564237,
+      "loss": 0.9808,
+      "step": 4536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018581609227098897,
+      "loss": 1.213,
+      "step": 4537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018580889127833557,
+      "loss": 1.1698,
+      "step": 4538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001858016885978238,
+      "loss": 1.0509,
+      "step": 4539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001857944842295954,
+      "loss": 1.0429,
+      "step": 4540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018578727817379195,
+      "loss": 1.1095,
+      "step": 4541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018578007043055532,
+      "loss": 0.9573,
+      "step": 4542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018577286100002723,
+      "loss": 1.1898,
+      "step": 4543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018576564988234947,
+      "loss": 1.1063,
+      "step": 4544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018575843707766394,
+      "loss": 1.069,
+      "step": 4545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001857512225861125,
+      "loss": 1.1042,
+      "step": 4546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000185744006407837,
+      "loss": 1.0356,
+      "step": 4547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018573678854297943,
+      "loss": 1.0026,
+      "step": 4548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001857295689916818,
+      "loss": 0.915,
+      "step": 4549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018572234775408606,
+      "loss": 0.9441,
+      "step": 4550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018571512483033425,
+      "loss": 1.0265,
+      "step": 4551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001857079002205685,
+      "loss": 1.061,
+      "step": 4552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018570067392493088,
+      "loss": 1.1932,
+      "step": 4553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018569344594356347,
+      "loss": 1.0485,
+      "step": 4554
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001856862162766086,
+      "loss": 0.983,
+      "step": 4555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018567898492420835,
+      "loss": 0.9319,
+      "step": 4556
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018567175188650498,
+      "loss": 0.937,
+      "step": 4557
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018566451716364076,
+      "loss": 1.052,
+      "step": 4558
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018565728075575807,
+      "loss": 1.2069,
+      "step": 4559
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018565004266299918,
+      "loss": 1.1218,
+      "step": 4560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018564280288550647,
+      "loss": 0.9264,
+      "step": 4561
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018563556142342237,
+      "loss": 1.0002,
+      "step": 4562
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001856283182768893,
+      "loss": 1.1013,
+      "step": 4563
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018562107344604974,
+      "loss": 1.1477,
+      "step": 4564
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018561382693104615,
+      "loss": 1.4099,
+      "step": 4565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018560657873202115,
+      "loss": 1.0476,
+      "step": 4566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001855993288491173,
+      "loss": 1.1845,
+      "step": 4567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018559207728247715,
+      "loss": 1.1679,
+      "step": 4568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018558482403224335,
+      "loss": 1.3009,
+      "step": 4569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018557756909855862,
+      "loss": 1.1398,
+      "step": 4570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018557031248156563,
+      "loss": 1.0888,
+      "step": 4571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018556305418140708,
+      "loss": 0.9198,
+      "step": 4572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018555579419822583,
+      "loss": 1.1994,
+      "step": 4573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018554853253216458,
+      "loss": 0.9132,
+      "step": 4574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018554126918336628,
+      "loss": 1.1749,
+      "step": 4575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018553400415197372,
+      "loss": 1.0265,
+      "step": 4576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001855267374381298,
+      "loss": 1.1786,
+      "step": 4577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018551946904197752,
+      "loss": 1.2081,
+      "step": 4578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001855121989636598,
+      "loss": 1.065,
+      "step": 4579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018550492720331962,
+      "loss": 1.0998,
+      "step": 4580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001854976537611001,
+      "loss": 1.2246,
+      "step": 4581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001854903786371442,
+      "loss": 0.9747,
+      "step": 4582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001854831018315951,
+      "loss": 1.0698,
+      "step": 4583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001854758233445959,
+      "loss": 1.0579,
+      "step": 4584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001854685431762898,
+      "loss": 1.1209,
+      "step": 4585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018546126132681998,
+      "loss": 1.0815,
+      "step": 4586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018545397779632964,
+      "loss": 0.8959,
+      "step": 4587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001854466925849621,
+      "loss": 1.1655,
+      "step": 4588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018543940569286065,
+      "loss": 1.1634,
+      "step": 4589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001854321171201686,
+      "loss": 0.8636,
+      "step": 4590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018542482686702933,
+      "loss": 0.9886,
+      "step": 4591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018541753493358623,
+      "loss": 1.0628,
+      "step": 4592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018541024131998274,
+      "loss": 1.1573,
+      "step": 4593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018540294602636235,
+      "loss": 1.0382,
+      "step": 4594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018539564905286854,
+      "loss": 0.8751,
+      "step": 4595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001853883503996448,
+      "loss": 1.0714,
+      "step": 4596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018538105006683472,
+      "loss": 1.0344,
+      "step": 4597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018537374805458194,
+      "loss": 0.9922,
+      "step": 4598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018536644436303003,
+      "loss": 1.234,
+      "step": 4599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018535913899232268,
+      "loss": 0.9574,
+      "step": 4600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018535183194260361,
+      "loss": 0.9507,
+      "step": 4601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001853445232140165,
+      "loss": 1.0475,
+      "step": 4602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001853372128067052,
+      "loss": 0.9533,
+      "step": 4603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001853299007208134,
+      "loss": 1.0798,
+      "step": 4604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018532258695648496,
+      "loss": 1.1451,
+      "step": 4605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018531527151386377,
+      "loss": 1.0064,
+      "step": 4606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001853079543930937,
+      "loss": 1.1268,
+      "step": 4607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018530063559431866,
+      "loss": 0.9909,
+      "step": 4608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018529331511768268,
+      "loss": 1.1366,
+      "step": 4609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001852859929633297,
+      "loss": 1.1717,
+      "step": 4610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018527866913140375,
+      "loss": 1.0152,
+      "step": 4611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018527134362204887,
+      "loss": 0.9414,
+      "step": 4612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018526401643540922,
+      "loss": 1.1035,
+      "step": 4613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018525668757162888,
+      "loss": 1.3913,
+      "step": 4614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.000185249357030852,
+      "loss": 1.1302,
+      "step": 4615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018524202481322278,
+      "loss": 0.9452,
+      "step": 4616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018523469091888547,
+      "loss": 1.159,
+      "step": 4617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018522735534798427,
+      "loss": 1.0117,
+      "step": 4618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018522001810066354,
+      "loss": 1.0016,
+      "step": 4619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018521267917706756,
+      "loss": 1.0861,
+      "step": 4620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001852053385773407,
+      "loss": 0.9642,
+      "step": 4621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018519799630162736,
+      "loss": 1.1335,
+      "step": 4622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018519065235007195,
+      "loss": 1.0216,
+      "step": 4623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018518330672281886,
+      "loss": 1.2261,
+      "step": 4624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851759594200127,
+      "loss": 1.2189,
+      "step": 4625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018516861044179793,
+      "loss": 0.9609,
+      "step": 4626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851612597883191,
+      "loss": 1.0782,
+      "step": 4627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018515390745972078,
+      "loss": 1.0108,
+      "step": 4628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018514655345614763,
+      "loss": 0.9595,
+      "step": 4629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001851391977777443,
+      "loss": 1.119,
+      "step": 4630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018513184042465546,
+      "loss": 1.0087,
+      "step": 4631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018512448139702584,
+      "loss": 0.999,
+      "step": 4632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018511712069500016,
+      "loss": 1.0767,
+      "step": 4633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018510975831872322,
+      "loss": 1.2375,
+      "step": 4634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018510239426833988,
+      "loss": 0.952,
+      "step": 4635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018509502854399497,
+      "loss": 1.2725,
+      "step": 4636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018508766114583332,
+      "loss": 1.0868,
+      "step": 4637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001850802920739999,
+      "loss": 1.1062,
+      "step": 4638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018507292132863965,
+      "loss": 0.9534,
+      "step": 4639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001850655489098976,
+      "loss": 0.9664,
+      "step": 4640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018505817481791862,
+      "loss": 1.0109,
+      "step": 4641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018505079905284792,
+      "loss": 0.9562,
+      "step": 4642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018504342161483047,
+      "loss": 1.0662,
+      "step": 4643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018503604250401145,
+      "loss": 0.9992,
+      "step": 4644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018502866172053598,
+      "loss": 1.2551,
+      "step": 4645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001850212792645493,
+      "loss": 1.1859,
+      "step": 4646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001850138951361965,
+      "loss": 0.9901,
+      "step": 4647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001850065093356229,
+      "loss": 0.9817,
+      "step": 4648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018499912186297377,
+      "loss": 0.9961,
+      "step": 4649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018499173271839444,
+      "loss": 1.07,
+      "step": 4650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018498434190203025,
+      "loss": 1.1372,
+      "step": 4651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018497694941402653,
+      "loss": 1.0429,
+      "step": 4652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018496955525452874,
+      "loss": 1.116,
+      "step": 4653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001849621594236823,
+      "loss": 1.2348,
+      "step": 4654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001849547619216327,
+      "loss": 1.0566,
+      "step": 4655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018494736274852545,
+      "loss": 1.0042,
+      "step": 4656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018493996190450607,
+      "loss": 1.0263,
+      "step": 4657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018493255938972013,
+      "loss": 1.0871,
+      "step": 4658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018492515520431328,
+      "loss": 1.0077,
+      "step": 4659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018491774934843114,
+      "loss": 1.0906,
+      "step": 4660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001849103418222194,
+      "loss": 1.0224,
+      "step": 4661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018490293262582368,
+      "loss": 1.1283,
+      "step": 4662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018489552175938981,
+      "loss": 0.978,
+      "step": 4663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018488810922306353,
+      "loss": 1.1241,
+      "step": 4664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018488069501699066,
+      "loss": 1.1431,
+      "step": 4665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018487327914131703,
+      "loss": 0.9387,
+      "step": 4666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018486586159618848,
+      "loss": 0.9891,
+      "step": 4667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018485844238175095,
+      "loss": 1.0454,
+      "step": 4668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018485102149815038,
+      "loss": 1.0176,
+      "step": 4669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001848435989455327,
+      "loss": 0.9466,
+      "step": 4670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018483617472404395,
+      "loss": 1.1289,
+      "step": 4671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018482874883383015,
+      "loss": 1.2432,
+      "step": 4672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018482132127503736,
+      "loss": 1.0482,
+      "step": 4673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018481389204781168,
+      "loss": 1.0666,
+      "step": 4674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018480646115229921,
+      "loss": 1.0361,
+      "step": 4675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001847990285886462,
+      "loss": 1.0841,
+      "step": 4676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018479159435699883,
+      "loss": 1.0365,
+      "step": 4677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018478415845750326,
+      "loss": 1.0976,
+      "step": 4678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018477672089030578,
+      "loss": 1.0151,
+      "step": 4679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018476928165555274,
+      "loss": 0.8767,
+      "step": 4680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001847618407533904,
+      "loss": 1.1336,
+      "step": 4681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018475439818396518,
+      "loss": 1.161,
+      "step": 4682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018474695394742345,
+      "loss": 1.0517,
+      "step": 4683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018473950804391164,
+      "loss": 0.9351,
+      "step": 4684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018473206047357619,
+      "loss": 0.9882,
+      "step": 4685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001847246112365636,
+      "loss": 1.151,
+      "step": 4686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018471716033302047,
+      "loss": 0.9413,
+      "step": 4687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018470970776309327,
+      "loss": 1.1192,
+      "step": 4688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018470225352692858,
+      "loss": 1.1268,
+      "step": 4689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018469479762467313,
+      "loss": 1.0347,
+      "step": 4690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018468734005647343,
+      "loss": 1.0324,
+      "step": 4691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001846798808224763,
+      "loss": 0.9927,
+      "step": 4692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018467241992282843,
+      "loss": 1.1174,
+      "step": 4693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018466495735767655,
+      "loss": 1.1233,
+      "step": 4694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018465749312716745,
+      "loss": 0.9605,
+      "step": 4695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018465002723144793,
+      "loss": 1.013,
+      "step": 4696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018464255967066493,
+      "loss": 0.9783,
+      "step": 4697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018463509044496523,
+      "loss": 1.1849,
+      "step": 4698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018462761955449585,
+      "loss": 0.8559,
+      "step": 4699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018462014699940365,
+      "loss": 1.1708,
+      "step": 4700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018461267277983568,
+      "loss": 1.1472,
+      "step": 4701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001846051968959389,
+      "loss": 1.0085,
+      "step": 4702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018459771934786043,
+      "loss": 1.0904,
+      "step": 4703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018459024013574733,
+      "loss": 1.1122,
+      "step": 4704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018458275925974666,
+      "loss": 1.0673,
+      "step": 4705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018457527672000566,
+      "loss": 0.9925,
+      "step": 4706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001845677925166714,
+      "loss": 0.9388,
+      "step": 4707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018456030664989122,
+      "loss": 1.0515,
+      "step": 4708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001845528191198123,
+      "loss": 0.9923,
+      "step": 4709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018454532992658192,
+      "loss": 0.9748,
+      "step": 4710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018453783907034738,
+      "loss": 1.1977,
+      "step": 4711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018453034655125606,
+      "loss": 0.8917,
+      "step": 4712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018452285236945527,
+      "loss": 1.0427,
+      "step": 4713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018451535652509252,
+      "loss": 1.227,
+      "step": 4714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001845078590183152,
+      "loss": 0.8939,
+      "step": 4715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018450035984927075,
+      "loss": 1.1388,
+      "step": 4716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018449285901810675,
+      "loss": 1.2605,
+      "step": 4717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018448535652497073,
+      "loss": 1.006,
+      "step": 4718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001844778523700102,
+      "loss": 0.9833,
+      "step": 4719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018447034655337283,
+      "loss": 1.0433,
+      "step": 4720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018446283907520622,
+      "loss": 1.0478,
+      "step": 4721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001844553299356581,
+      "loss": 1.0666,
+      "step": 4722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018444781913487608,
+      "loss": 1.1229,
+      "step": 4723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018444030667300797,
+      "loss": 1.2301,
+      "step": 4724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018443279255020152,
+      "loss": 1.1326,
+      "step": 4725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018442527676660453,
+      "loss": 0.9009,
+      "step": 4726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018441775932236484,
+      "loss": 1.0964,
+      "step": 4727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018441024021763032,
+      "loss": 1.1102,
+      "step": 4728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018440271945254887,
+      "loss": 0.9528,
+      "step": 4729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018439519702726843,
+      "loss": 0.9523,
+      "step": 4730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018438767294193695,
+      "loss": 0.8651,
+      "step": 4731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001843801471967024,
+      "loss": 1.1449,
+      "step": 4732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.0001843726197917129,
+      "loss": 1.0354,
+      "step": 4733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00018436509072711644,
+      "loss": 0.9132,
+      "step": 4734
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018435756000306116,
+      "loss": 1.1214,
+      "step": 4735
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018435002761969514,
+      "loss": 1.1259,
+      "step": 4736
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018434249357716654,
+      "loss": 1.1123,
+      "step": 4737
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018433495787562363,
+      "loss": 1.2122,
+      "step": 4738
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018432742051521455,
+      "loss": 1.3018,
+      "step": 4739
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018431988149608764,
+      "loss": 0.8766,
+      "step": 4740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001843123408183911,
+      "loss": 1.0685,
+      "step": 4741
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018430479848227335,
+      "loss": 0.8864,
+      "step": 4742
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018429725448788266,
+      "loss": 0.9887,
+      "step": 4743
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001842897088353675,
+      "loss": 0.9396,
+      "step": 4744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018428216152487626,
+      "loss": 1.0374,
+      "step": 4745
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018427461255655735,
+      "loss": 1.1702,
+      "step": 4746
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018426706193055934,
+      "loss": 1.0659,
+      "step": 4747
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001842595096470307,
+      "loss": 1.1396,
+      "step": 4748
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018425195570611998,
+      "loss": 1.1046,
+      "step": 4749
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001842444001079758,
+      "loss": 0.8581,
+      "step": 4750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018423684285274675,
+      "loss": 1.1714,
+      "step": 4751
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018422928394058144,
+      "loss": 1.3105,
+      "step": 4752
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018422172337162867,
+      "loss": 1.0695,
+      "step": 4753
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018421416114603706,
+      "loss": 1.0785,
+      "step": 4754
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018420659726395538,
+      "loss": 1.053,
+      "step": 4755
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018419903172553243,
+      "loss": 1.0726,
+      "step": 4756
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018419146453091701,
+      "loss": 1.1628,
+      "step": 4757
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018418389568025794,
+      "loss": 1.0887,
+      "step": 4758
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018417632517370419,
+      "loss": 0.9902,
+      "step": 4759
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018416875301140458,
+      "loss": 0.9361,
+      "step": 4760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018416117919350806,
+      "loss": 1.0536,
+      "step": 4761
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018415360372016365,
+      "loss": 0.9329,
+      "step": 4762
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018414602659152035,
+      "loss": 0.9858,
+      "step": 4763
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018413844780772718,
+      "loss": 1.0369,
+      "step": 4764
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001841308673689332,
+      "loss": 1.1727,
+      "step": 4765
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018412328527528756,
+      "loss": 0.9931,
+      "step": 4766
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018411570152693943,
+      "loss": 1.0927,
+      "step": 4767
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001841081161240379,
+      "loss": 1.1528,
+      "step": 4768
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018410052906673219,
+      "loss": 1.0447,
+      "step": 4769
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018409294035517156,
+      "loss": 1.0876,
+      "step": 4770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001840853499895053,
+      "loss": 1.1737,
+      "step": 4771
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018407775796988265,
+      "loss": 1.0693,
+      "step": 4772
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018407016429645303,
+      "loss": 1.0379,
+      "step": 4773
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018406256896936574,
+      "loss": 1.0742,
+      "step": 4774
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001840549719887702,
+      "loss": 1.023,
+      "step": 4775
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018404737335481582,
+      "loss": 1.1837,
+      "step": 4776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018403977306765212,
+      "loss": 0.8925,
+      "step": 4777
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018403217112742857,
+      "loss": 0.9992,
+      "step": 4778
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018402456753429468,
+      "loss": 0.9502,
+      "step": 4779
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018401696228840001,
+      "loss": 1.067,
+      "step": 4780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001840093553898942,
+      "loss": 1.1422,
+      "step": 4781
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018400174683892684,
+      "loss": 0.9397,
+      "step": 4782
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001839941366356476,
+      "loss": 1.0656,
+      "step": 4783
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018398652478020616,
+      "loss": 0.9574,
+      "step": 4784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018397891127275228,
+      "loss": 1.107,
+      "step": 4785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001839712961134357,
+      "loss": 1.1496,
+      "step": 4786
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018396367930240617,
+      "loss": 1.0467,
+      "step": 4787
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001839560608398136,
+      "loss": 1.1007,
+      "step": 4788
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018394844072580773,
+      "loss": 1.1267,
+      "step": 4789
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018394081896053855,
+      "loss": 1.0224,
+      "step": 4790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001839331955441559,
+      "loss": 0.9902,
+      "step": 4791
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018392557047680982,
+      "loss": 0.936,
+      "step": 4792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018391794375865024,
+      "loss": 1.0926,
+      "step": 4793
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018391031538982718,
+      "loss": 1.0352,
+      "step": 4794
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018390268537049064,
+      "loss": 1.0552,
+      "step": 4795
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001838950537007908,
+      "loss": 1.138,
+      "step": 4796
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018388742038087777,
+      "loss": 0.9599,
+      "step": 4797
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018387978541090162,
+      "loss": 1.0836,
+      "step": 4798
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018387214879101254,
+      "loss": 0.9826,
+      "step": 4799
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001838645105213608,
+      "loss": 1.0874,
+      "step": 4800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001838568706020966,
+      "loss": 1.1234,
+      "step": 4801
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018384922903337024,
+      "loss": 1.015,
+      "step": 4802
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018384158581533205,
+      "loss": 1.2707,
+      "step": 4803
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018383394094813232,
+      "loss": 1.1358,
+      "step": 4804
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018382629443192144,
+      "loss": 1.1848,
+      "step": 4805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001838186462668498,
+      "loss": 0.8829,
+      "step": 4806
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018381099645306792,
+      "loss": 1.2001,
+      "step": 4807
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018380334499072616,
+      "loss": 1.1219,
+      "step": 4808
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018379569187997513,
+      "loss": 1.1807,
+      "step": 4809
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018378803712096526,
+      "loss": 1.0133,
+      "step": 4810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018378038071384723,
+      "loss": 1.0338,
+      "step": 4811
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018377272265877154,
+      "loss": 1.0615,
+      "step": 4812
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018376506295588886,
+      "loss": 1.2173,
+      "step": 4813
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001837574016053499,
+      "loss": 1.1996,
+      "step": 4814
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018374973860730531,
+      "loss": 1.1752,
+      "step": 4815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018374207396190583,
+      "loss": 0.9252,
+      "step": 4816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018373440766930225,
+      "loss": 1.1493,
+      "step": 4817
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018372673972964535,
+      "loss": 0.8368,
+      "step": 4818
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018371907014308589,
+      "loss": 1.0964,
+      "step": 4819
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018371139890977484,
+      "loss": 0.9509,
+      "step": 4820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018370372602986303,
+      "loss": 0.9359,
+      "step": 4821
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018369605150350143,
+      "loss": 1.1754,
+      "step": 4822
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018368837533084095,
+      "loss": 1.4129,
+      "step": 4823
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001836806975120326,
+      "loss": 0.9111,
+      "step": 4824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001836730180472274,
+      "loss": 0.9382,
+      "step": 4825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001836653369365764,
+      "loss": 1.0265,
+      "step": 4826
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018365765418023068,
+      "loss": 1.2135,
+      "step": 4827
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018364996977834137,
+      "loss": 1.1144,
+      "step": 4828
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018364228373105964,
+      "loss": 1.0941,
+      "step": 4829
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018363459603853665,
+      "loss": 0.9802,
+      "step": 4830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018362690670092367,
+      "loss": 0.7953,
+      "step": 4831
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018361921571837187,
+      "loss": 1.0572,
+      "step": 4832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018361152309103257,
+      "loss": 0.9601,
+      "step": 4833
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018360382881905704,
+      "loss": 1.039,
+      "step": 4834
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001835961329025967,
+      "loss": 1.0858,
+      "step": 4835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001835884353418029,
+      "loss": 0.9476,
+      "step": 4836
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018358073613682706,
+      "loss": 1.232,
+      "step": 4837
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018357303528782057,
+      "loss": 1.1046,
+      "step": 4838
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018356533279493496,
+      "loss": 1.0403,
+      "step": 4839
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001835576286583217,
+      "loss": 1.0987,
+      "step": 4840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018354992287813237,
+      "loss": 1.4379,
+      "step": 4841
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018354221545451856,
+      "loss": 1.0462,
+      "step": 4842
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018353450638763179,
+      "loss": 1.012,
+      "step": 4843
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018352679567762374,
+      "loss": 1.2002,
+      "step": 4844
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001835190833246461,
+      "loss": 1.1096,
+      "step": 4845
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018351136932885057,
+      "loss": 1.3085,
+      "step": 4846
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018350365369038885,
+      "loss": 1.0249,
+      "step": 4847
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018349593640941278,
+      "loss": 1.1599,
+      "step": 4848
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018348821748607406,
+      "loss": 0.9811,
+      "step": 4849
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018348049692052456,
+      "loss": 1.0557,
+      "step": 4850
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018347277471291617,
+      "loss": 0.9026,
+      "step": 4851
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001834650508634008,
+      "loss": 1.0499,
+      "step": 4852
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018345732537213027,
+      "loss": 1.098,
+      "step": 4853
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018344959823925666,
+      "loss": 0.9954,
+      "step": 4854
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001834418694649319,
+      "loss": 1.16,
+      "step": 4855
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018343413904930804,
+      "loss": 1.2355,
+      "step": 4856
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018342640699253713,
+      "loss": 1.2513,
+      "step": 4857
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018341867329477125,
+      "loss": 1.2491,
+      "step": 4858
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018341093795616253,
+      "loss": 1.1147,
+      "step": 4859
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018340320097686311,
+      "loss": 0.9863,
+      "step": 4860
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018339546235702524,
+      "loss": 1.1249,
+      "step": 4861
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018338772209680105,
+      "loss": 1.1071,
+      "step": 4862
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018337998019634283,
+      "loss": 1.0356,
+      "step": 4863
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018337223665580284,
+      "loss": 1.0697,
+      "step": 4864
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018336449147533345,
+      "loss": 1.2059,
+      "step": 4865
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018335674465508697,
+      "loss": 1.1162,
+      "step": 4866
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018334899619521576,
+      "loss": 1.1799,
+      "step": 4867
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018334124609587228,
+      "loss": 0.968,
+      "step": 4868
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018333349435720894,
+      "loss": 1.0151,
+      "step": 4869
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018332574097937822,
+      "loss": 0.9342,
+      "step": 4870
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018331798596253265,
+      "loss": 1.1148,
+      "step": 4871
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018331022930682478,
+      "loss": 1.0901,
+      "step": 4872
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018330247101240712,
+      "loss": 0.8325,
+      "step": 4873
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001832947110794323,
+      "loss": 1.0795,
+      "step": 4874
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.000183286949508053,
+      "loss": 0.969,
+      "step": 4875
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018327918629842187,
+      "loss": 1.1549,
+      "step": 4876
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001832714214506916,
+      "loss": 1.2058,
+      "step": 4877
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018326365496501495,
+      "loss": 1.0363,
+      "step": 4878
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001832558868415446,
+      "loss": 0.9777,
+      "step": 4879
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001832481170804335,
+      "loss": 1.1861,
+      "step": 4880
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001832403456818343,
+      "loss": 1.1557,
+      "step": 4881
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018323257264590002,
+      "loss": 1.2102,
+      "step": 4882
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018322479797278348,
+      "loss": 0.894,
+      "step": 4883
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018321702166263762,
+      "loss": 1.1995,
+      "step": 4884
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001832092437156154,
+      "loss": 1.0316,
+      "step": 4885
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018320146413186982,
+      "loss": 1.1426,
+      "step": 4886
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018319368291155386,
+      "loss": 1.1344,
+      "step": 4887
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018318590005482067,
+      "loss": 1.114,
+      "step": 4888
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018317811556182325,
+      "loss": 1.0336,
+      "step": 4889
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018317032943271477,
+      "loss": 1.1899,
+      "step": 4890
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018316254166764835,
+      "loss": 1.1401,
+      "step": 4891
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001831547522667772,
+      "loss": 0.9672,
+      "step": 4892
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018314696123025454,
+      "loss": 1.0408,
+      "step": 4893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018313916855823358,
+      "loss": 1.1842,
+      "step": 4894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018313137425086766,
+      "loss": 1.0514,
+      "step": 4895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018312357830831003,
+      "loss": 1.0487,
+      "step": 4896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001831157807307141,
+      "loss": 0.9442,
+      "step": 4897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018310798151823318,
+      "loss": 1.1111,
+      "step": 4898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018310018067102074,
+      "loss": 0.9519,
+      "step": 4899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001830923781892302,
+      "loss": 0.9976,
+      "step": 4900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018308457407301503,
+      "loss": 1.1982,
+      "step": 4901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018307676832252876,
+      "loss": 0.9693,
+      "step": 4902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001830689609379249,
+      "loss": 1.2069,
+      "step": 4903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018306115191935702,
+      "loss": 1.0732,
+      "step": 4904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018305334126697873,
+      "loss": 1.094,
+      "step": 4905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018304552898094366,
+      "loss": 1.0828,
+      "step": 4906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018303771506140553,
+      "loss": 0.9955,
+      "step": 4907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018302989950851796,
+      "loss": 1.1319,
+      "step": 4908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018302208232243473,
+      "loss": 1.0917,
+      "step": 4909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001830142635033096,
+      "loss": 0.9125,
+      "step": 4910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018300644305129635,
+      "loss": 0.9084,
+      "step": 4911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001829986209665488,
+      "loss": 1.0543,
+      "step": 4912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018299079724922086,
+      "loss": 1.1413,
+      "step": 4913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018298297189946636,
+      "loss": 0.967,
+      "step": 4914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018297514491743928,
+      "loss": 1.1441,
+      "step": 4915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018296731630329353,
+      "loss": 1.176,
+      "step": 4916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018295948605718314,
+      "loss": 1.212,
+      "step": 4917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018295165417926207,
+      "loss": 1.1447,
+      "step": 4918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018294382066968442,
+      "loss": 0.9138,
+      "step": 4919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018293598552860432,
+      "loss": 0.9547,
+      "step": 4920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018292814875617576,
+      "loss": 1.0345,
+      "step": 4921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018292031035255303,
+      "loss": 1.2385,
+      "step": 4922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018291247031789018,
+      "loss": 1.0066,
+      "step": 4923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018290462865234154,
+      "loss": 1.0967,
+      "step": 4924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018289678535606126,
+      "loss": 1.2908,
+      "step": 4925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018288894042920367,
+      "loss": 1.0667,
+      "step": 4926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018288109387192307,
+      "loss": 1.0085,
+      "step": 4927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018287324568437381,
+      "loss": 0.8726,
+      "step": 4928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018286539586671025,
+      "loss": 1.0954,
+      "step": 4929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001828575444190868,
+      "loss": 1.094,
+      "step": 4930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001828496913416579,
+      "loss": 1.3039,
+      "step": 4931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018284183663457802,
+      "loss": 1.0271,
+      "step": 4932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018283398029800166,
+      "loss": 1.0925,
+      "step": 4933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018282612233208335,
+      "loss": 1.3114,
+      "step": 4934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018281826273697767,
+      "loss": 1.2437,
+      "step": 4935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001828104015128392,
+      "loss": 1.2265,
+      "step": 4936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018280253865982256,
+      "loss": 1.068,
+      "step": 4937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018279467417808246,
+      "loss": 1.158,
+      "step": 4938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018278680806777352,
+      "loss": 1.036,
+      "step": 4939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018277894032905055,
+      "loss": 0.9714,
+      "step": 4940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018277107096206826,
+      "loss": 1.1644,
+      "step": 4941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018276319996698143,
+      "loss": 1.0806,
+      "step": 4942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001827553273439449,
+      "loss": 0.9427,
+      "step": 4943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018274745309311357,
+      "loss": 1.0097,
+      "step": 4944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018273957721464223,
+      "loss": 0.9895,
+      "step": 4945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018273169970868587,
+      "loss": 1.2849,
+      "step": 4946
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001827238205753994,
+      "loss": 0.9999,
+      "step": 4947
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018271593981493783,
+      "loss": 1.036,
+      "step": 4948
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018270805742745617,
+      "loss": 0.945,
+      "step": 4949
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018270017341310947,
+      "loss": 1.2579,
+      "step": 4950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001826922877720528,
+      "loss": 1.0475,
+      "step": 4951
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018268440050444129,
+      "loss": 1.0739,
+      "step": 4952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018267651161043004,
+      "loss": 0.7608,
+      "step": 4953
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018266862109017425,
+      "loss": 0.9748,
+      "step": 4954
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018266072894382914,
+      "loss": 1.2174,
+      "step": 4955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001826528351715499,
+      "loss": 1.1791,
+      "step": 4956
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018264493977349184,
+      "loss": 1.186,
+      "step": 4957
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018263704274981026,
+      "loss": 1.0609,
+      "step": 4958
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001826291441006605,
+      "loss": 1.3517,
+      "step": 4959
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018262124382619792,
+      "loss": 1.0454,
+      "step": 4960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001826133419265779,
+      "loss": 1.0831,
+      "step": 4961
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018260543840195587,
+      "loss": 0.947,
+      "step": 4962
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001825975332524873,
+      "loss": 1.0952,
+      "step": 4963
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.00018258962647832775,
+      "loss": 1.027,
+      "step": 4964
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0001825817180796326,
+      "loss": 0.9574,
+      "step": 4965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001825738080565576,
+      "loss": 0.9706,
+      "step": 4966
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018256589640925817,
+      "loss": 0.9165,
+      "step": 4967
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018255798313789,
+      "loss": 1.271,
+      "step": 4968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001825500682426087,
+      "loss": 0.9345,
+      "step": 4969
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018254215172357004,
+      "loss": 1.0597,
+      "step": 4970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001825342335809297,
+      "loss": 1.1276,
+      "step": 4971
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001825263138148434,
+      "loss": 1.0672,
+      "step": 4972
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018251839242546696,
+      "loss": 1.1175,
+      "step": 4973
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018251046941295616,
+      "loss": 1.1561,
+      "step": 4974
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018250254477746687,
+      "loss": 0.9336,
+      "step": 4975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018249461851915498,
+      "loss": 1.2269,
+      "step": 4976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018248669063817636,
+      "loss": 0.9566,
+      "step": 4977
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018247876113468699,
+      "loss": 1.0835,
+      "step": 4978
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018247083000884278,
+      "loss": 1.0377,
+      "step": 4979
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001824628972607998,
+      "loss": 0.9816,
+      "step": 4980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001824549628907141,
+      "loss": 1.1682,
+      "step": 4981
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018244702689874167,
+      "loss": 1.1098,
+      "step": 4982
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018243908928503868,
+      "loss": 1.1223,
+      "step": 4983
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001824311500497612,
+      "loss": 1.0842,
+      "step": 4984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018242320919306548,
+      "loss": 1.3148,
+      "step": 4985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018241526671510764,
+      "loss": 1.0593,
+      "step": 4986
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018240732261604395,
+      "loss": 1.0955,
+      "step": 4987
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018239937689603066,
+      "loss": 1.1762,
+      "step": 4988
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018239142955522406,
+      "loss": 0.9485,
+      "step": 4989
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018238348059378048,
+      "loss": 1.0658,
+      "step": 4990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018237553001185624,
+      "loss": 1.0831,
+      "step": 4991
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018236757780960776,
+      "loss": 1.1091,
+      "step": 4992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018235962398719147,
+      "loss": 1.1574,
+      "step": 4993
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018235166854476382,
+      "loss": 0.8822,
+      "step": 4994
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018234371148248126,
+      "loss": 1.1203,
+      "step": 4995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018233575280050034,
+      "loss": 1.1275,
+      "step": 4996
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001823277924989776,
+      "loss": 1.0903,
+      "step": 4997
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001823198305780696,
+      "loss": 1.1029,
+      "step": 4998
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018231186703793296,
+      "loss": 1.1693,
+      "step": 4999
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018230390187872432,
+      "loss": 1.3235,
+      "step": 5000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001822959351006004,
+      "loss": 0.9393,
+      "step": 5001
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001822879667037178,
+      "loss": 1.0315,
+      "step": 5002
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018227999668823336,
+      "loss": 0.9243,
+      "step": 5003
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018227202505430383,
+      "loss": 1.1549,
+      "step": 5004
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000182264051802086,
+      "loss": 1.1413,
+      "step": 5005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018225607693173667,
+      "loss": 1.1477,
+      "step": 5006
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018224810044341276,
+      "loss": 0.9411,
+      "step": 5007
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018224012233727114,
+      "loss": 1.0106,
+      "step": 5008
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018223214261346872,
+      "loss": 1.1991,
+      "step": 5009
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001822241612721625,
+      "loss": 1.1394,
+      "step": 5010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018221617831350946,
+      "loss": 1.0648,
+      "step": 5011
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001822081937376666,
+      "loss": 1.176,
+      "step": 5012
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018220020754479102,
+      "loss": 0.9737,
+      "step": 5013
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018219221973503978,
+      "loss": 1.091,
+      "step": 5014
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018218423030856996,
+      "loss": 1.0917,
+      "step": 5015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001821762392655388,
+      "loss": 0.9669,
+      "step": 5016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018216824660610343,
+      "loss": 1.1468,
+      "step": 5017
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001821602523304211,
+      "loss": 1.1513,
+      "step": 5018
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000182152256438649,
+      "loss": 1.0026,
+      "step": 5019
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018214425893094444,
+      "loss": 0.9553,
+      "step": 5020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018213625980746475,
+      "loss": 1.0267,
+      "step": 5021
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018212825906836725,
+      "loss": 0.9991,
+      "step": 5022
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018212025671380933,
+      "loss": 1.1386,
+      "step": 5023
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018211225274394836,
+      "loss": 1.2251,
+      "step": 5024
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018210424715894182,
+      "loss": 1.3105,
+      "step": 5025
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018209623995894716,
+      "loss": 1.0039,
+      "step": 5026
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001820882311441219,
+      "loss": 0.9854,
+      "step": 5027
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018208022071462354,
+      "loss": 1.0546,
+      "step": 5028
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018207220867060966,
+      "loss": 1.1293,
+      "step": 5029
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018206419501223784,
+      "loss": 0.8839,
+      "step": 5030
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018205617973966574,
+      "loss": 1.0188,
+      "step": 5031
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018204816285305103,
+      "loss": 1.1426,
+      "step": 5032
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018204014435255135,
+      "loss": 1.0388,
+      "step": 5033
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018203212423832448,
+      "loss": 1.0178,
+      "step": 5034
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018202410251052815,
+      "loss": 1.1655,
+      "step": 5035
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018201607916932012,
+      "loss": 0.8869,
+      "step": 5036
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018200805421485824,
+      "loss": 1.1186,
+      "step": 5037
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018200002764730036,
+      "loss": 1.3209,
+      "step": 5038
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018199199946680438,
+      "loss": 0.9967,
+      "step": 5039
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018198396967352814,
+      "loss": 1.0946,
+      "step": 5040
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018197593826762968,
+      "loss": 1.0733,
+      "step": 5041
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018196790524926693,
+      "loss": 1.0316,
+      "step": 5042
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001819598706185979,
+      "loss": 1.0089,
+      "step": 5043
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018195183437578062,
+      "loss": 1.1112,
+      "step": 5044
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001819437965209732,
+      "loss": 0.9936,
+      "step": 5045
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001819357570543337,
+      "loss": 1.1809,
+      "step": 5046
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001819277159760203,
+      "loss": 1.1305,
+      "step": 5047
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018191967328619113,
+      "loss": 0.9925,
+      "step": 5048
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018191162898500442,
+      "loss": 1.2516,
+      "step": 5049
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018190358307261837,
+      "loss": 1.2492,
+      "step": 5050
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018189553554919126,
+      "loss": 0.9881,
+      "step": 5051
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001818874864148814,
+      "loss": 1.0361,
+      "step": 5052
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001818794356698471,
+      "loss": 1.1112,
+      "step": 5053
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001818713833142467,
+      "loss": 1.0033,
+      "step": 5054
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018186332934823859,
+      "loss": 1.0665,
+      "step": 5055
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018185527377198124,
+      "loss": 0.8754,
+      "step": 5056
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001818472165856331,
+      "loss": 1.0758,
+      "step": 5057
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018183915778935257,
+      "loss": 1.1977,
+      "step": 5058
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018183109738329823,
+      "loss": 1.1951,
+      "step": 5059
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018182303536762862,
+      "loss": 0.9966,
+      "step": 5060
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018181497174250236,
+      "loss": 1.0466,
+      "step": 5061
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018180690650807798,
+      "loss": 1.1192,
+      "step": 5062
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018179883966451418,
+      "loss": 1.1258,
+      "step": 5063
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018179077121196964,
+      "loss": 0.8937,
+      "step": 5064
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018178270115060303,
+      "loss": 1.0934,
+      "step": 5065
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018177462948057308,
+      "loss": 1.0867,
+      "step": 5066
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018176655620203862,
+      "loss": 1.1312,
+      "step": 5067
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018175848131515837,
+      "loss": 1.2954,
+      "step": 5068
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001817504048200912,
+      "loss": 1.0022,
+      "step": 5069
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018174232671699608,
+      "loss": 1.2674,
+      "step": 5070
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001817342470060317,
+      "loss": 0.9492,
+      "step": 5071
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018172616568735715,
+      "loss": 0.9794,
+      "step": 5072
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018171808276113132,
+      "loss": 1.0517,
+      "step": 5073
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001817099982275132,
+      "loss": 1.0851,
+      "step": 5074
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018170191208666185,
+      "loss": 1.0683,
+      "step": 5075
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018169382433873627,
+      "loss": 1.2993,
+      "step": 5076
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018168573498389564,
+      "loss": 1.2026,
+      "step": 5077
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018167764402229897,
+      "loss": 0.9663,
+      "step": 5078
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018166955145410544,
+      "loss": 0.9947,
+      "step": 5079
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018166145727947428,
+      "loss": 1.0615,
+      "step": 5080
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018165336149856463,
+      "loss": 1.4667,
+      "step": 5081
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018164526411153578,
+      "loss": 1.386,
+      "step": 5082
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018163716511854702,
+      "loss": 1.2424,
+      "step": 5083
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001816290645197576,
+      "loss": 1.1419,
+      "step": 5084
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018162096231532694,
+      "loss": 1.0527,
+      "step": 5085
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001816128585054143,
+      "loss": 1.1842,
+      "step": 5086
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018160475309017917,
+      "loss": 1.1771,
+      "step": 5087
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018159664606978097,
+      "loss": 1.0167,
+      "step": 5088
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018158853744437914,
+      "loss": 0.8006,
+      "step": 5089
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018158042721413316,
+      "loss": 1.24,
+      "step": 5090
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018157231537920264,
+      "loss": 1.173,
+      "step": 5091
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018156420193974704,
+      "loss": 1.1417,
+      "step": 5092
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018155608689592604,
+      "loss": 1.0672,
+      "step": 5093
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018154797024789917,
+      "loss": 1.026,
+      "step": 5094
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018153985199582617,
+      "loss": 1.2701,
+      "step": 5095
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018153173213986668,
+      "loss": 0.9285,
+      "step": 5096
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001815236106801804,
+      "loss": 1.0235,
+      "step": 5097
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018151548761692713,
+      "loss": 1.2762,
+      "step": 5098
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018150736295026662,
+      "loss": 1.1117,
+      "step": 5099
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001814992366803587,
+      "loss": 1.1423,
+      "step": 5100
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001814911088073632,
+      "loss": 1.0249,
+      "step": 5101
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018148297933143995,
+      "loss": 1.2224,
+      "step": 5102
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018147484825274893,
+      "loss": 1.0197,
+      "step": 5103
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001814667155714501,
+      "loss": 1.0841,
+      "step": 5104
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001814585812877033,
+      "loss": 0.9331,
+      "step": 5105
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018145044540166866,
+      "loss": 1.1291,
+      "step": 5106
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018144230791350616,
+      "loss": 1.0459,
+      "step": 5107
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018143416882337586,
+      "loss": 1.1536,
+      "step": 5108
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018142602813143785,
+      "loss": 0.9268,
+      "step": 5109
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001814178858378523,
+      "loss": 1.1725,
+      "step": 5110
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018140974194277932,
+      "loss": 1.3398,
+      "step": 5111
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018140159644637914,
+      "loss": 1.2883,
+      "step": 5112
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018139344934881195,
+      "loss": 1.0006,
+      "step": 5113
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000181385300650238,
+      "loss": 1.0595,
+      "step": 5114
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001813771503508176,
+      "loss": 1.0594,
+      "step": 5115
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018136899845071108,
+      "loss": 1.1472,
+      "step": 5116
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018136084495007872,
+      "loss": 0.929,
+      "step": 5117
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018135268984908096,
+      "loss": 1.0193,
+      "step": 5118
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018134453314787817,
+      "loss": 1.0749,
+      "step": 5119
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018133637484663085,
+      "loss": 1.1377,
+      "step": 5120
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018132821494549944,
+      "loss": 1.0988,
+      "step": 5121
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001813200534446444,
+      "loss": 1.0916,
+      "step": 5122
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018131189034422635,
+      "loss": 1.0007,
+      "step": 5123
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018130372564440582,
+      "loss": 1.1627,
+      "step": 5124
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812955593453434,
+      "loss": 1.13,
+      "step": 5125
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018128739144719972,
+      "loss": 1.1967,
+      "step": 5126
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018127922195013545,
+      "loss": 1.0939,
+      "step": 5127
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812710508543113,
+      "loss": 1.0729,
+      "step": 5128
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018126287815988797,
+      "loss": 1.1071,
+      "step": 5129
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812547038670262,
+      "loss": 1.2067,
+      "step": 5130
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812465279758868,
+      "loss": 1.0154,
+      "step": 5131
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018123835048663062,
+      "loss": 1.1405,
+      "step": 5132
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812301713994185,
+      "loss": 1.0591,
+      "step": 5133
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812219907144113,
+      "loss": 1.0817,
+      "step": 5134
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001812138084317699,
+      "loss": 1.2002,
+      "step": 5135
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018120562455165536,
+      "loss": 1.1907,
+      "step": 5136
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018119743907422852,
+      "loss": 0.801,
+      "step": 5137
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018118925199965048,
+      "loss": 1.0414,
+      "step": 5138
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018118106332808224,
+      "loss": 0.9662,
+      "step": 5139
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018117287305968488,
+      "loss": 1.1705,
+      "step": 5140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018116468119461953,
+      "loss": 1.0627,
+      "step": 5141
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018115648773304725,
+      "loss": 1.1717,
+      "step": 5142
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018114829267512928,
+      "loss": 1.2588,
+      "step": 5143
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018114009602102679,
+      "loss": 1.1559,
+      "step": 5144
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.000181131897770901,
+      "loss": 1.0467,
+      "step": 5145
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001811236979249132,
+      "loss": 1.121,
+      "step": 5146
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018111549648322463,
+      "loss": 0.887,
+      "step": 5147
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018110729344599665,
+      "loss": 0.9186,
+      "step": 5148
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018109908881339058,
+      "loss": 1.1331,
+      "step": 5149
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018109088258556785,
+      "loss": 1.1087,
+      "step": 5150
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018108267476268987,
+      "loss": 0.997,
+      "step": 5151
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018107446534491804,
+      "loss": 0.9738,
+      "step": 5152
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001810662543324139,
+      "loss": 1.0712,
+      "step": 5153
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001810580417253389,
+      "loss": 0.9957,
+      "step": 5154
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001810498275238546,
+      "loss": 1.0832,
+      "step": 5155
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018104161172812261,
+      "loss": 0.882,
+      "step": 5156
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018103339433830454,
+      "loss": 0.962,
+      "step": 5157
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018102517535456195,
+      "loss": 1.1866,
+      "step": 5158
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001810169547770566,
+      "loss": 1.2083,
+      "step": 5159
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001810087326059501,
+      "loss": 1.0922,
+      "step": 5160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018100050884140421,
+      "loss": 1.1968,
+      "step": 5161
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018099228348358076,
+      "loss": 1.0523,
+      "step": 5162
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018098405653264144,
+      "loss": 0.8998,
+      "step": 5163
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018097582798874812,
+      "loss": 0.8102,
+      "step": 5164
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018096759785206267,
+      "loss": 1.0265,
+      "step": 5165
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018095936612274696,
+      "loss": 1.0561,
+      "step": 5166
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001809511328009629,
+      "loss": 0.9431,
+      "step": 5167
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018094289788687245,
+      "loss": 0.978,
+      "step": 5168
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001809346613806376,
+      "loss": 0.9655,
+      "step": 5169
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018092642328242033,
+      "loss": 0.946,
+      "step": 5170
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018091818359238272,
+      "loss": 1.3012,
+      "step": 5171
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018090994231068682,
+      "loss": 0.9254,
+      "step": 5172
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018090169943749476,
+      "loss": 1.348,
+      "step": 5173
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018089345497296865,
+      "loss": 1.2187,
+      "step": 5174
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018088520891727065,
+      "loss": 1.1707,
+      "step": 5175
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018087696127056302,
+      "loss": 0.996,
+      "step": 5176
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001808687120330079,
+      "loss": 0.8737,
+      "step": 5177
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018086046120476763,
+      "loss": 1.2848,
+      "step": 5178
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018085220878600447,
+      "loss": 1.0615,
+      "step": 5179
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018084395477688077,
+      "loss": 0.999,
+      "step": 5180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018083569917755885,
+      "loss": 1.1528,
+      "step": 5181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001808274419882011,
+      "loss": 0.9523,
+      "step": 5182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018081918320896997,
+      "loss": 1.1572,
+      "step": 5183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001808109228400279,
+      "loss": 1.1107,
+      "step": 5184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018080266088153736,
+      "loss": 1.0931,
+      "step": 5185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018079439733366086,
+      "loss": 1.043,
+      "step": 5186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018078613219656094,
+      "loss": 1.1213,
+      "step": 5187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018077786547040017,
+      "loss": 1.0925,
+      "step": 5188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018076959715534124,
+      "loss": 1.176,
+      "step": 5189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018076132725154668,
+      "loss": 1.1489,
+      "step": 5190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018075305575917917,
+      "loss": 1.0427,
+      "step": 5191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018074478267840144,
+      "loss": 1.0884,
+      "step": 5192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018073650800937624,
+      "loss": 1.1805,
+      "step": 5193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0001807282317522663,
+      "loss": 1.0398,
+      "step": 5194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00018071995390723443,
+      "loss": 1.0303,
+      "step": 5195
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018071167447444348,
+      "loss": 1.0296,
+      "step": 5196
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018070339345405625,
+      "loss": 1.0937,
+      "step": 5197
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018069511084623568,
+      "loss": 1.0686,
+      "step": 5198
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018068682665114462,
+      "loss": 0.8634,
+      "step": 5199
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018067854086894607,
+      "loss": 0.959,
+      "step": 5200
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000180670253499803,
+      "loss": 1.1439,
+      "step": 5201
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018066196454387845,
+      "loss": 0.9008,
+      "step": 5202
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018065367400133542,
+      "loss": 1.0649,
+      "step": 5203
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000180645381872337,
+      "loss": 1.106,
+      "step": 5204
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018063708815704632,
+      "loss": 0.9728,
+      "step": 5205
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001806287928556265,
+      "loss": 0.9463,
+      "step": 5206
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001806204959682407,
+      "loss": 1.1378,
+      "step": 5207
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001806121974950521,
+      "loss": 1.1854,
+      "step": 5208
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018060389743622399,
+      "loss": 0.9399,
+      "step": 5209
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018059559579191957,
+      "loss": 1.3354,
+      "step": 5210
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018058729256230217,
+      "loss": 1.1201,
+      "step": 5211
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001805789877475351,
+      "loss": 1.1659,
+      "step": 5212
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018057068134778173,
+      "loss": 1.0865,
+      "step": 5213
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018056237336320542,
+      "loss": 0.9979,
+      "step": 5214
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018055406379396965,
+      "loss": 1.0368,
+      "step": 5215
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018054575264023776,
+      "loss": 1.1638,
+      "step": 5216
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018053743990217335,
+      "loss": 1.0703,
+      "step": 5217
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001805291255799398,
+      "loss": 0.9776,
+      "step": 5218
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018052080967370082,
+      "loss": 1.1273,
+      "step": 5219
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018051249218361984,
+      "loss": 1.0619,
+      "step": 5220
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018050417310986052,
+      "loss": 1.118,
+      "step": 5221
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001804958524525865,
+      "loss": 0.9226,
+      "step": 5222
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018048753021196142,
+      "loss": 1.0251,
+      "step": 5223
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018047920638814905,
+      "loss": 1.073,
+      "step": 5224
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000180470880981313,
+      "loss": 0.9881,
+      "step": 5225
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018046255399161711,
+      "loss": 1.0672,
+      "step": 5226
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018045422541922517,
+      "loss": 1.075,
+      "step": 5227
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018044589526430103,
+      "loss": 0.893,
+      "step": 5228
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018043756352700846,
+      "loss": 1.2571,
+      "step": 5229
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001804292302075114,
+      "loss": 1.0722,
+      "step": 5230
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018042089530597374,
+      "loss": 0.9935,
+      "step": 5231
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001804125588225595,
+      "loss": 0.9762,
+      "step": 5232
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018040422075743252,
+      "loss": 1.0294,
+      "step": 5233
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018039588111075697,
+      "loss": 0.9904,
+      "step": 5234
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018038753988269677,
+      "loss": 1.2173,
+      "step": 5235
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018037919707341603,
+      "loss": 1.0664,
+      "step": 5236
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018037085268307889,
+      "loss": 1.1672,
+      "step": 5237
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018036250671184942,
+      "loss": 0.8787,
+      "step": 5238
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018035415915989184,
+      "loss": 1.0563,
+      "step": 5239
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001803458100273703,
+      "loss": 0.8041,
+      "step": 5240
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018033745931444906,
+      "loss": 1.0611,
+      "step": 5241
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018032910702129234,
+      "loss": 1.1486,
+      "step": 5242
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018032075314806448,
+      "loss": 1.0868,
+      "step": 5243
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001803123976949298,
+      "loss": 1.006,
+      "step": 5244
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001803040406620526,
+      "loss": 0.9089,
+      "step": 5245
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001802956820495973,
+      "loss": 1.1985,
+      "step": 5246
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001802873218577283,
+      "loss": 1.239,
+      "step": 5247
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018027896008661006,
+      "loss": 0.8555,
+      "step": 5248
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018027059673640704,
+      "loss": 1.2922,
+      "step": 5249
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018026223180728375,
+      "loss": 1.0212,
+      "step": 5250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018025386529940473,
+      "loss": 1.0406,
+      "step": 5251
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018024549721293455,
+      "loss": 1.2045,
+      "step": 5252
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001802371275480378,
+      "loss": 1.2163,
+      "step": 5253
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018022875630487915,
+      "loss": 1.2418,
+      "step": 5254
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001802203834836232,
+      "loss": 1.063,
+      "step": 5255
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018021200908443466,
+      "loss": 1.0795,
+      "step": 5256
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018020363310747834,
+      "loss": 0.9983,
+      "step": 5257
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018019525555291885,
+      "loss": 1.2193,
+      "step": 5258
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001801868764209211,
+      "loss": 0.9722,
+      "step": 5259
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018017849571164984,
+      "loss": 1.2295,
+      "step": 5260
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001801701134252699,
+      "loss": 1.2529,
+      "step": 5261
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018016172956194625,
+      "loss": 0.943,
+      "step": 5262
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018015334412184374,
+      "loss": 0.9972,
+      "step": 5263
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018014495710512733,
+      "loss": 1.2291,
+      "step": 5264
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018013656851196194,
+      "loss": 1.2896,
+      "step": 5265
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018012817834251263,
+      "loss": 1.0732,
+      "step": 5266
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018011978659694444,
+      "loss": 1.1827,
+      "step": 5267
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018011139327542237,
+      "loss": 1.0962,
+      "step": 5268
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001801029983781116,
+      "loss": 1.0722,
+      "step": 5269
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001800946019051772,
+      "loss": 1.0445,
+      "step": 5270
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018008620385678433,
+      "loss": 1.1924,
+      "step": 5271
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018007780423309823,
+      "loss": 1.2389,
+      "step": 5272
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018006940303428405,
+      "loss": 0.9749,
+      "step": 5273
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001800610002605071,
+      "loss": 1.1935,
+      "step": 5274
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001800525959119326,
+      "loss": 1.1495,
+      "step": 5275
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018004418998872598,
+      "loss": 1.1005,
+      "step": 5276
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018003578249105244,
+      "loss": 1.1681,
+      "step": 5277
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018002737341907743,
+      "loss": 1.0398,
+      "step": 5278
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018001896277296636,
+      "loss": 0.8977,
+      "step": 5279
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018001055055288467,
+      "loss": 1.1774,
+      "step": 5280
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00018000213675899778,
+      "loss": 1.2004,
+      "step": 5281
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017999372139147123,
+      "loss": 1.1805,
+      "step": 5282
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017998530445047058,
+      "loss": 1.276,
+      "step": 5283
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001799768859361613,
+      "loss": 1.3426,
+      "step": 5284
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017996846584870908,
+      "loss": 0.9472,
+      "step": 5285
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017996004418827948,
+      "loss": 1.122,
+      "step": 5286
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017995162095503815,
+      "loss": 0.9311,
+      "step": 5287
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001799431961491508,
+      "loss": 1.2463,
+      "step": 5288
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017993476977078317,
+      "loss": 1.0942,
+      "step": 5289
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017992634182010096,
+      "loss": 1.0864,
+      "step": 5290
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017991791229727,
+      "loss": 1.3667,
+      "step": 5291
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017990948120245602,
+      "loss": 1.1293,
+      "step": 5292
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017990104853582493,
+      "loss": 1.1451,
+      "step": 5293
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017989261429754257,
+      "loss": 1.1943,
+      "step": 5294
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001798841784877748,
+      "loss": 1.0024,
+      "step": 5295
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017987574110668765,
+      "loss": 1.0929,
+      "step": 5296
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017986730215444703,
+      "loss": 0.9367,
+      "step": 5297
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017985886163121893,
+      "loss": 1.1416,
+      "step": 5298
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017985041953716936,
+      "loss": 0.9232,
+      "step": 5299
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001798419758724644,
+      "loss": 1.172,
+      "step": 5300
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017983353063727016,
+      "loss": 0.9323,
+      "step": 5301
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001798250838317527,
+      "loss": 1.0132,
+      "step": 5302
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017981663545607823,
+      "loss": 1.0493,
+      "step": 5303
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017980818551041288,
+      "loss": 1.0183,
+      "step": 5304
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017979973399492286,
+      "loss": 1.1444,
+      "step": 5305
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017979128090977445,
+      "loss": 1.1985,
+      "step": 5306
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017978282625513389,
+      "loss": 1.0715,
+      "step": 5307
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001797743700311675,
+      "loss": 1.1851,
+      "step": 5308
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017976591223804161,
+      "loss": 1.0208,
+      "step": 5309
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017975745287592253,
+      "loss": 1.0881,
+      "step": 5310
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017974899194497676,
+      "loss": 1.1678,
+      "step": 5311
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017974052944537067,
+      "loss": 1.1987,
+      "step": 5312
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017973206537727073,
+      "loss": 1.1641,
+      "step": 5313
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017972359974084337,
+      "loss": 0.9672,
+      "step": 5314
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017971513253625521,
+      "loss": 1.238,
+      "step": 5315
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017970666376367273,
+      "loss": 1.1202,
+      "step": 5316
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017969819342326254,
+      "loss": 1.0554,
+      "step": 5317
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001796897215151912,
+      "loss": 1.0019,
+      "step": 5318
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001796812480396254,
+      "loss": 0.9893,
+      "step": 5319
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017967277299673183,
+      "loss": 1.0158,
+      "step": 5320
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017966429638667717,
+      "loss": 1.0685,
+      "step": 5321
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017965581820962814,
+      "loss": 1.2338,
+      "step": 5322
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001796473384657515,
+      "loss": 1.0155,
+      "step": 5323
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017963885715521408,
+      "loss": 1.0776,
+      "step": 5324
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017963037427818268,
+      "loss": 1.1532,
+      "step": 5325
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017962188983482419,
+      "loss": 1.0241,
+      "step": 5326
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017961340382530546,
+      "loss": 1.0852,
+      "step": 5327
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017960491624979344,
+      "loss": 1.2021,
+      "step": 5328
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017959642710845507,
+      "loss": 0.9619,
+      "step": 5329
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001795879364014573,
+      "loss": 1.0214,
+      "step": 5330
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017957944412896721,
+      "loss": 1.0077,
+      "step": 5331
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017957095029115178,
+      "loss": 1.0453,
+      "step": 5332
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017956245488817812,
+      "loss": 1.1738,
+      "step": 5333
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001795539579202133,
+      "loss": 1.1166,
+      "step": 5334
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017954545938742448,
+      "loss": 1.1205,
+      "step": 5335
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017953695928997886,
+      "loss": 1.037,
+      "step": 5336
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017952845762804357,
+      "loss": 1.1792,
+      "step": 5337
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017951995440178587,
+      "loss": 1.1443,
+      "step": 5338
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017951144961137302,
+      "loss": 1.2817,
+      "step": 5339
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001795029432569723,
+      "loss": 0.969,
+      "step": 5340
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000179494435338751,
+      "loss": 1.2194,
+      "step": 5341
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017948592585687652,
+      "loss": 0.9238,
+      "step": 5342
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001794774148115163,
+      "loss": 1.1626,
+      "step": 5343
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001794689022028376,
+      "loss": 1.0383,
+      "step": 5344
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017946038803100795,
+      "loss": 1.0221,
+      "step": 5345
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017945187229619482,
+      "loss": 0.991,
+      "step": 5346
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001794433549985657,
+      "loss": 1.0429,
+      "step": 5347
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017943483613828815,
+      "loss": 0.978,
+      "step": 5348
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017942631571552973,
+      "loss": 1.0454,
+      "step": 5349
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017941779373045798,
+      "loss": 1.0434,
+      "step": 5350
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001794092701832406,
+      "loss": 1.0465,
+      "step": 5351
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017940074507404524,
+      "loss": 1.0879,
+      "step": 5352
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017939221840303953,
+      "loss": 0.9924,
+      "step": 5353
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017938369017039124,
+      "loss": 1.2397,
+      "step": 5354
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001793751603762681,
+      "loss": 1.1212,
+      "step": 5355
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017936662902083792,
+      "loss": 1.1158,
+      "step": 5356
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001793580961042685,
+      "loss": 0.9695,
+      "step": 5357
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017934956162672763,
+      "loss": 1.1896,
+      "step": 5358
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017934102558838326,
+      "loss": 1.1826,
+      "step": 5359
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017933248798940326,
+      "loss": 1.1816,
+      "step": 5360
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017932394882995555,
+      "loss": 1.0041,
+      "step": 5361
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001793154081102081,
+      "loss": 1.2976,
+      "step": 5362
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017930686583032894,
+      "loss": 0.9298,
+      "step": 5363
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017929832199048605,
+      "loss": 1.015,
+      "step": 5364
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017928977659084755,
+      "loss": 1.0959,
+      "step": 5365
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017928122963158143,
+      "loss": 0.8752,
+      "step": 5366
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017927268111285588,
+      "loss": 1.0116,
+      "step": 5367
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017926413103483904,
+      "loss": 1.1006,
+      "step": 5368
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001792555793976991,
+      "loss": 1.0362,
+      "step": 5369
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017924702620160423,
+      "loss": 1.0655,
+      "step": 5370
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001792384714467227,
+      "loss": 1.1512,
+      "step": 5371
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017922991513322277,
+      "loss": 1.1722,
+      "step": 5372
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017922135726127275,
+      "loss": 1.0129,
+      "step": 5373
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.000179212797831041,
+      "loss": 1.0492,
+      "step": 5374
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017920423684269582,
+      "loss": 0.9018,
+      "step": 5375
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017919567429640565,
+      "loss": 0.9617,
+      "step": 5376
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017918711019233893,
+      "loss": 1.1449,
+      "step": 5377
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017917854453066405,
+      "loss": 1.2253,
+      "step": 5378
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017916997731154955,
+      "loss": 1.2173,
+      "step": 5379
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017916140853516392,
+      "loss": 1.0077,
+      "step": 5380
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017915283820167573,
+      "loss": 1.123,
+      "step": 5381
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017914426631125357,
+      "loss": 1.1783,
+      "step": 5382
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017913569286406603,
+      "loss": 1.132,
+      "step": 5383
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017912711786028174,
+      "loss": 1.2058,
+      "step": 5384
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017911854130006939,
+      "loss": 1.0832,
+      "step": 5385
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017910996318359763,
+      "loss": 1.0976,
+      "step": 5386
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017910138351103524,
+      "loss": 1.2514,
+      "step": 5387
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017909280228255102,
+      "loss": 1.0479,
+      "step": 5388
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017908421949831366,
+      "loss": 1.1606,
+      "step": 5389
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017907563515849205,
+      "loss": 1.0737,
+      "step": 5390
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017906704926325503,
+      "loss": 0.9662,
+      "step": 5391
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017905846181277151,
+      "loss": 0.9989,
+      "step": 5392
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017904987280721035,
+      "loss": 1.0687,
+      "step": 5393
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017904128224674056,
+      "loss": 1.0706,
+      "step": 5394
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017903269013153105,
+      "loss": 1.1754,
+      "step": 5395
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017902409646175087,
+      "loss": 1.1239,
+      "step": 5396
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017901550123756906,
+      "loss": 1.1947,
+      "step": 5397
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017900690445915465,
+      "loss": 1.1362,
+      "step": 5398
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017899830612667675,
+      "loss": 1.1577,
+      "step": 5399
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017898970624030452,
+      "loss": 0.9227,
+      "step": 5400
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001789811048002071,
+      "loss": 0.958,
+      "step": 5401
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001789725018065537,
+      "loss": 1.0197,
+      "step": 5402
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017896389725951352,
+      "loss": 1.225,
+      "step": 5403
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001789552911592558,
+      "loss": 0.9654,
+      "step": 5404
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017894668350594981,
+      "loss": 0.9939,
+      "step": 5405
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017893807429976492,
+      "loss": 1.0919,
+      "step": 5406
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017892946354087044,
+      "loss": 1.1965,
+      "step": 5407
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001789208512294357,
+      "loss": 1.0331,
+      "step": 5408
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017891223736563017,
+      "loss": 1.0642,
+      "step": 5409
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001789036219496233,
+      "loss": 1.3196,
+      "step": 5410
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017889500498158447,
+      "loss": 0.9745,
+      "step": 5411
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017888638646168325,
+      "loss": 1.0726,
+      "step": 5412
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017887776639008914,
+      "loss": 0.7891,
+      "step": 5413
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017886914476697164,
+      "loss": 1.0582,
+      "step": 5414
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017886052159250045,
+      "loss": 1.0892,
+      "step": 5415
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017885189686684509,
+      "loss": 1.1771,
+      "step": 5416
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017884327059017528,
+      "loss": 1.0666,
+      "step": 5417
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017883464276266064,
+      "loss": 1.1728,
+      "step": 5418
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017882601338447093,
+      "loss": 0.981,
+      "step": 5419
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017881738245577582,
+      "loss": 1.0451,
+      "step": 5420
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017880874997674515,
+      "loss": 1.1216,
+      "step": 5421
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017880011594754872,
+      "loss": 1.026,
+      "step": 5422
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017879148036835632,
+      "loss": 1.0442,
+      "step": 5423
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001787828432393378,
+      "loss": 1.0171,
+      "step": 5424
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.0001787742045606631,
+      "loss": 1.0127,
+      "step": 5425
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00017876556433250212,
+      "loss": 0.9245,
+      "step": 5426
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017875692255502483,
+      "loss": 1.2063,
+      "step": 5427
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017874827922840117,
+      "loss": 1.1882,
+      "step": 5428
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017873963435280121,
+      "loss": 1.0823,
+      "step": 5429
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017873098792839493,
+      "loss": 1.2164,
+      "step": 5430
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017872233995535247,
+      "loss": 0.9793,
+      "step": 5431
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001787136904338439,
+      "loss": 1.1764,
+      "step": 5432
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001787050393640394,
+      "loss": 1.1638,
+      "step": 5433
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017869638674610907,
+      "loss": 1.1612,
+      "step": 5434
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017868773258022312,
+      "loss": 1.1549,
+      "step": 5435
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001786790768665518,
+      "loss": 1.2383,
+      "step": 5436
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017867041960526538,
+      "loss": 1.1508,
+      "step": 5437
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001786617607965341,
+      "loss": 1.1898,
+      "step": 5438
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017865310044052834,
+      "loss": 1.0283,
+      "step": 5439
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017864443853741842,
+      "loss": 1.0801,
+      "step": 5440
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001786357750873747,
+      "loss": 1.015,
+      "step": 5441
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001786271100905676,
+      "loss": 1.1033,
+      "step": 5442
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017861844354716757,
+      "loss": 1.192,
+      "step": 5443
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017860977545734508,
+      "loss": 1.1174,
+      "step": 5444
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001786011058212706,
+      "loss": 1.0209,
+      "step": 5445
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017859243463911475,
+      "loss": 1.0048,
+      "step": 5446
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000178583761911048,
+      "loss": 0.8542,
+      "step": 5447
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017857508763724096,
+      "loss": 1.0373,
+      "step": 5448
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017856641181786427,
+      "loss": 0.8747,
+      "step": 5449
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017855773445308858,
+      "loss": 1.0783,
+      "step": 5450
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017854905554308458,
+      "loss": 1.0861,
+      "step": 5451
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017854037508802296,
+      "loss": 0.9831,
+      "step": 5452
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017853169308807448,
+      "loss": 0.7887,
+      "step": 5453
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017852300954340994,
+      "loss": 1.1333,
+      "step": 5454
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017851432445420014,
+      "loss": 1.1139,
+      "step": 5455
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017850563782061585,
+      "loss": 1.1113,
+      "step": 5456
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017849694964282802,
+      "loss": 0.9841,
+      "step": 5457
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017848825992100746,
+      "loss": 0.9657,
+      "step": 5458
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001784795686553252,
+      "loss": 1.1101,
+      "step": 5459
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001784708758459521,
+      "loss": 1.1013,
+      "step": 5460
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001784621814930592,
+      "loss": 0.8582,
+      "step": 5461
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017845348559681752,
+      "loss": 1.0847,
+      "step": 5462
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017844478815739806,
+      "loss": 1.0512,
+      "step": 5463
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017843608917497198,
+      "loss": 1.1725,
+      "step": 5464
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001784273886497103,
+      "loss": 0.923,
+      "step": 5465
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017841868658178425,
+      "loss": 1.1114,
+      "step": 5466
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001784099829713649,
+      "loss": 1.2169,
+      "step": 5467
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017840127781862354,
+      "loss": 0.9902,
+      "step": 5468
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017839257112373135,
+      "loss": 0.9703,
+      "step": 5469
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017838386288685957,
+      "loss": 0.9546,
+      "step": 5470
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017837515310817955,
+      "loss": 1.1929,
+      "step": 5471
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017836644178786257,
+      "loss": 0.9677,
+      "step": 5472
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017835772892608,
+      "loss": 0.9785,
+      "step": 5473
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001783490145230032,
+      "loss": 1.1107,
+      "step": 5474
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017834029857880364,
+      "loss": 0.9171,
+      "step": 5475
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017833158109365269,
+      "loss": 1.1758,
+      "step": 5476
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017832286206772186,
+      "loss": 1.1245,
+      "step": 5477
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017831414150118265,
+      "loss": 1.0908,
+      "step": 5478
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001783054193942066,
+      "loss": 1.0532,
+      "step": 5479
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001782966957469652,
+      "loss": 1.3324,
+      "step": 5480
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017828797055963018,
+      "loss": 1.0278,
+      "step": 5481
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017827924383237304,
+      "loss": 0.759,
+      "step": 5482
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001782705155653655,
+      "loss": 1.0202,
+      "step": 5483
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017826178575877926,
+      "loss": 1.0403,
+      "step": 5484
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017825305441278598,
+      "loss": 0.9239,
+      "step": 5485
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001782443215275574,
+      "loss": 0.9861,
+      "step": 5486
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017823558710326537,
+      "loss": 0.9413,
+      "step": 5487
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017822685114008167,
+      "loss": 1.0657,
+      "step": 5488
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017821811363817806,
+      "loss": 1.1053,
+      "step": 5489
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001782093745977265,
+      "loss": 0.973,
+      "step": 5490
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017820063401889885,
+      "loss": 1.1191,
+      "step": 5491
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017819189190186702,
+      "loss": 1.1515,
+      "step": 5492
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000178183148246803,
+      "loss": 1.1109,
+      "step": 5493
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017817440305387873,
+      "loss": 0.9683,
+      "step": 5494
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017816565632326628,
+      "loss": 1.0855,
+      "step": 5495
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017815690805513768,
+      "loss": 0.938,
+      "step": 5496
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017814815824966504,
+      "loss": 1.0146,
+      "step": 5497
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001781394069070204,
+      "loss": 0.9292,
+      "step": 5498
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017813065402737593,
+      "loss": 0.9413,
+      "step": 5499
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001781218996109038,
+      "loss": 0.9139,
+      "step": 5500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017811314365777625,
+      "loss": 1.1135,
+      "step": 5501
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017810438616816545,
+      "loss": 1.076,
+      "step": 5502
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017809562714224368,
+      "loss": 1.3564,
+      "step": 5503
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001780868665801832,
+      "loss": 0.9669,
+      "step": 5504
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001780781044821564,
+      "loss": 0.8619,
+      "step": 5505
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017806934084833558,
+      "loss": 1.0067,
+      "step": 5506
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001780605756788931,
+      "loss": 0.9179,
+      "step": 5507
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017805180897400146,
+      "loss": 0.9309,
+      "step": 5508
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000178043040733833,
+      "loss": 1.1367,
+      "step": 5509
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017803427095856023,
+      "loss": 0.8489,
+      "step": 5510
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017802549964835565,
+      "loss": 1.1187,
+      "step": 5511
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017801672680339183,
+      "loss": 1.0821,
+      "step": 5512
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017800795242384128,
+      "loss": 1.2156,
+      "step": 5513
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017799917650987662,
+      "loss": 1.1986,
+      "step": 5514
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017799039906167042,
+      "loss": 0.9341,
+      "step": 5515
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001779816200793954,
+      "loss": 1.1932,
+      "step": 5516
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001779728395632242,
+      "loss": 1.085,
+      "step": 5517
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001779640575133296,
+      "loss": 0.7143,
+      "step": 5518
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017795527392988424,
+      "loss": 1.1876,
+      "step": 5519
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017794648881306094,
+      "loss": 1.1886,
+      "step": 5520
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017793770216303253,
+      "loss": 0.9728,
+      "step": 5521
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017792891397997183,
+      "loss": 1.0702,
+      "step": 5522
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017792012426405166,
+      "loss": 1.2672,
+      "step": 5523
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017791133301544497,
+      "loss": 0.971,
+      "step": 5524
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017790254023432466,
+      "loss": 1.1554,
+      "step": 5525
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001778937459208637,
+      "loss": 1.0933,
+      "step": 5526
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017788495007523502,
+      "loss": 0.9694,
+      "step": 5527
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017787615269761168,
+      "loss": 1.1527,
+      "step": 5528
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017786735378816671,
+      "loss": 1.1108,
+      "step": 5529
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017785855334707322,
+      "loss": 1.1516,
+      "step": 5530
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017784975137450428,
+      "loss": 1.2508,
+      "step": 5531
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017784094787063302,
+      "loss": 1.0016,
+      "step": 5532
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017783214283563262,
+      "loss": 1.0509,
+      "step": 5533
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017782333626967624,
+      "loss": 0.8749,
+      "step": 5534
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017781452817293715,
+      "loss": 1.2291,
+      "step": 5535
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017780571854558857,
+      "loss": 1.0468,
+      "step": 5536
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017779690738780387,
+      "loss": 1.134,
+      "step": 5537
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017778809469975624,
+      "loss": 1.0431,
+      "step": 5538
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017777928048161908,
+      "loss": 1.0228,
+      "step": 5539
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001777704647335658,
+      "loss": 1.0029,
+      "step": 5540
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017776164745576976,
+      "loss": 1.1468,
+      "step": 5541
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001777528286484044,
+      "loss": 1.0826,
+      "step": 5542
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017774400831164323,
+      "loss": 0.8608,
+      "step": 5543
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017773518644565968,
+      "loss": 0.994,
+      "step": 5544
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001777263630506273,
+      "loss": 0.9984,
+      "step": 5545
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001777175381267197,
+      "loss": 1.067,
+      "step": 5546
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017770871167411038,
+      "loss": 1.1142,
+      "step": 5547
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000177699883692973,
+      "loss": 1.1861,
+      "step": 5548
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017769105418348122,
+      "loss": 1.2503,
+      "step": 5549
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001776822231458087,
+      "loss": 1.0341,
+      "step": 5550
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017767339058012913,
+      "loss": 1.1805,
+      "step": 5551
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017766455648661627,
+      "loss": 1.1061,
+      "step": 5552
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017765572086544387,
+      "loss": 0.9837,
+      "step": 5553
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017764688371678573,
+      "loss": 1.3725,
+      "step": 5554
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017763804504081568,
+      "loss": 1.0272,
+      "step": 5555
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017762920483770756,
+      "loss": 1.0494,
+      "step": 5556
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017762036310763532,
+      "loss": 0.9651,
+      "step": 5557
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001776115198507728,
+      "loss": 1.1365,
+      "step": 5558
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000177602675067294,
+      "loss": 1.0588,
+      "step": 5559
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017759382875737282,
+      "loss": 1.1531,
+      "step": 5560
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017758498092118336,
+      "loss": 1.1082,
+      "step": 5561
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001775761315588996,
+      "loss": 1.113,
+      "step": 5562
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017756728067069562,
+      "loss": 1.1026,
+      "step": 5563
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017755842825674555,
+      "loss": 1.1899,
+      "step": 5564
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017754957431722346,
+      "loss": 1.2531,
+      "step": 5565
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017754071885230354,
+      "loss": 0.8415,
+      "step": 5566
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017753186186216,
+      "loss": 0.8035,
+      "step": 5567
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017752300334696695,
+      "loss": 1.0629,
+      "step": 5568
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017751414330689877,
+      "loss": 1.0918,
+      "step": 5569
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017750528174212966,
+      "loss": 1.07,
+      "step": 5570
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017749641865283394,
+      "loss": 1.2254,
+      "step": 5571
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000177487554039186,
+      "loss": 1.0314,
+      "step": 5572
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001774786879013601,
+      "loss": 1.0944,
+      "step": 5573
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017746982023953073,
+      "loss": 1.0742,
+      "step": 5574
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001774609510538723,
+      "loss": 1.2651,
+      "step": 5575
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001774520803445592,
+      "loss": 1.1412,
+      "step": 5576
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000177443208111766,
+      "loss": 1.0074,
+      "step": 5577
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017743433435566717,
+      "loss": 1.0576,
+      "step": 5578
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017742545907643728,
+      "loss": 1.0745,
+      "step": 5579
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017741658227425088,
+      "loss": 1.1326,
+      "step": 5580
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001774077039492826,
+      "loss": 1.0353,
+      "step": 5581
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001773988241017071,
+      "loss": 0.9767,
+      "step": 5582
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017738994273169898,
+      "loss": 1.1418,
+      "step": 5583
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017738105983943297,
+      "loss": 0.9964,
+      "step": 5584
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017737217542508382,
+      "loss": 1.0124,
+      "step": 5585
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017736328948882624,
+      "loss": 0.956,
+      "step": 5586
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017735440203083506,
+      "loss": 1.318,
+      "step": 5587
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017734551305128505,
+      "loss": 1.0043,
+      "step": 5588
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017733662255035114,
+      "loss": 0.9908,
+      "step": 5589
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017732773052820807,
+      "loss": 1.1456,
+      "step": 5590
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017731883698503088,
+      "loss": 1.1674,
+      "step": 5591
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017730994192099444,
+      "loss": 1.2841,
+      "step": 5592
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001773010453362737,
+      "loss": 1.1982,
+      "step": 5593
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001772921472310437,
+      "loss": 1.1453,
+      "step": 5594
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017728324760547944,
+      "loss": 1.0025,
+      "step": 5595
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017727434645975598,
+      "loss": 0.8678,
+      "step": 5596
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001772654437940484,
+      "loss": 0.9787,
+      "step": 5597
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017725653960853186,
+      "loss": 1.2992,
+      "step": 5598
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017724763390338142,
+      "loss": 1.1907,
+      "step": 5599
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017723872667877233,
+      "loss": 1.1398,
+      "step": 5600
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017722981793487976,
+      "loss": 1.1131,
+      "step": 5601
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017722090767187894,
+      "loss": 1.0779,
+      "step": 5602
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017721199588994517,
+      "loss": 0.9855,
+      "step": 5603
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001772030825892537,
+      "loss": 1.2,
+      "step": 5604
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001771941677699799,
+      "loss": 1.127,
+      "step": 5605
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017718525143229906,
+      "loss": 1.1377,
+      "step": 5606
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017717633357638665,
+      "loss": 0.9904,
+      "step": 5607
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.000177167414202418,
+      "loss": 0.9546,
+      "step": 5608
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001771584933105686,
+      "loss": 1.2316,
+      "step": 5609
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017714957090101392,
+      "loss": 1.1563,
+      "step": 5610
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017714064697392946,
+      "loss": 1.2066,
+      "step": 5611
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017713172152949075,
+      "loss": 0.9934,
+      "step": 5612
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017712279456787332,
+      "loss": 1.109,
+      "step": 5613
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017711386608925284,
+      "loss": 1.2446,
+      "step": 5614
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001771049360938049,
+      "loss": 1.2332,
+      "step": 5615
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017709600458170512,
+      "loss": 0.9036,
+      "step": 5616
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001770870715531292,
+      "loss": 0.876,
+      "step": 5617
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001770781370082529,
+      "loss": 1.0271,
+      "step": 5618
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001770692009472519,
+      "loss": 1.015,
+      "step": 5619
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017706026337030198,
+      "loss": 1.1445,
+      "step": 5620
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017705132427757895,
+      "loss": 1.009,
+      "step": 5621
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017704238366925864,
+      "loss": 0.8694,
+      "step": 5622
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017703344154551694,
+      "loss": 1.1672,
+      "step": 5623
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017702449790652972,
+      "loss": 1.1152,
+      "step": 5624
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017701555275247287,
+      "loss": 0.9432,
+      "step": 5625
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001770066060835224,
+      "loss": 0.9983,
+      "step": 5626
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017699765789985425,
+      "loss": 1.1347,
+      "step": 5627
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017698870820164446,
+      "loss": 1.0039,
+      "step": 5628
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017697975698906904,
+      "loss": 1.0753,
+      "step": 5629
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001769708042623041,
+      "loss": 0.83,
+      "step": 5630
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017696185002152566,
+      "loss": 0.9876,
+      "step": 5631
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001769528942669099,
+      "loss": 1.0933,
+      "step": 5632
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017694393699863302,
+      "loss": 1.1306,
+      "step": 5633
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017693497821687118,
+      "loss": 1.1993,
+      "step": 5634
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001769260179218006,
+      "loss": 1.1354,
+      "step": 5635
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017691705611359745,
+      "loss": 1.0217,
+      "step": 5636
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017690809279243813,
+      "loss": 1.1281,
+      "step": 5637
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001768991279584989,
+      "loss": 1.2071,
+      "step": 5638
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017689016161195607,
+      "loss": 1.0857,
+      "step": 5639
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017688119375298603,
+      "loss": 1.1992,
+      "step": 5640
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017687222438176519,
+      "loss": 0.74,
+      "step": 5641
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017686325349846997,
+      "loss": 0.9402,
+      "step": 5642
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017685428110327683,
+      "loss": 1.0199,
+      "step": 5643
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017684530719636223,
+      "loss": 1.1377,
+      "step": 5644
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017683633177790272,
+      "loss": 1.0833,
+      "step": 5645
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001768273548480748,
+      "loss": 1.2055,
+      "step": 5646
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017681837640705513,
+      "loss": 1.1241,
+      "step": 5647
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017680939645502023,
+      "loss": 1.2606,
+      "step": 5648
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017680041499214678,
+      "loss": 1.2777,
+      "step": 5649
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017679143201861143,
+      "loss": 1.0201,
+      "step": 5650
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001767824475345909,
+      "loss": 1.0788,
+      "step": 5651
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017677346154026187,
+      "loss": 1.0332,
+      "step": 5652
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001767644740358011,
+      "loss": 0.9869,
+      "step": 5653
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017675548502138542,
+      "loss": 1.1848,
+      "step": 5654
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017674649449719163,
+      "loss": 1.1007,
+      "step": 5655
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017673750246339655,
+      "loss": 1.1257,
+      "step": 5656
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00017672850892017704,
+      "loss": 1.104,
+      "step": 5657
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017671951386771006,
+      "loss": 0.8482,
+      "step": 5658
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001767105173061725,
+      "loss": 1.3181,
+      "step": 5659
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001767015192357413,
+      "loss": 1.2146,
+      "step": 5660
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001766925196565935,
+      "loss": 1.1905,
+      "step": 5661
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001766835185689061,
+      "loss": 1.0281,
+      "step": 5662
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017667451597285617,
+      "loss": 1.0079,
+      "step": 5663
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017666551186862074,
+      "loss": 1.1335,
+      "step": 5664
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017665650625637699,
+      "loss": 1.0131,
+      "step": 5665
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000176647499136302,
+      "loss": 0.8844,
+      "step": 5666
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000176638490508573,
+      "loss": 1.0893,
+      "step": 5667
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001766294803733671,
+      "loss": 1.1798,
+      "step": 5668
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017662046873086163,
+      "loss": 0.974,
+      "step": 5669
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001766114555812338,
+      "loss": 1.1539,
+      "step": 5670
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017660244092466087,
+      "loss": 1.0663,
+      "step": 5671
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017659342476132023,
+      "loss": 0.9601,
+      "step": 5672
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017658440709138914,
+      "loss": 1.0028,
+      "step": 5673
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017657538791504505,
+      "loss": 0.9234,
+      "step": 5674
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017656636723246535,
+      "loss": 1.1316,
+      "step": 5675
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017655734504382745,
+      "loss": 1.1561,
+      "step": 5676
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017654832134930882,
+      "loss": 1.1692,
+      "step": 5677
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000176539296149087,
+      "loss": 1.1021,
+      "step": 5678
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017653026944333946,
+      "loss": 0.9435,
+      "step": 5679
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001765212412322438,
+      "loss": 1.1507,
+      "step": 5680
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001765122115159776,
+      "loss": 1.1748,
+      "step": 5681
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001765031802947184,
+      "loss": 1.1235,
+      "step": 5682
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017649414756864396,
+      "loss": 0.9946,
+      "step": 5683
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017648511333793186,
+      "loss": 1.0166,
+      "step": 5684
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017647607760275987,
+      "loss": 1.0817,
+      "step": 5685
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017646704036330566,
+      "loss": 1.1136,
+      "step": 5686
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000176458001619747,
+      "loss": 0.9625,
+      "step": 5687
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017644896137226178,
+      "loss": 1.0469,
+      "step": 5688
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001764399196210277,
+      "loss": 0.9952,
+      "step": 5689
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017643087636622266,
+      "loss": 1.1972,
+      "step": 5690
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017642183160802456,
+      "loss": 1.1232,
+      "step": 5691
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001764127853466113,
+      "loss": 1.0663,
+      "step": 5692
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017640373758216077,
+      "loss": 1.0278,
+      "step": 5693
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017639468831485096,
+      "loss": 1.0645,
+      "step": 5694
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017638563754485996,
+      "loss": 1.2095,
+      "step": 5695
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017637658527236565,
+      "loss": 1.0504,
+      "step": 5696
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001763675314975462,
+      "loss": 1.1308,
+      "step": 5697
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017635847622057965,
+      "loss": 0.8186,
+      "step": 5698
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017634941944164414,
+      "loss": 1.0661,
+      "step": 5699
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017634036116091778,
+      "loss": 1.0177,
+      "step": 5700
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001763313013785788,
+      "loss": 0.9354,
+      "step": 5701
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001763222400948054,
+      "loss": 1.2379,
+      "step": 5702
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001763131773097757,
+      "loss": 1.0762,
+      "step": 5703
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017630411302366813,
+      "loss": 0.9057,
+      "step": 5704
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017629504723666088,
+      "loss": 1.1689,
+      "step": 5705
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017628597994893231,
+      "loss": 1.0044,
+      "step": 5706
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001762769111606608,
+      "loss": 0.9664,
+      "step": 5707
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017626784087202462,
+      "loss": 1.1407,
+      "step": 5708
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017625876908320236,
+      "loss": 1.0931,
+      "step": 5709
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001762496957943723,
+      "loss": 1.0636,
+      "step": 5710
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017624062100571298,
+      "loss": 0.978,
+      "step": 5711
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001762315447174029,
+      "loss": 1.1361,
+      "step": 5712
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001762224669296206,
+      "loss": 1.0969,
+      "step": 5713
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017621338764254463,
+      "loss": 1.07,
+      "step": 5714
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017620430685635357,
+      "loss": 1.1412,
+      "step": 5715
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017619522457122604,
+      "loss": 1.0391,
+      "step": 5716
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017618614078734069,
+      "loss": 1.1009,
+      "step": 5717
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017617705550487618,
+      "loss": 1.3592,
+      "step": 5718
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017616796872401124,
+      "loss": 1.1998,
+      "step": 5719
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017615888044492464,
+      "loss": 1.0255,
+      "step": 5720
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017614979066779508,
+      "loss": 0.9602,
+      "step": 5721
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017614069939280144,
+      "loss": 1.0604,
+      "step": 5722
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001761316066201224,
+      "loss": 1.0613,
+      "step": 5723
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017612251234993696,
+      "loss": 1.2629,
+      "step": 5724
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017611341658242395,
+      "loss": 1.1462,
+      "step": 5725
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001761043193177623,
+      "loss": 0.9908,
+      "step": 5726
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001760952205561309,
+      "loss": 1.0569,
+      "step": 5727
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017608612029770877,
+      "loss": 0.9876,
+      "step": 5728
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001760770185426749,
+      "loss": 1.12,
+      "step": 5729
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017606791529120832,
+      "loss": 1.1727,
+      "step": 5730
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001760588105434881,
+      "loss": 1.0253,
+      "step": 5731
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001760497042996933,
+      "loss": 1.0092,
+      "step": 5732
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001760405965600031,
+      "loss": 1.119,
+      "step": 5733
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001760314873245966,
+      "loss": 1.3608,
+      "step": 5734
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017602237659365298,
+      "loss": 1.1898,
+      "step": 5735
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017601326436735145,
+      "loss": 0.9915,
+      "step": 5736
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017600415064587127,
+      "loss": 0.9765,
+      "step": 5737
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017599503542939168,
+      "loss": 1.2369,
+      "step": 5738
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017598591871809198,
+      "loss": 1.1003,
+      "step": 5739
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017597680051215152,
+      "loss": 1.0388,
+      "step": 5740
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017596768081174963,
+      "loss": 1.0702,
+      "step": 5741
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001759585596170657,
+      "loss": 0.9745,
+      "step": 5742
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017594943692827915,
+      "loss": 1.1299,
+      "step": 5743
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001759403127455694,
+      "loss": 1.2237,
+      "step": 5744
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017593118706911594,
+      "loss": 1.1642,
+      "step": 5745
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001759220598990983,
+      "loss": 0.9262,
+      "step": 5746
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000175912931235696,
+      "loss": 1.2012,
+      "step": 5747
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017590380107908855,
+      "loss": 1.0302,
+      "step": 5748
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017589466942945556,
+      "loss": 1.1615,
+      "step": 5749
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001758855362869767,
+      "loss": 0.9813,
+      "step": 5750
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017587640165183155,
+      "loss": 0.8438,
+      "step": 5751
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017586726552419984,
+      "loss": 0.906,
+      "step": 5752
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017585812790426125,
+      "loss": 1.1338,
+      "step": 5753
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017584898879219552,
+      "loss": 1.1545,
+      "step": 5754
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017583984818818243,
+      "loss": 0.8327,
+      "step": 5755
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017583070609240175,
+      "loss": 0.8797,
+      "step": 5756
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017582156250503333,
+      "loss": 1.0798,
+      "step": 5757
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017581241742625702,
+      "loss": 1.1302,
+      "step": 5758
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001758032708562527,
+      "loss": 1.0439,
+      "step": 5759
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017579412279520025,
+      "loss": 1.0985,
+      "step": 5760
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001757849732432797,
+      "loss": 1.0609,
+      "step": 5761
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001757758222006709,
+      "loss": 0.9477,
+      "step": 5762
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017576666966755396,
+      "loss": 1.1028,
+      "step": 5763
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017575751564410884,
+      "loss": 1.1124,
+      "step": 5764
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017574836013051564,
+      "loss": 1.0863,
+      "step": 5765
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017573920312695442,
+      "loss": 1.069,
+      "step": 5766
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001757300446336053,
+      "loss": 1.1225,
+      "step": 5767
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017572088465064848,
+      "loss": 1.0053,
+      "step": 5768
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017571172317826404,
+      "loss": 0.9221,
+      "step": 5769
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017570256021663227,
+      "loss": 1.0121,
+      "step": 5770
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017569339576593339,
+      "loss": 1.105,
+      "step": 5771
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017568422982634766,
+      "loss": 1.1173,
+      "step": 5772
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001756750623980553,
+      "loss": 0.8422,
+      "step": 5773
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017566589348123673,
+      "loss": 1.0542,
+      "step": 5774
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001756567230760723,
+      "loss": 1.1987,
+      "step": 5775
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017564755118274232,
+      "loss": 1.144,
+      "step": 5776
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017563837780142726,
+      "loss": 1.0853,
+      "step": 5777
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017562920293230755,
+      "loss": 1.0521,
+      "step": 5778
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017562002657556364,
+      "loss": 1.1312,
+      "step": 5779
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017561084873137604,
+      "loss": 1.0725,
+      "step": 5780
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017560166939992528,
+      "loss": 1.1596,
+      "step": 5781
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017559248858139192,
+      "loss": 0.9812,
+      "step": 5782
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017558330627595654,
+      "loss": 1.034,
+      "step": 5783
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017557412248379975,
+      "loss": 0.8823,
+      "step": 5784
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001755649372051022,
+      "loss": 1.0446,
+      "step": 5785
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001755557504400446,
+      "loss": 1.0043,
+      "step": 5786
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017554656218880758,
+      "loss": 0.9819,
+      "step": 5787
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001755373724515719,
+      "loss": 0.9896,
+      "step": 5788
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017552818122851838,
+      "loss": 0.9911,
+      "step": 5789
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017551898851982773,
+      "loss": 1.1642,
+      "step": 5790
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017550979432568082,
+      "loss": 1.1319,
+      "step": 5791
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017550059864625847,
+      "loss": 1.0771,
+      "step": 5792
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001754914014817416,
+      "loss": 1.1749,
+      "step": 5793
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017548220283231108,
+      "loss": 0.9148,
+      "step": 5794
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017547300269814788,
+      "loss": 1.1947,
+      "step": 5795
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001754638010794329,
+      "loss": 0.9264,
+      "step": 5796
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001754545979763472,
+      "loss": 1.1521,
+      "step": 5797
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001754453933890718,
+      "loss": 1.1644,
+      "step": 5798
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017543618731778775,
+      "loss": 1.0853,
+      "step": 5799
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001754269797626761,
+      "loss": 1.0952,
+      "step": 5800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000175417770723918,
+      "loss": 1.1469,
+      "step": 5801
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017540856020169455,
+      "loss": 1.1617,
+      "step": 5802
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017539934819618696,
+      "loss": 1.0984,
+      "step": 5803
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001753901347075764,
+      "loss": 1.1108,
+      "step": 5804
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017538091973604415,
+      "loss": 1.1289,
+      "step": 5805
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017537170328177143,
+      "loss": 1.0915,
+      "step": 5806
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001753624853449395,
+      "loss": 1.258,
+      "step": 5807
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017535326592572976,
+      "loss": 0.9842,
+      "step": 5808
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017534404502432346,
+      "loss": 0.9784,
+      "step": 5809
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017533482264090204,
+      "loss": 1.3274,
+      "step": 5810
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017532559877564687,
+      "loss": 1.0141,
+      "step": 5811
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017531637342873941,
+      "loss": 1.1594,
+      "step": 5812
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017530714660036112,
+      "loss": 0.7819,
+      "step": 5813
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017529791829069344,
+      "loss": 0.8662,
+      "step": 5814
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017528868849991795,
+      "loss": 1.1009,
+      "step": 5815
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001752794572282162,
+      "loss": 1.0925,
+      "step": 5816
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001752702244757697,
+      "loss": 1.0624,
+      "step": 5817
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017526099024276016,
+      "loss": 0.9864,
+      "step": 5818
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017525175452936913,
+      "loss": 1.0252,
+      "step": 5819
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017524251733577834,
+      "loss": 0.959,
+      "step": 5820
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017523327866216944,
+      "loss": 1.0898,
+      "step": 5821
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017522403850872418,
+      "loss": 0.9923,
+      "step": 5822
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017521479687562428,
+      "loss": 1.1646,
+      "step": 5823
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017520555376305155,
+      "loss": 1.3684,
+      "step": 5824
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001751963091711878,
+      "loss": 0.9429,
+      "step": 5825
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017518706310021487,
+      "loss": 1.0294,
+      "step": 5826
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017517781555031464,
+      "loss": 1.0302,
+      "step": 5827
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000175168566521669,
+      "loss": 1.0719,
+      "step": 5828
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017515931601445986,
+      "loss": 0.9292,
+      "step": 5829
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001751500640288692,
+      "loss": 1.1403,
+      "step": 5830
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.000175140810565079,
+      "loss": 0.991,
+      "step": 5831
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017513155562327127,
+      "loss": 1.2742,
+      "step": 5832
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017512229920362807,
+      "loss": 1.0637,
+      "step": 5833
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017511304130633147,
+      "loss": 1.1072,
+      "step": 5834
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017510378193156357,
+      "loss": 1.1795,
+      "step": 5835
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001750945210795065,
+      "loss": 1.1281,
+      "step": 5836
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017508525875034237,
+      "loss": 1.2342,
+      "step": 5837
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017507599494425344,
+      "loss": 1.4036,
+      "step": 5838
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001750667296614219,
+      "loss": 1.2446,
+      "step": 5839
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017505746290203002,
+      "loss": 1.1147,
+      "step": 5840
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017504819466626005,
+      "loss": 1.2318,
+      "step": 5841
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001750389249542943,
+      "loss": 1.0564,
+      "step": 5842
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017502965376631514,
+      "loss": 1.0571,
+      "step": 5843
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001750203811025049,
+      "loss": 1.1128,
+      "step": 5844
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017501110696304596,
+      "loss": 1.1227,
+      "step": 5845
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017500183134812078,
+      "loss": 1.0777,
+      "step": 5846
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001749925542579118,
+      "loss": 0.9936,
+      "step": 5847
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017498327569260145,
+      "loss": 1.0946,
+      "step": 5848
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001749739956523723,
+      "loss": 1.1009,
+      "step": 5849
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001749647141374069,
+      "loss": 0.8947,
+      "step": 5850
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017495543114788773,
+      "loss": 1.2604,
+      "step": 5851
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017494614668399748,
+      "loss": 0.9012,
+      "step": 5852
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017493686074591873,
+      "loss": 0.9961,
+      "step": 5853
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017492757333383414,
+      "loss": 0.9642,
+      "step": 5854
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001749182844479264,
+      "loss": 1.0319,
+      "step": 5855
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017490899408837822,
+      "loss": 1.031,
+      "step": 5856
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017489970225537232,
+      "loss": 1.032,
+      "step": 5857
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001748904089490915,
+      "loss": 0.9944,
+      "step": 5858
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017488111416971854,
+      "loss": 1.1019,
+      "step": 5859
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001748718179174363,
+      "loss": 1.263,
+      "step": 5860
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001748625201924276,
+      "loss": 0.9921,
+      "step": 5861
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001748532209948753,
+      "loss": 1.1074,
+      "step": 5862
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017484392032496242,
+      "loss": 0.9821,
+      "step": 5863
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017483461818287182,
+      "loss": 1.2125,
+      "step": 5864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017482531456878645,
+      "loss": 0.9956,
+      "step": 5865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017481600948288936,
+      "loss": 0.8681,
+      "step": 5866
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001748067029253636,
+      "loss": 1.0267,
+      "step": 5867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017479739489639217,
+      "loss": 0.8845,
+      "step": 5868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017478808539615824,
+      "loss": 1.2938,
+      "step": 5869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017477877442484486,
+      "loss": 1.2179,
+      "step": 5870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017476946198263516,
+      "loss": 1.2042,
+      "step": 5871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001747601480697124,
+      "loss": 1.1783,
+      "step": 5872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001747508326862597,
+      "loss": 1.1088,
+      "step": 5873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017474151583246034,
+      "loss": 1.0247,
+      "step": 5874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017473219750849755,
+      "loss": 1.1489,
+      "step": 5875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017472287771455467,
+      "loss": 1.0035,
+      "step": 5876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017471355645081498,
+      "loss": 0.8516,
+      "step": 5877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017470423371746183,
+      "loss": 0.882,
+      "step": 5878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017469490951467864,
+      "loss": 1.1394,
+      "step": 5879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017468558384264876,
+      "loss": 1.097,
+      "step": 5880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017467625670155565,
+      "loss": 1.0198,
+      "step": 5881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017466692809158278,
+      "loss": 1.0911,
+      "step": 5882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017465759801291365,
+      "loss": 0.9918,
+      "step": 5883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017464826646573176,
+      "loss": 0.8384,
+      "step": 5884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017463893345022067,
+      "loss": 1.2693,
+      "step": 5885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017462959896656397,
+      "loss": 0.9895,
+      "step": 5886
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00017462026301494525,
+      "loss": 0.8482,
+      "step": 5887
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.0001746109255955482,
+      "loss": 0.9656,
+      "step": 5888
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001746015867085564,
+      "loss": 1.0438,
+      "step": 5889
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001745922463541536,
+      "loss": 0.9843,
+      "step": 5890
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017458290453252354,
+      "loss": 1.1084,
+      "step": 5891
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017457356124384995,
+      "loss": 1.104,
+      "step": 5892
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017456421648831655,
+      "loss": 0.8153,
+      "step": 5893
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017455487026610728,
+      "loss": 0.9479,
+      "step": 5894
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017454552257740587,
+      "loss": 1.0482,
+      "step": 5895
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017453617342239627,
+      "loss": 1.028,
+      "step": 5896
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017452682280126233,
+      "loss": 1.1528,
+      "step": 5897
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000174517470714188,
+      "loss": 1.2111,
+      "step": 5898
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017450811716135717,
+      "loss": 1.04,
+      "step": 5899
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017449876214295393,
+      "loss": 1.2458,
+      "step": 5900
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017448940565916222,
+      "loss": 1.0567,
+      "step": 5901
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001744800477101661,
+      "loss": 1.2226,
+      "step": 5902
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017447068829614962,
+      "loss": 0.9197,
+      "step": 5903
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017446132741729694,
+      "loss": 0.8229,
+      "step": 5904
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017445196507379212,
+      "loss": 1.0194,
+      "step": 5905
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017444260126581935,
+      "loss": 0.9525,
+      "step": 5906
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017443323599356281,
+      "loss": 1.0215,
+      "step": 5907
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001744238692572067,
+      "loss": 0.9981,
+      "step": 5908
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001744145010569353,
+      "loss": 1.1496,
+      "step": 5909
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017440513139293286,
+      "loss": 0.9982,
+      "step": 5910
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001743957602653837,
+      "loss": 0.9532,
+      "step": 5911
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001743863876744721,
+      "loss": 1.0079,
+      "step": 5912
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001743770136203825,
+      "loss": 1.0278,
+      "step": 5913
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017436763810329916,
+      "loss": 1.076,
+      "step": 5914
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017435826112340662,
+      "loss": 1.1458,
+      "step": 5915
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017434888268088927,
+      "loss": 0.9977,
+      "step": 5916
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017433950277593162,
+      "loss": 1.045,
+      "step": 5917
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017433012140871811,
+      "loss": 1.1093,
+      "step": 5918
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017432073857943331,
+      "loss": 0.9965,
+      "step": 5919
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001743113542882618,
+      "loss": 1.04,
+      "step": 5920
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017430196853538812,
+      "loss": 1.1596,
+      "step": 5921
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017429258132099692,
+      "loss": 1.1245,
+      "step": 5922
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017428319264527287,
+      "loss": 0.8767,
+      "step": 5923
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017427380250840057,
+      "loss": 1.0605,
+      "step": 5924
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001742644109105648,
+      "loss": 1.0995,
+      "step": 5925
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017425501785195023,
+      "loss": 0.9521,
+      "step": 5926
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001742456233327417,
+      "loss": 1.0323,
+      "step": 5927
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017423622735312388,
+      "loss": 1.1278,
+      "step": 5928
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001742268299132817,
+      "loss": 1.1276,
+      "step": 5929
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017421743101339997,
+      "loss": 1.0605,
+      "step": 5930
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017420803065366355,
+      "loss": 1.2018,
+      "step": 5931
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017419862883425737,
+      "loss": 1.0305,
+      "step": 5932
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017418922555536637,
+      "loss": 1.0372,
+      "step": 5933
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017417982081717546,
+      "loss": 0.8673,
+      "step": 5934
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017417041461986968,
+      "loss": 1.0403,
+      "step": 5935
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017416100696363403,
+      "loss": 0.9273,
+      "step": 5936
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017415159784865355,
+      "loss": 1.2276,
+      "step": 5937
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017414218727511335,
+      "loss": 0.9851,
+      "step": 5938
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001741327752431985,
+      "loss": 0.9365,
+      "step": 5939
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017412336175309418,
+      "loss": 0.8932,
+      "step": 5940
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001741139468049855,
+      "loss": 1.0055,
+      "step": 5941
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017410453039905764,
+      "loss": 1.0203,
+      "step": 5942
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017409511253549593,
+      "loss": 0.9693,
+      "step": 5943
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001740856932144855,
+      "loss": 1.1816,
+      "step": 5944
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017407627243621167,
+      "loss": 1.3036,
+      "step": 5945
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017406685020085976,
+      "loss": 1.244,
+      "step": 5946
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001740574265086151,
+      "loss": 1.1084,
+      "step": 5947
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017404800135966303,
+      "loss": 0.9984,
+      "step": 5948
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017403857475418896,
+      "loss": 1.0851,
+      "step": 5949
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017402914669237832,
+      "loss": 1.0278,
+      "step": 5950
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017401971717441654,
+      "loss": 1.0106,
+      "step": 5951
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017401028620048914,
+      "loss": 0.8676,
+      "step": 5952
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017400085377078156,
+      "loss": 0.9478,
+      "step": 5953
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017399141988547938,
+      "loss": 1.1767,
+      "step": 5954
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017398198454476814,
+      "loss": 0.9472,
+      "step": 5955
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017397254774883348,
+      "loss": 0.9163,
+      "step": 5956
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000173963109497861,
+      "loss": 1.0258,
+      "step": 5957
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001739536697920363,
+      "loss": 1.045,
+      "step": 5958
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001739442286315451,
+      "loss": 0.9793,
+      "step": 5959
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017393478601657314,
+      "loss": 0.94,
+      "step": 5960
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001739253419473061,
+      "loss": 1.0508,
+      "step": 5961
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017391589642392976,
+      "loss": 0.9897,
+      "step": 5962
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017390644944662994,
+      "loss": 0.8049,
+      "step": 5963
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017389700101559243,
+      "loss": 1.1551,
+      "step": 5964
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001738875511310031,
+      "loss": 0.942,
+      "step": 5965
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001738780997930478,
+      "loss": 1.1986,
+      "step": 5966
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001738686470019125,
+      "loss": 1.1275,
+      "step": 5967
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017385919275778307,
+      "loss": 1.1063,
+      "step": 5968
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001738497370608455,
+      "loss": 1.0414,
+      "step": 5969
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017384027991128575,
+      "loss": 1.0094,
+      "step": 5970
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001738308213092899,
+      "loss": 1.0525,
+      "step": 5971
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017382136125504397,
+      "loss": 1.2224,
+      "step": 5972
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017381189974873407,
+      "loss": 1.0498,
+      "step": 5973
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017380243679054625,
+      "loss": 1.1397,
+      "step": 5974
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001737929723806667,
+      "loss": 0.8568,
+      "step": 5975
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017378350651928157,
+      "loss": 1.2182,
+      "step": 5976
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017377403920657702,
+      "loss": 0.9648,
+      "step": 5977
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001737645704427393,
+      "loss": 0.9754,
+      "step": 5978
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017375510022795467,
+      "loss": 0.8765,
+      "step": 5979
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001737456285624094,
+      "loss": 1.0119,
+      "step": 5980
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017373615544628978,
+      "loss": 0.8842,
+      "step": 5981
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017372668087978216,
+      "loss": 0.9601,
+      "step": 5982
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001737172048630729,
+      "loss": 1.1913,
+      "step": 5983
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017370772739634842,
+      "loss": 0.9573,
+      "step": 5984
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001736982484797951,
+      "loss": 1.0236,
+      "step": 5985
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017368876811359939,
+      "loss": 1.1928,
+      "step": 5986
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001736792862979478,
+      "loss": 1.2313,
+      "step": 5987
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017366980303302684,
+      "loss": 1.0935,
+      "step": 5988
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017366031831902304,
+      "loss": 1.0616,
+      "step": 5989
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017365083215612292,
+      "loss": 1.1418,
+      "step": 5990
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017364134454451313,
+      "loss": 1.3658,
+      "step": 5991
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017363185548438024,
+      "loss": 1.0951,
+      "step": 5992
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017362236497591094,
+      "loss": 0.8919,
+      "step": 5993
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001736128730192919,
+      "loss": 0.9895,
+      "step": 5994
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017360337961470982,
+      "loss": 1.0658,
+      "step": 5995
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017359388476235145,
+      "loss": 0.9213,
+      "step": 5996
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001735843884624035,
+      "loss": 1.0722,
+      "step": 5997
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017357489071505286,
+      "loss": 1.0479,
+      "step": 5998
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017356539152048623,
+      "loss": 1.1868,
+      "step": 5999
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017355589087889056,
+      "loss": 1.0004,
+      "step": 6000
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001735463887904527,
+      "loss": 0.9747,
+      "step": 6001
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001735368852553595,
+      "loss": 0.8714,
+      "step": 6002
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017352738027379797,
+      "loss": 1.259,
+      "step": 6003
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017351787384595503,
+      "loss": 1.0485,
+      "step": 6004
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017350836597201767,
+      "loss": 0.8302,
+      "step": 6005
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017349885665217296,
+      "loss": 1.1843,
+      "step": 6006
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017348934588660788,
+      "loss": 1.0755,
+      "step": 6007
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017347983367550955,
+      "loss": 1.2059,
+      "step": 6008
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017347032001906504,
+      "loss": 0.8819,
+      "step": 6009
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017346080491746152,
+      "loss": 1.122,
+      "step": 6010
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017345128837088615,
+      "loss": 1.1437,
+      "step": 6011
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017344177037952608,
+      "loss": 1.1945,
+      "step": 6012
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017343225094356855,
+      "loss": 0.9548,
+      "step": 6013
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017342273006320085,
+      "loss": 1.214,
+      "step": 6014
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001734132077386102,
+      "loss": 1.0608,
+      "step": 6015
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001734036839699839,
+      "loss": 1.0271,
+      "step": 6016
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017339415875750934,
+      "loss": 1.1374,
+      "step": 6017
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001733846321013738,
+      "loss": 1.3438,
+      "step": 6018
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017337510400176472,
+      "loss": 1.0588,
+      "step": 6019
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017336557445886953,
+      "loss": 0.9936,
+      "step": 6020
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001733560434728757,
+      "loss": 0.9884,
+      "step": 6021
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001733465110439706,
+      "loss": 1.1276,
+      "step": 6022
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001733369771723418,
+      "loss": 1.1045,
+      "step": 6023
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017332744185817684,
+      "loss": 1.0651,
+      "step": 6024
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017331790510166325,
+      "loss": 0.9886,
+      "step": 6025
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017330836690298865,
+      "loss": 1.0243,
+      "step": 6026
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017329882726234064,
+      "loss": 0.8997,
+      "step": 6027
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017328928617990684,
+      "loss": 1.0895,
+      "step": 6028
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017327974365587495,
+      "loss": 1.0279,
+      "step": 6029
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017327019969043267,
+      "loss": 1.0419,
+      "step": 6030
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001732606542837677,
+      "loss": 1.0659,
+      "step": 6031
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017325110743606788,
+      "loss": 1.2278,
+      "step": 6032
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001732415591475209,
+      "loss": 1.0935,
+      "step": 6033
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017323200941831462,
+      "loss": 1.2799,
+      "step": 6034
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017322245824863687,
+      "loss": 1.1993,
+      "step": 6035
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017321290563867555,
+      "loss": 1.1159,
+      "step": 6036
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017320335158861855,
+      "loss": 1.088,
+      "step": 6037
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017319379609865375,
+      "loss": 1.1536,
+      "step": 6038
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017318423916896914,
+      "loss": 0.8928,
+      "step": 6039
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001731746807997527,
+      "loss": 1.0715,
+      "step": 6040
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001731651209911925,
+      "loss": 0.8642,
+      "step": 6041
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017315555974347648,
+      "loss": 1.0219,
+      "step": 6042
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017314599705679277,
+      "loss": 1.0374,
+      "step": 6043
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017313643293132946,
+      "loss": 1.0482,
+      "step": 6044
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001731268673672747,
+      "loss": 0.7455,
+      "step": 6045
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001731173003648166,
+      "loss": 1.0707,
+      "step": 6046
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017310773192414334,
+      "loss": 1.0101,
+      "step": 6047
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017309816204544317,
+      "loss": 1.1995,
+      "step": 6048
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017308859072890433,
+      "loss": 1.0904,
+      "step": 6049
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017307901797471503,
+      "loss": 1.2453,
+      "step": 6050
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017306944378306363,
+      "loss": 1.0854,
+      "step": 6051
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017305986815413843,
+      "loss": 1.0813,
+      "step": 6052
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017305029108812776,
+      "loss": 1.1566,
+      "step": 6053
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017304071258522006,
+      "loss": 0.9958,
+      "step": 6054
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017303113264560368,
+      "loss": 1.1289,
+      "step": 6055
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017302155126946707,
+      "loss": 1.1176,
+      "step": 6056
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017301196845699866,
+      "loss": 1.0992,
+      "step": 6057
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017300238420838704,
+      "loss": 0.9911,
+      "step": 6058
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017299279852382067,
+      "loss": 0.9831,
+      "step": 6059
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017298321140348812,
+      "loss": 1.0526,
+      "step": 6060
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017297362284757794,
+      "loss": 0.9954,
+      "step": 6061
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017296403285627877,
+      "loss": 1.0974,
+      "step": 6062
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001729544414297792,
+      "loss": 0.9096,
+      "step": 6063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017294484856826794,
+      "loss": 1.104,
+      "step": 6064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017293525427193368,
+      "loss": 1.108,
+      "step": 6065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017292565854096507,
+      "loss": 1.1547,
+      "step": 6066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017291606137555095,
+      "loss": 0.9107,
+      "step": 6067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017290646277588002,
+      "loss": 0.9181,
+      "step": 6068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017289686274214118,
+      "loss": 1.091,
+      "step": 6069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017288726127452316,
+      "loss": 1.1242,
+      "step": 6070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017287765837321485,
+      "loss": 0.9279,
+      "step": 6071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001728680540384052,
+      "loss": 1.1609,
+      "step": 6072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017285844827028303,
+      "loss": 1.2785,
+      "step": 6073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017284884106903736,
+      "loss": 1.1927,
+      "step": 6074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001728392324348571,
+      "loss": 1.1242,
+      "step": 6075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017282962236793136,
+      "loss": 1.0361,
+      "step": 6076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017282001086844906,
+      "loss": 0.9232,
+      "step": 6077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017281039793659927,
+      "loss": 0.9448,
+      "step": 6078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001728007835725711,
+      "loss": 1.0864,
+      "step": 6079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017279116777655374,
+      "loss": 1.1074,
+      "step": 6080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017278155054873618,
+      "loss": 1.1364,
+      "step": 6081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017277193188930775,
+      "loss": 1.0728,
+      "step": 6082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001727623117984575,
+      "loss": 0.9327,
+      "step": 6083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017275269027637474,
+      "loss": 1.1574,
+      "step": 6084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017274306732324872,
+      "loss": 1.0945,
+      "step": 6085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001727334429392687,
+      "loss": 0.9909,
+      "step": 6086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017272381712462402,
+      "loss": 1.0501,
+      "step": 6087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000172714189879504,
+      "loss": 1.0692,
+      "step": 6088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.000172704561204098,
+      "loss": 0.9427,
+      "step": 6089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017269493109859545,
+      "loss": 1.091,
+      "step": 6090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017268529956318576,
+      "loss": 1.0469,
+      "step": 6091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017267566659805834,
+      "loss": 0.8967,
+      "step": 6092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001726660322034027,
+      "loss": 1.1096,
+      "step": 6093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001726563963794084,
+      "loss": 1.3159,
+      "step": 6094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001726467591262649,
+      "loss": 1.1883,
+      "step": 6095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017263712044416179,
+      "loss": 1.0735,
+      "step": 6096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017262748033328867,
+      "loss": 1.0339,
+      "step": 6097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017261783879383513,
+      "loss": 0.9834,
+      "step": 6098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017260819582599088,
+      "loss": 1.0711,
+      "step": 6099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017259855142994556,
+      "loss": 1.1304,
+      "step": 6100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017258890560588888,
+      "loss": 0.9654,
+      "step": 6101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017257925835401054,
+      "loss": 1.0183,
+      "step": 6102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017256960967450033,
+      "loss": 0.9976,
+      "step": 6103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017255995956754806,
+      "loss": 1.0857,
+      "step": 6104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017255030803334354,
+      "loss": 1.1635,
+      "step": 6105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001725406550720766,
+      "loss": 1.1717,
+      "step": 6106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017253100068393708,
+      "loss": 0.9075,
+      "step": 6107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017252134486911495,
+      "loss": 0.9976,
+      "step": 6108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017251168762780013,
+      "loss": 1.038,
+      "step": 6109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001725020289601825,
+      "loss": 0.958,
+      "step": 6110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017249236886645214,
+      "loss": 1.0097,
+      "step": 6111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017248270734679903,
+      "loss": 1.0082,
+      "step": 6112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017247304440141318,
+      "loss": 1.1822,
+      "step": 6113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017246338003048468,
+      "loss": 1.2515,
+      "step": 6114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017245371423420366,
+      "loss": 1.0014,
+      "step": 6115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017244404701276023,
+      "loss": 1.1786,
+      "step": 6116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.0001724343783663445,
+      "loss": 1.3137,
+      "step": 6117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017242470829514672,
+      "loss": 1.2351,
+      "step": 6118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00017241503679935703,
+      "loss": 0.9746,
+      "step": 6119
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017240536387916572,
+      "loss": 1.1704,
+      "step": 6120
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017239568953476307,
+      "loss": 0.9373,
+      "step": 6121
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001723860137663393,
+      "loss": 1.2801,
+      "step": 6122
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017237633657408478,
+      "loss": 0.9028,
+      "step": 6123
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017236665795818988,
+      "loss": 1.0729,
+      "step": 6124
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017235697791884494,
+      "loss": 1.06,
+      "step": 6125
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017234729645624038,
+      "loss": 1.2892,
+      "step": 6126
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017233761357056664,
+      "loss": 1.1605,
+      "step": 6127
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017232792926201419,
+      "loss": 0.9728,
+      "step": 6128
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017231824353077348,
+      "loss": 0.9501,
+      "step": 6129
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017230855637703508,
+      "loss": 1.1183,
+      "step": 6130
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001722988678009895,
+      "loss": 1.1109,
+      "step": 6131
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017228917780282731,
+      "loss": 1.2849,
+      "step": 6132
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017227948638273916,
+      "loss": 1.0883,
+      "step": 6133
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001722697935409156,
+      "loss": 1.16,
+      "step": 6134
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001722600992775474,
+      "loss": 1.0927,
+      "step": 6135
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017225040359282513,
+      "loss": 1.3276,
+      "step": 6136
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017224070648693958,
+      "loss": 1.0677,
+      "step": 6137
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017223100796008147,
+      "loss": 1.0434,
+      "step": 6138
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017222130801244155,
+      "loss": 1.015,
+      "step": 6139
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017221160664421064,
+      "loss": 0.8287,
+      "step": 6140
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017220190385557955,
+      "loss": 1.0064,
+      "step": 6141
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017219219964673913,
+      "loss": 1.0628,
+      "step": 6142
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017218249401788031,
+      "loss": 0.9789,
+      "step": 6143
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017217278696919394,
+      "loss": 0.9789,
+      "step": 6144
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017216307850087098,
+      "loss": 1.1628,
+      "step": 6145
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017215336861310242,
+      "loss": 0.9382,
+      "step": 6146
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017214365730607923,
+      "loss": 1.1416,
+      "step": 6147
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017213394457999242,
+      "loss": 1.2738,
+      "step": 6148
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017212423043503304,
+      "loss": 1.1147,
+      "step": 6149
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017211451487139217,
+      "loss": 1.1209,
+      "step": 6150
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017210479788926092,
+      "loss": 0.9145,
+      "step": 6151
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017209507948883043,
+      "loss": 0.9528,
+      "step": 6152
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017208535967029188,
+      "loss": 1.1292,
+      "step": 6153
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017207563843383642,
+      "loss": 1.1104,
+      "step": 6154
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017206591577965526,
+      "loss": 1.0503,
+      "step": 6155
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001720561917079397,
+      "loss": 0.9937,
+      "step": 6156
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017204646621888096,
+      "loss": 0.9615,
+      "step": 6157
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017203673931267036,
+      "loss": 0.8598,
+      "step": 6158
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017202701098949924,
+      "loss": 1.0111,
+      "step": 6159
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001720172812495589,
+      "loss": 1.0274,
+      "step": 6160
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001720075500930408,
+      "loss": 0.8824,
+      "step": 6161
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017199781752013632,
+      "loss": 1.0491,
+      "step": 6162
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017198808353103686,
+      "loss": 1.0141,
+      "step": 6163
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001719783481259339,
+      "loss": 1.1687,
+      "step": 6164
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017196861130501905,
+      "loss": 1.021,
+      "step": 6165
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017195887306848366,
+      "loss": 1.2289,
+      "step": 6166
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017194913341651939,
+      "loss": 1.248,
+      "step": 6167
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017193939234931777,
+      "loss": 1.1736,
+      "step": 6168
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017192964986707048,
+      "loss": 1.0744,
+      "step": 6169
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017191990596996905,
+      "loss": 0.9254,
+      "step": 6170
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001719101606582052,
+      "loss": 1.0775,
+      "step": 6171
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017190041393197063,
+      "loss": 1.1147,
+      "step": 6172
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017189066579145705,
+      "loss": 1.0816,
+      "step": 6173
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017188091623685615,
+      "loss": 1.1148,
+      "step": 6174
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017187116526835977,
+      "loss": 1.014,
+      "step": 6175
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001718614128861597,
+      "loss": 1.0262,
+      "step": 6176
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001718516590904478,
+      "loss": 0.8279,
+      "step": 6177
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017184190388141586,
+      "loss": 1.1566,
+      "step": 6178
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017183214725925575,
+      "loss": 1.0556,
+      "step": 6179
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017182238922415946,
+      "loss": 1.0321,
+      "step": 6180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017181262977631888,
+      "loss": 1.1806,
+      "step": 6181
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017180286891592603,
+      "loss": 1.1475,
+      "step": 6182
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017179310664317284,
+      "loss": 0.9222,
+      "step": 6183
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017178334295825135,
+      "loss": 1.1073,
+      "step": 6184
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001717735778613536,
+      "loss": 1.0691,
+      "step": 6185
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017176381135267178,
+      "loss": 1.1502,
+      "step": 6186
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017175404343239785,
+      "loss": 0.977,
+      "step": 6187
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000171744274100724,
+      "loss": 0.9989,
+      "step": 6188
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001717345033578424,
+      "loss": 1.1427,
+      "step": 6189
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017172473120394526,
+      "loss": 0.8587,
+      "step": 6190
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017171495763922476,
+      "loss": 1.2055,
+      "step": 6191
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017170518266387317,
+      "loss": 1.0705,
+      "step": 6192
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017169540627808274,
+      "loss": 1.1464,
+      "step": 6193
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001716856284820458,
+      "loss": 1.175,
+      "step": 6194
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017167584927595467,
+      "loss": 1.2847,
+      "step": 6195
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017166606866000165,
+      "loss": 0.9116,
+      "step": 6196
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017165628663437922,
+      "loss": 1.0276,
+      "step": 6197
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017164650319927976,
+      "loss": 1.0171,
+      "step": 6198
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017163671835489566,
+      "loss": 1.1305,
+      "step": 6199
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017162693210141946,
+      "loss": 0.8119,
+      "step": 6200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001716171444390436,
+      "loss": 1.1572,
+      "step": 6201
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017160735536796062,
+      "loss": 0.9723,
+      "step": 6202
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017159756488836306,
+      "loss": 1.0799,
+      "step": 6203
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017158777300044356,
+      "loss": 1.011,
+      "step": 6204
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017157797970439464,
+      "loss": 0.9701,
+      "step": 6205
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000171568185000409,
+      "loss": 1.0499,
+      "step": 6206
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017155838888867925,
+      "loss": 1.1146,
+      "step": 6207
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001715485913693981,
+      "loss": 1.084,
+      "step": 6208
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001715387924427583,
+      "loss": 0.9742,
+      "step": 6209
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017152899210895256,
+      "loss": 1.1018,
+      "step": 6210
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017151919036817367,
+      "loss": 1.0855,
+      "step": 6211
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017150938722061435,
+      "loss": 0.9992,
+      "step": 6212
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017149958266646754,
+      "loss": 1.2116,
+      "step": 6213
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017148977670592604,
+      "loss": 0.7905,
+      "step": 6214
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017147996933918278,
+      "loss": 1.0622,
+      "step": 6215
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017147016056643058,
+      "loss": 0.994,
+      "step": 6216
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017146035038786246,
+      "loss": 1.143,
+      "step": 6217
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017145053880367135,
+      "loss": 0.9298,
+      "step": 6218
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017144072581405027,
+      "loss": 1.397,
+      "step": 6219
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017143091141919218,
+      "loss": 1.1365,
+      "step": 6220
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017142109561929023,
+      "loss": 1.1514,
+      "step": 6221
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017141127841453739,
+      "loss": 1.1279,
+      "step": 6222
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017140145980512684,
+      "loss": 1.0206,
+      "step": 6223
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017139163979125165,
+      "loss": 1.0305,
+      "step": 6224
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017138181837310506,
+      "loss": 1.0608,
+      "step": 6225
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017137199555088017,
+      "loss": 0.9796,
+      "step": 6226
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017136217132477025,
+      "loss": 0.9865,
+      "step": 6227
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017135234569496855,
+      "loss": 1.0394,
+      "step": 6228
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001713425186616683,
+      "loss": 1.0914,
+      "step": 6229
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001713326902250628,
+      "loss": 0.9593,
+      "step": 6230
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001713228603853454,
+      "loss": 1.2524,
+      "step": 6231
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017131302914270944,
+      "loss": 1.0377,
+      "step": 6232
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001713031964973483,
+      "loss": 0.9818,
+      "step": 6233
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001712933624494554,
+      "loss": 1.178,
+      "step": 6234
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017128352699922417,
+      "loss": 1.0127,
+      "step": 6235
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017127369014684803,
+      "loss": 1.0481,
+      "step": 6236
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017126385189252053,
+      "loss": 1.0582,
+      "step": 6237
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001712540122364352,
+      "loss": 1.2313,
+      "step": 6238
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017124417117878553,
+      "loss": 1.0021,
+      "step": 6239
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001712343287197651,
+      "loss": 1.142,
+      "step": 6240
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017122448485956755,
+      "loss": 1.208,
+      "step": 6241
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017121463959838645,
+      "loss": 1.1588,
+      "step": 6242
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017120479293641555,
+      "loss": 1.2193,
+      "step": 6243
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017119494487384843,
+      "loss": 1.0366,
+      "step": 6244
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017118509541087887,
+      "loss": 1.3069,
+      "step": 6245
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017117524454770058,
+      "loss": 1.0202,
+      "step": 6246
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017116539228450735,
+      "loss": 1.2144,
+      "step": 6247
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001711555386214929,
+      "loss": 1.2307,
+      "step": 6248
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017114568355885114,
+      "loss": 1.1346,
+      "step": 6249
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017113582709677588,
+      "loss": 1.0864,
+      "step": 6250
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000171125969235461,
+      "loss": 1.0054,
+      "step": 6251
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017111610997510044,
+      "loss": 0.8853,
+      "step": 6252
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017110624931588806,
+      "loss": 1.1447,
+      "step": 6253
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017109638725801787,
+      "loss": 1.0712,
+      "step": 6254
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017108652380168384,
+      "loss": 1.0576,
+      "step": 6255
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017107665894708,
+      "loss": 1.0116,
+      "step": 6256
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017106679269440035,
+      "loss": 0.8507,
+      "step": 6257
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017105692504383897,
+      "loss": 0.9391,
+      "step": 6258
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017104705599559,
+      "loss": 1.0645,
+      "step": 6259
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001710371855498475,
+      "loss": 1.1045,
+      "step": 6260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001710273137068057,
+      "loss": 1.0459,
+      "step": 6261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017101744046665873,
+      "loss": 1.0264,
+      "step": 6262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017100756582960082,
+      "loss": 0.9803,
+      "step": 6263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017099768979582616,
+      "loss": 0.9193,
+      "step": 6264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017098781236552903,
+      "loss": 1.1166,
+      "step": 6265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017097793353890373,
+      "loss": 1.0577,
+      "step": 6266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001709680533161446,
+      "loss": 1.1699,
+      "step": 6267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017095817169744595,
+      "loss": 1.1231,
+      "step": 6268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017094828868300215,
+      "loss": 0.9735,
+      "step": 6269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001709384042730076,
+      "loss": 1.1285,
+      "step": 6270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017092851846765673,
+      "loss": 1.0837,
+      "step": 6271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017091863126714402,
+      "loss": 1.11,
+      "step": 6272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001709087426716639,
+      "loss": 1.2627,
+      "step": 6273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017089885268141091,
+      "loss": 0.9842,
+      "step": 6274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017088896129657962,
+      "loss": 1.0862,
+      "step": 6275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001708790685173645,
+      "loss": 0.947,
+      "step": 6276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001708691743439602,
+      "loss": 0.937,
+      "step": 6277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017085927877656136,
+      "loss": 0.9329,
+      "step": 6278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017084938181536257,
+      "loss": 1.0466,
+      "step": 6279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017083948346055858,
+      "loss": 1.0217,
+      "step": 6280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000170829583712344,
+      "loss": 0.997,
+      "step": 6281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017081968257091361,
+      "loss": 0.7889,
+      "step": 6282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017080978003646216,
+      "loss": 1.1289,
+      "step": 6283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017079987610918443,
+      "loss": 0.9565,
+      "step": 6284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017078997078927521,
+      "loss": 0.9582,
+      "step": 6285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017078006407692938,
+      "loss": 1.1246,
+      "step": 6286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017077015597234175,
+      "loss": 1.1594,
+      "step": 6287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017076024647570725,
+      "loss": 1.114,
+      "step": 6288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017075033558722083,
+      "loss": 0.8555,
+      "step": 6289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017074042330707733,
+      "loss": 0.971,
+      "step": 6290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017073050963547183,
+      "loss": 1.0722,
+      "step": 6291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017072059457259932,
+      "loss": 1.0869,
+      "step": 6292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017071067811865476,
+      "loss": 1.1811,
+      "step": 6293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017070076027383326,
+      "loss": 0.9475,
+      "step": 6294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001706908410383299,
+      "loss": 1.0498,
+      "step": 6295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017068092041233978,
+      "loss": 1.1297,
+      "step": 6296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017067099839605807,
+      "loss": 1.0198,
+      "step": 6297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017066107498967984,
+      "loss": 1.0271,
+      "step": 6298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001706511501934004,
+      "loss": 1.184,
+      "step": 6299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017064122400741493,
+      "loss": 1.0868,
+      "step": 6300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017063129643191864,
+      "loss": 0.9238,
+      "step": 6301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017062136746710685,
+      "loss": 1.1498,
+      "step": 6302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017061143711317488,
+      "loss": 1.0862,
+      "step": 6303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017060150537031796,
+      "loss": 0.9215,
+      "step": 6304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017059157223873155,
+      "loss": 1.1249,
+      "step": 6305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000170581637718611,
+      "loss": 1.0411,
+      "step": 6306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017057170181015172,
+      "loss": 1.2043,
+      "step": 6307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017056176451354916,
+      "loss": 1.0186,
+      "step": 6308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017055182582899873,
+      "loss": 0.9291,
+      "step": 6309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.000170541885756696,
+      "loss": 0.8952,
+      "step": 6310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017053194429683647,
+      "loss": 1.0581,
+      "step": 6311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017052200144961565,
+      "loss": 0.9682,
+      "step": 6312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001705120572152292,
+      "loss": 1.2133,
+      "step": 6313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017050211159387263,
+      "loss": 1.3624,
+      "step": 6314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017049216458574158,
+      "loss": 1.1344,
+      "step": 6315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017048221619103178,
+      "loss": 1.2556,
+      "step": 6316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017047226640993883,
+      "loss": 0.9961,
+      "step": 6317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001704623152426585,
+      "loss": 0.9907,
+      "step": 6318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001704523626893865,
+      "loss": 1.0641,
+      "step": 6319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017044240875031862,
+      "loss": 0.9567,
+      "step": 6320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017043245342565063,
+      "loss": 1.0412,
+      "step": 6321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017042249671557835,
+      "loss": 1.1926,
+      "step": 6322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017041253862029766,
+      "loss": 1.1006,
+      "step": 6323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017040257914000443,
+      "loss": 1.1871,
+      "step": 6324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001703926182748945,
+      "loss": 1.005,
+      "step": 6325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001703826560251639,
+      "loss": 1.0697,
+      "step": 6326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017037269239100847,
+      "loss": 0.9401,
+      "step": 6327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001703627273726243,
+      "loss": 0.9881,
+      "step": 6328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017035276097020736,
+      "loss": 1.0537,
+      "step": 6329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001703427931839537,
+      "loss": 1.1526,
+      "step": 6330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017033282401405933,
+      "loss": 1.1875,
+      "step": 6331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001703228534607204,
+      "loss": 1.0249,
+      "step": 6332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017031288152413306,
+      "loss": 1.114,
+      "step": 6333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001703029082044934,
+      "loss": 1.3086,
+      "step": 6334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001702929335019976,
+      "loss": 1.1097,
+      "step": 6335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017028295741684186,
+      "loss": 0.9495,
+      "step": 6336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017027297994922248,
+      "loss": 1.1293,
+      "step": 6337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017026300109933558,
+      "loss": 0.9822,
+      "step": 6338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017025302086737755,
+      "loss": 1.1781,
+      "step": 6339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001702430392535447,
+      "loss": 1.0594,
+      "step": 6340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017023305625803333,
+      "loss": 1.1216,
+      "step": 6341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001702230718810398,
+      "loss": 0.8096,
+      "step": 6342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017021308612276054,
+      "loss": 1.0801,
+      "step": 6343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017020309898339194,
+      "loss": 1.1263,
+      "step": 6344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017019311046313047,
+      "loss": 1.1025,
+      "step": 6345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017018312056217257,
+      "loss": 1.0856,
+      "step": 6346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017017312928071477,
+      "loss": 1.0019,
+      "step": 6347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001701631366189536,
+      "loss": 1.1913,
+      "step": 6348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.0001701531425770856,
+      "loss": 0.895,
+      "step": 6349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00017014314715530735,
+      "loss": 1.0153,
+      "step": 6350
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017013315035381546,
+      "loss": 0.9961,
+      "step": 6351
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017012315217280655,
+      "loss": 1.1078,
+      "step": 6352
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017011315261247735,
+      "loss": 1.12,
+      "step": 6353
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017010315167302452,
+      "loss": 1.065,
+      "step": 6354
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001700931493546447,
+      "loss": 1.1383,
+      "step": 6355
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017008314565753477,
+      "loss": 0.7481,
+      "step": 6356
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001700731405818914,
+      "loss": 0.9098,
+      "step": 6357
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017006313412791144,
+      "loss": 0.8888,
+      "step": 6358
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017005312629579168,
+      "loss": 1.0653,
+      "step": 6359
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000170043117085729,
+      "loss": 1.0235,
+      "step": 6360
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017003310649792028,
+      "loss": 1.0614,
+      "step": 6361
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017002309453256244,
+      "loss": 0.9956,
+      "step": 6362
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017001308118985237,
+      "loss": 0.8884,
+      "step": 6363
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00017000306646998707,
+      "loss": 0.9479,
+      "step": 6364
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016999305037316352,
+      "loss": 1.0176,
+      "step": 6365
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016998303289957875,
+      "loss": 1.1099,
+      "step": 6366
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016997301404942977,
+      "loss": 1.1233,
+      "step": 6367
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016996299382291368,
+      "loss": 0.9086,
+      "step": 6368
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016995297222022754,
+      "loss": 1.1711,
+      "step": 6369
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016994294924156856,
+      "loss": 1.104,
+      "step": 6370
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016993292488713378,
+      "loss": 1.0144,
+      "step": 6371
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016992289915712045,
+      "loss": 0.9723,
+      "step": 6372
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016991287205172573,
+      "loss": 1.2805,
+      "step": 6373
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016990284357114695,
+      "loss": 1.2789,
+      "step": 6374
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016989281371558126,
+      "loss": 0.9525,
+      "step": 6375
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016988278248522597,
+      "loss": 1.0101,
+      "step": 6376
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016987274988027843,
+      "loss": 1.0468,
+      "step": 6377
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016986271590093595,
+      "loss": 0.942,
+      "step": 6378
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016985268054739593,
+      "loss": 0.9609,
+      "step": 6379
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016984264381985576,
+      "loss": 1.2126,
+      "step": 6380
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001698326057185128,
+      "loss": 1.1387,
+      "step": 6381
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001698225662435646,
+      "loss": 1.0504,
+      "step": 6382
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016981252539520857,
+      "loss": 1.0105,
+      "step": 6383
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001698024831736422,
+      "loss": 1.0347,
+      "step": 6384
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016979243957906305,
+      "loss": 1.1218,
+      "step": 6385
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016978239461166872,
+      "loss": 1.0741,
+      "step": 6386
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001697723482716567,
+      "loss": 0.8583,
+      "step": 6387
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016976230055922468,
+      "loss": 0.9091,
+      "step": 6388
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016975225147457026,
+      "loss": 0.9512,
+      "step": 6389
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016974220101789112,
+      "loss": 0.9725,
+      "step": 6390
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016973214918938488,
+      "loss": 1.1063,
+      "step": 6391
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016972209598924943,
+      "loss": 1.1302,
+      "step": 6392
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016971204141768233,
+      "loss": 1.1852,
+      "step": 6393
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016970198547488146,
+      "loss": 1.1268,
+      "step": 6394
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001696919281610446,
+      "loss": 1.1079,
+      "step": 6395
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016968186947636954,
+      "loss": 1.0917,
+      "step": 6396
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016967180942105423,
+      "loss": 0.9424,
+      "step": 6397
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001696617479952964,
+      "loss": 0.9353,
+      "step": 6398
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001696516851992941,
+      "loss": 1.1808,
+      "step": 6399
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001696416210332452,
+      "loss": 1.3,
+      "step": 6400
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001696315554973477,
+      "loss": 1.0523,
+      "step": 6401
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001696214885917995,
+      "loss": 0.9314,
+      "step": 6402
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001696114203167987,
+      "loss": 1.1464,
+      "step": 6403
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016960135067254332,
+      "loss": 1.1109,
+      "step": 6404
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016959127965923142,
+      "loss": 1.0916,
+      "step": 6405
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016958120727706116,
+      "loss": 0.8548,
+      "step": 6406
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016957113352623053,
+      "loss": 1.1721,
+      "step": 6407
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001695610584069378,
+      "loss": 1.1363,
+      "step": 6408
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001695509819193811,
+      "loss": 0.9734,
+      "step": 6409
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016954090406375862,
+      "loss": 0.9106,
+      "step": 6410
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001695308248402686,
+      "loss": 1.0849,
+      "step": 6411
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016952074424910935,
+      "loss": 1.0392,
+      "step": 6412
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001695106622904791,
+      "loss": 1.3265,
+      "step": 6413
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016950057896457615,
+      "loss": 0.9793,
+      "step": 6414
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016949049427159892,
+      "loss": 1.0151,
+      "step": 6415
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016948040821174564,
+      "loss": 1.2011,
+      "step": 6416
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016947032078521484,
+      "loss": 1.0325,
+      "step": 6417
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016946023199220487,
+      "loss": 1.1212,
+      "step": 6418
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001694501418329142,
+      "loss": 1.2042,
+      "step": 6419
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016944005030754123,
+      "loss": 1.0277,
+      "step": 6420
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001694299574162846,
+      "loss": 1.0665,
+      "step": 6421
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016941986315934272,
+      "loss": 1.2464,
+      "step": 6422
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016940976753691418,
+      "loss": 1.0654,
+      "step": 6423
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016939967054919753,
+      "loss": 0.9711,
+      "step": 6424
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016938957219639147,
+      "loss": 1.1803,
+      "step": 6425
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016937947247869453,
+      "loss": 0.8954,
+      "step": 6426
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016936937139630544,
+      "loss": 0.9664,
+      "step": 6427
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016935926894942288,
+      "loss": 1.1268,
+      "step": 6428
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001693491651382455,
+      "loss": 1.0898,
+      "step": 6429
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016933905996297215,
+      "loss": 0.9359,
+      "step": 6430
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016932895342380147,
+      "loss": 0.9838,
+      "step": 6431
