diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,14260 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.998632010943912,
+  "global_step": 1825,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.7499999999999997e-06,
+      "loss": 11.6636,
+      "step": 1
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.499999999999999e-06,
+      "loss": 10.7654,
+      "step": 2
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.1249999999999999e-05,
+      "loss": 10.5427,
+      "step": 3
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 11.7261,
+      "step": 4
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.875e-05,
+      "loss": 10.7838,
+      "step": 5
+    },
+    {
+      "epoch": 0.01,
+      "eval_loss": 14.503458976745605,
+      "eval_runtime": 140.7504,
+      "eval_samples_per_second": 34.408,
+      "eval_steps_per_second": 4.305,
+      "eval_wer": 1.0,
+      "step": 5
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.2499999999999998e-05,
+      "loss": 12.4901,
+      "step": 6
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.6249999999999998e-05,
+      "loss": 12.733,
+      "step": 7
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 12.1247,
+      "step": 8
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 3.375e-05,
+      "loss": 11.407,
+      "step": 9
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 3.75e-05,
+      "loss": 13.0582,
+      "step": 10
+    },
+    {
+      "epoch": 0.03,
+      "eval_loss": 13.665786743164062,
+      "eval_runtime": 138.4673,
+      "eval_samples_per_second": 34.976,
+      "eval_steps_per_second": 4.376,
+      "eval_wer": 1.0,
+      "step": 10
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.125e-05,
+      "loss": 8.5172,
+      "step": 11
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 9.4759,
+      "step": 12
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 4.875e-05,
+      "loss": 8.9879,
+      "step": 13
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.2499999999999995e-05,
+      "loss": 8.1922,
+      "step": 14
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.625e-05,
+      "loss": 7.3034,
+      "step": 15
+    },
+    {
+      "epoch": 0.04,
+      "eval_loss": 9.789834022521973,
+      "eval_runtime": 141.1991,
+      "eval_samples_per_second": 34.299,
+      "eval_steps_per_second": 4.292,
+      "eval_wer": 1.0,
+      "step": 15
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 5.625e-05,
+      "loss": 8.6055,
+      "step": 16
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 8.7149,
+      "step": 17
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.374999999999999e-05,
+      "loss": 7.4237,
+      "step": 18
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 6.75e-05,
+      "loss": 7.4301,
+      "step": 19
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 7.125e-05,
+      "loss": 6.1847,
+      "step": 20
+    },
+    {
+      "epoch": 0.05,
+      "eval_loss": 6.914824485778809,
+      "eval_runtime": 138.4688,
+      "eval_samples_per_second": 34.975,
+      "eval_steps_per_second": 4.376,
+      "eval_wer": 1.0,
+      "step": 20
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.5e-05,
+      "loss": 6.0554,
+      "step": 21
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.49584487534626e-05,
+      "loss": 5.8687,
+      "step": 22
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 7.491689750692521e-05,
+      "loss": 5.483,
+      "step": 23
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.48753462603878e-05,
+      "loss": 5.2607,
+      "step": 24
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.483379501385041e-05,
+      "loss": 5.3371,
+      "step": 25
+    },
+    {
+      "epoch": 0.07,
+      "eval_loss": 5.366059303283691,
+      "eval_runtime": 138.1639,
+      "eval_samples_per_second": 35.053,
+      "eval_steps_per_second": 4.386,
+      "eval_wer": 1.0,
+      "step": 25
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.479224376731301e-05,
+      "loss": 5.3036,
+      "step": 26
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 7.475069252077562e-05,
+      "loss": 4.8766,
+      "step": 27
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.470914127423822e-05,
+      "loss": 5.6001,
+      "step": 28
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.466759002770082e-05,
+      "loss": 4.7061,
+      "step": 29
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.462603878116343e-05,
+      "loss": 4.4274,
+      "step": 30
+    },
+    {
+      "epoch": 0.08,
+      "eval_loss": 4.694497108459473,
+      "eval_runtime": 138.2203,
+      "eval_samples_per_second": 35.038,
+      "eval_steps_per_second": 4.384,
+      "eval_wer": 1.0,
+      "step": 30
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 7.458448753462604e-05,
+      "loss": 4.6797,
+      "step": 31
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.454293628808864e-05,
+      "loss": 4.7467,
+      "step": 32
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.450138504155124e-05,
+      "loss": 4.1208,
+      "step": 33
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 7.445983379501384e-05,
+      "loss": 4.4005,
+      "step": 34
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 7.441828254847646e-05,
+      "loss": 4.0918,
+      "step": 35
+    },
+    {
+      "epoch": 0.1,
+      "eval_loss": 4.317225456237793,
+      "eval_runtime": 138.5437,
+      "eval_samples_per_second": 34.956,
+      "eval_steps_per_second": 4.374,
+      "eval_wer": 1.0,
+      "step": 35
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 7.437673130193904e-05,
+      "loss": 4.1334,
+      "step": 36
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 7.433518005540166e-05,
+      "loss": 4.1134,
+      "step": 37
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 7.429362880886426e-05,
+      "loss": 3.8943,
+      "step": 38
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.425207756232686e-05,
+      "loss": 4.1369,
+      "step": 39
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.421052631578946e-05,
+      "loss": 4.1734,
+      "step": 40
+    },
+    {
+      "epoch": 0.11,
+      "eval_loss": 4.075934886932373,
+      "eval_runtime": 147.7942,
+      "eval_samples_per_second": 32.769,
+      "eval_steps_per_second": 4.1,
+      "eval_wer": 1.0,
+      "step": 40
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.416897506925206e-05,
+      "loss": 3.8944,
+      "step": 41
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 7.412742382271468e-05,
+      "loss": 3.9561,
+      "step": 42
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.408587257617728e-05,
+      "loss": 3.8521,
+      "step": 43
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.404432132963988e-05,
+      "loss": 3.8403,
+      "step": 44
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 7.400277008310248e-05,
+      "loss": 3.7332,
+      "step": 45
+    },
+    {
+      "epoch": 0.12,
+      "eval_loss": 3.9038915634155273,
+      "eval_runtime": 141.3706,
+      "eval_samples_per_second": 34.257,
+      "eval_steps_per_second": 4.287,
+      "eval_wer": 1.0,
+      "step": 45
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.396121883656509e-05,
+      "loss": 3.6091,
+      "step": 46
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.39196675900277e-05,
+      "loss": 3.6816,
+      "step": 47
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.38781163434903e-05,
+      "loss": 3.7744,
+      "step": 48
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.38365650969529e-05,
+      "loss": 3.7214,
+      "step": 49
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.37950138504155e-05,
+      "loss": 3.6871,
+      "step": 50
+    },
+    {
+      "epoch": 0.14,
+      "eval_loss": 3.7777292728424072,
+      "eval_runtime": 138.4074,
+      "eval_samples_per_second": 34.991,
+      "eval_steps_per_second": 4.378,
+      "eval_wer": 1.0,
+      "step": 50
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.37534626038781e-05,
+      "loss": 3.5669,
+      "step": 51
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.371191135734071e-05,
+      "loss": 3.4696,
+      "step": 52
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 7.367036011080332e-05,
+      "loss": 3.4967,
+      "step": 53
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 7.362880886426592e-05,
+      "loss": 3.6514,
+      "step": 54
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 7.358725761772853e-05,
+      "loss": 3.4428,
+      "step": 55
+    },
+    {
+      "epoch": 0.15,
+      "eval_loss": 3.6718008518218994,
+      "eval_runtime": 145.2309,
+      "eval_samples_per_second": 33.347,
+      "eval_steps_per_second": 4.173,
+      "eval_wer": 1.0,
+      "step": 55
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 7.354570637119113e-05,
+      "loss": 3.443,
+      "step": 56
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.350415512465373e-05,
+      "loss": 3.4773,
+      "step": 57
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.346260387811634e-05,
+      "loss": 3.5355,
+      "step": 58
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.342105263157895e-05,
+      "loss": 3.3728,
+      "step": 59
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 7.337950138504155e-05,
+      "loss": 3.5514,
+      "step": 60
+    },
+    {
+      "epoch": 0.16,
+      "eval_loss": 3.5946505069732666,
+      "eval_runtime": 137.3959,
+      "eval_samples_per_second": 35.249,
+      "eval_steps_per_second": 4.411,
+      "eval_wer": 1.0,
+      "step": 60
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.333795013850415e-05,
+      "loss": 3.3587,
+      "step": 61
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.329639889196675e-05,
+      "loss": 3.4299,
+      "step": 62
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.325484764542937e-05,
+      "loss": 3.4615,
+      "step": 63
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.321329639889195e-05,
+      "loss": 3.4486,
+      "step": 64
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.317174515235457e-05,
+      "loss": 3.4307,
+      "step": 65
+    },
+    {
+      "epoch": 0.18,
+      "eval_loss": 3.514420747756958,
+      "eval_runtime": 151.8205,
+      "eval_samples_per_second": 31.9,
+      "eval_steps_per_second": 3.992,
+      "eval_wer": 1.0,
+      "step": 65
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.313019390581717e-05,
+      "loss": 3.631,
+      "step": 66
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 7.308864265927977e-05,
+      "loss": 3.4752,
+      "step": 67
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 7.304709141274237e-05,
+      "loss": 3.4899,
+      "step": 68
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 7.300554016620497e-05,
+      "loss": 3.3307,
+      "step": 69
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 7.296398891966759e-05,
+      "loss": 3.4102,
+      "step": 70
+    },
+    {
+      "epoch": 0.19,
+      "eval_loss": 3.4432265758514404,
+      "eval_runtime": 154.1428,
+      "eval_samples_per_second": 31.419,
+      "eval_steps_per_second": 3.931,
+      "eval_wer": 1.0,
+      "step": 70
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 7.292243767313019e-05,
+      "loss": 3.405,
+      "step": 71
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 7.288088642659279e-05,
+      "loss": 3.4745,
+      "step": 72
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 7.28393351800554e-05,
+      "loss": 3.4698,
+      "step": 73
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 7.2797783933518e-05,
+      "loss": 3.3873,
+      "step": 74
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 7.275623268698061e-05,
+      "loss": 3.4964,
+      "step": 75
+    },
+    {
+      "epoch": 0.21,
+      "eval_loss": 3.38900089263916,
+      "eval_runtime": 139.6645,
+      "eval_samples_per_second": 34.676,
+      "eval_steps_per_second": 4.339,
+      "eval_wer": 1.0,
+      "step": 75
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 7.27146814404432e-05,
+      "loss": 3.3407,
+      "step": 76
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 7.267313019390581e-05,
+      "loss": 3.3169,
+      "step": 77
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 7.263157894736842e-05,
+      "loss": 3.3345,
+      "step": 78
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.259002770083102e-05,
+      "loss": 3.3568,
+      "step": 79
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.254847645429362e-05,
+      "loss": 3.3936,
+      "step": 80
+    },
+    {
+      "epoch": 0.22,
+      "eval_loss": 3.3467209339141846,
+      "eval_runtime": 138.3392,
+      "eval_samples_per_second": 35.008,
+      "eval_steps_per_second": 4.381,
+      "eval_wer": 1.0,
+      "step": 80
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.250692520775622e-05,
+      "loss": 3.3996,
+      "step": 81
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 7.246537396121884e-05,
+      "loss": 3.3694,
+      "step": 82
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.242382271468144e-05,
+      "loss": 3.2897,
+      "step": 83
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.238227146814404e-05,
+      "loss": 3.2832,
+      "step": 84
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 7.234072022160664e-05,
+      "loss": 3.3051,
+      "step": 85
+    },
+    {
+      "epoch": 0.23,
+      "eval_loss": 3.3102469444274902,
+      "eval_runtime": 141.6188,
+      "eval_samples_per_second": 34.197,
+      "eval_steps_per_second": 4.279,
+      "eval_wer": 1.0,
+      "step": 85
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.229916897506924e-05,
+      "loss": 3.244,
+      "step": 86
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.225761772853186e-05,
+      "loss": 3.3285,
+      "step": 87
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.221606648199444e-05,
+      "loss": 3.2418,
+      "step": 88
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 7.217451523545706e-05,
+      "loss": 3.2292,
+      "step": 89
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.213296398891966e-05,
+      "loss": 3.278,
+      "step": 90
+    },
+    {
+      "epoch": 0.25,
+      "eval_loss": 3.2800509929656982,
+      "eval_runtime": 139.181,
+      "eval_samples_per_second": 34.796,
+      "eval_steps_per_second": 4.354,
+      "eval_wer": 1.0,
+      "step": 90
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.209141274238226e-05,
+      "loss": 3.1962,
+      "step": 91
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.204986149584486e-05,
+      "loss": 3.1822,
+      "step": 92
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.200831024930747e-05,
+      "loss": 3.2192,
+      "step": 93
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.196675900277008e-05,
+      "loss": 3.1372,
+      "step": 94
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.192520775623268e-05,
+      "loss": 3.2223,
+      "step": 95
+    },
+    {
+      "epoch": 0.26,
+      "eval_loss": 3.243964672088623,
+      "eval_runtime": 138.8764,
+      "eval_samples_per_second": 34.873,
+      "eval_steps_per_second": 4.364,
+      "eval_wer": 1.0,
+      "step": 95
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.188365650969528e-05,
+      "loss": 3.2247,
+      "step": 96
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.184210526315788e-05,
+      "loss": 3.1752,
+      "step": 97
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.180055401662049e-05,
+      "loss": 3.1476,
+      "step": 98
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.17590027700831e-05,
+      "loss": 3.2827,
+      "step": 99
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.17174515235457e-05,
+      "loss": 3.1888,
+      "step": 100
+    },
+    {
