diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -10,9965 +10,9965 @@
     {
       "epoch": 0.01,
       "learning_rate": 0.0,
-      "loss": 16.3331,
+      "loss": 16.3153,
       "step": 1
     },
     {
       "epoch": 0.02,
       "learning_rate": 0.0,
-      "loss": 14.9422,
+      "loss": 14.9214,
       "step": 2
     },
     {
       "epoch": 0.03,
       "learning_rate": 0.0,
-      "loss": 14.398,
+      "loss": 14.3728,
       "step": 3
     },
     {
       "epoch": 0.04,
       "learning_rate": 6e-07,
-      "loss": 14.8661,
+      "loss": 14.8368,
       "step": 4
     },
     {
       "epoch": 0.05,
       "learning_rate": 1.2e-06,
-      "loss": 14.7223,
+      "loss": 14.7156,
       "step": 5
     },
     {
       "epoch": 0.06,
-      "learning_rate": 1.2e-06,
-      "loss": 17.4397,
+      "learning_rate": 1.8e-06,
+      "loss": 17.4555,
       "step": 6
     },
     {
       "epoch": 0.06,
-      "learning_rate": 1.8e-06,
-      "loss": 15.274,
+      "learning_rate": 2.4e-06,
+      "loss": 15.2712,
       "step": 7
     },
     {
       "epoch": 0.07,
-      "learning_rate": 2.4e-06,
-      "loss": 15.7738,
+      "learning_rate": 2.9999999999999997e-06,
+      "loss": 15.7528,
       "step": 8
     },
     {
       "epoch": 0.08,
-      "learning_rate": 2.9999999999999997e-06,
-      "loss": 15.7839,
+      "learning_rate": 3.6e-06,
+      "loss": 15.7574,
       "step": 9
     },
     {
       "epoch": 0.09,
-      "learning_rate": 3.6e-06,
-      "loss": 14.7744,
+      "learning_rate": 4.2e-06,
+      "loss": 14.7553,
       "step": 10
     },
     {
       "epoch": 0.1,
-      "learning_rate": 4.2e-06,
-      "loss": 15.4308,
+      "learning_rate": 4.8e-06,
+      "loss": 15.4223,
       "step": 11
     },
     {
       "epoch": 0.11,
-      "learning_rate": 4.2e-06,
-      "loss": 15.0382,
+      "learning_rate": 5.399999999999999e-06,
+      "loss": 15.0585,
       "step": 12
     },
     {
       "epoch": 0.12,
-      "learning_rate": 4.8e-06,
-      "loss": 15.2342,
+      "learning_rate": 5.999999999999999e-06,
+      "loss": 15.2637,
       "step": 13
     },
     {
       "epoch": 0.13,
-      "learning_rate": 5.399999999999999e-06,
-      "loss": 14.877,
+      "learning_rate": 6.599999999999999e-06,
+      "loss": 14.8553,
       "step": 14
     },
     {
       "epoch": 0.14,
-      "learning_rate": 5.999999999999999e-06,
-      "loss": 14.9959,
+      "learning_rate": 7.2e-06,
+      "loss": 15.0055,
       "step": 15
     },
     {
       "epoch": 0.15,
-      "learning_rate": 6.599999999999999e-06,
-      "loss": 13.908,
+      "learning_rate": 7.799999999999998e-06,
+      "loss": 13.94,
       "step": 16
     },
     {
       "epoch": 0.16,
-      "learning_rate": 7.2e-06,
-      "loss": 14.1818,
+      "learning_rate": 8.4e-06,
+      "loss": 14.2361,
       "step": 17
     },
     {
       "epoch": 0.17,
-      "learning_rate": 7.799999999999998e-06,
-      "loss": 13.6791,
+      "learning_rate": 8.999999999999999e-06,
+      "loss": 13.7686,
       "step": 18
     },
     {
       "epoch": 0.17,
-      "learning_rate": 7.799999999999998e-06,
-      "loss": 13.8616,
+      "learning_rate": 9.6e-06,
+      "loss": 13.9449,
       "step": 19
     },
     {
       "epoch": 0.18,
-      "learning_rate": 7.799999999999998e-06,
-      "loss": 16.8555,
+      "learning_rate": 1.02e-05,
+      "loss": 17.0524,
       "step": 20
     },
     {
       "epoch": 0.19,
-      "learning_rate": 8.4e-06,
-      "loss": 15.283,
+      "learning_rate": 1.0799999999999998e-05,
+      "loss": 15.4083,
       "step": 21
     },
     {
       "epoch": 0.2,
-      "learning_rate": 8.999999999999999e-06,
-      "loss": 15.098,
+      "learning_rate": 1.14e-05,
+      "loss": 15.2864,
       "step": 22
     },
     {
       "epoch": 0.21,
-      "learning_rate": 9.6e-06,
-      "loss": 13.8708,
+      "learning_rate": 1.1999999999999999e-05,
+      "loss": 14.0752,
       "step": 23
     },
     {
       "epoch": 0.22,
-      "learning_rate": 1.02e-05,
-      "loss": 13.9717,
+      "learning_rate": 1.26e-05,
+      "loss": 14.1527,
       "step": 24
     },
     {
       "epoch": 0.23,
-      "learning_rate": 1.0799999999999998e-05,
-      "loss": 13.8972,
+      "learning_rate": 1.3199999999999997e-05,
+      "loss": 14.1257,
       "step": 25
     },
     {
       "epoch": 0.24,
-      "learning_rate": 1.0799999999999998e-05,
-      "loss": 14.4483,
+      "learning_rate": 1.3799999999999998e-05,
+      "loss": 14.8519,
       "step": 26
     },
     {
       "epoch": 0.25,
-      "learning_rate": 1.14e-05,
-      "loss": 14.4519,
+      "learning_rate": 1.44e-05,
+      "loss": 14.8596,
       "step": 27
     },
     {
       "epoch": 0.26,
-      "learning_rate": 1.1999999999999999e-05,
-      "loss": 14.5318,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 15.0555,
       "step": 28
     },
     {
       "epoch": 0.27,
-      "learning_rate": 1.26e-05,
-      "loss": 14.8161,
+      "learning_rate": 1.5599999999999996e-05,
+      "loss": 15.0283,
       "step": 29
     },
     {
       "epoch": 0.28,
-      "learning_rate": 1.3199999999999997e-05,
-      "loss": 12.7068,
+      "learning_rate": 1.6199999999999997e-05,
+      "loss": 12.862,
       "step": 30
     },
     {
       "epoch": 0.28,
-      "learning_rate": 1.3799999999999998e-05,
-      "loss": 13.2855,
+      "learning_rate": 1.68e-05,
+      "loss": 13.5678,
       "step": 31
     },
     {
       "epoch": 0.29,
-      "learning_rate": 1.44e-05,
-      "loss": 13.7449,
+      "learning_rate": 1.74e-05,
+      "loss": 14.1799,
       "step": 32
     },
     {
       "epoch": 0.3,
-      "learning_rate": 1.4999999999999999e-05,
-      "loss": 15.3445,
+      "learning_rate": 1.7999999999999997e-05,
+      "loss": 16.2557,
       "step": 33
     },
     {
       "epoch": 0.31,
-      "learning_rate": 1.5599999999999996e-05,
-      "loss": 13.6449,
+      "learning_rate": 1.8599999999999998e-05,
+      "loss": 14.2811,
       "step": 34
     },
     {
       "epoch": 0.32,
-      "learning_rate": 1.6199999999999997e-05,
-      "loss": 13.4612,
+      "learning_rate": 1.8599999999999998e-05,
+      "loss": 14.3547,
       "step": 35
     },
     {
       "epoch": 0.33,
-      "learning_rate": 1.68e-05,
-      "loss": 12.6859,
+      "learning_rate": 1.92e-05,
+      "loss": 13.5783,
       "step": 36
     },
     {
       "epoch": 0.34,
-      "learning_rate": 1.74e-05,
-      "loss": 12.8878,
+      "learning_rate": 1.98e-05,
+      "loss": 13.6662,
       "step": 37
     },
     {
       "epoch": 0.35,
-      "learning_rate": 1.74e-05,
-      "loss": 13.7221,
+      "learning_rate": 2.04e-05,
+      "loss": 14.466,
       "step": 38
     },
     {
       "epoch": 0.36,
-      "learning_rate": 1.7999999999999997e-05,
-      "loss": 15.6473,
+      "learning_rate": 2.1e-05,
+      "loss": 15.6346,
       "step": 39
     },
     {
       "epoch": 0.37,
-      "learning_rate": 1.8599999999999998e-05,
-      "loss": 13.2679,
+      "learning_rate": 2.1599999999999996e-05,
+      "loss": 13.9928,
       "step": 40
     },
     {
       "epoch": 0.38,
-      "learning_rate": 1.92e-05,
-      "loss": 12.2338,
+      "learning_rate": 2.2199999999999998e-05,
+      "loss": 13.0012,
       "step": 41
     },
     {
       "epoch": 0.39,
-      "learning_rate": 1.98e-05,
-      "loss": 12.5363,
+      "learning_rate": 2.28e-05,
+      "loss": 13.1536,
       "step": 42
     },
     {
       "epoch": 0.39,
-      "learning_rate": 2.04e-05,
-      "loss": 11.6428,
+      "learning_rate": 2.34e-05,
+      "loss": 12.2083,
       "step": 43
     },
     {
       "epoch": 0.4,
-      "learning_rate": 2.1e-05,
-      "loss": 13.3083,
+      "learning_rate": 2.3999999999999997e-05,
+      "loss": 13.6525,
       "step": 44
     },
     {
       "epoch": 0.41,
-      "learning_rate": 2.1599999999999996e-05,
-      "loss": 13.0461,
+      "learning_rate": 2.4599999999999998e-05,
+      "loss": 13.1448,
       "step": 45
     },
     {
       "epoch": 0.42,
-      "learning_rate": 2.2199999999999998e-05,
-      "loss": 12.597,
+      "learning_rate": 2.52e-05,
+      "loss": 12.6738,
       "step": 46
     },
     {
       "epoch": 0.43,
-      "learning_rate": 2.28e-05,
-      "loss": 11.8786,
+      "learning_rate": 2.5799999999999997e-05,
+      "loss": 11.7223,
       "step": 47
     },
     {
       "epoch": 0.44,
-      "learning_rate": 2.34e-05,
-      "loss": 12.2429,
+      "learning_rate": 2.6399999999999995e-05,
+      "loss": 11.6004,
       "step": 48
     },
     {
       "epoch": 0.45,
-      "learning_rate": 2.3999999999999997e-05,
-      "loss": 10.6743,
+      "learning_rate": 2.6999999999999996e-05,
+      "loss": 10.1678,
       "step": 49
     },
     {
       "epoch": 0.46,
-      "learning_rate": 2.4599999999999998e-05,
-      "loss": 13.4279,
+      "learning_rate": 2.7599999999999997e-05,
+      "loss": 12.2197,
       "step": 50
     },
     {
       "epoch": 0.47,
-      "learning_rate": 2.4599999999999998e-05,
-      "loss": 12.2301,
+      "learning_rate": 2.8199999999999998e-05,
+      "loss": 11.1081,
       "step": 51
     },
     {
       "epoch": 0.48,
-      "learning_rate": 2.52e-05,
-      "loss": 12.2183,
+      "learning_rate": 2.88e-05,
+      "loss": 10.6731,
       "step": 52
     },
     {
       "epoch": 0.49,
-      "learning_rate": 2.5799999999999997e-05,
-      "loss": 10.34,
+      "learning_rate": 2.94e-05,
+      "loss": 9.0168,
       "step": 53
     },
     {
       "epoch": 0.5,
-      "learning_rate": 2.6399999999999995e-05,
-      "loss": 11.4865,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 9.3134,
       "step": 54
     },
     {
       "epoch": 0.5,
-      "learning_rate": 2.6999999999999996e-05,
-      "loss": 11.7178,
+      "learning_rate": 3.06e-05,
+      "loss": 8.6994,
       "step": 55
     },
     {
       "epoch": 0.51,
-      "learning_rate": 2.7599999999999997e-05,
-      "loss": 10.5014,
+      "learning_rate": 3.119999999999999e-05,
+      "loss": 8.176,
       "step": 56
     },
     {
       "epoch": 0.52,
-      "learning_rate": 2.8199999999999998e-05,
-      "loss": 12.5073,
+      "learning_rate": 3.1799999999999994e-05,
+      "loss": 9.1473,
       "step": 57
     },
     {
       "epoch": 0.53,
-      "learning_rate": 2.88e-05,
-      "loss": 10.8103,
+      "learning_rate": 3.2399999999999995e-05,
+      "loss": 7.8212,
       "step": 58
     },
     {
       "epoch": 0.54,
-      "learning_rate": 2.94e-05,
-      "loss": 10.9225,
+      "learning_rate": 3.2999999999999996e-05,
+      "loss": 7.4427,
       "step": 59
     },
     {
       "epoch": 0.55,
-      "learning_rate": 2.9999999999999997e-05,
-      "loss": 11.6048,
+      "learning_rate": 3.36e-05,
+      "loss": 7.192,
       "step": 60
     },
     {
       "epoch": 0.56,
-      "learning_rate": 3.06e-05,
-      "loss": 9.8662,
+      "learning_rate": 3.42e-05,
+      "loss": 6.3576,
       "step": 61
     },
     {
       "epoch": 0.57,
-      "learning_rate": 3.119999999999999e-05,
-      "loss": 10.3016,
+      "learning_rate": 3.48e-05,
+      "loss": 6.3045,
       "step": 62
     },
     {
       "epoch": 0.58,
-      "learning_rate": 3.1799999999999994e-05,
-      "loss": 11.557,
+      "learning_rate": 3.539999999999999e-05,
+      "loss": 6.7366,
       "step": 63
     },
     {
       "epoch": 0.59,
-      "learning_rate": 3.2399999999999995e-05,
-      "loss": 10.9271,
+      "learning_rate": 3.5999999999999994e-05,
+      "loss": 6.1411,
       "step": 64
     },
     {
       "epoch": 0.6,
-      "learning_rate": 3.2999999999999996e-05,
-      "loss": 10.0703,
+      "learning_rate": 3.6599999999999995e-05,
+      "loss": 5.6342,
       "step": 65
     },
     {
       "epoch": 0.61,
-      "learning_rate": 3.36e-05,
-      "loss": 9.4592,
+      "learning_rate": 3.7199999999999996e-05,
+      "loss": 5.138,
       "step": 66
     },
     {
       "epoch": 0.61,
-      "learning_rate": 3.42e-05,
-      "loss": 9.7794,
+      "learning_rate": 3.78e-05,
+      "loss": 5.1519,
       "step": 67
     },
     {
       "epoch": 0.62,
-      "learning_rate": 3.48e-05,
-      "loss": 8.7059,
+      "learning_rate": 3.84e-05,
+      "loss": 4.8201,
       "step": 68
     },
     {
       "epoch": 0.63,
-      "learning_rate": 3.539999999999999e-05,
-      "loss": 10.2012,
+      "learning_rate": 3.9e-05,
+      "loss": 5.2521,
       "step": 69
     },
     {
       "epoch": 0.64,
-      "learning_rate": 3.5999999999999994e-05,
-      "loss": 9.7349,
+      "learning_rate": 3.96e-05,
+      "loss": 4.8285,
       "step": 70
     },
     {
       "epoch": 0.65,
-      "learning_rate": 3.6599999999999995e-05,
-      "loss": 9.2843,
+      "learning_rate": 4.02e-05,
+      "loss": 4.6744,
       "step": 71
     },
     {
       "epoch": 0.66,
-      "learning_rate": 3.7199999999999996e-05,
-      "loss": 9.5143,
+      "learning_rate": 4.08e-05,
+      "loss": 4.5332,
       "step": 72
     },
     {
       "epoch": 0.67,
-      "learning_rate": 3.78e-05,
-      "loss": 8.7418,
+      "learning_rate": 4.14e-05,
+      "loss": 4.4558,
       "step": 73
     },
     {
       "epoch": 0.68,
-      "learning_rate": 3.84e-05,
-      "loss": 8.0621,
+      "learning_rate": 4.2e-05,
+      "loss": 4.1843,
       "step": 74
     },
     {
       "epoch": 0.69,
-      "learning_rate": 3.9e-05,
-      "loss": 7.6643,
+      "learning_rate": 4.259999999999999e-05,
+      "loss": 4.0748,
       "step": 75
     },
     {
       "epoch": 0.7,
-      "learning_rate": 3.96e-05,
-      "loss": 9.118,
+      "learning_rate": 4.319999999999999e-05,
+      "loss": 4.3241,
       "step": 76
     },
     {
       "epoch": 0.71,
-      "learning_rate": 4.02e-05,
-      "loss": 8.1761,
+      "learning_rate": 4.3799999999999994e-05,
+      "loss": 4.1538,
       "step": 77
     },
     {
       "epoch": 0.72,
-      "learning_rate": 4.08e-05,
-      "loss": 8.5182,
+      "learning_rate": 4.4399999999999995e-05,
+      "loss": 4.2377,
       "step": 78
     },
     {
       "epoch": 0.72,
-      "learning_rate": 4.14e-05,
-      "loss": 7.0265,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 3.9475,
       "step": 79
     },
     {
       "epoch": 0.73,
-      "learning_rate": 4.2e-05,
-      "loss": 7.1477,
+      "learning_rate": 4.56e-05,
+      "loss": 3.9659,
       "step": 80
     },
     {
       "epoch": 0.74,
-      "learning_rate": 4.259999999999999e-05,
-      "loss": 7.0084,
+      "learning_rate": 4.62e-05,
+      "loss": 3.9191,
       "step": 81
     },
     {
       "epoch": 0.75,
-      "learning_rate": 4.319999999999999e-05,
-      "loss": 6.9016,
+      "learning_rate": 4.68e-05,
+      "loss": 3.9597,
       "step": 82
     },
     {
       "epoch": 0.76,
-      "learning_rate": 4.3799999999999994e-05,
-      "loss": 6.3869,
+      "learning_rate": 4.7399999999999993e-05,
+      "loss": 3.8435,
       "step": 83
     },
     {
       "epoch": 0.77,
-      "learning_rate": 4.4399999999999995e-05,
-      "loss": 5.6583,
+      "learning_rate": 4.7999999999999994e-05,
+      "loss": 3.7875,
       "step": 84
     },
     {
       "epoch": 0.78,
-      "learning_rate": 4.4999999999999996e-05,
-      "loss": 5.4926,
+      "learning_rate": 4.8599999999999995e-05,
+      "loss": 3.8098,
       "step": 85
     },
     {
       "epoch": 0.79,
-      "learning_rate": 4.56e-05,
-      "loss": 4.8137,
+      "learning_rate": 4.9199999999999997e-05,
+      "loss": 3.7099,
       "step": 86
     },
     {
       "epoch": 0.8,
-      "learning_rate": 4.62e-05,
-      "loss": 4.8259,
+      "learning_rate": 4.98e-05,
+      "loss": 3.7847,
       "step": 87
     },
     {
       "epoch": 0.81,
-      "learning_rate": 4.68e-05,
-      "loss": 5.211,
+      "learning_rate": 5.04e-05,
+      "loss": 4.1322,
       "step": 88
     },
     {
       "epoch": 0.82,
-      "learning_rate": 4.7399999999999993e-05,
-      "loss": 4.5014,
+      "learning_rate": 5.1e-05,
+      "loss": 3.6942,
       "step": 89
     },
     {
       "epoch": 0.83,
-      "learning_rate": 4.7999999999999994e-05,
-      "loss": 4.2795,
+      "learning_rate": 5.1599999999999994e-05,
+      "loss": 3.6789,
       "step": 90
     },
     {
       "epoch": 0.83,
-      "learning_rate": 4.8599999999999995e-05,
-      "loss": 4.0828,
+      "learning_rate": 5.2199999999999995e-05,
+      "loss": 3.6122,
       "step": 91
     },
     {
       "epoch": 0.84,
-      "learning_rate": 4.9199999999999997e-05,
-      "loss": 4.0464,
+      "learning_rate": 5.279999999999999e-05,
+      "loss": 3.6174,
       "step": 92
     },
     {
       "epoch": 0.85,
-      "learning_rate": 4.98e-05,
-      "loss": 3.9511,
+      "learning_rate": 5.339999999999999e-05,
+      "loss": 3.5916,
       "step": 93
     },
     {
       "epoch": 0.86,
-      "learning_rate": 5.04e-05,
-      "loss": 4.1563,
+      "learning_rate": 5.399999999999999e-05,
+      "loss": 3.7348,
       "step": 94
     },
     {
       "epoch": 0.87,
-      "learning_rate": 5.1e-05,
-      "loss": 3.8376,
+      "learning_rate": 5.459999999999999e-05,
+      "loss": 3.5631,
       "step": 95
     },
     {
       "epoch": 0.88,
-      "learning_rate": 5.1599999999999994e-05,
-      "loss": 3.7342,
+      "learning_rate": 5.519999999999999e-05,
+      "loss": 3.5056,
       "step": 96
     },
     {
       "epoch": 0.89,
-      "learning_rate": 5.2199999999999995e-05,
-      "loss": 3.7087,
+      "learning_rate": 5.5799999999999994e-05,
+      "loss": 3.503,
       "step": 97
     },
     {
       "epoch": 0.9,
-      "learning_rate": 5.279999999999999e-05,
-      "loss": 3.6373,
+      "learning_rate": 5.6399999999999995e-05,
+      "loss": 3.4679,
       "step": 98
     },
     {
       "epoch": 0.91,
-      "learning_rate": 5.339999999999999e-05,
-      "loss": 3.605,
+      "learning_rate": 5.6999999999999996e-05,
+      "loss": 3.4456,
       "step": 99
     },
     {
       "epoch": 0.92,
-      "learning_rate": 5.399999999999999e-05,
-      "loss": 3.6921,
+      "learning_rate": 5.76e-05,
+      "loss": 3.5372,
       "step": 100
     },
     {
       "epoch": 0.92,
-      "eval_loss": 3.814695358276367,
-      "eval_runtime": 21.1782,
-      "eval_samples_per_second": 77.769,
-      "eval_steps_per_second": 1.228,
+      "eval_loss": 3.619191884994507,
+      "eval_runtime": 21.6418,
+      "eval_samples_per_second": 76.103,
+      "eval_steps_per_second": 1.201,
       "eval_wer": 1.0,
       "step": 100
     },
     {
       "epoch": 0.93,
-      "learning_rate": 5.459999999999999e-05,
-      "loss": 3.6539,
+      "learning_rate": 5.82e-05,
+      "loss": 3.5018,
       "step": 101
     },
     {
       "epoch": 0.94,
-      "learning_rate": 5.519999999999999e-05,
-      "loss": 3.5837,
+      "learning_rate": 5.88e-05,
+      "loss": 3.4393,
       "step": 102
     },
     {
       "epoch": 0.94,
-      "learning_rate": 5.5799999999999994e-05,
-      "loss": 3.5867,
+      "learning_rate": 5.94e-05,
+      "loss": 3.4607,
       "step": 103
     },
     {
       "epoch": 0.95,
-      "learning_rate": 5.6399999999999995e-05,
-      "loss": 3.5088,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 3.4081,
       "step": 104
     },
     {
       "epoch": 0.96,
-      "learning_rate": 5.6999999999999996e-05,
-      "loss": 3.5013,
+      "learning_rate": 6.0599999999999996e-05,
+      "loss": 3.4106,
       "step": 105
     },
     {
       "epoch": 0.97,
-      "learning_rate": 5.76e-05,
-      "loss": 3.545,
+      "learning_rate": 6.12e-05,
+      "loss": 3.4513,
       "step": 106
     },
     {
       "epoch": 0.98,
-      "learning_rate": 5.82e-05,
-      "loss": 3.5638,
+      "learning_rate": 6.18e-05,
+      "loss": 3.4899,
       "step": 107
     },
     {
       "epoch": 0.99,
-      "learning_rate": 5.88e-05,
-      "loss": 3.4299,
+      "learning_rate": 6.239999999999999e-05,
+      "loss": 3.3631,
       "step": 108
     },
     {
       "epoch": 1.0,
-      "learning_rate": 5.94e-05,
-      "loss": 3.5303,
+      "learning_rate": 6.299999999999999e-05,
+      "loss": 3.4628,
       "step": 109
     },
     {
       "epoch": 1.01,
-      "learning_rate": 5.9999999999999995e-05,
-      "loss": 3.4169,
+      "learning_rate": 6.359999999999999e-05,
+      "loss": 3.3548,
       "step": 110
     },
     {
       "epoch": 1.02,
-      "learning_rate": 6.0599999999999996e-05,
-      "loss": 3.426,
+      "learning_rate": 6.419999999999999e-05,
+      "loss": 3.3648,
       "step": 111
     },
     {
       "epoch": 1.03,
-      "learning_rate": 6.12e-05,
-      "loss": 3.3772,
+      "learning_rate": 6.479999999999999e-05,
+      "loss": 3.3285,
       "step": 112
     },
     {
       "epoch": 1.04,
-      "learning_rate": 6.18e-05,
-      "loss": 3.3507,
+      "learning_rate": 6.539999999999999e-05,
+      "loss": 3.3107,
       "step": 113
     },
     {
       "epoch": 1.05,
-      "learning_rate": 6.239999999999999e-05,
-      "loss": 3.3428,
+      "learning_rate": 6.599999999999999e-05,
+      "loss": 3.307,
       "step": 114
     },
     {
       "epoch": 1.06,
-      "learning_rate": 6.299999999999999e-05,
-      "loss": 3.6732,
+      "learning_rate": 6.659999999999999e-05,
+      "loss": 3.6228,
       "step": 115
     },
     {
       "epoch": 1.06,
-      "learning_rate": 6.359999999999999e-05,
-      "loss": 3.3705,
+      "learning_rate": 6.72e-05,
+      "loss": 3.3355,
       "step": 116
     },
     {
       "epoch": 1.07,
-      "learning_rate": 6.419999999999999e-05,
-      "loss": 3.3319,
+      "learning_rate": 6.78e-05,
+      "loss": 3.2987,
       "step": 117
     },
     {
       "epoch": 1.08,
-      "learning_rate": 6.479999999999999e-05,
-      "loss": 3.2998,
+      "learning_rate": 6.84e-05,
+      "loss": 3.2757,
       "step": 118
     },
     {
       "epoch": 1.09,
-      "learning_rate": 6.539999999999999e-05,
-      "loss": 3.3073,
+      "learning_rate": 6.9e-05,
+      "loss": 3.2871,
       "step": 119
     },
     {
       "epoch": 1.1,
-      "learning_rate": 6.599999999999999e-05,
-      "loss": 3.2582,
+      "learning_rate": 6.96e-05,
+      "loss": 3.2383,
       "step": 120
     },
     {
       "epoch": 1.11,
-      "learning_rate": 6.659999999999999e-05,
-      "loss": 3.3608,
+      "learning_rate": 7.02e-05,
+      "loss": 3.337,
       "step": 121
     },
     {
       "epoch": 1.12,
-      "learning_rate": 6.72e-05,
-      "loss": 3.2902,
+      "learning_rate": 7.079999999999999e-05,
+      "loss": 3.266,
       "step": 122
     },
     {
       "epoch": 1.13,
-      "learning_rate": 6.78e-05,
-      "loss": 3.2695,
+      "learning_rate": 7.139999999999999e-05,
+      "loss": 3.2558,
       "step": 123
     },
     {
       "epoch": 1.14,
-      "learning_rate": 6.84e-05,
-      "loss": 3.2515,
+      "learning_rate": 7.199999999999999e-05,
+      "loss": 3.2389,
       "step": 124
     },
     {
       "epoch": 1.15,
-      "learning_rate": 6.9e-05,
-      "loss": 3.2504,
+      "learning_rate": 7.259999999999999e-05,
+      "loss": 3.239,
       "step": 125
     },
     {
       "epoch": 1.16,
-      "learning_rate": 6.96e-05,
-      "loss": 3.2494,
+      "learning_rate": 7.319999999999999e-05,
+      "loss": 3.2372,
       "step": 126
     },
     {
       "epoch": 1.17,
-      "learning_rate": 7.02e-05,
-      "loss": 3.2295,
+      "learning_rate": 7.379999999999999e-05,
+      "loss": 3.2169,
       "step": 127
     },
     {
       "epoch": 1.17,
-      "learning_rate": 7.079999999999999e-05,
-      "loss": 3.2915,
+      "learning_rate": 7.439999999999999e-05,
+      "loss": 3.2734,
       "step": 128
     },
     {
       "epoch": 1.18,
-      "learning_rate": 7.139999999999999e-05,
-      "loss": 3.2156,
+      "learning_rate": 7.5e-05,
+      "loss": 3.2043,
       "step": 129
     },
     {
       "epoch": 1.19,
-      "learning_rate": 7.199999999999999e-05,
-      "loss": 3.2271,
+      "learning_rate": 7.56e-05,
+      "loss": 3.214,
       "step": 130
     },
     {
       "epoch": 1.2,
-      "learning_rate": 7.259999999999999e-05,
-      "loss": 3.1876,
+      "learning_rate": 7.62e-05,
+      "loss": 3.1772,
       "step": 131
     },
     {
       "epoch": 1.21,
-      "learning_rate": 7.319999999999999e-05,
-      "loss": 3.2369,
+      "learning_rate": 7.68e-05,
+      "loss": 3.2279,
       "step": 132
     },
     {
       "epoch": 1.22,
-      "learning_rate": 7.379999999999999e-05,
-      "loss": 3.1826,
+      "learning_rate": 7.74e-05,
+      "loss": 3.1733,
       "step": 133
     },
     {
       "epoch": 1.23,
-      "learning_rate": 7.439999999999999e-05,
-      "loss": 3.2399,
+      "learning_rate": 7.8e-05,
+      "loss": 3.216,
       "step": 134
     },
     {
       "epoch": 1.24,
-      "learning_rate": 7.5e-05,
-      "loss": 3.2326,
+      "learning_rate": 7.86e-05,
+      "loss": 3.2469,
       "step": 135
     },
     {
       "epoch": 1.25,
-      "learning_rate": 7.56e-05,
-      "loss": 3.1849,
+      "learning_rate": 7.92e-05,
+      "loss": 3.1872,
       "step": 136
     },
     {
       "epoch": 1.26,
-      "learning_rate": 7.62e-05,
-      "loss": 3.197,
+      "learning_rate": 7.98e-05,
+      "loss": 3.193,
       "step": 137
     },
     {
       "epoch": 1.27,
-      "learning_rate": 7.68e-05,
-      "loss": 3.1575,
+      "learning_rate": 8.04e-05,
+      "loss": 3.1581,
       "step": 138
     },
     {
       "epoch": 1.28,
-      "learning_rate": 7.74e-05,
-      "loss": 3.1452,
+      "learning_rate": 8.1e-05,
+      "loss": 3.1466,
       "step": 139
     },
     {
       "epoch": 1.28,
-      "learning_rate": 7.8e-05,
-      "loss": 3.2644,
+      "learning_rate": 8.16e-05,
+      "loss": 3.244,
       "step": 140
     },
     {
       "epoch": 1.29,
-      "learning_rate": 7.86e-05,
-      "loss": 3.2408,
+      "learning_rate": 8.22e-05,
+      "loss": 3.2174,
       "step": 141
     },
     {
       "epoch": 1.3,
-      "learning_rate": 7.92e-05,
-      "loss": 3.1836,
+      "learning_rate": 8.28e-05,
+      "loss": 3.175,
       "step": 142
     },
     {
       "epoch": 1.31,
-      "learning_rate": 7.98e-05,
-      "loss": 3.1684,
+      "learning_rate": 8.34e-05,
+      "loss": 3.1617,
       "step": 143
     },
     {
       "epoch": 1.32,
-      "learning_rate": 8.04e-05,
-      "loss": 3.181,
+      "learning_rate": 8.4e-05,
+      "loss": 3.1759,
       "step": 144
     },
     {
       "epoch": 1.33,
-      "learning_rate": 8.1e-05,
-      "loss": 3.1455,
+      "learning_rate": 8.459999999999998e-05,
+      "loss": 3.1345,
       "step": 145
     },
     {
       "epoch": 1.34,
-      "learning_rate": 8.16e-05,
-      "loss": 3.1365,
+      "learning_rate": 8.519999999999998e-05,
+      "loss": 3.1084,
       "step": 146
     },
     {
       "epoch": 1.35,
-      "learning_rate": 8.22e-05,
-      "loss": 3.2451,
+      "learning_rate": 8.579999999999998e-05,
+      "loss": 3.2398,
       "step": 147
     },
     {
       "epoch": 1.36,
-      "learning_rate": 8.28e-05,
-      "loss": 3.1508,
+      "learning_rate": 8.639999999999999e-05,
+      "loss": 3.1668,
       "step": 148
     },
     {
       "epoch": 1.37,
-      "learning_rate": 8.34e-05,
-      "loss": 3.1395,
+      "learning_rate": 8.699999999999999e-05,
+      "loss": 3.1325,
       "step": 149
     },
     {
       "epoch": 1.38,
-      "learning_rate": 8.4e-05,
-      "loss": 3.1149,
+      "learning_rate": 8.759999999999999e-05,
+      "loss": 3.1138,
       "step": 150
     },
     {
       "epoch": 1.39,
-      "learning_rate": 8.459999999999998e-05,
-      "loss": 3.1598,
+      "learning_rate": 8.819999999999999e-05,
+      "loss": 3.1448,
       "step": 151
     },
     {
       "epoch": 1.39,
-      "learning_rate": 8.519999999999998e-05,
-      "loss": 3.1317,
+      "learning_rate": 8.879999999999999e-05,
+      "loss": 3.1121,
       "step": 152
     },
     {
       "epoch": 1.4,
-      "learning_rate": 8.579999999999998e-05,
-      "loss": 3.1269,
+      "learning_rate": 8.939999999999999e-05,
+      "loss": 3.1075,
       "step": 153
     },
     {
       "epoch": 1.41,
-      "learning_rate": 8.639999999999999e-05,
-      "loss": 3.1373,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 3.1384,
       "step": 154
     },
     {
       "epoch": 1.42,
-      "learning_rate": 8.699999999999999e-05,
-      "loss": 3.1243,
+      "learning_rate": 9.059999999999999e-05,
+      "loss": 3.1181,
       "step": 155
     },
     {
       "epoch": 1.43,
-      "learning_rate": 8.759999999999999e-05,
-      "loss": 3.1514,
+      "learning_rate": 9.12e-05,
+      "loss": 3.1343,
       "step": 156
     },
     {
       "epoch": 1.44,
-      "learning_rate": 8.819999999999999e-05,
-      "loss": 3.1339,
+      "learning_rate": 9.18e-05,
+      "loss": 3.1271,
       "step": 157
     },
     {
       "epoch": 1.45,
-      "learning_rate": 8.879999999999999e-05,
-      "loss": 3.1666,
+      "learning_rate": 9.24e-05,
+      "loss": 3.1349,
       "step": 158
     },
     {
       "epoch": 1.46,
-      "learning_rate": 8.939999999999999e-05,
-      "loss": 3.2427,
+      "learning_rate": 9.3e-05,
+      "loss": 3.2225,
       "step": 159
     },
     {
       "epoch": 1.47,
-      "learning_rate": 8.999999999999999e-05,
-      "loss": 3.1205,
+      "learning_rate": 9.36e-05,
+      "loss": 3.1083,
       "step": 160
     },
     {
       "epoch": 1.48,
-      "learning_rate": 9.059999999999999e-05,
-      "loss": 3.1025,
+      "learning_rate": 9.419999999999999e-05,
+      "loss": 3.098,
       "step": 161
     },
     {
       "epoch": 1.49,
-      "learning_rate": 9.12e-05,
-      "loss": 3.1352,
+      "learning_rate": 9.479999999999999e-05,
+      "loss": 3.1139,
       "step": 162
     },
     {
       "epoch": 1.5,
-      "learning_rate": 9.18e-05,
-      "loss": 3.1462,
+      "learning_rate": 9.539999999999999e-05,
+      "loss": 3.1217,
       "step": 163
     },
     {
       "epoch": 1.5,
-      "learning_rate": 9.24e-05,
-      "loss": 3.1189,
+      "learning_rate": 9.599999999999999e-05,
+      "loss": 3.0913,
       "step": 164
     },
     {
       "epoch": 1.51,
-      "learning_rate": 9.3e-05,
-      "loss": 3.1432,
+      "learning_rate": 9.659999999999999e-05,
+      "loss": 3.1138,
       "step": 165
     },
     {
       "epoch": 1.52,
-      "learning_rate": 9.36e-05,
-      "loss": 3.1346,
+      "learning_rate": 9.719999999999999e-05,
+      "loss": 3.0849,
       "step": 166
     },
     {
       "epoch": 1.53,
-      "learning_rate": 9.419999999999999e-05,
-      "loss": 3.124,
+      "learning_rate": 9.779999999999999e-05,
+      "loss": 3.1029,
       "step": 167
     },
     {
       "epoch": 1.54,
-      "learning_rate": 9.479999999999999e-05,
-      "loss": 3.086,
+      "learning_rate": 9.839999999999999e-05,
+      "loss": 3.08,
       "step": 168
     },
     {
       "epoch": 1.55,
-      "learning_rate": 9.539999999999999e-05,
-      "loss": 3.114,
+      "learning_rate": 9.9e-05,
+      "loss": 3.0984,
       "step": 169
     },
     {
       "epoch": 1.56,
-      "learning_rate": 9.599999999999999e-05,
-      "loss": 3.1096,
+      "learning_rate": 9.96e-05,
+      "loss": 3.0846,
       "step": 170
     },
     {
       "epoch": 1.57,
-      "learning_rate": 9.659999999999999e-05,
-      "loss": 3.1284,
+      "learning_rate": 0.0001002,
+      "loss": 3.097,
       "step": 171
     },
     {
       "epoch": 1.58,
-      "learning_rate": 9.719999999999999e-05,
-      "loss": 3.1583,
+      "learning_rate": 0.0001008,
+      "loss": 3.1146,
       "step": 172
     },
     {
       "epoch": 1.59,
-      "learning_rate": 9.779999999999999e-05,
-      "loss": 3.1135,
+      "learning_rate": 0.0001014,
+      "loss": 3.0922,
       "step": 173
     },
     {
       "epoch": 1.6,
-      "learning_rate": 9.839999999999999e-05,
-      "loss": 3.1133,
+      "learning_rate": 0.000102,
+      "loss": 3.0901,
       "step": 174
     },
     {
       "epoch": 1.61,
-      "learning_rate": 9.9e-05,
-      "loss": 3.0514,
+      "learning_rate": 0.0001026,
+      "loss": 3.0376,
       "step": 175
     },
     {
       "epoch": 1.61,
-      "learning_rate": 9.96e-05,
-      "loss": 3.0986,
+      "learning_rate": 0.00010319999999999999,
+      "loss": 3.0919,
       "step": 176
     },
     {
       "epoch": 1.62,
-      "learning_rate": 0.0001002,
-      "loss": 3.1156,
+      "learning_rate": 0.00010379999999999999,
+      "loss": 3.0961,
       "step": 177
     },
     {
       "epoch": 1.63,
-      "learning_rate": 0.0001008,
-      "loss": 3.1658,
+      "learning_rate": 0.00010439999999999999,
+      "loss": 3.0988,
       "step": 178
     },
     {
       "epoch": 1.64,
-      "learning_rate": 0.0001014,
-      "loss": 3.0773,
+      "learning_rate": 0.00010499999999999999,
+      "loss": 3.0418,
       "step": 179
     },
     {
       "epoch": 1.65,
-      "learning_rate": 0.000102,
-      "loss": 3.055,
+      "learning_rate": 0.00010559999999999998,
+      "loss": 3.0582,
       "step": 180
     },
     {
       "epoch": 1.66,
-      "learning_rate": 0.0001026,
-      "loss": 3.1044,
+      "learning_rate": 0.00010619999999999998,
+      "loss": 3.0874,
       "step": 181
     },
     {
       "epoch": 1.67,
-      "learning_rate": 0.00010319999999999999,
-      "loss": 3.0703,
+      "learning_rate": 0.00010679999999999998,
+      "loss": 3.051,
       "step": 182
     },
     {
       "epoch": 1.68,
-      "learning_rate": 0.00010379999999999999,
-      "loss": 3.1018,
+      "learning_rate": 0.00010739999999999998,
+      "loss": 3.0563,
       "step": 183
     },
     {
       "epoch": 1.69,
-      "learning_rate": 0.00010439999999999999,
-      "loss": 3.2493,
+      "learning_rate": 0.00010799999999999998,
+      "loss": 3.1318,
       "step": 184
     },
     {
       "epoch": 1.7,
-      "learning_rate": 0.00010499999999999999,
-      "loss": 3.0706,
+      "learning_rate": 0.00010859999999999998,
+      "loss": 3.0463,
       "step": 185
     },
     {
       "epoch": 1.71,
-      "learning_rate": 0.00010559999999999998,
-      "loss": 3.0781,
+      "learning_rate": 0.00010919999999999998,
+      "loss": 3.0536,
       "step": 186
     },
     {
       "epoch": 1.72,
-      "learning_rate": 0.00010619999999999998,
-      "loss": 3.0921,
+      "learning_rate": 0.00010979999999999999,
+      "loss": 3.0563,
       "step": 187
     },
     {
       "epoch": 1.72,
-      "learning_rate": 0.00010679999999999998,
-      "loss": 3.1075,
+      "learning_rate": 0.00011039999999999999,
+      "loss": 3.0522,
       "step": 188
     },
     {
       "epoch": 1.73,
-      "learning_rate": 0.00010739999999999998,
-      "loss": 3.089,
+      "learning_rate": 0.00011099999999999999,
+      "loss": 3.0587,
       "step": 189
     },
     {
       "epoch": 1.74,
-      "learning_rate": 0.00010799999999999998,
-      "loss": 3.0701,
+      "learning_rate": 0.00011159999999999999,
+      "loss": 3.023,
       "step": 190
     },
     {
       "epoch": 1.75,
-      "learning_rate": 0.00010859999999999998,
-      "loss": 3.2495,
+      "learning_rate": 0.00011219999999999999,
+      "loss": 3.0533,
       "step": 191
     },
     {
       "epoch": 1.76,
-      "learning_rate": 0.00010919999999999998,
-      "loss": 3.1257,
+      "learning_rate": 0.00011279999999999999,
+      "loss": 3.0331,
       "step": 192
     },
     {
       "epoch": 1.77,
-      "learning_rate": 0.00010979999999999999,
-      "loss": 3.0437,
+      "learning_rate": 0.00011339999999999999,
+      "loss": 2.984,
       "step": 193
     },
     {
       "epoch": 1.78,
-      "learning_rate": 0.00011039999999999999,
-      "loss": 3.1002,
+      "learning_rate": 0.00011399999999999999,
+      "loss": 3.0564,
       "step": 194
     },
     {
       "epoch": 1.79,
-      "learning_rate": 0.00011099999999999999,
-      "loss": 3.0903,
+      "learning_rate": 0.0001146,
+      "loss": 3.0372,
       "step": 195
     },
     {
       "epoch": 1.8,
-      "learning_rate": 0.00011159999999999999,
-      "loss": 3.0835,
+      "learning_rate": 0.0001152,
+      "loss": 2.998,
       "step": 196
     },
     {
       "epoch": 1.81,
-      "learning_rate": 0.00011219999999999999,
-      "loss": 3.1389,
+      "learning_rate": 0.0001158,
+      "loss": 3.0404,
       "step": 197
     },
     {
       "epoch": 1.82,
-      "learning_rate": 0.00011279999999999999,
-      "loss": 3.072,
+      "learning_rate": 0.0001164,
+      "loss": 3.0007,
       "step": 198
     },
     {
       "epoch": 1.83,
-      "learning_rate": 0.00011339999999999999,
-      "loss": 3.058,
+      "learning_rate": 0.000117,
+      "loss": 3.0027,
       "step": 199
     },
     {
       "epoch": 1.83,
-      "learning_rate": 0.00011399999999999999,
-      "loss": 3.0585,
+      "learning_rate": 0.0001176,
+      "loss": 2.9777,
       "step": 200
     },
     {
       "epoch": 1.83,
-      "eval_loss": 3.1415915489196777,
-      "eval_runtime": 20.908,
-      "eval_samples_per_second": 78.774,
-      "eval_steps_per_second": 1.244,
+      "eval_loss": 3.0280678272247314,
+      "eval_runtime": 20.8277,
+      "eval_samples_per_second": 79.077,
+      "eval_steps_per_second": 1.248,
       "eval_wer": 0.9998978653865795,
       "step": 200
     },
     {
       "epoch": 1.84,
-      "learning_rate": 0.0001146,
-      "loss": 3.0284,
+      "learning_rate": 0.0001182,
+      "loss": 2.9476,
       "step": 201
     },
     {
       "epoch": 1.85,
-      "learning_rate": 0.0001152,
-      "loss": 3.0219,
+      "learning_rate": 0.0001188,
+      "loss": 2.9329,
       "step": 202
     },
     {
       "epoch": 1.86,
-      "learning_rate": 0.0001158,
-      "loss": 3.1233,
+      "learning_rate": 0.0001194,
+      "loss": 2.9594,
       "step": 203
     },
     {
       "epoch": 1.87,
-      "learning_rate": 0.0001164,
-      "loss": 3.0807,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 3.0009,
       "step": 204
     },
     {
       "epoch": 1.88,
-      "learning_rate": 0.000117,
-      "loss": 3.0612,
+      "learning_rate": 0.00012059999999999999,
+      "loss": 3.0238,
       "step": 205
     },
     {
       "epoch": 1.89,
-      "learning_rate": 0.0001176,
-      "loss": 3.0313,
+      "learning_rate": 0.00012119999999999999,
+      "loss": 2.9371,
       "step": 206
     },
     {
       "epoch": 1.9,
-      "learning_rate": 0.0001182,
-      "loss": 3.0677,
+      "learning_rate": 0.00012179999999999999,
+      "loss": 3.017,
       "step": 207
     },
     {
       "epoch": 1.91,
-      "learning_rate": 0.0001188,
-      "loss": 3.0791,
+      "learning_rate": 0.0001224,
+      "loss": 3.0285,
       "step": 208
     },
     {
       "epoch": 1.92,
-      "learning_rate": 0.0001194,
-      "loss": 3.0257,
+      "learning_rate": 0.00012299999999999998,
+      "loss": 2.917,
       "step": 209
     },
     {
       "epoch": 1.93,
-      "learning_rate": 0.00011999999999999999,
-      "loss": 3.0242,
+      "learning_rate": 0.0001236,
+      "loss": 2.912,
       "step": 210
     },
     {
       "epoch": 1.94,
-      "learning_rate": 0.00012059999999999999,
-      "loss": 3.0427,
+      "learning_rate": 0.00012419999999999998,
+      "loss": 2.9682,
       "step": 211
     },
     {
       "epoch": 1.94,
-      "learning_rate": 0.00012119999999999999,
-      "loss": 3.0186,
+      "learning_rate": 0.00012479999999999997,
+      "loss": 2.9279,
       "step": 212
     },
     {
       "epoch": 1.95,
-      "learning_rate": 0.00012179999999999999,
-      "loss": 3.0672,
+      "learning_rate": 0.00012539999999999999,
+      "loss": 2.9678,
       "step": 213
     },
     {
       "epoch": 1.96,
-      "learning_rate": 0.0001224,
-      "loss": 3.0792,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 2.9455,
       "step": 214
     },
     {
       "epoch": 1.97,
-      "learning_rate": 0.00012299999999999998,
-      "loss": 3.0519,
+      "learning_rate": 0.0001266,
+      "loss": 2.914,
       "step": 215
     },
     {
       "epoch": 1.98,
-      "learning_rate": 0.0001236,
-      "loss": 3.06,
+      "learning_rate": 0.00012719999999999997,
+      "loss": 2.9427,
       "step": 216
     },
     {
       "epoch": 1.99,
-      "learning_rate": 0.00012419999999999998,