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016931884552093236,
+      "loss": 1.0097,
+      "step": 6432
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001693087362545636,
+      "loss": 1.1419,
+      "step": 6433
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016929862562489403,
+      "loss": 1.1362,
+      "step": 6434
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016928851363212256,
+      "loss": 1.2302,
+      "step": 6435
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016927840027644808,
+      "loss": 0.883,
+      "step": 6436
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001692682855580695,
+      "loss": 1.1676,
+      "step": 6437
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016925816947718577,
+      "loss": 1.0693,
+      "step": 6438
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016924805203399592,
+      "loss": 0.8884,
+      "step": 6439
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001692379332286989,
+      "loss": 1.0361,
+      "step": 6440
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001692278130614938,
+      "loss": 0.9688,
+      "step": 6441
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016921769153257967,
+      "loss": 1.1056,
+      "step": 6442
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016920756864215557,
+      "loss": 0.9146,
+      "step": 6443
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016919744439042065,
+      "loss": 1.153,
+      "step": 6444
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016918731877757406,
+      "loss": 1.0117,
+      "step": 6445
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016917719180381492,
+      "loss": 1.0989,
+      "step": 6446
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016916706346934248,
+      "loss": 1.0284,
+      "step": 6447
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016915693377435594,
+      "loss": 1.156,
+      "step": 6448
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016914680271905454,
+      "loss": 1.0757,
+      "step": 6449
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001691366703036376,
+      "loss": 0.9794,
+      "step": 6450
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016912653652830434,
+      "loss": 1.0365,
+      "step": 6451
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001691164013932542,
+      "loss": 0.9284,
+      "step": 6452
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016910626489868649,
+      "loss": 1.0027,
+      "step": 6453
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016909612704480055,
+      "loss": 1.0864,
+      "step": 6454
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016908598783179585,
+      "loss": 1.0488,
+      "step": 6455
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016907584725987183,
+      "loss": 1.2863,
+      "step": 6456
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001690657053292279,
+      "loss": 1.0894,
+      "step": 6457
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016905556204006362,
+      "loss": 1.046,
+      "step": 6458
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016904541739257845,
+      "loss": 1.1553,
+      "step": 6459
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016903527138697197,
+      "loss": 0.9321,
+      "step": 6460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016902512402344373,
+      "loss": 1.1855,
+      "step": 6461
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016901497530219333,
+      "loss": 1.1152,
+      "step": 6462
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016900482522342044,
+      "loss": 0.9883,
+      "step": 6463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016899467378732463,
+      "loss": 1.0374,
+      "step": 6464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016898452099410564,
+      "loss": 1.1306,
+      "step": 6465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016897436684396318,
+      "loss": 1.1017,
+      "step": 6466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016896421133709693,
+      "loss": 1.1062,
+      "step": 6467
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001689540544737067,
+      "loss": 0.9848,
+      "step": 6468
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016894389625399226,
+      "loss": 1.1464,
+      "step": 6469
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016893373667815336,
+      "loss": 0.9752,
+      "step": 6470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016892357574638998,
+      "loss": 0.9548,
+      "step": 6471
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016891341345890185,
+      "loss": 0.8579,
+      "step": 6472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016890324981588893,
+      "loss": 0.9989,
+      "step": 6473
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001688930848175511,
+      "loss": 1.0488,
+      "step": 6474
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016888291846408834,
+      "loss": 0.9742,
+      "step": 6475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016887275075570064,
+      "loss": 0.9536,
+      "step": 6476
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001688625816925879,
+      "loss": 1.1196,
+      "step": 6477
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016885241127495023,
+      "loss": 1.0965,
+      "step": 6478
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001688422395029877,
+      "loss": 1.1619,
+      "step": 6479
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001688320663769003,
+      "loss": 1.0591,
+      "step": 6480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016882189189688823,
+      "loss": 1.2155,
+      "step": 6481
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016881171606315158,
+      "loss": 0.9891,
+      "step": 6482
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016880153887589047,
+      "loss": 1.0593,
+      "step": 6483
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016879136033530516,
+      "loss": 0.8975,
+      "step": 6484
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016878118044159578,
+      "loss": 0.9712,
+      "step": 6485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016877099919496266,
+      "loss": 0.9695,
+      "step": 6486
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000168760816595606,
+      "loss": 1.2341,
+      "step": 6487
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016875063264372607,
+      "loss": 1.0566,
+      "step": 6488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016874044733952327,
+      "loss": 1.0797,
+      "step": 6489
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001687302606831979,
+      "loss": 1.0059,
+      "step": 6490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001687200726749503,
+      "loss": 1.1974,
+      "step": 6491
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016870988331498093,
+      "loss": 0.857,
+      "step": 6492
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016869969260349018,
+      "loss": 1.0159,
+      "step": 6493
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016868950054067848,
+      "loss": 1.0347,
+      "step": 6494
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016867930712674637,
+      "loss": 0.9208,
+      "step": 6495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001686691123618943,
+      "loss": 1.034,
+      "step": 6496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001686589162463228,
+      "loss": 1.0327,
+      "step": 6497
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016864871878023244,
+      "loss": 0.931,
+      "step": 6498
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016863851996382385,
+      "loss": 1.0105,
+      "step": 6499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001686283197972976,
+      "loss": 1.1466,
+      "step": 6500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016861811828085429,
+      "loss": 1.1697,
+      "step": 6501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016860791541469463,
+      "loss": 1.0285,
+      "step": 6502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016859771119901929,
+      "loss": 1.1426,
+      "step": 6503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016858750563402902,
+      "loss": 1.0329,
+      "step": 6504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016857729871992452,
+      "loss": 1.109,
+      "step": 6505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016856709045690656,
+      "loss": 0.9038,
+      "step": 6506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000168556880845176,
+      "loss": 0.9194,
+      "step": 6507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001685466698849336,
+      "loss": 1.0394,
+      "step": 6508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016853645757638024,
+      "loss": 0.8822,
+      "step": 6509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016852624391971673,
+      "loss": 1.1987,
+      "step": 6510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016851602891514408,
+      "loss": 0.9916,
+      "step": 6511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001685058125628632,
+      "loss": 1.1819,
+      "step": 6512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001684955948630749,
+      "loss": 1.13,
+      "step": 6513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016848537581598034,
+      "loss": 1.0955,
+      "step": 6514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016847515542178044,
+      "loss": 1.0816,
+      "step": 6515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001684649336806763,
+      "loss": 1.0925,
+      "step": 6516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016845471059286887,
+      "loss": 1.2272,
+      "step": 6517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016844448615855933,
+      "loss": 1.2114,
+      "step": 6518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016843426037794877,
+      "loss": 1.2079,
+      "step": 6519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016842403325123835,
+      "loss": 0.9816,
+      "step": 6520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016841380477862917,
+      "loss": 1.0855,
+      "step": 6521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016840357496032254,
+      "loss": 0.9346,
+      "step": 6522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016839334379651954,
+      "loss": 1.0753,
+      "step": 6523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016838311128742153,
+      "loss": 0.9557,
+      "step": 6524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016837287743322975,
+      "loss": 1.0937,
+      "step": 6525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016836264223414542,
+      "loss": 1.2162,
+      "step": 6526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016835240569036997,
+      "loss": 1.2053,
+      "step": 6527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016834216780210474,
+      "loss": 0.9626,
+      "step": 6528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016833192856955108,
+      "loss": 0.9756,
+      "step": 6529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016832168799291042,
+      "loss": 0.8479,
+      "step": 6530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016831144607238414,
+      "loss": 0.8881,
+      "step": 6531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016830120280817375,
+      "loss": 1.0126,
+      "step": 6532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016829095820048072,
+      "loss": 0.8766,
+      "step": 6533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016828071224950653,
+      "loss": 1.0371,
+      "step": 6534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001682704649554528,
+      "loss": 1.0352,
+      "step": 6535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016826021631852098,
+      "loss": 1.2205,
+      "step": 6536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016824996633891273,
+      "loss": 1.0843,
+      "step": 6537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001682397150168297,
+      "loss": 1.1883,
+      "step": 6538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016822946235247346,
+      "loss": 1.0257,
+      "step": 6539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001682192083460457,
+      "loss": 1.0453,
+      "step": 6540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016820895299774814,
+      "loss": 1.061,
+      "step": 6541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016819869630778253,
+      "loss": 1.0213,
+      "step": 6542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001681884382763505,
+      "loss": 0.9888,
+      "step": 6543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016817817890365396,
+      "loss": 1.1463,
+      "step": 6544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016816791818989466,
+      "loss": 1.0002,
+      "step": 6545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016815765613527436,
+      "loss": 0.9692,
+      "step": 6546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.000168147392739995,
+      "loss": 1.0006,
+      "step": 6547
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016813712800425846,
+      "loss": 1.0167,
+      "step": 6548
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016812686192826663,
+      "loss": 1.0333,
+      "step": 6549
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001681165945122214,
+      "loss": 1.0624,
+      "step": 6550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001681063257563248,
+      "loss": 0.889,
+      "step": 6551
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016809605566077879,
+      "loss": 0.9983,
+      "step": 6552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016808578422578533,
+      "loss": 1.1511,
+      "step": 6553
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016807551145154654,
+      "loss": 1.0716,
+      "step": 6554
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016806523733826445,
+      "loss": 1.0536,
+      "step": 6555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016805496188614113,
+      "loss": 1.1319,
+      "step": 6556
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016804468509537874,
+      "loss": 1.0556,
+      "step": 6557
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001680344069661794,
+      "loss": 1.1277,
+      "step": 6558
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016802412749874525,
+      "loss": 1.1838,
+      "step": 6559
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016801384669327856,
+      "loss": 0.9546,
+      "step": 6560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001680035645499815,
+      "loss": 1.1164,
+      "step": 6561
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016799328106905636,
+      "loss": 1.0586,
+      "step": 6562
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016798299625070534,
+      "loss": 0.9633,
+      "step": 6563
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016797271009513085,
+      "loss": 0.9626,
+      "step": 6564
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016796242260253512,
+      "loss": 1.0701,
+      "step": 6565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016795213377312056,
+      "loss": 1.1231,
+      "step": 6566
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016794184360708952,
+      "loss": 0.8466,
+      "step": 6567
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001679315521046444,
+      "loss": 0.9537,
+      "step": 6568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001679212592659877,
+      "loss": 1.2366,
+      "step": 6569
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016791096509132184,
+      "loss": 0.9679,
+      "step": 6570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016790066958084928,
+      "loss": 1.1781,
+      "step": 6571
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016789037273477254,
+      "loss": 1.1404,
+      "step": 6572
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001678800745532942,
+      "loss": 1.1278,
+      "step": 6573
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016786977503661677,
+      "loss": 0.8687,
+      "step": 6574
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001678594741849429,
+      "loss": 1.1014,
+      "step": 6575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016784917199847514,
+      "loss": 1.1256,
+      "step": 6576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001678388684774162,
+      "loss": 1.0808,
+      "step": 6577
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001678285636219687,
+      "loss": 0.971,
+      "step": 6578
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016781825743233538,
+      "loss": 1.2394,
+      "step": 6579
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016780794990871892,
+      "loss": 1.0907,
+      "step": 6580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.0001677976410513221,
+      "loss": 0.9874,
+      "step": 6581
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016778733086034766,
+      "loss": 0.9938,
+      "step": 6582
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016777701933599844,
+      "loss": 0.9287,
+      "step": 6583
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016776670647847727,
+      "loss": 0.9496,
+      "step": 6584
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016775639228798695,
+      "loss": 0.9064,
+      "step": 6585
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001677460767647304,
+      "loss": 1.0714,
+      "step": 6586
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016773575990891056,
+      "loss": 0.9643,
+      "step": 6587
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016772544172073028,
+      "loss": 1.054,
+      "step": 6588
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016771512220039255,
+      "loss": 1.2446,
+      "step": 6589
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016770480134810042,
+      "loss": 1.1668,
+      "step": 6590
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016769447916405682,
+      "loss": 1.0559,
+      "step": 6591
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016768415564846477,
+      "loss": 0.9828,
+      "step": 6592
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016767383080152742,
+      "loss": 1.1412,
+      "step": 6593
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016766350462344782,
+      "loss": 0.9936,
+      "step": 6594
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016765317711442908,
+      "loss": 0.9542,
+      "step": 6595
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016764284827467434,
+      "loss": 1.1594,
+      "step": 6596
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016763251810438678,
+      "loss": 1.014,
+      "step": 6597
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016762218660376957,
+      "loss": 1.0405,
+      "step": 6598
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016761185377302593,
+      "loss": 1.0678,
+      "step": 6599
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016760151961235916,
+      "loss": 1.0179,
+      "step": 6600
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016759118412197247,
+      "loss": 1.0251,
+      "step": 6601
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001675808473020692,
+      "loss": 1.0533,
+      "step": 6602
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016757050915285263,
+      "loss": 0.9035,
+      "step": 6603
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001675601696745262,
+      "loss": 1.2614,
+      "step": 6604
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016754982886729318,
+      "loss": 0.9716,
+      "step": 6605
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000167539486731357,
+      "loss": 1.0277,
+      "step": 6606
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016752914326692117,
+      "loss": 0.9813,
+      "step": 6607
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016751879847418905,
+      "loss": 1.1204,
+      "step": 6608
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016750845235336415,
+      "loss": 1.2392,
+      "step": 6609
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016749810490465002,
+      "loss": 1.0474,
+      "step": 6610
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016748775612825015,
+      "loss": 1.1282,
+      "step": 6611
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016747740602436807,
+      "loss": 1.1579,
+      "step": 6612
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016746705459320745,
+      "loss": 0.9922,
+      "step": 6613
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016745670183497183,
+      "loss": 1.165,
+      "step": 6614
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001674463477498649,
+      "loss": 1.1083,
+      "step": 6615
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001674359923380903,
+      "loss": 1.0754,
+      "step": 6616
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001674256355998517,
+      "loss": 0.8728,
+      "step": 6617
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016741527753535285,
+      "loss": 0.9607,
+      "step": 6618
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016740491814479748,
+      "loss": 1.0869,
+      "step": 6619
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016739455742838935,
+      "loss": 0.988,
+      "step": 6620
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001673841953863323,
+      "loss": 0.9896,
+      "step": 6621
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001673738320188301,
+      "loss": 1.3137,
+      "step": 6622
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001673634673260866,
+      "loss": 0.8684,
+      "step": 6623
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001673531013083057,
+      "loss": 1.2076,
+      "step": 6624
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016734273396569127,
+      "loss": 1.1009,
+      "step": 6625
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016733236529844727,
+      "loss": 1.2405,
+      "step": 6626
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016732199530677762,
+      "loss": 1.128,
+      "step": 6627
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016731162399088632,
+      "loss": 1.1696,
+      "step": 6628
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016730125135097735,
+      "loss": 0.9934,
+      "step": 6629
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016729087738725476,
+      "loss": 1.1867,
+      "step": 6630
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016728050209992258,
+      "loss": 0.9404,
+      "step": 6631
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016727012548918495,
+      "loss": 0.8601,
+      "step": 6632
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016725974755524593,
+      "loss": 1.2201,
+      "step": 6633
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016724936829830964,
+      "loss": 1.1424,
+      "step": 6634
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016723898771858027,
+      "loss": 1.1277,
+      "step": 6635
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000167228605816262,
+      "loss": 0.9382,
+      "step": 6636
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016721822259155902,
+      "loss": 1.2908,
+      "step": 6637
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016720783804467562,
+      "loss": 1.1573,
+      "step": 6638
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000167197452175816,
+      "loss": 1.0847,
+      "step": 6639
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001671870649851845,
+      "loss": 1.1682,
+      "step": 6640
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001671766764729854,
+      "loss": 1.1907,
+      "step": 6641
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016716628663942306,
+      "loss": 1.0794,
+      "step": 6642
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016715589548470185,
+      "loss": 1.0853,
+      "step": 6643
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016714550300902617,
+      "loss": 0.9445,
+      "step": 6644
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001671351092126004,
+      "loss": 0.8966,
+      "step": 6645
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016712471409562903,
+      "loss": 1.0295,
+      "step": 6646
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016711431765831653,
+      "loss": 1.0315,
+      "step": 6647
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016710391990086736,
+      "loss": 1.0924,
+      "step": 6648
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016709352082348606,
+      "loss": 1.0551,
+      "step": 6649
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016708312042637723,
+      "loss": 1.2186,
+      "step": 6650
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016707271870974538,
+      "loss": 0.9894,
+      "step": 6651
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016706231567379512,
+      "loss": 1.0772,
+      "step": 6652
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016705191131873107,
+      "loss": 1.1323,
+      "step": 6653
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016704150564475799,
+      "loss": 1.0636,
+      "step": 6654
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016703109865208037,
+      "loss": 1.3875,
+      "step": 6655
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016702069034090305,
+      "loss": 1.1885,
+      "step": 6656
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001670102807114308,
+      "loss": 1.1055,
+      "step": 6657
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016699986976386822,
+      "loss": 1.0986,
+      "step": 6658
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016698945749842023,
+      "loss": 1.1491,
+      "step": 6659
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016697904391529158,
+      "loss": 0.8781,
+      "step": 6660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001669686290146871,
+      "loss": 1.0362,
+      "step": 6661
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016695821279681165,
+      "loss": 0.9645,
+      "step": 6662
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016694779526187014,
+      "loss": 1.0117,
+      "step": 6663
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001669373764100675,
+      "loss": 0.9294,
+      "step": 6664
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016692695624160862,
+      "loss": 1.0212,
+      "step": 6665
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001669165347566985,
+      "loss": 0.9134,
+      "step": 6666
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016690611195554212,
+      "loss": 1.3922,
+      "step": 6667
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001668956878383445,
+      "loss": 1.0631,
+      "step": 6668
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016688526240531064,
+      "loss": 1.0471,
+      "step": 6669
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016687483565664565,
+      "loss": 1.0923,
+      "step": 6670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016686440759255464,
+      "loss": 0.9838,
+      "step": 6671
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016685397821324268,
+      "loss": 1.1168,
+      "step": 6672
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000166843547518915,
+      "loss": 1.1017,
+      "step": 6673
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016683311550977665,
+      "loss": 0.9792,
+      "step": 6674
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001668226821860329,
+      "loss": 0.9917,
+      "step": 6675
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000166812247547889,
+      "loss": 0.9745,
+      "step": 6676
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016680181159555013,
+      "loss": 1.2032,
+      "step": 6677
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016679137432922163,
+      "loss": 1.1006,
+      "step": 6678
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016678093574910873,
+      "loss": 1.1828,
+      "step": 6679
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016677049585541682,
+      "loss": 1.0962,
+      "step": 6680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016676005464835123,
+      "loss": 1.1108,
+      "step": 6681
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016674961212811732,
+      "loss": 1.1387,
+      "step": 6682
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016673916829492054,
+      "loss": 0.9482,
+      "step": 6683
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016672872314896627,
+      "loss": 1.0458,
+      "step": 6684
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016671827669045998,
+      "loss": 0.8732,
+      "step": 6685
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016670782891960718,
+      "loss": 1.1162,
+      "step": 6686
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016669737983661333,
+      "loss": 0.8777,
+      "step": 6687
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016668692944168397,
+      "loss": 1.0852,
+      "step": 6688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001666764777350247,
+      "loss": 1.1026,
+      "step": 6689
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001666660247168411,
+      "loss": 1.1444,
+      "step": 6690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016665557038733874,
+      "loss": 0.8637,
+      "step": 6691
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016664511474672331,
+      "loss": 0.7635,
+      "step": 6692
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001666346577952004,
+      "loss": 1.0651,
+      "step": 6693
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016662419953297576,
+      "loss": 1.1874,
+      "step": 6694
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016661373996025507,
+      "loss": 1.1692,
+      "step": 6695
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001666032790772441,
+      "loss": 1.1792,
+      "step": 6696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001665928168841486,
+      "loss": 0.9804,
+      "step": 6697
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016658235338117434,
+      "loss": 1.1658,
+      "step": 6698
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016657188856852715,
+      "loss": 0.9537,
+      "step": 6699
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001665614224464129,
+      "loss": 1.0414,
+      "step": 6700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016655095501503738,
+      "loss": 0.9876,
+      "step": 6701
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016654048627460662,
+      "loss": 1.1022,
+      "step": 6702
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016653001622532641,
+      "loss": 0.9884,
+      "step": 6703
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016651954486740278,
+      "loss": 0.8915,
+      "step": 6704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016650907220104163,
+      "loss": 1.051,
+      "step": 6705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016649859822644899,
+      "loss": 1.053,
+      "step": 6706
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001664881229438309,
+      "loss": 0.9738,
+      "step": 6707
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016647764635339341,
+      "loss": 1.0124,
+      "step": 6708
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001664671684553426,
+      "loss": 1.0364,
+      "step": 6709
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016645668924988448,
+      "loss": 0.9836,
+      "step": 6710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016644620873722528,
+      "loss": 0.9435,
+      "step": 6711
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016643572691757113,
+      "loss": 1.01,
+      "step": 6712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016642524379112817,
+      "loss": 1.0538,
+      "step": 6713
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016641475935810264,
+      "loss": 0.8316,
+      "step": 6714
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016640427361870076,
+      "loss": 1.203,
+      "step": 6715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001663937865731288,
+      "loss": 1.0828,
+      "step": 6716
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016638329822159296,
+      "loss": 0.9609,
+      "step": 6717
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016637280856429963,
+      "loss": 0.9556,
+      "step": 6718
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016636231760145512,
+      "loss": 1.3872,
+      "step": 6719
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016635182533326578,
+      "loss": 1.1456,
+      "step": 6720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016634133175993802,
+      "loss": 1.1593,
+      "step": 6721
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016633083688167818,
+      "loss": 0.9982,
+      "step": 6722
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016632034069869275,
+      "loss": 1.0666,
+      "step": 6723
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016630984321118817,
+      "loss": 1.1564,
+      "step": 6724
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016629934441937097,
+      "loss": 1.2837,
+      "step": 6725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001662888443234476,
+      "loss": 1.1264,
+      "step": 6726
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001662783429236246,
+      "loss": 1.1798,
+      "step": 6727
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016626784022010855,
+      "loss": 1.068,
+      "step": 6728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016625733621310608,
+      "loss": 1.1758,
+      "step": 6729
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016624683090282373,
+      "loss": 1.0629,
+      "step": 6730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001662363242894682,
+      "loss": 1.2654,
+      "step": 6731
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001662258163732461,
+      "loss": 1.2405,
+      "step": 6732
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016621530715436414,
+      "loss": 1.1873,
+      "step": 6733
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016620479663302906,
+      "loss": 0.9818,
+      "step": 6734
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016619428480944758,
+      "loss": 1.2444,
+      "step": 6735
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001661837716838265,
+      "loss": 1.0676,
+      "step": 6736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016617325725637257,
+      "loss": 1.2183,
+      "step": 6737
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016616274152729258,
+      "loss": 1.3055,
+      "step": 6738
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016615222449679347,
+      "loss": 1.0498,
+      "step": 6739
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016614170616508202,
+      "loss": 1.1367,
+      "step": 6740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016613118653236518,
+      "loss": 1.1816,
+      "step": 6741
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016612066559884987,
+      "loss": 0.9518,
+      "step": 6742
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016611014336474303,
+      "loss": 1.1933,
+      "step": 6743
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016609961983025156,
+      "loss": 1.0111,
+      "step": 6744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001660890949955826,
+      "loss": 1.1397,
+      "step": 6745
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016607856886094303,
+      "loss": 1.1895,
+      "step": 6746
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016606804142654,
+      "loss": 0.8756,
+      "step": 6747
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016605751269258053,
+      "loss": 0.8077,
+      "step": 6748
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016604698265927173,
+      "loss": 1.1597,
+      "step": 6749
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016603645132682072,
+      "loss": 1.0529,
+      "step": 6750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016602591869543468,
+      "loss": 1.3565,
+      "step": 6751
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016601538476532078,
+      "loss": 1.1075,
+      "step": 6752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016600484953668617,
+      "loss": 1.0968,
+      "step": 6753
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016599431300973815,
+      "loss": 1.0219,
+      "step": 6754
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016598377518468397,
+      "loss": 1.0663,
+      "step": 6755
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001659732360617308,
+      "loss": 1.1074,
+      "step": 6756
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001659626956410861,
+      "loss": 1.0524,
+      "step": 6757
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001659521539229571,
+      "loss": 0.999,
+      "step": 6758
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016594161090755115,
+      "loss": 0.962,
+      "step": 6759
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016593106659507572,
+      "loss": 1.088,
+      "step": 6760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001659205209857381,
+      "loss": 1.2847,
+      "step": 6761
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001659099740797458,
+      "loss": 1.0569,
+      "step": 6762
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001658994258773063,
+      "loss": 0.9956,
+      "step": 6763
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016588887637862694,
+      "loss": 0.8371,
+      "step": 6764
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001658783255839154,
+      "loss": 0.9795,
+      "step": 6765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016586777349337915,
+      "loss": 1.0445,
+      "step": 6766
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001658572201072257,
+      "loss": 1.0577,
+      "step": 6767
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001658466654256627,
+      "loss": 1.1464,
+      "step": 6768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016583610944889774,
+      "loss": 1.1416,
+      "step": 6769
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001658255521771384,
+      "loss": 1.0458,
+      "step": 6770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016581499361059243,
+      "loss": 0.9282,
+      "step": 6771
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016580443374946747,
+      "loss": 0.9288,
+      "step": 6772
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016579387259397127,
+      "loss": 1.006,
+      "step": 6773
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001657833101443115,
+      "loss": 1.111,
+      "step": 6774
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.000165772746400696,
+      "loss": 0.9999,
+      "step": 6775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001657621813633325,
+      "loss": 1.1847,
+      "step": 6776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016575161503242885,
+      "loss": 1.0698,
+      "step": 6777
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016574104740819286,
+      "loss": 1.1942,
+      "step": 6778
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016573047849083238,
+      "loss": 0.9282,
+      "step": 6779
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016571990828055536,
+      "loss": 1.062,
+      "step": 6780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016570933677756967,
+      "loss": 0.9909,
+      "step": 6781
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016569876398208328,
+      "loss": 1.0095,
+      "step": 6782
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016568818989430416,
+      "loss": 1.1322,
+      "step": 6783
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016567761451444024,
+      "loss": 1.3205,
+      "step": 6784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001656670378426996,
+      "loss": 1.0829,
+      "step": 6785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016565645987929028,
+      "loss": 0.9101,
+      "step": 6786
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016564588062442033,
+      "loss": 1.2146,
+      "step": 6787
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016563530007829784,
+      "loss": 1.0723,
+      "step": 6788
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016562471824113092,
+      "loss": 1.1581,
+      "step": 6789
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016561413511312773,
+      "loss": 1.2326,
+      "step": 6790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016560355069449647,
+      "loss": 0.9418,
+      "step": 6791
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001655929649854453,
+      "loss": 1.3289,
+      "step": 6792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016558237798618245,
+      "loss": 1.0747,
+      "step": 6793
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016557178969691609,
+      "loss": 0.9221,
+      "step": 6794
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016556120011785463,
+      "loss": 1.0297,
+      "step": 6795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001655506092492063,
+      "loss": 0.8623,
+      "step": 6796
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001655400170911794,
+      "loss": 1.047,
+      "step": 6797
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001655294236439823,
+      "loss": 1.1243,
+      "step": 6798
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016551882890782337,
+      "loss": 1.2033,
+      "step": 6799
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016550823288291104,
+      "loss": 0.9108,
+      "step": 6800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001654976355694537,
+      "loss": 1.1923,
+      "step": 6801
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016548703696765978,
+      "loss": 0.9626,
+      "step": 6802
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016547643707773776,
+      "loss": 1.1393,
+      "step": 6803
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016546583589989618,
+      "loss": 0.9707,
+      "step": 6804
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016545523343434356,
+      "loss": 0.8395,
+      "step": 6805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016544462968128836,
+      "loss": 0.9989,
+      "step": 6806
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001654340246409393,
+      "loss": 0.8705,
+      "step": 6807
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016542341831350487,
+      "loss": 1.1463,
+      "step": 6808
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016541281069919377,
+      "loss": 1.0916,
+      "step": 6809
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016540220179821457,
+      "loss": 0.989,
+      "step": 6810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016539159161077602,
+      "loss": 0.9684,
+      "step": 6811
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016538098013708677,
+      "loss": 1.248,
+      "step": 6812
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016537036737735564,
+      "loss": 1.0384,
+      "step": 6813
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016535975333179122,
+      "loss": 1.1208,
+      "step": 6814
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016534913800060242,
+      "loss": 1.039,
+      "step": 6815
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016533852138399802,
+      "loss": 1.1884,
+      "step": 6816
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001653279034821868,
+      "loss": 0.9945,
+      "step": 6817
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016531728429537766,
+      "loss": 1.2076,
+      "step": 6818
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001653066638237795,
+      "loss": 1.0505,
+      "step": 6819
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016529604206760116,
+      "loss": 0.9629,
+      "step": 6820
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016528541902705162,
+      "loss": 1.0009,
+      "step": 6821
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001652747947023398,
+      "loss": 1.2979,
+      "step": 6822
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001652641690936747,
+      "loss": 0.7973,
+      "step": 6823
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016525354220126533,
+      "loss": 0.8122,
+      "step": 6824
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001652429140253207,
+      "loss": 1.0641,
+      "step": 6825
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016523228456604986,
+      "loss": 0.9769,
+      "step": 6826
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016522165382366191,
+      "loss": 1.1472,
+      "step": 6827
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000165211021798366,
+      "loss": 1.1586,
+      "step": 6828
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016520038849037117,
+      "loss": 0.967,
+      "step": 6829
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016518975389988666,
+      "loss": 1.0813,
+      "step": 6830
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016517911802712158,
+      "loss": 1.0921,
+      "step": 6831
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001651684808722852,
+      "loss": 1.0567,
+      "step": 6832
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001651578424355867,
+      "loss": 0.9794,
+      "step": 6833
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001651472027172354,
+      "loss": 0.9881,
+      "step": 6834
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016513656171744052,
+      "loss": 0.9988,
+      "step": 6835
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001651259194364114,
+      "loss": 1.0997,
+      "step": 6836
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016511527587435737,
+      "loss": 1.1479,
+      "step": 6837
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001651046310314878,
+      "loss": 0.7897,
+      "step": 6838
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016509398490801203,
+      "loss": 1.0924,
+      "step": 6839
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001650833375041395,
+      "loss": 1.0075,
+      "step": 6840
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016507268882007966,
+      "loss": 0.9682,
+      "step": 6841
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016506203885604195,
+      "loss": 1.1156,
+      "step": 6842
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016505138761223584,
+      "loss": 1.1853,
+      "step": 6843
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016504073508887088,
+      "loss": 1.0785,
+      "step": 6844
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016503008128615655,
+      "loss": 1.0986,
+      "step": 6845
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016501942620430246,
+      "loss": 1.1878,
+      "step": 6846
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001650087698435182,
+      "loss": 1.136,
+      "step": 6847
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001649981122040133,
+      "loss": 0.995,
+      "step": 6848
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001649874532859975,
+      "loss": 1.0539,
+      "step": 6849
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016497679308968038,
+      "loss": 1.1035,
+      "step": 6850
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001649661316152717,
+      "loss": 0.9529,
+      "step": 6851
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016495546886298112,
+      "loss": 1.2711,
+      "step": 6852
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016494480483301836,
+      "loss": 1.0453,
+      "step": 6853
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016493413952559322,
+      "loss": 1.0486,
+      "step": 6854
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016492347294091547,
+      "loss": 1.1613,
+      "step": 6855
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016491280507919493,
+      "loss": 1.0548,
+      "step": 6856
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016490213594064142,
+      "loss": 1.1672,
+      "step": 6857
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016489146552546485,
+      "loss": 1.0753,
+      "step": 6858
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016488079383387505,
+      "loss": 1.1289,
+      "step": 6859
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001648701208660819,
+      "loss": 0.8561,
+      "step": 6860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016485944662229546,
+      "loss": 1.2813,
+      "step": 6861
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016484877110272558,
+      "loss": 0.9081,
+      "step": 6862
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016483809430758227,
+      "loss": 1.0287,
+      "step": 6863
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001648274162370756,
+      "loss": 1.1204,
+      "step": 6864
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016481673689141556,
+      "loss": 1.1544,
+      "step": 6865
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001648060562708122,
+      "loss": 0.9183,
+      "step": 6866
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016479537437547563,
+      "loss": 1.0216,
+      "step": 6867
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000164784691205616,
+      "loss": 1.0119,
+      "step": 6868
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016477400676144334,
+      "loss": 1.1335,
+      "step": 6869
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016476332104316792,
+      "loss": 0.8238,
+      "step": 6870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001647526340509999,
+      "loss": 1.0391,
+      "step": 6871
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016474194578514943,
+      "loss": 1.0512,
+      "step": 6872
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016473125624582685,
+      "loss": 1.0714,
+      "step": 6873
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016472056543324234,
+      "loss": 1.3355,
+      "step": 6874
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016470987334760623,
+      "loss": 1.02,
+      "step": 6875
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001646991799891288,
+      "loss": 1.0429,
+      "step": 6876
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016468848535802045,
+      "loss": 0.9682,
+      "step": 6877
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016467778945449148,
+      "loss": 1.0333,
+      "step": 6878
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016466709227875233,
+      "loss": 1.026,
+      "step": 6879
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016465639383101332,
+      "loss": 1.0554,
+      "step": 6880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000164645694111485,
+      "loss": 0.8989,
+      "step": 6881
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001646349931203778,
+      "loss": 0.9199,
+      "step": 6882
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016462429085790215,
+      "loss": 1.0974,
+      "step": 6883
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016461358732426862,
+      "loss": 1.1348,
+      "step": 6884
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016460288251968775,
+      "loss": 0.9504,
+      "step": 6885
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016459217644437007,
+      "loss": 1.0538,
+      "step": 6886
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001645814690985262,
+      "loss": 1.2328,
+      "step": 6887
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016457076048236675,
+      "loss": 1.0528,
+      "step": 6888
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016456005059610232,
+      "loss": 1.0872,
+      "step": 6889
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016454933943994363,
+      "loss": 1.1675,
+      "step": 6890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001645386270141013,
+      "loss": 1.1547,
+      "step": 6891
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001645279133187861,
+      "loss": 1.2158,
+      "step": 6892
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016451719835420877,
+      "loss": 0.9425,
+      "step": 6893
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016450648212058003,
+      "loss": 0.8548,
+      "step": 6894
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001644957646181107,
+      "loss": 1.1398,
+      "step": 6895
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001644850458470116,
+      "loss": 0.9798,
+      "step": 6896
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016447432580749354,
+      "loss": 1.0999,
+      "step": 6897
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016446360449976738,
+      "loss": 1.0648,
+      "step": 6898
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000164452881924044,
+      "loss": 1.1054,
+      "step": 6899
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016444215808053442,
+      "loss": 1.1524,
+      "step": 6900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016443143296944945,
+      "loss": 1.2108,
+      "step": 6901
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001644207065910001,
+      "loss": 1.1879,
+      "step": 6902
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001644099789453973,
+      "loss": 0.889,
+      "step": 6903
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001643992500328522,
+      "loss": 1.253,
+      "step": 6904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001643885198535757,
+      "loss": 1.1945,
+      "step": 6905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016437778840777892,
+      "loss": 0.994,
+      "step": 6906
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016436705569567293,
+      "loss": 1.049,
+      "step": 6907
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001643563217174689,
+      "loss": 1.2261,
+      "step": 6908
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001643455864733779,
+      "loss": 0.9536,
+      "step": 6909
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016433484996361113,
+      "loss": 0.9549,
+      "step": 6910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016432411218837974,
+      "loss": 1.0059,
+      "step": 6911
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016431337314789496,
+      "loss": 0.9237,
+      "step": 6912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016430263284236801,
+      "loss": 1.1815,
+      "step": 6913
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016429189127201022,
+      "loss": 0.9778,
+      "step": 6914
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001642811484370328,
+      "loss": 0.9056,
+      "step": 6915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016427040433764707,
+      "loss": 0.7968,
+      "step": 6916
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001642596589740644,
+      "loss": 1.0977,
+      "step": 6917
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016424891234649618,
+      "loss": 1.1645,
+      "step": 6918
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001642381644551537,
+      "loss": 1.078,
+      "step": 6919
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016422741530024843,
+      "loss": 0.9193,
+      "step": 6920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001642166648819918,
+      "loss": 1.0529,
+      "step": 6921
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016420591320059525,
+      "loss": 1.1679,
+      "step": 6922
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001641951602562703,
+      "loss": 1.0096,
+      "step": 6923
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016418440604922847,
+      "loss": 1.1104,
+      "step": 6924
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016417365057968123,
+      "loss": 1.148,
+      "step": 6925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001641628938478402,
+      "loss": 1.0314,
+      "step": 6926
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016415213585391694,
+      "loss": 1.0521,
+      "step": 6927
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016414137659812305,
+      "loss": 1.0482,
+      "step": 6928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016413061608067017,
+      "loss": 1.0223,
+      "step": 6929
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016411985430176998,
+      "loss": 0.8566,
+      "step": 6930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016410909126163417,
+      "loss": 0.8591,
+      "step": 6931
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016409832696047437,
+      "loss": 0.9644,
+      "step": 6932
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001640875613985024,
+      "loss": 1.1358,
+      "step": 6933
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016407679457593001,
+      "loss": 0.9233,
+      "step": 6934
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016406602649296896,
+      "loss": 1.0778,
+      "step": 6935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016405525714983105,
+      "loss": 1.0559,
+      "step": 6936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001640444865467281,
+      "loss": 0.9951,
+      "step": 6937
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016403371468387204,
+      "loss": 1.0784,
+      "step": 6938
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016402294156147467,
+      "loss": 1.1569,
+      "step": 6939
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001640121671797479,
+      "loss": 1.1229,
+      "step": 6940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016400139153890377,
+      "loss": 1.0739,
+      "step": 6941
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001639906146391541,
+      "loss": 1.1888,
+      "step": 6942
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016397983648071095,
+      "loss": 1.1333,
+      "step": 6943
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001639690570637863,
+      "loss": 1.0505,
+      "step": 6944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016395827638859218,
+      "loss": 0.9959,
+      "step": 6945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016394749445534067,
+      "loss": 0.9945,
+      "step": 6946
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016393671126424382,
+      "loss": 1.0102,
+      "step": 6947
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016392592681551373,
+      "loss": 1.0324,
+      "step": 6948
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016391514110936255,
+      "loss": 1.1331,
+      "step": 6949
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016390435414600244,
+      "loss": 0.9398,
+      "step": 6950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001638935659256456,
+      "loss": 0.8464,
+      "step": 6951
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016388277644850417,
+      "loss": 1.2181,
+      "step": 6952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016387198571479042,
+      "loss": 0.9974,
+      "step": 6953
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016386119372471662,
+      "loss": 1.0974,
+      "step": 6954
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000163850400478495,
+      "loss": 1.0152,
+      "step": 6955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001638396059763379,
+      "loss": 1.0191,
+      "step": 6956
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016382881021845762,
+      "loss": 0.9825,
+      "step": 6957
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016381801320506653,
+      "loss": 0.8815,
+      "step": 6958
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016380721493637702,
+      "loss": 1.0855,
+      "step": 6959
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016379641541260144,
+      "loss": 1.0995,
+      "step": 6960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001637856146339523,
+      "loss": 1.0172,
+      "step": 6961
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000163774812600642,
+      "loss": 1.0932,
+      "step": 6962
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016376400931288298,
+      "loss": 0.9878,
+      "step": 6963
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001637532047708878,
+      "loss": 1.0814,
+      "step": 6964
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000163742398974869,
+      "loss": 1.2458,
+      "step": 6965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016373159192503904,
+      "loss": 1.0629,
+      "step": 6966
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001637207836216106,
+      "loss": 0.793,
+      "step": 6967
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001637099740647962,
+      "loss": 0.9551,
+      "step": 6968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016369916325480844,
+      "loss": 0.994,
+      "step": 6969
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001636883511918601,
+      "loss": 1.2044,
+      "step": 6970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016367753787616373,
+      "loss": 1.1888,
+      "step": 6971
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016366672330793208,
+      "loss": 0.8744,
+      "step": 6972
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016365590748737785,
+      "loss": 0.9745,
+      "step": 6973
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016364509041471382,
+      "loss": 1.175,
+      "step": 6974
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016363427209015275,
+      "loss": 1.1389,
+      "step": 6975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001636234525139074,
+      "loss": 1.0735,
+      "step": 6976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016361263168619062,
+      "loss": 0.9527,
+      "step": 6977
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016360180960721526,
+      "loss": 0.9892,
+      "step": 6978
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001635909862771942,
+      "loss": 1.1603,
+      "step": 6979
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001635801616963403,
+      "loss": 1.0423,
+      "step": 6980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016356933586486648,
+      "loss": 0.9257,
+      "step": 6981
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016355850878298572,
+      "loss": 1.0434,
+      "step": 6982
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016354768045091097,
+      "loss": 0.9061,
+      "step": 6983
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016353685086885523,
+      "loss": 1.1387,
+      "step": 6984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001635260200370315,
+      "loss": 1.1322,
+      "step": 6985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016351518795565281,
+      "loss": 1.0143,
+      "step": 6986
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016350435462493228,
+      "loss": 1.0082,
+      "step": 6987
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000163493520045083,
+      "loss": 1.2397,
+      "step": 6988
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016348268421631797,
+      "loss": 0.9556,
+      "step": 6989
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016347184713885046,
+      "loss": 1.0226,
+      "step": 6990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001634610088128936,
+      "loss": 0.9195,
+      "step": 6991
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016345016923866054,
+      "loss": 1.0845,
+      "step": 6992
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016343932841636456,
+      "loss": 0.9254,
+      "step": 6993
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016342848634621885,
+      "loss": 0.8628,
+      "step": 6994
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016341764302843666,
+      "loss": 0.9921,
+      "step": 6995
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016340679846323135,
+      "loss": 1.0126,
+      "step": 6996
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001633959526508162,
+      "loss": 1.1637,
+      "step": 6997
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016338510559140446,
+      "loss": 1.0104,
+      "step": 6998
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016337425728520961,
+      "loss": 1.1025,
+      "step": 6999
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016336340773244498,
+      "loss": 1.0623,
+      "step": 7000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.000163352556933324,
+      "loss": 1.1031,
+      "step": 7001
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001633417048880601,
+      "loss": 1.1452,
+      "step": 7002
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016333085159686672,
+      "loss": 0.875,
+      "step": 7003
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016331999705995734,
+      "loss": 1.0471,
+      "step": 7004
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016330914127754553,
+      "loss": 1.178,
+      "step": 7005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016329828424984478,
+      "loss": 0.9915,
+      "step": 7006
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001632874259770686,
+      "loss": 0.9819,
+      "step": 7007
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016327656645943065,
+      "loss": 0.8797,
+      "step": 7008
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001632657056971445,
+      "loss": 1.1464,
+      "step": 7009
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001632548436904238,
+      "loss": 1.1339,
+      "step": 7010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001632439804394822,
+      "loss": 0.958,
+      "step": 7011
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016323311594453334,
+      "loss": 1.0547,
+      "step": 7012
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016322225020579099,
+      "loss": 1.0366,
+      "step": 7013
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016321138322346882,
+      "loss": 0.8503,
+      "step": 7014
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016320051499778062,
+      "loss": 1.1983,
+      "step": 7015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016318964552894017,
+      "loss": 1.2385,
+      "step": 7016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001631787748171612,
+      "loss": 1.0546,
+      "step": 7017
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016316790286265763,
+      "loss": 1.232,
+      "step": 7018
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001631570296656433,
+      "loss": 0.9328,
+      "step": 7019
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016314615522633206,
+      "loss": 1.1676,
+      "step": 7020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016313527954493778,
+      "loss": 0.9363,
+      "step": 7021
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016312440262167442,
+      "loss": 0.9048,
+      "step": 7022
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016311352445675596,
+      "loss": 1.1551,
+      "step": 7023
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001631026450503963,
+      "loss": 1.0713,
+      "step": 7024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016309176440280947,
+      "loss": 1.0984,
+      "step": 7025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016308088251420951,
+      "loss": 0.853,
+      "step": 7026
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016306999938481046,
+      "loss": 0.8826,
+      "step": 7027
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001630591150148264,
+      "loss": 1.1411,
+      "step": 7028
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016304822940447139,
+      "loss": 1.0935,
+      "step": 7029
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016303734255395958,
+      "loss": 1.0553,
+      "step": 7030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016302645446350508,
+      "loss": 1.0196,
+      "step": 7031
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016301556513332208,
+      "loss": 1.1935,
+      "step": 7032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001630046745636248,
+      "loss": 1.0997,
+      "step": 7033
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016299378275462742,
+      "loss": 1.1726,
+      "step": 7034
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001629828897065442,
+      "loss": 0.8965,
+      "step": 7035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016297199541958938,
+      "loss": 1.0166,
+      "step": 7036
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001629610998939773,
+      "loss": 0.7803,
+      "step": 7037
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001629502031299222,
+      "loss": 1.0157,
+      "step": 7038
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016293930512763848,
+      "loss": 1.1451,
+      "step": 7039
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016292840588734054,
+      "loss": 1.0984,
+      "step": 7040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016291750540924264,
+      "loss": 1.2327,
+      "step": 7041
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.0001629066036935593,
+      "loss": 0.9837,
+      "step": 7042
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00016289570074050493,
+      "loss": 1.1308,
+      "step": 7043
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016288479655029398,
+      "loss": 1.1334,
+      "step": 7044
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016287389112314092,
+      "loss": 1.0097,
+      "step": 7045
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001628629844592603,
+      "loss": 0.9787,
+      "step": 7046
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016285207655886662,
+      "loss": 1.0227,
+      "step": 7047
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016284116742217445,
+      "loss": 0.983,
+      "step": 7048
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016283025704939838,
+      "loss": 1.0964,
+      "step": 7049
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000162819345440753,
+      "loss": 1.0322,
+      "step": 7050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016280843259645295,
+      "loss": 1.0273,
+      "step": 7051
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001627975185167129,
+      "loss": 1.0104,
+      "step": 7052
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016278660320174747,
+      "loss": 1.0341,
+      "step": 7053
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016277568665177145,
+      "loss": 1.0887,
+      "step": 7054
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016276476886699953,
+      "loss": 1.1221,
+      "step": 7055
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016275384984764643,
+      "loss": 1.0747,
+      "step": 7056
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000162742929593927,
+      "loss": 1.1017,
+      "step": 7057
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016273200810605593,
+      "loss": 0.9566,
+      "step": 7058
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001627210853842482,
+      "loss": 1.0498,
+      "step": 7059
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001627101614287185,
+      "loss": 1.0269,
+      "step": 7060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001626992362396818,
+      "loss": 1.1275,
+      "step": 7061
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016268830981735298,
+      "loss": 0.8682,
+      "step": 7062
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016267738216194696,
+      "loss": 1.2916,
+      "step": 7063
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016266645327367866,
+      "loss": 1.1528,
+      "step": 7064
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001626555231527631,
+      "loss": 1.1666,
+      "step": 7065
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016264459179941528,
+      "loss": 1.2002,
+      "step": 7066
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016263365921385015,
+      "loss": 1.1855,
+      "step": 7067
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016262272539628277,
+      "loss": 1.1094,
+      "step": 7068
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016261179034692826,
+      "loss": 1.1106,
+      "step": 7069
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016260085406600167,
+      "loss": 1.2675,
+      "step": 7070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016258991655371813,
+      "loss": 0.9683,
+      "step": 7071
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001625789778102928,
+      "loss": 1.0301,
+      "step": 7072
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001625680378359408,
+      "loss": 1.0486,
+      "step": 7073
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016255709663087737,
+      "loss": 1.0558,
+      "step": 7074
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016254615419531766,
+      "loss": 0.977,
+      "step": 7075
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000162535210529477,
+      "loss": 0.8745,
+      "step": 7076
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016252426563357055,
+      "loss": 1.174,
+      "step": 7077
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016251331950781363,
+      "loss": 1.0744,
+      "step": 7078
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016250237215242156,
+      "loss": 1.2252,
+      "step": 7079
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001624914235676097,
+      "loss": 0.9575,
+      "step": 7080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016248047375359338,
+      "loss": 1.0903,
+      "step": 7081
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000162469522710588,
+      "loss": 1.182,
+      "step": 7082
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016245857043880894,
+      "loss": 1.1297,
+      "step": 7083
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001624476169384716,
+      "loss": 1.1655,
+      "step": 7084
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016243666220979157,
+      "loss": 0.8721,
+      "step": 7085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016242570625298419,
+      "loss": 1.1627,
+      "step": 7086
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016241474906826498,
+      "loss": 0.9376,
+      "step": 7087
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016240379065584953,
+      "loss": 0.9535,
+      "step": 7088
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016239283101595337,
+      "loss": 0.9763,
+      "step": 7089
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016238187014879207,
+      "loss": 0.9574,
+      "step": 7090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016237090805458119,
+      "loss": 1.2438,
+      "step": 7091
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001623599447335364,
+      "loss": 1.0313,
+      "step": 7092
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016234898018587337,
+      "loss": 1.0536,
+      "step": 7093
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001623380144118077,
+      "loss": 1.0489,
+      "step": 7094
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016232704741155516,
+      "loss": 0.9142,
+      "step": 7095
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016231607918533144,
+      "loss": 1.1536,
+      "step": 7096
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016230510973335225,
+      "loss": 1.2708,
+      "step": 7097
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016229413905583342,
+      "loss": 0.9709,
+      "step": 7098
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001622831671529907,
+      "loss": 1.1986,
+      "step": 7099
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016227219402503994,
+      "loss": 1.2334,
+      "step": 7100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001622612196721969,
+      "loss": 1.0113,
+      "step": 7101
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016225024409467756,
+      "loss": 1.2499,
+      "step": 7102
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016223926729269777,
+      "loss": 0.8304,
+      "step": 7103
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001622282892664734,
+      "loss": 0.9816,
+      "step": 7104
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001622173100162204,
+      "loss": 1.2019,
+      "step": 7105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001622063295421548,
+      "loss": 0.9744,
+      "step": 7106
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001621953478444925,
+      "loss": 1.0156,
+      "step": 7107
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016218436492344956,
+      "loss": 1.1216,
+      "step": 7108
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000162173380779242,
+      "loss": 1.0248,
+      "step": 7109
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001621623954120859,
+      "loss": 1.0743,
+      "step": 7110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016215140882219726,
+      "loss": 1.1355,
+      "step": 7111
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016214042100979228,
+      "loss": 0.9984,
+      "step": 7112
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016212943197508705,
+      "loss": 0.9559,
+      "step": 7113
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001621184417182977,
+      "loss": 1.1479,
+      "step": 7114
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016210745023964048,
+      "loss": 1.1351,
+      "step": 7115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001620964575393315,
+      "loss": 0.9945,
+      "step": 7116
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016208546361758705,
+      "loss": 1.1555,
+      "step": 7117
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016207446847462339,
+      "loss": 0.9212,
+      "step": 7118
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016206347211065674,
+      "loss": 1.0258,
+      "step": 7119
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016205247452590345,
+      "loss": 1.2403,
+      "step": 7120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001620414757205798,
+      "loss": 1.2359,
+      "step": 7121
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016203047569490216,
+      "loss": 1.0534,
+      "step": 7122
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016201947444908688,
+      "loss": 1.0221,
+      "step": 7123
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016200847198335038,
+      "loss": 1.1491,
+      "step": 7124
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016199746829790906,
+      "loss": 1.188,
+      "step": 7125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016198646339297937,
+      "loss": 0.9553,
+      "step": 7126
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001619754572687778,
+      "loss": 1.1621,
+      "step": 7127
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016196444992552081,
+      "loss": 1.2513,
+      "step": 7128
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001619534413634249,
+      "loss": 0.9419,
+      "step": 7129
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016194243158270663,
+      "loss": 1.1445,
+      "step": 7130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016193142058358256,
+      "loss": 1.0891,
+      "step": 7131
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001619204083662693,
+      "loss": 0.9983,
+      "step": 7132
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016190939493098344,
+      "loss": 1.0626,
+      "step": 7133
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016189838027794154,
+      "loss": 1.0645,
+      "step": 7134
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001618873644073604,
+      "loss": 0.9872,
+      "step": 7135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001618763473194566,
+      "loss": 0.9615,
+      "step": 7136
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016186532901444687,
+      "loss": 1.1478,
+      "step": 7137
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016185430949254795,
+      "loss": 0.9017,
+      "step": 7138
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001618432887539766,
+      "loss": 0.8848,
+      "step": 7139
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001618322667989496,
+      "loss": 1.0869,
+      "step": 7140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016182124362768368,
+      "loss": 1.4244,
+      "step": 7141
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016181021924039577,
+      "loss": 1.0872,
+      "step": 7142
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016179919363730266,
+      "loss": 1.0746,
+      "step": 7143
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016178816681862123,
+      "loss": 0.9405,
+      "step": 7144
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016177713878456838,
+      "loss": 0.8621,
+      "step": 7145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016176610953536107,
+      "loss": 1.3321,
+      "step": 7146
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016175507907121615,
+      "loss": 1.0201,
+      "step": 7147
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001617440473923507,
+      "loss": 1.0388,
+      "step": 7148
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016173301449898166,
+      "loss": 0.9874,
+      "step": 7149
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016172198039132602,
+      "loss": 1.2485,
+      "step": 7150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016171094506960086,
+      "loss": 0.8963,
+      "step": 7151
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016169990853402323,
+      "loss": 1.1481,
+      "step": 7152
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016168887078481023,
+      "loss": 1.1133,
+      "step": 7153
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016167783182217894,
+      "loss": 1.1349,
+      "step": 7154
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016166679164634655,
+      "loss": 1.1258,
+      "step": 7155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016165575025753017,
+      "loss": 1.0301,
+      "step": 7156
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016164470765594698,
+      "loss": 0.8259,
+      "step": 7157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016163366384181426,
+      "loss": 1.035,
+      "step": 7158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016162261881534916,
+      "loss": 1.143,
+      "step": 7159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016161157257676897,
+      "loss": 0.9007,
+      "step": 7160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000161600525126291,
+      "loss": 1.0591,
+      "step": 7161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016158947646413249,
+      "loss": 0.9739,
+      "step": 7162
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016157842659051078,
+      "loss": 1.2406,
+      "step": 7163
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016156737550564328,
+      "loss": 0.8153,
+      "step": 7164
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016155632320974728,
+      "loss": 1.0209,
+      "step": 7165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016154526970304025,
+      "loss": 1.0586,
+      "step": 7166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016153421498573954,
+      "loss": 1.0823,
+      "step": 7167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016152315905806268,
+      "loss": 1.1449,
+      "step": 7168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001615121019202271,
+      "loss": 1.0255,
+      "step": 7169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016150104357245027,
+      "loss": 0.89,
+      "step": 7170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016148998401494975,
+      "loss": 0.941,
+      "step": 7171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016147892324794305,
+      "loss": 0.8311,
+      "step": 7172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016146786127164771,
+      "loss": 1.1927,
+      "step": 7173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001614567980862814,
+      "loss": 1.0789,
+      "step": 7174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001614457336920617,
+      "loss": 1.2908,
+      "step": 7175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016143466808920618,
+      "loss": 0.9563,
+      "step": 7176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016142360127793256,
+      "loss": 0.9136,
+      "step": 7177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016141253325845854,
+      "loss": 0.9206,
+      "step": 7178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001614014640310018,
+      "loss": 1.1533,
+      "step": 7179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016139039359578007,
+      "loss": 1.0556,
+      "step": 7180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016137932195301114,
+      "loss": 1.1195,
+      "step": 7181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016136824910291273,
+      "loss": 0.9165,
+      "step": 7182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016135717504570267,
+      "loss": 1.2434,
+      "step": 7183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001613460997815988,
+      "loss": 1.0043,
+      "step": 7184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016133502331081893,
+      "loss": 1.1784,
+      "step": 7185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000161323945633581,
+      "loss": 1.0733,
+      "step": 7186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016131286675010283,
+      "loss": 1.018,
+      "step": 7187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016130178666060241,
+      "loss": 0.8699,
+      "step": 7188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016129070536529766,
+      "loss": 0.9193,
+      "step": 7189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016127962286440655,
+      "loss": 1.0839,
+      "step": 7190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016126853915814703,
+      "loss": 1.3155,
+      "step": 7191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001612574542467372,
+      "loss": 0.9362,
+      "step": 7192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016124636813039502,
+      "loss": 1.0877,
+      "step": 7193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016123528080933862,
+      "loss": 0.9963,
+      "step": 7194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000161224192283786,
+      "loss": 1.0218,
+      "step": 7195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016121310255395538,
+      "loss": 1.1028,
+      "step": 7196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016120201162006482,
+      "loss": 1.0657,
+      "step": 7197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001611909194823325,
+      "loss": 1.1628,
+      "step": 7198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001611798261409766,
+      "loss": 1.0845,
+      "step": 7199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016116873159621533,
+      "loss": 1.1761,
+      "step": 7200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016115763584826688,
+      "loss": 1.0852,
+      "step": 7201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001611465388973496,
+      "loss": 1.1166,
+      "step": 7202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016113544074368164,
+      "loss": 1.265,
+      "step": 7203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016112434138748142,
+      "loss": 0.8902,
+      "step": 7204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001611132408289672,
+      "loss": 0.9812,
+      "step": 7205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016110213906835732,
+      "loss": 1.1362,
+      "step": 7206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016109103610587017,
+      "loss": 1.0586,
+      "step": 7207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016107993194172416,
+      "loss": 0.8144,
+      "step": 7208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016106882657613766,
+      "loss": 1.0767,
+      "step": 7209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001610577200093292,
+      "loss": 1.0445,
+      "step": 7210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016104661224151714,
+      "loss": 1.0101,
+      "step": 7211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016103550327292002,
+      "loss": 1.1143,
+      "step": 7212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016102439310375637,
+      "loss": 0.8865,
+      "step": 7213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016101328173424474,
+      "loss": 1.2658,
+      "step": 7214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016100216916460362,
+      "loss": 1.1129,
+      "step": 7215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016099105539505161,
+      "loss": 0.9803,
+      "step": 7216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001609799404258074,
+      "loss": 1.1901,
+      "step": 7217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016096882425708952,
+      "loss": 1.0667,
+      "step": 7218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001609577068891167,
+      "loss": 1.0687,
+      "step": 7219
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016094658832210756,
+      "loss": 1.1219,
+      "step": 7220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016093546855628084,
+      "loss": 1.2281,
+      "step": 7221
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016092434759185523,
+      "loss": 1.117,
+      "step": 7222
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016091322542904953,
+      "loss": 1.0218,
+      "step": 7223
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016090210206808246,
+      "loss": 0.9374,
+      "step": 7224
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016089097750917288,
+      "loss": 1.0923,
+      "step": 7225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016087985175253951,
+      "loss": 0.8648,
+      "step": 7226
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016086872479840133,
+      "loss": 0.8765,
+      "step": 7227
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016085759664697706,
+      "loss": 1.0014,
+      "step": 7228
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016084646729848568,
+      "loss": 1.0842,
+      "step": 7229
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001608353367531461,
+      "loss": 1.3188,
+      "step": 7230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016082420501117723,
+      "loss": 1.1791,
+      "step": 7231
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016081307207279805,
+      "loss": 1.1738,
+      "step": 7232
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016080193793822754,
+      "loss": 1.1359,
+      "step": 7233
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001607908026076847,
+      "loss": 1.0835,
+      "step": 7234
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016077966608138859,
+      "loss": 0.9145,
+      "step": 7235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001607685283595582,
+      "loss": 1.0134,
+      "step": 7236
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016075738944241268,
+      "loss": 1.1283,
+      "step": 7237
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016074624933017112,
+      "loss": 0.9207,
+      "step": 7238
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001607351080230526,
+      "loss": 0.9082,
+      "step": 7239
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001607239655212763,
+      "loss": 0.8974,
+      "step": 7240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001607128218250614,
+      "loss": 1.2092,
+      "step": 7241
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001607016769346271,
+      "loss": 1.1151,
+      "step": 7242
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016069053085019256,
+      "loss": 1.124,
+      "step": 7243
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016067938357197713,
+      "loss": 0.9122,
+      "step": 7244
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016066823510019998,
+      "loss": 0.9498,
+      "step": 7245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016065708543508045,
+      "loss": 1.1575,
+      "step": 7246
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016064593457683783,
+      "loss": 1.1182,
+      "step": 7247
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016063478252569148,
+      "loss": 1.2143,
+      "step": 7248
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016062362928186073,
+      "loss": 1.0052,
+      "step": 7249
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000160612474845565,
+      "loss": 0.9294,
+      "step": 7250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016060131921702368,
+      "loss": 1.0045,
+      "step": 7251
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016059016239645618,
+      "loss": 1.1409,
+      "step": 7252
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000160579004384082,
+      "loss": 0.981,
+      "step": 7253
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016056784518012055,
+      "loss": 1.145,
+      "step": 7254
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001605566847847914,
+      "loss": 0.9688,
+      "step": 7255
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016054552319831407,
+      "loss": 1.0656,
+      "step": 7256
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001605343604209081,
+      "loss": 1.1316,
+      "step": 7257
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016052319645279298,
+      "loss": 0.9123,
+      "step": 7258
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016051203129418843,
+      "loss": 1.0221,
+      "step": 7259
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000160500864945314,
+      "loss": 0.9903,
+      "step": 7260
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016048969740638933,
+      "loss": 0.9722,
+      "step": 7261
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016047852867763412,
+      "loss": 1.0589,
+      "step": 7262
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.000160467358759268,
+      "loss": 1.3431,
+      "step": 7263
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016045618765151074,
+      "loss": 1.1365,
+      "step": 7264
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016044501535458205,
+      "loss": 0.965,
+      "step": 7265
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.0001604338418687017,
+      "loss": 1.2337,
+      "step": 7266
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016042266719408944,
+      "loss": 0.9212,
+      "step": 7267
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016041149133096512,
+      "loss": 1.1032,
+      "step": 7268
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016040031427954853,
+      "loss": 1.0901,
+      "step": 7269
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016038913604005957,
+      "loss": 1.1155,
+      "step": 7270
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016037795661271812,
+      "loss": 1.157,
+      "step": 7271
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016036677599774398,
+      "loss": 1.0356,
+      "step": 7272
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016035559419535716,
+      "loss": 0.9264,
+      "step": 7273
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00016034441120577758,
+      "loss": 1.1739,
+      "step": 7274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016033322702922522,
+      "loss": 1.0077,
+      "step": 7275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016032204166592008,
+      "loss": 0.8691,
+      "step": 7276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016031085511608213,
+      "loss": 0.9979,
+      "step": 7277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016029966737993146,
+      "loss": 1.0818,
+      "step": 7278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001602884784576881,
+      "loss": 1.087,
+      "step": 7279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016027728834957218,
+      "loss": 1.0407,
+      "step": 7280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016026609705580374,
+      "loss": 0.9965,
+      "step": 7281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016025490457660294,
+      "loss": 1.2279,
+      "step": 7282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016024371091219,
+      "loss": 1.0055,
+      "step": 7283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016023251606278501,
+      "loss": 0.9575,
+      "step": 7284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016022132002860824,
+      "loss": 0.9619,
+      "step": 7285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016021012280987985,
+      "loss": 1.1174,
+      "step": 7286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016019892440682015,
+      "loss": 1.0367,
+      "step": 7287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016018772481964934,
+      "loss": 1.1266,
+      "step": 7288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016017652404858777,
+      "loss": 1.0963,
+      "step": 7289
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016016532209385577,
+      "loss": 0.8922,
+      "step": 7290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016015411895567368,
+      "loss": 1.0959,
+      "step": 7291
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001601429146342618,
+      "loss": 1.0665,
+      "step": 7292
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016013170912984058,
+      "loss": 0.9607,
+      "step": 7293
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016012050244263045,
+      "loss": 1.0005,
+      "step": 7294
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016010929457285178,
+      "loss": 0.9571,
+      "step": 7295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016009808552072506,
+      "loss": 1.1615,
+      "step": 7296
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016008687528647078,
+      "loss": 1.0236,
+      "step": 7297
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016007566387030942,
+      "loss": 1.1788,
+      "step": 7298
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016006445127246156,
+      "loss": 1.32,
+      "step": 7299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001600532374931477,
+      "loss": 1.1693,
+      "step": 7300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016004202253258842,
+      "loss": 1.0113,
+      "step": 7301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016003080639100432,
+      "loss": 1.1399,
+      "step": 7302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016001958906861608,
+      "loss": 0.9202,
+      "step": 7303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00016000837056564423,
+      "loss": 0.9939,
+      "step": 7304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015999715088230957,
+      "loss": 1.0189,
+      "step": 7305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015998593001883266,
+      "loss": 0.9891,
+      "step": 7306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015997470797543434,
+      "loss": 0.8037,
+      "step": 7307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015996348475233525,
+      "loss": 1.0164,
+      "step": 7308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015995226034975617,
+      "loss": 0.9022,
+      "step": 7309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015994103476791793,
+      "loss": 1.1819,
+      "step": 7310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001599298080070413,
+      "loss": 0.9746,
+      "step": 7311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001599185800673471,
+      "loss": 0.8786,
+      "step": 7312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001599073509490562,
+      "loss": 1.125,
+      "step": 7313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001598961206523895,
+      "loss": 1.0042,
+      "step": 7314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015988488917756784,
+      "loss": 0.9174,
+      "step": 7315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015987365652481224,
+      "loss": 1.1737,
+      "step": 7316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015986242269434354,
+      "loss": 1.1459,
+      "step": 7317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015985118768638274,
+      "loss": 1.0819,
+      "step": 7318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015983995150115085,
+      "loss": 1.1958,
+      "step": 7319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001598287141388689,
+      "loss": 1.0552,
+      "step": 7320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001598174755997579,
+      "loss": 0.9185,
+      "step": 7321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001598062358840389,
+      "loss": 0.875,
+      "step": 7322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015979499499193305,
+      "loss": 1.0728,
+      "step": 7323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001597837529236614,
+      "loss": 1.0697,
+      "step": 7324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015977250967944508,
+      "loss": 1.2617,
+      "step": 7325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015976126525950524,
+      "loss": 1.2163,
+      "step": 7326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015975001966406314,
+      "loss": 0.8857,
+      "step": 7327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015973877289333989,
+      "loss": 0.9091,
+      "step": 7328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015972752494755672,
+      "loss": 0.9881,
+      "step": 7329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001597162758269349,
+      "loss": 0.9652,
+      "step": 7330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015970502553169574,
+      "loss": 1.0218,
+      "step": 7331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015969377406206044,
+      "loss": 1.1151,
+      "step": 7332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015968252141825037,
+      "loss": 1.1269,
+      "step": 7333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001596712676004869,
+      "loss": 0.9384,
+      "step": 7334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015966001260899132,
+      "loss": 1.1525,
+      "step": 7335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015964875644398506,
+      "loss": 1.0746,
+      "step": 7336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015963749910568957,
+      "loss": 0.9609,
+      "step": 7337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001596262405943262,
+      "loss": 0.9909,
+      "step": 7338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015961498091011643,
+      "loss": 1.0744,
+      "step": 7339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015960372005328175,
+      "loss": 1.158,
+      "step": 7340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015959245802404365,
+      "loss": 1.0462,
+      "step": 7341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015958119482262367,
+      "loss": 1.06,
+      "step": 7342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015956993044924334,
+      "loss": 1.088,
+      "step": 7343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015955866490412426,
+      "loss": 1.0447,
+      "step": 7344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015954739818748798,
+      "loss": 0.9872,
+      "step": 7345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001595361302995561,
+      "loss": 0.9298,
+      "step": 7346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015952486124055034,
+      "loss": 1.1546,
+      "step": 7347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015951359101069233,
+      "loss": 0.7666,
+      "step": 7348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015950231961020373,
+      "loss": 1.1078,
+      "step": 7349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015949104703930623,
+      "loss": 1.1961,
+      "step": 7350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015947977329822162,
+      "loss": 0.9981,
+      "step": 7351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001594684983871716,
+      "loss": 1.1674,
+      "step": 7352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015945722230637803,
+      "loss": 1.1838,
+      "step": 7353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001594459450560626,
+      "loss": 1.0486,
+      "step": 7354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015943466663644718,
+      "loss": 0.95,
+      "step": 7355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001594233870477537,
+      "loss": 1.0964,
+      "step": 7356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015941210629020388,
+      "loss": 1.151,
+      "step": 7357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015940082436401967,
+      "loss": 1.1262,
+      "step": 7358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015938954126942301,
+      "loss": 1.2865,
+      "step": 7359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015937825700663587,
+      "loss": 1.1643,
+      "step": 7360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015936697157588014,
+      "loss": 0.9802,
+      "step": 7361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015935568497737782,
+      "loss": 0.9534,
+      "step": 7362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015934439721135095,
+      "loss": 1.0903,
+      "step": 7363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015933310827802153,
+      "loss": 1.1389,
+      "step": 7364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015932181817761163,
+      "loss": 1.011,
+      "step": 7365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015931052691034328,
+      "loss": 1.1061,
+      "step": 7366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015929923447643865,
+      "loss": 1.1942,
+      "step": 7367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015928794087611986,
+      "loss": 1.0745,
+      "step": 7368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000159276646109609,
+      "loss": 1.0693,
+      "step": 7369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015926535017712822,
+      "loss": 1.0687,
+      "step": 7370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001592540530788998,
+      "loss": 1.1082,
+      "step": 7371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015924275481514588,
+      "loss": 0.7954,
+      "step": 7372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015923145538608875,
+      "loss": 1.0727,
+      "step": 7373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015922015479195062,
+      "loss": 1.0822,
+      "step": 7374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015920885303295377,
+      "loss": 1.0225,
+      "step": 7375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001591975501093206,
+      "loss": 0.9158,
+      "step": 7376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001591862460212733,
+      "loss": 1.0335,
+      "step": 7377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001591749407690343,
+      "loss": 1.0715,
+      "step": 7378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015916363435282597,
+      "loss": 1.1798,
+      "step": 7379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015915232677287072,
+      "loss": 1.326,
+      "step": 7380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001591410180293909,
+      "loss": 0.8958,
+      "step": 7381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015912970812260907,
+      "loss": 1.0423,
+      "step": 7382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001591183970527476,
+      "loss": 0.9912,
+      "step": 7383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015910708482002898,
+      "loss": 1.1459,
+      "step": 7384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015909577142467573,
+      "loss": 0.8908,
+      "step": 7385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015908445686691044,
+      "loss": 1.0323,
+      "step": 7386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015907314114695561,
+      "loss": 1.0117,
+      "step": 7387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015906182426503386,
+      "loss": 1.0078,
+      "step": 7388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015905050622136777,
+      "loss": 1.1014,
+      "step": 7389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015903918701617995,
+      "loss": 0.9966,
+      "step": 7390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015902786664969307,
+      "loss": 1.194,
+      "step": 7391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001590165451221298,
+      "loss": 1.1688,
+      "step": 7392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015900522243371282,
+      "loss": 1.1476,
+      "step": 7393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001589938985846649,
+      "loss": 0.9513,
+      "step": 7394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015898257357520868,
+      "loss": 0.9571,
+      "step": 7395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000158971247405567,
+      "loss": 0.9029,
+      "step": 7396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001589599200759626,
+      "loss": 1.0513,
+      "step": 7397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015894859158661835,
+      "loss": 0.8633,
+      "step": 7398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015893726193775702,
+      "loss": 1.1637,
+      "step": 7399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001589259311296015,
+      "loss": 0.8789,
+      "step": 7400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015891459916237465,
+      "loss": 1.1708,
+      "step": 7401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001589032660362994,
+      "loss": 0.9723,
+      "step": 7402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015889193175159863,
+      "loss": 0.9616,
+      "step": 7403
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001588805963084953,
+      "loss": 1.0232,
+      "step": 7404
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015886925970721235,
+      "loss": 1.0662,
+      "step": 7405
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015885792194797284,
+      "loss": 0.945,
+      "step": 7406
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001588465830309997,
+      "loss": 1.1049,
+      "step": 7407
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015883524295651604,
+      "loss": 0.9314,
+      "step": 7408
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015882390172474492,
+      "loss": 1.0149,
+      "step": 7409
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015881255933590933,
+      "loss": 1.2624,
+      "step": 7410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015880121579023247,
+      "loss": 0.8986,
+      "step": 7411
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001587898710879374,
+      "loss": 1.0564,
+      "step": 7412
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015877852522924732,
+      "loss": 0.9735,
+      "step": 7413
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001587671782143854,
+      "loss": 1.0821,
+      "step": 7414
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001587558300435748,
+      "loss": 1.0664,
+      "step": 7415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015874448071703876,
+      "loss": 0.9568,
+      "step": 7416
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015873313023500053,
+      "loss": 1.3285,
+      "step": 7417
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015872177859768333,
+      "loss": 1.3708,
+      "step": 7418
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015871042580531053,
+      "loss": 0.9078,
+      "step": 7419
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015869907185810534,
+      "loss": 1.0893,
+      "step": 7420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015868771675629114,
+      "loss": 0.8325,
+      "step": 7421
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015867636050009132,
+      "loss": 1.044,
+      "step": 7422
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015866500308972924,
+      "loss": 0.9048,
+      "step": 7423
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001586536445254282,
+      "loss": 1.1395,
+      "step": 7424
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001586422848074118,
+      "loss": 0.9532,
+      "step": 7425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015863092393590333,
+      "loss": 1.0759,
+      "step": 7426
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015861956191112633,
+      "loss": 1.277,
+      "step": 7427
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015860819873330425,
+      "loss": 0.966,
+      "step": 7428
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015859683440266068,
+      "loss": 1.0149,
+      "step": 7429
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001585854689194191,
+      "loss": 1.1034,
+      "step": 7430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015857410228380304,
+      "loss": 0.9885,
+      "step": 7431
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015856273449603613,
+      "loss": 1.1295,
+      "step": 7432
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015855136555634197,
+      "loss": 0.949,
+      "step": 7433
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001585399954649442,
+      "loss": 0.9997,
+      "step": 7434
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001585286242220664,
+      "loss": 1.1063,
+      "step": 7435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015851725182793234,
+      "loss": 0.8829,
+      "step": 7436
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015850587828276564,
+      "loss": 1.0604,
+      "step": 7437
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015849450358679004,
+      "loss": 1.0742,
+      "step": 7438
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001584831277402293,
+      "loss": 0.9885,
+      "step": 7439
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015847175074330713,
+      "loss": 0.9628,
+      "step": 7440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015846037259624736,
+      "loss": 1.0163,
+      "step": 7441
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001584489932992738,
+      "loss": 1.0282,
+      "step": 7442
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015843761285261028,
+      "loss": 1.1282,
+      "step": 7443
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001584262312564806,
+      "loss": 0.9459,
+      "step": 7444
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001584148485111087,
+      "loss": 1.0976,
+      "step": 7445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015840346461671847,
+      "loss": 1.0764,
+      "step": 7446
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001583920795735338,
+      "loss": 1.2682,
+      "step": 7447
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015838069338177863,
+      "loss": 1.1169,
+      "step": 7448
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015836930604167697,
+      "loss": 0.9972,
+      "step": 7449
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001583579175534528,
+      "loss": 1.0328,
+      "step": 7450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015834652791733007,
+      "loss": 1.0145,
+      "step": 7451
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015833513713353286,
+      "loss": 0.9069,
+      "step": 7452
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015832374520228525,
+      "loss": 1.1138,
+      "step": 7453
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001583123521238113,
+      "loss": 1.1441,
+      "step": 7454
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015830095789833511,
+      "loss": 1.0645,
+      "step": 7455
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015828956252608077,
+      "loss": 1.0769,
+      "step": 7456
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015827816600727247,
+      "loss": 1.3376,
+      "step": 7457
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015826676834213436,
+      "loss": 1.044,
+      "step": 7458
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015825536953089064,
+      "loss": 1.3206,
+      "step": 7459
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015824396957376554,
+      "loss": 0.8705,
+      "step": 7460
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015823256847098324,
+      "loss": 1.1293,
+      "step": 7461
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015822116622276808,
+      "loss": 1.0169,
+      "step": 7462
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015820976282934426,
+      "loss": 0.9172,
+      "step": 7463
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015819835829093614,
+      "loss": 1.0479,
+      "step": 7464
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015818695260776805,
+      "loss": 1.2701,
+      "step": 7465
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001581755457800643,
+      "loss": 1.0354,
+      "step": 7466
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001581641378080493,
+      "loss": 1.1315,
+      "step": 7467
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015815272869194739,
+      "loss": 1.0696,
+      "step": 7468
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015814131843198308,
+      "loss": 1.0778,
+      "step": 7469
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015812990702838072,
+      "loss": 1.0473,
+      "step": 7470
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015811849448136477,
+      "loss": 1.0354,
+      "step": 7471
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001581070807911598,
+      "loss": 1.1749,
+      "step": 7472
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015809566595799027,
+      "loss": 0.9568,
+      "step": 7473
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015808424998208068,
+      "loss": 1.0264,
+      "step": 7474
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001580728328636556,
+      "loss": 0.9169,
+      "step": 7475
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015806141460293963,
+      "loss": 1.0403,
+      "step": 7476
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015804999520015734,
+      "loss": 1.4276,
+      "step": 7477
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015803857465553338,
+      "loss": 1.21,
+      "step": 7478
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015802715296929235,
+      "loss": 1.0759,
+      "step": 7479
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015801573014165892,
+      "loss": 1.2058,
+      "step": 7480
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015800430617285782,
+      "loss": 0.9911,
+      "step": 7481
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015799288106311372,
+      "loss": 1.0162,
+      "step": 7482
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001579814548126514,
+      "loss": 1.0712,
+      "step": 7483
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001579700274216955,
+      "loss": 1.076,
+      "step": 7484
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015795859889047091,
+      "loss": 0.9463,
+      "step": 7485
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001579471692192024,
+      "loss": 1.0603,
+      "step": 7486
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015793573840811476,
+      "loss": 1.0193,
+      "step": 7487
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001579243064574329,
+      "loss": 1.0242,
+      "step": 7488
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001579128733673816,
+      "loss": 1.1028,
+      "step": 7489
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001579014391381858,
+      "loss": 1.1639,
+      "step": 7490
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001578900037700704,
+      "loss": 0.8285,
+      "step": 7491
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015787856726326037,
+      "loss": 0.9168,
+      "step": 7492
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001578671296179806,
+      "loss": 1.1468,
+      "step": 7493
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001578556908344561,
+      "loss": 1.0758,
+      "step": 7494
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015784425091291187,
+      "loss": 1.0839,
+      "step": 7495
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015783280985357293,
+      "loss": 1.0355,
+      "step": 7496
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001578213676566643,
+      "loss": 1.2176,
+      "step": 7497
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015780992432241114,
+      "loss": 0.9087,
+      "step": 7498
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015779847985103842,
+      "loss": 0.9005,
+      "step": 7499
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015778703424277132,
+      "loss": 0.9576,
+      "step": 7500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.000157775587497835,
+      "loss": 1.1588,
+      "step": 7501
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015776413961645454,
+      "loss": 0.9756,
+      "step": 7502
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015775269059885516,
+      "loss": 0.9822,
+      "step": 7503
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00015774124044526207,
+      "loss": 1.0435,
+      "step": 7504
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001577297891559005,
+      "loss": 1.0324,
+      "step": 7505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015771833673099565,
+      "loss": 1.0913,
+      "step": 7506
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015770688317077284,
+      "loss": 0.9659,
+      "step": 7507
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015769542847545736,
+      "loss": 1.1231,
+      "step": 7508
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015768397264527448,
+      "loss": 1.0541,
+      "step": 7509
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015767251568044957,
+      "loss": 1.0598,
+      "step": 7510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015766105758120797,
+      "loss": 1.0362,
+      "step": 7511
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015764959834777508,
+      "loss": 0.9693,
+      "step": 7512
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001576381379803763,
+      "loss": 1.0971,
+      "step": 7513
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015762667647923703,
+      "loss": 1.1319,
+      "step": 7514
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015761521384458272,
+      "loss": 1.0164,
+      "step": 7515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015760375007663893,
+      "loss": 1.1272,
+      "step": 7516
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015759228517563099,
+      "loss": 1.1433,
+      "step": 7517
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015758081914178456,
+      "loss": 1.1588,
+      "step": 7518
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001575693519753251,
+      "loss": 1.165,
+      "step": 7519
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015755788367647814,
+      "loss": 0.8944,
+      "step": 7520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015754641424546936,
+      "loss": 1.0519,
+      "step": 7521
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001575349436825243,
+      "loss": 1.0107,
+      "step": 7522
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001575234719878686,
+      "loss": 1.0414,
+      "step": 7523
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001575119991617279,
+      "loss": 1.06,
+      "step": 7524
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015750052520432787,
+      "loss": 1.0415,
+      "step": 7525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001574890501158942,
+      "loss": 1.0542,
+      "step": 7526
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001574775738966526,
+      "loss": 1.0591,
+      "step": 7527
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015746609654682887,
+      "loss": 1.0718,
+      "step": 7528
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015745461806664868,
+      "loss": 1.0228,
+      "step": 7529
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015744313845633787,
+      "loss": 0.9307,
+      "step": 7530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001574316577161222,
+      "loss": 1.0826,
+      "step": 7531
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015742017584622752,
+      "loss": 1.1769,
+      "step": 7532
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001574086928468797,
+      "loss": 1.0491,
+      "step": 7533
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015739720871830456,
+      "loss": 0.779,
+      "step": 7534
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015738572346072805,
+      "loss": 0.9975,
+      "step": 7535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015737423707437603,
+      "loss": 1.2415,
+      "step": 7536
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015736274955947444,
+      "loss": 0.7904,
+      "step": 7537
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015735126091624932,
+      "loss": 1.1293,
+      "step": 7538
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015733977114492658,
+      "loss": 1.0039,
+      "step": 7539
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015732828024573217,
+      "loss": 1.1605,
+      "step": 7540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015731678821889224,
+      "loss": 1.1577,
+      "step": 7541
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015730529506463277,
+      "loss": 1.0985,
+      "step": 7542
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001572938007831798,
+      "loss": 0.9317,
+      "step": 7543
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015728230537475949,
+      "loss": 1.06,
+      "step": 7544
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015727080883959794,
+      "loss": 1.1823,
+      "step": 7545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015725931117792125,
+      "loss": 0.8986,
+      "step": 7546
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001572478123899556,
+      "loss": 1.1681,
+      "step": 7547
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015723631247592716,
+      "loss": 0.9743,
+      "step": 7548
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001572248114360622,
+      "loss": 0.9514,
+      "step": 7549
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015721330927058683,
+      "loss": 0.9541,
+      "step": 7550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015720180597972736,
+      "loss": 1.4431,
+      "step": 7551
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001571903015637101,
+      "loss": 0.9353,
+      "step": 7552
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015717879602276122,
+      "loss": 0.9951,
+      "step": 7553
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015716728935710717,
+      "loss": 1.1986,
+      "step": 7554
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015715578156697422,
+      "loss": 1.2017,
+      "step": 7555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001571442726525887,
+      "loss": 0.8942,
+      "step": 7556
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015713276261417706,
+      "loss": 0.9655,
+      "step": 7557
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015712125145196567,
+      "loss": 1.0393,
+      "step": 7558
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015710973916618096,
+      "loss": 1.028,
+      "step": 7559
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001570982257570493,
+      "loss": 1.0413,
+      "step": 7560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001570867112247973,
+      "loss": 0.8554,
+      "step": 7561
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015707519556965133,
+      "loss": 1.3657,
+      "step": 7562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015706367879183796,
+      "loss": 1.1555,
+      "step": 7563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015705216089158372,
+      "loss": 1.0534,
+      "step": 7564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015704064186911518,
+      "loss": 1.0995,
+      "step": 7565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015702912172465885,
+      "loss": 1.0116,
+      "step": 7566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001570176004584414,
+      "loss": 1.2412,
+      "step": 7567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015700607807068946,
+      "loss": 1.0514,
+      "step": 7568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015699455456162962,
+      "loss": 1.0682,
+      "step": 7569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015698302993148857,
+      "loss": 1.2315,
+      "step": 7570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000156971504180493,
+      "loss": 1.0128,
+      "step": 7571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015695997730886965,
+      "loss": 0.9636,
+      "step": 7572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001569484493168452,
+      "loss": 1.0897,
+      "step": 7573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015693692020464647,
+      "loss": 0.978,
+      "step": 7574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015692538997250018,
+      "loss": 0.9414,
+      "step": 7575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015691385862063312,
+      "loss": 1.0122,
+      "step": 7576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001569023261492722,
+      "loss": 0.9152,
+      "step": 7577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015689079255864417,
+      "loss": 1.2563,
+      "step": 7578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015687925784897594,
+      "loss": 1.0549,
+      "step": 7579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015686772202049438,
+      "loss": 1.1091,
+      "step": 7580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001568561850734264,
+      "loss": 1.1123,
+      "step": 7581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015684464700799894,
+      "loss": 1.1687,
+      "step": 7582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000156833107824439,
+      "loss": 1.0666,
+      "step": 7583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015682156752297343,
+      "loss": 0.9787,
+      "step": 7584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015681002610382934,
+      "loss": 1.0772,
+      "step": 7585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001567984835672337,
+      "loss": 1.0552,
+      "step": 7586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001567869399134136,
+      "loss": 1.1063,
+      "step": 7587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015677539514259608,
+      "loss": 1.0105,
+      "step": 7588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015676384925500817,
+      "loss": 1.1051,
+      "step": 7589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015675230225087704,
+      "loss": 0.9489,
+      "step": 7590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015674075413042978,
+      "loss": 0.9697,
+      "step": 7591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015672920489389358,
+      "loss": 0.9022,
+      "step": 7592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015671765454149559,
+      "loss": 1.0876,
+      "step": 7593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015670610307346304,
+      "loss": 1.0957,
+      "step": 7594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015669455049002306,
+      "loss": 1.0617,
+      "step": 7595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015668299679140297,
+      "loss": 1.0147,
+      "step": 7596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015667144197783003,
+      "loss": 1.0736,
+      "step": 7597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015665988604953148,
+      "loss": 1.0209,
+      "step": 7598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015664832900673463,
+      "loss": 1.1329,
+      "step": 7599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015663677084966685,
+      "loss": 1.004,
+      "step": 7600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015662521157855546,
+      "loss": 0.7926,
+      "step": 7601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001566136511936278,
+      "loss": 0.9933,
+      "step": 7602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001566020896951113,
+      "loss": 0.9208,
+      "step": 7603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015659052708323339,
+      "loss": 1.1746,
+      "step": 7604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015657896335822147,
+      "loss": 1.1039,
+      "step": 7605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000156567398520303,
+      "loss": 1.0772,
+      "step": 7606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001565558325697055,
+      "loss": 0.9121,
+      "step": 7607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015654426550665643,
+      "loss": 0.905,
+      "step": 7608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015653269733138328,
+      "loss": 1.0411,
+      "step": 7609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001565211280441137,
+      "loss": 1.0256,
+      "step": 7610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001565095576450752,
+      "loss": 0.8842,
+      "step": 7611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015649798613449534,
+      "loss": 1.1977,
+      "step": 7612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015648641351260177,
+      "loss": 0.9848,
+      "step": 7613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015647483977962212,
+      "loss": 1.1443,
+      "step": 7614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015646326493578405,
+      "loss": 1.0733,
+      "step": 7615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001564516889813152,
+      "loss": 1.1223,
+      "step": 7616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015644011191644333,
+      "loss": 1.0675,
+      "step": 7617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015642853374139607,
+      "loss": 1.1044,
+      "step": 7618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015641695445640127,
+      "loss": 1.057,
+      "step": 7619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001564053740616866,
+      "loss": 1.0599,
+      "step": 7620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015639379255747994,
+      "loss": 0.9661,
+      "step": 7621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000156382209944009,
+      "loss": 1.0701,
+      "step": 7622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015637062622150168,
+      "loss": 0.9931,
+      "step": 7623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001563590413901858,
+      "loss": 1.1191,
+      "step": 7624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015634745545028923,
+      "loss": 1.0351,
+      "step": 7625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015633586840203988,
+      "loss": 1.1592,
+      "step": 7626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015632428024566568,
+      "loss": 1.2151,
+      "step": 7627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015631269098139453,
+      "loss": 1.1702,
+      "step": 7628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001563011006094544,
+      "loss": 1.1393,
+      "step": 7629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015628950913007333,
+      "loss": 1.2581,
+      "step": 7630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015627791654347926,
+      "loss": 0.9052,
+      "step": 7631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015626632284990017,
+      "loss": 1.0352,
+      "step": 7632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015625472804956425,
+      "loss": 0.99,
+      "step": 7633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015624313214269945,
+      "loss": 1.0945,
+      "step": 7634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015623153512953392,
+      "loss": 1.163,
+      "step": 7635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015621993701029575,
+      "loss": 0.9567,
+      "step": 7636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015620833778521307,
+      "loss": 0.9976,
+      "step": 7637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015619673745451403,
+      "loss": 1.0135,
+      "step": 7638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015618513601842686,
+      "loss": 1.0353,
+      "step": 7639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001561735334771797,
+      "loss": 1.1896,
+      "step": 7640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001561619298310008,
+      "loss": 1.1741,
+      "step": 7641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001561503250801184,
+      "loss": 1.0952,
+      "step": 7642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001561387192247608,
+      "loss": 0.8922,
+      "step": 7643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001561271122651562,
+      "loss": 1.0361,
+      "step": 7644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015611550420153297,
+      "loss": 1.0456,
+      "step": 7645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015610389503411943,
+      "loss": 1.1291,
+      "step": 7646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015609228476314396,
+      "loss": 0.9595,
+      "step": 7647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015608067338883488,
+      "loss": 0.969,
+      "step": 7648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001560690609114206,
+      "loss": 1.016,
+      "step": 7649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015605744733112955,
+      "loss": 1.1209,
+      "step": 7650
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001560458326481902,
+      "loss": 1.0561,
+      "step": 7651
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015603421686283093,
+      "loss": 1.3302,
+      "step": 7652
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015602259997528028,
+      "loss": 0.9904,
+      "step": 7653
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015601098198576678,
+      "loss": 1.0482,
+      "step": 7654
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001559993628945189,
+      "loss": 0.9287,
+      "step": 7655
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015598774270176517,
+      "loss": 1.094,
+      "step": 7656
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001559761214077342,
+      "loss": 1.0323,
+      "step": 7657
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015596449901265463,
+      "loss": 0.9299,
+      "step": 7658
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015595287551675496,
+      "loss": 1.2345,
+      "step": 7659
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001559412509202639,
+      "loss": 1.1671,
+      "step": 7660
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001559296252234101,
+      "loss": 0.9725,
+      "step": 7661
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001559179984264222,
+      "loss": 1.1131,
+      "step": 7662
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015590637052952894,
+      "loss": 1.1816,
+      "step": 7663
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015589474153295901,
+      "loss": 1.0325,
+      "step": 7664
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015588311143694116,
+      "loss": 0.8855,
+      "step": 7665
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001558714802417042,
+      "loss": 1.0475,
+      "step": 7666
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015585984794747683,
+      "loss": 1.3594,
+      "step": 7667
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001558482145544879,
+      "loss": 1.2175,
+      "step": 7668
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015583658006296624,
+      "loss": 1.1358,
+      "step": 7669
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015582494447314075,
+      "loss": 1.1566,
+      "step": 7670
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001558133077852402,
+      "loss": 1.1116,
+      "step": 7671
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015580166999949353,
+      "loss": 0.93,
+      "step": 7672
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001557900311161297,
+      "loss": 0.9876,
+      "step": 7673
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015577839113537756,
+      "loss": 0.9607,
+      "step": 7674
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015576675005746611,
+      "loss": 0.9791,
+      "step": 7675
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015575510788262434,
+      "loss": 1.2546,
+      "step": 7676
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001557434646110813,
+      "loss": 0.8389,
+      "step": 7677
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001557318202430659,
+      "loss": 0.9908,
+      "step": 7678
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015572017477880724,
+      "loss": 0.9975,
+      "step": 7679
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015570852821853443,
+      "loss": 0.9246,
+      "step": 7680
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015569688056247646,
+      "loss": 0.9496,
+      "step": 7681
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015568523181086255,
+      "loss": 1.0701,
+      "step": 7682
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001556735819639217,
+      "loss": 1.0039,
+      "step": 7683
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015566193102188323,
+      "loss": 0.8305,
+      "step": 7684
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015565027898497617,
+      "loss": 1.2202,
+      "step": 7685
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015563862585342975,
+      "loss": 1.0704,
+      "step": 7686
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015562697162747324,
+      "loss": 1.1182,
+      "step": 7687
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001556153163073358,
+      "loss": 1.0046,
+      "step": 7688
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015560365989324677,
+      "loss": 0.8138,
+      "step": 7689
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015559200238543535,
+      "loss": 1.0001,
+      "step": 7690
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015558034378413096,
+      "loss": 1.1881,
+      "step": 7691
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015556868408956275,
+      "loss": 0.9829,
+      "step": 7692
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015555702330196023,
+      "loss": 1.2481,
+      "step": 7693
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001555453614215527,
+      "loss": 1.0562,
+      "step": 7694
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015553369844856953,
+      "loss": 1.0138,
+      "step": 7695
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015552203438324016,
+      "loss": 1.3013,
+      "step": 7696
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015551036922579402,
+      "loss": 0.9674,
+      "step": 7697
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015549870297646052,
+      "loss": 1.0828,
+      "step": 7698
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001554870356354692,
+      "loss": 0.9806,
+      "step": 7699
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015547536720304951,
+      "loss": 0.9326,
+      "step": 7700
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015546369767943102,
+      "loss": 0.9817,
+      "step": 7701
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015545202706484322,
+      "loss": 0.994,
+      "step": 7702
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015544035535951567,
+      "loss": 1.1313,
+      "step": 7703
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015542868256367799,
+      "loss": 1.2117,
+      "step": 7704
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015541700867755974,
+      "loss": 1.0615,
+      "step": 7705
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015540533370139056,
+      "loss": 1.1876,
+      "step": 7706
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015539365763540014,
+      "loss": 1.3396,
+      "step": 7707
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015538198047981806,
+      "loss": 0.8238,
+      "step": 7708
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001553703022348741,
+      "loss": 0.8973,
+      "step": 7709
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001553586229007979,
+      "loss": 1.0286,
+      "step": 7710
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015534694247781926,
+      "loss": 1.0888,
+      "step": 7711
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015533526096616783,
+      "loss": 1.1054,
+      "step": 7712
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001553235783660735,
+      "loss": 1.0034,
+      "step": 7713
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.000155311894677766,
+      "loss": 1.1342,
+      "step": 7714
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015530020990147517,
+      "loss": 1.0381,
+      "step": 7715
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015528852403743083,
+      "loss": 1.0838,
+      "step": 7716
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001552768370858629,
+      "loss": 1.1263,
+      "step": 7717
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015526514904700117,
+      "loss": 1.0548,
+      "step": 7718
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015525345992107562,
+      "loss": 0.9219,
+      "step": 7719
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015524176970831615,
+      "loss": 1.1113,
+      "step": 7720
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001552300784089527,
+      "loss": 1.0824,
+      "step": 7721
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001552183860232152,
+      "loss": 1.0363,
+      "step": 7722
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001552066925513337,
+      "loss": 1.0815,
+      "step": 7723
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015519499799353817,
+      "loss": 0.9387,
+      "step": 7724
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015518330235005868,
+      "loss": 0.9724,
+      "step": 7725
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015517160562112528,
+      "loss": 1.1235,
+      "step": 7726
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015515990780696798,
+      "loss": 1.1069,
+      "step": 7727
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015514820890781693,
+      "loss": 1.0232,
+      "step": 7728
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001551365089239023,
+      "loss": 0.8609,
+      "step": 7729
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015512480785545408,
+      "loss": 1.0136,
+      "step": 7730
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015511310570270258,
+      "loss": 1.3,
+      "step": 7731
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001551014024658779,
+      "loss": 1.0663,
+      "step": 7732
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015508969814521025,
+      "loss": 1.0988,
+      "step": 7733
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001550779927409299,
+      "loss": 0.957,
+      "step": 7734
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015506628625326703,
+      "loss": 1.0183,
+      "step": 7735
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00015505457868245193,
+      "loss": 1.1899,
+      "step": 7736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015504287002871488,
+      "loss": 0.9594,
+      "step": 7737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015503116029228623,
+      "loss": 1.1417,
+      "step": 7738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015501944947339627,
+      "loss": 1.0232,
+      "step": 7739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001550077375722754,
+      "loss": 0.8589,
+      "step": 7740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001549960245891539,
+      "loss": 0.9944,
+      "step": 7741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015498431052426222,
+      "loss": 1.0082,
+      "step": 7742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015497259537783085,
+      "loss": 1.2129,
+      "step": 7743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001549608791500901,
+      "loss": 0.9803,
+      "step": 7744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015494916184127046,
+      "loss": 1.2281,
+      "step": 7745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015493744345160246,
+      "loss": 1.1089,
+      "step": 7746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015492572398131656,
+      "loss": 0.9761,
+      "step": 7747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015491400343064328,
+      "loss": 1.2698,
+      "step": 7748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001549022817998132,
+      "loss": 0.9005,
+      "step": 7749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015489055908905682,
+      "loss": 1.0182,
+      "step": 7750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001548788352986048,
+      "loss": 1.2094,
+      "step": 7751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015486711042868767,
+      "loss": 1.0312,
+      "step": 7752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015485538447953615,
+      "loss": 0.9115,
+      "step": 7753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015484365745138082,
+      "loss": 1.0586,
+      "step": 7754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015483192934445235,
+      "loss": 1.0081,
+      "step": 7755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015482020015898144,
+      "loss": 0.9725,
+      "step": 7756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015480846989519885,
+      "loss": 1.1476,
+      "step": 7757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015479673855333523,
+      "loss": 1.114,
+      "step": 7758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015478500613362142,
+      "loss": 0.9753,
+      "step": 7759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015477327263628814,
+      "loss": 1.0323,
+      "step": 7760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015476153806156623,
+      "loss": 1.179,
+      "step": 7761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015474980240968642,
+      "loss": 0.9186,
+      "step": 7762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015473806568087968,
+      "loss": 0.9897,
+      "step": 7763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015472632787537675,
+      "loss": 0.9982,
+      "step": 7764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001547145889934086,
+      "loss": 0.9754,
+      "step": 7765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001547028490352061,
+      "loss": 1.0854,
+      "step": 7766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015469110800100017,
+      "loss": 1.1945,
+      "step": 7767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015467936589102176,
+      "loss": 1.1103,
+      "step": 7768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015466762270550184,
+      "loss": 1.1571,
+      "step": 7769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001546558784446714,
+      "loss": 0.9774,
+      "step": 7770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001546441331087614,
+      "loss": 0.9372,
+      "step": 7771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015463238669800297,
+      "loss": 0.8276,
+      "step": 7772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001546206392126271,
+      "loss": 0.9702,
+      "step": 7773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015460889065286484,
+      "loss": 1.0904,
+      "step": 7774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015459714101894733,
+      "loss": 1.1021,
+      "step": 7775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015458539031110567,
+      "loss": 0.93,
+      "step": 7776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000154573638529571,
+      "loss": 1.0215,
+      "step": 7777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015456188567457445,
+      "loss": 1.0344,
+      "step": 7778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015455013174634725,
+      "loss": 1.0646,
+      "step": 7779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015453837674512055,
+      "loss": 1.0158,
+      "step": 7780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001545266206711256,
+      "loss": 1.1559,
+      "step": 7781
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015451486352459362,
+      "loss": 1.045,
+      "step": 7782
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001545031053057559,
+      "loss": 1.0574,
+      "step": 7783
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015449134601484368,
+      "loss": 1.0579,
+      "step": 7784
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001544795856520883,
+      "loss": 0.9911,
+      "step": 7785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001544678242177211,
+      "loss": 1.185,
+      "step": 7786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015445606171197342,
+      "loss": 1.0153,
+      "step": 7787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015444429813507658,
+      "loss": 0.9103,
+      "step": 7788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015443253348726202,
+      "loss": 1.2564,
+      "step": 7789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001544207677687611,
+      "loss": 1.0415,
+      "step": 7790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001544090009798053,
+      "loss": 0.835,
+      "step": 7791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001543972331206261,
+      "loss": 0.8869,
+      "step": 7792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015438546419145488,
+      "loss": 1.1516,
+      "step": 7793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001543736941925232,
+      "loss": 0.7971,
+      "step": 7794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015436192312406256,
+      "loss": 1.0138,
+      "step": 7795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001543501509863045,
+      "loss": 0.9477,
+      "step": 7796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001543383777794806,
+      "loss": 0.9875,
+      "step": 7797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015432660350382234,
+      "loss": 0.9683,
+      "step": 7798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015431482815956148,
+      "loss": 1.0297,
+      "step": 7799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001543030517469295,
+      "loss": 0.9994,
+      "step": 7800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001542912742661581,
+      "loss": 0.9849,
+      "step": 7801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015427949571747895,
+      "loss": 1.0203,
+      "step": 7802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015426771610112374,
+      "loss": 0.8886,
+      "step": 7803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015425593541732412,
+      "loss": 1.0249,
+      "step": 7804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015424415366631188,
+      "loss": 1.0428,
+      "step": 7805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015423237084831874,
+      "loss": 1.1414,
+      "step": 7806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015422058696357643,
+      "loss": 1.1215,
+      "step": 7807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015420880201231683,
+      "loss": 1.1131,
+      "step": 7808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015419701599477166,
+      "loss": 1.1201,
+      "step": 7809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001541852289111728,
+      "loss": 0.8555,
+      "step": 7810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015417344076175206,
+      "loss": 1.2418,
+      "step": 7811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015416165154674135,
+      "loss": 1.1681,
+      "step": 7812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015414986126637258,
+      "loss": 1.1566,
+      "step": 7813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015413806992087763,
+      "loss": 1.0093,
+      "step": 7814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015412627751048842,
+      "loss": 0.9661,
+      "step": 7815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015411448403543693,
+      "loss": 1.1596,
+      "step": 7816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015410268949595516,
+      "loss": 0.9531,
+      "step": 7817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001540908938922751,
+      "loss": 0.9571,
+      "step": 7818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001540790972246287,
+      "loss": 1.0372,
+      "step": 7819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015406729949324813,
+      "loss": 1.0357,
+      "step": 7820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015405550069836528,
+      "loss": 1.0609,
+      "step": 7821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001540437008402124,
+      "loss": 1.1604,
+      "step": 7822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001540318999190215,
+      "loss": 1.0786,
+      "step": 7823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015402009793502472,
+      "loss": 0.9295,
+      "step": 7824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001540082948884542,
+      "loss": 0.9856,
+      "step": 7825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001539964907795421,
+      "loss": 0.9441,
+      "step": 7826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015398468560852066,
+      "loss": 0.7704,
+      "step": 7827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015397287937562204,
+      "loss": 1.0512,
+      "step": 7828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015396107208107845,
+      "loss": 1.0732,
+      "step": 7829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015394926372512218,
+      "loss": 0.9576,
+      "step": 7830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001539374543079855,
+      "loss": 1.0935,
+      "step": 7831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001539256438299007,
+      "loss": 1.1784,
+      "step": 7832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015391383229110007,
+      "loss": 0.9425,
+      "step": 7833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001539020196918159,
+      "loss": 1.0817,
+      "step": 7834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015389020603228062,
+      "loss": 1.0168,
+      "step": 7835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015387839131272659,
+      "loss": 0.9679,
+      "step": 7836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001538665755333862,
+      "loss": 0.9814,
+      "step": 7837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015385475869449184,
+      "loss": 0.8781,
+      "step": 7838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000153842940796276,
+      "loss": 1.0959,
+      "step": 7839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015383112183897107,
+      "loss": 1.082,
+      "step": 7840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015381930182280958,
+      "loss": 1.0004,
+      "step": 7841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000153807480748024,
+      "loss": 0.9757,
+      "step": 7842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015379565861484688,
+      "loss": 1.0568,
+      "step": 7843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001537838354235107,
+      "loss": 1.2072,
+      "step": 7844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015377201117424812,
+      "loss": 1.1193,
+      "step": 7845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015376018586729163,
+      "loss": 0.9914,
+      "step": 7846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001537483595028739,
+      "loss": 1.0366,
+      "step": 7847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001537365320812275,
+      "loss": 0.9028,
+      "step": 7848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001537247036025851,
+      "loss": 1.0743,
+      "step": 7849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001537128740671794,
+      "loss": 1.005,
+      "step": 7850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015370104347524303,
+      "loss": 1.1747,
+      "step": 7851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015368921182700872,
+      "loss": 1.0407,
+      "step": 7852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001536773791227092,
+      "loss": 1.0039,
+      "step": 7853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015366554536257721,
+      "loss": 1.2568,
+      "step": 7854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015365371054684555,
+      "loss": 1.2883,
+      "step": 7855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015364187467574697,
+      "loss": 0.9686,
+      "step": 7856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001536300377495143,
+      "loss": 0.8667,
+      "step": 7857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015361819976838036,
+      "loss": 1.0492,
+      "step": 7858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015360636073257804,
+      "loss": 1.0486,
+      "step": 7859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015359452064234017,
+      "loss": 1.161,
+      "step": 7860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015358267949789966,
+      "loss": 1.117,
+      "step": 7861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015357083729948946,
+      "loss": 0.9928,
+      "step": 7862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015355899404734244,
+      "loss": 1.044,
+      "step": 7863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001535471497416916,
+      "loss": 0.9814,
+      "step": 7864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015353530438276986,
+      "loss": 1.2381,
+      "step": 7865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015352345797081034,
+      "loss": 1.1706,
+      "step": 7866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015351161050604592,
+      "loss": 1.2249,
+      "step": 7867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015349976198870973,
+      "loss": 1.0071,
+      "step": 7868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001534879124190348,
+      "loss": 0.9358,
+      "step": 7869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001534760617972542,
+      "loss": 1.0788,
+      "step": 7870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015346421012360105,
+      "loss": 0.9136,
+      "step": 7871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015345235739830848,
+      "loss": 1.1533,
+      "step": 7872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001534405036216096,
+      "loss": 1.0578,
+      "step": 7873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015342864879373762,
+      "loss": 1.2098,
+      "step": 7874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015341679291492566,
+      "loss": 1.2033,
+      "step": 7875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015340493598540696,
+      "loss": 0.9876,
+      "step": 7876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015339307800541475,
+      "loss": 1.0375,
+      "step": 7877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001533812189751823,
+      "loss": 0.9915,
+      "step": 7878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015336935889494283,
+      "loss": 1.0875,
+      "step": 7879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015335749776492966,
+      "loss": 0.9042,
+      "step": 7880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015334563558537604,
+      "loss": 1.1193,
+      "step": 7881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001533337723565154,
+      "loss": 1.1012,
+      "step": 7882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000153321908078581,
+      "loss": 1.0473,
+      "step": 7883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015331004275180625,
+      "loss": 0.8998,
+      "step": 7884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015329817637642454,
+      "loss": 0.9368,
+      "step": 7885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015328630895266926,
+      "loss": 1.0249,
+      "step": 7886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015327444048077385,
+      "loss": 1.1205,
+      "step": 7887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015326257096097182,
+      "loss": 0.8679,
+      "step": 7888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015325070039349655,
+      "loss": 1.0195,
+      "step": 7889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015323882877858153,
+      "loss": 0.9855,
+      "step": 7890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001532269561164604,
+      "loss": 1.1104,
+      "step": 7891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001532150824073666,
+      "loss": 1.1804,
+      "step": 7892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015320320765153367,
+      "loss": 1.0205,
+      "step": 7893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015319133184919524,
+      "loss": 1.1291,
+      "step": 7894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015317945500058485,
+      "loss": 1.002,
+      "step": 7895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001531675771059362,
+      "loss": 1.0011,
+      "step": 7896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015315569816548282,
+      "loss": 1.2324,
+      "step": 7897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001531438181794585,
+      "loss": 0.9268,
+      "step": 7898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015313193714809676,
+      "loss": 1.0137,
+      "step": 7899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015312005507163144,
+      "loss": 1.0559,
+      "step": 7900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001531081719502962,
+      "loss": 1.1143,
+      "step": 7901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001530962877843248,
+      "loss": 1.0971,
+      "step": 7902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015308440257395093,
+      "loss": 1.0933,
+      "step": 7903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015307251631940848,
+      "loss": 0.8695,
+      "step": 7904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015306062902093118,
+      "loss": 1.2545,
+      "step": 7905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001530487406787529,
+      "loss": 0.9565,
+      "step": 7906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015303685129310743,
+      "loss": 0.8302,
+      "step": 7907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015302496086422868,
+      "loss": 0.9442,
+      "step": 7908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015301306939235053,
+      "loss": 0.9958,
+      "step": 7909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015300117687770684,
+      "loss": 1.0326,
+      "step": 7910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015298928332053155,
+      "loss": 1.1479,
+      "step": 7911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015297738872105865,
+      "loss": 1.1242,
+      "step": 7912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015296549307952203,
+      "loss": 0.9646,
+      "step": 7913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015295359639615582,
+      "loss": 1.0889,
+      "step": 7914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015294169867119387,
+      "loss": 1.2422,
+      "step": 7915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015292979990487026,
+      "loss": 1.1651,
+      "step": 7916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015291790009741907,
+      "loss": 1.0667,
+      "step": 7917
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015290599924907433,
+      "loss": 0.9517,
+      "step": 7918
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001528940973600702,
+      "loss": 1.0192,
+      "step": 7919
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015288219443064066,
+      "loss": 0.933,
+      "step": 7920
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015287029046101997,
+      "loss": 1.1374,
+      "step": 7921
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015285838545144221,
+      "loss": 1.0947,
+      "step": 7922
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001528464794021416,
+      "loss": 0.9529,
+      "step": 7923
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015283457231335222,
+      "loss": 0.9941,
+      "step": 7924
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015282266418530847,
+      "loss": 1.0769,
+      "step": 7925
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015281075501824437,
+      "loss": 1.147,
+      "step": 7926
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015279884481239435,
+      "loss": 1.0899,
+      "step": 7927
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015278693356799258,
+      "loss": 1.3327,
+      "step": 7928
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015277502128527344,
+      "loss": 0.9711,
+      "step": 7929
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001527631079644711,
+      "loss": 0.7792,
+      "step": 7930
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015275119360582003,
+      "loss": 1.2596,
+      "step": 7931
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015273927820955456,
+      "loss": 1.0792,
+      "step": 7932
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.000152727361775909,
+      "loss": 1.1089,
+      "step": 7933
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001527154443051178,
+      "loss": 1.0541,
+      "step": 7934
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015270352579741542,
+      "loss": 1.0591,
+      "step": 7935
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015269160625303618,
+      "loss": 1.1296,
+      "step": 7936
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015267968567221464,
+      "loss": 0.9953,
+      "step": 7937
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015266776405518524,
+      "loss": 1.0971,
+      "step": 7938
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015265584140218243,
+      "loss": 1.0649,
+      "step": 7939
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001526439177134408,
+      "loss": 1.0463,
+      "step": 7940
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015263199298919488,
+      "loss": 1.0284,
+      "step": 7941
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015262006722967922,
+      "loss": 1.1045,
+      "step": 7942
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015260814043512836,
+      "loss": 1.0433,
+      "step": 7943
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015259621260577694,
+      "loss": 1.1458,
+      "step": 7944
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015258428374185956,
+      "loss": 1.3022,
+      "step": 7945
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001525723538436109,
+      "loss": 0.9136,
+      "step": 7946
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015256042291126557,
+      "loss": 1.1541,
+      "step": 7947
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015254849094505828,
+      "loss": 1.148,
+      "step": 7948
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015253655794522368,
+      "loss": 1.1796,
+      "step": 7949
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015252462391199663,
+      "loss": 1.0548,
+      "step": 7950
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001525126888456117,
+      "loss": 1.2836,
+      "step": 7951
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015250075274630379,
+      "loss": 1.0041,
+      "step": 7952
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015248881561430756,
+      "loss": 1.1504,
+      "step": 7953
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001524768774498579,
+      "loss": 1.0508,
+      "step": 7954
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015246493825318963,
+      "loss": 1.0374,
+      "step": 7955
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015245299802453758,
+      "loss": 0.9546,
+      "step": 7956
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001524410567641366,
+      "loss": 1.0041,
+      "step": 7957
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015242911447222156,
+      "loss": 1.0237,
+      "step": 7958
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001524171711490274,
+      "loss": 1.0606,
+      "step": 7959
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015240522679478905,
+      "loss": 0.8988,
+      "step": 7960
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001523932814097414,
+      "loss": 1.0352,
+      "step": 7961
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015238133499411946,
+      "loss": 1.1426,
+      "step": 7962
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015236938754815822,
+      "loss": 1.0001,
+      "step": 7963
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015235743907209267,
+      "loss": 1.0857,
+      "step": 7964
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015234548956615784,
+      "loss": 1.0243,
+      "step": 7965
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015233353903058877,
+      "loss": 1.1213,
+      "step": 7966
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00015232158746562052,
+      "loss": 0.8904,
+      "step": 7967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015230963487148822,
+      "loss": 0.9641,
+      "step": 7968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015229768124842691,
+      "loss": 0.8981,
+      "step": 7969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015228572659667177,
+      "loss": 1.1152,
+      "step": 7970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015227377091645797,
+      "loss": 1.1095,
+      "step": 7971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015226181420802058,
+      "loss": 1.2237,
+      "step": 7972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001522498564715949,
+      "loss": 1.0446,
+      "step": 7973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015223789770741605,
+      "loss": 1.0375,
+      "step": 7974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015222593791571932,
+      "loss": 1.1011,
+      "step": 7975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001522139770967399,
+      "loss": 1.2166,
+      "step": 7976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015220201525071314,
+      "loss": 0.9839,
+      "step": 7977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015219005237787422,
+      "loss": 1.1289,
+      "step": 7978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015217808847845855,
+      "loss": 1.2313,
+      "step": 7979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015216612355270143,
+      "loss": 1.0892,
+      "step": 7980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001521541576008382,
+      "loss": 0.9662,
+      "step": 7981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015214219062310418,
+      "loss": 1.1362,
+      "step": 7982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015213022261973484,
+      "loss": 1.0544,
+      "step": 7983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015211825359096558,
+      "loss": 0.9575,
+      "step": 7984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001521062835370318,
+      "loss": 1.0918,
+      "step": 7985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015209431245816893,
+      "loss": 1.086,
+      "step": 7986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015208234035461252,
+      "loss": 0.9424,
+      "step": 7987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000152070367226598,
+      "loss": 0.9979,
+      "step": 7988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015205839307436088,
+      "loss": 1.187,
+      "step": 7989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001520464178981367,
+      "loss": 0.936,
+      "step": 7990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015203444169816104,
+      "loss": 1.1548,
+      "step": 7991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015202246447466942,
+      "loss": 1.0112,
+      "step": 7992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015201048622789747,
+      "loss": 1.1391,
+      "step": 7993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001519985069580808,
+      "loss": 1.0639,
+      "step": 7994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015198652666545504,
+      "loss": 1.2208,
+      "step": 7995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015197454535025582,
+      "loss": 0.8918,
+      "step": 7996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015196256301271884,
+      "loss": 1.0383,
+      "step": 7997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015195057965307977,
+      "loss": 1.2405,
+      "step": 7998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001519385952715743,
+      "loss": 1.0239,
+      "step": 7999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015192660986843826,
+      "loss": 1.1216,
+      "step": 8000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001519146234439073,
+      "loss": 1.0466,
+      "step": 8001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015190263599821724,
+      "loss": 0.929,
+      "step": 8002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015189064753160384,
+      "loss": 0.9788,
+      "step": 8003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015187865804430298,
+      "loss": 0.9637,
+      "step": 8004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001518666675365504,
+      "loss": 1.0824,
+      "step": 8005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015185467600858203,
+      "loss": 1.0235,
+      "step": 8006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001518426834606337,
+      "loss": 1.1602,
+      "step": 8007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015183068989294132,
+      "loss": 1.0818,
+      "step": 8008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015181869530574081,
+      "loss": 1.0658,
+      "step": 8009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015180669969926808,
+      "loss": 1.1377,
+      "step": 8010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015179470307375911,
+      "loss": 1.0987,
+      "step": 8011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015178270542944987,
+      "loss": 0.8204,
+      "step": 8012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015177070676657633,
+      "loss": 1.1414,
+      "step": 8013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015175870708537448,
+      "loss": 1.0527,
+      "step": 8014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015174670638608043,
+      "loss": 0.9951,
+      "step": 8015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001517347046689302,
+      "loss": 1.0702,
+      "step": 8016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015172270193415985,
+      "loss": 1.1871,
+      "step": 8017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015171069818200548,
+      "loss": 1.1623,
+      "step": 8018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001516986934127032,
+      "loss": 1.268,
+      "step": 8019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001516866876264891,
+      "loss": 1.0466,
+      "step": 8020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015167468082359945,
+      "loss": 1.0506,
+      "step": 8021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015166267300427034,
+      "loss": 1.1438,
+      "step": 8022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015165066416873794,
+      "loss": 1.0144,
+      "step": 8023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015163865431723855,
+      "loss": 0.9499,
+      "step": 8024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001516266434500083,
+      "loss": 1.0223,
+      "step": 8025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015161463156728355,
+      "loss": 0.9526,
+      "step": 8026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015160261866930053,
+      "loss": 0.8962,
+      "step": 8027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015159060475629547,
+      "loss": 1.1378,
+      "step": 8028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015157858982850475,
+      "loss": 0.9751,
+      "step": 8029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001515665738861647,
+      "loss": 1.1025,
+      "step": 8030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015155455692951168,
+      "loss": 0.9596,
+      "step": 8031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015154253895878204,
+      "loss": 0.9616,
+      "step": 8032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015153051997421216,
+      "loss": 0.9691,
+      "step": 8033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001515184999760385,
+      "loss": 1.1254,
+      "step": 8034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015150647896449747,
+      "loss": 1.1173,
+      "step": 8035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015149445693982548,
+      "loss": 1.1308,
+      "step": 8036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015148243390225908,
+      "loss": 0.9558,
+      "step": 8037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001514704098520347,
+      "loss": 1.139,
+      "step": 8038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001514583847893889,
+      "loss": 0.9853,
+      "step": 8039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015144635871455818,
+      "loss": 1.3284,
+      "step": 8040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001514343316277791,
+      "loss": 1.1024,
+      "step": 8041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015142230352928823,
+      "loss": 1.2501,
+      "step": 8042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015141027441932216,
+      "loss": 1.3306,
+      "step": 8043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015139824429811759,
+      "loss": 1.003,
+      "step": 8044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015138621316591094,
+      "loss": 1.1626,
+      "step": 8045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001513741810229391,
+      "loss": 0.9498,
+      "step": 8046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001513621478694386,
+      "loss": 0.9983,
+      "step": 8047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015135011370564616,
+      "loss": 0.8995,
+      "step": 8048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015133807853179852,
+      "loss": 0.776,
+      "step": 8049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015132604234813237,
+      "loss": 1.0523,
+      "step": 8050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001513140051548845,
+      "loss": 1.0543,
+      "step": 8051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015130196695229165,
+      "loss": 0.9023,
+      "step": 8052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015128992774059063,
+      "loss": 0.9954,
+      "step": 8053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015127788752001823,
+      "loss": 1.1424,
+      "step": 8054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015126584629081134,
+      "loss": 1.175,
+      "step": 8055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015125380405320673,
+      "loss": 1.0259,
+      "step": 8056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015124176080744134,
+      "loss": 1.0712,
+      "step": 8057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015122971655375197,
+      "loss": 1.149,
+      "step": 8058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015121767129237562,
+      "loss": 1.055,
+      "step": 8059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015120562502354918,
+      "loss": 0.813,
+      "step": 8060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015119357774750965,
+      "loss": 0.9007,
+      "step": 8061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001511815294644939,
+      "loss": 1.1087,
+      "step": 8062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015116948017473902,
+      "loss": 1.0778,
+      "step": 8063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015115742987848195,
+      "loss": 1.2307,
+      "step": 8064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015114537857595974,
+      "loss": 0.8684,
+      "step": 8065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015113332626740946,
+      "loss": 0.9508,
+      "step": 8066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015112127295306818,
+      "loss": 0.8524,
+      "step": 8067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015110921863317295,
+      "loss": 1.135,
+      "step": 8068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001510971633079609,
+      "loss": 1.046,
+      "step": 8069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015108510697766912,
+      "loss": 1.0387,
+      "step": 8070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015107304964253487,
+      "loss": 0.8742,
+      "step": 8071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015106099130279517,
+      "loss": 0.9352,
+      "step": 8072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015104893195868731,
+      "loss": 1.0945,
+      "step": 8073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015103687161044848,
+      "loss": 1.1072,
+      "step": 8074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015102481025831588,
+      "loss": 1.1197,
+      "step": 8075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015101274790252676,
+      "loss": 1.2841,
+      "step": 8076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015100068454331836,
+      "loss": 1.0032,
+      "step": 8077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015098862018092808,
+      "loss": 0.9456,
+      "step": 8078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015097655481559308,
+      "loss": 1.1464,
+      "step": 8079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001509644884475508,
+      "loss": 1.115,
+      "step": 8080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001509524210770385,
+      "loss": 0.9585,
+      "step": 8081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001509403527042936,
+      "loss": 0.8887,
+      "step": 8082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015092828332955347,
+      "loss": 0.9961,
+      "step": 8083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015091621295305552,
+      "loss": 1.0728,
+      "step": 8084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015090414157503714,
+      "loss": 1.0482,
+      "step": 8085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001508920691957358,
+      "loss": 1.0626,
+      "step": 8086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015087999581538897,
+      "loss": 0.8475,
+      "step": 8087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015086792143423412,
+      "loss": 1.2003,
+      "step": 8088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015085584605250876,
+      "loss": 1.26,
+      "step": 8089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015084376967045043,
+      "loss": 1.0941,
+      "step": 8090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001508316922882966,
+      "loss": 0.9443,
+      "step": 8091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015081961390628497,
+      "loss": 1.2611,
+      "step": 8092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015080753452465296,
+      "loss": 1.2217,
+      "step": 8093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015079545414363826,
+      "loss": 1.1568,
+      "step": 8094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015078337276347846,
+      "loss": 0.9245,
+      "step": 8095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015077129038441127,
+      "loss": 0.9826,
+      "step": 8096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015075920700667426,
+      "loss": 1.2182,
+      "step": 8097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015074712263050515,
+      "loss": 1.1843,
+      "step": 8098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015073503725614165,
+      "loss": 0.9241,
+      "step": 8099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015072295088382144,
+      "loss": 1.1434,
+      "step": 8100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015071086351378231,
+      "loss": 1.0813,
+      "step": 8101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015069877514626198,
+      "loss": 1.0782,
+      "step": 8102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015068668578149827,
+      "loss": 1.1688,
+      "step": 8103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015067459541972893,
+      "loss": 1.3688,
+      "step": 8104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015066250406119177,
+      "loss": 1.2,
+      "step": 8105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015065041170612465,
+      "loss": 0.8955,
+      "step": 8106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015063831835476546,
+      "loss": 0.9852,
+      "step": 8107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000150626224007352,
+      "loss": 0.8772,
+      "step": 8108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015061412866412225,
+      "loss": 0.9784,
+      "step": 8109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015060203232531406,
+      "loss": 1.1775,
+      "step": 8110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001505899349911654,
+      "loss": 0.9153,
+      "step": 8111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001505778366619142,
+      "loss": 0.9989,
+      "step": 8112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015056573733779848,
+      "loss": 1.1863,
+      "step": 8113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015055363701905615,
+      "loss": 1.0627,
+      "step": 8114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001505415357059253,
+      "loss": 0.9536,
+      "step": 8115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015052943339864395,
+      "loss": 0.9558,
+      "step": 8116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015051733009745013,
+      "loss": 1.1205,
+      "step": 8117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001505052258025819,
+      "loss": 1.126,
+      "step": 8118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015049312051427737,
+      "loss": 1.4306,
+      "step": 8119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015048101423277465,
+      "loss": 0.8978,
+      "step": 8120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015046890695831188,
+      "loss": 1.1014,
+      "step": 8121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015045679869112722,
+      "loss": 1.0388,
+      "step": 8122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015044468943145877,
+      "loss": 1.0123,
+      "step": 8123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001504325791795448,
+      "loss": 1.2809,
+      "step": 8124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015042046793562348,
+      "loss": 0.9817,
+      "step": 8125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015040835569993306,
+      "loss": 1.1419,
+      "step": 8126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015039624247271176,
+      "loss": 1.101,
+      "step": 8127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015038412825419786,
+      "loss": 1.1805,
+      "step": 8128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015037201304462966,
+      "loss": 0.966,
+      "step": 8129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001503598968442454,
+      "loss": 1.1466,
+      "step": 8130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015034777965328348,
+      "loss": 1.1748,
+      "step": 8131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015033566147198223,
+      "loss": 0.8837,
+      "step": 8132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015032354230058004,
+      "loss": 0.932,
+      "step": 8133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001503114221393152,
+      "loss": 1.0012,
+      "step": 8134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001502993009884262,
+      "loss": 1.0235,
+      "step": 8135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015028717884815144,
+      "loss": 0.8809,
+      "step": 8136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015027505571872936,
+      "loss": 1.2004,
+      "step": 8137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001502629316003984,
+      "loss": 1.0172,
+      "step": 8138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015025080649339704,
+      "loss": 0.9793,
+      "step": 8139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015023868039796385,
+      "loss": 0.8634,
+      "step": 8140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015022655331433727,
+      "loss": 1.0203,
+      "step": 8141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015021442524275588,
+      "loss": 1.1803,
+      "step": 8142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001502022961834582,
+      "loss": 1.1208,
+      "step": 8143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015019016613668286,
+      "loss": 1.2504,
+      "step": 8144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015017803510266844,
+      "loss": 1.1197,
+      "step": 8145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015016590308165352,
+      "loss": 0.9681,
+      "step": 8146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015015377007387683,
+      "loss": 1.0998,
+      "step": 8147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001501416360795769,
+      "loss": 1.1062,
+      "step": 8148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001501295010989925,
+      "loss": 1.0406,
+      "step": 8149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001501173651323623,
+      "loss": 1.055,
+      "step": 8150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015010522817992497,
+      "loss": 0.9745,
+      "step": 8151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001500930902419193,
+      "loss": 1.0029,
+      "step": 8152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015008095131858407,
+      "loss": 0.9102,
+      "step": 8153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015006881141015795,
+      "loss": 1.0713,
+      "step": 8154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001500566705168798,
+      "loss": 1.0385,
+      "step": 8155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015004452863898843,
+      "loss": 1.1079,
+      "step": 8156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015003238577672266,
+      "loss": 1.0096,
+      "step": 8157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015002024193032132,
+      "loss": 1.1113,
+      "step": 8158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00015000809710002334,
+      "loss": 0.9903,
+      "step": 8159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001499959512860675,
+      "loss": 1.118,
+      "step": 8160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014998380448869283,
+      "loss": 1.1662,
+      "step": 8161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014997165670813816,
+      "loss": 1.1018,
+      "step": 8162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014995950794464253,
+      "loss": 1.2994,
+      "step": 8163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001499473581984448,
+      "loss": 1.0867,
+      "step": 8164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014993520746978405,
+      "loss": 1.1099,
+      "step": 8165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014992305575889923,
+      "loss": 1.1396,
+      "step": 8166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014991090306602938,
+      "loss": 0.954,
+      "step": 8167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014989874939141351,
+      "loss": 1.3896,
+      "step": 8168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014988659473529075,
+      "loss": 0.924,
+      "step": 8169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014987443909790014,
+      "loss": 1.1769,
+      "step": 8170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014986228247948076,
+      "loss": 1.0169,
+      "step": 8171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014985012488027177,
+      "loss": 0.9436,
+      "step": 8172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001498379663005123,
+      "loss": 1.0188,
+      "step": 8173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001498258067404415,
+      "loss": 1.0497,
+      "step": 8174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014981364620029856,
+      "loss": 1.1161,
+      "step": 8175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001498014846803227,
+      "loss": 0.9667,
+      "step": 8176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014978932218075302,
+      "loss": 1.4296,
+      "step": 8177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014977715870182894,
+      "loss": 0.9829,
+      "step": 8178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014976499424378955,
+      "loss": 0.9827,
+      "step": 8179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014975282880687422,
+      "loss": 1.0347,
+      "step": 8180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001497406623913222,
+      "loss": 1.2262,
+      "step": 8181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014972849499737286,
+      "loss": 1.1894,
+      "step": 8182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014971632662526545,
+      "loss": 1.0741,
+      "step": 8183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001497041572752394,
+      "loss": 1.0367,
+      "step": 8184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014969198694753398,
+      "loss": 1.0294,
+      "step": 8185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001496798156423887,
+      "loss": 1.094,
+      "step": 8186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001496676433600429,
+      "loss": 1.0247,
+      "step": 8187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.000149655470100736,
+      "loss": 0.9836,
+      "step": 8188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014964329586470752,
+      "loss": 1.047,
+      "step": 8189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014963112065219684,
+      "loss": 0.8035,
+      "step": 8190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014961894446344348,
+      "loss": 1.051,
+      "step": 8191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014960676729868695,
+      "loss": 1.1206,
+      "step": 8192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001495945891581668,
+      "loss": 1.1604,
+      "step": 8193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001495824100421225,
+      "loss": 0.9592,
+      "step": 8194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001495702299507937,
+      "loss": 0.9527,
+      "step": 8195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.0001495580488844199,
+      "loss": 0.9063,
+      "step": 8196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014954586684324078,
+      "loss": 1.0985,
+      "step": 8197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 0.00014953368382749588,
+      "loss": 1.0979,
+      "step": 8198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014952149983742493,
+      "loss": 1.1146,
+      "step": 8199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001495093148732675,
+      "loss": 0.9155,
+      "step": 8200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014949712893526333,
+      "loss": 0.994,
+      "step": 8201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014948494202365208,
+      "loss": 0.897,
+      "step": 8202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014947275413867351,
+      "loss": 1.0732,
+      "step": 8203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014946056528056728,
+      "loss": 0.9817,
+      "step": 8204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014944837544957323,
+      "loss": 1.0926,
+      "step": 8205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001494361846459311,
+      "loss": 1.0506,
+      "step": 8206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014942399286988067,
+      "loss": 0.7378,
+      "step": 8207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014941180012166173,
+      "loss": 0.8784,
+      "step": 8208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001493996064015142,
+      "loss": 0.8511,
+      "step": 8209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001493874117096778,
+      "loss": 1.0752,
+      "step": 8210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001493752160463925,
+      "loss": 1.0701,
+      "step": 8211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014936301941189818,
+      "loss": 1.1652,
+      "step": 8212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001493508218064347,
+      "loss": 1.0856,
+      "step": 8213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014933862323024202,
+      "loss": 1.1532,
+      "step": 8214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014932642368356008,
+      "loss": 1.1228,
+      "step": 8215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014931422316662884,
+      "loss": 1.3513,
+      "step": 8216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014930202167968827,
+      "loss": 0.9522,
+      "step": 8217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014928981922297842,
+      "loss": 1.0869,
+      "step": 8218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014927761579673927,
+      "loss": 1.2505,
+      "step": 8219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014926541140121086,
+      "loss": 0.8767,
+      "step": 8220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014925320603663327,
+      "loss": 1.0317,
+      "step": 8221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001492409997032466,
+      "loss": 1.0188,
+      "step": 8222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014922879240129085,
+      "loss": 1.0805,
+      "step": 8223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014921658413100626,
+      "loss": 1.2275,
+      "step": 8224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001492043748926329,
+      "loss": 0.9959,
+      "step": 8225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014919216468641094,
+      "loss": 1.1683,
+      "step": 8226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014917995351258057,
+      "loss": 1.1671,
+      "step": 8227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014916774137138195,
+      "loss": 1.1833,
+      "step": 8228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014915552826305533,
+      "loss": 1.0073,
+      "step": 8229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014914331418784088,
+      "loss": 1.0663,
+      "step": 8230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001491310991459789,
+      "loss": 0.9202,
+      "step": 8231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014911888313770967,
+      "loss": 0.8512,
+      "step": 8232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014910666616327345,
+      "loss": 1.0594,
+      "step": 8233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014909444822291056,
+      "loss": 0.9479,
+      "step": 8234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014908222931686132,
+      "loss": 0.9144,
+      "step": 8235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001490700094453661,
+      "loss": 1.1834,
+      "step": 8236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001490577886086652,
+      "loss": 1.0081,
+      "step": 8237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014904556680699906,
+      "loss": 0.9768,
+      "step": 8238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001490333440406081,
+      "loss": 1.0423,
+      "step": 8239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014902112030973266,
+      "loss": 1.0659,
+      "step": 8240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014900889561461327,
+      "loss": 1.0991,
+      "step": 8241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014899666995549034,
+      "loss": 0.8191,
+      "step": 8242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014898444333260436,
+      "loss": 1.115,
+      "step": 8243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014897221574619582,
+      "loss": 0.9653,
+      "step": 8244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014895998719650526,
+      "loss": 1.0025,
+      "step": 8245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001489477576837732,
+      "loss": 0.9949,
+      "step": 8246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014893552720824017,
+      "loss": 1.1114,
+      "step": 8247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014892329577014673,
+      "loss": 1.0839,
+      "step": 8248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014891106336973358,
+      "loss": 1.016,
+      "step": 8249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001488988300072412,
+      "loss": 0.8868,
+      "step": 8250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014888659568291032,
+      "loss": 1.3536,
+      "step": 8251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014887436039698153,
+      "loss": 1.1343,
+      "step": 8252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014886212414969553,
+      "loss": 0.9938,
+      "step": 8253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014884988694129295,
+      "loss": 1.0891,
+      "step": 8254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014883764877201457,
+      "loss": 0.9698,
+      "step": 8255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014882540964210106,
+      "loss": 1.0986,
+      "step": 8256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014881316955179318,
+      "loss": 1.0235,
+      "step": 8257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014880092850133174,
+      "loss": 1.2462,
+      "step": 8258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014878868649095744,
+      "loss": 0.9909,
+      "step": 8259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014877644352091112,
+      "loss": 1.1463,
+      "step": 8260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001487641995914336,
+      "loss": 0.9275,
+      "step": 8261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014875195470276574,
+      "loss": 0.9045,
+      "step": 8262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001487397088551483,
+      "loss": 1.0112,
+      "step": 8263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014872746204882227,
+      "loss": 1.1649,
+      "step": 8264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001487152142840285,
+      "loss": 0.9361,
+      "step": 8265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001487029655610079,
+      "loss": 1.1044,
+      "step": 8266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014869071588000142,
+      "loss": 1.0902,
+      "step": 8267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014867846524125,
+      "loss": 1.0113,
+      "step": 8268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001486662136449946,
+      "loss": 1.1295,
+      "step": 8269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001486539610914762,
+      "loss": 1.0506,
+      "step": 8270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014864170758093582,
+      "loss": 1.3095,
+      "step": 8271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001486294531136145,
+      "loss": 0.9787,
+      "step": 8272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014861719768975327,
+      "loss": 0.9913,
+      "step": 8273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014860494130959322,
+      "loss": 0.9387,
+      "step": 8274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014859268397337538,
+      "loss": 0.8239,
+      "step": 8275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014858042568134086,
+      "loss": 0.9437,
+      "step": 8276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014856816643373083,
+      "loss": 0.9945,
+      "step": 8277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001485559062307864,
+      "loss": 1.0763,
+      "step": 8278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014854364507274875,
+      "loss": 1.0936,
+      "step": 8279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000148531382959859,
+      "loss": 1.1221,
+      "step": 8280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001485191198923584,
+      "loss": 1.0944,
+      "step": 8281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001485068558704881,
+      "loss": 1.044,
+      "step": 8282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014849459089448942,
+      "loss": 0.9736,
+      "step": 8283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014848232496460356,
+      "loss": 1.2756,
+      "step": 8284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014847005808107185,
+      "loss": 0.8428,
+      "step": 8285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014845779024413543,
+      "loss": 1.0417,
+      "step": 8286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014844552145403577,
+      "loss": 1.342,
+      "step": 8287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014843325171101413,
+      "loss": 0.9789,
+      "step": 8288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014842098101531186,
+      "loss": 1.0014,
+      "step": 8289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001484087093671703,
+      "loss": 1.1461,
+      "step": 8290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001483964367668309,
+      "loss": 0.9353,
+      "step": 8291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014838416321453495,
+      "loss": 0.9413,
+      "step": 8292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000148371888710524,
+      "loss": 0.9822,
+      "step": 8293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014835961325503937,
+      "loss": 0.9091,
+      "step": 8294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014834733684832265,
+      "loss": 1.0169,
+      "step": 8295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014833505949061516,
+      "loss": 0.803,
+      "step": 8296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014832278118215853,
+      "loss": 0.9388,
+      "step": 8297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001483105019231942,
+      "loss": 0.9608,
+      "step": 8298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014829822171396375,
+      "loss": 0.9673,
+      "step": 8299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014828594055470867,
+      "loss": 0.938,
+      "step": 8300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001482736584456706,
+      "loss": 0.8625,
+      "step": 8301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014826137538709105,
+      "loss": 1.1199,
+      "step": 8302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001482490913792117,
+      "loss": 0.9999,
+      "step": 8303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014823680642227417,
+      "loss": 1.0263,
+      "step": 8304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014822452051652002,
+      "loss": 1.0356,
+      "step": 8305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014821223366219102,
+      "loss": 0.9638,
+      "step": 8306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014819994585952878,
+      "loss": 1.2174,
+      "step": 8307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014818765710877506,
+      "loss": 1.0878,
+      "step": 8308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014817536741017152,
+      "loss": 0.8828,
+      "step": 8309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014816307676395995,
+      "loss": 1.2105,
+      "step": 8310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014815078517038208,
+      "loss": 0.9953,
+      "step": 8311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001481384926296797,
+      "loss": 1.0617,
+      "step": 8312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014812619914209452,
+      "loss": 1.0179,
+      "step": 8313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001481139047078685,
+      "loss": 1.1347,
+      "step": 8314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014810160932724338,
+      "loss": 1.1971,
+      "step": 8315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014808931300046098,
+      "loss": 1.0083,
+      "step": 8316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014807701572776323,
+      "loss": 1.3426,
+      "step": 8317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014806471750939207,
+      "loss": 0.9234,
+      "step": 8318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014805241834558926,
+      "loss": 1.0242,
+      "step": 8319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014804011823659684,
+      "loss": 0.9667,
+      "step": 8320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014802781718265668,
+      "loss": 1.095,
+      "step": 8321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014801551518401078,
+      "loss": 1.12,
+      "step": 8322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014800321224090114,
+      "loss": 0.9338,
+      "step": 8323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014799090835356971,
+      "loss": 1.022,
+      "step": 8324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014797860352225853,
+      "loss": 1.1383,
+      "step": 8325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014796629774720965,
+      "loss": 0.9794,
+      "step": 8326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001479539910286651,
+      "loss": 1.1823,
+      "step": 8327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014794168336686693,
+      "loss": 1.0181,
+      "step": 8328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001479293747620573,
+      "loss": 0.9048,
+      "step": 8329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014791706521447828,
+      "loss": 1.1469,
+      "step": 8330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014790475472437203,
+      "loss": 1.1004,