+      "epoch": 0.27,
+      "eval_loss": 3.2899622917175293,
+      "eval_runtime": 140.6986,
+      "eval_samples_per_second": 34.421,
+      "eval_steps_per_second": 4.307,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.16759002770083e-05,
+      "loss": 3.1548,
+      "step": 101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.16343490304709e-05,
+      "loss": 3.1388,
+      "step": 102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.159279778393351e-05,
+      "loss": 3.199,
+      "step": 103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.155124653739611e-05,
+      "loss": 3.1247,
+      "step": 104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.150969529085872e-05,
+      "loss": 3.218,
+      "step": 105
+    },
+    {
+      "epoch": 0.29,
+      "eval_loss": 3.2627265453338623,
+      "eval_runtime": 145.0882,
+      "eval_samples_per_second": 33.38,
+      "eval_steps_per_second": 4.177,
+      "eval_wer": 1.0,
+      "step": 105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.146814404432133e-05,
+      "loss": 3.1647,
+      "step": 106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 7.142659279778393e-05,
+      "loss": 3.1678,
+      "step": 107
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.138504155124653e-05,
+      "loss": 3.0838,
+      "step": 108
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.134349030470913e-05,
+      "loss": 3.2472,
+      "step": 109
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.130193905817175e-05,
+      "loss": 3.1308,
+      "step": 110
+    },
+    {
+      "epoch": 0.3,
+      "eval_loss": 3.2152204513549805,
+      "eval_runtime": 148.0181,
+      "eval_samples_per_second": 32.719,
+      "eval_steps_per_second": 4.094,
+      "eval_wer": 1.0,
+      "step": 110
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 7.126038781163435e-05,
+      "loss": 3.1258,
+      "step": 111
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.121883656509695e-05,
+      "loss": 3.1609,
+      "step": 112
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.117728531855955e-05,
+      "loss": 3.134,
+      "step": 113
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.113573407202215e-05,
+      "loss": 3.0742,
+      "step": 114
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 7.109418282548477e-05,
+      "loss": 3.109,
+      "step": 115
+    },
+    {
+      "epoch": 0.31,
+      "eval_loss": 3.1685547828674316,
+      "eval_runtime": 138.9674,
+      "eval_samples_per_second": 34.85,
+      "eval_steps_per_second": 4.361,
+      "eval_wer": 1.0,
+      "step": 115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.105263157894735e-05,
+      "loss": 3.0879,
+      "step": 116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.101108033240997e-05,
+      "loss": 3.1192,
+      "step": 117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 7.096952908587257e-05,
+      "loss": 3.1015,
+      "step": 118
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.092797783933517e-05,
+      "loss": 3.147,
+      "step": 119
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.088642659279777e-05,
+      "loss": 3.1188,
+      "step": 120
+    },
+    {
+      "epoch": 0.33,
+      "eval_loss": 3.1734039783477783,
+      "eval_runtime": 138.2826,
+      "eval_samples_per_second": 35.022,
+      "eval_steps_per_second": 4.382,
+      "eval_wer": 1.0,
+      "step": 120
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.084487534626038e-05,
+      "loss": 3.106,
+      "step": 121
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 7.080332409972299e-05,
+      "loss": 3.1265,
+      "step": 122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.076177285318559e-05,
+      "loss": 3.0769,
+      "step": 123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.07202216066482e-05,
+      "loss": 3.0903,
+      "step": 124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.06786703601108e-05,
+      "loss": 3.1132,
+      "step": 125
+    },
+    {
+      "epoch": 0.34,
+      "eval_loss": 3.14314603805542,
+      "eval_runtime": 139.5613,
+      "eval_samples_per_second": 34.702,
+      "eval_steps_per_second": 4.342,
+      "eval_wer": 1.0,
+      "step": 125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 7.06371191135734e-05,
+      "loss": 3.0606,
+      "step": 126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.059556786703601e-05,
+      "loss": 3.108,
+      "step": 127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.055401662049861e-05,
+      "loss": 3.0936,
+      "step": 128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 7.051246537396121e-05,
+      "loss": 3.0532,
+      "step": 129
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.047091412742382e-05,
+      "loss": 3.0667,
+      "step": 130
+    },
+    {
+      "epoch": 0.36,
+      "eval_loss": 3.1685564517974854,
+      "eval_runtime": 139.9247,
+      "eval_samples_per_second": 34.611,
+      "eval_steps_per_second": 4.331,
+      "eval_wer": 1.0,
+      "step": 130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.042936288088642e-05,
+      "loss": 3.1034,
+      "step": 131
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.038781163434903e-05,
+      "loss": 3.0775,
+      "step": 132
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 7.034626038781162e-05,
+      "loss": 3.1185,
+      "step": 133
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.030470914127424e-05,
+      "loss": 3.1026,
+      "step": 134
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.026315789473684e-05,
+      "loss": 3.1167,
+      "step": 135
+    },
+    {
+      "epoch": 0.37,
+      "eval_loss": 3.1885087490081787,
+      "eval_runtime": 142.2545,
+      "eval_samples_per_second": 34.045,
+      "eval_steps_per_second": 4.26,
+      "eval_wer": 1.0,
+      "step": 135
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.022160664819944e-05,
+      "loss": 3.1172,
+      "step": 136
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 7.018005540166204e-05,
+      "loss": 3.06,
+      "step": 137
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.013850415512464e-05,
+      "loss": 3.0575,
+      "step": 138
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.009695290858726e-05,
+      "loss": 3.074,
+      "step": 139
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 7.005540166204986e-05,
+      "loss": 3.0592,
+      "step": 140
+    },
+    {
+      "epoch": 0.38,
+      "eval_loss": 3.1100046634674072,
+      "eval_runtime": 141.5413,
+      "eval_samples_per_second": 34.216,
+      "eval_steps_per_second": 4.281,
+      "eval_wer": 1.0,
+      "step": 140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 7.001385041551246e-05,
+      "loss": 3.0282,
+      "step": 141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.997229916897506e-05,
+      "loss": 3.0376,
+      "step": 142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.993074792243766e-05,
+      "loss": 3.0876,
+      "step": 143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 6.988919667590028e-05,
+      "loss": 3.0586,
+      "step": 144
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.984764542936287e-05,
+      "loss": 3.0531,
+      "step": 145
+    },
+    {
+      "epoch": 0.4,
+      "eval_loss": 3.1149468421936035,
+      "eval_runtime": 139.0059,
+      "eval_samples_per_second": 34.84,
+      "eval_steps_per_second": 4.36,
+      "eval_wer": 1.0,
+      "step": 145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.980609418282548e-05,
+      "loss": 3.1397,
+      "step": 146
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.976454293628808e-05,
+      "loss": 3.0942,
+      "step": 147
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 6.972299168975068e-05,
+      "loss": 3.1146,
+      "step": 148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.968144044321329e-05,
+      "loss": 3.125,
+      "step": 149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.963988919667589e-05,
+      "loss": 3.1224,
+      "step": 150
+    },
+    {
+      "epoch": 0.41,
+      "eval_loss": 3.1204922199249268,
+      "eval_runtime": 137.4026,
+      "eval_samples_per_second": 35.247,
+      "eval_steps_per_second": 4.41,
+      "eval_wer": 1.0,
+      "step": 150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 6.95983379501385e-05,
+      "loss": 3.0588,
+      "step": 151
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.95567867036011e-05,
+      "loss": 3.0511,
+      "step": 152
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.95152354570637e-05,
+      "loss": 3.0691,
+      "step": 153
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.947368421052631e-05,
+      "loss": 3.031,
+      "step": 154
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 6.943213296398891e-05,
+      "loss": 3.0651,
+      "step": 155
+    },
+    {
+      "epoch": 0.42,
+      "eval_loss": 3.110053777694702,
+      "eval_runtime": 137.5926,
+      "eval_samples_per_second": 35.198,
+      "eval_steps_per_second": 4.404,
+      "eval_wer": 1.0,
+      "step": 155
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.939058171745152e-05,
+      "loss": 3.0572,
+      "step": 156
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.934903047091412e-05,
+      "loss": 3.0695,
+      "step": 157
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 6.930747922437673e-05,
+      "loss": 3.0164,
+      "step": 158
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.926592797783933e-05,
+      "loss": 3.034,
+      "step": 159
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.922437673130194e-05,
+      "loss": 3.0077,
+      "step": 160
+    },
+    {
+      "epoch": 0.44,
+      "eval_loss": 3.0980353355407715,
+      "eval_runtime": 139.6081,
+      "eval_samples_per_second": 34.69,
+      "eval_steps_per_second": 4.341,
+      "eval_wer": 1.0,
+      "step": 160
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.918282548476453e-05,
+      "loss": 3.0347,
+      "step": 161
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 6.914127423822715e-05,
+      "loss": 3.0581,
+      "step": 162
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.909972299168975e-05,
+      "loss": 3.0572,
+      "step": 163
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.905817174515235e-05,
+      "loss": 3.0149,
+      "step": 164
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.901662049861495e-05,
+      "loss": 3.0027,
+      "step": 165
+    },
+    {
+      "epoch": 0.45,
+      "eval_loss": 3.1131813526153564,
+      "eval_runtime": 151.9666,
+      "eval_samples_per_second": 31.869,
+      "eval_steps_per_second": 3.988,
+      "eval_wer": 1.0,
+      "step": 165
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 6.897506925207755e-05,
+      "loss": 3.0422,
+      "step": 166
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.893351800554017e-05,
+      "loss": 3.0237,
+      "step": 167
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.889196675900277e-05,
+      "loss": 3.0209,
+      "step": 168
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 6.885041551246537e-05,
+      "loss": 3.0675,
+      "step": 169
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.880886426592797e-05,
+      "loss": 3.0423,
+      "step": 170
+    },
+    {
+      "epoch": 0.47,
+      "eval_loss": 3.08860445022583,
+      "eval_runtime": 144.8917,
+      "eval_samples_per_second": 33.425,
+      "eval_steps_per_second": 4.182,
+      "eval_wer": 1.0,
+      "step": 170
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.876731301939057e-05,
+      "loss": 3.0671,
+      "step": 171
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.872576177285319e-05,
+      "loss": 3.0283,
+      "step": 172
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 6.868421052631578e-05,
+      "loss": 3.0798,
+      "step": 173
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 6.864265927977839e-05,
+      "loss": 3.023,
+      "step": 174
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 6.860110803324099e-05,
+      "loss": 3.0462,
+      "step": 175
+    },
+    {
+      "epoch": 0.48,
+      "eval_loss": 3.086527109146118,
+      "eval_runtime": 138.5084,
+      "eval_samples_per_second": 34.965,
+      "eval_steps_per_second": 4.375,
+      "eval_wer": 1.0,
+      "step": 175
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 6.85595567867036e-05,
+      "loss": 3.0325,
+      "step": 176
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 6.85180055401662e-05,
+      "loss": 3.0584,
+      "step": 177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 6.84764542936288e-05,
+      "loss": 3.005,
+      "step": 178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 6.843490304709141e-05,
+      "loss": 3.0212,
+      "step": 179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 6.839335180055401e-05,
+      "loss": 3.0701,
+      "step": 180
+    },
+    {
+      "epoch": 0.49,
+      "eval_loss": 3.0862884521484375,
+      "eval_runtime": 139.973,
+      "eval_samples_per_second": 34.6,
+      "eval_steps_per_second": 4.329,
+      "eval_wer": 1.0,
+      "step": 180
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 6.835180055401662e-05,
+      "loss": 3.0448,
+      "step": 181
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 6.831024930747922e-05,
+      "loss": 3.0722,
+      "step": 182
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 6.826869806094182e-05,
+      "loss": 3.0851,
+      "step": 183
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 6.822714681440443e-05,
+      "loss": 3.0566,
+      "step": 184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.818559556786702e-05,
+      "loss": 3.0871,
+      "step": 185
+    },
+    {
+      "epoch": 0.51,
+      "eval_loss": 3.082496166229248,
+      "eval_runtime": 138.7766,
+      "eval_samples_per_second": 34.898,
+      "eval_steps_per_second": 4.367,
+      "eval_wer": 1.0,
+      "step": 185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.814404432132964e-05,
+      "loss": 3.0165,
+      "step": 186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.810249307479224e-05,
+      "loss": 3.0615,
+      "step": 187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 6.806094182825484e-05,
+      "loss": 3.035,
+      "step": 188
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 6.801939058171744e-05,
+      "loss": 3.0542,
+      "step": 189
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 6.797783933518004e-05,
+      "loss": 3.0585,
+      "step": 190
+    },
+    {
+      "epoch": 0.52,
+      "eval_loss": 3.071974754333496,
+      "eval_runtime": 140.1656,
+      "eval_samples_per_second": 34.552,
+      "eval_steps_per_second": 4.323,
+      "eval_wer": 1.0,
+      "step": 190
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 6.793628808864266e-05,
+      "loss": 3.096,
+      "step": 191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 6.789473684210526e-05,
+      "loss": 3.0398,
+      "step": 192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 6.785318559556786e-05,
+      "loss": 3.0531,
+      "step": 193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 6.781163434903046e-05,
+      "loss": 3.0613,
+      "step": 194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 6.777008310249306e-05,
+      "loss": 3.0274,
+      "step": 195
+    },
+    {
+      "epoch": 0.53,
+      "eval_loss": 3.0736279487609863,
+      "eval_runtime": 137.6029,
+      "eval_samples_per_second": 35.195,
+      "eval_steps_per_second": 4.404,
+      "eval_wer": 1.0,
+      "step": 195
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 6.772853185595568e-05,
+      "loss": 3.0259,
+      "step": 196
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 6.768698060941827e-05,
+      "loss": 3.0772,
+      "step": 197