-      "loss": 3.0171,
+      "learning_rate": 0.0001278,
+      "loss": 2.8951,
       "step": 217
     },
     {
       "epoch": 2.0,
-      "learning_rate": 0.00012479999999999997,
-      "loss": 3.0294,
+      "learning_rate": 0.00012839999999999998,
+      "loss": 2.9034,
       "step": 218
     },
     {
       "epoch": 2.01,
-      "learning_rate": 0.00012539999999999999,
-      "loss": 3.0277,
+      "learning_rate": 0.000129,
+      "loss": 2.8842,
       "step": 219
     },
     {
       "epoch": 2.02,
-      "learning_rate": 0.00012599999999999997,
-      "loss": 3.0193,
+      "learning_rate": 0.00012959999999999998,
+      "loss": 2.8631,
       "step": 220
     },
     {
       "epoch": 2.03,
-      "learning_rate": 0.0001266,
-      "loss": 3.0065,
+      "learning_rate": 0.0001302,
+      "loss": 2.8351,
       "step": 221
     },
     {
       "epoch": 2.04,
-      "learning_rate": 0.00012719999999999997,
-      "loss": 3.0114,
+      "learning_rate": 0.00013079999999999998,
+      "loss": 2.8251,
       "step": 222
     },
     {
       "epoch": 2.05,
-      "learning_rate": 0.0001278,
-      "loss": 3.027,
+      "learning_rate": 0.0001314,
+      "loss": 2.8393,
       "step": 223
     },
     {
       "epoch": 2.06,
-      "learning_rate": 0.00012839999999999998,
-      "loss": 3.0858,
+      "learning_rate": 0.00013199999999999998,
+      "loss": 2.8444,
       "step": 224
     },
     {
       "epoch": 2.06,
-      "learning_rate": 0.000129,
-      "loss": 3.0752,
+      "learning_rate": 0.0001326,
+      "loss": 2.8333,
       "step": 225
     },
     {
       "epoch": 2.07,
-      "learning_rate": 0.00012959999999999998,
-      "loss": 3.0549,
+      "learning_rate": 0.00013319999999999999,
+      "loss": 2.8476,
       "step": 226
     },
     {
       "epoch": 2.08,
-      "learning_rate": 0.0001302,
-      "loss": 3.0076,
+      "learning_rate": 0.0001338,
+      "loss": 2.7633,
       "step": 227
     },
     {
       "epoch": 2.09,
-      "learning_rate": 0.00013079999999999998,
-      "loss": 3.0495,
+      "learning_rate": 0.0001344,
+      "loss": 2.7843,
       "step": 228
     },
     {
       "epoch": 2.1,
-      "learning_rate": 0.0001314,
-      "loss": 2.9902,
+      "learning_rate": 0.000135,
+      "loss": 2.7302,
       "step": 229
     },
     {
       "epoch": 2.11,
-      "learning_rate": 0.00013199999999999998,
-      "loss": 3.0065,
+      "learning_rate": 0.0001356,
+      "loss": 2.6952,
       "step": 230
     },
     {
       "epoch": 2.12,
-      "learning_rate": 0.0001326,
-      "loss": 3.1297,
+      "learning_rate": 0.0001362,
+      "loss": 2.6943,
       "step": 231
     },
     {
       "epoch": 2.13,
-      "learning_rate": 0.00013319999999999999,
-      "loss": 3.0488,
+      "learning_rate": 0.0001368,
+      "loss": 2.6795,
       "step": 232
     },
     {
       "epoch": 2.14,
-      "learning_rate": 0.0001338,
-      "loss": 3.0229,
+      "learning_rate": 0.0001374,
+      "loss": 2.6582,
       "step": 233
     },
     {
       "epoch": 2.15,
-      "learning_rate": 0.0001344,
-      "loss": 3.0302,
+      "learning_rate": 0.000138,
+      "loss": 2.6236,
       "step": 234
     },
     {
       "epoch": 2.16,
-      "learning_rate": 0.000135,
-      "loss": 3.0351,
+      "learning_rate": 0.0001386,
+      "loss": 2.6159,
       "step": 235
     },
     {
       "epoch": 2.17,
-      "learning_rate": 0.0001356,
-      "loss": 3.02,
+      "learning_rate": 0.0001392,
+      "loss": 2.571,
       "step": 236
     },
     {
       "epoch": 2.17,
-      "learning_rate": 0.0001362,
-      "loss": 3.1479,
+      "learning_rate": 0.00013979999999999998,
+      "loss": 2.6935,
       "step": 237
     },
     {
       "epoch": 2.18,
-      "learning_rate": 0.0001368,
-      "loss": 3.0331,
+      "learning_rate": 0.0001404,
+      "loss": 2.5607,
       "step": 238
     },
     {
       "epoch": 2.19,
-      "learning_rate": 0.0001374,
-      "loss": 3.0238,
+      "learning_rate": 0.00014099999999999998,
+      "loss": 2.5766,
       "step": 239
     },
     {
       "epoch": 2.2,
-      "learning_rate": 0.000138,
-      "loss": 3.0261,
+      "learning_rate": 0.00014159999999999997,
+      "loss": 2.4837,
       "step": 240
     },
     {
       "epoch": 2.21,
-      "learning_rate": 0.0001386,
-      "loss": 3.0257,
+      "learning_rate": 0.0001422,
+      "loss": 2.5452,
       "step": 241
     },
     {
       "epoch": 2.22,
-      "learning_rate": 0.0001392,
-      "loss": 3.0236,
+      "learning_rate": 0.00014279999999999997,
+      "loss": 2.4572,
       "step": 242
     },
     {
       "epoch": 2.23,
-      "learning_rate": 0.00013979999999999998,
-      "loss": 3.0036,
+      "learning_rate": 0.0001434,
+      "loss": 2.4742,
       "step": 243
     },
     {
       "epoch": 2.24,
-      "learning_rate": 0.0001404,
-      "loss": 3.1014,
+      "learning_rate": 0.00014399999999999998,
+      "loss": 2.3591,
       "step": 244
     },
     {
       "epoch": 2.25,
-      "learning_rate": 0.00014099999999999998,
-      "loss": 3.0953,
+      "learning_rate": 0.0001446,
+      "loss": 2.3714,
       "step": 245
     },
     {
       "epoch": 2.26,
-      "learning_rate": 0.00014159999999999997,
-      "loss": 3.0452,
+      "learning_rate": 0.00014519999999999998,
+      "loss": 2.2839,
       "step": 246
     },
     {
       "epoch": 2.27,
-      "learning_rate": 0.0001422,
-      "loss": 4.1463,
+      "learning_rate": 0.0001458,
+      "loss": 2.2617,
       "step": 247
     },
     {
       "epoch": 2.28,
-      "learning_rate": 0.00014279999999999997,
-      "loss": 3.0316,
+      "learning_rate": 0.00014639999999999998,
+      "loss": 2.2579,
       "step": 248
     },
     {
       "epoch": 2.28,
-      "learning_rate": 0.0001434,
-      "loss": 3.0101,
+      "learning_rate": 0.000147,
+      "loss": 2.2222,
       "step": 249
     },
     {
       "epoch": 2.29,
-      "learning_rate": 0.00014399999999999998,
-      "loss": 3.0738,
+      "learning_rate": 0.00014759999999999998,
+      "loss": 2.1599,
       "step": 250
     },
     {
       "epoch": 2.3,
-      "learning_rate": 0.0001446,
-      "loss": 3.037,
+      "learning_rate": 0.0001482,
+      "loss": 2.1463,
       "step": 251
     },
     {
       "epoch": 2.31,
-      "learning_rate": 0.00014519999999999998,
-      "loss": 3.0305,
+      "learning_rate": 0.00014879999999999998,
+      "loss": 2.1426,
       "step": 252
     },
     {
       "epoch": 2.32,
-      "learning_rate": 0.0001458,
-      "loss": 3.0395,
+      "learning_rate": 0.0001494,
+      "loss": 2.1265,
       "step": 253
     },
     {
       "epoch": 2.33,
-      "learning_rate": 0.00014639999999999998,
-      "loss": 3.0146,
+      "learning_rate": 0.00015,
+      "loss": 2.0339,
       "step": 254
     },
     {
       "epoch": 2.34,
-      "learning_rate": 0.000147,
-      "loss": 3.0341,
+      "learning_rate": 0.00015059999999999997,
+      "loss": 1.9595,
       "step": 255
     },
     {
       "epoch": 2.35,
-      "learning_rate": 0.00014759999999999998,
-      "loss": 2.9889,
+      "learning_rate": 0.0001512,
+      "loss": 1.9296,
       "step": 256
     },
     {
       "epoch": 2.36,
-      "learning_rate": 0.0001482,
-      "loss": 3.0481,
+      "learning_rate": 0.00015179999999999998,
+      "loss": 1.9334,
       "step": 257
     },
     {
       "epoch": 2.37,
-      "learning_rate": 0.00014879999999999998,
-      "loss": 3.0284,
+      "learning_rate": 0.0001524,
+      "loss": 1.838,
       "step": 258
     },
     {
       "epoch": 2.38,
-      "learning_rate": 0.0001494,
-      "loss": 2.9996,
+      "learning_rate": 0.00015299999999999998,
+      "loss": 1.7801,
       "step": 259
     },
     {
       "epoch": 2.39,
-      "learning_rate": 0.00015,
-      "loss": 3.0224,
+      "learning_rate": 0.0001536,
+      "loss": 1.8893,
       "step": 260
     },
     {
       "epoch": 2.39,
-      "learning_rate": 0.00015059999999999997,
-      "loss": 3.0296,
+      "learning_rate": 0.00015419999999999998,
+      "loss": 1.7773,
       "step": 261
     },
     {
       "epoch": 2.4,
-      "learning_rate": 0.0001512,
-      "loss": 3.0457,
+      "learning_rate": 0.0001548,
+      "loss": 1.7869,
       "step": 262
     },
     {
       "epoch": 2.41,
-      "learning_rate": 0.00015179999999999998,
-      "loss": 3.0209,
+      "learning_rate": 0.00015539999999999998,
+      "loss": 1.6101,
       "step": 263
     },
     {
       "epoch": 2.42,
-      "learning_rate": 0.0001524,
-      "loss": 3.001,
+      "learning_rate": 0.000156,
+      "loss": 1.6763,
       "step": 264
     },
     {
       "epoch": 2.43,
-      "learning_rate": 0.00015299999999999998,
-      "loss": 3.0131,
+      "learning_rate": 0.00015659999999999998,
+      "loss": 1.6009,
       "step": 265
     },
     {
       "epoch": 2.44,
-      "learning_rate": 0.0001536,
-      "loss": 3.0192,
+      "learning_rate": 0.0001572,
+      "loss": 1.531,
       "step": 266
     },
     {
       "epoch": 2.45,
-      "learning_rate": 0.00015419999999999998,
-      "loss": 2.9596,
+      "learning_rate": 0.0001578,
+      "loss": 1.5178,
       "step": 267
     },
     {
       "epoch": 2.46,
-      "learning_rate": 0.0001548,
-      "loss": 3.0027,
+      "learning_rate": 0.0001584,
+      "loss": 1.7093,
       "step": 268
     },
     {
       "epoch": 2.47,
-      "learning_rate": 0.00015539999999999998,
-      "loss": 3.0379,
+      "learning_rate": 0.000159,
+      "loss": 1.4639,
       "step": 269
     },
     {
       "epoch": 2.48,
-      "learning_rate": 0.000156,
-      "loss": 2.9904,
+      "learning_rate": 0.0001596,
+      "loss": 1.3712,
       "step": 270
     },
     {
       "epoch": 2.49,
-      "learning_rate": 0.00015659999999999998,
-      "loss": 2.9954,
+      "learning_rate": 0.0001602,
+      "loss": 1.4015,
       "step": 271
     },
     {
       "epoch": 2.5,
-      "learning_rate": 0.0001572,
-      "loss": 3.0175,
+      "learning_rate": 0.0001608,
+      "loss": 1.3857,
       "step": 272
     },
     {
       "epoch": 2.5,
-      "learning_rate": 0.0001578,
-      "loss": 2.9823,
+      "learning_rate": 0.0001614,
+      "loss": 1.3684,
       "step": 273
     },
     {
       "epoch": 2.51,
-      "learning_rate": 0.0001584,
-      "loss": 2.976,
+      "learning_rate": 0.000162,
+      "loss": 1.3845,
       "step": 274
     },
     {
       "epoch": 2.52,
-      "learning_rate": 0.000159,
-      "loss": 3.0338,
+      "learning_rate": 0.0001626,
+      "loss": 1.36,
       "step": 275
     },
     {
       "epoch": 2.53,
-      "learning_rate": 0.0001596,
-      "loss": 2.9867,
+      "learning_rate": 0.0001632,
+      "loss": 1.2894,
       "step": 276
     },
     {
       "epoch": 2.54,
-      "learning_rate": 0.0001602,
-      "loss": 2.9976,
+      "learning_rate": 0.0001638,
+      "loss": 1.3132,
       "step": 277
     },
     {
       "epoch": 2.55,
-      "learning_rate": 0.0001608,
-      "loss": 3.0154,
+      "learning_rate": 0.0001644,
+      "loss": 1.1164,
       "step": 278
     },
     {
       "epoch": 2.56,
-      "learning_rate": 0.0001614,
-      "loss": 2.9624,
+      "learning_rate": 0.000165,
+      "loss": 1.2218,
       "step": 279
     },
     {
       "epoch": 2.57,
-      "learning_rate": 0.000162,
-      "loss": 3.0051,
+      "learning_rate": 0.0001656,
+      "loss": 1.2395,
       "step": 280
     },
     {
       "epoch": 2.58,
-      "learning_rate": 0.0001626,
-      "loss": 3.0145,
+      "learning_rate": 0.0001662,
+      "loss": 1.3201,
       "step": 281
     },
     {
       "epoch": 2.59,
-      "learning_rate": 0.0001632,
-      "loss": 3.011,
+      "learning_rate": 0.0001668,
+      "loss": 1.1685,
       "step": 282
     },
     {
       "epoch": 2.6,
-      "learning_rate": 0.0001638,
-      "loss": 2.9792,
+      "learning_rate": 0.0001674,
+      "loss": 1.1066,
       "step": 283
     },
     {
       "epoch": 2.61,
-      "learning_rate": 0.0001644,
-      "loss": 3.0356,
+      "learning_rate": 0.000168,
+      "loss": 1.1507,
       "step": 284
     },
     {
       "epoch": 2.61,
-      "learning_rate": 0.000165,
-      "loss": 3.0023,
+      "learning_rate": 0.0001686,
+      "loss": 1.0711,
       "step": 285
     },
     {
       "epoch": 2.62,
-      "learning_rate": 0.0001656,
-      "loss": 3.011,
+      "learning_rate": 0.00016919999999999997,
+      "loss": 1.1325,
       "step": 286
     },
     {
       "epoch": 2.63,
-      "learning_rate": 0.0001662,
-      "loss": 3.0044,
+      "learning_rate": 0.00016979999999999998,
+      "loss": 1.0844,
       "step": 287
     },
     {
       "epoch": 2.64,
-      "learning_rate": 0.0001668,
-      "loss": 3.023,
+      "learning_rate": 0.00017039999999999997,
+      "loss": 1.0547,
       "step": 288
     },
     {
       "epoch": 2.65,
-      "learning_rate": 0.0001674,
-      "loss": 2.9872,
+      "learning_rate": 0.00017099999999999998,
+      "loss": 1.082,
       "step": 289
     },
     {
       "epoch": 2.66,
-      "learning_rate": 0.000168,
-      "loss": 2.9911,
+      "learning_rate": 0.00017159999999999997,
+      "loss": 1.04,
       "step": 290
     },
     {
       "epoch": 2.67,
-      "learning_rate": 0.0001686,
-      "loss": 2.9894,
+      "learning_rate": 0.00017219999999999998,
+      "loss": 1.0513,
       "step": 291
     },
     {
       "epoch": 2.68,
-      "learning_rate": 0.00016919999999999997,
-      "loss": 3.0129,
+      "learning_rate": 0.00017279999999999997,
+      "loss": 1.1841,
       "step": 292
     },
     {
       "epoch": 2.69,
-      "learning_rate": 0.00016979999999999998,
-      "loss": 3.058,
+      "learning_rate": 0.00017339999999999996,
+      "loss": 1.0147,
       "step": 293
     },
     {
       "epoch": 2.7,
-      "learning_rate": 0.00017039999999999997,
-      "loss": 3.0095,
+      "learning_rate": 0.00017399999999999997,
+      "loss": 0.8994,
       "step": 294
     },
     {
       "epoch": 2.71,
-      "learning_rate": 0.00017099999999999998,
-      "loss": 3.0024,
+      "learning_rate": 0.00017459999999999996,
+      "loss": 0.9429,
       "step": 295
     },
     {
       "epoch": 2.72,
-      "learning_rate": 0.00017159999999999997,
-      "loss": 3.0121,
+      "learning_rate": 0.00017519999999999998,
+      "loss": 0.9682,
       "step": 296
     },
     {
       "epoch": 2.72,
-      "learning_rate": 0.00017219999999999998,
-      "loss": 2.98,
+      "learning_rate": 0.00017579999999999996,
+      "loss": 1.052,
       "step": 297
     },
     {
       "epoch": 2.73,
-      "learning_rate": 0.00017279999999999997,
-      "loss": 2.9751,
+      "learning_rate": 0.00017639999999999998,
+      "loss": 1.0992,
       "step": 298
     },
     {
       "epoch": 2.74,
-      "learning_rate": 0.00017339999999999996,
-      "loss": 2.9951,
+      "learning_rate": 0.00017699999999999997,
+      "loss": 1.1094,
       "step": 299
     },
     {
       "epoch": 2.75,
-      "learning_rate": 0.00017399999999999997,
-      "loss": 3.0291,
+      "learning_rate": 0.00017759999999999998,
+      "loss": 0.9653,
       "step": 300
     },
     {
       "epoch": 2.75,
-      "eval_loss": 2.9987223148345947,
-      "eval_runtime": 20.8174,
-      "eval_samples_per_second": 79.116,
-      "eval_steps_per_second": 1.249,
-      "eval_wer": 1.000408538453682,
+      "eval_loss": 0.9112017154693604,
+      "eval_runtime": 20.829,
+      "eval_samples_per_second": 79.072,
+      "eval_steps_per_second": 1.248,
+      "eval_wer": 0.7932795424369319,
       "step": 300
     },
     {
       "epoch": 2.76,
-      "learning_rate": 0.00017459999999999996,
-      "loss": 3.0167,
+      "learning_rate": 0.00017819999999999997,
+      "loss": 0.8969,
       "step": 301
     },
     {
       "epoch": 2.77,
-      "learning_rate": 0.00017519999999999998,
-      "loss": 3.0024,
+      "learning_rate": 0.00017879999999999998,
+      "loss": 1.0094,
       "step": 302
     },
     {
       "epoch": 2.78,
-      "learning_rate": 0.00017579999999999996,
-      "loss": 3.0008,
+      "learning_rate": 0.00017939999999999997,
+      "loss": 1.0534,
       "step": 303
     },
     {
       "epoch": 2.79,
-      "learning_rate": 0.00017639999999999998,
-      "loss": 2.9914,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.9988,
       "step": 304
     },
     {
       "epoch": 2.8,
-      "learning_rate": 0.00017699999999999997,
-      "loss": 3.0012,
+      "learning_rate": 0.00018059999999999997,
+      "loss": 0.9665,
       "step": 305
     },
     {
       "epoch": 2.81,
-      "learning_rate": 0.00017759999999999998,
-      "loss": 2.9832,
+      "learning_rate": 0.00018119999999999999,
+      "loss": 0.9126,
       "step": 306
     },
     {
       "epoch": 2.82,
-      "learning_rate": 0.00017819999999999997,
-      "loss": 2.9927,
+      "learning_rate": 0.00018179999999999997,
+      "loss": 0.8755,
       "step": 307
     },
     {
       "epoch": 2.83,
-      "learning_rate": 0.00017879999999999998,
-      "loss": 2.9741,
+      "learning_rate": 0.0001824,
+      "loss": 0.8298,
       "step": 308
     },
     {
       "epoch": 2.83,
-      "learning_rate": 0.00017939999999999997,
-      "loss": 2.9653,
+      "learning_rate": 0.00018299999999999998,
+      "loss": 0.916,
       "step": 309
     },
     {
       "epoch": 2.84,
-      "learning_rate": 0.00017999999999999998,
-      "loss": 2.9763,
+      "learning_rate": 0.0001836,
+      "loss": 0.9172,
       "step": 310
     },
     {
       "epoch": 2.85,
-      "learning_rate": 0.00018059999999999997,
-      "loss": 2.9964,
+      "learning_rate": 0.00018419999999999998,
+      "loss": 0.969,
       "step": 311
     },
     {
       "epoch": 2.86,
-      "learning_rate": 0.00018119999999999999,
-      "loss": 3.0337,
+      "learning_rate": 0.0001848,
+      "loss": 0.9171,
       "step": 312
     },
     {
       "epoch": 2.87,
-      "learning_rate": 0.00018179999999999997,
-      "loss": 3.0176,
+      "learning_rate": 0.00018539999999999998,
+      "loss": 0.8929,
       "step": 313
     },
     {
       "epoch": 2.88,
-      "learning_rate": 0.0001824,
-      "loss": 2.9892,
+      "learning_rate": 0.000186,
+      "loss": 0.7386,
       "step": 314
     },
     {
       "epoch": 2.89,
-      "learning_rate": 0.00018299999999999998,
-      "loss": 2.9574,
+      "learning_rate": 0.00018659999999999998,
+      "loss": 0.9016,
       "step": 315
     },
     {
       "epoch": 2.9,
-      "learning_rate": 0.0001836,
-      "loss": 2.9783,
+      "learning_rate": 0.0001872,
+      "loss": 0.9643,
       "step": 316
     },
     {
       "epoch": 2.91,
-      "learning_rate": 0.00018419999999999998,
-      "loss": 2.948,
+      "learning_rate": 0.00018779999999999998,
+      "loss": 0.9611,
       "step": 317
     },
     {
       "epoch": 2.92,
-      "learning_rate": 0.0001848,
-      "loss": 3.0035,
+      "learning_rate": 0.00018839999999999997,
+      "loss": 1.1287,
       "step": 318
     },
     {
       "epoch": 2.93,
-      "learning_rate": 0.00018539999999999998,
-      "loss": 3.0082,
+      "learning_rate": 0.00018899999999999999,
+      "loss": 0.8549,
       "step": 319
     },
     {
       "epoch": 2.94,
-      "learning_rate": 0.000186,
-      "loss": 2.9873,
+      "learning_rate": 0.00018959999999999997,
+      "loss": 0.7728,
       "step": 320
     },
     {
       "epoch": 2.94,
-      "learning_rate": 0.00018659999999999998,
-      "loss": 2.9693,
+      "learning_rate": 0.0001902,
+      "loss": 0.8282,
       "step": 321
     },
     {
       "epoch": 2.95,
-      "learning_rate": 0.0001872,
-      "loss": 2.9544,
+      "learning_rate": 0.00019079999999999998,
+      "loss": 0.8374,
       "step": 322
     },
     {
       "epoch": 2.96,
-      "learning_rate": 0.00018779999999999998,
-      "loss": 2.9226,
+      "learning_rate": 0.0001914,
+      "loss": 0.8338,
       "step": 323
     },
     {
       "epoch": 2.97,
-      "learning_rate": 0.00018839999999999997,
-      "loss": 2.9812,
+      "learning_rate": 0.00019199999999999998,
+      "loss": 0.8448,
       "step": 324
     },
     {
       "epoch": 2.98,
-      "learning_rate": 0.00018899999999999999,
-      "loss": 2.9695,
+      "learning_rate": 0.0001926,
+      "loss": 0.8177,
       "step": 325
     },
     {
       "epoch": 2.99,
-      "learning_rate": 0.00018959999999999997,
-      "loss": 2.9594,
+      "learning_rate": 0.00019319999999999998,
+      "loss": 0.8112,
       "step": 326
     },
     {
       "epoch": 3.0,
-      "learning_rate": 0.0001902,
-      "loss": 2.9308,
+      "learning_rate": 0.0001938,
+      "loss": 0.698,
       "step": 327
     },
     {
       "epoch": 3.01,
-      "learning_rate": 0.00019079999999999998,
-      "loss": 2.9654,
+      "learning_rate": 0.00019439999999999998,
+      "loss": 0.6733,
       "step": 328
     },
     {
       "epoch": 3.02,
-      "learning_rate": 0.0001914,
-      "loss": 2.9405,
+      "learning_rate": 0.000195,
+      "loss": 0.6288,
       "step": 329
     },
     {
       "epoch": 3.03,
-      "learning_rate": 0.00019199999999999998,
-      "loss": 2.9704,
+      "learning_rate": 0.00019559999999999998,
+      "loss": 0.6941,
       "step": 330
     },
     {
       "epoch": 3.04,
-      "learning_rate": 0.0001926,
-      "loss": 2.9496,
+      "learning_rate": 0.0001962,
+      "loss": 0.6773,
       "step": 331
     },
     {
       "epoch": 3.05,
-      "learning_rate": 0.00019319999999999998,
-      "loss": 2.9536,
+      "learning_rate": 0.00019679999999999999,
+      "loss": 0.7119,
       "step": 332
     },
     {
       "epoch": 3.06,
-      "learning_rate": 0.0001938,
-      "loss": 3.0699,
+      "learning_rate": 0.0001974,
+      "loss": 0.8595,
       "step": 333
     },
     {
       "epoch": 3.06,
-      "learning_rate": 0.00019439999999999998,
-      "loss": 3.0838,
+      "learning_rate": 0.000198,
+      "loss": 0.7988,
       "step": 334
     },
     {
       "epoch": 3.07,
-      "learning_rate": 0.000195,
-      "loss": 2.9375,
+      "learning_rate": 0.0001986,
+      "loss": 0.6729,
       "step": 335
     },
     {
       "epoch": 3.08,
-      "learning_rate": 0.00019559999999999998,
-      "loss": 2.9306,
+      "learning_rate": 0.0001992,
+      "loss": 0.6458,
       "step": 336
     },
     {
       "epoch": 3.09,
-      "learning_rate": 0.0001962,
-      "loss": 2.9351,
+      "learning_rate": 0.0001998,
+      "loss": 0.6262,
       "step": 337
     },
     {
       "epoch": 3.1,
-      "learning_rate": 0.00019679999999999999,
-      "loss": 2.9512,
+      "learning_rate": 0.0002004,
+      "loss": 0.6688,
       "step": 338
     },
     {
       "epoch": 3.11,
-      "learning_rate": 0.0001974,
-      "loss": 2.9305,
+      "learning_rate": 0.000201,
+      "loss": 0.7202,
       "step": 339
     },
     {
       "epoch": 3.12,
-      "learning_rate": 0.000198,
-      "loss": 2.9281,
+      "learning_rate": 0.0002016,
+      "loss": 0.6327,
       "step": 340
     },
     {
       "epoch": 3.13,
-      "learning_rate": 0.0001986,
-      "loss": 2.9215,
+      "learning_rate": 0.0002022,
+      "loss": 0.6829,
       "step": 341
     },
     {
       "epoch": 3.14,
-      "learning_rate": 0.0001992,
-      "loss": 2.9507,
+      "learning_rate": 0.0002028,
+      "loss": 0.6424,
       "step": 342
     },
     {
       "epoch": 3.15,
-      "learning_rate": 0.0001998,
-      "loss": 2.9108,
+      "learning_rate": 0.00020339999999999998,
+      "loss": 0.6687,
       "step": 343
     },
     {
       "epoch": 3.16,
-      "learning_rate": 0.0002004,
-      "loss": 2.9175,
+      "learning_rate": 0.000204,
+      "loss": 0.674,
       "step": 344
     },
     {
       "epoch": 3.17,
-      "learning_rate": 0.000201,
-      "loss": 2.9221,
+      "learning_rate": 0.00020459999999999999,
+      "loss": 0.6323,
       "step": 345
     },
     {
       "epoch": 3.17,
-      "learning_rate": 0.0002016,
-      "loss": 2.9165,
+      "learning_rate": 0.0002052,
+      "loss": 0.7028,
       "step": 346
     },
     {
       "epoch": 3.18,
-      "learning_rate": 0.0002022,
-      "loss": 2.9209,
+      "learning_rate": 0.0002058,
+      "loss": 0.5259,
       "step": 347
     },
     {
       "epoch": 3.19,
-      "learning_rate": 0.0002028,
-      "loss": 2.9363,
+      "learning_rate": 0.00020639999999999998,
+      "loss": 0.7341,
       "step": 348
     },
     {
       "epoch": 3.2,
-      "learning_rate": 0.00020339999999999998,
-      "loss": 2.9401,
+      "learning_rate": 0.00020699999999999996,
+      "loss": 0.6552,
       "step": 349
     },
     {
       "epoch": 3.21,
-      "learning_rate": 0.000204,
-      "loss": 2.9131,
+      "learning_rate": 0.00020759999999999998,
+      "loss": 0.7119,
       "step": 350
     },
     {
       "epoch": 3.22,
-      "learning_rate": 0.00020459999999999999,
-      "loss": 2.8862,
+      "learning_rate": 0.00020819999999999996,
+      "loss": 0.6971,
       "step": 351
     },
     {
       "epoch": 3.23,
-      "learning_rate": 0.0002052,
-      "loss": 2.8719,
+      "learning_rate": 0.00020879999999999998,
+      "loss": 0.7599,
       "step": 352
     },
     {
       "epoch": 3.24,
-      "learning_rate": 0.0002058,
-      "loss": 2.9431,
+      "learning_rate": 0.00020939999999999997,
+      "loss": 0.7553,
       "step": 353
     },
     {
       "epoch": 3.25,
-      "learning_rate": 0.00020639999999999998,
-      "loss": 2.9095,
+      "learning_rate": 0.00020999999999999998,
+      "loss": 0.6254,
       "step": 354
     },
     {
       "epoch": 3.26,
-      "learning_rate": 0.00020699999999999996,
-      "loss": 2.8315,
+      "learning_rate": 0.00021059999999999997,
+      "loss": 0.4797,
       "step": 355
     },
     {
       "epoch": 3.27,
-      "learning_rate": 0.00020759999999999998,
-      "loss": 2.9088,
+      "learning_rate": 0.00021119999999999996,
+      "loss": 0.5962,
       "step": 356
     },
     {
       "epoch": 3.28,
-      "learning_rate": 0.00020819999999999996,
-      "loss": 2.9362,
+      "learning_rate": 0.00021179999999999997,
+      "loss": 0.6244,
       "step": 357
     },
     {
       "epoch": 3.28,
-      "learning_rate": 0.00020879999999999998,
-      "loss": 2.8455,
+      "learning_rate": 0.00021239999999999996,
+      "loss": 0.5514,
       "step": 358
     },
     {
       "epoch": 3.29,
-      "learning_rate": 0.00020939999999999997,
-      "loss": 2.9655,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 0.5528,
       "step": 359
     },
     {
       "epoch": 3.3,
-      "learning_rate": 0.00020999999999999998,
-      "loss": 2.8797,
+      "learning_rate": 0.00021359999999999996,
+      "loss": 0.5929,
       "step": 360
     },
     {
       "epoch": 3.31,
-      "learning_rate": 0.00021059999999999997,
-      "loss": 2.904,
+      "learning_rate": 0.00021419999999999998,
+      "loss": 0.5428,
       "step": 361
     },
     {
       "epoch": 3.32,
-      "learning_rate": 0.00021119999999999996,
-      "loss": 2.9274,
+      "learning_rate": 0.00021479999999999996,
+      "loss": 0.6426,
       "step": 362
     },
     {
       "epoch": 3.33,
-      "learning_rate": 0.00021179999999999997,
-      "loss": 2.822,
+      "learning_rate": 0.00021539999999999998,
+      "loss": 0.6181,
       "step": 363
     },
     {
       "epoch": 3.34,
-      "learning_rate": 0.00021239999999999996,
-      "loss": 2.8912,
+      "learning_rate": 0.00021599999999999996,
+      "loss": 0.7736,
       "step": 364
     },
     {
       "epoch": 3.35,
-      "learning_rate": 0.00021299999999999997,
-      "loss": 2.9043,
+      "learning_rate": 0.00021659999999999998,
+      "loss": 0.6959,
       "step": 365
     },
     {
       "epoch": 3.36,
-      "learning_rate": 0.00021359999999999996,
-      "loss": 2.8948,
+      "learning_rate": 0.00021719999999999997,
+      "loss": 0.5946,
       "step": 366
     },
     {
       "epoch": 3.37,
-      "learning_rate": 0.00021419999999999998,
-      "loss": 2.8658,
+      "learning_rate": 0.00021779999999999998,
+      "loss": 0.5644,
       "step": 367
     },
     {
       "epoch": 3.38,
-      "learning_rate": 0.00021479999999999996,
-      "loss": 2.8782,
+      "learning_rate": 0.00021839999999999997,
+      "loss": 0.565,
       "step": 368
     },
     {
       "epoch": 3.39,
-      "learning_rate": 0.00021539999999999998,
-      "loss": 2.844,
+      "learning_rate": 0.00021899999999999998,
+      "loss": 0.6339,
       "step": 369
     },
     {
       "epoch": 3.39,
-      "learning_rate": 0.00021599999999999996,
-      "loss": 2.8464,
+      "learning_rate": 0.00021959999999999997,
+      "loss": 0.6975,
       "step": 370
     },
     {
       "epoch": 3.4,
-      "learning_rate": 0.00021659999999999998,
-      "loss": 2.9093,
+      "learning_rate": 0.00022019999999999999,
+      "loss": 0.6402,
       "step": 371
     },
     {
       "epoch": 3.41,
-      "learning_rate": 0.00021719999999999997,
-      "loss": 2.8937,
+      "learning_rate": 0.00022079999999999997,
+      "loss": 0.5821,
       "step": 372
     },
     {
       "epoch": 3.42,
-      "learning_rate": 0.00021779999999999998,
-      "loss": 2.8772,
+      "learning_rate": 0.0002214,
+      "loss": 0.565,
       "step": 373
     },
     {
       "epoch": 3.43,
-      "learning_rate": 0.00021839999999999997,
-      "loss": 2.8225,
+      "learning_rate": 0.00022199999999999998,
+      "loss": 0.5159,
       "step": 374
     },
     {
       "epoch": 3.44,
-      "learning_rate": 0.00021899999999999998,
-      "loss": 2.8238,
+      "learning_rate": 0.0002226,
+      "loss": 0.5854,
       "step": 375
     },
     {
       "epoch": 3.45,
-      "learning_rate": 0.00021959999999999997,
-      "loss": 2.8012,
+      "learning_rate": 0.00022319999999999998,
+      "loss": 0.5983,
       "step": 376
     },
     {
       "epoch": 3.46,
-      "learning_rate": 0.00022019999999999999,
-      "loss": 2.7986,
+      "learning_rate": 0.0002238,
+      "loss": 0.6607,
       "step": 377
     },
     {
       "epoch": 3.47,
-      "learning_rate": 0.00022079999999999997,
-      "loss": 2.9084,
+      "learning_rate": 0.00022439999999999998,
+      "loss": 0.5893,
       "step": 378
     },
     {
       "epoch": 3.48,
-      "learning_rate": 0.0002214,
-      "loss": 2.8282,
+      "learning_rate": 0.000225,
+      "loss": 0.5458,
       "step": 379
     },
     {
       "epoch": 3.49,
-      "learning_rate": 0.00022199999999999998,
-      "loss": 2.8076,
+      "learning_rate": 0.00022559999999999998,
+      "loss": 0.5717,
       "step": 380
     },
     {
       "epoch": 3.5,
-      "learning_rate": 0.0002226,
-      "loss": 2.7754,
+      "learning_rate": 0.00022619999999999997,
+      "loss": 0.6553,
       "step": 381
     },
     {
       "epoch": 3.5,
-      "learning_rate": 0.00022319999999999998,
-      "loss": 2.7515,
+      "learning_rate": 0.00022679999999999998,
+      "loss": 0.6054,
       "step": 382
     },
     {
       "epoch": 3.51,
-      "learning_rate": 0.0002238,
-      "loss": 2.7302,
+      "learning_rate": 0.00022739999999999997,
+      "loss": 0.6109,
       "step": 383
     },
     {
       "epoch": 3.52,
-      "learning_rate": 0.00022439999999999998,
-      "loss": 2.781,
+      "learning_rate": 0.00022799999999999999,
+      "loss": 0.6221,
       "step": 384
     },
     {
       "epoch": 3.53,
-      "learning_rate": 0.000225,
-      "loss": 2.7644,
+      "learning_rate": 0.00022859999999999997,
+      "loss": 0.5516,
       "step": 385
     },
     {
       "epoch": 3.54,
-      "learning_rate": 0.00022559999999999998,
-      "loss": 2.7296,
+      "learning_rate": 0.0002292,
+      "loss": 0.5633,
       "step": 386
     },
     {
       "epoch": 3.55,
-      "learning_rate": 0.00022619999999999997,
-      "loss": 2.7572,
+      "learning_rate": 0.00022979999999999997,
+      "loss": 0.6322,
       "step": 387
     },
     {
       "epoch": 3.56,
-      "learning_rate": 0.00022679999999999998,
-      "loss": 2.6636,
+      "learning_rate": 0.0002304,
+      "loss": 0.5794,
       "step": 388
     },
     {
       "epoch": 3.57,
-      "learning_rate": 0.00022739999999999997,
-      "loss": 2.7627,
+      "learning_rate": 0.00023099999999999998,
+      "loss": 0.7055,
       "step": 389
     },
     {
       "epoch": 3.58,
-      "learning_rate": 0.00022799999999999999,
-      "loss": 2.7373,
+      "learning_rate": 0.0002316,
+      "loss": 0.6361,
       "step": 390
     },
     {
       "epoch": 3.59,
-      "learning_rate": 0.00022859999999999997,
-      "loss": 2.7129,
+      "learning_rate": 0.00023219999999999998,
+      "loss": 0.5399,
       "step": 391
     },
     {
       "epoch": 3.6,
-      "learning_rate": 0.0002292,
-      "loss": 2.7321,
+      "learning_rate": 0.0002328,
+      "loss": 0.5314,
       "step": 392
     },
     {
       "epoch": 3.61,
-      "learning_rate": 0.00022979999999999997,
-      "loss": 2.7394,
+      "learning_rate": 0.00023339999999999998,
+      "loss": 0.6136,
       "step": 393
     },
     {
       "epoch": 3.61,
-      "learning_rate": 0.0002304,
-      "loss": 2.6314,
+      "learning_rate": 0.000234,
+      "loss": 0.684,
       "step": 394
     },
     {
       "epoch": 3.62,
-      "learning_rate": 0.00023099999999999998,
-      "loss": 2.5942,
+      "learning_rate": 0.00023459999999999998,
+      "loss": 0.6164,
       "step": 395
     },
     {
       "epoch": 3.63,
-      "learning_rate": 0.0002316,
-      "loss": 2.7123,
+      "learning_rate": 0.0002352,
+      "loss": 0.6316,
       "step": 396
     },
     {
       "epoch": 3.64,
-      "learning_rate": 0.00023219999999999998,
-      "loss": 2.6219,
+      "learning_rate": 0.00023579999999999999,
+      "loss": 0.5668,
       "step": 397
     },
     {
       "epoch": 3.65,
-      "learning_rate": 0.0002328,
-      "loss": 2.6275,
+      "learning_rate": 0.0002364,
+      "loss": 0.4738,
       "step": 398
     },
     {
       "epoch": 3.66,
-      "learning_rate": 0.00023339999999999998,
-      "loss": 2.5843,
+      "learning_rate": 0.000237,
+      "loss": 0.6115,
       "step": 399
     },
     {
       "epoch": 3.67,
-      "learning_rate": 0.000234,
-      "loss": 2.6102,
+      "learning_rate": 0.0002376,
+      "loss": 0.6105,
       "step": 400
     },
     {
       "epoch": 3.67,
-      "eval_loss": 2.2342309951782227,
-      "eval_runtime": 20.8876,
-      "eval_samples_per_second": 78.851,
-      "eval_steps_per_second": 1.245,
-      "eval_wer": 1.0003064038402614,
+      "eval_loss": 0.5851821303367615,
+      "eval_runtime": 21.0335,
+      "eval_samples_per_second": 78.304,
+      "eval_steps_per_second": 1.236,
+      "eval_wer": 0.6263915841078541,
       "step": 400
     },
     {
       "epoch": 3.68,
-      "learning_rate": 0.00023459999999999998,
-      "loss": 2.5973,
+      "learning_rate": 0.0002382,
+      "loss": 0.5114,
       "step": 401
     },
     {
       "epoch": 3.69,
-      "learning_rate": 0.0002352,
-      "loss": 2.5858,
+      "learning_rate": 0.0002388,
+      "loss": 0.6018,
       "step": 402
     },
     {
       "epoch": 3.7,
-      "learning_rate": 0.00023579999999999999,
-      "loss": 2.623,
+      "learning_rate": 0.0002394,
+      "loss": 0.5999,
       "step": 403
     },
     {
       "epoch": 3.71,
-      "learning_rate": 0.0002364,
-      "loss": 2.5821,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 0.5624,
       "step": 404
     },
     {
       "epoch": 3.72,
-      "learning_rate": 0.000237,
-      "loss": 2.6257,
+      "learning_rate": 0.0002406,
+      "loss": 0.4868,
       "step": 405
     },
     {
       "epoch": 3.72,
-      "learning_rate": 0.0002376,
-      "loss": 2.6132,
+      "learning_rate": 0.00024119999999999998,
+      "loss": 0.6016,
       "step": 406
     },
     {
       "epoch": 3.73,
-      "learning_rate": 0.0002382,
-      "loss": 2.5814,
+      "learning_rate": 0.0002418,
+      "loss": 0.6861,
       "step": 407
     },
     {
       "epoch": 3.74,
-      "learning_rate": 0.0002388,
-      "loss": 2.6228,
+      "learning_rate": 0.00024239999999999998,
+      "loss": 0.7149,
       "step": 408
     },
     {
       "epoch": 3.75,
-      "learning_rate": 0.0002394,
-      "loss": 2.5769,
+      "learning_rate": 0.000243,
+      "loss": 0.4899,
       "step": 409
     },
     {
       "epoch": 3.76,
-      "learning_rate": 0.00023999999999999998,
-      "loss": 2.5847,
+      "learning_rate": 0.00024359999999999999,
+      "loss": 0.5449,
       "step": 410
     },
     {
       "epoch": 3.77,
-      "learning_rate": 0.0002406,
-      "loss": 2.543,
+      "learning_rate": 0.00024419999999999997,
+      "loss": 0.4544,
       "step": 411
     },
     {
       "epoch": 3.78,
-      "learning_rate": 0.00024119999999999998,
-      "loss": 2.5247,
+      "learning_rate": 0.0002448,
+      "loss": 0.5238,
       "step": 412
     },
     {
       "epoch": 3.79,
-      "learning_rate": 0.0002418,
-      "loss": 2.5062,
+      "learning_rate": 0.00024539999999999995,
+      "loss": 0.6421,
       "step": 413
     },
     {
       "epoch": 3.8,
-      "learning_rate": 0.00024239999999999998,
-      "loss": 2.4728,
+      "learning_rate": 0.00024599999999999996,
+      "loss": 0.5785,
       "step": 414
     },
     {
       "epoch": 3.81,
-      "learning_rate": 0.000243,
-      "loss": 2.6206,
+      "learning_rate": 0.0002466,
+      "loss": 0.5638,
       "step": 415
     },
     {
       "epoch": 3.82,
-      "learning_rate": 0.00024359999999999999,
-      "loss": 2.537,
+      "learning_rate": 0.0002472,
+      "loss": 0.4528,
       "step": 416
     },
     {
       "epoch": 3.83,
-      "learning_rate": 0.00024419999999999997,
-      "loss": 2.5072,
+      "learning_rate": 0.00024779999999999995,
+      "loss": 0.5244,
       "step": 417
     },
     {
       "epoch": 3.83,
-      "learning_rate": 0.0002448,
-      "loss": 2.4686,
+      "learning_rate": 0.00024839999999999997,
+      "loss": 0.5412,
       "step": 418
     },
     {
       "epoch": 3.84,
-      "learning_rate": 0.00024539999999999995,
-      "loss": 2.4275,
+      "learning_rate": 0.000249,
+      "loss": 0.5474,
       "step": 419
     },
     {
       "epoch": 3.85,
-      "learning_rate": 0.00024599999999999996,
-      "loss": 2.4345,
+      "learning_rate": 0.00024959999999999994,
+      "loss": 0.5811,
       "step": 420
     },
     {
       "epoch": 3.86,
-      "learning_rate": 0.0002466,
-      "loss": 2.4363,
+      "learning_rate": 0.00025019999999999996,
+      "loss": 0.5832,
       "step": 421
     },
     {
       "epoch": 3.87,
-      "learning_rate": 0.0002472,
-      "loss": 2.6807,
+      "learning_rate": 0.00025079999999999997,
+      "loss": 0.4739,
       "step": 422
     },
     {
       "epoch": 3.88,
-      "learning_rate": 0.00024779999999999995,
-      "loss": 2.4657,
+      "learning_rate": 0.0002514,
+      "loss": 0.562,
       "step": 423
     },
     {
       "epoch": 3.89,
-      "learning_rate": 0.00024839999999999997,
-      "loss": 2.4013,
+      "learning_rate": 0.00025199999999999995,
+      "loss": 0.4931,
       "step": 424
     },
     {
       "epoch": 3.9,
-      "learning_rate": 0.000249,
-      "loss": 2.4115,
+      "learning_rate": 0.00025259999999999996,
+      "loss": 0.5044,
       "step": 425
     },
     {
       "epoch": 3.91,
-      "learning_rate": 0.00024959999999999994,
-      "loss": 2.4117,
+      "learning_rate": 0.0002532,
+      "loss": 0.5746,
       "step": 426
     },
     {
       "epoch": 3.92,
-      "learning_rate": 0.00025019999999999996,
-      "loss": 2.5098,
+      "learning_rate": 0.0002538,
+      "loss": 0.4839,
       "step": 427
     },
     {
       "epoch": 3.93,
-      "learning_rate": 0.00025079999999999997,
-      "loss": 2.4498,
+      "learning_rate": 0.00025439999999999995,
+      "loss": 0.4251,
       "step": 428
     },
     {
       "epoch": 3.94,
-      "learning_rate": 0.0002514,
-      "loss": 2.3888,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 0.538,
       "step": 429
     },
     {
       "epoch": 3.94,
-      "learning_rate": 0.00025199999999999995,
-      "loss": 2.429,
+      "learning_rate": 0.0002556,
+      "loss": 0.4613,
       "step": 430
     },
     {