+      "step": 8331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001478924432919806,
+      "loss": 1.0749,
+      "step": 8332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014788013091754628,
+      "loss": 0.9113,
+      "step": 8333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014786781760131118,
+      "loss": 1.1695,
+      "step": 8334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014785550334351757,
+      "loss": 0.901,
+      "step": 8335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014784318814440754,
+      "loss": 1.004,
+      "step": 8336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014783087200422344,
+      "loss": 1.1232,
+      "step": 8337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014781855492320754,
+      "loss": 1.1971,
+      "step": 8338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014780623690160201,
+      "loss": 0.965,
+      "step": 8339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014779391793964925,
+      "loss": 1.0853,
+      "step": 8340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014778159803759158,
+      "loss": 1.0334,
+      "step": 8341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001477692771956712,
+      "loss": 0.8872,
+      "step": 8342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014775695541413062,
+      "loss": 1.0434,
+      "step": 8343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014774463269321209,
+      "loss": 0.943,
+      "step": 8344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014773230903315803,
+      "loss": 1.0867,
+      "step": 8345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014771998443421092,
+      "loss": 1.2352,
+      "step": 8346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001477076588966131,
+      "loss": 1.0839,
+      "step": 8347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014769533242060703,
+      "loss": 0.9843,
+      "step": 8348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014768300500643517,
+      "loss": 1.0718,
+      "step": 8349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014767067665434004,
+      "loss": 1.0672,
+      "step": 8350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001476583473645641,
+      "loss": 0.7545,
+      "step": 8351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014764601713734984,
+      "loss": 1.1689,
+      "step": 8352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014763368597293984,
+      "loss": 1.136,
+      "step": 8353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014762135387157667,
+      "loss": 1.0259,
+      "step": 8354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014760902083350288,
+      "loss": 0.8991,
+      "step": 8355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000147596686858961,
+      "loss": 1.0621,
+      "step": 8356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014758435194819376,
+      "loss": 0.8939,
+      "step": 8357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014757201610144372,
+      "loss": 1.2175,
+      "step": 8358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014755967931895348,
+      "loss": 1.3294,
+      "step": 8359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001475473416009658,
+      "loss": 0.9458,
+      "step": 8360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014753500294772327,
+      "loss": 1.2376,
+      "step": 8361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014752266335946865,
+      "loss": 1.1995,
+      "step": 8362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014751032283644465,
+      "loss": 0.993,
+      "step": 8363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014749798137889402,
+      "loss": 1.033,
+      "step": 8364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014748563898705946,
+      "loss": 0.9458,
+      "step": 8365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001474732956611838,
+      "loss": 1.1443,
+      "step": 8366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014746095140150985,
+      "loss": 1.0576,
+      "step": 8367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014744860620828033,
+      "loss": 1.1697,
+      "step": 8368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014743626008173816,
+      "loss": 1.2382,
+      "step": 8369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014742391302212618,
+      "loss": 1.0279,
+      "step": 8370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014741156502968717,
+      "loss": 0.9779,
+      "step": 8371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014739921610466413,
+      "loss": 0.9524,
+      "step": 8372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014738686624729986,
+      "loss": 0.872,
+      "step": 8373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014737451545783739,
+      "loss": 1.2002,
+      "step": 8374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014736216373651954,
+      "loss": 1.3103,
+      "step": 8375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014734981108358936,
+      "loss": 1.2623,
+      "step": 8376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001473374574992898,
+      "loss": 0.8853,
+      "step": 8377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014732510298386383,
+      "loss": 1.1765,
+      "step": 8378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014731274753755448,
+      "loss": 0.9994,
+      "step": 8379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014730039116060478,
+      "loss": 1.1399,
+      "step": 8380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014728803385325782,
+      "loss": 1.1031,
+      "step": 8381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014727567561575658,
+      "loss": 1.1794,
+      "step": 8382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014726331644834424,
+      "loss": 1.3845,
+      "step": 8383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014725095635126383,
+      "loss": 1.1333,
+      "step": 8384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001472385953247585,
+      "loss": 1.0468,
+      "step": 8385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001472262333690714,
+      "loss": 0.9379,
+      "step": 8386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014721387048444572,
+      "loss": 0.9878,
+      "step": 8387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001472015066711246,
+      "loss": 0.9747,
+      "step": 8388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001471891419293512,
+      "loss": 1.1117,
+      "step": 8389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014717677625936877,
+      "loss": 0.9841,
+      "step": 8390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014716440966142057,
+      "loss": 0.9095,
+      "step": 8391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001471520421357498,
+      "loss": 1.198,
+      "step": 8392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001471396736825998,
+      "loss": 0.9411,
+      "step": 8393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014712730430221376,
+      "loss": 1.1243,
+      "step": 8394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014711493399483506,
+      "loss": 0.9814,
+      "step": 8395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.000147102562760707,
+      "loss": 0.9621,
+      "step": 8396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014709019060007286,
+      "loss": 0.9823,
+      "step": 8397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014707781751317612,
+      "loss": 1.0223,
+      "step": 8398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001470654435002601,
+      "loss": 0.9857,
+      "step": 8399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014705306856156822,
+      "loss": 1.0949,
+      "step": 8400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014704069269734385,
+      "loss": 1.0971,
+      "step": 8401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014702831590783041,
+      "loss": 1.1261,
+      "step": 8402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014701593819327143,
+      "loss": 1.1368,
+      "step": 8403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014700355955391033,
+      "loss": 1.3127,
+      "step": 8404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014699117998999054,
+      "loss": 0.9961,
+      "step": 8405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014697879950175568,
+      "loss": 0.9728,
+      "step": 8406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001469664180894492,
+      "loss": 1.2271,
+      "step": 8407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001469540357533147,
+      "loss": 1.2888,
+      "step": 8408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014694165249359565,
+      "loss": 1.0852,
+      "step": 8409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014692926831053575,
+      "loss": 0.869,
+      "step": 8410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014691688320437848,
+      "loss": 0.8346,
+      "step": 8411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014690449717536753,
+      "loss": 0.9345,
+      "step": 8412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014689211022374645,
+      "loss": 0.9816,
+      "step": 8413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014687972234975896,
+      "loss": 1.0105,
+      "step": 8414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014686733355364877,
+      "loss": 1.1697,
+      "step": 8415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014685494383565948,
+      "loss": 1.2613,
+      "step": 8416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014684255319603483,
+      "loss": 0.8703,
+      "step": 8417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014683016163501855,
+      "loss": 1.1433,
+      "step": 8418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014681776915285435,
+      "loss": 1.0332,
+      "step": 8419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014680537574978604,
+      "loss": 1.1039,
+      "step": 8420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014679298142605734,
+      "loss": 1.0416,
+      "step": 8421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001467805861819121,
+      "loss": 0.9451,
+      "step": 8422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001467681900175941,
+      "loss": 0.8691,
+      "step": 8423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001467557929333472,
+      "loss": 1.1512,
+      "step": 8424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014674339492941522,
+      "loss": 1.0932,
+      "step": 8425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014673099600604202,
+      "loss": 1.2842,
+      "step": 8426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014671859616347156,
+      "loss": 1.1709,
+      "step": 8427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00014670619540194766,
+      "loss": 0.9878,
+      "step": 8428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001466937937217143,
+      "loss": 1.0706,
+      "step": 8429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014668139112301536,
+      "loss": 1.0465,
+      "step": 8430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014666898760609486,
+      "loss": 1.2535,
+      "step": 8431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014665658317119674,
+      "loss": 1.1892,
+      "step": 8432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014664417781856502,
+      "loss": 1.0407,
+      "step": 8433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001466317715484437,
+      "loss": 1.019,
+      "step": 8434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001466193643610768,
+      "loss": 1.1209,
+      "step": 8435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014660695625670836,
+      "loss": 1.054,
+      "step": 8436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014659454723558248,
+      "loss": 0.9964,
+      "step": 8437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014658213729794325,
+      "loss": 0.9721,
+      "step": 8438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014656972644403474,
+      "loss": 1.0262,
+      "step": 8439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001465573146741011,
+      "loss": 0.9803,
+      "step": 8440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014654490198838643,
+      "loss": 1.0643,
+      "step": 8441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014653248838713493,
+      "loss": 1.1484,
+      "step": 8442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014652007387059077,
+      "loss": 1.0832,
+      "step": 8443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001465076584389981,
+      "loss": 1.1277,
+      "step": 8444
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014649524209260116,
+      "loss": 1.3075,
+      "step": 8445
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014648282483164423,
+      "loss": 1.174,
+      "step": 8446
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014647040665637147,
+      "loss": 0.8684,
+      "step": 8447
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001464579875670272,
+      "loss": 1.0471,
+      "step": 8448
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014644556756385565,
+      "loss": 0.9978,
+      "step": 8449
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001464331466471012,
+      "loss": 1.0521,
+      "step": 8450
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014642072481700811,
+      "loss": 1.2184,
+      "step": 8451
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014640830207382074,
+      "loss": 1.0247,
+      "step": 8452
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014639587841778343,
+      "loss": 1.0938,
+      "step": 8453
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014638345384914057,
+      "loss": 1.0665,
+      "step": 8454
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001463710283681365,
+      "loss": 1.0521,
+      "step": 8455
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014635860197501573,
+      "loss": 0.9382,
+      "step": 8456
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014634617467002258,
+      "loss": 1.0489,
+      "step": 8457
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001463337464534016,
+      "loss": 0.8914,
+      "step": 8458
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014632131732539714,
+      "loss": 1.1481,
+      "step": 8459
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014630888728625377,
+      "loss": 1.181,
+      "step": 8460
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001462964563362159,
+      "loss": 0.9286,
+      "step": 8461
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014628402447552816,
+      "loss": 1.1307,
+      "step": 8462
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014627159170443502,
+      "loss": 1.0863,
+      "step": 8463
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014625915802318103,
+      "loss": 1.0622,
+      "step": 8464
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014624672343201077,
+      "loss": 0.8324,
+      "step": 8465
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001462342879311688,
+      "loss": 0.9595,
+      "step": 8466
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014622185152089978,
+      "loss": 1.066,
+      "step": 8467
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014620941420144827,
+      "loss": 1.0664,
+      "step": 8468
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014619697597305898,
+      "loss": 1.0037,
+      "step": 8469
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014618453683597655,
+      "loss": 1.1219,
+      "step": 8470
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001461720967904456,
+      "loss": 0.9307,
+      "step": 8471
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001461596558367109,
+      "loss": 1.0413,
+      "step": 8472
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014614721397501713,
+      "loss": 1.078,
+      "step": 8473
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014613477120560902,
+      "loss": 0.8784,
+      "step": 8474
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014612232752873132,
+      "loss": 1.0866,
+      "step": 8475
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001461098829446288,
+      "loss": 1.1723,
+      "step": 8476
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014609743745354624,
+      "loss": 1.1013,
+      "step": 8477
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014608499105572844,
+      "loss": 1.1092,
+      "step": 8478
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014607254375142027,
+      "loss": 1.2423,
+      "step": 8479
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014606009554086647,
+      "loss": 1.0066,
+      "step": 8480
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014604764642431198,
+      "loss": 1.0489,
+      "step": 8481
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014603519640200163,
+      "loss": 1.0889,
+      "step": 8482
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014602274547418034,
+      "loss": 0.9702,
+      "step": 8483
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000146010293641093,
+      "loss": 1.0464,
+      "step": 8484
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001459978409029845,
+      "loss": 1.0323,
+      "step": 8485
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014598538726009992,
+      "loss": 0.8237,
+      "step": 8486
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014597293271268407,
+      "loss": 1.096,
+      "step": 8487
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014596047726098197,
+      "loss": 0.9655,
+      "step": 8488
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001459480209052387,
+      "loss": 1.1157,
+      "step": 8489
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001459355636456992,
+      "loss": 1.0233,
+      "step": 8490
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001459231054826085,
+      "loss": 1.0609,
+      "step": 8491
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014591064641621167,
+      "loss": 1.0593,
+      "step": 8492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014589818644675378,
+      "loss": 0.976,
+      "step": 8493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014588572557447992,
+      "loss": 0.8918,
+      "step": 8494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001458732637996352,
+      "loss": 1.0752,
+      "step": 8495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014586080112246474,
+      "loss": 1.2322,
+      "step": 8496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014584833754321367,
+      "loss": 0.9977,
+      "step": 8497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014583587306212713,
+      "loss": 1.1361,
+      "step": 8498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014582340767945034,
+      "loss": 1.142,
+      "step": 8499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001458109413954285,
+      "loss": 1.0112,
+      "step": 8500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014579847421030678,
+      "loss": 1.1298,
+      "step": 8501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001457860061243304,
+      "loss": 1.0616,
+      "step": 8502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014577353713774466,
+      "loss": 1.0798,
+      "step": 8503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014576106725079478,
+      "loss": 0.9672,
+      "step": 8504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014574859646372605,
+      "loss": 0.8769,
+      "step": 8505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001457361247767838,
+      "loss": 1.1413,
+      "step": 8506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014572365219021332,
+      "loss": 1.0155,
+      "step": 8507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014571117870425991,
+      "loss": 1.093,
+      "step": 8508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014569870431916903,
+      "loss": 1.0983,
+      "step": 8509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014568622903518596,
+      "loss": 1.1479,
+      "step": 8510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014567375285255612,
+      "loss": 0.9704,
+      "step": 8511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001456612757715249,
+      "loss": 0.8973,
+      "step": 8512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014564879779233774,
+      "loss": 0.807,
+      "step": 8513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014563631891524007,
+      "loss": 1.0909,
+      "step": 8514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014562383914047734,
+      "loss": 0.8635,
+      "step": 8515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014561135846829506,
+      "loss": 0.8347,
+      "step": 8516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001455988768989387,
+      "loss": 1.276,
+      "step": 8517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001455863944326538,
+      "loss": 1.0005,
+      "step": 8518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014557391106968584,
+      "loss": 1.2312,
+      "step": 8519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001455614268102804,
+      "loss": 1.1228,
+      "step": 8520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014554894165468306,
+      "loss": 1.1237,
+      "step": 8521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014553645560313932,
+      "loss": 0.8754,
+      "step": 8522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014552396865589494,
+      "loss": 0.8545,
+      "step": 8523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014551148081319535,
+      "loss": 1.0114,
+      "step": 8524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014549899207528632,
+      "loss": 1.1222,
+      "step": 8525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014548650244241348,
+      "loss": 1.019,
+      "step": 8526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014547401191482244,
+      "loss": 1.1552,
+      "step": 8527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014546152049275891,
+      "loss": 1.1192,
+      "step": 8528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014544902817646867,
+      "loss": 0.8688,
+      "step": 8529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014543653496619738,
+      "loss": 1.0911,
+      "step": 8530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014542404086219076,
+      "loss": 1.1706,
+      "step": 8531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001454115458646946,
+      "loss": 1.1649,
+      "step": 8532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014539904997395468,
+      "loss": 1.0862,
+      "step": 8533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014538655319021679,
+      "loss": 1.0022,
+      "step": 8534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014537405551372673,
+      "loss": 1.2124,
+      "step": 8535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014536155694473037,
+      "loss": 1.0415,
+      "step": 8536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001453490574834735,
+      "loss": 0.928,
+      "step": 8537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014533655713020203,
+      "loss": 1.0365,
+      "step": 8538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001453240558851618,
+      "loss": 1.1375,
+      "step": 8539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001453115537485987,
+      "loss": 0.9995,
+      "step": 8540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014529905072075872,
+      "loss": 1.0638,
+      "step": 8541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014528654680188774,
+      "loss": 1.1038,
+      "step": 8542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014527404199223172,
+      "loss": 1.0885,
+      "step": 8543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001452615362920366,
+      "loss": 1.3036,
+      "step": 8544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014524902970154842,
+      "loss": 1.0524,
+      "step": 8545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001452365222210132,
+      "loss": 1.1988,
+      "step": 8546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014522401385067688,
+      "loss": 0.9759,
+      "step": 8547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014521150459078557,
+      "loss": 0.8397,
+      "step": 8548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014519899444158526,
+      "loss": 0.9683,
+      "step": 8549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001451864834033221,
+      "loss": 1.0425,
+      "step": 8550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014517397147624213,
+      "loss": 0.8446,
+      "step": 8551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014516145866059146,
+      "loss": 0.9259,
+      "step": 8552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014514894495661627,
+      "loss": 1.0394,
+      "step": 8553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014513643036456262,
+      "loss": 0.8661,
+      "step": 8554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014512391488467676,
+      "loss": 1.1562,
+      "step": 8555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001451113985172048,
+      "loss": 1.083,
+      "step": 8556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014509888126239297,
+      "loss": 0.9826,
+      "step": 8557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014508636312048745,
+      "loss": 1.0277,
+      "step": 8558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014507384409173456,
+      "loss": 1.1914,
+      "step": 8559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014506132417638044,
+      "loss": 1.0183,
+      "step": 8560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014504880337467145,
+      "loss": 1.1782,
+      "step": 8561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001450362816868538,
+      "loss": 1.3118,
+      "step": 8562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001450237591131738,
+      "loss": 1.2412,
+      "step": 8563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001450112356538778,
+      "loss": 1.1367,
+      "step": 8564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014499871130921213,
+      "loss": 1.2217,
+      "step": 8565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014498618607942313,
+      "loss": 1.0952,
+      "step": 8566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014497365996475717,
+      "loss": 0.8654,
+      "step": 8567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014496113296546067,
+      "loss": 1.0538,
+      "step": 8568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014494860508178,
+      "loss": 1.0497,
+      "step": 8569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001449360763139616,
+      "loss": 0.8422,
+      "step": 8570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014492354666225193,
+      "loss": 0.9864,
+      "step": 8571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014491101612689738,
+      "loss": 1.0119,
+      "step": 8572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014489848470814447,
+      "loss": 1.1689,
+      "step": 8573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001448859524062397,
+      "loss": 1.0121,
+      "step": 8574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014487341922142955,
+      "loss": 0.9941,
+      "step": 8575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014486088515396065,
+      "loss": 0.9336,
+      "step": 8576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014484835020407935,
+      "loss": 0.9935,
+      "step": 8577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001448358143720324,
+      "loss": 1.1881,
+      "step": 8578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001448232776580663,
+      "loss": 1.0444,
+      "step": 8579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014481074006242764,
+      "loss": 0.8505,
+      "step": 8580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014479820158536306,
+      "loss": 1.0551,
+      "step": 8581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014478566222711917,
+      "loss": 0.975,
+      "step": 8582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001447731219879426,
+      "loss": 0.8537,
+      "step": 8583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014476058086808008,
+      "loss": 1.2203,
+      "step": 8584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014474803886777824,
+      "loss": 1.1104,
+      "step": 8585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014473549598728383,
+      "loss": 1.3031,
+      "step": 8586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014472295222684348,
+      "loss": 0.9897,
+      "step": 8587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000144710407586704,
+      "loss": 1.0908,
+      "step": 8588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014469786206711214,
+      "loss": 1.0321,
+      "step": 8589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014468531566831462,
+      "loss": 1.1984,
+      "step": 8590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014467276839055828,
+      "loss": 1.0827,
+      "step": 8591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001446602202340899,
+      "loss": 0.9989,
+      "step": 8592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014464767119915629,
+      "loss": 1.1363,
+      "step": 8593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014463512128600435,
+      "loss": 1.0507,
+      "step": 8594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014462257049488087,
+      "loss": 1.0479,
+      "step": 8595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014461001882603273,
+      "loss": 0.9209,
+      "step": 8596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014459746627970685,
+      "loss": 1.0858,
+      "step": 8597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001445849128561501,
+      "loss": 1.191,
+      "step": 8598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014457235855560948,
+      "loss": 1.0363,
+      "step": 8599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014455980337833185,
+      "loss": 1.0694,
+      "step": 8600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014454724732456423,
+      "loss": 0.9892,
+      "step": 8601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014453469039455357,
+      "loss": 1.1263,
+      "step": 8602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014452213258854684,
+      "loss": 1.1041,
+      "step": 8603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014450957390679106,
+      "loss": 1.0276,
+      "step": 8604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014449701434953333,
+      "loss": 1.1303,
+      "step": 8605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001444844539170206,
+      "loss": 1.1257,
+      "step": 8606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014447189260949998,
+      "loss": 1.1027,
+      "step": 8607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014445933042721857,
+      "loss": 1.0045,
+      "step": 8608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014444676737042342,
+      "loss": 0.8141,
+      "step": 8609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014443420343936164,
+      "loss": 1.0353,
+      "step": 8610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014442163863428043,
+      "loss": 1.099,
+      "step": 8611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014440907295542693,
+      "loss": 1.1503,
+      "step": 8612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014439650640304822,
+      "loss": 1.0745,
+      "step": 8613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014438393897739154,
+      "loss": 1.2278,
+      "step": 8614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014437137067870414,
+      "loss": 1.0434,
+      "step": 8615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014435880150723314,
+      "loss": 0.9646,
+      "step": 8616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014434623146322587,
+      "loss": 0.9736,
+      "step": 8617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001443336605469295,
+      "loss": 1.0556,
+      "step": 8618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014432108875859134,
+      "loss": 1.1171,
+      "step": 8619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014430851609845868,
+      "loss": 1.0134,
+      "step": 8620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014429594256677882,
+      "loss": 0.9015,
+      "step": 8621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014428336816379908,
+      "loss": 1.105,
+      "step": 8622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014427079288976678,
+      "loss": 0.8967,
+      "step": 8623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001442582167449293,
+      "loss": 1.0624,
+      "step": 8624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000144245639729534,
+      "loss": 1.0897,
+      "step": 8625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014423306184382825,
+      "loss": 1.1053,
+      "step": 8626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014422048308805952,
+      "loss": 1.157,
+      "step": 8627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001442079034624752,
+      "loss": 1.0295,
+      "step": 8628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001441953229673227,
+      "loss": 0.8721,
+      "step": 8629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001441827416028495,
+      "loss": 1.0303,
+      "step": 8630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014417015936930307,
+      "loss": 1.1656,
+      "step": 8631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014415757626693094,
+      "loss": 1.005,
+      "step": 8632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014414499229598054,
+      "loss": 1.2068,
+      "step": 8633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014413240745669948,
+      "loss": 0.951,
+      "step": 8634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014411982174933528,
+      "loss": 0.9956,
+      "step": 8635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014410723517413546,
+      "loss": 0.9096,
+      "step": 8636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014409464773134764,
+      "loss": 0.8731,
+      "step": 8637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014408205942121943,
+      "loss": 1.0361,
+      "step": 8638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014406947024399834,
+      "loss": 1.049,
+      "step": 8639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014405688019993214,
+      "loss": 1.151,
+      "step": 8640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001440442892892684,
+      "loss": 1.1652,
+      "step": 8641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014403169751225475,
+      "loss": 1.2933,
+      "step": 8642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014401910486913893,
+      "loss": 0.994,
+      "step": 8643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001440065113601686,
+      "loss": 1.1702,
+      "step": 8644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014399391698559152,
+      "loss": 0.9209,
+      "step": 8645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014398132174565538,
+      "loss": 1.1835,
+      "step": 8646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014396872564060795,
+      "loss": 1.0154,
+      "step": 8647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014395612867069697,
+      "loss": 1.1448,
+      "step": 8648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014394353083617024,
+      "loss": 1.1718,
+      "step": 8649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014393093213727556,
+      "loss": 1.1816,
+      "step": 8650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014391833257426076,
+      "loss": 0.9556,
+      "step": 8651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014390573214737362,
+      "loss": 1.0843,
+      "step": 8652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014389313085686205,
+      "loss": 1.052,
+      "step": 8653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014388052870297386,
+      "loss": 1.0931,
+      "step": 8654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014386792568595703,
+      "loss": 0.953,
+      "step": 8655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001438553218060593,
+      "loss": 1.1186,
+      "step": 8656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001438427170635288,
+      "loss": 1.0928,
+      "step": 8657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014383011145861326,
+      "loss": 1.23,
+      "step": 8658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00014381750499156074,
+      "loss": 1.0577,
+      "step": 8659
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001438048976626192,
+      "loss": 0.8177,
+      "step": 8660
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014379228947203665,
+      "loss": 0.9205,
+      "step": 8661
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014377968042006103,
+      "loss": 1.2063,
+      "step": 8662
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001437670705069404,
+      "loss": 0.9585,
+      "step": 8663
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014375445973292274,
+      "loss": 1.0751,
+      "step": 8664
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001437418480982562,
+      "loss": 1.1675,
+      "step": 8665
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014372923560318875,
+      "loss": 1.0372,
+      "step": 8666
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014371662224796857,
+      "loss": 1.0465,
+      "step": 8667
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014370400803284374,
+      "loss": 0.8532,
+      "step": 8668
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001436913929580623,
+      "loss": 0.9705,
+      "step": 8669
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001436787770238725,
+      "loss": 1.2483,
+      "step": 8670
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014366616023052243,
+      "loss": 0.7814,
+      "step": 8671
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014365354257826028,
+      "loss": 1.0018,
+      "step": 8672
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001436409240673342,
+      "loss": 1.1436,
+      "step": 8673
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014362830469799248,
+      "loss": 1.0689,
+      "step": 8674
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001436156844704833,
+      "loss": 0.8138,
+      "step": 8675
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014360306338505484,
+      "loss": 1.3342,
+      "step": 8676
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014359044144195547,
+      "loss": 1.1243,
+      "step": 8677
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014357781864143338,
+      "loss": 1.2203,
+      "step": 8678
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014356519498373685,
+      "loss": 1.0781,
+      "step": 8679
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014355257046911427,
+      "loss": 1.1507,
+      "step": 8680
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001435399450978139,
+      "loss": 0.9253,
+      "step": 8681
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014352731887008407,
+      "loss": 1.1718,
+      "step": 8682
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001435146917861732,
+      "loss": 1.302,
+      "step": 8683
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001435020638463296,
+      "loss": 1.1406,
+      "step": 8684
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014348943505080173,
+      "loss": 1.0935,
+      "step": 8685
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014347680539983793,
+      "loss": 1.158,
+      "step": 8686
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014346417489368663,
+      "loss": 0.8948,
+      "step": 8687
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014345154353259632,
+      "loss": 1.2383,
+      "step": 8688
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001434389113168154,
+      "loss": 1.2673,
+      "step": 8689
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014342627824659243,
+      "loss": 1.1692,
+      "step": 8690
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014341364432217583,
+      "loss": 1.0584,
+      "step": 8691
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014340100954381407,
+      "loss": 1.1401,
+      "step": 8692
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014338837391175582,
+      "loss": 0.8413,
+      "step": 8693
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001433757374262495,
+      "loss": 0.9417,
+      "step": 8694
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014336310008754372,
+      "loss": 1.0603,
+      "step": 8695
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000143350461895887,
+      "loss": 1.0003,
+      "step": 8696
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014333782285152802,
+      "loss": 1.271,
+      "step": 8697
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014332518295471532,
+      "loss": 1.37,
+      "step": 8698
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014331254220569756,
+      "loss": 1.1169,
+      "step": 8699
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014329990060472336,
+      "loss": 1.0842,
+      "step": 8700
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014328725815204144,
+      "loss": 1.0084,
+      "step": 8701
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014327461484790033,
+      "loss": 1.1753,
+      "step": 8702
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001432619706925489,
+      "loss": 0.9389,
+      "step": 8703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001432493256862358,
+      "loss": 1.0976,
+      "step": 8704
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001432366798292097,
+      "loss": 1.0017,
+      "step": 8705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014322403312171938,
+      "loss": 0.8909,
+      "step": 8706
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014321138556401363,
+      "loss": 1.1346,
+      "step": 8707
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001431987371563412,
+      "loss": 1.0448,
+      "step": 8708
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014318608789895088,
+      "loss": 1.0852,
+      "step": 8709
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014317343779209147,
+      "loss": 0.8925,
+      "step": 8710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014316078683601182,
+      "loss": 1.2447,
+      "step": 8711
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014314813503096078,
+      "loss": 0.9342,
+      "step": 8712
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001431354823771872,
+      "loss": 0.7806,
+      "step": 8713
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014312282887493996,
+      "loss": 1.2371,
+      "step": 8714
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001431101745244679,
+      "loss": 1.0256,
+      "step": 8715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014309751932602006,
+      "loss": 0.9426,
+      "step": 8716
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014308486327984523,
+      "loss": 1.2115,
+      "step": 8717
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014307220638619243,
+      "loss": 1.0551,
+      "step": 8718
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014305954864531057,
+      "loss": 0.9655,
+      "step": 8719
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014304689005744872,
+      "loss": 1.0338,
+      "step": 8720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014303423062285577,
+      "loss": 0.9374,
+      "step": 8721
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001430215703417808,
+      "loss": 0.9141,
+      "step": 8722
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014300890921447275,
+      "loss": 0.9158,
+      "step": 8723
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014299624724118081,
+      "loss": 1.008,
+      "step": 8724
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014298358442215388,
+      "loss": 1.0569,
+      "step": 8725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001429709207576412,
+      "loss": 1.1275,
+      "step": 8726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014295825624789172,
+      "loss": 1.1176,
+      "step": 8727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001429455908931546,
+      "loss": 0.8641,
+      "step": 8728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014293292469367902,
+      "loss": 1.1283,
+      "step": 8729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014292025764971403,
+      "loss": 1.0378,
+      "step": 8730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014290758976150887,
+      "loss": 0.8921,
+      "step": 8731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014289492102931268,
+      "loss": 1.2128,
+      "step": 8732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014288225145337466,
+      "loss": 1.1574,
+      "step": 8733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000142869581033944,
+      "loss": 0.8551,
+      "step": 8734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014285690977126997,
+      "loss": 1.2313,
+      "step": 8735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001428442376656018,
+      "loss": 1.0488,
+      "step": 8736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014283156471718873,
+      "loss": 1.1228,
+      "step": 8737
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014281889092628003,
+      "loss": 0.985,
+      "step": 8738
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000142806216293125,
+      "loss": 1.1344,
+      "step": 8739
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014279354081797298,
+      "loss": 1.1416,
+      "step": 8740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001427808645010733,
+      "loss": 0.9502,
+      "step": 8741
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014276818734267524,
+      "loss": 1.0283,
+      "step": 8742
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014275550934302823,
+      "loss": 1.2472,
+      "step": 8743
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014274283050238158,
+      "loss": 1.0536,
+      "step": 8744
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014273015082098473,
+      "loss": 1.158,
+      "step": 8745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014271747029908708,
+      "loss": 1.1025,
+      "step": 8746
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014270478893693806,
+      "loss": 1.0573,
+      "step": 8747
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001426921067347871,
+      "loss": 1.2187,
+      "step": 8748
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014267942369288364,
+      "loss": 1.0097,
+      "step": 8749
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014266673981147717,
+      "loss": 1.1809,
+      "step": 8750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014265405509081722,
+      "loss": 1.0785,
+      "step": 8751
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014264136953115323,
+      "loss": 1.0526,
+      "step": 8752
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014262868313273482,
+      "loss": 0.9222,
+      "step": 8753
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014261599589581142,
+      "loss": 0.942,
+      "step": 8754
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014260330782063266,
+      "loss": 1.1553,
+      "step": 8755
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014259061890744808,
+      "loss": 1.1275,
+      "step": 8756
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014257792915650728,
+      "loss": 0.9678,
+      "step": 8757
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014256523856805985,
+      "loss": 1.0653,
+      "step": 8758
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014255254714235548,
+      "loss": 1.1077,
+      "step": 8759
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014253985487964376,
+      "loss": 0.9253,
+      "step": 8760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014252716178017434,
+      "loss": 1.037,
+      "step": 8761
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001425144678441969,
+      "loss": 1.0415,
+      "step": 8762
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014250177307196114,
+      "loss": 1.1612,
+      "step": 8763
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014248907746371677,
+      "loss": 0.9493,
+      "step": 8764
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014247638101971344,
+      "loss": 0.9393,
+      "step": 8765
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000142463683740201,
+      "loss": 0.9355,
+      "step": 8766
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014245098562542917,
+      "loss": 1.0428,
+      "step": 8767
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014243828667564768,
+      "loss": 1.0139,
+      "step": 8768
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014242558689110633,
+      "loss": 1.1206,
+      "step": 8769
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014241288627205498,
+      "loss": 1.0184,
+      "step": 8770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014240018481874336,
+      "loss": 1.0437,
+      "step": 8771
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001423874825314214,
+      "loss": 0.9903,
+      "step": 8772
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014237477941033888,
+      "loss": 1.3264,
+      "step": 8773
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014236207545574573,
+      "loss": 0.915,
+      "step": 8774
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014234937066789178,
+      "loss": 1.1181,
+      "step": 8775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014233666504702698,
+      "loss": 1.108,
+      "step": 8776
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014232395859340122,
+      "loss": 0.974,
+      "step": 8777
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001423112513072644,
+      "loss": 1.0981,
+      "step": 8778
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014229854318886657,
+      "loss": 1.344,
+      "step": 8779
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014228583423845766,
+      "loss": 1.0742,
+      "step": 8780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014227312445628759,
+      "loss": 1.0645,
+      "step": 8781
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014226041384260643,
+      "loss": 1.0076,
+      "step": 8782
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001422477023976642,
+      "loss": 1.1465,
+      "step": 8783
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014223499012171086,
+      "loss": 1.092,
+      "step": 8784
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014222227701499656,
+      "loss": 1.1979,
+      "step": 8785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014220956307777128,
+      "loss": 1.1535,
+      "step": 8786
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014219684831028517,
+      "loss": 0.9293,
+      "step": 8787
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014218413271278827,
+      "loss": 1.1287,
+      "step": 8788
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014217141628553078,
+      "loss": 1.0569,
+      "step": 8789
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014215869902876272,
+      "loss": 1.0647,
+      "step": 8790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014214598094273434,
+      "loss": 1.1761,
+      "step": 8791
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001421332620276957,
+      "loss": 1.2805,
+      "step": 8792
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001421205422838971,
+      "loss": 1.0761,
+      "step": 8793
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014210782171158865,
+      "loss": 1.1802,
+      "step": 8794
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001420951003110206,
+      "loss": 0.9068,
+      "step": 8795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014208237808244315,
+      "loss": 1.1589,
+      "step": 8796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014206965502610655,
+      "loss": 0.9682,
+      "step": 8797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014205693114226114,
+      "loss": 0.8399,
+      "step": 8798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001420442064311571,
+      "loss": 0.9817,
+      "step": 8799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014203148089304477,
+      "loss": 0.9817,
+      "step": 8800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014201875452817445,
+      "loss": 1.0385,
+      "step": 8801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014200602733679645,
+      "loss": 1.0492,
+      "step": 8802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014199329931916115,
+      "loss": 1.1029,
+      "step": 8803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001419805704755189,
+      "loss": 0.8408,
+      "step": 8804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014196784080612002,
+      "loss": 1.0278,
+      "step": 8805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000141955110311215,
+      "loss": 0.9801,
+      "step": 8806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001419423789910542,
+      "loss": 1.1738,
+      "step": 8807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000141929646845888,
+      "loss": 1.1732,
+      "step": 8808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014191691387596694,
+      "loss": 1.1392,
+      "step": 8809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001419041800815414,
+      "loss": 1.1705,
+      "step": 8810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014189144546286183,
+      "loss": 0.8523,
+      "step": 8811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014187871002017884,
+      "loss": 0.8688,
+      "step": 8812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001418659737537428,
+      "loss": 0.9832,
+      "step": 8813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014185323666380436,
+      "loss": 1.1009,
+      "step": 8814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014184049875061393,
+      "loss": 1.1159,
+      "step": 8815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014182776001442216,
+      "loss": 0.8527,
+      "step": 8816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014181502045547956,
+      "loss": 0.899,
+      "step": 8817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014180228007403675,
+      "loss": 0.9108,
+      "step": 8818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001417895388703443,
+      "loss": 1.2217,
+      "step": 8819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001417767968446529,
+      "loss": 1.1836,
+      "step": 8820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014176405399721312,
+      "loss": 0.9468,
+      "step": 8821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014175131032827566,
+      "loss": 0.991,
+      "step": 8822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014173856583809112,
+      "loss": 0.9175,
+      "step": 8823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001417258205269102,
+      "loss": 0.9405,
+      "step": 8824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014171307439498368,
+      "loss": 1.0839,
+      "step": 8825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001417003274425622,
+      "loss": 0.9855,
+      "step": 8826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001416875796698965,
+      "loss": 0.9576,
+      "step": 8827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001416748310772373,
+      "loss": 0.9699,
+      "step": 8828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014166208166483548,
+      "loss": 1.0615,
+      "step": 8829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014164933143294173,
+      "loss": 0.8982,
+      "step": 8830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001416365803818068,
+      "loss": 0.884,
+      "step": 8831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001416238285116816,
+      "loss": 0.9787,
+      "step": 8832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014161107582281698,
+      "loss": 0.9887,
+      "step": 8833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014159832231546365,
+      "loss": 1.043,
+      "step": 8834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014158556798987256,
+      "loss": 0.9177,
+      "step": 8835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014157281284629458,
+      "loss": 1.0015,
+      "step": 8836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014156005688498063,
+      "loss": 1.2505,
+      "step": 8837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014154730010618153,
+      "loss": 1.059,
+      "step": 8838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014153454251014834,
+      "loss": 1.1509,
+      "step": 8839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014152178409713187,
+      "loss": 1.1158,
+      "step": 8840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014150902486738314,
+      "loss": 0.9831,
+      "step": 8841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001414962648211531,
+      "loss": 1.0421,
+      "step": 8842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014148350395869278,
+      "loss": 0.9221,
+      "step": 8843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014147074228025317,
+      "loss": 1.0129,
+      "step": 8844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014145797978608527,
+      "loss": 0.9193,
+      "step": 8845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014144521647644015,
+      "loss": 1.1712,
+      "step": 8846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014143245235156884,
+      "loss": 0.8871,
+      "step": 8847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014141968741172238,
+      "loss": 1.1163,
+      "step": 8848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014140692165715192,
+      "loss": 0.9542,
+      "step": 8849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014139415508810857,
+      "loss": 0.9821,
+      "step": 8850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014138138770484332,
+      "loss": 0.9617,
+      "step": 8851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014136861950760746,
+      "loss": 0.93,
+      "step": 8852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014135585049665207,
+      "loss": 1.2688,
+      "step": 8853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014134308067222833,
+      "loss": 1.0772,
+      "step": 8854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001413303100345874,
+      "loss": 0.9944,
+      "step": 8855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001413175385839805,
+      "loss": 1.0751,
+      "step": 8856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001413047663206588,
+      "loss": 1.0034,
+      "step": 8857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014129199324487362,
+      "loss": 1.0593,
+      "step": 8858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001412792193568761,
+      "loss": 1.0629,
+      "step": 8859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001412664446569176,
+      "loss": 0.9086,
+      "step": 8860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014125366914524934,
+      "loss": 1.0207,
+      "step": 8861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014124089282212263,
+      "loss": 0.984,
+      "step": 8862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014122811568778875,
+      "loss": 1.0602,
+      "step": 8863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014121533774249907,
+      "loss": 1.0156,
+      "step": 8864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001412025589865049,
+      "loss": 0.8239,
+      "step": 8865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001411897794200576,
+      "loss": 0.9322,
+      "step": 8866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001411769990434086,
+      "loss": 1.0843,
+      "step": 8867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014116421785680922,
+      "loss": 1.2499,
+      "step": 8868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014115143586051088,
+      "loss": 0.9426,
+      "step": 8869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014113865305476504,
+      "loss": 0.9898,
+      "step": 8870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014112586943982306,
+      "loss": 1.1123,
+      "step": 8871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014111308501593649,
+      "loss": 1.2739,
+      "step": 8872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014110029978335674,
+      "loss": 0.9192,
+      "step": 8873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001410875137423353,
+      "loss": 0.9746,
+      "step": 8874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001410747268931237,
+      "loss": 1.1228,
+      "step": 8875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001410619392359734,
+      "loss": 1.1269,
+      "step": 8876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.000141049150771136,
+      "loss": 1.1361,
+      "step": 8877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014103636149886297,
+      "loss": 1.1685,
+      "step": 8878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014102357141940595,
+      "loss": 1.0911,
+      "step": 8879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001410107805330165,
+      "loss": 1.1034,
+      "step": 8880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001409979888399462,
+      "loss": 0.9845,
+      "step": 8881
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014098519634044666,
+      "loss": 1.0691,
+      "step": 8882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014097240303476954,
+      "loss": 1.1171,
+      "step": 8883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014095960892316644,
+      "loss": 0.8755,
+      "step": 8884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014094681400588906,
+      "loss": 1.0112,
+      "step": 8885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014093401828318907,
+      "loss": 0.9274,
+      "step": 8886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001409212217553181,
+      "loss": 1.1454,
+      "step": 8887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014090842442252794,
+      "loss": 1.0453,
+      "step": 8888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014089562628507026,
+      "loss": 1.1919,
+      "step": 8889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00014088282734319685,
+      "loss": 1.1162,
+      "step": 8890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014087002759715942,
+      "loss": 1.0775,
+      "step": 8891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014085722704720977,
+      "loss": 0.9418,
+      "step": 8892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014084442569359964,
+      "loss": 0.9923,
+      "step": 8893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014083162353658086,
+      "loss": 1.056,
+      "step": 8894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014081882057640527,
+      "loss": 1.0708,
+      "step": 8895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014080601681332468,
+      "loss": 0.9759,
+      "step": 8896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014079321224759093,
+      "loss": 1.0067,
+      "step": 8897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001407804068794559,
+      "loss": 1.1637,
+      "step": 8898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001407676007091715,
+      "loss": 0.9964,
+      "step": 8899
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014075479373698956,
+      "loss": 1.0097,
+      "step": 8900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014074198596316205,
+      "loss": 1.054,
+      "step": 8901
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001407291773879409,
+      "loss": 0.9564,
+      "step": 8902
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000140716368011578,
+      "loss": 1.065,
+      "step": 8903
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014070355783432536,
+      "loss": 1.2527,
+      "step": 8904
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014069074685643494,
+      "loss": 0.974,
+      "step": 8905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014067793507815875,
+      "loss": 1.0225,
+      "step": 8906
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014066512249974873,
+      "loss": 0.9765,
+      "step": 8907
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014065230912145698,
+      "loss": 0.93,
+      "step": 8908
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001406394949435355,
+      "loss": 0.8517,
+      "step": 8909
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014062667996623634,
+      "loss": 1.1666,
+      "step": 8910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001406138641898116,
+      "loss": 1.1345,
+      "step": 8911
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014060104761451337,
+      "loss": 1.0313,
+      "step": 8912
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014058823024059374,
+      "loss": 1.1815,
+      "step": 8913
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014057541206830479,
+      "loss": 1.0759,
+      "step": 8914
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014056259309789867,
+      "loss": 1.0634,
+      "step": 8915
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001405497733296276,
+      "loss": 0.9995,
+      "step": 8916
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001405369527637436,
+      "loss": 0.9438,
+      "step": 8917
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000140524131400499,
+      "loss": 1.1826,
+      "step": 8918
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014051130924014592,
+      "loss": 0.9233,
+      "step": 8919
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014049848628293655,
+      "loss": 1.2458,
+      "step": 8920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001404856625291232,
+      "loss": 1.1924,
+      "step": 8921
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014047283797895803,
+      "loss": 0.8306,
+      "step": 8922
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014046001263269336,
+      "loss": 1.0199,
+      "step": 8923
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014044718649058142,
+      "loss": 0.89,
+      "step": 8924
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014043435955287452,
+      "loss": 1.0811,
+      "step": 8925
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014042153181982497,
+      "loss": 1.117,
+      "step": 8926
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014040870329168505,
+      "loss": 1.0395,
+      "step": 8927
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014039587396870714,
+      "loss": 1.0078,
+      "step": 8928
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014038304385114364,
+      "loss": 0.9137,
+      "step": 8929
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014037021293924676,
+      "loss": 0.9615,
+      "step": 8930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014035738123326905,
+      "loss": 1.0422,
+      "step": 8931
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014034454873346284,
+      "loss": 0.9605,
+      "step": 8932
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014033171544008052,
+      "loss": 1.1865,
+      "step": 8933
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014031888135337454,
+      "loss": 1.1678,
+      "step": 8934
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014030604647359739,
+      "loss": 1.0818,
+      "step": 8935
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001402932108010015,
+      "loss": 1.1397,
+      "step": 8936
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014028037433583932,
+      "loss": 1.0816,
+      "step": 8937
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014026753707836338,
+      "loss": 1.2692,
+      "step": 8938
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014025469902882615,
+      "loss": 1.1021,
+      "step": 8939
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001402418601874802,
+      "loss": 0.9654,
+      "step": 8940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014022902055457802,
+      "loss": 1.2768,
+      "step": 8941
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014021618013037226,
+      "loss": 1.1357,
+      "step": 8942
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014020333891511536,
+      "loss": 0.8875,
+      "step": 8943
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014019049690905999,
+      "loss": 1.026,
+      "step": 8944
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014017765411245873,
+      "loss": 0.7987,
+      "step": 8945
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014016481052556422,
+      "loss": 0.9831,
+      "step": 8946
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014015196614862904,
+      "loss": 1.0854,
+      "step": 8947
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014013912098190592,
+      "loss": 1.1424,
+      "step": 8948
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014012627502564743,
+      "loss": 1.1095,
+      "step": 8949
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014011342828010632,
+      "loss": 1.1479,
+      "step": 8950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014010058074553526,
+      "loss": 1.1951,
+      "step": 8951
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014008773242218692,
+      "loss": 1.0394,
+      "step": 8952
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001400748833103141,
+      "loss": 0.9254,
+      "step": 8953
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014006203341016953,
+      "loss": 1.0362,
+      "step": 8954
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001400491827220059,
+      "loss": 0.9872,
+      "step": 8955
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014003633124607605,
+      "loss": 0.958,
+      "step": 8956
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00014002347898263274,
+      "loss": 1.0551,
+      "step": 8957
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001400106259319288,
+      "loss": 1.0937,
+      "step": 8958
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013999777209421703,
+      "loss": 0.9086,
+      "step": 8959
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013998491746975024,
+      "loss": 1.157,
+      "step": 8960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001399720620587813,
+      "loss": 1.1111,
+      "step": 8961
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001399592058615631,
+      "loss": 1.1413,
+      "step": 8962
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013994634887834847,
+      "loss": 1.167,
+      "step": 8963
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013993349110939037,
+      "loss": 1.0968,
+      "step": 8964
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001399206325549417,
+      "loss": 1.0561,
+      "step": 8965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001399077732152553,
+      "loss": 0.9106,
+      "step": 8966
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013989491309058423,
+      "loss": 1.2559,
+      "step": 8967
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001398820521811814,
+      "loss": 1.0235,
+      "step": 8968
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013986919048729977,
+      "loss": 0.9221,
+      "step": 8969
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013985632800919231,
+      "loss": 0.9718,
+      "step": 8970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001398434647471121,
+      "loss": 1.0086,
+      "step": 8971
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001398306007013121,
+      "loss": 1.0132,
+      "step": 8972
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013981773587204535,
+      "loss": 1.0465,
+      "step": 8973
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001398048702595649,
+      "loss": 1.1343,
+      "step": 8974
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013979200386412386,
+      "loss": 1.0062,
+      "step": 8975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013977913668597523,
+      "loss": 1.1652,
+      "step": 8976
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001397662687253722,
+      "loss": 1.2418,
+      "step": 8977
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013975339998256785,
+      "loss": 1.022,
+      "step": 8978
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013974053045781525,
+      "loss": 1.0615,
+      "step": 8979
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001397276601513676,
+      "loss": 0.987,
+      "step": 8980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013971478906347806,
+      "loss": 1.1218,
+      "step": 8981
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001397019171943998,
+      "loss": 0.9753,
+      "step": 8982
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000139689044544386,
+      "loss": 1.1621,
+      "step": 8983
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013967617111368985,
+      "loss": 1.1195,
+      "step": 8984
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001396632969025646,
+      "loss": 1.1426,
+      "step": 8985
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013965042191126345,
+      "loss": 1.1055,
+      "step": 8986
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001396375461400397,
+      "loss": 0.9346,
+      "step": 8987
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013962466958914658,
+      "loss": 1.4724,
+      "step": 8988
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013961179225883735,
+      "loss": 0.9964,
+      "step": 8989
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013959891414936536,
+      "loss": 1.1551,
+      "step": 8990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001395860352609839,
+      "loss": 1.1556,
+      "step": 8991
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001395731555939463,
+      "loss": 0.9878,
+      "step": 8992
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001395602751485059,
+      "loss": 1.0747,
+      "step": 8993
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013954739392491605,
+      "loss": 0.9387,
+      "step": 8994
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001395345119234301,
+      "loss": 1.1292,
+      "step": 8995
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013952162914430152,
+      "loss": 0.9841,
+      "step": 8996
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013950874558778362,
+      "loss": 1.0436,
+      "step": 8997
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013949586125412987,
+      "loss": 0.9581,
+      "step": 8998
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013948297614359368,
+      "loss": 0.9567,
+      "step": 8999
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013947009025642857,
+      "loss": 1.0,
+      "step": 9000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001394572035928879,
+      "loss": 1.0062,
+      "step": 9001
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013944431615322522,
+      "loss": 1.1329,
+      "step": 9002
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000139431427937694,
+      "loss": 1.0391,
+      "step": 9003
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013941853894654775,
+      "loss": 0.9981,
+      "step": 9004
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013940564918004004,
+      "loss": 0.9385,
+      "step": 9005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013939275863842433,
+      "loss": 0.9856,
+      "step": 9006
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013937986732195421,
+      "loss": 0.9883,
+      "step": 9007
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001393669752308833,
+      "loss": 0.9083,
+      "step": 9008
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013935408236546515,
+      "loss": 1.0729,
+      "step": 9009
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013934118872595334,
+      "loss": 0.7882,
+      "step": 9010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013932829431260153,
+      "loss": 1.1211,
+      "step": 9011
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001393153991256633,
+      "loss": 1.1149,
+      "step": 9012
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013930250316539238,
+      "loss": 0.919,
+      "step": 9013
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013928960643204235,
+      "loss": 0.9194,
+      "step": 9014
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013927670892586696,
+      "loss": 0.9026,
+      "step": 9015
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001392638106471198,
+      "loss": 0.9715,
+      "step": 9016
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001392509115960547,
+      "loss": 1.1295,
+      "step": 9017
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013923801177292527,
+      "loss": 1.0103,
+      "step": 9018
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013922511117798534,
+      "loss": 0.9422,
+      "step": 9019
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013921220981148863,
+      "loss": 0.9608,
+      "step": 9020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013919930767368891,
+      "loss": 0.9147,
+      "step": 9021
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013918640476484,
+      "loss": 1.0543,
+      "step": 9022
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001391735010851956,
+      "loss": 1.1319,
+      "step": 9023
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013916059663500962,
+      "loss": 1.0971,
+      "step": 9024
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013914769141453587,
+      "loss": 1.2013,
+      "step": 9025
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013913478542402818,
+      "loss": 0.8728,
+      "step": 9026
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001391218786637404,
+      "loss": 1.3188,
+      "step": 9027
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013910897113392646,
+      "loss": 0.9789,
+      "step": 9028
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013909606283484018,
+      "loss": 1.071,
+      "step": 9029
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001390831537667355,
+      "loss": 0.9166,
+      "step": 9030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013907024392986636,
+      "loss": 1.3949,
+      "step": 9031
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013905733332448669,
+      "loss": 1.0281,
+      "step": 9032
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013904442195085038,
+      "loss": 1.0332,
+      "step": 9033
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013903150980921147,
+      "loss": 1.0072,
+      "step": 9034
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001390185968998239,
+      "loss": 0.9386,
+      "step": 9035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001390056832229417,
+      "loss": 0.8667,
+      "step": 9036
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013899276877881884,
+      "loss": 1.03,
+      "step": 9037
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001389798535677094,
+      "loss": 1.1125,
+      "step": 9038
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013896693758986734,
+      "loss": 1.2222,
+      "step": 9039
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013895402084554683,
+      "loss": 1.2283,
+      "step": 9040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013894110333500183,
+      "loss": 0.8721,
+      "step": 9041
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013892818505848652,
+      "loss": 0.938,
+      "step": 9042
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001389152660162549,
+      "loss": 1.0742,
+      "step": 9043
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013890234620856122,
+      "loss": 1.0794,
+      "step": 9044
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001388894256356595,
+      "loss": 0.9813,
+      "step": 9045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013887650429780392,
+      "loss": 1.1322,
+      "step": 9046
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013886358219524864,
+      "loss": 0.9576,
+      "step": 9047
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013885065932824788,
+      "loss": 0.9821,
+      "step": 9048
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013883773569705577,
+      "loss": 1.0898,
+      "step": 9049
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013882481130192656,
+      "loss": 1.0275,
+      "step": 9050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013881188614311448,
+      "loss": 0.9317,
+      "step": 9051
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013879896022087366,
+      "loss": 1.1121,
+      "step": 9052
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013878603353545854,
+      "loss": 0.955,
+      "step": 9053
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001387731060871232,
+      "loss": 1.0956,
+      "step": 9054
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001387601778761221,
+      "loss": 1.0604,
+      "step": 9055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013874724890270936,
+      "loss": 0.9696,
+      "step": 9056
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013873431916713944,
+      "loss": 1.0873,
+      "step": 9057
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013872138866966656,
+      "loss": 1.076,
+      "step": 9058
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013870845741054515,
+      "loss": 1.0819,
+      "step": 9059
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001386955253900295,
+      "loss": 0.8131,
+      "step": 9060
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013868259260837402,
+      "loss": 0.9445,
+      "step": 9061
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013866965906583303,
+      "loss": 0.8861,
+      "step": 9062
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013865672476266107,
+      "loss": 1.0845,
+      "step": 9063
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013864378969911245,
+      "loss": 1.2065,
+      "step": 9064
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013863085387544162,
+      "loss": 0.9767,
+      "step": 9065
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000138617917291903,
+      "loss": 1.1559,
+      "step": 9066
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013860497994875118,
+      "loss": 0.9611,
+      "step": 9067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013859204184624047,
+      "loss": 0.9901,
+      "step": 9068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013857910298462544,
+      "loss": 0.9352,
+      "step": 9069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001385661633641606,
+      "loss": 0.7975,
+      "step": 9070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013855322298510046,
+      "loss": 1.0236,
+      "step": 9071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013854028184769957,
+      "loss": 0.9881,
+      "step": 9072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013852733995221248,
+      "loss": 1.0903,
+      "step": 9073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013851439729889377,
+      "loss": 1.0078,
+      "step": 9074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013850145388799793,
+      "loss": 1.046,
+      "step": 9075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001384885097197797,
+      "loss": 1.0253,
+      "step": 9076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001384755647944936,
+      "loss": 0.9777,
+      "step": 9077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013846261911239428,
+      "loss": 0.937,
+      "step": 9078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013844967267373636,
+      "loss": 1.2047,
+      "step": 9079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013843672547877455,
+      "loss": 1.0173,
+      "step": 9080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013842377752776346,
+      "loss": 0.9933,
+      "step": 9081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001384108288209578,
+      "loss": 0.8499,
+      "step": 9082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013839787935861225,
+      "loss": 1.0603,
+      "step": 9083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001383849291409816,
+      "loss": 0.9826,
+      "step": 9084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001383719781683205,
+      "loss": 1.0721,
+      "step": 9085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001383590264408837,
+      "loss": 1.0693,
+      "step": 9086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000138346073958926,
+      "loss": 0.7774,
+      "step": 9087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013833312072270219,
+      "loss": 1.1097,
+      "step": 9088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013832016673246696,
+      "loss": 0.8807,
+      "step": 9089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013830721198847525,
+      "loss": 0.9006,
+      "step": 9090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001382942564909818,
+      "loss": 1.0809,
+      "step": 9091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001382813002402414,
+      "loss": 1.0949,
+      "step": 9092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000138268343236509,
+      "loss": 1.08,
+      "step": 9093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013825538548003938,
+      "loss": 1.0761,
+      "step": 9094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013824242697108748,
+      "loss": 0.9973,
+      "step": 9095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013822946770990818,
+      "loss": 0.9199,
+      "step": 9096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013821650769675641,
+      "loss": 1.1207,
+      "step": 9097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.000138203546931887,
+      "loss": 1.0969,
+      "step": 9098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013819058541555498,
+      "loss": 1.0417,
+      "step": 9099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013817762314801527,
+      "loss": 0.9507,
+      "step": 9100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013816466012952285,
+      "loss": 0.9353,
+      "step": 9101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013815169636033266,
+      "loss": 1.103,
+      "step": 9102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013813873184069977,
+      "loss": 1.231,
+      "step": 9103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013812576657087913,
+      "loss": 1.071,
+      "step": 9104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001381128005511258,
+      "loss": 1.0277,
+      "step": 9105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001380998337816948,
+      "loss": 1.0355,
+      "step": 9106
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013808686626284118,
+      "loss": 1.1789,
+      "step": 9107
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013807389799482003,
+      "loss": 0.8039,
+      "step": 9108
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013806092897788644,
+      "loss": 0.737,
+      "step": 9109
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001380479592122955,
+      "loss": 0.8518,
+      "step": 9110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013803498869830233,
+      "loss": 0.9358,
+      "step": 9111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013802201743616204,
+      "loss": 1.1061,
+      "step": 9112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001380090454261298,
+      "loss": 1.1652,
+      "step": 9113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013799607266846078,
+      "loss": 0.9513,
+      "step": 9114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001379830991634101,
+      "loss": 1.0956,
+      "step": 9115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013797012491123298,
+      "loss": 0.9983,
+      "step": 9116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013795714991218463,
+      "loss": 0.9836,
+      "step": 9117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013794417416652027,
+      "loss": 1.2935,
+      "step": 9118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013793119767449512,
+      "loss": 1.3161,
+      "step": 9119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00013791822043636445,
+      "loss": 1.074,
+      "step": 9120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001379052424523835,
+      "loss": 1.0479,
+      "step": 9121
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001378922637228075,
+      "loss": 1.0106,
+      "step": 9122
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013787928424789184,
+      "loss": 0.6505,
+      "step": 9123
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013786630402789175,
+      "loss": 0.9697,
+      "step": 9124
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013785332306306259,
+      "loss": 1.0541,
+      "step": 9125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013784034135365972,
+      "loss": 1.1149,
+      "step": 9126
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013782735889993842,
+      "loss": 0.9484,
+      "step": 9127
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013781437570215406,
+      "loss": 1.147,
+      "step": 9128
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013780139176056206,
+      "loss": 1.1295,
+      "step": 9129
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013778840707541785,
+      "loss": 1.168,
+      "step": 9130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013777542164697672,
+      "loss": 1.1697,
+      "step": 9131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013776243547549422,
+      "loss": 1.1133,
+      "step": 9132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013774944856122568,
+      "loss": 0.911,
+      "step": 9133
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013773646090442664,
+      "loss": 0.9015,
+      "step": 9134
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001377234725053525,
+      "loss": 1.087,
+      "step": 9135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001377104833642588,
+      "loss": 1.032,
+      "step": 9136
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013769749348140097,
+      "loss": 1.0818,
+      "step": 9137
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013768450285703458,
+      "loss": 1.0565,
+      "step": 9138
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013767151149141513,
+      "loss": 0.9284,
+      "step": 9139
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013765851938479814,
+      "loss": 0.9811,
+      "step": 9140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001376455265374392,
+      "loss": 0.9212,
+      "step": 9141
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013763253294959385,
+      "loss": 1.0958,
+      "step": 9142
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013761953862151773,
+      "loss": 1.204,
+      "step": 9143
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013760654355346637,
+      "loss": 0.9002,
+      "step": 9144
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001375935477456954,
+      "loss": 1.1547,
+      "step": 9145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013758055119846048,
+      "loss": 0.9658,
+      "step": 9146
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013756755391201723,
+      "loss": 0.8658,
+      "step": 9147
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013755455588662126,
+      "loss": 1.1609,
+      "step": 9148
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013754155712252832,
+      "loss": 1.112,
+      "step": 9149
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013752855761999406,
+      "loss": 0.9304,
+      "step": 9150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013751555737927416,
+      "loss": 0.8771,
+      "step": 9151
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013750255640062434,
+      "loss": 0.9917,
+      "step": 9152
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013748955468430041,
+      "loss": 0.9947,
+      "step": 9153
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000137476552230558,
+      "loss": 0.9779,
+      "step": 9154
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001374635490396529,
+      "loss": 1.0485,
+      "step": 9155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013745054511184093,
+      "loss": 1.1432,
+      "step": 9156
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013743754044737781,
+      "loss": 1.1093,
+      "step": 9157
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001374245350465194,
+      "loss": 1.172,
+      "step": 9158
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001374115289095215,
+      "loss": 0.9509,
+      "step": 9159
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001373985220366399,
+      "loss": 1.1585,
+      "step": 9160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001373855144281305,
+      "loss": 0.9506,
+      "step": 9161
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013737250608424913,
+      "loss": 0.9285,
+      "step": 9162
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013735949700525163,
+      "loss": 1.1508,
+      "step": 9163
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013734648719139397,
+      "loss": 1.0464,
+      "step": 9164
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000137333476642932,
+      "loss": 1.0485,
+      "step": 9165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013732046536012166,
+      "loss": 1.1251,
+      "step": 9166
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013730745334321887,
+      "loss": 0.9646,
+      "step": 9167
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013729444059247954,
+      "loss": 1.2047,
+      "step": 9168
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001372814271081597,
+      "loss": 0.9572,
+      "step": 9169
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013726841289051528,
+      "loss": 1.1324,
+      "step": 9170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013725539793980223,
+      "loss": 0.9441,
+      "step": 9171
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001372423822562767,
+      "loss": 1.1375,
+      "step": 9172
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013722936584019453,
+      "loss": 1.1033,
+      "step": 9173
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013721634869181187,
+      "loss": 1.1601,
+      "step": 9174
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001372033308113847,
+      "loss": 0.9174,
+      "step": 9175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013719031219916917,
+      "loss": 1.0706,
+      "step": 9176
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013717729285542122,
+      "loss": 1.0136,
+      "step": 9177
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013716427278039704,
+      "loss": 0.9478,
+      "step": 9178
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013715125197435275,
+      "loss": 1.0688,
+      "step": 9179
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001371382304375444,
+      "loss": 1.0621,
+      "step": 9180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013712520817022817,
+      "loss": 1.0362,
+      "step": 9181
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001371121851726602,
+      "loss": 1.0749,
+      "step": 9182
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001370991614450966,
+      "loss": 0.8821,
+      "step": 9183
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013708613698779364,
+      "loss": 0.9519,
+      "step": 9184
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013707311180100742,
+      "loss": 1.0315,
+      "step": 9185
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001370600858849942,
+      "loss": 1.0372,
+      "step": 9186
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001370470592400102,
+      "loss": 0.9703,
+      "step": 9187
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001370340318663116,
+      "loss": 1.2301,
+      "step": 9188
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013702100376415473,
+      "loss": 1.1038,
+      "step": 9189
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001370079749337958,
+      "loss": 1.0878,
+      "step": 9190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013699494537549106,
+      "loss": 0.9669,
+      "step": 9191
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013698191508949688,
+      "loss": 1.0649,
+      "step": 9192
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013696888407606952,
+      "loss": 1.2639,
+      "step": 9193
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013695585233546527,
+      "loss": 1.2367,
+      "step": 9194
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001369428198679405,
+      "loss": 0.9532,
+      "step": 9195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013692978667375157,
+      "loss": 1.2827,
+      "step": 9196
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013691675275315484,
+      "loss": 1.2318,
+      "step": 9197
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013690371810640665,
+      "loss": 0.9499,
+      "step": 9198
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013689068273376343,
+      "loss": 1.0694,
+      "step": 9199
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013687764663548158,
+      "loss": 0.9422,
+      "step": 9200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013686460981181747,
+      "loss": 1.0309,
+      "step": 9201
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001368515722630276,
+      "loss": 1.0335,
+      "step": 9202
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001368385339893684,
+      "loss": 1.0044,
+      "step": 9203
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001368254949910963,
+      "loss": 0.9328,
+      "step": 9204
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013681245526846783,
+      "loss": 1.0351,
+      "step": 9205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001367994148217394,
+      "loss": 1.0891,
+      "step": 9206
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001367863736511676,
+      "loss": 1.0376,
+      "step": 9207
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013677333175700894,
+      "loss": 0.9962,
+      "step": 9208
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013676028913951987,
+      "loss": 1.0332,
+      "step": 9209
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013674724579895706,
+      "loss": 1.1008,
+      "step": 9210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013673420173557696,
+      "loss": 0.9965,
+      "step": 9211
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013672115694963625,
+      "loss": 1.1648,
+      "step": 9212
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013670811144139144,
+      "loss": 0.9002,
+      "step": 9213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013669506521109915,
+      "loss": 1.013,
+      "step": 9214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013668201825901602,
+      "loss": 0.9329,
+      "step": 9215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001366689705853987,
+      "loss": 0.8297,
+      "step": 9216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013665592219050377,
+      "loss": 1.0895,
+      "step": 9217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013664287307458794,
+      "loss": 1.0648,
+      "step": 9218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001366298232379079,
+      "loss": 1.1078,
+      "step": 9219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013661677268072032,
+      "loss": 1.003,
+      "step": 9220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013660372140328187,
+      "loss": 0.7118,
+      "step": 9221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013659066940584932,
+      "loss": 1.2329,
+      "step": 9222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001365776166886794,
+      "loss": 1.0875,
+      "step": 9223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001365645632520288,
+      "loss": 1.0662,
+      "step": 9224
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013655150909615434,
+      "loss": 0.8341,
+      "step": 9225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013653845422131278,
+      "loss": 1.1066,
+      "step": 9226
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001365253986277609,
+      "loss": 0.9585,
+      "step": 9227
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013651234231575553,
+      "loss": 0.981,
+      "step": 9228
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013649928528555342,
+      "loss": 0.9719,
+      "step": 9229
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013648622753741147,
+      "loss": 0.9872,
+      "step": 9230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001364731690715865,
+      "loss": 1.2447,
+      "step": 9231
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001364601098883354,
+      "loss": 0.9454,
+      "step": 9232
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000136447049987915,
+      "loss": 0.9641,
+      "step": 9233
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013643398937058216,
+      "loss": 1.0525,
+      "step": 9234
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001364209280365939,
+      "loss": 0.983,
+      "step": 9235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000136407865986207,
+      "loss": 1.0773,
+      "step": 9236
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013639480321967845,
+      "loss": 1.1979,
+      "step": 9237
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001363817397372652,
+      "loss": 0.9133,
+      "step": 9238
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013636867553922424,
+      "loss": 1.1157,
+      "step": 9239
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013635561062581245,
+      "loss": 1.0942,
+      "step": 9240
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013634254499728693,
+      "loss": 1.2263,
+      "step": 9241
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001363294786539046,
+      "loss": 0.9839,
+      "step": 9242
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013631641159592253,
+      "loss": 1.3268,
+      "step": 9243
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013630334382359763,
+      "loss": 0.8699,
+      "step": 9244
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001362902753371871,
+      "loss": 1.0322,
+      "step": 9245
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001362772061369479,
+      "loss": 0.8465,
+      "step": 9246
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013626413622313714,
+      "loss": 1.154,
+      "step": 9247
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013625106559601188,
+      "loss": 1.1283,
+      "step": 9248
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013623799425582924,
+      "loss": 1.1667,
+      "step": 9249
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013622492220284632,
+      "loss": 1.0807,
+      "step": 9250
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013621184943732025,
+      "loss": 0.9763,
+      "step": 9251
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013619877595950815,
+      "loss": 0.9999,
+      "step": 9252
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013618570176966724,
+      "loss": 1.203,
+      "step": 9253
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013617262686805459,
+      "loss": 1.0611,
+      "step": 9254
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013615955125492748,
+      "loss": 1.1587,
+      "step": 9255
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013614647493054307,
+      "loss": 1.2286,
+      "step": 9256
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001361333978951585,
+      "loss": 0.9372,
+      "step": 9257
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013612032014903114,
+      "loss": 0.9055,
+      "step": 9258
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013610724169241812,
+      "loss": 1.1346,
+      "step": 9259
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001360941625255767,
+      "loss": 1.0034,
+      "step": 9260
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001360810826487642,
+      "loss": 1.1312,
+      "step": 9261
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013606800206223784,
+      "loss": 0.9683,
+      "step": 9262
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013605492076625497,
+      "loss": 0.9372,
+      "step": 9263
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013604183876107285,
+      "loss": 0.968,
+      "step": 9264
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013602875604694885,
+      "loss": 0.9733,
+      "step": 9265
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013601567262414027,
+      "loss": 1.2444,
+      "step": 9266
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013600258849290445,
+      "loss": 1.0591,
+      "step": 9267
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013598950365349883,
+      "loss": 0.904,
+      "step": 9268
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013597641810618073,
+      "loss": 0.9616,
+      "step": 9269
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001359633318512075,
+      "loss": 0.9995,
+      "step": 9270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001359502448888366,
+      "loss": 1.161,
+      "step": 9271
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013593715721932549,
+      "loss": 1.0042,
+      "step": 9272
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001359240688429315,
+      "loss": 1.0971,
+      "step": 9273
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013591097975991222,
+      "loss": 1.0226,
+      "step": 9274
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013589788997052496,
+      "loss": 0.85,
+      "step": 9275
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001358847994750273,
+      "loss": 1.1482,
+      "step": 9276
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013587170827367663,
+      "loss": 0.8821,
+      "step": 9277
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001358586163667306,
+      "loss": 1.0719,
+      "step": 9278
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001358455237544466,
+      "loss": 1.1108,
+      "step": 9279
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001358324304370822,
+      "loss": 1.3109,
+      "step": 9280
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013581933641489496,
+      "loss": 0.9431,
+      "step": 9281
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013580624168814243,
+      "loss": 1.0181,
+      "step": 9282
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001357931462570822,
+      "loss": 1.2187,
+      "step": 9283
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013578005012197184,
+      "loss": 1.2252,
+      "step": 9284
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013576695328306894,
+      "loss": 0.8892,
+      "step": 9285
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013575385574063112,
+      "loss": 0.9492,
+      "step": 9286
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013574075749491602,
+      "loss": 1.0206,
+      "step": 9287
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013572765854618127,
+      "loss": 1.157,
+      "step": 9288
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013571455889468457,
+      "loss": 1.0154,
+      "step": 9289
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013570145854068349,
+      "loss": 1.0485,
+      "step": 9290
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001356883574844358,
+      "loss": 0.9875,
+      "step": 9291
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013567525572619917,
+      "loss": 1.0909,
+      "step": 9292
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001356621532662313,
+      "loss": 1.1288,
+      "step": 9293
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013564905010478997,
+      "loss": 1.0255,
+      "step": 9294
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013563594624213282,
+      "loss": 0.9821,
+      "step": 9295
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013562284167851772,
+      "loss": 1.158,
+      "step": 9296
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013560973641420235,
+      "loss": 0.8824,
+      "step": 9297
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001355966304494445,
+      "loss": 1.1057,
+      "step": 9298
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013558352378450198,
+      "loss": 1.0127,
+      "step": 9299
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001355704164196326,
+      "loss": 1.0912,
+      "step": 9300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001355573083550942,
+      "loss": 1.073,
+      "step": 9301
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001355441995911446,
+      "loss": 1.0575,
+      "step": 9302
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013553109012804163,
+      "loss": 0.7913,
+      "step": 9303
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013551797996604317,
+      "loss": 1.2022,
+      "step": 9304
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013550486910540708,
+      "loss": 0.9629,
+      "step": 9305
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013549175754639124,
+      "loss": 0.991,
+      "step": 9306
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013547864528925364,
+      "loss": 1.2277,
+      "step": 9307
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001354655323342521,
+      "loss": 0.8947,
+      "step": 9308
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013545241868164457,
+      "loss": 0.9751,
+      "step": 9309
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013543930433168906,
+      "loss": 0.9419,
+      "step": 9310
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013542618928464345,
+      "loss": 0.9081,
+      "step": 9311
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013541307354076573,
+      "loss": 1.1072,
+      "step": 9312
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001353999571003139,
+      "loss": 1.148,
+      "step": 9313
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013538683996354596,
+      "loss": 0.8853,
+      "step": 9314
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013537372213071996,
+      "loss": 1.0663,
+      "step": 9315
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013536060360209385,
+      "loss": 1.1631,
+      "step": 9316
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013534748437792573,
+      "loss": 1.1051,
+      "step": 9317
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001353343644584736,
+      "loss": 1.0433,
+      "step": 9318
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013532124384399563,
+      "loss": 1.0546,
+      "step": 9319
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013530812253474982,
+      "loss": 0.9115,
+      "step": 9320
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013529500053099423,
+      "loss": 1.0253,
+      "step": 9321
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013528187783298705,
+      "loss": 0.9703,
+      "step": 9322
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001352687544409864,
+      "loss": 1.2404,
+      "step": 9323
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013525563035525034,
+      "loss": 0.8739,
+      "step": 9324
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013524250557603713,
+      "loss": 1.0058,
+      "step": 9325
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013522938010360482,
+      "loss": 0.9731,
+      "step": 9326
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001352162539382117,
+      "loss": 0.8438,
+      "step": 9327
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013520312708011585,
+      "loss": 1.0091,
+      "step": 9328
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001351899995295756,
+      "loss": 1.1835,
+      "step": 9329
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013517687128684905,
+      "loss": 1.2182,
+      "step": 9330
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001351637423521945,
+      "loss": 0.8829,
+      "step": 9331
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013515061272587013,
+      "loss": 0.9053,
+      "step": 9332
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001351374824081343,
+      "loss": 0.9643,
+      "step": 9333
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001351243513992452,
+      "loss": 1.0514,
+      "step": 9334
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013511121969946114,
+      "loss": 0.8794,
+      "step": 9335
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013509808730904042,
+      "loss": 1.0692,
+      "step": 9336
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001350849542282414,
+      "loss": 1.0085,
+      "step": 9337
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013507182045732234,
+      "loss": 0.9198,
+      "step": 9338
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001350586859965416,
+      "loss": 1.0134,
+      "step": 9339
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013504555084615757,
+      "loss": 0.94,
+      "step": 9340
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013503241500642855,
+      "loss": 1.053,
+      "step": 9341
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.000135019278477613,
+      "loss": 1.0655,
+      "step": 9342
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013500614125996923,
+      "loss": 0.6795,
+      "step": 9343
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013499300335375572,
+      "loss": 1.1302,
+      "step": 9344
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013497986475923088,
+      "loss": 1.0937,
+      "step": 9345
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013496672547665313,
+      "loss": 1.09,
+      "step": 9346
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001349535855062809,
+      "loss": 0.9587,
+      "step": 9347
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013494044484837272,
+      "loss": 0.9304,
+      "step": 9348
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013492730350318696,
+      "loss": 1.1214,
+      "step": 9349
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001349141614709822,
+      "loss": 1.3413,
+      "step": 9350
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013490101875201694,
+      "loss": 1.0608,
+      "step": 9351
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00013488787534654965,
+      "loss": 1.0568,
+      "step": 9352
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013487473125483888,
+      "loss": 1.1639,
+      "step": 9353
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013486158647714317,
+      "loss": 0.8455,
+      "step": 9354
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001348484410137211,
+      "loss": 1.0926,
+      "step": 9355
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001348352948648312,
+      "loss": 1.0015,
+      "step": 9356
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013482214803073209,
+      "loss": 1.065,
+      "step": 9357
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001348090005116824,
+      "loss": 1.2417,
+      "step": 9358
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013479585230794064,
+      "loss": 0.9333,
+      "step": 9359
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001347827034197655,
+      "loss": 0.8685,
+      "step": 9360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013476955384741563,
+      "loss": 1.1839,
+      "step": 9361
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013475640359114966,
+      "loss": 0.9754,
+      "step": 9362
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013474325265122621,
+      "loss": 0.969,
+      "step": 9363
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013473010102790407,
+      "loss": 1.0283,
+      "step": 9364
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013471694872144188,
+      "loss": 0.8848,
+      "step": 9365
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013470379573209828,
+      "loss": 1.1618,
+      "step": 9366
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013469064206013206,
+      "loss": 1.0308,
+      "step": 9367
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013467748770580194,
+      "loss": 1.0816,
+      "step": 9368
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013466433266936663,
+      "loss": 1.2239,
+      "step": 9369
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013465117695108495,
+      "loss": 0.988,
+      "step": 9370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013463802055121567,
+      "loss": 0.9562,
+      "step": 9371
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001346248634700175,
+      "loss": 1.2705,
+      "step": 9372
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001346117057077493,
+      "loss": 1.2587,
+      "step": 9373
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001345985472646699,
+      "loss": 1.113,
+      "step": 9374
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013458538814103808,
+      "loss": 1.0434,
+      "step": 9375
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013457222833711266,
+      "loss": 1.1007,
+      "step": 9376
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013455906785315256,
+      "loss": 1.0045,
+      "step": 9377
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013454590668941663,
+      "loss": 1.1981,
+      "step": 9378
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013453274484616373,
+      "loss": 1.0173,
+      "step": 9379
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013451958232365275,
+      "loss": 0.8798,
+      "step": 9380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013450641912214262,
+      "loss": 0.9502,
+      "step": 9381
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013449325524189217,
+      "loss": 1.0923,
+      "step": 9382
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001344800906831605,
+      "loss": 1.0881,
+      "step": 9383
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013446692544620643,
+      "loss": 0.8268,
+      "step": 9384
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013445375953128895,
+      "loss": 1.1087,
+      "step": 9385
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013444059293866704,
+      "loss": 1.0442,
+      "step": 9386
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013442742566859969,
+      "loss": 0.8633,
+      "step": 9387
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001344142577213459,
+      "loss": 0.928,
+      "step": 9388
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013440108909716468,
+      "loss": 0.8401,
+      "step": 9389
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000134387919796315,
+      "loss": 1.1125,
+      "step": 9390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000134374749819056,
+      "loss": 1.0546,
+      "step": 9391
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001343615791656467,
+      "loss": 0.9783,
+      "step": 9392
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001343484078363461,
+      "loss": 1.1728,
+      "step": 9393
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001343352358314134,
+      "loss": 1.2882,
+      "step": 9394
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013432206315110759,
+      "loss": 0.9807,
+      "step": 9395
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001343088897956878,
+      "loss": 0.9947,
+      "step": 9396
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013429571576541315,
+      "loss": 1.1874,
+      "step": 9397
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013428254106054283,
+      "loss": 1.0718,
+      "step": 9398
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001342693656813359,
+      "loss": 0.8721,
+      "step": 9399
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013425618962805157,
+      "loss": 0.7507,
+      "step": 9400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000134243012900949,
+      "loss": 1.0983,
+      "step": 9401
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013422983550028737,
+      "loss": 1.0129,
+      "step": 9402
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001342166574263259,
+      "loss": 1.19,
+      "step": 9403
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001342034786793238,
+      "loss": 0.9872,
+      "step": 9404
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013419029925954029,
+      "loss": 0.9032,
+      "step": 9405
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013417711916723457,
+      "loss": 0.9976,
+      "step": 9406
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013416393840266593,
+      "loss": 0.9747,
+      "step": 9407
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013415075696609363,
+      "loss": 1.0059,
+      "step": 9408
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013413757485777696,
+      "loss": 0.8815,
+      "step": 9409
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001341243920779752,
+      "loss": 1.1067,
+      "step": 9410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013411120862694768,
+      "loss": 1.0523,
+      "step": 9411
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013409802450495365,
+      "loss": 0.9702,
+      "step": 9412
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001340848397122525,
+      "loss": 1.0561,
+      "step": 9413
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013407165424910355,
+      "loss": 0.9835,
+      "step": 9414
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001340584681157662,
+      "loss": 1.1558,
+      "step": 9415
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013404528131249972,
+      "loss": 0.8821,
+      "step": 9416
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013403209383956366,
+      "loss": 1.0559,
+      "step": 9417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013401890569721725,
+      "loss": 1.0945,
+      "step": 9418
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013400571688571998,
+      "loss": 0.9192,
+      "step": 9419
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013399252740533125,
+      "loss": 1.1097,
+      "step": 9420
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013397933725631055,
+      "loss": 1.0126,
+      "step": 9421
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013396614643891724,
+      "loss": 0.9597,
+      "step": 9422
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001339529549534109,
+      "loss": 1.0852,
+      "step": 9423
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001339397628000509,
+      "loss": 1.0853,
+      "step": 9424
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013392656997909675,
+      "loss": 1.0508,
+      "step": 9425
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000133913376490808,
+      "loss": 0.9343,
+      "step": 9426
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013390018233544414,
+      "loss": 1.0705,
+      "step": 9427
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013388698751326468,
+      "loss": 1.0643,
+      "step": 9428
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013387379202452917,
+      "loss": 1.0113,
+      "step": 9429
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013386059586949716,
+      "loss": 1.0271,
+      "step": 9430
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013384739904842823,
+      "loss": 0.9874,
+      "step": 9431
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013383420156158198,
+      "loss": 1.0038,
+      "step": 9432
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013382100340921798,
+      "loss": 1.0022,
+      "step": 9433
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013380780459159585,
+      "loss": 0.8899,
+      "step": 9434
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001337946051089752,
+      "loss": 0.9418,
+      "step": 9435
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013378140496161564,
+      "loss": 1.0503,
+      "step": 9436
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013376820414977685,
+      "loss": 0.9574,
+      "step": 9437
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013375500267371853,
+      "loss": 0.8975,
+      "step": 9438
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013374180053370022,
+      "loss": 1.1549,
+      "step": 9439
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013372859772998175,
+      "loss": 0.9959,
+      "step": 9440
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013371539426282274,
+      "loss": 1.2227,
+      "step": 9441
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013370219013248292,
+      "loss": 0.9443,
+      "step": 9442
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000133688985339222,
+      "loss": 0.8819,
+      "step": 9443
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013367577988329977,
+      "loss": 1.0769,
+      "step": 9444
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001336625737649759,
+      "loss": 0.9677,
+      "step": 9445
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013364936698451025,
+      "loss": 1.2047,
+      "step": 9446
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013363615954216251,
+      "loss": 1.2083,
+      "step": 9447
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001336229514381925,
+      "loss": 1.1678,
+      "step": 9448
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013360974267286006,
+      "loss": 0.8622,
+      "step": 9449
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013359653324642495,
+      "loss": 0.7887,
+      "step": 9450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013358332315914704,
+      "loss": 1.1658,
+      "step": 9451
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013357011241128615,
+      "loss": 0.9287,
+      "step": 9452
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013355690100310213,
+      "loss": 1.1386,
+      "step": 9453
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013354368893485485,
+      "loss": 1.0891,
+      "step": 9454
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001335304762068042,
+      "loss": 0.8402,
+      "step": 9455
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013351726281921007,
+      "loss": 0.8859,
+      "step": 9456
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001335040487723324,
+      "loss": 0.8889,
+      "step": 9457
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013349083406643107,
+      "loss": 1.173,
+      "step": 9458
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013347761870176602,
+      "loss": 1.219,
+      "step": 9459
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001334644026785972,
+      "loss": 0.9066,
+      "step": 9460
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013345118599718457,
+      "loss": 1.0391,
+      "step": 9461
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013343796865778807,
+      "loss": 1.0133,
+      "step": 9462
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013342475066066776,
+      "loss": 0.9124,
+      "step": 9463
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013341153200608353,
+      "loss": 1.021,
+      "step": 9464
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001333983126942955,
+      "loss": 0.8601,
+      "step": 9465
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013338509272556363,
+      "loss": 0.8429,
+      "step": 9466
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013337187210014798,
+      "loss": 0.9847,
+      "step": 9467
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001333586508183086,
+      "loss": 0.8983,
+      "step": 9468
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013334542888030553,
+      "loss": 1.0027,
+      "step": 9469
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013333220628639885,
+      "loss": 1.1898,
+      "step": 9470
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001333189830368487,
+      "loss": 1.3177,
+      "step": 9471
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013330575913191506,
+      "loss": 1.2219,
+      "step": 9472
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013329253457185818,
+      "loss": 1.055,
+      "step": 9473
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001332793093569381,
+      "loss": 1.1509,
+      "step": 9474
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000133266083487415,
+      "loss": 1.0308,
+      "step": 9475
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013325285696354902,
+      "loss": 1.2074,
+      "step": 9476
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013323962978560033,
+      "loss": 1.2067,
+      "step": 9477
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013322640195382907,
+      "loss": 0.9339,
+      "step": 9478
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013321317346849548,
+      "loss": 1.0703,
+      "step": 9479
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001331999443298598,
+      "loss": 1.0527,
+      "step": 9480
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013318671453818216,
+      "loss": 1.0135,
+      "step": 9481
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001331734840937228,
+      "loss": 0.8912,
+      "step": 9482
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013316025299674202,
+      "loss": 1.0773,
+      "step": 9483
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013314702124750003,
+      "loss": 1.1366,
+      "step": 9484
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001331337888462571,
+      "loss": 1.1607,
+      "step": 9485
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013312055579327355,
+      "loss": 0.8936,
+      "step": 9486
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013310732208880963,
+      "loss": 1.0069,
+      "step": 9487
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013309408773312567,
+      "loss": 1.017,
+      "step": 9488
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013308085272648198,
+      "loss": 0.9361,
+      "step": 9489
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013306761706913887,
+      "loss": 1.0848,
+      "step": 9490
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001330543807613567,
+      "loss": 1.0729,
+      "step": 9491
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013304114380339587,
+      "loss": 0.9791,
+      "step": 9492
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013302790619551674,
+      "loss": 1.074,
+      "step": 9493
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013301466793797962,
+      "loss": 1.037,
+      "step": 9494
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013300142903104498,
+      "loss": 1.1383,
+      "step": 9495
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001329881894749732,
+      "loss": 1.1759,
+      "step": 9496
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001329749492700247,
+      "loss": 0.7509,
+      "step": 9497
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013296170841645995,
+      "loss": 0.9599,
+      "step": 9498
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013294846691453931,
+      "loss": 1.0926,
+      "step": 9499
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013293522476452336,
+      "loss": 1.0255,
+      "step": 9500
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013292198196667248,
+      "loss": 0.9651,
+      "step": 9501
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001329087385212472,
+      "loss": 0.9735,
+      "step": 9502
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000132895494428508,
+      "loss": 1.0568,
+      "step": 9503
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001328822496887154,
+      "loss": 0.7582,
+      "step": 9504
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001328690043021299,
+      "loss": 0.8844,
+      "step": 9505
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013285575826901206,
+      "loss": 0.9316,
+      "step": 9506
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013284251158962246,
+      "loss": 1.006,
+      "step": 9507
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001328292642642216,
+      "loss": 1.063,
+      "step": 9508
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013281601629307007,
+      "loss": 0.9213,
+      "step": 9509
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001328027676764285,
+      "loss": 0.8446,
+      "step": 9510
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013278951841455742,
+      "loss": 1.056,
+      "step": 9511
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013277626850771753,
+      "loss": 1.1433,
+      "step": 9512
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013276301795616936,
+      "loss": 0.8716,
+      "step": 9513
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001327497667601736,
+      "loss": 1.0045,
+      "step": 9514
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013273651491999092,
+      "loss": 0.82,
+      "step": 9515
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013272326243588193,
+      "loss": 1.1219,
+      "step": 9516
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013271000930810737,
+      "loss": 0.9427,
+      "step": 9517
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013269675553692785,
+      "loss": 1.3231,
+      "step": 9518
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013268350112260414,
+      "loss": 0.9012,
+      "step": 9519
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013267024606539694,
+      "loss": 1.0078,
+      "step": 9520
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013265699036556697,
+      "loss": 1.0605,
+      "step": 9521
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001326437340233749,
+      "loss": 0.96,
+      "step": 9522
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001326304770390816,
+      "loss": 0.9714,
+      "step": 9523
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013261721941294777,
+      "loss": 1.1024,
+      "step": 9524
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001326039611452342,
+      "loss": 0.9884,
+      "step": 9525
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013259070223620167,
+      "loss": 0.9883,
+      "step": 9526
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013257744268611098,
+      "loss": 1.1254,
+      "step": 9527
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013256418249522298,
+      "loss": 1.0896,
+      "step": 9528
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013255092166379845,
+      "loss": 0.8254,
+      "step": 9529
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001325376601920983,
+      "loss": 1.2063,
+      "step": 9530
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013252439808038324,
+      "loss": 1.2298,
+      "step": 9531
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001325111353289143,
+      "loss": 1.1903,
+      "step": 9532
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013249787193795226,
+      "loss": 0.772,
+      "step": 9533
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013248460790775806,
+      "loss": 0.9893,
+      "step": 9534
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013247134323859255,
+      "loss": 0.9703,
+      "step": 9535
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013245807793071674,
+      "loss": 1.1755,
+      "step": 9536
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013244481198439143,
+      "loss": 0.949,
+      "step": 9537
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013243154539987765,
+      "loss": 1.0723,
+      "step": 9538
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013241827817743633,
+      "loss": 1.0607,
+      "step": 9539
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013240501031732847,
+      "loss": 0.9579,
+      "step": 9540
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013239174181981495,
+      "loss": 1.0201,
+      "step": 9541
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013237847268515686,
+      "loss": 1.2037,
+      "step": 9542
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013236520291361515,
+      "loss": 1.2299,
+      "step": 9543
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013235193250545088,
+      "loss": 0.8874,
+      "step": 9544
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013233866146092503,
+      "loss": 0.9664,
+      "step": 9545
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013232538978029867,
+      "loss": 1.0624,
+      "step": 9546
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013231211746383287,
+      "loss": 1.1931,
+      "step": 9547
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001322988445117886,
+      "loss": 0.9678,
+      "step": 9548
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013228557092442712,
+      "loss": 1.1307,
+      "step": 9549
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013227229670200935,
+      "loss": 0.9439,
+      "step": 9550
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013225902184479645,
+      "loss": 0.9971,
+      "step": 9551
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013224574635304959,
+      "loss": 0.8737,
+      "step": 9552
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013223247022702982,
+      "loss": 1.0798,
+      "step": 9553
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001322191934669983,
+      "loss": 0.8484,
+      "step": 9554
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.0001322059160732162,
+      "loss": 1.092,
+      "step": 9555
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013219263804594472,
+      "loss": 1.0304,
+      "step": 9556
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013217935938544497,
+      "loss": 0.9473,
+      "step": 9557
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013216608009197816,
+      "loss": 1.1076,
+      "step": 9558
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013215280016580554,
+      "loss": 1.3252,
+      "step": 9559
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013213951960718825,
+      "loss": 0.8149,
+      "step": 9560
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013212623841638759,
+      "loss": 1.1563,
+      "step": 9561
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013211295659366473,
+      "loss": 1.0495,
+      "step": 9562
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000132099674139281,
+      "loss": 0.9587,
+      "step": 9563
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013208639105349762,
+      "loss": 1.1062,
+      "step": 9564
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013207310733657585,
+      "loss": 1.1649,
+      "step": 9565
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013205982298877703,
+      "loss": 0.9299,
+      "step": 9566
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013204653801036242,
+      "loss": 0.9362,
+      "step": 9567
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013203325240159337,
+      "loss": 1.1891,
+      "step": 9568
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013201996616273118,
+      "loss": 1.0323,
+      "step": 9569
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013200667929403722,
+      "loss": 0.9853,
+      "step": 9570
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013199339179577278,
+      "loss": 0.984,
+      "step": 9571
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013198010366819932,
+      "loss": 1.0059,
+      "step": 9572
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013196681491157818,
+      "loss": 1.0295,
+      "step": 9573
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013195352552617065,
+      "loss": 0.9869,
+      "step": 9574
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013194023551223828,
+      "loss": 1.1078,
+      "step": 9575
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013192694487004243,
+      "loss": 0.9643,
+      "step": 9576
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013191365359984446,
+      "loss": 1.0779,
+      "step": 9577
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013190036170190593,
+      "loss": 0.8956,
+      "step": 9578
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013188706917648818,
+      "loss": 1.1163,
+      "step": 9579
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013187377602385275,
+      "loss": 1.1648,
+      "step": 9580
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013186048224426107,
+      "loss": 0.8812,
+      "step": 9581
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013184718783797465,
+      "loss": 1.0241,
+      "step": 9582
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00013183389280525497,
+      "loss": 1.0647,
+      "step": 9583
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001318205971463636,
+      "loss": 0.8515,
+      "step": 9584
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000131807300861562,
+      "loss": 1.1787,
+      "step": 9585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013179400395111173,
+      "loss": 1.0022,
+      "step": 9586
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001317807064152743,
+      "loss": 1.1529,
+      "step": 9587
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013176740825431138,
+      "loss": 0.7834,
+      "step": 9588
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013175410946848445,
+      "loss": 0.9799,
+      "step": 9589
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013174081005805512,
+      "loss": 1.2392,
+      "step": 9590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000131727510023285,
+      "loss": 1.0203,
+      "step": 9591
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013171420936443568,
+      "loss": 0.9613,
+      "step": 9592
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013170090808176883,
+      "loss": 1.0195,
+      "step": 9593
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000131687606175546,
+      "loss": 1.2609,
+      "step": 9594
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013167430364602895,
+      "loss": 1.1174,
+      "step": 9595
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013166100049347924,
+      "loss": 0.9375,
+      "step": 9596
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013164769671815862,
+      "loss": 0.9758,
+      "step": 9597
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013163439232032872,
+      "loss": 1.1243,
+      "step": 9598
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001316210873002513,
+      "loss": 0.9313,
+      "step": 9599
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013160778165818794,
+      "loss": 1.16,
+      "step": 9600
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001315944753944005,
+      "loss": 0.9843,
+      "step": 9601
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013158116850915067,
+      "loss": 1.1672,
+      "step": 9602
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001315678610027002,
+      "loss": 1.0384,
+      "step": 9603
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013155455287531078,
+      "loss": 1.2441,
+      "step": 9604
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001315412441272443,
+      "loss": 0.8294,
+      "step": 9605
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013152793475876244,
+      "loss": 1.1,
+      "step": 9606
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013151462477012702,
+      "loss": 0.9767,
+      "step": 9607
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001315013141615999,
+      "loss": 0.9881,
+      "step": 9608
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013148800293344282,
+      "loss": 1.1735,
+      "step": 9609
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013147469108591768,
+      "loss": 0.9574,
+      "step": 9610
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013146137861928632,
+      "loss": 1.0339,
+      "step": 9611
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013144806553381052,
+      "loss": 0.9982,
+      "step": 9612
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013143475182975222,
+      "loss": 1.037,
+      "step": 9613
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013142143750737327,
+      "loss": 1.0577,
+      "step": 9614
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001314081225669356,
+      "loss": 0.9858,
+      "step": 9615
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013139480700870109,
+      "loss": 1.007,
+      "step": 9616
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001313814908329316,
+      "loss": 1.1583,
+      "step": 9617
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013136817403988917,
+      "loss": 1.0123,
+      "step": 9618
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013135485662983565,
+      "loss": 1.0007,
+      "step": 9619
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000131341538603033,
+      "loss": 0.9801,
+      "step": 9620
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013132821995974326,
+      "loss": 0.8989,
+      "step": 9621
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013131490070022837,
+      "loss": 1.0887,
+      "step": 9622
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013130158082475026,
+      "loss": 1.1733,
+      "step": 9623
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000131288260333571,
+      "loss": 0.915,
+      "step": 9624
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001312749392269526,
+      "loss": 0.8265,
+      "step": 9625
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013126161750515706,
+      "loss": 0.9668,
+      "step": 9626
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001312482951684464,
+      "loss": 0.9055,
+      "step": 9627
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013123497221708273,
+      "loss": 1.1114,
+      "step": 9628
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013122164865132807,
+      "loss": 1.1297,
+      "step": 9629
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013120832447144449,
+      "loss": 1.0812,
+      "step": 9630
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001311949996776941,
+      "loss": 0.9603,
+      "step": 9631
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000131181674270339,
+      "loss": 1.2533,
+      "step": 9632
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013116834824964123,
+      "loss": 1.0036,
+      "step": 9633
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013115502161586303,
+      "loss": 0.9546,
+      "step": 9634
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013114169436926647,
+      "loss": 0.9394,
+      "step": 9635
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013112836651011368,
+      "loss": 0.9977,
+      "step": 9636
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013111503803866684,
+      "loss": 1.0504,
+      "step": 9637
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013110170895518813,
+      "loss": 1.0566,
+      "step": 9638
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013108837925993974,
+      "loss": 0.9365,
+      "step": 9639
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013107504895318382,
+      "loss": 1.0115,
+      "step": 9640
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001310617180351826,
+      "loss": 0.8849,
+      "step": 9641
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013104838650619834,
+      "loss": 0.9936,
+      "step": 9642
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013103505436649318,
+      "loss": 1.0735,
+      "step": 9643
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013102172161632945,
+      "loss": 1.1433,
+      "step": 9644
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001310083882559694,
+      "loss": 1.022,
+      "step": 9645
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013099505428567523,
+      "loss": 1.0444,
+      "step": 9646
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013098171970570927,
+      "loss": 1.0589,
+      "step": 9647
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013096838451633378,
+      "loss": 1.0657,
+      "step": 9648
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013095504871781113,
+      "loss": 1.03,
+      "step": 9649
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013094171231040354,
+      "loss": 1.0415,
+      "step": 9650
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001309283752943734,
+      "loss": 0.8908,
+      "step": 9651
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013091503766998302,
+      "loss": 1.0573,
+      "step": 9652
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013090169943749476,
+      "loss": 0.9733,
+      "step": 9653
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013088836059717098,
+      "loss": 1.1141,
+      "step": 9654
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013087502114927411,
+      "loss": 0.8793,
+      "step": 9655
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001308616810940664,
+      "loss": 0.9971,
+      "step": 9656
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001308483404318104,
+      "loss": 1.0276,
+      "step": 9657
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013083499916276842,
+      "loss": 0.9826,
+      "step": 9658
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013082165728720292,
+      "loss": 1.1156,
+      "step": 9659
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013080831480537635,
+      "loss": 0.915,
+      "step": 9660
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013079497171755112,
+      "loss": 0.8794,
+      "step": 9661
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013078162802398975,
+      "loss": 0.9613,
+      "step": 9662
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001307682837249546,
+      "loss": 0.9903,
+      "step": 9663
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013075493882070826,
+      "loss": 1.0607,
+      "step": 9664
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013074159331151317,
+      "loss": 1.0961,
+      "step": 9665
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013072824719763184,
+      "loss": 0.9656,
+      "step": 9666
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001307149004793268,
+      "loss": 0.9255,
+      "step": 9667
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001307015531568606,
+      "loss": 1.0859,
+      "step": 9668
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013068820523049572,
+      "loss": 0.9784,
+      "step": 9669
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013067485670049478,
+      "loss": 0.901,
+      "step": 9670
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013066150756712027,
+      "loss": 0.9706,
+      "step": 9671
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013064815783063487,
+      "loss": 0.965,
+      "step": 9672
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013063480749130106,
+      "loss": 1.0904,
+      "step": 9673
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001306214565493815,
+      "loss": 1.0323,
+      "step": 9674
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013060810500513883,
+      "loss": 1.0788,
+      "step": 9675
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001305947528588356,
+      "loss": 0.8121,
+      "step": 9676
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013058140011073448,
+      "loss": 1.1016,
+      "step": 9677
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013056804676109818,
+      "loss": 0.8898,
+      "step": 9678
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013055469281018925,
+      "loss": 1.112,
+      "step": 9679
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013054133825827043,
+      "loss": 1.023,
+      "step": 9680
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013052798310560438,
+      "loss": 1.067,
+      "step": 9681
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001305146273524538,
+      "loss": 1.1047,
+      "step": 9682
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001305012709990814,
+      "loss": 0.9882,
+      "step": 9683
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001304879140457499,
+      "loss": 1.0725,
+      "step": 9684
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013047455649272207,
+      "loss": 0.8646,
+      "step": 9685
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013046119834026057,
+      "loss": 0.9273,
+      "step": 9686
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001304478395886282,
+      "loss": 0.9435,
+      "step": 9687
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013043448023808773,
+      "loss": 0.8054,
+      "step": 9688
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013042112028890194,
+      "loss": 1.0776,
+      "step": 9689
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013040775974133362,
+      "loss": 0.8406,
+      "step": 9690
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013039439859564557,
+      "loss": 0.9869,
+      "step": 9691
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013038103685210058,
+      "loss": 0.8806,
+      "step": 9692
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013036767451096148,
+      "loss": 1.1129,
+      "step": 9693
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013035431157249115,
+      "loss": 1.0311,
+      "step": 9694
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001303409480369524,
+      "loss": 1.1559,
+      "step": 9695
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013032758390460803,
+      "loss": 0.9486,
+      "step": 9696
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013031421917572108,
+      "loss": 1.0512,
+      "step": 9697
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013030085385055428,
+      "loss": 1.065,
+      "step": 9698
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001302874879293706,
+      "loss": 0.8797,
+      "step": 9699
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013027412141243292,
+      "loss": 0.8791,
+      "step": 9700
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013026075430000417,
+      "loss": 1.0452,
+      "step": 9701
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013024738659234727,
+      "loss": 1.0241,
+      "step": 9702
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013023401828972516,
+      "loss": 0.8896,
+      "step": 9703
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001302206493924008,
+      "loss": 1.0276,
+      "step": 9704
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013020727990063717,
+      "loss": 0.956,
+      "step": 9705
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013019390981469723,
+      "loss": 1.1202,
+      "step": 9706
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013018053913484398,
+      "loss": 1.1826,
+      "step": 9707
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013016716786134046,
+      "loss": 1.1062,
+      "step": 9708
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013015379599444957,
+      "loss": 0.8555,
+      "step": 9709
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013014042353443444,
+      "loss": 1.2242,
+      "step": 9710
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013012705048155807,
+      "loss": 1.2327,
+      "step": 9711
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001301136768360835,
+      "loss": 1.0084,
+      "step": 9712
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013010030259827377,
+      "loss": 1.1076,
+      "step": 9713
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013008692776839202,
+      "loss": 1.1587,
+      "step": 9714
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013007355234670125,
+      "loss": 1.059,
+      "step": 9715
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001300601763334646,
+      "loss": 1.1322,
+      "step": 9716
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001300467997289452,
+      "loss": 1.0939,
+      "step": 9717
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013003342253340612,
+      "loss": 0.9458,
+      "step": 9718
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013002004474711048,
+      "loss": 1.1108,
+      "step": 9719
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00013000666637032147,
+      "loss": 1.1526,
+      "step": 9720
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012999328740330223,
+      "loss": 0.9975,
+      "step": 9721
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001299799078463159,
+      "loss": 1.0953,
+      "step": 9722
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012996652769962566,
+      "loss": 0.9989,
+      "step": 9723
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012995314696349474,
+      "loss": 0.8442,
+      "step": 9724
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012993976563818626,
+      "loss": 0.9694,
+      "step": 9725
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012992638372396348,
+      "loss": 1.012,
+      "step": 9726
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012991300122108964,
+      "loss": 0.7721,
+      "step": 9727
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012989961812982795,
+      "loss": 1.0597,
+      "step": 9728
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001298862344504416,
+      "loss": 1.0054,
+      "step": 9729
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012987285018319395,
+      "loss": 0.9653,
+      "step": 9730
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012985946532834822,
+      "loss": 0.9271,
+      "step": 9731
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012984607988616767,
+      "loss": 0.8463,
+      "step": 9732
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001298326938569156,
+      "loss": 1.2654,
+      "step": 9733
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012981930724085535,
+      "loss": 0.8776,
+      "step": 9734
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012980592003825024,
+      "loss": 0.921,
+      "step": 9735
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012979253224936347,
+      "loss": 1.0207,
+      "step": 9736
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012977914387445855,
+      "loss": 1.0936,
+      "step": 9737
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001297657549137987,
+      "loss": 1.0078,
+      "step": 9738
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012975236536764737,
+      "loss": 1.2029,
+      "step": 9739
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012973897523626786,
+      "loss": 1.0536,
+      "step": 9740
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012972558451992362,
+      "loss": 0.9019,
+      "step": 9741
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012971219321887794,
+      "loss": 1.0903,
+      "step": 9742
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012969880133339437,
+      "loss": 0.7773,
+      "step": 9743
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001296854088637362,
+      "loss": 1.0037,
+      "step": 9744
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012967201581016698,
+      "loss": 1.0069,
+      "step": 9745
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012965862217295,
+      "loss": 1.0016,
+      "step": 9746
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012964522795234887,
+      "loss": 0.9725,
+      "step": 9747
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012963183314862697,
+      "loss": 1.3512,
+      "step": 9748
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001296184377620478,
+      "loss": 0.9512,
+      "step": 9749
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012960504179287477,
+      "loss": 0.9708,
+      "step": 9750
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012959164524137148,
+      "loss": 1.0735,
+      "step": 9751
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001295782481078014,
+      "loss": 1.0672,
+      "step": 9752
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012956485039242806,
+      "loss": 0.9382,
+      "step": 9753
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000129551452095515,
+      "loss": 1.1503,
+      "step": 9754
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001295380532173257,
+      "loss": 1.1908,
+      "step": 9755
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012952465375812378,
+      "loss": 0.9908,
+      "step": 9756
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012951125371817281,
+      "loss": 1.0503,
+      "step": 9757
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001294978530977364,
+      "loss": 1.03,
+      "step": 9758
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000129484451897078,
+      "loss": 1.1964,
+      "step": 9759
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012947105011646134,
+      "loss": 1.0476,
+      "step": 9760
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012945764775615,
+      "loss": 0.9865,
+      "step": 9761
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001294442448164076,
+      "loss": 0.918,
+      "step": 9762
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012943084129749774,
+      "loss": 0.9236,
+      "step": 9763
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012941743719968417,
+      "loss": 1.0077,
+      "step": 9764
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001294040325232304,
+      "loss": 1.0963,
+      "step": 9765
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012939062726840023,
+      "loss": 1.0136,
+      "step": 9766
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012937722143545728,
+      "loss": 0.9145,
+      "step": 9767
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012936381502466523,
+      "loss": 0.9143,
+      "step": 9768
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001293504080362878,
+      "loss": 1.1233,
+      "step": 9769
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012933700047058878,
+      "loss": 0.9223,
+      "step": 9770
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012932359232783176,
+      "loss": 1.0971,
+      "step": 9771
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012931018360828056,
+      "loss": 0.8751,
+      "step": 9772
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001292967743121989,
+      "loss": 0.9865,
+      "step": 9773
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001292833644398506,
+      "loss": 0.9834,
+      "step": 9774
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012926995399149932,
+      "loss": 1.1993,
+      "step": 9775
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012925654296740895,
+      "loss": 1.079,
+      "step": 9776
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012924313136784328,
+      "loss": 1.0482,
+      "step": 9777
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012922971919306604,
+      "loss": 1.0242,
+      "step": 9778
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012921630644334106,
+      "loss": 1.0526,
+      "step": 9779
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012920289311893225,
+      "loss": 1.1611,
+      "step": 9780
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012918947922010336,
+      "loss": 0.9623,
+      "step": 9781
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012917606474711826,
+      "loss": 1.2427,
+      "step": 9782
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012916264970024087,
+      "loss": 1.1657,
+      "step": 9783
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012914923407973499,
+      "loss": 1.1715,
+      "step": 9784
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012913581788586456,
+      "loss": 1.0144,
+      "step": 9785
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012912240111889343,
+      "loss": 1.1031,
+      "step": 9786
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012910898377908552,
+      "loss": 1.1393,
+      "step": 9787
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012909556586670478,
+      "loss": 0.9967,
+      "step": 9788
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012908214738201513,
+      "loss": 1.1155,
+      "step": 9789
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012906872832528046,
+      "loss": 1.0296,
+      "step": 9790
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001290553086967648,
+      "loss": 1.0544,
+      "step": 9791
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012904188849673207,
+      "loss": 1.0343,
+      "step": 9792
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012902846772544624,
+      "loss": 0.7549,
+      "step": 9793
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012901504638317132,
+      "loss": 1.1636,
+      "step": 9794
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012900162447017127,
+      "loss": 1.2345,
+      "step": 9795
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012898820198671015,
+      "loss": 1.1719,
+      "step": 9796
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012897477893305195,
+      "loss": 0.9514,
+      "step": 9797
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012896135530946067,
+      "loss": 1.0591,
+      "step": 9798
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012894793111620042,
+      "loss": 1.2719,
+      "step": 9799
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012893450635353523,
+      "loss": 0.9705,
+      "step": 9800
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012892108102172914,
+      "loss": 0.8723,
+      "step": 9801
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012890765512104624,
+      "loss": 0.8087,
+      "step": 9802
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001288942286517506,
+      "loss": 1.0552,
+      "step": 9803
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012888080161410636,
+      "loss": 1.0887,
+      "step": 9804
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012886737400837758,
+      "loss": 0.8425,
+      "step": 9805
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012885394583482844,
+      "loss": 1.0408,
+      "step": 9806
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.000128840517093723,
+      "loss": 0.9627,
+      "step": 9807
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012882708778532545,
+      "loss": 0.8572,
+      "step": 9808
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012881365790989995,
+      "loss": 1.0945,
+      "step": 9809
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012880022746771064,
+      "loss": 1.1027,
+      "step": 9810
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001287867964590217,
+      "loss": 1.24,
+      "step": 9811
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012877336488409735,
+      "loss": 0.9892,
+      "step": 9812
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00012875993274320173,
+      "loss": 1.1445,
+      "step": 9813
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001287465000365991,
+      "loss": 1.174,
+      "step": 9814
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012873306676455365,
+      "loss": 1.1244,
+      "step": 9815
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012871963292732965,
+      "loss": 1.1294,
+      "step": 9816
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001287061985251913,
+      "loss": 1.1792,
+      "step": 9817
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001286927635584029,
+      "loss": 1.0074,
+      "step": 9818
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012867932802722865,
+      "loss": 1.0286,
+      "step": 9819
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012866589193193285,
+      "loss": 1.1369,
+      "step": 9820
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012865245527277986,
+      "loss": 1.0769,
+      "step": 9821
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012863901805003388,
+      "loss": 1.1283,
+      "step": 9822
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012862558026395927,
+      "loss": 0.8875,
+      "step": 9823
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012861214191482034,
+      "loss": 1.0254,
+      "step": 9824
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012859870300288142,
+      "loss": 0.8343,
+      "step": 9825
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012858526352840685,
+      "loss": 1.3717,
+      "step": 9826
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012857182349166102,
+      "loss": 0.8061,
+      "step": 9827
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012855838289290821,
+      "loss": 1.0219,
+      "step": 9828
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012854494173241294,
+      "loss": 1.1616,
+      "step": 9829
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012853150001043944,
+      "loss": 0.9116,
+      "step": 9830
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001285180577272522,
+      "loss": 0.9843,
+      "step": 9831
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012850461488311558,
+      "loss": 1.1537,
+      "step": 9832
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012849117147829408,
+      "loss": 0.9197,
+      "step": 9833
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012847772751305205,
+      "loss": 0.885,
+      "step": 9834
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012846428298765394,
+      "loss": 1.056,
+      "step": 9835
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012845083790236425,
+      "loss": 1.0627,
+      "step": 9836
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001284373922574474,
+      "loss": 1.2232,
+      "step": 9837
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001284239460531679,
+      "loss": 1.0698,
+      "step": 9838
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001284104992897902,
+      "loss": 1.0823,
+      "step": 9839
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012839705196757885,
+      "loss": 0.9678,
+      "step": 9840
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001283836040867983,
+      "loss": 0.8666,
+      "step": 9841
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001283701556477131,
+      "loss": 0.881,
+      "step": 9842
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012835670665058778,
+      "loss": 0.9658,
+      "step": 9843
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012834325709568692,
+      "loss": 0.8586,
+      "step": 9844
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012832980698327496,
+      "loss": 1.0254,
+      "step": 9845
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012831635631361656,
+      "loss": 1.0612,
+      "step": 9846
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012830290508697627,
+      "loss": 0.8232,
+      "step": 9847
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012828945330361865,
+      "loss": 1.0233,
+      "step": 9848
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012827600096380833,
+      "loss": 1.0102,
+      "step": 9849
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012826254806780994,
+      "loss": 0.9361,
+      "step": 9850
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012824909461588803,
+      "loss": 1.0347,
+      "step": 9851
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001282356406083073,
+      "loss": 1.1415,
+      "step": 9852
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012822218604533232,
+      "loss": 1.1167,
+      "step": 9853
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012820873092722778,
+      "loss": 0.8624,
+      "step": 9854
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012819527525425833,
+      "loss": 1.0553,
+      "step": 9855
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012818181902668867,
+      "loss": 0.9488,
+      "step": 9856
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012816836224478344,
+      "loss": 1.142,
+      "step": 9857
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012815490490880734,
+      "loss": 1.0658,
+      "step": 9858
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001281414470190251,
+      "loss": 0.9587,
+      "step": 9859
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012812798857570146,
+      "loss": 1.0066,
+      "step": 9860
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012811452957910109,
+      "loss": 0.9314,
+      "step": 9861
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012810107002948876,
+      "loss": 0.9952,
+      "step": 9862
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012808760992712924,
+      "loss": 0.8536,
+      "step": 9863
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012807414927228723,
+      "loss": 0.8722,
+      "step": 9864
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012806068806522753,
+      "loss": 1.2688,
+      "step": 9865
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001280472263062149,
+      "loss": 1.0233,
+      "step": 9866
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001280337639955142,
+      "loss": 1.1992,
+      "step": 9867
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012802030113339015,
+      "loss": 1.0696,
+      "step": 9868
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012800683772010767,
+      "loss": 1.34,
+      "step": 9869
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012799337375593147,
+      "loss": 1.0449,
+      "step": 9870
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012797990924112647,
+      "loss": 0.8209,
+      "step": 9871
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012796644417595744,
+      "loss": 0.795,
+      "step": 9872
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012795297856068932,
+      "loss": 1.0929,
+      "step": 9873
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012793951239558688,
+      "loss": 1.1338,
+      "step": 9874
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001279260456809151,
+      "loss": 1.1065,
+      "step": 9875
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012791257841693882,
+      "loss": 1.1029,
+      "step": 9876
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012789911060392294,
+      "loss": 1.2072,
+      "step": 9877
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001278856422421324,
+      "loss": 0.9981,
+      "step": 9878
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012787217333183208,
+      "loss": 1.1607,
+      "step": 9879
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012785870387328692,
+      "loss": 1.1966,
+      "step": 9880
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001278452338667619,
+      "loss": 0.9105,
+      "step": 9881
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012783176331252193,
+      "loss": 0.9082,
+      "step": 9882
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000127818292210832,
+      "loss": 1.096,
+      "step": 9883
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012780482056195708,
+      "loss": 1.0882,
+      "step": 9884
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012779134836616218,
+      "loss": 1.2049,
+      "step": 9885
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001277778756237123,
+      "loss": 0.9482,
+      "step": 9886
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001277644023348724,
+      "loss": 0.8397,
+      "step": 9887
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001277509284999075,
+      "loss": 1.0816,
+      "step": 9888
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001277374541190827,
+      "loss": 0.9242,
+      "step": 9889
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012772397919266299,
+      "loss": 0.985,
+      "step": 9890
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012771050372091338,
+      "loss": 0.9386,
+      "step": 9891
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012769702770409905,
+      "loss": 1.1467,
+      "step": 9892
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012768355114248494,
+      "loss": 0.9444,
+      "step": 9893
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012767007403633623,
+      "loss": 0.8063,
+      "step": 9894
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012765659638591797,
+      "loss": 1.0709,
+      "step": 9895
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001276431181914953,
+      "loss": 1.1336,
+      "step": 9896
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012762963945333328,
+      "loss": 0.9851,
+      "step": 9897
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012761616017169708,
+      "loss": 0.9431,
+      "step": 9898
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012760268034685184,
+      "loss": 0.8945,
+      "step": 9899
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001275891999790627,
+      "loss": 0.9176,
+      "step": 9900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012757571906859476,
+      "loss": 1.0639,
+      "step": 9901
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012756223761571328,
+      "loss": 1.0512,
+      "step": 9902
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012754875562068338,
+      "loss": 0.7607,
+      "step": 9903
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012753527308377027,
+      "loss": 0.8913,
+      "step": 9904
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012752179000523917,
+      "loss": 0.9538,
+      "step": 9905
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012750830638535524,
+      "loss": 1.0155,
+      "step": 9906
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012749482222438376,
+      "loss": 1.1228,
+      "step": 9907
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012748133752258994,
+      "loss": 1.012,
+      "step": 9908
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012746785228023904,
+      "loss": 1.2595,
+      "step": 9909
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012745436649759629,
+      "loss": 1.152,
+      "step": 9910
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012744088017492693,
+      "loss": 0.9527,
+      "step": 9911
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012742739331249626,
+      "loss": 0.9631,
+      "step": 9912
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012741390591056962,
+      "loss": 1.1116,
+      "step": 9913
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012740041796941223,
+      "loss": 1.0891,
+      "step": 9914
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012738692948928944,
+      "loss": 1.065,
+      "step": 9915
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012737344047046657,
+      "loss": 1.1362,
+      "step": 9916
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012735995091320892,
+      "loss": 0.8168,
+      "step": 9917
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001273464608177818,
+      "loss": 1.1024,
+      "step": 9918
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001273329701844507,
+      "loss": 1.1353,
+      "step": 9919
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001273194790134808,
+      "loss": 1.0469,
+      "step": 9920
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012730598730513758,
+      "loss": 1.0548,
+      "step": 9921
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001272924950596864,
+      "loss": 0.9106,
+      "step": 9922
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012727900227739266,
+      "loss": 1.0226,
+      "step": 9923
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012726550895852173,
+      "loss": 1.0035,
+      "step": 9924
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012725201510333905,
+      "loss": 1.0252,
+      "step": 9925
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012723852071211002,
+      "loss": 1.1194,
+      "step": 9926
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012722502578510012,
+      "loss": 1.0361,
+      "step": 9927
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012721153032257477,
+      "loss": 0.8863,
+      "step": 9928
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012719803432479938,
+      "loss": 0.902,
+      "step": 9929
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001271845377920395,
+      "loss": 0.8486,
+      "step": 9930
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012717104072456053,
+      "loss": 1.023,
+      "step": 9931
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012715754312262804,
+      "loss": 0.9858,
+      "step": 9932
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012714404498650743,
+      "loss": 0.9503,
+      "step": 9933
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012713054631646427,
+      "loss": 1.0777,
+      "step": 9934
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012711704711276407,
+      "loss": 1.017,
+      "step": 9935
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001271035473756724,
+      "loss": 0.9515,
+      "step": 9936
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012709004710545466,
+      "loss": 1.1534,
+      "step": 9937
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012707654630237655,
+      "loss": 0.9322,
+      "step": 9938
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012706304496670358,
+      "loss": 0.9814,
+      "step": 9939
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001270495430987013,
+      "loss": 1.0291,
+      "step": 9940
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001270360406986353,
+      "loss": 0.918,
+      "step": 9941
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001270225377667712,
+      "loss": 1.2833,
+      "step": 9942
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012700903430337457,
+      "loss": 1.0983,
+      "step": 9943
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000126995530308711,
+      "loss": 0.9474,
+      "step": 9944
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012698202578304617,
+      "loss": 1.0655,
+      "step": 9945
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001269685207266457,
+      "loss": 0.9832,
+      "step": 9946
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012695501513977522,
+      "loss": 0.9923,
+      "step": 9947
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001269415090227004,
+      "loss": 1.1128,
+      "step": 9948
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012692800237568686,
+      "loss": 0.9207,
+      "step": 9949
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012691449519900032,
+      "loss": 0.9862,
+      "step": 9950
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012690098749290648,
+      "loss": 1.1188,
+      "step": 9951
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012688747925767096,
+      "loss": 1.1092,
+      "step": 9952
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012687397049355957,
+      "loss": 1.1287,
+      "step": 9953
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012686046120083792,
+      "loss": 1.0795,
+      "step": 9954
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012684695137977185,
+      "loss": 1.1521,
+      "step": 9955
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000126833441030627,
+      "loss": 1.0273,
+      "step": 9956
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012681993015366917,
+      "loss": 0.9042,
+      "step": 9957
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001268064187491641,
+      "loss": 1.0535,
+      "step": 9958
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001267929068173776,
+      "loss": 1.0988,
+      "step": 9959
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012677939435857532,
+      "loss": 1.0834,
+      "step": 9960
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012676588137302327,
+      "loss": 1.1395,
+      "step": 9961
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012675236786098703,
+      "loss": 1.0626,
+      "step": 9962
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012673885382273253,
+      "loss": 1.0896,
+      "step": 9963
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012672533925852557,
+      "loss": 0.8824,
+      "step": 9964
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000126711824168632,
+      "loss": 1.0204,
+      "step": 9965
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001266983085533176,
+      "loss": 0.8629,
+      "step": 9966
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012668479241284826,
+      "loss": 1.0714,
+      "step": 9967
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012667127574748986,
+      "loss": 0.9023,
+      "step": 9968
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012665775855750825,
+      "loss": 1.1856,
+      "step": 9969
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012664424084316928,
+      "loss": 0.8944,
+      "step": 9970
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012663072260473893,
+      "loss": 1.0538,
+      "step": 9971
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012661720384248302,
+      "loss": 0.9939,
+      "step": 9972
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012660368455666752,
+      "loss": 1.1139,
+      "step": 9973
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001265901647475583,
+      "loss": 1.0741,
+      "step": 9974
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012657664441542138,
+      "loss": 0.9838,
+      "step": 9975
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001265631235605226,
+      "loss": 1.0223,
+      "step": 9976
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012654960218312796,
+      "loss": 1.1486,
+      "step": 9977
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001265360802835035,
+      "loss": 1.0364,
+      "step": 9978
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012652255786191503,
+      "loss": 1.0525,
+      "step": 9979
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012650903491862866,
+      "loss": 0.9743,
+      "step": 9980
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012649551145391035,
+      "loss": 1.1105,
+      "step": 9981
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012648198746802617,
+      "loss": 0.9729,
+      "step": 9982
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000126468462961242,
+      "loss": 0.9927,
+      "step": 9983
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012645493793382402,
+      "loss": 0.8795,
+      "step": 9984
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012644141238603814,
+      "loss": 1.1359,
+      "step": 9985
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001264278863181505,
+      "loss": 1.0525,
+      "step": 9986
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012641435973042705,
+      "loss": 1.12,
+      "step": 9987
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012640083262313397,
+      "loss": 0.8648,
+      "step": 9988
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001263873049965373,
+      "loss": 1.153,
+      "step": 9989
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001263737768509031,
+      "loss": 1.0362,
+      "step": 9990
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001263602481864975,
+      "loss": 1.0616,
+      "step": 9991
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012634671900358662,
+      "loss": 0.9949,
+      "step": 9992
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012633318930243648,
+      "loss": 0.9625,
+      "step": 9993
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012631965908331335,
+      "loss": 1.0716,
+      "step": 9994
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012630612834648328,
+      "loss": 0.9932,
+      "step": 9995
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012629259709221244,
+      "loss": 1.1541,
+      "step": 9996
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000126279065320767,
+      "loss": 0.9574,
+      "step": 9997
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001262655330324131,
+      "loss": 1.0133,
+      "step": 9998
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000126252000227417,
+      "loss": 1.0742,
+      "step": 9999
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012623846690604476,
+      "loss": 0.9406,
+      "step": 10000
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012622493306856266,
+      "loss": 1.0816,
+      "step": 10001
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012621139871523695,
+      "loss": 1.0962,
+      "step": 10002
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012619786384633375,
+      "loss": 1.1709,
+      "step": 10003
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012618432846211937,
+      "loss": 1.2368,
+      "step": 10004
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012617079256286,
+      "loss": 0.8775,
+      "step": 10005
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001261572561488219,
+      "loss": 0.9393,
+      "step": 10006
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012614371922027142,
+      "loss": 0.9553,
+      "step": 10007
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001261301817774747,
+      "loss": 0.9167,
+      "step": 10008
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001261166438206981,
+      "loss": 1.0473,
+      "step": 10009
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012610310535020783,
+      "loss": 0.86,
+      "step": 10010
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001260895663662703,
+      "loss": 1.0423,
+      "step": 10011
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001260760268691518,
+      "loss": 1.1116,
+      "step": 10012
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012606248685911853,
+      "loss": 1.0399,
+      "step": 10013
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012604894633643697,
+      "loss": 0.8865,
+      "step": 10014
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012603540530137344,
+      "loss": 1.0975,
+      "step": 10015
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012602186375419418,
+      "loss": 0.9602,
+      "step": 10016
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012600832169516567,
+      "loss": 0.8957,
+      "step": 10017
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012599477912455425,
+      "loss": 0.8137,
+      "step": 10018
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012598123604262625,
+      "loss": 0.9623,
+      "step": 10019
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012596769244964813,
+      "loss": 1.0997,
+      "step": 10020
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012595414834588626,
+      "loss": 1.0537,
+      "step": 10021
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012594060373160708,
+      "loss": 1.0437,
+      "step": 10022
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012592705860707694,
+      "loss": 1.2475,
+      "step": 10023
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012591351297256236,
+      "loss": 1.2344,
+      "step": 10024
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012589996682832975,
+      "loss": 0.8695,
+      "step": 10025
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012588642017464556,
+      "loss": 0.9688,
+      "step": 10026
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012587287301177625,
+      "loss": 0.9807,
+      "step": 10027
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012585932533998827,
+      "loss": 0.9926,
+      "step": 10028
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012584577715954814,
+      "loss": 1.1399,
+      "step": 10029
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012583222847072233,
+      "loss": 1.1566,
+      "step": 10030
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012581867927377734,
+      "loss": 0.9919,
+      "step": 10031
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012580512956897972,
+      "loss": 1.0052,
+      "step": 10032
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012579157935659592,
+      "loss": 0.9127,
+      "step": 10033
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012577802863689255,
+      "loss": 0.9054,
+      "step": 10034
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001257644774101361,
+      "loss": 1.1701,
+      "step": 10035
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012575092567659315,
+      "loss": 1.1394,
+      "step": 10036
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012573737343653024,
+      "loss": 0.9686,
+      "step": 10037
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.000125723820690214,
+      "loss": 1.1792,
+      "step": 10038
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001257102674379109,
+      "loss": 0.9445,
+      "step": 10039
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012569671367988763,
+      "loss": 0.891,
+      "step": 10040
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012568315941641076,
+      "loss": 0.9414,
+      "step": 10041
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001256696046477469,
+      "loss": 0.8536,
+      "step": 10042
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012565604937416267,
+      "loss": 0.9546,
+      "step": 10043
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.0001256424935959247,
+      "loss": 1.0309,
+      "step": 10044
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012562893731329967,
+      "loss": 1.0815,
+      "step": 10045
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012561538052655418,
+      "loss": 1.1115,
+      "step": 10046
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012560182323595492,
+      "loss": 1.091,
+      "step": 10047
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012558826544176855,
+      "loss": 0.8942,
+      "step": 10048
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012557470714426173,
+      "loss": 1.042,
+      "step": 10049
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012556114834370117,
+      "loss": 0.895,
+      "step": 10050
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012554758904035365,
+      "loss": 1.0292,
+      "step": 10051
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012553402923448577,
+      "loss": 1.0224,
+      "step": 10052
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012552046892636427,
+      "loss": 0.9503,
+      "step": 10053
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001255069081162559,
+      "loss": 0.8982,
+      "step": 10054
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012549334680442742,
+      "loss": 1.1777,
+      "step": 10055
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012547978499114553,
+      "loss": 1.023,
+      "step": 10056
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012546622267667706,
+      "loss": 1.1461,
+      "step": 10057
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001254526598612887,
+      "loss": 0.95,
+      "step": 10058
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001254390965452473,
+      "loss": 0.9425,
+      "step": 10059
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012542553272881963,
+      "loss": 0.7976,
+      "step": 10060
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012541196841227246,
+      "loss": 1.1781,
+      "step": 10061
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012539840359587263,
+      "loss": 1.2203,
+      "step": 10062
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012538483827988697,
+      "loss": 1.0018,
+      "step": 10063
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012537127246458227,
+      "loss": 0.7699,
+      "step": 10064
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012535770615022538,
+      "loss": 1.0499,
+      "step": 10065
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012534413933708316,
+      "loss": 0.9865,
+      "step": 10066
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001253305720254225,
+      "loss": 1.0838,
+      "step": 10067
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001253170042155102,
+      "loss": 1.0079,
+      "step": 10068
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012530343590761318,
+      "loss": 0.9293,
+      "step": 10069
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012528986710199836,
+      "loss": 1.1217,
+      "step": 10070
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012527629779893255,
+      "loss": 1.0303,
+      "step": 10071
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012526272799868275,
+      "loss": 1.0222,
+      "step": 10072
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001252491577015158,
+      "loss": 0.9641,
+      "step": 10073
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001252355869076987,
+      "loss": 0.7662,
+      "step": 10074
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012522201561749832,
+      "loss": 0.7702,
+      "step": 10075
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012520844383118163,
+      "loss": 1.0618,
+      "step": 10076
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012519487154901562,
+      "loss": 1.05,
+      "step": 10077
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012518129877126725,
+      "loss": 0.9012,
+      "step": 10078
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012516772549820343,
+      "loss": 1.0114,
+      "step": 10079
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012515415173009123,
+      "loss": 1.2072,
+      "step": 10080
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012514057746719758,
+      "loss": 0.9668,
+      "step": 10081
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012512700270978953,
+      "loss": 0.9918,
+      "step": 10082
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012511342745813407,
+      "loss": 0.8206,
+      "step": 10083
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012509985171249826,
+      "loss": 0.9309,
+      "step": 10084
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001250862754731491,
+      "loss": 1.0542,
+      "step": 10085
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001250726987403536,
+      "loss": 1.0027,
+      "step": 10086
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001250591215143789,
+      "loss": 0.8039,
+      "step": 10087
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000125045543795492,
+      "loss": 0.9978,
+      "step": 10088
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012503196558396,
+      "loss": 0.893,
+      "step": 10089
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012501838688004993,
+      "loss": 1.0614,
+      "step": 10090
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000125004807684029,
+      "loss": 1.101,
+      "step": 10091
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012499122799616418,
+      "loss": 1.1446,
+      "step": 10092
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001249776478167227,
+      "loss": 1.0844,
+      "step": 10093
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001249640671459716,
+      "loss": 0.9853,
+      "step": 10094
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012495048598417803,
+      "loss": 1.0455,
+      "step": 10095
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012493690433160913,
+      "loss": 1.0451,
+      "step": 10096
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012492332218853208,
+      "loss": 1.0584,
+      "step": 10097
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012490973955521403,
+      "loss": 1.2193,
+      "step": 10098
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012489615643192212,
+      "loss": 1.1035,
+      "step": 10099
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012488257281892352,
+      "loss": 0.8512,
+      "step": 10100
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001248689887164855,
+      "loss": 1.1869,
+      "step": 10101
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012485540412487516,
+      "loss": 1.0811,
+      "step": 10102
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012484181904435978,
+      "loss": 1.0845,
+      "step": 10103
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012482823347520657,
+      "loss": 0.9951,
+      "step": 10104
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001248146474176827,
+      "loss": 1.088,
+      "step": 10105
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012480106087205548,
+      "loss": 0.9509,
+      "step": 10106
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001247874738385921,
+      "loss": 0.9704,
+      "step": 10107
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012477388631755985,
+      "loss": 1.2454,
+      "step": 10108
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000124760298309226,
+      "loss": 1.1733,
+      "step": 10109
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001247467098138578,
+      "loss": 0.9745,
+      "step": 10110
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012473312083172256,
+      "loss": 0.9588,
+      "step": 10111
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012471953136308753,
+      "loss": 0.9724,
+      "step": 10112
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012470594140822006,
+      "loss": 1.1093,
+      "step": 10113
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012469235096738748,
+      "loss": 1.0965,
+      "step": 10114
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012467876004085707,
+      "loss": 0.9906,
+      "step": 10115
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001246651686288962,
+      "loss": 1.017,
+      "step": 10116
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012465157673177216,
+      "loss": 1.0093,
+      "step": 10117
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012463798434975238,
+      "loss": 1.0243,
+      "step": 10118
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001246243914831041,
+      "loss": 1.0633,
+      "step": 10119
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012461079813209485,
+      "loss": 0.94,
+      "step": 10120
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012459720429699187,
+      "loss": 0.9408,
+      "step": 10121
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012458360997806263,
+      "loss": 1.111,
+      "step": 10122
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012457001517557448,
+      "loss": 0.9156,
+      "step": 10123
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001245564198897949,
+      "loss": 0.9577,
+      "step": 10124
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001245428241209912,
+      "loss": 0.9286,
+      "step": 10125
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012452922786943097,
+      "loss": 0.8953,
+      "step": 10126
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012451563113538146,
+      "loss": 0.9637,
+      "step": 10127
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012450203391911027,
+      "loss": 1.1481,
+      "step": 10128
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012448843622088476,
+      "loss": 1.1218,
+      "step": 10129
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012447483804097246,
+      "loss": 1.1331,
+      "step": 10130
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001244612393796408,
+      "loss": 0.8682,
+      "step": 10131
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012444764023715729,
+      "loss": 1.0025,
+      "step": 10132
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001244340406137894,
+      "loss": 1.0964,
+      "step": 10133
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001244204405098047,
+      "loss": 0.9942,
+      "step": 10134
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001244068399254706,
+      "loss": 1.1342,
+      "step": 10135
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001243932388610547,
+      "loss": 1.0847,
+      "step": 10136
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012437963731682454,
+      "loss": 0.8269,
+      "step": 10137
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001243660352930476,
+      "loss": 1.035,
+      "step": 10138
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001243524327899915,
+      "loss": 0.9451,
+      "step": 10139
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012433882980792372,
+      "loss": 1.108,
+      "step": 10140
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012432522634711192,
+      "loss": 1.1899,
+      "step": 10141
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001243116224078236,
+      "loss": 1.1207,
+      "step": 10142
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001242980179903264,
+      "loss": 1.0374,
+      "step": 10143
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001242844130948879,
+      "loss": 0.9349,
+      "step": 10144
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001242708077217757,
+      "loss": 1.0808,
+      "step": 10145
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012425720187125746,
+      "loss": 0.8089,
+      "step": 10146
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012424359554360077,
+      "loss": 1.0746,
+      "step": 10147
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012422998873907323,
+      "loss": 1.1382,
+      "step": 10148
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001242163814579426,
+      "loss": 0.989,
+      "step": 10149
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001242027737004764,
+      "loss": 1.0675,
+      "step": 10150
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012418916546694237,
+      "loss": 1.2152,
+      "step": 10151
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012417555675760817,
+      "loss": 1.132,
+      "step": 10152
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001241619475727415,
+      "loss": 1.0726,
+      "step": 10153
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012414833791261003,
+      "loss": 1.2066,
+      "step": 10154
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012413472777748146,
+      "loss": 1.1366,
+      "step": 10155
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001241211171676235,
+      "loss": 0.8953,
+      "step": 10156
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012410750608330388,
+      "loss": 0.9613,
+      "step": 10157
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012409389452479032,
+      "loss": 1.1926,
+      "step": 10158
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012408028249235057,
+      "loss": 1.2012,
+      "step": 10159
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001240666699862524,
+      "loss": 1.1262,
+      "step": 10160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001240530570067635,
+      "loss": 1.0397,
+      "step": 10161
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001240394435541517,
+      "loss": 1.0496,
+      "step": 10162
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012402582962868475,
+      "loss": 1.0675,
+      "step": 10163
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001240122152306305,
+      "loss": 1.2183,
+      "step": 10164
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001239986003602566,
+      "loss": 1.0751,
+      "step": 10165
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012398498501783097,
+      "loss": 0.9079,
+      "step": 10166
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001239713692036214,
+      "loss": 0.9076,
+      "step": 10167
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012395775291789568,
+      "loss": 1.1042,
+      "step": 10168
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012394413616092172,
+      "loss": 1.0002,
+      "step": 10169
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012393051893296728,
+      "loss": 0.9768,
+      "step": 10170
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012391690123430023,
+      "loss": 0.8907,
+      "step": 10171
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012390328306518845,
+      "loss": 0.8712,
+      "step": 10172
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001238896644258998,
+      "loss": 1.1004,
+      "step": 10173
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001238760453167021,
+      "loss": 1.0409,
+      "step": 10174
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012386242573786334,
+      "loss": 0.909,
+      "step": 10175
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001238488056896514,
+      "loss": 1.111,
+      "step": 10176
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012383518517233415,
+      "loss": 1.1593,
+      "step": 10177
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012382156418617947,
+      "loss": 1.1156,
+      "step": 10178
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012380794273145535,
+      "loss": 0.9002,
+      "step": 10179
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001237943208084297,
+      "loss": 0.9269,
+      "step": 10180
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012378069841737044,
+      "loss": 1.0831,
+      "step": 10181
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012376707555854555,
+      "loss": 1.216,
+      "step": 10182
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012375345223222304,
+      "loss": 0.979,
+      "step": 10183
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012373982843867076,
+      "loss": 1.1937,
+      "step": 10184
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012372620417815678,
+      "loss": 0.9717,
+      "step": 10185
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012371257945094905,
+      "loss": 1.1096,
+      "step": 10186
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001236989542573156,
+      "loss": 1.179,
+      "step": 10187
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001236853285975244,
+      "loss": 1.1415,
+      "step": 10188
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012367170247184355,
+      "loss": 1.1079,
+      "step": 10189
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012365807588054096,
+      "loss": 0.9674,
+      "step": 10190
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012364444882388467,
+      "loss": 1.1086,
+      "step": 10191
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012363082130214284,
+      "loss": 1.087,
+      "step": 10192
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012361719331558345,
+      "loss": 0.9926,
+      "step": 10193
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012360356486447452,
+      "loss": 1.072,
+      "step": 10194
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012358993594908423,
+      "loss": 1.0278,
+      "step": 10195
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012357630656968053,
+      "loss": 1.0598,
+      "step": 10196
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001235626767265316,
+      "loss": 0.9829,
+      "step": 10197
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012354904641990552,
+      "loss": 1.2252,
+      "step": 10198
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001235354156500704,
+      "loss": 0.9979,
+      "step": 10199
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012352178441729433,
+      "loss": 1.0757,
+      "step": 10200
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001235081527218455,
+      "loss": 0.9861,
+      "step": 10201
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012349452056399193,
+      "loss": 1.166,
+      "step": 10202
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012348088794400189,
+      "loss": 1.1226,
+      "step": 10203
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012346725486214345,
+      "loss": 0.8428,
+      "step": 10204
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001234536213186848,
+      "loss": 1.0048,
+      "step": 10205
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012343998731389412,
+      "loss": 1.0898,
+      "step": 10206
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012342635284803957,
+      "loss": 1.1269,
+      "step": 10207
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012341271792138935,
+      "loss": 1.0452,
+      "step": 10208
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012339908253421167,
+      "loss": 1.0772,
+      "step": 10209
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012338544668677475,
+      "loss": 0.9723,
+      "step": 10210
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012337181037934674,
+      "loss": 0.9315,
+      "step": 10211
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012335817361219594,
+      "loss": 0.9623,
+      "step": 10212
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012334453638559057,
+      "loss": 0.9988,
+      "step": 10213
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012333089869979882,
+      "loss": 0.812,
+      "step": 10214
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000123317260555089,
+      "loss": 1.0265,
+      "step": 10215
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012330362195172936,
+      "loss": 0.8948,
+      "step": 10216
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012328998288998815,
+      "loss": 1.0609,
+      "step": 10217
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012327634337013364,
+      "loss": 0.8472,
+      "step": 10218
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001232627033924342,
+      "loss": 0.9566,
+      "step": 10219
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.000123249062957158,
+      "loss": 0.9738,
+      "step": 10220
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012323542206457347,
+      "loss": 1.0792,
+      "step": 10221
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001232217807149489,
+      "loss": 1.1233,
+      "step": 10222
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012320813890855255,
+      "loss": 0.9186,
+      "step": 10223
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001231944966456528,
+      "loss": 1.0314,
+      "step": 10224
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012318085392651798,
+      "loss": 1.0757,