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 6.764542936288088e-05,
+      "loss": 3.0448,
+      "step": 198
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 6.760387811634348e-05,
+      "loss": 3.0747,
+      "step": 199
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.756232686980609e-05,
+      "loss": 3.0983,
+      "step": 200
+    },
+    {
+      "epoch": 0.55,
+      "eval_loss": 3.065845012664795,
+      "eval_runtime": 142.1471,
+      "eval_samples_per_second": 34.07,
+      "eval_steps_per_second": 4.263,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.752077562326869e-05,
+      "loss": 3.0774,
+      "step": 201
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 6.747922437673129e-05,
+      "loss": 3.0449,
+      "step": 202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.74376731301939e-05,
+      "loss": 3.0384,
+      "step": 203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.73961218836565e-05,
+      "loss": 3.0273,
+      "step": 204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.73545706371191e-05,
+      "loss": 3.0538,
+      "step": 205
+    },
+    {
+      "epoch": 0.56,
+      "eval_loss": 3.1240925788879395,
+      "eval_runtime": 139.0944,
+      "eval_samples_per_second": 34.818,
+      "eval_steps_per_second": 4.357,
+      "eval_wer": 1.0,
+      "step": 205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 6.731301939058171e-05,
+      "loss": 3.0331,
+      "step": 206
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 6.727146814404431e-05,
+      "loss": 3.008,
+      "step": 207
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 6.722991689750692e-05,
+      "loss": 3.0716,
+      "step": 208
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 6.718836565096953e-05,
+      "loss": 3.0585,
+      "step": 209
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 6.714681440443213e-05,
+      "loss": 3.0862,
+      "step": 210
+    },
+    {
+      "epoch": 0.57,
+      "eval_loss": 3.057255983352661,
+      "eval_runtime": 138.3216,
+      "eval_samples_per_second": 35.013,
+      "eval_steps_per_second": 4.381,
+      "eval_wer": 1.0,
+      "step": 210
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 6.710526315789473e-05,
+      "loss": 3.0637,
+      "step": 211
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 6.706371191135734e-05,
+      "loss": 3.0026,
+      "step": 212
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 6.702216066481993e-05,
+      "loss": 3.021,
+      "step": 213
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.698060941828255e-05,
+      "loss": 3.0396,
+      "step": 214
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.693905817174515e-05,
+      "loss": 3.0041,
+      "step": 215
+    },
+    {
+      "epoch": 0.59,
+      "eval_loss": 3.0608010292053223,
+      "eval_runtime": 141.5903,
+      "eval_samples_per_second": 34.204,
+      "eval_steps_per_second": 4.28,
+      "eval_wer": 1.0,
+      "step": 215
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.689750692520775e-05,
+      "loss": 3.0375,
+      "step": 216
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 6.685595567867035e-05,
+      "loss": 3.0124,
+      "step": 217
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.681440443213295e-05,
+      "loss": 3.0057,
+      "step": 218
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.677285318559557e-05,
+      "loss": 3.0525,
+      "step": 219
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.673130193905817e-05,
+      "loss": 3.027,
+      "step": 220
+    },
+    {
+      "epoch": 0.6,
+      "eval_loss": 3.061398983001709,
+      "eval_runtime": 138.0429,
+      "eval_samples_per_second": 35.083,
+      "eval_steps_per_second": 4.39,
+      "eval_wer": 1.0,
+      "step": 220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 6.668975069252077e-05,
+      "loss": 3.0518,
+      "step": 221
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.664819944598337e-05,
+      "loss": 2.999,
+      "step": 222
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.660664819944597e-05,
+      "loss": 3.0322,
+      "step": 223
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.656509695290859e-05,
+      "loss": 3.0088,
+      "step": 224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652354570637118e-05,
+      "loss": 2.9916,
+      "step": 225
+    },
+    {
+      "epoch": 0.62,
+      "eval_loss": 3.052710771560669,
+      "eval_runtime": 138.818,
+      "eval_samples_per_second": 34.887,
+      "eval_steps_per_second": 4.365,
+      "eval_wer": 1.0,
+      "step": 225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.648199445983379e-05,
+      "loss": 3.0266,
+      "step": 226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.64404432132964e-05,
+      "loss": 3.0052,
+      "step": 227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6398891966759e-05,
+      "loss": 3.0188,
+      "step": 228
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.63573407202216e-05,
+      "loss": 3.018,
+      "step": 229
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.63157894736842e-05,
+      "loss": 3.0157,
+      "step": 230
+    },
+    {
+      "epoch": 0.63,
+      "eval_loss": 3.051440954208374,
+      "eval_runtime": 153.6473,
+      "eval_samples_per_second": 31.52,
+      "eval_steps_per_second": 3.944,
+      "eval_wer": 1.0,
+      "step": 230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.627423822714681e-05,
+      "loss": 3.0212,
+      "step": 231
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.623268698060941e-05,
+      "loss": 3.0808,
+      "step": 232
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.619113573407202e-05,
+      "loss": 2.9837,
+      "step": 233
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.614958448753462e-05,
+      "loss": 3.0047,
+      "step": 234
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.610803324099722e-05,
+      "loss": 3.0429,
+      "step": 235
+    },
+    {
+      "epoch": 0.64,
+      "eval_loss": 3.039069652557373,
+      "eval_runtime": 140.8845,
+      "eval_samples_per_second": 34.376,
+      "eval_steps_per_second": 4.301,
+      "eval_wer": 1.0,
+      "step": 235
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.606648199445983e-05,
+      "loss": 3.025,
+      "step": 236
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.602493074792244e-05,
+      "loss": 3.0207,
+      "step": 237
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.598337950138504e-05,
+      "loss": 3.0576,
+      "step": 238
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 6.594182825484764e-05,
+      "loss": 3.0605,
+      "step": 239
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.590027700831024e-05,
+      "loss": 2.999,
+      "step": 240
+    },
+    {
+      "epoch": 0.66,
+      "eval_loss": 3.0461792945861816,
+      "eval_runtime": 138.1836,
+      "eval_samples_per_second": 35.048,
+      "eval_steps_per_second": 4.385,
+      "eval_wer": 1.0,
+      "step": 240
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.585872576177286e-05,
+      "loss": 3.0255,
+      "step": 241
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.581717451523544e-05,
+      "loss": 2.999,
+      "step": 242
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 6.577562326869806e-05,
+      "loss": 3.0669,
+      "step": 243
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.573407202216066e-05,
+      "loss": 3.0463,
+      "step": 244
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.569252077562326e-05,
+      "loss": 3.0053,
+      "step": 245
+    },
+    {
+      "epoch": 0.67,
+      "eval_loss": 3.043794631958008,
+      "eval_runtime": 153.1643,
+      "eval_samples_per_second": 31.62,
+      "eval_steps_per_second": 3.957,
+      "eval_wer": 1.0,
+      "step": 245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 6.565096952908586e-05,
+      "loss": 3.0505,
+      "step": 246
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.560941828254846e-05,
+      "loss": 3.0499,
+      "step": 247
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.556786703601108e-05,
+      "loss": 3.111,
+      "step": 248
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.552631578947368e-05,
+      "loss": 3.0701,
+      "step": 249
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 6.548476454293628e-05,
+      "loss": 2.9812,
+      "step": 250
+    },
+    {
+      "epoch": 0.68,
+      "eval_loss": 3.044740676879883,
+      "eval_runtime": 147.969,
+      "eval_samples_per_second": 32.73,
+      "eval_steps_per_second": 4.095,
+      "eval_wer": 1.0,
+      "step": 250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.544321329639888e-05,
+      "loss": 3.0341,
+      "step": 251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.540166204986149e-05,
+      "loss": 3.0316,
+      "step": 252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.53601108033241e-05,
+      "loss": 3.0118,
+      "step": 253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 6.531855955678669e-05,
+      "loss": 3.0496,
+      "step": 254
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.52770083102493e-05,
+      "loss": 3.0062,
+      "step": 255
+    },
+    {
+      "epoch": 0.7,
+      "eval_loss": 3.0659635066986084,
+      "eval_runtime": 138.8083,
+      "eval_samples_per_second": 34.89,
+      "eval_steps_per_second": 4.366,
+      "eval_wer": 1.0,
+      "step": 255
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.52354570637119e-05,
+      "loss": 3.011,
+      "step": 256
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 6.519390581717451e-05,
+      "loss": 3.0028,
+      "step": 257
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.515235457063711e-05,
+      "loss": 2.9794,
+      "step": 258
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.511080332409971e-05,
+      "loss": 3.0002,
+      "step": 259
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.506925207756233e-05,
+      "loss": 3.0045,
+      "step": 260
+    },
+    {
+      "epoch": 0.71,
+      "eval_loss": 3.0102784633636475,
+      "eval_runtime": 153.8367,
+      "eval_samples_per_second": 31.481,
+      "eval_steps_per_second": 3.939,
+      "eval_wer": 1.0,
+      "step": 260
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 6.502770083102493e-05,
+      "loss": 2.9959,
+      "step": 261
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.498614958448753e-05,
+      "loss": 3.005,
+      "step": 262
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.494459833795013e-05,
+      "loss": 3.021,
+      "step": 263
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 6.490304709141274e-05,
+      "loss": 2.988,
+      "step": 264
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.486149584487535e-05,
+      "loss": 2.9684,
+      "step": 265
+    },
+    {
+      "epoch": 0.73,
+      "eval_loss": 3.010617256164551,
+      "eval_runtime": 138.7834,
+      "eval_samples_per_second": 34.896,
+      "eval_steps_per_second": 4.367,
+      "eval_wer": 1.0,
+      "step": 265
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.481994459833795e-05,
+      "loss": 2.984,
+      "step": 266
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.477839335180055e-05,
+      "loss": 2.9466,
+      "step": 267
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 6.473684210526315e-05,
+      "loss": 2.9993,
+      "step": 268
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.469529085872577e-05,
+      "loss": 2.9867,
+      "step": 269
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.465373961218835e-05,
+      "loss": 2.9885,
+      "step": 270
+    },
+    {
+      "epoch": 0.74,
+      "eval_loss": 3.001441717147827,
+      "eval_runtime": 139.0471,
+      "eval_samples_per_second": 34.83,
+      "eval_steps_per_second": 4.358,
+      "eval_wer": 1.0,
+      "step": 270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.461218836565097e-05,
+      "loss": 3.024,
+      "step": 271
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 6.457063711911357e-05,
+      "loss": 3.0103,
+      "step": 272
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.452908587257617e-05,
+      "loss": 2.9813,
+      "step": 273
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.448753462603877e-05,
+      "loss": 3.0006,
+      "step": 274
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 6.444598337950137e-05,
+      "loss": 3.0062,
+      "step": 275
+    },
+    {
+      "epoch": 0.75,
+      "eval_loss": 2.9884729385375977,
+      "eval_runtime": 139.3991,
+      "eval_samples_per_second": 34.742,
+      "eval_steps_per_second": 4.347,
+      "eval_wer": 1.0,
+      "step": 275
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.440443213296399e-05,
+      "loss": 2.9705,
+      "step": 276
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.436288088642659e-05,
+      "loss": 3.0039,
+      "step": 277
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.432132963988919e-05,
+      "loss": 2.9914,
+      "step": 278
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 6.42797783933518e-05,
+      "loss": 3.0003,
+      "step": 279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.42382271468144e-05,
+      "loss": 2.9736,
+      "step": 280
+    },
+    {
+      "epoch": 0.77,
+      "eval_loss": 3.0330100059509277,
+      "eval_runtime": 140.1871,
+      "eval_samples_per_second": 34.547,
+      "eval_steps_per_second": 4.323,
+      "eval_wer": 1.0,
+      "step": 280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.419667590027701e-05,
+      "loss": 3.0202,
+      "step": 281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.41551246537396e-05,
+      "loss": 2.9919,
+      "step": 282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 6.411357340720221e-05,
+      "loss": 3.0131,
+      "step": 283
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.407202216066482e-05,
+      "loss": 2.9838,
+      "step": 284
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.403047091412742e-05,
+      "loss": 2.9766,
+      "step": 285
+    },
+    {
+      "epoch": 0.78,
+      "eval_loss": 2.990950345993042,
+      "eval_runtime": 140.6604,
+      "eval_samples_per_second": 34.43,
+      "eval_steps_per_second": 4.308,
+      "eval_wer": 1.0,
+      "step": 285
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 6.398891966759002e-05,
+      "loss": 2.9768,
+      "step": 286
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.394736842105262e-05,
+      "loss": 3.0161,
+      "step": 287
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.390581717451524e-05,
+      "loss": 2.9544,
+      "step": 288
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.386426592797784e-05,
+      "loss": 2.9571,
+      "step": 289
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 6.382271468144044e-05,
+      "loss": 2.9545,
+      "step": 290
+    },
+    {
+      "epoch": 0.79,
+      "eval_loss": 2.997222900390625,
+      "eval_runtime": 141.0522,
+      "eval_samples_per_second": 34.335,
+      "eval_steps_per_second": 4.296,
+      "eval_wer": 1.0,
+      "step": 290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.378116343490304e-05,
+      "loss": 2.9844,
+      "step": 291
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.373961218836564e-05,
+      "loss": 2.9664,
+      "step": 292
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.369806094182826e-05,
+      "loss": 2.9764,
+      "step": 293
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 6.365650969529084e-05,
+      "loss": 3.0138,
+      "step": 294
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 6.361495844875346e-05,
+      "loss": 2.9936,
+      "step": 295
+    },
+    {
+      "epoch": 0.81,
+      "eval_loss": 2.9872097969055176,
+      "eval_runtime": 148.4489,
+      "eval_samples_per_second": 32.624,