       "epoch": 3.95,
-      "learning_rate": 0.00025259999999999996,
-      "loss": 2.3786,
+      "learning_rate": 0.0002562,
+      "loss": 0.4259,
       "step": 431
     },
     {
       "epoch": 3.96,
-      "learning_rate": 0.0002532,
-      "loss": 2.4057,
+      "learning_rate": 0.00025679999999999995,
+      "loss": 0.6892,
       "step": 432
     },
     {
       "epoch": 3.97,
-      "learning_rate": 0.0002538,
-      "loss": 2.4641,
+      "learning_rate": 0.00025739999999999997,
+      "loss": 0.5854,
       "step": 433
     },
     {
       "epoch": 3.98,
-      "learning_rate": 0.00025439999999999995,
-      "loss": 2.6428,
+      "learning_rate": 0.000258,
+      "loss": 0.46,
       "step": 434
     },
     {
       "epoch": 3.99,
-      "learning_rate": 0.00025499999999999996,
-      "loss": 2.6125,
+      "learning_rate": 0.0002586,
+      "loss": 0.4968,
       "step": 435
     },
     {
       "epoch": 4.0,
-      "learning_rate": 0.0002556,
-      "loss": 2.7156,
+      "learning_rate": 0.00025919999999999996,
+      "loss": 0.5547,
       "step": 436
     },
     {
       "epoch": 4.01,
-      "learning_rate": 0.0002562,
-      "loss": 2.4681,
+      "learning_rate": 0.00025979999999999997,
+      "loss": 0.4119,
       "step": 437
     },
     {
       "epoch": 4.02,
-      "learning_rate": 0.00025679999999999995,
-      "loss": 2.4994,
+      "learning_rate": 0.0002604,
+      "loss": 0.3845,
       "step": 438
     },
     {
       "epoch": 4.03,
-      "learning_rate": 0.00025739999999999997,
-      "loss": 2.4267,
+      "learning_rate": 0.000261,
+      "loss": 0.3509,
       "step": 439
     },
     {
       "epoch": 4.04,
-      "learning_rate": 0.000258,
-      "loss": 2.462,
+      "learning_rate": 0.00026159999999999996,
+      "loss": 0.4111,
       "step": 440
     },
     {
       "epoch": 4.05,
-      "learning_rate": 0.0002586,
-      "loss": 2.4613,
+      "learning_rate": 0.0002622,
+      "loss": 0.4089,
       "step": 441
     },
     {
       "epoch": 4.06,
-      "learning_rate": 0.00025919999999999996,
-      "loss": 2.4945,
+      "learning_rate": 0.0002628,
+      "loss": 0.4633,
       "step": 442
     },
     {
       "epoch": 4.06,
-      "learning_rate": 0.00025979999999999997,
-      "loss": 2.4448,
+      "learning_rate": 0.00026339999999999995,
+      "loss": 0.4023,
       "step": 443
     },
     {
       "epoch": 4.07,
-      "learning_rate": 0.0002604,
-      "loss": 2.4539,
+      "learning_rate": 0.00026399999999999997,
+      "loss": 0.4379,
       "step": 444
     },
     {
       "epoch": 4.08,
-      "learning_rate": 0.000261,
-      "loss": 2.3582,
+      "learning_rate": 0.0002646,
+      "loss": 0.3484,
       "step": 445
     },
     {
       "epoch": 4.09,
-      "learning_rate": 0.00026159999999999996,
-      "loss": 2.4232,
+      "learning_rate": 0.0002652,
+      "loss": 0.4886,
       "step": 446
     },
     {
       "epoch": 4.1,
-      "learning_rate": 0.0002622,
-      "loss": 2.3736,
+      "learning_rate": 0.00026579999999999996,
+      "loss": 0.4894,
       "step": 447
     },
     {
       "epoch": 4.11,
-      "learning_rate": 0.0002628,
-      "loss": 2.3014,
+      "learning_rate": 0.00026639999999999997,
+      "loss": 0.5098,
       "step": 448
     },
     {
       "epoch": 4.12,
-      "learning_rate": 0.00026339999999999995,
-      "loss": 2.4751,
+      "learning_rate": 0.000267,
+      "loss": 0.4275,
       "step": 449
     },
     {
       "epoch": 4.13,
-      "learning_rate": 0.00026399999999999997,
-      "loss": 2.3984,
+      "learning_rate": 0.0002676,
+      "loss": 0.4012,
       "step": 450
     },
     {
       "epoch": 4.14,
-      "learning_rate": 0.0002646,
-      "loss": 2.3105,
+      "learning_rate": 0.00026819999999999996,
+      "loss": 0.3456,
       "step": 451
     },
     {
       "epoch": 4.15,
-      "learning_rate": 0.0002652,
-      "loss": 2.2458,
+      "learning_rate": 0.0002688,
+      "loss": 0.3676,
       "step": 452
     },
     {
       "epoch": 4.16,
-      "learning_rate": 0.00026579999999999996,
-      "loss": 2.3303,
+      "learning_rate": 0.0002694,
+      "loss": 0.3925,
       "step": 453
     },
     {
       "epoch": 4.17,
-      "learning_rate": 0.00026639999999999997,
-      "loss": 2.262,
+      "learning_rate": 0.00027,
+      "loss": 0.49,
       "step": 454
     },
     {
       "epoch": 4.17,
-      "learning_rate": 0.000267,
-      "loss": 2.2769,
+      "learning_rate": 0.00027059999999999996,
+      "loss": 0.4666,
       "step": 455
     },
     {
       "epoch": 4.18,
-      "learning_rate": 0.0002676,
-      "loss": 2.3735,
+      "learning_rate": 0.0002712,
+      "loss": 0.3403,
       "step": 456
     },
     {
       "epoch": 4.19,
-      "learning_rate": 0.00026819999999999996,
-      "loss": 2.3502,
+      "learning_rate": 0.0002718,
+      "loss": 0.359,
       "step": 457
     },
     {
       "epoch": 4.2,
-      "learning_rate": 0.0002688,
-      "loss": 2.2039,
+      "learning_rate": 0.0002724,
+      "loss": 0.3683,
       "step": 458
     },
     {
       "epoch": 4.21,
-      "learning_rate": 0.0002694,
-      "loss": 2.2015,
+      "learning_rate": 0.00027299999999999997,
+      "loss": 0.3627,
       "step": 459
     },
     {
       "epoch": 4.22,
-      "learning_rate": 0.00027,
-      "loss": 2.2316,
+      "learning_rate": 0.0002736,
+      "loss": 0.3815,
       "step": 460
     },
     {
       "epoch": 4.23,
-      "learning_rate": 0.00027059999999999996,
-      "loss": 2.2545,
+      "learning_rate": 0.0002742,
+      "loss": 0.54,
       "step": 461
     },
     {
       "epoch": 4.24,
-      "learning_rate": 0.0002712,
-      "loss": 2.3344,
+      "learning_rate": 0.0002748,
+      "loss": 0.3761,
       "step": 462
     },
     {
       "epoch": 4.25,
-      "learning_rate": 0.0002718,
-      "loss": 2.2358,
+      "learning_rate": 0.00027539999999999997,
+      "loss": 0.3497,
       "step": 463
     },
     {
       "epoch": 4.26,
-      "learning_rate": 0.0002724,
-      "loss": 2.2736,
+      "learning_rate": 0.000276,
+      "loss": 0.3253,
       "step": 464
     },
     {
       "epoch": 4.27,
-      "learning_rate": 0.00027299999999999997,
-      "loss": 2.2787,
+      "learning_rate": 0.0002766,
+      "loss": 0.5023,
       "step": 465
     },
     {
       "epoch": 4.28,
-      "learning_rate": 0.0002736,
-      "loss": 2.261,
+      "learning_rate": 0.0002772,
+      "loss": 0.4134,
       "step": 466
     },
     {
       "epoch": 4.28,
-      "learning_rate": 0.0002742,
-      "loss": 2.1453,
+      "learning_rate": 0.0002778,
+      "loss": 0.4535,
       "step": 467
     },
     {
       "epoch": 4.29,
-      "learning_rate": 0.0002748,
-      "loss": 2.2782,
+      "learning_rate": 0.0002784,
+      "loss": 0.4159,
       "step": 468
     },
     {
       "epoch": 4.3,
-      "learning_rate": 0.00027539999999999997,
-      "loss": 2.2013,
+      "learning_rate": 0.000279,
+      "loss": 0.4024,
       "step": 469
     },
     {
       "epoch": 4.31,
-      "learning_rate": 0.000276,
-      "loss": 2.1134,
+      "learning_rate": 0.00027959999999999997,
+      "loss": 0.3076,
       "step": 470
     },
     {
       "epoch": 4.32,
-      "learning_rate": 0.0002766,
-      "loss": 2.2219,
+      "learning_rate": 0.0002802,
+      "loss": 0.4876,
       "step": 471
     },
     {
       "epoch": 4.33,
-      "learning_rate": 0.0002772,
-      "loss": 2.1734,
+      "learning_rate": 0.0002808,
+      "loss": 0.4431,
       "step": 472
     },
     {
       "epoch": 4.34,
-      "learning_rate": 0.0002778,
-      "loss": 2.1974,
+      "learning_rate": 0.00028139999999999996,
+      "loss": 0.4771,
       "step": 473
     },
     {
       "epoch": 4.35,
-      "learning_rate": 0.0002784,
-      "loss": 2.1241,
+      "learning_rate": 0.00028199999999999997,
+      "loss": 0.4459,
       "step": 474
     },
     {
       "epoch": 4.36,
-      "learning_rate": 0.000279,
-      "loss": 2.2403,
+      "learning_rate": 0.0002826,
+      "loss": 0.4285,
       "step": 475
     },
     {
       "epoch": 4.37,
-      "learning_rate": 0.00027959999999999997,
-      "loss": 2.1909,
+      "learning_rate": 0.00028319999999999994,
+      "loss": 0.3663,
       "step": 476
     },
     {
       "epoch": 4.38,
-      "learning_rate": 0.0002802,
-      "loss": 2.1242,
+      "learning_rate": 0.00028379999999999996,
+      "loss": 0.3761,
       "step": 477
     },
     {
       "epoch": 4.39,
-      "learning_rate": 0.0002808,
-      "loss": 2.0393,
+      "learning_rate": 0.0002844,
+      "loss": 0.349,
       "step": 478
     },
     {
       "epoch": 4.39,
-      "learning_rate": 0.00028139999999999996,
-      "loss": 2.068,
+      "learning_rate": 0.000285,
+      "loss": 0.336,
       "step": 479
     },
     {
       "epoch": 4.4,
-      "learning_rate": 0.00028199999999999997,
-      "loss": 2.1905,
+      "learning_rate": 0.00028559999999999995,
+      "loss": 0.3992,
       "step": 480
     },
     {
       "epoch": 4.41,
-      "learning_rate": 0.0002826,
-      "loss": 2.1065,
+      "learning_rate": 0.00028619999999999996,
+      "loss": 0.3028,
       "step": 481
     },
     {
       "epoch": 4.42,
-      "learning_rate": 0.00028319999999999994,
-      "loss": 2.2154,
+      "learning_rate": 0.0002868,
+      "loss": 0.3384,
       "step": 482
     },
     {
       "epoch": 4.43,
-      "learning_rate": 0.00028379999999999996,
-      "loss": 2.6584,
+      "learning_rate": 0.00028739999999999994,
+      "loss": 0.3779,
       "step": 483
     },
     {
       "epoch": 4.44,
-      "learning_rate": 0.0002844,
-      "loss": 2.4039,
+      "learning_rate": 0.00028799999999999995,
+      "loss": 0.3832,
       "step": 484
     },
     {
       "epoch": 4.45,
-      "learning_rate": 0.000285,
-      "loss": 2.0366,
+      "learning_rate": 0.00028859999999999997,
+      "loss": 0.3877,
       "step": 485
     },
     {
       "epoch": 4.46,
-      "learning_rate": 0.00028559999999999995,
-      "loss": 2.2255,
+      "learning_rate": 0.0002892,
+      "loss": 0.4919,
       "step": 486
     },
     {
       "epoch": 4.47,
-      "learning_rate": 0.00028619999999999996,
-      "loss": 2.2209,
+      "learning_rate": 0.00028979999999999994,
+      "loss": 0.4293,
       "step": 487
     },
     {
       "epoch": 4.48,
-      "learning_rate": 0.0002868,
-      "loss": 2.0695,
+      "learning_rate": 0.00029039999999999996,
+      "loss": 0.3919,
       "step": 488
     },
     {
       "epoch": 4.49,
-      "learning_rate": 0.00028739999999999994,
-      "loss": 2.0602,
+      "learning_rate": 0.00029099999999999997,
+      "loss": 0.4876,
       "step": 489
     },
     {
       "epoch": 4.5,
-      "learning_rate": 0.00028799999999999995,
-      "loss": 2.0164,
+      "learning_rate": 0.0002916,
+      "loss": 0.3465,
       "step": 490
     },
     {
       "epoch": 4.5,
-      "learning_rate": 0.00028859999999999997,
-      "loss": 2.072,
+      "learning_rate": 0.00029219999999999995,
+      "loss": 0.5007,
       "step": 491
     },
     {
       "epoch": 4.51,
-      "learning_rate": 0.0002892,
-      "loss": 2.0403,
+      "learning_rate": 0.00029279999999999996,
+      "loss": 0.4381,
       "step": 492
     },
     {
       "epoch": 4.52,
-      "learning_rate": 0.00028979999999999994,
-      "loss": 2.0728,
+      "learning_rate": 0.0002934,
+      "loss": 0.402,
       "step": 493
     },
     {
       "epoch": 4.53,
-      "learning_rate": 0.00029039999999999996,
-      "loss": 1.9714,
+      "learning_rate": 0.000294,
+      "loss": 0.2915,
       "step": 494
     },
     {
       "epoch": 4.54,
-      "learning_rate": 0.00029099999999999997,
-      "loss": 1.9885,
+      "learning_rate": 0.00029459999999999995,
+      "loss": 0.2441,
       "step": 495
     },
     {
       "epoch": 4.55,
-      "learning_rate": 0.0002916,
-      "loss": 2.0967,
+      "learning_rate": 0.00029519999999999997,
+      "loss": 0.4288,
       "step": 496
     },
     {
       "epoch": 4.56,
-      "learning_rate": 0.00029219999999999995,
-      "loss": 2.1548,
+      "learning_rate": 0.0002958,
+      "loss": 0.4117,
       "step": 497
     },
     {
       "epoch": 4.57,
-      "learning_rate": 0.00029279999999999996,
-      "loss": 2.1185,
+      "learning_rate": 0.0002964,
+      "loss": 0.4262,
       "step": 498
     },
     {
       "epoch": 4.58,
-      "learning_rate": 0.0002934,
-      "loss": 2.1688,
+      "learning_rate": 0.00029699999999999996,
+      "loss": 0.3915,
       "step": 499
     },
     {
       "epoch": 4.59,
-      "learning_rate": 0.000294,
-      "loss": 2.1225,
+      "learning_rate": 0.00029759999999999997,
+      "loss": 0.4073,
       "step": 500
     },
     {
       "epoch": 4.59,
-      "eval_loss": 1.2467074394226074,
-      "eval_runtime": 20.7969,
-      "eval_samples_per_second": 79.194,
-      "eval_steps_per_second": 1.25,
-      "eval_wer": 0.9294249821264426,
+      "eval_loss": 0.5203901529312134,
+      "eval_runtime": 21.0005,
+      "eval_samples_per_second": 78.427,
+      "eval_steps_per_second": 1.238,
+      "eval_wer": 0.5455009702788275,
       "step": 500
     },
     {
       "epoch": 4.6,
-      "learning_rate": 0.00029459999999999995,
-      "loss": 2.0434,
+      "learning_rate": 0.0002982,
+      "loss": 0.4065,
       "step": 501
     },
     {
       "epoch": 4.61,
-      "learning_rate": 0.00029519999999999997,
-      "loss": 2.0797,
+      "learning_rate": 0.0002988,
+      "loss": 0.3701,
       "step": 502
     },
     {
       "epoch": 4.61,
-      "learning_rate": 0.0002958,
-      "loss": 2.1153,
+      "learning_rate": 0.00029939999999999996,
+      "loss": 0.4367,
       "step": 503
     },
     {
       "epoch": 4.62,
-      "learning_rate": 0.0002964,
-      "loss": 2.0404,
+      "learning_rate": 0.0003,
+      "loss": 0.4291,
       "step": 504
     },
     {
       "epoch": 4.63,
-      "learning_rate": 0.00029699999999999996,
-      "loss": 2.1289,
+      "learning_rate": 0.00029973568281938326,
+      "loss": 0.4421,
       "step": 505
     },
     {
       "epoch": 4.64,
-      "learning_rate": 0.00029759999999999997,
-      "loss": 1.9893,
+      "learning_rate": 0.0002994713656387665,
+      "loss": 0.3577,
       "step": 506
     },
     {
       "epoch": 4.65,
-      "learning_rate": 0.0002982,
-      "loss": 2.0455,
+      "learning_rate": 0.0002992070484581498,
+      "loss": 0.4039,
       "step": 507
     },
     {
       "epoch": 4.66,
-      "learning_rate": 0.0002988,
-      "loss": 2.0126,
+      "learning_rate": 0.000298942731277533,
+      "loss": 0.4106,
       "step": 508
     },
     {
       "epoch": 4.67,
-      "learning_rate": 0.00029939999999999996,
-      "loss": 2.0631,
+      "learning_rate": 0.00029867841409691624,
+      "loss": 0.4215,
       "step": 509
     },
     {
       "epoch": 4.68,
-      "learning_rate": 0.0003,
-      "loss": 1.9799,
+      "learning_rate": 0.0002984140969162995,
+      "loss": 0.4038,
       "step": 510
     },
     {
       "epoch": 4.69,
-      "learning_rate": 0.00029973568281938326,
-      "loss": 2.0514,
+      "learning_rate": 0.0002981497797356828,
+      "loss": 0.4627,
       "step": 511
     },
     {
       "epoch": 4.7,
-      "learning_rate": 0.0002994713656387665,
-      "loss": 1.9771,
+      "learning_rate": 0.00029788546255506604,
+      "loss": 0.3432,
       "step": 512
     },
     {
       "epoch": 4.71,
-      "learning_rate": 0.0002992070484581498,
-      "loss": 1.9922,
+      "learning_rate": 0.0002976211453744493,
+      "loss": 0.3631,
       "step": 513
     },
     {
       "epoch": 4.72,
-      "learning_rate": 0.000298942731277533,
-      "loss": 1.9445,
+      "learning_rate": 0.0002973568281938326,
+      "loss": 0.3139,
       "step": 514
     },
     {
       "epoch": 4.72,
-      "learning_rate": 0.00029867841409691624,
-      "loss": 2.0019,
+      "learning_rate": 0.00029709251101321584,
+      "loss": 0.4026,
       "step": 515
     },
     {
       "epoch": 4.73,
-      "learning_rate": 0.0002984140969162995,
-      "loss": 1.9189,
+      "learning_rate": 0.00029682819383259907,
+      "loss": 0.4275,
       "step": 516
     },
     {
       "epoch": 4.74,
-      "learning_rate": 0.0002981497797356828,
-      "loss": 1.976,
+      "learning_rate": 0.00029656387665198236,
+      "loss": 0.4779,
       "step": 517
     },
     {
       "epoch": 4.75,
-      "learning_rate": 0.00029788546255506604,
-      "loss": 1.9246,
+      "learning_rate": 0.0002962995594713656,
+      "loss": 0.3444,
       "step": 518
     },
     {
       "epoch": 4.76,
-      "learning_rate": 0.0002976211453744493,
-      "loss": 1.9739,
+      "learning_rate": 0.0002960352422907489,
+      "loss": 0.3387,
       "step": 519
     },
     {
       "epoch": 4.77,
-      "learning_rate": 0.0002973568281938326,
-      "loss": 1.9536,
+      "learning_rate": 0.00029577092511013216,
+      "loss": 0.2666,
       "step": 520
     },
     {
       "epoch": 4.78,
-      "learning_rate": 0.00029709251101321584,
-      "loss": 2.0236,
+      "learning_rate": 0.0002955066079295154,
+      "loss": 0.3659,
       "step": 521
     },
     {
       "epoch": 4.79,
-      "learning_rate": 0.00029682819383259907,
-      "loss": 1.9868,
+      "learning_rate": 0.0002952422907488987,
+      "loss": 0.4028,
       "step": 522
     },
     {
       "epoch": 4.8,
-      "learning_rate": 0.00029656387665198236,
-      "loss": 1.9315,
+      "learning_rate": 0.0002949779735682819,
+      "loss": 0.4018,
       "step": 523
     },
     {
       "epoch": 4.81,
-      "learning_rate": 0.0002962995594713656,
-      "loss": 2.1194,
+      "learning_rate": 0.0002947136563876652,
+      "loss": 0.3763,
       "step": 524
     },
     {
       "epoch": 4.82,
-      "learning_rate": 0.0002960352422907489,
-      "loss": 1.9697,
+      "learning_rate": 0.0002944493392070484,
+      "loss": 0.3136,
       "step": 525
     },
     {
       "epoch": 4.83,
-      "learning_rate": 0.00029577092511013216,
-      "loss": 1.9386,
+      "learning_rate": 0.0002941850220264317,
+      "loss": 0.3309,
       "step": 526
     },
     {
       "epoch": 4.83,
-      "learning_rate": 0.0002955066079295154,
-      "loss": 1.937,
+      "learning_rate": 0.00029392070484581494,
+      "loss": 0.4139,
       "step": 527
     },
     {
       "epoch": 4.84,
-      "learning_rate": 0.0002952422907488987,
-      "loss": 1.9104,
+      "learning_rate": 0.0002936563876651982,
+      "loss": 0.3973,
       "step": 528
     },
     {
       "epoch": 4.85,
-      "learning_rate": 0.0002949779735682819,
-      "loss": 1.8949,
+      "learning_rate": 0.0002933920704845815,
+      "loss": 0.3739,
       "step": 529
     },
     {
       "epoch": 4.86,
-      "learning_rate": 0.0002947136563876652,
-      "loss": 2.0053,
+      "learning_rate": 0.00029312775330396474,
+      "loss": 0.4663,
       "step": 530
     },
     {
       "epoch": 4.87,
-      "learning_rate": 0.0002944493392070484,
-      "loss": 2.0451,
+      "learning_rate": 0.00029286343612334797,
+      "loss": 0.475,
       "step": 531
     },
     {
       "epoch": 4.88,
-      "learning_rate": 0.0002941850220264317,
-      "loss": 1.9699,
+      "learning_rate": 0.00029259911894273126,
+      "loss": 0.3783,
       "step": 532
     },
     {
       "epoch": 4.89,
-      "learning_rate": 0.00029392070484581494,
-      "loss": 2.0467,
+      "learning_rate": 0.0002923348017621145,
+      "loss": 0.4308,
       "step": 533
     },
     {
       "epoch": 4.9,
-      "learning_rate": 0.0002936563876651982,
-      "loss": 2.0722,
+      "learning_rate": 0.00029207048458149777,
+      "loss": 0.4316,
       "step": 534
     },
     {
       "epoch": 4.91,
-      "learning_rate": 0.0002933920704845815,
-      "loss": 2.0914,
+      "learning_rate": 0.00029180616740088106,
+      "loss": 0.51,
       "step": 535
     },
     {
       "epoch": 4.92,
-      "learning_rate": 0.00029312775330396474,
-      "loss": 2.066,
+      "learning_rate": 0.0002915418502202643,
+      "loss": 0.4409,
       "step": 536
     },
     {
       "epoch": 4.93,
-      "learning_rate": 0.00029286343612334797,
-      "loss": 2.0059,
+      "learning_rate": 0.0002912775330396475,
+      "loss": 0.3423,
       "step": 537
     },
     {
       "epoch": 4.94,
-      "learning_rate": 0.00029259911894273126,
-      "loss": 1.9301,
+      "learning_rate": 0.0002910132158590308,
+      "loss": 0.3287,
       "step": 538
     },
     {
       "epoch": 4.94,
-      "learning_rate": 0.0002923348017621145,
-      "loss": 1.9485,
+      "learning_rate": 0.0002907488986784141,
+      "loss": 0.3699,
       "step": 539
     },
     {
       "epoch": 4.95,
-      "learning_rate": 0.00029207048458149777,
-      "loss": 1.9682,
+      "learning_rate": 0.0002904845814977973,
+      "loss": 0.3322,
       "step": 540
     },
     {
       "epoch": 4.96,
-      "learning_rate": 0.00029180616740088106,
-      "loss": 1.9335,
+      "learning_rate": 0.0002902202643171806,
+      "loss": 0.3896,
       "step": 541
     },
     {
       "epoch": 4.97,
-      "learning_rate": 0.0002915418502202643,
-      "loss": 1.9739,
+      "learning_rate": 0.00028995594713656384,
+      "loss": 0.4546,
       "step": 542
     },
     {
       "epoch": 4.98,
-      "learning_rate": 0.0002912775330396475,
-      "loss": 2.0254,
+      "learning_rate": 0.0002896916299559471,
+      "loss": 0.3377,
       "step": 543
     },
     {
       "epoch": 4.99,
-      "learning_rate": 0.0002910132158590308,
-      "loss": 2.0062,
+      "learning_rate": 0.00028942731277533035,
+      "loss": 0.3478,
       "step": 544
     },
     {
       "epoch": 5.0,
-      "learning_rate": 0.0002907488986784141,
-      "loss": 2.3722,
+      "learning_rate": 0.00028916299559471364,
+      "loss": 0.3015,
       "step": 545
     },
     {
       "epoch": 5.01,
-      "learning_rate": 0.0002904845814977973,
-      "loss": 2.074,
+      "learning_rate": 0.00028889867841409687,
+      "loss": 0.2193,
       "step": 546
     },
     {
       "epoch": 5.02,
-      "learning_rate": 0.0002902202643171806,
-      "loss": 2.0599,
+      "learning_rate": 0.00028863436123348016,
+      "loss": 0.3622,
       "step": 547
     },
     {
       "epoch": 5.03,
-      "learning_rate": 0.00028995594713656384,
-      "loss": 1.8885,
+      "learning_rate": 0.00028837004405286344,
+      "loss": 0.2089,
       "step": 548
     },
     {
       "epoch": 5.04,
-      "learning_rate": 0.0002896916299559471,
-      "loss": 1.9717,
+      "learning_rate": 0.00028810572687224667,
+      "loss": 0.4089,
       "step": 549
     },
     {
       "epoch": 5.05,
-      "learning_rate": 0.00028942731277533035,
-      "loss": 1.9851,
+      "learning_rate": 0.00028784140969162996,
+      "loss": 0.368,
       "step": 550
     },
     {
       "epoch": 5.06,
-      "learning_rate": 0.00028916299559471364,
-      "loss": 2.0565,
+      "learning_rate": 0.0002875770925110132,
+      "loss": 0.4156,
       "step": 551
     },
     {
       "epoch": 5.06,
-      "learning_rate": 0.00028889867841409687,
-      "loss": 2.0263,
+      "learning_rate": 0.0002873127753303964,
+      "loss": 0.2352,
       "step": 552
     },
     {
       "epoch": 5.07,
-      "learning_rate": 0.00028863436123348016,
-      "loss": 2.0672,
+      "learning_rate": 0.0002870484581497797,
+      "loss": 0.2344,
       "step": 553
     },
     {
       "epoch": 5.08,
-      "learning_rate": 0.00028837004405286344,
-      "loss": 2.0258,
+      "learning_rate": 0.000286784140969163,
+      "loss": 0.3489,
       "step": 554
     },
     {
       "epoch": 5.09,
-      "learning_rate": 0.00028810572687224667,
-      "loss": 2.031,
+      "learning_rate": 0.0002865198237885462,
+      "loss": 0.3032,
       "step": 555
     },
     {
       "epoch": 5.1,
-      "learning_rate": 0.00028784140969162996,
-      "loss": 2.0279,
+      "learning_rate": 0.0002862555066079295,
+      "loss": 0.2847,
       "step": 556
     },
     {
       "epoch": 5.11,
-      "learning_rate": 0.0002875770925110132,
-      "loss": 1.9515,
+      "learning_rate": 0.00028599118942731274,
+      "loss": 0.3175,
       "step": 557
     },
     {
       "epoch": 5.12,
-      "learning_rate": 0.0002873127753303964,
-      "loss": 2.0229,
+      "learning_rate": 0.000285726872246696,
+      "loss": 0.3816,
       "step": 558
     },
     {
       "epoch": 5.13,
-      "learning_rate": 0.0002870484581497797,
-      "loss": 1.9519,
+      "learning_rate": 0.00028546255506607925,
+      "loss": 0.2771,
       "step": 559
     },
     {
       "epoch": 5.14,
-      "learning_rate": 0.000286784140969163,
-      "loss": 1.8902,
+      "learning_rate": 0.00028519823788546254,
+      "loss": 0.2409,
       "step": 560
     },
     {
       "epoch": 5.15,
-      "learning_rate": 0.0002865198237885462,
-      "loss": 1.863,
+      "learning_rate": 0.00028493392070484577,
+      "loss": 0.2385,
       "step": 561
     },
     {
       "epoch": 5.16,
-      "learning_rate": 0.0002862555066079295,
-      "loss": 1.8731,
+      "learning_rate": 0.00028466960352422905,
+      "loss": 0.2434,
       "step": 562
     },
     {
       "epoch": 5.17,
-      "learning_rate": 0.00028599118942731274,
-      "loss": 1.8674,
+      "learning_rate": 0.00028440528634361234,
+      "loss": 0.2327,
       "step": 563
     },
     {
       "epoch": 5.17,
-      "learning_rate": 0.000285726872246696,
-      "loss": 1.8837,
+      "learning_rate": 0.00028414096916299557,
+      "loss": 0.3664,
       "step": 564
     },
     {
       "epoch": 5.18,
-      "learning_rate": 0.00028546255506607925,
-      "loss": 1.8195,
+      "learning_rate": 0.00028387665198237886,
+      "loss": 0.2255,
       "step": 565
     },
     {
       "epoch": 5.19,
-      "learning_rate": 0.00028519823788546254,
-      "loss": 1.841,
+      "learning_rate": 0.0002836123348017621,
+      "loss": 0.2442,
       "step": 566
     },
     {
       "epoch": 5.2,
-      "learning_rate": 0.00028493392070484577,
-      "loss": 1.9209,
+      "learning_rate": 0.0002833480176211453,
+      "loss": 0.2561,
       "step": 567
     },
     {
       "epoch": 5.21,
-      "learning_rate": 0.00028466960352422905,
-      "loss": 1.9295,
+      "learning_rate": 0.0002830837004405286,
+      "loss": 0.3358,
       "step": 568
     },
     {
       "epoch": 5.22,
-      "learning_rate": 0.00028440528634361234,
-      "loss": 1.929,
+      "learning_rate": 0.0002828193832599119,
+      "loss": 0.2582,
       "step": 569
     },
     {
       "epoch": 5.23,
-      "learning_rate": 0.00028414096916299557,
-      "loss": 1.952,
+      "learning_rate": 0.0002825550660792951,
+      "loss": 0.3909,
       "step": 570
     },
     {
       "epoch": 5.24,
-      "learning_rate": 0.00028387665198237886,
-      "loss": 1.9014,
+      "learning_rate": 0.0002822907488986784,
+      "loss": 0.2659,
       "step": 571
     },
     {
       "epoch": 5.25,
-      "learning_rate": 0.0002836123348017621,
-      "loss": 1.8508,
+      "learning_rate": 0.00028202643171806164,
+      "loss": 0.3133,
       "step": 572
     },
     {
       "epoch": 5.26,
-      "learning_rate": 0.0002833480176211453,
-      "loss": 1.8427,
+      "learning_rate": 0.00028176211453744487,
+      "loss": 0.3203,
       "step": 573
     },
     {
       "epoch": 5.27,
-      "learning_rate": 0.0002830837004405286,
-      "loss": 1.8917,
+      "learning_rate": 0.00028149779735682815,
+      "loss": 0.344,
       "step": 574
     },
     {
       "epoch": 5.28,
-      "learning_rate": 0.0002828193832599119,
-      "loss": 1.8682,
+      "learning_rate": 0.00028123348017621144,
+      "loss": 0.3378,
       "step": 575
     },
     {
       "epoch": 5.28,
-      "learning_rate": 0.0002825550660792951,
-      "loss": 1.9799,
+      "learning_rate": 0.00028096916299559467,
+      "loss": 0.3073,
       "step": 576
     },
     {
       "epoch": 5.29,
-      "learning_rate": 0.0002822907488986784,
-      "loss": 2.1559,
+      "learning_rate": 0.00028070484581497795,
+      "loss": 0.3369,
       "step": 577
     },
     {
       "epoch": 5.3,
-      "learning_rate": 0.00028202643171806164,
-      "loss": 1.913,
+      "learning_rate": 0.00028044052863436124,
+      "loss": 0.2831,
       "step": 578
     },
     {
       "epoch": 5.31,
-      "learning_rate": 0.00028176211453744487,
-      "loss": 1.9681,
+      "learning_rate": 0.00028017621145374447,
+      "loss": 0.2645,
       "step": 579
     },
     {
       "epoch": 5.32,
-      "learning_rate": 0.00028149779735682815,
-      "loss": 1.9536,
+      "learning_rate": 0.0002799118942731277,
+      "loss": 0.242,
       "step": 580
     },
     {
       "epoch": 5.33,
-      "learning_rate": 0.00028123348017621144,
-      "loss": 1.954,
+      "learning_rate": 0.000279647577092511,
+      "loss": 0.2721,
       "step": 581
     },
     {
       "epoch": 5.34,
-      "learning_rate": 0.00028096916299559467,
-      "loss": 1.9717,
+      "learning_rate": 0.0002793832599118942,
+      "loss": 0.4008,
       "step": 582
     },
     {
       "epoch": 5.35,
-      "learning_rate": 0.00028070484581497795,
-      "loss": 1.9943,
+      "learning_rate": 0.0002791189427312775,
+      "loss": 0.2655,
       "step": 583
     },
     {
       "epoch": 5.36,
-      "learning_rate": 0.00028044052863436124,
-      "loss": 2.0324,
+      "learning_rate": 0.0002788546255506608,
+      "loss": 0.2511,
       "step": 584
     },
     {
       "epoch": 5.37,
-      "learning_rate": 0.00028017621145374447,
-      "loss": 1.9779,
+      "learning_rate": 0.000278590308370044,
+      "loss": 0.2404,
       "step": 585
     },
     {
       "epoch": 5.38,
-      "learning_rate": 0.0002799118942731277,
-      "loss": 1.9241,
+      "learning_rate": 0.0002783259911894273,
+      "loss": 0.3033,
       "step": 586
     },
     {
       "epoch": 5.39,
-      "learning_rate": 0.000279647577092511,
-      "loss": 1.889,
+      "learning_rate": 0.0002780616740088106,
+      "loss": 0.3031,
       "step": 587
     },
     {
       "epoch": 5.39,
-      "learning_rate": 0.0002793832599118942,
-      "loss": 1.7891,
+      "learning_rate": 0.0002777973568281938,
+      "loss": 0.3271,
       "step": 588
     },
     {
       "epoch": 5.4,
-      "learning_rate": 0.0002791189427312775,
-      "loss": 1.9931,
+      "learning_rate": 0.00027753303964757705,
+      "loss": 0.4368,
       "step": 589
     },
     {
       "epoch": 5.41,
-      "learning_rate": 0.0002788546255506608,
-      "loss": 1.9316,
+      "learning_rate": 0.00027726872246696034,
+      "loss": 0.2294,
       "step": 590
     },
     {
       "epoch": 5.42,
-      "learning_rate": 0.000278590308370044,
-      "loss": 1.9971,
+      "learning_rate": 0.00027700440528634357,
+      "loss": 0.2604,
       "step": 591
     },
     {
       "epoch": 5.43,
-      "learning_rate": 0.0002783259911894273,
-      "loss": 1.968,
+      "learning_rate": 0.00027674008810572685,
+      "loss": 0.2362,
       "step": 592
     },
     {
       "epoch": 5.44,
-      "learning_rate": 0.0002780616740088106,
-      "loss": 1.9206,
+      "learning_rate": 0.00027647577092511014,
+      "loss": 0.2319,
       "step": 593
     },
     {
       "epoch": 5.45,
-      "learning_rate": 0.0002777973568281938,
-      "loss": 1.8991,
+      "learning_rate": 0.00027621145374449337,
+      "loss": 0.3253,
       "step": 594
     },
     {
       "epoch": 5.46,
-      "learning_rate": 0.00027753303964757705,
-      "loss": 1.9315,
+      "learning_rate": 0.0002759471365638766,
+      "loss": 0.352,
       "step": 595
     },
     {
       "epoch": 5.47,
-      "learning_rate": 0.00027726872246696034,
-      "loss": 2.0693,
+      "learning_rate": 0.0002756828193832599,
+      "loss": 0.3131,
       "step": 596
     },
     {
       "epoch": 5.48,
-      "learning_rate": 0.00027700440528634357,
-      "loss": 1.9013,
+      "learning_rate": 0.0002754185022026431,
+      "loss": 0.2439,
       "step": 597
     },
     {
       "epoch": 5.49,
-      "learning_rate": 0.00027674008810572685,
-      "loss": 1.9018,
+      "learning_rate": 0.0002751541850220264,
+      "loss": 0.2829,
       "step": 598
     },
     {
       "epoch": 5.5,
-      "learning_rate": 0.00027647577092511014,
-      "loss": 1.8696,
+      "learning_rate": 0.0002748898678414097,
+      "loss": 0.3586,
       "step": 599
     },
     {
       "epoch": 5.5,
-      "learning_rate": 0.00027621145374449337,
-      "loss": 1.8639,
+      "learning_rate": 0.0002746255506607929,
+      "loss": 0.3091,
       "step": 600
     },
     {
       "epoch": 5.5,
-      "eval_loss": 0.8659506440162659,
-      "eval_runtime": 21.0181,
-      "eval_samples_per_second": 78.361,
-      "eval_steps_per_second": 1.237,
-      "eval_wer": 0.7440506587682566,
+      "eval_loss": 0.4557536542415619,
+      "eval_runtime": 20.8951,
+      "eval_samples_per_second": 78.822,
+      "eval_steps_per_second": 1.244,
+      "eval_wer": 0.4821775099581248,
       "step": 600
     },
     {
       "epoch": 5.51,
-      "learning_rate": 0.0002759471365638766,
-      "loss": 1.8104,
+      "learning_rate": 0.0002743612334801762,
+      "loss": 0.4359,
       "step": 601
     },
     {
       "epoch": 5.52,
-      "learning_rate": 0.0002756828193832599,
-      "loss": 1.9019,
+      "learning_rate": 0.00027409691629955943,
+      "loss": 0.299,
       "step": 602
     },
     {
       "epoch": 5.53,
-      "learning_rate": 0.0002754185022026431,
-      "loss": 1.8551,
+      "learning_rate": 0.0002738325991189427,
+      "loss": 0.2056,
       "step": 603
     },
     {
       "epoch": 5.54,
-      "learning_rate": 0.0002751541850220264,
-      "loss": 1.8091,
+      "learning_rate": 0.00027356828193832595,
+      "loss": 0.2644,
       "step": 604
     },
     {
       "epoch": 5.55,
-      "learning_rate": 0.0002748898678414097,
-      "loss": 1.8517,
+      "learning_rate": 0.00027330396475770924,
+      "loss": 0.2268,
       "step": 605
     },
     {
       "epoch": 5.56,
-      "learning_rate": 0.0002746255506607929,
-      "loss": 1.8395,
+      "learning_rate": 0.00027303964757709247,
+      "loss": 0.2407,
       "step": 606
     },
     {
       "epoch": 5.57,
-      "learning_rate": 0.0002743612334801762,
-      "loss": 1.8981,
+      "learning_rate": 0.00027277533039647575,
+      "loss": 0.2776,
       "step": 607
     },
     {
       "epoch": 5.58,
-      "learning_rate": 0.00027409691629955943,
-      "loss": 1.9402,
+      "learning_rate": 0.00027251101321585904,
+      "loss": 0.3812,
       "step": 608
     },
     {
       "epoch": 5.59,
-      "learning_rate": 0.0002738325991189427,
-      "loss": 1.9369,
+      "learning_rate": 0.00027224669603524227,
+      "loss": 0.2837,
       "step": 609
     },
     {
       "epoch": 5.6,
-      "learning_rate": 0.00027356828193832595,
-      "loss": 1.9148,
+      "learning_rate": 0.0002719823788546255,
+      "loss": 0.2637,
       "step": 610
     },
     {
       "epoch": 5.61,
-      "learning_rate": 0.00027330396475770924,
-      "loss": 1.8603,
+      "learning_rate": 0.0002717180616740088,
+      "loss": 0.2823,
       "step": 611
     },
     {
       "epoch": 5.61,
-      "learning_rate": 0.00027303964757709247,
-      "loss": 1.8184,
+      "learning_rate": 0.00027145374449339207,
+      "loss": 0.2608,
       "step": 612
     },
     {
       "epoch": 5.62,
-      "learning_rate": 0.00027277533039647575,
-      "loss": 1.7831,
+      "learning_rate": 0.0002711894273127753,
+      "loss": 0.277,
       "step": 613
     },
     {
       "epoch": 5.63,
-      "learning_rate": 0.00027251101321585904,
-      "loss": 1.7893,
+      "learning_rate": 0.0002709251101321586,
+      "loss": 0.282,
       "step": 614
     },
     {
       "epoch": 5.64,
-      "learning_rate": 0.00027224669603524227,
-      "loss": 1.9005,
+      "learning_rate": 0.0002706607929515418,
+      "loss": 0.2229,
       "step": 615
     },
     {
       "epoch": 5.65,
-      "learning_rate": 0.0002719823788546255,
-      "loss": 1.8097,
+      "learning_rate": 0.0002703964757709251,
+      "loss": 0.2113,
       "step": 616
     },
     {
       "epoch": 5.66,
-      "learning_rate": 0.0002717180616740088,
-      "loss": 1.845,
+      "learning_rate": 0.00027013215859030833,
+      "loss": 0.2333,
       "step": 617
     },
     {
       "epoch": 5.67,
-      "learning_rate": 0.00027145374449339207,
-      "loss": 1.7442,
+      "learning_rate": 0.0002698678414096916,
+      "loss": 0.2448,
       "step": 618
     },
     {
       "epoch": 5.68,
-      "learning_rate": 0.0002711894273127753,
-      "loss": 1.7461,
+      "learning_rate": 0.00026960352422907485,
+      "loss": 0.3268,
       "step": 619
     },
     {
       "epoch": 5.69,
-      "learning_rate": 0.0002709251101321586,
-      "loss": 1.8321,
+      "learning_rate": 0.00026933920704845814,
+      "loss": 0.3127,
       "step": 620
     },
     {
       "epoch": 5.7,
-      "learning_rate": 0.0002706607929515418,
-      "loss": 1.8922,
+      "learning_rate": 0.00026907488986784137,
+      "loss": 0.2197,
       "step": 621
     },
     {
       "epoch": 5.71,
-      "learning_rate": 0.0002703964757709251,
-      "loss": 1.786,
+      "learning_rate": 0.00026881057268722465,
+      "loss": 0.2127,
       "step": 622
     },
     {
       "epoch": 5.72,
-      "learning_rate": 0.00027013215859030833,
-      "loss": 1.7509,
+      "learning_rate": 0.00026854625550660794,
+      "loss": 0.2957,
       "step": 623
     },
     {
       "epoch": 5.72,
-      "learning_rate": 0.0002698678414096916,
-      "loss": 1.8589,
+      "learning_rate": 0.00026828193832599117,
+      "loss": 0.2431,
       "step": 624
     },
     {
       "epoch": 5.73,
-      "learning_rate": 0.00026960352422907485,
-      "loss": 1.7976,
+      "learning_rate": 0.0002680176211453744,
+      "loss": 0.2184,
       "step": 625
     },
     {
       "epoch": 5.74,
-      "learning_rate": 0.00026933920704845814,
-      "loss": 1.8467,
+      "learning_rate": 0.0002677533039647577,
+      "loss": 0.3249,
       "step": 626
     },
     {
       "epoch": 5.75,
-      "learning_rate": 0.00026907488986784137,
-      "loss": 1.806,
+      "learning_rate": 0.00026748898678414097,
+      "loss": 0.2112,
       "step": 627
     },
     {
       "epoch": 5.76,
-      "learning_rate": 0.00026881057268722465,
-      "loss": 1.8451,
+      "learning_rate": 0.0002672246696035242,
+      "loss": 0.2661,
       "step": 628
     },
     {
       "epoch": 5.77,
-      "learning_rate": 0.00026854625550660794,
-      "loss": 1.7834,
+      "learning_rate": 0.0002669603524229075,
+      "loss": 0.2293,
       "step": 629
     },
     {
       "epoch": 5.78,
-      "learning_rate": 0.00026828193832599117,
-      "loss": 1.8057,
+      "learning_rate": 0.0002666960352422907,
+      "loss": 0.267,
       "step": 630
     },
     {
       "epoch": 5.79,
-      "learning_rate": 0.0002680176211453744,
-      "loss": 1.7974,
+      "learning_rate": 0.00026643171806167395,
+      "loss": 0.3231,
       "step": 631
     },
     {
       "epoch": 5.8,
-      "learning_rate": 0.0002677533039647577,
-      "loss": 1.8205,
+      "learning_rate": 0.00026616740088105723,
+      "loss": 0.2593,
       "step": 632
     },
     {
       "epoch": 5.81,
-      "learning_rate": 0.00026748898678414097,
-      "loss": 1.9248,
+      "learning_rate": 0.0002659030837004405,
+      "loss": 0.3079,
       "step": 633
     },
     {
       "epoch": 5.82,
-      "learning_rate": 0.0002672246696035242,
-      "loss": 1.8704,
+      "learning_rate": 0.00026563876651982375,
+      "loss": 0.2348,
       "step": 634
     },
     {
       "epoch": 5.83,
-      "learning_rate": 0.0002669603524229075,
-      "loss": 1.7954,
+      "learning_rate": 0.00026537444933920703,
+      "loss": 0.2311,
       "step": 635
     },
     {
       "epoch": 5.83,
-      "learning_rate": 0.0002666960352422907,
-      "loss": 1.901,
+      "learning_rate": 0.0002651101321585903,
+      "loss": 0.3229,
       "step": 636
     },
     {
       "epoch": 5.84,
-      "learning_rate": 0.00026643171806167395,
-      "loss": 1.817,
+      "learning_rate": 0.00026484581497797355,
+      "loss": 0.3854,