+      "step": 10225
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012316721075141644,
+      "loss": 0.984,
+      "step": 10226
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012315356712061656,
+      "loss": 0.7523,
+      "step": 10227
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001231399230343867,
+      "loss": 1.1745,
+      "step": 10228
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012312627849299523,
+      "loss": 0.9661,
+      "step": 10229
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012311263349671053,
+      "loss": 0.9101,
+      "step": 10230
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012309898804580105,
+      "loss": 0.9683,
+      "step": 10231
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012308534214053513,
+      "loss": 1.1596,
+      "step": 10232
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012307169578118122,
+      "loss": 1.0318,
+      "step": 10233
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001230580489680077,
+      "loss": 1.3271,
+      "step": 10234
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001230444017012831,
+      "loss": 1.0989,
+      "step": 10235
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001230307539812758,
+      "loss": 0.9347,
+      "step": 10236
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012301710580825423,
+      "loss": 1.0944,
+      "step": 10237
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012300345718248683,
+      "loss": 1.1876,
+      "step": 10238
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012298980810424213,
+      "loss": 1.067,
+      "step": 10239
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001229761585737886,
+      "loss": 0.7846,
+      "step": 10240
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001229625085913947,
+      "loss": 1.1448,
+      "step": 10241
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012294885815732892,
+      "loss": 1.1615,
+      "step": 10242
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001229352072718598,
+      "loss": 0.8767,
+      "step": 10243
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001229215559352558,
+      "loss": 0.9838,
+      "step": 10244
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001229079041477855,
+      "loss": 1.0824,
+      "step": 10245
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001228942519097174,
+      "loss": 1.1924,
+      "step": 10246
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012288059922132,
+      "loss": 1.1156,
+      "step": 10247
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012286694608286197,
+      "loss": 1.1264,
+      "step": 10248
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001228532924946117,
+      "loss": 1.2018,
+      "step": 10249
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001228396384568379,
+      "loss": 0.9888,
+      "step": 10250
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012282598396980908,
+      "loss": 0.9116,
+      "step": 10251
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001228123290337938,
+      "loss": 1.0441,
+      "step": 10252
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012279867364906071,
+      "loss": 1.1361,
+      "step": 10253
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012278501781587834,
+      "loss": 0.9644,
+      "step": 10254
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012277136153451539,
+      "loss": 1.0706,
+      "step": 10255
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012275770480524043,
+      "loss": 1.0637,
+      "step": 10256
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012274404762832204,
+      "loss": 0.9409,
+      "step": 10257
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012273039000402895,
+      "loss": 0.9211,
+      "step": 10258
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012271673193262975,
+      "loss": 0.9665,
+      "step": 10259
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001227030734143931,
+      "loss": 0.9619,
+      "step": 10260
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012268941444958765,
+      "loss": 0.9155,
+      "step": 10261
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001226757550384821,
+      "loss": 0.9925,
+      "step": 10262
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012266209518134512,
+      "loss": 1.0332,
+      "step": 10263
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012264843487844542,
+      "loss": 1.0931,
+      "step": 10264
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012263477413005164,
+      "loss": 1.0017,
+      "step": 10265
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012262111293643254,
+      "loss": 0.8661,
+      "step": 10266
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001226074512978568,
+      "loss": 1.0719,
+      "step": 10267
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001225937892145932,
+      "loss": 1.023,
+      "step": 10268
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001225801266869104,
+      "loss": 1.175,
+      "step": 10269
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012256646371507717,
+      "loss": 1.1814,
+      "step": 10270
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012255280029936224,
+      "loss": 0.9881,
+      "step": 10271
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012253913644003444,
+      "loss": 0.9426,
+      "step": 10272
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001225254721373625,
+      "loss": 1.1231,
+      "step": 10273
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012251180739161513,
+      "loss": 1.087,
+      "step": 10274
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012249814220306125,
+      "loss": 0.9862,
+      "step": 10275
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.00012248447657196954,
+      "loss": 1.0758,
+      "step": 10276
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012247081049860883,
+      "loss": 1.1419,
+      "step": 10277
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012245714398324793,
+      "loss": 0.953,
+      "step": 10278
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001224434770261557,
+      "loss": 1.0655,
+      "step": 10279
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012242980962760093,
+      "loss": 0.9409,
+      "step": 10280
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012241614178785248,
+      "loss": 0.8963,
+      "step": 10281
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012240247350717916,
+      "loss": 1.0326,
+      "step": 10282
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012238880478584985,
+      "loss": 1.1133,
+      "step": 10283
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001223751356241334,
+      "loss": 0.9093,
+      "step": 10284
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012236146602229874,
+      "loss": 1.3069,
+      "step": 10285
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012234779598061467,
+      "loss": 1.1067,
+      "step": 10286
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001223341254993501,
+      "loss": 0.9797,
+      "step": 10287
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012232045457877396,
+      "loss": 1.071,
+      "step": 10288
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012230678321915514,
+      "loss": 1.0811,
+      "step": 10289
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012229311142076255,
+      "loss": 1.0695,
+      "step": 10290
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012227943918386512,
+      "loss": 1.0128,
+      "step": 10291
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012226576650873177,
+      "loss": 1.0176,
+      "step": 10292
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012225209339563145,
+      "loss": 1.1442,
+      "step": 10293
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001222384198448331,
+      "loss": 0.9011,
+      "step": 10294
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001222247458566057,
+      "loss": 0.8329,
+      "step": 10295
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012221107143121823,
+      "loss": 1.2551,
+      "step": 10296
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001221973965689396,
+      "loss": 1.086,
+      "step": 10297
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012218372127003889,
+      "loss": 0.951,
+      "step": 10298
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012217004553478496,
+      "loss": 0.9405,
+      "step": 10299
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012215636936344695,
+      "loss": 0.9772,
+      "step": 10300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001221426927562938,
+      "loss": 1.0556,
+      "step": 10301
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012212901571359452,
+      "loss": 1.0989,
+      "step": 10302
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012211533823561815,
+      "loss": 1.0199,
+      "step": 10303
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012210166032263376,
+      "loss": 1.0052,
+      "step": 10304
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012208798197491032,
+      "loss": 0.9345,
+      "step": 10305
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012207430319271696,
+      "loss": 0.8953,
+      "step": 10306
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012206062397632267,
+      "loss": 0.893,
+      "step": 10307
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012204694432599662,
+      "loss": 1.0168,
+      "step": 10308
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012203326424200778,
+      "loss": 0.9919,
+      "step": 10309
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001220195837246253,
+      "loss": 0.8477,
+      "step": 10310
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012200590277411823,
+      "loss": 0.9706,
+      "step": 10311
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012199222139075575,
+      "loss": 1.119,
+      "step": 10312
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012197853957480689,
+      "loss": 0.9546,
+      "step": 10313
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012196485732654081,
+      "loss": 1.086,
+      "step": 10314
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012195117464622665,
+      "loss": 0.8929,
+      "step": 10315
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012193749153413351,
+      "loss": 0.9701,
+      "step": 10316
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012192380799053058,
+      "loss": 0.9776,
+      "step": 10317
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012191012401568698,
+      "loss": 1.0983,
+      "step": 10318
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012189643960987192,
+      "loss": 1.0669,
+      "step": 10319
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012188275477335449,
+      "loss": 0.942,
+      "step": 10320
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012186906950640395,
+      "loss": 1.0448,
+      "step": 10321
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012185538380928945,
+      "loss": 0.7215,
+      "step": 10322
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001218416976822802,
+      "loss": 1.0667,
+      "step": 10323
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001218280111256454,
+      "loss": 1.033,
+      "step": 10324
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012181432413965428,
+      "loss": 1.0002,
+      "step": 10325
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012180063672457602,
+      "loss": 1.1269,
+      "step": 10326
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001217869488806799,
+      "loss": 0.8757,
+      "step": 10327
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012177326060823513,
+      "loss": 1.0472,
+      "step": 10328
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012175957190751097,
+      "loss": 0.9837,
+      "step": 10329
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012174588277877665,
+      "loss": 1.2186,
+      "step": 10330
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001217321932223015,
+      "loss": 0.9885,
+      "step": 10331
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012171850323835472,
+      "loss": 1.093,
+      "step": 10332
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012170481282720563,
+      "loss": 1.0593,
+      "step": 10333
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012169112198912351,
+      "loss": 1.0457,
+      "step": 10334
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012167743072437767,
+      "loss": 0.8689,
+      "step": 10335
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012166373903323743,
+      "loss": 1.1548,
+      "step": 10336
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012165004691597203,
+      "loss": 0.9281,
+      "step": 10337
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012163635437285089,
+      "loss": 1.0803,
+      "step": 10338
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012162266140414327,
+      "loss": 0.8635,
+      "step": 10339
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012160896801011855,
+      "loss": 1.0328,
+      "step": 10340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012159527419104607,
+      "loss": 0.9723,
+      "step": 10341
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001215815799471952,
+      "loss": 1.1425,
+      "step": 10342
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012156788527883522,
+      "loss": 1.1342,
+      "step": 10343
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012155419018623567,
+      "loss": 0.9367,
+      "step": 10344
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012154049466966578,
+      "loss": 1.1204,
+      "step": 10345
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000121526798729395,
+      "loss": 1.1034,
+      "step": 10346
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001215131023656927,
+      "loss": 0.9162,
+      "step": 10347
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012149940557882837,
+      "loss": 1.2065,
+      "step": 10348
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012148570836907133,
+      "loss": 1.2085,
+      "step": 10349
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012147201073669106,
+      "loss": 0.9705,
+      "step": 10350
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012145831268195695,
+      "loss": 1.0115,
+      "step": 10351
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001214446142051385,
+      "loss": 1.1227,
+      "step": 10352
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012143091530650508,
+      "loss": 1.092,
+      "step": 10353
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012141721598632623,
+      "loss": 1.0685,
+      "step": 10354
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012140351624487134,
+      "loss": 1.2136,
+      "step": 10355
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012138981608240991,
+      "loss": 0.9328,
+      "step": 10356
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012137611549921146,
+      "loss": 0.9229,
+      "step": 10357
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012136241449554541,
+      "loss": 0.9731,
+      "step": 10358
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012134871307168136,
+      "loss": 0.9996,
+      "step": 10359
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012133501122788871,
+      "loss": 0.9374,
+      "step": 10360
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012132130896443703,
+      "loss": 0.979,
+      "step": 10361
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012130760628159583,
+      "loss": 1.1637,
+      "step": 10362
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012129390317963463,
+      "loss": 1.1607,
+      "step": 10363
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012128019965882301,
+      "loss": 1.1021,
+      "step": 10364
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012126649571943047,
+      "loss": 0.848,
+      "step": 10365
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012125279136172658,
+      "loss": 0.9064,
+      "step": 10366
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012123908658598093,
+      "loss": 1.0121,
+      "step": 10367
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012122538139246307,
+      "loss": 1.0373,
+      "step": 10368
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012121167578144261,
+      "loss": 1.0966,
+      "step": 10369
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012119796975318907,
+      "loss": 1.3561,
+      "step": 10370
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012118426330797213,
+      "loss": 1.0114,
+      "step": 10371
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012117055644606133,
+      "loss": 0.9639,
+      "step": 10372
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012115684916772629,
+      "loss": 0.8673,
+      "step": 10373
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012114314147323669,
+      "loss": 0.9148,
+      "step": 10374
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012112943336286211,
+      "loss": 1.0389,
+      "step": 10375
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012111572483687217,
+      "loss": 0.9955,
+      "step": 10376
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001211020158955366,
+      "loss": 1.0037,
+      "step": 10377
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012108830653912498,
+      "loss": 0.9389,
+      "step": 10378
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012107459676790698,
+      "loss": 1.2091,
+      "step": 10379
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012106088658215227,
+      "loss": 1.1942,
+      "step": 10380
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012104717598213056,
+      "loss": 0.9203,
+      "step": 10381
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012103346496811154,
+      "loss": 1.0797,
+      "step": 10382
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012101975354036487,
+      "loss": 0.9287,
+      "step": 10383
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012100604169916025,
+      "loss": 1.1514,
+      "step": 10384
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012099232944476744,
+      "loss": 1.2069,
+      "step": 10385
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012097861677745613,
+      "loss": 0.9774,
+      "step": 10386
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012096490369749603,
+      "loss": 0.9513,
+      "step": 10387
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012095119020515691,
+      "loss": 0.9099,
+      "step": 10388
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012093747630070844,
+      "loss": 1.1753,
+      "step": 10389
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001209237619844205,
+      "loss": 0.744,
+      "step": 10390
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012091004725656274,
+      "loss": 0.9908,
+      "step": 10391
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000120896332117405,
+      "loss": 1.0751,
+      "step": 10392
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000120882616567217,
+      "loss": 0.9153,
+      "step": 10393
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012086890060626857,
+      "loss": 1.0181,
+      "step": 10394
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012085518423482944,
+      "loss": 1.0135,
+      "step": 10395
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012084146745316949,
+      "loss": 0.8723,
+      "step": 10396
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012082775026155846,
+      "loss": 1.0354,
+      "step": 10397
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012081403266026625,
+      "loss": 1.0113,
+      "step": 10398
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012080031464956256,
+      "loss": 1.0381,
+      "step": 10399
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012078659622971735,
+      "loss": 1.0042,
+      "step": 10400
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012077287740100038,
+      "loss": 1.0608,
+      "step": 10401
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012075915816368155,
+      "loss": 1.1379,
+      "step": 10402
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012074543851803067,
+      "loss": 0.9431,
+      "step": 10403
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012073171846431761,
+      "loss": 1.0319,
+      "step": 10404
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012071799800281231,
+      "loss": 1.0992,
+      "step": 10405
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012070427713378456,
+      "loss": 1.0189,
+      "step": 10406
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012069055585750431,
+      "loss": 1.0718,
+      "step": 10407
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012067683417424142,
+      "loss": 1.0772,
+      "step": 10408
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012066311208426581,
+      "loss": 1.0136,
+      "step": 10409
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012064938958784742,
+      "loss": 1.016,
+      "step": 10410
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012063566668525614,
+      "loss": 1.0359,
+      "step": 10411
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012062194337676192,
+      "loss": 1.1757,
+      "step": 10412
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012060821966263465,
+      "loss": 0.9492,
+      "step": 10413
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001205944955431443,
+      "loss": 1.012,
+      "step": 10414
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012058077101856089,
+      "loss": 1.0186,
+      "step": 10415
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012056704608915426,
+      "loss": 1.0455,
+      "step": 10416
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012055332075519448,
+      "loss": 0.9885,
+      "step": 10417
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012053959501695145,
+      "loss": 1.1316,
+      "step": 10418
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012052586887469522,
+      "loss": 1.1083,
+      "step": 10419
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012051214232869575,
+      "loss": 1.2952,
+      "step": 10420
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012049841537922307,
+      "loss": 1.1606,
+      "step": 10421
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001204846880265471,
+      "loss": 1.0388,
+      "step": 10422
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012047096027093798,
+      "loss": 1.0103,
+      "step": 10423
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012045723211266567,
+      "loss": 1.0492,
+      "step": 10424
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012044350355200019,
+      "loss": 1.0499,
+      "step": 10425
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012042977458921159,
+      "loss": 1.1091,
+      "step": 10426
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012041604522456993,
+      "loss": 1.1418,
+      "step": 10427
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012040231545834531,
+      "loss": 1.1095,
+      "step": 10428
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012038858529080768,
+      "loss": 0.9537,
+      "step": 10429
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012037485472222722,
+      "loss": 1.0142,
+      "step": 10430
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012036112375287396,
+      "loss": 0.9937,
+      "step": 10431
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012034739238301798,
+      "loss": 0.9544,
+      "step": 10432
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001203336606129294,
+      "loss": 1.0174,
+      "step": 10433
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012031992844287836,
+      "loss": 0.8058,
+      "step": 10434
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001203061958731349,
+      "loss": 0.8984,
+      "step": 10435
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012029246290396914,
+      "loss": 0.9578,
+      "step": 10436
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012027872953565125,
+      "loss": 1.0004,
+      "step": 10437
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012026499576845138,
+      "loss": 1.2251,
+      "step": 10438
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012025126160263957,
+      "loss": 1.1218,
+      "step": 10439
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012023752703848612,
+      "loss": 0.9408,
+      "step": 10440
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012022379207626107,
+      "loss": 1.0117,
+      "step": 10441
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012021005671623464,
+      "loss": 0.9073,
+      "step": 10442
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012019632095867697,
+      "loss": 1.1099,
+      "step": 10443
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012018258480385829,
+      "loss": 0.9677,
+      "step": 10444
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012016884825204875,
+      "loss": 1.1731,
+      "step": 10445
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012015511130351857,
+      "loss": 0.9434,
+      "step": 10446
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012014137395853792,
+      "loss": 0.9061,
+      "step": 10447
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012012763621737706,
+      "loss": 0.9702,
+      "step": 10448
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012011389808030617,
+      "loss": 1.1027,
+      "step": 10449
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012010015954759553,
+      "loss": 0.8846,
+      "step": 10450
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012008642061951533,
+      "loss": 0.8823,
+      "step": 10451
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012007268129633582,
+      "loss": 1.0316,
+      "step": 10452
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012005894157832729,
+      "loss": 1.0181,
+      "step": 10453
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012004520146575995,
+      "loss": 0.9322,
+      "step": 10454
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012003146095890412,
+      "loss": 1.115,
+      "step": 10455
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012001772005802998,
+      "loss": 1.2343,
+      "step": 10456
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00012000397876340792,
+      "loss": 0.9493,
+      "step": 10457
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011999023707530819,
+      "loss": 1.0127,
+      "step": 10458
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011997649499400107,
+      "loss": 0.9627,
+      "step": 10459
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011996275251975688,
+      "loss": 0.9523,
+      "step": 10460
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011994900965284599,
+      "loss": 0.8569,
+      "step": 10461
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001199352663935386,
+      "loss": 1.0425,
+      "step": 10462
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011992152274210516,
+      "loss": 1.0694,
+      "step": 10463
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011990777869881591,
+      "loss": 1.0116,
+      "step": 10464
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011989403426394125,
+      "loss": 0.9053,
+      "step": 10465
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011988028943775155,
+      "loss": 1.1761,
+      "step": 10466
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011986654422051716,
+      "loss": 0.8866,
+      "step": 10467
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011985279861250838,
+      "loss": 1.0968,
+      "step": 10468
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011983905261399566,
+      "loss": 0.8524,
+      "step": 10469
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011982530622524937,
+      "loss": 0.8871,
+      "step": 10470
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011981155944653988,
+      "loss": 0.7919,
+      "step": 10471
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001197978122781376,
+      "loss": 1.1766,
+      "step": 10472
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011978406472031293,
+      "loss": 1.2377,
+      "step": 10473
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011977031677333632,
+      "loss": 1.2217,
+      "step": 10474
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011975656843747811,
+      "loss": 1.05,
+      "step": 10475
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011974281971300885,
+      "loss": 1.0929,
+      "step": 10476
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011972907060019887,
+      "loss": 0.7962,
+      "step": 10477
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011971532109931867,
+      "loss": 1.0262,
+      "step": 10478
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011970157121063866,
+      "loss": 1.0938,
+      "step": 10479
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011968782093442936,
+      "loss": 1.0034,
+      "step": 10480
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011967407027096118,
+      "loss": 1.0435,
+      "step": 10481
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011966031922050462,
+      "loss": 0.9542,
+      "step": 10482
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011964656778333017,
+      "loss": 1.126,
+      "step": 10483
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011963281595970832,
+      "loss": 0.9809,
+      "step": 10484
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011961906374990952,
+      "loss": 1.0447,
+      "step": 10485
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011960531115420435,
+      "loss": 0.9508,
+      "step": 10486
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011959155817286328,
+      "loss": 0.9835,
+      "step": 10487
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011957780480615685,
+      "loss": 1.0336,
+      "step": 10488
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011956405105435555,
+      "loss": 1.0051,
+      "step": 10489
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011955029691772999,
+      "loss": 1.1218,
+      "step": 10490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011953654239655064,
+      "loss": 0.9023,
+      "step": 10491
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011952278749108806,
+      "loss": 0.8924,
+      "step": 10492
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011950903220161285,
+      "loss": 1.0909,
+      "step": 10493
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011949527652839551,
+      "loss": 1.0214,
+      "step": 10494
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011948152047170671,
+      "loss": 0.9799,
+      "step": 10495
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011946776403181694,
+      "loss": 1.0276,
+      "step": 10496
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011945400720899687,
+      "loss": 1.1553,
+      "step": 10497
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000119440250003517,
+      "loss": 1.261,
+      "step": 10498
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011942649241564803,
+      "loss": 1.0755,
+      "step": 10499
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011941273444566053,
+      "loss": 0.9664,
+      "step": 10500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011939897609382511,
+      "loss": 0.9609,
+      "step": 10501
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011938521736041237,
+      "loss": 1.0509,
+      "step": 10502
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000119371458245693,
+      "loss": 1.1088,
+      "step": 10503
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011935769874993763,
+      "loss": 0.9341,
+      "step": 10504
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011934393887341686,
+      "loss": 0.7387,
+      "step": 10505
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011933017861640141,
+      "loss": 0.8858,
+      "step": 10506
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00011931641797916194,
+      "loss": 1.0021,
+      "step": 10507
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011930265696196906,
+      "loss": 1.0502,
+      "step": 10508
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001192888955650935,
+      "loss": 0.946,
+      "step": 10509
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011927513378880592,
+      "loss": 0.9212,
+      "step": 10510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011926137163337708,
+      "loss": 1.0617,
+      "step": 10511
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011924760909907755,
+      "loss": 0.9277,
+      "step": 10512
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011923384618617819,
+      "loss": 1.0363,
+      "step": 10513
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011922008289494961,
+      "loss": 0.9746,
+      "step": 10514
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011920631922566257,
+      "loss": 1.0719,
+      "step": 10515
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011919255517858779,
+      "loss": 1.148,
+      "step": 10516
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000119178790753996,
+      "loss": 1.0871,
+      "step": 10517
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000119165025952158,
+      "loss": 0.7088,
+      "step": 10518
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011915126077334449,
+      "loss": 0.8997,
+      "step": 10519
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011913749521782624,
+      "loss": 1.1469,
+      "step": 10520
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011912372928587406,
+      "loss": 1.0334,
+      "step": 10521
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011910996297775865,
+      "loss": 1.106,
+      "step": 10522
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011909619629375081,
+      "loss": 1.2094,
+      "step": 10523
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011908242923412143,
+      "loss": 1.1017,
+      "step": 10524
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011906866179914116,
+      "loss": 1.1875,
+      "step": 10525
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011905489398908091,
+      "loss": 1.1168,
+      "step": 10526
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011904112580421145,
+      "loss": 1.1962,
+      "step": 10527
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001190273572448036,
+      "loss": 1.1169,
+      "step": 10528
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011901358831112819,
+      "loss": 0.848,
+      "step": 10529
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001189998190034561,
+      "loss": 0.7207,
+      "step": 10530
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011898604932205808,
+      "loss": 1.1319,
+      "step": 10531
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011897227926720506,
+      "loss": 1.0827,
+      "step": 10532
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011895850883916786,
+      "loss": 0.9729,
+      "step": 10533
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011894473803821738,
+      "loss": 1.037,
+      "step": 10534
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011893096686462439,
+      "loss": 1.0039,
+      "step": 10535
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001189171953186599,
+      "loss": 1.023,
+      "step": 10536
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011890342340059473,
+      "loss": 0.9377,
+      "step": 10537
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011888965111069976,
+      "loss": 1.0824,
+      "step": 10538
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001188758784492459,
+      "loss": 0.8515,
+      "step": 10539
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011886210541650409,
+      "loss": 0.8949,
+      "step": 10540
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011884833201274523,
+      "loss": 0.9328,
+      "step": 10541
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011883455823824021,
+      "loss": 0.9678,
+      "step": 10542
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011882078409326002,
+      "loss": 1.1812,
+      "step": 10543
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011880700957807556,
+      "loss": 1.0291,
+      "step": 10544
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011879323469295776,
+      "loss": 1.0945,
+      "step": 10545
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011877945943817759,
+      "loss": 1.0353,
+      "step": 10546
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011876568381400601,
+      "loss": 0.9904,
+      "step": 10547
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011875190782071398,
+      "loss": 1.0447,
+      "step": 10548
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011873813145857249,
+      "loss": 1.0038,
+      "step": 10549
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011872435472785248,
+      "loss": 1.1867,
+      "step": 10550
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011871057762882499,
+      "loss": 0.9713,
+      "step": 10551
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011869680016176096,
+      "loss": 1.1278,
+      "step": 10552
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011868302232693147,
+      "loss": 1.0091,
+      "step": 10553
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011866924412460745,
+      "loss": 1.1648,
+      "step": 10554
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011865546555505995,
+      "loss": 1.0021,
+      "step": 10555
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011864168661856001,
+      "loss": 0.9733,
+      "step": 10556
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011862790731537865,
+      "loss": 1.0845,
+      "step": 10557
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011861412764578686,
+      "loss": 0.9555,
+      "step": 10558
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011860034761005577,
+      "loss": 1.0204,
+      "step": 10559
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011858656720845638,
+      "loss": 0.9322,
+      "step": 10560
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011857278644125976,
+      "loss": 1.026,
+      "step": 10561
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011855900530873696,
+      "loss": 1.0738,
+      "step": 10562
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011854522381115908,
+      "loss": 1.0666,
+      "step": 10563
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011853144194879721,
+      "loss": 1.1155,
+      "step": 10564
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011851765972192244,
+      "loss": 1.0323,
+      "step": 10565
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011850387713080584,
+      "loss": 1.1529,
+      "step": 10566
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001184900941757185,
+      "loss": 1.0514,
+      "step": 10567
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011847631085693158,
+      "loss": 0.9456,
+      "step": 10568
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011846252717471616,
+      "loss": 1.1149,
+      "step": 10569
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011844874312934338,
+      "loss": 1.0417,
+      "step": 10570
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011843495872108434,
+      "loss": 0.9443,
+      "step": 10571
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011842117395021026,
+      "loss": 1.1697,
+      "step": 10572
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011840738881699219,
+      "loss": 1.2027,
+      "step": 10573
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011839360332170133,
+      "loss": 1.1125,
+      "step": 10574
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011837981746460885,
+      "loss": 0.8501,
+      "step": 10575
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011836603124598591,
+      "loss": 1.0252,
+      "step": 10576
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011835224466610366,
+      "loss": 1.0691,
+      "step": 10577
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011833845772523329,
+      "loss": 0.908,
+      "step": 10578
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011832467042364602,
+      "loss": 0.9452,
+      "step": 10579
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011831088276161299,
+      "loss": 0.8394,
+      "step": 10580
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011829709473940547,
+      "loss": 0.831,
+      "step": 10581
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011828330635729465,
+      "loss": 0.9403,
+      "step": 10582
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001182695176155517,
+      "loss": 0.9813,
+      "step": 10583
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011825572851444789,
+      "loss": 0.9459,
+      "step": 10584
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011824193905425442,
+      "loss": 0.9782,
+      "step": 10585
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011822814923524257,
+      "loss": 1.0316,
+      "step": 10586
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011821435905768359,
+      "loss": 1.1551,
+      "step": 10587
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011820056852184864,
+      "loss": 1.1365,
+      "step": 10588
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011818677762800909,
+      "loss": 1.1413,
+      "step": 10589
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011817298637643614,
+      "loss": 1.1262,
+      "step": 10590
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011815919476740109,
+      "loss": 0.9118,
+      "step": 10591
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011814540280117523,
+      "loss": 1.2016,
+      "step": 10592
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011813161047802985,
+      "loss": 1.1988,
+      "step": 10593
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011811781779823616,
+      "loss": 1.1273,
+      "step": 10594
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011810402476206559,
+      "loss": 0.9425,
+      "step": 10595
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011809023136978936,
+      "loss": 1.0618,
+      "step": 10596
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011807643762167884,
+      "loss": 0.9429,
+      "step": 10597
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011806264351800526,
+      "loss": 1.0977,
+      "step": 10598
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001180488490590401,
+      "loss": 1.0985,
+      "step": 10599
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011803505424505457,
+      "loss": 0.9457,
+      "step": 10600
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011802125907632005,
+      "loss": 0.8878,
+      "step": 10601
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001180074635531079,
+      "loss": 1.0309,
+      "step": 10602
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011799366767568947,
+      "loss": 1.2224,
+      "step": 10603
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011797987144433613,
+      "loss": 1.1189,
+      "step": 10604
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011796607485931928,
+      "loss": 0.9663,
+      "step": 10605
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011795227792091024,
+      "loss": 0.8821,
+      "step": 10606
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011793848062938042,
+      "loss": 0.833,
+      "step": 10607
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011792468298500121,
+      "loss": 0.9739,
+      "step": 10608
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011791088498804403,
+      "loss": 0.9597,
+      "step": 10609
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011789708663878027,
+      "loss": 0.8515,
+      "step": 10610
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011788328793748134,
+      "loss": 0.9222,
+      "step": 10611
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011786948888441867,
+      "loss": 0.8158,
+      "step": 10612
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011785568947986367,
+      "loss": 1.0926,
+      "step": 10613
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011784188972408778,
+      "loss": 0.9883,
+      "step": 10614
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011782808961736246,
+      "loss": 1.0638,
+      "step": 10615
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011781428915995918,
+      "loss": 0.961,
+      "step": 10616
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011780048835214929,
+      "loss": 0.9749,
+      "step": 10617
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011778668719420436,
+      "loss": 0.9114,
+      "step": 10618
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011777288568639582,
+      "loss": 1.1254,
+      "step": 10619
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011775908382899515,
+      "loss": 1.129,
+      "step": 10620
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011774528162227376,
+      "loss": 1.0295,
+      "step": 10621
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011773147906650328,
+      "loss": 1.0195,
+      "step": 10622
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001177176761619551,
+      "loss": 1.1825,
+      "step": 10623
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011770387290890074,
+      "loss": 1.1614,
+      "step": 10624
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011769006930761173,
+      "loss": 1.0634,
+      "step": 10625
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011767626535835958,
+      "loss": 1.0111,
+      "step": 10626
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011766246106141583,
+      "loss": 1.0731,
+      "step": 10627
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011764865641705199,
+      "loss": 0.8587,
+      "step": 10628
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011763485142553959,
+      "loss": 1.1404,
+      "step": 10629
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011762104608715019,
+      "loss": 1.0332,
+      "step": 10630
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011760724040215532,
+      "loss": 1.0747,
+      "step": 10631
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011759343437082655,
+      "loss": 1.0172,
+      "step": 10632
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011757962799343547,
+      "loss": 1.0774,
+      "step": 10633
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011756582127025359,
+      "loss": 0.8431,
+      "step": 10634
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011755201420155255,
+      "loss": 1.0023,
+      "step": 10635
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001175382067876039,
+      "loss": 1.0717,
+      "step": 10636
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011752439902867925,
+      "loss": 1.0689,
+      "step": 10637
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011751059092505019,
+      "loss": 1.0979,
+      "step": 10638
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011749678247698832,
+      "loss": 1.2626,
+      "step": 10639
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011748297368476522,
+      "loss": 0.8307,
+      "step": 10640
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001174691645486526,
+      "loss": 0.8747,
+      "step": 10641
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.000117455355068922,
+      "loss": 1.083,
+      "step": 10642
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001174415452458451,
+      "loss": 0.929,
+      "step": 10643
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011742773507969348,
+      "loss": 1.1042,
+      "step": 10644
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011741392457073888,
+      "loss": 1.2657,
+      "step": 10645
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011740011371925285,
+      "loss": 1.0442,
+      "step": 10646
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011738630252550713,
+      "loss": 0.8988,
+      "step": 10647
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011737249098977332,
+      "loss": 0.8916,
+      "step": 10648
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011735867911232315,
+      "loss": 1.1579,
+      "step": 10649
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011734486689342825,
+      "loss": 0.7809,
+      "step": 10650
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011733105433336038,
+      "loss": 0.9442,
+      "step": 10651
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011731724143239113,
+      "loss": 1.186,
+      "step": 10652
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011730342819079227,
+      "loss": 0.9536,
+      "step": 10653
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001172896146088355,
+      "loss": 0.966,
+      "step": 10654
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011727580068679251,
+      "loss": 1.0761,
+      "step": 10655
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011726198642493505,
+      "loss": 0.9204,
+      "step": 10656
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011724817182353478,
+      "loss": 0.9558,
+      "step": 10657
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011723435688286352,
+      "loss": 1.0506,
+      "step": 10658
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011722054160319298,
+      "loss": 1.134,
+      "step": 10659
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011720672598479485,
+      "loss": 1.0209,
+      "step": 10660
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011719291002794096,
+      "loss": 0.9819,
+      "step": 10661
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011717909373290306,
+      "loss": 1.1168,
+      "step": 10662
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011716527709995284,
+      "loss": 0.9494,
+      "step": 10663
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011715146012936216,
+      "loss": 1.1067,
+      "step": 10664
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011713764282140274,
+      "loss": 0.874,
+      "step": 10665
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011712382517634641,
+      "loss": 1.124,
+      "step": 10666
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011711000719446492,
+      "loss": 0.7298,
+      "step": 10667
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011709618887603014,
+      "loss": 1.1476,
+      "step": 10668
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011708237022131381,
+      "loss": 0.8814,
+      "step": 10669
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011706855123058776,
+      "loss": 1.2985,
+      "step": 10670
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001170547319041238,
+      "loss": 1.1423,
+      "step": 10671
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011704091224219376,
+      "loss": 1.0113,
+      "step": 10672
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001170270922450695,
+      "loss": 0.8118,
+      "step": 10673
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011701327191302285,
+      "loss": 0.7508,
+      "step": 10674
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011699945124632561,
+      "loss": 0.9749,
+      "step": 10675
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011698563024524968,
+      "loss": 0.9305,
+      "step": 10676
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011697180891006689,
+      "loss": 0.9404,
+      "step": 10677
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011695798724104914,
+      "loss": 1.014,
+      "step": 10678
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011694416523846829,
+      "loss": 0.9997,
+      "step": 10679
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011693034290259619,
+      "loss": 1.0881,
+      "step": 10680
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011691652023370476,
+      "loss": 0.7313,
+      "step": 10681
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011690269723206587,
+      "loss": 1.19,
+      "step": 10682
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011688887389795142,
+      "loss": 0.9874,
+      "step": 10683
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011687505023163331,
+      "loss": 0.996,
+      "step": 10684
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011686122623338349,
+      "loss": 0.8625,
+      "step": 10685
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011684740190347383,
+      "loss": 1.1403,
+      "step": 10686
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011683357724217627,
+      "loss": 1.0588,
+      "step": 10687
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011681975224976276,
+      "loss": 1.16,
+      "step": 10688
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001168059269265052,
+      "loss": 1.0893,
+      "step": 10689
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011679210127267555,
+      "loss": 0.9836,
+      "step": 10690
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001167782752885458,
+      "loss": 1.149,
+      "step": 10691
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011676444897438783,
+      "loss": 1.2018,
+      "step": 10692
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011675062233047364,
+      "loss": 1.2029,
+      "step": 10693
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011673679535707522,
+      "loss": 0.9754,
+      "step": 10694
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011672296805446454,
+      "loss": 0.8774,
+      "step": 10695
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011670914042291357,
+      "loss": 1.1963,
+      "step": 10696
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011669531246269429,
+      "loss": 0.9249,
+      "step": 10697
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011668148417407871,
+      "loss": 1.1379,
+      "step": 10698
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011666765555733883,
+      "loss": 0.9444,
+      "step": 10699
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011665382661274664,
+      "loss": 1.0144,
+      "step": 10700
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001166399973405742,
+      "loss": 0.978,
+      "step": 10701
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001166261677410935,
+      "loss": 1.1061,
+      "step": 10702
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011661233781457654,
+      "loss": 1.0126,
+      "step": 10703
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011659850756129543,
+      "loss": 1.0906,
+      "step": 10704
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011658467698152213,
+      "loss": 0.9205,
+      "step": 10705
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011657084607552875,
+      "loss": 0.9617,
+      "step": 10706
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001165570148435873,
+      "loss": 0.8942,
+      "step": 10707
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001165431832859699,
+      "loss": 1.1495,
+      "step": 10708
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011652935140294852,
+      "loss": 1.1207,
+      "step": 10709
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011651551919479529,
+      "loss": 0.973,
+      "step": 10710
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001165016866617823,
+      "loss": 0.8945,
+      "step": 10711
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011648785380418164,
+      "loss": 0.9738,
+      "step": 10712
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011647402062226534,
+      "loss": 1.1814,
+      "step": 10713
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011646018711630557,
+      "loss": 1.0003,
+      "step": 10714
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011644635328657438,
+      "loss": 0.7805,
+      "step": 10715
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011643251913334392,
+      "loss": 1.0674,
+      "step": 10716
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001164186846568863,
+      "loss": 1.0355,
+      "step": 10717
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011640484985747364,
+      "loss": 1.0095,
+      "step": 10718
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011639101473537808,
+      "loss": 1.1756,
+      "step": 10719
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011637717929087171,
+      "loss": 1.0924,
+      "step": 10720
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011636334352422673,
+      "loss": 1.0858,
+      "step": 10721
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011634950743571524,
+      "loss": 1.0996,
+      "step": 10722
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011633567102560944,
+      "loss": 1.1194,
+      "step": 10723
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011632183429418147,
+      "loss": 0.9481,
+      "step": 10724
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011630799724170353,
+      "loss": 0.8963,
+      "step": 10725
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011629415986844771,
+      "loss": 1.1664,
+      "step": 10726
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001162803221746863,
+      "loss": 1.1121,
+      "step": 10727
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011626648416069141,
+      "loss": 1.0096,
+      "step": 10728
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011625264582673527,
+      "loss": 0.9925,
+      "step": 10729
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011623880717309004,
+      "loss": 0.9939,
+      "step": 10730
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011622496820002797,
+      "loss": 1.111,
+      "step": 10731
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011621112890782127,
+      "loss": 1.1508,
+      "step": 10732
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011619728929674209,
+      "loss": 1.1662,
+      "step": 10733
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011618344936706276,
+      "loss": 1.0575,
+      "step": 10734
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011616960911905545,
+      "loss": 1.2203,
+      "step": 10735
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011615576855299237,
+      "loss": 0.8414,
+      "step": 10736
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011614192766914584,
+      "loss": 1.0216,
+      "step": 10737
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00011612808646778806,
+      "loss": 1.1361,
+      "step": 10738
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001161142449491913,
+      "loss": 0.9389,
+      "step": 10739
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001161004031136278,
+      "loss": 1.1525,
+      "step": 10740
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011608656096136984,
+      "loss": 1.1421,
+      "step": 10741
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011607271849268975,
+      "loss": 1.1476,
+      "step": 10742
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011605887570785972,
+      "loss": 0.9887,
+      "step": 10743
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011604503260715208,
+      "loss": 1.0093,
+      "step": 10744
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011603118919083913,
+      "loss": 0.7197,
+      "step": 10745
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011601734545919316,
+      "loss": 1.0412,
+      "step": 10746
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011600350141248646,
+      "loss": 0.7374,
+      "step": 10747
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001159896570509914,
+      "loss": 1.0596,
+      "step": 10748
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001159758123749802,
+      "loss": 1.153,
+      "step": 10749
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011596196738472531,
+      "loss": 1.1622,
+      "step": 10750
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011594812208049897,
+      "loss": 1.1557,
+      "step": 10751
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011593427646257353,
+      "loss": 0.949,
+      "step": 10752
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011592043053122133,
+      "loss": 1.012,
+      "step": 10753
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011590658428671477,
+      "loss": 0.8094,
+      "step": 10754
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011589273772932612,
+      "loss": 1.1001,
+      "step": 10755
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011587889085932781,
+      "loss": 1.1638,
+      "step": 10756
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011586504367699217,
+      "loss": 1.1469,
+      "step": 10757
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011585119618259164,
+      "loss": 1.0075,
+      "step": 10758
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011583734837639848,
+      "loss": 0.8817,
+      "step": 10759
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001158235002586852,
+      "loss": 1.0621,
+      "step": 10760
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011580965182972412,
+      "loss": 1.037,
+      "step": 10761
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011579580308978764,
+      "loss": 1.1166,
+      "step": 10762
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001157819540391482,
+      "loss": 1.0101,
+      "step": 10763
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011576810467807817,
+      "loss": 0.9231,
+      "step": 10764
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011575425500685002,
+      "loss": 0.8752,
+      "step": 10765
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011574040502573611,
+      "loss": 0.9957,
+      "step": 10766
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001157265547350089,
+      "loss": 0.9563,
+      "step": 10767
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011571270413494082,
+      "loss": 0.9976,
+      "step": 10768
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011569885322580432,
+      "loss": 1.1378,
+      "step": 10769
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011568500200787184,
+      "loss": 0.9097,
+      "step": 10770
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011567115048141586,
+      "loss": 1.0201,
+      "step": 10771
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011565729864670878,
+      "loss": 1.1394,
+      "step": 10772
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001156434465040231,
+      "loss": 1.321,
+      "step": 10773
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011562959405363129,
+      "loss": 1.1296,
+      "step": 10774
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011561574129580585,
+      "loss": 1.0906,
+      "step": 10775
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011560188823081918,
+      "loss": 0.9573,
+      "step": 10776
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011558803485894389,
+      "loss": 0.9843,
+      "step": 10777
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001155741811804524,
+      "loss": 0.8713,
+      "step": 10778
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011556032719561722,
+      "loss": 0.9489,
+      "step": 10779
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011554647290471086,
+      "loss": 0.9154,
+      "step": 10780
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011553261830800586,
+      "loss": 1.1268,
+      "step": 10781
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011551876340577467,
+      "loss": 1.1033,
+      "step": 10782
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011550490819828991,
+      "loss": 0.8864,
+      "step": 10783
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011549105268582406,
+      "loss": 0.932,
+      "step": 10784
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011547719686864962,
+      "loss": 0.9868,
+      "step": 10785
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011546334074703919,
+      "loss": 1.1094,
+      "step": 10786
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011544948432126533,
+      "loss": 1.1263,
+      "step": 10787
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011543562759160056,
+      "loss": 1.1927,
+      "step": 10788
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011542177055831745,
+      "loss": 1.0903,
+      "step": 10789
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011540791322168855,
+      "loss": 0.9253,
+      "step": 10790
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011539405558198646,
+      "loss": 1.0479,
+      "step": 10791
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011538019763948375,
+      "loss": 1.0839,
+      "step": 10792
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000115366339394453,
+      "loss": 0.9219,
+      "step": 10793
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011535248084716686,
+      "loss": 0.9746,
+      "step": 10794
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011533862199789782,
+      "loss": 1.1141,
+      "step": 10795
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011532476284691855,
+      "loss": 1.0342,
+      "step": 10796
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011531090339450165,
+      "loss": 1.1045,
+      "step": 10797
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011529704364091975,
+      "loss": 1.0563,
+      "step": 10798
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001152831835864454,
+      "loss": 0.915,
+      "step": 10799
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011526932323135134,
+      "loss": 1.2671,
+      "step": 10800
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001152554625759101,
+      "loss": 0.9509,
+      "step": 10801
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011524160162039439,
+      "loss": 1.0111,
+      "step": 10802
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011522774036507682,
+      "loss": 0.9443,
+      "step": 10803
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011521387881023007,
+      "loss": 0.9457,
+      "step": 10804
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011520001695612674,
+      "loss": 1.1918,
+      "step": 10805
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011518615480303952,
+      "loss": 1.054,
+      "step": 10806
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011517229235124109,
+      "loss": 0.9662,
+      "step": 10807
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011515842960100411,
+      "loss": 0.9238,
+      "step": 10808
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011514456655260128,
+      "loss": 0.9889,
+      "step": 10809
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011513070320630528,
+      "loss": 1.1824,
+      "step": 10810
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011511683956238881,
+      "loss": 1.1983,
+      "step": 10811
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011510297562112451,
+      "loss": 1.0052,
+      "step": 10812
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011508911138278515,
+      "loss": 0.9174,
+      "step": 10813
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011507524684764342,
+      "loss": 0.97,
+      "step": 10814
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011506138201597201,
+      "loss": 0.9474,
+      "step": 10815
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011504751688804365,
+      "loss": 0.8587,
+      "step": 10816
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011503365146413112,
+      "loss": 0.8969,
+      "step": 10817
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011501978574450709,
+      "loss": 0.98,
+      "step": 10818
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001150059197294443,
+      "loss": 0.967,
+      "step": 10819
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011499205341921553,
+      "loss": 1.0328,
+      "step": 10820
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011497818681409351,
+      "loss": 0.936,
+      "step": 10821
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000114964319914351,
+      "loss": 0.9761,
+      "step": 10822
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011495045272026078,
+      "loss": 0.9104,
+      "step": 10823
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011493658523209554,
+      "loss": 1.1462,
+      "step": 10824
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011492271745012815,
+      "loss": 1.0355,
+      "step": 10825
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011490884937463133,
+      "loss": 0.8778,
+      "step": 10826
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011489498100587792,
+      "loss": 1.2202,
+      "step": 10827
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011488111234414062,
+      "loss": 1.1881,
+      "step": 10828
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011486724338969232,
+      "loss": 0.9112,
+      "step": 10829
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011485337414280577,
+      "loss": 0.924,
+      "step": 10830
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011483950460375376,
+      "loss": 0.7952,
+      "step": 10831
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011482563477280917,
+      "loss": 0.9571,
+      "step": 10832
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011481176465024477,
+      "loss": 1.1227,
+      "step": 10833
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001147978942363334,
+      "loss": 0.833,
+      "step": 10834
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001147840235313479,
+      "loss": 0.9792,
+      "step": 10835
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011477015253556107,
+      "loss": 1.0205,
+      "step": 10836
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011475628124924577,
+      "loss": 1.0264,
+      "step": 10837
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011474240967267489,
+      "loss": 1.0141,
+      "step": 10838
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001147285378061212,
+      "loss": 1.043,
+      "step": 10839
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011471466564985765,
+      "loss": 0.9824,
+      "step": 10840
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011470079320415703,
+      "loss": 1.0058,
+      "step": 10841
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011468692046929226,
+      "loss": 0.9049,
+      "step": 10842
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011467304744553618,
+      "loss": 1.0923,
+      "step": 10843
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011465917413316172,
+      "loss": 1.1022,
+      "step": 10844
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011464530053244169,
+      "loss": 0.9136,
+      "step": 10845
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011463142664364907,
+      "loss": 0.9327,
+      "step": 10846
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011461755246705673,
+      "loss": 1.0186,
+      "step": 10847
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011460367800293753,
+      "loss": 1.0599,
+      "step": 10848
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011458980325156444,
+      "loss": 1.0052,
+      "step": 10849
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011457592821321035,
+      "loss": 1.1326,
+      "step": 10850
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011456205288814816,
+      "loss": 0.9641,
+      "step": 10851
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011454817727665084,
+      "loss": 0.8893,
+      "step": 10852
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011453430137899129,
+      "loss": 0.9222,
+      "step": 10853
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011452042519544246,
+      "loss": 1.1436,
+      "step": 10854
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011450654872627728,
+      "loss": 1.0197,
+      "step": 10855
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011449267197176874,
+      "loss": 1.0222,
+      "step": 10856
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011447879493218979,
+      "loss": 0.8879,
+      "step": 10857
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011446491760781335,
+      "loss": 1.051,
+      "step": 10858
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001144510399989124,
+      "loss": 0.8394,
+      "step": 10859
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011443716210575992,
+      "loss": 1.0154,
+      "step": 10860
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011442328392862891,
+      "loss": 0.8953,
+      "step": 10861
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011440940546779228,
+      "loss": 0.8251,
+      "step": 10862
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011439552672352312,
+      "loss": 0.9723,
+      "step": 10863
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011438164769609436,
+      "loss": 0.7716,
+      "step": 10864
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000114367768385779,
+      "loss": 0.9905,
+      "step": 10865
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011435388879285008,
+      "loss": 0.9216,
+      "step": 10866
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011434000891758062,
+      "loss": 1.1727,
+      "step": 10867
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011432612876024352,
+      "loss": 1.2216,
+      "step": 10868
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011431224832111196,
+      "loss": 1.0421,
+      "step": 10869
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011429836760045888,
+      "loss": 0.9148,
+      "step": 10870
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011428448659855735,
+      "loss": 1.117,
+      "step": 10871
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011427060531568034,
+      "loss": 0.8503,
+      "step": 10872
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011425672375210099,
+      "loss": 0.9242,
+      "step": 10873
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001142428419080923,
+      "loss": 1.114,
+      "step": 10874
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011422895978392732,
+      "loss": 0.936,
+      "step": 10875
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001142150773798791,
+      "loss": 0.9826,
+      "step": 10876
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011420119469622077,
+      "loss": 1.0753,
+      "step": 10877
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011418731173322532,
+      "loss": 0.9444,
+      "step": 10878
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001141734284911659,
+      "loss": 0.768,
+      "step": 10879
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011415954497031555,
+      "loss": 1.0063,
+      "step": 10880
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011414566117094736,
+      "loss": 1.0413,
+      "step": 10881
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011413177709333445,
+      "loss": 0.9965,
+      "step": 10882
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011411789273774989,
+      "loss": 0.9579,
+      "step": 10883
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001141040081044668,
+      "loss": 1.1434,
+      "step": 10884
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011409012319375827,
+      "loss": 0.8676,
+      "step": 10885
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011407623800589746,
+      "loss": 0.898,
+      "step": 10886
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011406235254115745,
+      "loss": 0.7219,
+      "step": 10887
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011404846679981139,
+      "loss": 1.2073,
+      "step": 10888
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011403458078213239,
+      "loss": 0.9474,
+      "step": 10889
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011402069448839361,
+      "loss": 1.1508,
+      "step": 10890
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011400680791886815,
+      "loss": 1.1062,
+      "step": 10891
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011399292107382924,
+      "loss": 1.161,
+      "step": 10892
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011397903395354996,
+      "loss": 0.9828,
+      "step": 10893
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011396514655830352,
+      "loss": 0.9831,
+      "step": 10894
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011395125888836303,
+      "loss": 1.0876,
+      "step": 10895
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001139373709440017,
+      "loss": 1.1222,
+      "step": 10896
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011392348272549269,
+      "loss": 1.317,
+      "step": 10897
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011390959423310918,
+      "loss": 1.0522,
+      "step": 10898
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011389570546712438,
+      "loss": 0.9096,
+      "step": 10899
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011388181642781144,
+      "loss": 0.893,
+      "step": 10900
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001138679271154436,
+      "loss": 1.1715,
+      "step": 10901
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011385403753029404,
+      "loss": 0.8841,
+      "step": 10902
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011384014767263598,
+      "loss": 0.8527,
+      "step": 10903
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011382625754274262,
+      "loss": 1.137,
+      "step": 10904
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011381236714088716,
+      "loss": 1.0734,
+      "step": 10905
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011379847646734287,
+      "loss": 1.0035,
+      "step": 10906
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011378458552238296,
+      "loss": 1.1578,
+      "step": 10907
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011377069430628062,
+      "loss": 1.1292,
+      "step": 10908
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011375680281930919,
+      "loss": 1.0016,
+      "step": 10909
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011374291106174182,
+      "loss": 0.9916,
+      "step": 10910
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001137290190338518,
+      "loss": 1.1571,
+      "step": 10911
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011371512673591236,
+      "loss": 1.0603,
+      "step": 10912
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011370123416819682,
+      "loss": 1.1123,
+      "step": 10913
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011368734133097839,
+      "loss": 1.0657,
+      "step": 10914
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011367344822453036,
+      "loss": 1.0214,
+      "step": 10915
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.000113659554849126,
+      "loss": 0.9115,
+      "step": 10916
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011364566120503864,
+      "loss": 0.9868,
+      "step": 10917
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011363176729254146,
+      "loss": 0.9504,
+      "step": 10918
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011361787311190785,
+      "loss": 0.9889,
+      "step": 10919
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001136039786634111,
+      "loss": 0.9749,
+      "step": 10920
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011359008394732446,
+      "loss": 1.1475,
+      "step": 10921
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011357618896392128,
+      "loss": 0.6805,
+      "step": 10922
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011356229371347487,
+      "loss": 1.1306,
+      "step": 10923
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011354839819625854,
+      "loss": 1.0431,
+      "step": 10924
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001135345024125456,
+      "loss": 1.0106,
+      "step": 10925
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011352060636260944,
+      "loss": 0.943,
+      "step": 10926
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001135067100467233,
+      "loss": 1.0439,
+      "step": 10927
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001134928134651606,
+      "loss": 0.9178,
+      "step": 10928
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011347891661819465,
+      "loss": 1.1109,
+      "step": 10929
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011346501950609884,
+      "loss": 1.0729,
+      "step": 10930
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011345112212914646,
+      "loss": 1.0016,
+      "step": 10931
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011343722448761092,
+      "loss": 0.9535,
+      "step": 10932
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011342332658176555,
+      "loss": 1.0154,
+      "step": 10933
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011340942841188375,
+      "loss": 1.066,
+      "step": 10934
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001133955299782389,
+      "loss": 0.9262,
+      "step": 10935
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001133816312811044,
+      "loss": 1.0238,
+      "step": 10936
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011336773232075355,
+      "loss": 1.0359,
+      "step": 10937
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011335383309745982,
+      "loss": 0.8508,
+      "step": 10938
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011333993361149659,
+      "loss": 1.0388,
+      "step": 10939
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011332603386313729,
+      "loss": 1.0949,
+      "step": 10940
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011331213385265524,
+      "loss": 1.1372,
+      "step": 10941
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011329823358032396,
+      "loss": 1.0071,
+      "step": 10942
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011328433304641679,
+      "loss": 0.9728,
+      "step": 10943
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011327043225120718,
+      "loss": 0.908,
+      "step": 10944
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011325653119496857,
+      "loss": 1.0244,
+      "step": 10945
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011324262987797439,
+      "loss": 0.9434,
+      "step": 10946
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011322872830049807,
+      "loss": 1.2457,
+      "step": 10947
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011321482646281301,
+      "loss": 0.8253,
+      "step": 10948
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011320092436519276,
+      "loss": 0.9272,
+      "step": 10949
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001131870220079107,
+      "loss": 0.911,
+      "step": 10950
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011317311939124028,
+      "loss": 1.0975,
+      "step": 10951
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011315921651545502,
+      "loss": 0.9741,
+      "step": 10952
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011314531338082838,
+      "loss": 1.0181,
+      "step": 10953
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011313140998763377,
+      "loss": 1.002,
+      "step": 10954
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011311750633614475,
+      "loss": 1.0411,
+      "step": 10955
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011310360242663473,
+      "loss": 0.8425,
+      "step": 10956
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011308969825937725,
+      "loss": 1.1268,
+      "step": 10957
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011307579383464579,
+      "loss": 0.9134,
+      "step": 10958
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001130618891527139,
+      "loss": 0.8839,
+      "step": 10959
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011304798421385499,
+      "loss": 0.9391,
+      "step": 10960
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011303407901834263,
+      "loss": 0.8987,
+      "step": 10961
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011302017356645029,
+      "loss": 1.2178,
+      "step": 10962
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011300626785845156,
+      "loss": 0.9594,
+      "step": 10963
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011299236189461988,
+      "loss": 0.9474,
+      "step": 10964
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011297845567522886,
+      "loss": 1.1196,
+      "step": 10965
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011296454920055199,
+      "loss": 1.1006,
+      "step": 10966
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011295064247086283,
+      "loss": 1.1252,
+      "step": 10967
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001129367354864349,
+      "loss": 0.9555,
+      "step": 10968
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00011292282824754177,
+      "loss": 0.8958,
+      "step": 10969
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011290892075445701,
+      "loss": 0.9361,
+      "step": 10970
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001128950130074541,
+      "loss": 0.9939,
+      "step": 10971
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011288110500680675,
+      "loss": 1.0582,
+      "step": 10972
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011286719675278838,
+      "loss": 1.1007,
+      "step": 10973
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011285328824567268,
+      "loss": 1.0006,
+      "step": 10974
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011283937948573313,
+      "loss": 0.9279,
+      "step": 10975
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011282547047324342,
+      "loss": 1.1012,
+      "step": 10976
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000112811561208477,
+      "loss": 1.0705,
+      "step": 10977
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011279765169170762,
+      "loss": 1.0418,
+      "step": 10978
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001127837419232088,
+      "loss": 0.8837,
+      "step": 10979
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011276983190325416,
+      "loss": 0.8883,
+      "step": 10980
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011275592163211726,
+      "loss": 0.8537,
+      "step": 10981
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011274201111007179,
+      "loss": 0.9674,
+      "step": 10982
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011272810033739135,
+      "loss": 1.1847,
+      "step": 10983
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011271418931434952,
+      "loss": 1.15,
+      "step": 10984
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011270027804121996,
+      "loss": 0.8356,
+      "step": 10985
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011268636651827635,
+      "loss": 0.9372,
+      "step": 10986
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011267245474579225,
+      "loss": 1.1775,
+      "step": 10987
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011265854272404134,
+      "loss": 1.1786,
+      "step": 10988
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011264463045329728,
+      "loss": 1.1527,
+      "step": 10989
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001126307179338337,
+      "loss": 0.9213,
+      "step": 10990
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011261680516592428,
+      "loss": 0.8279,
+      "step": 10991
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011260289214984268,
+      "loss": 1.0788,
+      "step": 10992
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011258897888586255,
+      "loss": 0.838,
+      "step": 10993
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011257506537425757,
+      "loss": 0.9894,
+      "step": 10994
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011256115161530146,
+      "loss": 0.777,
+      "step": 10995
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011254723760926784,
+      "loss": 0.7614,
+      "step": 10996
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011253332335643043,
+      "loss": 0.9963,
+      "step": 10997
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011251940885706291,
+      "loss": 0.9722,
+      "step": 10998
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011250549411143903,
+      "loss": 0.939,
+      "step": 10999
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011249157911983239,
+      "loss": 0.9187,
+      "step": 11000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011247766388251682,
+      "loss": 1.1174,
+      "step": 11001
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011246374839976592,
+      "loss": 0.7155,
+      "step": 11002
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001124498326718535,
+      "loss": 1.0825,
+      "step": 11003
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001124359166990532,
+      "loss": 1.285,
+      "step": 11004
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011242200048163882,
+      "loss": 0.9532,
+      "step": 11005
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011240808401988403,
+      "loss": 1.1969,
+      "step": 11006
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011239416731406261,
+      "loss": 1.1214,
+      "step": 11007
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011238025036444825,
+      "loss": 0.9677,
+      "step": 11008
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011236633317131477,
+      "loss": 0.804,
+      "step": 11009
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011235241573493587,
+      "loss": 1.1503,
+      "step": 11010
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011233849805558532,
+      "loss": 1.113,
+      "step": 11011
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011232458013353686,
+      "loss": 0.9732,
+      "step": 11012
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001123106619690643,
+      "loss": 1.1367,
+      "step": 11013
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011229674356244135,
+      "loss": 0.8841,
+      "step": 11014
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011228282491394184,
+      "loss": 0.9103,
+      "step": 11015
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011226890602383955,
+      "loss": 1.0752,
+      "step": 11016
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011225498689240818,
+      "loss": 1.0374,
+      "step": 11017
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011224106751992163,
+      "loss": 1.0434,
+      "step": 11018
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011222714790665363,
+      "loss": 0.9519,
+      "step": 11019
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011221322805287798,
+      "loss": 1.0485,
+      "step": 11020
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001121993079588685,
+      "loss": 0.8866,
+      "step": 11021
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011218538762489901,
+      "loss": 1.0608,
+      "step": 11022
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011217146705124327,
+      "loss": 0.7718,
+      "step": 11023
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011215754623817518,
+      "loss": 1.135,
+      "step": 11024
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011214362518596848,
+      "loss": 1.0182,
+      "step": 11025
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011212970389489705,
+      "loss": 1.0341,
+      "step": 11026
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011211578236523468,
+      "loss": 0.9695,
+      "step": 11027
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011210186059725525,
+      "loss": 1.1346,
+      "step": 11028
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011208793859123257,
+      "loss": 0.9177,
+      "step": 11029
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011207401634744048,
+      "loss": 1.1598,
+      "step": 11030
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011206009386615285,
+      "loss": 1.1444,
+      "step": 11031
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011204617114764357,
+      "loss": 1.0688,
+      "step": 11032
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011203224819218641,
+      "loss": 1.1689,
+      "step": 11033
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011201832500005531,
+      "loss": 1.0177,
+      "step": 11034
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001120044015715241,
+      "loss": 1.1981,
+      "step": 11035
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011199047790686667,
+      "loss": 1.0291,
+      "step": 11036
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011197655400635689,
+      "loss": 1.1004,
+      "step": 11037
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011196262987026864,
+      "loss": 1.0121,
+      "step": 11038
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011194870549887584,
+      "loss": 1.1927,
+      "step": 11039
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011193478089245231,
+      "loss": 1.0729,
+      "step": 11040
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011192085605127205,
+      "loss": 0.9063,
+      "step": 11041
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011190693097560887,
+      "loss": 1.0909,
+      "step": 11042
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001118930056657367,
+      "loss": 1.0694,
+      "step": 11043
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011187908012192947,
+      "loss": 1.2268,
+      "step": 11044
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001118651543444611,
+      "loss": 1.1384,
+      "step": 11045
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011185122833360548,
+      "loss": 1.2383,
+      "step": 11046
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011183730208963657,
+      "loss": 1.0895,
+      "step": 11047
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011182337561282822,
+      "loss": 1.1549,
+      "step": 11048
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001118094489034545,
+      "loss": 0.9571,
+      "step": 11049
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011179552196178919,
+      "loss": 1.1607,
+      "step": 11050
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011178159478810637,
+      "loss": 1.0015,
+      "step": 11051
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011176766738267992,
+      "loss": 0.9501,
+      "step": 11052
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011175373974578378,
+      "loss": 0.961,
+      "step": 11053
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011173981187769194,
+      "loss": 1.2992,
+      "step": 11054
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011172588377867834,
+      "loss": 1.0989,
+      "step": 11055
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011171195544901693,
+      "loss": 0.8667,
+      "step": 11056
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011169802688898175,
+      "loss": 1.0585,
+      "step": 11057
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001116840980988467,
+      "loss": 0.9165,
+      "step": 11058
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011167016907888577,
+      "loss": 1.0069,
+      "step": 11059
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011165623982937298,
+      "loss": 1.0306,
+      "step": 11060
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011164231035058228,
+      "loss": 0.9382,
+      "step": 11061
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011162838064278771,
+      "loss": 1.066,
+      "step": 11062
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001116144507062632,
+      "loss": 0.9247,
+      "step": 11063
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011160052054128282,
+      "loss": 0.9471,
+      "step": 11064
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011158659014812054,
+      "loss": 1.1191,
+      "step": 11065
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011157265952705036,
+      "loss": 1.1521,
+      "step": 11066
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011155872867834629,
+      "loss": 0.9062,
+      "step": 11067
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011154479760228243,
+      "loss": 1.0959,
+      "step": 11068
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011153086629913267,
+      "loss": 0.9958,
+      "step": 11069
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011151693476917116,
+      "loss": 1.0292,
+      "step": 11070
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011150300301267185,
+      "loss": 1.0588,
+      "step": 11071
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011148907102990885,
+      "loss": 1.245,
+      "step": 11072
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001114751388211561,
+      "loss": 0.8535,
+      "step": 11073
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011146120638668778,
+      "loss": 0.96,
+      "step": 11074
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011144727372677784,
+      "loss": 1.1367,
+      "step": 11075
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011143334084170034,
+      "loss": 1.0538,
+      "step": 11076
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011141940773172939,
+      "loss": 0.9469,
+      "step": 11077
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011140547439713903,
+      "loss": 1.1243,
+      "step": 11078
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011139154083820331,
+      "loss": 0.9039,
+      "step": 11079
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011137760705519631,
+      "loss": 1.1113,
+      "step": 11080
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011136367304839212,
+      "loss": 0.9019,
+      "step": 11081
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011134973881806483,
+      "loss": 1.1028,
+      "step": 11082
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001113358043644885,
+      "loss": 1.0757,
+      "step": 11083
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011132186968793723,
+      "loss": 0.9899,
+      "step": 11084
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011130793478868515,
+      "loss": 1.1164,
+      "step": 11085
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011129399966700629,
+      "loss": 1.1501,
+      "step": 11086
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011128006432317482,
+      "loss": 1.1088,
+      "step": 11087
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011126612875746479,
+      "loss": 1.091,
+      "step": 11088
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011125219297015035,
+      "loss": 1.1731,
+      "step": 11089
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001112382569615056,
+      "loss": 1.0665,
+      "step": 11090
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011122432073180469,
+      "loss": 0.861,
+      "step": 11091
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011121038428132172,
+      "loss": 1.3035,
+      "step": 11092
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011119644761033078,
+      "loss": 1.1666,
+      "step": 11093
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011118251071910608,
+      "loss": 1.3405,
+      "step": 11094
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011116857360792174,
+      "loss": 0.9038,
+      "step": 11095
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011115463627705185,
+      "loss": 0.9769,
+      "step": 11096
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011114069872677063,
+      "loss": 0.8489,
+      "step": 11097
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011112676095735218,
+      "loss": 0.8316,
+      "step": 11098
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011111282296907066,
+      "loss": 1.3069,
+      "step": 11099
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011109888476220023,
+      "loss": 0.8984,
+      "step": 11100
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011108494633701512,
+      "loss": 0.8796,
+      "step": 11101
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011107100769378939,
+      "loss": 0.907,
+      "step": 11102
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011105706883279727,
+      "loss": 1.175,
+      "step": 11103
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011104312975431295,
+      "loss": 1.0128,
+      "step": 11104
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011102919045861058,
+      "loss": 1.0709,
+      "step": 11105
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011101525094596436,
+      "loss": 1.1495,
+      "step": 11106
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011100131121664847,
+      "loss": 1.0639,
+      "step": 11107
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011098737127093714,
+      "loss": 0.9594,
+      "step": 11108
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011097343110910452,
+      "loss": 1.0963,
+      "step": 11109
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011095949073142486,
+      "loss": 0.7919,
+      "step": 11110
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011094555013817231,
+      "loss": 0.9371,
+      "step": 11111
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011093160932962117,
+      "loss": 0.9858,
+      "step": 11112
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011091766830604551,
+      "loss": 1.0969,
+      "step": 11113
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001109037270677197,
+      "loss": 1.0722,
+      "step": 11114
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011088978561491788,
+      "loss": 0.8282,
+      "step": 11115
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001108758439479143,
+      "loss": 0.996,
+      "step": 11116
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001108619020669832,
+      "loss": 0.9893,
+      "step": 11117
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001108479599723988,
+      "loss": 0.9903,
+      "step": 11118
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001108340176644353,
+      "loss": 0.9219,
+      "step": 11119
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011082007514336707,
+      "loss": 1.212,
+      "step": 11120
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011080613240946825,
+      "loss": 0.9679,
+      "step": 11121
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001107921894630131,
+      "loss": 0.9941,
+      "step": 11122
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011077824630427594,
+      "loss": 0.8809,
+      "step": 11123
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000110764302933531,
+      "loss": 1.1497,
+      "step": 11124
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011075035935105252,
+      "loss": 1.0566,
+      "step": 11125
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011073641555711478,
+      "loss": 1.1024,
+      "step": 11126
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011072247155199206,
+      "loss": 1.1782,
+      "step": 11127
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011070852733595864,
+      "loss": 1.0712,
+      "step": 11128
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011069458290928883,
+      "loss": 1.0483,
+      "step": 11129
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011068063827225686,
+      "loss": 1.0634,
+      "step": 11130
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001106666934251371,
+      "loss": 1.0654,
+      "step": 11131
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011065274836820372,
+      "loss": 1.0933,
+      "step": 11132
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011063880310173113,
+      "loss": 0.9289,
+      "step": 11133
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011062485762599362,
+      "loss": 1.1019,
+      "step": 11134
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011061091194126548,
+      "loss": 0.9324,
+      "step": 11135
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011059696604782094,
+      "loss": 1.3003,
+      "step": 11136
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011058301994593448,
+      "loss": 0.9557,
+      "step": 11137
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011056907363588029,
+      "loss": 1.1163,
+      "step": 11138
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011055512711793272,
+      "loss": 0.971,
+      "step": 11139
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011054118039236612,
+      "loss": 0.8607,
+      "step": 11140
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011052723345945483,
+      "loss": 1.0756,
+      "step": 11141
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011051328631947315,
+      "loss": 0.9478,
+      "step": 11142
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011049933897269547,
+      "loss": 1.1699,
+      "step": 11143
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011048539141939606,
+      "loss": 1.0683,
+      "step": 11144
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011047144365984934,
+      "loss": 1.0478,
+      "step": 11145
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011045749569432962,
+      "loss": 1.1699,
+      "step": 11146
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011044354752311128,
+      "loss": 1.0146,
+      "step": 11147
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011042959914646867,
+      "loss": 1.0108,
+      "step": 11148
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011041565056467614,
+      "loss": 0.8646,
+      "step": 11149
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011040170177800806,
+      "loss": 1.103,
+      "step": 11150
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011038775278673881,
+      "loss": 1.1687,
+      "step": 11151
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001103738035911428,
+      "loss": 1.0715,
+      "step": 11152
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011035985419149434,
+      "loss": 0.9136,
+      "step": 11153
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001103459045880679,
+      "loss": 1.1183,
+      "step": 11154
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001103319547811378,
+      "loss": 1.0611,
+      "step": 11155
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011031800477097846,
+      "loss": 1.0127,
+      "step": 11156
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011030405455786425,
+      "loss": 1.0854,
+      "step": 11157
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011029010414206965,
+      "loss": 1.0695,
+      "step": 11158
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011027615352386894,
+      "loss": 1.1068,
+      "step": 11159
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011026220270353662,
+      "loss": 0.9693,
+      "step": 11160
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011024825168134707,
+      "loss": 0.9437,
+      "step": 11161
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001102343004575747,
+      "loss": 1.2447,
+      "step": 11162
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011022034903249396,
+      "loss": 1.0205,
+      "step": 11163
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011020639740637928,
+      "loss": 0.906,
+      "step": 11164
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000110192445579505,
+      "loss": 1.0196,
+      "step": 11165
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011017849355214567,
+      "loss": 1.0716,
+      "step": 11166
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011016454132457566,
+      "loss": 0.9028,
+      "step": 11167
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011015058889706942,
+      "loss": 0.9185,
+      "step": 11168
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011013663626990138,
+      "loss": 1.0114,
+      "step": 11169
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011012268344334604,
+      "loss": 1.054,
+      "step": 11170
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001101087304176778,
+      "loss": 1.0823,
+      "step": 11171
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011009477719317111,
+      "loss": 0.9374,
+      "step": 11172
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011008082377010045,
+      "loss": 0.8898,
+      "step": 11173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001100668701487403,
+      "loss": 0.7943,
+      "step": 11174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011005291632936512,
+      "loss": 1.0903,
+      "step": 11175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011003896231224936,
+      "loss": 1.0773,
+      "step": 11176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011002500809766752,
+      "loss": 1.0275,
+      "step": 11177
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00011001105368589406,
+      "loss": 0.9654,
+      "step": 11178
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010999709907720346,
+      "loss": 0.892,
+      "step": 11179
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010998314427187021,
+      "loss": 1.0811,
+      "step": 11180
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010996918927016886,
+      "loss": 1.0468,
+      "step": 11181
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010995523407237377,
+      "loss": 1.1668,
+      "step": 11182
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010994127867875959,
+      "loss": 1.1303,
+      "step": 11183
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010992732308960072,
+      "loss": 0.9918,
+      "step": 11184
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001099133673051717,
+      "loss": 0.7948,
+      "step": 11185
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010989941132574704,
+      "loss": 1.1122,
+      "step": 11186
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010988545515160127,
+      "loss": 0.9007,
+      "step": 11187
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010987149878300887,
+      "loss": 0.9548,
+      "step": 11188
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010985754222024436,
+      "loss": 1.1762,
+      "step": 11189
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010984358546358232,
+      "loss": 1.082,
+      "step": 11190
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010982962851329722,
+      "loss": 1.1243,
+      "step": 11191
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010981567136966362,
+      "loss": 0.9267,
+      "step": 11192
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001098017140329561,
+      "loss": 0.8715,
+      "step": 11193
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0001097877565034491,
+      "loss": 1.0122,
+      "step": 11194
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010977379878141723,
+      "loss": 0.9691,
+      "step": 11195
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010975984086713503,
+      "loss": 1.0524,
+      "step": 11196
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010974588276087705,
+      "loss": 1.276,
+      "step": 11197
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010973192446291782,
+      "loss": 0.9096,
+      "step": 11198
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00010971796597353195,
+      "loss": 1.0468,
+      "step": 11199
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000109704007292994,
+      "loss": 1.0107,
+      "step": 11200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010969004842157849,
+      "loss": 1.0962,
+      "step": 11201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010967608935955999,
+      "loss": 0.9959,
+      "step": 11202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010966213010721311,
+      "loss": 0.8517,
+      "step": 11203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010964817066481246,
+      "loss": 0.9887,
+      "step": 11204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010963421103263252,
+      "loss": 1.2275,
+      "step": 11205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010962025121094797,
+      "loss": 1.0234,
+      "step": 11206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010960629120003335,
+      "loss": 0.9845,
+      "step": 11207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010959233100016324,
+      "loss": 1.0017,
+      "step": 11208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001095783706116123,
+      "loss": 1.1543,
+      "step": 11209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001095644100346551,
+      "loss": 1.113,
+      "step": 11210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010955044926956619,
+      "loss": 1.0931,
+      "step": 11211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010953648831662027,
+      "loss": 0.9797,
+      "step": 11212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010952252717609188,
+      "loss": 0.9366,
+      "step": 11213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010950856584825566,
+      "loss": 0.9162,
+      "step": 11214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010949460433338624,
+      "loss": 0.861,
+      "step": 11215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010948064263175821,
+      "loss": 0.9391,
+      "step": 11216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010946668074364628,
+      "loss": 0.9337,
+      "step": 11217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010945271866932496,
+      "loss": 0.9023,
+      "step": 11218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010943875640906897,
+      "loss": 0.857,
+      "step": 11219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001094247939631529,
+      "loss": 0.9482,
+      "step": 11220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010941083133185146,
+      "loss": 0.8989,
+      "step": 11221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010939686851543918,
+      "loss": 1.1889,
+      "step": 11222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010938290551419081,
+      "loss": 1.0606,
+      "step": 11223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010936894232838098,
+      "loss": 1.0646,
+      "step": 11224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001093549789582843,
+      "loss": 0.764,
+      "step": 11225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010934101540417548,
+      "loss": 0.9973,
+      "step": 11226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010932705166632915,
+      "loss": 1.1859,
+      "step": 11227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010931308774501998,
+      "loss": 0.9458,
+      "step": 11228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010929912364052268,
+      "loss": 1.0034,
+      "step": 11229
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010928515935311188,
+      "loss": 0.8838,
+      "step": 11230
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010927119488306226,
+      "loss": 1.0252,
+      "step": 11231
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010925723023064851,
+      "loss": 1.0871,
+      "step": 11232
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010924326539614534,
+      "loss": 1.1682,
+      "step": 11233
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010922930037982737,
+      "loss": 1.1726,
+      "step": 11234
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010921533518196936,
+      "loss": 1.0268,
+      "step": 11235
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010920136980284597,
+      "loss": 1.1231,
+      "step": 11236
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010918740424273191,
+      "loss": 0.738,
+      "step": 11237
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010917343850190187,
+      "loss": 1.083,
+      "step": 11238
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001091594725806306,
+      "loss": 0.9381,
+      "step": 11239
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010914550647919275,
+      "loss": 1.0659,
+      "step": 11240
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010913154019786307,
+      "loss": 1.115,
+      "step": 11241
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010911757373691624,
+      "loss": 0.9717,
+      "step": 11242
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010910360709662701,
+      "loss": 0.893,
+      "step": 11243
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010908964027727014,
+      "loss": 1.1298,
+      "step": 11244
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010907567327912025,
+      "loss": 1.0459,
+      "step": 11245
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010906170610245219,
+      "loss": 0.8383,
+      "step": 11246
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001090477387475406,
+      "loss": 0.9332,
+      "step": 11247
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010903377121466027,
+      "loss": 1.0734,
+      "step": 11248
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010901980350408592,
+      "loss": 1.1153,
+      "step": 11249
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010900583561609232,
+      "loss": 1.0992,
+      "step": 11250
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010899186755095416,
+      "loss": 1.0684,
+      "step": 11251
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010897789930894629,
+      "loss": 0.9398,
+      "step": 11252
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010896393089034336,
+      "loss": 1.0974,
+      "step": 11253
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001089499622954202,
+      "loss": 0.9989,
+      "step": 11254
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010893599352445152,
+      "loss": 0.9031,
+      "step": 11255
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010892202457771215,
+      "loss": 1.0131,
+      "step": 11256
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010890805545547678,
+      "loss": 1.1443,
+      "step": 11257
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010889408615802023,
+      "loss": 1.1344,
+      "step": 11258
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010888011668561726,
+      "loss": 1.0649,
+      "step": 11259
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010886614703854268,
+      "loss": 1.1782,
+      "step": 11260
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010885217721707124,
+      "loss": 1.0292,
+      "step": 11261
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010883820722147773,
+      "loss": 1.0449,
+      "step": 11262
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010882423705203698,
+      "loss": 0.9772,
+      "step": 11263
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001088102667090237,
+      "loss": 0.9013,
+      "step": 11264
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010879629619271274,
+      "loss": 1.0346,
+      "step": 11265
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010878232550337892,
+      "loss": 0.9425,
+      "step": 11266
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010876835464129703,
+      "loss": 1.2409,
+      "step": 11267
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001087543836067418,
+      "loss": 0.8527,
+      "step": 11268
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010874041239998816,
+      "loss": 1.1656,
+      "step": 11269
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010872644102131085,
+      "loss": 0.9604,
+      "step": 11270
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010871246947098468,
+      "loss": 0.8385,
+      "step": 11271
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001086984977492845,
+      "loss": 0.9352,
+      "step": 11272
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010868452585648517,
+      "loss": 1.142,
+      "step": 11273
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010867055379286142,
+      "loss": 0.8948,
+      "step": 11274
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010865658155868817,
+      "loss": 1.0977,
+      "step": 11275
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010864260915424019,
+      "loss": 1.0469,
+      "step": 11276
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010862863657979237,
+      "loss": 0.8365,
+      "step": 11277
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010861466383561946,
+      "loss": 1.0,
+      "step": 11278
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010860069092199645,
+      "loss": 1.1662,
+      "step": 11279
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010858671783919804,
+      "loss": 1.0715,
+      "step": 11280
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010857274458749918,
+      "loss": 1.0906,
+      "step": 11281
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010855877116717467,
+      "loss": 0.8257,
+      "step": 11282
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001085447975784994,
+      "loss": 0.9489,
+      "step": 11283
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001085308238217482,
+      "loss": 1.1854,
+      "step": 11284
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010851684989719594,
+      "loss": 1.0991,
+      "step": 11285
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010850287580511753,
+      "loss": 1.0184,
+      "step": 11286
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010848890154578777,
+      "loss": 1.0959,
+      "step": 11287
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010847492711948157,
+      "loss": 0.9833,
+      "step": 11288
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010846095252647381,
+      "loss": 1.1249,
+      "step": 11289
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010844697776703938,
+      "loss": 0.7522,
+      "step": 11290
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001084330028414531,
+      "loss": 1.0126,
+      "step": 11291
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010841902774998995,
+      "loss": 0.9343,
+      "step": 11292
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010840505249292476,
+      "loss": 1.1504,
+      "step": 11293
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010839107707053244,
+      "loss": 0.9116,
+      "step": 11294
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010837710148308787,
+      "loss": 0.8878,
+      "step": 11295
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010836312573086598,
+      "loss": 1.0216,
+      "step": 11296
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010834914981414163,
+      "loss": 1.1571,
+      "step": 11297
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010833517373318975,
+      "loss": 1.3121,
+      "step": 11298
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010832119748828527,
+      "loss": 1.1314,
+      "step": 11299
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010830722107970308,
+      "loss": 0.9475,
+      "step": 11300
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010829324450771806,
+      "loss": 0.8981,
+      "step": 11301
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010827926777260522,
+      "loss": 1.0354,
+      "step": 11302
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010826529087463939,
+      "loss": 0.9651,
+      "step": 11303
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010825131381409552,
+      "loss": 1.004,
+      "step": 11304
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010823733659124857,
+      "loss": 0.9819,
+      "step": 11305
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010822335920637345,
+      "loss": 1.0331,
+      "step": 11306
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001082093816597451,
+      "loss": 0.9622,
+      "step": 11307
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010819540395163845,
+      "loss": 0.9585,
+      "step": 11308
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010818142608232849,
+      "loss": 0.8637,
+      "step": 11309
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010816744805209007,
+      "loss": 1.0589,
+      "step": 11310
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010815346986119819,
+      "loss": 1.0783,
+      "step": 11311
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010813949150992781,
+      "loss": 1.023,
+      "step": 11312
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010812551299855389,
+      "loss": 1.1658,
+      "step": 11313
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010811153432735132,
+      "loss": 1.1095,
+      "step": 11314
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010809755549659517,
+      "loss": 0.931,
+      "step": 11315
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010808357650656032,
+      "loss": 0.8357,
+      "step": 11316
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010806959735752174,
+      "loss": 0.9476,
+      "step": 11317
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010805561804975441,
+      "loss": 0.9685,
+      "step": 11318
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010804163858353336,
+      "loss": 1.0042,
+      "step": 11319
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010802765895913349,
+      "loss": 0.8484,
+      "step": 11320
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010801367917682979,
+      "loss": 1.006,
+      "step": 11321
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010799969923689727,
+      "loss": 0.9663,
+      "step": 11322
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010798571913961093,
+      "loss": 1.0288,
+      "step": 11323
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010797173888524567,
+      "loss": 1.0051,
+      "step": 11324
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010795775847407659,
+      "loss": 0.9657,
+      "step": 11325
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010794377790637862,
+      "loss": 0.9168,
+      "step": 11326
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010792979718242678,
+      "loss": 1.0449,
+      "step": 11327
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010791581630249603,
+      "loss": 0.9886,
+      "step": 11328
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010790183526686144,
+      "loss": 0.8141,
+      "step": 11329
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000107887854075798,
+      "loss": 0.7979,
+      "step": 11330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010787387272958065,
+      "loss": 0.8983,
+      "step": 11331
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001078598912284845,
+      "loss": 0.8987,
+      "step": 11332
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001078459095727845,
+      "loss": 1.1305,
+      "step": 11333
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001078319277627557,
+      "loss": 0.9228,
+      "step": 11334
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001078179457986731,
+      "loss": 0.9266,
+      "step": 11335
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010780396368081178,
+      "loss": 0.8759,
+      "step": 11336
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010778998140944665,
+      "loss": 1.2338,
+      "step": 11337
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010777599898485289,
+      "loss": 1.1704,
+      "step": 11338
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010776201640730541,
+      "loss": 1.0879,
+      "step": 11339
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010774803367707931,
+      "loss": 1.0482,
+      "step": 11340
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010773405079444963,
+      "loss": 1.0223,
+      "step": 11341
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010772006775969142,
+      "loss": 1.3105,
+      "step": 11342
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010770608457307965,
+      "loss": 1.0193,
+      "step": 11343
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010769210123488945,
+      "loss": 1.1377,
+      "step": 11344
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010767811774539584,
+      "loss": 0.9614,
+      "step": 11345
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001076641341048739,
+      "loss": 1.0435,
+      "step": 11346
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010765015031359862,
+      "loss": 1.0363,
+      "step": 11347
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010763616637184517,
+      "loss": 0.968,
+      "step": 11348
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010762218227988851,
+      "loss": 0.8382,
+      "step": 11349
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010760819803800376,
+      "loss": 1.1493,
+      "step": 11350
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000107594213646466,
+      "loss": 1.1981,
+      "step": 11351
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010758022910555026,
+      "loss": 1.0195,
+      "step": 11352
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010756624441553167,
+      "loss": 0.9818,
+      "step": 11353
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010755225957668522,
+      "loss": 0.9476,
+      "step": 11354
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001075382745892861,
+      "loss": 0.9529,
+      "step": 11355
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010752428945360931,
+      "loss": 1.1414,
+      "step": 11356
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010751030416992996,
+      "loss": 0.9681,
+      "step": 11357
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010749631873852317,
+      "loss": 0.8974,
+      "step": 11358
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010748233315966403,
+      "loss": 0.9311,
+      "step": 11359
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010746834743362757,
+      "loss": 0.9079,
+      "step": 11360
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010745436156068897,
+      "loss": 0.9997,
+      "step": 11361
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010744037554112329,
+      "loss": 0.9342,
+      "step": 11362
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010742638937520566,
+      "loss": 0.8359,
+      "step": 11363
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010741240306321112,
+      "loss": 1.0635,
+      "step": 11364
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010739841660541487,
+      "loss": 1.1392,
+      "step": 11365
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000107384430002092,
+      "loss": 1.085,
+      "step": 11366
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010737044325351757,
+      "loss": 0.9642,
+      "step": 11367
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010735645635996676,
+      "loss": 0.8839,
+      "step": 11368
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010734246932171467,
+      "loss": 1.0468,
+      "step": 11369
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010732848213903639,
+      "loss": 1.0737,
+      "step": 11370
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010731449481220714,
+      "loss": 1.0183,
+      "step": 11371
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010730050734150196,
+      "loss": 1.1138,
+      "step": 11372
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010728651972719603,
+      "loss": 1.0755,
+      "step": 11373
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010727253196956445,
+      "loss": 0.996,
+      "step": 11374
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001072585440688824,
+      "loss": 1.1428,
+      "step": 11375
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010724455602542502,
+      "loss": 0.9309,
+      "step": 11376
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010723056783946738,
+      "loss": 0.8853,
+      "step": 11377
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010721657951128474,
+      "loss": 1.1037,
+      "step": 11378
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010720259104115217,
+      "loss": 1.1095,
+      "step": 11379
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010718860242934486,
+      "loss": 0.9375,
+      "step": 11380
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010717461367613793,
+      "loss": 1.2483,
+      "step": 11381
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001071606247818066,
+      "loss": 1.0357,
+      "step": 11382
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010714663574662594,
+      "loss": 1.1083,
+      "step": 11383
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010713264657087121,
+      "loss": 0.9023,
+      "step": 11384
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001071186572548175,
+      "loss": 1.0154,
+      "step": 11385
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010710466779874005,
+      "loss": 1.058,
+      "step": 11386
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010709067820291395,
+      "loss": 0.9157,
+      "step": 11387
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010707668846761445,
+      "loss": 0.9374,
+      "step": 11388
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010706269859311669,
+      "loss": 0.9716,
+      "step": 11389
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010704870857969585,
+      "loss": 1.0577,
+      "step": 11390
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010703471842762711,
+      "loss": 0.9763,
+      "step": 11391
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010702072813718571,
+      "loss": 1.1092,
+      "step": 11392
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010700673770864673,
+      "loss": 0.9601,
+      "step": 11393
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001069927471422855,
+      "loss": 0.9192,
+      "step": 11394
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010697875643837707,
+      "loss": 0.8534,
+      "step": 11395
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010696476559719675,
+      "loss": 0.9079,
+      "step": 11396
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010695077461901968,
+      "loss": 0.8634,
+      "step": 11397
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010693678350412106,
+      "loss": 1.0613,
+      "step": 11398
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010692279225277618,
+      "loss": 1.2235,
+      "step": 11399
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001069088008652601,
+      "loss": 1.0832,
+      "step": 11400
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010689480934184816,
+      "loss": 1.0209,
+      "step": 11401
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001068808176828155,
+      "loss": 1.1766,
+      "step": 11402
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010686682588843737,
+      "loss": 0.9919,
+      "step": 11403
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.000106852833958989,
+      "loss": 1.0301,
+      "step": 11404
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010683884189474558,
+      "loss": 0.9871,
+      "step": 11405
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001068248496959823,
+      "loss": 1.0269,
+      "step": 11406
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010681085736297447,
+      "loss": 0.9529,
+      "step": 11407
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010679686489599725,
+      "loss": 0.9481,
+      "step": 11408
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010678287229532596,
+      "loss": 1.0624,
+      "step": 11409
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010676887956123569,
+      "loss": 1.1172,
+      "step": 11410
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001067548866940018,
+      "loss": 1.0402,
+      "step": 11411
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010674089369389948,
+      "loss": 0.966,
+      "step": 11412
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010672690056120399,
+      "loss": 0.9206,
+      "step": 11413
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010671290729619055,
+      "loss": 0.9666,
+      "step": 11414
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010669891389913442,
+      "loss": 1.1082,
+      "step": 11415
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010668492037031086,
+      "loss": 1.0787,
+      "step": 11416
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010667092670999513,
+      "loss": 0.9227,
+      "step": 11417
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010665693291846244,
+      "loss": 0.9015,
+      "step": 11418
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010664293899598807,
+      "loss": 1.0431,
+      "step": 11419
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010662894494284731,
+      "loss": 0.929,
+      "step": 11420
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010661495075931539,
+      "loss": 0.9991,
+      "step": 11421
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010660095644566761,
+      "loss": 0.8651,
+      "step": 11422
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010658696200217914,
+      "loss": 1.0785,
+      "step": 11423
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001065729674291254,
+      "loss": 0.9373,
+      "step": 11424
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010655897272678155,
+      "loss": 1.1863,
+      "step": 11425
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010654497789542291,
+      "loss": 1.031,
+      "step": 11426
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010653098293532473,
+      "loss": 0.8442,
+      "step": 11427
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010651698784676234,
+      "loss": 1.0814,
+      "step": 11428
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010650299263001097,
+      "loss": 0.8201,
+      "step": 11429
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010648899728534591,
+      "loss": 0.9075,
+      "step": 11430
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.00010647500181304249,
+      "loss": 0.9449,
+      "step": 11431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000106461006213376,
+      "loss": 1.0106,
+      "step": 11432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010644701048662164,
+      "loss": 1.1612,
+      "step": 11433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010643301463305483,
+      "loss": 1.0223,
+      "step": 11434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001064190186529508,
+      "loss": 1.0307,
+      "step": 11435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010640502254658485,
+      "loss": 0.8493,
+      "step": 11436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001063910263142323,
+      "loss": 1.1225,
+      "step": 11437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010637702995616847,
+      "loss": 1.1689,
+      "step": 11438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001063630334726686,
+      "loss": 0.9874,
+      "step": 11439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010634903686400812,
+      "loss": 1.087,
+      "step": 11440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010633504013046223,
+      "loss": 1.0531,
+      "step": 11441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010632104327230627,
+      "loss": 0.9583,
+      "step": 11442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001063070462898156,
+      "loss": 0.9987,
+      "step": 11443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010629304918326553,
+      "loss": 0.9478,
+      "step": 11444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010627905195293135,
+      "loss": 0.8721,
+      "step": 11445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010626505459908838,
+      "loss": 1.0771,
+      "step": 11446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010625105712201202,
+      "loss": 1.0802,
+      "step": 11447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010623705952197754,
+      "loss": 1.1589,
+      "step": 11448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010622306179926025,
+      "loss": 1.1682,
+      "step": 11449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010620906395413555,
+      "loss": 1.0962,
+      "step": 11450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010619506598687874,
+      "loss": 0.9533,
+      "step": 11451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010618106789776516,
+      "loss": 1.051,
+      "step": 11452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010616706968707015,
+      "loss": 0.9597,
+      "step": 11453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010615307135506907,
+      "loss": 1.0264,
+      "step": 11454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010613907290203727,
+      "loss": 0.9581,
+      "step": 11455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010612507432825005,
+      "loss": 0.9521,
+      "step": 11456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010611107563398284,
+      "loss": 0.9226,
+      "step": 11457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010609707681951093,
+      "loss": 0.9449,
+      "step": 11458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010608307788510969,
+      "loss": 0.8469,
+      "step": 11459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010606907883105451,
+      "loss": 0.9779,
+      "step": 11460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010605507965762074,
+      "loss": 0.8623,
+      "step": 11461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010604108036508371,
+      "loss": 1.1179,
+      "step": 11462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001060270809537188,
+      "loss": 0.9579,
+      "step": 11463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010601308142380139,
+      "loss": 1.1209,
+      "step": 11464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010599908177560685,
+      "loss": 0.9494,
+      "step": 11465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010598508200941053,
+      "loss": 1.0365,
+      "step": 11466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010597108212548785,
+      "loss": 0.9669,
+      "step": 11467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010595708212411417,
+      "loss": 1.0746,
+      "step": 11468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010594308200556483,
+      "loss": 0.819,
+      "step": 11469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010592908177011526,
+      "loss": 0.9543,
+      "step": 11470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010591508141804083,
+      "loss": 1.1863,
+      "step": 11471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010590108094961692,
+      "loss": 0.8839,
+      "step": 11472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001058870803651189,
+      "loss": 1.0059,
+      "step": 11473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010587307966482223,
+      "loss": 0.954,
+      "step": 11474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010585907884900221,
+      "loss": 1.0063,
+      "step": 11475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001058450779179343,
+      "loss": 1.0525,
+      "step": 11476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010583107687189388,
+      "loss": 1.0104,
+      "step": 11477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010581707571115637,
+      "loss": 0.8976,
+      "step": 11478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010580307443599711,
+      "loss": 0.9225,
+      "step": 11479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010578907304669159,
+      "loss": 0.9058,
+      "step": 11480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010577507154351514,
+      "loss": 0.9918,
+      "step": 11481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010576106992674322,
+      "loss": 0.9604,
+      "step": 11482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010574706819665121,
+      "loss": 0.8872,
+      "step": 11483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010573306635351456,
+      "loss": 1.0694,
+      "step": 11484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010571906439760866,
+      "loss": 0.8833,
+      "step": 11485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010570506232920894,
+      "loss": 0.9432,
+      "step": 11486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010569106014859078,
+      "loss": 1.0945,
+      "step": 11487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010567705785602966,
+      "loss": 1.0617,
+      "step": 11488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010566305545180096,
+      "loss": 0.7114,
+      "step": 11489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010564905293618014,
+      "loss": 1.1995,
+      "step": 11490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010563505030944263,
+      "loss": 1.0685,
+      "step": 11491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010562104757186382,
+      "loss": 0.7866,
+      "step": 11492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010560704472371919,
+      "loss": 0.9285,
+      "step": 11493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010559304176528415,
+      "loss": 0.9419,
+      "step": 11494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010557903869683416,
+      "loss": 1.0669,
+      "step": 11495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010556503551864458,
+      "loss": 0.97,
+      "step": 11496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010555103223099098,
+      "loss": 0.958,
+      "step": 11497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010553702883414874,
+      "loss": 0.8861,
+      "step": 11498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010552302532839327,
+      "loss": 1.1195,
+      "step": 11499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010550902171400007,
+      "loss": 1.0297,
+      "step": 11500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001054950179912446,
+      "loss": 1.089,
+      "step": 11501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010548101416040224,
+      "loss": 1.1941,
+      "step": 11502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010546701022174854,
+      "loss": 0.8979,
+      "step": 11503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010545300617555889,
+      "loss": 0.8764,
+      "step": 11504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010543900202210876,
+      "loss": 0.9632,
+      "step": 11505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010542499776167363,
+      "loss": 1.0194,
+      "step": 11506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010541099339452896,
+      "loss": 0.933,
+      "step": 11507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010539698892095021,
+      "loss": 1.1594,
+      "step": 11508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010538298434121283,
+      "loss": 1.0863,
+      "step": 11509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010536897965559231,
+      "loss": 1.0783,
+      "step": 11510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010535497486436414,
+      "loss": 1.2444,
+      "step": 11511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010534096996780375,
+      "loss": 0.9497,
+      "step": 11512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010532696496618664,
+      "loss": 0.9079,
+      "step": 11513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001053129598597883,
+      "loss": 0.8849,
+      "step": 11514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010529895464888418,
+      "loss": 1.0086,
+      "step": 11515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010528494933374981,
+      "loss": 1.1668,
+      "step": 11516
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010527094391466061,
+      "loss": 0.9739,
+      "step": 11517
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010525693839189214,
+      "loss": 1.0806,
+      "step": 11518
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001052429327657198,
+      "loss": 0.8918,
+      "step": 11519
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010522892703641916,
+      "loss": 0.9849,
+      "step": 11520
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010521492120426566,
+      "loss": 1.0575,
+      "step": 11521
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010520091526953481,
+      "loss": 0.9695,
+      "step": 11522
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010518690923250212,
+      "loss": 0.8973,
+      "step": 11523
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001051729030934431,
+      "loss": 1.0768,
+      "step": 11524
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010515889685263316,
+      "loss": 1.0882,
+      "step": 11525
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010514489051034795,
+      "loss": 1.2532,
+      "step": 11526
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010513088406686284,
+      "loss": 1.0352,
+      "step": 11527
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010511687752245343,
+      "loss": 0.9466,
+      "step": 11528
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010510287087739517,
+      "loss": 1.1363,
+      "step": 11529
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010508886413196361,
+      "loss": 1.0046,
+      "step": 11530
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010507485728643423,
+      "loss": 1.0871,
+      "step": 11531
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010506085034108256,
+      "loss": 0.8908,
+      "step": 11532
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010504684329618409,
+      "loss": 1.047,
+      "step": 11533
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010503283615201437,
+      "loss": 1.0186,
+      "step": 11534
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010501882890884893,
+      "loss": 1.0474,
+      "step": 11535
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010500482156696325,
+      "loss": 1.2425,
+      "step": 11536
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010499081412663293,
+      "loss": 1.0117,
+      "step": 11537
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010497680658813338,
+      "loss": 0.9702,
+      "step": 11538
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010496279895174022,
+      "loss": 1.1728,
+      "step": 11539
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010494879121772893,
+      "loss": 1.0666,
+      "step": 11540
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010493478338637512,
+      "loss": 0.9876,
+      "step": 11541
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001049207754579542,
+      "loss": 0.9236,
+      "step": 11542
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010490676743274181,
+      "loss": 0.9735,
+      "step": 11543
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010489275931101344,
+      "loss": 1.0047,
+      "step": 11544
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010487875109304463,
+      "loss": 0.9669,
+      "step": 11545
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010486474277911094,
+      "loss": 1.0291,
+      "step": 11546
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010485073436948791,
+      "loss": 0.9833,
+      "step": 11547
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010483672586445106,
+      "loss": 1.0566,
+      "step": 11548
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010482271726427598,
+      "loss": 1.0441,
+      "step": 11549
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010480870856923817,
+      "loss": 0.8977,
+      "step": 11550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001047946997796132,
+      "loss": 1.1456,
+      "step": 11551
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010478069089567666,
+      "loss": 0.9789,
+      "step": 11552
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010476668191770407,
+      "loss": 1.2984,
+      "step": 11553
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010475267284597095,
+      "loss": 0.8632,
+      "step": 11554
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010473866368075292,
+      "loss": 0.9641,
+      "step": 11555
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001047246544223255,
+      "loss": 1.1832,
+      "step": 11556
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010471064507096426,
+      "loss": 1.0777,
+      "step": 11557
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010469663562694481,
+      "loss": 0.894,
+      "step": 11558
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010468262609054265,
+      "loss": 1.027,
+      "step": 11559
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010466861646203339,
+      "loss": 0.9767,
+      "step": 11560
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010465460674169256,
+      "loss": 1.0147,
+      "step": 11561
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010464059692979576,
+      "loss": 0.8662,
+      "step": 11562
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010462658702661854,
+      "loss": 0.8588,
+      "step": 11563
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010461257703243652,
+      "loss": 1.0984,
+      "step": 11564
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010459856694752521,
+      "loss": 1.1712,
+      "step": 11565
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010458455677216026,
+      "loss": 1.0427,
+      "step": 11566
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010457054650661719,
+      "loss": 1.0768,
+      "step": 11567
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010455653615117163,
+      "loss": 0.9509,
+      "step": 11568
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001045425257060991,
+      "loss": 0.9967,
+      "step": 11569
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010452851517167528,
+      "loss": 1.0287,
+      "step": 11570
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010451450454817567,
+      "loss": 1.1819,
+      "step": 11571
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010450049383587588,
+      "loss": 1.1754,
+      "step": 11572
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010448648303505151,
+      "loss": 0.9414,
+      "step": 11573
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010447247214597816,
+      "loss": 0.9616,
+      "step": 11574
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001044584611689314,
+      "loss": 1.0805,
+      "step": 11575
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010444445010418686,
+      "loss": 1.0001,
+      "step": 11576
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010443043895202012,
+      "loss": 1.1908,
+      "step": 11577
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010441642771270675,
+      "loss": 1.0893,
+      "step": 11578
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001044024163865224,
+      "loss": 0.7549,
+      "step": 11579
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010438840497374263,
+      "loss": 1.093,
+      "step": 11580
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010437439347464306,
+      "loss": 0.9059,
+      "step": 11581
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010436038188949932,
+      "loss": 1.0551,
+      "step": 11582
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010434637021858702,
+      "loss": 1.0798,
+      "step": 11583
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010433235846218171,
+      "loss": 1.0095,
+      "step": 11584
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010431834662055904,
+      "loss": 0.9243,
+      "step": 11585
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010430433469399463,
+      "loss": 0.9663,
+      "step": 11586
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010429032268276409,
+      "loss": 1.0219,
+      "step": 11587
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000104276310587143,
+      "loss": 0.9835,
+      "step": 11588
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010426229840740705,
+      "loss": 0.997,
+      "step": 11589
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010424828614383177,
+      "loss": 0.9817,
+      "step": 11590
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010423427379669286,
+      "loss": 1.2005,
+      "step": 11591
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010422026136626588,
+      "loss": 0.9318,
+      "step": 11592
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010420624885282653,
+      "loss": 0.9871,
+      "step": 11593
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010419223625665036,
+      "loss": 1.0259,
+      "step": 11594
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010417822357801301,
+      "loss": 1.103,
+      "step": 11595
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010416421081719015,
+      "loss": 1.0018,
+      "step": 11596
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010415019797445738,
+      "loss": 1.0092,
+      "step": 11597
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010413618505009033,
+      "loss": 0.9691,
+      "step": 11598
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010412217204436468,
+      "loss": 1.0214,
+      "step": 11599
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010410815895755597,
+      "loss": 1.032,
+      "step": 11600
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010409414578993992,
+      "loss": 0.9237,
+      "step": 11601
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010408013254179214,
+      "loss": 0.9891,
+      "step": 11602
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001040661192133883,
+      "loss": 0.91,
+      "step": 11603
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000104052105805004,
+      "loss": 1.0594,
+      "step": 11604
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010403809231691486,
+      "loss": 0.8953,
+      "step": 11605
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010402407874939662,
+      "loss": 0.9267,
+      "step": 11606
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010401006510272484,
+      "loss": 1.1452,
+      "step": 11607
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010399605137717519,
+      "loss": 1.221,
+      "step": 11608
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010398203757302334,
+      "loss": 0.8643,
+      "step": 11609
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010396802369054495,
+      "loss": 1.3142,
+      "step": 11610
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010395400973001561,
+      "loss": 0.9217,
+      "step": 11611
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010393999569171104,
+      "loss": 0.9216,
+      "step": 11612
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010392598157590688,
+      "loss": 0.9007,
+      "step": 11613
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010391196738287876,
+      "loss": 0.9411,
+      "step": 11614
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010389795311290236,
+      "loss": 1.0472,
+      "step": 11615
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010388393876625335,
+      "loss": 1.038,
+      "step": 11616
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010386992434320737,
+      "loss": 1.0122,
+      "step": 11617
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010385590984404008,
+      "loss": 0.9425,
+      "step": 11618
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010384189526902716,
+      "loss": 1.0551,
+      "step": 11619
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010382788061844429,
+      "loss": 0.9443,
+      "step": 11620
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001038138658925671,
+      "loss": 0.9648,
+      "step": 11621
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010379985109167131,
+      "loss": 0.8765,
+      "step": 11622
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010378583621603254,
+      "loss": 0.991,
+      "step": 11623
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010377182126592649,
+      "loss": 1.0809,
+      "step": 11624
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010375780624162883,
+      "loss": 0.9911,
+      "step": 11625
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010374379114341521,
+      "loss": 0.9986,
+      "step": 11626
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010372977597156138,
+      "loss": 1.1404,
+      "step": 11627
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001037157607263429,
+      "loss": 1.037,
+      "step": 11628
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010370174540803556,
+      "loss": 0.9484,
+      "step": 11629
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010368773001691499,
+      "loss": 0.8646,
+      "step": 11630
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010367371455325688,
+      "loss": 0.9081,
+      "step": 11631
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001036596990173369,
+      "loss": 1.1051,
+      "step": 11632
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010364568340943077,
+      "loss": 1.043,
+      "step": 11633
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010363166772981412,
+      "loss": 0.899,
+      "step": 11634
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010361765197876272,
+      "loss": 0.9875,
+      "step": 11635
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010360363615655218,
+      "loss": 1.0103,
+      "step": 11636
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010358962026345824,
+      "loss": 1.0683,
+      "step": 11637
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010357560429975654,
+      "loss": 0.8942,
+      "step": 11638
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010356158826572287,
+      "loss": 1.0581,
+      "step": 11639
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010354757216163283,
+      "loss": 1.0716,
+      "step": 11640
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010353355598776214,
+      "loss": 1.1607,
+      "step": 11641
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010351953974438651,
+      "loss": 0.9294,
+      "step": 11642
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010350552343178163,
+      "loss": 1.1668,
+      "step": 11643
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010349150705022324,
+      "loss": 1.0741,
+      "step": 11644
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.000103477490599987,
+      "loss": 1.1975,
+      "step": 11645
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001034634740813486,
+      "loss": 1.1027,
+      "step": 11646
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001034494574945838,
+      "loss": 0.8578,
+      "step": 11647
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010343544083996824,
+      "loss": 1.0175,
+      "step": 11648
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010342142411777767,
+      "loss": 0.8438,
+      "step": 11649
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010340740732828779,
+      "loss": 1.0026,
+      "step": 11650
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.0001033933904717743,
+      "loss": 0.8523,
+      "step": 11651
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010337937354851295,
+      "loss": 0.9242,
+      "step": 11652
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010336535655877942,
+      "loss": 0.932,
+      "step": 11653
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010335133950284938,
+      "loss": 1.2041,
+      "step": 11654
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010333732238099862,
+      "loss": 1.059,
+      "step": 11655
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010332330519350286,
+      "loss": 0.8869,
+      "step": 11656
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010330928794063772,
+      "loss": 0.9828,
+      "step": 11657
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010329527062267903,
+      "loss": 0.9203,
+      "step": 11658
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010328125323990245,
+      "loss": 1.0141,
+      "step": 11659
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010326723579258373,
+      "loss": 0.8767,
+      "step": 11660
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010325321828099854,
+      "loss": 1.0767,
+      "step": 11661
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00010323920070542268,
+      "loss": 0.9006,
+      "step": 11662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010322518306613182,
+      "loss": 0.875,
+      "step": 11663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010321116536340172,
+      "loss": 0.9337,
+      "step": 11664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010319714759750808,
+      "loss": 0.9671,
+      "step": 11665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010318312976872664,
+      "loss": 1.1695,
+      "step": 11666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010316911187733311,
+      "loss": 1.1079,
+      "step": 11667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001031550939236033,
+      "loss": 1.0206,
+      "step": 11668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010314107590781284,
+      "loss": 0.8544,
+      "step": 11669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010312705783023752,
+      "loss": 1.075,
+      "step": 11670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010311303969115307,
+      "loss": 1.139,
+      "step": 11671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010309902149083522,
+      "loss": 0.9577,
+      "step": 11672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010308500322955972,
+      "loss": 1.0064,
+      "step": 11673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010307098490760225,
+      "loss": 1.0835,
+      "step": 11674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010305696652523867,
+      "loss": 1.0167,
+      "step": 11675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010304294808274459,
+      "loss": 0.8868,
+      "step": 11676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010302892958039583,
+      "loss": 1.3796,
+      "step": 11677
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010301491101846812,
+      "loss": 0.9779,
+      "step": 11678
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001030008923972372,
+      "loss": 1.1107,
+      "step": 11679
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010298687371697882,
+      "loss": 0.8999,
+      "step": 11680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010297285497796872,
+      "loss": 0.9704,
+      "step": 11681
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010295883618048263,
+      "loss": 0.8698,
+      "step": 11682
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010294481732479635,
+      "loss": 0.9781,
+      "step": 11683
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010293079841118556,
+      "loss": 0.9351,
+      "step": 11684
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010291677943992608,
+      "loss": 1.0299,
+      "step": 11685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010290276041129364,
+      "loss": 0.961,
+      "step": 11686
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010288874132556396,
+      "loss": 0.8968,
+      "step": 11687
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010287472218301285,
+      "loss": 1.0599,
+      "step": 11688
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010286070298391601,
+      "loss": 1.0717,
+      "step": 11689
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010284668372854924,
+      "loss": 1.0142,
+      "step": 11690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010283266441718833,
+      "loss": 0.9755,
+      "step": 11691
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010281864505010894,
+      "loss": 1.2316,
+      "step": 11692
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001028046256275869,
+      "loss": 0.9296,
+      "step": 11693
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010279060614989797,
+      "loss": 0.9575,
+      "step": 11694
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010277658661731788,
+      "loss": 1.0779,
+      "step": 11695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010276256703012243,
+      "loss": 1.0489,
+      "step": 11696
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010274854738858735,
+      "loss": 0.9854,
+      "step": 11697
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010273452769298844,
+      "loss": 0.9518,
+      "step": 11698
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010272050794360144,
+      "loss": 0.979,
+      "step": 11699
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010270648814070214,
+      "loss": 1.0352,
+      "step": 11700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010269246828456629,
+      "loss": 1.0171,
+      "step": 11701
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010267844837546967,
+      "loss": 0.9341,
+      "step": 11702
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010266442841368806,
+      "loss": 1.0444,
+      "step": 11703
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001026504083994972,
+      "loss": 0.8932,
+      "step": 11704
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010263638833317291,
+      "loss": 0.9592,
+      "step": 11705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010262236821499094,
+      "loss": 0.9666,
+      "step": 11706
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010260834804522702,
+      "loss": 1.1129,
+      "step": 11707
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010259432782415701,
+      "loss": 0.8619,
+      "step": 11708
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010258030755205665,
+      "loss": 0.9414,
+      "step": 11709
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010256628722920169,
+      "loss": 0.9565,
+      "step": 11710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010255226685586796,
+      "loss": 1.019,
+      "step": 11711
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001025382464323312,
+      "loss": 0.8678,
+      "step": 11712
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010252422595886724,
+      "loss": 0.9387,
+      "step": 11713
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001025102054357518,
+      "loss": 0.8811,
+      "step": 11714
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010249618486326069,
+      "loss": 0.8971,
+      "step": 11715
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010248216424166972,
+      "loss": 1.0877,
+      "step": 11716
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010246814357125463,
+      "loss": 1.0211,
+      "step": 11717
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010245412285229124,
+      "loss": 1.034,
+      "step": 11718
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010244010208505535,
+      "loss": 1.007,
+      "step": 11719
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010242608126982264,
+      "loss": 1.0048,
+      "step": 11720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010241206040686908,
+      "loss": 1.1865,
+      "step": 11721
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010239803949647031,
+      "loss": 1.213,
+      "step": 11722
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010238401853890219,
+      "loss": 1.1694,
+      "step": 11723
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001023699975344405,
+      "loss": 0.9254,
+      "step": 11724
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010235597648336104,
+      "loss": 0.9521,
+      "step": 11725
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010234195538593957,
+      "loss": 1.041,
+      "step": 11726
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010232793424245191,
+      "loss": 1.0106,
+      "step": 11727
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010231391305317383,
+      "loss": 1.1032,
+      "step": 11728
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001022998918183812,
+      "loss": 1.0928,
+      "step": 11729
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010228587053834973,
+      "loss": 0.974,
+      "step": 11730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010227184921335526,
+      "loss": 1.0257,
+      "step": 11731
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001022578278436736,
+      "loss": 1.0563,
+      "step": 11732
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010224380642958052,
+      "loss": 0.9539,
+      "step": 11733
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010222978497135181,
+      "loss": 0.9259,
+      "step": 11734
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010221576346926331,
+      "loss": 1.0132,
+      "step": 11735
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010220174192359085,
+      "loss": 0.99,
+      "step": 11736
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010218772033461015,
+      "loss": 1.0709,
+      "step": 11737
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010217369870259705,
+      "loss": 1.0715,
+      "step": 11738
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010215967702782738,
+      "loss": 1.0886,
+      "step": 11739
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010214565531057691,
+      "loss": 1.0519,
+      "step": 11740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010213163355112148,
+      "loss": 0.9396,
+      "step": 11741
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010211761174973688,
+      "loss": 1.0176,
+      "step": 11742
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010210358990669888,
+      "loss": 1.0434,
+      "step": 11743
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001020895680222834,
+      "loss": 1.1396,
+      "step": 11744
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010207554609676613,
+      "loss": 0.9656,
+      "step": 11745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010206152413042295,
+      "loss": 0.8955,
+      "step": 11746
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010204750212352963,
+      "loss": 0.9349,
+      "step": 11747
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010203348007636202,
+      "loss": 1.0874,
+      "step": 11748
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010201945798919593,
+      "loss": 0.9208,
+      "step": 11749
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010200543586230713,
+      "loss": 1.0348,
+      "step": 11750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010199141369597147,
+      "loss": 0.8174,
+      "step": 11751
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010197739149046479,
+      "loss": 0.8506,
+      "step": 11752
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010196336924606283,
+      "loss": 1.1748,
+      "step": 11753
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001019493469630415,
+      "loss": 0.9502,
+      "step": 11754
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010193532464167654,
+      "loss": 1.1634,
+      "step": 11755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010192130228224383,
+      "loss": 0.7935,
+      "step": 11756
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010190727988501913,
+      "loss": 1.1043,
+      "step": 11757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010189325745027831,
+      "loss": 1.1229,
+      "step": 11758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010187923497829719,
+      "loss": 1.0987,
+      "step": 11759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010186521246935156,
+      "loss": 1.0967,
+      "step": 11760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010185118992371725,
+      "loss": 1.0195,
+      "step": 11761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010183716734167008,
+      "loss": 0.9789,
+      "step": 11762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010182314472348589,
+      "loss": 0.81,
+      "step": 11763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001018091220694405,
+      "loss": 1.0728,
+      "step": 11764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010179509937980973,
+      "loss": 0.9641,
+      "step": 11765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001017810766548694,
+      "loss": 1.1069,
+      "step": 11766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010176705389489539,
+      "loss": 1.1137,
+      "step": 11767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010175303110016343,
+      "loss": 1.0391,
+      "step": 11768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010173900827094942,
+      "loss": 0.9503,
+      "step": 11769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010172498540752915,
+      "loss": 0.9044,
+      "step": 11770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010171096251017851,
+      "loss": 0.8975,
+      "step": 11771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010169693957917325,
+      "loss": 0.9874,
+      "step": 11772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010168291661478923,
+      "loss": 1.069,
+      "step": 11773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010166889361730232,
+      "loss": 0.9684,
+      "step": 11774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010165487058698832,
+      "loss": 1.0864,
+      "step": 11775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010164084752412304,
+      "loss": 0.8171,
+      "step": 11776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010162682442898236,
+      "loss": 0.8969,
+      "step": 11777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001016128013018421,
+      "loss": 0.7779,
+      "step": 11778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010159877814297804,
+      "loss": 0.9257,
+      "step": 11779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001015847549526661,
+      "loss": 0.8768,
+      "step": 11780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010157073173118208,
+      "loss": 1.0967,
+      "step": 11781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001015567084788018,
+      "loss": 0.9171,
+      "step": 11782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010154268519580112,
+      "loss": 1.1145,
+      "step": 11783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010152866188245585,
+      "loss": 0.9479,
+      "step": 11784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010151463853904186,
+      "loss": 1.0213,
+      "step": 11785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010150061516583495,
+      "loss": 0.9252,
+      "step": 11786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010148659176311102,
+      "loss": 1.1149,
+      "step": 11787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010147256833114586,
+      "loss": 0.9486,
+      "step": 11788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010145854487021532,
+      "loss": 0.9988,
+      "step": 11789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010144452138059526,
+      "loss": 1.1082,
+      "step": 11790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010143049786256149,
+      "loss": 1.0984,
+      "step": 11791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001014164743163899,
+      "loss": 1.1605,
+      "step": 11792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010140245074235624,
+      "loss": 0.8882,
+      "step": 11793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010138842714073646,
+      "loss": 1.0639,
+      "step": 11794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010137440351180636,
+      "loss": 1.041,
+      "step": 11795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010136037985584177,
+      "loss": 1.1157,
+      "step": 11796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010134635617311853,
+      "loss": 0.994,
+      "step": 11797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010133233246391256,
+      "loss": 0.9802,
+      "step": 11798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010131830872849959,
+      "loss": 0.8146,
+      "step": 11799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010130428496715558,
+      "loss": 0.9444,
+      "step": 11800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010129026118015629,
+      "loss": 1.0411,
+      "step": 11801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001012762373677776,
+      "loss": 0.8097,
+      "step": 11802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010126221353029537,
+      "loss": 1.0163,
+      "step": 11803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010124818966798541,
+      "loss": 0.9082,
+      "step": 11804
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010123416578112365,
+      "loss": 1.194,
+      "step": 11805
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010122014186998585,
+      "loss": 1.0354,
+      "step": 11806
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010120611793484789,
+      "loss": 0.9172,
+      "step": 11807
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010119209397598563,
+      "loss": 0.9471,
+      "step": 11808
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010117806999367492,
+      "loss": 0.9891,
+      "step": 11809
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001011640459881916,
+      "loss": 1.0293,
+      "step": 11810
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010115002195981158,
+      "loss": 0.9481,
+      "step": 11811
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010113599790881061,
+      "loss": 1.1348,
+      "step": 11812
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001011219738354646,
+      "loss": 1.0779,
+      "step": 11813
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001011079497400494,
+      "loss": 1.1428,
+      "step": 11814
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010109392562284088,
+      "loss": 1.0517,
+      "step": 11815
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010107990148411484,
+      "loss": 0.9266,
+      "step": 11816
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001010658773241472,
+      "loss": 1.179,
+      "step": 11817
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001010518531432138,
+      "loss": 0.9346,
+      "step": 11818
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010103782894159043,
+      "loss": 1.0524,
+      "step": 11819
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010102380471955302,
+      "loss": 0.891,
+      "step": 11820
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010100978047737741,
+      "loss": 0.838,
+      "step": 11821
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010099575621533946,
+      "loss": 1.1178,
+      "step": 11822
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010098173193371499,
+      "loss": 0.9578,
+      "step": 11823
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010096770763277987,
+      "loss": 1.1871,
+      "step": 11824
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010095368331280999,
+      "loss": 1.0299,
+      "step": 11825
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010093965897408117,
+      "loss": 0.9649,
+      "step": 11826
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010092563461686931,
+      "loss": 1.0789,
+      "step": 11827
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010091161024145024,
+      "loss": 1.0175,
+      "step": 11828
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010089758584809979,
+      "loss": 0.9947,
+      "step": 11829
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001008835614370939,
+      "loss": 1.0253,
+      "step": 11830
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010086953700870836,
+      "loss": 0.9816,
+      "step": 11831
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010085551256321904,
+      "loss": 1.0085,
+      "step": 11832
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010084148810090184,
+      "loss": 0.8181,
+      "step": 11833
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010082746362203259,
+      "loss": 1.1047,
+      "step": 11834
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010081343912688714,
+      "loss": 0.8274,
+      "step": 11835
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010079941461574138,
+      "loss": 0.7894,
+      "step": 11836
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010078539008887114,
+      "loss": 1.283,
+      "step": 11837
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010077136554655233,
+      "loss": 1.0298,
+      "step": 11838
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010075734098906073,
+      "loss": 1.0672,
+      "step": 11839
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010074331641667232,
+      "loss": 1.01,
+      "step": 11840
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010072929182966287,
+      "loss": 1.0184,
+      "step": 11841
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010071526722830826,
+      "loss": 1.1144,
+      "step": 11842
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010070124261288436,
+      "loss": 1.0514,
+      "step": 11843
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010068721798366707,
+      "loss": 0.9962,
+      "step": 11844
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010067319334093219,
+      "loss": 0.9444,
+      "step": 11845
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010065916868495564,
+      "loss": 0.9743,
+      "step": 11846
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010064514401601324,
+      "loss": 0.9796,
+      "step": 11847
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001006311193343809,
+      "loss": 0.8704,
+      "step": 11848
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010061709464033444,
+      "loss": 1.0512,
+      "step": 11849
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010060306993414977,
+      "loss": 0.9103,
+      "step": 11850
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010058904521610273,
+      "loss": 1.1089,
+      "step": 11851
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010057502048646913,
+      "loss": 0.9935,
+      "step": 11852
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010056099574552497,
+      "loss": 1.0209,
+      "step": 11853
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000100546970993546,
+      "loss": 0.932,
+      "step": 11854
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010053294623080815,
+      "loss": 1.1637,
+      "step": 11855
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010051892145758724,
+      "loss": 1.0905,
+      "step": 11856
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010050489667415918,
+      "loss": 1.1014,
+      "step": 11857
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010049087188079983,
+      "loss": 1.0686,
+      "step": 11858
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.000100476847077785,
+      "loss": 1.0222,
+      "step": 11859
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010046282226539064,
+      "loss": 1.2039,
+      "step": 11860
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010044879744389257,
+      "loss": 0.929,
+      "step": 11861
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010043477261356664,
+      "loss": 0.9027,
+      "step": 11862
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010042074777468877,
+      "loss": 0.8725,
+      "step": 11863
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010040672292753481,
+      "loss": 1.0636,
+      "step": 11864
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001003926980723806,
+      "loss": 0.8519,
+      "step": 11865
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010037867320950205,
+      "loss": 0.7262,
+      "step": 11866
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010036464833917502,
+      "loss": 1.0084,
+      "step": 11867
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010035062346167534,
+      "loss": 0.9748,
+      "step": 11868
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010033659857727893,
+      "loss": 0.7696,
+      "step": 11869
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010032257368626161,
+      "loss": 0.9713,
+      "step": 11870
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010030854878889928,
+      "loss": 1.1807,
+      "step": 11871
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010029452388546783,
+      "loss": 1.1014,
+      "step": 11872
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010028049897624308,
+      "loss": 1.079,
+      "step": 11873
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010026647406150094,
+      "loss": 0.8765,
+      "step": 11874
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010025244914151723,
+      "loss": 0.9245,
+      "step": 11875
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001002384242165679,
+      "loss": 0.7885,
+      "step": 11876
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010022439928692875,
+      "loss": 0.9693,
+      "step": 11877
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010021037435287567,
+      "loss": 1.0028,
+      "step": 11878
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010019634941468452,
+      "loss": 1.0991,
+      "step": 11879
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010018232447263123,
+      "loss": 1.1341,
+      "step": 11880
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010016829952699158,
+      "loss": 0.932,
+      "step": 11881
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001001542745780415,
+      "loss": 1.1673,
+      "step": 11882
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010014024962605682,
+      "loss": 0.9508,
+      "step": 11883
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010012622467131348,
+      "loss": 0.7924,
+      "step": 11884
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010011219971408725,
+      "loss": 0.9153,
+      "step": 11885
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010009817475465408,
+      "loss": 0.8786,
+      "step": 11886
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010008414979328982,
+      "loss": 0.915,
+      "step": 11887
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010007012483027036,
+      "loss": 0.7436,
+      "step": 11888
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010005609986587148,
+      "loss": 1.0682,
+      "step": 11889
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001000420749003692,
+      "loss": 1.2897,
+      "step": 11890
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010002804993403924,
+      "loss": 0.9794,
+      "step": 11891
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00010001402496715757,
+      "loss": 1.1552,
+      "step": 11892
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.0001,
+      "loss": 0.8767,
+      "step": 11893
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998597503284246e-05,
+      "loss": 0.8785,
+      "step": 11894
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.997195006596081e-05,
+      "loss": 1.0474,
+      "step": 11895
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.995792509963084e-05,
+      "loss": 1.1721,
+      "step": 11896
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99439001341285e-05,
+      "loss": 1.1557,
+      "step": 11897
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99298751697297e-05,
+      "loss": 0.8991,
+      "step": 11898
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99158502067102e-05,
+      "loss": 0.909,
+      "step": 11899
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990182524534594e-05,
+      "loss": 0.984,
+      "step": 11900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988780028591278e-05,
+      "loss": 0.9415,
+      "step": 11901
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.987377532868655e-05,
+      "loss": 1.4615,
+      "step": 11902
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.985975037394318e-05,
+      "loss": 1.0249,
+      "step": 11903
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.984572542195854e-05,
+      "loss": 1.0913,
+      "step": 11904
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.983170047300845e-05,
+      "loss": 1.0843,
+      "step": 11905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.981767552736882e-05,
+      "loss": 0.8826,
+      "step": 11906
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98036505853155e-05,
+      "loss": 1.0007,
+      "step": 11907
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.978962564712433e-05,
+      "loss": 1.2325,
+      "step": 11908
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.977560071307129e-05,
+      "loss": 1.0297,
+      "step": 11909
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.976157578343213e-05,
+      "loss": 1.082,
+      "step": 11910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.974755085848276e-05,
+      "loss": 1.0374,
+      "step": 11911
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.973352593849911e-05,
+      "loss": 1.2426,
+      "step": 11912
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.971950102375695e-05,
+      "loss": 1.0125,
+      "step": 11913
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.970547611453218e-05,
+      "loss": 0.943,
+      "step": 11914
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.969145121110073e-05,
+      "loss": 1.0516,
+      "step": 11915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.967742631373841e-05,
+      "loss": 1.0152,
+      "step": 11916
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96634014227211e-05,
+      "loss": 0.936,
+      "step": 11917
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.964937653832468e-05,
+      "loss": 1.0482,
+      "step": 11918
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9635351660825e-05,
+      "loss": 1.1702,
+      "step": 11919
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.962132679049794e-05,
+      "loss": 1.081,
+      "step": 11920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.960730192761942e-05,
+      "loss": 1.0943,
+      "step": 11921
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.959327707246522e-05,
+      "loss": 1.0577,
+      "step": 11922
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.957925222531124e-05,
+      "loss": 0.8626,
+      "step": 11923
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.956522738643339e-05,
+      "loss": 0.9924,
+      "step": 11924
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.955120255610746e-05,
+      "loss": 0.8218,
+      "step": 11925
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953717773460943e-05,
+      "loss": 1.0563,
+      "step": 11926
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.952315292221503e-05,
+      "loss": 0.9974,
+      "step": 11927
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.950912811920021e-05,
+      "loss": 0.8299,
+      "step": 11928
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.949510332584085e-05,
+      "loss": 0.8292,
+      "step": 11929
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.948107854241279e-05,
+      "loss": 1.0529,
+      "step": 11930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.946705376919186e-05,
+      "loss": 1.1264,
+      "step": 11931
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.945302900645402e-05,
+      "loss": 1.0883,
+      "step": 11932
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.943900425447505e-05,
+      "loss": 1.0864,
+      "step": 11933
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.942497951353086e-05,
+      "loss": 0.8734,
+      "step": 11934
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.941095478389733e-05,
+      "loss": 1.0182,
+      "step": 11935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.939693006585027e-05,
+      "loss": 0.9366,
+      "step": 11936
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.938290535966555e-05,
+      "loss": 1.1092,
+      "step": 11937
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936888066561913e-05,
+      "loss": 0.9553,
+      "step": 11938
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.935485598398677e-05,
+      "loss": 1.033,
+      "step": 11939
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.934083131504438e-05,
+      "loss": 1.034,
+      "step": 11940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.932680665906783e-05,
+      "loss": 0.9213,
+      "step": 11941
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.931278201633296e-05,
+      "loss": 0.984,
+      "step": 11942
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.929875738711564e-05,
+      "loss": 1.0407,
+      "step": 11943
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.928473277169177e-05,
+      "loss": 0.9438,
+      "step": 11944
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.927070817033717e-05,
+      "loss": 1.1006,
+      "step": 11945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.92566835833277e-05,
+      "loss": 0.8497,
+      "step": 11946
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.92426590109393e-05,
+      "loss": 1.0852,
+      "step": 11947
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.922863445344768e-05,
+      "loss": 0.8283,
+      "step": 11948
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921460991112891e-05,
+      "loss": 0.9835,
+      "step": 11949
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.920058538425864e-05,
+      "loss": 1.0072,
+      "step": 11950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918656087311288e-05,
+      "loss": 1.0361,
+      "step": 11951
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.917253637796745e-05,
+      "loss": 1.0361,
+      "step": 11952
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91585118990982e-05,
+      "loss": 0.9399,
+      "step": 11953
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.914448743678096e-05,
+      "loss": 0.861,
+      "step": 11954
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.913046299129168e-05,
+      "loss": 1.0725,
+      "step": 11955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.911643856290612e-05,
+      "loss": 0.8839,
+      "step": 11956
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.910241415190021e-05,
+      "loss": 1.0286,
+      "step": 11957
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90883897585498e-05,
+      "loss": 0.971,
+      "step": 11958
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907436538313073e-05,
+      "loss": 1.1433,
+      "step": 11959
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.906034102591883e-05,
+      "loss": 1.015,
+      "step": 11960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.904631668719005e-05,
+      "loss": 1.082,
+      "step": 11961
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.903229236722015e-05,
+      "loss": 0.9429,
+      "step": 11962
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.901826806628505e-05,
+      "loss": 1.1079,
+      "step": 11963
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.90042437846606e-05,
+      "loss": 0.9632,
+      "step": 11964
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89902195226226e-05,
+      "loss": 1.1207,
+      "step": 11965
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.897619528044699e-05,
+      "loss": 0.9224,
+      "step": 11966
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.896217105840959e-05,
+      "loss": 1.0582,
+      "step": 11967
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.894814685678624e-05,
+      "loss": 0.7663,
+      "step": 11968
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.893412267585282e-05,
+      "loss": 1.2834,
+      "step": 11969
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892009851588517e-05,
+      "loss": 0.8998,
+      "step": 11970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.890607437715913e-05,
+      "loss": 0.7503,
+      "step": 11971
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.889205025995064e-05,
+      "loss": 1.144,
+      "step": 11972
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.887802616453542e-05,
+      "loss": 0.8659,
+      "step": 11973
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.886400209118941e-05,
+      "loss": 0.9196,
+      "step": 11974
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.884997804018847e-05,
+      "loss": 0.9908,
+      "step": 11975
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.88359540118084e-05,
+      "loss": 0.9989,
+      "step": 11976
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.882193000632506e-05,
+      "loss": 0.9559,
+      "step": 11977
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.880790602401439e-05,
+      "loss": 1.0926,
+      "step": 11978
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.879388206515212e-05,
+      "loss": 0.9989,
+      "step": 11979
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.877985813001417e-05,
+      "loss": 0.9994,
+      "step": 11980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87658342188764e-05,
+      "loss": 0.8822,
+      "step": 11981
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87518103320146e-05,
+      "loss": 0.9818,
+      "step": 11982
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.873778646970464e-05,
+      "loss": 1.1293,
+      "step": 11983
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.872376263222241e-05,
+      "loss": 0.9631,
+      "step": 11984
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.870973881984373e-05,
+      "loss": 1.1721,
+      "step": 11985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.869571503284445e-05,
+      "loss": 1.1449,
+      "step": 11986
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868169127150044e-05,
+      "loss": 1.0613,
+      "step": 11987
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866766753608746e-05,
+      "loss": 1.1825,
+      "step": 11988
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.865364382688145e-05,
+      "loss": 1.2275,
+      "step": 11989
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.863962014415825e-05,
+      "loss": 1.005,
+      "step": 11990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.862559648819366e-05,
+      "loss": 0.8983,
+      "step": 11991
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.861157285926357e-05,
+      "loss": 0.8612,
+      "step": 11992
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.859754925764378e-05,
+      "loss": 1.0048,
+      "step": 11993
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.858352568361013e-05,
+      "loss": 1.119,
+      "step": 11994
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.856950213743855e-05,
+      "loss": 0.8799,
+      "step": 11995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.855547861940476e-05,
+      "loss": 0.9406,
+      "step": 11996
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.85414551297847e-05,
+      "loss": 0.93,
+      "step": 11997
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.852743166885417e-05,
+      "loss": 1.0763,
+      "step": 11998
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8513408236889e-05,
+      "loss": 1.0613,
+      "step": 11999
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849938483416503e-05,
+      "loss": 1.1215,
+      "step": 12000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.848536146095817e-05,
+      "loss": 1.0286,
+      "step": 12001
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.847133811754416e-05,
+      "loss": 1.0283,
+      "step": 12002
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.845731480419891e-05,
+      "loss": 1.0114,
+      "step": 12003
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.844329152119823e-05,
+      "loss": 0.8131,
+      "step": 12004
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.842926826881796e-05,
+      "loss": 1.1518,
+      "step": 12005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841524504733389e-05,
+      "loss": 1.1663,
+      "step": 12006
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.840122185702197e-05,
+      "loss": 1.1584,
+      "step": 12007
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838719869815794e-05,
+      "loss": 0.906,
+      "step": 12008
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.837317557101765e-05,
+      "loss": 1.0396,
+      "step": 12009
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.835915247587697e-05,
+      "loss": 1.2936,
+      "step": 12010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83451294130117e-05,
+      "loss": 0.8528,
+      "step": 12011
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.833110638269769e-05,
+      "loss": 0.9159,
+      "step": 12012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.831708338521078e-05,
+      "loss": 1.0258,
+      "step": 12013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830306042082677e-05,
+      "loss": 1.1036,
+      "step": 12014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.828903748982154e-05,
+      "loss": 0.935,
+      "step": 12015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.827501459247087e-05,
+      "loss": 0.9866,
+      "step": 12016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.826099172905059e-05,
+      "loss": 1.1775,
+      "step": 12017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.824696889983662e-05,
+      "loss": 0.8964,
+      "step": 12018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823294610510465e-05,
+      "loss": 0.9605,
+      "step": 12019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82189233451306e-05,
+      "loss": 0.9689,
+      "step": 12020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820490062019029e-05,
+      "loss": 1.2396,
+      "step": 12021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.819087793055953e-05,
+      "loss": 1.0544,
+      "step": 12022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.817685527651411e-05,
+      "loss": 0.9529,
+      "step": 12023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.816283265832994e-05,
+      "loss": 1.0009,
+      "step": 12024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.814881007628278e-05,
+      "loss": 1.1607,
+      "step": 12025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.813478753064846e-05,
+      "loss": 0.8534,
+      "step": 12026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812076502170285e-05,
+      "loss": 1.1479,
+      "step": 12027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810674254972171e-05,
+      "loss": 0.9384,
+      "step": 12028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.809272011498085e-05,
+      "loss": 1.0897,
+      "step": 12029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80786977177562e-05,
+      "loss": 1.0097,
+      "step": 12030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.806467535832347e-05,
+      "loss": 1.0246,
+      "step": 12031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.805065303695852e-05,
+      "loss": 1.0489,
+      "step": 12032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.803663075393718e-05,
+      "loss": 1.161,
+      "step": 12033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.802260850953523e-05,
+      "loss": 0.868,
+      "step": 12034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.800858630402852e-05,
+      "loss": 0.9496,
+      "step": 12035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79945641376929e-05,
+      "loss": 1.0584,
+      "step": 12036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.798054201080411e-05,
+      "loss": 1.1335,
+      "step": 12037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.796651992363803e-05,
+      "loss": 0.9879,
+      "step": 12038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.795249787647041e-05,
+      "loss": 0.9776,
+      "step": 12039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.793847586957706e-05,
+      "loss": 1.1506,
+      "step": 12040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.792445390323392e-05,
+      "loss": 0.8987,
+      "step": 12041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791043197771663e-05,
+      "loss": 1.0184,
+      "step": 12042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.789641009330111e-05,
+      "loss": 0.8013,
+      "step": 12043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788238825026317e-05,
+      "loss": 1.0545,
+      "step": 12044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.786836644887856e-05,
+      "loss": 0.9772,
+      "step": 12045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.78543446894231e-05,
+      "loss": 0.8973,
+      "step": 12046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.784032297217265e-05,
+      "loss": 1.1131,
+      "step": 12047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.782630129740297e-05,
+      "loss": 1.1163,
+      "step": 12048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.781227966538988e-05,
+      "loss": 0.9999,
+      "step": 12049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77982580764092e-05,
+      "loss": 0.9213,
+      "step": 12050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.77842365307367e-05,
+      "loss": 1.0169,
+      "step": 12051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.777021502864819e-05,
+      "loss": 0.9986,
+      "step": 12052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.775619357041952e-05,
+      "loss": 0.8971,
+      "step": 12053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.774217215632643e-05,
+      "loss": 0.9898,
+      "step": 12054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.772815078664475e-05,
+      "loss": 0.904,
+      "step": 12055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.771412946165031e-05,
+      "loss": 1.0731,
+      "step": 12056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.770010818161882e-05,
+      "loss": 0.8563,
+      "step": 12057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768608694682619e-05,
+      "loss": 0.9888,
+      "step": 12058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.767206575754811e-05,
+      "loss": 0.7909,
+      "step": 12059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.765804461406044e-05,
+      "loss": 1.0108,
+      "step": 12060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.764402351663901e-05,
+      "loss": 1.1041,
+      "step": 12061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.763000246555953e-05,
+      "loss": 1.0609,
+      "step": 12062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.761598146109781e-05,
+      "loss": 1.0227,
+      "step": 12063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.760196050352972e-05,
+      "loss": 0.9962,
+      "step": 12064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.758793959313093e-05,
+      "loss": 0.977,
+      "step": 12065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.757391873017734e-05,
+      "loss": 0.8321,
+      "step": 12066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.755989791494471e-05,
+      "loss": 0.9203,
+      "step": 12067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.754587714770878e-05,
+      "loss": 0.963,
+      "step": 12068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753185642874537e-05,
+      "loss": 1.0593,
+      "step": 12069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.751783575833031e-05,
+      "loss": 1.0714,
+      "step": 12070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.750381513673933e-05,
+      "loss": 0.9151,
+      "step": 12071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.748979456424822e-05,
+      "loss": 0.8804,
+      "step": 12072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.747577404113281e-05,
+      "loss": 1.0819,
+      "step": 12073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.746175356766882e-05,
+      "loss": 0.8458,
+      "step": 12074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.744773314413205e-05,
+      "loss": 0.9039,
+      "step": 12075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.743371277079833e-05,
+      "loss": 1.0818,
+      "step": 12076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.741969244794338e-05,
+      "loss": 0.9707,
+      "step": 12077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7405672175843e-05,
+      "loss": 1.003,
+      "step": 12078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7391651954773e-05,
+      "loss": 0.9783,
+      "step": 12079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.737763178500909e-05,
+      "loss": 1.1679,
+      "step": 12080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.736361166682714e-05,
+      "loss": 1.127,
+      "step": 12081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73495916005028e-05,
+      "loss": 0.9875,
+      "step": 12082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.733557158631198e-05,
+      "loss": 1.1482,
+      "step": 12083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.732155162453035e-05,
+      "loss": 1.032,
+      "step": 12084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.730753171543374e-05,
+      "loss": 0.9627,
+      "step": 12085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729351185929787e-05,
+      "loss": 1.1796,
+      "step": 12086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72794920563986e-05,
+      "loss": 0.9979,
+      "step": 12087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.726547230701157e-05,
+      "loss": 1.1944,
+      "step": 12088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.725145261141266e-05,
+      "loss": 0.8296,
+      "step": 12089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.72374329698776e-05,
+      "loss": 0.7971,
+      "step": 12090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.722341338268214e-05,
+      "loss": 0.8719,
+      "step": 12091
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.720939385010204e-05,
+      "loss": 1.0191,
+      "step": 12092
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.719537437241312e-05,
+      "loss": 0.904,
+      "step": 12093
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.718135494989107e-05,
+      "loss": 1.1004,
+      "step": 12094
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71673355828117e-05,
+      "loss": 0.9431,
+      "step": 12095
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.715331627145077e-05,
+      "loss": 1.0534,
+      "step": 12096
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7139297016084e-05,
+      "loss": 1.1513,
+      "step": 12097
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.712527781698716e-05,
+      "loss": 1.0139,
+      "step": 12098
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.711125867443605e-05,
+      "loss": 0.9789,
+      "step": 12099
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.709723958870639e-05,
+      "loss": 0.9888,