+      "eval_steps_per_second": 4.082,
+      "eval_wer": 1.0,
+      "step": 295
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 6.357340720221606e-05,
+      "loss": 3.0034,
+      "step": 296
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 6.353185595567866e-05,
+      "loss": 2.9694,
+      "step": 297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.349030470914126e-05,
+      "loss": 2.9703,
+      "step": 298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.344875346260387e-05,
+      "loss": 2.9506,
+      "step": 299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.340720221606648e-05,
+      "loss": 3.0832,
+      "step": 300
+    },
+    {
+      "epoch": 0.82,
+      "eval_loss": 2.9978036880493164,
+      "eval_runtime": 139.7284,
+      "eval_samples_per_second": 34.66,
+      "eval_steps_per_second": 4.337,
+      "eval_wer": 1.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 6.336565096952908e-05,
+      "loss": 3.0593,
+      "step": 301
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.332409972299168e-05,
+      "loss": 2.9744,
+      "step": 302
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.328254847645429e-05,
+      "loss": 3.006,
+      "step": 303
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.324099722991689e-05,
+      "loss": 3.0036,
+      "step": 304
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 6.31994459833795e-05,
+      "loss": 2.974,
+      "step": 305
+    },
+    {
+      "epoch": 0.83,
+      "eval_loss": 2.997765064239502,
+      "eval_runtime": 139.6069,
+      "eval_samples_per_second": 34.69,
+      "eval_steps_per_second": 4.341,
+      "eval_wer": 1.0,
+      "step": 305
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.315789473684209e-05,
+      "loss": 2.9871,
+      "step": 306
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.31163434903047e-05,
+      "loss": 2.9878,
+      "step": 307
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 6.30747922437673e-05,
+      "loss": 2.9892,
+      "step": 308
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.303324099722991e-05,
+      "loss": 2.9936,
+      "step": 309
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.299168975069251e-05,
+      "loss": 2.9846,
+      "step": 310
+    },
+    {
+      "epoch": 0.85,
+      "eval_loss": 2.9848742485046387,
+      "eval_runtime": 154.4288,
+      "eval_samples_per_second": 31.361,
+      "eval_steps_per_second": 3.924,
+      "eval_wer": 1.0,
+      "step": 310
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.295013850415511e-05,
+      "loss": 2.9937,
+      "step": 311
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 6.290858725761773e-05,
+      "loss": 2.9878,
+      "step": 312
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.286703601108033e-05,
+      "loss": 2.9779,
+      "step": 313
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.282548476454293e-05,
+      "loss": 3.0071,
+      "step": 314
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.278393351800553e-05,
+      "loss": 2.9554,
+      "step": 315
+    },
+    {
+      "epoch": 0.86,
+      "eval_loss": 2.9810054302215576,
+      "eval_runtime": 152.5186,
+      "eval_samples_per_second": 31.754,
+      "eval_steps_per_second": 3.973,
+      "eval_wer": 1.0,
+      "step": 315
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 6.274238227146815e-05,
+      "loss": 2.9442,
+      "step": 316
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.270083102493075e-05,
+      "loss": 2.9785,
+      "step": 317
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.265927977839335e-05,
+      "loss": 2.9699,
+      "step": 318
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 6.261772853185595e-05,
+      "loss": 3.0035,
+      "step": 319
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.257617728531855e-05,
+      "loss": 2.9524,
+      "step": 320
+    },
+    {
+      "epoch": 0.88,
+      "eval_loss": 2.9730944633483887,
+      "eval_runtime": 138.504,
+      "eval_samples_per_second": 34.966,
+      "eval_steps_per_second": 4.375,
+      "eval_wer": 1.0,
+      "step": 320
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.253462603878117e-05,
+      "loss": 2.9584,
+      "step": 321
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.249307479224375e-05,
+      "loss": 3.0005,
+      "step": 322
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 6.245152354570637e-05,
+      "loss": 2.9827,
+      "step": 323
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.240997229916897e-05,
+      "loss": 3.011,
+      "step": 324
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.236842105263157e-05,
+      "loss": 2.9426,
+      "step": 325
+    },
+    {
+      "epoch": 0.89,
+      "eval_loss": 2.982403516769409,
+      "eval_runtime": 139.6744,
+      "eval_samples_per_second": 34.674,
+      "eval_steps_per_second": 4.339,
+      "eval_wer": 1.0,
+      "step": 325
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.232686980609417e-05,
+      "loss": 2.9149,
+      "step": 326
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 6.228531855955678e-05,
+      "loss": 2.9827,
+      "step": 327
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.224376731301939e-05,
+      "loss": 2.9586,
+      "step": 328
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.220221606648199e-05,
+      "loss": 2.9289,
+      "step": 329
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 6.21606648199446e-05,
+      "loss": 2.9416,
+      "step": 330
+    },
+    {
+      "epoch": 0.9,
+      "eval_loss": 2.973137855529785,
+      "eval_runtime": 144.8803,
+      "eval_samples_per_second": 33.428,
+      "eval_steps_per_second": 4.183,
+      "eval_wer": 1.0,
+      "step": 330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.21191135734072e-05,
+      "loss": 2.9405,
+      "step": 331
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.20775623268698e-05,
+      "loss": 2.9243,
+      "step": 332
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.203601108033241e-05,
+      "loss": 2.9733,
+      "step": 333
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 6.1994459833795e-05,
+      "loss": 2.9676,
+      "step": 334
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 6.195290858725762e-05,
+      "loss": 2.9705,
+      "step": 335
+    },
+    {
+      "epoch": 0.92,
+      "eval_loss": 2.9829530715942383,
+      "eval_runtime": 138.2901,
+      "eval_samples_per_second": 35.021,
+      "eval_steps_per_second": 4.382,
+      "eval_wer": 1.0,
+      "step": 335
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 6.191135734072022e-05,
+      "loss": 2.9778,
+      "step": 336
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 6.186980609418282e-05,
+      "loss": 3.0158,
+      "step": 337
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 6.182825484764542e-05,
+      "loss": 2.9693,
+      "step": 338
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.178670360110802e-05,
+      "loss": 2.9627,
+      "step": 339
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.174515235457064e-05,
+      "loss": 2.9502,
+      "step": 340
+    },
+    {
+      "epoch": 0.93,
+      "eval_loss": 2.971325397491455,
+      "eval_runtime": 139.7434,
+      "eval_samples_per_second": 34.656,
+      "eval_steps_per_second": 4.337,
+      "eval_wer": 1.0,
+      "step": 340
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 6.170360110803324e-05,
+      "loss": 2.9047,
+      "step": 341
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.166204986149584e-05,
+      "loss": 2.9732,
+      "step": 342
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.162049861495844e-05,
+      "loss": 2.9257,
+      "step": 343
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.157894736842104e-05,
+      "loss": 2.9478,
+      "step": 344
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 6.153739612188366e-05,
+      "loss": 2.9393,
+      "step": 345
+    },
+    {
+      "epoch": 0.94,
+      "eval_loss": 2.978956699371338,
+      "eval_runtime": 139.4266,
+      "eval_samples_per_second": 34.735,
+      "eval_steps_per_second": 4.346,
+      "eval_wer": 1.0,
+      "step": 345
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.149584487534626e-05,
+      "loss": 2.8966,
+      "step": 346
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.145429362880886e-05,
+      "loss": 2.9612,
+      "step": 347
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.141274238227146e-05,
+      "loss": 2.9318,
+      "step": 348
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 6.137119113573406e-05,
+      "loss": 2.9819,
+      "step": 349
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.132963988919668e-05,
+      "loss": 2.9336,
+      "step": 350
+    },
+    {
+      "epoch": 0.96,
+      "eval_loss": 2.968351125717163,
+      "eval_runtime": 140.8713,
+      "eval_samples_per_second": 34.379,
+      "eval_steps_per_second": 4.302,
+      "eval_wer": 1.0,
+      "step": 350
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.128808864265927e-05,
+      "loss": 2.9953,
+      "step": 351
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 6.124653739612188e-05,
+      "loss": 2.9949,
+      "step": 352
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.120498614958448e-05,
+      "loss": 2.9588,
+      "step": 353
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.116343490304708e-05,
+      "loss": 2.9841,
+      "step": 354
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.112188365650969e-05,
+      "loss": 2.9542,
+      "step": 355
+    },
+    {
+      "epoch": 0.97,
+      "eval_loss": 2.96889066696167,
+      "eval_runtime": 141.6028,
+      "eval_samples_per_second": 34.201,
+      "eval_steps_per_second": 4.28,
+      "eval_wer": 1.0,
+      "step": 355
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 6.108033240997229e-05,
+      "loss": 2.9655,
+      "step": 356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.10387811634349e-05,
+      "loss": 2.9397,
+      "step": 357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.0997229916897504e-05,
+      "loss": 2.9519,
+      "step": 358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.0955678670360106e-05,
+      "loss": 2.8983,
+      "step": 359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 6.091412742382271e-05,
+      "loss": 2.9408,
+      "step": 360
+    },
+    {
+      "epoch": 0.98,
+      "eval_loss": 2.955634355545044,
+      "eval_runtime": 138.5742,
+      "eval_samples_per_second": 34.949,
+      "eval_steps_per_second": 4.373,
+      "eval_wer": 1.0,
+      "step": 360
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.0872576177285315e-05,
+      "loss": 2.9454,
+      "step": 361
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.0831024930747924e-05,
+      "loss": 2.9174,
+      "step": 362
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 6.078947368421052e-05,
+      "loss": 2.9425,
+      "step": 363
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.074792243767313e-05,
+      "loss": 2.9597,
+      "step": 364
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.070637119113573e-05,
+      "loss": 2.9544,
+      "step": 365
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 2.956270217895508,
+      "eval_runtime": 139.6182,
+      "eval_samples_per_second": 34.687,
+      "eval_steps_per_second": 4.34,
+      "eval_wer": 1.0,
+      "step": 365
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 6.0664819944598337e-05,
+      "loss": 4.4967,
+      "step": 366
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 6.062326869806093e-05,
+      "loss": 2.9612,
+      "step": 367
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 6.058171745152354e-05,
+      "loss": 3.0047,
+      "step": 368
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 6.054016620498615e-05,
+      "loss": 2.9431,
+      "step": 369
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 6.049861495844875e-05,
+      "loss": 2.9187,
+      "step": 370
+    },
+    {
+      "epoch": 1.01,
+      "eval_loss": 2.9623892307281494,
+      "eval_runtime": 141.8494,
+      "eval_samples_per_second": 34.142,
+      "eval_steps_per_second": 4.272,
+      "eval_wer": 1.0,
+      "step": 370
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 6.045706371191135e-05,
+      "loss": 2.9771,
+      "step": 371
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 6.041551246537395e-05,
+      "loss": 2.966,
+      "step": 372
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 6.037396121883656e-05,
+      "loss": 2.9662,
+      "step": 373
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 6.033240997229917e-05,
+      "loss": 3.001,
+      "step": 374
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 6.0290858725761764e-05,
+      "loss": 2.9935,
+      "step": 375
+    },
+    {
+      "epoch": 1.03,
+      "eval_loss": 2.949976921081543,
+      "eval_runtime": 138.4199,
+      "eval_samples_per_second": 34.988,
+      "eval_steps_per_second": 4.378,
+      "eval_wer": 1.0,
+      "step": 375
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 6.024930747922437e-05,
+      "loss": 2.951,
+      "step": 376
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 6.0207756232686974e-05,
+      "loss": 2.9511,
+      "step": 377
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 6.016620498614958e-05,
+      "loss": 2.9524,
+      "step": 378
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 6.0124653739612183e-05,
+      "loss": 2.9503,
+      "step": 379
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 6.0083102493074785e-05,
+      "loss": 2.9803,
+      "step": 380
+    },
+    {
+      "epoch": 1.04,
+      "eval_loss": 2.95582914352417,
+      "eval_runtime": 139.2934,
+      "eval_samples_per_second": 34.768,
+      "eval_steps_per_second": 4.351,
+      "eval_wer": 1.0,
+      "step": 380
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 6.004155124653739e-05,
+      "loss": 2.9333,
+      "step": 381
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 5.9999999999999995e-05,
+      "loss": 2.9503,
+      "step": 382
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 5.9958448753462596e-05,
+      "loss": 2.9337,
+      "step": 383
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 5.9916897506925205e-05,
+      "loss": 2.9673,
+      "step": 384
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 5.9875346260387806e-05,
+      "loss": 2.9867,
+      "step": 385
+    },
+    {
+      "epoch": 1.05,
+      "eval_loss": 2.9473366737365723,
+      "eval_runtime": 144.168,
+      "eval_samples_per_second": 33.593,
+      "eval_steps_per_second": 4.203,
+      "eval_wer": 1.0,
+      "step": 385
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 5.9833795013850414e-05,
+      "loss": 2.9225,
+      "step": 386
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 5.979224376731301e-05,
+      "loss": 2.9381,
+      "step": 387
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 5.975069252077562e-05,
+      "loss": 2.9096,
+      "step": 388
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 5.9709141274238226e-05,
+      "loss": 2.9099,
+      "step": 389
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 5.966759002770083e-05,
+      "loss": 2.8925,
+      "step": 390
+    },
+    {
+      "epoch": 1.07,
+      "eval_loss": 2.944408655166626,
+      "eval_runtime": 145.7009,
+      "eval_samples_per_second": 33.239,
+      "eval_steps_per_second": 4.159,
+      "eval_wer": 1.0,
+      "step": 390
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 5.962603878116343e-05,
+      "loss": 2.9145,
+      "step": 391
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 5.958448753462603e-05,
+      "loss": 2.9205,
+      "step": 392
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 5.954293628808864e-05,