       "step": 637
     },
     {
       "epoch": 5.85,
-      "learning_rate": 0.00026616740088105723,
-      "loss": 1.9251,
+      "learning_rate": 0.0002645814977973568,
+      "loss": 0.3085,
       "step": 638
     },
     {
       "epoch": 5.86,
-      "learning_rate": 0.0002659030837004405,
-      "loss": 1.8708,
+      "learning_rate": 0.00026431718061674007,
+      "loss": 0.2557,
       "step": 639
     },
     {
       "epoch": 5.87,
-      "learning_rate": 0.00026563876651982375,
-      "loss": 1.8829,
+      "learning_rate": 0.0002640528634361233,
+      "loss": 0.2796,
       "step": 640
     },
     {
       "epoch": 5.88,
-      "learning_rate": 0.00026537444933920703,
-      "loss": 1.807,
+      "learning_rate": 0.0002637885462555066,
+      "loss": 0.2291,
       "step": 641
     },
     {
       "epoch": 5.89,
-      "learning_rate": 0.0002651101321585903,
-      "loss": 1.8007,
+      "learning_rate": 0.00026352422907488987,
+      "loss": 0.2752,
       "step": 642
     },
     {
       "epoch": 5.9,
-      "learning_rate": 0.00026484581497797355,
-      "loss": 1.7205,
+      "learning_rate": 0.0002632599118942731,
+      "loss": 0.2125,
       "step": 643
     },
     {
       "epoch": 5.91,
-      "learning_rate": 0.0002645814977973568,
-      "loss": 1.8491,
+      "learning_rate": 0.0002629955947136564,
+      "loss": 0.3172,
       "step": 644
     },
     {
       "epoch": 5.92,
-      "learning_rate": 0.00026431718061674007,
-      "loss": 1.8746,
+      "learning_rate": 0.0002627312775330396,
+      "loss": 0.3995,
       "step": 645
     },
     {
       "epoch": 5.93,
-      "learning_rate": 0.0002640528634361233,
-      "loss": 1.9822,
+      "learning_rate": 0.00026246696035242285,
+      "loss": 0.2922,
       "step": 646
     },
     {
       "epoch": 5.94,
-      "learning_rate": 0.0002637885462555066,
-      "loss": 1.8358,
+      "learning_rate": 0.00026220264317180613,
+      "loss": 0.2256,
       "step": 647
     },
     {
       "epoch": 5.94,
-      "learning_rate": 0.00026352422907488987,
-      "loss": 1.8416,
+      "learning_rate": 0.0002619383259911894,
+      "loss": 0.2659,
       "step": 648
     },
     {
       "epoch": 5.95,
-      "learning_rate": 0.0002632599118942731,
-      "loss": 1.8368,
+      "learning_rate": 0.00026167400881057265,
+      "loss": 0.2338,
       "step": 649
     },
     {
       "epoch": 5.96,
-      "learning_rate": 0.0002629955947136564,
-      "loss": 1.8595,
+      "learning_rate": 0.00026140969162995593,
+      "loss": 0.3681,
       "step": 650
     },
     {
       "epoch": 5.97,
-      "learning_rate": 0.0002627312775330396,
-      "loss": 1.8277,
+      "learning_rate": 0.0002611453744493392,
+      "loss": 0.2911,
       "step": 651
     },
     {
       "epoch": 5.98,
-      "learning_rate": 0.00026246696035242285,
-      "loss": 1.7826,
+      "learning_rate": 0.00026088105726872245,
+      "loss": 0.2522,
       "step": 652
     },
     {
       "epoch": 5.99,
-      "learning_rate": 0.00026220264317180613,
-      "loss": 1.743,
+      "learning_rate": 0.0002606167400881057,
+      "loss": 0.2695,
       "step": 653
     },
     {
       "epoch": 6.0,
-      "learning_rate": 0.0002619383259911894,
-      "loss": 2.1512,
+      "learning_rate": 0.00026035242290748897,
+      "loss": 0.3656,
       "step": 654
     },
     {
       "epoch": 6.01,
-      "learning_rate": 0.00026167400881057265,
-      "loss": 1.7216,
+      "learning_rate": 0.0002600881057268722,
+      "loss": 0.1868,
       "step": 655
     },
     {
       "epoch": 6.02,
-      "learning_rate": 0.00026140969162995593,
-      "loss": 1.7787,
+      "learning_rate": 0.0002598237885462555,
+      "loss": 0.2143,
       "step": 656
     },
     {
       "epoch": 6.03,
-      "learning_rate": 0.0002611453744493392,
-      "loss": 1.7447,
+      "learning_rate": 0.00025955947136563877,
+      "loss": 0.1579,
       "step": 657
     },
     {
       "epoch": 6.04,
-      "learning_rate": 0.00026088105726872245,
-      "loss": 1.6895,
+      "learning_rate": 0.000259295154185022,
+      "loss": 0.226,
       "step": 658
     },
     {
       "epoch": 6.05,
-      "learning_rate": 0.0002606167400881057,
-      "loss": 1.8221,
+      "learning_rate": 0.0002590308370044053,
+      "loss": 0.2669,
       "step": 659
     },
     {
       "epoch": 6.06,
-      "learning_rate": 0.00026035242290748897,
-      "loss": 1.8217,
+      "learning_rate": 0.0002587665198237885,
+      "loss": 0.2189,
       "step": 660
     },
     {
       "epoch": 6.06,
-      "learning_rate": 0.0002600881057268722,
-      "loss": 1.8453,
+      "learning_rate": 0.0002585022026431718,
+      "loss": 0.3625,
       "step": 661
     },
     {
       "epoch": 6.07,
-      "learning_rate": 0.0002598237885462555,
-      "loss": 1.7842,
+      "learning_rate": 0.00025823788546255503,
+      "loss": 0.208,
       "step": 662
     },
     {
       "epoch": 6.08,
-      "learning_rate": 0.00025955947136563877,
-      "loss": 1.8008,
+      "learning_rate": 0.0002579735682819383,
+      "loss": 0.2366,
       "step": 663
     },
     {
       "epoch": 6.09,
-      "learning_rate": 0.000259295154185022,
-      "loss": 1.6501,
+      "learning_rate": 0.00025770925110132155,
+      "loss": 0.2485,
       "step": 664
     },
     {
       "epoch": 6.1,
-      "learning_rate": 0.0002590308370044053,
-      "loss": 1.781,
+      "learning_rate": 0.00025744493392070483,
+      "loss": 0.2683,
       "step": 665
     },
     {
       "epoch": 6.11,
-      "learning_rate": 0.0002587665198237885,
-      "loss": 1.6887,
+      "learning_rate": 0.0002571806167400881,
+      "loss": 0.2011,
       "step": 666
     },
     {
       "epoch": 6.12,
-      "learning_rate": 0.0002585022026431718,
-      "loss": 1.7087,
+      "learning_rate": 0.00025691629955947135,
+      "loss": 0.2022,
       "step": 667
     },
     {
       "epoch": 6.13,
-      "learning_rate": 0.00025823788546255503,
-      "loss": 1.7554,
+      "learning_rate": 0.0002566519823788546,
+      "loss": 0.1996,
       "step": 668
     },
     {
       "epoch": 6.14,
-      "learning_rate": 0.0002579735682819383,
-      "loss": 1.7168,
+      "learning_rate": 0.00025638766519823787,
+      "loss": 0.2015,
       "step": 669
     },
     {
       "epoch": 6.15,
-      "learning_rate": 0.00025770925110132155,
-      "loss": 1.6717,
+      "learning_rate": 0.0002561233480176211,
+      "loss": 0.1825,
       "step": 670
     },
     {
       "epoch": 6.16,
-      "learning_rate": 0.00025744493392070483,
-      "loss": 1.7776,
+      "learning_rate": 0.0002558590308370044,
+      "loss": 0.2698,
       "step": 671
     },
     {
       "epoch": 6.17,
-      "learning_rate": 0.0002571806167400881,
-      "loss": 1.7223,
+      "learning_rate": 0.00025559471365638767,
+      "loss": 0.2194,
       "step": 672
     },
     {
       "epoch": 6.17,
-      "learning_rate": 0.00025691629955947135,
-      "loss": 1.7859,
+      "learning_rate": 0.0002553303964757709,
+      "loss": 0.2564,
       "step": 673
     },
     {
       "epoch": 6.18,
-      "learning_rate": 0.0002566519823788546,
-      "loss": 1.7444,
+      "learning_rate": 0.00025506607929515413,
+      "loss": 0.2014,
       "step": 674
     },
     {
       "epoch": 6.19,
-      "learning_rate": 0.00025638766519823787,
-      "loss": 1.7097,
+      "learning_rate": 0.0002548017621145374,
+      "loss": 0.141,
       "step": 675
     },
     {
       "epoch": 6.2,
-      "learning_rate": 0.0002561233480176211,
-      "loss": 1.6929,
+      "learning_rate": 0.0002545374449339207,
+      "loss": 0.1765,
       "step": 676
     },
     {
       "epoch": 6.21,
-      "learning_rate": 0.0002558590308370044,
-      "loss": 1.7599,
+      "learning_rate": 0.00025427312775330393,
+      "loss": 0.2134,
       "step": 677
     },
     {
       "epoch": 6.22,
-      "learning_rate": 0.00025559471365638767,
-      "loss": 1.6638,
+      "learning_rate": 0.0002540088105726872,
+      "loss": 0.2748,
       "step": 678
     },
     {
       "epoch": 6.23,
-      "learning_rate": 0.0002553303964757709,
-      "loss": 1.659,
+      "learning_rate": 0.00025374449339207045,
+      "loss": 0.276,
       "step": 679
     },
     {
       "epoch": 6.24,
-      "learning_rate": 0.00025506607929515413,
-      "loss": 1.7977,
+      "learning_rate": 0.00025348017621145373,
+      "loss": 0.1729,
       "step": 680
     },
     {
       "epoch": 6.25,
-      "learning_rate": 0.0002548017621145374,
-      "loss": 1.7007,
+      "learning_rate": 0.000253215859030837,
+      "loss": 0.1909,
       "step": 681
     },
     {
       "epoch": 6.26,
-      "learning_rate": 0.0002545374449339207,
-      "loss": 1.6674,
+      "learning_rate": 0.00025295154185022025,
+      "loss": 0.2264,
       "step": 682
     },
     {
       "epoch": 6.27,
-      "learning_rate": 0.00025427312775330393,
-      "loss": 1.6497,
+      "learning_rate": 0.0002526872246696035,
+      "loss": 0.1984,
       "step": 683
     },
     {
       "epoch": 6.28,
-      "learning_rate": 0.0002540088105726872,
-      "loss": 1.7454,
+      "learning_rate": 0.00025242290748898676,
+      "loss": 0.251,
       "step": 684
     },
     {
       "epoch": 6.28,
-      "learning_rate": 0.00025374449339207045,
-      "loss": 1.7096,
+      "learning_rate": 0.00025215859030837005,
+      "loss": 0.221,
       "step": 685
     },
     {
       "epoch": 6.29,
-      "learning_rate": 0.00025348017621145373,
-      "loss": 1.8057,
+      "learning_rate": 0.0002518942731277533,
+      "loss": 0.1975,
       "step": 686
     },
     {
       "epoch": 6.3,
-      "learning_rate": 0.000253215859030837,
-      "loss": 1.7705,
+      "learning_rate": 0.00025162995594713657,
+      "loss": 0.2279,
       "step": 687
     },
     {
       "epoch": 6.31,
-      "learning_rate": 0.00025295154185022025,
-      "loss": 1.676,
+      "learning_rate": 0.0002513656387665198,
+      "loss": 0.1775,
       "step": 688
     },
     {
       "epoch": 6.32,
-      "learning_rate": 0.0002526872246696035,
-      "loss": 1.7219,
+      "learning_rate": 0.00025110132158590303,
+      "loss": 0.2394,
       "step": 689
     },
     {
       "epoch": 6.33,
-      "learning_rate": 0.00025242290748898676,
-      "loss": 1.7371,
+      "learning_rate": 0.0002508370044052863,
+      "loss": 0.2505,
       "step": 690
     },
     {
       "epoch": 6.34,
-      "learning_rate": 0.00025215859030837005,
-      "loss": 1.8171,
+      "learning_rate": 0.0002505726872246696,
+      "loss": 0.268,
       "step": 691
     },
     {
       "epoch": 6.35,
-      "learning_rate": 0.0002518942731277533,
-      "loss": 1.7531,
+      "learning_rate": 0.00025030837004405283,
+      "loss": 0.3044,
       "step": 692
     },
     {
       "epoch": 6.36,
-      "learning_rate": 0.00025162995594713657,
-      "loss": 1.7269,
+      "learning_rate": 0.0002500440528634361,
+      "loss": 0.2008,
       "step": 693
     },
     {
       "epoch": 6.37,
-      "learning_rate": 0.0002513656387665198,
-      "loss": 1.6751,
+      "learning_rate": 0.00024977973568281935,
+      "loss": 0.2192,
       "step": 694
     },
     {
       "epoch": 6.38,
-      "learning_rate": 0.00025110132158590303,
-      "loss": 1.6892,
+      "learning_rate": 0.00024951541850220263,
+      "loss": 0.184,
       "step": 695
     },
     {
       "epoch": 6.39,
-      "learning_rate": 0.0002508370044052863,
-      "loss": 1.6933,
+      "learning_rate": 0.00024925110132158586,
+      "loss": 0.1961,
       "step": 696
     },
     {
       "epoch": 6.39,
-      "learning_rate": 0.0002505726872246696,
-      "loss": 1.698,
+      "learning_rate": 0.00024898678414096915,
+      "loss": 0.2749,
       "step": 697
     },
     {
       "epoch": 6.4,
-      "learning_rate": 0.00025030837004405283,
-      "loss": 1.7008,
+      "learning_rate": 0.0002487224669603524,
+      "loss": 0.2141,
       "step": 698
     },
     {
       "epoch": 6.41,
-      "learning_rate": 0.0002500440528634361,
-      "loss": 1.7438,
+      "learning_rate": 0.00024845814977973566,
+      "loss": 0.1919,
       "step": 699
     },
     {
       "epoch": 6.42,
-      "learning_rate": 0.00024977973568281935,
-      "loss": 1.6772,
+      "learning_rate": 0.00024819383259911895,
+      "loss": 0.1484,
       "step": 700
     },
     {
       "epoch": 6.42,
-      "eval_loss": 0.6171106696128845,
-      "eval_runtime": 20.8666,
-      "eval_samples_per_second": 78.93,
-      "eval_steps_per_second": 1.246,
-      "eval_wer": 0.6269022571749566,
+      "eval_loss": 0.44725894927978516,
+      "eval_runtime": 21.3065,
+      "eval_samples_per_second": 77.3,
+      "eval_steps_per_second": 1.22,
+      "eval_wer": 0.4623633949545501,
       "step": 700
     },
     {
       "epoch": 6.43,
-      "learning_rate": 0.00024951541850220263,
-      "loss": 1.658,
+      "learning_rate": 0.0002479295154185022,
+      "loss": 0.2316,
       "step": 701
     },
     {
       "epoch": 6.44,
-      "learning_rate": 0.00024925110132158586,
-      "loss": 1.679,
+      "learning_rate": 0.00024766519823788547,
+      "loss": 0.2719,
       "step": 702
     },
     {
       "epoch": 6.45,
-      "learning_rate": 0.00024898678414096915,
-      "loss": 1.7094,
+      "learning_rate": 0.0002474008810572687,
+      "loss": 0.2269,
       "step": 703
     },
     {
       "epoch": 6.46,
-      "learning_rate": 0.0002487224669603524,
-      "loss": 1.7007,
+      "learning_rate": 0.00024713656387665193,
+      "loss": 0.2922,
       "step": 704
     },
     {
       "epoch": 6.47,
-      "learning_rate": 0.00024845814977973566,
-      "loss": 1.7109,
+      "learning_rate": 0.0002468722466960352,
+      "loss": 0.1721,
       "step": 705
     },
     {
       "epoch": 6.48,
-      "learning_rate": 0.00024819383259911895,
-      "loss": 1.8622,
+      "learning_rate": 0.0002466079295154185,
+      "loss": 0.1877,
       "step": 706
     },
     {
       "epoch": 6.49,
-      "learning_rate": 0.0002479295154185022,
-      "loss": 1.9508,
+      "learning_rate": 0.00024634361233480173,
+      "loss": 0.1773,
       "step": 707
     },
     {
       "epoch": 6.5,
-      "learning_rate": 0.00024766519823788547,
-      "loss": 1.8974,
+      "learning_rate": 0.000246079295154185,
+      "loss": 0.234,
       "step": 708
     },
     {
       "epoch": 6.5,
-      "learning_rate": 0.0002474008810572687,
-      "loss": 1.783,
+      "learning_rate": 0.00024581497797356825,
+      "loss": 0.2008,
       "step": 709
     },
     {
       "epoch": 6.51,
-      "learning_rate": 0.00024713656387665193,
-      "loss": 1.842,
+      "learning_rate": 0.00024555066079295153,
+      "loss": 0.2841,
       "step": 710
     },
     {
       "epoch": 6.52,
-      "learning_rate": 0.0002468722466960352,
-      "loss": 1.803,
+      "learning_rate": 0.00024528634361233476,
+      "loss": 0.2233,
       "step": 711
     },
     {
       "epoch": 6.53,
-      "learning_rate": 0.0002466079295154185,
-      "loss": 1.7776,
+      "learning_rate": 0.00024502202643171805,
+      "loss": 0.1634,
       "step": 712
     },
     {
       "epoch": 6.54,
-      "learning_rate": 0.00024634361233480173,
-      "loss": 1.8012,
+      "learning_rate": 0.0002447577092511013,
+      "loss": 0.1706,
       "step": 713
     },
     {
       "epoch": 6.55,
-      "learning_rate": 0.000246079295154185,
-      "loss": 1.7424,
+      "learning_rate": 0.00024449339207048456,
+      "loss": 0.1843,
       "step": 714
     },
     {
       "epoch": 6.56,
-      "learning_rate": 0.00024581497797356825,
-      "loss": 1.8033,
+      "learning_rate": 0.00024422907488986785,
+      "loss": 0.2538,
       "step": 715
     },
     {
       "epoch": 6.57,
-      "learning_rate": 0.00024555066079295153,
-      "loss": 1.784,
+      "learning_rate": 0.00024396475770925108,
+      "loss": 0.2078,
       "step": 716
     },
     {
       "epoch": 6.58,
-      "learning_rate": 0.00024528634361233476,
-      "loss": 1.8119,
+      "learning_rate": 0.00024370044052863436,
+      "loss": 0.2662,
       "step": 717
     },
     {
       "epoch": 6.59,
-      "learning_rate": 0.00024502202643171805,
-      "loss": 1.7796,
+      "learning_rate": 0.0002434361233480176,
+      "loss": 0.1827,
       "step": 718
     },
     {
       "epoch": 6.6,
-      "learning_rate": 0.0002447577092511013,
-      "loss": 1.8066,
+      "learning_rate": 0.00024317180616740085,
+      "loss": 0.2322,
       "step": 719
     },
     {
       "epoch": 6.61,
-      "learning_rate": 0.00024449339207048456,
-      "loss": 1.8161,
+      "learning_rate": 0.0002429074889867841,
+      "loss": 0.2418,
       "step": 720
     },
     {
       "epoch": 6.61,
-      "learning_rate": 0.00024422907488986785,
-      "loss": 1.6484,
+      "learning_rate": 0.00024264317180616737,
+      "loss": 0.2417,
       "step": 721
     },
     {
       "epoch": 6.62,
-      "learning_rate": 0.00024396475770925108,
-      "loss": 1.7214,
+      "learning_rate": 0.00024237885462555066,
+      "loss": 0.2523,
       "step": 722
     },
     {
       "epoch": 6.63,
-      "learning_rate": 0.00024370044052863436,
-      "loss": 1.7112,
+      "learning_rate": 0.00024211453744493391,
+      "loss": 0.2274,
       "step": 723
     },
     {
       "epoch": 6.64,
-      "learning_rate": 0.0002434361233480176,
-      "loss": 1.8136,
+      "learning_rate": 0.00024185022026431717,
+      "loss": 0.2396,
       "step": 724
     },
     {
       "epoch": 6.65,
-      "learning_rate": 0.00024317180616740085,
-      "loss": 1.7552,
+      "learning_rate": 0.0002415859030837004,
+      "loss": 0.1321,
       "step": 725
     },
     {
       "epoch": 6.66,
-      "learning_rate": 0.0002429074889867841,
-      "loss": 1.7602,
+      "learning_rate": 0.00024132158590308366,
+      "loss": 0.1649,
       "step": 726
     },
     {
       "epoch": 6.67,
-      "learning_rate": 0.00024264317180616737,
-      "loss": 1.8713,
+      "learning_rate": 0.00024105726872246695,
+      "loss": 0.2182,
       "step": 727
     },
     {
       "epoch": 6.68,
-      "learning_rate": 0.00024237885462555066,
-      "loss": 1.8833,
+      "learning_rate": 0.0002407929515418502,
+      "loss": 0.2395,
       "step": 728
     },
     {
       "epoch": 6.69,
-      "learning_rate": 0.00024211453744493391,
-      "loss": 1.813,
+      "learning_rate": 0.00024052863436123346,
+      "loss": 0.337,
       "step": 729
     },
     {
       "epoch": 6.7,
-      "learning_rate": 0.00024185022026431717,
-      "loss": 1.9528,
+      "learning_rate": 0.00024026431718061672,
+      "loss": 0.2058,
       "step": 730
     },
     {
       "epoch": 6.71,
-      "learning_rate": 0.0002415859030837004,
-      "loss": 1.8044,
+      "learning_rate": 0.00023999999999999998,
+      "loss": 0.2289,
       "step": 731
     },
     {
       "epoch": 6.72,
-      "learning_rate": 0.00024132158590308366,
-      "loss": 1.8028,
+      "learning_rate": 0.0002397356828193832,
+      "loss": 0.1534,
       "step": 732
     },
     {
       "epoch": 6.72,
-      "learning_rate": 0.00024105726872246695,
-      "loss": 1.8632,
+      "learning_rate": 0.0002394713656387665,
+      "loss": 0.1888,
       "step": 733
     },
     {
       "epoch": 6.73,
-      "learning_rate": 0.0002407929515418502,
-      "loss": 1.927,
+      "learning_rate": 0.00023920704845814975,
+      "loss": 0.2931,
       "step": 734
     },
     {
       "epoch": 6.74,
-      "learning_rate": 0.00024052863436123346,
-      "loss": 1.8165,
+      "learning_rate": 0.000238942731277533,
+      "loss": 0.2858,
       "step": 735
     },
     {
       "epoch": 6.75,
-      "learning_rate": 0.00024026431718061672,
-      "loss": 2.0022,
+      "learning_rate": 0.00023867841409691627,
+      "loss": 0.184,
       "step": 736
     },
     {
       "epoch": 6.76,
-      "learning_rate": 0.00023999999999999998,
-      "loss": 1.7592,
+      "learning_rate": 0.00023841409691629955,
+      "loss": 0.1908,
       "step": 737
     },
     {
       "epoch": 6.77,
-      "learning_rate": 0.0002397356828193832,
-      "loss": 1.7187,
+      "learning_rate": 0.0002381497797356828,
+      "loss": 0.1862,
       "step": 738
     },
     {
       "epoch": 6.78,
-      "learning_rate": 0.0002394713656387665,
-      "loss": 1.743,
+      "learning_rate": 0.00023788546255506604,
+      "loss": 0.1371,
       "step": 739
     },
     {
       "epoch": 6.79,
-      "learning_rate": 0.00023920704845814975,
-      "loss": 1.7974,
+      "learning_rate": 0.0002376211453744493,
+      "loss": 0.2329,
       "step": 740
     },
     {
       "epoch": 6.8,
-      "learning_rate": 0.000238942731277533,
-      "loss": 1.7439,
+      "learning_rate": 0.00023735682819383256,
+      "loss": 0.2378,
       "step": 741
     },
     {
       "epoch": 6.81,
-      "learning_rate": 0.00023867841409691627,
-      "loss": 1.7994,
+      "learning_rate": 0.00023709251101321585,
+      "loss": 0.2129,
       "step": 742
     },
     {
       "epoch": 6.82,
-      "learning_rate": 0.00023841409691629955,
-      "loss": 1.7032,
+      "learning_rate": 0.0002368281938325991,
+      "loss": 0.1315,
       "step": 743
     },
     {
       "epoch": 6.83,
-      "learning_rate": 0.0002381497797356828,
-      "loss": 1.7411,
+      "learning_rate": 0.00023656387665198236,
+      "loss": 0.1781,
       "step": 744
     },
     {
       "epoch": 6.83,
-      "learning_rate": 0.00023788546255506604,
-      "loss": 1.7417,
+      "learning_rate": 0.00023629955947136562,
+      "loss": 0.2376,
       "step": 745
     },
     {
       "epoch": 6.84,
-      "learning_rate": 0.0002376211453744493,
-      "loss": 1.7555,
+      "learning_rate": 0.0002360352422907489,
+      "loss": 0.231,
       "step": 746
     },
     {
       "epoch": 6.85,
-      "learning_rate": 0.00023735682819383256,
-      "loss": 1.7636,
+      "learning_rate": 0.00023577092511013214,
+      "loss": 0.2704,
       "step": 747
     },
     {
       "epoch": 6.86,
-      "learning_rate": 0.00023709251101321585,
-      "loss": 1.7361,
+      "learning_rate": 0.0002355066079295154,
+      "loss": 0.2497,
       "step": 748
     },
     {
       "epoch": 6.87,
-      "learning_rate": 0.0002368281938325991,
-      "loss": 1.8248,
+      "learning_rate": 0.00023524229074889865,
+      "loss": 0.1828,
       "step": 749
     },
     {
       "epoch": 6.88,
-      "learning_rate": 0.00023656387665198236,
-      "loss": 1.7235,
+      "learning_rate": 0.0002349779735682819,
+      "loss": 0.1667,
       "step": 750
     },
     {
       "epoch": 6.89,
-      "learning_rate": 0.00023629955947136562,
-      "loss": 1.7269,
+      "learning_rate": 0.0002347136563876652,
+      "loss": 0.1568,
       "step": 751
     },
     {
       "epoch": 6.9,
-      "learning_rate": 0.0002360352422907489,
-      "loss": 1.7182,
+      "learning_rate": 0.00023444933920704845,
+      "loss": 0.2251,
       "step": 752
     },
     {
       "epoch": 6.91,
-      "learning_rate": 0.00023577092511013214,
-      "loss": 1.628,
+      "learning_rate": 0.0002341850220264317,
+      "loss": 0.1459,
       "step": 753
     },
     {
       "epoch": 6.92,
-      "learning_rate": 0.0002355066079295154,
-      "loss": 1.832,
+      "learning_rate": 0.00023392070484581494,
+      "loss": 0.3429,
       "step": 754
     },
     {
       "epoch": 6.93,
-      "learning_rate": 0.00023524229074889865,
-      "loss": 1.7814,
+      "learning_rate": 0.0002336563876651982,
+      "loss": 0.1364,
       "step": 755
     },
     {
       "epoch": 6.94,
-      "learning_rate": 0.0002349779735682819,
-      "loss": 1.7063,
+      "learning_rate": 0.00023339207048458146,
+      "loss": 0.2423,
       "step": 756
     },
     {
       "epoch": 6.94,
-      "learning_rate": 0.0002347136563876652,
-      "loss": 1.7003,
+      "learning_rate": 0.00023312775330396474,
+      "loss": 0.166,
       "step": 757
     },
     {
       "epoch": 6.95,
-      "learning_rate": 0.00023444933920704845,
-      "loss": 1.7253,
+      "learning_rate": 0.000232863436123348,
+      "loss": 0.2024,
       "step": 758
     },
     {
       "epoch": 6.96,
-      "learning_rate": 0.0002341850220264317,
-      "loss": 1.6989,
+      "learning_rate": 0.00023259911894273126,
+      "loss": 0.2621,
       "step": 759
     },
     {
       "epoch": 6.97,
-      "learning_rate": 0.00023392070484581494,
-      "loss": 1.8083,
+      "learning_rate": 0.00023233480176211452,
+      "loss": 0.2241,
       "step": 760
     },
     {
       "epoch": 6.98,
-      "learning_rate": 0.0002336563876651982,
-      "loss": 1.753,
+      "learning_rate": 0.00023207048458149775,
+      "loss": 0.1972,
       "step": 761
     },
     {
       "epoch": 6.99,
-      "learning_rate": 0.00023339207048458146,
-      "loss": 1.7658,
+      "learning_rate": 0.00023180616740088104,
+      "loss": 0.2454,
       "step": 762
     },
     {
       "epoch": 7.0,
-      "learning_rate": 0.00023312775330396474,
-      "loss": 2.1726,
+      "learning_rate": 0.0002315418502202643,
+      "loss": 0.2874,
       "step": 763
     },
     {
       "epoch": 7.01,
-      "learning_rate": 0.000232863436123348,
-      "loss": 1.7021,
+      "learning_rate": 0.00023127753303964755,
+      "loss": 0.1194,
       "step": 764
     },
     {
       "epoch": 7.02,
-      "learning_rate": 0.00023259911894273126,
-      "loss": 1.6523,
+      "learning_rate": 0.0002310132158590308,
+      "loss": 0.1803,
       "step": 765
     },
     {
       "epoch": 7.03,
-      "learning_rate": 0.00023233480176211452,
-      "loss": 1.6579,
+      "learning_rate": 0.0002307488986784141,
+      "loss": 0.1668,
       "step": 766
     },
     {
       "epoch": 7.04,
-      "learning_rate": 0.00023207048458149775,
-      "loss": 1.6989,
+      "learning_rate": 0.00023048458149779735,
+      "loss": 0.1583,
       "step": 767
     },
     {
       "epoch": 7.05,
-      "learning_rate": 0.00023180616740088104,
-      "loss": 1.7047,
+      "learning_rate": 0.00023022026431718058,
+      "loss": 0.2183,
       "step": 768
     },
     {
       "epoch": 7.06,
-      "learning_rate": 0.0002315418502202643,
-      "loss": 1.7122,
+      "learning_rate": 0.00022995594713656384,
+      "loss": 0.2507,
       "step": 769
     },
     {
       "epoch": 7.06,
-      "learning_rate": 0.00023127753303964755,
-      "loss": 1.6775,
+      "learning_rate": 0.0002296916299559471,
+      "loss": 0.1699,
       "step": 770
     },
     {
       "epoch": 7.07,
-      "learning_rate": 0.0002310132158590308,
-      "loss": 1.7112,
+      "learning_rate": 0.00022942731277533039,
+      "loss": 0.1684,
       "step": 771
     },
     {
       "epoch": 7.08,
-      "learning_rate": 0.0002307488986784141,
-      "loss": 1.763,
+      "learning_rate": 0.00022916299559471364,
+      "loss": 0.2108,
       "step": 772
     },
     {
       "epoch": 7.09,
-      "learning_rate": 0.00023048458149779735,
-      "loss": 1.7288,
+      "learning_rate": 0.0002288986784140969,
+      "loss": 0.1444,
       "step": 773
     },
     {
       "epoch": 7.1,
-      "learning_rate": 0.00023022026431718058,
-      "loss": 1.6951,
+      "learning_rate": 0.00022863436123348016,
+      "loss": 0.2075,
       "step": 774
     },
     {
       "epoch": 7.11,
-      "learning_rate": 0.00022995594713656384,
-      "loss": 1.6846,
+      "learning_rate": 0.00022837004405286345,
+      "loss": 0.2776,
       "step": 775
     },
     {
       "epoch": 7.12,
-      "learning_rate": 0.0002296916299559471,
-      "loss": 1.8127,
+      "learning_rate": 0.00022810572687224668,
+      "loss": 0.2262,
       "step": 776
     },
     {
       "epoch": 7.13,
-      "learning_rate": 0.00022942731277533039,
-      "loss": 1.7197,
+      "learning_rate": 0.00022784140969162993,
+      "loss": 0.1307,
       "step": 777
     },
     {
       "epoch": 7.14,
-      "learning_rate": 0.00022916299559471364,
-      "loss": 1.7206,
+      "learning_rate": 0.0002275770925110132,
+      "loss": 0.1265,
       "step": 778
     },
     {
       "epoch": 7.15,
-      "learning_rate": 0.0002288986784140969,
-      "loss": 1.7696,
+      "learning_rate": 0.00022731277533039645,
+      "loss": 0.1651,
       "step": 779
     },
     {
       "epoch": 7.16,
-      "learning_rate": 0.00022863436123348016,
-      "loss": 1.6943,
+      "learning_rate": 0.0002270484581497797,
+      "loss": 0.1558,
       "step": 780
     },
     {
       "epoch": 7.17,
-      "learning_rate": 0.00022837004405286345,
-      "loss": 1.7233,
+      "learning_rate": 0.000226784140969163,
+      "loss": 0.2029,
       "step": 781
     },
     {
       "epoch": 7.17,
-      "learning_rate": 0.00022810572687224668,
-      "loss": 1.6809,
+      "learning_rate": 0.00022651982378854625,
+      "loss": 0.2891,
       "step": 782
     },
     {
       "epoch": 7.18,
-      "learning_rate": 0.00022784140969162993,
-      "loss": 1.765,
+      "learning_rate": 0.00022625550660792948,
+      "loss": 0.1478,
       "step": 783
     },
     {
       "epoch": 7.19,
-      "learning_rate": 0.0002275770925110132,
-      "loss": 1.7707,
+      "learning_rate": 0.00022599118942731274,
+      "loss": 0.1648,
       "step": 784
     },
     {
       "epoch": 7.2,
-      "learning_rate": 0.00022731277533039645,
-      "loss": 1.6673,
+      "learning_rate": 0.000225726872246696,
+      "loss": 0.1438,
       "step": 785
     },
     {
       "epoch": 7.21,
-      "learning_rate": 0.0002270484581497797,
-      "loss": 1.6652,
+      "learning_rate": 0.00022546255506607929,
+      "loss": 0.1161,
       "step": 786
     },
     {
       "epoch": 7.22,
-      "learning_rate": 0.000226784140969163,
-      "loss": 1.6569,
+      "learning_rate": 0.00022519823788546254,
+      "loss": 0.2302,
       "step": 787
     },
     {
       "epoch": 7.23,
-      "learning_rate": 0.00022651982378854625,
-      "loss": 1.7319,
+      "learning_rate": 0.0002249339207048458,
+      "loss": 0.2175,
       "step": 788
     },
     {
       "epoch": 7.24,
-      "learning_rate": 0.00022625550660792948,
-      "loss": 1.7634,
+      "learning_rate": 0.00022466960352422906,
+      "loss": 0.2133,
       "step": 789
     },
     {
       "epoch": 7.25,
-      "learning_rate": 0.00022599118942731274,
-      "loss": 1.6778,
+      "learning_rate": 0.0002244052863436123,
+      "loss": 0.1251,
       "step": 790
     },
     {
       "epoch": 7.26,
-      "learning_rate": 0.000225726872246696,
-      "loss": 1.7072,
+      "learning_rate": 0.00022414096916299558,
+      "loss": 0.194,
       "step": 791
     },
     {
       "epoch": 7.27,
-      "learning_rate": 0.00022546255506607929,
-      "loss": 1.7117,
+      "learning_rate": 0.00022387665198237883,
+      "loss": 0.1543,
       "step": 792
     },
     {
       "epoch": 7.28,
-      "learning_rate": 0.00022519823788546254,
-      "loss": 1.6583,
+      "learning_rate": 0.0002236123348017621,
+      "loss": 0.1873,
       "step": 793
     },
     {
       "epoch": 7.28,
-      "learning_rate": 0.0002249339207048458,
-      "loss": 1.619,
+      "learning_rate": 0.00022334801762114535,
+      "loss": 0.185,
       "step": 794
     },
     {
       "epoch": 7.29,
-      "learning_rate": 0.00022466960352422906,
-      "loss": 1.7795,
+      "learning_rate": 0.00022308370044052864,
+      "loss": 0.1767,
       "step": 795
     },
     {
       "epoch": 7.3,
-      "learning_rate": 0.0002244052863436123,
-      "loss": 1.7374,
+      "learning_rate": 0.0002228193832599119,
+      "loss": 0.1263,
       "step": 796
     },
     {
       "epoch": 7.31,
-      "learning_rate": 0.00022414096916299558,
-      "loss": 1.7857,
+      "learning_rate": 0.00022255506607929512,
+      "loss": 0.112,
       "step": 797
     },
     {
       "epoch": 7.32,
-      "learning_rate": 0.00022387665198237883,
-      "loss": 1.7872,
+      "learning_rate": 0.00022229074889867838,
+      "loss": 0.2028,
       "step": 798
     },
     {
       "epoch": 7.33,
-      "learning_rate": 0.0002236123348017621,
-      "loss": 1.7709,
+      "learning_rate": 0.00022202643171806164,
+      "loss": 0.2288,
       "step": 799
     },
     {
       "epoch": 7.34,
-      "learning_rate": 0.00022334801762114535,
-      "loss": 1.7139,
+      "learning_rate": 0.0002217621145374449,
+      "loss": 0.1801,
       "step": 800
     },
     {
       "epoch": 7.34,
-      "eval_loss": 0.6518137454986572,
-      "eval_runtime": 20.8939,
-      "eval_samples_per_second": 78.827,
-      "eval_steps_per_second": 1.244,
-      "eval_wer": 0.6261873148810132,
+      "eval_loss": 0.40969517827033997,
+      "eval_runtime": 21.0436,
+      "eval_samples_per_second": 78.266,
+      "eval_steps_per_second": 1.236,
+      "eval_wer": 0.428454703298948,
       "step": 800
     },
     {
       "epoch": 7.35,
-      "learning_rate": 0.00022308370044052864,
-      "loss": 1.7601,
+      "learning_rate": 0.00022149779735682818,
+      "loss": 0.1673,
       "step": 801
     },
     {
       "epoch": 7.36,
-      "learning_rate": 0.0002228193832599119,
-      "loss": 1.656,
+      "learning_rate": 0.00022123348017621144,
+      "loss": 0.1298,
       "step": 802
     },
     {
       "epoch": 7.37,
-      "learning_rate": 0.00022255506607929512,
-      "loss": 1.7154,
+      "learning_rate": 0.0002209691629955947,
+      "loss": 0.1701,
       "step": 803
     },
     {
       "epoch": 7.38,
-      "learning_rate": 0.00022229074889867838,
-      "loss": 1.6442,
+      "learning_rate": 0.00022070484581497796,
+      "loss": 0.2269,
       "step": 804
     },
     {
       "epoch": 7.39,
-      "learning_rate": 0.00022202643171806164,
-      "loss": 1.7488,
+      "learning_rate": 0.0002204405286343612,
+      "loss": 0.1996,
       "step": 805
     },
     {
       "epoch": 7.39,
-      "learning_rate": 0.0002217621145374449,
-      "loss": 1.6336,
+      "learning_rate": 0.00022017621145374448,
+      "loss": 0.1237,
       "step": 806
     },
     {
       "epoch": 7.4,
-      "learning_rate": 0.00022149779735682818,
-      "loss": 1.7226,
+      "learning_rate": 0.00021991189427312773,
+      "loss": 0.2968,
       "step": 807
     },
     {
       "epoch": 7.41,
-      "learning_rate": 0.00022123348017621144,
-      "loss": 1.744,
+      "learning_rate": 0.000219647577092511,
+      "loss": 0.1781,
       "step": 808
     },
     {
       "epoch": 7.42,
-      "learning_rate": 0.0002209691629955947,
-      "loss": 1.6906,
+      "learning_rate": 0.00021938325991189425,
+      "loss": 0.1663,
       "step": 809
     },
     {
       "epoch": 7.43,
-      "learning_rate": 0.00022070484581497796,
-      "loss": 1.7315,
+      "learning_rate": 0.00021911894273127753,
+      "loss": 0.1419,
       "step": 810
     },
     {
       "epoch": 7.44,
-      "learning_rate": 0.0002204405286343612,
-      "loss": 1.6515,
+      "learning_rate": 0.0002188546255506608,
+      "loss": 0.1551,
       "step": 811
     },
     {
       "epoch": 7.45,
-      "learning_rate": 0.00022017621145374448,
-      "loss": 1.678,
+      "learning_rate": 0.00021859030837004402,
+      "loss": 0.2213,
       "step": 812
     },
     {
       "epoch": 7.46,
-      "learning_rate": 0.00021991189427312773,
-      "loss": 1.696,
+      "learning_rate": 0.00021832599118942728,
+      "loss": 0.2327,
       "step": 813
     },
     {
       "epoch": 7.47,
-      "learning_rate": 0.000219647577092511,
-      "loss": 1.7406,
+      "learning_rate": 0.00021806167400881054,
+      "loss": 0.0938,
       "step": 814
     },
     {
       "epoch": 7.48,
-      "learning_rate": 0.00021938325991189425,
-      "loss": 1.6292,
+      "learning_rate": 0.00021779735682819383,
+      "loss": 0.1363,
       "step": 815
     },
     {
       "epoch": 7.49,
-      "learning_rate": 0.00021911894273127753,
-      "loss": 1.7183,
+      "learning_rate": 0.00021753303964757708,
+      "loss": 0.1534,
       "step": 816
     },
     {
       "epoch": 7.5,
-      "learning_rate": 0.0002188546255506608,
-      "loss": 1.6959,
+      "learning_rate": 0.00021726872246696034,
+      "loss": 0.1775,
       "step": 817
     },
     {
       "epoch": 7.5,
-      "learning_rate": 0.00021859030837004402,
-      "loss": 1.7002,
+      "learning_rate": 0.0002170044052863436,
+      "loss": 0.1932,
       "step": 818
     },
     {
       "epoch": 7.51,
-      "learning_rate": 0.00021832599118942728,
-      "loss": 1.6983,
+      "learning_rate": 0.00021674008810572683,
+      "loss": 0.2554,
       "step": 819
     },
     {
       "epoch": 7.52,
-      "learning_rate": 0.00021806167400881054,
-      "loss": 1.7621,
+      "learning_rate": 0.00021647577092511012,
+      "loss": 0.1672,
       "step": 820
     },
     {
       "epoch": 7.53,
-      "learning_rate": 0.00021779735682819383,
-      "loss": 1.6705,
+      "learning_rate": 0.00021621145374449337,
+      "loss": 0.1025,
       "step": 821
     },
     {
       "epoch": 7.54,
-      "learning_rate": 0.00021753303964757708,
-      "loss": 1.6668,
+      "learning_rate": 0.00021594713656387663,
+      "loss": 0.169,
       "step": 822
     },
     {
       "epoch": 7.55,
-      "learning_rate": 0.00021726872246696034,
-      "loss": 1.8106,
+      "learning_rate": 0.0002156828193832599,
+      "loss": 0.178,
       "step": 823
     },
     {
       "epoch": 7.56,
-      "learning_rate": 0.0002170044052863436,
-      "loss": 1.6704,
+      "learning_rate": 0.00021541850220264315,
+      "loss": 0.1614,
       "step": 824
     },
     {
       "epoch": 7.57,
-      "learning_rate": 0.00021674008810572683,
-      "loss": 1.7363,
+      "learning_rate": 0.00021515418502202643,
+      "loss": 0.2649,
       "step": 825
     },
     {
       "epoch": 7.58,
-      "learning_rate": 0.00021647577092511012,
-      "loss": 1.7024,
+      "learning_rate": 0.00021488986784140967,
+      "loss": 0.1525,
       "step": 826
     },
     {
       "epoch": 7.59,
-      "learning_rate": 0.00021621145374449337,
-      "loss": 1.7021,
+      "learning_rate": 0.00021462555066079292,
+      "loss": 0.1949,
       "step": 827
     },
     {
       "epoch": 7.6,
-      "learning_rate": 0.00021594713656387663,
-      "loss": 1.6228,
+      "learning_rate": 0.00021436123348017618,
+      "loss": 0.1786,
       "step": 828
     },
     {
       "epoch": 7.61,
-      "learning_rate": 0.0002156828193832599,
-      "loss": 1.6462,
+      "learning_rate": 0.00021409691629955944,
+      "loss": 0.1504,
       "step": 829
     },
     {
       "epoch": 7.61,
-      "learning_rate": 0.00021541850220264315,
-      "loss": 1.6967,
+      "learning_rate": 0.00021383259911894272,
+      "loss": 0.216,
       "step": 830
     },
     {
       "epoch": 7.62,
-      "learning_rate": 0.00021515418502202643,
-      "loss": 1.6666,
+      "learning_rate": 0.00021356828193832598,
+      "loss": 0.2152,
       "step": 831
     },
     {
       "epoch": 7.63,
-      "learning_rate": 0.00021488986784140967,
-      "loss": 1.6013,
+      "learning_rate": 0.00021330396475770924,
+      "loss": 0.2064,
       "step": 832
     },
     {
       "epoch": 7.64,
-      "learning_rate": 0.00021462555066079292,
-      "loss": 1.7204,
+      "learning_rate": 0.00021303964757709247,
+      "loss": 0.1383,
       "step": 833
     },
     {
       "epoch": 7.65,
-      "learning_rate": 0.00021436123348017618,
-      "loss": 1.642,
+      "learning_rate": 0.00021277533039647573,
+      "loss": 0.1251,
       "step": 834
     },
     {
       "epoch": 7.66,
-      "learning_rate": 0.00021409691629955944,
-      "loss": 1.6011,
+      "learning_rate": 0.00021251101321585902,
+      "loss": 0.1835,
       "step": 835
     },
     {
       "epoch": 7.67,
-      "learning_rate": 0.00021383259911894272,
-      "loss": 1.6703,
+      "learning_rate": 0.00021224669603524227,
+      "loss": 0.1845,
       "step": 836
     },
     {
       "epoch": 7.68,
-      "learning_rate": 0.00021356828193832598,
-      "loss": 1.6139,
+      "learning_rate": 0.00021198237885462553,
+      "loss": 0.1915,
       "step": 837
     },
     {
       "epoch": 7.69,
-      "learning_rate": 0.00021330396475770924,
-      "loss": 1.7284,
+      "learning_rate": 0.0002117180616740088,
+      "loss": 0.2982,
       "step": 838
     },
     {