+      "step": 12100
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.708322056007397e-05,
+      "loss": 1.0638,
+      "step": 12101
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.706920158881448e-05,
+      "loss": 1.0678,
+      "step": 12102
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.705518267520368e-05,
+      "loss": 0.9023,
+      "step": 12103
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.704116381951742e-05,
+      "loss": 0.9847,
+      "step": 12104
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.702714502203132e-05,
+      "loss": 1.0051,
+      "step": 12105
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70131262830212e-05,
+      "loss": 1.0358,
+      "step": 12106
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.699910760276284e-05,
+      "loss": 1.1553,
+      "step": 12107
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.698508898153191e-05,
+      "loss": 0.9598,
+      "step": 12108
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.697107041960417e-05,
+      "loss": 1.0743,
+      "step": 12109
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.695705191725545e-05,
+      "loss": 1.142,
+      "step": 12110
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.694303347476137e-05,
+      "loss": 1.1003,
+      "step": 12111
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.692901509239774e-05,
+      "loss": 1.0921,
+      "step": 12112
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.691499677044033e-05,
+      "loss": 0.9688,
+      "step": 12113
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69009785091648e-05,
+      "loss": 0.9681,
+      "step": 12114
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.688696030884694e-05,
+      "loss": 0.9829,
+      "step": 12115
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68729421697625e-05,
+      "loss": 0.8576,
+      "step": 12116
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.685892409218717e-05,
+      "loss": 0.7304,
+      "step": 12117
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.684490607639673e-05,
+      "loss": 0.8606,
+      "step": 12118
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68308881226669e-05,
+      "loss": 0.9028,
+      "step": 12119
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68168702312734e-05,
+      "loss": 0.9258,
+      "step": 12120
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.680285240249193e-05,
+      "loss": 1.124,
+      "step": 12121
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.67888346365983e-05,
+      "loss": 0.95,
+      "step": 12122
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.67748169338682e-05,
+      "loss": 1.0251,
+      "step": 12123
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.676079929457736e-05,
+      "loss": 1.1581,
+      "step": 12124
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.674678171900147e-05,
+      "loss": 0.8965,
+      "step": 12125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673276420741628e-05,
+      "loss": 0.9156,
+      "step": 12126
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.67187467600976e-05,
+      "loss": 0.8849,
+      "step": 12127
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.670472937732099e-05,
+      "loss": 1.1834,
+      "step": 12128
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.669071205936229e-05,
+      "loss": 0.9847,
+      "step": 12129
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.667669480649719e-05,
+      "loss": 0.9088,
+      "step": 12130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.66626776190014e-05,
+      "loss": 0.9328,
+      "step": 12131
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.664866049715061e-05,
+      "loss": 1.1192,
+      "step": 12132
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.663464344122064e-05,
+      "loss": 1.0407,
+      "step": 12133
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662062645148707e-05,
+      "loss": 0.7136,
+      "step": 12134
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.66066095282257e-05,
+      "loss": 0.7669,
+      "step": 12135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.659259267171223e-05,
+      "loss": 1.0596,
+      "step": 12136
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.657857588222236e-05,
+      "loss": 0.6987,
+      "step": 12137
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.656455916003176e-05,
+      "loss": 1.035,
+      "step": 12138
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.655054250541624e-05,
+      "loss": 0.8595,
+      "step": 12139
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.653652591865141e-05,
+      "loss": 0.9641,
+      "step": 12140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652250940001302e-05,
+      "loss": 0.9369,
+      "step": 12141
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.65084929497768e-05,
+      "loss": 1.1927,
+      "step": 12142
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.649447656821839e-05,
+      "loss": 0.8757,
+      "step": 12143
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.648046025561349e-05,
+      "loss": 1.0324,
+      "step": 12144
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.646644401223789e-05,
+      "loss": 0.7821,
+      "step": 12145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.645242783836721e-05,
+      "loss": 1.0649,
+      "step": 12146
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.643841173427718e-05,
+      "loss": 1.0439,
+      "step": 12147
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.642439570024347e-05,
+      "loss": 0.9615,
+      "step": 12148
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641037973654178e-05,
+      "loss": 0.8946,
+      "step": 12149
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.639636384344787e-05,
+      "loss": 1.0258,
+      "step": 12150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.638234802123732e-05,
+      "loss": 1.0122,
+      "step": 12151
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.636833227018588e-05,
+      "loss": 0.9853,
+      "step": 12152
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.635431659056928e-05,
+      "loss": 1.2594,
+      "step": 12153
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634030098266313e-05,
+      "loss": 1.0904,
+      "step": 12154
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.632628544674312e-05,
+      "loss": 1.0413,
+      "step": 12155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631226998308505e-05,
+      "loss": 0.9013,
+      "step": 12156
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.629825459196446e-05,
+      "loss": 0.8457,
+      "step": 12157
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.628423927365711e-05,
+      "loss": 0.9502,
+      "step": 12158
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.627022402843867e-05,
+      "loss": 0.8073,
+      "step": 12159
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.625620885658481e-05,
+      "loss": 1.0554,
+      "step": 12160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.624219375837118e-05,
+      "loss": 0.9296,
+      "step": 12161
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.622817873407352e-05,
+      "loss": 1.0059,
+      "step": 12162
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.621416378396746e-05,
+      "loss": 0.8504,
+      "step": 12163
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.620014890832871e-05,
+      "loss": 0.9992,
+      "step": 12164
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.618613410743291e-05,
+      "loss": 1.0808,
+      "step": 12165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.617211938155573e-05,
+      "loss": 1.0764,
+      "step": 12166
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.615810473097287e-05,
+      "loss": 1.2064,
+      "step": 12167
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.614409015595995e-05,
+      "loss": 0.9765,
+      "step": 12168
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.613007565679266e-05,
+      "loss": 1.1285,
+      "step": 12169
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.61160612337467e-05,
+      "loss": 0.9673,
+      "step": 12170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610204688709768e-05,
+      "loss": 0.9639,
+      "step": 12171
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.608803261712125e-05,
+      "loss": 0.9317,
+      "step": 12172
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.607401842409317e-05,
+      "loss": 1.0124,
+      "step": 12173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.606000430828898e-05,
+      "loss": 1.0558,
+      "step": 12174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60459902699844e-05,
+      "loss": 0.8973,
+      "step": 12175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.603197630945509e-05,
+      "loss": 1.0292,
+      "step": 12176
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.601796242697668e-05,
+      "loss": 1.0168,
+      "step": 12177
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60039486228248e-05,
+      "loss": 1.0103,
+      "step": 12178
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.598993489727522e-05,
+      "loss": 0.8878,
+      "step": 12179
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59759212506034e-05,
+      "loss": 0.8431,
+      "step": 12180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.596190768308515e-05,
+      "loss": 0.9167,
+      "step": 12181
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.594789419499605e-05,
+      "loss": 1.1191,
+      "step": 12182
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.593388078661174e-05,
+      "loss": 0.996,
+      "step": 12183
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591986745820786e-05,
+      "loss": 0.9175,
+      "step": 12184
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.590585421006009e-05,
+      "loss": 0.9521,
+      "step": 12185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.589184104244404e-05,
+      "loss": 0.9969,
+      "step": 12186
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.587782795563536e-05,
+      "loss": 1.0236,
+      "step": 12187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58638149499097e-05,
+      "loss": 1.0592,
+      "step": 12188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.584980202554266e-05,
+      "loss": 1.091,
+      "step": 12189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58357891828099e-05,
+      "loss": 1.2225,
+      "step": 12190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.582177642198701e-05,
+      "loss": 1.3067,
+      "step": 12191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580776374334966e-05,
+      "loss": 0.9388,
+      "step": 12192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.579375114717351e-05,
+      "loss": 0.999,
+      "step": 12193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.577973863373414e-05,
+      "loss": 0.9106,
+      "step": 12194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.576572620330715e-05,
+      "loss": 1.0563,
+      "step": 12195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.575171385616827e-05,
+      "loss": 0.9203,
+      "step": 12196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.573770159259299e-05,
+      "loss": 0.9315,
+      "step": 12197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.572368941285701e-05,
+      "loss": 1.0576,
+      "step": 12198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.570967731723596e-05,
+      "loss": 0.9429,
+      "step": 12199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56956653060054e-05,
+      "loss": 1.1154,
+      "step": 12200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.568165337944096e-05,
+      "loss": 0.722,
+      "step": 12201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.566764153781833e-05,
+      "loss": 1.0717,
+      "step": 12202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.565362978141301e-05,
+      "loss": 1.0243,
+      "step": 12203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.563961811050067e-05,
+      "loss": 1.0533,
+      "step": 12204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.562560652535695e-05,
+      "loss": 0.9849,
+      "step": 12205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56115950262574e-05,
+      "loss": 1.0336,
+      "step": 12206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.559758361347761e-05,
+      "loss": 1.0062,
+      "step": 12207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.558357228729327e-05,
+      "loss": 0.8597,
+      "step": 12208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556956104797992e-05,
+      "loss": 1.1513,
+      "step": 12209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.555554989581316e-05,
+      "loss": 1.068,
+      "step": 12210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.554153883106863e-05,
+      "loss": 1.136,
+      "step": 12211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.552752785402188e-05,
+      "loss": 0.9773,
+      "step": 12212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.551351696494854e-05,
+      "loss": 0.8397,
+      "step": 12213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.549950616412415e-05,
+      "loss": 0.9923,
+      "step": 12214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.548549545182435e-05,
+      "loss": 1.0733,
+      "step": 12215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.547148482832477e-05,
+      "loss": 0.8813,
+      "step": 12216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.545747429390091e-05,
+      "loss": 0.9094,
+      "step": 12217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.544346384882838e-05,
+      "loss": 0.9726,
+      "step": 12218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.542945349338283e-05,
+      "loss": 0.8514,
+      "step": 12219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.541544322783975e-05,
+      "loss": 0.9523,
+      "step": 12220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.540143305247478e-05,
+      "loss": 0.7248,
+      "step": 12221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.538742296756352e-05,
+      "loss": 0.9578,
+      "step": 12222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537341297338148e-05,
+      "loss": 0.847,
+      "step": 12223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535940307020424e-05,
+      "loss": 0.9801,
+      "step": 12224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.534539325830749e-05,
+      "loss": 1.1719,
+      "step": 12225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.533138353796664e-05,
+      "loss": 1.095,
+      "step": 12226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.531737390945736e-05,
+      "loss": 0.8834,
+      "step": 12227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.530336437305523e-05,
+      "loss": 1.0283,
+      "step": 12228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.528935492903575e-05,
+      "loss": 1.0211,
+      "step": 12229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.527534557767451e-05,
+      "loss": 1.0804,
+      "step": 12230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52613363192471e-05,
+      "loss": 1.17,
+      "step": 12231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.524732715402906e-05,
+      "loss": 1.3095,
+      "step": 12232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.523331808229599e-05,
+      "loss": 0.9161,
+      "step": 12233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.521930910432338e-05,
+      "loss": 0.9047,
+      "step": 12234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.520530022038681e-05,
+      "loss": 1.1044,
+      "step": 12235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519129143076186e-05,
+      "loss": 1.0691,
+      "step": 12236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.517728273572403e-05,
+      "loss": 0.9308,
+      "step": 12237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.516327413554895e-05,
+      "loss": 1.2205,
+      "step": 12238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.514926563051211e-05,
+      "loss": 1.0093,
+      "step": 12239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.513525722088907e-05,
+      "loss": 1.009,
+      "step": 12240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.512124890695536e-05,
+      "loss": 0.9961,
+      "step": 12241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51072406889866e-05,
+      "loss": 0.956,
+      "step": 12242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509323256725821e-05,
+      "loss": 1.1177,
+      "step": 12243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50792245420458e-05,
+      "loss": 1.1239,
+      "step": 12244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.506521661362493e-05,
+      "loss": 0.9799,
+      "step": 12245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.505120878227108e-05,
+      "loss": 1.016,
+      "step": 12246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.503720104825978e-05,
+      "loss": 0.931,
+      "step": 12247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502319341186666e-05,
+      "loss": 1.0125,
+      "step": 12248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50091858733671e-05,
+      "loss": 1.1195,
+      "step": 12249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.499517843303675e-05,
+      "loss": 0.9983,
+      "step": 12250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49811710911511e-05,
+      "loss": 0.7651,
+      "step": 12251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.496716384798564e-05,
+      "loss": 1.0649,
+      "step": 12252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.495315670381591e-05,
+      "loss": 1.058,
+      "step": 12253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.493914965891748e-05,
+      "loss": 1.0257,
+      "step": 12254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49251427135658e-05,
+      "loss": 0.9982,
+      "step": 12255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491113586803644e-05,
+      "loss": 1.0558,
+      "step": 12256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.489712912260486e-05,
+      "loss": 0.9376,
+      "step": 12257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.48831224775466e-05,
+      "loss": 0.8902,
+      "step": 12258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.486911593313718e-05,
+      "loss": 1.0926,
+      "step": 12259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485510948965208e-05,
+      "loss": 0.7957,
+      "step": 12260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.484110314736682e-05,
+      "loss": 0.9485,
+      "step": 12261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.482709690655696e-05,
+      "loss": 1.0215,
+      "step": 12262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.481309076749792e-05,
+      "loss": 1.0444,
+      "step": 12263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47990847304652e-05,
+      "loss": 0.9338,
+      "step": 12264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.478507879573439e-05,
+      "loss": 0.9257,
+      "step": 12265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.477107296358087e-05,
+      "loss": 0.829,
+      "step": 12266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.475706723428021e-05,
+      "loss": 0.964,
+      "step": 12267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474306160810791e-05,
+      "loss": 0.9704,
+      "step": 12268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.472905608533941e-05,
+      "loss": 0.8174,
+      "step": 12269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.47150506662502e-05,
+      "loss": 1.1687,
+      "step": 12270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.470104535111583e-05,
+      "loss": 0.9406,
+      "step": 12271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.468704014021171e-05,
+      "loss": 1.1603,
+      "step": 12272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.467303503381335e-05,
+      "loss": 0.8069,
+      "step": 12273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465903003219627e-05,
+      "loss": 1.1624,
+      "step": 12274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46450251356359e-05,
+      "loss": 0.973,
+      "step": 12275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463102034440773e-05,
+      "loss": 0.9807,
+      "step": 12276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.461701565878719e-05,
+      "loss": 0.8368,
+      "step": 12277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46030110790498e-05,
+      "loss": 1.0031,
+      "step": 12278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458900660547108e-05,
+      "loss": 0.9543,
+      "step": 12279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45750022383264e-05,
+      "loss": 1.0214,
+      "step": 12280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.456099797789126e-05,
+      "loss": 1.2415,
+      "step": 12281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454699382444117e-05,
+      "loss": 1.0281,
+      "step": 12282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453298977825149e-05,
+      "loss": 0.8983,
+      "step": 12283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.451898583959776e-05,
+      "loss": 1.1543,
+      "step": 12284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.450498200875546e-05,
+      "loss": 0.8769,
+      "step": 12285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449097828599995e-05,
+      "loss": 0.7842,
+      "step": 12286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.447697467160672e-05,
+      "loss": 0.9591,
+      "step": 12287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.446297116585132e-05,
+      "loss": 1.1037,
+      "step": 12288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.444896776900903e-05,
+      "loss": 1.095,
+      "step": 12289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44349644813554e-05,
+      "loss": 0.949,
+      "step": 12290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442096130316589e-05,
+      "loss": 0.7713,
+      "step": 12291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440695823471588e-05,
+      "loss": 1.1427,
+      "step": 12292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439295527628081e-05,
+      "loss": 0.9748,
+      "step": 12293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437895242813622e-05,
+      "loss": 0.9188,
+      "step": 12294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436494969055739e-05,
+      "loss": 1.0241,
+      "step": 12295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435094706381987e-05,
+      "loss": 0.9444,
+      "step": 12296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.433694454819905e-05,
+      "loss": 1.0567,
+      "step": 12297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.432294214397037e-05,
+      "loss": 0.9659,
+      "step": 12298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.430893985140926e-05,
+      "loss": 1.0575,
+      "step": 12299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42949376707911e-05,
+      "loss": 1.1817,
+      "step": 12300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428093560239135e-05,
+      "loss": 0.8963,
+      "step": 12301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426693364648546e-05,
+      "loss": 1.2468,
+      "step": 12302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.425293180334881e-05,
+      "loss": 0.9698,
+      "step": 12303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42389300732568e-05,
+      "loss": 1.3224,
+      "step": 12304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42249284564849e-05,
+      "loss": 1.1425,
+      "step": 12305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.421092695330844e-05,
+      "loss": 0.914,
+      "step": 12306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.419692556400289e-05,
+      "loss": 1.045,
+      "step": 12307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.418292428884368e-05,
+      "loss": 1.1698,
+      "step": 12308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.416892312810614e-05,
+      "loss": 1.1579,
+      "step": 12309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41549220820657e-05,
+      "loss": 0.8947,
+      "step": 12310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.414092115099784e-05,
+      "loss": 0.8582,
+      "step": 12311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41269203351778e-05,
+      "loss": 0.8236,
+      "step": 12312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.411291963488109e-05,
+      "loss": 1.0271,
+      "step": 12313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.409891905038311e-05,
+      "loss": 0.8885,
+      "step": 12314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.40849185819592e-05,
+      "loss": 1.0829,
+      "step": 12315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.407091822988473e-05,
+      "loss": 1.0805,
+      "step": 12316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.40569179944352e-05,
+      "loss": 1.064,
+      "step": 12317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.404291787588584e-05,
+      "loss": 1.1656,
+      "step": 12318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402891787451214e-05,
+      "loss": 1.0594,
+      "step": 12319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.401491799058948e-05,
+      "loss": 1.0824,
+      "step": 12320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400091822439318e-05,
+      "loss": 0.8174,
+      "step": 12321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.398691857619866e-05,
+      "loss": 1.189,
+      "step": 12322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.397291904628123e-05,
+      "loss": 1.0841,
+      "step": 12323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.395891963491633e-05,
+      "loss": 0.7855,
+      "step": 12324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.394492034237931e-05,
+      "loss": 1.0625,
+      "step": 12325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.393092116894552e-05,
+      "loss": 0.8105,
+      "step": 12326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.391692211489033e-05,
+      "loss": 1.0162,
+      "step": 12327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.390292318048912e-05,
+      "loss": 0.9644,
+      "step": 12328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.388892436601718e-05,
+      "loss": 0.9028,
+      "step": 12329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.387492567174996e-05,
+      "loss": 0.8186,
+      "step": 12330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.386092709796278e-05,
+      "loss": 0.9357,
+      "step": 12331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.384692864493095e-05,
+      "loss": 1.0081,
+      "step": 12332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.383293031292985e-05,
+      "loss": 1.0636,
+      "step": 12333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.381893210223489e-05,
+      "loss": 1.1554,
+      "step": 12334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.380493401312128e-05,
+      "loss": 0.7459,
+      "step": 12335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.379093604586446e-05,
+      "loss": 1.1398,
+      "step": 12336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.377693820073977e-05,
+      "loss": 1.0682,
+      "step": 12337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37629404780225e-05,
+      "loss": 0.9804,
+      "step": 12338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.374894287798798e-05,
+      "loss": 0.824,
+      "step": 12339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.373494540091163e-05,
+      "loss": 0.9362,
+      "step": 12340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.372094804706867e-05,
+      "loss": 1.0791,
+      "step": 12341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.370695081673451e-05,
+      "loss": 0.7784,
+      "step": 12342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.369295371018442e-05,
+      "loss": 1.0012,
+      "step": 12343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.367895672769374e-05,
+      "loss": 1.0296,
+      "step": 12344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.366495986953782e-05,
+      "loss": 0.8819,
+      "step": 12345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.365096313599191e-05,
+      "loss": 0.9966,
+      "step": 12346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.363696652733138e-05,
+      "loss": 0.9712,
+      "step": 12347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.362297004383158e-05,
+      "loss": 1.2055,
+      "step": 12348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.360897368576772e-05,
+      "loss": 1.2054,
+      "step": 12349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.359497745341516e-05,
+      "loss": 1.1391,
+      "step": 12350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.358098134704924e-05,
+      "loss": 0.9838,
+      "step": 12351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.356698536694518e-05,
+      "loss": 1.1877,
+      "step": 12352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.355298951337836e-05,
+      "loss": 0.7845,
+      "step": 12353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.353899378662405e-05,
+      "loss": 0.9836,
+      "step": 12354
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.352499818695752e-05,
+      "loss": 0.9729,
+      "step": 12355
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.351100271465407e-05,
+      "loss": 0.8742,
+      "step": 12356
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.349700736998908e-05,
+      "loss": 1.1099,
+      "step": 12357
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.348301215323768e-05,
+      "loss": 0.9695,
+      "step": 12358
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346901706467527e-05,
+      "loss": 0.9487,
+      "step": 12359
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.345502210457711e-05,
+      "loss": 0.8716,
+      "step": 12360
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.344102727321848e-05,
+      "loss": 0.9241,
+      "step": 12361
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34270325708746e-05,
+      "loss": 1.0452,
+      "step": 12362
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.341303799782087e-05,
+      "loss": 0.9539,
+      "step": 12363
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.339904355433243e-05,
+      "loss": 0.9141,
+      "step": 12364
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.338504924068462e-05,
+      "loss": 0.7465,
+      "step": 12365
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.337105505715271e-05,
+      "loss": 0.9418,
+      "step": 12366
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.335706100401194e-05,
+      "loss": 0.9615,
+      "step": 12367
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.334306708153761e-05,
+      "loss": 1.0935,
+      "step": 12368
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.332907329000491e-05,
+      "loss": 1.0426,
+      "step": 12369
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.331507962968915e-05,
+      "loss": 0.976,
+      "step": 12370
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.330108610086561e-05,
+      "loss": 1.1519,
+      "step": 12371
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.328709270380949e-05,
+      "loss": 0.9384,
+      "step": 12372
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.327309943879604e-05,
+      "loss": 0.893,
+      "step": 12373
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325910630610057e-05,
+      "loss": 0.9751,
+      "step": 12374
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.324511330599822e-05,
+      "loss": 0.8978,
+      "step": 12375
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.323112043876432e-05,
+      "loss": 0.9508,
+      "step": 12376
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32171277046741e-05,
+      "loss": 0.9689,
+      "step": 12377
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320313510400276e-05,
+      "loss": 0.9149,
+      "step": 12378
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.318914263702553e-05,
+      "loss": 0.851,
+      "step": 12379
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.317515030401773e-05,
+      "loss": 1.0726,
+      "step": 12380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.316115810525446e-05,
+      "loss": 0.9822,
+      "step": 12381
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.314716604101101e-05,
+      "loss": 0.9294,
+      "step": 12382
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.313317411156264e-05,
+      "loss": 0.9716,
+      "step": 12383
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31191823171845e-05,
+      "loss": 0.9857,
+      "step": 12384
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.310519065815183e-05,
+      "loss": 0.9073,
+      "step": 12385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309119913473993e-05,
+      "loss": 0.9185,
+      "step": 12386
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.307720774722385e-05,
+      "loss": 0.8854,
+      "step": 12387
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.306321649587895e-05,
+      "loss": 1.0783,
+      "step": 12388
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.304922538098035e-05,
+      "loss": 1.0212,
+      "step": 12389
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.303523440280329e-05,
+      "loss": 0.9702,
+      "step": 12390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.302124356162297e-05,
+      "loss": 1.0203,
+      "step": 12391
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.300725285771454e-05,
+      "loss": 0.9468,
+      "step": 12392
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.299326229135326e-05,
+      "loss": 0.9132,
+      "step": 12393
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.297927186281434e-05,
+      "loss": 0.7533,
+      "step": 12394
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.296528157237291e-05,
+      "loss": 1.1599,
+      "step": 12395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.295129142030417e-05,
+      "loss": 0.9254,
+      "step": 12396
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.293730140688336e-05,
+      "loss": 0.8934,
+      "step": 12397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.292331153238557e-05,
+      "loss": 1.222,
+      "step": 12398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.290932179708606e-05,
+      "loss": 0.9091,
+      "step": 12399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.289533220126e-05,
+      "loss": 0.9202,
+      "step": 12400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.288134274518252e-05,
+      "loss": 0.869,
+      "step": 12401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.28673534291288e-05,
+      "loss": 0.9755,
+      "step": 12402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285336425337409e-05,
+      "loss": 0.9885,
+      "step": 12403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.283937521819343e-05,
+      "loss": 1.0153,
+      "step": 12404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.282538632386207e-05,
+      "loss": 1.0149,
+      "step": 12405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.281139757065517e-05,
+      "loss": 0.8079,
+      "step": 12406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.279740895884785e-05,
+      "loss": 0.9742,
+      "step": 12407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.27834204887153e-05,
+      "loss": 0.916,
+      "step": 12408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.276943216053263e-05,
+      "loss": 1.0598,
+      "step": 12409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2755443974575e-05,
+      "loss": 0.968,
+      "step": 12410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.274145593111763e-05,
+      "loss": 1.0955,
+      "step": 12411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272746803043556e-05,
+      "loss": 1.0617,
+      "step": 12412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2713480272804e-05,
+      "loss": 0.8577,
+      "step": 12413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.269949265849808e-05,
+      "loss": 0.7692,
+      "step": 12414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268550518779288e-05,
+      "loss": 1.2456,
+      "step": 12415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.267151786096359e-05,
+      "loss": 1.0351,
+      "step": 12416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.265753067828536e-05,
+      "loss": 1.0365,
+      "step": 12417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.264354364003327e-05,
+      "loss": 0.8829,
+      "step": 12418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.262955674648245e-05,
+      "loss": 1.0841,
+      "step": 12419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.261556999790805e-05,
+      "loss": 1.02,
+      "step": 12420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.260158339458514e-05,
+      "loss": 1.0104,
+      "step": 12421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.258759693678889e-05,
+      "loss": 1.0259,
+      "step": 12422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.257361062479439e-05,
+      "loss": 0.8044,
+      "step": 12423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.255962445887673e-05,
+      "loss": 1.0772,
+      "step": 12424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.254563843931104e-05,
+      "loss": 0.9791,
+      "step": 12425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.253165256637247e-05,
+      "loss": 0.9985,
+      "step": 12426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2517666840336e-05,
+      "loss": 1.0539,
+      "step": 12427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.250368126147684e-05,
+      "loss": 0.9777,
+      "step": 12428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.248969583007005e-05,
+      "loss": 1.1894,
+      "step": 12429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.247571054639073e-05,
+      "loss": 1.1053,
+      "step": 12430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.246172541071397e-05,
+      "loss": 0.8326,
+      "step": 12431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.244774042331481e-05,
+      "loss": 0.9799,
+      "step": 12432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.243375558446837e-05,
+      "loss": 0.9166,
+      "step": 12433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.241977089444977e-05,
+      "loss": 0.9633,
+      "step": 12434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.240578635353403e-05,
+      "loss": 0.9458,
+      "step": 12435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.239180196199624e-05,
+      "loss": 1.0749,
+      "step": 12436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.237781772011153e-05,
+      "loss": 1.212,
+      "step": 12437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236383362815485e-05,
+      "loss": 1.0719,
+      "step": 12438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.234984968640136e-05,
+      "loss": 1.1481,
+      "step": 12439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.233586589512614e-05,
+      "loss": 1.0209,
+      "step": 12440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.232188225460418e-05,
+      "loss": 1.0079,
+      "step": 12441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.230789876511056e-05,
+      "loss": 1.0957,
+      "step": 12442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.229391542692039e-05,
+      "loss": 1.1832,
+      "step": 12443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.227993224030862e-05,
+      "loss": 0.9785,
+      "step": 12444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.226594920555037e-05,
+      "loss": 0.8613,
+      "step": 12445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22519663229207e-05,
+      "loss": 1.1536,
+      "step": 12446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22379835926946e-05,
+      "loss": 1.0968,
+      "step": 12447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.222400101514712e-05,
+      "loss": 0.9584,
+      "step": 12448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.221001859055336e-05,
+      "loss": 0.9194,
+      "step": 12449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.219603631918825e-05,
+      "loss": 1.0481,
+      "step": 12450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218205420132692e-05,
+      "loss": 0.9766,
+      "step": 12451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.216807223724432e-05,
+      "loss": 1.0126,
+      "step": 12452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.215409042721552e-05,
+      "loss": 0.9585,
+      "step": 12453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214010877151554e-05,
+      "loss": 0.9095,
+      "step": 12454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.212612727041937e-05,
+      "loss": 0.9749,
+      "step": 12455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211214592420203e-05,
+      "loss": 1.0373,
+      "step": 12456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.209816473313859e-05,
+      "loss": 1.0409,
+      "step": 12457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2084183697504e-05,
+      "loss": 1.0175,
+      "step": 12458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.207020281757326e-05,
+      "loss": 0.9289,
+      "step": 12459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.205622209362142e-05,
+      "loss": 0.9782,
+      "step": 12460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.204224152592343e-05,
+      "loss": 0.9136,
+      "step": 12461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.202826111475434e-05,
+      "loss": 0.8886,
+      "step": 12462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.201428086038912e-05,
+      "loss": 1.1804,
+      "step": 12463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.200030076310275e-05,
+      "loss": 1.0899,
+      "step": 12464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19863208231702e-05,
+      "loss": 0.9286,
+      "step": 12465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.197234104086656e-05,
+      "loss": 1.0249,
+      "step": 12466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.195836141646666e-05,
+      "loss": 0.981,
+      "step": 12467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.194438195024557e-05,
+      "loss": 1.1153,
+      "step": 12468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.193040264247829e-05,
+      "loss": 0.9371,
+      "step": 12469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.191642349343972e-05,
+      "loss": 0.817,
+      "step": 12470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190244450340483e-05,
+      "loss": 0.9539,
+      "step": 12471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.188846567264869e-05,
+      "loss": 0.9565,
+      "step": 12472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.187448700144614e-05,
+      "loss": 0.9405,
+      "step": 12473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.186050849007221e-05,
+      "loss": 1.2132,
+      "step": 12474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.184653013880182e-05,
+      "loss": 0.9788,
+      "step": 12475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.183255194790995e-05,
+      "loss": 1.1935,
+      "step": 12476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.181857391767157e-05,
+      "loss": 0.9938,
+      "step": 12477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.180459604836156e-05,
+      "loss": 0.8581,
+      "step": 12478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17906183402549e-05,
+      "loss": 0.8057,
+      "step": 12479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.177664079362657e-05,
+      "loss": 1.0216,
+      "step": 12480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176266340875144e-05,
+      "loss": 0.8879,
+      "step": 12481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.174868618590449e-05,
+      "loss": 0.9059,
+      "step": 12482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.173470912536066e-05,
+      "loss": 0.934,
+      "step": 12483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.172073222739482e-05,
+      "loss": 0.945,
+      "step": 12484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.170675549228194e-05,
+      "loss": 0.9413,
+      "step": 12485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.169277892029696e-05,
+      "loss": 1.0332,
+      "step": 12486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.167880251171475e-05,
+      "loss": 1.0529,
+      "step": 12487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166482626681024e-05,
+      "loss": 1.0729,
+      "step": 12488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16508501858584e-05,
+      "loss": 0.8744,
+      "step": 12489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.163687426913405e-05,
+      "loss": 1.0463,
+      "step": 12490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.162289851691213e-05,
+      "loss": 0.9722,
+      "step": 12491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.160892292946759e-05,
+      "loss": 1.0086,
+      "step": 12492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.159494750707526e-05,
+      "loss": 0.93,
+      "step": 12493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.158097225001003e-05,
+      "loss": 1.116,
+      "step": 12494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15669971585469e-05,
+      "loss": 0.9914,
+      "step": 12495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.155302223296064e-05,
+      "loss": 1.0666,
+      "step": 12496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15390474735262e-05,
+      "loss": 0.8477,
+      "step": 12497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.152507288051845e-05,
+      "loss": 1.1135,
+      "step": 12498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.151109845421226e-05,
+      "loss": 1.0782,
+      "step": 12499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.149712419488252e-05,
+      "loss": 0.8996,
+      "step": 12500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.148315010280408e-05,
+      "loss": 1.2485,
+      "step": 12501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.14691761782518e-05,
+      "loss": 0.9588,
+      "step": 12502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.145520242150063e-05,
+      "loss": 0.9293,
+      "step": 12503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.144122883282535e-05,
+      "loss": 0.8688,
+      "step": 12504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.142725541250083e-05,
+      "loss": 1.0572,
+      "step": 12505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.141328216080198e-05,
+      "loss": 0.8184,
+      "step": 12506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.139930907800358e-05,
+      "loss": 1.0215,
+      "step": 12507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138533616438052e-05,
+      "loss": 1.2577,
+      "step": 12508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137136342020768e-05,
+      "loss": 1.0505,
+      "step": 12509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.135739084575984e-05,
+      "loss": 0.8896,
+      "step": 12510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.134341844131184e-05,
+      "loss": 0.8446,
+      "step": 12511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132944620713861e-05,
+      "loss": 0.9568,
+      "step": 12512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.131547414351486e-05,
+      "loss": 1.0374,
+      "step": 12513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.130150225071549e-05,
+      "loss": 1.0346,
+      "step": 12514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.128753052901534e-05,
+      "loss": 1.0238,
+      "step": 12515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127355897868919e-05,
+      "loss": 0.9906,
+      "step": 12516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.125958760001189e-05,
+      "loss": 1.0955,
+      "step": 12517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.124561639325822e-05,
+      "loss": 0.949,
+      "step": 12518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123164535870301e-05,
+      "loss": 0.9552,
+      "step": 12519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12176744966211e-05,
+      "loss": 0.7285,
+      "step": 12520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.120370380728727e-05,
+      "loss": 0.9487,
+      "step": 12521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.118973329097632e-05,
+      "loss": 1.0031,
+      "step": 12522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.117576294796307e-05,
+      "loss": 1.0283,
+      "step": 12523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.116179277852229e-05,
+      "loss": 1.0543,
+      "step": 12524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.114782278292876e-05,
+      "loss": 1.1067,
+      "step": 12525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.113385296145735e-05,
+      "loss": 0.8018,
+      "step": 12526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.111988331438274e-05,
+      "loss": 1.0514,
+      "step": 12527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.110591384197979e-05,
+      "loss": 1.0606,
+      "step": 12528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.109194454452327e-05,
+      "loss": 1.0318,
+      "step": 12529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.107797542228789e-05,
+      "loss": 1.0273,
+      "step": 12530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.106400647554849e-05,
+      "loss": 0.8392,
+      "step": 12531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.105003770457983e-05,
+      "loss": 0.9487,
+      "step": 12532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.103606910965666e-05,
+      "loss": 0.8794,
+      "step": 12533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.102210069105372e-05,
+      "loss": 1.1106,
+      "step": 12534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.100813244904585e-05,
+      "loss": 0.8439,
+      "step": 12535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09941643839077e-05,
+      "loss": 0.9282,
+      "step": 12536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.098019649591408e-05,
+      "loss": 0.9997,
+      "step": 12537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.096622878533976e-05,
+      "loss": 1.0446,
+      "step": 12538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.095226125245941e-05,
+      "loss": 1.1035,
+      "step": 12539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.093829389754786e-05,
+      "loss": 0.6865,
+      "step": 12540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.092432672087977e-05,
+      "loss": 0.8829,
+      "step": 12541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.091035972272988e-05,
+      "loss": 0.9503,
+      "step": 12542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0896392903373e-05,
+      "loss": 0.9074,
+      "step": 12543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.088242626308378e-05,
+      "loss": 1.0982,
+      "step": 12544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086845980213697e-05,
+      "loss": 0.8446,
+      "step": 12545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.085449352080728e-05,
+      "loss": 0.732,
+      "step": 12546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.084052741936944e-05,
+      "loss": 1.099,
+      "step": 12547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.082656149809811e-05,
+      "loss": 0.9327,
+      "step": 12548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08125957572681e-05,
+      "loss": 0.9939,
+      "step": 12549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.079863019715404e-05,
+      "loss": 0.9345,
+      "step": 12550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.078466481803067e-05,
+      "loss": 0.9769,
+      "step": 12551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.077069962017267e-05,
+      "loss": 0.9833,
+      "step": 12552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.075673460385468e-05,
+      "loss": 0.9164,
+      "step": 12553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07427697693515e-05,
+      "loss": 0.9318,
+      "step": 12554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.072880511693778e-05,
+      "loss": 0.9315,
+      "step": 12555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.071484064688816e-05,
+      "loss": 1.0945,
+      "step": 12556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.070087635947733e-05,
+      "loss": 1.17,
+      "step": 12557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.068691225498004e-05,
+      "loss": 0.8337,
+      "step": 12558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.067294833367086e-05,
+      "loss": 0.9679,
+      "step": 12559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065898459582453e-05,
+      "loss": 0.9899,
+      "step": 12560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.064502104171573e-05,
+      "loss": 0.9836,
+      "step": 12561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.063105767161905e-05,
+      "loss": 0.9695,
+      "step": 12562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.061709448580922e-05,
+      "loss": 0.9072,
+      "step": 12563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.060313148456084e-05,
+      "loss": 0.8225,
+      "step": 12564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.058916866814858e-05,
+      "loss": 0.9536,
+      "step": 12565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.057520603684711e-05,
+      "loss": 1.0531,
+      "step": 12566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.056124359093105e-05,
+      "loss": 1.1487,
+      "step": 12567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.054728133067505e-05,
+      "loss": 0.9177,
+      "step": 12568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.053331925635377e-05,
+      "loss": 0.9197,
+      "step": 12569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05193573682418e-05,
+      "loss": 1.0642,
+      "step": 12570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.050539566661377e-05,
+      "loss": 1.0278,
+      "step": 12571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.049143415174437e-05,
+      "loss": 1.2302,
+      "step": 12572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.047747282390815e-05,
+      "loss": 0.9604,
+      "step": 12573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.046351168337977e-05,
+      "loss": 0.9653,
+      "step": 12574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044955073043385e-05,
+      "loss": 0.9784,
+      "step": 12575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.043558996534494e-05,
+      "loss": 0.8406,
+      "step": 12576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.04216293883877e-05,
+      "loss": 1.1763,
+      "step": 12577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.040766899983677e-05,
+      "loss": 0.8361,
+      "step": 12578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.039370879996668e-05,
+      "loss": 0.9884,
+      "step": 12579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037974878905203e-05,
+      "loss": 1.0246,
+      "step": 12580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036578896736752e-05,
+      "loss": 1.1459,
+      "step": 12581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.035182933518756e-05,
+      "loss": 0.9853,
+      "step": 12582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.033786989278693e-05,
+      "loss": 1.2119,
+      "step": 12583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.032391064044002e-05,
+      "loss": 1.0882,
+      "step": 12584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.030995157842155e-05,
+      "loss": 0.9125,
+      "step": 12585
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.029599270700605e-05,
+      "loss": 0.9872,
+      "step": 12586
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028203402646806e-05,
+      "loss": 0.8929,
+      "step": 12587
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.026807553708216e-05,
+      "loss": 1.0113,
+      "step": 12588
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.025411723912298e-05,
+      "loss": 0.9994,
+      "step": 12589
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.0240159132865e-05,
+      "loss": 1.007,
+      "step": 12590
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02262012185828e-05,
+      "loss": 1.0232,
+      "step": 12591
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.021224349655095e-05,
+      "loss": 1.0849,
+      "step": 12592
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.019828596704394e-05,
+      "loss": 0.8758,
+      "step": 12593
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.018432863033638e-05,
+      "loss": 0.7562,
+      "step": 12594
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.017037148670281e-05,
+      "loss": 1.0548,
+      "step": 12595
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01564145364177e-05,
+      "loss": 0.9028,
+      "step": 12596
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.014245777975565e-05,
+      "loss": 1.1672,
+      "step": 12597
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.012850121699118e-05,
+      "loss": 0.9968,
+      "step": 12598
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.011454484839876e-05,
+      "loss": 1.0837,
+      "step": 12599
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.010058867425296e-05,
+      "loss": 1.1498,
+      "step": 12600
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.008663269482832e-05,
+      "loss": 1.1311,
+      "step": 12601
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00726769103993e-05,
+      "loss": 0.9005,
+      "step": 12602
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.005872132124042e-05,
+      "loss": 0.8929,
+      "step": 12603
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.004476592762623e-05,
+      "loss": 1.0099,
+      "step": 12604
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003081072983117e-05,
+      "loss": 0.7847,
+      "step": 12605
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.001685572812982e-05,
+      "loss": 1.0563,
+      "step": 12606
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000290092279657e-05,
+      "loss": 1.1212,
+      "step": 12607
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.998894631410595e-05,
+      "loss": 1.0388,
+      "step": 12608
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.997499190233252e-05,
+      "loss": 1.0333,
+      "step": 12609
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996103768775067e-05,
+      "loss": 0.9783,
+      "step": 12610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994708367063489e-05,
+      "loss": 1.065,
+      "step": 12611
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.993312985125972e-05,
+      "loss": 1.1427,
+      "step": 12612
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991917622989956e-05,
+      "loss": 0.9283,
+      "step": 12613
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.990522280682891e-05,
+      "loss": 1.27,
+      "step": 12614
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.989126958232226e-05,
+      "loss": 1.0543,
+      "step": 12615
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.987731655665398e-05,
+      "loss": 0.8358,
+      "step": 12616
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.986336373009862e-05,
+      "loss": 0.8521,
+      "step": 12617
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.984941110293061e-05,
+      "loss": 1.0322,
+      "step": 12618
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.983545867542437e-05,
+      "loss": 0.9487,
+      "step": 12619
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.982150644785434e-05,
+      "loss": 0.951,
+      "step": 12620
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.980755442049502e-05,
+      "loss": 0.9321,
+      "step": 12621
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.979360259362075e-05,
+      "loss": 0.9254,
+      "step": 12622
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.977965096750605e-05,
+      "loss": 1.0037,
+      "step": 12623
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.976569954242532e-05,
+      "loss": 0.8758,
+      "step": 12624
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.975174831865296e-05,
+      "loss": 1.2049,
+      "step": 12625
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.973779729646343e-05,
+      "loss": 0.9133,
+      "step": 12626
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97238464761311e-05,
+      "loss": 0.9183,
+      "step": 12627
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.970989585793039e-05,
+      "loss": 0.9968,
+      "step": 12628
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.969594544213578e-05,
+      "loss": 1.0028,
+      "step": 12629
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968199522902156e-05,
+      "loss": 0.918,
+      "step": 12630
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.966804521886222e-05,
+      "loss": 0.8638,
+      "step": 12631
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.965409541193213e-05,
+      "loss": 0.8562,
+      "step": 12632
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.964014580850567e-05,
+      "loss": 0.9841,
+      "step": 12633
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.962619640885721e-05,
+      "loss": 1.0422,
+      "step": 12634
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.96122472132612e-05,
+      "loss": 0.8311,
+      "step": 12635
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.959829822199197e-05,
+      "loss": 0.9587,
+      "step": 12636
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95843494353239e-05,
+      "loss": 1.0167,
+      "step": 12637
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.957040085353138e-05,
+      "loss": 0.9285,
+      "step": 12638
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.955645247688874e-05,
+      "loss": 1.1938,
+      "step": 12639
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.954250430567039e-05,
+      "loss": 0.8731,
+      "step": 12640
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.952855634015068e-05,
+      "loss": 0.9565,
+      "step": 12641
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.951460858060396e-05,
+      "loss": 1.0395,
+      "step": 12642
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.950066102730456e-05,
+      "loss": 0.8699,
+      "step": 12643
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.948671368052688e-05,
+      "loss": 0.9512,
+      "step": 12644
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.947276654054518e-05,
+      "loss": 1.0392,
+      "step": 12645
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.945881960763387e-05,
+      "loss": 0.9312,
+      "step": 12646
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.94448728820673e-05,
+      "loss": 0.9025,
+      "step": 12647
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943092636411973e-05,
+      "loss": 0.9455,
+      "step": 12648
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.941698005406557e-05,
+      "loss": 1.009,
+      "step": 12649
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.940303395217907e-05,
+      "loss": 1.1088,
+      "step": 12650
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938908805873455e-05,
+      "loss": 1.1718,
+      "step": 12651
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.937514237400643e-05,
+      "loss": 1.1633,
+      "step": 12652
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.936119689826888e-05,
+      "loss": 0.9761,
+      "step": 12653
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.934725163179627e-05,
+      "loss": 0.8928,
+      "step": 12654
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.933330657486297e-05,
+      "loss": 0.9926,
+      "step": 12655
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.931936172774317e-05,
+      "loss": 0.989,
+      "step": 12656
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.930541709071118e-05,
+      "loss": 0.8305,
+      "step": 12657
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.929147266404138e-05,
+      "loss": 1.2055,
+      "step": 12658
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927752844800796e-05,
+      "loss": 1.0545,
+      "step": 12659
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.926358444288526e-05,
+      "loss": 1.0339,
+      "step": 12660
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.924964064894753e-05,
+      "loss": 0.9535,
+      "step": 12661
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.923569706646902e-05,
+      "loss": 1.1221,
+      "step": 12662
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922175369572407e-05,
+      "loss": 0.8751,
+      "step": 12663
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.920781053698691e-05,
+      "loss": 1.0674,
+      "step": 12664
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.919386759053178e-05,
+      "loss": 1.0423,
+      "step": 12665
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.917992485663295e-05,
+      "loss": 0.9186,
+      "step": 12666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.91659823355647e-05,
+      "loss": 0.9417,
+      "step": 12667
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.915204002760122e-05,
+      "loss": 1.1652,
+      "step": 12668
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.913809793301681e-05,
+      "loss": 1.0734,
+      "step": 12669
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912415605208573e-05,
+      "loss": 0.948,
+      "step": 12670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.911021438508213e-05,
+      "loss": 0.8137,
+      "step": 12671
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.909627293228033e-05,
+      "loss": 0.9448,
+      "step": 12672
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90823316939545e-05,
+      "loss": 1.0269,
+      "step": 12673
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.906839067037887e-05,
+      "loss": 0.8799,
+      "step": 12674
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.905444986182771e-05,
+      "loss": 1.0974,
+      "step": 12675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.904050926857516e-05,
+      "loss": 1.197,
+      "step": 12676
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.902656889089548e-05,
+      "loss": 0.7987,
+      "step": 12677
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901262872906288e-05,
+      "loss": 0.8499,
+      "step": 12678
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.899868878335154e-05,
+      "loss": 1.0958,
+      "step": 12679
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898474905403565e-05,
+      "loss": 1.1203,
+      "step": 12680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.897080954138945e-05,
+      "loss": 1.032,
+      "step": 12681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.895687024568707e-05,
+      "loss": 1.0156,
+      "step": 12682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.894293116720274e-05,
+      "loss": 1.0042,
+      "step": 12683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.892899230621066e-05,
+      "loss": 0.8873,
+      "step": 12684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.891505366298492e-05,
+      "loss": 0.7425,
+      "step": 12685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890111523779975e-05,
+      "loss": 0.9661,
+      "step": 12686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.888717703092936e-05,
+      "loss": 0.8308,
+      "step": 12687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.887323904264785e-05,
+      "loss": 1.107,
+      "step": 12688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.885930127322938e-05,
+      "loss": 1.1192,
+      "step": 12689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.884536372294817e-05,
+      "loss": 1.0703,
+      "step": 12690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.883142639207828e-05,
+      "loss": 0.8806,
+      "step": 12691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.881748928089394e-05,
+      "loss": 0.883,
+      "step": 12692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.880355238966923e-05,
+      "loss": 0.8597,
+      "step": 12693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87896157186783e-05,
+      "loss": 1.0618,
+      "step": 12694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.877567926819534e-05,
+      "loss": 0.9992,
+      "step": 12695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.876174303849442e-05,
+      "loss": 1.0291,
+      "step": 12696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.874780702984965e-05,
+      "loss": 0.9037,
+      "step": 12697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.873387124253525e-05,
+      "loss": 0.9555,
+      "step": 12698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87199356768252e-05,
+      "loss": 0.8424,
+      "step": 12699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.870600033299372e-05,
+      "loss": 0.9538,
+      "step": 12700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869206521131489e-05,
+      "loss": 1.2945,
+      "step": 12701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867813031206278e-05,
+      "loss": 0.9382,
+      "step": 12702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86641956355115e-05,
+      "loss": 0.9568,
+      "step": 12703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86502611819352e-05,
+      "loss": 0.9418,
+      "step": 12704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86363269516079e-05,
+      "loss": 1.0632,
+      "step": 12705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.862239294480371e-05,
+      "loss": 0.9454,
+      "step": 12706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.860845916179674e-05,
+      "loss": 1.022,
+      "step": 12707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8594525602861e-05,
+      "loss": 1.0942,
+      "step": 12708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858059226827062e-05,
+      "loss": 1.1051,
+      "step": 12709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.856665915829969e-05,
+      "loss": 0.9016,
+      "step": 12710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.85527262732222e-05,
+      "loss": 0.7354,
+      "step": 12711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853879361331226e-05,
+      "loss": 0.9398,
+      "step": 12712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.852486117884392e-05,
+      "loss": 0.9348,
+      "step": 12713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.851092897009118e-05,
+      "loss": 0.9925,
+      "step": 12714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.849699698732819e-05,
+      "loss": 0.982,
+      "step": 12715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.848306523082888e-05,
+      "loss": 1.0128,
+      "step": 12716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.846913370086734e-05,
+      "loss": 1.1205,
+      "step": 12717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.845520239771762e-05,
+      "loss": 0.961,
+      "step": 12718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.844127132165373e-05,
+      "loss": 1.0039,
+      "step": 12719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.842734047294965e-05,
+      "loss": 1.232,
+      "step": 12720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.841340985187951e-05,
+      "loss": 0.9998,
+      "step": 12721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83994794587172e-05,
+      "loss": 1.0327,
+      "step": 12722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83855492937368e-05,
+      "loss": 0.9249,
+      "step": 12723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.837161935721232e-05,
+      "loss": 0.9505,
+      "step": 12724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.835768964941773e-05,
+      "loss": 1.137,
+      "step": 12725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.834376017062703e-05,
+      "loss": 0.9958,
+      "step": 12726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.832983092111424e-05,
+      "loss": 0.8409,
+      "step": 12727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.831590190115333e-05,
+      "loss": 1.023,
+      "step": 12728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.830197311101827e-05,
+      "loss": 0.9523,
+      "step": 12729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82880445509831e-05,
+      "loss": 0.9346,
+      "step": 12730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.827411622132169e-05,
+      "loss": 1.0419,
+      "step": 12731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.826018812230807e-05,
+      "loss": 1.3719,
+      "step": 12732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.824626025421626e-05,
+      "loss": 0.9011,
+      "step": 12733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.823233261732011e-05,
+      "loss": 0.9793,
+      "step": 12734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.821840521189364e-05,
+      "loss": 1.1291,
+      "step": 12735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.820447803821082e-05,
+      "loss": 1.163,
+      "step": 12736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819055109654552e-05,
+      "loss": 1.0734,
+      "step": 12737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81766243871718e-05,
+      "loss": 1.0014,
+      "step": 12738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.816269791036347e-05,
+      "loss": 0.9089,
+      "step": 12739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.814877166639453e-05,
+      "loss": 0.972,
+      "step": 12740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.813484565553893e-05,
+      "loss": 0.8851,
+      "step": 12741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.812091987807056e-05,
+      "loss": 0.9216,
+      "step": 12742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81069943342633e-05,
+      "loss": 0.9395,
+      "step": 12743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.809306902439117e-05,
+      "loss": 0.8779,
+      "step": 12744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.807914394872799e-05,
+      "loss": 0.8624,
+      "step": 12745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.806521910754768e-05,
+      "loss": 0.9137,
+      "step": 12746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80512945011242e-05,
+      "loss": 0.7839,
+      "step": 12747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.803737012973137e-05,
+      "loss": 0.9931,
+      "step": 12748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.802344599364312e-05,
+      "loss": 0.9197,
+      "step": 12749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.800952209313335e-05,
+      "loss": 0.7914,
+      "step": 12750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.799559842847592e-05,
+      "loss": 0.9522,
+      "step": 12751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.79816749999447e-05,
+      "loss": 1.0674,
+      "step": 12752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.796775180781363e-05,
+      "loss": 0.9506,
+      "step": 12753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.795382885235646e-05,
+      "loss": 0.759,
+      "step": 12754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.793990613384714e-05,
+      "loss": 1.0995,
+      "step": 12755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.792598365255953e-05,
+      "loss": 1.0702,
+      "step": 12756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791206140876746e-05,
+      "loss": 1.0781,
+      "step": 12757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.789813940274479e-05,
+      "loss": 0.8801,
+      "step": 12758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.788421763476535e-05,
+      "loss": 0.8918,
+      "step": 12759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.787029610510297e-05,
+      "loss": 1.0895,
+      "step": 12760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785637481403157e-05,
+      "loss": 1.0464,
+      "step": 12761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.784245376182484e-05,
+      "loss": 0.9038,
+      "step": 12762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.782853294875673e-05,
+      "loss": 1.0004,
+      "step": 12763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.781461237510102e-05,
+      "loss": 0.9745,
+      "step": 12764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.780069204113153e-05,
+      "loss": 0.9118,
+      "step": 12765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.778677194712202e-05,
+      "loss": 1.0763,
+      "step": 12766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.777285209334639e-05,
+      "loss": 1.0956,
+      "step": 12767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.775893248007839e-05,
+      "loss": 1.0071,
+      "step": 12768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.774501310759183e-05,
+      "loss": 1.048,
+      "step": 12769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77310939761605e-05,
+      "loss": 1.1324,
+      "step": 12770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.771717508605817e-05,
+      "loss": 1.1135,
+      "step": 12771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.770325643755863e-05,
+      "loss": 1.1226,
+      "step": 12772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.768933803093573e-05,
+      "loss": 0.9785,
+      "step": 12773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.767541986646315e-05,
+      "loss": 0.9216,
+      "step": 12774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76615019444147e-05,
+      "loss": 0.9246,
+      "step": 12775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.764758426506417e-05,
+      "loss": 0.975,
+      "step": 12776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763366682868524e-05,
+      "loss": 0.8077,
+      "step": 12777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.761974963555174e-05,
+      "loss": 0.9768,
+      "step": 12778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.760583268593743e-05,
+      "loss": 1.1285,
+      "step": 12779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.759191598011598e-05,
+      "loss": 1.0118,
+      "step": 12780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.757799951836123e-05,
+      "loss": 1.0406,
+      "step": 12781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.756408330094683e-05,
+      "loss": 1.0512,
+      "step": 12782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.755016732814651e-05,
+      "loss": 0.9227,
+      "step": 12783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75362516002341e-05,
+      "loss": 1.0036,
+      "step": 12784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75223361174832e-05,
+      "loss": 0.9871,
+      "step": 12785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.75084208801676e-05,
+      "loss": 0.829,
+      "step": 12786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.749450588856101e-05,
+      "loss": 1.0292,
+      "step": 12787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74805911429371e-05,
+      "loss": 0.8781,
+      "step": 12788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.746667664356956e-05,
+      "loss": 1.1223,
+      "step": 12789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.745276239073219e-05,
+      "loss": 1.0324,
+      "step": 12790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.743884838469856e-05,
+      "loss": 0.8581,
+      "step": 12791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.742493462574242e-05,
+      "loss": 0.9752,
+      "step": 12792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.741102111413748e-05,
+      "loss": 1.1451,
+      "step": 12793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.739710785015736e-05,
+      "loss": 0.7317,
+      "step": 12794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.738319483407573e-05,
+      "loss": 1.0176,
+      "step": 12795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.736928206616632e-05,
+      "loss": 0.8078,
+      "step": 12796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735536954670275e-05,
+      "loss": 0.7531,
+      "step": 12797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.734145727595866e-05,
+      "loss": 1.1097,
+      "step": 12798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.732754525420779e-05,
+      "loss": 0.9427,
+      "step": 12799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.731363348172368e-05,
+      "loss": 0.9894,
+      "step": 12800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.729972195878006e-05,
+      "loss": 1.0052,
+      "step": 12801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72858106856505e-05,
+      "loss": 0.891,
+      "step": 12802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.727189966260869e-05,
+      "loss": 0.9703,
+      "step": 12803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.725798888992825e-05,
+      "loss": 0.7844,
+      "step": 12804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.724407836788276e-05,
+      "loss": 0.8318,
+      "step": 12805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.723016809674586e-05,
+      "loss": 0.8941,
+      "step": 12806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.721625807679125e-05,
+      "loss": 0.9612,
+      "step": 12807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.720234830829239e-05,
+      "loss": 0.9775,
+      "step": 12808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.718843879152298e-05,
+      "loss": 0.9109,
+      "step": 12809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.717452952675665e-05,
+      "loss": 1.2145,
+      "step": 12810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71606205142669e-05,
+      "loss": 1.0702,
+      "step": 12811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.714671175432733e-05,
+      "loss": 0.8663,
+      "step": 12812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.713280324721163e-05,
+      "loss": 1.1832,
+      "step": 12813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.711889499319327e-05,
+      "loss": 0.9365,
+      "step": 12814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.710498699254588e-05,
+      "loss": 0.927,
+      "step": 12815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709107924554303e-05,
+      "loss": 1.1876,
+      "step": 12816
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.707717175245825e-05,
+      "loss": 1.0341,
+      "step": 12817
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.706326451356511e-05,
+      "loss": 0.9448,
+      "step": 12818
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.704935752913718e-05,
+      "loss": 0.9347,
+      "step": 12819
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.703545079944802e-05,
+      "loss": 0.8139,
+      "step": 12820
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.702154432477115e-05,
+      "loss": 1.0536,
+      "step": 12821
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.700763810538014e-05,
+      "loss": 0.939,
+      "step": 12822
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699373214154846e-05,
+      "loss": 0.9701,
+      "step": 12823
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.697982643354976e-05,
+      "loss": 0.8854,
+      "step": 12824
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.696592098165741e-05,
+      "loss": 1.0811,
+      "step": 12825
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.695201578614503e-05,
+      "loss": 1.1146,
+      "step": 12826
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.693811084728616e-05,
+      "loss": 0.9784,
+      "step": 12827
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692420616535422e-05,
+      "loss": 0.7817,
+      "step": 12828
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.691030174062273e-05,
+      "loss": 1.0658,
+      "step": 12829
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68963975733653e-05,
+      "loss": 1.0654,
+      "step": 12830
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688249366385528e-05,
+      "loss": 1.1618,
+      "step": 12831
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.686859001236623e-05,
+      "loss": 0.9635,
+      "step": 12832
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.685468661917167e-05,
+      "loss": 0.9334,
+      "step": 12833
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.6840783484545e-05,
+      "loss": 0.934,
+      "step": 12834
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68268806087597e-05,
+      "loss": 0.9356,
+      "step": 12835
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.681297799208933e-05,
+      "loss": 1.1391,
+      "step": 12836
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.679907563480726e-05,
+      "loss": 1.0629,
+      "step": 12837
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.678517353718698e-05,
+      "loss": 1.108,
+      "step": 12838
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.6771271699502e-05,
+      "loss": 0.8058,
+      "step": 12839
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.675737012202565e-05,
+      "loss": 0.9842,
+      "step": 12840
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.674346880503144e-05,
+      "loss": 1.0476,
+      "step": 12841
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.672956774879283e-05,
+      "loss": 0.9524,
+      "step": 12842
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671566695358324e-05,
+      "loss": 0.924,
+      "step": 12843
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.670176641967606e-05,
+      "loss": 0.9601,
+      "step": 12844
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668786614734478e-05,
+      "loss": 1.305,
+      "step": 12845
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.667396613686273e-05,
+      "loss": 1.0745,
+      "step": 12846
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.666006638850345e-05,
+      "loss": 1.1052,
+      "step": 12847
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.664616690254019e-05,
+      "loss": 0.8636,
+      "step": 12848
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663226767924646e-05,
+      "loss": 1.1242,
+      "step": 12849
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.661836871889566e-05,
+      "loss": 0.8495,
+      "step": 12850
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660447002176112e-05,
+      "loss": 1.0581,
+      "step": 12851
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.659057158811623e-05,
+      "loss": 0.9642,
+      "step": 12852
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.657667341823448e-05,
+      "loss": 0.972,
+      "step": 12853
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65627755123891e-05,
+      "loss": 0.9757,
+      "step": 12854
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.654887787085355e-05,
+      "loss": 1.0951,
+      "step": 12855
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65349804939012e-05,
+      "loss": 1.0219,
+      "step": 12856
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.652108338180536e-05,
+      "loss": 1.0049,
+      "step": 12857
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.650718653483939e-05,
+      "loss": 0.8472,
+      "step": 12858
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649328995327671e-05,
+      "loss": 1.0221,
+      "step": 12859
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64793936373906e-05,
+      "loss": 1.1026,
+      "step": 12860
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64654975874544e-05,
+      "loss": 1.0273,
+      "step": 12861
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.64516018037415e-05,
+      "loss": 0.8334,
+      "step": 12862
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.643770628652517e-05,
+      "loss": 1.0091,
+      "step": 12863
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.642381103607872e-05,
+      "loss": 0.8475,
+      "step": 12864
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.640991605267556e-05,
+      "loss": 0.8669,
+      "step": 12865
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.639602133658893e-05,
+      "loss": 1.0484,
+      "step": 12866
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.638212688809217e-05,
+      "loss": 0.9537,
+      "step": 12867
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.636823270745858e-05,
+      "loss": 1.0545,
+      "step": 12868
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63543387949614e-05,
+      "loss": 1.193,
+      "step": 12869
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634044515087405e-05,
+      "loss": 0.9443,
+      "step": 12870
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.632655177546966e-05,
+      "loss": 1.0088,
+      "step": 12871
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.631265866902162e-05,
+      "loss": 0.9422,
+      "step": 12872
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.629876583180321e-05,
+      "loss": 1.0513,
+      "step": 12873
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.628487326408764e-05,
+      "loss": 0.9723,
+      "step": 12874
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.627098096614821e-05,
+      "loss": 0.9997,
+      "step": 12875
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.625708893825822e-05,
+      "loss": 1.0648,
+      "step": 12876
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.624319718069084e-05,
+      "loss": 1.0148,
+      "step": 12877
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.622930569371937e-05,
+      "loss": 0.9164,
+      "step": 12878
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.621541447761708e-05,
+      "loss": 1.1853,
+      "step": 12879
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.620152353265715e-05,
+      "loss": 1.0794,
+      "step": 12880
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.618763285911282e-05,
+      "loss": 0.8248,
+      "step": 12881
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61737424572574e-05,
+      "loss": 1.0175,
+      "step": 12882
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.615985232736403e-05,
+      "loss": 1.079,
+      "step": 12883
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.614596246970597e-05,
+      "loss": 0.9933,
+      "step": 12884
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.613207288455643e-05,
+      "loss": 0.669,
+      "step": 12885
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.611818357218857e-05,
+      "loss": 0.9166,
+      "step": 12886
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.610429453287563e-05,
+      "loss": 1.0162,
+      "step": 12887
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.609040576689084e-05,
+      "loss": 0.8949,
+      "step": 12888
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.607651727450733e-05,
+      "loss": 0.8949,
+      "step": 12889
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.606262905599835e-05,
+      "loss": 0.9913,
+      "step": 12890
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604874111163701e-05,
+      "loss": 0.944,
+      "step": 12891
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60348534416965e-05,
+      "loss": 1.0779,
+      "step": 12892
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.602096604645009e-05,
+      "loss": 1.0219,
+      "step": 12893
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600707892617078e-05,
+      "loss": 0.952,
+      "step": 12894
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.599319208113186e-05,
+      "loss": 0.9924,
+      "step": 12895
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.597930551160644e-05,
+      "loss": 1.0529,
+      "step": 12896
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.596541921786765e-05,
+      "loss": 1.1602,
+      "step": 12897
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595153320018862e-05,
+      "loss": 1.0011,
+      "step": 12898
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.593764745884259e-05,
+      "loss": 0.9217,
+      "step": 12899
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.592376199410257e-05,
+      "loss": 1.1369,
+      "step": 12900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.590987680624174e-05,
+      "loss": 0.7526,
+      "step": 12901
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589599189553323e-05,
+      "loss": 0.9587,
+      "step": 12902
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.588210726225015e-05,
+      "loss": 0.9183,
+      "step": 12903
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.586822290666555e-05,
+      "loss": 1.0668,
+      "step": 12904
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.585433882905265e-05,
+      "loss": 1.2444,
+      "step": 12905
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.584045502968447e-05,
+      "loss": 0.9638,
+      "step": 12906
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.582657150883411e-05,
+      "loss": 0.8302,
+      "step": 12907
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58126882667747e-05,
+      "loss": 0.9601,
+      "step": 12908
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.579880530377926e-05,
+      "loss": 0.796,
+      "step": 12909
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.578492262012088e-05,
+      "loss": 0.9475,
+      "step": 12910
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57710402160727e-05,
+      "loss": 0.8924,
+      "step": 12911
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.575715809190771e-05,
+      "loss": 0.8647,
+      "step": 12912
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.574327624789904e-05,
+      "loss": 1.0541,
+      "step": 12913
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.572939468431967e-05,
+      "loss": 0.8901,
+      "step": 12914
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.571551340144267e-05,
+      "loss": 0.769,
+      "step": 12915
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.570163239954117e-05,
+      "loss": 1.0022,
+      "step": 12916
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.568775167888806e-05,
+      "loss": 1.2377,
+      "step": 12917
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567387123975648e-05,
+      "loss": 0.9808,
+      "step": 12918
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.565999108241945e-05,
+      "loss": 0.8416,
+      "step": 12919
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.564611120714995e-05,
+      "loss": 0.9207,
+      "step": 12920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.563223161422099e-05,
+      "loss": 0.9218,
+      "step": 12921
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.561835230390569e-05,
+      "loss": 0.8686,
+      "step": 12922
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.560447327647689e-05,
+      "loss": 0.9261,
+      "step": 12923
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.559059453220772e-05,
+      "loss": 1.2967,
+      "step": 12924
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.557671607137114e-05,
+      "loss": 1.0208,
+      "step": 12925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.556283789424011e-05,
+      "loss": 1.0195,
+      "step": 12926
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.55489600010876e-05,
+      "loss": 0.8632,
+      "step": 12927
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.553508239218669e-05,
+      "loss": 0.942,
+      "step": 12928
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.552120506781022e-05,
+      "loss": 0.9147,
+      "step": 12929
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.550732802823125e-05,
+      "loss": 1.0028,
+      "step": 12930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.549345127372273e-05,
+      "loss": 1.0228,
+      "step": 12931
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547957480455757e-05,
+      "loss": 1.0216,
+      "step": 12932
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546569862100876e-05,
+      "loss": 0.885,
+      "step": 12933
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54518227233492e-05,
+      "loss": 0.9735,
+      "step": 12934
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.543794711185185e-05,
+      "loss": 0.953,
+      "step": 12935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54240717867897e-05,
+      "loss": 1.0766,
+      "step": 12936
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54101967484356e-05,
+      "loss": 0.9742,
+      "step": 12937
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.539632199706247e-05,
+      "loss": 0.8964,
+      "step": 12938
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.538244753294332e-05,
+      "loss": 1.1175,
+      "step": 12939
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.536857335635094e-05,
+      "loss": 0.9802,
+      "step": 12940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.53546994675583e-05,
+      "loss": 0.9101,
+      "step": 12941
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.534082586683833e-05,
+      "loss": 0.906,
+      "step": 12942
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.532695255446383e-05,
+      "loss": 0.9304,
+      "step": 12943
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531307953070774e-05,
+      "loss": 1.0567,
+      "step": 12944
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529920679584301e-05,
+      "loss": 0.8992,
+      "step": 12945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.528533435014237e-05,
+      "loss": 0.8367,
+      "step": 12946
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.527146219387881e-05,
+      "loss": 1.1428,
+      "step": 12947
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.525759032732515e-05,
+      "loss": 1.0086,
+      "step": 12948
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.524371875075425e-05,
+      "loss": 0.7737,
+      "step": 12949
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.522984746443894e-05,
+      "loss": 0.9818,
+      "step": 12950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.521597646865214e-05,
+      "loss": 1.2262,
+      "step": 12951
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.520210576366661e-05,
+      "loss": 1.086,
+      "step": 12952
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518823534975524e-05,
+      "loss": 1.1383,
+      "step": 12953
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.517436522719085e-05,
+      "loss": 1.1015,
+      "step": 12954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.516049539624626e-05,
+      "loss": 1.1618,
+      "step": 12955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.514662585719429e-05,
+      "loss": 1.1553,
+      "step": 12956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.51327566103077e-05,
+      "loss": 0.977,
+      "step": 12957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.511888765585939e-05,
+      "loss": 0.9689,
+      "step": 12958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510501899412213e-05,
+      "loss": 0.9821,
+      "step": 12959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50911506253687e-05,
+      "loss": 1.0463,
+      "step": 12960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.507728254987186e-05,
+      "loss": 0.8976,
+      "step": 12961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.506341476790449e-05,
+      "loss": 0.8107,
+      "step": 12962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.504954727973926e-05,
+      "loss": 0.8461,
+      "step": 12963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.503568008564901e-05,
+      "loss": 0.8644,
+      "step": 12964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.502181318590652e-05,
+      "loss": 1.2256,
+      "step": 12965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.50079465807845e-05,
+      "loss": 1.094,
+      "step": 12966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.499408027055568e-05,
+      "loss": 1.1316,
+      "step": 12967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.498021425549296e-05,
+      "loss": 0.9528,
+      "step": 12968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.496634853586891e-05,
+      "loss": 1.0657,
+      "step": 12969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.495248311195633e-05,
+      "loss": 0.8574,
+      "step": 12970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.493861798402802e-05,
+      "loss": 1.0406,
+      "step": 12971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.492475315235662e-05,
+      "loss": 0.9224,
+      "step": 12972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.491088861721486e-05,
+      "loss": 1.0098,
+      "step": 12973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48970243788755e-05,
+      "loss": 1.0384,
+      "step": 12974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.488316043761121e-05,
+      "loss": 0.9102,
+      "step": 12975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.486929679369475e-05,
+      "loss": 0.8691,
+      "step": 12976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.485543344739874e-05,
+      "loss": 0.9484,
+      "step": 12977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.48415703989959e-05,
+      "loss": 0.7748,
+      "step": 12978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.482770764875896e-05,
+      "loss": 1.0484,
+      "step": 12979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.481384519696049e-05,
+      "loss": 0.9585,
+      "step": 12980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.479998304387329e-05,
+      "loss": 0.8701,
+      "step": 12981
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.478612118976998e-05,
+      "loss": 0.9937,
+      "step": 12982
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47722596349232e-05,
+      "loss": 1.2384,
+      "step": 12983
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.475839837960561e-05,
+      "loss": 0.9142,
+      "step": 12984
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.474453742408992e-05,
+      "loss": 1.1036,
+      "step": 12985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.473067676864867e-05,
+      "loss": 0.9954,
+      "step": 12986
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.471681641355459e-05,
+      "loss": 0.8941,
+      "step": 12987
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47029563590803e-05,
+      "loss": 0.8608,
+      "step": 12988
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.468909660549837e-05,
+      "loss": 0.9813,
+      "step": 12989
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.467523715308146e-05,
+      "loss": 1.0479,
+      "step": 12990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.466137800210222e-05,
+      "loss": 0.9405,
+      "step": 12991
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.464751915283318e-05,
+      "loss": 1.0626,
+      "step": 12992
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.463366060554698e-05,
+      "loss": 1.0863,
+      "step": 12993
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.461980236051627e-05,
+      "loss": 1.0371,
+      "step": 12994
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.460594441801355e-05,
+      "loss": 1.1231,
+      "step": 12995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.459208677831146e-05,
+      "loss": 0.9046,
+      "step": 12996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.457822944168257e-05,
+      "loss": 1.0786,
+      "step": 12997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.456437240839946e-05,
+      "loss": 1.0597,
+      "step": 12998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45505156787347e-05,
+      "loss": 0.984,
+      "step": 12999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.453665925296082e-05,
+      "loss": 0.785,
+      "step": 13000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.45228031313504e-05,
+      "loss": 0.9384,
+      "step": 13001
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.450894731417601e-05,
+      "loss": 1.1034,
+      "step": 13002
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.449509180171011e-05,
+      "loss": 1.1341,
+      "step": 13003
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.448123659422532e-05,
+      "loss": 0.8834,
+      "step": 13004
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.446738169199418e-05,
+      "loss": 0.7419,
+      "step": 13005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.445352709528915e-05,
+      "loss": 1.0097,
+      "step": 13006
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.443967280438277e-05,
+      "loss": 1.003,
+      "step": 13007
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.442581881954764e-05,
+      "loss": 0.921,
+      "step": 13008
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.441196514105612e-05,
+      "loss": 0.9697,
+      "step": 13009
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43981117691808e-05,
+      "loss": 0.8297,
+      "step": 13010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43842587041942e-05,
+      "loss": 0.9967,
+      "step": 13011
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.437040594636873e-05,
+      "loss": 1.139,
+      "step": 13012
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.435655349597689e-05,
+      "loss": 0.9504,
+      "step": 13013
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.434270135329127e-05,
+      "loss": 0.986,
+      "step": 13014
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.432884951858418e-05,
+      "loss": 0.9944,
+      "step": 13015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.431499799212817e-05,
+      "loss": 1.0276,
+      "step": 13016
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.430114677419571e-05,
+      "loss": 1.0805,
+      "step": 13017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42872958650592e-05,
+      "loss": 1.0936,
+      "step": 13018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.427344526499111e-05,
+      "loss": 1.1462,
+      "step": 13019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.425959497426392e-05,
+      "loss": 0.9156,
+      "step": 13020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.424574499315e-05,
+      "loss": 0.7897,
+      "step": 13021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.423189532192185e-05,
+      "loss": 0.8371,
+      "step": 13022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.421804596085184e-05,
+      "loss": 1.0734,
+      "step": 13023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.420419691021237e-05,
+      "loss": 1.0345,
+      "step": 13024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.419034817027594e-05,
+      "loss": 1.0067,
+      "step": 13025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.417649974131482e-05,
+      "loss": 1.0371,
+      "step": 13026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.416265162360151e-05,
+      "loss": 0.8972,
+      "step": 13027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.414880381740841e-05,
+      "loss": 1.0422,
+      "step": 13028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413495632300784e-05,
+      "loss": 0.9021,
+      "step": 13029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.41211091406722e-05,
+      "loss": 0.9642,
+      "step": 13030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410726227067391e-05,
+      "loss": 1.0931,
+      "step": 13031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.409341571328526e-05,
+      "loss": 1.0828,
+      "step": 13032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.407956946877867e-05,
+      "loss": 0.9142,
+      "step": 13033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.406572353742651e-05,
+      "loss": 0.9994,
+      "step": 13034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.405187791950107e-05,
+      "loss": 1.0287,
+      "step": 13035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.403803261527468e-05,
+      "loss": 1.0011,
+      "step": 13036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.402418762501982e-05,
+      "loss": 1.1758,
+      "step": 13037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.401034294900862e-05,
+      "loss": 1.1114,
+      "step": 13038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.399649858751353e-05,
+      "loss": 1.0494,
+      "step": 13039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.398265454080687e-05,
+      "loss": 0.6936,
+      "step": 13040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396881080916091e-05,
+      "loss": 0.7772,
+      "step": 13041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.395496739284796e-05,
+      "loss": 0.8283,
+      "step": 13042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.39411242921403e-05,
+      "loss": 1.0229,
+      "step": 13043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392728150731029e-05,
+      "loss": 0.8318,
+      "step": 13044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.391343903863018e-05,
+      "loss": 0.8948,
+      "step": 13045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.389959688637223e-05,
+      "loss": 0.8997,
+      "step": 13046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.388575505080873e-05,
+      "loss": 0.9343,
+      "step": 13047
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.387191353221198e-05,
+      "loss": 1.0523,
+      "step": 13048
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.385807233085418e-05,
+      "loss": 1.0294,
+      "step": 13049
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.384423144700763e-05,
+      "loss": 1.0274,
+      "step": 13050
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.383039088094458e-05,
+      "loss": 0.9609,
+      "step": 13051
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381655063293728e-05,
+      "loss": 0.8804,
+      "step": 13052
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.380271070325789e-05,
+      "loss": 0.8413,
+      "step": 13053
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37888710921788e-05,
+      "loss": 1.0426,
+      "step": 13054
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.377503179997204e-05,
+      "loss": 0.8731,
+      "step": 13055
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.376119282690997e-05,
+      "loss": 0.954,
+      "step": 13056
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.374735417326477e-05,
+      "loss": 0.8983,
+      "step": 13057
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.373351583930862e-05,
+      "loss": 0.8871,
+      "step": 13058
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37196778253137e-05,
+      "loss": 0.9269,
+      "step": 13059
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.370584013155231e-05,
+      "loss": 1.0532,
+      "step": 13060
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.369200275829649e-05,
+      "loss": 0.801,
+      "step": 13061
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.367816570581853e-05,
+      "loss": 0.9705,
+      "step": 13062
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.366432897439058e-05,
+      "loss": 0.9796,
+      "step": 13063
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.365049256428477e-05,
+      "loss": 0.8331,
+      "step": 13064
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.363665647577333e-05,
+      "loss": 1.0679,
+      "step": 13065
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36228207091283e-05,
+      "loss": 0.9859,
+      "step": 13066
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.360898526462195e-05,
+      "loss": 0.9691,
+      "step": 13067
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359515014252638e-05,
+      "loss": 1.0924,
+      "step": 13068
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.358131534311372e-05,
+      "loss": 0.8267,
+      "step": 13069
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.356748086665609e-05,
+      "loss": 0.9856,
+      "step": 13070
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.355364671342565e-05,
+      "loss": 1.1073,
+      "step": 13071
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.353981288369445e-05,
+      "loss": 1.0528,
+      "step": 13072
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.352597937773467e-05,
+      "loss": 0.9687,
+      "step": 13073
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.351214619581841e-05,
+      "loss": 1.0045,
+      "step": 13074
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349831333821772e-05,
+      "loss": 1.0586,
+      "step": 13075
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.348448080520471e-05,
+      "loss": 0.868,
+      "step": 13076
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.347064859705153e-05,
+      "loss": 0.8846,
+      "step": 13077
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.345681671403014e-05,
+      "loss": 0.8457,
+      "step": 13078
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34429851564127e-05,
+      "loss": 0.9191,
+      "step": 13079
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.342915392447128e-05,
+      "loss": 1.1016,
+      "step": 13080
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.341532301847788e-05,
+      "loss": 1.0715,
+      "step": 13081
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.340149243870457e-05,
+      "loss": 1.0371,
+      "step": 13082
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.338766218542347e-05,
+      "loss": 1.0157,
+      "step": 13083
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.337383225890652e-05,
+      "loss": 0.8207,
+      "step": 13084
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.336000265942581e-05,
+      "loss": 0.8922,
+      "step": 13085
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.334617338725336e-05,
+      "loss": 0.9829,
+      "step": 13086
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33323444426612e-05,
+      "loss": 1.2074,
+      "step": 13087
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.331851582592133e-05,
+      "loss": 0.8624,
+      "step": 13088
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330468753730572e-05,
+      "loss": 0.9444,
+      "step": 13089
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.329085957708644e-05,
+      "loss": 0.9774,
+      "step": 13090
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.327703194553549e-05,
+      "loss": 0.9269,
+      "step": 13091
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.326320464292479e-05,
+      "loss": 1.1593,
+      "step": 13092
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.324937766952638e-05,
+      "loss": 0.6997,
+      "step": 13093
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.323555102561221e-05,
+      "loss": 0.9583,
+      "step": 13094
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.322172471145423e-05,
+      "loss": 1.141,
+      "step": 13095
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.320789872732445e-05,
+      "loss": 0.9952,
+      "step": 13096
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.319407307349482e-05,
+      "loss": 1.2047,
+      "step": 13097
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.318024775023728e-05,
+      "loss": 0.7026,
+      "step": 13098
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.316642275782373e-05,
+      "loss": 1.0387,
+      "step": 13099
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31525980965262e-05,
+      "loss": 0.9386,
+      "step": 13100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.313877376661652e-05,
+      "loss": 0.9258,
+      "step": 13101
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.312494976836668e-05,
+      "loss": 1.0242,
+      "step": 13102
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31111261020486e-05,
+      "loss": 1.0203,
+      "step": 13103
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.309730276793414e-05,
+      "loss": 0.9772,
+      "step": 13104
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.308347976629523e-05,
+      "loss": 0.9157,
+      "step": 13105
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.306965709740385e-05,
+      "loss": 1.3718,
+      "step": 13106
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.305583476153172e-05,
+      "loss": 1.0321,
+      "step": 13107
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.304201275895087e-05,
+      "loss": 0.8442,
+      "step": 13108
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.302819108993312e-05,
+      "loss": 0.9426,
+      "step": 13109
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.301436975475034e-05,
+      "loss": 0.9816,
+      "step": 13110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.300054875367443e-05,
+      "loss": 1.1328,
+      "step": 13111
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.298672808697718e-05,
+      "loss": 0.7887,
+      "step": 13112
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29729077549305e-05,
+      "loss": 1.253,
+      "step": 13113
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.295908775780626e-05,
+      "loss": 0.8098,
+      "step": 13114
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.294526809587624e-05,
+      "loss": 0.9151,
+      "step": 13115
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.293144876941228e-05,
+      "loss": 1.0409,
+      "step": 13116
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.291762977868625e-05,
+      "loss": 1.0356,
+      "step": 13117
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290381112396987e-05,
+      "loss": 0.9146,
+      "step": 13118
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.288999280553506e-05,
+      "loss": 0.9697,
+      "step": 13119
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.287617482365362e-05,
+      "loss": 0.944,
+      "step": 13120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.286235717859728e-05,
+      "loss": 0.9429,
+      "step": 13121
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.284853987063785e-05,
+      "loss": 0.8823,
+      "step": 13122
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28347229000472e-05,
+      "loss": 0.8832,
+      "step": 13123
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.282090626709698e-05,
+      "loss": 0.9816,
+      "step": 13124
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280708997205904e-05,
+      "loss": 1.0951,
+      "step": 13125
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.279327401520516e-05,
+      "loss": 0.9406,
+      "step": 13126
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.277945839680706e-05,
+      "loss": 1.1612,
+      "step": 13127
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.276564311713646e-05,
+      "loss": 1.2774,
+      "step": 13128
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.275182817646523e-05,
+      "loss": 0.9162,
+      "step": 13129
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.273801357506496e-05,
+      "loss": 0.9751,
+      "step": 13130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.272419931320752e-05,
+      "loss": 1.0648,
+      "step": 13131
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.271038539116452e-05,
+      "loss": 0.8902,
+      "step": 13132
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.269657180920774e-05,
+      "loss": 0.8687,
+      "step": 13133
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.268275856760889e-05,
+      "loss": 0.7622,
+      "step": 13134
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.266894566663966e-05,
+      "loss": 0.9318,
+      "step": 13135
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.265513310657173e-05,
+      "loss": 1.0906,
+      "step": 13136
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.264132088767687e-05,
+      "loss": 1.051,
+      "step": 13137
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26275090102267e-05,
+      "loss": 0.9242,
+      "step": 13138
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26136974744929e-05,
+      "loss": 0.9225,
+      "step": 13139
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.259988628074718e-05,
+      "loss": 1.0644,
+      "step": 13140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.258607542926114e-05,
+      "loss": 1.0745,
+      "step": 13141
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.257226492030653e-05,
+      "loss": 0.8512,
+      "step": 13142
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.255845475415494e-05,
+      "loss": 1.0332,
+      "step": 13143
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254464493107803e-05,
+      "loss": 0.912,
+      "step": 13144
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25308354513474e-05,
+      "loss": 0.9316,
+      "step": 13145
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.251702631523479e-05,
+      "loss": 1.0017,
+      "step": 13146
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25032175230117e-05,
+      "loss": 0.8187,
+      "step": 13147
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.248940907494982e-05,
+      "loss": 1.102,
+      "step": 13148
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.247560097132076e-05,
+      "loss": 1.0305,
+      "step": 13149
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246179321239611e-05,
+      "loss": 0.9411,
+      "step": 13150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.244798579844748e-05,
+      "loss": 0.9153,
+      "step": 13151
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.243417872974643e-05,
+      "loss": 1.0623,
+      "step": 13152
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.242037200656455e-05,
+      "loss": 0.9904,
+      "step": 13153
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.240656562917347e-05,
+      "loss": 1.1547,
+      "step": 13154
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23927595978447e-05,
+      "loss": 1.062,
+      "step": 13155
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.237895391284985e-05,
+      "loss": 0.9582,
+      "step": 13156
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.236514857446045e-05,
+      "loss": 0.8598,
+      "step": 13157
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.235134358294802e-05,
+      "loss": 1.0965,
+      "step": 13158
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.233753893858418e-05,
+      "loss": 0.9744,
+      "step": 13159
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.232373464164043e-05,
+      "loss": 1.0002,
+      "step": 13160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.230993069238828e-05,
+      "loss": 1.1831,
+      "step": 13161
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.229612709109927e-05,
+      "loss": 0.9373,
+      "step": 13162
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.228232383804494e-05,
+      "loss": 1.0438,
+      "step": 13163
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.226852093349676e-05,
+      "loss": 1.1718,
+      "step": 13164
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.225471837772623e-05,
+      "loss": 0.9949,
+      "step": 13165
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.224091617100492e-05,
+      "loss": 1.0116,
+      "step": 13166
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.222711431360422e-05,
+      "loss": 0.9389,
+      "step": 13167
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.221331280579564e-05,
+      "loss": 0.8998,
+      "step": 13168
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.219951164785073e-05,
+      "loss": 0.9629,
+      "step": 13169
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.218571084004086e-05,
+      "loss": 0.8954,
+      "step": 13170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217191038263753e-05,
+      "loss": 0.9711,
+      "step": 13171
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.215811027591223e-05,
+      "loss": 0.941,
+      "step": 13172
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.214431052013634e-05,
+      "loss": 0.824,
+      "step": 13173
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.213051111558137e-05,
+      "loss": 1.0506,
+      "step": 13174
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21167120625187e-05,
+      "loss": 0.7985,
+      "step": 13175
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.210291336121974e-05,
+      "loss": 0.9741,
+      "step": 13176
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208911501195599e-05,
+      "loss": 1.0758,
+      "step": 13177
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.207531701499881e-05,
+      "loss": 1.1114,
+      "step": 13178
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206151937061962e-05,
+      "loss": 0.9162,
+      "step": 13179
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.204772207908982e-05,
+      "loss": 1.0065,
+      "step": 13180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203392514068074e-05,
+      "loss": 0.9762,
+      "step": 13181
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.202012855566387e-05,
+      "loss": 0.9416,
+      "step": 13182
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.200633232431055e-05,
+      "loss": 0.9323,
+      "step": 13183
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.199253644689213e-05,
+      "loss": 1.0266,
+      "step": 13184
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.197874092367997e-05,
+      "loss": 1.0589,
+      "step": 13185
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.196494575494547e-05,
+      "loss": 1.032,
+      "step": 13186
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195115094095993e-05,
+      "loss": 0.9418,
+      "step": 13187
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193735648199472e-05,
+      "loss": 1.0637,
+      "step": 13188
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19235623783212e-05,
+      "loss": 1.0491,
+      "step": 13189
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.190976863021066e-05,
+      "loss": 0.9815,
+      "step": 13190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.189597523793441e-05,
+      "loss": 0.9618,
+      "step": 13191
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.188218220176386e-05,
+      "loss": 1.054,
+      "step": 13192
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186838952197018e-05,
+      "loss": 0.8605,
+      "step": 13193
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.185459719882477e-05,
+      "loss": 0.9562,
+      "step": 13194
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.184080523259892e-05,
+      "loss": 0.9463,
+      "step": 13195
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.182701362356387e-05,
+      "loss": 0.9013,
+      "step": 13196
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.181322237199094e-05,
+      "loss": 0.9971,
+      "step": 13197
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179943147815138e-05,
+      "loss": 1.0838,
+      "step": 13198
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.178564094231645e-05,
+      "loss": 1.0736,
+      "step": 13199
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.177185076475744e-05,
+      "loss": 1.1178,
+      "step": 13200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.175806094574559e-05,
+      "loss": 0.9917,
+      "step": 13201
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.174427148555212e-05,
+      "loss": 0.9572,
+      "step": 13202
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.173048238444834e-05,
+      "loss": 1.1097,
+      "step": 13203
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171669364270538e-05,
+      "loss": 0.7965,
+      "step": 13204
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.170290526059452e-05,
+      "loss": 0.8529,
+      "step": 13205
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.168911723838702e-05,
+      "loss": 1.0915,
+      "step": 13206
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.167532957635401e-05,
+      "loss": 0.9116,
+      "step": 13207
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166154227476672e-05,
+      "loss": 1.0529,
+      "step": 13208
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.16477553338964e-05,
+      "loss": 1.1114,
+      "step": 13209
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163396875401412e-05,
+      "loss": 0.9568,
+      "step": 13210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.162018253539116e-05,
+      "loss": 1.0719,
+      "step": 13211
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.160639667829869e-05,
+      "loss": 1.0947,
+      "step": 13212
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.159261118300782e-05,
+      "loss": 1.1197,
+      "step": 13213
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.157882604978975e-05,
+      "loss": 0.9147,
+      "step": 13214
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156504127891568e-05,
+      "loss": 0.9881,
+      "step": 13215
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.155125687065663e-05,
+      "loss": 0.921,
+      "step": 13216
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153747282528387e-05,
+      "loss": 1.0372,
+      "step": 13217
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.152368914306846e-05,
+      "loss": 0.9083,
+      "step": 13218
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.15099058242815e-05,
+      "loss": 0.8976,
+      "step": 13219
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14961228691942e-05,
+      "loss": 1.0249,
+      "step": 13220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14823402780776e-05,
+      "loss": 1.0614,
+      "step": 13221
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.146855805120277e-05,
+      "loss": 0.9927,
+      "step": 13222
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145477618884093e-05,
+      "loss": 0.9789,
+      "step": 13223
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.144099469126305e-05,
+      "loss": 0.914,
+      "step": 13224
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.142721355874026e-05,
+      "loss": 1.3094,
+      "step": 13225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.141343279154367e-05,
+      "loss": 1.01,
+      "step": 13226
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.139965238994424e-05,
+      "loss": 0.9074,
+      "step": 13227
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.138587235421314e-05,
+      "loss": 0.8285,
+      "step": 13228
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.13720926846214e-05,
+      "loss": 0.9176,
+      "step": 13229
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.135831338144002e-05,
+      "loss": 0.9905,
+      "step": 13230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.134453444494007e-05,
+      "loss": 0.9979,
+      "step": 13231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.133075587539259e-05,
+      "loss": 0.991,
+      "step": 13232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.131697767306855e-05,
+      "loss": 0.8598,
+      "step": 13233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.130319983823903e-05,
+      "loss": 1.0243,
+      "step": 13234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.128942237117504e-05,
+      "loss": 0.9498,
+      "step": 13235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.127564527214753e-05,
+      "loss": 1.1093,
+      "step": 13236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.126186854142752e-05,
+      "loss": 0.9508,
+      "step": 13237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.124809217928606e-05,
+      "loss": 0.8082,
+      "step": 13238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.123431618599401e-05,
+      "loss": 0.8718,
+      "step": 13239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.122054056182245e-05,
+      "loss": 1.0516,
+      "step": 13240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.120676530704227e-05,
+      "loss": 0.8855,
+      "step": 13241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.119299042192448e-05,
+      "loss": 1.0021,
+      "step": 13242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.117921590674001e-05,
+      "loss": 0.9045,
+      "step": 13243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11654417617598e-05,
+      "loss": 0.9566,
+      "step": 13244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.115166798725478e-05,
+      "loss": 0.8782,
+      "step": 13245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.113789458349592e-05,
+      "loss": 1.1372,
+      "step": 13246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.112412155075411e-05,
+      "loss": 1.0615,
+      "step": 13247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.111034888930026e-05,
+      "loss": 1.0326,
+      "step": 13248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.109657659940533e-05,
+      "loss": 0.9659,
+      "step": 13249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.108280468134012e-05,
+      "loss": 0.9026,
+      "step": 13250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.106903313537561e-05,
+      "loss": 0.8182,
+      "step": 13251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105526196178269e-05,
+      "loss": 0.8159,
+      "step": 13252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.104149116083216e-05,
+      "loss": 1.1121,
+      "step": 13253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.102772073279496e-05,
+      "loss": 0.926,
+      "step": 13254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.101395067794195e-05,
+      "loss": 0.8116,
+      "step": 13255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.100018099654393e-05,
+      "loss": 1.2419,
+      "step": 13256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09864116888718e-05,
+      "loss": 0.8709,
+      "step": 13257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.097264275519642e-05,
+      "loss": 1.036,
+      "step": 13258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.095887419578859e-05,
+      "loss": 0.9345,
+      "step": 13259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09451060109191e-05,
+      "loss": 0.795,
+      "step": 13260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093133820085888e-05,
+      "loss": 1.0573,
+      "step": 13261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.091757076587861e-05,
+      "loss": 0.976,
+      "step": 13262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09038037062492e-05,
+      "loss": 1.0397,
+      "step": 13263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.089003702224139e-05,
+      "loss": 1.0252,
+      "step": 13264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.087627071412598e-05,
+      "loss": 0.7051,
+      "step": 13265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.086250478217378e-05,
+      "loss": 0.9551,
+      "step": 13266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.084873922665555e-05,
+      "loss": 0.8208,
+      "step": 13267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.083497404784201e-05,
+      "loss": 1.1945,
+      "step": 13268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.0821209246004e-05,
+      "loss": 0.7766,
+      "step": 13269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.080744482141224e-05,
+      "loss": 0.9208,
+      "step": 13270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.079368077433747e-05,
+      "loss": 0.8407,
+      "step": 13271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.077991710505044e-05,
+      "loss": 0.9867,
+      "step": 13272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.076615381382183e-05,
+      "loss": 0.8837,
+      "step": 13273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.075239090092243e-05,
+      "loss": 1.0606,
+      "step": 13274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.073862836662297e-05,
+      "loss": 1.0173,
+      "step": 13275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072486621119409e-05,
+      "loss": 1.0758,
+      "step": 13276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.071110443490652e-05,
+      "loss": 1.2472,
+      "step": 13277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.069734303803098e-05,
+      "loss": 0.889,
+      "step": 13278
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.068358202083809e-05,
+      "loss": 0.8987,
+      "step": 13279
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.066982138359858e-05,
+      "loss": 0.8974,
+      "step": 13280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.065606112658315e-05,
+      "loss": 1.1132,
+      "step": 13281
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.06423012500624e-05,
+      "loss": 1.1385,
+      "step": 13282
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.062854175430703e-05,
+      "loss": 1.0086,
+      "step": 13283
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.061478263958766e-05,
+      "loss": 0.9295,
+      "step": 13284
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.060102390617493e-05,
+      "loss": 1.0991,
+      "step": 13285
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.058726555433951e-05,
+      "loss": 0.9008,
+      "step": 13286
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.057350758435198e-05,
+      "loss": 1.0417,
+      "step": 13287
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.0559749996483e-05,
+      "loss": 0.9505,
+      "step": 13288
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054599279100317e-05,
+      "loss": 0.9344,
+      "step": 13289
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.053223596818307e-05,
+      "loss": 0.9982,
+      "step": 13290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.051847952829328e-05,
+      "loss": 1.1866,
+      "step": 13291
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05047234716045e-05,
+      "loss": 0.9992,
+      "step": 13292
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.049096779838719e-05,
+      "loss": 0.7957,
+      "step": 13293
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.047721250891195e-05,
+      "loss": 1.1511,
+      "step": 13294
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.046345760344943e-05,
+      "loss": 1.0614,
+      "step": 13295
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044970308227004e-05,
+      "loss": 1.0535,
+      "step": 13296
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.043594894564445e-05,
+      "loss": 0.962,
+      "step": 13297
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.042219519384318e-05,
+      "loss": 1.0265,
+      "step": 13298
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.040844182713673e-05,
+      "loss": 0.9694,
+      "step": 13299
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.039468884579566e-05,
+      "loss": 0.9157,
+      "step": 13300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.038093625009052e-05,
+      "loss": 0.9265,
+      "step": 13301
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036718404029171e-05,
+      "loss": 0.832,
+      "step": 13302
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.035343221666984e-05,
+      "loss": 0.8478,
+      "step": 13303
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03396807794954e-05,
+      "loss": 0.9746,
+      "step": 13304
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.032592972903885e-05,
+      "loss": 0.9541,
+      "step": 13305
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.031217906557069e-05,
+      "loss": 0.9195,
+      "step": 13306
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.029842878936137e-05,
+      "loss": 0.932,
+      "step": 13307
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.028467890068134e-05,
+      "loss": 0.954,
+      "step": 13308
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.027092939980115e-05,
+      "loss": 0.7853,
+      "step": 13309
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.025718028699119e-05,
+      "loss": 0.9288,
+      "step": 13310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.024343156252188e-05,
+      "loss": 1.1445,
+      "step": 13311
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.022968322666371e-05,
+      "loss": 0.9539,
+      "step": 13312
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.021593527968709e-05,
+      "loss": 0.8509,
+      "step": 13313
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02021877218624e-05,
+      "loss": 0.8789,
+      "step": 13314
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.018844055346015e-05,
+      "loss": 0.8822,
+      "step": 13315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.017469377475067e-05,
+      "loss": 1.0718,
+      "step": 13316
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016094738600435e-05,
+      "loss": 1.1635,
+      "step": 13317
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.014720138749166e-05,
+      "loss": 0.8621,
+      "step": 13318
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.013345577948287e-05,
+      "loss": 1.0962,
+      "step": 13319
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.011971056224845e-05,
+      "loss": 1.1149,
+      "step": 13320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010596573605876e-05,
+      "loss": 0.9716,
+      "step": 13321
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00922213011841e-05,
+      "loss": 1.1685,
+      "step": 13322
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.007847725789488e-05,
+      "loss": 0.9434,
+      "step": 13323
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006473360646142e-05,
+      "loss": 0.9296,
+      "step": 13324
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.005099034715404e-05,
+      "loss": 0.9711,
+      "step": 13325
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.003724748024315e-05,
+      "loss": 0.8925,
+      "step": 13326
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.002350500599895e-05,
+      "loss": 1.0073,
+      "step": 13327
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.000976292469183e-05,
+      "loss": 0.8986,
+      "step": 13328
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.999602123659211e-05,
+      "loss": 1.0513,
+      "step": 13329
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.998227994197004e-05,
+      "loss": 1.0507,
+      "step": 13330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.996853904109592e-05,
+      "loss": 0.9067,
+      "step": 13331
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.995479853424008e-05,
+      "loss": 0.7839,
+      "step": 13332
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.994105842167273e-05,
+      "loss": 0.9945,
+      "step": 13333
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.992731870366418e-05,
+      "loss": 0.9606,
+      "step": 13334
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991357938048469e-05,
+      "loss": 0.8905,
+      "step": 13335
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.989984045240449e-05,
+      "loss": 0.9255,
+      "step": 13336
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.988610191969381e-05,
+      "loss": 0.8905,
+      "step": 13337
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.987236378262296e-05,
+      "loss": 0.9619,
+      "step": 13338
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.985862604146209e-05,
+      "loss": 1.0285,
+      "step": 13339
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.984488869648145e-05,
+      "loss": 1.1302,
+      "step": 13340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.983115174795128e-05,
+      "loss": 1.1537,
+      "step": 13341
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.981741519614172e-05,
+      "loss": 1.0015,
+      "step": 13342
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980367904132302e-05,
+      "loss": 1.0664,
+      "step": 13343
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.978994328376539e-05,
+      "loss": 1.0595,
+      "step": 13344
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.977620792373894e-05,
+      "loss": 1.1743,
+      "step": 13345
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97624729615139e-05,
+      "loss": 1.0313,
+      "step": 13346
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.974873839736044e-05,
+      "loss": 0.9453,
+      "step": 13347
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.973500423154865e-05,
+      "loss": 0.8782,
+      "step": 13348
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.972127046434878e-05,
+      "loss": 0.771,
+      "step": 13349
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.970753709603087e-05,
+      "loss": 0.8409,
+      "step": 13350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.969380412686514e-05,
+      "loss": 1.0414,
+      "step": 13351
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.968007155712169e-05,
+      "loss": 0.9807,
+      "step": 13352
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96663393870706e-05,
+      "loss": 1.2105,
+      "step": 13353
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965260761698201e-05,
+      "loss": 0.8513,
+      "step": 13354
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.963887624712607e-05,
+      "loss": 1.2427,
+      "step": 13355
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96251452777728e-05,
+      "loss": 0.91,
+      "step": 13356
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.961141470919233e-05,
+      "loss": 0.9059,
+      "step": 13357
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.959768454165476e-05,
+      "loss": 1.0464,
+      "step": 13358
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.958395477543009e-05,
+      "loss": 0.9229,
+      "step": 13359
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957022541078841e-05,
+      "loss": 0.9448,
+      "step": 13360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.955649644799985e-05,
+      "loss": 0.9657,
+      "step": 13361
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.954276788733437e-05,
+      "loss": 0.9155,
+      "step": 13362
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952903972906204e-05,
+      "loss": 1.035,
+      "step": 13363
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.951531197345292e-05,
+      "loss": 1.0953,
+      "step": 13364
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.950158462077697e-05,
+      "loss": 0.8567,
+      "step": 13365
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.948785767130425e-05,
+      "loss": 0.9071,
+      "step": 13366
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.94741311253048e-05,
+      "loss": 1.0972,
+      "step": 13367
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.946040498304856e-05,
+      "loss": 1.108,
+      "step": 13368
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944667924480555e-05,
+      "loss": 0.7771,
+      "step": 13369
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.943295391084577e-05,
+      "loss": 0.6306,
+      "step": 13370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.941922898143914e-05,
+      "loss": 1.0587,
+      "step": 13371
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940550445685573e-05,
+      "loss": 1.0158,
+      "step": 13372
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.939178033736538e-05,
+      "loss": 0.896,
+      "step": 13373
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937805662323812e-05,
+      "loss": 0.9354,
+      "step": 13374
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.936433331474388e-05,
+      "loss": 0.7878,
+      "step": 13375
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93506104121526e-05,
+      "loss": 1.2785,
+      "step": 13376
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.933688791573417e-05,
+      "loss": 1.0268,
+      "step": 13377
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93231658257586e-05,
+      "loss": 0.7963,
+      "step": 13378
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.930944414249571e-05,
+      "loss": 1.1845,
+      "step": 13379
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.929572286621545e-05,
+      "loss": 0.9275,
+      "step": 13380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.928200199718774e-05,
+      "loss": 0.9169,
+      "step": 13381
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92682815356824e-05,
+      "loss": 0.9685,
+      "step": 13382
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925456148196934e-05,
+      "loss": 0.9569,
+      "step": 13383
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.924084183631848e-05,
+      "loss": 1.0221,
+      "step": 13384
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.922712259899963e-05,
+      "loss": 1.1274,
+      "step": 13385
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.921340377028267e-05,
+      "loss": 1.3304,
+      "step": 13386
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.919968535043745e-05,
+      "loss": 1.0992,
+      "step": 13387
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.918596733973378e-05,
+      "loss": 0.9095,
+      "step": 13388
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.917224973844152e-05,
+      "loss": 0.9783,
+      "step": 13389
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.915853254683053e-05,
+      "loss": 0.9418,
+      "step": 13390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.914481576517057e-05,
+      "loss": 0.998,
+      "step": 13391
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.913109939373148e-05,
+      "loss": 1.0872,
+      "step": 13392
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.911738343278304e-05,
+      "loss": 1.0778,
+      "step": 13393
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.910366788259502e-05,
+      "loss": 0.9136,
+      "step": 13394
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.908995274343729e-05,
+      "loss": 0.8539,
+      "step": 13395
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.907623801557952e-05,
+      "loss": 0.9335,
+      "step": 13396
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.906252369929154e-05,
+      "loss": 1.0371,
+      "step": 13397
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.904880979484315e-05,
+      "loss": 1.0754,
+      "step": 13398
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.903509630250401e-05,
+      "loss": 0.949,
+      "step": 13399
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.902138322254388e-05,
+      "loss": 0.9321,
+      "step": 13400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.900767055523258e-05,
+      "loss": 0.9219,
+      "step": 13401
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.899395830083976e-05,
+      "loss": 0.9409,
+      "step": 13402
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.898024645963514e-05,
+      "loss": 0.9683,
+      "step": 13403
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.896653503188849e-05,
+      "loss": 1.131,
+      "step": 13404
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.895282401786945e-05,
+      "loss": 0.9508,
+      "step": 13405
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.893911341784772e-05,
+      "loss": 1.151,
+      "step": 13406
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.892540323209305e-05,
+      "loss": 1.1043,
+      "step": 13407
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.891169346087506e-05,
+      "loss": 0.9973,
+      "step": 13408
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.889798410446343e-05,
+      "loss": 0.9719,
+      "step": 13409
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.888427516312784e-05,
+      "loss": 1.1157,
+      "step": 13410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88705666371379e-05,
+      "loss": 1.0606,
+      "step": 13411
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.885685852676332e-05,
+      "loss": 0.9011,
+      "step": 13412
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.884315083227373e-05,
+      "loss": 0.8148,
+      "step": 13413
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882944355393871e-05,
+      "loss": 0.9805,
+      "step": 13414
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.881573669202793e-05,
+      "loss": 0.9337,
+      "step": 13415
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880203024681096e-05,
+      "loss": 1.062,
+      "step": 13416
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878832421855742e-05,
+      "loss": 0.9322,
+      "step": 13417
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.877461860753697e-05,
+      "loss": 1.0967,
+      "step": 13418
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.876091341401907e-05,
+      "loss": 0.8579,
+      "step": 13419
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.874720863827343e-05,
+      "loss": 0.9819,
+      "step": 13420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.873350428056957e-05,
+      "loss": 1.0126,
+      "step": 13421
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.871980034117703e-05,
+      "loss": 0.9745,
+      "step": 13422
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870609682036537e-05,
+      "loss": 0.9488,
+      "step": 13423
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86923937184042e-05,
+      "loss": 0.8634,
+      "step": 13424
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8678691035563e-05,
+      "loss": 1.1663,
+      "step": 13425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.866498877211131e-05,
+      "loss": 1.0488,
+      "step": 13426
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.865128692831869e-05,
+      "loss": 0.9742,
+      "step": 13427
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86375855044546e-05,
+      "loss": 0.7244,
+      "step": 13428
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.862388450078855e-05,
+      "loss": 0.8704,
+      "step": 13429
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86101839175901e-05,
+      "loss": 0.9318,
+      "step": 13430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.859648375512869e-05,
+      "loss": 0.958,
+      "step": 13431
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.858278401367381e-05,
+      "loss": 0.9479,
+      "step": 13432
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.856908469349495e-05,
+      "loss": 1.1612,
+      "step": 13433
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.855538579486152e-05,
+      "loss": 1.0293,
+      "step": 13434
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.854168731804305e-05,
+      "loss": 0.9685,
+      "step": 13435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.852798926330897e-05,
+      "loss": 1.0557,
+      "step": 13436
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.851429163092868e-05,
+      "loss": 0.929,
+      "step": 13437
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850059442117165e-05,
+      "loss": 1.0209,
+      "step": 13438
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84868976343073e-05,
+      "loss": 1.2334,
+      "step": 13439
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8473201270605e-05,
+      "loss": 0.9212,
+      "step": 13440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.845950533033427e-05,
+      "loss": 1.1551,
+      "step": 13441
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.844580981376437e-05,
+      "loss": 0.6706,
+      "step": 13442
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.843211472116476e-05,
+      "loss": 1.1451,
+      "step": 13443
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.841842005280486e-05,
+      "loss": 1.015,
+      "step": 13444
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840472580895396e-05,
+      "loss": 1.0366,
+      "step": 13445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.839103198988146e-05,
+      "loss": 1.1256,
+      "step": 13446
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.837733859585675e-05,
+      "loss": 0.8026,
+      "step": 13447
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.836364562714914e-05,
+      "loss": 0.9811,
+      "step": 13448
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.834995308402798e-05,
+      "loss": 0.9544,
+      "step": 13449
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.833626096676263e-05,
+      "loss": 0.975,
+      "step": 13450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.832256927562236e-05,
+      "loss": 0.7805,
+      "step": 13451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.830887801087648e-05,
+      "loss": 0.9896,
+      "step": 13452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.829518717279438e-05,
+      "loss": 1.008,
+      "step": 13453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.828149676164529e-05,
+      "loss": 0.9351,
+      "step": 13454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.826780677769852e-05,
+      "loss": 0.9131,
+      "step": 13455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825411722122337e-05,
+      "loss": 0.8807,
+      "step": 13456
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.824042809248905e-05,
+      "loss": 0.9356,
+      "step": 13457
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.822673939176493e-05,
+      "loss": 0.9117,
+      "step": 13458
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.821305111932014e-05,
+      "loss": 1.1139,
+      "step": 13459
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8199363275424e-05,
+      "loss": 1.0084,
+      "step": 13460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.818567586034577e-05,
+      "loss": 0.9668,
+      "step": 13461
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817198887435463e-05,
+      "loss": 0.9341,
+      "step": 13462
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81583023177198e-05,
+      "loss": 0.9996,
+      "step": 13463
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814461619071059e-05,
+      "loss": 0.9873,
+      "step": 13464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.813093049359607e-05,
+      "loss": 0.8969,
+      "step": 13465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.811724522664552e-05,
+      "loss": 1.0395,
+      "step": 13466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810356039012812e-05,
+      "loss": 0.8453,
+      "step": 13467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.808987598431303e-05,
+      "loss": 0.9358,
+      "step": 13468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.807619200946942e-05,
+      "loss": 0.9035,
+      "step": 13469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80625084658665e-05,
+      "loss": 0.8508,
+      "step": 13470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.804882535377337e-05,
+      "loss": 1.0593,
+      "step": 13471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80351426734592e-05,
+      "loss": 1.0506,
+      "step": 13472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.802146042519313e-05,
+      "loss": 0.8681,
+      "step": 13473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.800777860924429e-05,
+      "loss": 0.9077,
+      "step": 13474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.799409722588176e-05,
+      "loss": 0.8833,
+      "step": 13475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.798041627537473e-05,
+      "loss": 0.9775,
+      "step": 13476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.796673575799224e-05,
+      "loss": 1.1556,
+      "step": 13477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79530556740034e-05,
+      "loss": 0.8147,
+      "step": 13478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.793937602367734e-05,
+      "loss": 0.8841,
+      "step": 13479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.792569680728305e-05,
+      "loss": 0.9737,
+      "step": 13480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791201802508973e-05,
+      "loss": 0.9949,
+      "step": 13481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.789833967736628e-05,
+      "loss": 0.8595,
+      "step": 13482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.788466176438187e-05,
+      "loss": 0.9666,
+      "step": 13483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.787098428640553e-05,
+      "loss": 0.9562,
+      "step": 13484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.785730724370625e-05,
+      "loss": 1.0498,
+      "step": 13485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.784363063655306e-05,
+      "loss": 1.0285,
+      "step": 13486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.782995446521506e-05,
+      "loss": 0.8855,
+      "step": 13487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.781627872996115e-05,
+      "loss": 0.9106,
+      "step": 13488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.780260343106041e-05,
+      "loss": 0.9359,
+      "step": 13489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.778892856878181e-05,
+      "loss": 1.0596,
+      "step": 13490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.777525414339431e-05,
+      "loss": 0.9657,
+      "step": 13491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.776158015516688e-05,
+      "loss": 1.1952,
+      "step": 13492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.774790660436858e-05,
+      "loss": 0.9517,
+      "step": 13493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.773423349126824e-05,
+      "loss": 0.8751,
+      "step": 13494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77205608161349e-05,
+      "loss": 1.0742,
+      "step": 13495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.770688857923748e-05,
+      "loss": 1.1062,
+      "step": 13496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.769321678084489e-05,
+      "loss": 0.8475,
+      "step": 13497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.767954542122605e-05,
+      "loss": 1.04,
+      "step": 13498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.766587450064991e-05,
+      "loss": 0.8978,
+      "step": 13499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.765220401938535e-05,
+      "loss": 1.0041,
+      "step": 13500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.763853397770131e-05,
+      "loss": 0.8786,
+      "step": 13501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.762486437586662e-05,
+      "loss": 0.9924,
+      "step": 13502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.761119521415016e-05,
+      "loss": 0.9357,
+      "step": 13503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75975264928209e-05,
+      "loss": 1.1026,
+      "step": 13504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.758385821214755e-05,
+      "loss": 0.9765,
+      "step": 13505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.757019037239908e-05,
+      "loss": 1.0357,
+      "step": 13506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.755652297384434e-05,
+      "loss": 0.9568,
+      "step": 13507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.754285601675208e-05,
+      "loss": 0.906,
+      "step": 13508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.752918950139119e-05,
+      "loss": 0.9264,
+      "step": 13509
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.751552342803053e-05,
+      "loss": 1.0982,
+      "step": 13510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750185779693877e-05,
+      "loss": 0.8946,
+      "step": 13511
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.748819260838486e-05,
+      "loss": 0.9793,
+      "step": 13512
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747452786263755e-05,
+      "loss": 0.9531,
+      "step": 13513
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.746086355996558e-05,
+      "loss": 0.9339,
+      "step": 13514
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.744719970063775e-05,
+      "loss": 0.9729,
+      "step": 13515
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.743353628492286e-05,
+      "loss": 0.8369,
+      "step": 13516
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.741987331308964e-05,
+      "loss": 0.7962,
+      "step": 13517
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.740621078540684e-05,
+      "loss": 0.8234,
+      "step": 13518
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.739254870214323e-05,
+      "loss": 1.0165,
+      "step": 13519
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73788870635675e-05,
+      "loss": 1.1048,
+      "step": 13520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.736522586994837e-05,
+      "loss": 1.0599,
+      "step": 13521
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.735156512155461e-05,
+      "loss": 0.9963,
+      "step": 13522
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.733790481865488e-05,
+      "loss": 0.9609,
+      "step": 13523
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.732424496151793e-05,
+      "loss": 0.7784,
+      "step": 13524
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731058555041237e-05,
+      "loss": 1.0723,
+      "step": 13525
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.729692658560691e-05,
+      "loss": 1.2126,
+      "step": 13526
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72832680673703e-05,
+      "loss": 1.0556,
+      "step": 13527
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.726960999597107e-05,
+      "loss": 1.0113,
+      "step": 13528
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.725595237167796e-05,
+      "loss": 0.8358,
+      "step": 13529
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.724229519475962e-05,
+      "loss": 1.1215,
+      "step": 13530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.722863846548464e-05,
+      "loss": 1.0184,
+      "step": 13531
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721498218412164e-05,
+      "loss": 0.7976,
+      "step": 13532
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.720132635093934e-05,
+      "loss": 1.0103,
+      "step": 13533
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718767096620621e-05,
+      "loss": 0.9966,
+      "step": 13534
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.717401603019094e-05,
+      "loss": 0.8983,
+      "step": 13535
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.716036154316213e-05,
+      "loss": 1.1015,
+      "step": 13536
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.714670750538831e-05,
+      "loss": 0.8275,
+      "step": 13537
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.713305391713806e-05,
+      "loss": 0.828,
+      "step": 13538
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.711940077867998e-05,
+      "loss": 1.0593,
+      "step": 13539
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.710574809028262e-05,
+      "loss": 0.8827,
+      "step": 13540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.709209585221451e-05,
+      "loss": 0.9856,
+      "step": 13541
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707844406474421e-05,
+      "loss": 1.0324,
+      "step": 13542
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.706479272814023e-05,
+      "loss": 0.9291,
+      "step": 13543
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.705114184267108e-05,
+      "loss": 0.9816,
+      "step": 13544
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.703749140860532e-05,
+      "loss": 1.0225,
+      "step": 13545
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.702384142621142e-05,
+      "loss": 0.8659,
+      "step": 13546
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.701019189575789e-05,
+      "loss": 0.9693,
+      "step": 13547
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69965428175132e-05,
+      "loss": 1.1134,
+      "step": 13548
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69828941917458e-05,
+      "loss": 1.205,
+      "step": 13549
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.696924601872426e-05,
+      "loss": 1.0293,
+      "step": 13550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69555982987169e-05,
+      "loss": 0.904,
+      "step": 13551
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.694195103199227e-05,
+      "loss": 0.8995,
+      "step": 13552
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69283042188188e-05,
+      "loss": 1.0229,
+      "step": 13553
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69146578594649e-05,
+      "loss": 0.897,
+      "step": 13554
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.690101195419896e-05,
+      "loss": 1.0384,
+      "step": 13555
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68873665032895e-05,
+      "loss": 0.9708,
+      "step": 13556
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.68737215070048e-05,
+      "loss": 0.9701,
+      "step": 13557
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686007696561333e-05,
+      "loss": 0.9334,
+      "step": 13558
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.684643287938348e-05,
+      "loss": 0.9656,
+      "step": 13559
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.683278924858359e-05,
+      "loss": 0.8824,
+      "step": 13560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.681914607348204e-05,
+      "loss": 0.9033,
+      "step": 13561
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.680550335434724e-05,
+      "loss": 1.157,
+      "step": 13562
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679186109144749e-05,
+      "loss": 1.0014,
+      "step": 13563
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.677821928505114e-05,
+      "loss": 0.9315,
+      "step": 13564
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.676457793542654e-05,
+      "loss": 1.2981,
+      "step": 13565
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6750937042842e-05,
+      "loss": 0.8915,
+      "step": 13566
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.673729660756585e-05,
+      "loss": 1.2251,
+      "step": 13567
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.672365662986636e-05,
+      "loss": 1.0253,
+      "step": 13568
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.671001711001188e-05,
+      "loss": 0.8621,
+      "step": 13569
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.669637804827069e-05,
+      "loss": 1.1373,
+      "step": 13570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.668273944491103e-05,
+      "loss": 1.046,
+      "step": 13571
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.666910130020119e-05,
+      "loss": 1.2961,
+      "step": 13572
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66554636144095e-05,
+      "loss": 0.7915,
+      "step": 13573
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664182638780407e-05,
+      "loss": 0.863,
+      "step": 13574
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.662818962065326e-05,
+      "loss": 0.9865,
+      "step": 13575
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.661455331322528e-05,
+      "loss": 1.0048,
+      "step": 13576
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.660091746578835e-05,
+      "loss": 0.8917,
+      "step": 13577
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.658728207861064e-05,
+      "loss": 0.8677,
+      "step": 13578
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.657364715196047e-05,
+      "loss": 1.0836,
+      "step": 13579
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65600126861059e-05,
+      "loss": 0.9699,
+      "step": 13580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.654637868131521e-05,
+      "loss": 0.9429,
+      "step": 13581
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653274513785659e-05,
+      "loss": 1.072,
+      "step": 13582
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.651911205599815e-05,
+      "loss": 1.1074,
+      "step": 13583
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.650547943600807e-05,
+      "loss": 0.9809,
+      "step": 13584
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.649184727815455e-05,
+      "loss": 0.9735,
+      "step": 13585
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.647821558270568e-05,
+      "loss": 0.9523,
+      "step": 13586
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.646458434992961e-05,
+      "loss": 1.2124,
+      "step": 13587
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645095358009451e-05,
+      "loss": 1.0681,
+      "step": 13588
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643732327346842e-05,
+      "loss": 0.9221,
+      "step": 13589
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.642369343031951e-05,
+      "loss": 0.9355,
+      "step": 13590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.641006405091583e-05,
+      "loss": 1.0022,
+      "step": 13591
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.639643513552548e-05,
+      "loss": 0.9487,
+      "step": 13592
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63828066844166e-05,
+      "loss": 0.9796,
+      "step": 13593
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.636917869785718e-05,
+      "loss": 1.0677,
+      "step": 13594
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63555511761153e-05,
+      "loss": 0.915,
+      "step": 13595
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63419241194591e-05,
+      "loss": 0.9379,
+      "step": 13596
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63282975281565e-05,
+      "loss": 1.087,
+      "step": 13597
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63146714024756e-05,
+      "loss": 1.0104,
+      "step": 13598
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.630104574268442e-05,
+      "loss": 0.94,
+      "step": 13599
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.628742054905096e-05,
+      "loss": 0.8449,
+      "step": 13600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.627379582184322e-05,
+      "loss": 0.9197,
+      "step": 13601
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.626017156132928e-05,
+      "loss": 0.769,
+      "step": 13602
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6246547767777e-05,
+      "loss": 1.0805,
+      "step": 13603
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.623292444145444e-05,
+      "loss": 0.8777,
+      "step": 13604
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.621930158262958e-05,
+      "loss": 0.9697,
+      "step": 13605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.620567919157033e-05,
+      "loss": 0.8077,
+      "step": 13606
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.619205726854466e-05,
+      "loss": 1.1082,
+      "step": 13607
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.617843581382055e-05,
+      "loss": 1.1165,
+      "step": 13608
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.616481482766588e-05,
+      "loss": 0.9095,
+      "step": 13609
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.615119431034861e-05,
+      "loss": 1.0572,
+      "step": 13610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.613757426213666e-05,
+      "loss": 0.9192,
+      "step": 13611
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.61239546832979e-05,
+      "loss": 1.1028,
+      "step": 13612
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.611033557410026e-05,
+      "loss": 0.9877,
+      "step": 13613
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.609671693481159e-05,
+      "loss": 0.8156,
+      "step": 13614
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.608309876569978e-05,
+      "loss": 0.9001,
+      "step": 13615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.606948106703276e-05,
+      "loss": 0.874,
+      "step": 13616
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605586383907833e-05,
+      "loss": 1.0541,
+      "step": 13617
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.60422470821043e-05,
+      "loss": 1.0218,
+      "step": 13618
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.602863079637864e-05,
+      "loss": 0.836,
+      "step": 13619
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.601501498216904e-05,
+      "loss": 1.1487,
+      "step": 13620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600139963974341e-05,
+      "loss": 0.8883,
+      "step": 13621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.598778476936957e-05,
+      "loss": 0.9546,
+      "step": 13622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.597417037131526e-05,
+      "loss": 1.0367,
+      "step": 13623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.596055644584829e-05,
+      "loss": 1.0734,
+      "step": 13624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.594694299323652e-05,
+      "loss": 0.8232,
+      "step": 13625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.593333001374763e-05,
+      "loss": 0.8752,
+      "step": 13626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591971750764943e-05,
+      "loss": 1.0124,
+      "step": 13627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.59061054752097e-05,
+      "loss": 1.0561,
+      "step": 13628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589249391669616e-05,
+      "loss": 0.8062,
+      "step": 13629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.58788828323765e-05,
+      "loss": 1.0142,
+      "step": 13630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.586527222251857e-05,
+      "loss": 1.0275,
+      "step": 13631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.585166208739e-05,
+      "loss": 0.9526,
+      "step": 13632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.583805242725853e-05,
+      "loss": 0.9053,
+      "step": 13633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582444324239186e-05,
+      "loss": 1.1012,
+      "step": 13634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.581083453305766e-05,
+      "loss": 1.0602,
+      "step": 13635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.579722629952364e-05,
+      "loss": 0.9712,
+      "step": 13636
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.578361854205743e-05,
+      "loss": 1.1532,
+      "step": 13637
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.577001126092675e-05,
+      "loss": 1.1158,
+      "step": 13638
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.575640445639926e-05,
+      "loss": 1.1189,
+      "step": 13639
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.574279812874256e-05,
+      "loss": 0.964,
+      "step": 13640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.572919227822429e-05,
+      "loss": 0.8933,
+      "step": 13641
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.571558690511213e-05,
+      "loss": 1.0719,
+      "step": 13642
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.570198200967362e-05,
+      "loss": 0.9453,
+      "step": 13643
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568837759217642e-05,
+      "loss": 0.9653,
+      "step": 13644
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.567477365288813e-05,
+      "loss": 0.9296,
+      "step": 13645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.56611701920763e-05,
+      "loss": 0.8765,
+      "step": 13646
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.564756721000852e-05,
+      "loss": 1.0155,
+      "step": 13647
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.563396470695244e-05,
+      "loss": 0.8884,
+      "step": 13648
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.562036268317548e-05,
+      "loss": 1.0287,
+      "step": 13649
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.56067611389453e-05,
+      "loss": 0.9263,
+      "step": 13650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559316007452942e-05,
+      "loss": 1.1217,
+      "step": 13651
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.557955949019535e-05,
+      "loss": 0.9587,
+      "step": 13652
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.556595938621058e-05,
+      "loss": 1.0176,
+      "step": 13653
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.555235976284272e-05,
+      "loss": 0.8487,
+      "step": 13654
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.553876062035921e-05,
+      "loss": 1.0952,
+      "step": 13655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.552516195902758e-05,
+      "loss": 0.9611,
+      "step": 13656
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.551156377911526e-05,
+      "loss": 0.9808,
+      "step": 13657
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.549796608088976e-05,
+      "loss": 1.0961,
+      "step": 13658
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.548436886461858e-05,
+      "loss": 0.9492,
+      "step": 13659
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.547077213056908e-05,
+      "loss": 0.9499,
+      "step": 13660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.545717587900879e-05,
+      "loss": 0.86,
+      "step": 13661
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.544358011020514e-05,
+      "loss": 1.0121,
+      "step": 13662
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.542998482442554e-05,
+      "loss": 1.0049,
+      "step": 13663
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.541639002193739e-05,
+      "loss": 0.9512,
+      "step": 13664
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540279570300818e-05,
+      "loss": 0.8649,
+      "step": 13665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.53892018679052e-05,
+      "loss": 0.9051,
+      "step": 13666
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.537560851689589e-05,
+      "loss": 0.8995,
+      "step": 13667
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.536201565024767e-05,
+      "loss": 0.9071,
+      "step": 13668
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.534842326822785e-05,
+      "loss": 0.8911,
+      "step": 13669
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.53348313711038e-05,
+      "loss": 0.9568,
+      "step": 13670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532123995914295e-05,
+      "loss": 0.9495,
+      "step": 13671
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.530764903261251e-05,
+      "loss": 1.0348,
+      "step": 13672
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.529405859177992e-05,
+      "loss": 0.9155,
+      "step": 13673
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.528046863691248e-05,
+      "loss": 1.0884,
+      "step": 13674
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.526687916827747e-05,
+      "loss": 0.9576,
+      "step": 13675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.525329018614224e-05,
+      "loss": 0.8239,
+      "step": 13676
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.523970169077402e-05,
+      "loss": 0.8809,
+      "step": 13677
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.522611368244016e-05,
+      "loss": 0.9586,
+      "step": 13678
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521252616140792e-05,
+      "loss": 1.0645,
+      "step": 13679
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.519893912794455e-05,
+      "loss": 1.1488,
+      "step": 13680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.518535258231731e-05,
+      "loss": 1.0105,
+      "step": 13681
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.517176652479348e-05,
+      "loss": 0.8708,
+      "step": 13682
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.515818095564023e-05,
+      "loss": 1.008,
+      "step": 13683
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514459587512485e-05,
+      "loss": 0.9708,
+      "step": 13684
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.513101128351454e-05,
+      "loss": 0.9903,
+      "step": 13685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.511742718107649e-05,
+      "loss": 1.1134,
+      "step": 13686
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.510384356807789e-05,
+      "loss": 0.9421,
+      "step": 13687
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.509026044478602e-05,
+      "loss": 1.0328,
+      "step": 13688
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.507667781146792e-05,
+      "loss": 1.0527,
+      "step": 13689
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.506309566839088e-05,
+      "loss": 1.1176,
+      "step": 13690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.504951401582199e-05,
+      "loss": 0.876,
+      "step": 13691
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.503593285402843e-05,
+      "loss": 1.1021,
+      "step": 13692
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.502235218327731e-05,
+      "loss": 1.0168,
+      "step": 13693
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.500877200383584e-05,
+      "loss": 0.9244,
+      "step": 13694
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.499519231597103e-05,
+      "loss": 1.075,
+      "step": 13695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.498161311995007e-05,
+      "loss": 1.132,
+      "step": 13696
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.496803441604005e-05,
+      "loss": 0.9577,
+      "step": 13697
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.495445620450805e-05,
+      "loss": 0.8422,
+      "step": 13698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.494087848562117e-05,
+      "loss": 1.0383,
+      "step": 13699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.492730125964642e-05,
+      "loss": 0.8359,
+      "step": 13700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.491372452685094e-05,
+      "loss": 1.1489,
+      "step": 13701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.490014828750179e-05,
+      "loss": 0.9185,
+      "step": 13702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488657254186596e-05,
+      "loss": 0.8965,
+      "step": 13703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48729972902105e-05,
+      "loss": 0.8824,
+      "step": 13704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.485942253280246e-05,
+      "loss": 1.1415,
+      "step": 13705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.484584826990879e-05,
+      "loss": 1.0006,
+      "step": 13706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.483227450179657e-05,
+      "loss": 1.0316,
+      "step": 13707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48187012287328e-05,
+      "loss": 0.9312,
+      "step": 13708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48051284509844e-05,
+      "loss": 0.7931,
+      "step": 13709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479155616881834e-05,
+      "loss": 0.9434,
+      "step": 13710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.477798438250172e-05,
+      "loss": 0.9469,
+      "step": 13711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476441309230134e-05,
+      "loss": 1.131,
+      "step": 13712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.47508422984842e-05,
+      "loss": 1.0269,
+      "step": 13713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473727200131729e-05,
+      "loss": 1.0849,
+      "step": 13714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.472370220106747e-05,
+      "loss": 1.1199,
+      "step": 13715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471013289800167e-05,
+      "loss": 0.872,
+      "step": 13716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.469656409238685e-05,
+      "loss": 0.9914,
+      "step": 13717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.46829957844898e-05,
+      "loss": 0.8329,
+      "step": 13718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.466942797457752e-05,
+      "loss": 1.141,
+      "step": 13719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.465586066291685e-05,
+      "loss": 0.8942,
+      "step": 13720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.464229384977465e-05,
+      "loss": 1.0484,
+      "step": 13721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.462872753541779e-05,
+      "loss": 0.8671,
+      "step": 13722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.461516172011307e-05,
+      "loss": 0.9062,
+      "step": 13723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.460159640412738e-05,
+      "loss": 0.9115,
+      "step": 13724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.458803158772758e-05,
+      "loss": 0.9415,
+      "step": 13725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.45744672711804e-05,
+      "loss": 0.9406,
+      "step": 13726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.456090345475272e-05,
+      "loss": 1.0837,
+      "step": 13727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.454734013871133e-05,
+      "loss": 1.2074,
+      "step": 13728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.453377732332298e-05,
+      "loss": 0.91,
+      "step": 13729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452021500885448e-05,
+      "loss": 0.8621,
+      "step": 13730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.450665319557262e-05,
+      "loss": 1.075,
+      "step": 13731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.449309188374414e-05,
+      "loss": 1.0106,
+      "step": 13732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.447953107363574e-05,
+      "loss": 0.9346,
+      "step": 13733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.446597076551429e-05,
+      "loss": 1.1445,
+      "step": 13734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.445241095964637e-05,
+      "loss": 0.9905,
+      "step": 13735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443885165629881e-05,
+      "loss": 0.9297,
+      "step": 13736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.44252928557383e-05,
+      "loss": 0.8939,
+      "step": 13737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.441173455823148e-05,
+      "loss": 0.7482,
+      "step": 13738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439817676404508e-05,
+      "loss": 0.8875,
+      "step": 13739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.438461947344586e-05,
+      "loss": 1.0261,
+      "step": 13740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.437106268670034e-05,
+      "loss": 0.8788,
+      "step": 13741
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.435750640407531e-05,
+      "loss": 1.0905,
+      "step": 13742
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.434395062583734e-05,
+      "loss": 1.037,
+      "step": 13743
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.433039535225312e-05,
+      "loss": 0.9398,
+      "step": 13744
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.431684058358928e-05,
+      "loss": 1.1569,
+      "step": 13745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.430328632011239e-05,
+      "loss": 1.1267,
+      "step": 13746
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.428973256208911e-05,
+      "loss": 0.8798,
+      "step": 13747
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.427617930978606e-05,
+      "loss": 0.9137,
+      "step": 13748
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.426262656346978e-05,
+      "loss": 1.0233,
+      "step": 13749
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.424907432340686e-05,
+      "loss": 0.8912,
+      "step": 13750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.423552258986394e-05,
+      "loss": 1.0276,
+      "step": 13751
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.422197136310747e-05,
+      "loss": 0.7147,
+      "step": 13752
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420842064340408e-05,
+      "loss": 0.9531,
+      "step": 13753
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.419487043102033e-05,
+      "loss": 1.0185,
+      "step": 13754
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.418132072622268e-05,
+      "loss": 1.0863,
+      "step": 13755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.416777152927768e-05,
+      "loss": 1.0224,
+      "step": 13756
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.41542228404519e-05,
+      "loss": 0.9961,
+      "step": 13757
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.414067466001174e-05,
+      "loss": 0.9653,
+      "step": 13758
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.412712698822377e-05,
+      "loss": 0.8999,
+      "step": 13759
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.411357982535448e-05,
+      "loss": 1.0188,
+      "step": 13760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410003317167026e-05,
+      "loss": 1.0357,
+      "step": 13761
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.408648702743761e-05,
+      "loss": 1.0252,
+      "step": 13762
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.407294139292307e-05,
+      "loss": 0.8881,
+      "step": 13763
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.405939626839293e-05,
+      "loss": 0.8797,
+      "step": 13764
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.404585165411377e-05,
+      "loss": 0.8895,
+      "step": 13765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.403230755035188e-05,
+      "loss": 0.9685,
+      "step": 13766
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.401876395737377e-05,
+      "loss": 1.1447,
+      "step": 13767
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.400522087544582e-05,
+      "loss": 1.0649,
+      "step": 13768
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.399167830483434e-05,
+      "loss": 0.9282,
+      "step": 13769
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.397813624580583e-05,
+      "loss": 1.0136,
+      "step": 13770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.396459469862662e-05,
+      "loss": 0.9507,
+      "step": 13771
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.395105366356304e-05,
+      "loss": 0.9984,
+      "step": 13772
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.393751314088146e-05,
+      "loss": 0.972,
+      "step": 13773
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.392397313084827e-05,
+      "loss": 0.985,
+      "step": 13774
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.391043363372969e-05,
+      "loss": 0.9467,
+      "step": 13775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.389689464979214e-05,
+      "loss": 0.8663,
+      "step": 13776
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.388335617930194e-05,
+      "loss": 1.1034,
+      "step": 13777
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.386981822252532e-05,
+      "loss": 1.0232,
+      "step": 13778
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.385628077972857e-05,
+      "loss": 0.9364,
+      "step": 13779
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.384274385117808e-05,
+      "loss": 1.029,
+      "step": 13780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.382920743713999e-05,
+      "loss": 0.8153,
+      "step": 13781
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381567153788064e-05,
+      "loss": 0.7834,
+      "step": 13782
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.380213615366627e-05,
+      "loss": 0.7925,
+      "step": 13783
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.378860128476309e-05,
+      "loss": 0.8333,
+      "step": 13784
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.377506693143733e-05,
+      "loss": 0.9665,
+      "step": 13785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.376153309395528e-05,
+      "loss": 1.0336,
+      "step": 13786
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.374799977258304e-05,
+      "loss": 1.091,
+      "step": 13787
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.373446696758692e-05,
+      "loss": 0.8851,
+      "step": 13788
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.372093467923302e-05,
+      "loss": 1.1452,
+      "step": 13789
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.370740290778758e-05,
+      "loss": 0.84,
+      "step": 13790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.369387165351675e-05,
+      "loss": 0.7585,
+      "step": 13791
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.368034091668667e-05,
+      "loss": 0.9846,
+      "step": 13792
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.366681069756352e-05,
+      "loss": 0.9117,
+      "step": 13793
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.365328099641344e-05,
+      "loss": 1.0845,
+      "step": 13794
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.363975181350252e-05,
+      "loss": 1.1183,
+      "step": 13795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.36262231490969e-05,
+      "loss": 0.9529,
+      "step": 13796
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.361269500346274e-05,
+      "loss": 0.8999,
+      "step": 13797
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.359916737686603e-05,
+      "loss": 0.9845,
+      "step": 13798
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.358564026957294e-05,
+      "loss": 0.928,
+      "step": 13799
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.357211368184956e-05,
+      "loss": 0.9185,
+      "step": 13800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.355858761396189e-05,
+      "loss": 1.0003,
+      "step": 13801
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.354506206617598e-05,
+      "loss": 1.016,
+      "step": 13802
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.353153703875801e-05,
+      "loss": 1.0504,
+      "step": 13803
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.351801253197385e-05,
+      "loss": 0.953,
+      "step": 13804
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.350448854608962e-05,
+      "loss": 1.1891,
+      "step": 13805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.349096508137135e-05,
+      "loss": 0.9557,
+      "step": 13806
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.347744213808499e-05,
+      "loss": 0.934,
+      "step": 13807
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.346391971649658e-05,
+      "loss": 1.0075,
+      "step": 13808
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.345039781687206e-05,
+      "loss": 1.117,
+      "step": 13809
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.343687643947741e-05,
+      "loss": 0.9523,
+      "step": 13810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.342335558457867e-05,
+      "loss": 1.0267,
+      "step": 13811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.34098352524417e-05,
+      "loss": 0.8139,
+      "step": 13812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.339631544333249e-05,
+      "loss": 0.9269,
+      "step": 13813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.338279615751701e-05,
+      "loss": 0.9332,
+      "step": 13814
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.33692773952611e-05,
+      "loss": 0.8598,
+      "step": 13815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.33557591568307e-05,
+      "loss": 1.0028,
+      "step": 13816
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.33422414424918e-05,
+      "loss": 0.9603,
+      "step": 13817
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.332872425251018e-05,
+      "loss": 0.9344,
+      "step": 13818
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.331520758715176e-05,
+      "loss": 0.9673,
+      "step": 13819
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.330169144668246e-05,
+      "loss": 0.8948,
+      "step": 13820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.328817583136804e-05,
+      "loss": 0.9681,
+      "step": 13821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.327466074147445e-05,
+      "loss": 0.779,
+      "step": 13822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326114617726749e-05,
+      "loss": 0.9529,
+      "step": 13823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3247632139013e-05,
+      "loss": 1.0341,
+      "step": 13824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.323411862697676e-05,
+      "loss": 1.1518,
+      "step": 13825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.322060564142467e-05,
+      "loss": 0.9239,
+      "step": 13826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.320709318262244e-05,
+      "loss": 0.9252,
+      "step": 13827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31935812508359e-05,
+      "loss": 1.0071,
+      "step": 13828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.318006984633085e-05,
+      "loss": 1.0391,
+      "step": 13829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.316655896937302e-05,
+      "loss": 0.816,
+      "step": 13830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31530486202282e-05,
+      "loss": 1.0787,
+      "step": 13831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.31395387991621e-05,
+      "loss": 1.0035,
+      "step": 13832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.312602950644045e-05,
+      "loss": 1.0038,
+      "step": 13833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.311252074232906e-05,
+      "loss": 1.0016,
+      "step": 13834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.309901250709357e-05,
+      "loss": 0.9666,
+      "step": 13835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.308550480099969e-05,
+      "loss": 0.8668,
+      "step": 13836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.307199762431319e-05,
+      "loss": 0.8563,
+      "step": 13837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.305849097729963e-05,
+      "loss": 0.8767,
+      "step": 13838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.30449848602248e-05,
+      "loss": 1.2238,
+      "step": 13839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.303147927335432e-05,
+      "loss": 0.8713,
+      "step": 13840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.301797421695385e-05,
+      "loss": 1.0787,
+      "step": 13841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.300446969128902e-05,
+      "loss": 1.0221,
+      "step": 13842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29909656966255e-05,
+      "loss": 0.9619,
+      "step": 13843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.297746223322883e-05,
+      "loss": 1.0366,
+      "step": 13844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29639593013647e-05,
+      "loss": 1.091,
+      "step": 13845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.295045690129873e-05,
+      "loss": 1.0021,
+      "step": 13846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293695503329645e-05,
+      "loss": 0.8492,
+      "step": 13847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.292345369762344e-05,
+      "loss": 0.9612,
+      "step": 13848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.290995289454536e-05,
+      "loss": 1.1193,
+      "step": 13849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.289645262432765e-05,
+      "loss": 0.8831,
+      "step": 13850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.288295288723594e-05,
+      "loss": 0.9496,
+      "step": 13851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.286945368353575e-05,
+      "loss": 1.1134,
+      "step": 13852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285595501349258e-05,
+      "loss": 1.0495,
+      "step": 13853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.284245687737201e-05,
+      "loss": 0.9646,
+      "step": 13854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.282895927543949e-05,
+      "loss": 1.2116,
+      "step": 13855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.281546220796052e-05,
+      "loss": 0.8424,
+      "step": 13856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.280196567520063e-05,
+      "loss": 0.9136,
+      "step": 13857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.278846967742527e-05,
+      "loss": 1.0398,
+      "step": 13858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27749742148999e-05,
+      "loss": 1.1428,
+      "step": 13859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.276147928789e-05,
+      "loss": 1.0021,
+      "step": 13860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.274798489666097e-05,
+      "loss": 0.9379,
+      "step": 13861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27344910414783e-05,
+      "loss": 0.9185,
+      "step": 13862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.272099772260738e-05,
+      "loss": 0.7957,
+      "step": 13863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.270750494031362e-05,
+      "loss": 1.0686,
+      "step": 13864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.269401269486242e-05,
+      "loss": 0.9885,
+      "step": 13865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.268052098651923e-05,
+      "loss": 0.9277,
+      "step": 13866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266702981554934e-05,
+      "loss": 0.951,
+      "step": 13867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.265353918221817e-05,
+      "loss": 1.1076,
+      "step": 13868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.264004908679112e-05,
+      "loss": 0.8879,
+      "step": 13869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.262655952953346e-05,
+      "loss": 0.8554,
+      "step": 13870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.261307051071054e-05,
+      "loss": 0.8456,
+      "step": 13871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.259958203058779e-05,
+      "loss": 1.1111,
+      "step": 13872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.258609408943038e-05,
+      "loss": 0.9922,
+      "step": 13873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.257260668750374e-05,
+      "loss": 1.0459,
+      "step": 13874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.25591198250731e-05,
+      "loss": 1.099,
+      "step": 13875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.254563350240375e-05,
+      "loss": 0.8953,
+      "step": 13876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2532147719761e-05,
+      "loss": 1.1554,
+      "step": 13877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251866247741009e-05,
+      "loss": 0.8209,
+      "step": 13878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.250517777561623e-05,
+      "loss": 1.203,
+      "step": 13879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249169361464477e-05,
+      "loss": 0.9082,
+      "step": 13880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.247820999476086e-05,
+      "loss": 0.8857,
+      "step": 13881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.246472691622974e-05,
+      "loss": 0.8789,
+      "step": 13882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.245124437931666e-05,
+      "loss": 0.9433,
+      "step": 13883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.243776238428675e-05,
+      "loss": 0.8768,
+      "step": 13884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.242428093140525e-05,
+      "loss": 0.9811,
+      "step": 13885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.241080002093736e-05,
+      "loss": 0.9837,
+      "step": 13886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.23973196531482e-05,
+      "loss": 0.9291,
+      "step": 13887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.238383982830292e-05,
+      "loss": 0.9005,
+      "step": 13888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.237036054666675e-05,
+      "loss": 1.1581,
+      "step": 13889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.235688180850471e-05,
+      "loss": 1.0045,
+      "step": 13890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.234340361408202e-05,
+      "loss": 0.8384,
+      "step": 13891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.23299259636638e-05,
+      "loss": 0.9095,
+      "step": 13892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.231644885751507e-05,
+      "loss": 0.9891,
+      "step": 13893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230297229590098e-05,
+      "loss": 0.9279,
+      "step": 13894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.228949627908663e-05,
+      "loss": 0.9492,
+      "step": 13895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.227602080733704e-05,
+      "loss": 1.0512,
+      "step": 13896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.226254588091734e-05,
+      "loss": 1.0992,
+      "step": 13897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.224907150009251e-05,
+      "loss": 0.9453,
+      "step": 13898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.223559766512764e-05,
+      "loss": 1.2,
+      "step": 13899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.222212437628775e-05,
+      "loss": 0.6884,
+      "step": 13900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.220865163383784e-05,
+      "loss": 0.926,
+      "step": 13901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.21951794380429e-05,
+      "loss": 1.0951,
+      "step": 13902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.218170778916802e-05,
+      "loss": 0.9511,
+      "step": 13903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.216823668747809e-05,
+      "loss": 1.1344,
+      "step": 13904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.215476613323814e-05,
+      "loss": 0.9654,
+      "step": 13905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.214129612671312e-05,
+      "loss": 0.8282,
+      "step": 13906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.212782666816796e-05,
+      "loss": 0.687,
+      "step": 13907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.211435775786764e-05,
+      "loss": 1.06,
+      "step": 13908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.210088939607708e-05,
+      "loss": 1.0011,
+      "step": 13909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.208742158306122e-05,
+      "loss": 0.8772,
+      "step": 13910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.20739543190849e-05,
+      "loss": 0.8563,
+      "step": 13911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.206048760441316e-05,
+      "loss": 1.092,
+      "step": 13912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.204702143931072e-05,
+      "loss": 0.8992,
+      "step": 13913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.203355582404256e-05,
+      "loss": 1.1054,
+      "step": 13914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.202009075887357e-05,
+      "loss": 0.9786,
+      "step": 13915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.200662624406854e-05,
+      "loss": 1.0414,
+      "step": 13916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.199316227989236e-05,
+      "loss": 1.307,
+      "step": 13917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.197969886660984e-05,
+      "loss": 1.0314,
+      "step": 13918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.196623600448579e-05,
+      "loss": 1.0511,
+      "step": 13919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.19527736937851e-05,
+      "loss": 0.8026,
+      "step": 13920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.19393119347725e-05,
+      "loss": 1.0714,
+      "step": 13921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.19258507277128e-05,
+      "loss": 0.9017,
+      "step": 13922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191239007287081e-05,
+      "loss": 0.9097,
+      "step": 13923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.189892997051127e-05,
+      "loss": 0.8389,
+      "step": 13924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.18854704208989e-05,
+      "loss": 0.9314,
+      "step": 13925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.187201142429855e-05,
+      "loss": 1.2077,
+      "step": 13926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.18585529809749e-05,
+      "loss": 1.0562,
+      "step": 13927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.184509509119267e-05,
+      "loss": 0.8954,
+      "step": 13928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.183163775521662e-05,
+      "loss": 0.9657,
+      "step": 13929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.181818097331137e-05,
+      "loss": 1.3147,
+      "step": 13930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180472474574168e-05,
+      "loss": 0.9666,
+      "step": 13931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.179126907277225e-05,
+      "loss": 0.9721,
+      "step": 13932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.177781395466772e-05,
+      "loss": 1.0066,
+      "step": 13933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.17643593916927e-05,
+      "loss": 1.0687,
+      "step": 13934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1750905384112e-05,
+      "loss": 0.8844,
+      "step": 13935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.173745193219007e-05,
+      "loss": 0.9058,
+      "step": 13936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.172399903619164e-05,
+      "loss": 1.0805,
+      "step": 13937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.171054669638135e-05,
+      "loss": 1.233,
+      "step": 13938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.169709491302376e-05,
+      "loss": 1.0199,
+      "step": 13939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.168364368638348e-05,
+      "loss": 0.9531,
+      "step": 13940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.167019301672509e-05,
+      "loss": 0.6826,
+      "step": 13941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.165674290431312e-05,
+      "loss": 1.063,
+      "step": 13942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.164329334941224e-05,
+      "loss": 0.9491,
+      "step": 13943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.162984435228692e-05,
+      "loss": 0.979,
+      "step": 13944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.16163959132017e-05,
+      "loss": 0.8676,
+      "step": 13945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160294803242119e-05,
+      "loss": 0.9931,
+      "step": 13946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.158950071020982e-05,
+      "loss": 0.8644,
+      "step": 13947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.157605394683212e-05,
+      "loss": 0.8046,
+      "step": 13948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.156260774255262e-05,
+      "loss": 0.8352,
+      "step": 13949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.154916209763579e-05,
+      "loss": 0.932,
+      "step": 13950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.153571701234607e-05,
+      "loss": 0.9767,
+      "step": 13951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1522272486948e-05,
+      "loss": 0.852,
+      "step": 13952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.150882852170595e-05,
+      "loss": 1.0319,
+      "step": 13953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14953851168844e-05,
+      "loss": 0.8043,
+      "step": 13954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.148194227274783e-05,
+      "loss": 0.9135,
+      "step": 13955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.146849998956058e-05,
+      "loss": 0.8389,
+      "step": 13956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.145505826758707e-05,
+      "loss": 1.0127,
+      "step": 13957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.144161710709178e-05,