+      "loss": 2.9231,
+      "step": 393
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 5.950138504155125e-05,
+      "loss": 2.9068,
+      "step": 394
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 5.945983379501384e-05,
+      "loss": 2.9633,
+      "step": 395
+    },
+    {
+      "epoch": 1.08,
+      "eval_loss": 2.9489939212799072,
+      "eval_runtime": 150.5099,
+      "eval_samples_per_second": 32.177,
+      "eval_steps_per_second": 4.026,
+      "eval_wer": 1.0,
+      "step": 395
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 5.941828254847645e-05,
+      "loss": 2.9099,
+      "step": 396
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 5.937673130193905e-05,
+      "loss": 2.9325,
+      "step": 397
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 5.933518005540166e-05,
+      "loss": 2.9444,
+      "step": 398
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 5.9293628808864254e-05,
+      "loss": 2.9118,
+      "step": 399
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 5.925207756232686e-05,
+      "loss": 2.9191,
+      "step": 400
+    },
+    {
+      "epoch": 1.1,
+      "eval_loss": 2.9361679553985596,
+      "eval_runtime": 140.9786,
+      "eval_samples_per_second": 34.353,
+      "eval_steps_per_second": 4.299,
+      "eval_wer": 1.0,
+      "step": 400
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 5.921052631578947e-05,
+      "loss": 2.9139,
+      "step": 401
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 5.916897506925207e-05,
+      "loss": 2.982,
+      "step": 402
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 5.9127423822714674e-05,
+      "loss": 2.9486,
+      "step": 403
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 5.9085872576177276e-05,
+      "loss": 2.8854,
+      "step": 404
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 5.9044321329639884e-05,
+      "loss": 2.9081,
+      "step": 405
+    },
+    {
+      "epoch": 1.11,
+      "eval_loss": 2.9393563270568848,
+      "eval_runtime": 138.9404,
+      "eval_samples_per_second": 34.857,
+      "eval_steps_per_second": 4.362,
+      "eval_wer": 1.0,
+      "step": 405
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 5.900277008310249e-05,
+      "loss": 2.9866,
+      "step": 406
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 5.896121883656509e-05,
+      "loss": 2.8782,
+      "step": 407
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 5.8919667590027695e-05,
+      "loss": 2.8814,
+      "step": 408
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 5.8878116343490303e-05,
+      "loss": 2.9144,
+      "step": 409
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 5.8836565096952905e-05,
+      "loss": 2.9381,
+      "step": 410
+    },
+    {
+      "epoch": 1.12,
+      "eval_loss": 2.9845528602600098,
+      "eval_runtime": 140.1554,
+      "eval_samples_per_second": 34.555,
+      "eval_steps_per_second": 4.324,
+      "eval_wer": 1.0,
+      "step": 410
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 5.8795013850415506e-05,
+      "loss": 2.9658,
+      "step": 411
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 5.875346260387811e-05,
+      "loss": 2.9623,
+      "step": 412
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 5.8711911357340716e-05,
+      "loss": 2.9131,
+      "step": 413
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 5.8670360110803325e-05,
+      "loss": 2.9526,
+      "step": 414
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 5.862880886426592e-05,
+      "loss": 2.9271,
+      "step": 415
+    },
+    {
+      "epoch": 1.14,
+      "eval_loss": 2.963838815689087,
+      "eval_runtime": 138.9896,
+      "eval_samples_per_second": 34.844,
+      "eval_steps_per_second": 4.36,
+      "eval_wer": 1.0,
+      "step": 415
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 5.858725761772853e-05,
+      "loss": 2.9842,
+      "step": 416
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 5.854570637119113e-05,
+      "loss": 2.96,
+      "step": 417
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 5.850415512465374e-05,
+      "loss": 2.9847,
+      "step": 418
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 5.846260387811633e-05,
+      "loss": 2.9779,
+      "step": 419
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 5.842105263157894e-05,
+      "loss": 2.959,
+      "step": 420
+    },
+    {
+      "epoch": 1.15,
+      "eval_loss": 2.9835448265075684,
+      "eval_runtime": 140.9902,
+      "eval_samples_per_second": 34.35,
+      "eval_steps_per_second": 4.298,
+      "eval_wer": 1.0,
+      "step": 420
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 5.837950138504155e-05,
+      "loss": 2.978,
+      "step": 421
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 5.833795013850415e-05,
+      "loss": 2.9544,
+      "step": 422
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 5.829639889196675e-05,
+      "loss": 2.9387,
+      "step": 423
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 5.825484764542935e-05,
+      "loss": 2.9354,
+      "step": 424
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 5.821329639889196e-05,
+      "loss": 2.9486,
+      "step": 425
+    },
+    {
+      "epoch": 1.16,
+      "eval_loss": 2.936093330383301,
+      "eval_runtime": 150.3452,
+      "eval_samples_per_second": 32.213,
+      "eval_steps_per_second": 4.031,
+      "eval_wer": 1.0,
+      "step": 425
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 5.817174515235457e-05,
+      "loss": 2.9728,
+      "step": 426
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 5.8130193905817165e-05,
+      "loss": 2.9476,
+      "step": 427
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 5.808864265927977e-05,
+      "loss": 2.9478,
+      "step": 428
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 5.8047091412742374e-05,
+      "loss": 2.9442,
+      "step": 429
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 5.800554016620498e-05,
+      "loss": 2.9246,
+      "step": 430
+    },
+    {
+      "epoch": 1.18,
+      "eval_loss": 2.961500644683838,
+      "eval_runtime": 140.2999,
+      "eval_samples_per_second": 34.519,
+      "eval_steps_per_second": 4.319,
+      "eval_wer": 1.0,
+      "step": 430
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 5.7963988919667584e-05,
+      "loss": 2.9591,
+      "step": 431
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 5.7922437673130186e-05,
+      "loss": 2.9663,
+      "step": 432
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 5.7880886426592794e-05,
+      "loss": 2.9332,
+      "step": 433
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 5.7839335180055396e-05,
+      "loss": 2.9235,
+      "step": 434
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 5.7797783933518e-05,
+      "loss": 2.923,
+      "step": 435
+    },
+    {
+      "epoch": 1.19,
+      "eval_loss": 2.9313013553619385,
+      "eval_runtime": 138.3532,
+      "eval_samples_per_second": 35.005,
+      "eval_steps_per_second": 4.38,
+      "eval_wer": 1.0,
+      "step": 435
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 5.7756232686980605e-05,
+      "loss": 2.9673,
+      "step": 436
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 5.771468144044321e-05,
+      "loss": 2.9402,
+      "step": 437
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 5.7673130193905815e-05,
+      "loss": 2.8955,
+      "step": 438
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 5.763157894736841e-05,
+      "loss": 2.9274,
+      "step": 439
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 5.759002770083102e-05,
+      "loss": 2.8908,
+      "step": 440
+    },
+    {
+      "epoch": 1.21,
+      "eval_loss": 2.936167001724243,
+      "eval_runtime": 138.4925,
+      "eval_samples_per_second": 34.969,
+      "eval_steps_per_second": 4.376,
+      "eval_wer": 1.0,
+      "step": 440
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 5.7548476454293627e-05,
+      "loss": 2.9045,
+      "step": 441
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 5.750692520775623e-05,
+      "loss": 2.9286,
+      "step": 442
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 5.746537396121883e-05,
+      "loss": 2.9256,
+      "step": 443
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 5.742382271468143e-05,
+      "loss": 2.9056,
+      "step": 444
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 5.738227146814404e-05,
+      "loss": 2.8976,
+      "step": 445
+    },
+    {
+      "epoch": 1.22,
+      "eval_loss": 2.92244553565979,
+      "eval_runtime": 143.3663,
+      "eval_samples_per_second": 33.781,
+      "eval_steps_per_second": 4.227,
+      "eval_wer": 1.0,
+      "step": 445
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 5.734072022160665e-05,
+      "loss": 2.8916,
+      "step": 446
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 5.729916897506925e-05,
+      "loss": 2.9318,
+      "step": 447
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 5.725761772853185e-05,
+      "loss": 2.9276,
+      "step": 448
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 5.721606648199445e-05,
+      "loss": 2.9009,
+      "step": 449
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 5.717451523545706e-05,
+      "loss": 2.9278,
+      "step": 450
+    },
+    {
+      "epoch": 1.23,
+      "eval_loss": 2.9276182651519775,
+      "eval_runtime": 142.4434,
+      "eval_samples_per_second": 33.999,
+      "eval_steps_per_second": 4.254,
+      "eval_wer": 1.0,
+      "step": 450
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 5.713296398891967e-05,
+      "loss": 2.8915,
+      "step": 451
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 5.7091412742382264e-05,
+      "loss": 2.898,
+      "step": 452
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 5.704986149584487e-05,
+      "loss": 2.89,
+      "step": 453
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 5.700831024930747e-05,
+      "loss": 2.8834,
+      "step": 454
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 5.696675900277008e-05,
+      "loss": 2.8429,
+      "step": 455
+    },
+    {
+      "epoch": 1.25,
+      "eval_loss": 2.92988657951355,
+      "eval_runtime": 145.7611,
+      "eval_samples_per_second": 33.226,
+      "eval_steps_per_second": 4.157,
+      "eval_wer": 1.0,
+      "step": 455
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 5.6925207756232676e-05,
+      "loss": 2.8771,
+      "step": 456
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 5.6883656509695285e-05,
+      "loss": 2.9245,
+      "step": 457
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 5.684210526315789e-05,
+      "loss": 2.9192,
+      "step": 458
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 5.6800554016620494e-05,
+      "loss": 2.9421,
+      "step": 459
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 5.6759002770083096e-05,
+      "loss": 2.867,
+      "step": 460
+    },
+    {
+      "epoch": 1.26,
+      "eval_loss": 2.9258229732513428,
+      "eval_runtime": 143.8058,
+      "eval_samples_per_second": 33.677,
+      "eval_steps_per_second": 4.214,
+      "eval_wer": 1.0,
+      "step": 460
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 5.6717451523545704e-05,
+      "loss": 2.879,
+      "step": 461
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 5.6675900277008306e-05,
+      "loss": 2.8976,
+      "step": 462
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 5.6634349030470914e-05,
+      "loss": 2.8255,
+      "step": 463
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 5.659279778393351e-05,
+      "loss": 2.8579,
+      "step": 464
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 5.655124653739612e-05,
+      "loss": 2.9734,
+      "step": 465
+    },
+    {
+      "epoch": 1.27,
+      "eval_loss": 2.928051233291626,
+      "eval_runtime": 139.3845,
+      "eval_samples_per_second": 34.746,
+      "eval_steps_per_second": 4.348,
+      "eval_wer": 0.9999716151007664,
+      "step": 465
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.6509695290858725e-05,
+      "loss": 2.996,
+      "step": 466
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.646814404432133e-05,
+      "loss": 2.9522,
+      "step": 467
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.642659279778393e-05,
+      "loss": 2.9662,
+      "step": 468
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 5.638504155124653e-05,
+      "loss": 2.9523,
+      "step": 469
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.634349030470914e-05,
+      "loss": 2.934,
+      "step": 470
+    },
+    {
+      "epoch": 1.29,
+      "eval_loss": 2.9229252338409424,
+      "eval_runtime": 140.441,
+      "eval_samples_per_second": 34.484,
+      "eval_steps_per_second": 4.315,
+      "eval_wer": 1.0,
+      "step": 470
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.6301939058171747e-05,
+      "loss": 2.9282,
+      "step": 471
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 5.626038781163434e-05,
+      "loss": 2.9611,
+      "step": 472
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.621883656509695e-05,
+      "loss": 2.955,
+      "step": 473
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.617728531855955e-05,
+      "loss": 2.9393,
+      "step": 474
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.613573407202216e-05,
+      "loss": 2.9521,
+      "step": 475
+    },
+    {
+      "epoch": 1.3,
+      "eval_loss": 2.913442611694336,
+      "eval_runtime": 149.0709,
+      "eval_samples_per_second": 32.488,
+      "eval_steps_per_second": 4.065,
+      "eval_wer": 1.0,
+      "step": 475
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 5.6094182825484754e-05,
+      "loss": 2.9447,
+      "step": 476
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.605263157894736e-05,
+      "loss": 2.93,
+      "step": 477
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.601108033240997e-05,
+      "loss": 2.9706,
+      "step": 478
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.596952908587257e-05,
+      "loss": 2.9684,
+      "step": 479
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 5.5927977839335174e-05,
+      "loss": 2.9098,
+      "step": 480
+    },
+    {
+      "epoch": 1.31,
+      "eval_loss": 2.9051098823547363,
+      "eval_runtime": 144.0225,
+      "eval_samples_per_second": 33.627,
+      "eval_steps_per_second": 4.208,
+      "eval_wer": 0.9992903775191598,
+      "step": 480
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5886426592797775e-05,
+      "loss": 2.9175,
+      "step": 481
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.5844875346260384e-05,
+      "loss": 2.9216,
+      "step": 482
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 5.580332409972299e-05,
+      "loss": 2.936,
+      "step": 483
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.5761772853185587e-05,
+      "loss": 2.9125,
+      "step": 484
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.5720221606648195e-05,
+      "loss": 2.9112,
+      "step": 485
+    },
+    {
+      "epoch": 1.33,
+      "eval_loss": 2.9028286933898926,
+      "eval_runtime": 154.3777,
+      "eval_samples_per_second": 31.371,
+      "eval_steps_per_second": 3.925,
+      "eval_wer": 0.9998864604030656,
+      "step": 485
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.5678670360110796e-05,
+      "loss": 2.9107,
+      "step": 486
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 5.5637119113573405e-05,
+      "loss": 2.931,
+      "step": 487
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.5595567867036006e-05,
+      "loss": 2.9035,
+      "step": 488
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.555401662049861e-05,
+      "loss": 2.8951,
+      "step": 489