       "epoch": 7.7,
-      "learning_rate": 0.00021303964757709247,
-      "loss": 1.7131,
+      "learning_rate": 0.00021145374449339208,
+      "loss": 0.1006,
       "step": 839
     },
     {
       "epoch": 7.71,
-      "learning_rate": 0.00021277533039647573,
-      "loss": 1.6328,
+      "learning_rate": 0.00021118942731277533,
+      "loss": 0.1313,
       "step": 840
     },
     {
       "epoch": 7.72,
-      "learning_rate": 0.00021251101321585902,
-      "loss": 1.6618,
+      "learning_rate": 0.00021092511013215856,
+      "loss": 0.1281,
       "step": 841
     },
     {
       "epoch": 7.72,
-      "learning_rate": 0.00021224669603524227,
-      "loss": 1.6476,
+      "learning_rate": 0.00021066079295154182,
+      "loss": 0.1432,
       "step": 842
     },
     {
       "epoch": 7.73,
-      "learning_rate": 0.00021198237885462553,
-      "loss": 1.7134,
+      "learning_rate": 0.00021039647577092508,
+      "loss": 0.1616,
       "step": 843
     },
     {
       "epoch": 7.74,
-      "learning_rate": 0.0002117180616740088,
-      "loss": 1.6916,
+      "learning_rate": 0.00021013215859030837,
+      "loss": 0.2699,
       "step": 844
     },
     {
       "epoch": 7.75,
-      "learning_rate": 0.00021145374449339208,
-      "loss": 1.8085,
+      "learning_rate": 0.00020986784140969162,
+      "loss": 0.1141,
       "step": 845
     },
     {
       "epoch": 7.76,
-      "learning_rate": 0.00021118942731277533,
-      "loss": 1.6404,
+      "learning_rate": 0.00020960352422907488,
+      "loss": 0.1538,
       "step": 846
     },
     {
       "epoch": 7.77,
-      "learning_rate": 0.00021092511013215856,
-      "loss": 1.6518,
+      "learning_rate": 0.00020933920704845814,
+      "loss": 0.1377,
       "step": 847
     },
     {
       "epoch": 7.78,
-      "learning_rate": 0.00021066079295154182,
-      "loss": 1.6264,
+      "learning_rate": 0.00020907488986784137,
+      "loss": 0.1717,
       "step": 848
     },
     {
       "epoch": 7.79,
-      "learning_rate": 0.00021039647577092508,
-      "loss": 1.6682,
+      "learning_rate": 0.00020881057268722463,
+      "loss": 0.2159,
       "step": 849
     },
     {
       "epoch": 7.8,
-      "learning_rate": 0.00021013215859030837,
-      "loss": 1.5866,
+      "learning_rate": 0.00020854625550660791,
+      "loss": 0.1836,
       "step": 850
     },
     {
       "epoch": 7.81,
-      "learning_rate": 0.00020986784140969162,
-      "loss": 1.7107,
+      "learning_rate": 0.00020828193832599117,
+      "loss": 0.2245,
       "step": 851
     },
     {
       "epoch": 7.82,
-      "learning_rate": 0.00020960352422907488,
-      "loss": 1.6214,
+      "learning_rate": 0.00020801762114537443,
+      "loss": 0.1304,
       "step": 852
     },
     {
       "epoch": 7.83,
-      "learning_rate": 0.00020933920704845814,
-      "loss": 1.6158,
+      "learning_rate": 0.0002077533039647577,
+      "loss": 0.1473,
       "step": 853
     },
     {
       "epoch": 7.83,
-      "learning_rate": 0.00020907488986784137,
-      "loss": 1.6769,
+      "learning_rate": 0.00020748898678414097,
+      "loss": 0.1814,
       "step": 854
     },
     {
       "epoch": 7.84,
-      "learning_rate": 0.00020881057268722463,
-      "loss": 1.5869,
+      "learning_rate": 0.0002072246696035242,
+      "loss": 0.2117,
       "step": 855
     },
     {
       "epoch": 7.85,
-      "learning_rate": 0.00020854625550660791,
-      "loss": 1.6152,
+      "learning_rate": 0.00020696035242290746,
+      "loss": 0.1704,
       "step": 856
     },
     {
       "epoch": 7.86,
-      "learning_rate": 0.00020828193832599117,
-      "loss": 1.7082,
+      "learning_rate": 0.00020669603524229072,
+      "loss": 0.1865,
       "step": 857
     },
     {
       "epoch": 7.87,
-      "learning_rate": 0.00020801762114537443,
-      "loss": 1.6231,
+      "learning_rate": 0.00020643171806167398,
+      "loss": 0.1798,
       "step": 858
     },
     {
       "epoch": 7.88,
-      "learning_rate": 0.0002077533039647577,
-      "loss": 1.6374,
+      "learning_rate": 0.00020616740088105727,
+      "loss": 0.1951,
       "step": 859
     },
     {
       "epoch": 7.89,
-      "learning_rate": 0.00020748898678414097,
-      "loss": 1.6668,
+      "learning_rate": 0.00020590308370044052,
+      "loss": 0.1616,
       "step": 860
     },
     {
       "epoch": 7.9,
-      "learning_rate": 0.0002072246696035242,
-      "loss": 1.6537,
+      "learning_rate": 0.00020563876651982378,
+      "loss": 0.1541,
       "step": 861
     },
     {
       "epoch": 7.91,
-      "learning_rate": 0.00020696035242290746,
-      "loss": 1.7657,
+      "learning_rate": 0.000205374449339207,
+      "loss": 0.2718,
       "step": 862
     },
     {
       "epoch": 7.92,
-      "learning_rate": 0.00020669603524229072,
-      "loss": 1.6784,
+      "learning_rate": 0.00020511013215859027,
+      "loss": 0.2051,
       "step": 863
     },
     {
       "epoch": 7.93,
-      "learning_rate": 0.00020643171806167398,
-      "loss": 1.7346,
+      "learning_rate": 0.00020484581497797356,
+      "loss": 0.1709,
       "step": 864
     },
     {
       "epoch": 7.94,
-      "learning_rate": 0.00020616740088105727,
-      "loss": 1.6958,
+      "learning_rate": 0.00020458149779735681,
+      "loss": 0.1119,
       "step": 865
     },
     {
       "epoch": 7.94,
-      "learning_rate": 0.00020590308370044052,
-      "loss": 1.7143,
+      "learning_rate": 0.00020431718061674007,
+      "loss": 0.1663,
       "step": 866
     },
     {
       "epoch": 7.95,
-      "learning_rate": 0.00020563876651982378,
-      "loss": 1.6537,
+      "learning_rate": 0.00020405286343612333,
+      "loss": 0.2377,
       "step": 867
     },
     {
       "epoch": 7.96,
-      "learning_rate": 0.000205374449339207,
-      "loss": 1.6154,
+      "learning_rate": 0.0002037885462555066,
+      "loss": 0.2317,
       "step": 868
     },
     {
       "epoch": 7.97,
-      "learning_rate": 0.00020511013215859027,
-      "loss": 1.7157,
+      "learning_rate": 0.00020352422907488987,
+      "loss": 0.2326,
       "step": 869
     },
     {
       "epoch": 7.98,
-      "learning_rate": 0.00020484581497797356,
-      "loss": 1.6345,
+      "learning_rate": 0.0002032599118942731,
+      "loss": 0.1692,
       "step": 870
     },
     {
       "epoch": 7.99,
-      "learning_rate": 0.00020458149779735681,
-      "loss": 1.6598,
+      "learning_rate": 0.00020299559471365636,
+      "loss": 0.1838,
       "step": 871
     },
     {
       "epoch": 8.0,
-      "learning_rate": 0.00020431718061674007,
-      "loss": 2.1348,
+      "learning_rate": 0.00020273127753303962,
+      "loss": 0.2223,
       "step": 872
     },
     {
       "epoch": 8.01,
-      "learning_rate": 0.00020405286343612333,
-      "loss": 1.6902,
+      "learning_rate": 0.00020246696035242288,
+      "loss": 0.1387,
       "step": 873
     },
     {
       "epoch": 8.02,
-      "learning_rate": 0.0002037885462555066,
-      "loss": 1.7089,
+      "learning_rate": 0.00020220264317180616,
+      "loss": 0.1372,
       "step": 874
     },
     {
       "epoch": 8.03,
-      "learning_rate": 0.00020352422907488987,
-      "loss": 1.5706,
+      "learning_rate": 0.00020193832599118942,
+      "loss": 0.1138,
       "step": 875
     },
     {
       "epoch": 8.04,
-      "learning_rate": 0.0002032599118942731,
-      "loss": 1.5755,
+      "learning_rate": 0.00020167400881057268,
+      "loss": 0.1358,
       "step": 876
     },
     {
       "epoch": 8.05,
-      "learning_rate": 0.00020299559471365636,
-      "loss": 1.615,
+      "learning_rate": 0.0002014096916299559,
+      "loss": 0.1682,
       "step": 877
     },
     {
       "epoch": 8.06,
-      "learning_rate": 0.00020273127753303962,
-      "loss": 1.6768,
+      "learning_rate": 0.00020114537444933917,
+      "loss": 0.1421,
       "step": 878
     },
     {
       "epoch": 8.06,
-      "learning_rate": 0.00020246696035242288,
-      "loss": 1.7643,
+      "learning_rate": 0.00020088105726872246,
+      "loss": 0.1382,
       "step": 879
     },
     {
       "epoch": 8.07,
-      "learning_rate": 0.00020220264317180616,
-      "loss": 1.7449,
+      "learning_rate": 0.0002006167400881057,
+      "loss": 0.1307,
       "step": 880
     },
     {
       "epoch": 8.08,
-      "learning_rate": 0.00020193832599118942,
-      "loss": 1.6319,
+      "learning_rate": 0.00020035242290748897,
+      "loss": 0.0951,
       "step": 881
     },
     {
       "epoch": 8.09,
-      "learning_rate": 0.00020167400881057268,
-      "loss": 1.6433,
+      "learning_rate": 0.00020008810572687223,
+      "loss": 0.1337,
       "step": 882
     },
     {
       "epoch": 8.1,
-      "learning_rate": 0.0002014096916299559,
-      "loss": 1.6226,
+      "learning_rate": 0.00019982378854625552,
+      "loss": 0.1175,
       "step": 883
     },
     {
       "epoch": 8.11,
-      "learning_rate": 0.00020114537444933917,
-      "loss": 1.755,
+      "learning_rate": 0.00019955947136563875,
+      "loss": 0.2996,
       "step": 884
     },
     {
       "epoch": 8.12,
-      "learning_rate": 0.00020088105726872246,
-      "loss": 1.6302,
+      "learning_rate": 0.000199295154185022,
+      "loss": 0.1636,
       "step": 885
     },
     {
       "epoch": 8.13,
-      "learning_rate": 0.0002006167400881057,
-      "loss": 1.6521,
+      "learning_rate": 0.00019903083700440526,
+      "loss": 0.109,
       "step": 886
     },
     {
       "epoch": 8.14,
-      "learning_rate": 0.00020035242290748897,
-      "loss": 1.6209,
+      "learning_rate": 0.00019876651982378852,
+      "loss": 0.156,
       "step": 887
     },
     {
       "epoch": 8.15,
-      "learning_rate": 0.00020008810572687223,
-      "loss": 1.5949,
+      "learning_rate": 0.0001985022026431718,
+      "loss": 0.0764,
       "step": 888
     },
     {
       "epoch": 8.16,
-      "learning_rate": 0.00019982378854625552,
-      "loss": 1.6218,
+      "learning_rate": 0.00019823788546255506,
+      "loss": 0.1908,
       "step": 889
     },
     {
       "epoch": 8.17,
-      "learning_rate": 0.00019955947136563875,
-      "loss": 1.7067,
+      "learning_rate": 0.00019797356828193832,
+      "loss": 0.1818,
       "step": 890
     },
     {
       "epoch": 8.17,
-      "learning_rate": 0.000199295154185022,
-      "loss": 1.6497,
+      "learning_rate": 0.00019770925110132155,
+      "loss": 0.2334,
       "step": 891
     },
     {
       "epoch": 8.18,
-      "learning_rate": 0.00019903083700440526,
-      "loss": 1.6171,
+      "learning_rate": 0.0001974449339207048,
+      "loss": 0.1428,
       "step": 892
     },
     {
       "epoch": 8.19,
-      "learning_rate": 0.00019876651982378852,
-      "loss": 1.6179,
+      "learning_rate": 0.00019718061674008807,
+      "loss": 0.1531,
       "step": 893
     },
     {
       "epoch": 8.2,
-      "learning_rate": 0.0001985022026431718,
-      "loss": 1.6303,
+      "learning_rate": 0.00019691629955947135,
+      "loss": 0.1504,
       "step": 894
     },
     {
       "epoch": 8.21,
-      "learning_rate": 0.00019823788546255506,
-      "loss": 1.6474,
+      "learning_rate": 0.0001966519823788546,
+      "loss": 0.1974,
       "step": 895
     },
     {
       "epoch": 8.22,
-      "learning_rate": 0.00019797356828193832,
-      "loss": 1.6022,
+      "learning_rate": 0.00019638766519823787,
+      "loss": 0.1785,
       "step": 896
     },
     {
       "epoch": 8.23,
-      "learning_rate": 0.00019770925110132155,
-      "loss": 1.5697,
+      "learning_rate": 0.00019612334801762113,
+      "loss": 0.1646,
       "step": 897
     },
     {
       "epoch": 8.24,
-      "learning_rate": 0.0001974449339207048,
-      "loss": 1.7455,
+      "learning_rate": 0.00019585903083700436,
+      "loss": 0.1288,
       "step": 898
     },
     {
       "epoch": 8.25,
-      "learning_rate": 0.00019718061674008807,
-      "loss": 1.6656,
+      "learning_rate": 0.00019559471365638765,
+      "loss": 0.1231,
       "step": 899
     },
     {
       "epoch": 8.26,
-      "learning_rate": 0.00019691629955947135,
-      "loss": 1.6243,
+      "learning_rate": 0.0001953303964757709,
+      "loss": 0.1467,
       "step": 900
     },
     {
       "epoch": 8.26,
-      "eval_loss": 0.5664854645729065,
-      "eval_runtime": 21.1139,
-      "eval_samples_per_second": 78.006,
-      "eval_steps_per_second": 1.231,
-      "eval_wer": 0.5923807578388316,
+      "eval_loss": 0.4279836416244507,
+      "eval_runtime": 23.3682,
+      "eval_samples_per_second": 70.48,
+      "eval_steps_per_second": 1.113,
+      "eval_wer": 0.42712695332448164,
       "step": 900
     },
     {
       "epoch": 8.27,
-      "learning_rate": 0.0001966519823788546,
-      "loss": 1.5774,
+      "learning_rate": 0.00019506607929515416,
+      "loss": 0.146,
       "step": 901
     },
     {
       "epoch": 8.28,
-      "learning_rate": 0.00019638766519823787,
-      "loss": 1.6349,
+      "learning_rate": 0.00019480176211453742,
+      "loss": 0.1686,
       "step": 902
     },
     {
       "epoch": 8.28,
-      "learning_rate": 0.00019612334801762113,
-      "loss": 1.6093,
+      "learning_rate": 0.0001945374449339207,
+      "loss": 0.2508,
       "step": 903
     },
     {
       "epoch": 8.29,
-      "learning_rate": 0.00019585903083700436,
-      "loss": 1.7645,
+      "learning_rate": 0.00019427312775330396,
+      "loss": 0.1797,
       "step": 904
     },
     {
       "epoch": 8.3,
-      "learning_rate": 0.00019559471365638765,
-      "loss": 1.6351,
+      "learning_rate": 0.00019400881057268722,
+      "loss": 0.1325,
       "step": 905
     },
     {
       "epoch": 8.31,
-      "learning_rate": 0.0001953303964757709,
-      "loss": 1.6893,
+      "learning_rate": 0.00019374449339207045,
+      "loss": 0.2011,
       "step": 906
     },
     {
       "epoch": 8.32,
-      "learning_rate": 0.00019506607929515416,
-      "loss": 1.5826,
+      "learning_rate": 0.0001934801762114537,
+      "loss": 0.1362,
       "step": 907
     },
     {
       "epoch": 8.33,
-      "learning_rate": 0.00019480176211453742,
-      "loss": 1.5624,
+      "learning_rate": 0.000193215859030837,
+      "loss": 0.1004,
       "step": 908
     },
     {
       "epoch": 8.34,
-      "learning_rate": 0.0001945374449339207,
-      "loss": 1.5514,
+      "learning_rate": 0.00019295154185022025,
+      "loss": 0.1455,
       "step": 909
     },
     {
       "epoch": 8.35,
-      "learning_rate": 0.00019427312775330396,
-      "loss": 1.7248,
+      "learning_rate": 0.0001926872246696035,
+      "loss": 0.2208,
       "step": 910
     },
     {
       "epoch": 8.36,
-      "learning_rate": 0.00019400881057268722,
-      "loss": 1.625,
+      "learning_rate": 0.00019242290748898677,
+      "loss": 0.1366,
       "step": 911
     },
     {
       "epoch": 8.37,
-      "learning_rate": 0.00019374449339207045,
-      "loss": 1.6302,
+      "learning_rate": 0.00019215859030837003,
+      "loss": 0.1142,
       "step": 912
     },
     {
       "epoch": 8.38,
-      "learning_rate": 0.0001934801762114537,
-      "loss": 1.6895,
+      "learning_rate": 0.00019189427312775329,
+      "loss": 0.1532,
       "step": 913
     },
     {
       "epoch": 8.39,
-      "learning_rate": 0.000193215859030837,
-      "loss": 1.6782,
+      "learning_rate": 0.00019162995594713654,
+      "loss": 0.1615,
       "step": 914
     },
     {
       "epoch": 8.39,
-      "learning_rate": 0.00019295154185022025,
-      "loss": 1.6153,
+      "learning_rate": 0.0001913656387665198,
+      "loss": 0.1826,
       "step": 915
     },
     {
       "epoch": 8.4,
-      "learning_rate": 0.0001926872246696035,
-      "loss": 1.5905,
+      "learning_rate": 0.00019110132158590306,
+      "loss": 0.179,
       "step": 916
     },
     {
       "epoch": 8.41,
-      "learning_rate": 0.00019242290748898677,
-      "loss": 1.7377,
+      "learning_rate": 0.00019083700440528632,
+      "loss": 0.1497,
       "step": 917
     },
     {
       "epoch": 8.42,
-      "learning_rate": 0.00019215859030837003,
-      "loss": 1.6076,
+      "learning_rate": 0.0001905726872246696,
+      "loss": 0.1173,
       "step": 918
     },
     {
       "epoch": 8.43,
-      "learning_rate": 0.00019189427312775329,
-      "loss": 1.6388,
+      "learning_rate": 0.00019030837004405286,
+      "loss": 0.1372,
       "step": 919
     },
     {
       "epoch": 8.44,
-      "learning_rate": 0.00019162995594713654,
-      "loss": 1.6896,
+      "learning_rate": 0.0001900440528634361,
+      "loss": 0.1266,
       "step": 920
     },
     {
       "epoch": 8.45,
-      "learning_rate": 0.0001913656387665198,
-      "loss": 1.6318,
+      "learning_rate": 0.00018977973568281935,
+      "loss": 0.1337,
       "step": 921
     },
     {
       "epoch": 8.46,
-      "learning_rate": 0.00019110132158590306,
-      "loss": 1.6079,
+      "learning_rate": 0.0001895154185022026,
+      "loss": 0.2144,
       "step": 922
     },
     {
       "epoch": 8.47,
-      "learning_rate": 0.00019083700440528632,
-      "loss": 1.6639,
+      "learning_rate": 0.0001892511013215859,
+      "loss": 0.1001,
       "step": 923
     },
     {
       "epoch": 8.48,
-      "learning_rate": 0.0001905726872246696,
-      "loss": 1.6478,
+      "learning_rate": 0.00018898678414096915,
+      "loss": 0.1445,
       "step": 924
     },
     {
       "epoch": 8.49,
-      "learning_rate": 0.00019030837004405286,
-      "loss": 1.6415,
+      "learning_rate": 0.0001887224669603524,
+      "loss": 0.1624,
       "step": 925
     },
     {
       "epoch": 8.5,
-      "learning_rate": 0.0001900440528634361,
-      "loss": 1.6518,
+      "learning_rate": 0.00018845814977973567,
+      "loss": 0.0907,
       "step": 926
     },
     {
       "epoch": 8.5,
-      "learning_rate": 0.00018977973568281935,
-      "loss": 1.5873,
+      "learning_rate": 0.0001881938325991189,
+      "loss": 0.1221,
       "step": 927
     },
     {
       "epoch": 8.51,
-      "learning_rate": 0.0001895154185022026,
-      "loss": 1.5699,
+      "learning_rate": 0.00018792951541850219,
+      "loss": 0.1408,
       "step": 928
     },
     {
       "epoch": 8.52,
-      "learning_rate": 0.0001892511013215859,
-      "loss": 1.9128,
+      "learning_rate": 0.00018766519823788544,
+      "loss": 0.0703,
       "step": 929
     },
     {
       "epoch": 8.53,
-      "learning_rate": 0.00018898678414096915,
-      "loss": 1.6879,
+      "learning_rate": 0.0001874008810572687,
+      "loss": 0.1218,
       "step": 930
     },
     {
       "epoch": 8.54,
-      "learning_rate": 0.0001887224669603524,
-      "loss": 1.6297,
+      "learning_rate": 0.00018713656387665196,
+      "loss": 0.1388,
       "step": 931
     },
     {
       "epoch": 8.55,
-      "learning_rate": 0.00018845814977973567,
-      "loss": 1.6972,
+      "learning_rate": 0.00018687224669603525,
+      "loss": 0.1538,
       "step": 932
     },
     {
       "epoch": 8.56,
-      "learning_rate": 0.0001881938325991189,
-      "loss": 1.6173,
+      "learning_rate": 0.0001866079295154185,
+      "loss": 0.102,
       "step": 933
     },
     {
       "epoch": 8.57,
-      "learning_rate": 0.00018792951541850219,
-      "loss": 1.6243,
+      "learning_rate": 0.00018634361233480176,
+      "loss": 0.192,
       "step": 934
     },
     {
       "epoch": 8.58,
-      "learning_rate": 0.00018766519823788544,
-      "loss": 1.6772,
+      "learning_rate": 0.000186079295154185,
+      "loss": 0.2181,
       "step": 935
     },
     {
       "epoch": 8.59,
-      "learning_rate": 0.0001874008810572687,
-      "loss": 1.6862,
+      "learning_rate": 0.00018581497797356825,
+      "loss": 0.1124,
       "step": 936
     },
     {
       "epoch": 8.6,
-      "learning_rate": 0.00018713656387665196,
-      "loss": 1.6254,
+      "learning_rate": 0.0001855506607929515,
+      "loss": 0.1091,
       "step": 937
     },
     {
       "epoch": 8.61,
-      "learning_rate": 0.00018687224669603525,
-      "loss": 1.6108,
+      "learning_rate": 0.0001852863436123348,
+      "loss": 0.1319,
       "step": 938
     },
     {
       "epoch": 8.61,
-      "learning_rate": 0.0001866079295154185,
-      "loss": 1.617,
+      "learning_rate": 0.00018502202643171805,
+      "loss": 0.1344,
       "step": 939
     },
     {
       "epoch": 8.62,
-      "learning_rate": 0.00018634361233480176,
-      "loss": 1.6502,
+      "learning_rate": 0.0001847577092511013,
+      "loss": 0.2203,
       "step": 940
     },
     {
       "epoch": 8.63,
-      "learning_rate": 0.000186079295154185,
-      "loss": 1.8229,
+      "learning_rate": 0.00018449339207048457,
+      "loss": 0.1618,
       "step": 941
     },
     {
       "epoch": 8.64,
-      "learning_rate": 0.00018581497797356825,
-      "loss": 1.6594,
+      "learning_rate": 0.0001842290748898678,
+      "loss": 0.1264,
       "step": 942
     },
     {
       "epoch": 8.65,
-      "learning_rate": 0.0001855506607929515,
-      "loss": 1.6047,
+      "learning_rate": 0.00018396475770925108,
+      "loss": 0.1237,
       "step": 943
     },
     {
       "epoch": 8.66,
-      "learning_rate": 0.0001852863436123348,
-      "loss": 1.5735,
+      "learning_rate": 0.00018370044052863434,
+      "loss": 0.1423,
       "step": 944
     },
     {
       "epoch": 8.67,
-      "learning_rate": 0.00018502202643171805,
-      "loss": 1.6185,
+      "learning_rate": 0.0001834361233480176,
+      "loss": 0.1398,
       "step": 945
     },
     {
       "epoch": 8.68,
-      "learning_rate": 0.0001847577092511013,
-      "loss": 1.5985,
+      "learning_rate": 0.00018317180616740086,
+      "loss": 0.1445,
       "step": 946
     },
     {
       "epoch": 8.69,
-      "learning_rate": 0.00018449339207048457,
-      "loss": 1.6531,
+      "learning_rate": 0.00018290748898678414,
+      "loss": 0.2474,
       "step": 947
     },
     {
       "epoch": 8.7,
-      "learning_rate": 0.0001842290748898678,
-      "loss": 1.6693,
+      "learning_rate": 0.0001826431718061674,
+      "loss": 0.129,
       "step": 948
     },
     {
       "epoch": 8.71,
-      "learning_rate": 0.00018396475770925108,
-      "loss": 1.6113,
+      "learning_rate": 0.00018237885462555063,
+      "loss": 0.115,
       "step": 949
     },
     {
       "epoch": 8.72,
-      "learning_rate": 0.00018370044052863434,
-      "loss": 1.6527,
+      "learning_rate": 0.0001821145374449339,
+      "loss": 0.1407,
       "step": 950
     },
     {
       "epoch": 8.72,
-      "learning_rate": 0.0001834361233480176,
-      "loss": 1.6604,
+      "learning_rate": 0.00018185022026431715,
+      "loss": 0.1804,
       "step": 951
     },
     {
       "epoch": 8.73,
-      "learning_rate": 0.00018317180616740086,
-      "loss": 1.6552,
+      "learning_rate": 0.00018158590308370044,
+      "loss": 0.1579,
       "step": 952
     },
     {
       "epoch": 8.74,
-      "learning_rate": 0.00018290748898678414,
-      "loss": 1.6615,
+      "learning_rate": 0.0001813215859030837,
+      "loss": 0.1809,
       "step": 953
     },
     {
       "epoch": 8.75,
-      "learning_rate": 0.0001826431718061674,
-      "loss": 1.8437,
+      "learning_rate": 0.00018105726872246695,
+      "loss": 0.1342,
       "step": 954
     },
     {
       "epoch": 8.76,
-      "learning_rate": 0.00018237885462555063,
-      "loss": 1.6606,
+      "learning_rate": 0.0001807929515418502,
+      "loss": 0.1326,
       "step": 955
     },
     {
       "epoch": 8.77,
-      "learning_rate": 0.0001821145374449339,
-      "loss": 1.6069,
+      "learning_rate": 0.00018052863436123344,
+      "loss": 0.0835,
       "step": 956
     },
     {
       "epoch": 8.78,
-      "learning_rate": 0.00018185022026431715,
-      "loss": 1.5687,
+      "learning_rate": 0.00018026431718061673,
+      "loss": 0.1275,
       "step": 957
     },
     {
       "epoch": 8.79,
-      "learning_rate": 0.00018158590308370044,
-      "loss": 1.6024,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 0.206,
       "step": 958
     },
     {
       "epoch": 8.8,
-      "learning_rate": 0.0001813215859030837,
-      "loss": 1.6132,
+      "learning_rate": 0.00017973568281938324,
+      "loss": 0.227,
       "step": 959
     },
     {
       "epoch": 8.81,
-      "learning_rate": 0.00018105726872246695,
-      "loss": 1.6314,
+      "learning_rate": 0.0001794713656387665,
+      "loss": 0.2326,
       "step": 960
     },
     {
       "epoch": 8.82,
-      "learning_rate": 0.0001807929515418502,
-      "loss": 1.6969,
+      "learning_rate": 0.00017920704845814976,
+      "loss": 0.1439,
       "step": 961
     },
     {
       "epoch": 8.83,
-      "learning_rate": 0.00018052863436123344,
-      "loss": 1.657,
+      "learning_rate": 0.00017894273127753304,
+      "loss": 0.1149,
       "step": 962
     },
     {
       "epoch": 8.83,
-      "learning_rate": 0.00018026431718061673,
-      "loss": 1.6352,
+      "learning_rate": 0.0001786784140969163,
+      "loss": 0.1489,
       "step": 963
     },
     {
       "epoch": 8.84,
-      "learning_rate": 0.00017999999999999998,
-      "loss": 1.5957,
+      "learning_rate": 0.00017841409691629953,
+      "loss": 0.167,
       "step": 964
     },
     {
       "epoch": 8.85,
-      "learning_rate": 0.00017973568281938324,
-      "loss": 1.6209,
+      "learning_rate": 0.0001781497797356828,
+      "loss": 0.2292,
       "step": 965
     },
     {
       "epoch": 8.86,
-      "learning_rate": 0.0001794713656387665,
-      "loss": 1.696,
+      "learning_rate": 0.00017788546255506605,
+      "loss": 0.167,
       "step": 966
     },
     {
       "epoch": 8.87,
-      "learning_rate": 0.00017920704845814976,
-      "loss": 1.6295,
+      "learning_rate": 0.00017762114537444933,
+      "loss": 0.1512,
       "step": 967
     },
     {
       "epoch": 8.88,
-      "learning_rate": 0.00017894273127753304,
-      "loss": 1.5429,
+      "learning_rate": 0.0001773568281938326,
+      "loss": 0.1023,
       "step": 968
     },
     {
       "epoch": 8.89,
-      "learning_rate": 0.0001786784140969163,
-      "loss": 1.6564,
+      "learning_rate": 0.00017709251101321585,
+      "loss": 0.1232,
       "step": 969
     },
     {
       "epoch": 8.9,
-      "learning_rate": 0.00017841409691629953,
-      "loss": 1.6178,
+      "learning_rate": 0.0001768281938325991,
+      "loss": 0.1245,
       "step": 970
     },
     {
       "epoch": 8.91,
-      "learning_rate": 0.0001781497797356828,
-      "loss": 1.6342,
+      "learning_rate": 0.00017656387665198234,
+      "loss": 0.1469,
       "step": 971
     },
     {
       "epoch": 8.92,
-      "learning_rate": 0.00017788546255506605,
-      "loss": 1.5766,
+      "learning_rate": 0.00017629955947136563,
+      "loss": 0.1341,
       "step": 972
     },
     {
       "epoch": 8.93,
-      "learning_rate": 0.00017762114537444933,
-      "loss": 1.666,
+      "learning_rate": 0.00017603524229074888,
+      "loss": 0.0992,
       "step": 973
     },
     {
       "epoch": 8.94,
-      "learning_rate": 0.0001773568281938326,
-      "loss": 1.6663,
+      "learning_rate": 0.00017577092511013214,
+      "loss": 0.1408,
       "step": 974
     },
     {
       "epoch": 8.94,
-      "learning_rate": 0.00017709251101321585,
-      "loss": 1.6561,
+      "learning_rate": 0.0001755066079295154,
+      "loss": 0.1319,
       "step": 975
     },
     {
       "epoch": 8.95,
-      "learning_rate": 0.0001768281938325991,
-      "loss": 1.6486,
+      "learning_rate": 0.00017524229074889869,
+      "loss": 0.1218,
       "step": 976
     },
     {
       "epoch": 8.96,
-      "learning_rate": 0.00017656387665198234,
-      "loss": 1.6171,
+      "learning_rate": 0.00017497797356828194,
+      "loss": 0.1517,
       "step": 977
     },
     {
       "epoch": 8.97,
-      "learning_rate": 0.00017629955947136563,
-      "loss": 1.583,
+      "learning_rate": 0.00017471365638766517,
+      "loss": 0.117,
       "step": 978
     },
     {
       "epoch": 8.98,
-      "learning_rate": 0.00017603524229074888,
-      "loss": 1.6525,
+      "learning_rate": 0.00017444933920704843,
+      "loss": 0.2165,
       "step": 979
     },
     {
       "epoch": 8.99,
-      "learning_rate": 0.00017577092511013214,
-      "loss": 1.5893,
+      "learning_rate": 0.0001741850220264317,
+      "loss": 0.1642,
       "step": 980
     },
     {
       "epoch": 9.0,
-      "learning_rate": 0.0001755066079295154,
-      "loss": 2.003,
+      "learning_rate": 0.00017392070484581495,
+      "loss": 0.1552,
       "step": 981
     },
     {
       "epoch": 9.01,
-      "learning_rate": 0.00017524229074889869,
-      "loss": 1.7363,
+      "learning_rate": 0.00017365638766519823,
+      "loss": 0.1073,
       "step": 982
     },
     {
       "epoch": 9.02,
-      "learning_rate": 0.00017497797356828194,
-      "loss": 1.6075,
+      "learning_rate": 0.0001733920704845815,
+      "loss": 0.1445,
       "step": 983
     },
     {
       "epoch": 9.03,
-      "learning_rate": 0.00017471365638766517,
-      "loss": 1.529,
+      "learning_rate": 0.00017312775330396475,
+      "loss": 0.0925,
       "step": 984
     },
     {
       "epoch": 9.04,
-      "learning_rate": 0.00017444933920704843,
-      "loss": 1.5293,
+      "learning_rate": 0.00017286343612334798,
+      "loss": 0.1236,
       "step": 985
     },
     {
       "epoch": 9.05,
-      "learning_rate": 0.0001741850220264317,
-      "loss": 1.6018,
+      "learning_rate": 0.00017259911894273124,
+      "loss": 0.1546,
       "step": 986
     },
     {
       "epoch": 9.06,
-      "learning_rate": 0.00017392070484581495,
-      "loss": 1.5752,
+      "learning_rate": 0.00017233480176211452,
+      "loss": 0.151,
       "step": 987
     },
     {
       "epoch": 9.06,
-      "learning_rate": 0.00017365638766519823,
-      "loss": 1.8495,
+      "learning_rate": 0.00017207048458149778,
+      "loss": 0.1273,
       "step": 988
     },
     {
       "epoch": 9.07,
-      "learning_rate": 0.0001733920704845815,
-      "loss": 1.6298,
+      "learning_rate": 0.00017180616740088104,
+      "loss": 0.1166,
       "step": 989
     },
     {
       "epoch": 9.08,
-      "learning_rate": 0.00017312775330396475,
-      "loss": 1.5699,
+      "learning_rate": 0.0001715418502202643,
+      "loss": 0.0858,
       "step": 990
     },
     {
       "epoch": 9.09,
-      "learning_rate": 0.00017286343612334798,
-      "loss": 1.5898,
+      "learning_rate": 0.00017127753303964758,
+      "loss": 0.1049,
       "step": 991
     },
     {
       "epoch": 9.1,
-      "learning_rate": 0.00017259911894273124,
-      "loss": 1.6082,
+      "learning_rate": 0.00017101321585903082,
+      "loss": 0.1177,
       "step": 992
     },
     {
       "epoch": 9.11,
-      "learning_rate": 0.00017233480176211452,
-      "loss": 1.5693,
+      "learning_rate": 0.00017074889867841407,
+      "loss": 0.1521,
       "step": 993
     },
     {
       "epoch": 9.12,
-      "learning_rate": 0.00017207048458149778,
-      "loss": 1.6238,
+      "learning_rate": 0.00017048458149779733,
+      "loss": 0.178,
       "step": 994
     },
     {
       "epoch": 9.13,
-      "learning_rate": 0.00017180616740088104,
-      "loss": 1.5579,
+      "learning_rate": 0.0001702202643171806,
+      "loss": 0.0872,
       "step": 995
     },
     {
       "epoch": 9.14,
-      "learning_rate": 0.0001715418502202643,
-      "loss": 1.5698,
+      "learning_rate": 0.00016995594713656388,
+      "loss": 0.1187,
       "step": 996
     },
     {
       "epoch": 9.15,
-      "learning_rate": 0.00017127753303964758,
-      "loss": 1.5861,
+      "learning_rate": 0.00016969162995594713,
+      "loss": 0.129,
       "step": 997
     },
     {
       "epoch": 9.16,
-      "learning_rate": 0.00017101321585903082,
-      "loss": 1.6554,
+      "learning_rate": 0.0001694273127753304,
+      "loss": 0.141,
       "step": 998
     },
     {
       "epoch": 9.17,
-      "learning_rate": 0.00017074889867841407,
-      "loss": 1.5908,
+      "learning_rate": 0.00016916299559471365,
+      "loss": 0.1063,
       "step": 999
     },
     {
       "epoch": 9.17,
-      "learning_rate": 0.00017048458149779733,
-      "loss": 1.6852,
+      "learning_rate": 0.00016889867841409688,
+      "loss": 0.1898,
       "step": 1000
     },
     {
       "epoch": 9.17,
-      "eval_loss": 0.5044031739234924,
-      "eval_runtime": 21.0229,
-      "eval_samples_per_second": 78.343,
-      "eval_steps_per_second": 1.237,
-      "eval_wer": 0.5258911245020937,
+      "eval_loss": 0.4175008535385132,
+      "eval_runtime": 21.0086,
+      "eval_samples_per_second": 78.396,
+      "eval_steps_per_second": 1.238,
+      "eval_wer": 0.40332958839750793,
       "step": 1000
     },
     {
       "epoch": 9.18,
-      "learning_rate": 0.0001702202643171806,
-      "loss": 1.6246,
+      "learning_rate": 0.00016863436123348017,
+      "loss": 0.0936,
       "step": 1001
     },
     {
       "epoch": 9.19,
-      "learning_rate": 0.00016995594713656388,
-      "loss": 1.5697,
+      "learning_rate": 0.00016837004405286342,
+      "loss": 0.1079,
       "step": 1002
     },
     {
       "epoch": 9.2,
-      "learning_rate": 0.00016969162995594713,
-      "loss": 1.5659,
+      "learning_rate": 0.00016810572687224668,
+      "loss": 0.1115,
       "step": 1003
     },
     {
       "epoch": 9.21,
-      "learning_rate": 0.0001694273127753304,
-      "loss": 1.581,
+      "learning_rate": 0.00016784140969162994,
+      "loss": 0.1477,
       "step": 1004
     },
     {
       "epoch": 9.22,
-      "learning_rate": 0.00016916299559471365,
-      "loss": 1.5852,
+      "learning_rate": 0.0001675770925110132,
+      "loss": 0.1056,
       "step": 1005
     },
     {
       "epoch": 9.23,
-      "learning_rate": 0.00016889867841409688,
-      "loss": 1.5913,
+      "learning_rate": 0.00016731277533039648,
+      "loss": 0.19,
       "step": 1006
     },
     {
       "epoch": 9.24,
-      "learning_rate": 0.00016863436123348017,
-      "loss": 1.6227,
+      "learning_rate": 0.00016704845814977971,
+      "loss": 0.0954,
       "step": 1007
     },
     {
       "epoch": 9.25,
-      "learning_rate": 0.00016837004405286342,
-      "loss": 1.5573,
+      "learning_rate": 0.00016678414096916297,
+      "loss": 0.1186,
       "step": 1008
     },
     {
       "epoch": 9.26,
-      "learning_rate": 0.00016810572687224668,
-      "loss": 1.578,
+      "learning_rate": 0.00016651982378854623,
+      "loss": 0.1029,
       "step": 1009
     },
     {
       "epoch": 9.27,
-      "learning_rate": 0.00016784140969162994,
-      "loss": 1.5683,
+      "learning_rate": 0.0001662555066079295,
+      "loss": 0.1568,
       "step": 1010
     },
     {
       "epoch": 9.28,
-      "learning_rate": 0.0001675770925110132,
-      "loss": 1.5681,
+      "learning_rate": 0.00016599118942731277,
+      "loss": 0.1359,
       "step": 1011
     },
     {
       "epoch": 9.28,
-      "learning_rate": 0.00016731277533039648,
-      "loss": 1.6919,
+      "learning_rate": 0.00016572687224669603,
+      "loss": 0.1774,
       "step": 1012
     },
     {
       "epoch": 9.29,
-      "learning_rate": 0.00016704845814977971,
-      "loss": 1.6148,
+      "learning_rate": 0.0001654625550660793,
+      "loss": 0.1233,
       "step": 1013
     },
     {
       "epoch": 9.3,
-      "learning_rate": 0.00016678414096916297,
-      "loss": 1.5858,
+      "learning_rate": 0.00016519823788546252,
+      "loss": 0.1182,
       "step": 1014
     },
     {
       "epoch": 9.31,
-      "learning_rate": 0.00016651982378854623,
-      "loss": 1.5836,
+      "learning_rate": 0.00016493392070484578,
+      "loss": 0.1272,
       "step": 1015
     },
     {
       "epoch": 9.32,
-      "learning_rate": 0.0001662555066079295,
-      "loss": 1.5601,
+      "learning_rate": 0.00016466960352422907,
+      "loss": 0.1182,
       "step": 1016
     },
     {
       "epoch": 9.33,
-      "learning_rate": 0.00016599118942731277,
-      "loss": 1.6487,
+      "learning_rate": 0.00016440528634361232,
+      "loss": 0.1077,
       "step": 1017
     },
     {
       "epoch": 9.34,
-      "learning_rate": 0.00016572687224669603,
-      "loss": 1.6194,
+      "learning_rate": 0.00016414096916299558,
+      "loss": 0.1901,
       "step": 1018
     },
     {
       "epoch": 9.35,
-      "learning_rate": 0.0001654625550660793,
-      "loss": 1.5555,
+      "learning_rate": 0.00016387665198237884,
+      "loss": 0.1208,
       "step": 1019
     },
     {
       "epoch": 9.36,
-      "learning_rate": 0.00016519823788546252,
-      "loss": 1.6004,
+      "learning_rate": 0.00016361233480176212,
+      "loss": 0.097,
       "step": 1020
     },
     {
       "epoch": 9.37,
-      "learning_rate": 0.00016493392070484578,
-      "loss": 1.5824,
+      "learning_rate": 0.00016334801762114536,
+      "loss": 0.116,
       "step": 1021
     },
     {
       "epoch": 9.38,
-      "learning_rate": 0.00016466960352422907,
-      "loss": 1.5762,
+      "learning_rate": 0.00016308370044052861,
+      "loss": 0.1512,
       "step": 1022
     },
     {
       "epoch": 9.39,
-      "learning_rate": 0.00016440528634361232,
-      "loss": 1.5608,
+      "learning_rate": 0.00016281938325991187,
+      "loss": 0.138,
       "step": 1023
     },
     {
       "epoch": 9.39,
-      "learning_rate": 0.00016414096916299558,
-      "loss": 1.5811,
+      "learning_rate": 0.00016255506607929513,
+      "loss": 0.1542,
       "step": 1024
     },
     {
       "epoch": 9.4,
-      "learning_rate": 0.00016387665198237884,
-      "loss": 1.6083,
+      "learning_rate": 0.00016229074889867842,
+      "loss": 0.1023,
       "step": 1025
     },
     {
       "epoch": 9.41,
-      "learning_rate": 0.00016361233480176212,
-      "loss": 1.6198,
+      "learning_rate": 0.00016202643171806167,
+      "loss": 0.1271,
       "step": 1026
     },
     {
       "epoch": 9.42,
-      "learning_rate": 0.00016334801762114536,
-      "loss": 1.6063,
+      "learning_rate": 0.00016176211453744493,
+      "loss": 0.11,
       "step": 1027
     },
     {
       "epoch": 9.43,
-      "learning_rate": 0.00016308370044052861,
-      "loss": 1.567,
+      "learning_rate": 0.0001614977973568282,
+      "loss": 0.1337,
       "step": 1028
     },
     {
       "epoch": 9.44,
-      "learning_rate": 0.00016281938325991187,
-      "loss": 1.6433,
+      "learning_rate": 0.00016123348017621142,
+      "loss": 0.1571,
       "step": 1029
     },
     {
       "epoch": 9.45,
-      "learning_rate": 0.00016255506607929513,
-      "loss": 1.5944,
+      "learning_rate": 0.00016096916299559468,
+      "loss": 0.2177,
       "step": 1030
     },
     {
       "epoch": 9.46,
-      "learning_rate": 0.00016229074889867842,
-      "loss": 1.5009,
+      "learning_rate": 0.00016070484581497796,
+      "loss": 0.109,
       "step": 1031
     },
     {
       "epoch": 9.47,
-      "learning_rate": 0.00016202643171806167,
-      "loss": 1.611,
+      "learning_rate": 0.00016044052863436122,
+      "loss": 0.1037,
       "step": 1032
     },
     {
       "epoch": 9.48,
-      "learning_rate": 0.00016176211453744493,
-      "loss": 1.5432,
+      "learning_rate": 0.00016017621145374448,
+      "loss": 0.1184,
       "step": 1033
     },
     {
       "epoch": 9.49,
-      "learning_rate": 0.0001614977973568282,
-      "loss": 1.5492,
+      "learning_rate": 0.00015991189427312774,
+      "loss": 0.0989,
       "step": 1034
     },
     {
       "epoch": 9.5,
-      "learning_rate": 0.00016123348017621142,
-      "loss": 1.6536,
+      "learning_rate": 0.00015964757709251102,
+      "loss": 0.0992,
       "step": 1035
     },
     {
       "epoch": 9.5,
-      "learning_rate": 0.00016096916299559468,
-      "loss": 1.5603,
+      "learning_rate": 0.00015938325991189425,
+      "loss": 0.1249,
       "step": 1036
     },
     {
       "epoch": 9.51,
-      "learning_rate": 0.00016070484581497796,
-      "loss": 1.5582,
+      "learning_rate": 0.0001591189427312775,
+      "loss": 0.1107,
       "step": 1037
     },
     {
       "epoch": 9.52,
-      "learning_rate": 0.00016044052863436122,
-      "loss": 1.6228,
+      "learning_rate": 0.00015885462555066077,
+      "loss": 0.0996,
       "step": 1038
     },
     {
       "epoch": 9.53,
-      "learning_rate": 0.00016017621145374448,
-      "loss": 1.5943,
+      "learning_rate": 0.00015859030837004403,