+      "loss": 0.9435,
+      "step": 13958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1428176508339e-05,
+      "loss": 0.7218,
+      "step": 13959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.141473647159315e-05,
+      "loss": 0.8458,
+      "step": 13960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14012969971186e-05,
+      "loss": 1.0542,
+      "step": 13961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.138785808517969e-05,
+      "loss": 0.9484,
+      "step": 13962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.137441973604078e-05,
+      "loss": 1.0096,
+      "step": 13963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.136098194996617e-05,
+      "loss": 1.1418,
+      "step": 13964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.134754472722017e-05,
+      "loss": 0.8767,
+      "step": 13965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133410806806717e-05,
+      "loss": 0.9043,
+      "step": 13966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132067197277139e-05,
+      "loss": 0.9263,
+      "step": 13967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.130723644159715e-05,
+      "loss": 1.1697,
+      "step": 13968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.129380147480874e-05,
+      "loss": 0.8131,
+      "step": 13969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.128036707267038e-05,
+      "loss": 0.9191,
+      "step": 13970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.126693323544635e-05,
+      "loss": 0.9755,
+      "step": 13971
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.125349996340091e-05,
+      "loss": 1.2159,
+      "step": 13972
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124006725679828e-05,
+      "loss": 0.928,
+      "step": 13973
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.122663511590267e-05,
+      "loss": 0.8555,
+      "step": 13974
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.121320354097833e-05,
+      "loss": 1.0136,
+      "step": 13975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.119977253228937e-05,
+      "loss": 1.1258,
+      "step": 13976
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.118634209010005e-05,
+      "loss": 0.9406,
+      "step": 13977
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.117291221467457e-05,
+      "loss": 1.0287,
+      "step": 13978
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.115948290627702e-05,
+      "loss": 0.9892,
+      "step": 13979
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.114605416517157e-05,
+      "loss": 0.8452,
+      "step": 13980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.113262599162245e-05,
+      "loss": 0.9486,
+      "step": 13981
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.111919838589366e-05,
+      "loss": 1.1004,
+      "step": 13982
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.110577134824945e-05,
+      "loss": 0.9783,
+      "step": 13983
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.10923448789538e-05,
+      "loss": 0.8062,
+      "step": 13984
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.107891897827088e-05,
+      "loss": 1.0788,
+      "step": 13985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.106549364646481e-05,
+      "loss": 0.8055,
+      "step": 13986
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.105206888379961e-05,
+      "loss": 0.948,
+      "step": 13987
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.103864469053931e-05,
+      "loss": 0.9503,
+      "step": 13988
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.102522106694808e-05,
+      "loss": 0.9815,
+      "step": 13989
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.101179801328986e-05,
+      "loss": 1.1356,
+      "step": 13990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.099837552982874e-05,
+      "loss": 0.9695,
+      "step": 13991
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.098495361682871e-05,
+      "loss": 0.9582,
+      "step": 13992
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.097153227455379e-05,
+      "loss": 1.0485,
+      "step": 13993
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.095811150326793e-05,
+      "loss": 0.9455,
+      "step": 13994
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.094469130323523e-05,
+      "loss": 0.9359,
+      "step": 13995
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.093127167471953e-05,
+      "loss": 0.9101,
+      "step": 13996
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.09178526179849e-05,
+      "loss": 0.8924,
+      "step": 13997
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.090443413329524e-05,
+      "loss": 0.9505,
+      "step": 13998
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.089101622091447e-05,
+      "loss": 0.9761,
+      "step": 13999
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.087759888110659e-05,
+      "loss": 0.8426,
+      "step": 14000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.086418211413548e-05,
+      "loss": 0.8619,
+      "step": 14001
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085076592026504e-05,
+      "loss": 0.9847,
+      "step": 14002
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.083735029975914e-05,
+      "loss": 1.1258,
+      "step": 14003
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.082393525288177e-05,
+      "loss": 0.8246,
+      "step": 14004
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.081052077989667e-05,
+      "loss": 0.9768,
+      "step": 14005
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.079710688106782e-05,
+      "loss": 0.9436,
+      "step": 14006
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.078369355665895e-05,
+      "loss": 1.0625,
+      "step": 14007
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0770280806934e-05,
+      "loss": 1.0523,
+      "step": 14008
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.075686863215677e-05,
+      "loss": 0.9831,
+      "step": 14009
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.074345703259106e-05,
+      "loss": 1.1843,
+      "step": 14010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.073004600850066e-05,
+      "loss": 1.0077,
+      "step": 14011
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.071663556014945e-05,
+      "loss": 0.8147,
+      "step": 14012
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07032256878011e-05,
+      "loss": 0.9721,
+      "step": 14013
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.068981639171946e-05,
+      "loss": 0.8777,
+      "step": 14014
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.067640767216828e-05,
+      "loss": 1.1993,
+      "step": 14015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.066299952941126e-05,
+      "loss": 1.0037,
+      "step": 14016
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.064959196371219e-05,
+      "loss": 0.9151,
+      "step": 14017
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.063618497533479e-05,
+      "loss": 0.8932,
+      "step": 14018
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.062277856454275e-05,
+      "loss": 0.8271,
+      "step": 14019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.06093727315998e-05,
+      "loss": 0.9693,
+      "step": 14020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.059596747676962e-05,
+      "loss": 1.0604,
+      "step": 14021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.058256280031587e-05,
+      "loss": 1.0518,
+      "step": 14022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.056915870250225e-05,
+      "loss": 0.9299,
+      "step": 14023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.055575518359244e-05,
+      "loss": 0.9166,
+      "step": 14024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.054235224385002e-05,
+      "loss": 0.9284,
+      "step": 14025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.05289498835387e-05,
+      "loss": 1.1945,
+      "step": 14026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.051554810292203e-05,
+      "loss": 0.921,
+      "step": 14027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.050214690226364e-05,
+      "loss": 1.09,
+      "step": 14028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048874628182722e-05,
+      "loss": 1.0129,
+      "step": 14029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.047534624187621e-05,
+      "loss": 0.9505,
+      "step": 14030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.046194678267432e-05,
+      "loss": 1.0811,
+      "step": 14031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.044854790448506e-05,
+      "loss": 1.0894,
+      "step": 14032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043514960757198e-05,
+      "loss": 0.9918,
+      "step": 14033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.04217518921986e-05,
+      "loss": 0.8105,
+      "step": 14034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.040835475862854e-05,
+      "loss": 0.9847,
+      "step": 14035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039495820712525e-05,
+      "loss": 1.0313,
+      "step": 14036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.038156223795224e-05,
+      "loss": 0.8525,
+      "step": 14037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.036816685137308e-05,
+      "loss": 0.9376,
+      "step": 14038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.035477204765114e-05,
+      "loss": 0.7561,
+      "step": 14039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.034137782704997e-05,
+      "loss": 0.9012,
+      "step": 14040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.032798418983306e-05,
+      "loss": 1.0176,
+      "step": 14041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.031459113626379e-05,
+      "loss": 1.0845,
+      "step": 14042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.030119866660564e-05,
+      "loss": 0.9431,
+      "step": 14043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.028780678112207e-05,
+      "loss": 1.1811,
+      "step": 14044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.027441548007641e-05,
+      "loss": 1.0387,
+      "step": 14045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.026102476373215e-05,
+      "loss": 1.0203,
+      "step": 14046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.024763463235267e-05,
+      "loss": 0.9164,
+      "step": 14047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.023424508620131e-05,
+      "loss": 0.9172,
+      "step": 14048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.02208561255415e-05,
+      "loss": 1.1169,
+      "step": 14049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.020746775063654e-05,
+      "loss": 0.9752,
+      "step": 14050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.01940799617498e-05,
+      "loss": 0.7599,
+      "step": 14051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.018069275914468e-05,
+      "loss": 1.1527,
+      "step": 14052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.01673061430844e-05,
+      "loss": 0.9639,
+      "step": 14053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.015392011383234e-05,
+      "loss": 0.9213,
+      "step": 14054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.014053467165182e-05,
+      "loss": 0.7822,
+      "step": 14055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.012714981680607e-05,
+      "loss": 1.0437,
+      "step": 14056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.011376554955838e-05,
+      "loss": 0.9164,
+      "step": 14057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.010038187017209e-05,
+      "loss": 0.9834,
+      "step": 14058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008699877891039e-05,
+      "loss": 0.9447,
+      "step": 14059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.007361627603653e-05,
+      "loss": 0.911,
+      "step": 14060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.006023436181377e-05,
+      "loss": 0.9756,
+      "step": 14061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004685303650528e-05,
+      "loss": 0.8797,
+      "step": 14062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.003347230037433e-05,
+      "loss": 1.1188,
+      "step": 14063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.002009215368412e-05,
+      "loss": 0.9634,
+      "step": 14064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.000671259669778e-05,
+      "loss": 1.0696,
+      "step": 14065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.999333362967852e-05,
+      "loss": 1.0124,
+      "step": 14066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.997995525288954e-05,
+      "loss": 1.0902,
+      "step": 14067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99665774665939e-05,
+      "loss": 0.9273,
+      "step": 14068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.995320027105481e-05,
+      "loss": 0.9945,
+      "step": 14069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99398236665354e-05,
+      "loss": 0.9219,
+      "step": 14070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.992644765329877e-05,
+      "loss": 0.8926,
+      "step": 14071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.991307223160803e-05,
+      "loss": 0.9177,
+      "step": 14072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.989969740172626e-05,
+      "loss": 0.9749,
+      "step": 14073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.988632316391653e-05,
+      "loss": 0.8624,
+      "step": 14074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.987294951844199e-05,
+      "loss": 1.1278,
+      "step": 14075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.985957646556557e-05,
+      "loss": 1.091,
+      "step": 14076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.984620400555044e-05,
+      "loss": 1.0503,
+      "step": 14077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.983283213865959e-05,
+      "loss": 1.0199,
+      "step": 14078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.981946086515603e-05,
+      "loss": 1.1672,
+      "step": 14079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.980609018530275e-05,
+      "loss": 0.984,
+      "step": 14080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979272009936284e-05,
+      "loss": 1.0162,
+      "step": 14081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.977935060759921e-05,
+      "loss": 0.8798,
+      "step": 14082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.976598171027486e-05,
+      "loss": 0.9338,
+      "step": 14083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.975261340765278e-05,
+      "loss": 0.9816,
+      "step": 14084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.973924569999585e-05,
+      "loss": 0.9231,
+      "step": 14085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.97258785875671e-05,
+      "loss": 0.9785,
+      "step": 14086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.971251207062943e-05,
+      "loss": 0.9335,
+      "step": 14087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969914614944572e-05,
+      "loss": 0.9474,
+      "step": 14088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.968578082427894e-05,
+      "loss": 0.9338,
+      "step": 14089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967241609539197e-05,
+      "loss": 1.0166,
+      "step": 14090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965905196304763e-05,
+      "loss": 0.8666,
+      "step": 14091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.964568842750892e-05,
+      "loss": 0.7625,
+      "step": 14092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.963232548903853e-05,
+      "loss": 1.0842,
+      "step": 14093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.961896314789944e-05,
+      "loss": 1.1065,
+      "step": 14094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.960560140435448e-05,
+      "loss": 0.9228,
+      "step": 14095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95922402586664e-05,
+      "loss": 1.0315,
+      "step": 14096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.957887971109805e-05,
+      "loss": 0.8339,
+      "step": 14097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.95655197619123e-05,
+      "loss": 1.025,
+      "step": 14098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.955216041137181e-05,
+      "loss": 0.9666,
+      "step": 14099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.953880165973946e-05,
+      "loss": 0.9566,
+      "step": 14100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.952544350727799e-05,
+      "loss": 1.0396,
+      "step": 14101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.951208595425011e-05,
+      "loss": 0.7156,
+      "step": 14102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.949872900091861e-05,
+      "loss": 0.9517,
+      "step": 14103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.948537264754622e-05,
+      "loss": 0.9908,
+      "step": 14104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.947201689439565e-05,
+      "loss": 1.153,
+      "step": 14105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.945866174172961e-05,
+      "loss": 1.1788,
+      "step": 14106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.944530718981079e-05,
+      "loss": 1.0073,
+      "step": 14107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.943195323890185e-05,
+      "loss": 0.9093,
+      "step": 14108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.94185998892655e-05,
+      "loss": 0.8366,
+      "step": 14109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.940524714116443e-05,
+      "loss": 0.9525,
+      "step": 14110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.939189499486121e-05,
+      "loss": 0.953,
+      "step": 14111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.93785434506185e-05,
+      "loss": 0.886,
+      "step": 14112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.936519250869897e-05,
+      "loss": 0.9845,
+      "step": 14113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.935184216936515e-05,
+      "loss": 1.0154,
+      "step": 14114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.933849243287976e-05,
+      "loss": 1.0131,
+      "step": 14115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.932514329950526e-05,
+      "loss": 0.8872,
+      "step": 14116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.931179476950429e-05,
+      "loss": 0.8974,
+      "step": 14117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.929844684313945e-05,
+      "loss": 0.9525,
+      "step": 14118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.928509952067321e-05,
+      "loss": 0.9879,
+      "step": 14119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.927175280236815e-05,
+      "loss": 1.1223,
+      "step": 14120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.925840668848686e-05,
+      "loss": 1.1265,
+      "step": 14121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.924506117929175e-05,
+      "loss": 1.0995,
+      "step": 14122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92317162750454e-05,
+      "loss": 0.8932,
+      "step": 14123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.92183719760103e-05,
+      "loss": 1.1908,
+      "step": 14124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.920502828244889e-05,
+      "loss": 0.8706,
+      "step": 14125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.919168519462365e-05,
+      "loss": 0.9618,
+      "step": 14126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917834271279709e-05,
+      "loss": 1.1594,
+      "step": 14127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916500083723159e-05,
+      "loss": 1.1319,
+      "step": 14128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.915165956818963e-05,
+      "loss": 0.9817,
+      "step": 14129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913831890593362e-05,
+      "loss": 1.1635,
+      "step": 14130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.912497885072594e-05,
+      "loss": 0.8605,
+      "step": 14131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.911163940282902e-05,
+      "loss": 1.001,
+      "step": 14132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.909830056250527e-05,
+      "loss": 1.0259,
+      "step": 14133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.908496233001702e-05,
+      "loss": 1.0733,
+      "step": 14134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.907162470562663e-05,
+      "loss": 0.9798,
+      "step": 14135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.90582876895965e-05,
+      "loss": 1.0406,
+      "step": 14136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.904495128218891e-05,
+      "loss": 1.0204,
+      "step": 14137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.903161548366625e-05,
+      "loss": 1.136,
+      "step": 14138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.901828029429076e-05,
+      "loss": 0.8442,
+      "step": 14139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.900494571432478e-05,
+      "loss": 0.8729,
+      "step": 14140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.899161174403064e-05,
+      "loss": 1.0359,
+      "step": 14141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.897827838367056e-05,
+      "loss": 0.975,
+      "step": 14142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89649456335068e-05,
+      "loss": 1.0264,
+      "step": 14143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89516134938017e-05,
+      "loss": 0.9535,
+      "step": 14144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89382819648174e-05,
+      "loss": 0.872,
+      "step": 14145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89249510468162e-05,
+      "loss": 1.0489,
+      "step": 14146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.89116207400603e-05,
+      "loss": 0.8947,
+      "step": 14147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88982910448119e-05,
+      "loss": 0.8445,
+      "step": 14148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.888496196133317e-05,
+      "loss": 0.8119,
+      "step": 14149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.887163348988635e-05,
+      "loss": 0.9233,
+      "step": 14150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.885830563073355e-05,
+      "loss": 0.8931,
+      "step": 14151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.884497838413698e-05,
+      "loss": 0.8555,
+      "step": 14152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.883165175035878e-05,
+      "loss": 0.8785,
+      "step": 14153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.881832572966102e-05,
+      "loss": 0.9413,
+      "step": 14154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88050003223059e-05,
+      "loss": 1.1802,
+      "step": 14155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.879167552855554e-05,
+      "loss": 1.0209,
+      "step": 14156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.877835134867196e-05,
+      "loss": 1.043,
+      "step": 14157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.876502778291731e-05,
+      "loss": 0.8473,
+      "step": 14158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.875170483155362e-05,
+      "loss": 1.0925,
+      "step": 14159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.873838249484296e-05,
+      "loss": 1.1469,
+      "step": 14160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.872506077304746e-05,
+      "loss": 1.1956,
+      "step": 14161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8711739666429e-05,
+      "loss": 0.8709,
+      "step": 14162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.869841917524975e-05,
+      "loss": 0.9136,
+      "step": 14163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.868509929977168e-05,
+      "loss": 1.0407,
+      "step": 14164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.867178004025676e-05,
+      "loss": 0.934,
+      "step": 14165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.865846139696697e-05,
+      "loss": 0.7213,
+      "step": 14166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.864514337016438e-05,
+      "loss": 1.0335,
+      "step": 14167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.863182596011087e-05,
+      "loss": 1.0987,
+      "step": 14168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.86185091670684e-05,
+      "loss": 0.8881,
+      "step": 14169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.860519299129897e-05,
+      "loss": 0.9882,
+      "step": 14170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.859187743306443e-05,
+      "loss": 0.905,
+      "step": 14171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.857856249262672e-05,
+      "loss": 0.85,
+      "step": 14172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.85652481702478e-05,
+      "loss": 0.9114,
+      "step": 14173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.85519344661895e-05,
+      "loss": 0.9069,
+      "step": 14174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.853862138071372e-05,
+      "loss": 0.8581,
+      "step": 14175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.852530891408234e-05,
+      "loss": 0.9401,
+      "step": 14176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.851199706655717e-05,
+      "loss": 0.7703,
+      "step": 14177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849868583840012e-05,
+      "loss": 1.071,
+      "step": 14178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.848537522987299e-05,
+      "loss": 1.0328,
+      "step": 14179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.847206524123759e-05,
+      "loss": 0.6905,
+      "step": 14180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845875587275576e-05,
+      "loss": 1.0019,
+      "step": 14181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.844544712468924e-05,
+      "loss": 1.1001,
+      "step": 14182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.843213899729983e-05,
+      "loss": 0.8743,
+      "step": 14183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.841883149084938e-05,
+      "loss": 0.7348,
+      "step": 14184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.84055246055995e-05,
+      "loss": 0.8974,
+      "step": 14185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.839221834181207e-05,
+      "loss": 1.0659,
+      "step": 14186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.837891269974876e-05,
+      "loss": 0.9605,
+      "step": 14187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.836560767967131e-05,
+      "loss": 0.9033,
+      "step": 14188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.835230328184138e-05,
+      "loss": 0.8932,
+      "step": 14189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833899950652076e-05,
+      "loss": 0.9881,
+      "step": 14190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.832569635397108e-05,
+      "loss": 0.7663,
+      "step": 14191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8312393824454e-05,
+      "loss": 0.847,
+      "step": 14192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.829909191823121e-05,
+      "loss": 0.9365,
+      "step": 14193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.828579063556434e-05,
+      "loss": 1.0478,
+      "step": 14194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8272489976715e-05,
+      "loss": 0.8967,
+      "step": 14195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.82591899419449e-05,
+      "loss": 0.8558,
+      "step": 14196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824589053151558e-05,
+      "loss": 0.9823,
+      "step": 14197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.823259174568863e-05,
+      "loss": 0.9995,
+      "step": 14198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.821929358472571e-05,
+      "loss": 1.0092,
+      "step": 14199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.820599604888828e-05,
+      "loss": 0.9647,
+      "step": 14200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.819269913843806e-05,
+      "loss": 0.9664,
+      "step": 14201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.817940285363644e-05,
+      "loss": 0.8399,
+      "step": 14202
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.816610719474503e-05,
+      "loss": 0.8029,
+      "step": 14203
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815281216202538e-05,
+      "loss": 1.0889,
+      "step": 14204
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.813951775573896e-05,
+      "loss": 0.9365,
+      "step": 14205
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812622397614726e-05,
+      "loss": 0.895,
+      "step": 14206
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.811293082351186e-05,
+      "loss": 1.0501,
+      "step": 14207
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.80996382980941e-05,
+      "loss": 1.0827,
+      "step": 14208
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.808634640015553e-05,
+      "loss": 0.9196,
+      "step": 14209
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.80730551299576e-05,
+      "loss": 1.1299,
+      "step": 14210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.805976448776173e-05,
+      "loss": 1.027,
+      "step": 14211
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.804647447382933e-05,
+      "loss": 0.8387,
+      "step": 14212
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.803318508842187e-05,
+      "loss": 1.464,
+      "step": 14213
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.801989633180068e-05,
+      "loss": 0.973,
+      "step": 14214
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.80066082042272e-05,
+      "loss": 0.8114,
+      "step": 14215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.79933207059628e-05,
+      "loss": 0.9547,
+      "step": 14216
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.798003383726883e-05,
+      "loss": 0.95,
+      "step": 14217
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796674759840662e-05,
+      "loss": 1.0008,
+      "step": 14218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.795346198963759e-05,
+      "loss": 0.8427,
+      "step": 14219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7940177011223e-05,
+      "loss": 1.1447,
+      "step": 14220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.792689266342416e-05,
+      "loss": 1.1973,
+      "step": 14221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.791360894650243e-05,
+      "loss": 0.9908,
+      "step": 14222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790032586071902e-05,
+      "loss": 1.0231,
+      "step": 14223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.788704340633531e-05,
+      "loss": 1.028,
+      "step": 14224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.787376158361245e-05,
+      "loss": 0.9153,
+      "step": 14225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.786048039281177e-05,
+      "loss": 1.1299,
+      "step": 14226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.784719983419452e-05,
+      "loss": 0.9608,
+      "step": 14227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.783391990802187e-05,
+      "loss": 0.9448,
+      "step": 14228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.782064061455504e-05,
+      "loss": 0.9514,
+      "step": 14229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.780736195405533e-05,
+      "loss": 0.9373,
+      "step": 14230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.77940839267838e-05,
+      "loss": 0.9459,
+      "step": 14231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.778080653300171e-05,
+      "loss": 1.054,
+      "step": 14232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.776752977297022e-05,
+      "loss": 0.7622,
+      "step": 14233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.775425364695045e-05,
+      "loss": 0.8925,
+      "step": 14234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.774097815520352e-05,
+      "loss": 0.9885,
+      "step": 14235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.772770329799066e-05,
+      "loss": 0.891,
+      "step": 14236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.77144290755729e-05,
+      "loss": 0.9075,
+      "step": 14237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.770115548821137e-05,
+      "loss": 1.053,
+      "step": 14238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.768788253616717e-05,
+      "loss": 0.9391,
+      "step": 14239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.767461021970134e-05,
+      "loss": 0.8688,
+      "step": 14240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.766133853907498e-05,
+      "loss": 0.9411,
+      "step": 14241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.764806749454915e-05,
+      "loss": 1.2498,
+      "step": 14242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.763479708638485e-05,
+      "loss": 0.9179,
+      "step": 14243
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.762152731484315e-05,
+      "loss": 1.1155,
+      "step": 14244
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.760825818018508e-05,
+      "loss": 1.1642,
+      "step": 14245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.759498968267157e-05,
+      "loss": 0.9707,
+      "step": 14246
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75817218225637e-05,
+      "loss": 0.8888,
+      "step": 14247
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.756845460012236e-05,
+      "loss": 0.9077,
+      "step": 14248
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.755518801560859e-05,
+      "loss": 1.0013,
+      "step": 14249
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.754192206928331e-05,
+      "loss": 0.9029,
+      "step": 14250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.752865676140745e-05,
+      "loss": 0.9755,
+      "step": 14251
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.751539209224194e-05,
+      "loss": 1.2135,
+      "step": 14252
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.750212806204777e-05,
+      "loss": 0.8939,
+      "step": 14253
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.748886467108572e-05,
+      "loss": 1.0251,
+      "step": 14254
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.747560191961676e-05,
+      "loss": 0.8893,
+      "step": 14255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.746233980790177e-05,
+      "loss": 0.9754,
+      "step": 14256
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744907833620158e-05,
+      "loss": 0.8695,
+      "step": 14257
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.743581750477703e-05,
+      "loss": 0.8901,
+      "step": 14258
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.742255731388904e-05,
+      "loss": 1.1526,
+      "step": 14259
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.740929776379835e-05,
+      "loss": 0.9764,
+      "step": 14260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.739603885476582e-05,
+      "loss": 1.0918,
+      "step": 14261
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.738278058705226e-05,
+      "loss": 0.8941,
+      "step": 14262
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.736952296091842e-05,
+      "loss": 0.9371,
+      "step": 14263
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.73562659766251e-05,
+      "loss": 1.0089,
+      "step": 14264
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.734300963443307e-05,
+      "loss": 1.0207,
+      "step": 14265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732975393460308e-05,
+      "loss": 1.0975,
+      "step": 14266
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.731649887739589e-05,
+      "loss": 1.0279,
+      "step": 14267
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.730324446307217e-05,
+      "loss": 0.8985,
+      "step": 14268
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.728999069189263e-05,
+      "loss": 0.9104,
+      "step": 14269
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72767375641181e-05,
+      "loss": 0.8616,
+      "step": 14270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72634850800091e-05,
+      "loss": 0.9417,
+      "step": 14271
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.72502332398264e-05,
+      "loss": 1.0954,
+      "step": 14272
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.723698204383066e-05,
+      "loss": 1.0456,
+      "step": 14273
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.722373149228252e-05,
+      "loss": 0.9403,
+      "step": 14274
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.721048158544256e-05,
+      "loss": 0.8124,
+      "step": 14275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.719723232357154e-05,
+      "loss": 0.8967,
+      "step": 14276
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.718398370692993e-05,
+      "loss": 1.12,
+      "step": 14277
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.717073573577841e-05,
+      "loss": 0.8556,
+      "step": 14278
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.715748841037758e-05,
+      "loss": 0.9725,
+      "step": 14279
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.714424173098795e-05,
+      "loss": 0.9008,
+      "step": 14280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.713099569787009e-05,
+      "loss": 1.0299,
+      "step": 14281
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.711775031128462e-05,
+      "loss": 0.8592,
+      "step": 14282
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.710450557149202e-05,
+      "loss": 0.8582,
+      "step": 14283
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.709126147875282e-05,
+      "loss": 0.6882,
+      "step": 14284
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.707801803332754e-05,
+      "loss": 0.8757,
+      "step": 14285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.706477523547668e-05,
+      "loss": 0.997,
+      "step": 14286
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.705153308546067e-05,
+      "loss": 0.964,
+      "step": 14287
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.703829158354009e-05,
+      "loss": 0.9636,
+      "step": 14288
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.702505072997532e-05,
+      "loss": 0.6582,
+      "step": 14289
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.701181052502683e-05,
+      "loss": 0.9505,
+      "step": 14290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699857096895506e-05,
+      "loss": 0.9518,
+      "step": 14291
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.698533206202037e-05,
+      "loss": 1.0498,
+      "step": 14292
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.697209380448333e-05,
+      "loss": 0.9159,
+      "step": 14293
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.695885619660414e-05,
+      "loss": 1.0549,
+      "step": 14294
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.694561923864329e-05,
+      "loss": 0.8444,
+      "step": 14295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.693238293086118e-05,
+      "loss": 1.1868,
+      "step": 14296
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.691914727351806e-05,
+      "loss": 0.9048,
+      "step": 14297
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.690591226687434e-05,
+      "loss": 0.8786,
+      "step": 14298
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.689267791119042e-05,
+      "loss": 0.828,
+      "step": 14299
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.687944420672648e-05,
+      "loss": 0.9152,
+      "step": 14300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68662111537429e-05,
+      "loss": 0.9991,
+      "step": 14301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68529787525e-05,
+      "loss": 0.9301,
+      "step": 14302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.683974700325801e-05,
+      "loss": 0.9472,
+      "step": 14303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.68265159062772e-05,
+      "loss": 0.9456,
+      "step": 14304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.681328546181788e-05,
+      "loss": 0.9543,
+      "step": 14305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.680005567014022e-05,
+      "loss": 0.9837,
+      "step": 14306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.67868265315045e-05,
+      "loss": 1.1866,
+      "step": 14307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.677359804617094e-05,
+      "loss": 0.96,
+      "step": 14308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.676037021439971e-05,
+      "loss": 1.0103,
+      "step": 14309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.674714303645099e-05,
+      "loss": 1.0203,
+      "step": 14310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.673391651258502e-05,
+      "loss": 1.0481,
+      "step": 14311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.672069064306192e-05,
+      "loss": 0.9332,
+      "step": 14312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.670746542814187e-05,
+      "loss": 1.0725,
+      "step": 14313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.669424086808497e-05,
+      "loss": 0.9212,
+      "step": 14314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.668101696315135e-05,
+      "loss": 1.1268,
+      "step": 14315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66677937136012e-05,
+      "loss": 1.0871,
+      "step": 14316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.665457111969451e-05,
+      "loss": 0.8949,
+      "step": 14317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.664134918169143e-05,
+      "loss": 1.1959,
+      "step": 14318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662812789985206e-05,
+      "loss": 1.0151,
+      "step": 14319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66149072744364e-05,
+      "loss": 1.1412,
+      "step": 14320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.66016873057045e-05,
+      "loss": 0.9958,
+      "step": 14321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.65884679939165e-05,
+      "loss": 0.8869,
+      "step": 14322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.657524933933228e-05,
+      "loss": 0.9011,
+      "step": 14323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.656203134221195e-05,
+      "loss": 1.0017,
+      "step": 14324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.654881400281547e-05,
+      "loss": 1.1875,
+      "step": 14325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.653559732140284e-05,
+      "loss": 0.9988,
+      "step": 14326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652238129823398e-05,
+      "loss": 0.8805,
+      "step": 14327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.650916593356895e-05,
+      "loss": 0.9484,
+      "step": 14328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.64959512276676e-05,
+      "loss": 0.9504,
+      "step": 14329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.648273718078993e-05,
+      "loss": 1.0655,
+      "step": 14330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.646952379319582e-05,
+      "loss": 0.8977,
+      "step": 14331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.645631106514519e-05,
+      "loss": 1.0787,
+      "step": 14332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644309899689793e-05,
+      "loss": 0.8774,
+      "step": 14333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.64298875887139e-05,
+      "loss": 0.9699,
+      "step": 14334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.641667684085299e-05,
+      "loss": 0.8542,
+      "step": 14335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.640346675357507e-05,
+      "loss": 0.8442,
+      "step": 14336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.639025732713998e-05,
+      "loss": 1.1454,
+      "step": 14337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.637704856180748e-05,
+      "loss": 0.8208,
+      "step": 14338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.636384045783752e-05,
+      "loss": 0.8476,
+      "step": 14339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.635063301548977e-05,
+      "loss": 0.9033,
+      "step": 14340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.63374262350241e-05,
+      "loss": 0.9518,
+      "step": 14341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.632422011670027e-05,
+      "loss": 1.0675,
+      "step": 14342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6311014660778e-05,
+      "loss": 0.9974,
+      "step": 14343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.629780986751708e-05,
+      "loss": 1.1297,
+      "step": 14344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.62846057371773e-05,
+      "loss": 0.8493,
+      "step": 14345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.627140227001827e-05,
+      "loss": 1.1559,
+      "step": 14346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.625819946629979e-05,
+      "loss": 0.9507,
+      "step": 14347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624499732628154e-05,
+      "loss": 1.0944,
+      "step": 14348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.623179585022316e-05,
+      "loss": 0.9335,
+      "step": 14349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.621859503838436e-05,
+      "loss": 0.8405,
+      "step": 14350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.620539489102483e-05,
+      "loss": 0.9611,
+      "step": 14351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.619219540840417e-05,
+      "loss": 1.0501,
+      "step": 14352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.617899659078204e-05,
+      "loss": 1.0068,
+      "step": 14353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.616579843841804e-05,
+      "loss": 0.8317,
+      "step": 14354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.615260095157179e-05,
+      "loss": 1.2033,
+      "step": 14355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.613940413050288e-05,
+      "loss": 1.0927,
+      "step": 14356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.612620797547087e-05,
+      "loss": 0.9315,
+      "step": 14357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.611301248673536e-05,
+      "loss": 0.7953,
+      "step": 14358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60998176645559e-05,
+      "loss": 0.8178,
+      "step": 14359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.608662350919203e-05,
+      "loss": 0.9353,
+      "step": 14360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.607343002090325e-05,
+      "loss": 0.9472,
+      "step": 14361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.606023719994915e-05,
+      "loss": 0.9445,
+      "step": 14362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.604704504658911e-05,
+      "loss": 1.021,
+      "step": 14363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.603385356108273e-05,
+      "loss": 1.0335,
+      "step": 14364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.602066274368947e-05,
+      "loss": 1.0905,
+      "step": 14365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.600747259466876e-05,
+      "loss": 1.1093,
+      "step": 14366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.599428311428002e-05,
+      "loss": 1.0622,
+      "step": 14367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.598109430278279e-05,
+      "loss": 0.8111,
+      "step": 14368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596790616043637e-05,
+      "loss": 1.054,
+      "step": 14369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.595471868750024e-05,
+      "loss": 1.0109,
+      "step": 14370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.594153188423384e-05,
+      "loss": 0.8766,
+      "step": 14371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.592834575089646e-05,
+      "loss": 0.8411,
+      "step": 14372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.59151602877475e-05,
+      "loss": 0.9284,
+      "step": 14373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.590197549504637e-05,
+      "loss": 1.0389,
+      "step": 14374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.588879137305235e-05,
+      "loss": 1.1085,
+      "step": 14375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.587560792202483e-05,
+      "loss": 0.9905,
+      "step": 14376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.586242514222306e-05,
+      "loss": 0.8463,
+      "step": 14377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.584924303390638e-05,
+      "loss": 1.0391,
+      "step": 14378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.583606159733411e-05,
+      "loss": 0.8104,
+      "step": 14379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.582288083276546e-05,
+      "loss": 0.9724,
+      "step": 14380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.580970074045975e-05,
+      "loss": 0.9124,
+      "step": 14381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579652132067624e-05,
+      "loss": 0.9292,
+      "step": 14382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.578334257367412e-05,
+      "loss": 1.1684,
+      "step": 14383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.577016449971261e-05,
+      "loss": 1.069,
+      "step": 14384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.575698709905104e-05,
+      "loss": 1.2204,
+      "step": 14385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.574381037194844e-05,
+      "loss": 0.9158,
+      "step": 14386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.57306343186641e-05,
+      "loss": 0.9791,
+      "step": 14387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.571745893945722e-05,
+      "loss": 1.0162,
+      "step": 14388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.570428423458687e-05,
+      "loss": 0.9669,
+      "step": 14389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.56911102043122e-05,
+      "loss": 0.7555,
+      "step": 14390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.567793684889248e-05,
+      "loss": 1.0151,
+      "step": 14391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.566476416858662e-05,
+      "loss": 0.8932,
+      "step": 14392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.565159216365389e-05,
+      "loss": 1.2164,
+      "step": 14393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.563842083435334e-05,
+      "loss": 0.8689,
+      "step": 14394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5625250180944e-05,
+      "loss": 0.9651,
+      "step": 14395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.561208020368498e-05,
+      "loss": 0.8726,
+      "step": 14396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.559891090283537e-05,
+      "loss": 0.944,
+      "step": 14397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.558574227865413e-05,
+      "loss": 0.7956,
+      "step": 14398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.557257433140034e-05,
+      "loss": 0.9713,
+      "step": 14399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.555940706133299e-05,
+      "loss": 1.1074,
+      "step": 14400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.554624046871108e-05,
+      "loss": 0.914,
+      "step": 14401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.553307455379361e-05,
+      "loss": 0.8775,
+      "step": 14402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.551990931683953e-05,
+      "loss": 1.0107,
+      "step": 14403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.550674475810781e-05,
+      "loss": 0.9927,
+      "step": 14404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.549358087785744e-05,
+      "loss": 1.2035,
+      "step": 14405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.548041767634729e-05,
+      "loss": 0.9561,
+      "step": 14406
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.54672551538363e-05,
+      "loss": 0.9175,
+      "step": 14407
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.545409331058341e-05,
+      "loss": 0.9895,
+      "step": 14408
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544093214684745e-05,
+      "loss": 0.9753,
+      "step": 14409
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.542777166288734e-05,
+      "loss": 1.1396,
+      "step": 14410
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.541461185896197e-05,
+      "loss": 0.9619,
+      "step": 14411
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.540145273533013e-05,
+      "loss": 1.0375,
+      "step": 14412
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.538829429225069e-05,
+      "loss": 0.8561,
+      "step": 14413
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.537513652998253e-05,
+      "loss": 0.9499,
+      "step": 14414
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.536197944878437e-05,
+      "loss": 0.8345,
+      "step": 14415
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.534882304891506e-05,
+      "loss": 0.8209,
+      "step": 14416
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.533566733063339e-05,
+      "loss": 0.8556,
+      "step": 14417
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.53225122941981e-05,
+      "loss": 1.068,
+      "step": 14418
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.530935793986794e-05,
+      "loss": 0.8114,
+      "step": 14419
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.529620426790175e-05,
+      "loss": 0.9403,
+      "step": 14420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.528305127855816e-05,
+      "loss": 1.0409,
+      "step": 14421
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.526989897209596e-05,
+      "loss": 0.8386,
+      "step": 14422
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.525674734877378e-05,
+      "loss": 1.0075,
+      "step": 14423
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524359640885038e-05,
+      "loss": 1.1147,
+      "step": 14424
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.523044615258442e-05,
+      "loss": 0.9199,
+      "step": 14425
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.521729658023452e-05,
+      "loss": 0.8836,
+      "step": 14426
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.520414769205938e-05,
+      "loss": 1.0332,
+      "step": 14427
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.519099948831766e-05,
+      "loss": 1.0423,
+      "step": 14428
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.517785196926792e-05,
+      "loss": 0.8486,
+      "step": 14429
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.516470513516879e-05,
+      "loss": 1.048,
+      "step": 14430
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.515155898627894e-05,
+      "loss": 0.7879,
+      "step": 14431
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.513841352285684e-05,
+      "loss": 1.0337,
+      "step": 14432
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.512526874516113e-05,
+      "loss": 0.7953,
+      "step": 14433
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.511212465345039e-05,
+      "loss": 0.9287,
+      "step": 14434
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.509898124798309e-05,
+      "loss": 0.8653,
+      "step": 14435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.508583852901778e-05,
+      "loss": 0.8103,
+      "step": 14436
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.507269649681307e-05,
+      "loss": 1.0799,
+      "step": 14437
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505955515162732e-05,
+      "loss": 1.1117,
+      "step": 14438
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.504641449371909e-05,
+      "loss": 0.974,
+      "step": 14439
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.50332745233469e-05,
+      "loss": 1.1531,
+      "step": 14440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.502013524076913e-05,
+      "loss": 1.0512,
+      "step": 14441
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.50069966462443e-05,
+      "loss": 1.1511,
+      "step": 14442
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.499385874003076e-05,
+      "loss": 0.9721,
+      "step": 14443
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.498072152238703e-05,
+      "loss": 1.1191,
+      "step": 14444
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.496758499357148e-05,
+      "loss": 0.8994,
+      "step": 14445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.495444915384248e-05,
+      "loss": 0.8325,
+      "step": 14446
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.494131400345842e-05,
+      "loss": 0.9582,
+      "step": 14447
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.492817954267771e-05,
+      "loss": 0.8693,
+      "step": 14448
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.491504577175861e-05,
+      "loss": 0.9053,
+      "step": 14449
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.490191269095957e-05,
+      "loss": 1.0563,
+      "step": 14450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.488878030053888e-05,
+      "loss": 0.9008,
+      "step": 14451
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.487564860075483e-05,
+      "loss": 0.8733,
+      "step": 14452
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.486251759186572e-05,
+      "loss": 0.836,
+      "step": 14453
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.484938727412991e-05,
+      "loss": 0.9284,
+      "step": 14454
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.483625764780554e-05,
+      "loss": 0.8815,
+      "step": 14455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.482312871315098e-05,
+      "loss": 0.9314,
+      "step": 14456
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.481000047042445e-05,
+      "loss": 1.0866,
+      "step": 14457
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.479687291988414e-05,
+      "loss": 0.9332,
+      "step": 14458
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.47837460617883e-05,
+      "loss": 0.9373,
+      "step": 14459
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.477061989639519e-05,
+      "loss": 0.9247,
+      "step": 14460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.475749442396289e-05,
+      "loss": 1.2271,
+      "step": 14461
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.474436964474965e-05,
+      "loss": 0.9489,
+      "step": 14462
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.473124555901365e-05,
+      "loss": 0.9837,
+      "step": 14463
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.471812216701297e-05,
+      "loss": 1.1676,
+      "step": 14464
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.470499946900582e-05,
+      "loss": 0.9401,
+      "step": 14465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469187746525023e-05,
+      "loss": 0.8851,
+      "step": 14466
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.46787561560044e-05,
+      "loss": 0.8794,
+      "step": 14467
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.466563554152641e-05,
+      "loss": 0.9684,
+      "step": 14468
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.465251562207431e-05,
+      "loss": 0.8538,
+      "step": 14469
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.463939639790617e-05,
+      "loss": 1.0352,
+      "step": 14470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.462627786928009e-05,
+      "loss": 1.0757,
+      "step": 14471
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.461316003645403e-05,
+      "loss": 0.8841,
+      "step": 14472
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.460004289968611e-05,
+      "loss": 0.9259,
+      "step": 14473
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45869264592343e-05,
+      "loss": 0.876,
+      "step": 14474
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.45738107153566e-05,
+      "loss": 1.1622,
+      "step": 14475
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.456069566831095e-05,
+      "loss": 1.075,
+      "step": 14476
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.454758131835545e-05,
+      "loss": 1.0545,
+      "step": 14477
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.453446766574793e-05,
+      "loss": 1.1318,
+      "step": 14478
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.452135471074639e-05,
+      "loss": 1.0975,
+      "step": 14479
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.450824245360876e-05,
+      "loss": 1.1869,
+      "step": 14480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.449513089459295e-05,
+      "loss": 1.0173,
+      "step": 14481
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.448202003395684e-05,
+      "loss": 1.0008,
+      "step": 14482
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.446890987195841e-05,
+      "loss": 1.0365,
+      "step": 14483
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.445580040885541e-05,
+      "loss": 1.0057,
+      "step": 14484
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.44426916449058e-05,
+      "loss": 1.0593,
+      "step": 14485
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.442958358036741e-05,
+      "loss": 1.0321,
+      "step": 14486
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.441647621549804e-05,
+      "loss": 1.1045,
+      "step": 14487
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.440336955055555e-05,
+      "loss": 0.9285,
+      "step": 14488
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.439026358579768e-05,
+      "loss": 1.0182,
+      "step": 14489
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43771583214823e-05,
+      "loss": 0.8423,
+      "step": 14490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.436405375786719e-05,
+      "loss": 0.9466,
+      "step": 14491
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.435094989521006e-05,
+      "loss": 0.9777,
+      "step": 14492
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.43378467337687e-05,
+      "loss": 1.1207,
+      "step": 14493
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.432474427380086e-05,
+      "loss": 0.8017,
+      "step": 14494
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.431164251556422e-05,
+      "loss": 0.9788,
+      "step": 14495
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.429854145931652e-05,
+      "loss": 0.9333,
+      "step": 14496
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.42854411053155e-05,
+      "loss": 1.0197,
+      "step": 14497
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.427234145381875e-05,
+      "loss": 0.9501,
+      "step": 14498
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425924250508398e-05,
+      "loss": 1.0842,
+      "step": 14499
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.424614425936892e-05,
+      "loss": 0.9823,
+      "step": 14500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.423304671693107e-05,
+      "loss": 0.9994,
+      "step": 14501
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.421994987802818e-05,
+      "loss": 0.7595,
+      "step": 14502
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.420685374291782e-05,
+      "loss": 1.033,
+      "step": 14503
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.419375831185756e-05,
+      "loss": 0.7762,
+      "step": 14504
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.418066358510502e-05,
+      "loss": 1.2032,
+      "step": 14505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.416756956291782e-05,
+      "loss": 0.8946,
+      "step": 14506
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.415447624555342e-05,
+      "loss": 0.9034,
+      "step": 14507
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.414138363326945e-05,
+      "loss": 0.9604,
+      "step": 14508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.412829172632338e-05,
+      "loss": 0.7916,
+      "step": 14509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.411520052497276e-05,
+      "loss": 1.0674,
+      "step": 14510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410211002947508e-05,
+      "loss": 1.0145,
+      "step": 14511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.408902024008783e-05,
+      "loss": 1.0325,
+      "step": 14512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40759311570685e-05,
+      "loss": 1.0155,
+      "step": 14513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.406284278067454e-05,
+      "loss": 0.9983,
+      "step": 14514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40497551111634e-05,
+      "loss": 0.8098,
+      "step": 14515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.403666814879253e-05,
+      "loss": 0.9744,
+      "step": 14516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.402358189381934e-05,
+      "loss": 0.7428,
+      "step": 14517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.401049634650118e-05,
+      "loss": 1.0645,
+      "step": 14518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.399741150709553e-05,
+      "loss": 0.8946,
+      "step": 14519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398432737585975e-05,
+      "loss": 0.8663,
+      "step": 14520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.397124395305117e-05,
+      "loss": 0.9461,
+      "step": 14521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.395816123892714e-05,
+      "loss": 1.0558,
+      "step": 14522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.394507923374507e-05,
+      "loss": 0.8205,
+      "step": 14523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.393199793776217e-05,
+      "loss": 0.7028,
+      "step": 14524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.391891735123582e-05,
+      "loss": 0.9385,
+      "step": 14525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.390583747442333e-05,
+      "loss": 0.8989,
+      "step": 14526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.389275830758192e-05,
+      "loss": 1.1034,
+      "step": 14527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.387967985096886e-05,
+      "loss": 1.0965,
+      "step": 14528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.38666021048415e-05,
+      "loss": 0.9993,
+      "step": 14529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.385352506945696e-05,
+      "loss": 0.8166,
+      "step": 14530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.384044874507255e-05,
+      "loss": 0.9743,
+      "step": 14531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.382737313194542e-05,
+      "loss": 1.0419,
+      "step": 14532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.38142982303328e-05,
+      "loss": 1.0311,
+      "step": 14533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.380122404049188e-05,
+      "loss": 0.8879,
+      "step": 14534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.378815056267978e-05,
+      "loss": 1.1577,
+      "step": 14535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.37750777971537e-05,
+      "loss": 0.8355,
+      "step": 14536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.37620057441708e-05,
+      "loss": 0.9498,
+      "step": 14537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374893440398814e-05,
+      "loss": 0.9943,
+      "step": 14538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.373586377686289e-05,
+      "loss": 0.8669,
+      "step": 14539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.372279386305213e-05,
+      "loss": 1.174,
+      "step": 14540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.370972466281291e-05,
+      "loss": 1.0894,
+      "step": 14541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.369665617640237e-05,
+      "loss": 1.0775,
+      "step": 14542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.368358840407753e-05,
+      "loss": 0.9716,
+      "step": 14543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.367052134609542e-05,
+      "loss": 0.9752,
+      "step": 14544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.365745500271307e-05,
+      "loss": 0.8018,
+      "step": 14545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.364438937418756e-05,
+      "loss": 0.9909,
+      "step": 14546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.363132446077578e-05,
+      "loss": 0.8805,
+      "step": 14547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36182602627348e-05,
+      "loss": 0.9596,
+      "step": 14548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.360519678032157e-05,
+      "loss": 1.001,
+      "step": 14549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.359213401379303e-05,
+      "loss": 1.0569,
+      "step": 14550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.357907196340618e-05,
+      "loss": 1.0098,
+      "step": 14551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.356601062941787e-05,
+      "loss": 0.8993,
+      "step": 14552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.355295001208504e-05,
+      "loss": 1.0617,
+      "step": 14553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.353989011166464e-05,
+      "loss": 0.9172,
+      "step": 14554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.352683092841352e-05,
+      "loss": 1.0407,
+      "step": 14555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351377246258854e-05,
+      "loss": 1.0024,
+      "step": 14556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.350071471444661e-05,
+      "loss": 1.1115,
+      "step": 14557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.348765768424449e-05,
+      "loss": 1.081,
+      "step": 14558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.347460137223911e-05,
+      "loss": 0.7313,
+      "step": 14559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.346154577868724e-05,
+      "loss": 0.8911,
+      "step": 14560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.344849090384568e-05,
+      "loss": 0.9539,
+      "step": 14561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.343543674797121e-05,
+      "loss": 0.7251,
+      "step": 14562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.342238331132065e-05,
+      "loss": 1.0034,
+      "step": 14563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.340933059415069e-05,
+      "loss": 1.1934,
+      "step": 14564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.339627859671814e-05,
+      "loss": 1.0098,
+      "step": 14565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.338322731927973e-05,
+      "loss": 0.9954,
+      "step": 14566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337017676209213e-05,
+      "loss": 0.92,
+      "step": 14567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.335712692541204e-05,
+      "loss": 0.8737,
+      "step": 14568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.334407780949627e-05,
+      "loss": 0.8723,
+      "step": 14569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.333102941460132e-05,
+      "loss": 0.8239,
+      "step": 14570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.331798174098398e-05,
+      "loss": 0.9103,
+      "step": 14571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.330493478890087e-05,
+      "loss": 0.9786,
+      "step": 14572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.32918885586086e-05,
+      "loss": 0.8541,
+      "step": 14573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.327884305036379e-05,
+      "loss": 1.108,
+      "step": 14574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.326579826442305e-05,
+      "loss": 0.944,
+      "step": 14575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.325275420104295e-05,
+      "loss": 0.8139,
+      "step": 14576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.323971086048012e-05,
+      "loss": 1.0178,
+      "step": 14577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.322666824299109e-05,
+      "loss": 0.9599,
+      "step": 14578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.321362634883241e-05,
+      "loss": 0.8067,
+      "step": 14579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.320058517826063e-05,
+      "loss": 0.8262,
+      "step": 14580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.318754473153221e-05,
+      "loss": 0.9505,
+      "step": 14581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317450500890372e-05,
+      "loss": 0.8634,
+      "step": 14582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.316146601063164e-05,
+      "loss": 0.9859,
+      "step": 14583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.314842773697243e-05,
+      "loss": 0.9175,
+      "step": 14584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.313539018818256e-05,
+      "loss": 1.0158,
+      "step": 14585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.312235336451847e-05,
+      "loss": 0.9967,
+      "step": 14586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.310931726623659e-05,
+      "loss": 1.1713,
+      "step": 14587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.309628189359335e-05,
+      "loss": 0.8511,
+      "step": 14588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.308324724684518e-05,
+      "loss": 0.9812,
+      "step": 14589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.307021332624843e-05,
+      "loss": 0.8651,
+      "step": 14590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.305718013205949e-05,
+      "loss": 0.9866,
+      "step": 14591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.304414766453476e-05,
+      "loss": 1.0366,
+      "step": 14592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.30311159239305e-05,
+      "loss": 1.0085,
+      "step": 14593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.301808491050313e-05,
+      "loss": 1.0191,
+      "step": 14594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.300505462450895e-05,
+      "loss": 0.839,
+      "step": 14595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.299202506620424e-05,
+      "loss": 0.813,
+      "step": 14596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.297899623584532e-05,
+      "loss": 1.0173,
+      "step": 14597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.296596813368843e-05,
+      "loss": 0.7654,
+      "step": 14598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.295294075998982e-05,
+      "loss": 0.9409,
+      "step": 14599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.293991411500582e-05,
+      "loss": 0.992,
+      "step": 14600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29268881989926e-05,
+      "loss": 0.9557,
+      "step": 14601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29138630122064e-05,
+      "loss": 0.9025,
+      "step": 14602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.290083855490342e-05,
+      "loss": 0.9447,
+      "step": 14603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.288781482733983e-05,
+      "loss": 1.0067,
+      "step": 14604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.287479182977184e-05,
+      "loss": 1.0405,
+      "step": 14605
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.286176956245562e-05,
+      "loss": 0.9799,
+      "step": 14606
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.284874802564727e-05,
+      "loss": 0.8649,
+      "step": 14607
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283572721960295e-05,
+      "loss": 1.0178,
+      "step": 14608
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.28227071445788e-05,
+      "loss": 0.8428,
+      "step": 14609
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.280968780083087e-05,
+      "loss": 1.0134,
+      "step": 14610
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.279666918861531e-05,
+      "loss": 0.9088,
+      "step": 14611
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.278365130818818e-05,
+      "loss": 1.1482,
+      "step": 14612
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.277063415980549e-05,
+      "loss": 0.8869,
+      "step": 14613
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.275761774372333e-05,
+      "loss": 1.1066,
+      "step": 14614
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.274460206019778e-05,
+      "loss": 1.2454,
+      "step": 14615
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.273158710948475e-05,
+      "loss": 0.9286,
+      "step": 14616
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.271857289184034e-05,
+      "loss": 0.8626,
+      "step": 14617
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.270555940752047e-05,
+      "loss": 0.9397,
+      "step": 14618
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.269254665678116e-05,
+      "loss": 0.9573,
+      "step": 14619
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267953463987838e-05,
+      "loss": 0.7371,
+      "step": 14620
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.266652335706802e-05,
+      "loss": 0.9688,
+      "step": 14621
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.265351280860603e-05,
+      "loss": 1.0186,
+      "step": 14622
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.264050299474837e-05,
+      "loss": 0.7634,
+      "step": 14623
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.262749391575092e-05,
+      "loss": 0.8811,
+      "step": 14624
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.261448557186953e-05,
+      "loss": 1.1284,
+      "step": 14625
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.260147796336014e-05,
+      "loss": 1.0059,
+      "step": 14626
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.258847109047853e-05,
+      "loss": 0.8942,
+      "step": 14627
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.257546495348061e-05,
+      "loss": 0.8929,
+      "step": 14628
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.256245955262221e-05,
+      "loss": 0.7327,
+      "step": 14629
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.254945488815911e-05,
+      "loss": 1.0412,
+      "step": 14630
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.253645096034712e-05,
+      "loss": 0.993,
+      "step": 14631
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.252344776944206e-05,
+      "loss": 1.0103,
+      "step": 14632
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251044531569962e-05,
+      "loss": 0.9318,
+      "step": 14633
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.249744359937564e-05,
+      "loss": 0.9142,
+      "step": 14634
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.248444262072586e-05,
+      "loss": 1.0575,
+      "step": 14635
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.247144238000598e-05,
+      "loss": 0.9121,
+      "step": 14636
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.245844287747168e-05,
+      "loss": 1.0197,
+      "step": 14637
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.244544411337877e-05,
+      "loss": 1.2652,
+      "step": 14638
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.24324460879828e-05,
+      "loss": 0.8422,
+      "step": 14639
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241944880153955e-05,
+      "loss": 1.0158,
+      "step": 14640
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.240645225430462e-05,
+      "loss": 1.1044,
+      "step": 14641
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.239345644653364e-05,
+      "loss": 0.8749,
+      "step": 14642
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23804613784823e-05,
+      "loss": 0.9899,
+      "step": 14643
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.236746705040614e-05,
+      "loss": 1.0008,
+      "step": 14644
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23544734625608e-05,
+      "loss": 0.7065,
+      "step": 14645
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.234148061520187e-05,
+      "loss": 0.92,
+      "step": 14646
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23284885085849e-05,
+      "loss": 0.8601,
+      "step": 14647
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231549714296544e-05,
+      "loss": 0.8729,
+      "step": 14648
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.230250651859905e-05,
+      "loss": 0.8678,
+      "step": 14649
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.228951663574123e-05,
+      "loss": 0.8668,
+      "step": 14650
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.22765274946475e-05,
+      "loss": 1.1083,
+      "step": 14651
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.226353909557338e-05,
+      "loss": 1.0042,
+      "step": 14652
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.225055143877434e-05,
+      "loss": 0.9272,
+      "step": 14653
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223756452450582e-05,
+      "loss": 0.8003,
+      "step": 14654
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222457835302332e-05,
+      "loss": 0.8876,
+      "step": 14655
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.221159292458219e-05,
+      "loss": 1.0167,
+      "step": 14656
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.219860823943793e-05,
+      "loss": 1.0908,
+      "step": 14657
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.218562429784595e-05,
+      "loss": 1.0773,
+      "step": 14658
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.217264110006163e-05,
+      "loss": 1.0842,
+      "step": 14659
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.215965864634031e-05,
+      "loss": 0.9465,
+      "step": 14660
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.214667693693744e-05,
+      "loss": 1.0066,
+      "step": 14661
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.213369597210826e-05,
+      "loss": 1.0706,
+      "step": 14662
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.21207157521082e-05,
+      "loss": 0.969,
+      "step": 14663
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.210773627719251e-05,
+      "loss": 0.9396,
+      "step": 14664
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.209475754761656e-05,
+      "loss": 0.8768,
+      "step": 14665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.208177956363561e-05,
+      "loss": 0.8872,
+      "step": 14666
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206880232550492e-05,
+      "loss": 0.8744,
+      "step": 14667
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.205582583347974e-05,
+      "loss": 1.0777,
+      "step": 14668
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20428500878154e-05,
+      "loss": 0.8868,
+      "step": 14669
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202987508876706e-05,
+      "loss": 1.0512,
+      "step": 14670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201690083658993e-05,
+      "loss": 1.0745,
+      "step": 14671
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.200392733153929e-05,
+      "loss": 0.9303,
+      "step": 14672
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.19909545738702e-05,
+      "loss": 0.9096,
+      "step": 14673
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.197798256383798e-05,
+      "loss": 0.8046,
+      "step": 14674
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.19650113016977e-05,
+      "loss": 0.8075,
+      "step": 14675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.195204078770452e-05,
+      "loss": 1.0824,
+      "step": 14676
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.193907102211358e-05,
+      "loss": 1.036,
+      "step": 14677
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.192610200518001e-05,
+      "loss": 1.0497,
+      "step": 14678
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.191313373715884e-05,
+      "loss": 0.975,
+      "step": 14679
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.190016621830522e-05,
+      "loss": 0.8051,
+      "step": 14680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188719944887424e-05,
+      "loss": 1.0834,
+      "step": 14681
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.187423342912089e-05,
+      "loss": 1.086,
+      "step": 14682
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.186126815930026e-05,
+      "loss": 0.9834,
+      "step": 14683
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184830363966736e-05,
+      "loss": 0.9995,
+      "step": 14684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.183533987047716e-05,
+      "loss": 0.9684,
+      "step": 14685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.182237685198475e-05,
+      "loss": 0.9521,
+      "step": 14686
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.180941458444504e-05,
+      "loss": 0.9936,
+      "step": 14687
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179645306811302e-05,
+      "loss": 0.929,
+      "step": 14688
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.178349230324365e-05,
+      "loss": 1.0453,
+      "step": 14689
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.177053229009183e-05,
+      "loss": 1.1052,
+      "step": 14690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.175757302891251e-05,
+      "loss": 0.9484,
+      "step": 14691
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.174461451996063e-05,
+      "loss": 0.9801,
+      "step": 14692
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.173165676349103e-05,
+      "loss": 1.0163,
+      "step": 14693
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.171869975975861e-05,
+      "loss": 0.9088,
+      "step": 14694
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.170574350901828e-05,
+      "loss": 0.7776,
+      "step": 14695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169278801152478e-05,
+      "loss": 1.0493,
+      "step": 14696
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.167983326753303e-05,
+      "loss": 0.9277,
+      "step": 14697
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.166687927729787e-05,
+      "loss": 0.9743,
+      "step": 14698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.165392604107401e-05,
+      "loss": 0.8852,
+      "step": 14699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.164097355911631e-05,
+      "loss": 1.0101,
+      "step": 14700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.162802183167954e-05,
+      "loss": 0.9559,
+      "step": 14701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.161507085901843e-05,
+      "loss": 1.3145,
+      "step": 14702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160212064138773e-05,
+      "loss": 0.8441,
+      "step": 14703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.158917117904222e-05,
+      "loss": 0.9907,
+      "step": 14704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.157622247223655e-05,
+      "loss": 1.1305,
+      "step": 14705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.156327452122549e-05,
+      "loss": 0.6442,
+      "step": 14706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155032732626365e-05,
+      "loss": 1.0233,
+      "step": 14707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.153738088760572e-05,
+      "loss": 1.0304,
+      "step": 14708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152443520550641e-05,
+      "loss": 1.0377,
+      "step": 14709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.151149028022031e-05,
+      "loss": 1.2553,
+      "step": 14710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.149854611200206e-05,
+      "loss": 0.9318,
+      "step": 14711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.148560270110628e-05,
+      "loss": 0.8869,
+      "step": 14712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.147266004778754e-05,
+      "loss": 0.8081,
+      "step": 14713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.145971815230043e-05,
+      "loss": 0.8596,
+      "step": 14714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144677701489956e-05,
+      "loss": 0.974,
+      "step": 14715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.143383663583942e-05,
+      "loss": 1.2075,
+      "step": 14716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.14208970153746e-05,
+      "loss": 1.0079,
+      "step": 14717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140795815375958e-05,
+      "loss": 1.0325,
+      "step": 14718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139502005124886e-05,
+      "loss": 1.1243,
+      "step": 14719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.138208270809698e-05,
+      "loss": 0.9251,
+      "step": 14720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.136914612455842e-05,
+      "loss": 0.9567,
+      "step": 14721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.135621030088758e-05,
+      "loss": 0.8307,
+      "step": 14722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.134327523733894e-05,
+      "loss": 0.9875,
+      "step": 14723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.133034093416696e-05,
+      "loss": 1.1025,
+      "step": 14724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1317407391626e-05,
+      "loss": 0.702,
+      "step": 14725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.130447460997055e-05,
+      "loss": 1.0043,
+      "step": 14726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.129154258945489e-05,
+      "loss": 0.9693,
+      "step": 14727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.127861133033345e-05,
+      "loss": 0.9593,
+      "step": 14728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.12656808328606e-05,
+      "loss": 0.9996,
+      "step": 14729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.125275109729065e-05,
+      "loss": 1.0864,
+      "step": 14730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.123982212387793e-05,
+      "loss": 1.0637,
+      "step": 14731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.12268939128768e-05,
+      "loss": 0.9628,
+      "step": 14732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.121396646454149e-05,
+      "loss": 0.7849,
+      "step": 14733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.120103977912632e-05,
+      "loss": 1.0178,
+      "step": 14734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.118811385688559e-05,
+      "loss": 0.8564,
+      "step": 14735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.117518869807348e-05,
+      "loss": 0.7449,
+      "step": 14736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.116226430294424e-05,
+      "loss": 0.8481,
+      "step": 14737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.114934067175215e-05,
+      "loss": 0.9374,
+      "step": 14738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.113641780475137e-05,
+      "loss": 0.8278,
+      "step": 14739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.11234957021961e-05,
+      "loss": 1.0943,
+      "step": 14740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.111057436434056e-05,
+      "loss": 0.8883,
+      "step": 14741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.109765379143881e-05,
+      "loss": 0.9716,
+      "step": 14742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.10847339837451e-05,
+      "loss": 0.9551,
+      "step": 14743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.107181494151352e-05,
+      "loss": 1.1751,
+      "step": 14744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.105889666499819e-05,
+      "loss": 1.0843,
+      "step": 14745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.10459791544532e-05,
+      "loss": 1.1067,
+      "step": 14746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.103306241013268e-05,
+      "loss": 1.1006,
+      "step": 14747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1020146432290625e-05,
+      "loss": 1.0606,
+      "step": 14748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1007231221181206e-05,
+      "loss": 0.8601,
+      "step": 14749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.099431677705833e-05,
+      "loss": 0.833,
+      "step": 14750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.098140310017613e-05,
+      "loss": 1.0015,
+      "step": 14751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.096849019078857e-05,
+      "loss": 0.7767,
+      "step": 14752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.095557804914965e-05,
+      "loss": 1.12,
+      "step": 14753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.094266667551334e-05,
+      "loss": 0.7793,
+      "step": 14754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.092975607013366e-05,
+      "loss": 0.6846,
+      "step": 14755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.09168462332645e-05,
+      "loss": 0.987,
+      "step": 14756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.090393716515983e-05,
+      "loss": 0.9642,
+      "step": 14757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.089102886607358e-05,
+      "loss": 0.9094,
+      "step": 14758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0878121336259605e-05,
+      "loss": 1.0431,
+      "step": 14759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.086521457597182e-05,
+      "loss": 0.8268,
+      "step": 14760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.085230858546415e-05,
+      "loss": 1.0074,
+      "step": 14761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.083940336499039e-05,
+      "loss": 0.8933,
+      "step": 14762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.082649891480441e-05,
+      "loss": 1.0072,
+      "step": 14763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081359523516006e-05,
+      "loss": 0.8903,
+      "step": 14764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0800692326311095e-05,
+      "loss": 1.0093,
+      "step": 14765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078779018851137e-05,
+      "loss": 1.1562,
+      "step": 14766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.077488882201468e-05,
+      "loss": 0.8405,
+      "step": 14767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.076198822707475e-05,
+      "loss": 0.8312,
+      "step": 14768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.074908840394534e-05,
+      "loss": 0.9021,
+      "step": 14769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0736189352880234e-05,
+      "loss": 1.0362,
+      "step": 14770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.072329107413308e-05,
+      "loss": 0.9548,
+      "step": 14771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.071039356795769e-05,
+      "loss": 0.9937,
+      "step": 14772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.069749683460765e-05,
+      "loss": 0.8729,
+      "step": 14773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0684600874336693e-05,
+      "loss": 1.2949,
+      "step": 14774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.067170568739851e-05,
+      "loss": 1.1265,
+      "step": 14775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.065881127404668e-05,
+      "loss": 0.8672,
+      "step": 14776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064591763453485e-05,
+      "loss": 0.8499,
+      "step": 14777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.063302476911672e-05,
+      "loss": 1.1,
+      "step": 14778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.062013267804578e-05,
+      "loss": 1.0193,
+      "step": 14779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0607241361575695e-05,
+      "loss": 1.0776,
+      "step": 14780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.059435081996001e-05,
+      "loss": 0.8334,
+      "step": 14781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.058146105345227e-05,
+      "loss": 1.0019,
+      "step": 14782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056857206230601e-05,
+      "loss": 0.9976,
+      "step": 14783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.05556838467748e-05,
+      "loss": 0.8907,
+      "step": 14784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.054279640711211e-05,
+      "loss": 0.9351,
+      "step": 14785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.052990974357145e-05,
+      "loss": 0.9841,
+      "step": 14786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.051702385640633e-05,
+      "loss": 0.9379,
+      "step": 14787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.050413874587013e-05,
+      "loss": 0.8984,
+      "step": 14788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049125441221639e-05,
+      "loss": 0.9193,
+      "step": 14789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.047837085569852e-05,
+      "loss": 0.7877,
+      "step": 14790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.046548807656991e-05,
+      "loss": 0.877,
+      "step": 14791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0452606075084e-05,
+      "loss": 1.0307,
+      "step": 14792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.043972485149414e-05,
+      "loss": 1.0103,
+      "step": 14793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0426844406053706e-05,
+      "loss": 1.0422,
+      "step": 14794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.041396473901614e-05,
+      "loss": 0.8344,
+      "step": 14795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.040108585063464e-05,
+      "loss": 1.0866,
+      "step": 14796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.038820774116265e-05,
+      "loss": 0.919,
+      "step": 14797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.037533041085346e-05,
+      "loss": 0.8288,
+      "step": 14798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.036245385996033e-05,
+      "loss": 0.9264,
+      "step": 14799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.034957808873655e-05,
+      "loss": 0.9604,
+      "step": 14800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033670309743542e-05,
+      "loss": 0.985,
+      "step": 14801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.032382888631016e-05,
+      "loss": 0.9282,
+      "step": 14802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0310955455614024e-05,
+      "loss": 0.7973,
+      "step": 14803
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.029808280560023e-05,
+      "loss": 0.7863,
+      "step": 14804
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0285210936521955e-05,
+      "loss": 0.8153,
+      "step": 14805
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0272339848632385e-05,
+      "loss": 0.8443,
+      "step": 14806
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0259469542184776e-05,
+      "loss": 0.9947,
+      "step": 14807
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.024660001743219e-05,
+      "loss": 1.0274,
+      "step": 14808
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.023373127462781e-05,
+      "loss": 1.033,
+      "step": 14809
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.022086331402479e-05,
+      "loss": 1.0227,
+      "step": 14810
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020799613587615e-05,
+      "loss": 0.9407,
+      "step": 14811
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0195129740435106e-05,
+      "loss": 0.8501,
+      "step": 14812
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.018226412795469e-05,
+      "loss": 0.9495,
+      "step": 14813
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.016939929868793e-05,
+      "loss": 0.9139,
+      "step": 14814
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.015653525288795e-05,
+      "loss": 0.8816,
+      "step": 14815
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.01436719908077e-05,
+      "loss": 0.8913,
+      "step": 14816
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013080951270025e-05,
+      "loss": 0.9896,
+      "step": 14817
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.011794781881866e-05,
+      "loss": 1.0222,
+      "step": 14818
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.010508690941579e-05,
+      "loss": 1.1008,
+      "step": 14819
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.009222678474469e-05,
+      "loss": 0.8703,
+      "step": 14820
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0079367445058354e-05,
+      "loss": 0.8861,
+      "step": 14821
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.006650889060964e-05,
+      "loss": 0.9091,
+      "step": 14822
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.005365112165151e-05,
+      "loss": 0.8409,
+      "step": 14823
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.004079413843693e-05,
+      "loss": 0.876,
+      "step": 14824
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0027937941218704e-05,
+      "loss": 0.8239,
+      "step": 14825
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0015082530249786e-05,
+      "loss": 0.947,
+      "step": 14826
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0002227905783024e-05,
+      "loss": 0.9202,
+      "step": 14827
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9989374068071234e-05,
+      "loss": 0.9563,
+      "step": 14828
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.997652101736726e-05,
+      "loss": 1.0029,
+      "step": 14829
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9963668753923964e-05,
+      "loss": 1.0614,
+      "step": 14830
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995081727799412e-05,
+      "loss": 0.9927,
+      "step": 14831
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9937966589830507e-05,
+      "loss": 0.8054,
+      "step": 14832
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.992511668968592e-05,
+      "loss": 0.9036,
+      "step": 14833
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9912267577813075e-05,
+      "loss": 1.0329,
+      "step": 14834
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.989941925446476e-05,
+      "loss": 0.914,
+      "step": 14835
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.988657171989371e-05,
+      "loss": 1.1124,
+      "step": 14836
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9873724974352585e-05,
+      "loss": 0.7712,
+      "step": 14837
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.986087901809413e-05,
+      "loss": 0.9919,
+      "step": 14838
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9848033851370966e-05,
+      "loss": 0.8674,
+      "step": 14839
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983518947443578e-05,
+      "loss": 1.0399,
+      "step": 14840
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9822345887541296e-05,
+      "loss": 0.8915,
+      "step": 14841
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9809503090940024e-05,
+      "loss": 0.9291,
+      "step": 14842
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9796661084884644e-05,
+      "loss": 0.9956,
+      "step": 14843
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.978381986962779e-05,
+      "loss": 0.9865,
+      "step": 14844
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.977097944542198e-05,
+      "loss": 0.7961,
+      "step": 14845
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.975813981251981e-05,
+      "loss": 0.8646,
+      "step": 14846
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9745300971173864e-05,
+      "loss": 0.9561,
+      "step": 14847
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9732462921636655e-05,
+      "loss": 1.0056,
+      "step": 14848
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9719625664160694e-05,
+      "loss": 0.8901,
+      "step": 14849
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9706789198998545e-05,
+      "loss": 0.9543,
+      "step": 14850
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.969395352640264e-05,
+      "loss": 1.0358,
+      "step": 14851
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.968111864662544e-05,
+      "loss": 0.9954,
+      "step": 14852
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9668284559919504e-05,
+      "loss": 0.9913,
+      "step": 14853
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.96554512665372e-05,
+      "loss": 0.955,
+      "step": 14854
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9642618766730964e-05,
+      "loss": 0.9773,
+      "step": 14855
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.962978706075326e-05,
+      "loss": 0.9059,
+      "step": 14856
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9616956148856406e-05,
+      "loss": 0.8419,
+      "step": 14857
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.960412603129289e-05,
+      "loss": 0.9892,
+      "step": 14858
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.959129670831497e-05,
+      "loss": 0.9371,
+      "step": 14859
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.957846818017507e-05,
+      "loss": 0.75,
+      "step": 14860
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.956564044712551e-05,
+      "loss": 0.9617,
+      "step": 14861
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.955281350941862e-05,
+      "loss": 1.0243,
+      "step": 14862
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.953998736730665e-05,
+      "loss": 1.0203,
+      "step": 14863
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9527162021042004e-05,
+      "loss": 1.1738,
+      "step": 14864
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9514337470876825e-05,
+      "loss": 0.913,
+      "step": 14865
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.950151371706345e-05,
+      "loss": 0.9846,
+      "step": 14866
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9488690759854125e-05,
+      "loss": 0.7475,
+      "step": 14867
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947586859950103e-05,
+      "loss": 0.8127,
+      "step": 14868
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.94630472362564e-05,
+      "loss": 1.0311,
+      "step": 14869
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9450226670372456e-05,
+      "loss": 0.9942,
+      "step": 14870
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9437406902101335e-05,
+      "loss": 0.9861,
+      "step": 14871
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.942458793169523e-05,
+      "loss": 0.9755,
+      "step": 14872
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.941176975940631e-05,
+      "loss": 0.9749,
+      "step": 14873
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.939895238548665e-05,
+      "loss": 0.9818,
+      "step": 14874
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.938613581018838e-05,
+      "loss": 0.98,
+      "step": 14875
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.937332003376366e-05,
+      "loss": 0.8931,
+      "step": 14876
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.936050505646452e-05,
+      "loss": 0.9158,
+      "step": 14877
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934769087854304e-05,
+      "loss": 1.1076,
+      "step": 14878
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9334877500251306e-05,
+      "loss": 0.9126,
+      "step": 14879
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.932206492184129e-05,
+      "loss": 0.9002,
+      "step": 14880
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.93092531435651e-05,
+      "loss": 0.9565,
+      "step": 14881
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.929644216567466e-05,
+      "loss": 0.9791,
+      "step": 14882
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.928363198842202e-05,
+      "loss": 1.0886,
+      "step": 14883
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.927082261205914e-05,
+      "loss": 0.9337,
+      "step": 14884
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.925801403683797e-05,
+      "loss": 0.9481,
+      "step": 14885
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9245206263010435e-05,
+      "loss": 1.1761,
+      "step": 14886
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9232399290828555e-05,
+      "loss": 0.8988,
+      "step": 14887
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9219593120544104e-05,
+      "loss": 0.98,
+      "step": 14888
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920678775240909e-05,
+      "loss": 1.1002,
+      "step": 14889
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.919398318667536e-05,
+      "loss": 0.9237,
+      "step": 14890
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.918117942359477e-05,
+      "loss": 0.9369,
+      "step": 14891
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.916837646341915e-05,
+      "loss": 1.044,
+      "step": 14892
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9155574306400395e-05,
+      "loss": 0.8505,
+      "step": 14893
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.914277295279027e-05,
+      "loss": 1.0257,
+      "step": 14894
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.91299724028406e-05,
+      "loss": 0.7386,
+      "step": 14895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9117172656803185e-05,
+      "loss": 0.9611,
+      "step": 14896
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910437371492975e-05,
+      "loss": 0.9291,
+      "step": 14897
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.909157557747206e-05,
+      "loss": 0.7436,
+      "step": 14898
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.907877824468191e-05,
+      "loss": 0.9739,
+      "step": 14899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906598171681097e-05,
+      "loss": 0.8988,
+      "step": 14900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9053185994110974e-05,
+      "loss": 1.1183,
+      "step": 14901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904039107683358e-05,
+      "loss": 1.0015,
+      "step": 14902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.902759696523046e-05,
+      "loss": 1.08,
+      "step": 14903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.901480365955336e-05,
+      "loss": 1.0154,
+      "step": 14904
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.900201116005381e-05,
+      "loss": 1.0396,
+      "step": 14905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.89892194669835e-05,
+      "loss": 0.94,
+      "step": 14906
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.897642858059407e-05,
+      "loss": 1.0248,
+      "step": 14907
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8963638501137045e-05,
+      "loss": 0.8262,
+      "step": 14908
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8950849228864025e-05,
+      "loss": 0.9558,
+      "step": 14909
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8938060764026634e-05,
+      "loss": 0.9978,
+      "step": 14910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8925273106876336e-05,
+      "loss": 1.1616,
+      "step": 14911
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.891248625766471e-05,
+      "loss": 1.0226,
+      "step": 14912
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889970021664328e-05,
+      "loss": 0.8665,
+      "step": 14913
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.888691498406354e-05,
+      "loss": 1.0811,
+      "step": 14914
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.887413056017692e-05,
+      "loss": 1.095,
+      "step": 14915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8861346945235e-05,
+      "loss": 1.0188,
+      "step": 14916
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.884856413948913e-05,
+      "loss": 0.9687,
+      "step": 14917
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.88357821431908e-05,
+      "loss": 1.1955,
+      "step": 14918
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.882300095659144e-05,
+      "loss": 0.9724,
+      "step": 14919
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.881022057994241e-05,
+      "loss": 0.8542,
+      "step": 14920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8797441013495104e-05,
+      "loss": 0.8958,
+      "step": 14921
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8784662257500966e-05,
+      "loss": 0.8225,
+      "step": 14922
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.877188431221128e-05,
+      "loss": 0.7841,
+      "step": 14923
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8759107177877424e-05,
+      "loss": 0.9375,
+      "step": 14924
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.87463308547507e-05,
+      "loss": 0.9852,
+      "step": 14925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.87335553430824e-05,
+      "loss": 1.0033,
+      "step": 14926
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.872078064312392e-05,
+      "loss": 0.8947,
+      "step": 14927
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870800675512641e-05,
+      "loss": 0.9237,
+      "step": 14928
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.86952336793412e-05,
+      "loss": 0.9224,
+      "step": 14929
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.868246141601954e-05,
+      "loss": 0.9376,
+      "step": 14930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8669689965412634e-05,
+      "loss": 0.9002,
+      "step": 14931
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.865691932777169e-05,
+      "loss": 0.8247,
+      "step": 14932
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.864414950334796e-05,
+      "loss": 0.898,
+      "step": 14933
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.863138049239254e-05,
+      "loss": 0.9615,
+      "step": 14934
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8618612295156675e-05,
+      "loss": 0.9524,
+      "step": 14935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8605844911891494e-05,
+      "loss": 0.8591,
+      "step": 14936
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8593078342848105e-05,
+      "loss": 0.9996,
+      "step": 14937
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.858031258827761e-05,
+      "loss": 0.8208,
+      "step": 14938
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.85675476484312e-05,
+      "loss": 0.9728,
+      "step": 14939
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.855478352355988e-05,
+      "loss": 0.8797,
+      "step": 14940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854202021391474e-05,
+      "loss": 0.7158,
+      "step": 14941
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852925771974686e-05,
+      "loss": 1.1185,
+      "step": 14942
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.851649604130723e-05,
+      "loss": 1.1109,
+      "step": 14943
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850373517884688e-05,
+      "loss": 0.9286,
+      "step": 14944
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.849097513261689e-05,
+      "loss": 0.8294,
+      "step": 14945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.847821590286815e-05,
+      "loss": 0.7668,
+      "step": 14946
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.846545748985172e-05,
+      "loss": 1.066,
+      "step": 14947
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8452699893818474e-05,
+      "loss": 1.1157,
+      "step": 14948
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.843994311501939e-05,
+      "loss": 0.9276,
+      "step": 14949
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.842718715370545e-05,
+      "loss": 0.9107,
+      "step": 14950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.841443201012746e-05,
+      "loss": 1.0019,
+      "step": 14951
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.840167768453637e-05,
+      "loss": 0.9693,
+      "step": 14952
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838892417718308e-05,
+      "loss": 1.1877,
+      "step": 14953
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8376171488318376e-05,
+      "loss": 0.9141,
+      "step": 14954
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.83634196181932e-05,
+      "loss": 0.8911,
+      "step": 14955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.835066856705833e-05,
+      "loss": 0.89,
+      "step": 14956
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833791833516453e-05,
+      "loss": 0.9772,
+      "step": 14957
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8325168922762654e-05,
+      "loss": 0.9517,
+      "step": 14958
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8312420330103514e-05,
+      "loss": 0.7436,
+      "step": 14959
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.829967255743785e-05,
+      "loss": 1.1255,
+      "step": 14960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.828692560501633e-05,
+      "loss": 0.9865,
+      "step": 14961
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8274179473089796e-05,
+      "loss": 0.8626,
+      "step": 14962
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826143416190892e-05,
+      "loss": 0.9065,
+      "step": 14963
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8248689671724375e-05,
+      "loss": 0.7942,
+      "step": 14964
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.82359460027869e-05,
+      "loss": 0.9497,
+      "step": 14965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8223203155347106e-05,
+      "loss": 0.9412,
+      "step": 14966
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.82104611296557e-05,
+      "loss": 1.1138,
+      "step": 14967
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.81977199259633e-05,
+      "loss": 1.028,
+      "step": 14968
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8184979544520465e-05,
+      "loss": 0.7968,
+      "step": 14969
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.817223998557789e-05,
+      "loss": 0.8673,
+      "step": 14970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.815950124938607e-05,
+      "loss": 0.7529,
+      "step": 14971
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.814676333619569e-05,
+      "loss": 1.0822,
+      "step": 14972
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8134026246257225e-05,
+      "loss": 0.9673,
+      "step": 14973
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.812128997982118e-05,
+      "loss": 1.0167,
+      "step": 14974
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.810855453713814e-05,
+      "loss": 0.8977,
+      "step": 14975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8095819918458674e-05,
+      "loss": 1.0943,
+      "step": 14976
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.80830861240331e-05,
+      "loss": 0.9187,
+      "step": 14977
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.807035315411199e-05,
+      "loss": 1.1026,
+      "step": 14978
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.805762100894583e-05,
+      "loss": 0.8774,
+      "step": 14979
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.804488968878503e-05,
+      "loss": 0.7611,
+      "step": 14980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8032159193879985e-05,
+      "loss": 0.8826,
+      "step": 14981
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.801942952448115e-05,
+      "loss": 0.9824,
+      "step": 14982
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8006700680838864e-05,
+      "loss": 0.7437,
+      "step": 14983
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.799397266320358e-05,
+      "loss": 1.0385,
+      "step": 14984
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.798124547182562e-05,
+      "loss": 0.9414,
+      "step": 14985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.796851910695527e-05,
+      "loss": 0.9568,
+      "step": 14986
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.79557935688429e-05,
+      "loss": 0.861,
+      "step": 14987
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.794306885773888e-05,
+      "loss": 0.8508,
+      "step": 14988
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.793034497389345e-05,
+      "loss": 0.8373,
+      "step": 14989
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.79176219175569e-05,
+      "loss": 1.1076,
+      "step": 14990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.790489968897943e-05,
+      "loss": 0.9574,
+      "step": 14991
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789217828841136e-05,
+      "loss": 0.6231,
+      "step": 14992
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787945771610296e-05,
+      "loss": 0.952,
+      "step": 14993
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.786673797230431e-05,
+      "loss": 1.0126,
+      "step": 14994
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7854019057265686e-05,
+      "loss": 0.8626,
+      "step": 14995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.78413009712373e-05,
+      "loss": 0.9098,
+      "step": 14996
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.782858371446927e-05,
+      "loss": 1.0447,
+      "step": 14997
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.781586728721172e-05,
+      "loss": 0.9576,
+      "step": 14998
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.780315168971486e-05,
+      "loss": 0.8371,
+      "step": 14999
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.779043692222872e-05,
+      "loss": 0.9644,
+      "step": 15000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777772298500347e-05,
+      "loss": 1.287,
+      "step": 15001
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7765009878289166e-05,
+      "loss": 0.8317,
+      "step": 15002
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.775229760233583e-05,
+      "loss": 1.1286,
+      "step": 15003
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.773958615739357e-05,
+      "loss": 1.0303,
+      "step": 15004
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.772687554371241e-05,
+      "loss": 0.8913,
+      "step": 15005
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771416576154238e-05,
+      "loss": 1.087,
+      "step": 15006
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.770145681113342e-05,
+      "loss": 0.784,
+      "step": 15007
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7688748692735596e-05,
+      "loss": 0.9479,
+      "step": 15008
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7676041406598834e-05,
+      "loss": 0.5999,
+      "step": 15009
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7663334952973046e-05,
+      "loss": 0.9618,
+      "step": 15010
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7650629332108255e-05,
+      "loss": 0.953,
+      "step": 15011
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7637924544254296e-05,
+      "loss": 1.0739,
+      "step": 15012
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.762522058966113e-05,
+      "loss": 1.1128,
+      "step": 15013
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.761251746857865e-05,
+      "loss": 0.9005,
+      "step": 15014
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7599815181256646e-05,
+      "loss": 1.0873,
+      "step": 15015
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7587113727945074e-05,
+      "loss": 0.9064,
+      "step": 15016
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7574413108893666e-05,
+      "loss": 0.8635,
+      "step": 15017
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7561713324352365e-05,
+      "loss": 0.847,
+      "step": 15018
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7549014374570884e-05,
+      "loss": 0.9155,
+      "step": 15019
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753631625979901e-05,
+      "loss": 0.8708,
+      "step": 15020
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.752361898028653e-05,
+      "loss": 0.9773,
+      "step": 15021
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.751092253628331e-05,
+      "loss": 0.8678,
+      "step": 15022
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.74982269280389e-05,
+      "loss": 0.978,
+      "step": 15023
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.74855321558031e-05,
+      "loss": 0.9184,
+      "step": 15024
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7472838219825684e-05,
+      "loss": 1.027,
+      "step": 15025
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.746014512035628e-05,
+      "loss": 0.9719,
+      "step": 15026
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7447452857644526e-05,
+      "loss": 1.0501,
+      "step": 15027
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7434761431940156e-05,
+      "loss": 0.8376,
+      "step": 15028
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7422070843492734e-05,
+      "loss": 0.8959,
+      "step": 15029
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7409381092551964e-05,
+      "loss": 0.7877,
+      "step": 15030
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7396692179367396e-05,
+      "loss": 0.9401,
+      "step": 15031
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7384004104188606e-05,
+      "loss": 1.0098,
+      "step": 15032
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.737131686726523e-05,
+      "loss": 0.9873,
+      "step": 15033
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7358630468846755e-05,
+      "loss": 0.8814,
+      "step": 15034
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7345944909182814e-05,
+      "loss": 1.0101,
+      "step": 15035
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7333260188522855e-05,
+      "loss": 0.7962,
+      "step": 15036
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.732057630711638e-05,
+      "loss": 0.8697,
+      "step": 15037
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.730789326521291e-05,
+      "loss": 0.9814,
+      "step": 15038
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.729521106306199e-05,
+      "loss": 1.0237,
+      "step": 15039
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7282529700912945e-05,
+      "loss": 0.9175,
+      "step": 15040
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.726984917901527e-05,
+      "loss": 0.911,
+      "step": 15041
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.725716949761843e-05,
+      "loss": 1.0046,
+      "step": 15042
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7244490656971815e-05,
+      "loss": 0.8989,
+      "step": 15043
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7231812657324756e-05,
+      "loss": 0.8551,
+      "step": 15044
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.721913549892674e-05,
+      "loss": 1.1407,
+      "step": 15045
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7206459182027004e-05,
+      "loss": 0.9316,
+      "step": 15046
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7193783706875e-05,
+      "loss": 0.9224,
+      "step": 15047
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7181109073720007e-05,
+      "loss": 0.8935,
+      "step": 15048
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7168435282811306e-05,
+      "loss": 0.8839,
+      "step": 15049
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71557623343982e-05,
+      "loss": 0.9209,
+      "step": 15050
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.714309022873004e-05,
+      "loss": 0.9971,
+      "step": 15051
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.713041896605602e-05,
+      "loss": 0.8942,
+      "step": 15052
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.711774854662535e-05,
+      "loss": 1.0308,
+      "step": 15053
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.710507897068734e-05,
+      "loss": 0.7691,
+      "step": 15054
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.709241023849117e-05,
+      "loss": 1.0296,
+      "step": 15055
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.707974235028601e-05,
+      "loss": 0.9616,
+      "step": 15056
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.706707530632103e-05,
+      "loss": 1.0867,
+      "step": 15057
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7054409106845385e-05,
+      "loss": 1.044,
+      "step": 15058
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7041743752108314e-05,
+      "loss": 0.8709,
+      "step": 15059
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.702907924235885e-05,
+      "loss": 1.1131,
+      "step": 15060
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701641557784609e-05,
+      "loss": 0.836,
+      "step": 15061
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.700375275881923e-05,
+      "loss": 1.063,
+      "step": 15062
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6991090785527225e-05,
+      "loss": 1.0224,
+      "step": 15063
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.697842965821924e-05,
+      "loss": 0.8647,
+      "step": 15064
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696576937714427e-05,
+      "loss": 1.06,
+      "step": 15065
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.695310994255131e-05,
+      "loss": 0.7131,
+      "step": 15066
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6940451354689396e-05,
+      "loss": 1.0456,
+      "step": 15067
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6927793613807624e-05,
+      "loss": 0.9312,
+      "step": 15068
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.69151367201548e-05,
+      "loss": 0.9665,
+      "step": 15069
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.690248067397996e-05,
+      "loss": 0.9496,
+      "step": 15070
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6889825475532096e-05,
+      "loss": 0.9771,
+      "step": 15071
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6877171125060083e-05,
+      "loss": 0.8719,
+      "step": 15072
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.686451762281282e-05,
+      "loss": 0.9258,
+      "step": 15073
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6851864969039246e-05,
+      "loss": 0.9153,
+      "step": 15074
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.683921316398817e-05,
+      "loss": 0.8692,
+      "step": 15075
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.682656220790854e-05,
+      "loss": 0.9807,
+      "step": 15076
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681391210104916e-05,
+      "loss": 0.9635,
+      "step": 15077
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.680126284365882e-05,
+      "loss": 1.049,
+      "step": 15078
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6788614435986396e-05,
+      "loss": 0.9244,
+      "step": 15079
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6775966878280616e-05,
+      "loss": 0.809,
+      "step": 15080
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.676332017079034e-05,
+      "loss": 0.8569,
+      "step": 15081
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.675067431376426e-05,
+      "loss": 0.7759,
+      "step": 15082
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6738029307451104e-05,
+      "loss": 1.0922,
+      "step": 15083
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.672538515209964e-05,
+      "loss": 0.9578,
+      "step": 15084
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671274184795865e-05,
+      "loss": 0.8215,
+      "step": 15085
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.670009939527666e-05,
+      "loss": 1.0297,
+      "step": 15086
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668745779430245e-05,
+      "loss": 0.9223,
+      "step": 15087
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.66748170452847e-05,
+      "loss": 1.0356,
+      "step": 15088
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.666217714847202e-05,
+      "loss": 0.944,
+      "step": 15089
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6649538104113e-05,
+      "loss": 0.9742,
+      "step": 15090
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.663689991245632e-05,
+      "loss": 0.9055,
+      "step": 15091
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.662426257375051e-05,
+      "loss": 1.0806,
+      "step": 15092
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6611626088244194e-05,
+      "loss": 0.8495,
+      "step": 15093
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.659899045618593e-05,
+      "loss": 1.0477,
+      "step": 15094
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.65863556778242e-05,
+      "loss": 1.1044,
+      "step": 15095
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.657372175340756e-05,
+      "loss": 0.9011,
+      "step": 15096
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.656108868318457e-05,
+      "loss": 0.8407,
+      "step": 15097
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.65484564674037e-05,
+      "loss": 1.022,
+      "step": 15098
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.65358251063134e-05,
+      "loss": 1.184,
+      "step": 15099
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652319460016209e-05,
+      "loss": 1.1091,
+      "step": 15100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.651056494919831e-05,
+      "loss": 0.9609,
+      "step": 15101
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.649793615367043e-05,
+      "loss": 0.91,
+      "step": 15102
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.648530821382683e-05,
+      "loss": 0.9153,
+      "step": 15103
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.647268112991591e-05,
+      "loss": 1.0781,
+      "step": 15104
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.646005490218612e-05,
+      "loss": 0.8021,
+      "step": 15105
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6447429530885764e-05,
+      "loss": 0.8994,
+      "step": 15106
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6434805016263145e-05,
+      "loss": 0.9243,
+      "step": 15107
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.642218135856666e-05,
+      "loss": 0.9541,
+      "step": 15108
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640955855804455e-05,
+      "loss": 1.0619,
+      "step": 15109
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.639693661494516e-05,
+      "loss": 1.0096,
+      "step": 15110
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6384315529516754e-05,
+      "loss": 0.9395,
+      "step": 15111
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.637169530200753e-05,
+      "loss": 1.057,
+      "step": 15112
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6359075932665775e-05,
+      "loss": 1.0047,
+      "step": 15113
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.634645742173979e-05,
+      "loss": 0.7722,
+      "step": 15114
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.63338397694776e-05,
+      "loss": 1.0255,
+      "step": 15115
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.632122297612751e-05,
+      "loss": 0.8077,
+      "step": 15116
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.63086070419377e-05,
+      "loss": 0.9125,
+      "step": 15117
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6295991967156314e-05,
+      "loss": 0.9816,
+      "step": 15118
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628337775203143e-05,
+      "loss": 0.9676,
+      "step": 15119
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.627076439681125e-05,
+      "loss": 1.0177,
+      "step": 15120
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.625815190174382e-05,
+      "loss": 0.9513,
+      "step": 15121
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6245540267077266e-05,
+      "loss": 1.1337,
+      "step": 15122
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.623292949305966e-05,
+      "loss": 1.0606,
+      "step": 15123
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6220319579938996e-05,
+      "loss": 1.1478,
+      "step": 15124
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.620771052796339e-05,
+      "loss": 1.1495,
+      "step": 15125
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.619510233738079e-05,
+      "loss": 0.8507,
+      "step": 15126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618249500843926e-05,
+      "loss": 0.9758,
+      "step": 15127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.616988854138677e-05,
+      "loss": 0.9164,
+      "step": 15128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.615728293647125e-05,
+      "loss": 0.8468,
+      "step": 15129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.614467819394066e-05,
+      "loss": 0.9249,
+      "step": 15130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6132074314043035e-05,
+      "loss": 0.9997,
+      "step": 15131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.611947129702615e-05,
+      "loss": 1.0956,
+      "step": 15132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.610686914313796e-05,
+      "loss": 0.9094,
+      "step": 15133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.609426785262639e-05,
+      "loss": 1.064,
+      "step": 15134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608166742573928e-05,
+      "loss": 0.8789,
+      "step": 15135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.606906786272445e-05,
+      "loss": 0.8685,
+      "step": 15136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.605646916382978e-05,
+      "loss": 1.0709,
+      "step": 15137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.604387132930302e-05,
+      "loss": 1.1145,
+      "step": 15138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.603127435939207e-05,
+      "loss": 1.0216,
+      "step": 15139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.601867825434465e-05,
+      "loss": 0.8619,
+      "step": 15140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6006083014408484e-05,
+      "loss": 1.0278,
+      "step": 15141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5993488639831415e-05,
+      "loss": 0.9699,
+      "step": 15142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.598089513086108e-05,
+      "loss": 1.0695,
+      "step": 15143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.596830248774527e-05,
+      "loss": 1.0451,
+      "step": 15144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.595571071073167e-05,
+      "loss": 0.9766,
+      "step": 15145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.594311980006788e-05,
+      "loss": 1.0458,
+      "step": 15146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.593052975600168e-05,
+      "loss": 0.7242,
+      "step": 15147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.591794057878063e-05,
+      "loss": 0.9519,
+      "step": 15148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.590535226865238e-05,
+      "loss": 0.8154,
+      "step": 15149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.589276482586454e-05,
+      "loss": 1.1744,
+      "step": 15150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.588017825066474e-05,
+      "loss": 0.926,
+      "step": 15151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.586759254330054e-05,
+      "loss": 0.8165,
+      "step": 15152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5855007704019456e-05,
+      "loss": 0.7663,
+      "step": 15153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5842423733069105e-05,
+      "loss": 1.0993,
+      "step": 15154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5829840630696926e-05,
+      "loss": 0.8581,
+      "step": 15155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.581725839715053e-05,
+      "loss": 1.031,
+      "step": 15156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5804677032677354e-05,
+      "loss": 0.9101,
+      "step": 15157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.579209653752483e-05,
+      "loss": 0.8302,
+      "step": 15158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5779516911940466e-05,
+      "loss": 0.8414,
+      "step": 15159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5766938156171774e-05,
+      "loss": 1.0298,
+      "step": 15160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5754360270466036e-05,
+      "loss": 1.0055,
+      "step": 15161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.574178325507071e-05,
+      "loss": 0.9312,
+      "step": 15162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572920711023324e-05,
+      "loss": 0.9615,
+      "step": 15163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5716631836200974e-05,
+      "loss": 1.0362,
+      "step": 15164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.570405743322124e-05,
+      "loss": 0.8924,
+      "step": 15165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5691483901541354e-05,
+      "loss": 0.8987,
+      "step": 15166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.567891124140867e-05,
+      "loss": 0.8832,
+      "step": 15167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.566633945307052e-05,
+      "loss": 0.906,
+      "step": 15168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5653768536774175e-05,
+      "loss": 0.9481,
+      "step": 15169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.564119849276686e-05,
+      "loss": 1.1675,
+      "step": 15170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.56286293212959e-05,
+      "loss": 1.0208,
+      "step": 15171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5616061022608445e-05,
+      "loss": 0.9409,
+      "step": 15172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.560349359695181e-05,
+      "loss": 0.8656,
+      "step": 15173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.559092704457313e-05,
+      "loss": 0.952,
+      "step": 15174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.557836136571957e-05,
+      "loss": 0.9811,
+      "step": 15175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5565796560638326e-05,
+      "loss": 0.7904,
+      "step": 15176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5553232629576635e-05,
+      "loss": 0.9803,
+      "step": 15177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554066957278147e-05,
+      "loss": 0.9562,
+      "step": 15178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.552810739050002e-05,
+      "loss": 1.0042,
+      "step": 15179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.551554608297942e-05,
+      "loss": 1.1374,
+      "step": 15180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.550298565046672e-05,
+      "loss": 1.0967,
+      "step": 15181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549042609320893e-05,
+      "loss": 0.8554,
+      "step": 15182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54778674114532e-05,
+      "loss": 0.8267,
+      "step": 15183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.546530960544645e-05,
+      "loss": 1.0341,
+      "step": 15184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5452752675435796e-05,
+      "loss": 1.0049,
+      "step": 15185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.544019662166817e-05,
+      "loss": 0.9424,
+      "step": 15186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.542764144439052e-05,
+      "loss": 0.8329,
+      "step": 15187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5415087143849886e-05,
+      "loss": 0.81,
+      "step": 15188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.540253372029315e-05,
+      "loss": 0.8802,
+      "step": 15189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.538998117396728e-05,
+      "loss": 1.0187,
+      "step": 15190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.537742950511917e-05,
+      "loss": 1.1668,
+      "step": 15191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.536487871399567e-05,
+      "loss": 1.0603,
+      "step": 15192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5352328800843724e-05,
+      "loss": 0.8423,
+      "step": 15193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.533977976591014e-05,
+      "loss": 0.7521,
+      "step": 15194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.532723160944174e-05,
+      "loss": 0.8925,
+      "step": 15195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.531468433168539e-05,
+      "loss": 1.0168,
+      "step": 15196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53021379328879e-05,
+      "loss": 1.073,
+      "step": 15197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.528959241329603e-05,
+      "loss": 1.0526,
+      "step": 15198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5277047773156544e-05,
+      "loss": 0.8341,
+      "step": 15199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5264504012716225e-05,
+      "loss": 0.8611,
+      "step": 15200
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5251961132221754e-05,
+      "loss": 0.9699,
+      "step": 15201
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523941913191994e-05,
+      "loss": 1.1392,
+      "step": 15202
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.522687801205742e-05,
+      "loss": 0.8277,
+      "step": 15203
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.521433777288084e-05,
+      "loss": 1.0344,
+      "step": 15204
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520179841463693e-05,
+      "loss": 0.8311,
+      "step": 15205
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.518925993757239e-05,
+      "loss": 0.9649,
+      "step": 15206
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.517672234193372e-05,
+      "loss": 1.0507,
+      "step": 15207
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.516418562796761e-05,
+      "loss": 1.0124,
+      "step": 15208
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5151649795920624e-05,
+      "loss": 0.8376,
+      "step": 15209
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.51391148460394e-05,
+      "loss": 0.9117,
+      "step": 15210
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5126580778570466e-05,
+      "loss": 0.9968,
+      "step": 15211
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511404759376032e-05,
+      "loss": 0.6694,
+      "step": 15212
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.510151529185553e-05,
+      "loss": 0.7748,
+      "step": 15213
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.508898387310265e-05,
+      "loss": 0.8441,
+      "step": 15214
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.507645333774814e-05,
+      "loss": 0.8509,
+      "step": 15215
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.506392368603841e-05,
+      "loss": 0.7903,
+      "step": 15216
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5051394918220025e-05,
+      "loss": 0.8853,
+      "step": 15217
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.503886703453933e-05,
+      "loss": 0.8932,
+      "step": 15218
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5026340035242827e-05,
+      "loss": 1.0842,
+      "step": 15219
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.501381392057691e-05,
+      "loss": 0.9995,
+      "step": 15220
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5001288690787886e-05,
+      "loss": 0.9099,
+      "step": 15221
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.498876434612219e-05,
+      "loss": 0.8061,
+      "step": 15222
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.497624088682625e-05,
+      "loss": 0.6812,
+      "step": 15223
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.496371831314624e-05,
+      "loss": 0.8562,
+      "step": 15224
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.495119662532857e-05,
+      "loss": 0.8875,
+      "step": 15225
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4938675823619555e-05,
+      "loss": 1.113,
+      "step": 15226
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.492615590826547e-05,
+      "loss": 0.9905,
+      "step": 15227
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.491363687951252e-05,
+      "loss": 0.9524,
+      "step": 15228
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490111873760705e-05,
+      "loss": 1.028,
+      "step": 15229
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.48886014827952e-05,
+      "loss": 1.0278,
+      "step": 15230
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487608511532326e-05,
+      "loss": 1.014,
+      "step": 15231
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.48635696354374e-05,
+      "loss": 0.9577,
+      "step": 15232
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4851055043383744e-05,
+      "loss": 0.9315,
+      "step": 15233
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4838541339408554e-05,
+      "loss": 0.9341,
+      "step": 15234
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4826028523757886e-05,
+      "loss": 0.9429,
+      "step": 15235
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.481351659667793e-05,
+      "loss": 0.9286,
+      "step": 15236
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.480100555841478e-05,
+      "loss": 1.0759,
+      "step": 15237
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.478849540921447e-05,
+      "loss": 0.8323,
+      "step": 15238
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4775986149323154e-05,
+      "loss": 0.8577,
+      "step": 15239
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476347777898686e-05,
+      "loss": 1.0337,
+      "step": 15240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4750970298451574e-05,
+      "loss": 1.103,
+      "step": 15241
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.473846370796338e-05,
+      "loss": 0.87,
+      "step": 15242
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4725958007768304e-05,
+      "loss": 0.904,
+      "step": 15243
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4713453198112294e-05,
+      "loss": 0.9538,
+      "step": 15244
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4700949279241295e-05,
+      "loss": 1.0524,
+      "step": 15245
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.468844625140131e-05,
+      "loss": 0.9752,
+      "step": 15246
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4675944114838216e-05,
+      "loss": 0.761,
+      "step": 15247
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.466344286979801e-05,
+      "loss": 0.8074,
+      "step": 15248
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4650942516526536e-05,
+      "loss": 0.8294,
+      "step": 15249
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.463844305526964e-05,
+      "loss": 0.9853,
+      "step": 15250
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4625944486273275e-05,
+      "loss": 1.1138,
+      "step": 15251
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4613446809783244e-05,
+      "loss": 0.9374,
+      "step": 15252
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4600950026045326e-05,
+      "loss": 0.8898,
+      "step": 15253
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.458845413530543e-05,
+      "loss": 1.1138,
+      "step": 15254
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.457595913780926e-05,
+      "loss": 0.9091,
+      "step": 15255
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456346503380268e-05,
+      "loss": 0.771,
+      "step": 15256
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.455097182353137e-05,
+      "loss": 1.0118,
+      "step": 15257
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.453847950724109e-05,
+      "loss": 0.8908,
+      "step": 15258
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.452598808517757e-05,
+      "loss": 0.8332,
+      "step": 15259
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4513497557586554e-05,
+      "loss": 1.0546,
+      "step": 15260
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4501007924713707e-05,
+      "loss": 0.6114,
+      "step": 15261
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.448851918680464e-05,
+      "loss": 1.1376,
+      "step": 15262
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.447603134410512e-05,
+      "loss": 1.0727,
+      "step": 15263
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.446354439686066e-05,
+      "loss": 1.1813,
+      "step": 15264
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4451058345316985e-05,
+      "loss": 0.8865,
+      "step": 15265
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4438573189719636e-05,
+      "loss": 1.0793,
+      "step": 15266
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.442608893031418e-05,
+      "loss": 0.9583,
+      "step": 15267
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.44136055673462e-05,
+      "loss": 0.9666,
+      "step": 15268
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.440112310106134e-05,
+      "loss": 0.8778,
+      "step": 15269
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.438864153170495e-05,
+      "loss": 1.0301,
+      "step": 15270
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.437616085952265e-05,
+      "loss": 0.9166,
+      "step": 15271
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4363681084759955e-05,
+      "loss": 0.7672,
+      "step": 15272
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.43512022076623e-05,
+      "loss": 0.9206,
+      "step": 15273
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4338724228475145e-05,
+      "loss": 0.9656,
+      "step": 15274
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.432624714744391e-05,
+      "loss": 1.133,
+      "step": 15275
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.431377096481404e-05,
+      "loss": 0.8959,
+      "step": 15276
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4301295680830976e-05,
+      "loss": 0.8933,
+      "step": 15277
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.428882129574009e-05,
+      "loss": 1.0564,
+      "step": 15278
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4276347809786695e-05,
+      "loss": 1.0214,
+      "step": 15279
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.426387522321623e-05,
+      "loss": 0.8877,
+      "step": 15280
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.425140353627395e-05,
+      "loss": 0.9003,
+      "step": 15281
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.423893274920524e-05,
+      "loss": 0.9366,
+      "step": 15282
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.422646286225539e-05,
+      "loss": 0.8972,
+      "step": 15283
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4213993875669614e-05,
+      "loss": 0.7989,
+      "step": 15284
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.420152578969326e-05,
+      "loss": 0.8991,
+      "step": 15285
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.418905860457155e-05,
+      "loss": 0.9992,
+      "step": 15286
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417659232054967e-05,
+      "loss": 0.9824,
+      "step": 15287
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.416412693787285e-05,
+      "loss": 1.144,
+      "step": 15288
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.415166245678636e-05,
+      "loss": 0.9493,
+      "step": 15289
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.413919887753529e-05,
+      "loss": 0.894,
+      "step": 15290
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4126736200364813e-05,
+      "loss": 0.899,
+      "step": 15291
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4114274425520106e-05,
+      "loss": 0.8477,
+      "step": 15292
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.410181355324622e-05,
+      "loss": 0.839,
+      "step": 15293
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.408935358378835e-05,
+      "loss": 0.9984,
+      "step": 15294
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407689451739155e-05,
+      "loss": 0.9102,
+      "step": 15295
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4064436354300826e-05,
+      "loss": 1.0196,
+      "step": 15296
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.405197909476133e-05,
+      "loss": 1.0124,
+      "step": 15297
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.403952273901804e-05,
+      "loss": 1.0895,
+      "step": 15298
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402706728731595e-05,
+      "loss": 1.0272,
+      "step": 15299
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.401461273990013e-05,
+      "loss": 0.9203,
+      "step": 15300
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.400215909701547e-05,
+      "loss": 1.0288,
+      "step": 15301
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3989706358907035e-05,
+      "loss": 0.9921,
+      "step": 15302
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.39772545258197e-05,
+      "loss": 0.9953,
+      "step": 15303
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.396480359799838e-05,
+      "loss": 0.8501,
+      "step": 15304
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3952353575688016e-05,
+      "loss": 1.1598,
+      "step": 15305
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.393990445913354e-05,
+      "loss": 0.8608,
+      "step": 15306
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.392745624857978e-05,
+      "loss": 1.0036,
+      "step": 15307
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391500894427155e-05,
+      "loss": 0.9346,
+      "step": 15308
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390256254645378e-05,
+      "loss": 1.0624,
+      "step": 15309
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3890117055371195e-05,
+      "loss": 0.7849,
+      "step": 15310
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.387767247126869e-05,
+      "loss": 0.8778,
+      "step": 15311
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.386522879439102e-05,
+      "loss": 0.775,
+      "step": 15312
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.385278602498288e-05,
+      "loss": 1.1132,
+      "step": 15313
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.384034416328908e-05,
+      "loss": 1.1219,
+      "step": 15314
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.382790320955443e-05,
+      "loss": 1.0246,
+      "step": 15315
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3815463164023484e-05,
+      "loss": 0.9392,
+      "step": 15316
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.380302402694104e-05,
+      "loss": 1.0288,
+      "step": 15317
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.379058579855172e-05,
+      "loss": 0.8837,
+      "step": 15318
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.377814847910026e-05,
+      "loss": 0.8546,
+      "step": 15319
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3765712068831244e-05,
+      "loss": 0.9964,
+      "step": 15320
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3753276567989266e-05,
+      "loss": 1.0275,
+      "step": 15321
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3740841976818965e-05,
+      "loss": 0.9725,
+      "step": 15322
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3728408295565005e-05,
+      "loss": 1.0844,
+      "step": 15323
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.371597552447186e-05,
+      "loss": 0.7838,
+      "step": 15324
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3703543663784074e-05,
+      "loss": 0.9034,
+      "step": 15325
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369111271374627e-05,
+      "loss": 0.8941,
+      "step": 15326
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3678682674602866e-05,
+      "loss": 1.0886,
+      "step": 15327
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.366625354659844e-05,
+      "loss": 0.9675,
+      "step": 15328
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3653825329977446e-05,
+      "loss": 0.9759,
+      "step": 15329
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.364139802498429e-05,
+      "loss": 0.7672,
+      "step": 15330
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.362897163186351e-05,
+      "loss": 1.0132,
+      "step": 15331
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.361654615085949e-05,
+      "loss": 1.0236,
+      "step": 15332
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.36041215822166e-05,
+      "loss": 0.837,
+      "step": 15333
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3591697926179275e-05,
+      "loss": 1.0576,
+      "step": 15334
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357927518299191e-05,
+      "loss": 0.9022,
+      "step": 15335
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.356685335289884e-05,
+      "loss": 0.8649,
+      "step": 15336
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355443243614434e-05,
+      "loss": 1.0403,
+      "step": 15337
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3542012432972834e-05,
+      "loss": 0.9124,
+      "step": 15338
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3529593343628525e-05,
+      "loss": 0.8518,
+      "step": 15339
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3517175168355796e-05,
+      "loss": 0.8522,
+      "step": 15340
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3504757907398836e-05,
+      "loss": 0.8241,
+      "step": 15341
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.349234156100189e-05,
+      "loss": 0.8719,
+      "step": 15342
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.347992612940926e-05,
+      "loss": 0.9878,
+      "step": 15343
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.34675116128651e-05,
+      "loss": 0.8809,
+      "step": 15344
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.345509801161357e-05,
+      "loss": 0.8848,
+      "step": 15345
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3442685325898937e-05,
+      "loss": 0.8519,
+      "step": 15346
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.343027355596526e-05,
+      "loss": 1.0621,
+      "step": 15347
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.341786270205678e-05,
+      "loss": 1.1266,
+      "step": 15348
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.340545276441755e-05,
+      "loss": 0.9942,
+      "step": 15349
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.339304374329165e-05,
+      "loss": 1.0201,
+      "step": 15350
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.338063563892322e-05,
+      "loss": 0.9075,
+      "step": 15351
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.336822845155633e-05,
+      "loss": 0.9535,
+      "step": 15352
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.335582218143501e-05,
+      "loss": 0.9216,
+      "step": 15353
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.334341682880326e-05,
+      "loss": 0.9281,
+      "step": 15354
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3331012393905165e-05,
+      "loss": 1.0927,
+      "step": 15355
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.331860887698463e-05,
+      "loss": 0.954,
+      "step": 15356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3306206278285734e-05,
+      "loss": 1.0618,
+      "step": 15357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.329380459805237e-05,
+      "loss": 1.0438,
+      "step": 15358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.328140383652844e-05,
+      "loss": 0.8361,
+      "step": 15359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3269003993957936e-05,
+      "loss": 0.8906,
+      "step": 15360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.325660507058482e-05,
+      "loss": 0.9294,
+      "step": 15361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.324420706665282e-05,
+      "loss": 0.8543,
+      "step": 15362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.323180998240592e-05,
+      "loss": 0.8762,
+      "step": 15363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.32194138180879e-05,
+      "loss": 0.9189,
+      "step": 15364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.320701857394268e-05,
+      "loss": 0.9265,
+      "step": 15365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319462425021402e-05,
+      "loss": 0.9213,
+      "step": 15366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3182230847145666e-05,
+      "loss": 0.917,
+      "step": 15367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.316983836498146e-05,
+      "loss": 0.9368,
+      "step": 15368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.31574468039652e-05,
+      "loss": 0.8948,
+      "step": 15369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.314505616434055e-05,
+      "loss": 0.9311,
+      "step": 15370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.313266644635124e-05,
+      "loss": 0.8771,
+      "step": 15371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3120277650241035e-05,
+      "loss": 1.1718,
+      "step": 15372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3107889776253536e-05,
+      "loss": 1.0715,
+      "step": 15373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3095502824632516e-05,
+      "loss": 1.0679,
+      "step": 15374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.308311679562157e-05,
+      "loss": 0.9512,
+      "step": 15375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.307073168946428e-05,
+      "loss": 0.9033,
+      "step": 15376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.305834750640435e-05,
+      "loss": 0.8931,
+      "step": 15377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.304596424668535e-05,
+      "loss": 1.0719,
+      "step": 15378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30335819105508e-05,
+      "loss": 1.0102,
+      "step": 15379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.302120049824431e-05,
+      "loss": 0.9919,
+      "step": 15380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3008820010009464e-05,
+      "loss": 0.9661,
+      "step": 15381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2996440446089735e-05,
+      "loss": 0.8953,
+      "step": 15382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.298406180672863e-05,
+      "loss": 0.9734,
+      "step": 15383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.29716840921696e-05,
+      "loss": 0.9874,
+      "step": 15384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.295930730265617e-05,
+      "loss": 0.9328,
+      "step": 15385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2946931438431813e-05,
+      "loss": 0.9647,
+      "step": 15386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2934556499739904e-05,
+      "loss": 0.8101,
+      "step": 15387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2922182486823855e-05,
+      "loss": 0.9491,
+      "step": 15388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.290980939992713e-05,
+      "loss": 1.105,
+      "step": 15389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2897437239293056e-05,
+      "loss": 0.9634,
+      "step": 15390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2885066005164976e-05,
+      "loss": 0.8416,
+      "step": 15391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2872695697786276e-05,
+      "loss": 1.1514,
+      "step": 15392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286032631740023e-05,
+      "loss": 0.8279,
+      "step": 15393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.284795786425022e-05,
+      "loss": 0.7942,
+      "step": 15394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.283559033857948e-05,
+      "loss": 0.7245,
+      "step": 15395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2823223740631246e-05,
+      "loss": 0.8204,
+      "step": 15396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.281085807064881e-05,
+      "loss": 0.8879,
+      "step": 15397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.279849332887543e-05,
+      "loss": 1.0927,
+      "step": 15398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2786129515554307e-05,
+      "loss": 1.0063,
+      "step": 15399
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2773766630928564e-05,
+      "loss": 1.0791,
+      "step": 15400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2761404675241496e-05,
+      "loss": 0.7832,
+      "step": 15401
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.274904364873616e-05,
+      "loss": 0.8597,
+      "step": 15402
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.273668355165577e-05,
+      "loss": 1.2406,
+      "step": 15403
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.272432438424344e-05,
+      "loss": 0.7987,
+      "step": 15404
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.27119661467422e-05,
+      "loss": 0.972,
+      "step": 15405
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.269960883939522e-05,
+      "loss": 0.9403,
+      "step": 15406
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268725246244555e-05,
+      "loss": 0.92,
+      "step": 15407
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.267489701613619e-05,
+      "loss": 0.9397,
+      "step": 15408
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2662542500710246e-05,
+      "loss": 0.9538,
+      "step": 15409
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2650188916410644e-05,
+      "loss": 1.0056,
+      "step": 15410
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.263783626348048e-05,
+      "loss": 1.015,
+      "step": 15411
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.262548454216267e-05,
+      "loss": 0.973,
+      "step": 15412
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.261313375270014e-05,
+      "loss": 0.9236,
+      "step": 15413
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.260078389533587e-05,
+      "loss": 0.8738,
+      "step": 15414
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2588434970312825e-05,
+      "loss": 0.923,
+      "step": 15415
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2576086977873864e-05,
+      "loss": 0.8992,
+      "step": 15416
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.256373991826183e-05,
+      "loss": 0.8356,
+      "step": 15417
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.255139379171967e-05,
+      "loss": 0.9824,
+      "step": 15418
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.253904859849017e-05,
+      "loss": 0.9201,
+      "step": 15419
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.25267043388162e-05,
+      "loss": 0.7542,
+      "step": 15420
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.251436101294056e-05,
+      "loss": 0.9068,
+      "step": 15421
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2502018621106e-05,
+      "loss": 1.0378,
+      "step": 15422
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.248967716355537e-05,
+      "loss": 1.0346,
+      "step": 15423
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247733664053138e-05,
+      "loss": 0.9488,
+      "step": 15424
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.246499705227675e-05,
+      "loss": 0.9056,
+      "step": 15425
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.245265839903426e-05,
+      "loss": 1.0739,
+      "step": 15426
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.244032068104654e-05,
+      "loss": 0.8063,
+      "step": 15427
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2427983898556335e-05,
+      "loss": 1.0724,
+      "step": 15428
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2415648051806275e-05,
+      "loss": 0.9014,
+      "step": 15429
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.240331314103899e-05,
+      "loss": 0.9078,
+      "step": 15430
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2390979166497136e-05,
+      "loss": 0.9887,
+      "step": 15431
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.237864612842333e-05,
+      "loss": 0.9742,
+      "step": 15432
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.236631402706018e-05,
+      "loss": 1.0891,
+      "step": 15433
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2353982862650166e-05,
+      "loss": 0.9531,
+      "step": 15434
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.234165263543595e-05,
+      "loss": 0.8356,
+      "step": 15435
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.232932334566001e-05,
+      "loss": 0.89,
+      "step": 15436
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231699499356484e-05,
+      "loss": 0.8493,
+      "step": 15437
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.230466757939301e-05,
+      "loss": 1.0511,
+      "step": 15438
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2292341103386923e-05,
+      "loss": 1.0359,
+      "step": 15439
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.228001556578912e-05,
+      "loss": 0.9706,
+      "step": 15440
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2267690966841984e-05,
+      "loss": 0.7956,
+      "step": 15441
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.225536730678794e-05,
+      "loss": 1.1487,
+      "step": 15442
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2243044585869394e-05,
+      "loss": 1.0315,
+      "step": 15443
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.22307228043288e-05,
+      "loss": 0.8599,
+      "step": 15444
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.221840196240848e-05,
+      "loss": 0.8597,
+      "step": 15445
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.220608206035074e-05,
+      "loss": 1.0288,
+      "step": 15446
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.219376309839799e-05,
+      "loss": 0.8262,
+      "step": 15447
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.218144507679248e-05,
+      "loss": 1.2424,
+      "step": 15448
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2169127995776604e-05,
+      "loss": 0.8926,
+      "step": 15449
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.215681185559248e-05,
+      "loss": 0.779,
+      "step": 15450
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214449665648247e-05,
+      "loss": 0.8143,
+      "step": 15451
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.213218239868882e-05,
+      "loss": 0.882,
+      "step": 15452
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.211986908245374e-05,
+      "loss": 1.1997,
+      "step": 15453
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.210755670801939e-05,
+      "loss": 1.1646,
+      "step": 15454
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2095245275628016e-05,
+      "loss": 1.0117,
+      "step": 15455
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208293478552171e-05,
+      "loss": 0.9726,
+      "step": 15456
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2070625237942706e-05,
+      "loss": 1.0671,
+      "step": 15457
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.20583166331331e-05,
+      "loss": 0.9391,
+      "step": 15458
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2046008971334935e-05,
+      "loss": 0.9821,
+      "step": 15459
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2033702252790364e-05,
+      "loss": 0.9629,
+      "step": 15460
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202139647774149e-05,
+      "loss": 1.0149,
+      "step": 15461
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.200909164643033e-05,
+      "loss": 1.0146,
+      "step": 15462
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.199678775909889e-05,
+      "loss": 1.0051,
+      "step": 15463
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.198448481598923e-05,
+      "loss": 0.8161,
+      "step": 15464
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.197218281734333e-05,
+      "loss": 1.0399,
+      "step": 15465
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1959881763403194e-05,
+      "loss": 0.9448,
+      "step": 15466
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.194758165441077e-05,
+      "loss": 0.9236,
+      "step": 15467
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193528249060796e-05,
+      "loss": 1.1895,
+      "step": 15468
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.192298427223676e-05,
+      "loss": 0.768,
+      "step": 15469
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.191068699953905e-05,
+      "loss": 0.9784,
+      "step": 15470
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.189839067275666e-05,
+      "loss": 0.9968,
+      "step": 15471
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.188609529213153e-05,
+      "loss": 0.9544,
+      "step": 15472
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1873800857905474e-05,
+      "loss": 0.8988,
+      "step": 15473
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.186150737032035e-05,
+      "loss": 0.7815,
+      "step": 15474
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.184921482961797e-05,
+      "loss": 1.1179,
+      "step": 15475
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.183692323604006e-05,
+      "loss": 1.0113,
+      "step": 15476
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.182463258982846e-05,
+      "loss": 0.9564,
+      "step": 15477
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181234289122494e-05,
+      "loss": 0.9056,
+      "step": 15478
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.180005414047122e-05,
+      "loss": 0.7833,
+      "step": 15479
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1787766337808983e-05,
+      "loss": 0.9578,
+      "step": 15480
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1775479483479984e-05,
+      "loss": 0.8398,
+      "step": 15481
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176319357772589e-05,
+      "loss": 1.0388,
+      "step": 15482
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.175090862078831e-05,
+      "loss": 0.9616,
+      "step": 15483
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1738624612908973e-05,
+      "loss": 1.0258,
+      "step": 15484
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.172634155432943e-05,
+      "loss": 0.8971,
+      "step": 15485
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1714059445291354e-05,
+      "loss": 1.1906,
+      "step": 15486
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1701778286036304e-05,
+      "loss": 1.0441,
+      "step": 15487
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.168949807680581e-05,
+      "loss": 0.8505,
+      "step": 15488
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1677218817841464e-05,
+      "loss": 0.8998,
+      "step": 15489
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.166494050938484e-05,
+      "loss": 1.0576,
+      "step": 15490
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1652663151677404e-05,
+      "loss": 0.8576,
+      "step": 15491
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1640386744960655e-05,
+      "loss": 0.9936,
+      "step": 15492
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.162811128947602e-05,
+      "loss": 0.7469,
+      "step": 15493
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1615836785465036e-05,
+      "loss": 0.8119,
+      "step": 15494
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.160356323316917e-05,
+      "loss": 0.9246,
+      "step": 15495
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159129063282973e-05,
+      "loss": 1.0301,
+      "step": 15496
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.157901898468815e-05,
+      "loss": 0.7803,
+      "step": 15497
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156674828898589e-05,
+      "loss": 0.9276,
+      "step": 15498
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.155447854596424e-05,
+      "loss": 1.0723,
+      "step": 15499
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1542209755864556e-05,
+      "loss": 0.8389,
+      "step": 15500
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1529941918928204e-05,
+      "loss": 1.0287,
+      "step": 15501
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.151767503539642e-05,
+      "loss": 1.0356,
+      "step": 15502
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1505409105510586e-05,
+      "loss": 0.799,
+      "step": 15503
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149314412951192e-05,
+      "loss": 0.8973,
+      "step": 15504
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.148088010764163e-05,
+      "loss": 0.7459,
+      "step": 15505
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.146861704014099e-05,
+      "loss": 0.8938,
+      "step": 15506
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.145635492725127e-05,
+      "loss": 0.8272,
+      "step": 15507
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1444093769213617e-05,
+      "loss": 0.9633,
+      "step": 15508
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.143183356626916e-05,
+      "loss": 0.934,
+      "step": 15509
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.141957431865915e-05,
+      "loss": 1.0004,
+      "step": 15510
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.140731602662463e-05,
+      "loss": 0.9103,
+      "step": 15511
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1395058690406815e-05,
+      "loss": 0.9498,
+      "step": 15512
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.138280231024676e-05,
+      "loss": 0.8114,
+      "step": 15513
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137054688638549e-05,
+      "loss": 0.9686,
+      "step": 15514
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.13582924190642e-05,
+      "loss": 0.8947,
+      "step": 15515
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.134603890852384e-05,
+      "loss": 0.9709,
+      "step": 15516
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.133378635500543e-05,
+      "loss": 1.1168,
+      "step": 15517
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1321534758750037e-05,
+      "loss": 0.9503,
+      "step": 15518
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.130928411999858e-05,
+      "loss": 0.8658,
+      "step": 15519
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12970344389921e-05,
+      "loss": 0.7365,
+      "step": 15520
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1284785715971525e-05,
+      "loss": 0.9147,
+      "step": 15521
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.127253795117774e-05,
+      "loss": 0.8426,
+      "step": 15522
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.126029114485168e-05,
+      "loss": 1.1877,
+      "step": 15523
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12480452972343e-05,
+      "loss": 0.9915,
+      "step": 15524
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.123580040856644e-05,
+      "loss": 0.8591,
+      "step": 15525
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.12235564790889e-05,
+      "loss": 0.9657,
+      "step": 15526
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1211313509042594e-05,
+      "loss": 0.9112,
+      "step": 15527
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1199071498668316e-05,
+      "loss": 0.8482,
+      "step": 15528
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1186830448206825e-05,
+      "loss": 0.7246,
+      "step": 15529
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1174590357898975e-05,
+      "loss": 1.0328,
+      "step": 15530
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.116235122798545e-05,
+      "loss": 1.025,
+      "step": 15531
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.115011305870707e-05,
+      "loss": 0.9223,
+      "step": 15532
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.113787585030454e-05,
+      "loss": 0.8226,
+      "step": 15533
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1125639603018494e-05,
+      "loss": 1.0155,
+      "step": 15534
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1113404317089686e-05,
+      "loss": 0.801,
+      "step": 15535
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.11011699927588e-05,
+      "loss": 0.9167,
+      "step": 15536
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.108893663026646e-05,
+      "loss": 0.9138,
+      "step": 15537
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.107670422985329e-05,
+      "loss": 0.8086,
+      "step": 15538
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.106447279175987e-05,
+      "loss": 0.9369,
+      "step": 15539
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.105224231622683e-05,
+      "loss": 0.9561,
+      "step": 15540
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1040012803494795e-05,
+      "loss": 0.9816,
+      "step": 15541
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.10277842538042e-05,
+      "loss": 0.9693,
+      "step": 15542
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1015556667395636e-05,
+      "loss": 0.8965,
+      "step": 15543
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.100333004450968e-05,
+      "loss": 0.8791,
+      "step": 15544
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.099110438538677e-05,
+      "loss": 0.8524,
+      "step": 15545
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.097887969026733e-05,
+      "loss": 1.0578,
+      "step": 15546
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.096665595939194e-05,
+      "loss": 0.8272,
+      "step": 15547
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.095443319300094e-05,
+      "loss": 1.0037,
+      "step": 15548
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.094221139133482e-05,
+      "loss": 0.6993,
+      "step": 15549
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.092999055463396e-05,
+      "loss": 1.0015,
+      "step": 15550
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.09177706831387e-05,
+      "loss": 0.9571,
+      "step": 15551
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0905551777089434e-05,
+      "loss": 1.0348,
+      "step": 15552
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.089333383672656e-05,
+      "loss": 1.0241,
+      "step": 15553
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0881116862290355e-05,
+      "loss": 0.8989,
+      "step": 15554
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0868900854021095e-05,
+      "loss": 1.0355,
+      "step": 15555
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.085668581215915e-05,
+      "loss": 0.8869,
+      "step": 15556
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.084447173694469e-05,
+      "loss": 1.0375,
+      "step": 15557
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.083225862861811e-05,
+      "loss": 0.8487,
+      "step": 15558
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.082004648741947e-05,
+      "loss": 0.9525,
+      "step": 15559
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.080783531358906e-05,
+      "loss": 0.8548,
+      "step": 15560
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0795625107367105e-05,
+      "loss": 0.8421,
+      "step": 15561
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.078341586899378e-05,
+      "loss": 1.0029,
+      "step": 15562
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077120759870915e-05,
+      "loss": 0.9824,
+      "step": 15563
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075900029675346e-05,
+      "loss": 1.0275,
+      "step": 15564
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0746793963366744e-05,
+      "loss": 0.9577,
+      "step": 15565
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.073458859878917e-05,
+      "loss": 0.7562,
+      "step": 15566
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0722384203260786e-05,
+      "loss": 0.7876,
+      "step": 15567
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071018077702161e-05,
+      "loss": 0.958,
+      "step": 15568
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.069797832031171e-05,
+      "loss": 0.8987,
+      "step": 15569
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0685776833371176e-05,
+      "loss": 0.8885,
+      "step": 15570
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.067357631643995e-05,
+      "loss": 1.1612,
+      "step": 15571
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.066137676975799e-05,
+      "loss": 0.8325,
+      "step": 15572
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0649178193565314e-05,
+      "loss": 0.8223,
+      "step": 15573
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.063698058810187e-05,
+      "loss": 0.8796,
+      "step": 15574
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0624783953607504e-05,
+      "loss": 1.1375,
+      "step": 15575
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.061258829032223e-05,
+      "loss": 1.0099,
+      "step": 15576
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.060039359848583e-05,
+      "loss": 1.2824,
+      "step": 15577
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0588199878338284e-05,
+      "loss": 0.9618,
+      "step": 15578
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.057600713011938e-05,
+      "loss": 0.9592,
+      "step": 15579
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056381535406892e-05,
+      "loss": 0.9306,
+      "step": 15580
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.055162455042679e-05,
+      "loss": 1.021,
+      "step": 15581
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.053943471943271e-05,
+      "loss": 0.8889,
+      "step": 15582
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0527245861326536e-05,
+      "loss": 0.9402,
+      "step": 15583
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0515057976347945e-05,
+      "loss": 0.9706,
+      "step": 15584
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.050287106473669e-05,
+      "loss": 1.0635,
+      "step": 15585
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0490685126732493e-05,
+      "loss": 0.9979,
+      "step": 15586
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0478500162575134e-05,
+      "loss": 1.0746,
+      "step": 15587
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0466316172504133e-05,
+      "loss": 0.7179,
+      "step": 15588
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045413315675924e-05,
+      "loss": 1.0064,
+      "step": 15589
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.044195111558012e-05,
+      "loss": 1.1253,
+      "step": 15590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.042977004920635e-05,
+      "loss": 1.1199,
+      "step": 15591
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.041758995787751e-05,
+      "loss": 0.812,
+      "step": 15592
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0405410841833253e-05,
+      "loss": 0.9108,
+      "step": 15593
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0393232701313056e-05,
+      "loss": 0.8013,
+      "step": 15594
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0381055536556546e-05,
+      "loss": 1.0211,
+      "step": 15595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.036887934780321e-05,
+      "loss": 0.8309,
+      "step": 15596
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.035670413529251e-05,
+      "loss": 1.0509,
+      "step": 15597
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.034452989926397e-05,
+      "loss": 0.91,
+      "step": 15598
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0332356639957114e-05,
+      "loss": 0.9417,
+      "step": 15599
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.032018435761132e-05,
+      "loss": 0.7763,
+      "step": 15600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030801305246604e-05,
+      "loss": 0.8917,
+      "step": 15601
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.029584272476065e-05,
+      "loss": 1.107,
+      "step": 15602
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.028367337473454e-05,
+      "loss": 0.9136,
+      "step": 15603
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0271505002627207e-05,
+      "loss": 0.7926,
+      "step": 15604
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025933760867781e-05,
+      "loss": 0.978,
+      "step": 15605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0247171193125785e-05,
+      "loss": 0.9601,
+      "step": 15606
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.023500575621047e-05,
+      "loss": 1.1273,
+      "step": 15607
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.022284129817112e-05,
+      "loss": 1.045,
+      "step": 15608
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0210677819246974e-05,
+      "loss": 0.8686,
+      "step": 15609
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019851531967736e-05,
+      "loss": 0.9206,
+      "step": 15610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.018635379970145e-05,
+      "loss": 0.8933,
+      "step": 15611
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0174193259558514e-05,
+      "loss": 0.9478,
+      "step": 15612
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.016203369948773e-05,
+      "loss": 0.8107,
+      "step": 15613
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.014987511972824e-05,
+      "loss": 0.9564,
+      "step": 15614
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0137717520519235e-05,
+      "loss": 0.9179,
+      "step": 15615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.012556090209988e-05,
+      "loss": 1.161,
+      "step": 15616
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0113405264709276e-05,
+      "loss": 0.6868,
+      "step": 15617
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.010125060858648e-05,
+      "loss": 0.9307,
+      "step": 15618
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.008909693397066e-05,
+      "loss": 0.8992,
+      "step": 15619
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.007694424110081e-05,
+      "loss": 0.957,
+      "step": 15620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006479253021597e-05,
+      "loss": 1.0679,
+      "step": 15621
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.005264180155522e-05,
+      "loss": 0.9664,
+      "step": 15622
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0040492055357494e-05,
+      "loss": 1.1852,
+      "step": 15623
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0028343291861846e-05,
+      "loss": 0.7689,
+      "step": 15624
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.001619551130721e-05,
+      "loss": 0.8669,
+      "step": 15625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0004048713932496e-05,
+      "loss": 0.981,
+      "step": 15626
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.999190289997672e-05,
+      "loss": 0.9521,
+      "step": 15627
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.997975806967868e-05,
+      "loss": 1.0069,
+      "step": 15628
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996761422327737e-05,
+      "loss": 1.0074,
+      "step": 15629
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.995547136101161e-05,
+      "loss": 0.7623,
+      "step": 15630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99433294831202e-05,
+      "loss": 0.8079,
+      "step": 15631
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.993118858984204e-05,
+      "loss": 0.8862,
+      "step": 15632
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991904868141599e-05,
+      "loss": 1.1068,
+      "step": 15633
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.99069097580807e-05,
+      "loss": 0.8111,
+      "step": 15634
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9894771820075015e-05,
+      "loss": 0.8936,
+      "step": 15635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9882634867637724e-05,
+      "loss": 0.7978,
+      "step": 15636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9870498901007524e-05,
+      "loss": 0.9297,
+      "step": 15637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.98583639204231e-05,
+      "loss": 1.068,
+      "step": 15638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984622992612321e-05,
+      "loss": 0.7983,
+      "step": 15639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.983409691834646e-05,
+      "loss": 1.088,
+      "step": 15640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9821964897331586e-05,
+      "loss": 1.0072,
+      "step": 15641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9809833863317165e-05,
+      "loss": 0.8999,
+      "step": 15642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979770381654181e-05,
+      "loss": 0.9582,
+      "step": 15643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.978557475724412e-05,
+      "loss": 1.0424,
+      "step": 15644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.977344668566275e-05,
+      "loss": 0.7406,
+      "step": 15645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.976131960203618e-05,
+      "loss": 1.0089,
+      "step": 15646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974919350660299e-05,
+      "loss": 0.8901,
+      "step": 15647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9737068399601624e-05,
+      "loss": 0.9716,
+      "step": 15648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9724944281270656e-05,
+      "loss": 0.9033,
+      "step": 15649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971282115184862e-05,
+      "loss": 1.0346,
+      "step": 15650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9700699011573825e-05,
+      "loss": 0.9312,
+      "step": 15651
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.96885778606848e-05,
+      "loss": 1.1594,
+      "step": 15652
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.967645769942e-05,
+      "loss": 0.9622,
+      "step": 15653
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.966433852801778e-05,
+      "loss": 0.9657,
+      "step": 15654
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.965222034671652e-05,
+      "loss": 0.8894,
+      "step": 15655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.964010315575462e-05,
+      "loss": 1.1016,
+      "step": 15656
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9627986955370374e-05,
+      "loss": 1.2012,
+      "step": 15657
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9615871745802164e-05,
+      "loss": 0.7505,
+      "step": 15658
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9603757527288286e-05,
+      "loss": 0.7908,
+      "step": 15659
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9591644300066966e-05,
+      "loss": 1.2041,
+      "step": 15660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957953206437651e-05,
+      "loss": 0.9869,
+      "step": 15661
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95674208204552e-05,
+      "loss": 0.8383,
+      "step": 15662
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9555310568541244e-05,
+      "loss": 0.8509,
+      "step": 15663
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.95432013088728e-05,
+      "loss": 1.069,
+      "step": 15664
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9531093041688134e-05,
+      "loss": 0.844,
+      "step": 15665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.951898576722538e-05,
+      "loss": 1.037,
+      "step": 15666
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9506879485722676e-05,
+      "loss": 1.0536,
+      "step": 15667
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949477419741814e-05,
+      "loss": 0.9588,
+      "step": 15668
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9482669902549894e-05,
+      "loss": 1.1078,
+      "step": 15669
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.947056660135607e-05,
+      "loss": 0.8004,
+      "step": 15670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9458464294074724e-05,
+      "loss": 0.8905,
+      "step": 15671
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.944636298094385e-05,
+      "loss": 0.7623,
+      "step": 15672
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.943426266220156e-05,
+      "loss": 0.9424,
+      "step": 15673
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.942216333808579e-05,
+      "loss": 0.8806,
+      "step": 15674
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9410065008834615e-05,
+      "loss": 0.8595,
+      "step": 15675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.939796767468596e-05,
+      "loss": 0.9592,
+      "step": 15676
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938587133587777e-05,
+      "loss": 0.8239,
+      "step": 15677
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.937377599264798e-05,
+      "loss": 0.879,
+      "step": 15678
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936168164523459e-05,
+      "loss": 1.1194,
+      "step": 15679
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.934958829387536e-05,
+      "loss": 0.9496,
+      "step": 15680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.933749593880824e-05,
+      "loss": 0.8329,
+      "step": 15681
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9325404580271106e-05,
+      "loss": 0.9504,
+      "step": 15682
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.931331421850177e-05,
+      "loss": 0.8601,
+      "step": 15683
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.930122485373801e-05,
+      "loss": 0.877,
+      "step": 15684
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.92891364862177e-05,
+      "loss": 0.9825,
+      "step": 15685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.927704911617854e-05,
+      "loss": 0.8727,
+      "step": 15686
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9264962743858364e-05,
+      "loss": 0.847,
+      "step": 15687
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9252877369494875e-05,
+      "loss": 0.9909,
+      "step": 15688
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9240792993325745e-05,
+      "loss": 0.9261,
+      "step": 15689
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.922870961558877e-05,
+      "loss": 0.8266,
+      "step": 15690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9216627236521516e-05,
+      "loss": 0.8961,
+      "step": 15691
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9204545856361764e-05,
+      "loss": 0.8781,
+      "step": 15692
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.919246547534708e-05,
+      "loss": 0.7955,
+      "step": 15693
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.918038609371507e-05,
+      "loss": 0.8405,
+      "step": 15694
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.916830771170337e-05,
+      "loss": 1.0078,
+      "step": 15695
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9156230329549624e-05,
+      "loss": 0.8639,
+      "step": 15696
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.914415394749126e-05,
+      "loss": 0.7904,
+      "step": 15697
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.913207856576587e-05,
+      "loss": 0.9017,
+      "step": 15698
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.912000418461104e-05,
+      "loss": 0.9129,
+      "step": 15699
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910793080426422e-05,
+      "loss": 1.0341,
+      "step": 15700
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.909585842496287e-05,
+      "loss": 0.769,
+      "step": 15701
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908378704694452e-05,
+      "loss": 1.0465,
+      "step": 15702
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.907171667044653e-05,
+      "loss": 1.0003,
+      "step": 15703
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9059647295706414e-05,
+      "loss": 1.0819,
+      "step": 15704
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9047578922961524e-05,
+      "loss": 0.9781,
+      "step": 15705
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.903551155244922e-05,
+      "loss": 0.9683,
+      "step": 15706
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.902344518440689e-05,
+      "loss": 0.8483,
+      "step": 15707
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9011379819071935e-05,
+      "loss": 0.8817,
+      "step": 15708
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899931545668164e-05,
+      "loss": 0.8813,
+      "step": 15709
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.898725209747326e-05,
+      "loss": 1.0316,
+      "step": 15710
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8975189741684154e-05,
+      "loss": 1.0336,
+      "step": 15711
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.896312838955156e-05,
+      "loss": 0.8943,
+      "step": 15712
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.895106804131273e-05,
+      "loss": 0.8762,
+      "step": 15713
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.893900869720485e-05,
+      "loss": 0.9575,
+      "step": 15714
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.892695035746516e-05,
+      "loss": 0.9696,
+      "step": 15715
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8914893022330874e-05,
+      "loss": 0.8463,
+      "step": 15716
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890283669203915e-05,
+      "loss": 0.9508,
+      "step": 15717
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.889078136682708e-05,
+      "loss": 1.0109,
+      "step": 15718
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8878727046931863e-05,
+      "loss": 0.8869,
+      "step": 15719
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.886667373259054e-05,
+      "loss": 0.9529,
+      "step": 15720
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.885462142404027e-05,
+      "loss": 0.9146,
+      "step": 15721
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.884257012151808e-05,
+      "loss": 1.0174,
+      "step": 15722
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8830519825261e-05,
+      "loss": 0.8733,
+      "step": 15723
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881847053550608e-05,
+      "loss": 0.9221,
+      "step": 15724
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.880642225249041e-05,
+      "loss": 0.9946,
+      "step": 15725
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.879437497645083e-05,
+      "loss": 1.0137,
+      "step": 15726
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.878232870762438e-05,
+      "loss": 0.906,
+      "step": 15727
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.877028344624804e-05,
+      "loss": 0.8086,
+      "step": 15728
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.875823919255873e-05,
+      "loss": 0.9969,
+      "step": 15729
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874619594679329e-05,
+      "loss": 1.0174,
+      "step": 15730
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87341537091887e-05,
+      "loss": 0.8688,
+      "step": 15731
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.872211247998176e-05,
+      "loss": 1.0213,
+      "step": 15732
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.87100722594094e-05,
+      "loss": 1.1059,
+      "step": 15733
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.869803304770839e-05,
+      "loss": 0.7689,
+      "step": 15734
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.868599484511552e-05,
+      "loss": 0.8984,
+      "step": 15735
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.867395765186766e-05,
+      "loss": 1.1334,
+      "step": 15736
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.86619214682015e-05,
+      "loss": 0.8159,
+      "step": 15737
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.864988629435386e-05,
+      "loss": 1.1112,
+      "step": 15738
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.863785213056145e-05,
+      "loss": 1.0152,
+      "step": 15739
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.862581897706092e-05,
+      "loss": 0.7282,
+      "step": 15740
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8613786834089024e-05,
+      "loss": 1.078,
+      "step": 15741
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.86017557018825e-05,
+      "loss": 0.8965,
+      "step": 15742
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8589725580677835e-05,
+      "loss": 0.8528,
+      "step": 15743
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8577696470711764e-05,
+      "loss": 1.023,
+      "step": 15744
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.856566837222092e-05,
+      "loss": 1.0195,
+      "step": 15745
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8553641285441864e-05,
+      "loss": 0.9702,
+      "step": 15746
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.854161521061112e-05,
+      "loss": 0.7377,
+      "step": 15747
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.852959014796532e-05,
+      "loss": 0.8148,
+      "step": 15748
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.851756609774094e-05,
+      "loss": 0.8528,
+      "step": 15749
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.850554306017454e-05,
+      "loss": 1.0932,
+      "step": 15750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849352103550259e-05,
+      "loss": 0.9252,
+      "step": 15751
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8481500023961525e-05,
+      "loss": 0.9674,
+      "step": 15752
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.846948002578783e-05,
+      "loss": 0.9954,
+      "step": 15753
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.845746104121798e-05,
+      "loss": 0.977,
+      "step": 15754
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.844544307048835e-05,
+      "loss": 1.1087,
+      "step": 15755
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.843342611383533e-05,
+      "loss": 0.9017,
+      "step": 15756
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.842141017149526e-05,
+      "loss": 1.0456,
+      "step": 15757
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.840939524370457e-05,
+      "loss": 0.8571,
+      "step": 15758
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8397381330699545e-05,
+      "loss": 0.8182,
+      "step": 15759
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838536843271647e-05,
+      "loss": 0.8672,
+      "step": 15760
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.837335654999168e-05,
+      "loss": 1.2573,
+      "step": 15761
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.836134568276148e-05,
+      "loss": 0.8366,
+      "step": 15762
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.834933583126208e-05,
+      "loss": 0.9992,
+      "step": 15763
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.833732699572968e-05,
+      "loss": 0.9254,
+      "step": 15764
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8325319176400576e-05,
+      "loss": 1.0238,
+      "step": 15765
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.831331237351088e-05,
+      "loss": 0.7815,
+      "step": 15766
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8301306587296834e-05,
+      "loss": 1.0024,
+      "step": 15767
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.828930181799457e-05,
+      "loss": 1.0317,
+      "step": 15768
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.827729806584017e-05,
+      "loss": 0.9641,
+      "step": 15769
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826529533106979e-05,
+      "loss": 0.9709,
+      "step": 15770
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.82532936139196e-05,
+      "loss": 0.9925,
+      "step": 15771
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8241292914625526e-05,
+      "loss": 0.7759,
+      "step": 15772
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.822929323342369e-05,
+      "loss": 0.9361,
+      "step": 15773
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.821729457055015e-05,
+      "loss": 0.8724,
+      "step": 15774
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8205296926240916e-05,
+      "loss": 0.8108,
+      "step": 15775
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.819330030073196e-05,
+      "loss": 1.0079,
+      "step": 15776
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.818130469425921e-05,
+      "loss": 0.9485,
+      "step": 15777
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8169310107058677e-05,
+      "loss": 0.9896,
+      "step": 15778
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.815731653936631e-05,
+      "loss": 0.7987,
+      "step": 15779
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8145323991418e-05,
+      "loss": 0.9544,
+      "step": 15780
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.81333324634496e-05,
+      "loss": 0.9291,
+      "step": 15781
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8121341955697064e-05,
+      "loss": 0.8676,
+      "step": 15782
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8109352468396154e-05,
+      "loss": 0.9712,
+      "step": 15783
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.809736400178279e-05,
+      "loss": 0.979,
+      "step": 15784
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.808537655609274e-05,
+      "loss": 0.847,
+      "step": 15785
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.807339013156176e-05,
+      "loss": 1.0798,
+      "step": 15786
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.806140472842567e-05,
+      "loss": 1.0075,
+      "step": 15787
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.804942034692028e-05,
+      "loss": 0.9878,
+      "step": 15788
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8037436987281194e-05,
+      "loss": 1.0238,
+      "step": 15789
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8025454649744185e-05,
+      "loss": 0.9011,
+      "step": 15790
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801347333454499e-05,
+      "loss": 1.3881,
+      "step": 15791
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.800149304191923e-05,
+      "loss": 0.9086,
+      "step": 15792
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7989513772102537e-05,
+      "loss": 1.0728,
+      "step": 15793
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797753552533061e-05,
+      "loss": 1.0301,
+      "step": 15794
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7965558301838977e-05,
+      "loss": 1.1361,
+      "step": 15795
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795358210186333e-05,
+      "loss": 0.8411,
+      "step": 15796
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.794160692563917e-05,
+      "loss": 0.9318,
+      "step": 15797
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.792963277340203e-05,
+      "loss": 0.9823,
+      "step": 15798
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7917659645387516e-05,
+      "loss": 1.1233,
+      "step": 15799
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7905687541831054e-05,
+      "loss": 0.9857,
+      "step": 15800
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.789371646296823e-05,
+      "loss": 0.9576,
+      "step": 15801
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7881746409034465e-05,
+      "loss": 0.8567,
+      "step": 15802
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7869777380265166e-05,
+      "loss": 1.0616,
+      "step": 15803
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.785780937689585e-05,
+      "loss": 0.8064,
+      "step": 15804
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.784584239916187e-05,
+      "loss": 0.8733,
+      "step": 15805
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7833876447298595e-05,
+      "loss": 0.7472,
+      "step": 15806
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.782191152154144e-05,
+      "loss": 0.8618,
+      "step": 15807
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7809947622125784e-05,
+      "loss": 1.0524,
+      "step": 15808
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7797984749286905e-05,
+      "loss": 1.0415,
+      "step": 15809
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.77860229032601e-05,
+      "loss": 0.9812,
+      "step": 15810
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.777406208428071e-05,
+      "loss": 0.9547,