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.5512465373961216e-05,
+      "loss": 2.8799,
+      "step": 490
+    },
+    {
+      "epoch": 1.34,
+      "eval_loss": 2.910067081451416,
+      "eval_runtime": 139.5277,
+      "eval_samples_per_second": 34.71,
+      "eval_steps_per_second": 4.343,
+      "eval_wer": 0.9986375248367868,
+      "step": 490
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 5.5470914127423824e-05,
+      "loss": 2.8484,
+      "step": 491
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.542936288088642e-05,
+      "loss": 2.9236,
+      "step": 492
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.538781163434903e-05,
+      "loss": 2.859,
+      "step": 493
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 5.534626038781163e-05,
+      "loss": 2.8699,
+      "step": 494
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 5.530470914127424e-05,
+      "loss": 2.857,
+      "step": 495
+    },
+    {
+      "epoch": 1.36,
+      "eval_loss": 2.9004716873168945,
+      "eval_runtime": 139.9768,
+      "eval_samples_per_second": 34.599,
+      "eval_steps_per_second": 4.329,
+      "eval_wer": 0.999205222821459,
+      "step": 495
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 5.526315789473683e-05,
+      "loss": 2.8653,
+      "step": 496
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 5.522160664819944e-05,
+      "loss": 2.8896,
+      "step": 497
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 5.518005540166205e-05,
+      "loss": 2.8958,
+      "step": 498
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 5.513850415512465e-05,
+      "loss": 2.8699,
+      "step": 499
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 5.509695290858725e-05,
+      "loss": 2.8525,
+      "step": 500
+    },
+    {
+      "epoch": 1.37,
+      "eval_loss": 2.893693685531616,
+      "eval_runtime": 139.5951,
+      "eval_samples_per_second": 34.693,
+      "eval_steps_per_second": 4.341,
+      "eval_wer": 1.0,
+      "step": 500
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 5.505540166204985e-05,
+      "loss": 2.8441,
+      "step": 501
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 5.501385041551246e-05,
+      "loss": 2.8695,
+      "step": 502
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 5.497229916897507e-05,
+      "loss": 2.8454,
+      "step": 503
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 5.4930747922437664e-05,
+      "loss": 2.8558,
+      "step": 504
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 5.488919667590027e-05,
+      "loss": 2.8682,
+      "step": 505
+    },
+    {
+      "epoch": 1.38,
+      "eval_loss": 2.8903708457946777,
+      "eval_runtime": 139.876,
+      "eval_samples_per_second": 34.624,
+      "eval_steps_per_second": 4.332,
+      "eval_wer": 0.9999716151007664,
+      "step": 505
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 5.4847645429362874e-05,
+      "loss": 2.8851,
+      "step": 506
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 5.480609418282548e-05,
+      "loss": 2.8332,
+      "step": 507
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 5.476454293628808e-05,
+      "loss": 2.8778,
+      "step": 508
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 5.4722991689750686e-05,
+      "loss": 2.8975,
+      "step": 509
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 5.4681440443213294e-05,
+      "loss": 2.8899,
+      "step": 510
+    },
+    {
+      "epoch": 1.4,
+      "eval_loss": 2.891392469406128,
+      "eval_runtime": 143.7998,
+      "eval_samples_per_second": 33.679,
+      "eval_steps_per_second": 4.214,
+      "eval_wer": 0.9964235026965654,
+      "step": 510
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 5.4639889196675895e-05,
+      "loss": 2.8337,
+      "step": 511
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 5.45983379501385e-05,
+      "loss": 2.8885,
+      "step": 512
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 5.4556786703601105e-05,
+      "loss": 2.8154,
+      "step": 513
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 5.451523545706371e-05,
+      "loss": 2.8331,
+      "step": 514
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 5.4473684210526315e-05,
+      "loss": 2.7475,
+      "step": 515
+    },
+    {
+      "epoch": 1.41,
+      "eval_loss": 2.8841705322265625,
+      "eval_runtime": 143.1862,
+      "eval_samples_per_second": 33.823,
+      "eval_steps_per_second": 4.232,
+      "eval_wer": 0.9949758728356515,
+      "step": 515
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 5.443213296398891e-05,
+      "loss": 2.9662,
+      "step": 516
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.439058171745152e-05,
+      "loss": 2.9371,
+      "step": 517
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.4349030470914126e-05,
+      "loss": 2.951,
+      "step": 518
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.430747922437673e-05,
+      "loss": 2.9043,
+      "step": 519
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.426592797783933e-05,
+      "loss": 2.9263,
+      "step": 520
+    },
+    {
+      "epoch": 1.42,
+      "eval_loss": 2.885201930999756,
+      "eval_runtime": 140.0653,
+      "eval_samples_per_second": 34.577,
+      "eval_steps_per_second": 4.327,
+      "eval_wer": 0.9971898949758728,
+      "step": 520
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.422437673130193e-05,
+      "loss": 2.9212,
+      "step": 521
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.418282548476454e-05,
+      "loss": 2.907,
+      "step": 522
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.414127423822715e-05,
+      "loss": 2.8914,
+      "step": 523
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.409972299168974e-05,
+      "loss": 2.9022,
+      "step": 524
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.405817174515235e-05,
+      "loss": 2.8603,
+      "step": 525
+    },
+    {
+      "epoch": 1.44,
+      "eval_loss": 2.8762059211730957,
+      "eval_runtime": 139.2528,
+      "eval_samples_per_second": 34.778,
+      "eval_steps_per_second": 4.352,
+      "eval_wer": 0.9966221969912007,
+      "step": 525
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.401662049861495e-05,
+      "loss": 2.8961,
+      "step": 526
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.397506925207756e-05,
+      "loss": 2.8595,
+      "step": 527
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.3933518005540155e-05,
+      "loss": 2.8559,
+      "step": 528
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.389196675900276e-05,
+      "loss": 2.8631,
+      "step": 529
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.385041551246537e-05,
+      "loss": 2.864,
+      "step": 530
+    },
+    {
+      "epoch": 1.45,
+      "eval_loss": 2.8680307865142822,
+      "eval_runtime": 139.0826,
+      "eval_samples_per_second": 34.821,
+      "eval_steps_per_second": 4.357,
+      "eval_wer": 0.997757592960545,
+      "step": 530
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.380886426592797e-05,
+      "loss": 2.8688,
+      "step": 531
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.3767313019390575e-05,
+      "loss": 2.8695,
+      "step": 532
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.3725761772853176e-05,
+      "loss": 2.8508,
+      "step": 533
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.3684210526315784e-05,
+      "loss": 2.8775,
+      "step": 534
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.364265927977839e-05,
+      "loss": 2.8632,
+      "step": 535
+    },
+    {
+      "epoch": 1.47,
+      "eval_loss": 2.860178232192993,
+      "eval_runtime": 141.1033,
+      "eval_samples_per_second": 34.322,
+      "eval_steps_per_second": 4.295,
+      "eval_wer": 0.9963951177973318,
+      "step": 535
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.360110803324099e-05,
+      "loss": 2.8687,
+      "step": 536
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.3559556786703596e-05,
+      "loss": 2.8723,
+      "step": 537
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.35180055401662e-05,
+      "loss": 2.8284,
+      "step": 538
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3476454293628806e-05,
+      "loss": 2.8968,
+      "step": 539
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.343490304709141e-05,
+      "loss": 2.9289,
+      "step": 540
+    },
+    {
+      "epoch": 1.48,
+      "eval_loss": 2.8583950996398926,
+      "eval_runtime": 138.2843,
+      "eval_samples_per_second": 35.022,
+      "eval_steps_per_second": 4.382,
+      "eval_wer": 0.9952313369287539,
+      "step": 540
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.339335180055401e-05,
+      "loss": 2.8136,
+      "step": 541
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.335180055401662e-05,
+      "loss": 2.8341,
+      "step": 542
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.3310249307479225e-05,
+      "loss": 2.8136,
+      "step": 543
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.326869806094182e-05,
+      "loss": 2.7994,
+      "step": 544
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.322714681440443e-05,
+      "loss": 2.8689,
+      "step": 545
+    },
+    {
+      "epoch": 1.49,
+      "eval_loss": 2.8587305545806885,
+      "eval_runtime": 149.9846,
+      "eval_samples_per_second": 32.29,
+      "eval_steps_per_second": 4.04,
+      "eval_wer": 0.9956003406187908,
+      "step": 545
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.318559556786703e-05,
+      "loss": 2.8277,
+      "step": 546
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.314404432132964e-05,
+      "loss": 2.8735,
+      "step": 547
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.310249307479223e-05,
+      "loss": 2.7942,
+      "step": 548
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.306094182825484e-05,
+      "loss": 2.8556,
+      "step": 549
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.301939058171745e-05,
+      "loss": 2.8304,
+      "step": 550
+    },
+    {
+      "epoch": 1.51,
+      "eval_loss": 2.851149082183838,
+      "eval_runtime": 142.6259,
+      "eval_samples_per_second": 33.956,
+      "eval_steps_per_second": 4.249,
+      "eval_wer": 0.9993187624183935,
+      "step": 550
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.297783933518005e-05,
+      "loss": 2.8442,
+      "step": 551
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.293628808864266e-05,
+      "loss": 2.8402,
+      "step": 552
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.2894736842105254e-05,
+      "loss": 2.7748,
+      "step": 553
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 5.285318559556786e-05,
+      "loss": 2.8555,
+      "step": 554
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 5.281163434903047e-05,
+      "loss": 2.8024,
+      "step": 555
+    },
+    {
+      "epoch": 1.52,
+      "eval_loss": 2.845991849899292,
+      "eval_runtime": 140.0808,
+      "eval_samples_per_second": 34.573,
+      "eval_steps_per_second": 4.326,
+      "eval_wer": 1.0,
+      "step": 555
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 5.277008310249307e-05,
+      "loss": 2.7621,
+      "step": 556
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 5.2728531855955674e-05,
+      "loss": 2.7169,
+      "step": 557
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 5.2686980609418275e-05,
+      "loss": 2.8385,
+      "step": 558
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 5.264542936288088e-05,
+      "loss": 2.7846,
+      "step": 559
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 5.260387811634349e-05,
+      "loss": 2.7649,
+      "step": 560
+    },
+    {
+      "epoch": 1.53,
+      "eval_loss": 2.8460400104522705,
+      "eval_runtime": 140.2323,
+      "eval_samples_per_second": 34.536,
+      "eval_steps_per_second": 4.321,
+      "eval_wer": 1.0000283848992335,
+      "step": 560
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 5.2562326869806086e-05,
+      "loss": 2.7549,
+      "step": 561
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 5.2520775623268695e-05,
+      "loss": 2.7807,
+      "step": 562
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 5.2479224376731296e-05,
+      "loss": 2.8133,
+      "step": 563
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 5.2437673130193904e-05,
+      "loss": 2.8473,
+      "step": 564
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 5.2396121883656506e-05,
+      "loss": 2.8756,
+      "step": 565
+    },
+    {
+      "epoch": 1.55,
+      "eval_loss": 2.834791660308838,
+      "eval_runtime": 146.243,
+      "eval_samples_per_second": 33.116,
+      "eval_steps_per_second": 4.144,
+      "eval_wer": 0.9987226795344877,
+      "step": 565
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 5.235457063711911e-05,
+      "loss": 2.913,
+      "step": 566
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 5.2313019390581716e-05,
+      "loss": 2.8613,
+      "step": 567
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 5.227146814404432e-05,
+      "loss": 2.892,
+      "step": 568
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 5.222991689750692e-05,
+      "loss": 2.8786,
+      "step": 569
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 5.218836565096953e-05,
+      "loss": 2.8808,
+      "step": 570
+    },
+    {
+      "epoch": 1.56,
+      "eval_loss": 2.8539350032806396,
+      "eval_runtime": 140.0537,
+      "eval_samples_per_second": 34.58,
+      "eval_steps_per_second": 4.327,
+      "eval_wer": 0.9993187624183935,
+      "step": 570
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 5.214681440443213e-05,
+      "loss": 2.8789,
+      "step": 571
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 5.210526315789474e-05,
+      "loss": 2.8492,
+      "step": 572
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 5.206371191135733e-05,
+      "loss": 2.8207,
+      "step": 573
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 5.202216066481994e-05,
+      "loss": 2.869,
+      "step": 574
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 5.198060941828255e-05,
+      "loss": 2.9027,
+      "step": 575
+    },
+    {
+      "epoch": 1.57,
+      "eval_loss": 2.8282339572906494,
+      "eval_runtime": 142.3231,
+      "eval_samples_per_second": 34.028,
+      "eval_steps_per_second": 4.258,
+      "eval_wer": 0.9975021288674425,
+      "step": 575
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 5.193905817174515e-05,
+      "loss": 2.8345,
+      "step": 576
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 5.189750692520775e-05,
+      "loss": 2.8725,
+      "step": 577
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 5.185595567867035e-05,
+      "loss": 2.8619,
+      "step": 578
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 5.181440443213296e-05,
+      "loss": 2.7877,
+      "step": 579
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 5.177285318559557e-05,
+      "loss": 2.8586,
+      "step": 580
+    },
+    {
+      "epoch": 1.59,
+      "eval_loss": 2.8288161754608154,
+      "eval_runtime": 140.1893,
+      "eval_samples_per_second": 34.546,
+      "eval_steps_per_second": 4.323,
+      "eval_wer": 0.9975872835651434,
+      "step": 580
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 5.1731301939058164e-05,
+      "loss": 2.7929,
+      "step": 581
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 5.168975069252077e-05,
+      "loss": 2.8394,
+      "step": 582
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.1648199445983374e-05,
+      "loss": 2.7996,
+      "step": 583
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.160664819944598e-05,
+      "loss": 2.8532,
+      "step": 584
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.156509695290858e-05,
+      "loss": 2.8193,