+      "loss": 0.0809,
       "step": 1039
     },
     {
       "epoch": 9.54,
-      "learning_rate": 0.00015991189427312774,
-      "loss": 1.5878,
+      "learning_rate": 0.00015832599118942731,
+      "loss": 0.1099,
       "step": 1040
     },
     {
       "epoch": 9.55,
-      "learning_rate": 0.00015964757709251102,
-      "loss": 1.5324,
+      "learning_rate": 0.00015806167400881057,
+      "loss": 0.1211,
       "step": 1041
     },
     {
       "epoch": 9.56,
-      "learning_rate": 0.00015938325991189425,
-      "loss": 1.5536,
+      "learning_rate": 0.00015779735682819383,
+      "loss": 0.1359,
       "step": 1042
     },
     {
       "epoch": 9.57,
-      "learning_rate": 0.0001591189427312775,
-      "loss": 1.5317,
+      "learning_rate": 0.00015753303964757706,
+      "loss": 0.1824,
       "step": 1043
     },
     {
       "epoch": 9.58,
-      "learning_rate": 0.00015885462555066077,
-      "loss": 1.5843,
+      "learning_rate": 0.00015726872246696032,
+      "loss": 0.1613,
       "step": 1044
     },
     {
       "epoch": 9.59,
-      "learning_rate": 0.00015859030837004403,
-      "loss": 1.6868,
+      "learning_rate": 0.0001570044052863436,
+      "loss": 0.1307,
       "step": 1045
     },
     {
       "epoch": 9.6,
-      "learning_rate": 0.00015832599118942731,
-      "loss": 1.6115,
+      "learning_rate": 0.00015674008810572686,
+      "loss": 0.1238,
       "step": 1046
     },
     {
       "epoch": 9.61,
-      "learning_rate": 0.00015806167400881057,
-      "loss": 1.5433,
+      "learning_rate": 0.00015647577092511012,
+      "loss": 0.0608,
       "step": 1047
     },
     {
       "epoch": 9.61,
-      "learning_rate": 0.00015779735682819383,
-      "loss": 1.6344,
+      "learning_rate": 0.00015621145374449338,
+      "loss": 0.1266,
       "step": 1048
     },
     {
       "epoch": 9.62,
-      "learning_rate": 0.00015753303964757706,
-      "loss": 1.5386,
+      "learning_rate": 0.00015594713656387664,
+      "loss": 0.1303,
       "step": 1049
     },
     {
       "epoch": 9.63,
-      "learning_rate": 0.00015726872246696032,
-      "loss": 1.6277,
+      "learning_rate": 0.00015568281938325987,
+      "loss": 0.2192,
       "step": 1050
     },
     {
       "epoch": 9.64,
-      "learning_rate": 0.0001570044052863436,
-      "loss": 1.6116,
+      "learning_rate": 0.00015541850220264315,
+      "loss": 0.0966,
       "step": 1051
     },
     {
       "epoch": 9.65,
-      "learning_rate": 0.00015674008810572686,
-      "loss": 1.5838,
+      "learning_rate": 0.0001551541850220264,
+      "loss": 0.0697,
       "step": 1052
     },
     {
       "epoch": 9.66,
-      "learning_rate": 0.00015647577092511012,
-      "loss": 1.5311,
+      "learning_rate": 0.00015488986784140967,
+      "loss": 0.0606,
       "step": 1053
     },
     {
       "epoch": 9.67,
-      "learning_rate": 0.00015621145374449338,
-      "loss": 1.5433,
+      "learning_rate": 0.00015462555066079293,
+      "loss": 0.1334,
       "step": 1054
     },
     {
       "epoch": 9.68,
-      "learning_rate": 0.00015594713656387664,
-      "loss": 1.5621,
+      "learning_rate": 0.00015436123348017621,
+      "loss": 0.1269,
       "step": 1055
     },
     {
       "epoch": 9.69,
-      "learning_rate": 0.00015568281938325987,
-      "loss": 1.5606,
+      "learning_rate": 0.00015409691629955947,
+      "loss": 0.1558,
       "step": 1056
     },
     {
       "epoch": 9.7,
-      "learning_rate": 0.00015541850220264315,
-      "loss": 1.596,
+      "learning_rate": 0.00015383259911894273,
+      "loss": 0.0826,
       "step": 1057
     },
     {
       "epoch": 9.71,
-      "learning_rate": 0.0001551541850220264,
-      "loss": 1.5365,
+      "learning_rate": 0.00015356828193832596,
+      "loss": 0.1122,
       "step": 1058
     },
     {
       "epoch": 9.72,
-      "learning_rate": 0.00015488986784140967,
-      "loss": 1.5808,
+      "learning_rate": 0.00015330396475770922,
+      "loss": 0.164,
       "step": 1059
     },
     {
       "epoch": 9.72,
-      "learning_rate": 0.00015462555066079293,
-      "loss": 1.556,
+      "learning_rate": 0.0001530396475770925,
+      "loss": 0.0995,
       "step": 1060
     },
     {
       "epoch": 9.73,
-      "learning_rate": 0.00015436123348017621,
-      "loss": 1.5998,
+      "learning_rate": 0.00015277533039647576,
+      "loss": 0.176,
       "step": 1061
     },
     {
       "epoch": 9.74,
-      "learning_rate": 0.00015409691629955947,
-      "loss": 1.6132,
+      "learning_rate": 0.00015251101321585902,
+      "loss": 0.1804,
       "step": 1062
     },
     {
       "epoch": 9.75,
-      "learning_rate": 0.00015383259911894273,
-      "loss": 1.6509,
+      "learning_rate": 0.00015224669603524228,
+      "loss": 0.1517,
       "step": 1063
     },
     {
       "epoch": 9.76,
-      "learning_rate": 0.00015356828193832596,
-      "loss": 1.5865,
+      "learning_rate": 0.00015198237885462556,
+      "loss": 0.0847,
       "step": 1064
     },
     {
       "epoch": 9.77,
-      "learning_rate": 0.00015330396475770922,
-      "loss": 1.6284,
+      "learning_rate": 0.0001517180616740088,
+      "loss": 0.1293,
       "step": 1065
     },
     {
       "epoch": 9.78,
-      "learning_rate": 0.0001530396475770925,
-      "loss": 1.5832,
+      "learning_rate": 0.00015145374449339205,
+      "loss": 0.1332,
       "step": 1066
     },
     {
       "epoch": 9.79,
-      "learning_rate": 0.00015277533039647576,
-      "loss": 1.5351,
+      "learning_rate": 0.0001511894273127753,
+      "loss": 0.1422,
       "step": 1067
     },
     {
       "epoch": 9.8,
-      "learning_rate": 0.00015251101321585902,
-      "loss": 1.5859,
+      "learning_rate": 0.00015092511013215857,
+      "loss": 0.146,
       "step": 1068
     },
     {
       "epoch": 9.81,
-      "learning_rate": 0.00015224669603524228,
-      "loss": 1.6021,
+      "learning_rate": 0.00015066079295154186,
+      "loss": 0.1195,
       "step": 1069
     },
     {
       "epoch": 9.82,
-      "learning_rate": 0.00015198237885462556,
-      "loss": 1.5758,
+      "learning_rate": 0.0001503964757709251,
+      "loss": 0.0889,
       "step": 1070
     },
     {
       "epoch": 9.83,
-      "learning_rate": 0.0001517180616740088,
-      "loss": 1.5604,
+      "learning_rate": 0.00015013215859030837,
+      "loss": 0.1303,
       "step": 1071
     },
     {
       "epoch": 9.83,
-      "learning_rate": 0.00015145374449339205,
-      "loss": 1.5247,
+      "learning_rate": 0.00014986784140969163,
+      "loss": 0.1619,
       "step": 1072
     },
     {
       "epoch": 9.84,
-      "learning_rate": 0.0001511894273127753,
-      "loss": 1.561,
+      "learning_rate": 0.0001496035242290749,
+      "loss": 0.1817,
       "step": 1073
     },
     {
       "epoch": 9.85,
-      "learning_rate": 0.00015092511013215857,
-      "loss": 1.5541,
+      "learning_rate": 0.00014933920704845812,
+      "loss": 0.1508,
       "step": 1074
     },
     {
       "epoch": 9.86,
-      "learning_rate": 0.00015066079295154186,
-      "loss": 1.6636,
+      "learning_rate": 0.0001490748898678414,
+      "loss": 0.1742,
       "step": 1075
     },
     {
       "epoch": 9.87,
-      "learning_rate": 0.0001503964757709251,
-      "loss": 1.5859,
+      "learning_rate": 0.00014881057268722466,
+      "loss": 0.1021,
       "step": 1076
     },
     {
       "epoch": 9.88,
-      "learning_rate": 0.00015013215859030837,
-      "loss": 1.6205,
+      "learning_rate": 0.00014854625550660792,
+      "loss": 0.0953,
       "step": 1077
     },
     {
       "epoch": 9.89,
-      "learning_rate": 0.00014986784140969163,
-      "loss": 1.5415,
+      "learning_rate": 0.00014828193832599118,
+      "loss": 0.1044,
       "step": 1078
     },
     {
       "epoch": 9.9,
-      "learning_rate": 0.0001496035242290749,
-      "loss": 1.5073,
+      "learning_rate": 0.00014801762114537444,
+      "loss": 0.1086,
       "step": 1079
     },
     {
       "epoch": 9.91,
-      "learning_rate": 0.00014933920704845812,
-      "loss": 1.5491,
+      "learning_rate": 0.0001477533039647577,
+      "loss": 0.163,
       "step": 1080
     },
     {
       "epoch": 9.92,
-      "learning_rate": 0.0001490748898678414,
-      "loss": 1.5948,
+      "learning_rate": 0.00014748898678414095,
+      "loss": 0.1219,
       "step": 1081
     },
     {
       "epoch": 9.93,
-      "learning_rate": 0.00014881057268722466,
-      "loss": 1.6682,
+      "learning_rate": 0.0001472246696035242,
+      "loss": 0.1392,
       "step": 1082
     },
     {
       "epoch": 9.94,
-      "learning_rate": 0.00014854625550660792,
-      "loss": 1.5404,
+      "learning_rate": 0.00014696035242290747,
+      "loss": 0.1,
       "step": 1083
     },
     {
       "epoch": 9.94,
-      "learning_rate": 0.00014828193832599118,
-      "loss": 1.5593,
+      "learning_rate": 0.00014669603524229075,
+      "loss": 0.1296,
       "step": 1084
     },
     {
       "epoch": 9.95,
-      "learning_rate": 0.00014801762114537444,
-      "loss": 1.5701,
+      "learning_rate": 0.00014643171806167399,
+      "loss": 0.1318,
       "step": 1085
     },
     {
       "epoch": 9.96,
-      "learning_rate": 0.0001477533039647577,
-      "loss": 1.6395,
+      "learning_rate": 0.00014616740088105724,
+      "loss": 0.1062,
       "step": 1086
     },
     {
       "epoch": 9.97,
-      "learning_rate": 0.00014748898678414095,
-      "loss": 1.628,
+      "learning_rate": 0.00014590308370044053,
+      "loss": 0.1823,
       "step": 1087
     },
     {
       "epoch": 9.98,
-      "learning_rate": 0.0001472246696035242,
-      "loss": 1.5785,
+      "learning_rate": 0.00014563876651982376,
+      "loss": 0.1352,
       "step": 1088
     },
     {
       "epoch": 9.99,
-      "learning_rate": 0.00014696035242290747,
-      "loss": 1.5428,
+      "learning_rate": 0.00014537444933920705,
+      "loss": 0.1155,
       "step": 1089
     },
     {
       "epoch": 10.0,
-      "learning_rate": 0.00014669603524229075,
-      "loss": 2.0497,
+      "learning_rate": 0.0001451101321585903,
+      "loss": 0.2058,
       "step": 1090
     },
     {
       "epoch": 10.01,
-      "learning_rate": 0.00014643171806167399,
-      "loss": 1.5731,
+      "learning_rate": 0.00014484581497797356,
+      "loss": 0.0672,
       "step": 1091
     },
     {
       "epoch": 10.02,
-      "learning_rate": 0.00014616740088105724,
-      "loss": 1.5251,
+      "learning_rate": 0.00014458149779735682,
+      "loss": 0.1145,
       "step": 1092
     },
     {
       "epoch": 10.03,
-      "learning_rate": 0.00014590308370044053,
-      "loss": 1.5524,
+      "learning_rate": 0.00014431718061674008,
+      "loss": 0.0891,
       "step": 1093
     },
     {
       "epoch": 10.04,
-      "learning_rate": 0.00014563876651982376,
-      "loss": 1.5208,
+      "learning_rate": 0.00014405286343612334,
+      "loss": 0.107,
       "step": 1094
     },
     {
       "epoch": 10.05,
-      "learning_rate": 0.00014537444933920705,
-      "loss": 1.5138,
+      "learning_rate": 0.0001437885462555066,
+      "loss": 0.1782,
       "step": 1095
     },
     {
       "epoch": 10.06,
-      "learning_rate": 0.0001451101321585903,
-      "loss": 1.5124,
+      "learning_rate": 0.00014352422907488985,
+      "loss": 0.1101,
       "step": 1096
     },
     {
       "epoch": 10.06,
-      "learning_rate": 0.00014484581497797356,
-      "loss": 1.6408,
+      "learning_rate": 0.0001432599118942731,
+      "loss": 0.1396,
       "step": 1097
     },
     {
       "epoch": 10.07,
-      "learning_rate": 0.00014458149779735682,
-      "loss": 1.51,
+      "learning_rate": 0.00014299559471365637,
+      "loss": 0.1019,
       "step": 1098
     },
     {
       "epoch": 10.08,
-      "learning_rate": 0.00014431718061674008,
-      "loss": 1.5119,
+      "learning_rate": 0.00014273127753303963,
+      "loss": 0.0881,
       "step": 1099
     },
     {
       "epoch": 10.09,
-      "learning_rate": 0.00014405286343612334,
-      "loss": 1.5237,
+      "learning_rate": 0.00014246696035242288,
+      "loss": 0.089,
       "step": 1100
     },
     {
       "epoch": 10.09,
-      "eval_loss": 0.4720684289932251,
-      "eval_runtime": 20.9681,
-      "eval_samples_per_second": 78.548,
-      "eval_steps_per_second": 1.24,
-      "eval_wer": 0.4916760290062302,
+      "eval_loss": 0.39230093359947205,
+      "eval_runtime": 21.0264,
+      "eval_samples_per_second": 78.33,
+      "eval_steps_per_second": 1.237,
+      "eval_wer": 0.3910734347870493,
       "step": 1100
     },
     {
       "epoch": 10.1,
-      "learning_rate": 0.0001437885462555066,
-      "loss": 1.5285,
+      "learning_rate": 0.00014220264317180617,
+      "loss": 0.105,
       "step": 1101
     },
     {
       "epoch": 10.11,
-      "learning_rate": 0.00014352422907488985,
-      "loss": 1.5283,
+      "learning_rate": 0.00014193832599118943,
+      "loss": 0.1626,
       "step": 1102
     },
     {
       "epoch": 10.12,
-      "learning_rate": 0.0001432599118942731,
-      "loss": 1.5577,
+      "learning_rate": 0.00014167400881057266,
+      "loss": 0.0937,
       "step": 1103
     },
     {
       "epoch": 10.13,
-      "learning_rate": 0.00014299559471365637,
-      "loss": 1.5479,
+      "learning_rate": 0.00014140969162995594,
+      "loss": 0.1065,
       "step": 1104
     },
     {
       "epoch": 10.14,
-      "learning_rate": 0.00014273127753303963,
-      "loss": 1.5162,
+      "learning_rate": 0.0001411453744493392,
+      "loss": 0.1224,
       "step": 1105
     },
     {
       "epoch": 10.15,
-      "learning_rate": 0.00014246696035242288,
-      "loss": 1.5513,
+      "learning_rate": 0.00014088105726872243,
+      "loss": 0.1029,
       "step": 1106
     },
     {
       "epoch": 10.16,
-      "learning_rate": 0.00014220264317180617,
-      "loss": 1.5192,
+      "learning_rate": 0.00014061674008810572,
+      "loss": 0.1696,
       "step": 1107
     },
     {
       "epoch": 10.17,
-      "learning_rate": 0.00014193832599118943,
-      "loss": 1.5322,
+      "learning_rate": 0.00014035242290748898,
+      "loss": 0.1045,
       "step": 1108
     },
     {
       "epoch": 10.17,
-      "learning_rate": 0.00014167400881057266,
-      "loss": 1.5899,
+      "learning_rate": 0.00014008810572687224,
+      "loss": 0.1136,
       "step": 1109
     },
     {
       "epoch": 10.18,
-      "learning_rate": 0.00014140969162995594,
-      "loss": 1.5304,
+      "learning_rate": 0.0001398237885462555,
+      "loss": 0.1154,
       "step": 1110
     },
     {
       "epoch": 10.19,
-      "learning_rate": 0.0001411453744493392,
-      "loss": 1.575,
+      "learning_rate": 0.00013955947136563875,
+      "loss": 0.1182,
       "step": 1111
     },
     {
       "epoch": 10.2,
-      "learning_rate": 0.00014088105726872243,
-      "loss": 1.5226,
+      "learning_rate": 0.000139295154185022,
+      "loss": 0.0919,
       "step": 1112
     },
     {
       "epoch": 10.21,
-      "learning_rate": 0.00014061674008810572,
-      "loss": 1.5685,
+      "learning_rate": 0.0001390308370044053,
+      "loss": 0.1345,
       "step": 1113
     },
     {
       "epoch": 10.22,
-      "learning_rate": 0.00014035242290748898,
-      "loss": 1.57,
+      "learning_rate": 0.00013876651982378853,
+      "loss": 0.1609,
       "step": 1114
     },
     {
       "epoch": 10.23,
-      "learning_rate": 0.00014008810572687224,
-      "loss": 1.5094,
+      "learning_rate": 0.00013850220264317178,
+      "loss": 0.1547,
       "step": 1115
     },
     {
       "epoch": 10.24,
-      "learning_rate": 0.0001398237885462555,
-      "loss": 1.6545,
+      "learning_rate": 0.00013823788546255507,
+      "loss": 0.0904,
       "step": 1116
     },
     {
       "epoch": 10.25,
-      "learning_rate": 0.00013955947136563875,
-      "loss": 1.5069,
+      "learning_rate": 0.0001379735682819383,
+      "loss": 0.0758,
       "step": 1117
     },
     {
       "epoch": 10.26,
-      "learning_rate": 0.000139295154185022,
-      "loss": 1.5598,
+      "learning_rate": 0.00013770925110132156,
+      "loss": 0.1107,
       "step": 1118
     },
     {
       "epoch": 10.27,
-      "learning_rate": 0.0001390308370044053,
-      "loss": 1.6012,
+      "learning_rate": 0.00013744493392070484,
+      "loss": 0.1254,
       "step": 1119
     },
     {
       "epoch": 10.28,
-      "learning_rate": 0.00013876651982378853,
-      "loss": 1.5345,
+      "learning_rate": 0.0001371806167400881,
+      "loss": 0.1265,
       "step": 1120
     },
     {
       "epoch": 10.28,
-      "learning_rate": 0.00013850220264317178,
-      "loss": 1.5899,
+      "learning_rate": 0.00013691629955947136,
+      "loss": 0.1658,
       "step": 1121
     },
     {
       "epoch": 10.29,
-      "learning_rate": 0.00013823788546255507,
-      "loss": 1.7795,
+      "learning_rate": 0.00013665198237885462,
+      "loss": 0.0688,
       "step": 1122
     },
     {
       "epoch": 10.3,
-      "learning_rate": 0.0001379735682819383,
-      "loss": 1.5573,
+      "learning_rate": 0.00013638766519823788,
+      "loss": 0.1023,
       "step": 1123
     },
     {
       "epoch": 10.31,
-      "learning_rate": 0.00013770925110132156,
-      "loss": 1.5792,
+      "learning_rate": 0.00013612334801762113,
+      "loss": 0.08,
       "step": 1124
     },
     {
       "epoch": 10.32,
-      "learning_rate": 0.00013744493392070484,
-      "loss": 1.5095,
+      "learning_rate": 0.0001358590308370044,
+      "loss": 0.1163,
       "step": 1125
     },
     {
       "epoch": 10.33,
-      "learning_rate": 0.0001371806167400881,
-      "loss": 1.5703,
+      "learning_rate": 0.00013559471365638765,
+      "loss": 0.1488,
       "step": 1126
     },
     {
       "epoch": 10.34,
-      "learning_rate": 0.00013691629955947136,
-      "loss": 1.514,
+      "learning_rate": 0.0001353303964757709,
+      "loss": 0.1208,
       "step": 1127
     },
     {
       "epoch": 10.35,
-      "learning_rate": 0.00013665198237885462,
-      "loss": 1.6368,
+      "learning_rate": 0.00013506607929515417,
+      "loss": 0.1321,
       "step": 1128
     },
     {
       "epoch": 10.36,
-      "learning_rate": 0.00013638766519823788,
-      "loss": 1.592,
+      "learning_rate": 0.00013480176211453743,
+      "loss": 0.1185,
       "step": 1129
     },
     {
       "epoch": 10.37,
-      "learning_rate": 0.00013612334801762113,
-      "loss": 1.5122,
+      "learning_rate": 0.00013453744493392068,
+      "loss": 0.1034,
       "step": 1130
     },
     {
       "epoch": 10.38,
-      "learning_rate": 0.0001358590308370044,
-      "loss": 1.5336,
+      "learning_rate": 0.00013427312775330397,
+      "loss": 0.0996,
       "step": 1131
     },
     {
       "epoch": 10.39,
-      "learning_rate": 0.00013559471365638765,
-      "loss": 1.5602,
+      "learning_rate": 0.0001340088105726872,
+      "loss": 0.1105,
       "step": 1132
     },
     {
       "epoch": 10.39,
-      "learning_rate": 0.0001353303964757709,
-      "loss": 1.5211,
+      "learning_rate": 0.00013374449339207048,
+      "loss": 0.0845,
       "step": 1133
     },
     {
       "epoch": 10.4,
-      "learning_rate": 0.00013506607929515417,
-      "loss": 1.5703,
+      "learning_rate": 0.00013348017621145374,
+      "loss": 0.1784,
       "step": 1134
     },
     {
       "epoch": 10.41,
-      "learning_rate": 0.00013480176211453743,
-      "loss": 1.5531,
+      "learning_rate": 0.00013321585903083697,
+      "loss": 0.0856,
       "step": 1135
     },
     {
       "epoch": 10.42,
-      "learning_rate": 0.00013453744493392068,
-      "loss": 1.5494,
+      "learning_rate": 0.00013295154185022026,
+      "loss": 0.0666,
       "step": 1136
     },
     {
       "epoch": 10.43,
-      "learning_rate": 0.00013427312775330397,
-      "loss": 1.5692,
+      "learning_rate": 0.00013268722466960352,
+      "loss": 0.0667,
       "step": 1137
     },
     {
       "epoch": 10.44,
-      "learning_rate": 0.0001340088105726872,
-      "loss": 1.5314,
+      "learning_rate": 0.00013242290748898678,
+      "loss": 0.1157,
       "step": 1138
     },
     {
       "epoch": 10.45,
-      "learning_rate": 0.00013374449339207048,
-      "loss": 1.5327,
+      "learning_rate": 0.00013215859030837003,
+      "loss": 0.1463,
       "step": 1139
     },
     {
       "epoch": 10.46,
-      "learning_rate": 0.00013348017621145374,
-      "loss": 1.5638,
+      "learning_rate": 0.0001318942731277533,
+      "loss": 0.1591,
       "step": 1140
     },
     {
       "epoch": 10.47,
-      "learning_rate": 0.00013321585903083697,
-      "loss": 1.6203,
+      "learning_rate": 0.00013162995594713655,
+      "loss": 0.1064,
       "step": 1141
     },
     {
       "epoch": 10.48,
-      "learning_rate": 0.00013295154185022026,
-      "loss": 1.5667,
+      "learning_rate": 0.0001313656387665198,
+      "loss": 0.1144,
       "step": 1142
     },
     {
       "epoch": 10.49,
-      "learning_rate": 0.00013268722466960352,
-      "loss": 1.553,
+      "learning_rate": 0.00013110132158590307,
+      "loss": 0.0763,
       "step": 1143
     },
     {
       "epoch": 10.5,
-      "learning_rate": 0.00013242290748898678,
-      "loss": 1.5383,
+      "learning_rate": 0.00013083700440528632,
+      "loss": 0.096,
       "step": 1144
     },
     {
       "epoch": 10.5,
-      "learning_rate": 0.00013215859030837003,
-      "loss": 1.518,
+      "learning_rate": 0.0001305726872246696,
+      "loss": 0.1363,
       "step": 1145
     },
     {
       "epoch": 10.51,
-      "learning_rate": 0.0001318942731277533,
-      "loss": 1.5163,
+      "learning_rate": 0.00013030837004405284,
+      "loss": 0.1399,
       "step": 1146
     },
     {
       "epoch": 10.52,
-      "learning_rate": 0.00013162995594713655,
-      "loss": 1.7565,
+      "learning_rate": 0.0001300440528634361,
+      "loss": 0.1278,
       "step": 1147
     },
     {
       "epoch": 10.53,
-      "learning_rate": 0.0001313656387665198,
-      "loss": 1.5791,
+      "learning_rate": 0.00012977973568281938,
+      "loss": 0.1156,
       "step": 1148
     },
     {
       "epoch": 10.54,
-      "learning_rate": 0.00013110132158590307,
-      "loss": 1.5326,
+      "learning_rate": 0.00012951541850220264,
+      "loss": 0.1058,
       "step": 1149
     },
     {
       "epoch": 10.55,
-      "learning_rate": 0.00013083700440528632,
-      "loss": 1.4817,
+      "learning_rate": 0.0001292511013215859,
+      "loss": 0.1207,
       "step": 1150
     },
     {
       "epoch": 10.56,
-      "learning_rate": 0.0001305726872246696,
-      "loss": 1.5087,
+      "learning_rate": 0.00012898678414096916,
+      "loss": 0.1166,
       "step": 1151
     },
     {
       "epoch": 10.57,
-      "learning_rate": 0.00013030837004405284,
-      "loss": 1.5143,
+      "learning_rate": 0.00012872246696035242,
+      "loss": 0.1246,
       "step": 1152
     },
     {
       "epoch": 10.58,
-      "learning_rate": 0.0001300440528634361,
-      "loss": 1.6335,
+      "learning_rate": 0.00012845814977973567,
+      "loss": 0.1535,
       "step": 1153
     },
     {
       "epoch": 10.59,
-      "learning_rate": 0.00012977973568281938,
-      "loss": 1.5427,
+      "learning_rate": 0.00012819383259911893,
+      "loss": 0.0768,
       "step": 1154
     },
     {
       "epoch": 10.6,
-      "learning_rate": 0.00012951541850220264,
-      "loss": 1.5492,
+      "learning_rate": 0.0001279295154185022,
+      "loss": 0.1135,
       "step": 1155
     },
     {
       "epoch": 10.61,
-      "learning_rate": 0.0001292511013215859,
-      "loss": 1.5266,
+      "learning_rate": 0.00012766519823788545,
+      "loss": 0.0971,
       "step": 1156
     },
     {
       "epoch": 10.61,
-      "learning_rate": 0.00012898678414096916,
-      "loss": 1.5393,
+      "learning_rate": 0.0001274008810572687,
+      "loss": 0.0929,
       "step": 1157
     },
     {
       "epoch": 10.62,
-      "learning_rate": 0.00012872246696035242,
-      "loss": 1.5511,
+      "learning_rate": 0.00012713656387665197,
+      "loss": 0.1718,
       "step": 1158
     },
     {
       "epoch": 10.63,
-      "learning_rate": 0.00012845814977973567,
-      "loss": 1.6114,
+      "learning_rate": 0.00012687224669603522,
+      "loss": 0.1988,
       "step": 1159
     },
     {
       "epoch": 10.64,
-      "learning_rate": 0.00012819383259911893,
-      "loss": 1.5549,
+      "learning_rate": 0.0001266079295154185,
+      "loss": 0.0932,
       "step": 1160
     },
     {
       "epoch": 10.65,
-      "learning_rate": 0.0001279295154185022,
-      "loss": 1.5257,
+      "learning_rate": 0.00012634361233480174,
+      "loss": 0.0779,
       "step": 1161
     },
     {
       "epoch": 10.66,
-      "learning_rate": 0.00012766519823788545,
-      "loss": 1.5248,
+      "learning_rate": 0.00012607929515418503,
+      "loss": 0.103,
       "step": 1162
     },
     {
       "epoch": 10.67,
-      "learning_rate": 0.0001274008810572687,
-      "loss": 1.5507,
+      "learning_rate": 0.00012581497797356828,
+      "loss": 0.1053,
       "step": 1163
     },
     {
       "epoch": 10.68,
-      "learning_rate": 0.00012713656387665197,
-      "loss": 1.5748,
+      "learning_rate": 0.00012555066079295151,
+      "loss": 0.121,
       "step": 1164
     },
     {
       "epoch": 10.69,
-      "learning_rate": 0.00012687224669603522,
-      "loss": 1.5115,
+      "learning_rate": 0.0001252863436123348,
+      "loss": 0.1839,
       "step": 1165
     },
     {
       "epoch": 10.7,
-      "learning_rate": 0.0001266079295154185,
-      "loss": 1.6113,
+      "learning_rate": 0.00012502202643171806,
+      "loss": 0.0829,
       "step": 1166
     },
     {
       "epoch": 10.71,
-      "learning_rate": 0.00012634361233480174,
-      "loss": 1.5139,
+      "learning_rate": 0.00012475770925110132,
+      "loss": 0.0764,
       "step": 1167
     },
     {
       "epoch": 10.72,
-      "learning_rate": 0.00012607929515418503,
-      "loss": 1.5388,
+      "learning_rate": 0.00012449339207048457,
+      "loss": 0.143,
       "step": 1168
     },
     {
       "epoch": 10.72,
-      "learning_rate": 0.00012581497797356828,
-      "loss": 1.5804,
+      "learning_rate": 0.00012422907488986783,
+      "loss": 0.1292,
       "step": 1169
     },
     {
       "epoch": 10.73,
-      "learning_rate": 0.00012555066079295151,
-      "loss": 1.5486,
+      "learning_rate": 0.0001239647577092511,
+      "loss": 0.1887,
       "step": 1170
     },
     {
       "epoch": 10.74,
-      "learning_rate": 0.0001252863436123348,
-      "loss": 1.6409,
+      "learning_rate": 0.00012370044052863435,
+      "loss": 0.1439,
       "step": 1171
     },
     {
       "epoch": 10.75,
-      "learning_rate": 0.00012502202643171806,
-      "loss": 1.6016,
+      "learning_rate": 0.0001234361233480176,
+      "loss": 0.1357,
       "step": 1172
     },
     {
       "epoch": 10.76,
-      "learning_rate": 0.00012475770925110132,
-      "loss": 1.5654,
+      "learning_rate": 0.00012317180616740086,
+      "loss": 0.1228,
       "step": 1173
     },
     {
       "epoch": 10.77,
-      "learning_rate": 0.00012449339207048457,
-      "loss": 1.5399,
+      "learning_rate": 0.00012290748898678412,
+      "loss": 0.0908,
       "step": 1174
     },
     {
       "epoch": 10.78,
-      "learning_rate": 0.00012422907488986783,
-      "loss": 1.5156,
+      "learning_rate": 0.00012264317180616738,
+      "loss": 0.1005,
       "step": 1175
     },
     {
       "epoch": 10.79,
-      "learning_rate": 0.0001239647577092511,
-      "loss": 1.5279,
+      "learning_rate": 0.00012237885462555064,
+      "loss": 0.1364,
       "step": 1176
     },
     {
       "epoch": 10.8,
-      "learning_rate": 0.00012370044052863435,
-      "loss": 1.541,
+      "learning_rate": 0.00012211453744493392,
+      "loss": 0.1338,
       "step": 1177
     },
     {
       "epoch": 10.81,
-      "learning_rate": 0.0001234361233480176,
-      "loss": 1.6171,
+      "learning_rate": 0.00012185022026431718,
+      "loss": 0.0755,
       "step": 1178
     },
     {
       "epoch": 10.82,
-      "learning_rate": 0.00012317180616740086,
-      "loss": 1.5107,
+      "learning_rate": 0.00012158590308370043,
+      "loss": 0.066,
       "step": 1179
     },
     {
       "epoch": 10.83,
-      "learning_rate": 0.00012290748898678412,
-      "loss": 1.5255,
+      "learning_rate": 0.00012132158590308369,
+      "loss": 0.1124,
       "step": 1180
     },
     {
       "epoch": 10.83,
-      "learning_rate": 0.00012264317180616738,
-      "loss": 1.4897,
+      "learning_rate": 0.00012105726872246696,
+      "loss": 0.1175,
       "step": 1181
     },
     {
       "epoch": 10.84,
-      "learning_rate": 0.00012237885462555064,
-      "loss": 1.4565,
+      "learning_rate": 0.0001207929515418502,
+      "loss": 0.1069,
       "step": 1182
     },
     {
       "epoch": 10.85,
-      "learning_rate": 0.00012211453744493392,
-      "loss": 1.5377,
+      "learning_rate": 0.00012052863436123347,
+      "loss": 0.1581,
       "step": 1183
     },
     {
       "epoch": 10.86,
-      "learning_rate": 0.00012185022026431718,
-      "loss": 1.5567,
+      "learning_rate": 0.00012026431718061673,
+      "loss": 0.1614,
       "step": 1184
     },
     {
       "epoch": 10.87,
-      "learning_rate": 0.00012158590308370043,
-      "loss": 1.5353,
+      "learning_rate": 0.00011999999999999999,
+      "loss": 0.0848,
       "step": 1185
     },
     {
       "epoch": 10.88,
-      "learning_rate": 0.00012132158590308369,
-      "loss": 1.4727,
+      "learning_rate": 0.00011973568281938325,
+      "loss": 0.0833,
       "step": 1186
     },
     {
       "epoch": 10.89,
-      "learning_rate": 0.00012105726872246696,
-      "loss": 1.553,
+      "learning_rate": 0.0001194713656387665,
+      "loss": 0.149,
       "step": 1187
     },
     {
       "epoch": 10.9,
-      "learning_rate": 0.0001207929515418502,
-      "loss": 1.5276,
+      "learning_rate": 0.00011920704845814978,
+      "loss": 0.151,
       "step": 1188
     },
     {
       "epoch": 10.91,
-      "learning_rate": 0.00012052863436123347,
-      "loss": 1.4707,
+      "learning_rate": 0.00011894273127753302,
+      "loss": 0.1447,
       "step": 1189
     },
     {
       "epoch": 10.92,
-      "learning_rate": 0.00012026431718061673,
-      "loss": 1.5177,
+      "learning_rate": 0.00011867841409691628,
+      "loss": 0.1953,
       "step": 1190
     },
     {
       "epoch": 10.93,
-      "learning_rate": 0.00011999999999999999,
-      "loss": 1.5896,
+      "learning_rate": 0.00011841409691629955,
+      "loss": 0.1053,
       "step": 1191
     },
     {
       "epoch": 10.94,
-      "learning_rate": 0.00011973568281938325,
-      "loss": 1.5048,
+      "learning_rate": 0.00011814977973568281,
+      "loss": 0.1165,
       "step": 1192
     },
     {
       "epoch": 10.94,
-      "learning_rate": 0.0001194713656387665,
-      "loss": 1.5094,
+      "learning_rate": 0.00011788546255506607,
+      "loss": 0.0916,
       "step": 1193
     },
     {
       "epoch": 10.95,
-      "learning_rate": 0.00011920704845814978,
-      "loss": 1.4948,
+      "learning_rate": 0.00011762114537444933,
+      "loss": 0.0704,
       "step": 1194
     },
     {
       "epoch": 10.96,
-      "learning_rate": 0.00011894273127753302,
-      "loss": 1.4661,
+      "learning_rate": 0.0001173568281938326,
+      "loss": 0.1679,
       "step": 1195
     },
     {
       "epoch": 10.97,
-      "learning_rate": 0.00011867841409691628,
-      "loss": 1.6128,
+      "learning_rate": 0.00011709251101321586,
+      "loss": 0.1749,
       "step": 1196
     },
     {
       "epoch": 10.98,
-      "learning_rate": 0.00011841409691629955,
-      "loss": 1.4812,
+      "learning_rate": 0.0001168281938325991,
+      "loss": 0.1957,
       "step": 1197
     },
     {
       "epoch": 10.99,
-      "learning_rate": 0.00011814977973568281,
-      "loss": 1.5494,
+      "learning_rate": 0.00011656387665198237,
+      "loss": 0.1103,
       "step": 1198
     },
     {
       "epoch": 11.0,
-      "learning_rate": 0.00011788546255506607,
-      "loss": 1.9485,
+      "learning_rate": 0.00011629955947136563,
+      "loss": 0.1216,
       "step": 1199
     },
     {
       "epoch": 11.01,
-      "learning_rate": 0.00011762114537444933,
-      "loss": 1.5667,
+      "learning_rate": 0.00011603524229074888,
+      "loss": 0.0691,
       "step": 1200
     },
     {
       "epoch": 11.01,
-      "eval_loss": 0.43370336294174194,
-      "eval_runtime": 20.9972,
-      "eval_samples_per_second": 78.439,
-      "eval_steps_per_second": 1.238,
-      "eval_wer": 0.463180471861914,
+      "eval_loss": 0.3922382891178131,
+      "eval_runtime": 20.9593,
+      "eval_samples_per_second": 78.581,
+      "eval_steps_per_second": 1.24,
+      "eval_wer": 0.37983862731079565,
       "step": 1200
     },
     {
       "epoch": 11.02,
-      "learning_rate": 0.0001173568281938326,
-      "loss": 1.4949,
+      "learning_rate": 0.00011577092511013215,
+      "loss": 0.0627,
       "step": 1201
     },
     {
       "epoch": 11.03,
-      "learning_rate": 0.00011709251101321586,
-      "loss": 1.5519,
+      "learning_rate": 0.0001155066079295154,
+      "loss": 0.0974,
       "step": 1202
     },
     {
       "epoch": 11.04,
-      "learning_rate": 0.0001168281938325991,
-      "loss": 1.5931,
+      "learning_rate": 0.00011524229074889868,
+      "loss": 0.0954,
       "step": 1203
     },
     {
       "epoch": 11.05,
-      "learning_rate": 0.00011656387665198237,
-      "loss": 1.4918,
+      "learning_rate": 0.00011497797356828192,
+      "loss": 0.153,
       "step": 1204
     },
     {
       "epoch": 11.06,
-      "learning_rate": 0.00011629955947136563,
-      "loss": 1.498,
+      "learning_rate": 0.00011471365638766519,
+      "loss": 0.1071,
       "step": 1205
     },
     {
       "epoch": 11.06,
-      "learning_rate": 0.00011603524229074888,
-      "loss": 1.5518,
+      "learning_rate": 0.00011444933920704845,
+      "loss": 0.0703,
       "step": 1206
     },
     {
       "epoch": 11.07,
-      "learning_rate": 0.00011577092511013215,
-      "loss": 1.5334,
+      "learning_rate": 0.00011418502202643172,
+      "loss": 0.0838,
       "step": 1207
     },
     {
       "epoch": 11.08,
-      "learning_rate": 0.0001155066079295154,
-      "loss": 1.5252,
+      "learning_rate": 0.00011392070484581497,
+      "loss": 0.0828,
       "step": 1208
     },
     {
       "epoch": 11.09,
-      "learning_rate": 0.00011524229074889868,
-      "loss": 1.4789,
+      "learning_rate": 0.00011365638766519823,
+      "loss": 0.1275,
       "step": 1209
     },
     {
       "epoch": 11.1,
-      "learning_rate": 0.00011497797356828192,
-      "loss": 1.5152,
+      "learning_rate": 0.0001133920704845815,
+      "loss": 0.1507,
       "step": 1210
     },
     {
       "epoch": 11.11,
-      "learning_rate": 0.00011471365638766519,
-      "loss": 1.5421,
+      "learning_rate": 0.00011312775330396474,
+      "loss": 0.2026,
       "step": 1211
     },
     {
       "epoch": 11.12,
-      "learning_rate": 0.00011444933920704845,
-      "loss": 1.6477,
+      "learning_rate": 0.000112863436123348,
+      "loss": 0.0651,
       "step": 1212
     },
     {
       "epoch": 11.13,
-      "learning_rate": 0.00011418502202643172,
-      "loss": 1.5492,
+      "learning_rate": 0.00011259911894273127,
+      "loss": 0.1351,
       "step": 1213
     },
     {
       "epoch": 11.14,
-      "learning_rate": 0.00011392070484581497,
-      "loss": 1.5514,
+      "learning_rate": 0.00011233480176211453,
+      "loss": 0.1235,
       "step": 1214
     },
     {
       "epoch": 11.15,
-      "learning_rate": 0.00011365638766519823,
-      "loss": 1.4917,
+      "learning_rate": 0.00011207048458149779,
+      "loss": 0.0679,
       "step": 1215
     },
     {
       "epoch": 11.16,
-      "learning_rate": 0.0001133920704845815,
-      "loss": 1.4686,
+      "learning_rate": 0.00011180616740088105,
+      "loss": 0.1111,
       "step": 1216
     },
     {
       "epoch": 11.17,
-      "learning_rate": 0.00011312775330396474,
-      "loss": 1.5125,
+      "learning_rate": 0.00011154185022026432,
+      "loss": 0.1301,
       "step": 1217
     },
     {
       "epoch": 11.17,
-      "learning_rate": 0.000112863436123348,
-      "loss": 1.5181,
+      "learning_rate": 0.00011127753303964756,
+      "loss": 0.1177,
       "step": 1218
     },
     {
       "epoch": 11.18,
-      "learning_rate": 0.00011259911894273127,
-      "loss": 1.5211,
+      "learning_rate": 0.00011101321585903082,
+      "loss": 0.0484,
       "step": 1219
     },
     {
       "epoch": 11.19,
-      "learning_rate": 0.00011233480176211453,
-      "loss": 1.5371,
+      "learning_rate": 0.00011074889867841409,
+      "loss": 0.0639,
       "step": 1220
     },
     {
       "epoch": 11.2,
-      "learning_rate": 0.00011207048458149779,
-      "loss": 1.498,
+      "learning_rate": 0.00011048458149779735,
+      "loss": 0.0846,
       "step": 1221
     },
     {
       "epoch": 11.21,
-      "learning_rate": 0.00011180616740088105,
-      "loss": 1.5204,
+      "learning_rate": 0.0001102202643171806,
+      "loss": 0.1091,
       "step": 1222
     },
     {
       "epoch": 11.22,
-      "learning_rate": 0.00011154185022026432,
-      "loss": 1.5045,
+      "learning_rate": 0.00010995594713656387,
+      "loss": 0.1301,
       "step": 1223
     },
     {
       "epoch": 11.23,
-      "learning_rate": 0.00011127753303964756,
-      "loss": 1.5553,
+      "learning_rate": 0.00010969162995594712,
+      "loss": 0.2167,
       "step": 1224
     },
     {
       "epoch": 11.24,
-      "learning_rate": 0.00011101321585903082,
-      "loss": 1.647,
+      "learning_rate": 0.0001094273127753304,
+      "loss": 0.0951,
       "step": 1225
     },
     {
       "epoch": 11.25,
-      "learning_rate": 0.00011074889867841409,
-      "loss": 1.5047,
+      "learning_rate": 0.00010916299559471364,
+      "loss": 0.0737,
       "step": 1226
     },
     {
       "epoch": 11.26,
-      "learning_rate": 0.00011048458149779735,
-      "loss": 1.4503,
+      "learning_rate": 0.00010889867841409691,
+      "loss": 0.1023,
       "step": 1227
     },
     {
       "epoch": 11.27,
-      "learning_rate": 0.0001102202643171806,
-      "loss": 1.518,
+      "learning_rate": 0.00010863436123348017,
+      "loss": 0.0808,
       "step": 1228
     },
     {
       "epoch": 11.28,
-      "learning_rate": 0.00010995594713656387,
-      "loss": 1.5057,
+      "learning_rate": 0.00010837004405286342,
+      "loss": 0.156,
       "step": 1229
     },
     {
       "epoch": 11.28,
-      "learning_rate": 0.00010969162995594712,
-      "loss": 1.5258,
+      "learning_rate": 0.00010810572687224669,
+      "loss": 0.141,
       "step": 1230
     },
     {
       "epoch": 11.29,
-      "learning_rate": 0.0001094273127753304,
-      "loss": 1.6164,
+      "learning_rate": 0.00010784140969162995,
+      "loss": 0.0832,
       "step": 1231
     },
     {
       "epoch": 11.3,
-      "learning_rate": 0.00010916299559471364,
-      "loss": 1.5406,
+      "learning_rate": 0.00010757709251101322,
+      "loss": 0.0698,
       "step": 1232
     },
     {
       "epoch": 11.31,
-      "learning_rate": 0.00010889867841409691,
-      "loss": 1.5368,
+      "learning_rate": 0.00010731277533039646,
+      "loss": 0.059,
       "step": 1233
     },
     {
       "epoch": 11.32,
-      "learning_rate": 0.00010863436123348017,
-      "loss": 1.503,
+      "learning_rate": 0.00010704845814977972,
+      "loss": 0.0643,
       "step": 1234
     },
     {
       "epoch": 11.33,
-      "learning_rate": 0.00010837004405286342,
-      "loss": 1.5159,
+      "learning_rate": 0.00010678414096916299,
+      "loss": 0.1568,
       "step": 1235
     },
     {
       "epoch": 11.34,
-      "learning_rate": 0.00010810572687224669,
-      "loss": 1.5041,
+      "learning_rate": 0.00010651982378854624,
+      "loss": 0.0931,
       "step": 1236
     },
     {
       "epoch": 11.35,
-      "learning_rate": 0.00010784140969162995,
-      "loss": 1.5158,
+      "learning_rate": 0.00010625550660792951,
+      "loss": 0.1392,
       "step": 1237
     },
     {
       "epoch": 11.36,
-      "learning_rate": 0.00010757709251101322,
-      "loss": 1.5301,
+      "learning_rate": 0.00010599118942731277,