+      "step": 15811
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7762102292583944e-05,
+      "loss": 0.9527,
+      "step": 15812
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7750143528405126e-05,
+      "loss": 0.9108,
+      "step": 15813
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7738185791979426e-05,
+      "loss": 0.9242,
+      "step": 15814
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.772622908354205e-05,
+      "loss": 0.8972,
+      "step": 15815
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.771427340332819e-05,
+      "loss": 0.8971,
+      "step": 15816
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7702318751573115e-05,
+      "loss": 0.9677,
+      "step": 15817
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.76903651285118e-05,
+      "loss": 0.9172,
+      "step": 15818
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7678412534379466e-05,
+      "loss": 1.0555,
+      "step": 15819
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.766646096941125e-05,
+      "loss": 0.9168,
+      "step": 15820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.76545104338422e-05,
+      "loss": 0.9797,
+      "step": 15821
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.764256092790739e-05,
+      "loss": 0.9824,
+      "step": 15822
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.763061245184181e-05,
+      "loss": 0.659,
+      "step": 15823
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7618665005880546e-05,
+      "loss": 0.8041,
+      "step": 15824
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7606718590258625e-05,
+      "loss": 0.9252,
+      "step": 15825
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7594773205211004e-05,
+      "loss": 0.8917,
+      "step": 15826
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758282885097262e-05,
+      "loss": 1.1437,
+      "step": 15827
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.757088552777847e-05,
+      "loss": 0.8362,
+      "step": 15828
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.755894323586342e-05,
+      "loss": 0.8701,
+      "step": 15829
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.754700197546245e-05,
+      "loss": 0.7779,
+      "step": 15830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7535061746810394e-05,
+      "loss": 1.0193,
+      "step": 15831
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75231225501421e-05,
+      "loss": 1.0263,
+      "step": 15832
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751118438569242e-05,
+      "loss": 1.0649,
+      "step": 15833
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749924725369628e-05,
+      "loss": 0.977,
+      "step": 15834
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7487311154388314e-05,
+      "loss": 0.7443,
+      "step": 15835
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7475376088003386e-05,
+      "loss": 0.962,
+      "step": 15836
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.746344205477631e-05,
+      "loss": 1.0649,
+      "step": 15837
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7451509054941765e-05,
+      "loss": 0.8334,
+      "step": 15838
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7439577088734456e-05,
+      "loss": 0.8718,
+      "step": 15839
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7427646156389135e-05,
+      "loss": 0.8366,
+      "step": 15840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7415716258140444e-05,
+      "loss": 0.937,
+      "step": 15841
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7403787394223085e-05,
+      "loss": 0.9151,
+      "step": 15842
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.739185956487169e-05,
+      "loss": 0.8548,
+      "step": 15843
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737993277032081e-05,
+      "loss": 0.942,
+      "step": 15844
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736800701080515e-05,
+      "loss": 1.0051,
+      "step": 15845
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.735608228655919e-05,
+      "loss": 0.8467,
+      "step": 15846
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.734415859781758e-05,
+      "loss": 1.0155,
+      "step": 15847
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.733223594481482e-05,
+      "loss": 0.8892,
+      "step": 15848
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.732031432778537e-05,
+      "loss": 0.8097,
+      "step": 15849
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.730839374696384e-05,
+      "loss": 0.9448,
+      "step": 15850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729647420258464e-05,
+      "loss": 1.0215,
+      "step": 15851
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72845556948822e-05,
+      "loss": 0.9153,
+      "step": 15852
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.727263822409099e-05,
+      "loss": 1.0091,
+      "step": 15853
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.726072179044547e-05,
+      "loss": 0.7746,
+      "step": 15854
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.724880639417999e-05,
+      "loss": 0.9576,
+      "step": 15855
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72368920355289e-05,
+      "loss": 0.9661,
+      "step": 15856
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722497871472662e-05,
+      "loss": 1.073,
+      "step": 15857
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7213066432007414e-05,
+      "loss": 0.9249,
+      "step": 15858
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.720115518760567e-05,
+      "loss": 0.9368,
+      "step": 15859
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7189244981755644e-05,
+      "loss": 1.0211,
+      "step": 15860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717733581469157e-05,
+      "loss": 0.9803,
+      "step": 15861
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716542768664775e-05,
+      "loss": 1.0256,
+      "step": 15862
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.715352059785847e-05,
+      "loss": 0.9802,
+      "step": 15863
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.714161454855781e-05,
+      "loss": 0.8125,
+      "step": 15864
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.712970953898006e-05,
+      "loss": 0.8765,
+      "step": 15865
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7117805569359343e-05,
+      "loss": 1.1169,
+      "step": 15866
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.710590263992985e-05,
+      "loss": 0.967,
+      "step": 15867
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.70940007509257e-05,
+      "loss": 0.8817,
+      "step": 15868
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.708209990258095e-05,
+      "loss": 0.9868,
+      "step": 15869
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.707020009512974e-05,
+      "loss": 0.8713,
+      "step": 15870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.705830132880616e-05,
+      "loss": 1.0126,
+      "step": 15871
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704640360384424e-05,
+      "loss": 0.9566,
+      "step": 15872
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.703450692047795e-05,
+      "loss": 1.0563,
+      "step": 15873
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.702261127894139e-05,
+      "loss": 0.7707,
+      "step": 15874
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7010716679468455e-05,
+      "loss": 0.917,
+      "step": 15875
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.69988231222932e-05,
+      "loss": 0.9097,
+      "step": 15876
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6986930607649535e-05,
+      "loss": 0.928,
+      "step": 15877
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6975039135771335e-05,
+      "loss": 1.0591,
+      "step": 15878
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6963148706892555e-05,
+      "loss": 0.9546,
+      "step": 15879
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6951259321247145e-05,
+      "loss": 1.0186,
+      "step": 15880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.693937097906882e-05,
+      "loss": 0.8501,
+      "step": 15881
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692748368059151e-05,
+      "loss": 1.0657,
+      "step": 15882
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.691559742604906e-05,
+      "loss": 1.0409,
+      "step": 15883
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6903712215675245e-05,
+      "loss": 0.7979,
+      "step": 15884
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.68918280497038e-05,
+      "loss": 0.9281,
+      "step": 15885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6879944928368566e-05,
+      "loss": 1.0616,
+      "step": 15886
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.686806285190322e-05,
+      "loss": 1.0103,
+      "step": 15887
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6856181820541546e-05,
+      "loss": 0.8088,
+      "step": 15888
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.684430183451719e-05,
+      "loss": 0.844,
+      "step": 15889
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6832422894063824e-05,
+      "loss": 0.7913,
+      "step": 15890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682054499941516e-05,
+      "loss": 0.8887,
+      "step": 15891
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.680866815080478e-05,
+      "loss": 0.8832,
+      "step": 15892
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6796792348466356e-05,
+      "loss": 1.0344,
+      "step": 15893
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6784917592633445e-05,
+      "loss": 0.8733,
+      "step": 15894
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.677304388353961e-05,
+      "loss": 0.6458,
+      "step": 15895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.676117122141847e-05,
+      "loss": 0.9894,
+      "step": 15896
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6749299606503516e-05,
+      "loss": 1.2555,
+      "step": 15897
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6737429039028226e-05,
+      "loss": 1.0614,
+      "step": 15898
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6725559519226135e-05,
+      "loss": 0.7187,
+      "step": 15899
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.671369104733075e-05,
+      "loss": 1.0809,
+      "step": 15900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.67018236235755e-05,
+      "loss": 0.9167,
+      "step": 15901
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668995724819376e-05,
+      "loss": 0.7381,
+      "step": 15902
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.667809192141902e-05,
+      "loss": 1.0965,
+      "step": 15903
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6666227643484605e-05,
+      "loss": 1.0918,
+      "step": 15904
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.665436441462396e-05,
+      "loss": 0.88,
+      "step": 15905
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6642502235070384e-05,
+      "loss": 0.9819,
+      "step": 15906
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6630641105057184e-05,
+      "loss": 1.0938,
+      "step": 15907
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.661878102481772e-05,
+      "loss": 0.9818,
+      "step": 15908
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6606921994585274e-05,
+      "loss": 0.9504,
+      "step": 15909
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.659506401459305e-05,
+      "loss": 0.9485,
+      "step": 15910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.658320708507439e-05,
+      "loss": 0.8638,
+      "step": 15911
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6571351206262415e-05,
+      "loss": 0.8029,
+      "step": 15912
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6559496378390423e-05,
+      "loss": 0.8504,
+      "step": 15913
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6547642601691565e-05,
+      "loss": 1.0196,
+      "step": 15914
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.653578987639896e-05,
+      "loss": 0.8926,
+      "step": 15915
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65239382027458e-05,
+      "loss": 0.8714,
+      "step": 15916
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.651208758096521e-05,
+      "loss": 0.9529,
+      "step": 15917
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6500238011290295e-05,
+      "loss": 0.9532,
+      "step": 15918
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.648838949395409e-05,
+      "loss": 0.9251,
+      "step": 15919
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.647654202918971e-05,
+      "loss": 0.8518,
+      "step": 15920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6464695617230126e-05,
+      "loss": 0.8689,
+      "step": 15921
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.645285025830843e-05,
+      "loss": 0.9171,
+      "step": 15922
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.64410059526576e-05,
+      "loss": 0.7703,
+      "step": 15923
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.642916270051057e-05,
+      "loss": 1.04,
+      "step": 15924
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6417320502100316e-05,
+      "loss": 1.0153,
+      "step": 15925
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.640547935765986e-05,
+      "loss": 1.0294,
+      "step": 15926
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.639363926742197e-05,
+      "loss": 0.9624,
+      "step": 15927
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638180023161962e-05,
+      "loss": 0.8597,
+      "step": 15928
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.636996225048571e-05,
+      "loss": 0.831,
+      "step": 15929
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.635812532425306e-05,
+      "loss": 0.882,
+      "step": 15930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63462894531545e-05,
+      "loss": 0.8253,
+      "step": 15931
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.633445463742281e-05,
+      "loss": 0.8814,
+      "step": 15932
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63226208772908e-05,
+      "loss": 0.8668,
+      "step": 15933
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.631078817299129e-05,
+      "loss": 0.7832,
+      "step": 15934
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.629895652475701e-05,
+      "loss": 0.8458,
+      "step": 15935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628712593282062e-05,
+      "loss": 0.6781,
+      "step": 15936
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.627529639741491e-05,
+      "loss": 1.037,
+      "step": 15937
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6263467918772496e-05,
+      "loss": 0.9571,
+      "step": 15938
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.625164049712612e-05,
+      "loss": 0.8338,
+      "step": 15939
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.62398141327084e-05,
+      "loss": 0.9206,
+      "step": 15940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.62279888257519e-05,
+      "loss": 0.9237,
+      "step": 15941
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621616457648932e-05,
+      "loss": 1.054,
+      "step": 15942
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6204341385153186e-05,
+      "loss": 0.9672,
+      "step": 15943
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.619251925197603e-05,
+      "loss": 0.8759,
+      "step": 15944
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.618069817719043e-05,
+      "loss": 0.9739,
+      "step": 15945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.616887816102895e-05,
+      "loss": 0.9849,
+      "step": 15946
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.615705920372405e-05,
+      "loss": 0.9924,
+      "step": 15947
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.614524130550817e-05,
+      "loss": 1.0015,
+      "step": 15948
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.613342446661383e-05,
+      "loss": 1.0009,
+      "step": 15949
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.612160868727341e-05,
+      "loss": 1.1231,
+      "step": 15950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.610979396771939e-05,
+      "loss": 0.8147,
+      "step": 15951
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.609798030818413e-05,
+      "loss": 0.8357,
+      "step": 15952
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6086167708899975e-05,
+      "loss": 0.8691,
+      "step": 15953
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607435617009934e-05,
+      "loss": 0.9257,
+      "step": 15954
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.606254569201454e-05,
+      "loss": 0.9505,
+      "step": 15955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6050736274877825e-05,
+      "loss": 0.9207,
+      "step": 15956
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.603892791892157e-05,
+      "loss": 0.9737,
+      "step": 15957
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.602712062437797e-05,
+      "loss": 0.876,
+      "step": 15958
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.601531439147937e-05,
+      "loss": 1.0182,
+      "step": 15959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.600350922045792e-05,
+      "loss": 0.9169,
+      "step": 15960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5991705111545826e-05,
+      "loss": 0.9498,
+      "step": 15961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.597990206497529e-05,
+      "loss": 0.9423,
+      "step": 15962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.596810008097853e-05,
+      "loss": 1.0123,
+      "step": 15963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595629915978764e-05,
+      "loss": 0.9533,
+      "step": 15964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.594449930163471e-05,
+      "loss": 1.166,
+      "step": 15965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593270050675192e-05,
+      "loss": 0.8581,
+      "step": 15966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.592090277537129e-05,
+      "loss": 0.8975,
+      "step": 15967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.590910610772493e-05,
+      "loss": 1.1201,
+      "step": 15968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5897310504044864e-05,
+      "loss": 0.9606,
+      "step": 15969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5885515964563066e-05,
+      "loss": 0.9327,
+      "step": 15970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.587372248951156e-05,
+      "loss": 0.9701,
+      "step": 15971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5861930079122416e-05,
+      "loss": 0.8118,
+      "step": 15972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5850138733627434e-05,
+      "loss": 0.8654,
+      "step": 15973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.583834845325866e-05,
+      "loss": 1.0083,
+      "step": 15974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5826559238247946e-05,
+      "loss": 0.9302,
+      "step": 15975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5814771088827235e-05,
+      "loss": 0.9758,
+      "step": 15976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580298400522839e-05,
+      "loss": 0.8907,
+      "step": 15977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5791197987683196e-05,
+      "loss": 1.097,
+      "step": 15978
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.577941303642355e-05,
+      "loss": 1.0862,
+      "step": 15979
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.576762915168128e-05,
+      "loss": 0.7522,
+      "step": 15980
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575584633368815e-05,
+      "loss": 1.0507,
+      "step": 15981
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.574406458267587e-05,
+      "loss": 1.0,
+      "step": 15982
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.573228389887628e-05,
+      "loss": 0.9471,
+      "step": 15983
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.572050428252104e-05,
+      "loss": 0.9184,
+      "step": 15984
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.570872573384191e-05,
+      "loss": 0.8864,
+      "step": 15985
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5696948253070525e-05,
+      "loss": 1.0043,
+      "step": 15986
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.568517184043855e-05,
+      "loss": 0.9483,
+      "step": 15987
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567339649617762e-05,
+      "loss": 0.8337,
+      "step": 15988
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.566162222051946e-05,
+      "loss": 0.8871,
+      "step": 15989
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5649849013695525e-05,
+      "loss": 0.8363,
+      "step": 15990
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.563807687593744e-05,
+      "loss": 0.7426,
+      "step": 15991
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562630580747682e-05,
+      "loss": 0.8122,
+      "step": 15992
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.561453580854516e-05,
+      "loss": 0.9487,
+      "step": 15993
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.560276687937393e-05,
+      "loss": 0.8782,
+      "step": 15994
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.559099902019471e-05,
+      "loss": 0.8962,
+      "step": 15995
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.55792322312389e-05,
+      "loss": 0.8394,
+      "step": 15996
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.556746651273802e-05,
+      "loss": 0.9031,
+      "step": 15997
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555570186492347e-05,
+      "loss": 0.9381,
+      "step": 15998
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.554393828802661e-05,
+      "loss": 0.9286,
+      "step": 15999
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.553217578227892e-05,
+      "loss": 0.6948,
+      "step": 16000
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.552041434791172e-05,
+      "loss": 0.8027,
+      "step": 16001
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.550865398515634e-05,
+      "loss": 0.8873,
+      "step": 16002
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549689469424414e-05,
+      "loss": 1.0276,
+      "step": 16003
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.54851364754064e-05,
+      "loss": 0.8654,
+      "step": 16004
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.547337932887443e-05,
+      "loss": 1.0829,
+      "step": 16005
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.546162325487949e-05,
+      "loss": 1.1578,
+      "step": 16006
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544986825365277e-05,
+      "loss": 0.7187,
+      "step": 16007
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.543811432542554e-05,
+      "loss": 1.0593,
+      "step": 16008
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542636147042901e-05,
+      "loss": 1.05,
+      "step": 16009
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.541460968889435e-05,
+      "loss": 0.8982,
+      "step": 16010
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540285898105268e-05,
+      "loss": 0.9674,
+      "step": 16011
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5391109347135174e-05,
+      "loss": 0.857,
+      "step": 16012
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.537936078737292e-05,
+      "loss": 0.9223,
+      "step": 16013
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5367613301997046e-05,
+      "loss": 0.8377,
+      "step": 16014
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5355866891238605e-05,
+      "loss": 0.8653,
+      "step": 16015
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.534412155532862e-05,
+      "loss": 0.9385,
+      "step": 16016
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5332377294498194e-05,
+      "loss": 0.9055,
+      "step": 16017
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532063410897829e-05,
+      "loss": 1.1174,
+      "step": 16018
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5308891998999844e-05,
+      "loss": 0.8315,
+      "step": 16019
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529715096479393e-05,
+      "loss": 0.8974,
+      "step": 16020
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.52854110065914e-05,
+      "loss": 0.9235,
+      "step": 16021
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.527367212462327e-05,
+      "loss": 0.9074,
+      "step": 16022
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.526193431912038e-05,
+      "loss": 0.9338,
+      "step": 16023
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525019759031359e-05,
+      "loss": 0.9538,
+      "step": 16024
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5238461938433796e-05,
+      "loss": 0.9772,
+      "step": 16025
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5226727363711874e-05,
+      "loss": 0.9181,
+      "step": 16026
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5214993866378605e-05,
+      "loss": 0.8733,
+      "step": 16027
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.520326144666477e-05,
+      "loss": 1.1535,
+      "step": 16028
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.519153010480118e-05,
+      "loss": 0.8783,
+      "step": 16029
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5179799841018554e-05,
+      "loss": 0.9251,
+      "step": 16030
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.516807065554768e-05,
+      "loss": 0.8858,
+      "step": 16031
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515634254861922e-05,
+      "loss": 0.9387,
+      "step": 16032
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.514461552046386e-05,
+      "loss": 0.9551,
+      "step": 16033
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51328895713123e-05,
+      "loss": 1.0657,
+      "step": 16034
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5121164701395247e-05,
+      "loss": 0.7874,
+      "step": 16035
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.510944091094319e-05,
+      "loss": 0.9792,
+      "step": 16036
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5097718200186814e-05,
+      "loss": 0.7564,
+      "step": 16037
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5085996569356734e-05,
+      "loss": 0.821,
+      "step": 16038
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5074276018683484e-05,
+      "loss": 0.9335,
+      "step": 16039
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5062556548397595e-05,
+      "loss": 1.0387,
+      "step": 16040
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.505083815872956e-05,
+      "loss": 0.9406,
+      "step": 16041
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503912084990992e-05,
+      "loss": 1.1132,
+      "step": 16042
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.502740462216919e-05,
+      "loss": 0.8907,
+      "step": 16043
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.501568947573779e-05,
+      "loss": 0.881,
+      "step": 16044
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.50039754108461e-05,
+      "loss": 0.9132,
+      "step": 16045
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4992262427724644e-05,
+      "loss": 0.8483,
+      "step": 16046
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498055052660376e-05,
+      "loss": 0.9041,
+      "step": 16047
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.4968839707713783e-05,
+      "loss": 0.8614,
+      "step": 16048
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.495712997128514e-05,
+      "loss": 0.8555,
+      "step": 16049
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.494542131754809e-05,
+      "loss": 0.8946,
+      "step": 16050
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4933713746733007e-05,
+      "loss": 0.897,
+      "step": 16051
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.492200725907016e-05,
+      "loss": 1.0455,
+      "step": 16052
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.491030185478976e-05,
+      "loss": 0.8566,
+      "step": 16053
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4898597534122086e-05,
+      "loss": 0.9624,
+      "step": 16054
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4886894297297434e-05,
+      "loss": 1.0064,
+      "step": 16055
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.487519214454592e-05,
+      "loss": 0.9486,
+      "step": 16056
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.486349107609773e-05,
+      "loss": 0.9368,
+      "step": 16057
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485179109218307e-05,
+      "loss": 1.039,
+      "step": 16058
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4840092193032015e-05,
+      "loss": 0.9527,
+      "step": 16059
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482839437887475e-05,
+      "loss": 0.9682,
+      "step": 16060
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.481669764994134e-05,
+      "loss": 0.9543,
+      "step": 16061
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4805002006461826e-05,
+      "loss": 0.9504,
+      "step": 16062
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4793307448666335e-05,
+      "loss": 0.7972,
+      "step": 16063
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4781613976784856e-05,
+      "loss": 0.9079,
+      "step": 16064
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.476992159104735e-05,
+      "loss": 0.8866,
+      "step": 16065
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.47582302916839e-05,
+      "loss": 0.918,
+      "step": 16066
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4746540078924394e-05,
+      "loss": 0.9441,
+      "step": 16067
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.473485095299884e-05,
+      "loss": 1.0143,
+      "step": 16068
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.472316291413715e-05,
+      "loss": 0.9917,
+      "step": 16069
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.471147596256917e-05,
+      "loss": 0.8268,
+      "step": 16070
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.469979009852482e-05,
+      "loss": 0.9063,
+      "step": 16071
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.468810532223401e-05,
+      "loss": 0.9403,
+      "step": 16072
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4676421633926534e-05,
+      "loss": 0.89,
+      "step": 16073
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.466473903383217e-05,
+      "loss": 0.9032,
+      "step": 16074
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.46530575221808e-05,
+      "loss": 0.8055,
+      "step": 16075
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4641377099202106e-05,
+      "loss": 1.0732,
+      "step": 16076
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4629697765125934e-05,
+      "loss": 0.8522,
+      "step": 16077
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.461801952018198e-05,
+      "loss": 0.994,
+      "step": 16078
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.460634236459989e-05,
+      "loss": 1.079,
+      "step": 16079
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.459466629860943e-05,
+      "loss": 0.9306,
+      "step": 16080
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4582991322440315e-05,
+      "loss": 0.9238,
+      "step": 16081
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4571317436322044e-05,
+      "loss": 1.0771,
+      "step": 16082
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.455964464048436e-05,
+      "loss": 0.9108,
+      "step": 16083
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.45479729351568e-05,
+      "loss": 0.9159,
+      "step": 16084
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.453630232056901e-05,
+      "loss": 0.9508,
+      "step": 16085
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4524632796950516e-05,
+      "loss": 0.76,
+      "step": 16086
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.451296436453082e-05,
+      "loss": 1.1055,
+      "step": 16087
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.450129702353948e-05,
+      "loss": 0.9509,
+      "step": 16088
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4489630774206014e-05,
+      "loss": 0.779,
+      "step": 16089
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.447796561675988e-05,
+      "loss": 0.7969,
+      "step": 16090
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.446630155143049e-05,
+      "loss": 0.9601,
+      "step": 16091
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.445463857844734e-05,
+      "loss": 0.9784,
+      "step": 16092
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.444297669803981e-05,
+      "loss": 0.9636,
+      "step": 16093
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.443131591043724e-05,
+      "loss": 0.9134,
+      "step": 16094
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.44196562158691e-05,
+      "loss": 0.9385,
+      "step": 16095
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.440799761456463e-05,
+      "loss": 1.0026,
+      "step": 16096
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.439634010675325e-05,
+      "loss": 0.9375,
+      "step": 16097
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.438468369266422e-05,
+      "loss": 1.0405,
+      "step": 16098
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4373028372526784e-05,
+      "loss": 0.8204,
+      "step": 16099
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4361374146570236e-05,
+      "loss": 0.9914,
+      "step": 16100
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.434972101502385e-05,
+      "loss": 1.0009,
+      "step": 16101
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4338068978116806e-05,
+      "loss": 0.8869,
+      "step": 16102
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.432641803607826e-05,
+      "loss": 0.9491,
+      "step": 16103
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.431476818913748e-05,
+      "loss": 0.8857,
+      "step": 16104
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4303119437523523e-05,
+      "loss": 0.9844,
+      "step": 16105
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.429147178146563e-05,
+      "loss": 0.9369,
+      "step": 16106
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.427982522119277e-05,
+      "loss": 1.0214,
+      "step": 16107
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.42681797569341e-05,
+      "loss": 0.8728,
+      "step": 16108
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.425653538891874e-05,
+      "loss": 0.9657,
+      "step": 16109
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.424489211737567e-05,
+      "loss": 0.8079,
+      "step": 16110
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4233249942533895e-05,
+      "loss": 0.8619,
+      "step": 16111
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.422160886462248e-05,
+      "loss": 0.7705,
+      "step": 16112
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.420996888387033e-05,
+      "loss": 0.6664,
+      "step": 16113
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4198330000506494e-05,
+      "loss": 1.0849,
+      "step": 16114
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.418669221475985e-05,
+      "loss": 0.9946,
+      "step": 16115
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.417505552685929e-05,
+      "loss": 0.9858,
+      "step": 16116
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.416341993703373e-05,
+      "loss": 0.9625,
+      "step": 16117
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.415178544551211e-05,
+      "loss": 1.0055,
+      "step": 16118
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4140152052523195e-05,
+      "loss": 1.0551,
+      "step": 16119
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.412851975829582e-05,
+      "loss": 0.8646,
+      "step": 16120
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.411688856305885e-05,
+      "loss": 1.1331,
+      "step": 16121
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4105258467040975e-05,
+      "loss": 0.7431,
+      "step": 16122
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4093629470471074e-05,
+      "loss": 0.9958,
+      "step": 16123
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4082001573577825e-05,
+      "loss": 1.0344,
+      "step": 16124
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4070374776589906e-05,
+      "loss": 0.8914,
+      "step": 16125
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.405874907973612e-05,
+      "loss": 0.8678,
+      "step": 16126
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.404712448324507e-05,
+      "loss": 0.9304,
+      "step": 16127
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.40355009873454e-05,
+      "loss": 0.8382,
+      "step": 16128
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.402387859226581e-05,
+      "loss": 1.0015,
+      "step": 16129
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.401225729823484e-05,
+      "loss": 1.0125,
+      "step": 16130
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.400063710548114e-05,
+      "loss": 0.8744,
+      "step": 16131
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3989018014233275e-05,
+      "loss": 1.121,
+      "step": 16132
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3977400024719725e-05,
+      "loss": 1.0366,
+      "step": 16133
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.396578313716907e-05,
+      "loss": 0.8038,
+      "step": 16134
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3954167351809825e-05,
+      "loss": 0.8858,
+      "step": 16135
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3942552668870474e-05,
+      "loss": 1.0635,
+      "step": 16136
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.39309390885794e-05,
+      "loss": 0.8285,
+      "step": 16137
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.391932661116516e-05,
+      "loss": 1.0652,
+      "step": 16138
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.390771523685605e-05,
+      "loss": 1.0343,
+      "step": 16139
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3896104965880566e-05,
+      "loss": 0.8907,
+      "step": 16140
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.388449579846705e-05,
+      "loss": 0.8896,
+      "step": 16141
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.38728877348438e-05,
+      "loss": 0.9744,
+      "step": 16142
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.386128077523923e-05,
+      "loss": 0.7228,
+      "step": 16143
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.384967491988162e-05,
+      "loss": 1.0656,
+      "step": 16144
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.38380701689992e-05,
+      "loss": 0.983,
+      "step": 16145
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.382646652282028e-05,
+      "loss": 1.0064,
+      "step": 16146
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.381486398157314e-05,
+      "loss": 0.8874,
+      "step": 16147
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3803262545485976e-05,
+      "loss": 0.7449,
+      "step": 16148
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.379166221478697e-05,
+      "loss": 0.9722,
+      "step": 16149
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.378006298970427e-05,
+      "loss": 0.9632,
+      "step": 16150
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.376846487046609e-05,
+      "loss": 0.9021,
+      "step": 16151
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.37568678573006e-05,
+      "loss": 0.872,
+      "step": 16152
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3745271950435785e-05,
+      "loss": 0.8,
+      "step": 16153
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.373367715009982e-05,
+      "loss": 1.0162,
+      "step": 16154
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3722083456520793e-05,
+      "loss": 0.9358,
+      "step": 16155
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3710490869926723e-05,
+      "loss": 0.9159,
+      "step": 16156
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36988993905456e-05,
+      "loss": 1.0664,
+      "step": 16157
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.368730901860551e-05,
+      "loss": 0.9647,
+      "step": 16158
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.367571975433433e-05,
+      "loss": 0.8915,
+      "step": 16159
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.366413159796014e-05,
+      "loss": 0.858,
+      "step": 16160
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3652544549710806e-05,
+      "loss": 0.9645,
+      "step": 16161
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.364095860981422e-05,
+      "loss": 1.0631,
+      "step": 16162
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.362937377849832e-05,
+      "loss": 0.9465,
+      "step": 16163
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3617790055991006e-05,
+      "loss": 0.8803,
+      "step": 16164
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.36062074425201e-05,
+      "loss": 0.8764,
+      "step": 16165
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3594625938313384e-05,
+      "loss": 0.8418,
+      "step": 16166
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.358304554359876e-05,
+      "loss": 0.8866,
+      "step": 16167
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.357146625860391e-05,
+      "loss": 0.9015,
+      "step": 16168
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.35598880835567e-05,
+      "loss": 0.7897,
+      "step": 16169
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3548311018684816e-05,
+      "loss": 0.8672,
+      "step": 16170
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.353673506421596e-05,
+      "loss": 0.8771,
+      "step": 16171
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.35251602203779e-05,
+      "loss": 0.9685,
+      "step": 16172
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.351358648739826e-05,
+      "loss": 1.1005,
+      "step": 16173
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3502013865504676e-05,
+      "loss": 1.109,
+      "step": 16174
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3490442354924844e-05,
+      "loss": 0.992,
+      "step": 16175
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.347887195588631e-05,
+      "loss": 1.0288,
+      "step": 16176
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.346730266861673e-05,
+      "loss": 0.9451,
+      "step": 16177
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.345573449334364e-05,
+      "loss": 0.8952,
+      "step": 16178
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3444167430294536e-05,
+      "loss": 0.9332,
+      "step": 16179
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3432601479697e-05,
+      "loss": 0.9952,
+      "step": 16180
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3421036641778556e-05,
+      "loss": 0.7604,
+      "step": 16181
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3409472916766644e-05,
+      "loss": 0.9193,
+      "step": 16182
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3397910304888704e-05,
+      "loss": 1.1346,
+      "step": 16183
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.338634880637222e-05,
+      "loss": 0.9267,
+      "step": 16184
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3374788421444554e-05,
+      "loss": 0.9278,
+      "step": 16185
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3363229150333174e-05,
+      "loss": 1.0042,
+      "step": 16186
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.335167099326539e-05,
+      "loss": 0.9622,
+      "step": 16187
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.334011395046853e-05,
+      "loss": 0.9382,
+      "step": 16188
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.332855802217e-05,
+      "loss": 0.777,
+      "step": 16189
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.331700320859706e-05,
+      "loss": 0.9391,
+      "step": 16190
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.330544950997695e-05,
+      "loss": 0.9275,
+      "step": 16191
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3293896926537016e-05,
+      "loss": 1.048,
+      "step": 16192
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.328234545850442e-05,
+      "loss": 1.0328,
+      "step": 16193
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.327079510610644e-05,
+      "loss": 0.9236,
+      "step": 16194
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3259245869570255e-05,
+      "loss": 1.0097,
+      "step": 16195
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.324769774912299e-05,
+      "loss": 0.7527,
+      "step": 16196
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.323615074499183e-05,
+      "loss": 1.0825,
+      "step": 16197
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.322460485740398e-05,
+      "loss": 0.9832,
+      "step": 16198
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.321306008658641e-05,
+      "loss": 0.6973,
+      "step": 16199
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.320151643276628e-05,
+      "loss": 0.8291,
+      "step": 16200
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.318997389617067e-05,
+      "loss": 1.1178,
+      "step": 16201
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3178432477026595e-05,
+      "loss": 0.9407,
+      "step": 16202
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3166892175561036e-05,
+      "loss": 0.894,
+      "step": 16203
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3155352992001077e-05,
+      "loss": 1.0598,
+      "step": 16204
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.31438149265736e-05,
+      "loss": 0.8632,
+      "step": 16205
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.313227797950565e-05,
+      "loss": 0.9257,
+      "step": 16206
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.312074215102411e-05,
+      "loss": 1.0558,
+      "step": 16207
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.310920744135585e-05,
+      "loss": 0.9231,
+      "step": 16208
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3097673850727805e-05,
+      "loss": 0.8403,
+      "step": 16209
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.308614137936687e-05,
+      "loss": 1.102,
+      "step": 16210
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.307461002749986e-05,
+      "loss": 0.7533,
+      "step": 16211
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.306307979535354e-05,
+      "loss": 0.932,
+      "step": 16212
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.305155068315481e-05,
+      "loss": 0.8781,
+      "step": 16213
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.3040022691130355e-05,
+      "loss": 0.7948,
+      "step": 16214
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.302849581950704e-05,
+      "loss": 0.9428,
+      "step": 16215
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.301697006851146e-05,
+      "loss": 0.8802,
+      "step": 16216
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.300544543837039e-05,
+      "loss": 0.8652,
+      "step": 16217
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2993921929310576e-05,
+      "loss": 0.8507,
+      "step": 16218
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.298239954155863e-05,
+      "loss": 1.0341,
+      "step": 16219
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2970878275341163e-05,
+      "loss": 1.0089,
+      "step": 16220
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.295935813088487e-05,
+      "loss": 1.0929,
+      "step": 16221
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.294783910841629e-05,
+      "loss": 0.8123,
+      "step": 16222
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2936321208162055e-05,
+      "loss": 0.9447,
+      "step": 16223
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2924804430348706e-05,
+      "loss": 0.9722,
+      "step": 16224
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.291328877520273e-05,
+      "loss": 0.9477,
+      "step": 16225
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.290177424295068e-05,
+      "loss": 1.0375,
+      "step": 16226
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.289026083381907e-05,
+      "loss": 0.8495,
+      "step": 16227
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.287874854803437e-05,
+      "loss": 0.9046,
+      "step": 16228
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2867237385822935e-05,
+      "loss": 0.8801,
+      "step": 16229
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.28557273474113e-05,
+      "loss": 0.9748,
+      "step": 16230
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.284421843302578e-05,
+      "loss": 1.0498,
+      "step": 16231
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.283271064289284e-05,
+      "loss": 0.9465,
+      "step": 16232
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.282120397723879e-05,
+      "loss": 1.0728,
+      "step": 16233
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.280969843628993e-05,
+      "loss": 0.8516,
+      "step": 16234
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.279819402027265e-05,
+      "loss": 0.873,
+      "step": 16235
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2786690729413214e-05,
+      "loss": 0.8816,
+      "step": 16236
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2775188563937843e-05,
+      "loss": 0.9665,
+      "step": 16237
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.276368752407285e-05,
+      "loss": 0.9504,
+      "step": 16238
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.27521876100444e-05,
+      "loss": 1.0556,
+      "step": 16239
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.274068882207878e-05,
+      "loss": 1.0351,
+      "step": 16240
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.272919116040211e-05,
+      "loss": 0.7733,
+      "step": 16241
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.271769462524051e-05,
+      "loss": 0.9288,
+      "step": 16242
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2706199216820185e-05,
+      "loss": 0.9894,
+      "step": 16243
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2694704935367256e-05,
+      "loss": 0.8595,
+      "step": 16244
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.268321178110779e-05,
+      "loss": 0.9449,
+      "step": 16245
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.267171975426782e-05,
+      "loss": 1.1114,
+      "step": 16246
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.266022885507347e-05,
+      "loss": 0.8994,
+      "step": 16247
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.264873908375072e-05,
+      "loss": 0.8831,
+      "step": 16248
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.263725044052554e-05,
+      "loss": 0.9009,
+      "step": 16249
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2625762925624e-05,
+      "loss": 0.9192,
+      "step": 16250
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.261427653927196e-05,
+      "loss": 0.848,
+      "step": 16251
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2602791281695443e-05,
+      "loss": 0.8704,
+      "step": 16252
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.259130715312033e-05,
+      "loss": 1.0165,
+      "step": 16253
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2579824153772476e-05,
+      "loss": 0.9771,
+      "step": 16254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2568342283877785e-05,
+      "loss": 0.8996,
+      "step": 16255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.255686154366214e-05,
+      "loss": 0.7704,
+      "step": 16256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.254538193335133e-05,
+      "loss": 1.1084,
+      "step": 16257
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.253390345317116e-05,
+      "loss": 1.1161,
+      "step": 16258
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2522426103347394e-05,
+      "loss": 0.9611,
+      "step": 16259
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.251094988410579e-05,
+      "loss": 0.8802,
+      "step": 16260
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.249947479567218e-05,
+      "loss": 1.1396,
+      "step": 16261
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.248800083827214e-05,
+      "loss": 0.9161,
+      "step": 16262
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.247652801213141e-05,
+      "loss": 0.9387,
+      "step": 16263
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.246505631747573e-05,
+      "loss": 1.055,
+      "step": 16264
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2453585754530676e-05,
+      "loss": 1.0352,
+      "step": 16265
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2442116323521864e-05,
+      "loss": 0.8418,
+      "step": 16266
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2430648024674955e-05,
+      "loss": 1.0254,
+      "step": 16267
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.241918085821547e-05,
+      "loss": 0.8862,
+      "step": 16268
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2407714824369024e-05,
+      "loss": 1.0145,
+      "step": 16269
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.239624992336114e-05,
+      "loss": 0.7741,
+      "step": 16270
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.238478615541728e-05,
+      "loss": 0.9543,
+      "step": 16271
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2373323520762964e-05,
+      "loss": 0.8073,
+      "step": 16272
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.236186201962372e-05,
+      "loss": 0.8315,
+      "step": 16273
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.235040165222495e-05,
+      "loss": 0.9981,
+      "step": 16274
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2338942418792026e-05,
+      "loss": 1.0128,
+      "step": 16275
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.232748431955045e-05,
+      "loss": 1.0103,
+      "step": 16276
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.231602735472552e-05,
+      "loss": 0.7707,
+      "step": 16277
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.230457152454266e-05,
+      "loss": 1.0631,
+      "step": 16278
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.229311682922718e-05,
+      "loss": 1.0977,
+      "step": 16279
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.2281663269004346e-05,
+      "loss": 1.1003,
+      "step": 16280
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.227021084409953e-05,
+      "loss": 0.8141,
+      "step": 16281
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.225875955473796e-05,
+      "loss": 1.0787,
+      "step": 16282
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.224730940114485e-05,
+      "loss": 0.9387,
+      "step": 16283
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.22358603835455e-05,
+      "loss": 1.0892,
+      "step": 16284
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.222441250216502e-05,
+      "loss": 1.1599,
+      "step": 16285
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.221296575722869e-05,
+      "loss": 0.9623,
+      "step": 16286
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.220152014896161e-05,
+      "loss": 0.759,
+      "step": 16287
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.219007567758889e-05,
+      "loss": 0.8638,
+      "step": 16288
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2178632343335675e-05,
+      "loss": 0.9049,
+      "step": 16289
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.216719014642709e-05,
+      "loss": 0.8336,
+      "step": 16290
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.215574908708817e-05,
+      "loss": 0.8474,
+      "step": 16291
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2144309165543916e-05,
+      "loss": 0.973,
+      "step": 16292
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.213287038201943e-05,
+      "loss": 0.6396,
+      "step": 16293
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.212143273673968e-05,
+      "loss": 0.9926,
+      "step": 16294
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2109996229929605e-05,
+      "loss": 0.9074,
+      "step": 16295
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2098560861814215e-05,
+      "loss": 1.0061,
+      "step": 16296
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20871266326184e-05,
+      "loss": 0.7901,
+      "step": 16297
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.207569354256713e-05,
+      "loss": 0.9758,
+      "step": 16298
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2064261591885254e-05,
+      "loss": 0.7256,
+      "step": 16299
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20528307807976e-05,
+      "loss": 0.8727,
+      "step": 16300
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20414011095291e-05,
+      "loss": 0.9012,
+      "step": 16301
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2029972578304487e-05,
+      "loss": 0.8106,
+      "step": 16302
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.2018545187348645e-05,
+      "loss": 0.8152,
+      "step": 16303
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.20071189368863e-05,
+      "loss": 0.8776,
+      "step": 16304
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.199569382714218e-05,
+      "loss": 1.0245,
+      "step": 16305
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.198426985834105e-05,
+      "loss": 0.8223,
+      "step": 16306
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1972847030707696e-05,
+      "loss": 0.8936,
+      "step": 16307
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1961425344466646e-05,
+      "loss": 0.8506,
+      "step": 16308
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.195000479984265e-05,
+      "loss": 1.1464,
+      "step": 16309
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.193858539706038e-05,
+      "loss": 0.8,
+      "step": 16310
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1927167136344424e-05,
+      "loss": 0.9257,
+      "step": 16311
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.191575001791934e-05,
+      "loss": 0.7966,
+      "step": 16312
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1904334042009776e-05,
+      "loss": 1.0127,
+      "step": 16313
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.18929192088402e-05,
+      "loss": 0.976,
+      "step": 16314
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1881505518635234e-05,
+      "loss": 0.9394,
+      "step": 16315
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1870092971619326e-05,
+      "loss": 1.1406,
+      "step": 16316
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.185868156801694e-05,
+      "loss": 0.9508,
+      "step": 16317
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1847271308052583e-05,
+      "loss": 0.8579,
+      "step": 16318
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.183586219195072e-05,
+      "loss": 1.0127,
+      "step": 16319
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1824454219935716e-05,
+      "loss": 0.828,
+      "step": 16320
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.181304739223194e-05,
+      "loss": 0.7533,
+      "step": 16321
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.180164170906385e-05,
+      "loss": 0.913,
+      "step": 16322
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.179023717065571e-05,
+      "loss": 0.9479,
+      "step": 16323
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.177883377723196e-05,
+      "loss": 0.9845,
+      "step": 16324
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1767431529016766e-05,
+      "loss": 0.9203,
+      "step": 16325
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1756030426234474e-05,
+      "loss": 1.0174,
+      "step": 16326
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1744630469109366e-05,
+      "loss": 0.9055,
+      "step": 16327
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.173323165786567e-05,
+      "loss": 0.887,
+      "step": 16328
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1721833992727535e-05,
+      "loss": 0.91,
+      "step": 16329
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.171043747391925e-05,
+      "loss": 0.8553,
+      "step": 16330
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.169904210166491e-05,
+      "loss": 0.9183,
+      "step": 16331
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1687647876188715e-05,
+      "loss": 0.9956,
+      "step": 16332
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1676254797714766e-05,
+      "loss": 0.9583,
+      "step": 16333
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.166486286646713e-05,
+      "loss": 0.8222,
+      "step": 16334
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.165347208266992e-05,
+      "loss": 0.8829,
+      "step": 16335
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.164208244654723e-05,
+      "loss": 0.9518,
+      "step": 16336
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.163069395832305e-05,
+      "loss": 0.931,
+      "step": 16337
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.161930661822137e-05,
+      "loss": 0.7416,
+      "step": 16338
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.160792042646623e-05,
+      "loss": 1.1626,
+      "step": 16339
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1596535383281565e-05,
+      "loss": 0.8986,
+      "step": 16340
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1585151488891305e-05,
+      "loss": 1.1432,
+      "step": 16341
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.157376874351941e-05,
+      "loss": 0.9566,
+      "step": 16342
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1562387147389735e-05,
+      "loss": 0.9522,
+      "step": 16343
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.155100670072621e-05,
+      "loss": 1.0151,
+      "step": 16344
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1539627403752666e-05,
+      "loss": 0.7605,
+      "step": 16345
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1528249256692875e-05,
+      "loss": 0.9107,
+      "step": 16346
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1516872259770747e-05,
+      "loss": 0.8364,
+      "step": 16347
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1505496413209966e-05,
+      "loss": 1.0321,
+      "step": 16348
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.149412171723438e-05,
+      "loss": 0.9435,
+      "step": 16349
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1482748172067696e-05,
+      "loss": 0.7619,
+      "step": 16350
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1471375777933594e-05,
+      "loss": 0.896,
+      "step": 16351
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1460004535055795e-05,
+      "loss": 0.8236,
+      "step": 16352
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1448634443658065e-05,
+      "loss": 1.1064,
+      "step": 16353
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.143726550396388e-05,
+      "loss": 0.957,
+      "step": 16354
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.142589771619696e-05,
+      "loss": 0.8691,
+      "step": 16355
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1414531080580934e-05,
+      "loss": 0.9932,
+      "step": 16356
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.140316559733935e-05,
+      "loss": 0.9529,
+      "step": 16357
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.139180126669574e-05,
+      "loss": 1.0256,
+      "step": 16358
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.138043808887371e-05,
+      "loss": 0.8709,
+      "step": 16359
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.136907606409668e-05,
+      "loss": 1.0737,
+      "step": 16360
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.135771519258824e-05,
+      "loss": 1.018,
+      "step": 16361
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.134635547457181e-05,
+      "loss": 0.8285,
+      "step": 16362
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.133499691027079e-05,
+      "loss": 0.9441,
+      "step": 16363
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.132363949990866e-05,
+      "loss": 0.8045,
+      "step": 16364
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1312283243708836e-05,
+      "loss": 1.0763,
+      "step": 16365
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.130092814189468e-05,
+      "loss": 1.042,
+      "step": 16366
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.128957419468952e-05,
+      "loss": 0.8241,
+      "step": 16367
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1278221402316674e-05,
+      "loss": 1.062,
+      "step": 16368
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.126686976499947e-05,
+      "loss": 0.919,
+      "step": 16369
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.125551928296129e-05,
+      "loss": 0.9333,
+      "step": 16370
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.124416995642523e-05,
+      "loss": 1.0822,
+      "step": 16371
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1232821785614615e-05,
+      "loss": 1.0066,
+      "step": 16372
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.12214747707527e-05,
+      "loss": 0.607,
+      "step": 16373
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.121012891206263e-05,
+      "loss": 0.7365,
+      "step": 16374
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.119878420976756e-05,
+      "loss": 0.8614,
+      "step": 16375
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1187440664090704e-05,
+      "loss": 0.9543,
+      "step": 16376
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1176098275255115e-05,
+      "loss": 0.8282,
+      "step": 16377
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.116475704348398e-05,
+      "loss": 1.0929,
+      "step": 16378
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.115341696900033e-05,
+      "loss": 0.8856,
+      "step": 16379
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.114207805202719e-05,
+      "loss": 0.9346,
+      "step": 16380
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1130740292787636e-05,
+      "loss": 0.8311,
+      "step": 16381
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1119403691504735e-05,
+      "loss": 1.1359,
+      "step": 16382
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.110806824840141e-05,
+      "loss": 0.8836,
+      "step": 16383
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1096733963700615e-05,
+      "loss": 0.918,
+      "step": 16384
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.108540083762536e-05,
+      "loss": 0.9954,
+      "step": 16385
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.107406887039852e-05,
+      "loss": 0.9002,
+      "step": 16386
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1062738062242976e-05,
+      "loss": 0.9621,
+      "step": 16387
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.105140841338168e-05,
+      "loss": 0.8282,
+      "step": 16388
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.1040079924037386e-05,
+      "loss": 0.9673,
+      "step": 16389
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.102875259443303e-05,
+      "loss": 0.8546,
+      "step": 16390
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.101742642479136e-05,
+      "loss": 0.767,
+      "step": 16391
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.100610141533514e-05,
+      "loss": 0.8518,
+      "step": 16392
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0994777566287204e-05,
+      "loss": 1.0065,
+      "step": 16393
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.098345487787021e-05,
+      "loss": 0.9114,
+      "step": 16394
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.097213335030694e-05,
+      "loss": 0.8667,
+      "step": 16395
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.096081298382008e-05,
+      "loss": 0.8861,
+      "step": 16396
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0949493778632244e-05,
+      "loss": 0.9064,
+      "step": 16397
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.093817573496612e-05,
+      "loss": 0.9605,
+      "step": 16398
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0926858853044416e-05,
+      "loss": 0.9447,
+      "step": 16399
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.091554313308957e-05,
+      "loss": 1.0733,
+      "step": 16400
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0904228575324257e-05,
+      "loss": 1.1567,
+      "step": 16401
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0892915179971046e-05,
+      "loss": 0.7704,
+      "step": 16402
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.088160294725246e-05,
+      "loss": 1.0231,
+      "step": 16403
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0870291877390956e-05,
+      "loss": 0.9785,
+      "step": 16404
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.085898197060911e-05,
+      "loss": 0.9021,
+      "step": 16405
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.08476732271293e-05,
+      "loss": 1.037,
+      "step": 16406
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0836365647174045e-05,
+      "loss": 0.8189,
+      "step": 16407
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0825059230965737e-05,
+      "loss": 1.1207,
+      "step": 16408
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.081375397872672e-05,
+      "loss": 0.8412,
+      "step": 16409
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.080244989067942e-05,
+      "loss": 1.0153,
+      "step": 16410
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.079114696704622e-05,
+      "loss": 0.9156,
+      "step": 16411
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0779845208049413e-05,
+      "loss": 0.9086,
+      "step": 16412
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0768544613911294e-05,
+      "loss": 1.0942,
+      "step": 16413
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.075724518485413e-05,
+      "loss": 1.0732,
+      "step": 16414
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0745946921100196e-05,
+      "loss": 0.9099,
+      "step": 16415
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.073464982287181e-05,
+      "loss": 0.8835,
+      "step": 16416
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0723353890391034e-05,
+      "loss": 0.8123,
+      "step": 16417
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0712059123880155e-05,
+      "loss": 0.9485,
+      "step": 16418
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0700765523561344e-05,
+      "loss": 0.8085,
+      "step": 16419
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0689473089656727e-05,
+      "loss": 0.9713,
+      "step": 16420
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.067818182238838e-05,
+      "loss": 0.9213,
+      "step": 16421
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.06668917219785e-05,
+      "loss": 1.0429,
+      "step": 16422
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.065560278864905e-05,
+      "loss": 1.0114,
+      "step": 16423
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.064431502262219e-05,
+      "loss": 0.9447,
+      "step": 16424
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.063302842411989e-05,
+      "loss": 0.9332,
+      "step": 16425
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.062174299336414e-05,
+      "loss": 0.8187,
+      "step": 16426
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.061045873057696e-05,
+      "loss": 0.907,
+      "step": 16427
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0599175635980324e-05,
+      "loss": 0.9191,
+      "step": 16428
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.058789370979615e-05,
+      "loss": 0.7799,
+      "step": 16429
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.057661295224633e-05,
+      "loss": 0.8285,
+      "step": 16430
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0565333363552815e-05,
+      "loss": 0.8064,
+      "step": 16431
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0554054943937426e-05,
+      "loss": 0.8772,
+      "step": 16432
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0542777693622024e-05,
+      "loss": 0.9985,
+      "step": 16433
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.053150161282839e-05,
+      "loss": 1.1489,
+      "step": 16434
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0520226701778374e-05,
+      "loss": 0.9219,
+      "step": 16435
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.050895296069377e-05,
+      "loss": 0.8178,
+      "step": 16436
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.049768038979631e-05,
+      "loss": 0.8753,
+      "step": 16437
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0486408989307677e-05,
+      "loss": 1.0032,
+      "step": 16438
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.047513875944966e-05,
+      "loss": 0.9563,
+      "step": 16439
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.046386970044388e-05,
+      "loss": 1.0057,
+      "step": 16440
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.045260181251205e-05,
+      "loss": 0.8475,
+      "step": 16441
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0441335095875786e-05,
+      "loss": 1.2263,
+      "step": 16442
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0430069550756665e-05,
+      "loss": 1.0559,
+      "step": 16443
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.041880517737633e-05,
+      "loss": 0.9311,
+      "step": 16444
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.04075419759564e-05,
+      "loss": 0.7442,
+      "step": 16445
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0396279946718284e-05,
+      "loss": 1.0005,
+      "step": 16446
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.038501908988358e-05,
+      "loss": 0.9406,
+      "step": 16447
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.037375940567384e-05,
+      "loss": 0.8532,
+      "step": 16448
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0362500894310476e-05,
+      "loss": 1.0112,
+      "step": 16449
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.035124355601493e-05,
+      "loss": 0.8603,
+      "step": 16450
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.033998739100869e-05,
+      "loss": 1.0406,
+      "step": 16451
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.032873239951311e-05,
+      "loss": 1.0161,
+      "step": 16452
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.031747858174964e-05,
+      "loss": 0.8757,
+      "step": 16453
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.03062259379396e-05,
+      "loss": 0.9663,
+      "step": 16454
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0294974468304305e-05,
+      "loss": 0.8901,
+      "step": 16455
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0283724173065115e-05,
+      "loss": 1.0529,
+      "step": 16456
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0272475052443295e-05,
+      "loss": 1.1318,
+      "step": 16457
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.026122710666015e-05,
+      "loss": 0.917,
+      "step": 16458
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.024998033593691e-05,
+      "loss": 0.9961,
+      "step": 16459
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.023873474049474e-05,
+      "loss": 0.862,
+      "step": 16460
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0227490320554916e-05,
+      "loss": 0.9172,
+      "step": 16461
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.021624707633865e-05,
+      "loss": 1.0883,
+      "step": 16462
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.020500500806698e-05,
+      "loss": 0.8899,
+      "step": 16463
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0193764115961083e-05,
+      "loss": 0.7944,
+      "step": 16464
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.018252440024212e-05,
+      "loss": 0.93,
+      "step": 16465
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.017128586113114e-05,
+      "loss": 0.6431,
+      "step": 16466
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.016004849884916e-05,
+      "loss": 0.8629,
+      "step": 16467
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.014881231361729e-05,
+      "loss": 0.9409,
+      "step": 16468
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0137577305656485e-05,
+      "loss": 0.8179,
+      "step": 16469
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.01263434751878e-05,
+      "loss": 0.7434,
+      "step": 16470
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0115110822432175e-05,
+      "loss": 0.9774,
+      "step": 16471
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.010387934761052e-05,
+      "loss": 0.9853,
+      "step": 16472
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.009264905094378e-05,
+      "loss": 0.9945,
+      "step": 16473
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0081419932652905e-05,
+      "loss": 1.0442,
+      "step": 16474
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.007019199295873e-05,
+      "loss": 0.782,
+      "step": 16475
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0058965232082116e-05,
+      "loss": 0.9546,
+      "step": 16476
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.004773965024383e-05,
+      "loss": 1.0061,
+      "step": 16477
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0036515247664786e-05,
+      "loss": 0.8533,
+      "step": 16478
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.002529202456572e-05,
+      "loss": 0.7976,
+      "step": 16479
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.0014069981167345e-05,
+      "loss": 0.9153,
+      "step": 16480
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 4.000284911769045e-05,
+      "loss": 1.061,
+      "step": 16481
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.999162943435577e-05,
+      "loss": 0.6866,
+      "step": 16482
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.998041093138397e-05,
+      "loss": 1.0225,
+      "step": 16483
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.996919360899568e-05,
+      "loss": 0.9106,
+      "step": 16484
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9957977467411615e-05,
+      "loss": 0.9033,
+      "step": 16485
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.994676250685232e-05,
+      "loss": 0.9341,
+      "step": 16486
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9935548727538474e-05,
+      "loss": 0.9676,
+      "step": 16487
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.992433612969061e-05,
+      "loss": 0.666,
+      "step": 16488
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.991312471352924e-05,
+      "loss": 1.0151,
+      "step": 16489
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9901914479274935e-05,
+      "loss": 0.8848,
+      "step": 16490
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.989070542714828e-05,
+      "loss": 1.0678,
+      "step": 16491
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.987949755736958e-05,
+      "loss": 1.0521,
+      "step": 16492
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9868290870159405e-05,
+      "loss": 0.8158,
+      "step": 16493
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.985708536573821e-05,
+      "loss": 0.947,
+      "step": 16494
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.984588104432637e-05,
+      "loss": 0.9737,
+      "step": 16495
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9834677906144236e-05,
+      "loss": 1.0222,
+      "step": 16496
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9823475951412235e-05,
+      "loss": 1.0474,
+      "step": 16497
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9812275180350665e-05,
+      "loss": 0.876,
+      "step": 16498
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9801075593179896e-05,
+      "loss": 0.9768,
+      "step": 16499
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97898771901202e-05,
+      "loss": 0.9403,
+      "step": 16500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9778679971391785e-05,
+      "loss": 0.9397,
+      "step": 16501
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.976748393721501e-05,
+      "loss": 0.9175,
+      "step": 16502
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.975628908781001e-05,
+      "loss": 0.8979,
+      "step": 16503
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.974509542339705e-05,
+      "loss": 0.8845,
+      "step": 16504
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.97339029441963e-05,
+      "loss": 0.7187,
+      "step": 16505
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.972271165042785e-05,
+      "loss": 1.0887,
+      "step": 16506
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9711521542311894e-05,
+      "loss": 0.9952,
+      "step": 16507
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.970033262006859e-05,
+      "loss": 0.94,
+      "step": 16508
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.96891448839179e-05,
+      "loss": 0.8051,
+      "step": 16509
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.9677958334079946e-05,
+      "loss": 0.7823,
+      "step": 16510
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 3.96667729707748e-05,
+      "loss": 0.8771,
+      "step": 16511
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9655588794222456e-05,
+      "loss": 0.8856,
+      "step": 16512
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.964440580464286e-05,
+      "loss": 0.872,
+      "step": 16513
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.963322400225605e-05,
+      "loss": 0.8827,
+      "step": 16514
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9622043387281914e-05,
+      "loss": 0.853,
+      "step": 16515
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9610863959940436e-05,
+      "loss": 0.9256,
+      "step": 16516
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.959968572045148e-05,
+      "loss": 1.0066,
+      "step": 16517
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9588508669034886e-05,
+      "loss": 0.8482,
+      "step": 16518
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.957733280591055e-05,
+      "loss": 0.9871,
+      "step": 16519
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9566158131298326e-05,
+      "loss": 0.8242,
+      "step": 16520
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9554984645417984e-05,
+      "loss": 1.0007,
+      "step": 16521
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.954381234848931e-05,
+      "loss": 0.9104,
+      "step": 16522
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9532641240732016e-05,
+      "loss": 0.7277,
+      "step": 16523
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.952147132236593e-05,
+      "loss": 0.8742,
+      "step": 16524
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.951030259361072e-05,
+      "loss": 1.1166,
+      "step": 16525
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.949913505468603e-05,
+      "loss": 0.9748,
+      "step": 16526
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9487968705811573e-05,
+      "loss": 0.9083,
+      "step": 16527
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.947680354720702e-05,
+      "loss": 0.9269,
+      "step": 16528
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9465639579091954e-05,
+      "loss": 0.7128,
+      "step": 16529
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.945447680168592e-05,
+      "loss": 1.0983,
+      "step": 16530
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.94433152152086e-05,
+      "loss": 0.8935,
+      "step": 16531
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.943215481987943e-05,
+      "loss": 1.0794,
+      "step": 16532
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.942099561591802e-05,
+      "loss": 0.9585,
+      "step": 16533
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9409837603543844e-05,
+      "loss": 1.0833,
+      "step": 16534
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9398680782976336e-05,
+      "loss": 1.155,
+      "step": 16535
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.938752515443499e-05,
+      "loss": 0.9747,
+      "step": 16536
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.937637071813931e-05,
+      "loss": 0.7504,
+      "step": 16537
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.936521747430855e-05,
+      "loss": 0.9167,
+      "step": 16538
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.935406542316217e-05,
+      "loss": 0.7766,
+      "step": 16539
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9342914564919575e-05,
+      "loss": 0.9299,
+      "step": 16540
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.933176489980005e-05,
+      "loss": 0.7867,
+      "step": 16541
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.932061642802293e-05,
+      "loss": 0.9495,
+      "step": 16542
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.930946914980744e-05,
+      "loss": 0.9127,
+      "step": 16543
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9298323065372924e-05,
+      "loss": 0.8283,
+      "step": 16544
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.928717817493862e-05,
+      "loss": 0.8052,
+      "step": 16545
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.927603447872373e-05,
+      "loss": 0.9118,
+      "step": 16546
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.926489197694742e-05,
+      "loss": 0.9717,
+      "step": 16547
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9253750669828915e-05,
+      "loss": 0.8983,
+      "step": 16548
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.924261055758731e-05,
+      "loss": 1.0389,
+      "step": 16549
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.923147164044181e-05,
+      "loss": 0.9234,
+      "step": 16550
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.922033391861145e-05,
+      "loss": 0.909,
+      "step": 16551
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.92091973923153e-05,
+      "loss": 1.0823,
+      "step": 16552
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.919806206177244e-05,
+      "loss": 0.9076,
+      "step": 16553
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.918692792720199e-05,
+      "loss": 1.0146,
+      "step": 16554
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9175794988822775e-05,
+      "loss": 0.9649,
+      "step": 16555
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9164663246853895e-05,
+      "loss": 0.9643,
+      "step": 16556
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.915353270151433e-05,
+      "loss": 0.8592,
+      "step": 16557
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.914240335302297e-05,
+      "loss": 0.98,
+      "step": 16558
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.913127520159871e-05,
+      "loss": 0.8273,
+      "step": 16559
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9120148247460496e-05,
+      "loss": 1.0876,
+      "step": 16560
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.910902249082714e-05,
+      "loss": 0.7995,
+      "step": 16561
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.909789793191755e-05,
+      "loss": 0.8487,
+      "step": 16562
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9086774570950516e-05,
+      "loss": 0.8371,
+      "step": 16563
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.907565240814478e-05,
+      "loss": 0.8247,
+      "step": 16564
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.90645314437192e-05,
+      "loss": 0.929,
+      "step": 16565
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9053411677892446e-05,
+      "loss": 0.8678,
+      "step": 16566
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.904229311088333e-05,
+      "loss": 0.745,
+      "step": 16567
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.9031175742910506e-05,
+      "loss": 0.9708,
+      "step": 16568
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.902005957419263e-05,
+      "loss": 0.8095,
+      "step": 16569
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.90089446049484e-05,
+      "loss": 0.9733,
+      "step": 16570
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8997830835396446e-05,
+      "loss": 0.8809,
+      "step": 16571
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8986718265755306e-05,
+      "loss": 0.868,
+      "step": 16572
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.897560689624362e-05,
+      "loss": 0.8349,
+      "step": 16573
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.896449672707999e-05,
+      "loss": 1.0211,
+      "step": 16574
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.895338775848289e-05,
+      "loss": 0.8897,
+      "step": 16575
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.894227999067083e-05,
+      "loss": 0.8407,
+      "step": 16576
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.893117342386236e-05,
+      "loss": 0.8072,
+      "step": 16577
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.892006805827585e-05,
+      "loss": 0.827,
+      "step": 16578
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8908963894129844e-05,
+      "loss": 0.9099,
+      "step": 16579
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.889786093164271e-05,
+      "loss": 0.8466,
+      "step": 16580
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.888675917103283e-05,
+      "loss": 0.9118,
+      "step": 16581
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.887565861251857e-05,
+      "loss": 0.8678,
+      "step": 16582
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.886455925631838e-05,
+      "loss": 1.153,
+      "step": 16583
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8853461102650435e-05,
+      "loss": 1.0056,
+      "step": 16584
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.88423641517331e-05,
+      "loss": 0.9261,
+      "step": 16585
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.883126840378471e-05,
+      "loss": 0.8461,
+      "step": 16586
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8820173859023435e-05,
+      "loss": 0.9434,
+      "step": 16587
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.880908051766755e-05,
+      "loss": 0.8858,
+      "step": 16588
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.879798837993521e-05,
+      "loss": 0.9688,
+      "step": 16589
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.878689744604462e-05,
+      "loss": 0.8847,
+      "step": 16590
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8775807716213996e-05,
+      "loss": 1.0255,
+      "step": 16591
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.876471919066143e-05,
+      "loss": 1.0137,
+      "step": 16592
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.875363186960499e-05,
+      "loss": 0.9639,
+      "step": 16593
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8742545753262836e-05,
+      "loss": 0.9133,
+      "step": 16594
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8731460841852965e-05,
+      "loss": 0.7724,
+      "step": 16595
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8720377135593475e-05,
+      "loss": 0.8661,
+      "step": 16596
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8709294634702376e-05,
+      "loss": 0.7999,
+      "step": 16597
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8698213339397595e-05,
+      "loss": 0.8826,
+      "step": 16598
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.868713324989714e-05,
+      "loss": 0.8705,
+      "step": 16599
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8676054366419055e-05,
+      "loss": 0.9769,
+      "step": 16600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8664976689181084e-05,
+      "loss": 1.0399,
+      "step": 16601
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.865390021840122e-05,
+      "loss": 1.143,
+      "step": 16602
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.864282495429735e-05,
+      "loss": 1.1405,
+      "step": 16603
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.863175089708732e-05,
+      "loss": 0.8801,
+      "step": 16604
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.862067804698889e-05,
+      "loss": 0.7897,
+      "step": 16605
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8609606404219945e-05,
+      "loss": 0.9626,
+      "step": 16606
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.859853596899819e-05,
+      "loss": 0.7571,
+      "step": 16607
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.858746674154147e-05,
+      "loss": 0.8067,
+      "step": 16608
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.857639872206745e-05,
+      "loss": 0.9801,
+      "step": 16609
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.856533191079383e-05,
+      "loss": 1.0499,
+      "step": 16610
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8554266307938345e-05,
+      "loss": 0.9679,
+      "step": 16611
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8543201913718594e-05,
+      "loss": 0.8843,
+      "step": 16612
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.853213872835228e-05,
+      "loss": 0.8403,
+      "step": 16613
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.852107675205699e-05,
+      "loss": 1.0341,
+      "step": 16614
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8510015985050274e-05,
+      "loss": 0.8322,
+      "step": 16615
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.849895642754975e-05,
+      "loss": 1.103,
+      "step": 16616
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.848789807977294e-05,
+      "loss": 0.8174,
+      "step": 16617
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.847684094193733e-05,
+      "loss": 0.8931,
+      "step": 16618
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8465785014260436e-05,
+      "loss": 0.8816,
+      "step": 16619
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8454730296959775e-05,
+      "loss": 0.7937,
+      "step": 16620
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.844367679025274e-05,
+      "loss": 0.9413,
+      "step": 16621
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.843262449435674e-05,
+      "loss": 1.0647,
+      "step": 16622
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.842157340948923e-05,
+      "loss": 1.0071,
+      "step": 16623
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.841052353586752e-05,
+      "loss": 0.8574,
+      "step": 16624
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.839947487370903e-05,
+      "loss": 0.9242,
+      "step": 16625
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8388427423231044e-05,
+      "loss": 0.9342,
+      "step": 16626
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.837738118465084e-05,
+      "loss": 1.0316,
+      "step": 16627
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8366336158185725e-05,
+      "loss": 1.1218,
+      "step": 16628
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8355292344053026e-05,
+      "loss": 0.7779,
+      "step": 16629
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.834424974246985e-05,
+      "loss": 0.8477,
+      "step": 16630
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8333208353653484e-05,
+      "loss": 0.8083,
+      "step": 16631
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.832216817782106e-05,
+      "loss": 0.9414,
+      "step": 16632
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8311129215189804e-05,
+      "loss": 0.9037,
+      "step": 16633
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8300091465976804e-05,
+      "loss": 0.7691,
+      "step": 16634
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8289054930399156e-05,
+      "loss": 1.1882,
+      "step": 16635
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.827801960867397e-05,
+      "loss": 1.12,
+      "step": 16636
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.826698550101836e-05,
+      "loss": 0.8235,
+      "step": 16637
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.825595260764932e-05,
+      "loss": 0.7554,
+      "step": 16638
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8244920928783836e-05,
+      "loss": 1.0724,
+      "step": 16639
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8233890464638953e-05,
+      "loss": 0.8363,
+      "step": 16640
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8222861215431605e-05,
+      "loss": 0.9859,
+      "step": 16641
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.821183318137877e-05,
+      "loss": 0.7651,
+      "step": 16642
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.820080636269737e-05,
+      "loss": 0.8421,
+      "step": 16643
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.818978075960424e-05,
+      "loss": 0.7893,
+      "step": 16644
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.817875637231631e-05,
+      "loss": 1.064,
+      "step": 16645
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.816773320105046e-05,
+      "loss": 0.9898,
+      "step": 16646
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.815671124602343e-05,
+      "loss": 0.7785,
+      "step": 16647
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.814569050745205e-05,
+      "loss": 1.1599,
+      "step": 16648
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.813467098555314e-05,
+      "loss": 0.9008,
+      "step": 16649
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.812365268054343e-05,
+      "loss": 0.9916,
+      "step": 16650
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.811263559263965e-05,
+      "loss": 0.8791,
+      "step": 16651
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.810161972205847e-05,
+      "loss": 0.8992,
+      "step": 16652
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8090605069016595e-05,
+      "loss": 0.8679,
+      "step": 16653
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.807959163373073e-05,
+      "loss": 0.9961,
+      "step": 16654
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.806857941641746e-05,
+      "loss": 1.0615,
+      "step": 16655
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8057568417293376e-05,
+      "loss": 1.045,
+      "step": 16656
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8046558636575124e-05,
+      "loss": 0.9139,
+      "step": 16657
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8035550074479196e-05,
+      "loss": 1.048,
+      "step": 16658
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.8024542731222215e-05,
+      "loss": 1.0002,
+      "step": 16659
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.801353660702064e-05,
+      "loss": 1.1085,
+      "step": 16660
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.800253170209094e-05,
+      "loss": 0.9377,
+      "step": 16661
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7991528016649646e-05,
+      "loss": 0.8929,
+      "step": 16662
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.798052555091316e-05,
+      "loss": 0.9176,
+      "step": 16663
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.796952430509787e-05,
+      "loss": 0.9636,
+      "step": 16664
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.79585242794202e-05,
+      "loss": 0.9102,
+      "step": 16665
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.794752547409658e-05,
+      "loss": 1.0479,
+      "step": 16666
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7936527889343285e-05,
+      "loss": 1.0666,
+      "step": 16667
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7925531525376623e-05,
+      "loss": 0.7858,
+      "step": 16668
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7914536382412956e-05,
+      "loss": 1.0705,
+      "step": 16669
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.790354246066849e-05,
+      "loss": 0.9976,
+      "step": 16670
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.789254976035955e-05,
+      "loss": 1.04,
+      "step": 16671
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.788155828170232e-05,
+      "loss": 0.9685,
+      "step": 16672
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7870568024912976e-05,
+      "loss": 0.7961,
+      "step": 16673
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.785957899020776e-05,
+      "loss": 1.0696,
+      "step": 16674
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7848591177802784e-05,
+      "loss": 0.9563,
+      "step": 16675
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.783760458791415e-05,
+      "loss": 0.9344,
+      "step": 16676
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.782661922075803e-05,
+      "loss": 0.9211,
+      "step": 16677
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.781563507655044e-05,
+      "loss": 0.8362,
+      "step": 16678
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7804652155507513e-05,
+      "loss": 0.9461,
+      "step": 16679
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.779367045784523e-05,
+      "loss": 1.1318,
+      "step": 16680
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7782689983779596e-05,
+      "loss": 0.7816,
+      "step": 16681
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.77717107335266e-05,
+      "loss": 1.1219,
+      "step": 16682
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.776073270730225e-05,
+      "loss": 0.8403,
+      "step": 16683
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.774975590532246e-05,
+      "loss": 0.8327,
+      "step": 16684
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.773878032780309e-05,
+      "loss": 0.8437,
+      "step": 16685
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7727805974960105e-05,
+      "loss": 1.1745,
+      "step": 16686
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.771683284700931e-05,
+      "loss": 0.9879,
+      "step": 16687
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7705860944166604e-05,
+      "loss": 1.122,
+      "step": 16688
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.769489026664778e-05,
+      "loss": 0.9523,
+      "step": 16689
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7683920814668586e-05,
+      "loss": 0.8681,
+      "step": 16690
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.767295258844483e-05,
+      "loss": 0.8708,
+      "step": 16691
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.766198558819233e-05,
+      "loss": 0.8827,
+      "step": 16692
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7651019814126654e-05,
+      "loss": 1.0257,
+      "step": 16693
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.764005526646358e-05,
+      "loss": 1.0241,
+      "step": 16694
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7629091945418816e-05,
+      "loss": 0.9716,
+      "step": 16695
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.761812985120797e-05,
+      "loss": 0.9385,
+      "step": 16696
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.760716898404667e-05,
+      "loss": 0.8275,
+      "step": 16697
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.759620934415048e-05,
+      "loss": 1.1648,
+      "step": 16698
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7585250931735006e-05,
+      "loss": 0.8085,
+      "step": 16699
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7574293747015844e-05,
+      "loss": 0.8527,
+      "step": 16700
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7563337790208476e-05,
+      "loss": 0.9225,
+      "step": 16701
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7552383061528375e-05,
+      "loss": 0.7814,
+      "step": 16702
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.754142956119109e-05,
+      "loss": 0.9661,
+      "step": 16703
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7530477289412e-05,
+      "loss": 0.9167,
+      "step": 16704
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7519526246406624e-05,
+      "loss": 0.9656,
+      "step": 16705
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.750857643239032e-05,
+      "loss": 0.8596,
+      "step": 16706
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.749762784757843e-05,
+      "loss": 1.0366,
+      "step": 16707
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.74866804921864e-05,
+      "loss": 1.047,
+      "step": 16708
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.747573436642951e-05,
+      "loss": 0.8126,
+      "step": 16709
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.746478947052305e-05,
+      "loss": 0.7849,
+      "step": 16710
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.745384580468233e-05,
+      "loss": 0.97,
+      "step": 16711
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.744290336912265e-05,
+      "loss": 0.7414,
+      "step": 16712
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.743196216405922e-05,
+      "loss": 0.9419,
+      "step": 16713
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.742102218970721e-05,
+      "loss": 1.0121,
+      "step": 16714
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.741008344628188e-05,
+      "loss": 1.0459,
+      "step": 16715
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7399145933998324e-05,
+      "loss": 1.1977,
+      "step": 16716
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.738820965307176e-05,
+      "loss": 1.0423,
+      "step": 16717
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.737727460371726e-05,
+      "loss": 1.0986,
+      "step": 16718
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.736634078614988e-05,
+      "loss": 0.9005,
+      "step": 16719
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.735540820058476e-05,
+      "loss": 0.777,
+      "step": 16720
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.734447684723692e-05,
+      "loss": 1.0523,
+      "step": 16721
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.733354672632134e-05,
+      "loss": 1.048,
+      "step": 16722
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7322617838053064e-05,
+      "loss": 0.8833,
+      "step": 16723
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.731169018264702e-05,
+      "loss": 1.1288,
+      "step": 16724
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.730076376031821e-05,
+      "loss": 1.0109,
+      "step": 16725
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7289838571281535e-05,
+      "loss": 0.9595,
+      "step": 16726
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.727891461575184e-05,
+      "loss": 1.0064,
+      "step": 16727
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.726799189394404e-05,
+      "loss": 0.8725,
+      "step": 16728
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.725707040607303e-05,
+      "loss": 0.8674,
+      "step": 16729
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7246150152353586e-05,
+      "loss": 0.8726,
+      "step": 16730
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7235231133000494e-05,
+      "loss": 0.9483,
+      "step": 16731
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.722431334822857e-05,
+      "loss": 1.0611,
+      "step": 16732
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.721339679825252e-05,
+      "loss": 0.9226,
+      "step": 16733
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.720248148328713e-05,
+      "loss": 0.9707,
+      "step": 16734
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.719156740354709e-05,
+      "loss": 0.8641,
+      "step": 16735
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.718065455924702e-05,
+      "loss": 0.8914,
+      "step": 16736
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.716974295060163e-05,
+      "loss": 0.9019,
+      "step": 16737
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.71588325778256e-05,
+      "loss": 0.9348,
+      "step": 16738
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7147923441133404e-05,
+      "loss": 0.7904,
+      "step": 16739
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7137015540739736e-05,
+      "loss": 1.0587,
+      "step": 16740
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.712610887685909e-05,
+      "loss": 0.9751,
+      "step": 16741
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 3.7115203449706056e-05,
+      "loss": 0.8399,
+      "step": 16742
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.710429925949511e-05,
+      "loss": 0.7857,
+      "step": 16743
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7093396306440707e-05,
+      "loss": 0.8794,
+      "step": 16744
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7082494590757355e-05,
+      "loss": 0.8922,
+      "step": 16745
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.707159411265949e-05,
+      "loss": 0.9488,
+      "step": 16746
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.706069487236152e-05,
+      "loss": 1.1121,
+      "step": 16747
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.704979687007779e-05,
+      "loss": 0.8184,
+      "step": 16748
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.703890010602274e-05,
+      "loss": 0.8126,
+      "step": 16749
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.7028004580410614e-05,
+      "loss": 0.9893,
+      "step": 16750
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.701711029345583e-05,
+      "loss": 0.9947,
+      "step": 16751
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.700621724537261e-05,
+      "loss": 0.9303,
+      "step": 16752
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6995325436375216e-05,
+      "loss": 1.1419,
+      "step": 16753
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.698443486667793e-05,
+      "loss": 0.901,
+      "step": 16754
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.697354553649497e-05,
+      "loss": 0.9578,
+      "step": 16755
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.696265744604046e-05,
+      "loss": 0.8026,
+      "step": 16756
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.695177059552861e-05,
+      "loss": 0.9311,
+      "step": 16757
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.694088498517362e-05,
+      "loss": 1.0099,
+      "step": 16758
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6930000615189554e-05,
+      "loss": 0.9524,
+      "step": 16759
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6919117485790475e-05,
+      "loss": 0.8474,
+      "step": 16760
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6908235597190546e-05,
+      "loss": 0.8683,
+      "step": 16761
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.68973549496037e-05,
+      "loss": 0.8252,
+      "step": 16762
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.688647554324407e-05,
+      "loss": 0.9723,
+      "step": 16763
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.68755973783256e-05,
+      "loss": 0.8155,
+      "step": 16764
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.686472045506223e-05,
+      "loss": 1.0033,
+      "step": 16765
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.685384477366798e-05,
+      "loss": 0.902,
+      "step": 16766
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.684297033435673e-05,
+      "loss": 0.745,
+      "step": 16767
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.683209713734237e-05,
+      "loss": 0.7603,
+      "step": 16768
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.682122518283881e-05,
+      "loss": 0.8602,
+      "step": 16769
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.681035447105986e-05,
+      "loss": 0.86,
+      "step": 16770
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6799485002219414e-05,
+      "loss": 0.9294,
+      "step": 16771
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.678861677653122e-05,
+      "loss": 0.991,
+      "step": 16772
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.677774979420904e-05,
+      "loss": 0.9404,
+      "step": 16773
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6766884055466656e-05,
+      "loss": 0.93,
+      "step": 16774
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.675601956051783e-05,
+      "loss": 1.0493,
+      "step": 16775
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6745156309576225e-05,
+      "loss": 0.8159,
+      "step": 16776
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.67342943028555e-05,
+      "loss": 0.9785,
+      "step": 16777
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6723433540569364e-05,
+      "loss": 0.8961,
+      "step": 16778
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.671257402293139e-05,
+      "loss": 0.8519,
+      "step": 16779
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.670171575015525e-05,
+      "loss": 1.0154,
+      "step": 16780
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66908587224545e-05,
+      "loss": 0.8235,
+      "step": 16781
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.668000294004266e-05,
+      "loss": 0.6524,
+      "step": 16782
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.666914840313331e-05,
+      "loss": 0.9362,
+      "step": 16783
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.665829511193996e-05,
+      "loss": 0.9145,
+      "step": 16784
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.664744306667602e-05,
+      "loss": 0.9813,
+      "step": 16785
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.663659226755505e-05,
+      "loss": 0.9804,
+      "step": 16786
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.66257427147904e-05,
+      "loss": 0.9752,
+      "step": 16787
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.661489440859556e-05,
+      "loss": 0.8094,
+      "step": 16788
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.660404734918387e-05,
+      "loss": 1.0307,
+      "step": 16789
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6593201536768665e-05,
+      "loss": 0.9318,
+      "step": 16790
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6582356971563315e-05,
+      "loss": 0.7419,
+      "step": 16791
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.657151365378116e-05,
+      "loss": 0.9315,
+      "step": 16792
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6560671583635467e-05,
+      "loss": 0.7554,
+      "step": 16793
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.654983076133945e-05,
+      "loss": 0.983,
+      "step": 16794
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.653899118710643e-05,
+      "loss": 0.8545,
+      "step": 16795
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.652815286114953e-05,
+      "loss": 0.8676,
+      "step": 16796
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6517315783682036e-05,
+      "loss": 0.925,
+      "step": 16797
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.650647995491707e-05,
+      "loss": 0.8203,
+      "step": 16798
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.649564537506772e-05,
+      "loss": 0.8752,
+      "step": 16799
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6484812044347196e-05,
+      "loss": 1.1031,
+      "step": 16800
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.647397996296854e-05,
+      "loss": 0.7802,
+      "step": 16801
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.646314913114479e-05,
+      "loss": 0.9612,
+      "step": 16802
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.645231954908902e-05,
+      "loss": 0.833,
+      "step": 16803
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.644149121701428e-05,
+      "loss": 0.815,
+      "step": 16804
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.643066413513353e-05,
+      "loss": 1.0174,
+      "step": 16805
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.641983830365975e-05,
+      "loss": 0.7714,
+      "step": 16806
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.640901372280583e-05,
+      "loss": 1.0098,
+      "step": 16807
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6398190392784735e-05,
+      "loss": 0.8461,
+      "step": 16808
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.638736831380939e-05,
+      "loss": 0.9143,
+      "step": 16809
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.637654748609263e-05,
+      "loss": 0.8449,
+      "step": 16810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.636572790984727e-05,
+      "loss": 0.8451,
+      "step": 16811
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.63549095852862e-05,
+      "loss": 0.8863,
+      "step": 16812
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.634409251262217e-05,
+      "loss": 0.919,
+      "step": 16813
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.633327669206794e-05,
+      "loss": 0.9573,
+      "step": 16814
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.63224621238363e-05,
+      "loss": 0.7899,
+      "step": 16815
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.631164880813992e-05,
+      "loss": 0.8903,
+      "step": 16816
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.630083674519157e-05,
+      "loss": 0.8143,
+      "step": 16817
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.629002593520386e-05,
+      "loss": 0.9858,
+      "step": 16818
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6279216378389445e-05,
+      "loss": 0.8227,
+      "step": 16819
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.626840807496096e-05,
+      "loss": 0.9292,
+      "step": 16820
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6257601025131026e-05,
+      "loss": 0.8419,
+      "step": 16821
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.624679522911221e-05,
+      "loss": 0.6751,
+      "step": 16822
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.623599068711702e-05,
+      "loss": 0.885,
+      "step": 16823
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.622518739935803e-05,
+      "loss": 1.1007,
+      "step": 16824
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6214385366047696e-05,
+      "loss": 0.877,
+      "step": 16825
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.620358458739859e-05,
+      "loss": 1.1748,
+      "step": 16826
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.619278506362301e-05,
+      "loss": 0.86,
+      "step": 16827
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.618198679493348e-05,
+      "loss": 0.9027,
+      "step": 16828
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.61711897815424e-05,
+      "loss": 1.0165,
+      "step": 16829
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.616039402366215e-05,
+      "loss": 1.0286,
+      "step": 16830
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.614959952150502e-05,
+      "loss": 0.8165,
+      "step": 16831
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.613880627528342e-05,
+      "loss": 0.7926,
+      "step": 16832
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.612801428520959e-05,
+      "loss": 0.926,
+      "step": 16833
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.611722355149585e-05,
+      "loss": 0.9028,
+      "step": 16834
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6106434074354446e-05,
+      "loss": 0.9799,
+      "step": 16835
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.609564585399756e-05,
+      "loss": 0.8044,
+      "step": 16836
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.608485889063744e-05,
+      "loss": 0.9257,
+      "step": 16837
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.607407318448629e-05,
+      "loss": 1.0588,
+      "step": 16838
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.606328873575623e-05,
+      "loss": 0.9847,
+      "step": 16839
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.605250554465936e-05,
+      "loss": 0.8275,
+      "step": 16840
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.604172361140785e-05,
+      "loss": 0.9762,
+      "step": 16841
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.603094293621371e-05,
+      "loss": 1.057,
+      "step": 16842
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6020163519289076e-05,
+      "loss": 0.8436,
+      "step": 16843
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.6009385360845936e-05,
+      "loss": 0.8092,
+      "step": 16844
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.599860846109626e-05,
+      "loss": 1.066,
+      "step": 16845
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.59878328202521e-05,
+      "loss": 0.7928,
+      "step": 16846
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.597705843852538e-05,
+      "loss": 0.8467,
+      "step": 16847
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.596628531612799e-05,
+      "loss": 1.0602,
+      "step": 16848
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5955513453271915e-05,
+      "loss": 1.0775,
+      "step": 16849
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5944742850168965e-05,
+      "loss": 0.7938,
+      "step": 16850
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5933973507031074e-05,
+      "loss": 0.893,
+      "step": 16851
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5923205424070015e-05,
+      "loss": 1.0068,
+      "step": 16852
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.591243860149759e-05,
+      "loss": 0.7157,
+      "step": 16853
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.590167303952561e-05,
+      "loss": 1.0935,
+      "step": 16854
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.589090873836586e-05,
+      "loss": 1.0454,
+      "step": 16855
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.588014569823004e-05,
+      "loss": 1.0099,
+      "step": 16856
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5869383919329824e-05,
+      "loss": 0.9716,
+      "step": 16857
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.585862340187697e-05,
+      "loss": 0.7397,
+      "step": 16858
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.58478641460831e-05,
+      "loss": 0.8829,
+      "step": 16859
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.583710615215982e-05,
+      "loss": 0.9844,
+      "step": 16860
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.582634942031878e-05,
+      "loss": 0.8486,
+      "step": 16861
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.581559395077154e-05,
+      "loss": 1.0037,
+      "step": 16862
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.58048397437297e-05,
+      "loss": 1.0154,
+      "step": 16863
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.579408679940477e-05,
+      "loss": 0.8283,
+      "step": 16864
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.578333511800821e-05,
+      "loss": 1.0501,
+      "step": 16865
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.577258469975156e-05,
+      "loss": 1.0349,
+      "step": 16866
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.576183554484631e-05,
+      "loss": 0.9422,
+      "step": 16867
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.575108765350386e-05,
+      "loss": 0.8639,
+      "step": 16868
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.574034102593557e-05,
+      "loss": 0.9596,
+      "step": 16869
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5729595662352935e-05,
+      "loss": 0.9783,
+      "step": 16870
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.57188515629672e-05,
+      "loss": 1.1083,
+      "step": 16871
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.570810872798983e-05,
+      "loss": 0.8693,
+      "step": 16872
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5697367157632e-05,
+      "loss": 0.9387,
+      "step": 16873
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.568662685210505e-05,
+      "loss": 0.8711,
+      "step": 16874
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.567588781162029e-05,
+      "loss": 0.8271,
+      "step": 16875
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.566515003638892e-05,
+      "loss": 0.9142,
+      "step": 16876
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.565441352662211e-05,
+      "loss": 0.9882,
+      "step": 16877
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5643678282531125e-05,
+      "loss": 0.8742,
+      "step": 16878
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5632944304327055e-05,
+      "loss": 0.7208,
+      "step": 16879
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.56222115922211e-05,
+      "loss": 1.0472,
+      "step": 16880
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.561148014642434e-05,
+      "loss": 0.8876,
+      "step": 16881
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.560074996714783e-05,
+      "loss": 0.8656,
+      "step": 16882
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.559002105460267e-05,
+      "loss": 0.9207,
+      "step": 16883
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.557929340899992e-05,
+      "loss": 1.0387,
+      "step": 16884
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.556856703055058e-05,
+      "loss": 0.8173,
+      "step": 16885
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5557841919465586e-05,
+      "loss": 1.0103,
+      "step": 16886
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.554711807595598e-05,
+      "loss": 1.0627,
+      "step": 16887
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5536395500232614e-05,
+      "loss": 0.7035,
+      "step": 16888
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.552567419250647e-05,
+      "loss": 0.8702,
+      "step": 16889
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.551495415298842e-05,
+      "loss": 1.0166,
+      "step": 16890
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.55042353818893e-05,
+      "loss": 0.9663,
+      "step": 16891
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.549351787941998e-05,
+      "loss": 0.9944,
+      "step": 16892
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.548280164579126e-05,
+      "loss": 0.9087,
+      "step": 16893
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.54720866812139e-05,
+      "loss": 0.9239,
+      "step": 16894
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.546137298589872e-05,
+      "loss": 0.933,
+      "step": 16895
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5450660560056396e-05,
+      "loss": 0.9496,
+      "step": 16896
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.54399494038977e-05,
+      "loss": 0.8958,
+      "step": 16897
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5429239517633296e-05,
+      "loss": 0.7536,
+      "step": 16898
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.541853090147381e-05,
+      "loss": 0.8679,
+      "step": 16899
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.540782355562992e-05,
+      "loss": 0.9735,
+      "step": 16900
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.539711748031227e-05,
+      "loss": 0.8815,
+      "step": 16901
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5386412675731406e-05,
+      "loss": 0.8776,
+      "step": 16902
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.537570914209786e-05,
+      "loss": 0.8489,
+      "step": 16903
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.536500687962224e-05,
+      "loss": 0.9916,
+      "step": 16904
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5354305888515025e-05,
+      "loss": 1.0037,
+      "step": 16905
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5343606168986686e-05,
+      "loss": 0.6727,
+      "step": 16906
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.533290772124772e-05,
+      "loss": 0.8435,
+      "step": 16907
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.532221054550853e-05,
+      "loss": 0.9974,
+      "step": 16908
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.531151464197957e-05,
+      "loss": 0.9501,
+      "step": 16909
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5300820010871214e-05,
+      "loss": 0.9944,
+      "step": 16910
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.529012665239378e-05,
+      "loss": 0.9091,
+      "step": 16911
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.527943456675765e-05,
+      "loss": 1.0781,
+      "step": 16912
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5268743754173173e-05,
+      "loss": 0.8111,
+      "step": 16913
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.525805421485058e-05,
+      "loss": 0.8682,
+      "step": 16914
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.524736594900016e-05,
+      "loss": 0.9851,
+      "step": 16915
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5236678956832094e-05,
+      "loss": 0.7861,
+      "step": 16916
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5225993238556644e-05,
+      "loss": 0.9403,
+      "step": 16917
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.521530879438407e-05,
+      "loss": 1.0767,
+      "step": 16918
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.520462562452438e-05,
+      "loss": 0.892,
+      "step": 16919
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5193943729187794e-05,
+      "loss": 1.1775,
+      "step": 16920
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5183263108584465e-05,
+      "loss": 0.9196,
+      "step": 16921
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.517258376292443e-05,
+      "loss": 0.9707,
+      "step": 16922
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.516190569241772e-05,
+      "loss": 0.7891,
+      "step": 16923
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5151228897274456e-05,
+      "loss": 0.9322,
+      "step": 16924
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.514055337770455e-05,
+      "loss": 0.9448,
+      "step": 16925
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.51298791339181e-05,
+      "loss": 0.9069,
+      "step": 16926
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5119206166125005e-05,
+      "loss": 0.7715,
+      "step": 16927
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.510853447453517e-05,
+      "loss": 1.1046,
+      "step": 16928
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5097864059358554e-05,
+      "loss": 1.021,
+      "step": 16929
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.508719492080507e-05,
+      "loss": 1.0094,
+      "step": 16930
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.507652705908454e-05,
+      "loss": 0.9256,
+      "step": 16931
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5065860474406765e-05,
+      "loss": 1.0748,
+      "step": 16932
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5055195166981645e-05,
+      "loss": 0.8992,
+      "step": 16933
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.504453113701888e-05,
+      "loss": 0.8686,
+      "step": 16934
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.5033868384728296e-05,
+      "loss": 0.7407,
+      "step": 16935
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.502320691031961e-05,
+      "loss": 1.0473,
+      "step": 16936
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.501254671400249e-05,
+      "loss": 1.0329,
+      "step": 16937
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.500188779598669e-05,
+      "loss": 1.0515,
+      "step": 16938
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4991230156481836e-05,
+      "loss": 0.9542,
+      "step": 16939
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4980573795697534e-05,
+      "loss": 0.9288,
+      "step": 16940
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.496991871384346e-05,
+      "loss": 0.9425,
+      "step": 16941
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.495926491112914e-05,
+      "loss": 0.9393,
+      "step": 16942
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.494861238776418e-05,
+      "loss": 0.878,
+      "step": 16943
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.49379611439581e-05,
+      "loss": 0.9319,
+      "step": 16944
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4927311179920366e-05,
+      "loss": 1.0158,
+      "step": 16945
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4916662495860494e-05,
+      "loss": 0.9471,
+      "step": 16946
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4906015091988e-05,
+      "loss": 0.9058,
+      "step": 16947
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4895368968512255e-05,
+      "loss": 0.8768,
+      "step": 16948
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.488472412564264e-05,
+      "loss": 1.0404,
+      "step": 16949
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.487408056358863e-05,
+      "loss": 1.1105,
+      "step": 16950
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.486343828255951e-05,
+      "loss": 0.9395,
+      "step": 16951
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.485279728276463e-05,
+      "loss": 0.7186,
+      "step": 16952
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4842157564413315e-05,
+      "loss": 0.972,
+      "step": 16953
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.483151912771481e-05,
+      "loss": 0.8856,
+      "step": 16954
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.482088197287844e-05,
+      "loss": 1.0919,
+      "step": 16955
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4810246100113386e-05,
+      "loss": 0.8822,
+      "step": 16956
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.479961150962884e-05,
+      "loss": 0.8977,
+      "step": 16957
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.478897820163404e-05,
+      "loss": 0.9289,
+      "step": 16958
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.477834617633808e-05,
+      "loss": 0.9052,
+      "step": 16959
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.476771543395017e-05,
+      "loss": 1.1687,
+      "step": 16960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.475708597467935e-05,
+      "loss": 0.9673,
+      "step": 16961
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.47464577987347e-05,
+      "loss": 1.0267,
+      "step": 16962
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4735830906325303e-05,
+      "loss": 0.9768,
+      "step": 16963
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4725205297660244e-05,
+      "loss": 0.9588,
+      "step": 16964
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4714580972948406e-05,
+      "loss": 0.9739,
+      "step": 16965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.470395793239883e-05,
+      "loss": 0.8118,
+      "step": 16966
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4693336176220516e-05,
+      "loss": 1.0094,
+      "step": 16967
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.468271570462235e-05,
+      "loss": 0.9532,
+      "step": 16968
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.467209651781319e-05,
+      "loss": 0.9219,
+      "step": 16969
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.466147861600201e-05,
+      "loss": 0.9332,
+      "step": 16970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4650861999397577e-05,
+      "loss": 0.9065,
+      "step": 16971
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.4640246668208795e-05,
+      "loss": 0.8364,
+      "step": 16972
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 3.462963262264444e-05,
+      "loss": 0.834,
+      "step": 16973
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.461901986291323e-05,
+      "loss": 0.8774,
+      "step": 16974
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.460840838922398e-05,
+      "loss": 1.0986,
+      "step": 16975
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.459779820178544e-05,
+      "loss": 0.9683,
+      "step": 16976
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4587189300806265e-05,
+      "loss": 1.0645,
+      "step": 16977
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.457658168649512e-05,
+      "loss": 0.9218,
+      "step": 16978
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.456597535906072e-05,
+      "loss": 0.938,
+      "step": 16979
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.455537031871161e-05,
+      "loss": 0.8953,
+      "step": 16980
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4544766565656505e-05,
+      "loss": 0.9296,
+      "step": 16981
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.453416410010385e-05,
+      "loss": 0.7719,
+      "step": 16982
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4523562922262245e-05,
+      "loss": 1.0182,
+      "step": 16983
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4512963032340265e-05,
+      "loss": 0.7979,
+      "step": 16984
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.450236443054635e-05,
+      "loss": 1.0119,
+      "step": 16985
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.449176711708898e-05,
+      "loss": 0.942,
+      "step": 16986
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4481171092176645e-05,
+      "loss": 1.0096,
+      "step": 16987
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.44705763560177e-05,
+      "loss": 1.0299,
+      "step": 16988
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.445998290882062e-05,
+      "loss": 0.8834,
+      "step": 16989
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.444939075079374e-05,
+      "loss": 0.8755,
+      "step": 16990
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.443879988214538e-05,
+      "loss": 0.8421,
+      "step": 16991
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.442821030308388e-05,
+      "loss": 0.8996,
+      "step": 16992
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4417622013817595e-05,
+      "loss": 0.7664,
+      "step": 16993
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4407035014554743e-05,
+      "loss": 1.0648,
+      "step": 16994
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.439644930550354e-05,
+      "loss": 1.0046,
+      "step": 16995
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4385864886872275e-05,
+      "loss": 0.8638,
+      "step": 16996
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4375281758869114e-05,
+      "loss": 0.9662,
+      "step": 16997
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.436469992170218e-05,
+      "loss": 0.9215,
+      "step": 16998
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.435411937557971e-05,
+      "loss": 0.9215,
+      "step": 16999
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.434354012070973e-05,
+      "loss": 1.1011,
+      "step": 17000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.433296215730041e-05,
+      "loss": 0.8419,
+      "step": 17001
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.432238548555978e-05,
+      "loss": 0.8922,
+      "step": 17002
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4311810105695874e-05,
+      "loss": 1.101,
+      "step": 17003
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.430123601791674e-05,
+      "loss": 0.9533,
+      "step": 17004
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.429066322243033e-05,
+      "loss": 0.8787,
+      "step": 17005
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4280091719444665e-05,
+      "loss": 0.9602,
+      "step": 17006
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4269521509167644e-05,
+      "loss": 0.9611,
+      "step": 17007
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.425895259180718e-05,
+      "loss": 1.0113,
+      "step": 17008
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.424838496757117e-05,
+      "loss": 1.022,
+      "step": 17009
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.423781863666755e-05,
+      "loss": 0.9974,
+      "step": 17010
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.422725359930402e-05,
+      "loss": 1.0972,
+      "step": 17011
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.421668985568849e-05,
+      "loss": 0.8186,
+      "step": 17012
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4206127406028745e-05,
+      "loss": 0.9264,
+      "step": 17013
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.419556625053254e-05,
+      "loss": 0.8719,
+      "step": 17014
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.418500638940757e-05,
+      "loss": 1.1056,
+      "step": 17015
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.417444782286161e-05,
+      "loss": 0.8955,
+      "step": 17016
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.416389055110229e-05,
+      "loss": 0.9597,
+      "step": 17017
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.415333457433733e-05,
+      "loss": 0.9675,
+      "step": 17018
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4142779892774336e-05,
+      "loss": 0.9547,
+      "step": 17019
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.413222650662088e-05,
+      "loss": 0.9841,
+      "step": 17020
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4121674416084584e-05,
+      "loss": 0.839,
+      "step": 17021
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4111123621373045e-05,
+      "loss": 1.0216,
+      "step": 17022
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.410057412269375e-05,
+      "loss": 0.9039,
+      "step": 17023
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.409002592025422e-05,
+      "loss": 0.8459,
+      "step": 17024
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.407947901426191e-05,
+      "loss": 0.8093,
+      "step": 17025
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.406893340492429e-05,
+      "loss": 0.9829,
+      "step": 17026
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.405838909244887e-05,
+      "loss": 0.864,
+      "step": 17027
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4047846077042934e-05,
+      "loss": 0.7035,
+      "step": 17028
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.403730435891391e-05,
+      "loss": 1.0715,
+      "step": 17029
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.402676393826919e-05,
+      "loss": 1.0185,
+      "step": 17030
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.401622481531609e-05,
+      "loss": 0.8138,
+      "step": 17031
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.4005686990261855e-05,
+      "loss": 0.7534,
+      "step": 17032
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.399515046331384e-05,
+      "loss": 0.8058,
+      "step": 17033
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.398461523467924e-05,
+      "loss": 0.9913,
+      "step": 17034
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3974081304565345e-05,
+      "loss": 1.1332,
+      "step": 17035
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.396354867317931e-05,
+      "loss": 0.9096,
+      "step": 17036
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.39530173407283e-05,
+      "loss": 0.8373,
+      "step": 17037
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.394248730741948e-05,
+      "loss": 1.0486,
+      "step": 17038
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.393195857346003e-05,
+      "loss": 1.1872,
+      "step": 17039
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3921431139056994e-05,
+      "loss": 0.7149,
+      "step": 17040
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.391090500441743e-05,
+      "loss": 1.0373,
+      "step": 17041
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3900380169748435e-05,
+      "loss": 0.9381,
+      "step": 17042
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.388985663525702e-05,
+      "loss": 1.0397,
+      "step": 17043
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.387933440115014e-05,
+      "loss": 1.0659,
+      "step": 17044
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.386881346763483e-05,
+      "loss": 0.9672,
+      "step": 17045
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.385829383491798e-05,
+      "loss": 0.8575,
+      "step": 17046
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3847775503206556e-05,
+      "loss": 0.9649,
+      "step": 17047
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3837258472707445e-05,
+      "loss": 0.8059,
+      "step": 17048
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.382674274362747e-05,
+      "loss": 0.8019,
+      "step": 17049
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3816228316173547e-05,
+      "loss": 1.028,
+      "step": 17050
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.380571519055242e-05,
+      "loss": 0.9034,
+      "step": 17051
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.379520336697096e-05,
+      "loss": 1.0547,
+      "step": 17052
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.378469284563589e-05,
+      "loss": 1.051,
+      "step": 17053
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.377418362675393e-05,
+      "loss": 0.8736,
+      "step": 17054
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3763675710531814e-05,
+      "loss": 0.8835,
+      "step": 17055
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.375316909717632e-05,
+      "loss": 0.8544,
+      "step": 17056
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.374266378689396e-05,
+      "loss": 0.9113,
+      "step": 17057
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.373215977989145e-05,
+      "loss": 0.788,
+      "step": 17058
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3721657076375424e-05,
+      "loss": 0.9247,
+      "step": 17059
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3711155676552456e-05,
+      "loss": 0.8155,
+      "step": 17060
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.370065558062906e-05,
+      "loss": 0.7968,
+      "step": 17061
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3690156788811844e-05,
+      "loss": 1.0237,
+      "step": 17062
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.367965930130727e-05,
+      "loss": 0.8118,
+      "step": 17063
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.366916311832185e-05,
+      "loss": 0.8837,
+      "step": 17064
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.365866824006204e-05,
+      "loss": 1.0035,
+      "step": 17065
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3648174666734235e-05,
+      "loss": 0.9887,
+      "step": 17066
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3637682398544915e-05,
+      "loss": 0.9229,
+      "step": 17067
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3627191435700376e-05,
+      "loss": 0.7473,
+      "step": 17068
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3616701778407065e-05,
+      "loss": 1.0684,
+      "step": 17069
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.360621342687127e-05,
+      "loss": 1.003,
+      "step": 17070
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.359572638129925e-05,
+      "loss": 0.8083,
+      "step": 17071
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3585240641897344e-05,
+      "loss": 0.8618,
+      "step": 17072
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.357475620887186e-05,
+      "loss": 0.9412,
+      "step": 17073
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.356427308242889e-05,
+      "loss": 0.935,
+      "step": 17074
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.355379126277471e-05,
+      "loss": 0.9511,
+      "step": 17075
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.354331075011553e-05,
+      "loss": 0.8417,
+      "step": 17076
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.353283154465746e-05,
+      "loss": 0.948,
+      "step": 17077
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3522353646606596e-05,
+      "loss": 0.9469,
+      "step": 17078
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.35118770561691e-05,
+      "loss": 0.9543,
+      "step": 17079
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3501401773551e-05,
+      "loss": 1.0175,
+      "step": 17080
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3490927798958395e-05,
+      "loss": 0.9448,
+      "step": 17081
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.348045513259728e-05,
+      "loss": 0.8664,
+      "step": 17082
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3469983774673596e-05,
+      "loss": 1.0405,
+      "step": 17083
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.345951372539339e-05,
+      "loss": 0.9831,
+      "step": 17084
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.34490449849626e-05,
+      "loss": 0.9595,
+      "step": 17085
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.343857755358714e-05,
+      "loss": 0.8675,
+      "step": 17086
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3428111431472853e-05,
+      "loss": 0.8699,
+      "step": 17087
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.341764661882569e-05,
+      "loss": 0.8683,
+      "step": 17088
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.340718311585145e-05,
+      "loss": 0.8756,
+      "step": 17089
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3396720922755944e-05,
+      "loss": 1.0209,
+      "step": 17090
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.338626003974494e-05,
+      "loss": 0.7761,
+      "step": 17091
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3375800467024246e-05,
+      "loss": 0.9516,
+      "step": 17092
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.336534220479961e-05,
+      "loss": 1.0942,
+      "step": 17093
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3354885253276736e-05,
+      "loss": 0.9191,
+      "step": 17094
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.334442961266127e-05,
+      "loss": 0.8163,
+      "step": 17095
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.333397528315893e-05,
+      "loss": 0.9061,
+      "step": 17096
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3323522264975285e-05,
+      "loss": 0.8347,
+      "step": 17097
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.331307055831604e-05,
+      "loss": 1.0711,
+      "step": 17098
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.330262016338671e-05,
+      "loss": 0.8037,
+      "step": 17099
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.329217108039285e-05,
+      "loss": 0.9233,
+      "step": 17100
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.328172330954001e-05,
+      "loss": 0.9768,
+      "step": 17101
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3271276851033784e-05,
+      "loss": 0.8372,
+      "step": 17102
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.32608317050795e-05,
+      "loss": 0.9071,
+      "step": 17103
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.325038787188267e-05,
+      "loss": 0.8125,
+      "step": 17104
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3239945351648794e-05,
+      "loss": 0.7814,
+      "step": 17105
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.32295041445832e-05,
+      "loss": 1.1009,
+      "step": 17106
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.321906425089127e-05,
+      "loss": 0.7716,
+      "step": 17107
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.32086256707784e-05,
+      "loss": 0.8456,
+      "step": 17108
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.319818840444987e-05,
+      "loss": 0.8753,
+      "step": 17109
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.318775245211102e-05,
+      "loss": 1.1152,
+      "step": 17110
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.317731781396711e-05,
+      "loss": 0.9367,
+      "step": 17111
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.316688449022336e-05,
+      "loss": 0.7145,
+      "step": 17112
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3156452481085056e-05,
+      "loss": 1.0955,
+      "step": 17113
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.314602178675731e-05,
+      "loss": 0.8662,
+      "step": 17114
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.313559240744538e-05,
+      "loss": 1.0521,
+      "step": 17115
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3125164343354374e-05,
+      "loss": 0.848,
+      "step": 17116
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.311473759468938e-05,
+      "loss": 0.9227,
+      "step": 17117
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.310431216165551e-05,
+      "loss": 0.9409,
+      "step": 17118
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.309388804445793e-05,
+      "loss": 0.8507,
+      "step": 17119
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.308346524330153e-05,
+      "loss": 0.7987,
+      "step": 17120
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3073043758391374e-05,
+      "loss": 1.1328,
+      "step": 17121
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3062623589932515e-05,
+      "loss": 0.8506,
+      "step": 17122
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.305220473812988e-05,
+      "loss": 0.7817,
+      "step": 17123
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3041787203188354e-05,
+      "loss": 0.7805,
+      "step": 17124
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3031370985312934e-05,
+      "loss": 1.0501,
+      "step": 17125
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.3020956084708445e-05,
+      "loss": 0.8529,
+      "step": 17126
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.30105425015798e-05,
+      "loss": 1.0024,
+      "step": 17127
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.300013023613181e-05,
+      "loss": 1.0098,
+      "step": 17128
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.298971928856924e-05,
+      "loss": 0.8858,
+      "step": 17129
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.297930965909691e-05,
+      "loss": 0.891,
+      "step": 17130
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.296890134791962e-05,
+      "loss": 1.0761,
+      "step": 17131
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2958494355242064e-05,
+      "loss": 0.7788,
+      "step": 17132
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.294808868126894e-05,
+      "loss": 0.9744,
+      "step": 17133
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.29376843262049e-05,
+      "loss": 0.9765,
+      "step": 17134
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2927281290254664e-05,
+      "loss": 0.8662,
+      "step": 17135
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.291687957362281e-05,
+      "loss": 0.951,
+      "step": 17136
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.290647917651395e-05,
+      "loss": 0.9111,
+      "step": 17137
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.289608009913264e-05,
+      "loss": 0.9465,
+      "step": 17138
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.28856823416835e-05,
+      "loss": 1.0161,
+      "step": 17139
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2875285904371e-05,
+      "loss": 0.7342,
+      "step": 17140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.28648907873996e-05,
+      "loss": 0.969,
+      "step": 17141
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.285449699097387e-05,
+      "loss": 0.8815,
+      "step": 17142
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2844104515298155e-05,
+      "loss": 1.0508,
+      "step": 17143
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.283371336057696e-05,
+      "loss": 0.7788,
+      "step": 17144
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.282332352701464e-05,
+      "loss": 0.8277,
+      "step": 17145
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.281293501481553e-05,
+      "loss": 0.7798,
+      "step": 17146
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2802547824183995e-05,
+      "loss": 0.9672,
+      "step": 17147
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2792161955324443e-05,
+      "loss": 0.7482,
+      "step": 17148
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2781777408441e-05,
+      "loss": 0.9923,
+      "step": 17149
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2771394183738014e-05,
+      "loss": 1.0662,
+      "step": 17150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.276101228141976e-05,
+      "loss": 0.7834,
+      "step": 17151
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2750631701690404e-05,
+      "loss": 0.9527,
+      "step": 17152
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2740252444754094e-05,
+      "loss": 0.8789,
+      "step": 17153
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2729874510815086e-05,
+      "loss": 1.0601,
+      "step": 17154
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.271949790007741e-05,
+      "loss": 0.8882,
+      "step": 17155
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2709122612745255e-05,
+      "loss": 0.8698,
+      "step": 17156
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.269874864902269e-05,
+      "loss": 0.8981,
+      "step": 17157
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2688376009113707e-05,
+      "loss": 0.8176,
+      "step": 17158
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.267800469322241e-05,
+      "loss": 0.9199,
+      "step": 17159
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.266763470155275e-05,
+      "loss": 0.8717,
+      "step": 17160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2657266034308755e-05,
+      "loss": 0.8595,
+      "step": 17161
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.264689869169435e-05,
+      "loss": 0.9165,
+      "step": 17162
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2636532673913425e-05,
+      "loss": 0.9822,
+      "step": 17163
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.262616798116991e-05,
+      "loss": 0.8557,
+      "step": 17164
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.261580461366776e-05,
+      "loss": 0.9154,
+      "step": 17165
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2605442571610656e-05,
+      "loss": 0.8514,
+      "step": 17166
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.259508185520252e-05,
+      "loss": 0.9604,
+      "step": 17167
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.258472246464717e-05,
+      "loss": 0.8183,
+      "step": 17168
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2574364400148325e-05,
+      "loss": 0.9547,
+      "step": 17169
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2564007661909724e-05,
+      "loss": 0.7718,
+      "step": 17170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2553652250135126e-05,
+      "loss": 0.9087,
+      "step": 17171
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2543298165028166e-05,
+      "loss": 0.7011,
+      "step": 17172
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.253294540679257e-05,
+      "loss": 0.9296,
+      "step": 17173
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2522593975631956e-05,
+      "loss": 0.9386,
+      "step": 17174
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.251224387174988e-05,
+      "loss": 0.7693,
+      "step": 17175
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.250189509535001e-05,
+      "loss": 0.8526,
+      "step": 17176
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.249154764663584e-05,
+      "loss": 0.9982,
+      "step": 17177
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.248120152581097e-05,
+      "loss": 0.9428,
+      "step": 17178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.247085673307887e-05,
+      "loss": 1.0575,
+      "step": 17179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.246051326864299e-05,
+      "loss": 0.9915,
+      "step": 17180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.245017113270685e-05,
+      "loss": 0.879,
+      "step": 17181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.243983032547385e-05,
+      "loss": 0.8499,
+      "step": 17182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.242949084714737e-05,
+      "loss": 1.0865,
+      "step": 17183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.24191526979308e-05,
+      "loss": 0.9466,
+      "step": 17184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2408815878027544e-05,
+      "loss": 0.8576,
+      "step": 17185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.239848038764087e-05,
+      "loss": 0.9778,
+      "step": 17186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.238814622697407e-05,
+      "loss": 0.8749,
+      "step": 17187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.237781339623046e-05,
+      "loss": 0.9717,
+      "step": 17188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.236748189561324e-05,
+      "loss": 0.8711,
+      "step": 17189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.235715172532567e-05,
+      "loss": 1.1334,
+      "step": 17190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.234682288557095e-05,
+      "loss": 0.9471,
+      "step": 17191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.233649537655219e-05,
+      "loss": 1.0351,
+      "step": 17192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2326169198472556e-05,
+      "loss": 1.1122,
+      "step": 17193
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2315844351535255e-05,
+      "loss": 0.9218,
+      "step": 17194
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.230552083594322e-05,
+      "loss": 0.836,
+      "step": 17195
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.229519865189959e-05,
+      "loss": 0.918,
+      "step": 17196
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.228487779960745e-05,
+      "loss": 0.8776,
+      "step": 17197
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2274558279269754e-05,
+      "loss": 0.9698,
+      "step": 17198
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.22642400910895e-05,
+      "loss": 0.8424,
+      "step": 17199
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.225392323526961e-05,
+      "loss": 1.0409,
+      "step": 17200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2243607712013046e-05,
+      "loss": 0.9155,
+      "step": 17201
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.2233293521522756e-05,
+      "loss": 0.8417,
+      "step": 17202
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.222298066400158e-05,
+      "loss": 1.0424,
+      "step": 17203
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 3.221266913965234e-05,
+      "loss": 0.92,
+      "step": 17204
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.220235894867794e-05,
+      "loss": 0.8628,
+      "step": 17205
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.219205009128109e-05,
+      "loss": 0.7963,
+      "step": 17206
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.218174256766464e-05,
+      "loss": 1.0554,
+      "step": 17207
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2171436378031326e-05,
+      "loss": 0.8672,
+      "step": 17208
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.216113152258382e-05,
+      "loss": 0.9301,
+      "step": 17209
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.215082800152485e-05,
+      "loss": 0.9676,
+      "step": 17210
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.214052581505715e-05,
+      "loss": 0.9431,
+      "step": 17211
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.213022496338325e-05,
+      "loss": 1.0679,
+      "step": 17212
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.211992544670582e-05,
+      "loss": 1.0711,
+      "step": 17213
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2109627265227485e-05,
+      "loss": 0.7857,
+      "step": 17214
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2099330419150766e-05,
+      "loss": 1.2097,
+      "step": 17215
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2089034908678186e-05,
+      "loss": 0.8312,
+      "step": 17216
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.207874073401232e-05,
+      "loss": 0.9444,
+      "step": 17217
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.206844789535558e-05,
+      "loss": 0.961,
+      "step": 17218
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.205815639291051e-05,
+      "loss": 1.0483,
+      "step": 17219
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2047866226879485e-05,
+      "loss": 0.921,
+      "step": 17220
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.203757739746489e-05,
+      "loss": 1.0111,
+      "step": 17221
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.2027289904869184e-05,
+      "loss": 1.007,
+      "step": 17222
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.201700374929464e-05,
+      "loss": 1.0183,
+      "step": 17223
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.200671893094366e-05,
+      "loss": 0.944,
+      "step": 17224
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1996435450018515e-05,
+      "loss": 0.7971,
+      "step": 17225
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.198615330672143e-05,
+      "loss": 1.0411,
+      "step": 17226
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.197587250125474e-05,
+      "loss": 1.1217,
+      "step": 17227
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.196559303382064e-05,
+      "loss": 0.8848,
+      "step": 17228
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.195531490462128e-05,
+      "loss": 0.8188,
+      "step": 17229
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1945038113858864e-05,
+      "loss": 0.9869,
+      "step": 17230
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.193476266173556e-05,
+      "loss": 0.7502,
+      "step": 17231
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.192448854845348e-05,
+      "loss": 1.017,
+      "step": 17232
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1914215774214664e-05,
+      "loss": 0.843,
+      "step": 17233
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.190394433922125e-05,
+      "loss": 0.9082,
+      "step": 17234
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1893674243675195e-05,
+      "loss": 0.9046,
+      "step": 17235
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1883405487778594e-05,
+      "loss": 0.9677,
+      "step": 17236
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.18731380717334e-05,
+      "loss": 0.9028,
+      "step": 17237
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.186287199574154e-05,
+      "loss": 0.8239,
+      "step": 17238
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.185260726000496e-05,
+      "loss": 0.9812,
+      "step": 17239
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1842343864725666e-05,
+      "loss": 0.8981,
+      "step": 17240
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1832081810105385e-05,
+      "loss": 1.1242,
+      "step": 17241
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.182182109634607e-05,
+      "loss": 0.856,
+      "step": 17242
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.18115617236495e-05,
+      "loss": 0.8954,
+      "step": 17243
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.180130369221751e-05,
+      "loss": 0.9633,
+      "step": 17244
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1791047002251884e-05,
+      "loss": 1.0262,
+      "step": 17245
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.17807916539543e-05,
+      "loss": 0.974,
+      "step": 17246
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1770537647526545e-05,
+      "loss": 1.0949,
+      "step": 17247
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1760284983170316e-05,
+      "loss": 0.8904,
+      "step": 17248
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.175003366108728e-05,
+      "loss": 0.8565,
+      "step": 17249
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.173978368147903e-05,
+      "loss": 0.9271,
+      "step": 17250
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.172953504454724e-05,
+      "loss": 0.8839,
+      "step": 17251
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.171928775049346e-05,
+      "loss": 0.9676,
+      "step": 17252
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1709041799519315e-05,
+      "loss": 1.0068,
+      "step": 17253
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.169879719182629e-05,
+      "loss": 1.1164,
+      "step": 17254
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.168855392761587e-05,
+      "loss": 1.0372,
+      "step": 17255
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.167831200708958e-05,
+      "loss": 0.9413,
+      "step": 17256
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1668071430448954e-05,
+      "loss": 1.0164,
+      "step": 17257
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.165783219789526e-05,
+      "loss": 1.0247,
+      "step": 17258
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1647594309630016e-05,
+      "loss": 0.9362,
+      "step": 17259
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.163735776585458e-05,
+      "loss": 1.0307,
+      "step": 17260
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1627122566770305e-05,
+      "loss": 0.9807,
+      "step": 17261
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.161688871257849e-05,
+      "loss": 0.7438,
+      "step": 17262
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.160665620348048e-05,
+      "loss": 0.7298,
+      "step": 17263
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.159642503967748e-05,
+      "loss": 1.0842,
+      "step": 17264
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1586195221370826e-05,
+      "loss": 1.0629,
+      "step": 17265
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1575966748761687e-05,
+      "loss": 0.9528,
+      "step": 17266
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.156573962205123e-05,
+      "loss": 0.7708,
+      "step": 17267
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.155551384144069e-05,
+      "loss": 0.707,
+      "step": 17268
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.154528940713113e-05,
+      "loss": 0.9882,
+      "step": 17269
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.153506631932373e-05,
+      "loss": 0.9865,
+      "step": 17270
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.152484457821958e-05,
+      "loss": 0.7929,
+      "step": 17271
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.151462418401967e-05,
+      "loss": 0.9056,
+      "step": 17272
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.150440513692511e-05,
+      "loss": 0.7675,
+      "step": 17273
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.149418743713688e-05,
+      "loss": 0.9323,
+      "step": 17274
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1483971084855936e-05,
+      "loss": 0.9281,
+      "step": 17275
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.147375608028324e-05,
+      "loss": 0.8158,
+      "step": 17276
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1463542423619794e-05,
+      "loss": 0.8906,
+      "step": 17277
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.145333011506644e-05,
+      "loss": 0.9276,
+      "step": 17278
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1443119154824005e-05,
+      "loss": 0.9241,
+      "step": 17279
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.143290954309345e-05,
+      "loss": 0.8132,
+      "step": 17280
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.142270128007549e-05,
+      "loss": 1.0125,
+      "step": 17281
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1412494365971015e-05,
+      "loss": 0.9513,
+      "step": 17282
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.140228880098074e-05,
+      "loss": 0.9657,
+      "step": 17283
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1392084585305394e-05,
+      "loss": 0.8433,
+      "step": 17284
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1381881719145724e-05,
+      "loss": 0.8931,
+      "step": 17285
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.137168020270247e-05,
+      "loss": 0.8266,
+      "step": 17286
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1361480036176174e-05,
+      "loss": 0.9128,
+      "step": 17287
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.135128121976757e-05,
+      "loss": 0.8661,
+      "step": 17288
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1341083753677215e-05,
+      "loss": 1.024,
+      "step": 17289
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1330887638105745e-05,
+      "loss": 0.9664,
+      "step": 17290
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.132069287325368e-05,
+      "loss": 0.7358,
+      "step": 17291
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.131049945932153e-05,
+      "loss": 0.9976,
+      "step": 17292
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.130030739650983e-05,
+      "loss": 0.8383,
+      "step": 17293
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.129011668501909e-05,
+      "loss": 0.9567,
+      "step": 17294
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.127992732504972e-05,
+      "loss": 0.9758,
+      "step": 17295
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1269739316802116e-05,
+      "loss": 0.8606,
+      "step": 17296
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.125955266047674e-05,
+      "loss": 0.9504,
+      "step": 17297
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.124936735627392e-05,
+      "loss": 0.9048,
+      "step": 17298
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.123918340439403e-05,
+      "loss": 0.9869,
+      "step": 17299
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.122900080503738e-05,
+      "loss": 0.7828,
+      "step": 17300
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.121881955840421e-05,
+      "loss": 1.0573,
+      "step": 17301
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.120863966469484e-05,
+      "loss": 0.8736,
+      "step": 17302
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1198461124109554e-05,
+      "loss": 1.2072,
+      "step": 17303
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.118828393684845e-05,
+      "loss": 0.9188,
+      "step": 17304
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.117810810311176e-05,
+      "loss": 0.8396,
+      "step": 17305
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.116793362309969e-05,
+      "loss": 0.9456,
+      "step": 17306
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1157760497012336e-05,
+      "loss": 0.9295,
+      "step": 17307
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1147588725049795e-05,
+      "loss": 0.8952,
+      "step": 17308
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1137418307412116e-05,
+      "loss": 1.0834,
+      "step": 17309
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1127249244299396e-05,
+      "loss": 0.9811,
+      "step": 17310
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.111708153591166e-05,
+      "loss": 0.7516,
+      "step": 17311
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.110691518244892e-05,
+      "loss": 1.006,
+      "step": 17312
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.109675018411108e-05,
+      "loss": 0.8934,
+      "step": 17313
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1086586541098175e-05,
+      "loss": 0.8742,
+      "step": 17314
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1076424253610025e-05,
+      "loss": 0.9465,
+      "step": 17315
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1066263321846624e-05,
+      "loss": 0.6916,
+      "step": 17316
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.105610374600778e-05,
+      "loss": 1.0678,
+      "step": 17317
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.104594552629331e-05,
+      "loss": 0.9359,
+      "step": 17318
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.103578866290309e-05,
+      "loss": 0.9041,
+      "step": 17319
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.102563315603686e-05,
+      "loss": 0.9982,
+      "step": 17320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.1015479005894364e-05,
+      "loss": 0.8045,
+      "step": 17321
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.100532621267537e-05,
+      "loss": 0.9442,
+      "step": 17322
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.099517477657959e-05,
+      "loss": 0.9804,
+      "step": 17323
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.098502469780669e-05,
+      "loss": 0.9893,
+      "step": 17324
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0974875976556284e-05,
+      "loss": 0.9232,
+      "step": 17325
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.096472861302806e-05,
+      "loss": 1.0788,
+      "step": 17326
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.095458260742156e-05,
+      "loss": 0.8649,
+      "step": 17327
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0944437959936404e-05,
+      "loss": 0.949,
+      "step": 17328
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.093429467077212e-05,
+      "loss": 0.8866,
+      "step": 17329
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.092415274012819e-05,
+      "loss": 1.0143,
+      "step": 17330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0914012168204164e-05,
+      "loss": 1.0289,
+      "step": 17331
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.090387295519947e-05,
+      "loss": 0.8614,
+      "step": 17332
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.089373510131354e-05,
+      "loss": 0.9973,
+      "step": 17333
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0883598606745826e-05,
+      "loss": 0.8946,
+      "step": 17334
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0873463471695654e-05,
+      "loss": 0.884,
+      "step": 17335
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.086332969636244e-05,
+      "loss": 0.8831,
+      "step": 17336
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0853197280945504e-05,
+      "loss": 0.9756,
+      "step": 17337
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.084306622564409e-05,
+      "loss": 0.8965,
+      "step": 17338
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.083293653065753e-05,
+      "loss": 0.9638,
+      "step": 17339
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0822808196185097e-05,
+      "loss": 0.9005,
+      "step": 17340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.081268122242598e-05,
+      "loss": 0.826,
+      "step": 17341
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0802555609579355e-05,
+      "loss": 0.8432,
+      "step": 17342
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.079243135784444e-05,
+      "loss": 1.0683,
+      "step": 17343
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0782308467420326e-05,
+      "loss": 1.2178,
+      "step": 17344
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.077218693850621e-05,
+      "loss": 0.8537,
+      "step": 17345
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0762066771301116e-05,
+      "loss": 0.8856,
+      "step": 17346
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.075194796600409e-05,
+      "loss": 0.9414,
+      "step": 17347
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.074183052281422e-05,
+      "loss": 0.9364,
+      "step": 17348
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.073171444193055e-05,
+      "loss": 0.8309,
+      "step": 17349
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0721599723551954e-05,
+      "loss": 0.899,
+      "step": 17350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.071148636787746e-05,
+      "loss": 0.984,
+      "step": 17351
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.070137437510597e-05,
+      "loss": 0.7947,
+      "step": 17352
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.069126374543643e-05,
+      "loss": 0.9432,
+      "step": 17353
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.068115447906769e-05,
+      "loss": 0.926,
+      "step": 17354
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.067104657619855e-05,
+      "loss": 0.9683,
+      "step": 17355
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0660940037027875e-05,
+      "loss": 1.0016,
+      "step": 17356
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.065083486175451e-05,
+      "loss": 0.8446,
+      "step": 17357
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.064073105057716e-05,
+      "loss": 0.9444,
+      "step": 17358
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.063062860369456e-05,
+      "loss": 0.8918,
+      "step": 17359
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0620527521305475e-05,
+      "loss": 0.9975,
+      "step": 17360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.061042780360853e-05,
+      "loss": 0.7689,
+      "step": 17361
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0600329450802466e-05,
+      "loss": 0.9936,
+      "step": 17362
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0590232463085867e-05,
+      "loss": 1.0388,
+      "step": 17363
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0580136840657304e-05,
+      "loss": 1.0498,
+      "step": 17364
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.057004258371541e-05,
+      "loss": 0.9129,
+      "step": 17365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.055994969245879e-05,
+      "loss": 1.0549,
+      "step": 17366
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.054985816708584e-05,
+      "loss": 0.8927,
+      "step": 17367
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.053976800779513e-05,
+      "loss": 0.898,
+      "step": 17368
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0529679214785176e-05,
+      "loss": 0.8102,
+      "step": 17369
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.051959178825438e-05,
+      "loss": 0.9332,
+      "step": 17370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.050950572840112e-05,
+      "loss": 0.93,
+      "step": 17371
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.049942103542386e-05,
+      "loss": 0.9213,
+      "step": 17372
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0489337709520917e-05,
+      "loss": 0.9929,
+      "step": 17373
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0479255750890667e-05,
+      "loss": 0.8653,
+      "step": 17374
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.046917515973141e-05,
+      "loss": 1.0873,
+      "step": 17375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.04590959362414e-05,
+      "loss": 0.8406,
+      "step": 17376
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.044901808061894e-05,
+      "loss": 0.8755,
+      "step": 17377
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0438941593062244e-05,
+      "loss": 0.9555,
+      "step": 17378
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.042886647376948e-05,
+      "loss": 1.2982,
+      "step": 17379
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0418792722938893e-05,
+      "loss": 0.8321,
+      "step": 17380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0408720340768572e-05,
+      "loss": 0.9536,
+      "step": 17381
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0398649327456684e-05,
+      "loss": 0.7583,
+      "step": 17382
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0388579683201324e-05,
+      "loss": 0.9521,
+      "step": 17383
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0378511408200506e-05,
+      "loss": 0.9739,
+      "step": 17384
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0368444502652315e-05,
+      "loss": 0.8806,
+      "step": 17385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0358378966754797e-05,
+      "loss": 1.0102,
+      "step": 17386
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0348314800705914e-05,
+      "loss": 0.9086,
+      "step": 17387
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0338252004703583e-05,
+      "loss": 0.8515,
+      "step": 17388
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0328190578945815e-05,
+      "loss": 0.9921,
+      "step": 17389
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0318130523630437e-05,
+      "loss": 0.8703,
+      "step": 17390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.030807183895542e-05,
+      "loss": 1.1324,
+      "step": 17391
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.029801452511857e-05,
+      "loss": 0.8381,
+      "step": 17392
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0287958582317676e-05,
+      "loss": 0.9532,
+      "step": 17393
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.027790401075059e-05,
+      "loss": 0.9206,
+      "step": 17394
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0267850810615127e-05,
+      "loss": 0.9088,
+      "step": 17395
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.025779898210892e-05,
+      "loss": 0.9199,
+      "step": 17396
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0247748525429787e-05,
+      "loss": 0.8263,
+      "step": 17397
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0237699440775337e-05,
+      "loss": 0.9369,
+      "step": 17398
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.022765172834332e-05,
+      "loss": 0.8438,
+      "step": 17399
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.021760538833133e-05,
+      "loss": 0.9656,
+      "step": 17400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0207560420936942e-05,
+      "loss": 0.9146,
+      "step": 17401
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0197516826357785e-05,
+      "loss": 0.8323,
+      "step": 17402
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0187474604791456e-05,
+      "loss": 0.7985,
+      "step": 17403
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0177433756435425e-05,
+      "loss": 0.8433,
+      "step": 17404
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0167394281487184e-05,
+      "loss": 1.0956,
+      "step": 17405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0157356180144257e-05,
+      "loss": 0.6897,
+      "step": 17406
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.014731945260405e-05,
+      "loss": 0.9981,
+      "step": 17407
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0137284099064044e-05,
+      "loss": 1.0066,
+      "step": 17408
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0127250119721594e-05,
+      "loss": 0.8436,
+      "step": 17409
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.011721751477403e-05,
+      "loss": 0.8262,
+      "step": 17410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0107186284418743e-05,
+      "loss": 0.8883,
+      "step": 17411
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0097156428853103e-05,
+      "loss": 0.9408,
+      "step": 17412
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.008712794827426e-05,
+      "loss": 0.9986,
+      "step": 17413
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0077100842879545e-05,
+      "loss": 0.9791,
+      "step": 17414
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.006707511286624e-05,
+      "loss": 0.9838,
+      "step": 17415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0057050758431482e-05,
+      "loss": 0.9138,
+      "step": 17416
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0047027779772486e-05,
+      "loss": 0.9342,
+      "step": 17417
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.0037006177086346e-05,
+      "loss": 0.8099,
+      "step": 17418
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.002698595057023e-05,
+      "loss": 0.69,
+      "step": 17419
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.001696710042128e-05,
+      "loss": 0.9839,
+      "step": 17420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 3.00069496268365e-05,
+      "loss": 0.8532,
+      "step": 17421
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9996933530012938e-05,
+      "loss": 0.8855,
+      "step": 17422
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9986918810147646e-05,
+      "loss": 0.9353,
+      "step": 17423
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.997690546743761e-05,
+      "loss": 0.7904,
+      "step": 17424
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.996689350207973e-05,
+      "loss": 0.7972,
+      "step": 17425
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9956882914271024e-05,
+      "loss": 0.9432,
+      "step": 17426
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9946873704208334e-05,
+      "loss": 0.8626,
+      "step": 17427
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9936865872088594e-05,
+      "loss": 1.1085,
+      "step": 17428
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9926859418108644e-05,
+      "loss": 0.7829,
+      "step": 17429
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9916854342465262e-05,
+      "loss": 0.8381,
+      "step": 17430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9906850645355276e-05,
+      "loss": 0.7949,
+      "step": 17431
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9896848326975513e-05,
+      "loss": 0.7781,
+      "step": 17432
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9886847387522678e-05,
+      "loss": 1.0768,
+      "step": 17433
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.9876847827193444e-05,
+      "loss": 0.861,
+      "step": 17434
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2.986684964618457e-05,
+      "loss": 0.9522,
+      "step": 17435
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9856852844692674e-05,
+      "loss": 1.0861,
+      "step": 17436
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9846857422914433e-05,
+      "loss": 0.8805,
+      "step": 17437
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9836863381046443e-05,
+      "loss": 0.7331,
+      "step": 17438
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.982687071928524e-05,
+      "loss": 0.9898,
+      "step": 17439
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.981687943782745e-05,
+      "loss": 0.9152,
+      "step": 17440
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9806889536869564e-05,
+      "loss": 0.9809,
+      "step": 17441
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.979690101660807e-05,
+      "loss": 0.9792,
+      "step": 17442
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9786913877239487e-05,
+      "loss": 0.9808,
+      "step": 17443
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.97769281189602e-05,
+      "loss": 1.1123,
+      "step": 17444
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.97669437419667e-05,
+      "loss": 1.0235,
+      "step": 17445
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.975696074645534e-05,
+      "loss": 0.824,
+      "step": 17446
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9746979132622453e-05,
+      "loss": 0.9171,
+      "step": 17447
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9736998900664415e-05,
+      "loss": 0.7919,
+      "step": 17448
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9727020050777565e-05,
+      "loss": 1.0991,
+      "step": 17449
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.971704258315816e-05,
+      "loss": 1.0143,
+      "step": 17450
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9707066498002412e-05,
+      "loss": 0.9672,
+      "step": 17451
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9697091795506638e-05,
+      "loss": 0.9117,
+      "step": 17452
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9687118475866947e-05,
+      "loss": 1.0827,
+      "step": 17453
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.967714653927959e-05,
+      "loss": 0.8357,
+      "step": 17454
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9667175985940687e-05,
+      "loss": 0.916,
+      "step": 17455
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9657206816046335e-05,
+      "loss": 1.0331,
+      "step": 17456
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.964723902979264e-05,
+      "loss": 0.8831,
+      "step": 17457
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9637272627375734e-05,
+      "loss": 0.9851,
+      "step": 17458
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9627307608991538e-05,
+      "loss": 0.9825,
+      "step": 17459
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.961734397483612e-05,
+      "loss": 0.9465,
+      "step": 17460
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.960738172510551e-05,
+      "loss": 0.7757,
+      "step": 17461
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9597420859995618e-05,
+      "loss": 1.1243,
+      "step": 17462
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9587461379702374e-05,
+      "loss": 0.8795,
+      "step": 17463
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.957750328442166e-05,
+      "loss": 0.9229,
+      "step": 17464
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.956754657434937e-05,
+      "loss": 0.9494,
+      "step": 17465
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9557591249681405e-05,
+      "loss": 1.1569,
+      "step": 17466
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.954763731061353e-05,
+      "loss": 0.9651,
+      "step": 17467
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.953768475734151e-05,
+      "loss": 0.9026,
+      "step": 17468
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.952773359006119e-05,
+      "loss": 0.8772,
+      "step": 17469
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.951778380896827e-05,
+      "loss": 0.9203,
+      "step": 17470
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9507835414258432e-05,
+      "loss": 0.813,
+      "step": 17471
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9497888406127417e-05,
+      "loss": 0.8578,
+      "step": 17472
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.948794278477083e-05,
+      "loss": 1.0542,
+      "step": 17473
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9477998550384356e-05,
+      "loss": 1.0328,
+      "step": 17474
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9468055703163568e-05,
+      "loss": 0.961,
+      "step": 17475
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9458114243304004e-05,
+      "loss": 0.9786,
+      "step": 17476
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9448174171001254e-05,
+      "loss": 0.9304,
+      "step": 17477
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.943823548645087e-05,
+      "loss": 1.0254,
+      "step": 17478
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9428298189848302e-05,
+      "loss": 0.9919,
+      "step": 17479
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9418362281389e-05,
+      "loss": 0.8455,
+      "step": 17480
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9408427761268464e-05,
+      "loss": 1.0431,
+      "step": 17481
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9398494629682026e-05,
+      "loss": 1.162,
+      "step": 17482
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9388562886825188e-05,
+      "loss": 0.9729,
+      "step": 17483
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.937863253289317e-05,
+      "loss": 0.9551,
+      "step": 17484
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9368703568081356e-05,
+      "loss": 1.0687,
+      "step": 17485
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9358775992585096e-05,
+      "loss": 0.88,
+      "step": 17486
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.934884980659962e-05,
+      "loss": 0.7996,
+      "step": 17487
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.933892501032015e-05,
+      "loss": 0.9104,
+      "step": 17488
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9329001603941985e-05,
+      "loss": 0.9985,
+      "step": 17489
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9319079587660225e-05,
+      "loss": 0.9089,
+      "step": 17490
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9309158961670124e-05,
+      "loss": 0.9238,
+      "step": 17491
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9299239726166772e-05,
+      "loss": 1.0608,
+      "step": 17492
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9289321881345254e-05,
+      "loss": 1.0174,
+      "step": 17493
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9279405427400696e-05,
+      "loss": 0.7815,
+      "step": 17494
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.926949036452816e-05,
+      "loss": 0.9263,
+      "step": 17495
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9259576692922673e-05,
+      "loss": 0.9655,
+      "step": 17496
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.92496644127792e-05,
+      "loss": 0.9329,
+      "step": 17497
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9239753524292756e-05,
+      "loss": 0.8326,
+      "step": 17498
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9229844027658248e-05,
+      "loss": 0.8676,
+      "step": 17499
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.921993592307064e-05,
+      "loss": 0.8042,
+      "step": 17500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.921002921072481e-05,
+      "loss": 0.9395,
+      "step": 17501
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.920012389081559e-05,
+      "loss": 0.8878,
+      "step": 17502
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9190219963537836e-05,
+      "loss": 1.0178,
+      "step": 17503
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9180317429086422e-05,
+      "loss": 0.845,
+      "step": 17504
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.917041628765601e-05,
+      "loss": 0.9319,
+      "step": 17505
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9160516539441464e-05,
+      "loss": 0.8713,
+      "step": 17506
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9150618184637425e-05,
+      "loss": 0.8748,
+      "step": 17507
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.914072122343866e-05,
+      "loss": 1.0466,
+      "step": 17508
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.913082565603983e-05,
+      "loss": 0.9207,
+      "step": 17509
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9120931482635517e-05,
+      "loss": 0.8973,
+      "step": 17510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9111038703420402e-05,
+      "loss": 0.827,
+      "step": 17511
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9101147318589095e-05,
+      "loss": 0.9136,
+      "step": 17512
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.909125732833612e-05,
+      "loss": 1.0282,
+      "step": 17513
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9081368732855994e-05,
+      "loss": 0.9642,
+      "step": 17514
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.907148153234328e-05,
+      "loss": 0.685,
+      "step": 17515
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9061595726992396e-05,
+      "loss": 1.0357,
+      "step": 17516
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9051711316997864e-05,
+      "loss": 0.7067,
+      "step": 17517
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9041828302554074e-05,
+      "loss": 0.955,
+      "step": 17518
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9031946683855406e-05,
+      "loss": 1.1738,
+      "step": 17519
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9022066461096264e-05,
+      "loss": 0.8438,
+      "step": 17520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.901218763447099e-05,
+      "loss": 1.0475,
+      "step": 17521
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.9002310204173856e-05,
+      "loss": 0.9998,
+      "step": 17522
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8992434170399187e-05,
+      "loss": 0.7424,
+      "step": 17523
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.898255953334127e-05,
+      "loss": 0.7314,
+      "step": 17524
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8972686293194308e-05,
+      "loss": 0.8505,
+      "step": 17525
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.896281445015251e-05,
+      "loss": 1.059,
+      "step": 17526
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8952944004410022e-05,
+      "loss": 0.8873,
+      "step": 17527
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8943074956161033e-05,
+      "loss": 0.983,
+      "step": 17528
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8933207305599718e-05,
+      "loss": 0.7271,
+      "step": 17529
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.892334105292006e-05,
+      "loss": 0.855,
+      "step": 17530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.891347619831618e-05,
+      "loss": 1.0104,
+      "step": 17531
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8903612741982154e-05,
+      "loss": 0.9666,
+      "step": 17532
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8893750684111975e-05,
+      "loss": 0.8575,
+      "step": 17533
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8883890024899586e-05,
+      "loss": 0.7799,
+      "step": 17534
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8874030764539006e-05,
+      "loss": 0.9166,
+      "step": 17535
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8864172903224117e-05,
+      "loss": 0.935,
+      "step": 17536
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8854316441148877e-05,
+      "loss": 0.7514,
+      "step": 17537
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8844461378507126e-05,
+      "loss": 0.9471,
+      "step": 17538
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.883460771549269e-05,
+      "loss": 1.0837,
+      "step": 17539
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.882475545229941e-05,
+      "loss": 0.7819,
+      "step": 17540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8814904589121137e-05,
+      "loss": 0.9106,
+      "step": 17541
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.880505512615158e-05,
+      "loss": 0.7988,
+      "step": 17542
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.879520706358446e-05,
+      "loss": 0.8742,
+      "step": 17543
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8785360401613538e-05,
+      "loss": 1.0613,
+      "step": 17544
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8775515140432452e-05,
+      "loss": 1.0255,
+      "step": 17545
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8765671280234907e-05,
+      "loss": 0.7244,
+      "step": 17546
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8755828821214502e-05,
+      "loss": 0.7875,
+      "step": 17547
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8745987763564806e-05,
+      "loss": 0.9629,
+      "step": 17548
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8736148107479467e-05,
+      "loss": 0.8427,
+      "step": 17549
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.872630985315199e-05,
+      "loss": 1.0302,
+      "step": 17550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8716473000775856e-05,
+      "loss": 0.9227,
+      "step": 17551
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8706637550544634e-05,
+      "loss": 1.1018,
+      "step": 17552
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8696803502651714e-05,
+      "loss": 0.9212,
+      "step": 17553
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8686970857290584e-05,
+      "loss": 0.921,
+      "step": 17554
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8677139614654636e-05,
+      "loss": 1.052,
+      "step": 17555
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8667309774937223e-05,
+      "loss": 0.9105,
+      "step": 17556
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8657481338331715e-05,
+      "loss": 1.0016,
+      "step": 17557
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.864765430503147e-05,
+      "loss": 0.8797,
+      "step": 17558
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.863782867522976e-05,
+      "loss": 0.8937,
+      "step": 17559
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.862800444911983e-05,
+      "loss": 0.8678,
+      "step": 17560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8618181626894967e-05,
+      "loss": 0.9625,
+      "step": 17561
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.860836020874833e-05,
+      "loss": 1.0574,
+      "step": 17562
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.859854019487318e-05,
+      "loss": 0.9499,
+      "step": 17563
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.858872158546263e-05,
+      "loss": 1.0569,
+      "step": 17564
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8578904380709793e-05,
+      "loss": 0.9446,
+      "step": 17565
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8569088580807823e-05,
+      "loss": 1.0026,
+      "step": 17566
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8559274185949768e-05,
+      "loss": 0.9575,
+      "step": 17567
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8549461196328663e-05,
+      "loss": 0.9119,
+      "step": 17568
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8539649612137532e-05,
+      "loss": 0.8583,
+      "step": 17569
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.852983943356943e-05,
+      "loss": 0.7659,
+      "step": 17570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8520030660817263e-05,
+      "loss": 0.914,
+      "step": 17571
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.851022329407398e-05,
+      "loss": 0.74,
+      "step": 17572
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.850041733353247e-05,
+      "loss": 1.0038,
+      "step": 17573
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.849061277938564e-05,
+      "loss": 0.8963,
+      "step": 17574
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.848080963182641e-05,
+      "loss": 1.0294,
+      "step": 17575
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8471007891047474e-05,
+      "loss": 1.0939,
+      "step": 17576
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.846120755724171e-05,
+      "loss": 0.8048,
+      "step": 17577
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8451408630601906e-05,
+      "loss": 0.9604,
+      "step": 17578
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8441611111320776e-05,
+      "loss": 0.8519,
+      "step": 17579
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.843181499959102e-05,
+      "loss": 0.981,
+      "step": 17580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.842202029560538e-05,
+      "loss": 0.8899,
+      "step": 17581
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8412226999556445e-05,
+      "loss": 0.9855,
+      "step": 17582
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8402435111636937e-05,
+      "loss": 1.0079,
+      "step": 17583
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.839264463203941e-05,
+      "loss": 0.9055,
+      "step": 17584
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.838285556095642e-05,
+      "loss": 0.7797,
+      "step": 17585
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.837306789858054e-05,
+      "loss": 0.7957,
+      "step": 17586
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8363281645104346e-05,
+      "loss": 0.8155,
+      "step": 17587
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.835349680072027e-05,
+      "loss": 0.7674,
+      "step": 17588
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8343713365620772e-05,
+      "loss": 0.9666,
+      "step": 17589
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8333931339998356e-05,
+      "loss": 0.9411,
+      "step": 17590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.832415072404535e-05,
+      "loss": 0.8348,
+      "step": 17591
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8314371517954252e-05,
+      "loss": 0.9416,
+      "step": 17592
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8304593721917285e-05,
+      "loss": 0.8364,
+      "step": 17593
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8294817336126843e-05,
+      "loss": 1.1432,
+      "step": 17594
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.828504236077526e-05,
+      "loss": 0.9622,
+      "step": 17595
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8275268796054777e-05,
+      "loss": 0.8857,
+      "step": 17596
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8265496642157597e-05,
+      "loss": 0.9282,
+      "step": 17597
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.825572589927602e-05,
+      "loss": 0.8562,
+      "step": 17598
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8245956567602172e-05,
+      "loss": 0.9362,
+      "step": 17599
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8236188647328253e-05,
+      "loss": 0.9728,
+      "step": 17600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8226422138646403e-05,
+      "loss": 1.0754,
+      "step": 17601
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8216657041748675e-05,
+      "loss": 0.9847,
+      "step": 17602
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8206893356827168e-05,
+      "loss": 1.0254,
+      "step": 17603
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.819713108407399e-05,
+      "loss": 0.7965,
+      "step": 17604
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8187370223681132e-05,
+      "loss": 0.9227,
+      "step": 17605
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8177610775840546e-05,
+      "loss": 1.0694,
+      "step": 17606
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8167852740744262e-05,
+      "loss": 1.0144,
+      "step": 17607
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8158096118584165e-05,
+      "loss": 0.8238,
+      "step": 17608
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8148340909552228e-05,
+      "loss": 0.7246,
+      "step": 17609
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.81385871138403e-05,
+      "loss": 0.6519,
+      "step": 17610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.812883473164021e-05,
+      "loss": 0.8644,
+      "step": 17611
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.811908376314385e-05,
+      "loss": 0.8778,
+      "step": 17612
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8109334208542993e-05,
+      "loss": 0.7395,
+      "step": 17613
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8099586068029383e-05,
+      "loss": 1.0419,
+      "step": 17614
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8089839341794823e-05,
+      "loss": 0.8605,
+      "step": 17615
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8080094030030956e-05,
+      "loss": 0.8388,
+      "step": 17616
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8070350132929557e-05,
+      "loss": 0.9816,
+      "step": 17617
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8060607650682246e-05,
+      "loss": 0.822,
+      "step": 17618
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8050866583480627e-05,
+      "loss": 0.9534,
+      "step": 17619
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8041126931516338e-05,
+      "loss": 0.8954,
+      "step": 17620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.803138869498102e-05,
+      "loss": 0.8801,
+      "step": 17621
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.802165187406609e-05,
+      "loss": 0.8656,
+      "step": 17622
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.801191646896315e-05,
+      "loss": 1.0343,
+      "step": 17623
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.8002182479863726e-05,
+      "loss": 0.97,
+      "step": 17624
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7992449906959238e-05,
+      "loss": 0.9045,
+      "step": 17625
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7982718750441106e-05,
+      "loss": 0.8785,
+      "step": 17626
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7972989010500804e-05,
+      "loss": 0.8277,
+      "step": 17627
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.796326068732964e-05,
+      "loss": 0.936,
+      "step": 17628
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7953533781119055e-05,
+      "loss": 0.9965,
+      "step": 17629
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7943808292060325e-05,
+      "loss": 0.9936,
+      "step": 17630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7934084220344725e-05,
+      "loss": 0.8283,
+      "step": 17631
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.792436156616357e-05,
+      "loss": 0.929,
+      "step": 17632
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7914640329708118e-05,
+      "loss": 0.9522,
+      "step": 17633
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7904920511169563e-05,
+      "loss": 0.8486,
+      "step": 17634
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.789520211073907e-05,
+      "loss": 1.0778,
+      "step": 17635
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7885485128607846e-05,
+      "loss": 0.8541,
+      "step": 17636
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7875769564966968e-05,
+      "loss": 0.7703,
+      "step": 17637
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.786605542000764e-05,
+      "loss": 0.7712,
+      "step": 17638
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7856342693920813e-05,
+      "loss": 0.892,
+      "step": 17639
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7846631386897583e-05,
+      "loss": 1.047,
+      "step": 17640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.783692149912902e-05,
+      "loss": 0.9032,
+      "step": 17641
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7827213030806086e-05,
+      "loss": 1.0246,
+      "step": 17642
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7817505982119706e-05,
+      "loss": 1.1849,
+      "step": 17643
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7807800353260883e-05,
+      "loss": 0.8597,
+      "step": 17644
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.779809614442046e-05,
+      "loss": 0.8762,
+      "step": 17645
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7788393355789387e-05,
+      "loss": 0.9565,
+      "step": 17646
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7778691987558492e-05,
+      "loss": 0.8931,
+      "step": 17647
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.776899203991855e-05,
+      "loss": 0.8034,
+      "step": 17648
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7759293513060414e-05,
+      "loss": 0.9782,
+      "step": 17649
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7749596407174872e-05,
+      "loss": 0.9384,
+      "step": 17650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7739900722452627e-05,
+      "loss": 0.8988,
+      "step": 17651
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7730206459084374e-05,
+      "loss": 1.0962,
+      "step": 17652
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7720513617260856e-05,
+      "loss": 0.9587,
+      "step": 17653
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.771082219717267e-05,
+      "loss": 0.9353,
+      "step": 17654
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7701132199010505e-05,
+      "loss": 0.8148,
+      "step": 17655
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.769144362296494e-05,
+      "loss": 0.8727,
+      "step": 17656
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7681756469226518e-05,
+      "loss": 0.9848,