+      "step": 585
+    },
+    {
+      "epoch": 1.6,
+      "eval_loss": 2.81013822555542,
+      "eval_runtime": 142.9864,
+      "eval_samples_per_second": 33.87,
+      "eval_steps_per_second": 4.238,
+      "eval_wer": 1.0050525120635823,
+      "step": 585
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 5.1523545706371185e-05,
+      "loss": 2.8696,
+      "step": 586
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.1481994459833794e-05,
+      "loss": 2.8035,
+      "step": 587
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.1440443213296395e-05,
+      "loss": 2.8114,
+      "step": 588
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 5.1398891966758997e-05,
+      "loss": 2.8353,
+      "step": 589
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.13573407202216e-05,
+      "loss": 2.811,
+      "step": 590
+    },
+    {
+      "epoch": 1.62,
+      "eval_loss": 2.7964913845062256,
+      "eval_runtime": 139.0817,
+      "eval_samples_per_second": 34.821,
+      "eval_steps_per_second": 4.357,
+      "eval_wer": 1.0014192449616803,
+      "step": 590
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.1315789473684206e-05,
+      "loss": 2.751,
+      "step": 591
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.1274238227146815e-05,
+      "loss": 2.8043,
+      "step": 592
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 5.123268698060941e-05,
+      "loss": 2.7356,
+      "step": 593
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.119113573407202e-05,
+      "loss": 2.7633,
+      "step": 594
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.1149584487534626e-05,
+      "loss": 2.7332,
+      "step": 595
+    },
+    {
+      "epoch": 1.63,
+      "eval_loss": 2.7884466648101807,
+      "eval_runtime": 152.6632,
+      "eval_samples_per_second": 31.723,
+      "eval_steps_per_second": 3.97,
+      "eval_wer": 1.0025830258302584,
+      "step": 595
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.110803324099723e-05,
+      "loss": 2.7712,
+      "step": 596
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 5.106648199445983e-05,
+      "loss": 2.7643,
+      "step": 597
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.102493074792243e-05,
+      "loss": 2.7129,
+      "step": 598
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.098337950138504e-05,
+      "loss": 2.694,
+      "step": 599
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 5.094182825484765e-05,
+      "loss": 2.7717,
+      "step": 600
+    },
+    {
+      "epoch": 1.64,
+      "eval_loss": 2.7883124351501465,
+      "eval_runtime": 139.7359,
+      "eval_samples_per_second": 34.658,
+      "eval_steps_per_second": 4.337,
+      "eval_wer": 1.0060459835367583,
+      "step": 600
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.090027700831024e-05,
+      "loss": 2.8192,
+      "step": 601
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.085872576177285e-05,
+      "loss": 2.7632,
+      "step": 602
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.081717451523545e-05,
+      "loss": 2.8272,
+      "step": 603
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 5.077562326869806e-05,
+      "loss": 2.7025,
+      "step": 604
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.0734072022160655e-05,
+      "loss": 2.6901,
+      "step": 605
+    },
+    {
+      "epoch": 1.66,
+      "eval_loss": 2.7800726890563965,
+      "eval_runtime": 149.872,
+      "eval_samples_per_second": 32.314,
+      "eval_steps_per_second": 4.043,
+      "eval_wer": 0.9974453590689754,
+      "step": 605
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.069252077562326e-05,
+      "loss": 2.7073,
+      "step": 606
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.065096952908587e-05,
+      "loss": 2.7538,
+      "step": 607
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 5.060941828254847e-05,
+      "loss": 2.7703,
+      "step": 608
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.0567867036011074e-05,
+      "loss": 2.7492,
+      "step": 609
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.0526315789473676e-05,
+      "loss": 2.6905,
+      "step": 610
+    },
+    {
+      "epoch": 1.67,
+      "eval_loss": 2.811263084411621,
+      "eval_runtime": 143.6498,
+      "eval_samples_per_second": 33.714,
+      "eval_steps_per_second": 4.219,
+      "eval_wer": 0.996820891285836,
+      "step": 610
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 5.0484764542936284e-05,
+      "loss": 2.7315,
+      "step": 611
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 5.044321329639889e-05,
+      "loss": 2.7007,
+      "step": 612
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 5.040166204986149e-05,
+      "loss": 2.7963,
+      "step": 613
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 5.0360110803324095e-05,
+      "loss": 2.7419,
+      "step": 614
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 5.03185595567867e-05,
+      "loss": 2.7442,
+      "step": 615
+    },
+    {
+      "epoch": 1.68,
+      "eval_loss": 2.8113441467285156,
+      "eval_runtime": 138.6884,
+      "eval_samples_per_second": 34.92,
+      "eval_steps_per_second": 4.37,
+      "eval_wer": 1.0006812375816065,
+      "step": 615
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 5.0277008310249305e-05,
+      "loss": 2.9024,
+      "step": 616
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 5.023545706371191e-05,
+      "loss": 2.8703,
+      "step": 617
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 5.019390581717451e-05,
+      "loss": 2.8659,
+      "step": 618
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 5.015235457063712e-05,
+      "loss": 2.8224,
+      "step": 619
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 5.011080332409972e-05,
+      "loss": 2.8431,
+      "step": 620
+    },
+    {
+      "epoch": 1.7,
+      "eval_loss": 2.8152356147766113,
+      "eval_runtime": 150.0291,
+      "eval_samples_per_second": 32.28,
+      "eval_steps_per_second": 4.039,
+      "eval_wer": 1.0343173431734318,
+      "step": 620
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 5.006925207756232e-05,
+      "loss": 2.8095,
+      "step": 621
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 5.002770083102493e-05,
+      "loss": 2.814,
+      "step": 622
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.998614958448753e-05,
+      "loss": 2.8354,
+      "step": 623
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.994459833795014e-05,
+      "loss": 2.7698,
+      "step": 624
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.990304709141273e-05,
+      "loss": 2.8028,
+      "step": 625
+    },
+    {
+      "epoch": 1.71,
+      "eval_loss": 2.7790422439575195,
+      "eval_runtime": 139.093,
+      "eval_samples_per_second": 34.818,
+      "eval_steps_per_second": 4.357,
+      "eval_wer": 1.025035481124042,
+      "step": 625
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 4.986149584487534e-05,
+      "loss": 2.7882,
+      "step": 626
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.981994459833795e-05,
+      "loss": 2.8561,
+      "step": 627
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.977839335180055e-05,
+      "loss": 2.7917,
+      "step": 628
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 4.973684210526315e-05,
+      "loss": 2.7962,
+      "step": 629
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.9695290858725754e-05,
+      "loss": 2.7151,
+      "step": 630
+    },
+    {
+      "epoch": 1.73,
+      "eval_loss": 2.7653234004974365,
+      "eval_runtime": 138.9273,
+      "eval_samples_per_second": 34.86,
+      "eval_steps_per_second": 4.362,
+      "eval_wer": 1.0286687482259438,
+      "step": 630
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.965373961218836e-05,
+      "loss": 2.7514,
+      "step": 631
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.961218836565097e-05,
+      "loss": 2.7411,
+      "step": 632
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 4.9570637119113565e-05,
+      "loss": 2.7896,
+      "step": 633
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.952908587257617e-05,
+      "loss": 2.7212,
+      "step": 634
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.9487534626038775e-05,
+      "loss": 2.7405,
+      "step": 635
+    },
+    {
+      "epoch": 1.74,
+      "eval_loss": 2.77144193649292,
+      "eval_runtime": 138.7362,
+      "eval_samples_per_second": 34.908,
+      "eval_steps_per_second": 4.368,
+      "eval_wer": 1.1303434572807267,
+      "step": 635
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.944598337950138e-05,
+      "loss": 2.7499,
+      "step": 636
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 4.940443213296398e-05,
+      "loss": 2.7469,
+      "step": 637
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.9362880886426586e-05,
+      "loss": 2.7052,
+      "step": 638
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.9321329639889194e-05,
+      "loss": 2.7415,
+      "step": 639
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 4.9279778393351796e-05,
+      "loss": 2.7566,
+      "step": 640
+    },
+    {
+      "epoch": 1.75,
+      "eval_loss": 2.7487998008728027,
+      "eval_runtime": 140.2245,
+      "eval_samples_per_second": 34.537,
+      "eval_steps_per_second": 4.322,
+      "eval_wer": 1.031195004257735,
+      "step": 640
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.92382271468144e-05,
+      "loss": 2.7278,
+      "step": 641
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.9196675900277e-05,
+      "loss": 2.6935,
+      "step": 642
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.915512465373961e-05,
+      "loss": 2.7563,
+      "step": 643
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 4.9113573407202216e-05,
+      "loss": 2.7078,
+      "step": 644
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.907202216066481e-05,
+      "loss": 2.7337,
+      "step": 645
+    },
+    {
+      "epoch": 1.77,
+      "eval_loss": 2.7498059272766113,
+      "eval_runtime": 146.1642,
+      "eval_samples_per_second": 33.134,
+      "eval_steps_per_second": 4.146,
+      "eval_wer": 1.0175986375248367,
+      "step": 645
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.903047091412742e-05,
+      "loss": 2.6966,
+      "step": 646
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.898891966759003e-05,
+      "loss": 2.7468,
+      "step": 647
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 4.894736842105263e-05,
+      "loss": 2.7255,
+      "step": 648
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.890581717451523e-05,
+      "loss": 2.6999,
+      "step": 649
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.886426592797783e-05,
+      "loss": 2.7486,
+      "step": 650
+    },
+    {
+      "epoch": 1.78,
+      "eval_loss": 2.7496137619018555,
+      "eval_runtime": 150.8065,
+      "eval_samples_per_second": 32.114,
+      "eval_steps_per_second": 4.018,
+      "eval_wer": 1.0760147601476016,
+      "step": 650
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 4.882271468144044e-05,
+      "loss": 2.6494,
+      "step": 651
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.878116343490305e-05,
+      "loss": 2.7285,
+      "step": 652
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.873961218836564e-05,
+      "loss": 2.7431,
+      "step": 653
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.869806094182825e-05,
+      "loss": 2.689,
+      "step": 654
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 4.865650969529085e-05,
+      "loss": 2.6918,
+      "step": 655
+    },
+    {
+      "epoch": 1.79,
+      "eval_loss": 2.7391245365142822,
+      "eval_runtime": 147.4606,
+      "eval_samples_per_second": 32.843,
+      "eval_steps_per_second": 4.11,
+      "eval_wer": 1.035310814646608,
+      "step": 655
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.861495844875346e-05,
+      "loss": 2.7717,
+      "step": 656
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.8573407202216056e-05,
+      "loss": 2.6603,
+      "step": 657
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.8531855955678664e-05,
+      "loss": 2.7325,
+      "step": 658
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 4.849030470914127e-05,
+      "loss": 2.6566,
+      "step": 659
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.8448753462603874e-05,
+      "loss": 2.7142,
+      "step": 660
+    },
+    {
+      "epoch": 1.81,
+      "eval_loss": 2.7500298023223877,
+      "eval_runtime": 139.4821,
+      "eval_samples_per_second": 34.721,
+      "eval_steps_per_second": 4.345,
+      "eval_wer": 1.0283281294351405,
+      "step": 660
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.840720221606648e-05,
+      "loss": 2.7783,
+      "step": 661
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 4.836565096952908e-05,
+      "loss": 2.6976,
+      "step": 662
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.8324099722991685e-05,
+      "loss": 2.6649,
+      "step": 663
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.828254847645429e-05,
+      "loss": 2.7258,
+      "step": 664
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.8240997229916895e-05,
+      "loss": 2.7057,
+      "step": 665
+    },
+    {
+      "epoch": 1.82,
+      "eval_loss": 2.761242389678955,
+      "eval_runtime": 138.8799,
+      "eval_samples_per_second": 34.872,
+      "eval_steps_per_second": 4.363,
+      "eval_wer": 1.0127164348566562,
+      "step": 665
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 4.8199445983379496e-05,
+      "loss": 2.8234,
+      "step": 666
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.81578947368421e-05,
+      "loss": 2.8261,
+      "step": 667
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.8116343490304706e-05,
+      "loss": 2.7821,
+      "step": 668
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.8074792243767314e-05,
+      "loss": 2.8253,
+      "step": 669
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 4.803324099722991e-05,
+      "loss": 2.8348,
+      "step": 670
+    },
+    {
+      "epoch": 1.83,
+      "eval_loss": 2.7440872192382812,
+      "eval_runtime": 139.2265,
+      "eval_samples_per_second": 34.785,
+      "eval_steps_per_second": 4.353,
+      "eval_wer": 1.005648594947488,
+      "step": 670
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.799168975069252e-05,
+      "loss": 2.7736,
+      "step": 671
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.795013850415512e-05,
+      "loss": 2.7853,
+      "step": 672
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 4.790858725761773e-05,
+      "loss": 2.7833,
+      "step": 673
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.786703601108033e-05,
+      "loss": 2.7472,
+      "step": 674
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.782548476454293e-05,
+      "loss": 2.705,
+      "step": 675
+    },
+    {
+      "epoch": 1.85,
+      "eval_loss": 2.7473442554473877,
+      "eval_runtime": 138.9306,
+      "eval_samples_per_second": 34.859,
+      "eval_steps_per_second": 4.362,
+      "eval_wer": 1.051887595799035,
+      "step": 675
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.778393351800554e-05,
+      "loss": 2.7458,
+      "step": 676
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 4.774238227146815e-05,
+      "loss": 2.7273,
+      "step": 677
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.770083102493074e-05,
+      "loss": 2.7563,
+      "step": 678
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.765927977839335e-05,
+      "loss": 2.707,
+      "step": 679
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.761772853185595e-05,
+      "loss": 2.7547,
+      "step": 680
+    },
+    {
+      "epoch": 1.86,