+      "loss": 0.0717,
       "step": 1238
     },
     {
       "epoch": 11.37,
-      "learning_rate": 0.00010731277533039646,
-      "loss": 1.495,
+      "learning_rate": 0.00010572687224669604,
+      "loss": 0.0716,
       "step": 1239
     },
     {
       "epoch": 11.38,
-      "learning_rate": 0.00010704845814977972,
-      "loss": 1.4934,
+      "learning_rate": 0.00010546255506607928,
+      "loss": 0.1098,
       "step": 1240
     },
     {
       "epoch": 11.39,
-      "learning_rate": 0.00010678414096916299,
-      "loss": 1.4698,
+      "learning_rate": 0.00010519823788546254,
+      "loss": 0.0822,
       "step": 1241
     },
     {
       "epoch": 11.39,
-      "learning_rate": 0.00010651982378854624,
-      "loss": 1.5279,
+      "learning_rate": 0.00010493392070484581,
+      "loss": 0.1944,
       "step": 1242
     },
     {
       "epoch": 11.4,
-      "learning_rate": 0.00010625550660792951,
-      "loss": 1.472,
+      "learning_rate": 0.00010466960352422907,
+      "loss": 0.1619,
       "step": 1243
     },
     {
       "epoch": 11.41,
-      "learning_rate": 0.00010599118942731277,
-      "loss": 1.4987,
+      "learning_rate": 0.00010440528634361231,
+      "loss": 0.0784,
       "step": 1244
     },
     {
       "epoch": 11.42,
-      "learning_rate": 0.00010572687224669604,
-      "loss": 1.5118,
+      "learning_rate": 0.00010414096916299559,
+      "loss": 0.0667,
       "step": 1245
     },
     {
       "epoch": 11.43,
-      "learning_rate": 0.00010546255506607928,
-      "loss": 1.5138,
+      "learning_rate": 0.00010387665198237884,
+      "loss": 0.067,
       "step": 1246
     },
     {
       "epoch": 11.44,
-      "learning_rate": 0.00010519823788546254,
-      "loss": 1.4867,
+      "learning_rate": 0.0001036123348017621,
+      "loss": 0.0904,
       "step": 1247
     },
     {
       "epoch": 11.45,
-      "learning_rate": 0.00010493392070484581,
-      "loss": 1.5252,
+      "learning_rate": 0.00010334801762114536,
+      "loss": 0.0971,
       "step": 1248
     },
     {
       "epoch": 11.46,
-      "learning_rate": 0.00010466960352422907,
-      "loss": 1.5149,
+      "learning_rate": 0.00010308370044052863,
+      "loss": 0.1179,
       "step": 1249
     },
     {
       "epoch": 11.47,
-      "learning_rate": 0.00010440528634361231,
-      "loss": 1.51,
+      "learning_rate": 0.00010281938325991189,
+      "loss": 0.057,
       "step": 1250
     },
     {
       "epoch": 11.48,
-      "learning_rate": 0.00010414096916299559,
-      "loss": 1.5318,
+      "learning_rate": 0.00010255506607929514,
+      "loss": 0.1131,
       "step": 1251
     },
     {
       "epoch": 11.49,
-      "learning_rate": 0.00010387665198237884,
-      "loss": 1.4992,
+      "learning_rate": 0.00010229074889867841,
+      "loss": 0.1203,
       "step": 1252
     },
     {
       "epoch": 11.5,
-      "learning_rate": 0.0001036123348017621,
-      "loss": 1.4932,
+      "learning_rate": 0.00010202643171806167,
+      "loss": 0.1213,
       "step": 1253
     },
     {
       "epoch": 11.5,
-      "learning_rate": 0.00010334801762114536,
-      "loss": 1.526,
+      "learning_rate": 0.00010176211453744494,
+      "loss": 0.0837,
       "step": 1254
     },
     {
       "epoch": 11.51,
-      "learning_rate": 0.00010308370044052863,
-      "loss": 1.4746,
+      "learning_rate": 0.00010149779735682818,
+      "loss": 0.0799,
       "step": 1255
     },
     {
       "epoch": 11.52,
-      "learning_rate": 0.00010281938325991189,
-      "loss": 1.5285,
+      "learning_rate": 0.00010123348017621144,
+      "loss": 0.0948,
       "step": 1256
     },
     {
       "epoch": 11.53,
-      "learning_rate": 0.00010255506607929514,
-      "loss": 1.5386,
+      "learning_rate": 0.00010096916299559471,
+      "loss": 0.0967,
       "step": 1257
     },
     {
       "epoch": 11.54,
-      "learning_rate": 0.00010229074889867841,
-      "loss": 1.556,
+      "learning_rate": 0.00010070484581497796,
+      "loss": 0.0805,
       "step": 1258
     },
     {
       "epoch": 11.55,
-      "learning_rate": 0.00010202643171806167,
-      "loss": 1.4758,
+      "learning_rate": 0.00010044052863436123,
+      "loss": 0.0755,
       "step": 1259
     },
     {
       "epoch": 11.56,
-      "learning_rate": 0.00010176211453744494,
-      "loss": 1.4726,
+      "learning_rate": 0.00010017621145374449,
+      "loss": 0.1151,
       "step": 1260
     },
     {
       "epoch": 11.57,
-      "learning_rate": 0.00010149779735682818,
-      "loss": 1.5111,
+      "learning_rate": 9.991189427312776e-05,
+      "loss": 0.1114,
       "step": 1261
     },
     {
       "epoch": 11.58,
-      "learning_rate": 0.00010123348017621144,
-      "loss": 1.7773,
+      "learning_rate": 9.9647577092511e-05,
+      "loss": 0.0772,
       "step": 1262
     },
     {
       "epoch": 11.59,
-      "learning_rate": 0.00010096916299559471,
-      "loss": 1.5433,
+      "learning_rate": 9.938325991189426e-05,
+      "loss": 0.0725,
       "step": 1263
     },
     {
       "epoch": 11.6,
-      "learning_rate": 0.00010070484581497796,
-      "loss": 1.4969,
+      "learning_rate": 9.911894273127753e-05,
+      "loss": 0.0835,
       "step": 1264
     },
     {
       "epoch": 11.61,
-      "learning_rate": 0.00010044052863436123,
-      "loss": 1.5735,
+      "learning_rate": 9.885462555066078e-05,
+      "loss": 0.1527,
       "step": 1265
     },
     {
       "epoch": 11.61,
-      "learning_rate": 0.00010017621145374449,
-      "loss": 1.4888,
+      "learning_rate": 9.859030837004403e-05,
+      "loss": 0.0821,
       "step": 1266
     },
     {
       "epoch": 11.62,
-      "learning_rate": 9.991189427312776e-05,
-      "loss": 1.542,
+      "learning_rate": 9.83259911894273e-05,
+      "loss": 0.1087,
       "step": 1267
     },
     {
       "epoch": 11.63,
-      "learning_rate": 9.9647577092511e-05,
-      "loss": 1.4666,
+      "learning_rate": 9.806167400881056e-05,
+      "loss": 0.0937,
       "step": 1268
     },
     {
       "epoch": 11.64,
-      "learning_rate": 9.938325991189426e-05,
-      "loss": 1.5336,
+      "learning_rate": 9.779735682819382e-05,
+      "loss": 0.1045,
       "step": 1269
     },
     {
       "epoch": 11.65,
-      "learning_rate": 9.911894273127753e-05,
-      "loss": 1.5172,
+      "learning_rate": 9.753303964757708e-05,
+      "loss": 0.0832,
       "step": 1270
     },
     {
       "epoch": 11.66,
-      "learning_rate": 9.885462555066078e-05,
-      "loss": 1.4628,
+      "learning_rate": 9.726872246696035e-05,
+      "loss": 0.1076,
       "step": 1271
     },
     {
       "epoch": 11.67,
-      "learning_rate": 9.859030837004403e-05,
-      "loss": 1.525,
+      "learning_rate": 9.700440528634361e-05,
+      "loss": 0.0973,
       "step": 1272
     },
     {
       "epoch": 11.68,
-      "learning_rate": 9.83259911894273e-05,
-      "loss": 1.5023,
+      "learning_rate": 9.674008810572686e-05,
+      "loss": 0.1237,
       "step": 1273
     },
     {
       "epoch": 11.69,
-      "learning_rate": 9.806167400881056e-05,
-      "loss": 1.5739,
+      "learning_rate": 9.647577092511013e-05,
+      "loss": 0.1068,
       "step": 1274
     },
     {
       "epoch": 11.7,
-      "learning_rate": 9.779735682819382e-05,
-      "loss": 1.5225,
+      "learning_rate": 9.621145374449339e-05,
+      "loss": 0.0644,
       "step": 1275
     },
     {
       "epoch": 11.71,
-      "learning_rate": 9.753303964757708e-05,
-      "loss": 1.4997,
+      "learning_rate": 9.594713656387664e-05,
+      "loss": 0.0831,
       "step": 1276
     },
     {
       "epoch": 11.72,
-      "learning_rate": 9.726872246696035e-05,
-      "loss": 1.504,
+      "learning_rate": 9.56828193832599e-05,
+      "loss": 0.1116,
       "step": 1277
     },
     {
       "epoch": 11.72,
-      "learning_rate": 9.700440528634361e-05,
-      "loss": 1.5095,
+      "learning_rate": 9.541850220264316e-05,
+      "loss": 0.1238,
       "step": 1278
     },
     {
       "epoch": 11.73,
-      "learning_rate": 9.674008810572686e-05,
-      "loss": 1.5436,
+      "learning_rate": 9.515418502202643e-05,
+      "loss": 0.0571,
       "step": 1279
     },
     {
       "epoch": 11.74,
-      "learning_rate": 9.647577092511013e-05,
-      "loss": 1.5135,
+      "learning_rate": 9.488986784140968e-05,
+      "loss": 0.1301,
       "step": 1280
     },
     {
       "epoch": 11.75,
-      "learning_rate": 9.621145374449339e-05,
-      "loss": 1.61,
+      "learning_rate": 9.462555066079295e-05,
+      "loss": 0.0946,
       "step": 1281
     },
     {
       "epoch": 11.76,
-      "learning_rate": 9.594713656387664e-05,
-      "loss": 1.5256,
+      "learning_rate": 9.43612334801762e-05,
+      "loss": 0.0857,
       "step": 1282
     },
     {
       "epoch": 11.77,
-      "learning_rate": 9.56828193832599e-05,
-      "loss": 1.4422,
+      "learning_rate": 9.409691629955945e-05,
+      "loss": 0.0892,
       "step": 1283
     },
     {
       "epoch": 11.78,
-      "learning_rate": 9.541850220264316e-05,
-      "loss": 1.4856,
+      "learning_rate": 9.383259911894272e-05,
+      "loss": 0.0984,
       "step": 1284
     },
     {
       "epoch": 11.79,
-      "learning_rate": 9.515418502202643e-05,
-      "loss": 1.4937,
+      "learning_rate": 9.356828193832598e-05,
+      "loss": 0.0902,
       "step": 1285
     },
     {
       "epoch": 11.8,
-      "learning_rate": 9.488986784140968e-05,
-      "loss": 1.499,
+      "learning_rate": 9.330396475770925e-05,
+      "loss": 0.0897,
       "step": 1286
     },
     {
       "epoch": 11.81,
-      "learning_rate": 9.462555066079295e-05,
-      "loss": 1.5462,
+      "learning_rate": 9.30396475770925e-05,
+      "loss": 0.0901,
       "step": 1287
     },
     {
       "epoch": 11.82,
-      "learning_rate": 9.43612334801762e-05,
-      "loss": 1.5391,
+      "learning_rate": 9.277533039647575e-05,
+      "loss": 0.072,
       "step": 1288
     },
     {
       "epoch": 11.83,
-      "learning_rate": 9.409691629955945e-05,
-      "loss": 1.4635,
+      "learning_rate": 9.251101321585903e-05,
+      "loss": 0.0792,
       "step": 1289
     },
     {
       "epoch": 11.83,
-      "learning_rate": 9.383259911894272e-05,
-      "loss": 1.4987,
+      "learning_rate": 9.224669603524228e-05,
+      "loss": 0.0913,
       "step": 1290
     },
     {
       "epoch": 11.84,
-      "learning_rate": 9.356828193832598e-05,
-      "loss": 1.609,
+      "learning_rate": 9.198237885462554e-05,
+      "loss": 0.1143,
       "step": 1291
     },
     {
       "epoch": 11.85,
-      "learning_rate": 9.330396475770925e-05,
-      "loss": 1.5287,
+      "learning_rate": 9.17180616740088e-05,
+      "loss": 0.1502,
       "step": 1292
     },
     {
       "epoch": 11.86,
-      "learning_rate": 9.30396475770925e-05,
-      "loss": 1.4973,
+      "learning_rate": 9.145374449339207e-05,
+      "loss": 0.0671,
       "step": 1293
     },
     {
       "epoch": 11.87,
-      "learning_rate": 9.277533039647575e-05,
-      "loss": 1.5303,
+      "learning_rate": 9.118942731277532e-05,
+      "loss": 0.0795,
       "step": 1294
     },
     {
       "epoch": 11.88,
-      "learning_rate": 9.251101321585903e-05,
-      "loss": 1.5246,
+      "learning_rate": 9.092511013215858e-05,
+      "loss": 0.1114,
       "step": 1295
     },
     {
       "epoch": 11.89,
-      "learning_rate": 9.224669603524228e-05,
-      "loss": 1.5383,
+      "learning_rate": 9.066079295154185e-05,
+      "loss": 0.1039,
       "step": 1296
     },
     {
       "epoch": 11.9,
-      "learning_rate": 9.198237885462554e-05,
-      "loss": 1.5202,
+      "learning_rate": 9.03964757709251e-05,
+      "loss": 0.1041,
       "step": 1297
     },
     {
       "epoch": 11.91,
-      "learning_rate": 9.17180616740088e-05,
-      "loss": 1.5355,
+      "learning_rate": 9.013215859030836e-05,
+      "loss": 0.0682,
       "step": 1298
     },
     {
       "epoch": 11.92,
-      "learning_rate": 9.145374449339207e-05,
-      "loss": 1.5254,
+      "learning_rate": 8.986784140969162e-05,
+      "loss": 0.1559,
       "step": 1299
     },
     {
       "epoch": 11.93,
-      "learning_rate": 9.118942731277532e-05,
-      "loss": 1.5733,
+      "learning_rate": 8.960352422907488e-05,
+      "loss": 0.0835,
       "step": 1300
     },
     {
       "epoch": 11.93,
-      "eval_loss": 0.4091590940952301,
-      "eval_runtime": 20.813,
-      "eval_samples_per_second": 79.133,
-      "eval_steps_per_second": 1.249,
-      "eval_wer": 0.4390767030946788,
+      "eval_loss": 0.40420442819595337,
+      "eval_runtime": 21.0898,
+      "eval_samples_per_second": 78.095,
+      "eval_steps_per_second": 1.233,
+      "eval_wer": 0.3671739352466551,
       "step": 1300
     },
     {
       "epoch": 11.94,
-      "learning_rate": 9.092511013215858e-05,
-      "loss": 1.5063,
+      "learning_rate": 8.933920704845815e-05,
+      "loss": 0.0942,
       "step": 1301
     },
     {
       "epoch": 11.94,
-      "learning_rate": 9.066079295154185e-05,
-      "loss": 1.5239,
+      "learning_rate": 8.90748898678414e-05,
+      "loss": 0.0853,
       "step": 1302
     },
     {
       "epoch": 11.95,
-      "learning_rate": 9.03964757709251e-05,
-      "loss": 1.5066,
+      "learning_rate": 8.881057268722467e-05,
+      "loss": 0.0522,
       "step": 1303
     },
     {
       "epoch": 11.96,
-      "learning_rate": 9.013215859030836e-05,
-      "loss": 1.4776,
+      "learning_rate": 8.854625550660793e-05,
+      "loss": 0.0964,
       "step": 1304
     },
     {
       "epoch": 11.97,
-      "learning_rate": 8.986784140969162e-05,
-      "loss": 1.4728,
+      "learning_rate": 8.828193832599117e-05,
+      "loss": 0.118,
       "step": 1305
     },
     {
       "epoch": 11.98,
-      "learning_rate": 8.960352422907488e-05,
-      "loss": 1.5691,
+      "learning_rate": 8.801762114537444e-05,
+      "loss": 0.108,
       "step": 1306
     },
     {
       "epoch": 11.99,
-      "learning_rate": 8.933920704845815e-05,
-      "loss": 1.5315,
+      "learning_rate": 8.77533039647577e-05,
+      "loss": 0.0886,
       "step": 1307
     },
     {
       "epoch": 12.0,
-      "learning_rate": 8.90748898678414e-05,
-      "loss": 2.0071,
+      "learning_rate": 8.748898678414097e-05,
+      "loss": 0.1153,
       "step": 1308
     },
     {
       "epoch": 12.01,
-      "learning_rate": 8.881057268722467e-05,
-      "loss": 1.5192,
+      "learning_rate": 8.722466960352422e-05,
+      "loss": 0.0874,
       "step": 1309
     },
     {
       "epoch": 12.02,
-      "learning_rate": 8.854625550660793e-05,
-      "loss": 1.4924,
+      "learning_rate": 8.696035242290747e-05,
+      "loss": 0.0461,
       "step": 1310
     },
     {
       "epoch": 12.03,
-      "learning_rate": 8.828193832599117e-05,
-      "loss": 1.4775,
+      "learning_rate": 8.669603524229075e-05,
+      "loss": 0.1146,
       "step": 1311
     },
     {
       "epoch": 12.04,
-      "learning_rate": 8.801762114537444e-05,
-      "loss": 1.491,
+      "learning_rate": 8.643171806167399e-05,
+      "loss": 0.0929,
       "step": 1312
     },
     {
       "epoch": 12.05,
-      "learning_rate": 8.77533039647577e-05,
-      "loss": 1.4746,
+      "learning_rate": 8.616740088105726e-05,
+      "loss": 0.094,
       "step": 1313
     },
     {
       "epoch": 12.06,
-      "learning_rate": 8.748898678414097e-05,
-      "loss": 1.5117,
+      "learning_rate": 8.590308370044052e-05,
+      "loss": 0.1288,
       "step": 1314
     },
     {
       "epoch": 12.06,
-      "learning_rate": 8.722466960352422e-05,
-      "loss": 1.6019,
+      "learning_rate": 8.563876651982379e-05,
+      "loss": 0.0947,
       "step": 1315
     },
     {
       "epoch": 12.07,
-      "learning_rate": 8.696035242290747e-05,
-      "loss": 1.491,
+      "learning_rate": 8.537444933920704e-05,
+      "loss": 0.0686,
       "step": 1316
     },
     {
       "epoch": 12.08,
-      "learning_rate": 8.669603524229075e-05,
-      "loss": 1.4998,
+      "learning_rate": 8.51101321585903e-05,
+      "loss": 0.0751,
       "step": 1317
     },
     {
       "epoch": 12.09,
-      "learning_rate": 8.643171806167399e-05,
-      "loss": 1.4868,
+      "learning_rate": 8.484581497797357e-05,
+      "loss": 0.0614,
       "step": 1318
     },
     {
       "epoch": 12.1,
-      "learning_rate": 8.616740088105726e-05,
-      "loss": 1.4857,
+      "learning_rate": 8.458149779735682e-05,
+      "loss": 0.0941,
       "step": 1319
     },
     {
       "epoch": 12.11,
-      "learning_rate": 8.590308370044052e-05,
-      "loss": 1.4969,
+      "learning_rate": 8.431718061674008e-05,
+      "loss": 0.1342,
       "step": 1320
     },
     {
       "epoch": 12.12,
-      "learning_rate": 8.563876651982379e-05,
-      "loss": 1.511,
+      "learning_rate": 8.405286343612334e-05,
+      "loss": 0.1111,
       "step": 1321
     },
     {
       "epoch": 12.13,
-      "learning_rate": 8.537444933920704e-05,
-      "loss": 1.4798,
+      "learning_rate": 8.37885462555066e-05,
+      "loss": 0.097,
       "step": 1322
     },
     {
       "epoch": 12.14,
-      "learning_rate": 8.51101321585903e-05,
-      "loss": 1.5374,
+      "learning_rate": 8.352422907488986e-05,
+      "loss": 0.0647,
       "step": 1323
     },
     {
       "epoch": 12.15,
-      "learning_rate": 8.484581497797357e-05,
-      "loss": 1.4924,
+      "learning_rate": 8.325991189427312e-05,
+      "loss": 0.0987,
       "step": 1324
     },
     {
       "epoch": 12.16,
-      "learning_rate": 8.458149779735682e-05,
-      "loss": 1.4797,
+      "learning_rate": 8.299559471365639e-05,
+      "loss": 0.0943,
       "step": 1325
     },
     {
       "epoch": 12.17,
-      "learning_rate": 8.431718061674008e-05,
-      "loss": 1.4956,
+      "learning_rate": 8.273127753303965e-05,
+      "loss": 0.0956,
       "step": 1326
     },
     {
       "epoch": 12.17,
-      "learning_rate": 8.405286343612334e-05,
-      "loss": 1.5135,
+      "learning_rate": 8.246696035242289e-05,
+      "loss": 0.1403,
       "step": 1327
     },
     {
       "epoch": 12.18,
-      "learning_rate": 8.37885462555066e-05,
-      "loss": 1.4989,
+      "learning_rate": 8.220264317180616e-05,
+      "loss": 0.0665,
       "step": 1328
     },
     {
       "epoch": 12.19,
-      "learning_rate": 8.352422907488986e-05,
-      "loss": 1.4955,
+      "learning_rate": 8.193832599118942e-05,
+      "loss": 0.0931,
       "step": 1329
     },
     {
       "epoch": 12.2,
-      "learning_rate": 8.325991189427312e-05,
-      "loss": 1.4798,
+      "learning_rate": 8.167400881057268e-05,
+      "loss": 0.0924,
       "step": 1330
     },
     {
       "epoch": 12.21,
-      "learning_rate": 8.299559471365639e-05,
-      "loss": 1.5305,
+      "learning_rate": 8.140969162995594e-05,
+      "loss": 0.0767,
       "step": 1331
     },
     {
       "epoch": 12.22,
-      "learning_rate": 8.273127753303965e-05,
-      "loss": 1.5479,
+      "learning_rate": 8.114537444933921e-05,
+      "loss": 0.0922,
       "step": 1332
     },
     {
       "epoch": 12.23,
-      "learning_rate": 8.246696035242289e-05,
-      "loss": 1.4823,
+      "learning_rate": 8.088105726872247e-05,
+      "loss": 0.0979,
       "step": 1333
     },
     {
       "epoch": 12.24,
-      "learning_rate": 8.220264317180616e-05,
-      "loss": 1.5585,
+      "learning_rate": 8.061674008810571e-05,
+      "loss": 0.0693,
       "step": 1334
     },
     {
       "epoch": 12.25,
-      "learning_rate": 8.193832599118942e-05,
-      "loss": 1.4817,
+      "learning_rate": 8.035242290748898e-05,
+      "loss": 0.0416,
       "step": 1335
     },
     {
       "epoch": 12.26,
-      "learning_rate": 8.167400881057268e-05,
-      "loss": 1.4635,
+      "learning_rate": 8.008810572687224e-05,
+      "loss": 0.0869,
       "step": 1336
     },
     {
       "epoch": 12.27,
-      "learning_rate": 8.140969162995594e-05,
-      "loss": 1.4679,
+      "learning_rate": 7.982378854625551e-05,
+      "loss": 0.0863,
       "step": 1337
     },
     {
       "epoch": 12.28,
-      "learning_rate": 8.114537444933921e-05,
-      "loss": 1.5119,
+      "learning_rate": 7.955947136563876e-05,
+      "loss": 0.0746,
       "step": 1338
     },
     {
       "epoch": 12.28,
-      "learning_rate": 8.088105726872247e-05,
-      "loss": 1.5056,
+      "learning_rate": 7.929515418502201e-05,
+      "loss": 0.096,
       "step": 1339
     },
     {
       "epoch": 12.29,
-      "learning_rate": 8.061674008810571e-05,
-      "loss": 1.5951,
+      "learning_rate": 7.903083700440529e-05,
+      "loss": 0.0959,
       "step": 1340
     },
     {
       "epoch": 12.3,
-      "learning_rate": 8.035242290748898e-05,
-      "loss": 1.4861,
+      "learning_rate": 7.876651982378853e-05,
+      "loss": 0.0536,
       "step": 1341
     },
     {
       "epoch": 12.31,
-      "learning_rate": 8.008810572687224e-05,
-      "loss": 1.4816,
+      "learning_rate": 7.85022026431718e-05,
+      "loss": 0.0934,
       "step": 1342
     },
     {
       "epoch": 12.32,
-      "learning_rate": 7.982378854625551e-05,
-      "loss": 1.5003,
+      "learning_rate": 7.823788546255506e-05,
+      "loss": 0.1157,
       "step": 1343
     },
     {
       "epoch": 12.33,
-      "learning_rate": 7.955947136563876e-05,
-      "loss": 1.4585,
+      "learning_rate": 7.797356828193832e-05,
+      "loss": 0.1415,
       "step": 1344
     },
     {
       "epoch": 12.34,
-      "learning_rate": 7.929515418502201e-05,
-      "loss": 1.49,
+      "learning_rate": 7.770925110132158e-05,
+      "loss": 0.1288,
       "step": 1345
     },
     {
       "epoch": 12.35,
-      "learning_rate": 7.903083700440529e-05,
-      "loss": 1.6917,
+      "learning_rate": 7.744493392070484e-05,
+      "loss": 0.1046,
       "step": 1346
     },
     {
       "epoch": 12.36,
-      "learning_rate": 7.876651982378853e-05,
-      "loss": 1.5359,
+      "learning_rate": 7.718061674008811e-05,
+      "loss": 0.0473,
       "step": 1347
     },
     {
       "epoch": 12.37,
-      "learning_rate": 7.85022026431718e-05,
-      "loss": 1.482,
+      "learning_rate": 7.691629955947137e-05,
+      "loss": 0.1119,
       "step": 1348
     },
     {
       "epoch": 12.38,
-      "learning_rate": 7.823788546255506e-05,
-      "loss": 1.5104,
+      "learning_rate": 7.665198237885461e-05,
+      "loss": 0.0883,
       "step": 1349
     },
     {
       "epoch": 12.39,
-      "learning_rate": 7.797356828193832e-05,
-      "loss": 1.46,
+      "learning_rate": 7.638766519823788e-05,
+      "loss": 0.114,
       "step": 1350
     },
     {
       "epoch": 12.39,
-      "learning_rate": 7.770925110132158e-05,
-      "loss": 1.5011,
+      "learning_rate": 7.612334801762114e-05,
+      "loss": 0.1219,
       "step": 1351
     },
     {
       "epoch": 12.4,
-      "learning_rate": 7.744493392070484e-05,
-      "loss": 1.5202,
+      "learning_rate": 7.58590308370044e-05,
+      "loss": 0.1246,
       "step": 1352
     },
     {
       "epoch": 12.41,
-      "learning_rate": 7.718061674008811e-05,
-      "loss": 1.4979,
+      "learning_rate": 7.559471365638766e-05,
+      "loss": 0.0612,
       "step": 1353
     },
     {
       "epoch": 12.42,
-      "learning_rate": 7.691629955947137e-05,
-      "loss": 1.4655,
+      "learning_rate": 7.533039647577093e-05,
+      "loss": 0.0325,
       "step": 1354
     },
     {
       "epoch": 12.43,
-      "learning_rate": 7.665198237885461e-05,
-      "loss": 1.5359,
+      "learning_rate": 7.506607929515419e-05,
+      "loss": 0.1092,
       "step": 1355
     },
     {
       "epoch": 12.44,
-      "learning_rate": 7.638766519823788e-05,
-      "loss": 1.488,
+      "learning_rate": 7.480176211453744e-05,
+      "loss": 0.0529,
       "step": 1356
     },
     {
       "epoch": 12.45,
-      "learning_rate": 7.612334801762114e-05,
-      "loss": 1.4798,
+      "learning_rate": 7.45374449339207e-05,
+      "loss": 0.1272,
       "step": 1357
     },
     {
       "epoch": 12.46,
-      "learning_rate": 7.58590308370044e-05,
-      "loss": 1.6298,
+      "learning_rate": 7.427312775330396e-05,
+      "loss": 0.0708,
       "step": 1358
     },
     {
       "epoch": 12.47,
-      "learning_rate": 7.559471365638766e-05,
-      "loss": 1.5306,
+      "learning_rate": 7.400881057268722e-05,
+      "loss": 0.0657,
       "step": 1359
     },
     {
       "epoch": 12.48,
-      "learning_rate": 7.533039647577093e-05,
-      "loss": 1.5554,
+      "learning_rate": 7.374449339207048e-05,
+      "loss": 0.0741,
       "step": 1360
     },
     {
       "epoch": 12.49,
-      "learning_rate": 7.506607929515419e-05,
-      "loss": 1.539,
+      "learning_rate": 7.348017621145373e-05,
+      "loss": 0.0564,
       "step": 1361
     },
     {
       "epoch": 12.5,
-      "learning_rate": 7.480176211453744e-05,
-      "loss": 1.5268,
+      "learning_rate": 7.321585903083699e-05,
+      "loss": 0.0876,
       "step": 1362
     },
     {
       "epoch": 12.5,
-      "learning_rate": 7.45374449339207e-05,
-      "loss": 1.4924,
+      "learning_rate": 7.295154185022026e-05,
+      "loss": 0.155,
       "step": 1363
     },
     {
       "epoch": 12.51,
-      "learning_rate": 7.427312775330396e-05,
-      "loss": 1.4871,
+      "learning_rate": 7.268722466960352e-05,
+      "loss": 0.0842,
       "step": 1364
     },
     {
       "epoch": 12.52,
-      "learning_rate": 7.400881057268722e-05,
-      "loss": 1.5619,
+      "learning_rate": 7.242290748898678e-05,
+      "loss": 0.0642,
       "step": 1365
     },
     {
       "epoch": 12.53,
-      "learning_rate": 7.374449339207048e-05,
-      "loss": 1.5432,
+      "learning_rate": 7.215859030837004e-05,
+      "loss": 0.0495,
       "step": 1366
     },
     {
       "epoch": 12.54,
-      "learning_rate": 7.348017621145373e-05,
-      "loss": 1.497,
+      "learning_rate": 7.18942731277533e-05,
+      "loss": 0.052,
       "step": 1367
     },
     {
       "epoch": 12.55,
-      "learning_rate": 7.321585903083699e-05,
-      "loss": 1.5244,
+      "learning_rate": 7.162995594713656e-05,
+      "loss": 0.0544,
       "step": 1368
     },
     {
       "epoch": 12.56,
-      "learning_rate": 7.295154185022026e-05,
-      "loss": 1.483,
+      "learning_rate": 7.136563876651981e-05,
+      "loss": 0.0739,
       "step": 1369
     },
     {
       "epoch": 12.57,
-      "learning_rate": 7.268722466960352e-05,
-      "loss": 1.472,
+      "learning_rate": 7.110132158590308e-05,
+      "loss": 0.1053,
       "step": 1370
     },
     {
       "epoch": 12.58,
-      "learning_rate": 7.242290748898678e-05,
-      "loss": 1.6716,
+      "learning_rate": 7.083700440528633e-05,
+      "loss": 0.1097,
       "step": 1371
     },
     {
       "epoch": 12.59,
-      "learning_rate": 7.215859030837004e-05,
-      "loss": 1.5015,
+      "learning_rate": 7.05726872246696e-05,
+      "loss": 0.0714,
       "step": 1372
     },
     {
       "epoch": 12.6,
-      "learning_rate": 7.18942731277533e-05,
-      "loss": 1.4747,
+      "learning_rate": 7.030837004405286e-05,
+      "loss": 0.075,
       "step": 1373
     },
     {
       "epoch": 12.61,
-      "learning_rate": 7.162995594713656e-05,
-      "loss": 1.4988,
+      "learning_rate": 7.004405286343612e-05,
+      "loss": 0.0922,
       "step": 1374
     },
     {
       "epoch": 12.61,
-      "learning_rate": 7.136563876651981e-05,
-      "loss": 1.504,
+      "learning_rate": 6.977973568281938e-05,
+      "loss": 0.0845,
       "step": 1375
     },
     {
       "epoch": 12.62,
-      "learning_rate": 7.110132158590308e-05,
-      "loss": 1.4581,
+      "learning_rate": 6.951541850220265e-05,
+      "loss": 0.0936,
       "step": 1376
     },
     {
       "epoch": 12.63,
-      "learning_rate": 7.083700440528633e-05,
-      "loss": 1.5009,
+      "learning_rate": 6.925110132158589e-05,
+      "loss": 0.1342,
       "step": 1377
     },
     {
       "epoch": 12.64,
-      "learning_rate": 7.05726872246696e-05,
-      "loss": 1.5111,
+      "learning_rate": 6.898678414096915e-05,
+      "loss": 0.0383,
       "step": 1378
     },
     {
       "epoch": 12.65,
-      "learning_rate": 7.030837004405286e-05,
-      "loss": 1.5095,
+      "learning_rate": 6.872246696035242e-05,
+      "loss": 0.0901,
       "step": 1379
     },
     {
       "epoch": 12.66,
-      "learning_rate": 7.004405286343612e-05,
-      "loss": 1.4913,
+      "learning_rate": 6.845814977973568e-05,
+      "loss": 0.1072,
       "step": 1380
     },
     {
       "epoch": 12.67,
-      "learning_rate": 6.977973568281938e-05,
-      "loss": 1.5231,
+      "learning_rate": 6.819383259911894e-05,
+      "loss": 0.06,
       "step": 1381
     },
     {
       "epoch": 12.68,
-      "learning_rate": 6.951541850220265e-05,
-      "loss": 1.4415,
+      "learning_rate": 6.79295154185022e-05,
+      "loss": 0.1152,
       "step": 1382
     },
     {
       "epoch": 12.69,
-      "learning_rate": 6.925110132158589e-05,
-      "loss": 1.4707,
+      "learning_rate": 6.766519823788545e-05,
+      "loss": 0.0881,
       "step": 1383
     },
     {
       "epoch": 12.7,
-      "learning_rate": 6.898678414096915e-05,
-      "loss": 1.5285,
+      "learning_rate": 6.740088105726871e-05,
+      "loss": 0.0693,
       "step": 1384
     },
     {
       "epoch": 12.71,
-      "learning_rate": 6.872246696035242e-05,
-      "loss": 1.4876,
+      "learning_rate": 6.713656387665198e-05,
+      "loss": 0.0707,
       "step": 1385
     },
     {
       "epoch": 12.72,
-      "learning_rate": 6.845814977973568e-05,
-      "loss": 1.4891,
+      "learning_rate": 6.687224669603524e-05,
+      "loss": 0.0939,
       "step": 1386
     },
     {
       "epoch": 12.72,
-      "learning_rate": 6.819383259911894e-05,
-      "loss": 1.4758,
+      "learning_rate": 6.660792951541849e-05,
+      "loss": 0.046,
       "step": 1387
     },
     {
       "epoch": 12.73,
-      "learning_rate": 6.79295154185022e-05,
-      "loss": 1.4973,
+      "learning_rate": 6.634361233480176e-05,
+      "loss": 0.0852,
       "step": 1388
     },
     {
       "epoch": 12.74,
-      "learning_rate": 6.766519823788545e-05,
-      "loss": 1.501,
+      "learning_rate": 6.607929515418502e-05,
+      "loss": 0.0771,
       "step": 1389
     },
     {
       "epoch": 12.75,
-      "learning_rate": 6.740088105726871e-05,
-      "loss": 1.5019,
+      "learning_rate": 6.581497797356827e-05,
+      "loss": 0.0512,
       "step": 1390
     },
     {
       "epoch": 12.76,
-      "learning_rate": 6.713656387665198e-05,
-      "loss": 1.5108,
+      "learning_rate": 6.555066079295153e-05,
+      "loss": 0.0529,
       "step": 1391
     },
     {
       "epoch": 12.77,
-      "learning_rate": 6.687224669603524e-05,
-      "loss": 1.4814,
+      "learning_rate": 6.52863436123348e-05,
+      "loss": 0.0934,
       "step": 1392
     },
     {
       "epoch": 12.78,
-      "learning_rate": 6.660792951541849e-05,
-      "loss": 1.489,
+      "learning_rate": 6.502202643171805e-05,
+      "loss": 0.1117,
       "step": 1393
     },
     {
       "epoch": 12.79,
-      "learning_rate": 6.634361233480176e-05,
-      "loss": 1.437,
+      "learning_rate": 6.475770925110132e-05,
+      "loss": 0.0989,
       "step": 1394
     },
     {
       "epoch": 12.8,
-      "learning_rate": 6.607929515418502e-05,
-      "loss": 1.4809,
+      "learning_rate": 6.449339207048458e-05,
+      "loss": 0.107,
       "step": 1395
     },
     {
       "epoch": 12.81,
-      "learning_rate": 6.581497797356827e-05,
-      "loss": 1.5436,
+      "learning_rate": 6.422907488986784e-05,
+      "loss": 0.0736,
       "step": 1396
     },
     {
       "epoch": 12.82,
-      "learning_rate": 6.555066079295153e-05,
-      "loss": 1.494,
+      "learning_rate": 6.39647577092511e-05,
+      "loss": 0.1064,
       "step": 1397
     },
     {
       "epoch": 12.83,
-      "learning_rate": 6.52863436123348e-05,
-      "loss": 1.4668,
+      "learning_rate": 6.370044052863435e-05,
+      "loss": 0.0705,
       "step": 1398
     },
     {
       "epoch": 12.83,
-      "learning_rate": 6.502202643171805e-05,
-      "loss": 1.4604,
+      "learning_rate": 6.343612334801761e-05,
+      "loss": 0.0911,
       "step": 1399
     },
     {
       "epoch": 12.84,
-      "learning_rate": 6.475770925110132e-05,
-      "loss": 1.4852,
+      "learning_rate": 6.317180616740087e-05,
+      "loss": 0.1074,
       "step": 1400
     },
     {
       "epoch": 12.84,
-      "eval_loss": 0.40769898891448975,
-      "eval_runtime": 20.9928,
-      "eval_samples_per_second": 78.455,
-      "eval_steps_per_second": 1.239,
-      "eval_wer": 0.42825043407210706,
+      "eval_loss": 0.3926457166671753,
+      "eval_runtime": 21.1988,
+      "eval_samples_per_second": 77.693,
+      "eval_steps_per_second": 1.226,
+      "eval_wer": 0.3617608007353692,
       "step": 1400
     },
     {
       "epoch": 12.85,
-      "learning_rate": 6.449339207048458e-05,
-      "loss": 1.4956,
+      "learning_rate": 6.290748898678414e-05,
+      "loss": 0.1029,
       "step": 1401
     },
     {
       "epoch": 12.86,
-      "learning_rate": 6.422907488986784e-05,
-      "loss": 1.4979,
+      "learning_rate": 6.26431718061674e-05,
+      "loss": 0.0875,
       "step": 1402
     },
     {
       "epoch": 12.87,
-      "learning_rate": 6.39647577092511e-05,
-      "loss": 1.4797,
+      "learning_rate": 6.237885462555066e-05,
+      "loss": 0.0602,
       "step": 1403
     },
     {
       "epoch": 12.88,
-      "learning_rate": 6.370044052863435e-05,
-      "loss": 1.462,
+      "learning_rate": 6.211453744493392e-05,
+      "loss": 0.0829,
       "step": 1404
     },
     {
       "epoch": 12.89,
-      "learning_rate": 6.343612334801761e-05,
-      "loss": 1.5033,
+      "learning_rate": 6.185022026431717e-05,
+      "loss": 0.0624,
       "step": 1405
     },
     {
       "epoch": 12.9,
-      "learning_rate": 6.317180616740087e-05,
-      "loss": 1.4981,
+      "learning_rate": 6.158590308370043e-05,
+      "loss": 0.0922,
       "step": 1406
     },
     {
       "epoch": 12.91,
-      "learning_rate": 6.290748898678414e-05,
-      "loss": 1.479,
+      "learning_rate": 6.132158590308369e-05,
+      "loss": 0.0736,
       "step": 1407
     },
     {
       "epoch": 12.92,
-      "learning_rate": 6.26431718061674e-05,
-      "loss": 1.5963,
+      "learning_rate": 6.105726872246696e-05,
+      "loss": 0.1069,
       "step": 1408
     },
     {
       "epoch": 12.93,
-      "learning_rate": 6.237885462555066e-05,
-      "loss": 1.563,
+      "learning_rate": 6.0792951541850214e-05,
+      "loss": 0.0589,
       "step": 1409
     },
     {
       "epoch": 12.94,
-      "learning_rate": 6.211453744493392e-05,
-      "loss": 1.5173,
+      "learning_rate": 6.052863436123348e-05,
+      "loss": 0.0432,
       "step": 1410
     },
     {
       "epoch": 12.94,
-      "learning_rate": 6.185022026431717e-05,
-      "loss": 1.5251,
+      "learning_rate": 6.0264317180616737e-05,
+      "loss": 0.0759,
       "step": 1411
     },
     {
       "epoch": 12.95,
-      "learning_rate": 6.158590308370043e-05,
-      "loss": 1.4968,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 0.0788,
       "step": 1412
     },
     {
       "epoch": 12.96,
-      "learning_rate": 6.132158590308369e-05,
-      "loss": 1.4802,
+      "learning_rate": 5.973568281938325e-05,
+      "loss": 0.0862,
       "step": 1413
     },
     {
       "epoch": 12.97,
-      "learning_rate": 6.105726872246696e-05,
-      "loss": 1.4974,
+      "learning_rate": 5.947136563876651e-05,
+      "loss": 0.1013,
       "step": 1414
     },
     {
       "epoch": 12.98,
-      "learning_rate": 6.0792951541850214e-05,
-      "loss": 1.5357,
+      "learning_rate": 5.9207048458149776e-05,
+      "loss": 0.085,
       "step": 1415
     },
     {
       "epoch": 12.99,
-      "learning_rate": 6.052863436123348e-05,
-      "loss": 1.4987,
+      "learning_rate": 5.8942731277533034e-05,
+      "loss": 0.1192,
       "step": 1416
     },
     {
       "epoch": 13.0,
-      "learning_rate": 6.0264317180616737e-05,
-      "loss": 1.9541,
+      "learning_rate": 5.86784140969163e-05,
+      "loss": 0.1241,
       "step": 1417
     },
     {
       "epoch": 13.01,
-      "learning_rate": 5.9999999999999995e-05,
-      "loss": 1.5341,
+      "learning_rate": 5.841409691629955e-05,
+      "loss": 0.0705,
       "step": 1418
     },
     {
       "epoch": 13.02,
-      "learning_rate": 5.973568281938325e-05,
-      "loss": 1.5126,
+      "learning_rate": 5.8149779735682815e-05,
+      "loss": 0.0744,
       "step": 1419
     },
     {
       "epoch": 13.03,
-      "learning_rate": 5.947136563876651e-05,
-      "loss": 1.4718,
+      "learning_rate": 5.7885462555066073e-05,
+      "loss": 0.0638,
       "step": 1420
     },
     {
       "epoch": 13.04,
-      "learning_rate": 5.9207048458149776e-05,
-      "loss": 1.4824,
+      "learning_rate": 5.762114537444934e-05,
+      "loss": 0.1206,
       "step": 1421
     },
     {
       "epoch": 13.05,
-      "learning_rate": 5.8942731277533034e-05,
-      "loss": 1.453,
+      "learning_rate": 5.7356828193832597e-05,
+      "loss": 0.1191,
       "step": 1422
     },
     {
       "epoch": 13.06,
-      "learning_rate": 5.86784140969163e-05,
-      "loss": 1.4687,
+      "learning_rate": 5.709251101321586e-05,
+      "loss": 0.1695,
       "step": 1423
     },
     {
       "epoch": 13.06,
-      "learning_rate": 5.841409691629955e-05,
-      "loss": 1.5294,
+      "learning_rate": 5.682819383259911e-05,
+      "loss": 0.0543,
       "step": 1424
     },
     {
       "epoch": 13.07,
-      "learning_rate": 5.8149779735682815e-05,
-      "loss": 1.4715,
+      "learning_rate": 5.656387665198237e-05,
+      "loss": 0.0639,
       "step": 1425
     },
     {
       "epoch": 13.08,
-      "learning_rate": 5.7885462555066073e-05,
-      "loss": 1.4908,
+      "learning_rate": 5.6299559471365636e-05,
+      "loss": 0.1181,
       "step": 1426
     },
     {
       "epoch": 13.09,
-      "learning_rate": 5.762114537444934e-05,
-      "loss": 1.4764,
+      "learning_rate": 5.6035242290748894e-05,
+      "loss": 0.0891,
       "step": 1427
     },
     {
       "epoch": 13.1,
-      "learning_rate": 5.7356828193832597e-05,
-      "loss": 1.5062,
+      "learning_rate": 5.577092511013216e-05,
+      "loss": 0.1052,
       "step": 1428
     },
     {
       "epoch": 13.11,
-      "learning_rate": 5.709251101321586e-05,
-      "loss": 1.512,
+      "learning_rate": 5.550660792951541e-05,
+      "loss": 0.1114,
       "step": 1429
     },
     {
       "epoch": 13.12,
-      "learning_rate": 5.682819383259911e-05,
-      "loss": 1.534,
+      "learning_rate": 5.5242290748898675e-05,
+      "loss": 0.1153,
       "step": 1430
     },
     {
       "epoch": 13.13,
-      "learning_rate": 5.656387665198237e-05,
-      "loss": 1.4882,
+      "learning_rate": 5.497797356828193e-05,
+      "loss": 0.0727,
       "step": 1431
     },
     {
       "epoch": 13.14,
-      "learning_rate": 5.6299559471365636e-05,
-      "loss": 1.4592,
+      "learning_rate": 5.47136563876652e-05,
+      "loss": 0.0531,
       "step": 1432
     },
     {
       "epoch": 13.15,
-      "learning_rate": 5.6035242290748894e-05,
-      "loss": 1.4677,
+      "learning_rate": 5.4449339207048456e-05,
+      "loss": 0.083,
       "step": 1433
     },
     {
       "epoch": 13.16,
-      "learning_rate": 5.577092511013216e-05,
-      "loss": 1.4501,
+      "learning_rate": 5.418502202643171e-05,
+      "loss": 0.1119,
       "step": 1434
     },
     {
       "epoch": 13.17,
-      "learning_rate": 5.550660792951541e-05,
-      "loss": 1.4669,
+      "learning_rate": 5.392070484581497e-05,
+      "loss": 0.1323,
       "step": 1435
     },
     {
       "epoch": 13.17,
-      "learning_rate": 5.5242290748898675e-05,
-      "loss": 1.4568,
+      "learning_rate": 5.365638766519823e-05,
+      "loss": 0.0957,
       "step": 1436
     },
     {
       "epoch": 13.18,
-      "learning_rate": 5.497797356828193e-05,
-      "loss": 1.5195,
+      "learning_rate": 5.3392070484581496e-05,
+      "loss": 0.0792,
       "step": 1437
     },
     {
       "epoch": 13.19,
-      "learning_rate": 5.47136563876652e-05,
-      "loss": 1.4671,
+      "learning_rate": 5.3127753303964754e-05,
+      "loss": 0.0516,
       "step": 1438
     },
     {