+      "step": 17657
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7672070737985834e-05,
+      "loss": 0.8019,
+      "step": 17658
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7662386429433385e-05,
+      "loss": 0.7198,
+      "step": 17659
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.765270354375963e-05,
+      "loss": 0.9274,
+      "step": 17660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.764302208115508e-05,
+      "loss": 0.7704,
+      "step": 17661
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.763334204181013e-05,
+      "loss": 0.9646,
+      "step": 17662
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.7623663425915237e-05,
+      "loss": 0.9269,
+      "step": 17663
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.761398623366074e-05,
+      "loss": 1.006,
+      "step": 17664
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.760431046523697e-05,
+      "loss": 0.6684,
+      "step": 17665
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2.759463612083427e-05,
+      "loss": 1.065,
+      "step": 17666
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7584963200642976e-05,
+      "loss": 1.0526,
+      "step": 17667
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7575291704853323e-05,
+      "loss": 1.0015,
+      "step": 17668
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7565621633655502e-05,
+      "loss": 0.9687,
+      "step": 17669
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7555952987239808e-05,
+      "loss": 1.1154,
+      "step": 17670
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7546285765796365e-05,
+      "loss": 0.7276,
+      "step": 17671
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.753661996951532e-05,
+      "loss": 1.1899,
+      "step": 17672
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7526955598586857e-05,
+      "loss": 1.1286,
+      "step": 17673
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7517292653201e-05,
+      "loss": 0.8255,
+      "step": 17674
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7507631133547886e-05,
+      "loss": 0.8667,
+      "step": 17675
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.749797103981753e-05,
+      "loss": 0.7884,
+      "step": 17676
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.74883123721999e-05,
+      "loss": 0.9086,
+      "step": 17677
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7478655130885034e-05,
+      "loss": 0.881,
+      "step": 17678
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7468999316062915e-05,
+      "loss": 1.1461,
+      "step": 17679
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7459344927923435e-05,
+      "loss": 0.9451,
+      "step": 17680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7449691966656498e-05,
+      "loss": 0.9866,
+      "step": 17681
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.744004043245194e-05,
+      "loss": 0.9389,
+      "step": 17682
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7430390325499654e-05,
+      "loss": 0.7806,
+      "step": 17683
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7420741645989513e-05,
+      "loss": 0.9088,
+      "step": 17684
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.741109439411117e-05,
+      "loss": 0.9692,
+      "step": 17685
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7401448570054454e-05,
+      "loss": 0.9093,
+      "step": 17686
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7391804174009138e-05,
+      "loss": 0.8562,
+      "step": 17687
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7382161206164892e-05,
+      "loss": 0.932,
+      "step": 17688
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.737251966671136e-05,
+      "loss": 0.9945,
+      "step": 17689
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.736287955583825e-05,
+      "loss": 0.7295,
+      "step": 17690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.735324087373512e-05,
+      "loss": 0.9105,
+      "step": 17691
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.734360362059164e-05,
+      "loss": 1.0908,
+      "step": 17692
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7333967796597315e-05,
+      "loss": 0.931,
+      "step": 17693
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7324333401941683e-05,
+      "loss": 1.106,
+      "step": 17694
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7314700436814255e-05,
+      "loss": 1.0446,
+      "step": 17695
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.730506890140456e-05,
+      "loss": 0.8333,
+      "step": 17696
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7295438795902007e-05,
+      "loss": 0.8258,
+      "step": 17697
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7285810120496003e-05,
+      "loss": 1.0138,
+      "step": 17698
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7276182875376e-05,
+      "loss": 0.9953,
+      "step": 17699
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.726655706073129e-05,
+      "loss": 0.9417,
+      "step": 17700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7256932676751323e-05,
+      "loss": 0.9382,
+      "step": 17701
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7247309723625282e-05,
+      "loss": 1.0719,
+      "step": 17702
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.723768820154251e-05,
+      "loss": 0.884,
+      "step": 17703
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7228068110692295e-05,
+      "loss": 0.9232,
+      "step": 17704
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7218449451263827e-05,
+      "loss": 0.9472,
+      "step": 17705
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.720883222344629e-05,
+      "loss": 1.08,
+      "step": 17706
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7199216427428887e-05,
+      "loss": 1.1716,
+      "step": 17707
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.718960206340073e-05,
+      "loss": 1.0634,
+      "step": 17708
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7179989131550977e-05,
+      "loss": 1.1314,
+      "step": 17709
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7170377632068688e-05,
+      "loss": 0.8975,
+      "step": 17710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7160767565142885e-05,
+      "loss": 1.0106,
+      "step": 17711
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.715115893096264e-05,
+      "loss": 0.7933,
+      "step": 17712
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7141551729716974e-05,
+      "loss": 0.7937,
+      "step": 17713
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.713194596159484e-05,
+      "loss": 0.9322,
+      "step": 17714
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.712234162678514e-05,
+      "loss": 0.8774,
+      "step": 17715
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.711273872547687e-05,
+      "loss": 0.7715,
+      "step": 17716
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7103137257858868e-05,
+      "loss": 0.9216,
+      "step": 17717
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.709353722411997e-05,
+      "loss": 0.9675,
+      "step": 17718
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7083938624449078e-05,
+      "loss": 0.9337,
+      "step": 17719
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7074341459034935e-05,
+      "loss": 0.9221,
+      "step": 17720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7064745728066365e-05,
+      "loss": 0.8146,
+      "step": 17721
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7055151431732094e-05,
+      "loss": 0.9295,
+      "step": 17722
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7045558570220807e-05,
+      "loss": 1.0046,
+      "step": 17723
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7035967143721274e-05,
+      "loss": 0.9317,
+      "step": 17724
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7026377152422067e-05,
+      "loss": 0.899,
+      "step": 17725
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.70167885965119e-05,
+      "loss": 1.0046,
+      "step": 17726
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.7007201476179355e-05,
+      "loss": 0.8643,
+      "step": 17727
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6997615791612962e-05,
+      "loss": 0.9429,
+      "step": 17728
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6988031543001314e-05,
+      "loss": 0.7869,
+      "step": 17729
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.697844873053299e-05,
+      "loss": 0.8254,
+      "step": 17730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6968867354396364e-05,
+      "loss": 0.8707,
+      "step": 17731
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6959287414779955e-05,
+      "loss": 0.988,
+      "step": 17732
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.694970891187225e-05,
+      "loss": 0.8605,
+      "step": 17733
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6940131845861604e-05,
+      "loss": 0.8108,
+      "step": 17734
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6930556216936375e-05,
+      "loss": 0.98,
+      "step": 17735
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.692098202528498e-05,
+      "loss": 0.9787,
+      "step": 17736
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.691140927109568e-05,
+      "loss": 0.9424,
+      "step": 17737
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6901837954556843e-05,
+      "loss": 0.8599,
+      "step": 17738
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6892268075856675e-05,
+      "loss": 0.8439,
+      "step": 17739
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6882699635183417e-05,
+      "loss": 1.0099,
+      "step": 17740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6873132632725297e-05,
+      "loss": 0.8119,
+      "step": 17741
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6863567068670538e-05,
+      "loss": 1.0713,
+      "step": 17742
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6854002943207246e-05,
+      "loss": 1.087,
+      "step": 17743
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6844440256523518e-05,
+      "loss": 0.952,
+      "step": 17744
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6834879008807535e-05,
+      "loss": 0.9516,
+      "step": 17745
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.682531920024728e-05,
+      "loss": 1.0262,
+      "step": 17746
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.68157608310309e-05,
+      "loss": 0.9161,
+      "step": 17747
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6806203901346295e-05,
+      "loss": 0.8215,
+      "step": 17748
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6796648411381485e-05,
+      "loss": 0.9121,
+      "step": 17749
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6787094361324472e-05,
+      "loss": 0.6965,
+      "step": 17750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6777541751363155e-05,
+      "loss": 0.8718,
+      "step": 17751
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6767990581685387e-05,
+      "loss": 0.9908,
+      "step": 17752
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.675844085247913e-05,
+      "loss": 0.7205,
+      "step": 17753
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6748892563932136e-05,
+      "loss": 0.8631,
+      "step": 17754
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.67393457162323e-05,
+      "loss": 1.0835,
+      "step": 17755
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6729800309567367e-05,
+      "loss": 0.7745,
+      "step": 17756
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6720256344125074e-05,
+      "loss": 0.883,
+      "step": 17757
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.671071382009316e-05,
+      "loss": 0.9015,
+      "step": 17758
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6701172737659385e-05,
+      "loss": 0.8447,
+      "step": 17759
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6691633097011372e-05,
+      "loss": 0.7828,
+      "step": 17760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.668209489833674e-05,
+      "loss": 1.0405,
+      "step": 17761
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.667255814182318e-05,
+      "loss": 1.0972,
+      "step": 17762
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6663022827658223e-05,
+      "loss": 0.9543,
+      "step": 17763
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6653488956029417e-05,
+      "loss": 1.0049,
+      "step": 17764
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.664395652712435e-05,
+      "loss": 0.9149,
+      "step": 17765
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6634425541130446e-05,
+      "loss": 0.826,
+      "step": 17766
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6624895998235267e-05,
+      "loss": 0.9673,
+      "step": 17767
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.661536789862622e-05,
+      "loss": 0.8435,
+      "step": 17768
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6605841242490694e-05,
+      "loss": 0.9838,
+      "step": 17769
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6596316030016122e-05,
+      "loss": 1.0284,
+      "step": 17770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6586792261389826e-05,
+      "loss": 1.0361,
+      "step": 17771
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6577269936799177e-05,
+      "loss": 0.9067,
+      "step": 17772
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6567749056431467e-05,
+      "loss": 0.9794,
+      "step": 17773
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.655822962047394e-05,
+      "loss": 0.8681,
+      "step": 17774
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6548711629113866e-05,
+      "loss": 0.819,
+      "step": 17775
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6539195082538514e-05,
+      "loss": 1.1073,
+      "step": 17776
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6529679980934975e-05,
+      "loss": 0.9098,
+      "step": 17777
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.652016632449046e-05,
+      "loss": 0.8684,
+      "step": 17778
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6510654113392142e-05,
+      "loss": 0.7639,
+      "step": 17779
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6501143347827074e-05,
+      "loss": 0.8104,
+      "step": 17780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6491634027982325e-05,
+      "loss": 0.9806,
+      "step": 17781
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6482126154044996e-05,
+      "loss": 0.9497,
+      "step": 17782
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.647261972620203e-05,
+      "loss": 0.8422,
+      "step": 17783
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6463114744640504e-05,
+      "loss": 0.9433,
+      "step": 17784
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6453611209547346e-05,
+      "loss": 0.8963,
+      "step": 17785
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.644410912110945e-05,
+      "loss": 0.7796,
+      "step": 17786
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6434608479513745e-05,
+      "loss": 0.939,
+      "step": 17787
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.642510928494716e-05,
+      "loss": 0.7436,
+      "step": 17788
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6415611537596496e-05,
+      "loss": 0.8704,
+      "step": 17789
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6406115237648587e-05,
+      "loss": 0.9251,
+      "step": 17790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6396620385290184e-05,
+      "loss": 0.9495,
+      "step": 17791
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6387126980708087e-05,
+      "loss": 0.7161,
+      "step": 17792
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6377635024089087e-05,
+      "loss": 0.8813,
+      "step": 17793
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6368144515619776e-05,
+      "loss": 0.7778,
+      "step": 17794
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.635865545548689e-05,
+      "loss": 0.7021,
+      "step": 17795
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6349167843877097e-05,
+      "loss": 1.102,
+      "step": 17796
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6339681680977003e-05,
+      "loss": 0.8167,
+      "step": 17797
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6330196966973175e-05,
+      "loss": 0.8184,
+      "step": 17798
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6320713702052212e-05,
+      "loss": 0.9156,
+      "step": 17799
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.631123188640061e-05,
+      "loss": 0.784,
+      "step": 17800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.630175152020493e-05,
+      "loss": 0.8227,
+      "step": 17801
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6292272603651625e-05,
+      "loss": 0.9686,
+      "step": 17802
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.628279513692712e-05,
+      "loss": 0.8807,
+      "step": 17803
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6273319120217845e-05,
+      "loss": 0.7846,
+      "step": 17804
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6263844553710236e-05,
+      "loss": 0.9109,
+      "step": 17805
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.625437143759063e-05,
+      "loss": 0.9569,
+      "step": 17806
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6244899772045338e-05,
+      "loss": 1.0416,
+      "step": 17807
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6235429557260715e-05,
+      "loss": 0.9828,
+      "step": 17808
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.622596079342301e-05,
+      "loss": 1.0585,
+      "step": 17809
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.621649348071845e-05,
+      "loss": 0.8989,
+      "step": 17810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6207027619333312e-05,
+      "loss": 0.9613,
+      "step": 17811
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6197563209453735e-05,
+      "loss": 0.8028,
+      "step": 17812
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6188100251265945e-05,
+      "loss": 0.9594,
+      "step": 17813
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6178638744956042e-05,
+      "loss": 0.9441,
+      "step": 17814
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6169178690710106e-05,
+      "loss": 0.6903,
+      "step": 17815
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6159720088714268e-05,
+      "loss": 1.0031,
+      "step": 17816
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6150262939154525e-05,
+      "loss": 0.8481,
+      "step": 17817
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.614080724221697e-05,
+      "loss": 0.978,
+      "step": 17818
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6131352998087545e-05,
+      "loss": 0.793,
+      "step": 17819
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.61219002069522e-05,
+      "loss": 0.9586,
+      "step": 17820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6112448868996897e-05,
+      "loss": 0.8659,
+      "step": 17821
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6102998984407612e-05,
+      "loss": 0.8669,
+      "step": 17822
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6093550553370084e-05,
+      "loss": 0.9133,
+      "step": 17823
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.608410357607024e-05,
+      "loss": 0.9724,
+      "step": 17824
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.607465805269391e-05,
+      "loss": 0.9122,
+      "step": 17825
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6065213983426896e-05,
+      "loss": 0.9468,
+      "step": 17826
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6055771368454896e-05,
+      "loss": 0.809,
+      "step": 17827
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6046330207963733e-05,
+      "loss": 0.7561,
+      "step": 17828
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.603689050213902e-05,
+      "loss": 0.8558,
+      "step": 17829
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.602745225116654e-05,
+      "loss": 0.6961,
+      "step": 17830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.601801545523187e-05,
+      "loss": 0.9235,
+      "step": 17831
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.6008580114520632e-05,
+      "loss": 0.9951,
+      "step": 17832
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5999146229218475e-05,
+      "loss": 0.9509,
+      "step": 17833
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5989713799510884e-05,
+      "loss": 0.8759,
+      "step": 17834
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5980282825583468e-05,
+      "loss": 0.9024,
+      "step": 17835
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5970853307621713e-05,
+      "loss": 0.8622,
+      "step": 17836
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5961425245811055e-05,
+      "loss": 1.0158,
+      "step": 17837
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5951998640336972e-05,
+      "loss": 1.049,
+      "step": 17838
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.594257349138495e-05,
+      "loss": 0.976,
+      "step": 17839
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5933149799140266e-05,
+      "loss": 1.0538,
+      "step": 17840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5923727563788336e-05,
+      "loss": 0.8705,
+      "step": 17841
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.591430678551452e-05,
+      "loss": 1.0288,
+      "step": 17842
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5904887464504114e-05,
+      "loss": 0.952,
+      "step": 17843
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5895469600942346e-05,
+      "loss": 0.8307,
+      "step": 17844
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5886053195014538e-05,
+      "loss": 1.0632,
+      "step": 17845
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5876638246905838e-05,
+      "loss": 1.111,
+      "step": 17846
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.586722475680151e-05,
+      "loss": 0.8769,
+      "step": 17847
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5857812724886677e-05,
+      "loss": 0.9482,
+      "step": 17848
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.584840215134645e-05,
+      "loss": 0.9705,
+      "step": 17849
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5838993036365976e-05,
+      "loss": 0.9246,
+      "step": 17850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5829585380130338e-05,
+      "loss": 1.0068,
+      "step": 17851
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5820179182824567e-05,
+      "loss": 0.8843,
+      "step": 17852
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5810774444633644e-05,
+      "loss": 0.8965,
+      "step": 17853
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5801371165742637e-05,
+      "loss": 0.9231,
+      "step": 17854
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5791969346336463e-05,
+      "loss": 0.5869,
+      "step": 17855
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5782568986600065e-05,
+      "loss": 1.0351,
+      "step": 17856
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5773170086718312e-05,
+      "loss": 0.9073,
+      "step": 17857
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5763772646876116e-05,
+      "loss": 0.8397,
+      "step": 17858
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5754376667258352e-05,
+      "loss": 0.7192,
+      "step": 17859
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.57449821480498e-05,
+      "loss": 0.8192,
+      "step": 17860
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.573558908943523e-05,
+      "loss": 0.8567,
+      "step": 17861
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5726197491599457e-05,
+      "loss": 0.7998,
+      "step": 17862
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5716807354727156e-05,
+      "loss": 1.0576,
+      "step": 17863
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.570741867900309e-05,
+      "loss": 0.7694,
+      "step": 17864
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5698031464611905e-05,
+      "loss": 0.8135,
+      "step": 17865
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5688645711738224e-05,
+      "loss": 1.0696,
+      "step": 17866
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.567926142056668e-05,
+      "loss": 0.82,
+      "step": 17867
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5669878591281926e-05,
+      "loss": 1.0721,
+      "step": 17868
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5660497224068415e-05,
+      "loss": 0.9874,
+      "step": 17869
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.565111731911073e-05,
+      "loss": 0.7979,
+      "step": 17870
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.564173887659339e-05,
+      "loss": 0.8457,
+      "step": 17871
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5632361896700864e-05,
+      "loss": 0.8672,
+      "step": 17872
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.562298637961755e-05,
+      "loss": 0.709,
+      "step": 17873
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5613612325527926e-05,
+      "loss": 1.0054,
+      "step": 17874
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.560423973461632e-05,
+      "loss": 0.9706,
+      "step": 17875
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5594868607067145e-05,
+      "loss": 0.865,
+      "step": 17876
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5585498943064724e-05,
+      "loss": 0.9522,
+      "step": 17877
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.55761307427933e-05,
+      "loss": 0.8292,
+      "step": 17878
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5566764006437215e-05,
+      "loss": 0.9449,
+      "step": 17879
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.555739873418066e-05,
+      "loss": 1.0612,
+      "step": 17880
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.554803492620791e-05,
+      "loss": 1.0324,
+      "step": 17881
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5538672582703106e-05,
+      "loss": 0.9785,
+      "step": 17882
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.552931170385039e-05,
+      "loss": 1.0155,
+      "step": 17883
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5519952289833904e-05,
+      "loss": 0.752,
+      "step": 17884
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5510594340837824e-05,
+      "loss": 0.9746,
+      "step": 17885
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.550123785704609e-05,
+      "loss": 1.1113,
+      "step": 17886
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.549188283864281e-05,
+      "loss": 1.02,
+      "step": 17887
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5482529285812028e-05,
+      "loss": 0.9427,
+      "step": 17888
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5473177198737696e-05,
+      "loss": 0.938,
+      "step": 17889
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5463826577603733e-05,
+      "loss": 0.8398,
+      "step": 17890
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5454477422594125e-05,
+      "loss": 0.7705,
+      "step": 17891
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5445129733892724e-05,
+      "loss": 1.064,
+      "step": 17892
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5435783511683443e-05,
+      "loss": 0.8611,
+      "step": 17893
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5426438756150105e-05,
+      "loss": 0.8753,
+      "step": 17894
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.541709546747648e-05,
+      "loss": 0.884,
+      "step": 17895
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.540775364584639e-05,
+      "loss": 0.9435,
+      "step": 17896
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2.5398413291443613e-05,
+      "loss": 0.9222,
+      "step": 17897
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5389074404451842e-05,
+      "loss": 0.8315,
+      "step": 17898
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.537973698505477e-05,
+      "loss": 0.925,
+      "step": 17899
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5370401033436042e-05,
+      "loss": 0.9907,
+      "step": 17900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5361066549779356e-05,
+      "loss": 0.9005,
+      "step": 17901
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5351733534268285e-05,
+      "loss": 0.765,
+      "step": 17902
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5342401987086385e-05,
+      "loss": 0.8738,
+      "step": 17903
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.533307190841723e-05,
+      "loss": 1.181,
+      "step": 17904
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.532374329844437e-05,
+      "loss": 0.8899,
+      "step": 17905
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5314416157351274e-05,
+      "loss": 0.9756,
+      "step": 17906
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.530509048532138e-05,
+      "loss": 0.8858,
+      "step": 17907
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.529576628253819e-05,
+      "loss": 0.9945,
+      "step": 17908
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5286443549185036e-05,
+      "loss": 0.8158,
+      "step": 17909
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5277122285445352e-05,
+      "loss": 0.9481,
+      "step": 17910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.526780249150248e-05,
+      "loss": 0.9738,
+      "step": 17911
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5258484167539687e-05,
+      "loss": 0.7923,
+      "step": 17912
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5249167313740308e-05,
+      "loss": 0.8899,
+      "step": 17913
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.523985193028766e-05,
+      "loss": 0.9717,
+      "step": 17914
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.523053801736486e-05,
+      "loss": 0.9792,
+      "step": 17915
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.522122557515516e-05,
+      "loss": 0.7728,
+      "step": 17916
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.521191460384178e-05,
+      "loss": 0.7749,
+      "step": 17917
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5202605103607835e-05,
+      "loss": 0.8472,
+      "step": 17918
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5193297074636413e-05,
+      "loss": 0.9429,
+      "step": 17919
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5183990517110644e-05,
+      "loss": 0.8914,
+      "step": 17920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.517468543121355e-05,
+      "loss": 0.8299,
+      "step": 17921
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5165381817128218e-05,
+      "loss": 0.8609,
+      "step": 17922
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5156079675037612e-05,
+      "loss": 0.9088,
+      "step": 17923
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.514677900512469e-05,
+      "loss": 0.8079,
+      "step": 17924
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5137479807572427e-05,
+      "loss": 0.7756,
+      "step": 17925
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.512818208256371e-05,
+      "loss": 0.8659,
+      "step": 17926
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5118885830281468e-05,
+      "loss": 0.9455,
+      "step": 17927
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.510959105090852e-05,
+      "loss": 0.9184,
+      "step": 17928
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5100297744627686e-05,
+      "loss": 0.8225,
+      "step": 17929
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.509100591162178e-05,
+      "loss": 0.8599,
+      "step": 17930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5081715552073637e-05,
+      "loss": 0.9837,
+      "step": 17931
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5072426666165872e-05,
+      "loss": 0.7615,
+      "step": 17932
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.506313925408127e-05,
+      "loss": 0.8313,
+      "step": 17933
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.505385331600253e-05,
+      "loss": 0.8682,
+      "step": 17934
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5044568852112292e-05,
+      "loss": 0.9324,
+      "step": 17935
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.503528586259313e-05,
+      "loss": 0.9686,
+      "step": 17936
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5026004347627717e-05,
+      "loss": 0.7878,
+      "step": 17937
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5016724307398553e-05,
+      "loss": 0.8921,
+      "step": 17938
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.5007445742088242e-05,
+      "loss": 0.7955,
+      "step": 17939
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.499816865187925e-05,
+      "loss": 0.8339,
+      "step": 17940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4988893036954043e-05,
+      "loss": 1.021,
+      "step": 17941
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4979618897495128e-05,
+      "loss": 1.0742,
+      "step": 17942
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.497034623368486e-05,
+      "loss": 0.8232,
+      "step": 17943
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4961075045705697e-05,
+      "loss": 0.7265,
+      "step": 17944
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4951805333739974e-05,
+      "loss": 0.9433,
+      "step": 17945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4942537097969987e-05,
+      "loss": 0.895,
+      "step": 17946
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.493327033857812e-05,
+      "loss": 1.0672,
+      "step": 17947
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4924005055746604e-05,
+      "loss": 1.023,
+      "step": 17948
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4914741249657658e-05,
+      "loss": 0.9661,
+      "step": 17949
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.490547892049354e-05,
+      "loss": 0.9597,
+      "step": 17950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.489621806843646e-05,
+      "loss": 0.8523,
+      "step": 17951
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.488695869366856e-05,
+      "loss": 0.9279,
+      "step": 17952
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4877700796371928e-05,
+      "loss": 1.1827,
+      "step": 17953
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.486844437672873e-05,
+      "loss": 0.8031,
+      "step": 17954
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4859189434920994e-05,
+      "loss": 0.7608,
+      "step": 17955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.48499359711308e-05,
+      "loss": 0.8195,
+      "step": 17956
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.484068398554016e-05,
+      "loss": 0.7939,
+      "step": 17957
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4831433478331012e-05,
+      "loss": 0.9582,
+      "step": 17958
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4822184449685347e-05,
+      "loss": 0.9102,
+      "step": 17959
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4812936899785146e-05,
+      "loss": 0.8373,
+      "step": 17960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.480369082881221e-05,
+      "loss": 1.0095,
+      "step": 17961
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4794446236948445e-05,
+      "loss": 1.043,
+      "step": 17962
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4785203124375732e-05,
+      "loss": 0.9947,
+      "step": 17963
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4775961491275868e-05,
+      "loss": 0.8597,
+      "step": 17964
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.47667213378306e-05,
+      "loss": 0.8764,
+      "step": 17965
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4757482664221688e-05,
+      "loss": 0.9169,
+      "step": 17966
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.474824547063086e-05,
+      "loss": 0.8435,
+      "step": 17967
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.473900975723985e-05,
+      "loss": 0.7658,
+      "step": 17968
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4729775524230304e-05,
+      "loss": 0.7759,
+      "step": 17969
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4720542771783817e-05,
+      "loss": 0.8231,
+      "step": 17970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4711311500082067e-05,
+      "loss": 1.1093,
+      "step": 17971
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4702081709306555e-05,
+      "loss": 0.9725,
+      "step": 17972
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4692853399638917e-05,
+      "loss": 1.0554,
+      "step": 17973
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4683626571260622e-05,
+      "loss": 0.7831,
+      "step": 17974
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4674401224353137e-05,
+      "loss": 0.9934,
+      "step": 17975
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4665177359097957e-05,
+      "loss": 0.9101,
+      "step": 17976
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4655954975676564e-05,
+      "loss": 1.1185,
+      "step": 17977
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4646734074270273e-05,
+      "loss": 1.0341,
+      "step": 17978
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.463751465506048e-05,
+      "loss": 1.0452,
+      "step": 17979
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4628296718228582e-05,
+      "loss": 0.9641,
+      "step": 17980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4619080263955875e-05,
+      "loss": 0.7996,
+      "step": 17981
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4609865292423584e-05,
+      "loss": 1.0069,
+      "step": 17982
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4600651803813056e-05,
+      "loss": 0.756,
+      "step": 17983
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.459143979830546e-05,
+      "loss": 0.9472,
+      "step": 17984
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4582229276082037e-05,
+      "loss": 0.906,
+      "step": 17985
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.457302023732394e-05,
+      "loss": 0.8003,
+      "step": 17986
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4563812682212272e-05,
+      "loss": 0.996,
+      "step": 17987
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.455460661092821e-05,
+      "loss": 0.918,
+      "step": 17988
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4545402023652784e-05,
+      "loss": 0.8242,
+      "step": 17989
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.453619892056711e-05,
+      "loss": 0.7449,
+      "step": 17990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.452699730185216e-05,
+      "loss": 0.8794,
+      "step": 17991
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4517797167688928e-05,
+      "loss": 0.9867,
+      "step": 17992
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.450859851825842e-05,
+      "loss": 1.1069,
+      "step": 17993
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4499401353741546e-05,
+      "loss": 0.9365,
+      "step": 17994
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.44902056743192e-05,
+      "loss": 0.9875,
+      "step": 17995
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4481011480172266e-05,
+      "loss": 1.0408,
+      "step": 17996
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4471818771481648e-05,
+      "loss": 0.9901,
+      "step": 17997
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.446262754842812e-05,
+      "loss": 1.1412,
+      "step": 17998
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4453437811192435e-05,
+      "loss": 0.9016,
+      "step": 17999
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4444249559955446e-05,
+      "loss": 0.9054,
+      "step": 18000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.44350627948978e-05,
+      "loss": 0.9903,
+      "step": 18001
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.442587751620027e-05,
+      "loss": 0.8805,
+      "step": 18002
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4416693724043493e-05,
+      "loss": 1.0471,
+      "step": 18003
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4407511418608097e-05,
+      "loss": 0.8356,
+      "step": 18004
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4398330600074713e-05,
+      "loss": 1.0015,
+      "step": 18005
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.438915126862399e-05,
+      "loss": 0.937,
+      "step": 18006
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4379973424436375e-05,
+      "loss": 0.9424,
+      "step": 18007
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4370797067692487e-05,
+      "loss": 0.9563,
+      "step": 18008
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4361622198572743e-05,
+      "loss": 0.879,
+      "step": 18009
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4352448817257688e-05,
+      "loss": 1.0722,
+      "step": 18010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.434327692392774e-05,
+      "loss": 0.7012,
+      "step": 18011
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4334106518763266e-05,
+      "loss": 1.0309,
+      "step": 18012
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4324937601944685e-05,
+      "loss": 0.8671,
+      "step": 18013
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4315770173652386e-05,
+      "loss": 0.8864,
+      "step": 18014
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4306604234066645e-05,
+      "loss": 1.0201,
+      "step": 18015
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4297439783367727e-05,
+      "loss": 0.9488,
+      "step": 18016
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4288276821735966e-05,
+      "loss": 0.806,
+      "step": 18017
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4279115349351543e-05,
+      "loss": 1.1028,
+      "step": 18018
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4269955366394703e-05,
+      "loss": 0.8907,
+      "step": 18019
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.426079687304561e-05,
+      "loss": 1.0186,
+      "step": 18020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4251639869484375e-05,
+      "loss": 0.9482,
+      "step": 18021
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.424248435589116e-05,
+      "loss": 1.0017,
+      "step": 18022
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4233330332446092e-05,
+      "loss": 0.9975,
+      "step": 18023
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4224177799329117e-05,
+      "loss": 0.8796,
+      "step": 18024
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4215026756720328e-05,
+      "loss": 1.078,
+      "step": 18025
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4205877204799755e-05,
+      "loss": 0.9265,
+      "step": 18026
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4196729143747333e-05,
+      "loss": 0.8687,
+      "step": 18027
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4187582573742994e-05,
+      "loss": 0.939,
+      "step": 18028
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4178437494966688e-05,
+      "loss": 0.8976,
+      "step": 18029
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4169293907598246e-05,
+      "loss": 0.849,
+      "step": 18030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.416015181181759e-05,
+      "loss": 0.99,
+      "step": 18031
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4151011207804507e-05,
+      "loss": 0.9661,
+      "step": 18032
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4141872095738772e-05,
+      "loss": 0.9208,
+      "step": 18033
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.413273447580019e-05,
+      "loss": 0.8646,
+      "step": 18034
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4123598348168453e-05,
+      "loss": 0.9119,
+      "step": 18035
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.411446371302333e-05,
+      "loss": 0.9791,
+      "step": 18036
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.410533057054446e-05,
+      "loss": 1.1035,
+      "step": 18037
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4096198920911472e-05,
+      "loss": 1.0739,
+      "step": 18038
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4087068764304043e-05,
+      "loss": 0.9441,
+      "step": 18039
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.407794010090172e-05,
+      "loss": 0.9663,
+      "step": 18040
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4068812930884054e-05,
+      "loss": 0.9778,
+      "step": 18041
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4059687254430595e-05,
+      "loss": 0.9906,
+      "step": 18042
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4050563071720867e-05,
+      "loss": 0.9848,
+      "step": 18043
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.404144038293432e-05,
+      "loss": 0.9633,
+      "step": 18044
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4032319188250374e-05,
+      "loss": 1.0218,
+      "step": 18045
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4023199487848502e-05,
+      "loss": 0.7125,
+      "step": 18046
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4014081281908017e-05,
+      "loss": 1.0337,
+      "step": 18047
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.4004964570608335e-05,
+      "loss": 0.9389,
+      "step": 18048
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3995849354128763e-05,
+      "loss": 0.9287,
+      "step": 18049
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.398673563264856e-05,
+      "loss": 0.8408,
+      "step": 18050
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.397762340634705e-05,
+      "loss": 0.8887,
+      "step": 18051
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3968512675403444e-05,
+      "loss": 0.8082,
+      "step": 18052
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3959403439996907e-05,
+      "loss": 0.9516,
+      "step": 18053
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.39502957003067e-05,
+      "loss": 0.7839,
+      "step": 18054
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3941189456511904e-05,
+      "loss": 1.0876,
+      "step": 18055
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3932084708791692e-05,
+      "loss": 0.8554,
+      "step": 18056
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3922981457325132e-05,
+      "loss": 1.0499,
+      "step": 18057
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3913879702291243e-05,
+      "loss": 0.7605,
+      "step": 18058
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3904779443869108e-05,
+      "loss": 0.892,
+      "step": 18059
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.389568068223773e-05,
+      "loss": 0.8882,
+      "step": 18060
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.388658341757607e-05,
+      "loss": 0.8197,
+      "step": 18061
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3877487650063036e-05,
+      "loss": 0.8397,
+      "step": 18062
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3868393379877608e-05,
+      "loss": 1.0321,
+      "step": 18063
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3859300607198598e-05,
+      "loss": 0.928,
+      "step": 18064
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.385020933220492e-05,
+      "loss": 1.0319,
+      "step": 18065
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3841119555075388e-05,
+      "loss": 0.841,
+      "step": 18066
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.383203127598875e-05,
+      "loss": 0.8253,
+      "step": 18067
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.382294449512381e-05,
+      "loss": 0.9054,
+      "step": 18068
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.381385921265936e-05,
+      "loss": 0.8534,
+      "step": 18069
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3804775428773995e-05,
+      "loss": 1.0957,
+      "step": 18070
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3795693143646448e-05,
+      "loss": 0.9813,
+      "step": 18071
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3786612357455395e-05,
+      "loss": 0.9973,
+      "step": 18072
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.377753307037942e-05,
+      "loss": 1.0149,
+      "step": 18073
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3768455282597125e-05,
+      "loss": 1.0635,
+      "step": 18074
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.375937899428704e-05,
+      "loss": 0.9884,
+      "step": 18075
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3750304205627703e-05,
+      "loss": 0.9219,
+      "step": 18076
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3741230916797675e-05,
+      "loss": 0.8246,
+      "step": 18077
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3732159127975373e-05,
+      "loss": 0.8184,
+      "step": 18078
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3723088839339225e-05,
+      "loss": 0.999,
+      "step": 18079
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.37140200510677e-05,
+      "loss": 0.9761,
+      "step": 18080
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3704952763339115e-05,
+      "loss": 0.8664,
+      "step": 18081
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3695886976331883e-05,
+      "loss": 1.0975,
+      "step": 18082
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3686822690224308e-05,
+      "loss": 0.9082,
+      "step": 18083
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.367775990519464e-05,
+      "loss": 1.0678,
+      "step": 18084
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3668698621421216e-05,
+      "loss": 1.0327,
+      "step": 18085
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3659638839082242e-05,
+      "loss": 0.7777,
+      "step": 18086
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3650580558355874e-05,
+      "loss": 0.9483,
+      "step": 18087
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3641523779420348e-05,
+      "loss": 1.0161,
+      "step": 18088
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.363246850245381e-05,
+      "loss": 0.9026,
+      "step": 18089
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3623414727634373e-05,
+      "loss": 0.9009,
+      "step": 18090
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.361436245514007e-05,
+      "loss": 0.7367,
+      "step": 18091
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3605311685149046e-05,
+      "loss": 0.9652,
+      "step": 18092
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3596262417839255e-05,
+      "loss": 0.813,
+      "step": 18093
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3587214653388746e-05,
+      "loss": 0.8195,
+      "step": 18094
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3578168391975474e-05,
+      "loss": 0.8708,
+      "step": 18095
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3569123633777347e-05,
+      "loss": 0.848,
+      "step": 18096
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3560080378972326e-05,
+      "loss": 0.9155,
+      "step": 18097
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3551038627738265e-05,
+      "loss": 0.8866,
+      "step": 18098
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3541998380252996e-05,
+      "loss": 0.8093,
+      "step": 18099
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3532959636694373e-05,
+      "loss": 0.9607,
+      "step": 18100
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.352392239724016e-05,
+      "loss": 1.0085,
+      "step": 18101
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.351488666206817e-05,
+      "loss": 1.0221,
+      "step": 18102
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.350585243135609e-05,
+      "loss": 0.8531,
+      "step": 18103
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3496819705281615e-05,
+      "loss": 0.9438,
+      "step": 18104
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3487788484022432e-05,
+      "loss": 0.8709,
+      "step": 18105
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3478758767756215e-05,
+      "loss": 0.8155,
+      "step": 18106
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.346973055666056e-05,
+      "loss": 1.014,
+      "step": 18107
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3460703850913014e-05,
+      "loss": 0.9071,
+      "step": 18108
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.34516786506912e-05,
+      "loss": 0.8976,
+      "step": 18109
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.344265495617256e-05,
+      "loss": 0.7828,
+      "step": 18110
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3433632767534687e-05,
+      "loss": 0.8905,
+      "step": 18111
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3424612084954977e-05,
+      "loss": 0.8708,
+      "step": 18112
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3415592908610872e-05,
+      "loss": 1.0783,
+      "step": 18113
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3406575238679785e-05,
+      "loss": 1.0167,
+      "step": 18114
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3397559075339158e-05,
+      "loss": 1.0242,
+      "step": 18115
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3388544418766233e-05,
+      "loss": 0.9972,
+      "step": 18116
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.33795312691384e-05,
+      "loss": 0.7882,
+      "step": 18117
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3370519626632902e-05,
+      "loss": 1.0071,
+      "step": 18118
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3361509491427047e-05,
+      "loss": 0.8295,
+      "step": 18119
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3352500863698036e-05,
+      "loss": 0.9116,
+      "step": 18120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3343493743623033e-05,
+      "loss": 1.0176,
+      "step": 18121
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.333448813137925e-05,
+      "loss": 1.0058,
+      "step": 18122
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.332548402714385e-05,
+      "loss": 1.0386,
+      "step": 18123
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.331648143109393e-05,
+      "loss": 1.1531,
+      "step": 18124
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.330748034340651e-05,
+      "loss": 0.9216,
+      "step": 18125
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3298480764258722e-05,
+      "loss": 1.0587,
+      "step": 18126
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3289482693827525e-05,
+      "loss": 1.1621,
+      "step": 18127
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2.3280486132289958e-05,
+      "loss": 0.9412,
+      "step": 18128
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.327149107982297e-05,
+      "loss": 0.7383,
+      "step": 18129
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3262497536603468e-05,
+      "loss": 0.9494,
+      "step": 18130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.325350550280838e-05,
+      "loss": 0.9048,
+      "step": 18131
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3244514978614595e-05,
+      "loss": 0.9713,
+      "step": 18132
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.323552596419889e-05,
+      "loss": 1.0816,
+      "step": 18133
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3226538459738133e-05,
+      "loss": 1.0282,
+      "step": 18134
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.321755246540912e-05,
+      "loss": 0.8174,
+      "step": 18135
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3208567981388585e-05,
+      "loss": 0.8365,
+      "step": 18136
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3199585007853218e-05,
+      "loss": 0.9032,
+      "step": 18137
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3190603544979785e-05,
+      "loss": 0.8115,
+      "step": 18138
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3181623592944877e-05,
+      "loss": 0.7376,
+      "step": 18139
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3172645151925197e-05,
+      "loss": 1.1013,
+      "step": 18140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3163668222097313e-05,
+      "loss": 0.8731,
+      "step": 18141
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3154692803637778e-05,
+      "loss": 1.0486,
+      "step": 18142
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.31457188967232e-05,
+      "loss": 0.984,
+      "step": 18143
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.313674650153006e-05,
+      "loss": 0.8623,
+      "step": 18144
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3127775618234814e-05,
+      "loss": 0.715,
+      "step": 18145
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.311880624701399e-05,
+      "loss": 0.8478,
+      "step": 18146
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3109838388043937e-05,
+      "loss": 0.8395,
+      "step": 18147
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3100872041501133e-05,
+      "loss": 0.9575,
+      "step": 18148
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.30919072075619e-05,
+      "loss": 0.9646,
+      "step": 18149
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3082943886402554e-05,
+      "loss": 0.7324,
+      "step": 18150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.307398207819943e-05,
+      "loss": 0.8556,
+      "step": 18151
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3065021783128825e-05,
+      "loss": 0.9332,
+      "step": 18152
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.305606300136698e-05,
+      "loss": 0.9457,
+      "step": 18153
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.304710573309008e-05,
+      "loss": 0.8944,
+      "step": 18154
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3038149978474368e-05,
+      "loss": 0.8431,
+      "step": 18155
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.302919573769593e-05,
+      "loss": 0.9846,
+      "step": 18156
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3020243010930975e-05,
+      "loss": 0.9216,
+      "step": 18157
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.3011291798355573e-05,
+      "loss": 1.1039,
+      "step": 18158
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.300234210014576e-05,
+      "loss": 0.8161,
+      "step": 18159
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.299339391647759e-05,
+      "loss": 0.8572,
+      "step": 18160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2984447247527152e-05,
+      "loss": 1.057,
+      "step": 18161
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2975502093470312e-05,
+      "loss": 0.9317,
+      "step": 18162
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.29665584544831e-05,
+      "loss": 0.7878,
+      "step": 18163
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.295761633074137e-05,
+      "loss": 0.9726,
+      "step": 18164
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2948675722421086e-05,
+      "loss": 1.0703,
+      "step": 18165
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2939736629698084e-05,
+      "loss": 0.8084,
+      "step": 18166
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.293079905274814e-05,
+      "loss": 0.878,
+      "step": 18167
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.292186299174712e-05,
+      "loss": 0.9507,
+      "step": 18168
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2912928446870807e-05,
+      "loss": 0.9144,
+      "step": 18169
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.290399541829491e-05,
+      "loss": 1.1118,
+      "step": 18170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2895063906195115e-05,
+      "loss": 0.7431,
+      "step": 18171
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2886133910747167e-05,
+      "loss": 1.0103,
+      "step": 18172
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2877205432126657e-05,
+      "loss": 0.9331,
+      "step": 18173
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2868278470509274e-05,
+      "loss": 0.8327,
+      "step": 18174
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2859353026070574e-05,
+      "loss": 0.8874,
+      "step": 18175
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2850429098986093e-05,
+      "loss": 0.8717,
+      "step": 18176
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2841506689431423e-05,
+      "loss": 0.7724,
+      "step": 18177
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.283258579758204e-05,
+      "loss": 0.6796,
+      "step": 18178
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2823666423613388e-05,
+      "loss": 1.0777,
+      "step": 18179
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2814748567700927e-05,
+      "loss": 0.8676,
+      "step": 18180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.280583223002012e-05,
+      "loss": 0.9379,
+      "step": 18181
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2796917410746322e-05,
+      "loss": 1.033,
+      "step": 18182
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2788004110054862e-05,
+      "loss": 1.0154,
+      "step": 18183
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2779092328121067e-05,
+      "loss": 0.991,
+      "step": 18184
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2770182065120237e-05,
+      "loss": 0.9146,
+      "step": 18185
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2761273321227682e-05,
+      "loss": 0.8481,
+      "step": 18186
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2752366096618595e-05,
+      "loss": 0.6815,
+      "step": 18187
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2743460391468153e-05,
+      "loss": 0.8656,
+      "step": 18188
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2734556205951596e-05,
+      "loss": 0.9046,
+      "step": 18189
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.272565354024404e-05,
+      "loss": 0.9006,
+      "step": 18190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.271675239452057e-05,
+      "loss": 0.9847,
+      "step": 18191
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.270785276895633e-05,
+      "loss": 0.8705,
+      "step": 18192
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.26989546637263e-05,
+      "loss": 0.7439,
+      "step": 18193
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2690058079005595e-05,
+      "loss": 0.7782,
+      "step": 18194
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2681163014969154e-05,
+      "loss": 0.8501,
+      "step": 18195
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.267226947179193e-05,
+      "loss": 0.7805,
+      "step": 18196
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.266337744964888e-05,
+      "loss": 0.9234,
+      "step": 18197
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2654486948714936e-05,
+      "loss": 0.7975,
+      "step": 18198
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2645597969164955e-05,
+      "loss": 0.9954,
+      "step": 18199
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2636710511173752e-05,
+      "loss": 1.0642,
+      "step": 18200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2627824574916203e-05,
+      "loss": 0.93,
+      "step": 18201
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2618940160567025e-05,
+      "loss": 1.0409,
+      "step": 18202
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.261005726830103e-05,
+      "loss": 0.8318,
+      "step": 18203
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2601175898292936e-05,
+      "loss": 0.8842,
+      "step": 18204
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2592296050717388e-05,
+      "loss": 0.8077,
+      "step": 18205
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2583417725749123e-05,
+      "loss": 0.7411,
+      "step": 18206
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2574540923562748e-05,
+      "loss": 0.8433,
+      "step": 18207
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2565665644332834e-05,
+      "loss": 0.7268,
+      "step": 18208
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2556791888234032e-05,
+      "loss": 0.7124,
+      "step": 18209
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2547919655440807e-05,
+      "loss": 0.8701,
+      "step": 18210
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.253904894612774e-05,
+      "loss": 0.743,
+      "step": 18211
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2530179760469306e-05,
+      "loss": 1.0169,
+      "step": 18212
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2521312098639914e-05,
+      "loss": 0.9235,
+      "step": 18213
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.251244596081402e-05,
+      "loss": 0.8914,
+      "step": 18214
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.250358134716606e-05,
+      "loss": 1.0117,
+      "step": 18215
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.249471825787036e-05,
+      "loss": 0.9275,
+      "step": 18216
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2485856693101247e-05,
+      "loss": 1.1111,
+      "step": 18217
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2476996653033065e-05,
+      "loss": 0.8395,
+      "step": 18218
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2468138137840043e-05,
+      "loss": 1.1033,
+      "step": 18219
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2459281147696476e-05,
+      "loss": 0.8555,
+      "step": 18220
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2450425682776565e-05,
+      "loss": 0.7736,
+      "step": 18221
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2441571743254463e-05,
+      "loss": 0.9939,
+      "step": 18222
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2432719329304385e-05,
+      "loss": 0.97,
+      "step": 18223
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2423868441100427e-05,
+      "loss": 1.0037,
+      "step": 18224
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2415019078816656e-05,
+      "loss": 0.76,
+      "step": 18225
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2406171242627204e-05,
+      "loss": 0.9048,
+      "step": 18226
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2397324932706033e-05,
+      "loss": 0.9646,
+      "step": 18227
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2388480149227232e-05,
+      "loss": 0.9728,
+      "step": 18228
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.237963689236472e-05,
+      "loss": 0.9122,
+      "step": 18229
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2370795162292437e-05,
+      "loss": 0.9996,
+      "step": 18230
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2361954959184315e-05,
+      "loss": 1.0559,
+      "step": 18231
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.235311628321428e-05,
+      "loss": 0.9056,
+      "step": 18232
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2344279134556155e-05,
+      "loss": 0.9239,
+      "step": 18233
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2335443513383746e-05,
+      "loss": 0.9527,
+      "step": 18234
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2326609419870893e-05,
+      "loss": 0.7723,
+      "step": 18235
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.231777685419133e-05,
+      "loss": 0.9022,
+      "step": 18236
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.230894581651879e-05,
+      "loss": 0.9851,
+      "step": 18237
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2300116307027008e-05,
+      "loss": 0.9179,
+      "step": 18238
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2291288325889613e-05,
+      "loss": 0.9056,
+      "step": 18239
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2282461873280325e-05,
+      "loss": 0.8205,
+      "step": 18240
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2273636949372713e-05,
+      "loss": 1.0732,
+      "step": 18241
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2264813554340326e-05,
+      "loss": 0.9729,
+      "step": 18242
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.225599168835677e-05,
+      "loss": 0.9642,
+      "step": 18243
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2247171351595597e-05,
+      "loss": 0.9762,
+      "step": 18244
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2238352544230255e-05,
+      "loss": 1.0002,
+      "step": 18245
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.222953526643421e-05,
+      "loss": 0.9572,
+      "step": 18246
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.222071951838093e-05,
+      "loss": 0.9811,
+      "step": 18247
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2211905300243764e-05,
+      "loss": 0.8122,
+      "step": 18248
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2203092612196185e-05,
+      "loss": 0.8771,
+      "step": 18249
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2194281454411424e-05,
+      "loss": 0.9956,
+      "step": 18250
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2185471827062855e-05,
+      "loss": 0.7958,
+      "step": 18251
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2176663730323777e-05,
+      "loss": 0.8882,
+      "step": 18252
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2167857164367422e-05,
+      "loss": 0.7432,
+      "step": 18253
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2159052129366998e-05,
+      "loss": 0.8674,
+      "step": 18254
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.215024862549575e-05,
+      "loss": 0.8803,
+      "step": 18255
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.214144665292678e-05,
+      "loss": 1.0314,
+      "step": 18256
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2132646211833285e-05,
+      "loss": 0.8538,
+      "step": 18257
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2123847302388345e-05,
+      "loss": 0.8074,
+      "step": 18258
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2115049924764996e-05,
+      "loss": 1.0116,
+      "step": 18259
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2106254079136313e-05,
+      "loss": 1.0509,
+      "step": 18260
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2097459765675342e-05,
+      "loss": 0.9559,
+      "step": 18261
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.208866698455504e-05,
+      "loss": 0.8682,
+      "step": 18262
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.207987573594833e-05,
+      "loss": 1.0768,
+      "step": 18263
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2071086020028187e-05,
+      "loss": 0.9072,
+      "step": 18264
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.206229783696746e-05,
+      "loss": 0.9209,
+      "step": 18265
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.205351118693906e-05,
+      "loss": 1.0242,
+      "step": 18266
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2044726070115785e-05,
+      "loss": 0.953,
+      "step": 18267
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2035942486670425e-05,
+      "loss": 1.0226,
+      "step": 18268
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2027160436775794e-05,
+      "loss": 0.9785,
+      "step": 18269
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2018379920604625e-05,
+      "loss": 0.8289,
+      "step": 18270
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2009600938329587e-05,
+      "loss": 0.9285,
+      "step": 18271
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.2000823490123435e-05,
+      "loss": 0.8043,
+      "step": 18272
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1992047576158747e-05,
+      "loss": 0.9044,
+      "step": 18273
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.19832731966082e-05,
+      "loss": 1.1046,
+      "step": 18274
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1974500351644377e-05,
+      "loss": 0.944,
+      "step": 18275
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1965729041439798e-05,
+      "loss": 0.7508,
+      "step": 18276
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.195695926616702e-05,
+      "loss": 0.8938,
+      "step": 18277
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1948191025998578e-05,
+      "loss": 0.893,
+      "step": 18278
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1939424321106915e-05,
+      "loss": 0.9634,
+      "step": 18279
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1930659151664445e-05,
+      "loss": 0.9579,
+      "step": 18280
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.192189551784363e-05,
+      "loss": 0.7741,
+      "step": 18281
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1913133419816822e-05,
+      "loss": 1.1112,
+      "step": 18282
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.190437285775635e-05,
+      "loss": 0.9228,
+      "step": 18283
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1895613831834584e-05,
+      "loss": 0.9595,
+      "step": 18284
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.188685634222376e-05,
+      "loss": 0.9088,
+      "step": 18285
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.187810038909619e-05,
+      "loss": 0.9397,
+      "step": 18286
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1869345972624088e-05,
+      "loss": 0.9284,
+      "step": 18287
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.186059309297961e-05,
+      "loss": 0.8485,
+      "step": 18288
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1851841750334956e-05,
+      "loss": 0.8319,
+      "step": 18289
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1843091944862305e-05,
+      "loss": 0.9699,
+      "step": 18290
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1834343676733713e-05,
+      "loss": 0.9338,
+      "step": 18291
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1825596946121286e-05,
+      "loss": 1.0217,
+      "step": 18292
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.181685175319702e-05,
+      "loss": 1.022,
+      "step": 18293
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1808108098132986e-05,
+      "loss": 1.0748,
+      "step": 18294
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1799365981101205e-05,
+      "loss": 1.2077,
+      "step": 18295
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1790625402273524e-05,
+      "loss": 0.9185,
+      "step": 18296
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1781886361821945e-05,
+      "loss": 1.0522,
+      "step": 18297
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.177314885991837e-05,
+      "loss": 0.8611,
+      "step": 18298
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1764412896734653e-05,
+      "loss": 0.8238,
+      "step": 18299
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1755678472442586e-05,
+      "loss": 0.826,
+      "step": 18300
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.174694558721405e-05,
+      "loss": 0.866,
+      "step": 18301
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1738214241220756e-05,
+      "loss": 0.8932,
+      "step": 18302
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1729484434634505e-05,
+      "loss": 0.8903,
+      "step": 18303
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1720756167626987e-05,
+      "loss": 1.1503,
+      "step": 18304
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1712029440369852e-05,
+      "loss": 1.0846,
+      "step": 18305
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1703304253034783e-05,
+      "loss": 0.7619,
+      "step": 18306
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1694580605793437e-05,
+      "loss": 0.7444,
+      "step": 18307
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.168585849881738e-05,
+      "loss": 0.9342,
+      "step": 18308
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1677137932278147e-05,
+      "loss": 0.8521,
+      "step": 18309
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.166841890634732e-05,
+      "loss": 0.8502,
+      "step": 18310
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1659701421196367e-05,
+      "loss": 0.8125,
+      "step": 18311
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1650985476996788e-05,
+      "loss": 0.9216,
+      "step": 18312
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1642271073920018e-05,
+      "loss": 0.8547,
+      "step": 18313
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1633558212137427e-05,
+      "loss": 0.9718,
+      "step": 18314
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.162484689182046e-05,
+      "loss": 0.9779,
+      "step": 18315
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1616137113140457e-05,
+      "loss": 0.7947,
+      "step": 18316
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1607428876268675e-05,
+      "loss": 0.955,
+      "step": 18317
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1598722181376497e-05,
+      "loss": 1.0049,
+      "step": 18318
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1590017028635102e-05,
+      "loss": 1.0475,
+      "step": 18319
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1581313418215777e-05,
+      "loss": 0.9477,
+      "step": 18320
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1572611350289717e-05,
+      "loss": 0.7524,
+      "step": 18321
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1563910825028034e-05,
+      "loss": 0.9318,
+      "step": 18322
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1555211842601918e-05,
+      "loss": 0.7932,
+      "step": 18323
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1546514403182495e-05,
+      "loss": 0.8198,
+      "step": 18324
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.153781850694082e-05,
+      "loss": 0.9757,
+      "step": 18325
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.152912415404791e-05,
+      "loss": 1.0113,
+      "step": 18326
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.152043134467483e-05,
+      "loss": 0.8972,
+      "step": 18327
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.151174007899256e-05,
+      "loss": 0.8739,
+      "step": 18328
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1503050357172006e-05,
+      "loss": 1.0717,
+      "step": 18329
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1494362179384175e-05,
+      "loss": 0.999,
+      "step": 18330
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1485675545799877e-05,
+      "loss": 0.8897,
+      "step": 18331
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.147699045659006e-05,
+      "loss": 0.7825,
+      "step": 18332
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.146830691192553e-05,
+      "loss": 0.9682,
+      "step": 18333
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.145962491197705e-05,
+      "loss": 1.0399,
+      "step": 18334
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1450944456915422e-05,
+      "loss": 0.9518,
+      "step": 18335
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1442265546911434e-05,
+      "loss": 0.9669,
+      "step": 18336
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1433588182135755e-05,
+      "loss": 0.9672,
+      "step": 18337
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1424912362759087e-05,
+      "loss": 0.7639,
+      "step": 18338
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1416238088952047e-05,
+      "loss": 0.8067,
+      "step": 18339
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1407565360885274e-05,
+      "loss": 1.0634,
+      "step": 18340
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1398894178729423e-05,
+      "loss": 0.8777,
+      "step": 18341
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1390224542654957e-05,
+      "loss": 0.7856,
+      "step": 18342
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.138155645283244e-05,
+      "loss": 0.8198,
+      "step": 18343
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1372889909432424e-05,
+      "loss": 0.9388,
+      "step": 18344
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1364224912625343e-05,
+      "loss": 0.8953,
+      "step": 18345
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.135556146258161e-05,
+      "loss": 0.8561,
+      "step": 18346
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1346899559471677e-05,
+      "loss": 1.3147,
+      "step": 18347
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1338239203465894e-05,
+      "loss": 0.8481,
+      "step": 18348
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1329580394734638e-05,
+      "loss": 1.0571,
+      "step": 18349
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1320923133448222e-05,
+      "loss": 1.1179,
+      "step": 18350
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1312267419776886e-05,
+      "loss": 0.9342,
+      "step": 18351
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1303613253890942e-05,
+      "loss": 0.9256,
+      "step": 18352
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1294960635960613e-05,
+      "loss": 0.8773,
+      "step": 18353
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1286309566156105e-05,
+      "loss": 1.0028,
+      "step": 18354
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.127766004464752e-05,
+      "loss": 0.8757,
+      "step": 18355
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1269012071605067e-05,
+      "loss": 0.9943,
+      "step": 18356
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.1260365647198798e-05,
+      "loss": 1.1018,
+      "step": 18357
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.125172077159887e-05,
+      "loss": 0.9041,
+      "step": 18358
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2.12430774449752e-05,
+      "loss": 0.8785,
+      "step": 18359
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1234435667497887e-05,
+      "loss": 0.9382,
+      "step": 18360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1225795439336913e-05,
+      "loss": 1.0313,
+      "step": 18361
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1217156760662228e-05,
+      "loss": 1.0283,
+      "step": 18362
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1208519631643718e-05,
+      "loss": 0.8293,
+      "step": 18363
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1199884052451326e-05,
+      "loss": 0.9751,
+      "step": 18364
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.119125002325485e-05,
+      "loss": 0.9201,
+      "step": 18365
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1182617544224194e-05,
+      "loss": 0.9984,
+      "step": 18366
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1173986615529118e-05,
+      "loss": 0.7423,
+      "step": 18367
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.116535723733938e-05,
+      "loss": 0.7957,
+      "step": 18368
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1156729409824728e-05,
+      "loss": 0.9399,
+      "step": 18369
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1148103133154918e-05,
+      "loss": 0.9501,
+      "step": 18370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.113947840749958e-05,
+      "loss": 0.8306,
+      "step": 18371
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1130855233028357e-05,
+      "loss": 0.7184,
+      "step": 18372
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.11222336099109e-05,
+      "loss": 0.9682,
+      "step": 18373
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.111361353831679e-05,
+      "loss": 1.0533,
+      "step": 18374
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.110499501841554e-05,
+      "loss": 0.995,
+      "step": 18375
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.109637805037673e-05,
+      "loss": 0.908,
+      "step": 18376
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1087762634369813e-05,
+      "loss": 0.799,
+      "step": 18377
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1079148770564304e-05,
+      "loss": 0.8465,
+      "step": 18378
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1070536459129597e-05,
+      "loss": 0.718,
+      "step": 18379
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1061925700235098e-05,
+      "loss": 1.0548,
+      "step": 18380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1053316494050202e-05,
+      "loss": 1.0561,
+      "step": 18381
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.104470884074422e-05,
+      "loss": 1.002,
+      "step": 18382
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1036102740486508e-05,
+      "loss": 0.8573,
+      "step": 18383
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1027498193446326e-05,
+      "loss": 1.0315,
+      "step": 18384
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1018895199792886e-05,
+      "loss": 0.9609,
+      "step": 18385
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1010293759695453e-05,
+      "loss": 0.8505,
+      "step": 18386
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.1001693873323268e-05,
+      "loss": 1.1783,
+      "step": 18387
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0993095540845377e-05,
+      "loss": 0.8745,
+      "step": 18388
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.098449876243096e-05,
+      "loss": 0.9843,
+      "step": 18389
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.097590353824914e-05,
+      "loss": 0.8814,
+      "step": 18390
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0967309868468976e-05,
+      "loss": 0.8083,
+      "step": 18391
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.095871775325946e-05,
+      "loss": 0.9166,
+      "step": 18392
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.095012719278966e-05,
+      "loss": 0.7635,
+      "step": 18393
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0941538187228505e-05,
+      "loss": 0.8532,
+      "step": 18394
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0932950736744972e-05,
+      "loss": 0.8425,
+      "step": 18395
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0924364841507972e-05,
+      "loss": 1.0202,
+      "step": 18396
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0915780501686355e-05,
+      "loss": 1.1403,
+      "step": 18397
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0907197717449002e-05,
+      "loss": 1.0264,
+      "step": 18398
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.089861648896475e-05,
+      "loss": 1.2351,
+      "step": 18399
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0890036816402393e-05,
+      "loss": 0.913,
+      "step": 18400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0881458699930655e-05,
+      "loss": 0.9006,
+      "step": 18401
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0872882139718277e-05,
+      "loss": 0.8975,
+      "step": 18402
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.086430713593397e-05,
+      "loss": 0.8433,
+      "step": 18403
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0855733688746458e-05,
+      "loss": 0.9714,
+      "step": 18404
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0847161798324266e-05,
+      "loss": 0.9571,
+      "step": 18405
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.083859146483608e-05,
+      "loss": 0.9846,
+      "step": 18406
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.083002268845047e-05,
+      "loss": 0.8455,
+      "step": 18407
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0821455469335983e-05,
+      "loss": 0.7579,
+      "step": 18408
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0812889807661106e-05,
+      "loss": 0.9772,
+      "step": 18409
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0804325703594374e-05,
+      "loss": 0.9372,
+      "step": 18410
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0795763157304193e-05,
+      "loss": 1.0563,
+      "step": 18411
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.078720216895903e-05,
+      "loss": 1.0426,
+      "step": 18412
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0778642738727272e-05,
+      "loss": 0.935,
+      "step": 18413
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.077008486677724e-05,
+      "loss": 0.9073,
+      "step": 18414
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0761528553277297e-05,
+      "loss": 0.9194,
+      "step": 18415
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.075297379839578e-05,
+      "loss": 0.8985,
+      "step": 18416
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.074442060230093e-05,
+      "loss": 0.9074,
+      "step": 18417
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0735868965160953e-05,
+      "loss": 0.7314,
+      "step": 18418
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.072731888714412e-05,
+      "loss": 1.0467,
+      "step": 18419
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0718770368418593e-05,
+      "loss": 0.9715,
+      "step": 18420
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0710223409152475e-05,
+      "loss": 0.7996,
+      "step": 18421
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0701678009513957e-05,
+      "loss": 0.7412,
+      "step": 18422
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0693134169671057e-05,
+      "loss": 0.871,
+      "step": 18423
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0684591889791893e-05,
+      "loss": 0.9751,
+      "step": 18424
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0676051170044476e-05,
+      "loss": 0.7905,
+      "step": 18425
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0667512010596746e-05,
+      "loss": 1.0215,
+      "step": 18426
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.065897441161675e-05,
+      "loss": 0.8699,
+      "step": 18427
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.065043837327236e-05,
+      "loss": 0.7903,
+      "step": 18428
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.064190389573153e-05,
+      "loss": 0.9491,
+      "step": 18429
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0633370979162104e-05,
+      "loss": 0.9326,
+      "step": 18430
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0624839623731894e-05,
+      "loss": 0.8911,
+      "step": 18431
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.061630982960875e-05,
+      "loss": 0.8255,
+      "step": 18432
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0607781596960507e-05,
+      "loss": 0.8218,
+      "step": 18433
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0599254925954804e-05,
+      "loss": 0.9829,
+      "step": 18434
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0590729816759402e-05,
+      "loss": 0.7696,
+      "step": 18435
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0582206269542026e-05,
+      "loss": 0.9101,
+      "step": 18436
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.057368428447032e-05,
+      "loss": 0.8162,
+      "step": 18437
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0565163861711865e-05,
+      "loss": 0.8699,
+      "step": 18438
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0556645001434317e-05,
+      "loss": 0.9106,
+      "step": 18439
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0548127703805186e-05,
+      "loss": 0.8577,
+      "step": 18440
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0539611968992067e-05,
+      "loss": 0.959,
+      "step": 18441
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0531097797162445e-05,
+      "loss": 0.9719,
+      "step": 18442
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0522585188483745e-05,
+      "loss": 0.9451,
+      "step": 18443
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.051407414312345e-05,
+      "loss": 0.8207,
+      "step": 18444
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0505564661249e-05,
+      "loss": 0.947,
+      "step": 18445
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.049705674302773e-05,
+      "loss": 1.0722,
+      "step": 18446
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0488550388627026e-05,
+      "loss": 0.9706,
+      "step": 18447
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.048004559821416e-05,
+      "loss": 0.9622,
+      "step": 18448
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0471542371956443e-05,
+      "loss": 0.9279,
+      "step": 18449
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0463040710021187e-05,
+      "loss": 0.7976,
+      "step": 18450
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0454540612575523e-05,
+      "loss": 0.9761,
+      "step": 18451
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0446042079786697e-05,
+      "loss": 0.936,
+      "step": 18452
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.043754511182191e-05,
+      "loss": 0.8202,
+      "step": 18453
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0429049708848247e-05,
+      "loss": 0.8736,
+      "step": 18454
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.042055587103281e-05,
+      "loss": 0.8281,
+      "step": 18455
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.041206359854271e-05,
+      "loss": 0.8435,
+      "step": 18456
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.040357289154494e-05,
+      "loss": 1.1328,
+      "step": 18457
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.039508375020658e-05,
+      "loss": 0.9118,
+      "step": 18458
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.038659617469456e-05,
+      "loss": 0.8301,
+      "step": 18459
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0378110165175824e-05,
+      "loss": 0.9937,
+      "step": 18460
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.036962572181731e-05,
+      "loss": 1.024,
+      "step": 18461
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.036114284478593e-05,
+      "loss": 0.837,
+      "step": 18462
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0352661534248517e-05,
+      "loss": 1.0441,
+      "step": 18463
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0344181790371873e-05,
+      "loss": 0.9915,
+      "step": 18464
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0335703613322853e-05,
+      "loss": 0.8519,
+      "step": 18465
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.032722700326819e-05,
+      "loss": 0.7879,
+      "step": 18466
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0318751960374614e-05,
+      "loss": 1.05,
+      "step": 18467
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.031027848480881e-05,
+      "loss": 1.0411,
+      "step": 18468
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0301806576737482e-05,
+      "loss": 0.9493,
+      "step": 18469
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0293336236327288e-05,
+      "loss": 0.7936,
+      "step": 18470
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0284867463744815e-05,
+      "loss": 1.0117,
+      "step": 18471
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0276400259156613e-05,
+      "loss": 0.8424,
+      "step": 18472
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.02679346227293e-05,
+      "loss": 0.9237,
+      "step": 18473
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0259470554629333e-05,
+      "loss": 0.7785,
+      "step": 18474
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0251008055023245e-05,
+      "loss": 0.9176,
+      "step": 18475
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0242547124077472e-05,
+      "loss": 0.8635,
+      "step": 18476
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.023408776195842e-05,
+      "loss": 0.8626,
+      "step": 18477
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0225629968832515e-05,
+      "loss": 0.8542,
+      "step": 18478
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0217173744866157e-05,
+      "loss": 0.8556,
+      "step": 18479
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.020871909022558e-05,
+      "loss": 0.8447,
+      "step": 18480
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.020026600507715e-05,
+      "loss": 0.861,
+      "step": 18481
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0191814489587158e-05,
+      "loss": 0.9187,
+      "step": 18482
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0183364543921813e-05,
+      "loss": 0.7815,
+      "step": 18483
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0174916168247306e-05,
+      "loss": 1.03,
+      "step": 18484
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.016646936272987e-05,
+      "loss": 1.0298,
+      "step": 18485
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.015802412753559e-05,
+      "loss": 0.7289,
+      "step": 18486
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0149580462830654e-05,
+      "loss": 1.0059,
+      "step": 18487
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.014113836878111e-05,
+      "loss": 0.8916,
+      "step": 18488
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0132697845552984e-05,
+      "loss": 0.9652,
+      "step": 18489
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0124258893312366e-05,
+      "loss": 0.9285,
+      "step": 18490
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0115821512225186e-05,
+      "loss": 0.8338,
+      "step": 18491
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.010738570245747e-05,
+      "loss": 0.8246,
+      "step": 18492
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.009895146417512e-05,
+      "loss": 0.7446,
+      "step": 18493
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0090518797544e-05,
+      "loss": 0.9047,
+      "step": 18494
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0082087702730022e-05,
+      "loss": 0.9539,
+      "step": 18495
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.007365817989907e-05,
+      "loss": 0.9738,
+      "step": 18496
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0065230229216847e-05,
+      "loss": 0.9512,
+      "step": 18497
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0056803850849192e-05,
+      "loss": 0.9564,
+      "step": 18498
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0048379044961873e-05,
+      "loss": 0.7113,
+      "step": 18499
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.003995581172057e-05,
+      "loss": 0.8922,
+      "step": 18500
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0031534151290943e-05,
+      "loss": 0.8949,
+      "step": 18501
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.002311406383872e-05,
+      "loss": 1.0313,
+      "step": 18502
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.001469554952945e-05,
+      "loss": 0.99,
+      "step": 18503
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2.0006278608528773e-05,
+      "loss": 0.9854,
+      "step": 18504
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9997863241002245e-05,
+      "loss": 0.9755,
+      "step": 18505
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.998944944711535e-05,
+      "loss": 0.8437,
+      "step": 18506
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9981037227033627e-05,
+      "loss": 0.7795,
+      "step": 18507
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.997262658092257e-05,
+      "loss": 1.041,
+      "step": 18508
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.996421750894758e-05,
+      "loss": 0.9154,
+      "step": 18509
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9955810011274046e-05,
+      "loss": 0.8898,
+      "step": 18510
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.994740408806739e-05,
+      "loss": 0.8334,
+      "step": 18511
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9938999739492927e-05,
+      "loss": 0.8878,
+      "step": 18512
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9930596965715986e-05,
+      "loss": 0.8166,
+      "step": 18513
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9922195766901798e-05,
+      "loss": 0.9053,
+      "step": 18514
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9913796143215668e-05,
+      "loss": 0.7441,
+      "step": 18515
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9905398094822823e-05,
+      "loss": 0.9624,
+      "step": 18516
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9897001621888434e-05,
+      "loss": 0.8423,
+      "step": 18517
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.988860672457763e-05,
+      "loss": 0.9952,
+      "step": 18518
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9880213403055592e-05,
+      "loss": 0.8783,
+      "step": 18519
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9871821657487366e-05,
+      "loss": 0.921,
+      "step": 18520
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9863431488038065e-05,
+      "loss": 0.894,
+      "step": 18521
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.985504289487271e-05,
+      "loss": 0.8863,
+      "step": 18522
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9846655878156263e-05,
+      "loss": 0.902,
+      "step": 18523
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.983827043805373e-05,
+      "loss": 1.0008,
+      "step": 18524
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9829886574730093e-05,
+      "loss": 0.8806,
+      "step": 18525
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.982150428835018e-05,
+      "loss": 1.0617,
+      "step": 18526
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9813123579078907e-05,
+      "loss": 0.7988,
+      "step": 18527
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9804744447081148e-05,
+      "loss": 0.9446,
+      "step": 18528
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.97963668925217e-05,
+      "loss": 1.0649,
+      "step": 18529
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9787990915565323e-05,
+      "loss": 0.8064,
+      "step": 18530
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.977961651637682e-05,
+      "loss": 0.9879,
+      "step": 18531
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9771243695120867e-05,
+      "loss": 0.7867,
+      "step": 18532
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9762872451962212e-05,
+      "loss": 1.1712,
+      "step": 18533
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9754502787065475e-05,
+      "loss": 1.0196,
+      "step": 18534
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9746134700595286e-05,
+      "loss": 1.18,
+      "step": 18535
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9737768192716277e-05,
+      "loss": 0.7972,
+      "step": 18536
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9729403263592973e-05,
+      "loss": 1.098,
+      "step": 18537
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.972103991338996e-05,
+      "loss": 0.8057,
+      "step": 18538
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9712678142271725e-05,
+      "loss": 0.9251,
+      "step": 18539
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9704317950402707e-05,
+      "loss": 0.8482,
+      "step": 18540
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.969595933794739e-05,
+      "loss": 0.9158,
+      "step": 18541
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9687602305070242e-05,
+      "loss": 0.9867,
+      "step": 18542
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.967924685193552e-05,
+      "loss": 0.9026,
+      "step": 18543
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9670892978707644e-05,
+      "loss": 0.9239,
+      "step": 18544
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9662540685550957e-05,
+      "loss": 0.8216,
+      "step": 18545
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9654189972629735e-05,
+      "loss": 0.651,
+      "step": 18546
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9645840840108175e-05,
+      "loss": 0.8765,
+      "step": 18547
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.96374932881506e-05,
+      "loss": 0.75,
+      "step": 18548
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9629147316921125e-05,
+      "loss": 1.0759,
+      "step": 18549
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.962080292658397e-05,
+      "loss": 0.9077,
+      "step": 18550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9612460117303255e-05,
+      "loss": 0.8374,
+      "step": 18551
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9604118889243052e-05,
+      "loss": 0.8657,
+      "step": 18552
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9595779242567446e-05,
+      "loss": 0.7669,
+      "step": 18553
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9587441177440525e-05,
+      "loss": 0.8874,
+      "step": 18554
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9579104694026262e-05,
+      "loss": 0.7314,
+      "step": 18555
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.957076979248863e-05,
+      "loss": 1.0675,
+      "step": 18556
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9562436472991552e-05,
+      "loss": 0.8814,
+      "step": 18557
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.955410473569901e-05,
+      "loss": 1.0113,
+      "step": 18558
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9545774580774846e-05,
+      "loss": 0.7571,
+      "step": 18559
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9537446008382898e-05,
+      "loss": 0.8999,
+      "step": 18560
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9529119018687005e-05,
+      "loss": 0.9105,
+      "step": 18561
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9520793611850996e-05,
+      "loss": 0.8374,
+      "step": 18562
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9512469788038602e-05,
+      "loss": 0.9197,
+      "step": 18563
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.950414754741352e-05,
+      "loss": 0.7494,
+      "step": 18564
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9495826890139503e-05,
+      "loss": 0.8554,
+      "step": 18565
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.948750781638017e-05,
+      "loss": 0.9417,
+      "step": 18566
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9479190326299213e-05,
+      "loss": 0.9649,
+      "step": 18567
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9470874420060202e-05,
+      "loss": 0.7903,
+      "step": 18568
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9462560097826687e-05,
+      "loss": 0.9073,
+      "step": 18569
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9454247359762234e-05,
+      "loss": 0.7758,
+      "step": 18570
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9445936206030412e-05,
+      "loss": 1.0416,
+      "step": 18571
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9437626636794592e-05,
+      "loss": 1.0902,
+      "step": 18572
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9429318652218276e-05,
+      "loss": 0.7317,
+      "step": 18573
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.942101225246491e-05,
+      "loss": 0.8869,
+      "step": 18574
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9412707437697862e-05,
+      "loss": 1.0506,
+      "step": 18575
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9404404208080475e-05,
+      "loss": 0.8708,
+      "step": 18576
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.939610256377604e-05,
+      "loss": 0.9703,
+      "step": 18577
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9387802504947903e-05,
+      "loss": 1.0277,
+      "step": 18578
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.937950403175933e-05,
+      "loss": 0.7813,
+      "step": 18579
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9371207144373538e-05,
+      "loss": 0.808,
+      "step": 18580
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9362911842953678e-05,
+      "loss": 0.9427,
+      "step": 18581
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9354618127663005e-05,
+      "loss": 0.7398,
+      "step": 18582
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9346325998664584e-05,
+      "loss": 0.9037,
+      "step": 18583
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9338035456121573e-05,
+      "loss": 1.3102,
+      "step": 18584
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9329746500197023e-05,
+      "loss": 0.7291,
+      "step": 18585
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9321459131053943e-05,
+      "loss": 0.8177,
+      "step": 18586
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9313173348855384e-05,
+      "loss": 0.7989,
+      "step": 18587
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9304889153764383e-05,
+      "loss": 0.9092,
+      "step": 18588
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9296606545943774e-05,
+      "loss": 0.8576,
+      "step": 18589
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.928832552555653e-05,
+      "loss": 0.8793,
+      "step": 18590
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.928004609276556e-05,
+      "loss": 0.9453,
+      "step": 18591
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.92717682477337e-05,
+      "loss": 0.933,
+      "step": 18592
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.926349199062376e-05,
+      "loss": 0.9068,
+      "step": 18593
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9255217321598574e-05,
+      "loss": 0.9956,
+      "step": 18594
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9246944240820843e-05,
+      "loss": 0.8596,
+      "step": 18595
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.923867274845337e-05,
+      "loss": 0.8745,
+      "step": 18596
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9230402844658812e-05,
+      "loss": 0.9054,
+      "step": 18597
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.922213452959982e-05,
+      "loss": 0.7464,
+      "step": 18598
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.921386780343909e-05,
+      "loss": 0.9071,
+      "step": 18599
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9205602666339163e-05,
+      "loss": 1.0383,
+      "step": 18600
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9197339118462677e-05,
+      "loss": 1.0628,
+      "step": 18601
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9189077159972136e-05,
+      "loss": 0.8824,
+      "step": 18602
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9180816791030044e-05,
+      "loss": 0.768,
+      "step": 18603
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9172558011798913e-05,
+      "loss": 0.8638,
+      "step": 18604
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9164300822441194e-05,
+      "loss": 0.8681,
+      "step": 18605
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9156045223119257e-05,
+      "loss": 0.992,
+      "step": 18606
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9147791213995524e-05,
+      "loss": 0.8769,
+      "step": 18607
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9139538795232382e-05,
+      "loss": 1.0077,
+      "step": 18608
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9131287966992117e-05,
+      "loss": 0.8791,
+      "step": 18609
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9123038729437005e-05,
+      "loss": 1.0917,
+      "step": 18610
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9114791082729366e-05,
+      "loss": 0.7538,
+      "step": 18611
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9106545027031363e-05,
+      "loss": 1.0776,
+      "step": 18612
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9098300562505266e-05,
+      "loss": 0.835,
+      "step": 18613
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.90900576893132e-05,
+      "loss": 0.9521,
+      "step": 18614
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9081816407617294e-05,
+      "loss": 0.78,
+      "step": 18615
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9073576717579656e-05,
+      "loss": 0.7424,
+      "step": 18616
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9065338619362437e-05,
+      "loss": 0.8958,
+      "step": 18617
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.905710211312757e-05,
+      "loss": 0.7831,
+      "step": 18618
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9048867199037103e-05,
+      "loss": 1.0462,
+      "step": 18619
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9040633877253056e-05,
+      "loss": 0.8932,
+      "step": 18620
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9032402147937357e-05,
+      "loss": 0.9423,
+      "step": 18621
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9024172011251908e-05,
+      "loss": 0.9568,
+      "step": 18622
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9015943467358587e-05,
+      "loss": 1.025,
+      "step": 18623
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.9007716516419262e-05,
+      "loss": 0.9121,
+      "step": 18624
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8999491158595795e-05,
+      "loss": 0.8558,
+      "step": 18625
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8991267394049938e-05,
+      "loss": 0.8176,
+      "step": 18626
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.898304522294343e-05,
+      "loss": 1.0513,
+      "step": 18627
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.897482464543806e-05,
+      "loss": 0.7338,
+      "step": 18628
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.896660566169548e-05,
+      "loss": 0.9834,
+      "step": 18629
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.895838827187739e-05,
+      "loss": 0.8098,
+      "step": 18630
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8950172476145413e-05,
+      "loss": 0.8832,
+      "step": 18631
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.894195827466112e-05,
+      "loss": 1.1147,
+      "step": 18632
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.893374566758611e-05,
+      "loss": 0.8801,
+      "step": 18633
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8925534655082e-05,
+      "loss": 1.0174,
+      "step": 18634
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8917325237310156e-05,
+      "loss": 0.8256,
+      "step": 18635
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.890911741443214e-05,
+      "loss": 0.7516,
+      "step": 18636
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8900911186609426e-05,
+      "loss": 0.9203,
+      "step": 18637
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8892706554003382e-05,
+      "loss": 1.0068,
+      "step": 18638
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8884503516775387e-05,
+      "loss": 0.9639,
+      "step": 18639
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8876302075086828e-05,
+      "loss": 0.7853,
+      "step": 18640
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8868102229098995e-05,
+      "loss": 0.8101,
+      "step": 18641
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8859903978973213e-05,
+      "loss": 0.9313,
+      "step": 18642
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.885170732487074e-05,
+      "loss": 0.7797,
+      "step": 18643
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8843512266952746e-05,
+      "loss": 1.137,
+      "step": 18644
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8835318805380507e-05,
+      "loss": 0.8697,
+      "step": 18645
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.882712694031512e-05,
+      "loss": 0.8612,
+      "step": 18646
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8818936671917775e-05,
+      "loss": 0.9843,
+      "step": 18647
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8810748000349544e-05,
+      "loss": 1.0149,
+      "step": 18648
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8802560925771482e-05,
+      "loss": 0.8913,
+      "step": 18649
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.879437544834467e-05,
+      "loss": 1.0252,
+      "step": 18650
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8786191568230106e-05,
+      "loss": 1.1164,
+      "step": 18651
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8778009285588715e-05,
+      "loss": 0.9531,
+      "step": 18652
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8769828600581496e-05,
+      "loss": 0.9287,
+      "step": 18653
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8761649513369374e-05,
+      "loss": 0.7951,
+      "step": 18654
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8753472024113194e-05,
+      "loss": 0.8302,
+      "step": 18655
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8745296132973812e-05,
+      "loss": 1.0022,
+      "step": 18656
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.873712184011207e-05,
+      "loss": 0.8884,
+      "step": 18657
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8728949145688722e-05,
+      "loss": 1.1734,
+      "step": 18658
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8720778049864574e-05,
+      "loss": 0.7235,
+      "step": 18659
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8712608552800315e-05,
+      "loss": 0.8052,
+      "step": 18660
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.870444065465662e-05,
+      "loss": 1.0181,
+      "step": 18661
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.869627435559418e-05,
+      "loss": 1.0287,
+      "step": 18662
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8688109655773676e-05,
+      "loss": 0.8931,
+      "step": 18663
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.86799465553556e-05,
+      "loss": 0.7571,
+      "step": 18664
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8671785054500602e-05,
+      "loss": 0.9324,
+      "step": 18665
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8663625153369156e-05,
+      "loss": 0.859,
+      "step": 18666
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8655466852121838e-05,
+      "loss": 1.1315,
+      "step": 18667
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8647310150919083e-05,
+      "loss": 1.0148,
+      "step": 18668
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.863915504992131e-05,
+      "loss": 0.8187,
+      "step": 18669
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.863100154928895e-05,
+      "loss": 0.7532,
+      "step": 18670
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8622849649182417e-05,
+      "loss": 1.0244,
+      "step": 18671
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8614699349762022e-05,
+      "loss": 0.7583,
+      "step": 18672
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.860655065118807e-05,
+      "loss": 0.9817,
+      "step": 18673
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.859840355362089e-05,
+      "loss": 0.9201,
+      "step": 18674
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8590258057220676e-05,
+      "loss": 0.8848,
+      "step": 18675
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.858211416214771e-05,
+      "loss": 0.9491,
+      "step": 18676
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8573971868562156e-05,
+      "loss": 0.7841,
+      "step": 18677
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.856583117662415e-05,
+      "loss": 1.0037,
+      "step": 18678
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.855769208649384e-05,
+      "loss": 1.012,
+      "step": 18679
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8549554598331365e-05,
+      "loss": 0.851,
+      "step": 18680
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8541418712296698e-05,
+      "loss": 1.0702,
+      "step": 18681
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8533284428549923e-05,
+      "loss": 0.8998,
+      "step": 18682
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8525151747251056e-05,
+      "loss": 0.8428,
+      "step": 18683
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.851702066856006e-05,
+      "loss": 0.8743,
+      "step": 18684
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8508891192636824e-05,
+      "loss": 0.7556,
+      "step": 18685
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8500763319641323e-05,
+      "loss": 1.0961,
+      "step": 18686
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8492637049733374e-05,
+      "loss": 0.9836,
+      "step": 18687
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.848451238307288e-05,
+      "loss": 1.0489,
+      "step": 18688
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8476389319819622e-05,
+      "loss": 0.9142,
+      "step": 18689
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8468267860133335e-05,
+      "loss": 1.0393,
+      "step": 18690
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.846014800417386e-05,
+      "loss": 0.7021,
+      "step": 18691
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.845202975210083e-05,
+      "loss": 0.774,
+      "step": 18692
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8443913104073983e-05,
+      "loss": 0.9434,
+      "step": 18693
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8435798060252972e-05,
+      "loss": 0.8346,
+      "step": 18694
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8427684620797382e-05,
+      "loss": 0.9057,
+      "step": 18695
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8419572785866834e-05,
+      "loss": 0.9173,
+      "step": 18696
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8411462555620895e-05,
+      "loss": 1.0179,
+      "step": 18697
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.840335393021905e-05,
+      "loss": 0.9636,
+      "step": 18698
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.839524690982082e-05,
+      "loss": 0.938,
+      "step": 18699
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8387141494585703e-05,
+      "loss": 0.9509,
+      "step": 18700
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8379037684673105e-05,
+      "loss": 0.8337,
+      "step": 18701
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8370935480242402e-05,
+      "loss": 0.7955,
+      "step": 18702
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8362834881453005e-05,
+      "loss": 1.0381,
+      "step": 18703
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8354735888464213e-05,
+      "loss": 0.9432,
+      "step": 18704
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8346638501435386e-05,
+      "loss": 1.0778,
+      "step": 18705
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8338542720525763e-05,
+      "loss": 0.8343,
+      "step": 18706
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8330448545894575e-05,
+      "loss": 0.9125,
+      "step": 18707
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8322355977701067e-05,
+      "loss": 1.0342,
+      "step": 18708
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8314265016104417e-05,
+      "loss": 0.9743,
+      "step": 18709
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8306175661263724e-05,
+      "loss": 0.8253,
+      "step": 18710
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.829808791333818e-05,
+      "loss": 0.9615,
+      "step": 18711
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.829000177248681e-05,
+      "loss": 0.8032,
+      "step": 18712
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8281917238868707e-05,
+      "loss": 1.0462,
+      "step": 18713
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.827383431264289e-05,
+      "loss": 0.9274,
+      "step": 18714
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.826575299396831e-05,
+      "loss": 0.9647,
+      "step": 18715
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8257673283003953e-05,
+      "loss": 0.84,
+      "step": 18716
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8249595179908773e-05,
+      "loss": 0.877,
+      "step": 18717
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.824151868484164e-05,
+      "loss": 0.9723,
+      "step": 18718
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8233443797961403e-05,
+      "loss": 0.7983,
+      "step": 18719
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.822537051942693e-05,
+      "loss": 0.9351,
+      "step": 18720
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.821729884939699e-05,
+      "loss": 0.8067,
+      "step": 18721
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8209228788030386e-05,
+      "loss": 0.9026,
+      "step": 18722
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8201160335485835e-05,
+      "loss": 1.1581,
+      "step": 18723
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.819309349192202e-05,
+      "loss": 0.8619,
+      "step": 18724
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.818502825749764e-05,
+      "loss": 0.7693,
+      "step": 18725
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8176964632371397e-05,
+      "loss": 1.1015,
+      "step": 18726
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.816890261670179e-05,
+      "loss": 0.94,
+      "step": 18727
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8160842210647443e-05,
+      "loss": 0.8137,
+      "step": 18728
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8152783414366946e-05,
+      "loss": 1.0156,
+      "step": 18729
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8144726228018783e-05,
+      "loss": 0.9454,
+      "step": 18730
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8136670651761434e-05,
+      "loss": 0.9609,
+      "step": 18731
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.812861668575333e-05,
+      "loss": 1.0582,
+      "step": 18732
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8120564330152923e-05,
+      "loss": 0.8559,
+      "step": 18733
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.811251358511863e-05,
+      "loss": 0.8788,
+      "step": 18734
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8104464450808767e-05,
+      "loss": 0.8251,
+      "step": 18735
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8096416927381644e-05,
+      "loss": 0.829,
+      "step": 18736
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.808837101499561e-05,
+      "loss": 0.9321,
+      "step": 18737
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.808032671380887e-05,
+      "loss": 0.7974,
+      "step": 18738
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8072284023979726e-05,
+      "loss": 0.9025,
+      "step": 18739
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.806424294566632e-05,
+      "loss": 0.8526,
+      "step": 18740
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.805620347902681e-05,
+      "loss": 0.8656,
+      "step": 18741
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8048165624219394e-05,
+      "loss": 1.1507,
+      "step": 18742
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8040129381402137e-05,
+      "loss": 1.0476,
+      "step": 18743
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8032094750733087e-05,
+      "loss": 0.9208,
+      "step": 18744
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8024061732370313e-05,
+      "loss": 0.8091,
+      "step": 18745
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8016030326471845e-05,
+      "loss": 0.7829,
+      "step": 18746
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.8008000533195647e-05,
+      "loss": 0.782,
+      "step": 18747
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7999972352699634e-05,
+      "loss": 0.9664,
+      "step": 18748
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.799194578514176e-05,
+      "loss": 0.8915,
+      "step": 18749
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.798392083067988e-05,
+      "loss": 1.0417,
+      "step": 18750
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.797589748947187e-05,
+      "loss": 1.1333,
+      "step": 18751
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7967875761675536e-05,
+      "loss": 0.9941,
+      "step": 18752
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.795985564744864e-05,
+      "loss": 1.0745,
+      "step": 18753
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.795183714694898e-05,
+      "loss": 1.0019,
+      "step": 18754
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7943820260334276e-05,
+      "loss": 0.8103,
+      "step": 18755
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7935804987762162e-05,
+      "loss": 0.9004,
+      "step": 18756
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7927791329390375e-05,
+      "loss": 1.0774,
+      "step": 18757
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7919779285376483e-05,
+      "loss": 1.1255,
+      "step": 18758
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7911768855878142e-05,
+      "loss": 0.9269,
+      "step": 18759
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7903760041052875e-05,
+      "loss": 0.964,
+      "step": 18760
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.78957528410582e-05,
+      "loss": 0.8693,
+      "step": 18761
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7887747256051636e-05,
+      "loss": 0.967,
+      "step": 18762
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7879743286190686e-05,
+      "loss": 0.9706,
+      "step": 18763
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7871740931632775e-05,
+      "loss": 1.0682,
+      "step": 18764
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.786374019253526e-05,
+      "loss": 0.7378,
+      "step": 18765
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7855741069055575e-05,
+      "loss": 1.0679,
+      "step": 18766
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7847743561351004e-05,
+      "loss": 1.0232,
+      "step": 18767
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.783974766957893e-05,
+      "loss": 0.9442,
+      "step": 18768
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7831753393896588e-05,
+      "loss": 0.9469,
+      "step": 18769
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7823760734461202e-05,
+      "loss": 0.9394,
+      "step": 18770
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.781576969143003e-05,
+      "loss": 1.0556,
+      "step": 18771
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7807780264960282e-05,
+      "loss": 1.0382,
+      "step": 18772
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7799792455209018e-05,
+      "loss": 0.778,
+      "step": 18773
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7791806262333433e-05,
+      "loss": 1.0088,
+      "step": 18774
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7783821686490576e-05,
+      "loss": 1.0619,
+      "step": 18775
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7775838727837537e-05,
+      "loss": 0.8656,
+      "step": 18776
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7767857386531317e-05,
+      "loss": 0.9472,
+      "step": 18777
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7759877662728898e-05,
+      "loss": 0.8825,
+      "step": 18778
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.775189955658725e-05,
+      "loss": 1.0469,
+      "step": 18779
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7743923068263345e-05,
+      "loss": 0.8998,
+      "step": 18780
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.773594819791403e-05,
+      "loss": 1.094,
+      "step": 18781
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7727974945696167e-05,
+      "loss": 0.7759,
+      "step": 18782
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7720003311766642e-05,
+      "loss": 1.0652,
+      "step": 18783
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7712033296282184e-05,
+      "loss": 0.9869,
+      "step": 18784
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.770406489939963e-05,
+      "loss": 0.8947,
+      "step": 18785
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7696098121275693e-05,
+      "loss": 0.8366,
+      "step": 18786
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.768813296206705e-05,
+      "loss": 1.0804,
+      "step": 18787
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7680169421930404e-05,
+      "loss": 0.8467,
+      "step": 18788
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.767220750102243e-05,
+      "loss": 1.0082,
+      "step": 18789
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7664247199499672e-05,
+      "loss": 0.8438,
+      "step": 18790
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7656288517518726e-05,
+      "loss": 0.8245,
+      "step": 18791
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.764833145523619e-05,
+      "loss": 1.0321,
+      "step": 18792
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7640376012808536e-05,
+      "loss": 0.8988,
+      "step": 18793
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7632422190392238e-05,
+      "loss": 0.8203,
+      "step": 18794
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7624469988143777e-05,
+      "loss": 0.6906,
+      "step": 18795
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7616519406219544e-05,
+      "loss": 0.9515,
+      "step": 18796
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.760857044477596e-05,
+      "loss": 0.8129,
+      "step": 18797
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.760062310396936e-05,
+      "loss": 0.9984,
+      "step": 18798
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7592677383956056e-05,
+      "loss": 0.7918,
+      "step": 18799
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7584733284892374e-05,
+      "loss": 0.8771,
+      "step": 18800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.757679080693455e-05,
+      "loss": 1.0284,
+      "step": 18801
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7568849950238797e-05,
+      "loss": 0.9548,
+      "step": 18802
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7560910714961355e-05,
+      "loss": 0.9828,
+      "step": 18803
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7552973101258343e-05,
+      "loss": 0.9393,
+      "step": 18804
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7545037109285946e-05,
+      "loss": 0.8073,
+      "step": 18805
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7537102739200216e-05,
+      "loss": 1.0966,
+      "step": 18806
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7529169991157236e-05,
+      "loss": 0.8595,
+      "step": 18807
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7521238865313038e-05,
+      "loss": 0.7244,
+      "step": 18808
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.751330936182366e-05,
+      "loss": 0.7877,
+      "step": 18809
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.750538148084505e-05,
+      "loss": 0.9179,
+      "step": 18810
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7497455222533133e-05,
+      "loss": 0.704,
+      "step": 18811
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7489530587043856e-05,
+      "loss": 1.0468,
+      "step": 18812
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7481607574533044e-05,
+      "loss": 0.9621,
+      "step": 18813
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.747368618515661e-05,
+      "loss": 0.7743,
+      "step": 18814
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7465766419070328e-05,
+      "loss": 0.9786,
+      "step": 18815
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7457848276429957e-05,
+      "loss": 0.8042,
+      "step": 18816
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7449931757391303e-05,
+      "loss": 0.8477,
+      "step": 18817
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.7442016862110056e-05,
+      "loss": 0.9095,
+      "step": 18818
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.743410359074187e-05,
+      "loss": 0.9352,
+      "step": 18819
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.742619194344245e-05,
+      "loss": 1.0971,
+      "step": 18820
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7418281920367373e-05,
+      "loss": 0.9322,
+      "step": 18821
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7410373521672286e-05,
+      "loss": 0.8724,
+      "step": 18822
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7402466747512704e-05,
+      "loss": 0.9151,
+      "step": 18823
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7394561598044134e-05,
+      "loss": 0.9357,
+      "step": 18824
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7386658073422113e-05,
+      "loss": 0.814,
+      "step": 18825
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.73787561738021e-05,
+      "loss": 1.0884,
+      "step": 18826
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7370855899339522e-05,
+      "loss": 0.822,
+      "step": 18827
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7362957250189736e-05,
+      "loss": 0.8931,
+      "step": 18828
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.735506022650817e-05,
+      "loss": 0.91,
+      "step": 18829
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7347164828450102e-05,
+      "loss": 1.1652,
+      "step": 18830
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.733927105617089e-05,
+      "loss": 0.9364,
+      "step": 18831
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.733137890982578e-05,
+      "loss": 0.9345,
+      "step": 18832
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7323488389569976e-05,
+      "loss": 1.0569,
+      "step": 18833
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.731559949555871e-05,
+      "loss": 0.8359,
+      "step": 18834
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7307712227947215e-05,
+      "loss": 0.9265,
+      "step": 18835
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7299826586890523e-05,
+      "loss": 0.9343,
+      "step": 18836
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7291942572543807e-05,
+      "loss": 0.796,
+      "step": 18837
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7284060185062167e-05,
+      "loss": 0.8878,
+      "step": 18838
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7276179424600613e-05,
+      "loss": 0.7956,
+      "step": 18839
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7268300291314165e-05,
+      "loss": 0.8335,
+      "step": 18840
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7260422785357787e-05,
+      "loss": 0.9939,
+      "step": 18841
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7252546906886457e-05,
+      "loss": 0.901,
+      "step": 18842
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7244672656055106e-05,
+      "loss": 0.9672,
+      "step": 18843
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7236800033018595e-05,
+      "loss": 0.8714,
+      "step": 18844
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.722892903793176e-05,
+      "loss": 0.8671,
+      "step": 18845
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.722105967094948e-05,
+      "loss": 0.9313,
+      "step": 18846
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7213191932226503e-05,
+      "loss": 1.0398,
+      "step": 18847
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7205325821917574e-05,
+      "loss": 0.9451,
+      "step": 18848
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7197461340177467e-05,
+      "loss": 0.8007,
+      "step": 18849
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.718959848716083e-05,
+      "loss": 0.8626,
+      "step": 18850
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7181737263022358e-05,
+      "loss": 0.9301,
+      "step": 18851
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7173877667916682e-05,
+      "loss": 0.9353,
+      "step": 18852
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.716601970199836e-05,
+      "loss": 0.8483,
+      "step": 18853
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7158163365421977e-05,
+      "loss": 0.9041,
+      "step": 18854
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7150308658342107e-05,
+      "loss": 0.9238,
+      "step": 18855
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7142455580913207e-05,
+      "loss": 0.8157,
+      "step": 18856
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7134604133289746e-05,
+      "loss": 1.0938,
+      "step": 18857
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7126754315626202e-05,
+      "loss": 0.8882,
+      "step": 18858
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7118906128076927e-05,
+      "loss": 0.7673,
+      "step": 18859
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.711105957079634e-05,
+      "loss": 0.9511,
+      "step": 18860
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.710321464393877e-05,
+      "loss": 0.9999,
+      "step": 18861
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7095371347658496e-05,
+      "loss": 0.8201,
+      "step": 18862
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7087529682109837e-05,
+      "loss": 0.9084,
+      "step": 18863
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.707968964744703e-05,
+      "loss": 0.8553,
+      "step": 18864
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7071851243824245e-05,
+      "loss": 1.0469,
+      "step": 18865
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.706401447139573e-05,
+      "loss": 0.9335,
+      "step": 18866
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.705617933031557e-05,
+      "loss": 0.8007,
+      "step": 18867
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.704834582073794e-05,
+      "loss": 0.9437,
+      "step": 18868
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7040513942816906e-05,
+      "loss": 0.9462,
+      "step": 18869
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7032683696706487e-05,
+      "loss": 1.0683,
+      "step": 18870
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.702485508256072e-05,
+      "loss": 0.9003,
+      "step": 18871
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.701702810053364e-05,
+      "loss": 0.9495,
+      "step": 18872
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.7009202750779164e-05,
+      "loss": 0.9544,
+      "step": 18873
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.700137903345118e-05,
+      "loss": 0.9407,
+      "step": 18874
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.699355694870366e-05,
+      "loss": 0.7219,
+      "step": 18875
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6985736496690397e-05,
+      "loss": 0.752,
+      "step": 18876
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.697791767756527e-05,
+      "loss": 0.9778,
+      "step": 18877
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6970100491482054e-05,
+      "loss": 0.6893,
+      "step": 18878
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6962284938594486e-05,
+      "loss": 0.9034,
+      "step": 18879
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6954471019056316e-05,
+      "loss": 0.9014,
+      "step": 18880
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6946658733021293e-05,
+      "loss": 1.041,
+      "step": 18881
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6938848080643e-05,
+      "loss": 0.9036,
+      "step": 18882
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.693103906207514e-05,
+      "loss": 1.0664,
+      "step": 18883
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.692323167747125e-05,
+      "loss": 0.9283,
+      "step": 18884
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6915425926984986e-05,
+      "loss": 0.9503,
+      "step": 18885
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6907621810769825e-05,
+      "loss": 0.8872,
+      "step": 18886
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.689981932897927e-05,
+      "loss": 0.7254,
+      "step": 18887
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6892018481766814e-05,
+      "loss": 1.0593,
+      "step": 18888
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6884219269285928e-05,
+      "loss": 0.7602,
+      "step": 18889
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6876421691689992e-05,
+      "loss": 0.822,
+      "step": 18890
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6868625749132362e-05,
+      "loss": 0.86,
+      "step": 18891
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6860831441766433e-05,
+      "loss": 0.8793,
+      "step": 18892
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6853038769745467e-05,
+      "loss": 0.8997,
+      "step": 18893
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6845247733222803e-05,
+      "loss": 0.9324,
+      "step": 18894
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6837458332351664e-05,
+      "loss": 0.7617,
+      "step": 18895
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.682967056728524e-05,
+      "loss": 1.1642,
+      "step": 18896
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.682188443817676e-05,
+      "loss": 1.0617,
+      "step": 18897
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6814099945179352e-05,
+      "loss": 0.8755,
+      "step": 18898
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6806317088446122e-05,
+      "loss": 1.0152,
+      "step": 18899
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6798535868130184e-05,
+      "loss": 1.0675,
+      "step": 18900
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6790756284384612e-05,
+      "loss": 1.1391,
+      "step": 18901
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.67829783373624e-05,
+      "loss": 0.9598,
+      "step": 18902
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.677520202721653e-05,
+      "loss": 0.8486,
+      "step": 18903
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6767427354099994e-05,
+      "loss": 1.0064,
+      "step": 18904
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6759654318165686e-05,
+      "loss": 0.938,
+      "step": 18905
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6751882919566576e-05,
+      "loss": 0.7903,
+      "step": 18906
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.674411315845541e-05,
+      "loss": 0.9111,
+      "step": 18907
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.673634503498508e-05,
+      "loss": 0.7674,
+      "step": 18908
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6728578549308416e-05,
+      "loss": 0.9693,
+      "step": 18909
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6720813701578152e-05,
+      "loss": 0.8693,
+      "step": 18910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6713050491947003e-05,
+      "loss": 1.0421,
+      "step": 18911
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6705288920567707e-05,
+      "loss": 0.8316,
+      "step": 18912
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6697528987592892e-05,
+      "loss": 0.9702,
+      "step": 18913
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6689770693175256e-05,
+      "loss": 0.9456,
+      "step": 18914
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6682014037467363e-05,
+      "loss": 0.7394,
+      "step": 18915
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6674259020621785e-05,
+      "loss": 0.9986,
+      "step": 18916
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.666650564279105e-05,
+      "loss": 0.9676,
+      "step": 18917
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6658753904127734e-05,
+      "loss": 1.0901,
+      "step": 18918
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.665100380478425e-05,
+      "loss": 0.9017,
+      "step": 18919
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6643255344913043e-05,
+      "loss": 0.7718,
+      "step": 18920
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6635508524666565e-05,
+      "loss": 0.857,
+      "step": 18921
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6627763344197157e-05,
+      "loss": 0.9612,
+      "step": 18922
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6620019803657195e-05,
+      "loss": 0.8668,
+      "step": 18923
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.661227790319899e-05,
+      "loss": 0.8722,
+      "step": 18924
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6604537642974793e-05,
+      "loss": 0.9345,
+      "step": 18925
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.659679902313689e-05,
+      "loss": 0.9301,
+      "step": 18926
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.65890620438375e-05,
+      "loss": 1.0348,
+      "step": 18927
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6581326705228774e-05,
+      "loss": 0.9887,
+      "step": 18928
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6573593007462907e-05,
+      "loss": 0.9004,
+      "step": 18929
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.656586095069197e-05,
+      "loss": 0.8509,
+      "step": 18930
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6558130535068116e-05,
+      "loss": 0.943,
+      "step": 18931
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6550401760743373e-05,
+      "loss": 0.7067,
+      "step": 18932
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6542674627869737e-05,
+      "loss": 1.043,
+      "step": 18933
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6534949136599232e-05,
+      "loss": 1.0649,
+      "step": 18934
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6527225287083835e-05,
+      "loss": 0.7517,
+      "step": 18935
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6519503079475453e-05,
+      "loss": 1.038,
+      "step": 18936
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6511782513925946e-05,
+      "loss": 0.9943,
+      "step": 18937
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.650406359058725e-05,
+      "loss": 0.8829,
+      "step": 18938
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.649634630961112e-05,
+      "loss": 1.0912,
+      "step": 18939
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.648863067114943e-05,
+      "loss": 0.8952,
+      "step": 18940
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6480916675353898e-05,
+      "loss": 0.8617,
+      "step": 18941
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6473204322376258e-05,
+      "loss": 1.0695,
+      "step": 18942
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6465493612368233e-05,
+      "loss": 0.8984,
+      "step": 18943
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6457784545481493e-05,
+      "loss": 1.007,
+      "step": 18944
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6450077121867635e-05,
+      "loss": 1.2241,
+      "step": 18945
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6442371341678286e-05,
+      "loss": 0.921,
+      "step": 18946
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6434667205065057e-05,
+      "loss": 0.8959,
+      "step": 18947
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.642696471217946e-05,
+      "loss": 1.0478,
+      "step": 18948
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6419263863172997e-05,
+      "loss": 0.9713,
+      "step": 18949
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.641156465819712e-05,
+      "loss": 0.9977,
+      "step": 18950
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6403867097403304e-05,
+      "loss": 0.9251,
+      "step": 18951
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6396171180942986e-05,
+      "loss": 0.8767,
+      "step": 18952
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6388476908967476e-05,
+      "loss": 0.9837,
+      "step": 18953
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.638078428162816e-05,
+      "loss": 0.9603,
+      "step": 18954
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6373093299076368e-05,
+      "loss": 0.9125,
+      "step": 18955
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6365403961463367e-05,
+      "loss": 0.8352,
+      "step": 18956
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6357716268940372e-05,
+      "loss": 0.6348,
+      "step": 18957
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6350030221658642e-05,
+      "loss": 0.8709,
+      "step": 18958
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6342345819769333e-05,
+      "loss": 0.9236,
+      "step": 18959
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.633466306342364e-05,
+      "loss": 0.822,
+      "step": 18960
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6326981952772647e-05,
+      "loss": 1.0077,
+      "step": 18961
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6319302487967434e-05,
+      "loss": 0.8352,
+      "step": 18962
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6311624669159065e-05,
+      "loss": 0.9078,
+      "step": 18963
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.630394849649859e-05,
+      "loss": 0.9488,
+      "step": 18964
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6296273970136977e-05,
+      "loss": 0.8918,
+      "step": 18965
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.628860109022515e-05,
+      "loss": 0.9252,
+      "step": 18966
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6280929856914106e-05,
+      "loss": 0.8809,
+      "step": 18967
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6273260270354674e-05,
+      "loss": 0.953,
+      "step": 18968
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6265592330697767e-05,
+      "loss": 0.7745,
+      "step": 18969
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6257926038094172e-05,
+      "loss": 0.8832,
+      "step": 18970
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.625026139269469e-05,
+      "loss": 1.0234,
+      "step": 18971
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.624259839465011e-05,
+      "loss": 0.9395,
+      "step": 18972
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6234937044111152e-05,
+      "loss": 0.9711,
+      "step": 18973
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6227277341228487e-05,
+      "loss": 1.0603,
+      "step": 18974
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.621961928615282e-05,
+      "loss": 1.0676,
+      "step": 18975
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.621196287903475e-05,
+      "loss": 0.8581,
+      "step": 18976
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6204308120024913e-05,
+      "loss": 0.8632,
+      "step": 18977
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6196655009273866e-05,
+      "loss": 0.7927,
+      "step": 18978
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6189003546932113e-05,
+      "loss": 1.0408,
+      "step": 18979
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6181353733150183e-05,
+      "loss": 0.949,
+      "step": 18980
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6173705568078578e-05,
+      "loss": 0.9962,
+      "step": 18981
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.616605905186771e-05,
+      "loss": 1.0944,
+      "step": 18982
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.615841418466797e-05,
+      "loss": 0.9512,
+      "step": 18983
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6150770966629757e-05,
+      "loss": 0.8266,
+      "step": 18984
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.614312939790339e-05,
+      "loss": 0.9761,
+      "step": 18985
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.613548947863921e-05,
+      "loss": 0.9637,
+      "step": 18986
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6127851208987478e-05,
+      "loss": 0.8859,
+      "step": 18987
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6120214589098405e-05,
+      "loss": 0.8748,
+      "step": 18988
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.611257961912227e-05,
+      "loss": 0.7607,
+      "step": 18989
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6104946299209213e-05,
+      "loss": 0.9522,
+      "step": 18990
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.609731462950935e-05,
+      "loss": 0.8675,
+      "step": 18991
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.608968461017287e-05,
+      "loss": 0.8041,
+      "step": 18992
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6082056241349786e-05,
+      "loss": 0.7688,
+      "step": 18993
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6074429523190204e-05,
+      "loss": 0.9258,
+      "step": 18994
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.606680445584411e-05,
+      "loss": 0.9595,
+      "step": 18995
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6059181039461478e-05,
+      "loss": 0.7684,
+      "step": 18996
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6051559274192275e-05,
+      "loss": 1.0925,
+      "step": 18997
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.604393916018646e-05,
+      "loss": 0.9681,
+      "step": 18998
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6036320697593844e-05,
+      "loss": 0.8685,
+      "step": 18999
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6028703886564323e-05,
+      "loss": 1.1037,
+      "step": 19000
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6021088727247735e-05,
+      "loss": 0.8347,
+      "step": 19001
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.601347521979385e-05,
+      "loss": 0.8728,
+      "step": 19002
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.6005863364352412e-05,
+      "loss": 0.9845,
+      "step": 19003
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5998253161073185e-05,
+      "loss": 0.9026,
+      "step": 19004
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.599064461010581e-05,
+      "loss": 0.7667,
+      "step": 19005
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5983037711600003e-05,
+      "loss": 1.0307,
+      "step": 19006
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5975432465705352e-05,
+      "loss": 0.8393,
+      "step": 19007
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5967828872571455e-05,
+      "loss": 0.9196,
+      "step": 19008
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5960226932347876e-05,
+      "loss": 0.8853,
+      "step": 19009
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5952626645184178e-05,
+      "loss": 0.901,
+      "step": 19010
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5945028011229824e-05,
+      "loss": 0.982,
+      "step": 19011
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5937431030634277e-05,
+      "loss": 0.806,
+      "step": 19012
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5929835703546993e-05,
+      "loss": 0.8629,
+      "step": 19013
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.592224203011734e-05,
+      "loss": 0.634,
+      "step": 19014
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5914650010494747e-05,
+      "loss": 0.9949,
+      "step": 19015
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.590705964482846e-05,
+      "loss": 1.0364,
+      "step": 19016
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5899470933267824e-05,
+      "loss": 0.8162,
+      "step": 19017
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5891883875962134e-05,
+      "loss": 0.9023,
+      "step": 19018
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.588429847306061e-05,
+      "loss": 0.9539,
+      "step": 19019
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5876714724712426e-05,
+      "loss": 0.9399,
+      "step": 19020
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5869132631066808e-05,
+      "loss": 0.9044,
+      "step": 19021
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5861552192272833e-05,
+      "loss": 0.8725,
+      "step": 19022
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.585397340847967e-05,
+      "loss": 0.7912,
+      "step": 19023
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5846396279836372e-05,
+      "loss": 0.9623,
+      "step": 19024
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.583882080649195e-05,
+      "loss": 0.9744,
+      "step": 19025
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5831246988595427e-05,
+      "loss": 0.8671,
+      "step": 19026
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.582367482629583e-05,
+      "loss": 0.9286,
+      "step": 19027
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5816104319742053e-05,
+      "loss": 0.9267,
+      "step": 19028
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5808535469082996e-05,
+      "loss": 1.0892,
+      "step": 19029
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.580096827446759e-05,
+      "loss": 0.7887,
+      "step": 19030
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5793402736044617e-05,
+      "loss": 0.9903,
+      "step": 19031
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5785838853962952e-05,
+      "loss": 0.9878,
+      "step": 19032
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5778276628371357e-05,
+      "loss": 0.8508,
+      "step": 19033
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5770716059418543e-05,
+      "loss": 0.8298,
+      "step": 19034
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5763157147253294e-05,
+      "loss": 0.8668,
+      "step": 19035
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5755599892024243e-05,
+      "loss": 0.9075,
+      "step": 19036
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.574804429388004e-05,
+      "loss": 0.8532,
+      "step": 19037
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5740490352969328e-05,
+      "loss": 0.8244,
+      "step": 19038
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5732938069440673e-05,
+      "loss": 0.9407,
+      "step": 19039
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5725387443442664e-05,
+      "loss": 0.9542,
+      "step": 19040
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.571783847512378e-05,
+      "loss": 0.9157,
+      "step": 19041
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.571029116463252e-05,
+      "loss": 0.8875,
+      "step": 19042
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5702745512117324e-05,
+      "loss": 0.7515,
+      "step": 19043
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5695201517726678e-05,
+      "loss": 0.7983,
+      "step": 19044
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5687659181608916e-05,
+      "loss": 0.7658,
+      "step": 19045
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5680118503912387e-05,
+      "loss": 0.8775,
+      "step": 19046
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5672579484785466e-05,
+      "loss": 0.8566,
+      "step": 19047
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5665042124376405e-05,
+      "loss": 0.8479,
+      "step": 19048
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.5657506422833467e-05,
+      "loss": 0.817,
+      "step": 19049
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.56499723803049e-05,
+      "loss": 1.0679,
+      "step": 19050
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.564243999693886e-05,
+      "loss": 1.04,
+      "step": 19051
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.563490927288357e-05,
+      "loss": 0.8491,
+      "step": 19052
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5627380208287114e-05,
+      "loss": 0.9383,
+      "step": 19053
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5619852803297585e-05,
+      "loss": 0.7062,
+      "step": 19054
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5612327058063048e-05,
+      "loss": 0.9477,
+      "step": 19055
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5604802972731582e-05,
+      "loss": 0.9554,
+      "step": 19056
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5597280547451132e-05,
+      "loss": 1.0432,
+      "step": 19057
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5589759782369695e-05,
+      "loss": 0.8798,
+      "step": 19058
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.558224067763516e-05,
+      "loss": 0.8995,
+      "step": 19059
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.557472323339546e-05,
+      "loss": 0.7191,
+      "step": 19060
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5567207449798515e-05,
+      "loss": 1.0292,
+      "step": 19061
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5559693326992052e-05,
+      "loss": 0.9239,
+      "step": 19062
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5552180865123933e-05,
+      "loss": 1.0357,
+      "step": 19063
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.554467006434195e-05,
+      "loss": 1.0047,
+      "step": 19064
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5537160924793802e-05,
+      "loss": 0.9152,
+      "step": 19065
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.552965344662719e-05,
+      "loss": 1.0313,
+      "step": 19066
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5522147629989825e-05,
+      "loss": 1.0567,
+      "step": 19067
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.551464347502929e-05,
+      "loss": 0.9353,
+      "step": 19068
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5507140981893253e-05,
+      "loss": 0.8225,
+      "step": 19069
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5499640150729255e-05,
+      "loss": 0.8225,
+      "step": 19070
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5492140981684823e-05,
+      "loss": 1.0865,
+      "step": 19071
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.548464347490748e-05,
+      "loss": 0.8627,
+      "step": 19072
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.547714763054472e-05,
+      "loss": 0.9768,
+      "step": 19073
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.546965344874398e-05,
+      "loss": 1.1597,
+      "step": 19074
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5462160929652625e-05,
+      "loss": 0.9269,
+      "step": 19075
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.545467007341811e-05,
+      "loss": 0.866,
+      "step": 19076
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5447180880187696e-05,
+      "loss": 0.909,
+      "step": 19077
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5439693350108775e-05,
+      "loss": 0.9133,
+      "step": 19078
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5432207483328587e-05,
+      "loss": 0.7789,
+      "step": 19079
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.542472327999436e-05,
+      "loss": 0.9899,
+      "step": 19080
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5417240740253336e-05,
+      "loss": 0.8483,
+      "step": 19081
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5409759864252704e-05,
+      "loss": 0.9055,
+      "step": 19082
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5402280652139565e-05,
+      "loss": 1.0315,
+      "step": 19083
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5394803104061106e-05,
+      "loss": 0.7095,
+      "step": 19084
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.538732722016434e-05,
+      "loss": 0.9063,
+      "step": 19085
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.537985300059637e-05,
+      "loss": 0.9502,
+      "step": 19086
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5372380445504196e-05,
+      "loss": 0.8021,
+      "step": 19087
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5364909555034768e-05,
+      "loss": 0.6465,
+      "step": 19088
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5357440329335083e-05,
+      "loss": 0.8079,
+      "step": 19089
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.534997276855207e-05,
+      "loss": 0.8783,
+      "step": 19090
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5342506872832586e-05,
+      "loss": 0.9634,
+      "step": 19091
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.533504264232347e-05,
+      "loss": 0.9698,
+      "step": 19092
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5327580077171587e-05,
+      "loss": 0.7366,
+      "step": 19093
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5320119177523707e-05,
+      "loss": 0.9491,
+      "step": 19094
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5312659943526562e-05,
+      "loss": 0.7959,
+      "step": 19095
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5305202375326922e-05,
+      "loss": 0.8633,
+      "step": 19096
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5297746473071418e-05,
+      "loss": 0.857,
+      "step": 19097
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5290292236906767e-05,
+      "loss": 0.9239,
+      "step": 19098
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5282839666979566e-05,
+      "loss": 0.8124,
+      "step": 19099
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5275388763436382e-05,
+      "loss": 0.7702,
+      "step": 19100
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5267939526423824e-05,
+      "loss": 0.8949,
+      "step": 19101
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5260491956088374e-05,
+      "loss": 0.8238,
+      "step": 19102
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5253046052576558e-05,
+      "loss": 0.8881,
+      "step": 19103
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5245601816034838e-05,
+      "loss": 0.9241,
+      "step": 19104
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5238159246609596e-05,
+      "loss": 0.8404,
+      "step": 19105
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5230718344447258e-05,
+      "loss": 0.9484,
+      "step": 19106
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5223279109694244e-05,
+      "loss": 0.9353,
+      "step": 19107
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5215841542496767e-05,
+      "loss": 0.7876,
+      "step": 19108
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.520840564300119e-05,
+      "loss": 0.8171,
+      "step": 19109
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5200971411353793e-05,
+      "loss": 0.9175,
+      "step": 19110
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5193538847700784e-05,
+      "loss": 0.8657,
+      "step": 19111
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5186107952188345e-05,
+      "loss": 0.8439,
+      "step": 19112
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.517867872496267e-05,
+      "loss": 0.9721,
+      "step": 19113
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5171251166169864e-05,
+      "loss": 0.8683,
+      "step": 19114
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5163825275956067e-05,
+      "loss": 0.8018,
+      "step": 19115
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5156401054467317e-05,
+      "loss": 0.8561,
+      "step": 19116
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5148978501849642e-05,
+      "loss": 0.6933,
+      "step": 19117
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5141557618249036e-05,
+      "loss": 0.9784,
+      "step": 19118
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5134138403811526e-05,
+      "loss": 1.037,
+      "step": 19119
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5126720858682996e-05,
+      "loss": 0.9733,
+      "step": 19120
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5119304983009342e-05,
+      "loss": 1.0261,
+      "step": 19121
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5111890776936478e-05,
+      "loss": 1.0785,
+      "step": 19122
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5104478240610187e-05,
+      "loss": 0.6994,
+      "step": 19123
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5097067374176355e-05,
+      "loss": 1.0143,
+      "step": 19124
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5089658177780653e-05,
+      "loss": 0.8496,
+      "step": 19125
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5082250651568862e-05,
+      "loss": 0.7961,
+      "step": 19126
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5074844795686727e-05,
+      "loss": 1.0809,
+      "step": 19127
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5067440610279881e-05,
+      "loss": 0.9464,
+      "step": 19128
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5060038095493944e-05,
+      "loss": 0.825,
+      "step": 19129
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5052637251474577e-05,
+      "loss": 0.8266,
+      "step": 19130
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5045238078367307e-05,
+      "loss": 0.9823,
+      "step": 19131
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5037840576317719e-05,
+      "loss": 0.7981,
+      "step": 19132
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5030444745471294e-05,
+      "loss": 0.7878,
+      "step": 19133
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5023050585973486e-05,
+      "loss": 1.0171,
+      "step": 19134
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5015658097969765e-05,
+      "loss": 0.7383,
+      "step": 19135
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5008267281605571e-05,
+      "loss": 0.8595,
+      "step": 19136
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.5000878137026242e-05,
+      "loss": 0.784,
+      "step": 19137
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4993490664377108e-05,
+      "loss": 0.9215,
+      "step": 19138
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.498610486380353e-05,
+      "loss": 0.9282,
+      "step": 19139
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4978720735450757e-05,
+      "loss": 0.9733,
+      "step": 19140
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4971338279464009e-05,
+      "loss": 1.0019,
+      "step": 19141
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4963957495988557e-05,
+      "loss": 0.7857,
+      "step": 19142
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4956578385169528e-05,
+      "loss": 1.0125,
+      "step": 19143
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4949200947152109e-05,
+      "loss": 0.893,
+      "step": 19144
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4941825182081392e-05,
+      "loss": 0.957,
+      "step": 19145
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4934451090102452e-05,
+      "loss": 0.7704,
+      "step": 19146
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4927078671360362e-05,
+      "loss": 0.8101,
+      "step": 19147
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4919707926000092e-05,
+      "loss": 0.9764,
+      "step": 19148
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4912338854166696e-05,
+      "loss": 1.074,
+      "step": 19149
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4904971456005067e-05,
+      "loss": 0.8515,
+      "step": 19150
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4897605731660124e-05,
+      "loss": 1.2033,
+      "step": 19151
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4890241681276762e-05,
+      "loss": 1.0487,
+      "step": 19152
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4882879304999875e-05,
+      "loss": 0.8748,
+      "step": 19153
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4875518602974193e-05,
+      "loss": 0.8528,
+      "step": 19154
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4868159575344543e-05,
+      "loss": 0.7985,
+      "step": 19155
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.486080222225571e-05,
+      "loss": 0.7985,
+      "step": 19156
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.485344654385239e-05,
+      "loss": 0.9782,
+      "step": 19157
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4846092540279233e-05,
+      "loss": 0.7771,
+      "step": 19158
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4838740211680935e-05,
+      "loss": 0.925,
+      "step": 19159
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4831389558202091e-05,
+      "loss": 0.9245,
+      "step": 19160
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.482404057998732e-05,
+      "loss": 0.9944,
+      "step": 19161
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.481669327718116e-05,
+      "loss": 0.8196,
+      "step": 19162
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4809347649928095e-05,
+      "loss": 0.9946,
+      "step": 19163
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4802003698372646e-05,
+      "loss": 0.9624,
+      "step": 19164
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4794661422659306e-05,
+      "loss": 1.1008,
+      "step": 19165
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4787320822932449e-05,
+      "loss": 0.9894,
+      "step": 19166
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4779981899336482e-05,
+      "loss": 0.8863,
+      "step": 19167
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.477264465201572e-05,
+      "loss": 1.0211,
+      "step": 19168
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4765309081114532e-05,
+      "loss": 0.9069,
+      "step": 19169
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.475797518677724e-05,
+      "loss": 0.8005,
+      "step": 19170
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4750642969148009e-05,
+      "loss": 0.8301,
+      "step": 19171
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4743312428371125e-05,
+      "loss": 0.8,
+      "step": 19172
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4735983564590783e-05,
+      "loss": 0.8659,
+      "step": 19173
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4728656377951134e-05,
+      "loss": 0.8396,
+      "step": 19174
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4721330868596261e-05,
+      "loss": 0.9579,
+      "step": 19175
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4714007036670319e-05,
+      "loss": 0.7991,
+      "step": 19176
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4706684882317323e-05,
+      "loss": 1.0237,
+      "step": 19177
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4699364405681338e-05,
+      "loss": 0.9046,
+      "step": 19178
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4692045606906334e-05,
+      "loss": 0.8826,
+      "step": 19179
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4684728486136256e-05,
+      "loss": 0.8543,
+      "step": 19180
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4677413043515042e-05,
+      "loss": 0.7168,
+      "step": 19181
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4670099279186623e-05,
+      "loss": 0.7223,
+      "step": 19182
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4662787193294835e-05,
+      "loss": 0.8729,
+      "step": 19183
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.465547678598347e-05,
+      "loss": 0.9878,
+      "step": 19184
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4648168057396383e-05,
+      "loss": 1.0125,
+      "step": 19185
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.464086100767732e-05,
+      "loss": 1.0287,
+      "step": 19186
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4633555636969964e-05,
+      "loss": 0.8774,
+      "step": 19187
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4626251945418079e-05,
+      "loss": 1.0636,
+      "step": 19188
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4618949933165272e-05,
+      "loss": 0.9089,
+      "step": 19189
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4611649600355226e-05,
+      "loss": 0.9256,
+      "step": 19190
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4604350947131506e-05,
+      "loss": 0.9627,
+      "step": 19191
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4597053973637664e-05,
+      "loss": 1.0542,
+      "step": 19192
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4589758680017263e-05,
+      "loss": 0.8873,
+      "step": 19193
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4582465066413764e-05,
+      "loss": 0.8833,
+      "step": 19194
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4575173132970688e-05,
+      "loss": 0.7052,
+      "step": 19195
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4567882879831429e-05,
+      "loss": 0.7928,
+      "step": 19196
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4560594307139364e-05,
+      "loss": 0.8344,
+      "step": 19197
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4553307415037898e-05,
+      "loss": 0.892,
+      "step": 19198
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4546022203670385e-05,
+      "loss": 1.0031,
+      "step": 19199
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4538738673180052e-05,
+      "loss": 0.9115,
+      "step": 19200
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.453145682371021e-05,
+      "loss": 0.9173,
+      "step": 19201
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.452417665540411e-05,
+      "loss": 0.8479,
+      "step": 19202
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4516898168404925e-05,
+      "loss": 0.8062,
+      "step": 19203
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4509621362855808e-05,
+      "loss": 0.9511,
+      "step": 19204
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4502346238899945e-05,
+      "loss": 0.8423,
+      "step": 19205
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4495072796680375e-05,
+      "loss": 1.0211,
+      "step": 19206
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448780103634022e-05,
+      "loss": 0.7479,
+      "step": 19207
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4480530958022498e-05,
+      "loss": 0.783,
+      "step": 19208
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4473262561870193e-05,
+      "loss": 0.9077,
+      "step": 19209
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4465995848026271e-05,
+      "loss": 0.7586,
+      "step": 19210
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4458730816633726e-05,
+      "loss": 0.9732,
+      "step": 19211
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4451467467835411e-05,
+      "loss": 0.8589,
+      "step": 19212
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4444205801774202e-05,
+      "loss": 0.8627,
+      "step": 19213
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4436945818592917e-05,
+      "loss": 0.8683,
+      "step": 19214
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4429687518434387e-05,
+      "loss": 0.7771,
+      "step": 19215
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4422430901441419e-05,
+      "loss": 0.7581,
+      "step": 19216
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4415175967756656e-05,
+      "loss": 1.1432,
+      "step": 19217
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.440792271752287e-05,
+      "loss": 0.8602,
+      "step": 19218
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4400671150882728e-05,
+      "loss": 0.9739,
+      "step": 19219
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4393421267978857e-05,
+      "loss": 0.8497,
+      "step": 19220
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4386173068953846e-05,
+      "loss": 0.9738,
+      "step": 19221
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4378926553950301e-05,
+      "loss": 0.9367,
+      "step": 19222
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4371681723110719e-05,
+      "loss": 0.8949,
+      "step": 19223
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4364438576577655e-05,
+      "loss": 0.997,
+      "step": 19224
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.435719711449356e-05,
+      "loss": 1.1416,
+      "step": 19225
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.434995733700083e-05,
+      "loss": 0.9514,
+      "step": 19226
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4342719244241931e-05,
+      "loss": 0.8837,
+      "step": 19227
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4335482836359227e-05,
+      "loss": 0.9334,
+      "step": 19228
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4328248113495047e-05,
+      "loss": 0.8728,
+      "step": 19229
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4321015075791678e-05,
+      "loss": 0.9902,
+      "step": 19230
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4313783723391428e-05,
+      "loss": 1.0068,
+      "step": 19231
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4306554056436527e-05,
+      "loss": 0.8182,
+      "step": 19232
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4299326075069174e-05,
+      "loss": 1.0013,
+      "step": 19233
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4292099779431523e-05,
+      "loss": 0.9098,
+      "step": 19234
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4284875169665745e-05,
+      "loss": 0.8656,
+      "step": 19235
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4277652245913964e-05,
+      "loss": 0.9641,
+      "step": 19236
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.427043100831823e-05,
+      "loss": 0.8524,
+      "step": 19237
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4263211457020565e-05,
+      "loss": 1.056,
+      "step": 19238
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.425599359216302e-05,
+      "loss": 0.9743,
+      "step": 19239
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4248777413887527e-05,
+      "loss": 1.0195,
+      "step": 19240
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.424156292233607e-05,
+      "loss": 0.8371,
+      "step": 19241
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.423435011765054e-05,
+      "loss": 0.8345,
+      "step": 19242
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.42271389999728e-05,
+      "loss": 0.8208,
+      "step": 19243
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.421992956944469e-05,
+      "loss": 0.9018,
+      "step": 19244
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4212721826208086e-05,
+      "loss": 0.9981,
+      "step": 19245
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4205515770404654e-05,
+      "loss": 0.8706,
+      "step": 19246
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4198311402176201e-05,
+      "loss": 0.8791,
+      "step": 19247
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4191108721664448e-05,
+      "loss": 0.8863,
+      "step": 19248
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4183907729011058e-05,
+      "loss": 1.0191,
+      "step": 19249
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4176708424357643e-05,
+      "loss": 0.9066,
+      "step": 19250
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4169510807845865e-05,
+      "loss": 0.8442,
+      "step": 19251
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4162314879617256e-05,
+      "loss": 0.8472,
+      "step": 19252
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.415512063981339e-05,
+      "loss": 0.8564,
+      "step": 19253
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4147928088575779e-05,
+      "loss": 0.9047,
+      "step": 19254
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4140737226045853e-05,
+      "loss": 0.9208,
+      "step": 19255
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.413354805236512e-05,
+      "loss": 0.9244,
+      "step": 19256
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4126360567674945e-05,
+      "loss": 0.7672,
+      "step": 19257
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.411917477211674e-05,
+      "loss": 0.7622,
+      "step": 19258
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.411199066583182e-05,
+      "loss": 0.9834,
+      "step": 19259
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4104808248961499e-05,
+      "loss": 1.0216,
+      "step": 19260
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4097627521647061e-05,
+      "loss": 1.251,
+      "step": 19261
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4090448484029794e-05,
+      "loss": 0.9169,
+      "step": 19262
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.408327113625082e-05,
+      "loss": 0.9811,
+      "step": 19263
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4076095478451378e-05,
+      "loss": 0.8411,
+      "step": 19264
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4068921510772615e-05,
+      "loss": 0.906,
+      "step": 19265
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4061749233355636e-05,
+      "loss": 0.9104,
+      "step": 19266
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4054578646341476e-05,
+      "loss": 1.0171,
+      "step": 19267
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4047409749871255e-05,
+      "loss": 0.8441,
+      "step": 19268
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4040242544085913e-05,
+      "loss": 0.9089,
+      "step": 19269
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4033077029126485e-05,
+      "loss": 1.0156,
+      "step": 19270
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.40259132051339e-05,
+      "loss": 0.7168,
+      "step": 19271
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4018751072249037e-05,
+      "loss": 0.7559,
+      "step": 19272
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4011590630612803e-05,
+      "loss": 0.7959,
+      "step": 19273
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4004431880366064e-05,
+      "loss": 1.0135,
+      "step": 19274
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3997274821649609e-05,
+      "loss": 0.9352,
+      "step": 19275
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3990119454604223e-05,
+      "loss": 0.6538,
+      "step": 19276
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3982965779370626e-05,
+      "loss": 0.9023,
+      "step": 19277
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3975813796089565e-05,
+      "loss": 0.9101,
+      "step": 19278
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3968663504901713e-05,
+      "loss": 0.7411,
+      "step": 19279
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3961514905947692e-05,
+      "loss": 0.7141,
+      "step": 19280
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3954367999368134e-05,
+      "loss": 0.9759,
+      "step": 19281
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.3947222785303627e-05,
+      "loss": 0.9433,
+      "step": 19282
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3940079263894712e-05,
+      "loss": 0.8457,
+      "step": 19283
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.393293743528188e-05,
+      "loss": 0.892,
+      "step": 19284
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3925797299605647e-05,
+      "loss": 0.8783,
+      "step": 19285
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3918658857006418e-05,
+      "loss": 0.7128,
+      "step": 19286
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3911522107624642e-05,
+      "loss": 1.0049,
+      "step": 19287
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3904387051600698e-05,
+      "loss": 0.9677,
+      "step": 19288
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3897253689074885e-05,
+      "loss": 0.7553,
+      "step": 19289
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3890122020187558e-05,
+      "loss": 0.9219,
+      "step": 19290
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3882992045079025e-05,
+      "loss": 0.9036,
+      "step": 19291
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3875863763889451e-05,
+      "loss": 0.9349,
+      "step": 19292
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3868737176759106e-05,
+      "loss": 0.7516,
+      "step": 19293
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3861612283828163e-05,
+      "loss": 0.8939,
+      "step": 19294
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3854489085236777e-05,
+      "loss": 0.9049,
+      "step": 19295
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3847367581125014e-05,
+      "loss": 0.8611,
+      "step": 19296
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3840247771633019e-05,
+      "loss": 0.9125,
+      "step": 19297
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3833129656900779e-05,
+      "loss": 0.951,
+      "step": 19298
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3826013237068359e-05,
+      "loss": 0.8701,
+      "step": 19299
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3818898512275713e-05,
+      "loss": 1.01,
+      "step": 19300
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3811785482662764e-05,
+      "loss": 0.5452,
+      "step": 19301
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3804674148369468e-05,
+      "loss": 0.995,
+      "step": 19302
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3797564509535676e-05,
+      "loss": 0.8612,
+      "step": 19303
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3790456566301257e-05,
+      "loss": 1.0315,
+      "step": 19304
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3783350318806021e-05,
+      "loss": 0.754,
+      "step": 19305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.377624576718971e-05,
+      "loss": 0.9779,
+      "step": 19306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3769142911592103e-05,
+      "loss": 1.0305,
+      "step": 19307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3762041752152943e-05,
+      "loss": 0.9093,
+      "step": 19308
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.375494228901184e-05,
+      "loss": 1.0849,
+      "step": 19309
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3747844522308474e-05,
+      "loss": 0.7887,
+      "step": 19310
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3740748452182484e-05,
+      "loss": 1.0019,
+      "step": 19311
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3733654078773428e-05,
+      "loss": 1.056,
+      "step": 19312
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3726561402220817e-05,
+      "loss": 0.9184,
+      "step": 19313
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3719470422664227e-05,
+      "loss": 0.9523,
+      "step": 19314
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.371238114024308e-05,
+      "loss": 1.0018,
+      "step": 19315
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3705293555096865e-05,
+      "loss": 0.8727,
+      "step": 19316
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3698207667364982e-05,
+      "loss": 0.8821,
+      "step": 19317
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3691123477186785e-05,
+      "loss": 0.8785,
+      "step": 19318
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3684040984701652e-05,
+      "loss": 1.1396,
+      "step": 19319
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3676960190048893e-05,
+      "loss": 0.7208,
+      "step": 19320
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3669881093367775e-05,
+      "loss": 1.0707,
+      "step": 19321
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3662803694797554e-05,
+      "loss": 0.8867,
+      "step": 19322
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3655727994477408e-05,
+      "loss": 0.9707,
+      "step": 19323
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3648653992546579e-05,
+      "loss": 0.8602,
+      "step": 19324
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3641581689144156e-05,
+      "loss": 1.0296,
+      "step": 19325
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3634511084409262e-05,
+      "loss": 1.0825,
+      "step": 19326
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3627442178480986e-05,
+      "loss": 0.8814,
+      "step": 19327
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3620374971498384e-05,
+      "loss": 1.0148,
+      "step": 19328
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3613309463600465e-05,
+      "loss": 0.7171,
+      "step": 19329
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3606245654926175e-05,
+      "loss": 0.9582,
+      "step": 19330
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3599183545614503e-05,
+      "loss": 0.8814,
+      "step": 19331
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3592123135804314e-05,
+      "loss": 0.9125,
+      "step": 19332
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3585064425634542e-05,
+      "loss": 0.7779,
+      "step": 19333
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3578007415243998e-05,
+      "loss": 0.6376,
+      "step": 19334
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3570952104771462e-05,
+      "loss": 0.9967,
+      "step": 19335
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3563898494355765e-05,
+      "loss": 0.7346,
+      "step": 19336
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3556846584135663e-05,
+      "loss": 1.007,
+      "step": 19337
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3549796374249801e-05,
+      "loss": 0.815,
+      "step": 19338
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.354274786483689e-05,
+      "loss": 1.1587,
+      "step": 19339
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3535701056035588e-05,
+      "loss": 1.1276,
+      "step": 19340
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3528655947984504e-05,
+      "loss": 0.9993,
+      "step": 19341
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3521612540822204e-05,
+      "loss": 0.8696,
+      "step": 19342
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3514570834687202e-05,
+      "loss": 0.873,
+      "step": 19343
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3507530829718052e-05,
+      "loss": 0.689,
+      "step": 19344
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.350049252605322e-05,
+      "loss": 1.1485,
+      "step": 19345
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3493455923831167e-05,
+      "loss": 0.8566,
+      "step": 19346
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3486421023190243e-05,
+      "loss": 1.0887,
+      "step": 19347
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3479387824268896e-05,
+      "loss": 0.8902,
+      "step": 19348
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3472356327205405e-05,
+      "loss": 0.9526,
+      "step": 19349
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3465326532138135e-05,
+      "loss": 0.8859,
+      "step": 19350
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.345829843920534e-05,
+      "loss": 0.9067,
+      "step": 19351
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3451272048545238e-05,
+      "loss": 1.1578,
+      "step": 19352
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3444247360296058e-05,
+      "loss": 0.8641,
+      "step": 19353
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3437224374596014e-05,
+      "loss": 0.9273,
+      "step": 19354
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3430203091583171e-05,
+      "loss": 0.8671,
+      "step": 19355
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3423183511395677e-05,
+      "loss": 0.8235,
+      "step": 19356
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.341616563417163e-05,
+      "loss": 0.9385,
+      "step": 19357
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3409149460049053e-05,
+      "loss": 0.9366,
+      "step": 19358
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3402134989165926e-05,
+      "loss": 1.0927,
+      "step": 19359
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.339512222166026e-05,
+      "loss": 0.7534,
+      "step": 19360
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3388111157669969e-05,
+      "loss": 0.8727,
+      "step": 19361
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3381101797332995e-05,
+      "loss": 0.9492,
+      "step": 19362
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3374094140787196e-05,
+      "loss": 1.0375,
+      "step": 19363
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3367088188170373e-05,
+      "loss": 0.8243,
+      "step": 19364
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3360083939620406e-05,
+      "loss": 0.7614,
+      "step": 19365
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3353081395274991e-05,
+      "loss": 0.729,
+      "step": 19366
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3346080555271933e-05,
+      "loss": 0.9407,
+      "step": 19367
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3339081419748922e-05,
+      "loss": 0.7831,
+      "step": 19368
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.333208398884359e-05,
+      "loss": 0.8173,
+      "step": 19369
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.332508826269363e-05,
+      "loss": 0.8988,
+      "step": 19370
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3318094241436619e-05,
+      "loss": 0.8053,
+      "step": 19371
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3311101925210124e-05,
+      "loss": 0.8772,
+      "step": 19372
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.330411131415169e-05,
+      "loss": 0.8202,
+      "step": 19373
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3297122408398843e-05,
+      "loss": 0.9941,
+      "step": 19374
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.329013520808905e-05,
+      "loss": 0.9339,
+      "step": 19375
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3283149713359699e-05,
+      "loss": 0.9107,
+      "step": 19376
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.327616592434827e-05,
+      "loss": 0.7693,
+      "step": 19377
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3269183841192067e-05,
+      "loss": 0.9178,
+      "step": 19378
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3262203464028477e-05,
+      "loss": 1.06,
+      "step": 19379
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3255224792994792e-05,
+      "loss": 0.8002,
+      "step": 19380
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3248247828228245e-05,
+      "loss": 0.8807,
+      "step": 19381
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3241272569866103e-05,
+      "loss": 1.086,
+      "step": 19382
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3234299018045615e-05,
+      "loss": 0.8301,
+      "step": 19383
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3227327172903858e-05,
+      "loss": 0.791,
+      "step": 19384
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3220357034578013e-05,
+      "loss": 0.9131,
+      "step": 19385
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3213388603205213e-05,
+      "loss": 1.0007,
+      "step": 19386
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3206421878922481e-05,
+      "loss": 1.1217,
+      "step": 19387
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3199456861866877e-05,
+      "loss": 0.958,
+      "step": 19388
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3192493552175366e-05,
+      "loss": 0.8349,
+      "step": 19389
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3185531949984953e-05,
+      "loss": 0.949,
+      "step": 19390
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3178572055432582e-05,
+      "loss": 0.8593,
+      "step": 19391
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.317161386865513e-05,
+      "loss": 0.8323,
+      "step": 19392
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3164657389789458e-05,
+      "loss": 0.826,
+      "step": 19393
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3157702618972423e-05,
+      "loss": 0.9591,
+      "step": 19394
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3150749556340791e-05,
+      "loss": 1.0109,
+      "step": 19395
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3143798202031377e-05,
+      "loss": 1.0011,
+      "step": 19396
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3136848556180892e-05,
+      "loss": 1.1463,
+      "step": 19397
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3129900618926005e-05,
+      "loss": 0.7073,
+      "step": 19398
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3122954390403408e-05,
+      "loss": 0.7354,
+      "step": 19399
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3116009870749769e-05,
+      "loss": 0.8219,
+      "step": 19400
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.310906706010162e-05,
+      "loss": 0.9717,
+      "step": 19401
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3102125958595546e-05,
+      "loss": 0.7889,
+      "step": 19402
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3095186566368112e-05,
+      "loss": 0.922,
+      "step": 19403
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3088248883555797e-05,
+      "loss": 0.7171,
+      "step": 19404
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.308131291029504e-05,
+      "loss": 0.9809,
+      "step": 19405
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3074378646722308e-05,
+      "loss": 0.8261,
+      "step": 19406
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3067446092973967e-05,
+      "loss": 0.8409,
+      "step": 19407
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3060515249186411e-05,
+      "loss": 0.798,
+      "step": 19408
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3053586115495953e-05,
+      "loss": 1.0224,
+      "step": 19409
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.304665869203886e-05,
+      "loss": 0.9658,
+      "step": 19410
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3039732978951458e-05,
+      "loss": 0.7976,
+      "step": 19411
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3032808976369915e-05,
+      "loss": 0.765,
+      "step": 19412
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3025886684430467e-05,
+      "loss": 0.9803,
+      "step": 19413
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.301896610326926e-05,
+      "loss": 0.8309,
+      "step": 19414
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3012047233022406e-05,
+      "loss": 0.9996,
+      "step": 19415
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.3005130073826034e-05,
+      "loss": 0.865,
+      "step": 19416
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2998214625816185e-05,
+      "loss": 0.8901,
+      "step": 19417
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2991300889128866e-05,
+      "loss": 0.9654,
+      "step": 19418
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2984388863900088e-05,
+      "loss": 0.7069,
+      "step": 19419
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2977478550265832e-05,
+      "loss": 0.9634,
+      "step": 19420
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2970569948361999e-05,
+      "loss": 1.0586,
+      "step": 19421
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2963663058324471e-05,
+      "loss": 1.0418,
+      "step": 19422
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2956757880289138e-05,
+      "loss": 0.9099,
+      "step": 19423
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2949854414391793e-05,
+      "loss": 0.9132,
+      "step": 19424
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2942952660768259e-05,
+      "loss": 1.0014,
+      "step": 19425
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2936052619554274e-05,
+      "loss": 0.8626,
+      "step": 19426
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2929154290885548e-05,
+      "loss": 0.998,
+      "step": 19427
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2922257674897776e-05,
+      "loss": 0.7481,
+      "step": 19428
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.291536277172667e-05,
+      "loss": 0.8863,
+      "step": 19429
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2908469581507765e-05,
+      "loss": 0.979,
+      "step": 19430
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2901578104376721e-05,
+      "loss": 0.805,
+      "step": 19431
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.289468834046903e-05,
+      "loss": 0.8621,
+      "step": 19432
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2887800289920283e-05,
+      "loss": 0.8745,
+      "step": 19433
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2880913952865926e-05,
+      "loss": 0.9279,
+      "step": 19434
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2874029329441395e-05,
+      "loss": 0.8934,
+      "step": 19435
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2867146419782127e-05,
+      "loss": 0.8575,
+      "step": 19436
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2860265224023549e-05,
+      "loss": 0.8774,
+      "step": 19437
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2853385742300983e-05,
+      "loss": 0.9576,
+      "step": 19438
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2846507974749722e-05,
+      "loss": 0.9963,
+      "step": 19439
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2839631921505092e-05,
+      "loss": 0.8253,
+      "step": 19440
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2832757582702316e-05,
+      "loss": 0.9838,
+      "step": 19441
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.282588495847663e-05,
+      "loss": 0.6603,
+      "step": 19442
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.281901404896323e-05,
+      "loss": 0.9423,
+      "step": 19443
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2812144854297226e-05,
+      "loss": 0.8432,
+      "step": 19444
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2805277374613744e-05,
+      "loss": 0.8636,
+      "step": 19445
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2798411610047944e-05,
+      "loss": 0.9429,
+      "step": 19446
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2791547560734762e-05,
+      "loss": 0.9238,
+      "step": 19447
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2784685226809257e-05,
+      "loss": 0.8118,
+      "step": 19448
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2777824608406453e-05,
+      "loss": 0.8266,
+      "step": 19449
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2770965705661254e-05,
+      "loss": 0.8427,
+      "step": 19450
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2764108518708585e-05,
+      "loss": 1.1387,
+      "step": 19451
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2757253047683303e-05,
+      "loss": 0.7513,
+      "step": 19452
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2750399292720283e-05,
+      "loss": 0.8593,
+      "step": 19453
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2743547253954358e-05,
+      "loss": 0.9793,
+      "step": 19454
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2736696931520276e-05,
+      "loss": 0.9572,
+      "step": 19455
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2729848325552762e-05,
+      "loss": 0.9952,
+      "step": 19456
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2723001436186576e-05,
+      "loss": 0.8735,
+      "step": 19457
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2716156263556367e-05,
+      "loss": 1.0244,
+      "step": 19458
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2709312807796802e-05,
+      "loss": 0.9282,
+      "step": 19459
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2702471069042465e-05,
+      "loss": 1.1845,
+      "step": 19460
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2695631047427937e-05,
+      "loss": 1.0127,
+      "step": 19461
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2688792743087786e-05,
+      "loss": 0.9274,
+      "step": 19462
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2681956156156505e-05,
+      "loss": 0.8483,
+      "step": 19463
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2675121286768543e-05,
+      "loss": 0.9391,
+      "step": 19464
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2668288135058371e-05,
+      "loss": 0.8728,
+      "step": 19465
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2661456701160413e-05,
+      "loss": 0.9493,
+      "step": 19466
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2654626985209029e-05,
+      "loss": 1.0346,
+      "step": 19467
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2647798987338521e-05,
+      "loss": 0.8659,
+      "step": 19468
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2640972707683252e-05,
+      "loss": 0.9615,
+      "step": 19469
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2634148146377455e-05,
+      "loss": 1.028,
+      "step": 19470
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2627325303555404e-05,
+      "loss": 0.8931,
+      "step": 19471
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2620504179351278e-05,
+      "loss": 0.864,
+      "step": 19472
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2613684773899237e-05,
+      "loss": 0.9174,
+      "step": 19473
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2606867087333463e-05,
+      "loss": 0.7607,
+      "step": 19474
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2600051119788025e-05,
+      "loss": 0.7782,
+      "step": 19475
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2593236871396984e-05,
+      "loss": 0.7641,
+      "step": 19476
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2586424342294412e-05,
+      "loss": 0.8737,
+      "step": 19477
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2579613532614264e-05,
+      "loss": 0.8258,
+      "step": 19478
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2572804442490571e-05,
+      "loss": 1.0292,
+      "step": 19479
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2565997072057233e-05,
+      "loss": 0.9144,
+      "step": 19480
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2559191421448125e-05,
+      "loss": 0.9143,
+      "step": 19481
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2552387490797136e-05,
+      "loss": 0.9667,
+      "step": 19482
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2545585280238126e-05,
+      "loss": 1.0379,
+      "step": 19483
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2538784789904878e-05,
+      "loss": 0.8541,
+      "step": 19484
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2531986019931119e-05,
+      "loss": 0.8722,
+      "step": 19485
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.252518897045064e-05,
+      "loss": 0.7658,
+      "step": 19486
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2518393641597093e-05,
+      "loss": 0.8725,
+      "step": 19487
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2511600033504179e-05,
+      "loss": 0.7749,
+      "step": 19488
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2504808146305502e-05,
+      "loss": 1.1,
+      "step": 19489
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2498017980134657e-05,
+      "loss": 0.8881,
+      "step": 19490
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2491229535125203e-05,
+      "loss": 0.7986,
+      "step": 19491
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2484442811410724e-05,
+      "loss": 1.0035,
+      "step": 19492
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2477657809124631e-05,
+      "loss": 0.9972,
+      "step": 19493
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2470874528400433e-05,
+      "loss": 1.0115,
+      "step": 19494
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2464092969371555e-05,
+      "loss": 0.9705,
+      "step": 19495
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.24573131321714e-05,
+      "loss": 0.9592,
+      "step": 19496
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2450535016933296e-05,
+      "loss": 1.0346,
+      "step": 19497
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.244375862379057e-05,
+      "loss": 0.7721,
+      "step": 19498
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2436983952876524e-05,
+      "loss": 0.8802,
+      "step": 19499
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2430211004324444e-05,
+      "loss": 1.0039,
+      "step": 19500
+    }
+  ],
+  "max_steps": 23093,
+  "num_train_epochs": 1,
+  "total_flos": 1.1150375552909312e+16,
+  "trial_name": null,
+  "trial_params": null
+}