       "epoch": 13.2,
-      "learning_rate": 5.4449339207048456e-05,
-      "loss": 1.4576,
+      "learning_rate": 5.286343612334802e-05,
+      "loss": 0.0712,
       "step": 1439
     },
     {
       "epoch": 13.21,
-      "learning_rate": 5.418502202643171e-05,
-      "loss": 1.4991,
+      "learning_rate": 5.259911894273127e-05,
+      "loss": 0.1206,
       "step": 1440
     },
     {
       "epoch": 13.22,
-      "learning_rate": 5.392070484581497e-05,
-      "loss": 1.4885,
+      "learning_rate": 5.2334801762114535e-05,
+      "loss": 0.0845,
       "step": 1441
     },
     {
       "epoch": 13.23,
-      "learning_rate": 5.365638766519823e-05,
-      "loss": 1.4851,
+      "learning_rate": 5.207048458149779e-05,
+      "loss": 0.1177,
       "step": 1442
     },
     {
       "epoch": 13.24,
-      "learning_rate": 5.3392070484581496e-05,
-      "loss": 1.4983,
+      "learning_rate": 5.180616740088105e-05,
+      "loss": 0.07,
       "step": 1443
     },
     {
       "epoch": 13.25,
-      "learning_rate": 5.3127753303964754e-05,
-      "loss": 1.479,
+      "learning_rate": 5.1541850220264316e-05,
+      "loss": 0.0582,
       "step": 1444
     },
     {
       "epoch": 13.26,
-      "learning_rate": 5.286343612334802e-05,
-      "loss": 1.4725,
+      "learning_rate": 5.127753303964757e-05,
+      "loss": 0.0897,
       "step": 1445
     },
     {
       "epoch": 13.27,
-      "learning_rate": 5.259911894273127e-05,
-      "loss": 1.4963,
+      "learning_rate": 5.101321585903083e-05,
+      "loss": 0.0982,
       "step": 1446
     },
     {
       "epoch": 13.28,
-      "learning_rate": 5.2334801762114535e-05,
-      "loss": 1.5176,
+      "learning_rate": 5.074889867841409e-05,
+      "loss": 0.0792,
       "step": 1447
     },
     {
       "epoch": 13.28,
-      "learning_rate": 5.207048458149779e-05,
-      "loss": 1.5161,
+      "learning_rate": 5.0484581497797356e-05,
+      "loss": 0.096,
       "step": 1448
     },
     {
       "epoch": 13.29,
-      "learning_rate": 5.180616740088105e-05,
-      "loss": 1.7019,
+      "learning_rate": 5.0220264317180614e-05,
+      "loss": 0.0888,
       "step": 1449
     },
     {
       "epoch": 13.3,
-      "learning_rate": 5.1541850220264316e-05,
-      "loss": 1.5181,
+      "learning_rate": 4.995594713656388e-05,
+      "loss": 0.0749,
       "step": 1450
     },
     {
       "epoch": 13.31,
-      "learning_rate": 5.127753303964757e-05,
-      "loss": 1.5238,
+      "learning_rate": 4.969162995594713e-05,
+      "loss": 0.0932,
       "step": 1451
     },
     {
       "epoch": 13.32,
-      "learning_rate": 5.101321585903083e-05,
-      "loss": 1.4692,
+      "learning_rate": 4.942731277533039e-05,
+      "loss": 0.0686,
       "step": 1452
     },
     {
       "epoch": 13.33,
-      "learning_rate": 5.074889867841409e-05,
-      "loss": 1.5011,
+      "learning_rate": 4.916299559471365e-05,
+      "loss": 0.0926,
       "step": 1453
     },
     {
       "epoch": 13.34,
-      "learning_rate": 5.0484581497797356e-05,
-      "loss": 1.5006,
+      "learning_rate": 4.889867841409691e-05,
+      "loss": 0.1093,
       "step": 1454
     },
     {
       "epoch": 13.35,
-      "learning_rate": 5.0220264317180614e-05,
-      "loss": 1.4952,
+      "learning_rate": 4.8634361233480176e-05,
+      "loss": 0.0715,
       "step": 1455
     },
     {
       "epoch": 13.36,
-      "learning_rate": 4.995594713656388e-05,
-      "loss": 1.4832,
+      "learning_rate": 4.837004405286343e-05,
+      "loss": 0.0563,
       "step": 1456
     },
     {
       "epoch": 13.37,
-      "learning_rate": 4.969162995594713e-05,
-      "loss": 1.5061,
+      "learning_rate": 4.810572687224669e-05,
+      "loss": 0.0645,
       "step": 1457
     },
     {
       "epoch": 13.38,
-      "learning_rate": 4.942731277533039e-05,
-      "loss": 1.4432,
+      "learning_rate": 4.784140969162995e-05,
+      "loss": 0.0663,
       "step": 1458
     },
     {
       "epoch": 13.39,
-      "learning_rate": 4.916299559471365e-05,
-      "loss": 1.4745,
+      "learning_rate": 4.7577092511013216e-05,
+      "loss": 0.0783,
       "step": 1459
     },
     {
       "epoch": 13.39,
-      "learning_rate": 4.889867841409691e-05,
-      "loss": 1.4222,
+      "learning_rate": 4.7312775330396474e-05,
+      "loss": 0.1201,
       "step": 1460
     },
     {
       "epoch": 13.4,
-      "learning_rate": 4.8634361233480176e-05,
-      "loss": 1.5123,
+      "learning_rate": 4.7048458149779725e-05,
+      "loss": 0.1478,
       "step": 1461
     },
     {
       "epoch": 13.41,
-      "learning_rate": 4.837004405286343e-05,
-      "loss": 1.4816,
+      "learning_rate": 4.678414096916299e-05,
+      "loss": 0.0513,
       "step": 1462
     },
     {
       "epoch": 13.42,
-      "learning_rate": 4.810572687224669e-05,
-      "loss": 1.4997,
+      "learning_rate": 4.651982378854625e-05,
+      "loss": 0.0752,
       "step": 1463
     },
     {
       "epoch": 13.43,
-      "learning_rate": 4.784140969162995e-05,
-      "loss": 1.4949,
+      "learning_rate": 4.625550660792951e-05,
+      "loss": 0.0629,
       "step": 1464
     },
     {
       "epoch": 13.44,
-      "learning_rate": 4.7577092511013216e-05,
-      "loss": 1.5357,
+      "learning_rate": 4.599118942731277e-05,
+      "loss": 0.0942,
       "step": 1465
     },
     {
       "epoch": 13.45,
-      "learning_rate": 4.7312775330396474e-05,
-      "loss": 1.4516,
+      "learning_rate": 4.5726872246696036e-05,
+      "loss": 0.0861,
       "step": 1466
     },
     {
       "epoch": 13.46,
-      "learning_rate": 4.7048458149779725e-05,
-      "loss": 1.5238,
+      "learning_rate": 4.546255506607929e-05,
+      "loss": 0.1298,
       "step": 1467
     },
     {
       "epoch": 13.47,
-      "learning_rate": 4.678414096916299e-05,
-      "loss": 1.5551,
+      "learning_rate": 4.519823788546255e-05,
+      "loss": 0.0574,
       "step": 1468
     },
     {
       "epoch": 13.48,
-      "learning_rate": 4.651982378854625e-05,
-      "loss": 1.4897,
+      "learning_rate": 4.493392070484581e-05,
+      "loss": 0.0848,
       "step": 1469
     },
     {
       "epoch": 13.49,
-      "learning_rate": 4.625550660792951e-05,
-      "loss": 1.4675,
+      "learning_rate": 4.4669603524229076e-05,
+      "loss": 0.1034,
       "step": 1470
     },
     {
       "epoch": 13.5,
-      "learning_rate": 4.599118942731277e-05,
-      "loss": 1.4645,
+      "learning_rate": 4.4405286343612334e-05,
+      "loss": 0.0673,
       "step": 1471
     },
     {
       "epoch": 13.5,
-      "learning_rate": 4.5726872246696036e-05,
-      "loss": 1.5253,
+      "learning_rate": 4.4140969162995585e-05,
+      "loss": 0.1241,
       "step": 1472
     },
     {
       "epoch": 13.51,
-      "learning_rate": 4.546255506607929e-05,
-      "loss": 1.4864,
+      "learning_rate": 4.387665198237885e-05,
+      "loss": 0.1121,
       "step": 1473
     },
     {
       "epoch": 13.52,
-      "learning_rate": 4.519823788546255e-05,
-      "loss": 1.5194,
+      "learning_rate": 4.361233480176211e-05,
+      "loss": 0.0975,
       "step": 1474
     },
     {
       "epoch": 13.53,
-      "learning_rate": 4.493392070484581e-05,
-      "loss": 1.456,
+      "learning_rate": 4.334801762114537e-05,
+      "loss": 0.0679,
       "step": 1475
     },
     {
       "epoch": 13.54,
-      "learning_rate": 4.4669603524229076e-05,
-      "loss": 1.481,
+      "learning_rate": 4.308370044052863e-05,
+      "loss": 0.0639,
       "step": 1476
     },
     {
       "epoch": 13.55,
-      "learning_rate": 4.4405286343612334e-05,
-      "loss": 1.5375,
+      "learning_rate": 4.2819383259911896e-05,
+      "loss": 0.0671,
       "step": 1477
     },
     {
       "epoch": 13.56,
-      "learning_rate": 4.4140969162995585e-05,
-      "loss": 1.4726,
+      "learning_rate": 4.255506607929515e-05,
+      "loss": 0.0584,
       "step": 1478
     },
     {
       "epoch": 13.57,
-      "learning_rate": 4.387665198237885e-05,
-      "loss": 1.5097,
+      "learning_rate": 4.229074889867841e-05,
+      "loss": 0.0966,
       "step": 1479
     },
     {
       "epoch": 13.58,
-      "learning_rate": 4.361233480176211e-05,
-      "loss": 1.4905,
+      "learning_rate": 4.202643171806167e-05,
+      "loss": 0.0964,
       "step": 1480
     },
     {
       "epoch": 13.59,
-      "learning_rate": 4.334801762114537e-05,
-      "loss": 1.514,
+      "learning_rate": 4.176211453744493e-05,
+      "loss": 0.0737,
       "step": 1481
     },
     {
       "epoch": 13.6,
-      "learning_rate": 4.308370044052863e-05,
-      "loss": 1.4854,
+      "learning_rate": 4.1497797356828194e-05,
+      "loss": 0.0551,
       "step": 1482
     },
     {
       "epoch": 13.61,
-      "learning_rate": 4.2819383259911896e-05,
-      "loss": 1.5097,
+      "learning_rate": 4.1233480176211445e-05,
+      "loss": 0.0905,
       "step": 1483
     },
     {
       "epoch": 13.61,
-      "learning_rate": 4.255506607929515e-05,
-      "loss": 1.4881,
+      "learning_rate": 4.096916299559471e-05,
+      "loss": 0.0629,
       "step": 1484
     },
     {
       "epoch": 13.62,
-      "learning_rate": 4.229074889867841e-05,
-      "loss": 1.427,
+      "learning_rate": 4.070484581497797e-05,
+      "loss": 0.1176,
       "step": 1485
     },
     {
       "epoch": 13.63,
-      "learning_rate": 4.202643171806167e-05,
-      "loss": 1.5309,
+      "learning_rate": 4.044052863436123e-05,
+      "loss": 0.0816,
       "step": 1486
     },
     {
       "epoch": 13.64,
-      "learning_rate": 4.176211453744493e-05,
-      "loss": 1.5017,
+      "learning_rate": 4.017621145374449e-05,
+      "loss": 0.0773,
       "step": 1487
     },
     {
       "epoch": 13.65,
-      "learning_rate": 4.1497797356828194e-05,
-      "loss": 1.4745,
+      "learning_rate": 3.9911894273127756e-05,
+      "loss": 0.1114,
       "step": 1488
     },
     {
       "epoch": 13.66,
-      "learning_rate": 4.1233480176211445e-05,
-      "loss": 1.4644,
+      "learning_rate": 3.964757709251101e-05,
+      "loss": 0.0774,
       "step": 1489
     },
     {
       "epoch": 13.67,
-      "learning_rate": 4.096916299559471e-05,
-      "loss": 1.4728,
+      "learning_rate": 3.9383259911894265e-05,
+      "loss": 0.0617,
       "step": 1490
     },
     {
       "epoch": 13.68,
-      "learning_rate": 4.070484581497797e-05,
-      "loss": 1.4631,
+      "learning_rate": 3.911894273127753e-05,
+      "loss": 0.1109,
       "step": 1491
     },
     {
       "epoch": 13.69,
-      "learning_rate": 4.044052863436123e-05,
-      "loss": 1.5044,
+      "learning_rate": 3.885462555066079e-05,
+      "loss": 0.128,
       "step": 1492
     },
     {
       "epoch": 13.7,
-      "learning_rate": 4.017621145374449e-05,
-      "loss": 1.5122,
+      "learning_rate": 3.8590308370044053e-05,
+      "loss": 0.067,
       "step": 1493
     },
     {
       "epoch": 13.71,
-      "learning_rate": 3.9911894273127756e-05,
-      "loss": 1.4658,
+      "learning_rate": 3.8325991189427305e-05,
+      "loss": 0.0573,
       "step": 1494
     },
     {
       "epoch": 13.72,
-      "learning_rate": 3.964757709251101e-05,
-      "loss": 1.4657,
+      "learning_rate": 3.806167400881057e-05,
+      "loss": 0.0664,
       "step": 1495
     },
     {
       "epoch": 13.72,
-      "learning_rate": 3.9383259911894265e-05,
-      "loss": 1.4691,
+      "learning_rate": 3.779735682819383e-05,
+      "loss": 0.0809,
       "step": 1496
     },
     {
       "epoch": 13.73,
-      "learning_rate": 3.911894273127753e-05,
-      "loss": 1.4515,
+      "learning_rate": 3.753303964757709e-05,
+      "loss": 0.0854,
       "step": 1497
     },
     {
       "epoch": 13.74,
-      "learning_rate": 3.885462555066079e-05,
-      "loss": 1.4578,
+      "learning_rate": 3.726872246696035e-05,
+      "loss": 0.1014,
       "step": 1498
     },
     {
       "epoch": 13.75,
-      "learning_rate": 3.8590308370044053e-05,
-      "loss": 1.4747,
+      "learning_rate": 3.700440528634361e-05,
+      "loss": 0.0587,
       "step": 1499
     },
     {
       "epoch": 13.76,
-      "learning_rate": 3.8325991189427305e-05,
-      "loss": 1.4843,
+      "learning_rate": 3.674008810572687e-05,
+      "loss": 0.0626,
       "step": 1500
     },
     {
       "epoch": 13.76,
-      "eval_loss": 0.3978853225708008,
-      "eval_runtime": 20.9676,
-      "eval_samples_per_second": 78.55,
-      "eval_steps_per_second": 1.24,
-      "eval_wer": 0.4155857420079665,
+      "eval_loss": 0.3855108618736267,
+      "eval_runtime": 21.2392,
+      "eval_samples_per_second": 77.545,
+      "eval_steps_per_second": 1.224,
+      "eval_wer": 0.3549177816361965,
       "step": 1500
     },
     {
       "epoch": 13.77,
-      "learning_rate": 3.806167400881057e-05,
-      "loss": 1.4807,
+      "learning_rate": 3.647577092511013e-05,
+      "loss": 0.0861,
       "step": 1501
     },
     {
       "epoch": 13.78,
-      "learning_rate": 3.779735682819383e-05,
-      "loss": 1.449,
+      "learning_rate": 3.621145374449339e-05,
+      "loss": 0.0489,
       "step": 1502
     },
     {
       "epoch": 13.79,
-      "learning_rate": 3.753303964757709e-05,
-      "loss": 1.4596,
+      "learning_rate": 3.594713656387665e-05,
+      "loss": 0.0971,
       "step": 1503
     },
     {
       "epoch": 13.8,
-      "learning_rate": 3.726872246696035e-05,
-      "loss": 1.5024,
+      "learning_rate": 3.568281938325991e-05,
+      "loss": 0.0904,
       "step": 1504
     },
     {
       "epoch": 13.81,
-      "learning_rate": 3.700440528634361e-05,
-      "loss": 1.7214,
+      "learning_rate": 3.5418502202643165e-05,
+      "loss": 0.0469,
       "step": 1505
     },
     {
       "epoch": 13.82,
-      "learning_rate": 3.674008810572687e-05,
-      "loss": 1.4686,
+      "learning_rate": 3.515418502202643e-05,
+      "loss": 0.064,
       "step": 1506
     },
     {
       "epoch": 13.83,
-      "learning_rate": 3.647577092511013e-05,
-      "loss": 1.4975,
+      "learning_rate": 3.488986784140969e-05,
+      "loss": 0.0371,
       "step": 1507
     },
     {
       "epoch": 13.83,
-      "learning_rate": 3.621145374449339e-05,
-      "loss": 1.4693,
+      "learning_rate": 3.4625550660792946e-05,
+      "loss": 0.0625,
       "step": 1508
     },
     {
       "epoch": 13.84,
-      "learning_rate": 3.594713656387665e-05,
-      "loss": 1.4859,
+      "learning_rate": 3.436123348017621e-05,
+      "loss": 0.0996,
       "step": 1509
     },
     {
       "epoch": 13.85,
-      "learning_rate": 3.568281938325991e-05,
-      "loss": 1.4867,
+      "learning_rate": 3.409691629955947e-05,
+      "loss": 0.0834,
       "step": 1510
     },
     {
       "epoch": 13.86,
-      "learning_rate": 3.5418502202643165e-05,
-      "loss": 1.5533,
+      "learning_rate": 3.383259911894273e-05,
+      "loss": 0.1429,
       "step": 1511
     },
     {
       "epoch": 13.87,
-      "learning_rate": 3.515418502202643e-05,
-      "loss": 1.4755,
+      "learning_rate": 3.356828193832599e-05,
+      "loss": 0.0737,
       "step": 1512
     },
     {
       "epoch": 13.88,
-      "learning_rate": 3.488986784140969e-05,
-      "loss": 1.4938,
+      "learning_rate": 3.3303964757709243e-05,
+      "loss": 0.0666,
       "step": 1513
     },
     {
       "epoch": 13.89,
-      "learning_rate": 3.4625550660792946e-05,
-      "loss": 1.4909,
+      "learning_rate": 3.303964757709251e-05,
+      "loss": 0.1045,
       "step": 1514
     },
     {
       "epoch": 13.9,
-      "learning_rate": 3.436123348017621e-05,
-      "loss": 1.5392,
+      "learning_rate": 3.2775330396475767e-05,
+      "loss": 0.0813,
       "step": 1515
     },
     {
       "epoch": 13.91,
-      "learning_rate": 3.409691629955947e-05,
-      "loss": 1.4822,
+      "learning_rate": 3.2511013215859025e-05,
+      "loss": 0.0813,
       "step": 1516
     },
     {
       "epoch": 13.92,
-      "learning_rate": 3.383259911894273e-05,
-      "loss": 1.4621,
+      "learning_rate": 3.224669603524229e-05,
+      "loss": 0.086,
       "step": 1517
     },
     {
       "epoch": 13.93,
-      "learning_rate": 3.356828193832599e-05,
-      "loss": 1.56,
+      "learning_rate": 3.198237885462555e-05,
+      "loss": 0.0834,
       "step": 1518
     },
     {
       "epoch": 13.94,
-      "learning_rate": 3.3303964757709243e-05,
-      "loss": 1.489,
+      "learning_rate": 3.1718061674008806e-05,
+      "loss": 0.0717,
       "step": 1519
     },
     {
       "epoch": 13.94,
-      "learning_rate": 3.303964757709251e-05,
-      "loss": 1.4878,
+      "learning_rate": 3.145374449339207e-05,
+      "loss": 0.0979,
       "step": 1520
     },
     {
       "epoch": 13.95,
-      "learning_rate": 3.2775330396475767e-05,
-      "loss": 1.5491,
+      "learning_rate": 3.118942731277533e-05,
+      "loss": 0.0542,
       "step": 1521
     },
     {
       "epoch": 13.96,
-      "learning_rate": 3.2511013215859025e-05,
-      "loss": 1.4727,
+      "learning_rate": 3.092511013215859e-05,
+      "loss": 0.0675,
       "step": 1522
     },
     {
       "epoch": 13.97,
-      "learning_rate": 3.224669603524229e-05,
-      "loss": 1.4504,
+      "learning_rate": 3.0660792951541845e-05,
+      "loss": 0.0983,
       "step": 1523
     },
     {
       "epoch": 13.98,
-      "learning_rate": 3.198237885462555e-05,
-      "loss": 1.5349,
+      "learning_rate": 3.0396475770925107e-05,
+      "loss": 0.065,
       "step": 1524
     },
     {
       "epoch": 13.99,
-      "learning_rate": 3.1718061674008806e-05,
-      "loss": 1.5072,
+      "learning_rate": 3.0132158590308368e-05,
+      "loss": 0.1228,
       "step": 1525
     },
     {
       "epoch": 14.0,
-      "learning_rate": 3.145374449339207e-05,
-      "loss": 1.9805,
+      "learning_rate": 2.9867841409691626e-05,
+      "loss": 0.075,
       "step": 1526
     },
     {
       "epoch": 14.01,
-      "learning_rate": 3.118942731277533e-05,
-      "loss": 1.5201,
+      "learning_rate": 2.9603524229074888e-05,
+      "loss": 0.0631,
       "step": 1527
     },
     {
       "epoch": 14.02,
-      "learning_rate": 3.092511013215859e-05,
-      "loss": 1.4773,
+      "learning_rate": 2.933920704845815e-05,
+      "loss": 0.0711,
       "step": 1528
     },
     {
       "epoch": 14.03,
-      "learning_rate": 3.0660792951541845e-05,
-      "loss": 1.4412,
+      "learning_rate": 2.9074889867841408e-05,
+      "loss": 0.0903,
       "step": 1529
     },
     {
       "epoch": 14.04,
-      "learning_rate": 3.0396475770925107e-05,
-      "loss": 1.4534,
+      "learning_rate": 2.881057268722467e-05,
+      "loss": 0.0709,
       "step": 1530
     },
     {
       "epoch": 14.05,
-      "learning_rate": 3.0132158590308368e-05,
-      "loss": 1.4556,
+      "learning_rate": 2.854625550660793e-05,
+      "loss": 0.1471,
       "step": 1531
     },
     {
       "epoch": 14.06,
-      "learning_rate": 2.9867841409691626e-05,
-      "loss": 1.4723,
+      "learning_rate": 2.8281938325991185e-05,
+      "loss": 0.1172,
       "step": 1532
     },
     {
       "epoch": 14.06,
-      "learning_rate": 2.9603524229074888e-05,
-      "loss": 1.6283,
+      "learning_rate": 2.8017621145374447e-05,
+      "loss": 0.0668,
       "step": 1533
     },
     {
       "epoch": 14.07,
-      "learning_rate": 2.933920704845815e-05,
-      "loss": 1.48,
+      "learning_rate": 2.7753303964757705e-05,
+      "loss": 0.0574,
       "step": 1534
     },
     {
       "epoch": 14.08,
-      "learning_rate": 2.9074889867841408e-05,
-      "loss": 1.4533,
+      "learning_rate": 2.7488986784140967e-05,
+      "loss": 0.0741,
       "step": 1535
     },
     {
       "epoch": 14.09,
-      "learning_rate": 2.881057268722467e-05,
-      "loss": 1.4636,
+      "learning_rate": 2.7224669603524228e-05,
+      "loss": 0.0842,
       "step": 1536
     },
     {
       "epoch": 14.1,
-      "learning_rate": 2.854625550660793e-05,
-      "loss": 1.4532,
+      "learning_rate": 2.6960352422907486e-05,
+      "loss": 0.0811,
       "step": 1537
     },
     {
       "epoch": 14.11,
-      "learning_rate": 2.8281938325991185e-05,
-      "loss": 1.4393,
+      "learning_rate": 2.6696035242290748e-05,
+      "loss": 0.1016,
       "step": 1538
     },
     {
       "epoch": 14.12,
-      "learning_rate": 2.8017621145374447e-05,
-      "loss": 1.552,
+      "learning_rate": 2.643171806167401e-05,
+      "loss": 0.0683,
       "step": 1539
     },
     {
       "epoch": 14.13,
-      "learning_rate": 2.7753303964757705e-05,
-      "loss": 1.4657,
+      "learning_rate": 2.6167400881057268e-05,
+      "loss": 0.0604,
       "step": 1540
     },
     {
       "epoch": 14.14,
-      "learning_rate": 2.7488986784140967e-05,
-      "loss": 1.4808,
+      "learning_rate": 2.5903083700440526e-05,
+      "loss": 0.0784,
       "step": 1541
     },
     {
       "epoch": 14.15,
-      "learning_rate": 2.7224669603524228e-05,
-      "loss": 1.4669,
+      "learning_rate": 2.5638766519823784e-05,
+      "loss": 0.0695,
       "step": 1542
     },
     {
       "epoch": 14.16,
-      "learning_rate": 2.6960352422907486e-05,
-      "loss": 1.4588,
+      "learning_rate": 2.5374449339207045e-05,
+      "loss": 0.0921,
       "step": 1543
     },
     {
       "epoch": 14.17,
-      "learning_rate": 2.6696035242290748e-05,
-      "loss": 1.4943,
+      "learning_rate": 2.5110132158590307e-05,
+      "loss": 0.1504,
       "step": 1544
     },
     {
       "epoch": 14.17,
-      "learning_rate": 2.643171806167401e-05,
-      "loss": 1.5422,
+      "learning_rate": 2.4845814977973565e-05,
+      "loss": 0.0956,
       "step": 1545
     },
     {
       "epoch": 14.18,
-      "learning_rate": 2.6167400881057268e-05,
-      "loss": 1.4966,
+      "learning_rate": 2.4581497797356827e-05,
+      "loss": 0.0606,
       "step": 1546
     },
     {
       "epoch": 14.19,
-      "learning_rate": 2.5903083700440526e-05,
-      "loss": 1.499,
+      "learning_rate": 2.4317180616740088e-05,
+      "loss": 0.0636,
       "step": 1547
     },
     {
       "epoch": 14.2,
-      "learning_rate": 2.5638766519823784e-05,
-      "loss": 1.4905,
+      "learning_rate": 2.4052863436123346e-05,
+      "loss": 0.0941,
       "step": 1548
     },
     {
       "epoch": 14.21,
-      "learning_rate": 2.5374449339207045e-05,
-      "loss": 1.4897,
+      "learning_rate": 2.3788546255506608e-05,
+      "loss": 0.0797,
       "step": 1549
     },
     {
       "epoch": 14.22,
-      "learning_rate": 2.5110132158590307e-05,
-      "loss": 1.4807,
+      "learning_rate": 2.3524229074889863e-05,
+      "loss": 0.0798,
       "step": 1550
     },
     {
       "epoch": 14.23,
-      "learning_rate": 2.4845814977973565e-05,
-      "loss": 1.4428,
+      "learning_rate": 2.3259911894273124e-05,
+      "loss": 0.1081,
       "step": 1551
     },
     {
       "epoch": 14.24,
-      "learning_rate": 2.4581497797356827e-05,
-      "loss": 1.5415,
+      "learning_rate": 2.2995594713656386e-05,
+      "loss": 0.0587,
       "step": 1552
     },
     {
       "epoch": 14.25,
-      "learning_rate": 2.4317180616740088e-05,
-      "loss": 1.4673,
+      "learning_rate": 2.2731277533039644e-05,
+      "loss": 0.0644,
       "step": 1553
     },
     {
       "epoch": 14.26,
-      "learning_rate": 2.4052863436123346e-05,
-      "loss": 1.4751,
+      "learning_rate": 2.2466960352422905e-05,
+      "loss": 0.0619,
       "step": 1554
     },
     {
       "epoch": 14.27,
-      "learning_rate": 2.3788546255506608e-05,
-      "loss": 1.4662,
+      "learning_rate": 2.2202643171806167e-05,
+      "loss": 0.0929,
       "step": 1555
     },
     {
       "epoch": 14.28,
-      "learning_rate": 2.3524229074889863e-05,
-      "loss": 1.4568,
+      "learning_rate": 2.1938325991189425e-05,
+      "loss": 0.1102,
       "step": 1556
     },
     {
       "epoch": 14.28,
-      "learning_rate": 2.3259911894273124e-05,
-      "loss": 1.4995,
+      "learning_rate": 2.1674008810572687e-05,
+      "loss": 0.0764,
       "step": 1557
     },
     {
       "epoch": 14.29,
-      "learning_rate": 2.3259911894273124e-05,
-      "loss": 1.6136,
+      "learning_rate": 2.1409691629955948e-05,
+      "loss": 0.091,
       "step": 1558
     },
     {
       "epoch": 14.3,
-      "learning_rate": 2.2995594713656386e-05,
-      "loss": 1.4834,
+      "learning_rate": 2.1145374449339206e-05,
+      "loss": 0.0308,
       "step": 1559
     },
     {
       "epoch": 14.31,
-      "learning_rate": 2.2731277533039644e-05,
-      "loss": 1.4505,
+      "learning_rate": 2.0881057268722464e-05,
+      "loss": 0.0834,
       "step": 1560
     },
     {
       "epoch": 14.32,
-      "learning_rate": 2.2466960352422905e-05,
-      "loss": 1.4459,
+      "learning_rate": 2.0616740088105722e-05,
+      "loss": 0.0994,
       "step": 1561
     },
     {
       "epoch": 14.33,
-      "learning_rate": 2.2202643171806167e-05,
-      "loss": 1.4566,
+      "learning_rate": 2.0352422907488984e-05,
+      "loss": 0.0796,
       "step": 1562
     },
     {
       "epoch": 14.34,
-      "learning_rate": 2.1938325991189425e-05,
-      "loss": 1.5203,
+      "learning_rate": 2.0088105726872246e-05,
+      "loss": 0.1123,
       "step": 1563
     },
     {
       "epoch": 14.35,
-      "learning_rate": 2.1674008810572687e-05,
-      "loss": 1.503,
+      "learning_rate": 1.9823788546255504e-05,
+      "loss": 0.1205,
       "step": 1564
     },
     {
       "epoch": 14.36,
-      "learning_rate": 2.1409691629955948e-05,
-      "loss": 1.4821,
+      "learning_rate": 1.9559471365638765e-05,
+      "loss": 0.0692,
       "step": 1565
     },
     {
       "epoch": 14.37,
-      "learning_rate": 2.1145374449339206e-05,
-      "loss": 1.4728,
+      "learning_rate": 1.9295154185022027e-05,
+      "loss": 0.0622,
       "step": 1566
     },
     {
       "epoch": 14.38,
-      "learning_rate": 2.0881057268722464e-05,
-      "loss": 1.4625,
+      "learning_rate": 1.9030837004405285e-05,
+      "loss": 0.0505,
       "step": 1567
     },
     {
       "epoch": 14.39,
-      "learning_rate": 2.0616740088105722e-05,
-      "loss": 1.4744,
+      "learning_rate": 1.8766519823788546e-05,
+      "loss": 0.0714,
       "step": 1568
     },
     {
       "epoch": 14.39,
-      "learning_rate": 2.0352422907488984e-05,
-      "loss": 1.4708,
+      "learning_rate": 1.8502202643171805e-05,
+      "loss": 0.0781,
       "step": 1569
     },
     {
       "epoch": 14.4,
-      "learning_rate": 2.0088105726872246e-05,
-      "loss": 1.5415,
+      "learning_rate": 1.8237885462555066e-05,
+      "loss": 0.1112,
       "step": 1570
     },
     {
       "epoch": 14.41,
-      "learning_rate": 1.9823788546255504e-05,
-      "loss": 1.477,
+      "learning_rate": 1.7973568281938324e-05,
+      "loss": 0.0626,
       "step": 1571
     },
     {
       "epoch": 14.42,
-      "learning_rate": 1.9559471365638765e-05,
-      "loss": 1.4894,
+      "learning_rate": 1.7709251101321582e-05,
+      "loss": 0.0567,
       "step": 1572
     },
     {
       "epoch": 14.43,
-      "learning_rate": 1.9295154185022027e-05,
-      "loss": 1.5156,
+      "learning_rate": 1.7444933920704844e-05,
+      "loss": 0.0589,
       "step": 1573
     },
     {
       "epoch": 14.44,
-      "learning_rate": 1.9030837004405285e-05,
-      "loss": 1.4633,
+      "learning_rate": 1.7180616740088105e-05,
+      "loss": 0.0768,
       "step": 1574
     },
     {
       "epoch": 14.45,
-      "learning_rate": 1.8766519823788546e-05,
-      "loss": 1.4774,
+      "learning_rate": 1.6916299559471364e-05,
+      "loss": 0.0675,
       "step": 1575
     },
     {
       "epoch": 14.46,
-      "learning_rate": 1.8502202643171805e-05,
-      "loss": 1.4512,
+      "learning_rate": 1.6651982378854622e-05,
+      "loss": 0.1247,
       "step": 1576
     },
     {
       "epoch": 14.47,
-      "learning_rate": 1.8237885462555066e-05,
-      "loss": 1.547,
+      "learning_rate": 1.6387665198237883e-05,
+      "loss": 0.058,
       "step": 1577
     },
     {
       "epoch": 14.48,
-      "learning_rate": 1.7973568281938324e-05,
-      "loss": 1.4858,
+      "learning_rate": 1.6123348017621145e-05,
+      "loss": 0.058,
       "step": 1578
     },
     {
       "epoch": 14.49,
-      "learning_rate": 1.7709251101321582e-05,
-      "loss": 1.5058,
+      "learning_rate": 1.5859030837004403e-05,
+      "loss": 0.0732,
       "step": 1579
     },
     {
       "epoch": 14.5,
-      "learning_rate": 1.7444933920704844e-05,
-      "loss": 1.506,
+      "learning_rate": 1.5594713656387664e-05,
+      "loss": 0.0917,
       "step": 1580
     },
     {
       "epoch": 14.5,
-      "learning_rate": 1.7180616740088105e-05,
-      "loss": 1.4962,
+      "learning_rate": 1.5330396475770923e-05,
+      "loss": 0.0833,
       "step": 1581
     },
     {
       "epoch": 14.51,
-      "learning_rate": 1.6916299559471364e-05,
-      "loss": 1.4682,
+      "learning_rate": 1.5066079295154184e-05,
+      "loss": 0.1035,
       "step": 1582
     },
     {
       "epoch": 14.52,
-      "learning_rate": 1.6651982378854622e-05,
-      "loss": 1.5919,
+      "learning_rate": 1.4801762114537444e-05,
+      "loss": 0.0728,
       "step": 1583
     },
     {
       "epoch": 14.53,
-      "learning_rate": 1.6387665198237883e-05,
-      "loss": 1.5002,
+      "learning_rate": 1.4537444933920704e-05,
+      "loss": 0.0332,
       "step": 1584
     },
     {
       "epoch": 14.54,
-      "learning_rate": 1.6123348017621145e-05,
-      "loss": 1.4964,
+      "learning_rate": 1.4273127753303965e-05,
+      "loss": 0.05,
       "step": 1585
     },
     {
       "epoch": 14.55,
-      "learning_rate": 1.5859030837004403e-05,
-      "loss": 1.4897,
+      "learning_rate": 1.4008810572687224e-05,
+      "loss": 0.0695,
       "step": 1586
     },
     {
       "epoch": 14.56,
-      "learning_rate": 1.5594713656387664e-05,
-      "loss": 1.4735,
+      "learning_rate": 1.3744493392070483e-05,
+      "loss": 0.0585,
       "step": 1587
     },
     {
       "epoch": 14.57,
-      "learning_rate": 1.5330396475770923e-05,
-      "loss": 1.4534,
+      "learning_rate": 1.3480176211453743e-05,
+      "loss": 0.0883,
       "step": 1588
     },
     {
       "epoch": 14.58,
-      "learning_rate": 1.5066079295154184e-05,
-      "loss": 1.5232,
+      "learning_rate": 1.3215859030837005e-05,
+      "loss": 0.0768,
       "step": 1589
     },
     {
       "epoch": 14.59,
-      "learning_rate": 1.4801762114537444e-05,
-      "loss": 1.5007,
+      "learning_rate": 1.2951541850220263e-05,
+      "loss": 0.0621,
       "step": 1590
     },
     {
       "epoch": 14.6,
-      "learning_rate": 1.4537444933920704e-05,
-      "loss": 1.4783,
+      "learning_rate": 1.2687224669603523e-05,
+      "loss": 0.0465,
       "step": 1591
     },
     {
       "epoch": 14.61,
-      "learning_rate": 1.4273127753303965e-05,
-      "loss": 1.4968,
+      "learning_rate": 1.2422907488986783e-05,
+      "loss": 0.0535,
       "step": 1592
     },
     {
       "epoch": 14.61,
-      "learning_rate": 1.4008810572687224e-05,
-      "loss": 1.4699,
+      "learning_rate": 1.2158590308370044e-05,
+      "loss": 0.0762,
       "step": 1593
     },
     {
       "epoch": 14.62,
-      "learning_rate": 1.3744493392070483e-05,
-      "loss": 1.4453,
+      "learning_rate": 1.1894273127753304e-05,
+      "loss": 0.0649,
       "step": 1594
     },
     {
       "epoch": 14.63,
-      "learning_rate": 1.3480176211453743e-05,
-      "loss": 1.432,
+      "learning_rate": 1.1629955947136562e-05,
+      "loss": 0.1025,
       "step": 1595
     },
     {
       "epoch": 14.64,
-      "learning_rate": 1.3215859030837005e-05,
-      "loss": 1.5281,
+      "learning_rate": 1.1365638766519822e-05,
+      "loss": 0.0999,
       "step": 1596
     },
     {
       "epoch": 14.65,
-      "learning_rate": 1.2951541850220263e-05,
-      "loss": 1.4998,
+      "learning_rate": 1.1101321585903083e-05,
+      "loss": 0.0421,
       "step": 1597
     },
     {
       "epoch": 14.66,
-      "learning_rate": 1.2687224669603523e-05,
-      "loss": 1.4647,
+      "learning_rate": 1.0837004405286343e-05,
+      "loss": 0.0392,
       "step": 1598
     },
     {
       "epoch": 14.67,
-      "learning_rate": 1.2422907488986783e-05,
-      "loss": 1.4877,
+      "learning_rate": 1.0572687224669603e-05,
+      "loss": 0.094,
       "step": 1599
     },
     {
       "epoch": 14.68,
-      "learning_rate": 1.2158590308370044e-05,
-      "loss": 1.5273,
+      "learning_rate": 1.0308370044052861e-05,
+      "loss": 0.1026,
       "step": 1600
     },
     {
       "epoch": 14.68,
-      "eval_loss": 0.3971070349216461,
-      "eval_runtime": 21.0663,
-      "eval_samples_per_second": 78.182,
-      "eval_steps_per_second": 1.234,
-      "eval_wer": 0.4182412419568992,
+      "eval_loss": 0.3872491419315338,
+      "eval_runtime": 22.349,
+      "eval_samples_per_second": 73.695,
+      "eval_steps_per_second": 1.163,
+      "eval_wer": 0.350628127872536,
       "step": 1600
     },
     {
       "epoch": 14.69,
-      "learning_rate": 1.1894273127753304e-05,
-      "loss": 1.4634,
+      "learning_rate": 1.0044052863436123e-05,
+      "loss": 0.0907,
       "step": 1601
     },
     {
       "epoch": 14.7,
-      "learning_rate": 1.1629955947136562e-05,
-      "loss": 1.5482,
+      "learning_rate": 9.779735682819383e-06,
+      "loss": 0.0632,
       "step": 1602
     },
     {
       "epoch": 14.71,
-      "learning_rate": 1.1365638766519822e-05,
-      "loss": 1.4853,
+      "learning_rate": 9.515418502202642e-06,
+      "loss": 0.0581,
       "step": 1603
     },
     {
       "epoch": 14.72,
-      "learning_rate": 1.1101321585903083e-05,
-      "loss": 1.5104,
+      "learning_rate": 9.251101321585902e-06,
+      "loss": 0.0925,
       "step": 1604
     },
     {
       "epoch": 14.72,
-      "learning_rate": 1.0837004405286343e-05,
-      "loss": 1.4923,
+      "learning_rate": 8.986784140969162e-06,
+      "loss": 0.0775,
       "step": 1605
     },
     {
       "epoch": 14.73,
-      "learning_rate": 1.0572687224669603e-05,
-      "loss": 1.5053,
+      "learning_rate": 8.722466960352422e-06,
+      "loss": 0.0449,
       "step": 1606
     },
     {
       "epoch": 14.74,
-      "learning_rate": 1.0308370044052861e-05,
-      "loss": 1.4633,
+      "learning_rate": 8.458149779735682e-06,
+      "loss": 0.1388,
       "step": 1607
     },
     {
       "epoch": 14.75,
-      "learning_rate": 1.0044052863436123e-05,
-      "loss": 1.5385,
+      "learning_rate": 8.193832599118942e-06,
+      "loss": 0.0296,
       "step": 1608
     },
     {
       "epoch": 14.76,
-      "learning_rate": 9.779735682819383e-06,
-      "loss": 1.4954,
+      "learning_rate": 7.929515418502201e-06,
+      "loss": 0.0694,
       "step": 1609
     },
     {
       "epoch": 14.77,
-      "learning_rate": 9.515418502202642e-06,
-      "loss": 1.4997,
+      "learning_rate": 7.665198237885461e-06,
+      "loss": 0.1095,
       "step": 1610
     },
     {
       "epoch": 14.78,
-      "learning_rate": 9.251101321585902e-06,
-      "loss": 1.4769,
+      "learning_rate": 7.400881057268722e-06,
+      "loss": 0.1115,
       "step": 1611
     },
     {
       "epoch": 14.79,
-      "learning_rate": 8.986784140969162e-06,
-      "loss": 1.4584,
+      "learning_rate": 7.136563876651983e-06,
+      "loss": 0.0552,
       "step": 1612
     },
     {
       "epoch": 14.8,
-      "learning_rate": 8.722466960352422e-06,
-      "loss": 1.5095,
+      "learning_rate": 6.872246696035242e-06,
+      "loss": 0.1568,
       "step": 1613
     },
     {
       "epoch": 14.81,
-      "learning_rate": 8.458149779735682e-06,
-      "loss": 1.5448,
+      "learning_rate": 6.607929515418502e-06,
+      "loss": 0.1188,
       "step": 1614
     },
     {
       "epoch": 14.82,
-      "learning_rate": 8.193832599118942e-06,
-      "loss": 1.4963,
+      "learning_rate": 6.343612334801761e-06,
+      "loss": 0.0586,
       "step": 1615
     },
     {
       "epoch": 14.83,
-      "learning_rate": 7.929515418502201e-06,
-      "loss": 1.4533,
+      "learning_rate": 6.079295154185022e-06,
+      "loss": 0.1038,
       "step": 1616
     },
     {
       "epoch": 14.83,
-      "learning_rate": 7.665198237885461e-06,
-      "loss": 1.4947,
+      "learning_rate": 5.814977973568281e-06,
+      "loss": 0.0311,
       "step": 1617
     },
     {
       "epoch": 14.84,
-      "learning_rate": 7.400881057268722e-06,
-      "loss": 1.4949,
+      "learning_rate": 5.550660792951542e-06,
+      "loss": 0.0831,
       "step": 1618
     },
     {
       "epoch": 14.85,
-      "learning_rate": 7.136563876651983e-06,
-      "loss": 1.4551,
+      "learning_rate": 5.2863436123348015e-06,
+      "loss": 0.079,
       "step": 1619
     },
     {
       "epoch": 14.86,
-      "learning_rate": 6.872246696035242e-06,
-      "loss": 1.7636,
+      "learning_rate": 5.022026431718061e-06,
+      "loss": 0.1095,
       "step": 1620
     },
     {
       "epoch": 14.87,
-      "learning_rate": 6.607929515418502e-06,
-      "loss": 1.5273,
+      "learning_rate": 4.757709251101321e-06,
+      "loss": 0.0673,
       "step": 1621
     },
     {
       "epoch": 14.88,
-      "learning_rate": 6.343612334801761e-06,
-      "loss": 1.5131,
+      "learning_rate": 4.493392070484581e-06,
+      "loss": 0.0778,
       "step": 1622
     },
     {
       "epoch": 14.89,
-      "learning_rate": 6.079295154185022e-06,
-      "loss": 1.4993,
+      "learning_rate": 4.229074889867841e-06,
+      "loss": 0.0852,
       "step": 1623
     },
     {
       "epoch": 14.9,
-      "learning_rate": 5.814977973568281e-06,
-      "loss": 1.4614,
+      "learning_rate": 3.964757709251101e-06,
+      "loss": 0.0879,
       "step": 1624
     },
     {
       "epoch": 14.91,
-      "learning_rate": 5.550660792951542e-06,
-      "loss": 1.4642,
+      "learning_rate": 3.700440528634361e-06,
+      "loss": 0.1133,
       "step": 1625
     },
     {
       "epoch": 14.92,
-      "learning_rate": 5.2863436123348015e-06,
-      "loss": 1.4975,
+      "learning_rate": 3.436123348017621e-06,
+      "loss": 0.0932,
       "step": 1626
     },
     {
       "epoch": 14.93,
-      "learning_rate": 5.022026431718061e-06,
-      "loss": 1.5468,
+      "learning_rate": 3.1718061674008807e-06,
+      "loss": 0.0496,
       "step": 1627
     },
     {
       "epoch": 14.94,
-      "learning_rate": 4.757709251101321e-06,
-      "loss": 1.4468,
+      "learning_rate": 2.9074889867841405e-06,
+      "loss": 0.0493,
       "step": 1628
     },
     {
       "epoch": 14.94,
-      "learning_rate": 4.493392070484581e-06,
-      "loss": 1.4752,
+      "learning_rate": 2.6431718061674008e-06,
+      "loss": 0.0337,
       "step": 1629
     },
     {
       "epoch": 14.95,
-      "learning_rate": 4.229074889867841e-06,
-      "loss": 1.5024,
+      "learning_rate": 2.3788546255506606e-06,
+      "loss": 0.0782,
       "step": 1630
     },
     {
       "epoch": 14.96,
-      "learning_rate": 3.964757709251101e-06,
-      "loss": 1.5055,
+      "learning_rate": 2.1145374449339204e-06,
+      "loss": 0.0856,
       "step": 1631
     },
     {
       "epoch": 14.97,
-      "learning_rate": 3.700440528634361e-06,
-      "loss": 1.4639,
+      "learning_rate": 1.8502202643171805e-06,
+      "loss": 0.0924,
       "step": 1632
     },
     {
       "epoch": 14.98,
-      "learning_rate": 3.436123348017621e-06,
-      "loss": 1.624,
+      "learning_rate": 1.5859030837004403e-06,
+      "loss": 0.1104,
       "step": 1633
     },
     {
       "epoch": 14.99,
-      "learning_rate": 3.1718061674008807e-06,
-      "loss": 1.4662,
+      "learning_rate": 1.3215859030837004e-06,
+      "loss": 0.0893,
       "step": 1634
     },
     {
       "epoch": 15.0,
-      "learning_rate": 2.9074889867841405e-06,
-      "loss": 1.9516,
+      "learning_rate": 1.0572687224669602e-06,
+      "loss": 0.0913,
       "step": 1635
     },
     {
       "epoch": 15.0,
       "step": 1635,
       "total_flos": 7.077752219977646e+18,
-      "train_loss": 2.507998379363197,
-      "train_runtime": 1214.3899,
-      "train_samples_per_second": 42.96,
-      "train_steps_per_second": 1.346
+      "train_loss": 1.095775873795015,
+      "train_runtime": 1262.0742,
+      "train_samples_per_second": 41.337,
+      "train_steps_per_second": 1.295
     }
   ],
   "max_steps": 1635,