+      "eval_loss": 2.721618890762329,
+      "eval_runtime": 142.5446,
+      "eval_samples_per_second": 33.975,
+      "eval_steps_per_second": 4.251,
+      "eval_wer": 1.0218279875106444,
+      "step": 680
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 4.757617728531856e-05,
+      "loss": 2.7561,
+      "step": 681
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.7534626038781154e-05,
+      "loss": 2.7061,
+      "step": 682
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.749307479224376e-05,
+      "loss": 2.7269,
+      "step": 683
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 4.745152354570637e-05,
+      "loss": 2.7382,
+      "step": 684
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.740997229916897e-05,
+      "loss": 2.7045,
+      "step": 685
+    },
+    {
+      "epoch": 1.88,
+      "eval_loss": 2.726120948791504,
+      "eval_runtime": 139.9212,
+      "eval_samples_per_second": 34.612,
+      "eval_steps_per_second": 4.331,
+      "eval_wer": 1.1413851830826,
+      "step": 685
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.7368421052631574e-05,
+      "loss": 2.7189,
+      "step": 686
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.7326869806094176e-05,
+      "loss": 2.7367,
+      "step": 687
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 4.7285318559556784e-05,
+      "loss": 2.7164,
+      "step": 688
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.724376731301939e-05,
+      "loss": 2.7778,
+      "step": 689
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.720221606648199e-05,
+      "loss": 2.7121,
+      "step": 690
+    },
+    {
+      "epoch": 1.89,
+      "eval_loss": 2.722256898880005,
+      "eval_runtime": 151.9107,
+      "eval_samples_per_second": 31.881,
+      "eval_steps_per_second": 3.989,
+      "eval_wer": 1.028725518024411,
+      "step": 690
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.7160664819944595e-05,
+      "loss": 2.7102,
+      "step": 691
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 4.71191135734072e-05,
+      "loss": 2.7099,
+      "step": 692
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.7077562326869805e-05,
+      "loss": 2.7525,
+      "step": 693
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.70360110803324e-05,
+      "loss": 2.716,
+      "step": 694
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 4.699445983379501e-05,
+      "loss": 2.6877,
+      "step": 695
+    },
+    {
+      "epoch": 1.9,
+      "eval_loss": 2.7282636165618896,
+      "eval_runtime": 139.5522,
+      "eval_samples_per_second": 34.704,
+      "eval_steps_per_second": 4.342,
+      "eval_wer": 1.027419812659665,
+      "step": 695
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.6952908587257616e-05,
+      "loss": 2.6912,
+      "step": 696
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.691135734072022e-05,
+      "loss": 2.7023,
+      "step": 697
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.686980609418282e-05,
+      "loss": 2.6557,
+      "step": 698
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 4.682825484764543e-05,
+      "loss": 2.6696,
+      "step": 699
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.678670360110803e-05,
+      "loss": 2.6879,
+      "step": 700
+    },
+    {
+      "epoch": 1.92,
+      "eval_loss": 2.7450602054595947,
+      "eval_runtime": 142.4033,
+      "eval_samples_per_second": 34.009,
+      "eval_steps_per_second": 4.256,
+      "eval_wer": 1.132188475730911,
+      "step": 700
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.674515235457064e-05,
+      "loss": 2.76,
+      "step": 701
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.670360110803323e-05,
+      "loss": 2.6698,
+      "step": 702
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 4.666204986149584e-05,
+      "loss": 2.6865,
+      "step": 703
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.662049861495845e-05,
+      "loss": 2.5945,
+      "step": 704
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.657894736842105e-05,
+      "loss": 2.6958,
+      "step": 705
+    },
+    {
+      "epoch": 1.93,
+      "eval_loss": 2.7165942192077637,
+      "eval_runtime": 142.805,
+      "eval_samples_per_second": 33.913,
+      "eval_steps_per_second": 4.244,
+      "eval_wer": 1.0363610559182515,
+      "step": 705
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 4.653739612188365e-05,
+      "loss": 2.6767,
+      "step": 706
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.649584487534625e-05,
+      "loss": 2.7,
+      "step": 707
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.645429362880886e-05,
+      "loss": 2.7405,
+      "step": 708
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.641274238227147e-05,
+      "loss": 2.6583,
+      "step": 709
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 4.6371191135734065e-05,
+      "loss": 2.6692,
+      "step": 710
+    },
+    {
+      "epoch": 1.94,
+      "eval_loss": 2.714829206466675,
+      "eval_runtime": 139.5631,
+      "eval_samples_per_second": 34.701,
+      "eval_steps_per_second": 4.342,
+      "eval_wer": 1.0074368435992052,
+      "step": 710
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.632963988919667e-05,
+      "loss": 2.6923,
+      "step": 711
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.6288088642659275e-05,
+      "loss": 2.6492,
+      "step": 712
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.624653739612188e-05,
+      "loss": 2.6941,
+      "step": 713
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 4.620498614958448e-05,
+      "loss": 2.7656,
+      "step": 714
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.6163434903047086e-05,
+      "loss": 2.5786,
+      "step": 715
+    },
+    {
+      "epoch": 1.96,
+      "eval_loss": 2.71008563041687,
+      "eval_runtime": 139.9583,
+      "eval_samples_per_second": 34.603,
+      "eval_steps_per_second": 4.33,
+      "eval_wer": 1.0503831961396537,
+      "step": 715
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.6121883656509694e-05,
+      "loss": 2.7607,
+      "step": 716
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 4.6080332409972296e-05,
+      "loss": 2.735,
+      "step": 717
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.60387811634349e-05,
+      "loss": 2.7587,
+      "step": 718
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.59972299168975e-05,
+      "loss": 2.7283,
+      "step": 719
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.595567867036011e-05,
+      "loss": 2.6919,
+      "step": 720
+    },
+    {
+      "epoch": 1.97,
+      "eval_loss": 2.6963367462158203,
+      "eval_runtime": 147.8298,
+      "eval_samples_per_second": 32.761,
+      "eval_steps_per_second": 4.099,
+      "eval_wer": 1.0453874538745387,
+      "step": 720
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 4.5914127423822715e-05,
+      "loss": 2.701,
+      "step": 721
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.587257617728531e-05,
+      "loss": 2.7177,
+      "step": 722
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.583102493074792e-05,
+      "loss": 2.7435,
+      "step": 723
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.578947368421052e-05,
+      "loss": 2.6731,
+      "step": 724
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 4.574792243767313e-05,
+      "loss": 2.7256,
+      "step": 725
+    },
+    {
+      "epoch": 1.98,
+      "eval_loss": 2.720106363296509,
+      "eval_runtime": 139.368,
+      "eval_samples_per_second": 34.75,
+      "eval_steps_per_second": 4.348,
+      "eval_wer": 1.0348566562588704,
+      "step": 725
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.570637119113573e-05,
+      "loss": 2.6877,
+      "step": 726
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.566481994459833e-05,
+      "loss": 2.6487,
+      "step": 727
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 4.562326869806094e-05,
+      "loss": 2.654,
+      "step": 728
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.558171745152355e-05,
+      "loss": 2.6796,
+      "step": 729
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.554016620498614e-05,
+      "loss": 2.6507,
+      "step": 730
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 2.7099249362945557,
+      "eval_runtime": 154.6571,
+      "eval_samples_per_second": 31.314,
+      "eval_steps_per_second": 3.918,
+      "eval_wer": 1.1339199545841612,
+      "step": 730
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 4.549861495844875e-05,
+      "loss": 4.2501,
+      "step": 731
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.545706371191135e-05,
+      "loss": 2.763,
+      "step": 732
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.541551246537396e-05,
+      "loss": 2.7169,
+      "step": 733
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.5373961218836555e-05,
+      "loss": 2.7634,
+      "step": 734
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 4.5332409972299164e-05,
+      "loss": 2.7833,
+      "step": 735
+    },
+    {
+      "epoch": 2.01,
+      "eval_loss": 2.7110776901245117,
+      "eval_runtime": 149.2864,
+      "eval_samples_per_second": 32.441,
+      "eval_steps_per_second": 4.059,
+      "eval_wer": 1.0124042009650867,
+      "step": 735
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.529085872576177e-05,
+      "loss": 2.7391,
+      "step": 736
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5249307479224373e-05,
+      "loss": 2.7287,
+      "step": 737
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5207756232686975e-05,
+      "loss": 2.7202,
+      "step": 738
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 4.5166204986149576e-05,
+      "loss": 2.73,
+      "step": 739
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.5124653739612185e-05,
+      "loss": 2.7521,
+      "step": 740
+    },
+    {
+      "epoch": 2.03,
+      "eval_loss": 2.70237398147583,
+      "eval_runtime": 141.0255,
+      "eval_samples_per_second": 34.341,
+      "eval_steps_per_second": 4.297,
+      "eval_wer": 1.0275333522565995,
+      "step": 740
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.508310249307479e-05,
+      "loss": 2.7051,
+      "step": 741
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 4.504155124653739e-05,
+      "loss": 2.7435,
+      "step": 742
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 2.7306,
+      "step": 743
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.49584487534626e-05,
+      "loss": 2.7458,
+      "step": 744
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.4916897506925206e-05,
+      "loss": 2.6732,
+      "step": 745
+    },
+    {
+      "epoch": 2.04,
+      "eval_loss": 2.7057559490203857,
+      "eval_runtime": 147.4805,
+      "eval_samples_per_second": 32.838,
+      "eval_steps_per_second": 4.109,
+      "eval_wer": 1.0646608004541585,
+      "step": 745
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 4.48753462603878e-05,
+      "loss": 2.7154,
+      "step": 746
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.483379501385041e-05,
+      "loss": 2.6876,
+      "step": 747
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.479224376731302e-05,
+      "loss": 2.7409,
+      "step": 748
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.475069252077562e-05,
+      "loss": 2.6881,
+      "step": 749
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 4.470914127423822e-05,
+      "loss": 2.719,
+      "step": 750
+    },
+    {
+      "epoch": 2.05,
+      "eval_loss": 2.720003604888916,
+      "eval_runtime": 138.8115,
+      "eval_samples_per_second": 34.889,
+      "eval_steps_per_second": 4.366,
+      "eval_wer": 1.0210899801305706,
+      "step": 750
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.466759002770083e-05,
+      "loss": 2.7685,
+      "step": 751
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.462603878116343e-05,
+      "loss": 2.7488,
+      "step": 752
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 4.458448753462604e-05,
+      "loss": 2.6918,
+      "step": 753
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.454293628808863e-05,
+      "loss": 2.7196,
+      "step": 754
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.450138504155124e-05,
+      "loss": 2.701,
+      "step": 755
+    },
+    {
+      "epoch": 2.07,
+      "eval_loss": 2.702404260635376,
+      "eval_runtime": 140.4098,
+      "eval_samples_per_second": 34.492,
+      "eval_steps_per_second": 4.316,
+      "eval_wer": 1.080811808118081,
+      "step": 755
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.445983379501385e-05,
+      "loss": 2.6268,
+      "step": 756
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 4.441828254847645e-05,
+      "loss": 2.6787,
+      "step": 757
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.437673130193905e-05,
+      "loss": 2.6857,
+      "step": 758
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.4335180055401654e-05,
+      "loss": 2.7008,
+      "step": 759
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.429362880886426e-05,
+      "loss": 2.6444,
+      "step": 760
+    },
+    {
+      "epoch": 2.08,
+      "eval_loss": 2.6812727451324463,
+      "eval_runtime": 140.0447,
+      "eval_samples_per_second": 34.582,
+      "eval_steps_per_second": 4.327,
+      "eval_wer": 1.058245813227363,
+      "step": 760
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 4.425207756232687e-05,
+      "loss": 2.6699,
+      "step": 761
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4210526315789466e-05,
+      "loss": 2.6386,
+      "step": 762
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4168975069252074e-05,
+      "loss": 2.6532,
+      "step": 763
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 4.4127423822714675e-05,
+      "loss": 2.6089,
+      "step": 764
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.4085872576177284e-05,
+      "loss": 2.5592,
+      "step": 765
+    },
+    {
+      "epoch": 2.1,
+      "eval_loss": 2.678250551223755,
+      "eval_runtime": 143.9966,
+      "eval_samples_per_second": 33.633,
+      "eval_steps_per_second": 4.208,
+      "eval_wer": 1.1009934714731762,
+      "step": 765
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.404432132963989e-05,
+      "loss": 2.5787,
+      "step": 766
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.400277008310249e-05,
+      "loss": 2.6157,
+      "step": 767
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 4.3961218836565095e-05,
+      "loss": 2.6288,
+      "step": 768
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.3919667590027696e-05,
+      "loss": 2.7,
+      "step": 769
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.3878116343490305e-05,
+      "loss": 2.6444,
+      "step": 770
+    },
+    {
+      "epoch": 2.11,
+      "eval_loss": 2.6706833839416504,
+      "eval_runtime": 157.6333,
+      "eval_samples_per_second": 30.723,
+      "eval_steps_per_second": 3.844,
+      "eval_wer": 1.0946068691456146,
+      "step": 770
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.38365650969529e-05,
+      "loss": 2.6467,
+      "step": 771
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 4.379501385041551e-05,
+      "loss": 2.6146,
+      "step": 772
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.3753462603878116e-05,
+      "loss": 2.622,
+      "step": 773
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.371191135734072e-05,
+      "loss": 2.7354,
+      "step": 774
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 4.367036011080332e-05,
+      "loss": 2.6944,
+      "step": 775
+    },
+    {
+      "epoch": 2.12,
+      "eval_loss": 2.701211452484131,
+      "eval_runtime": 149.7771,
+      "eval_samples_per_second": 32.335,
+      "eval_steps_per_second": 4.046,
+      "eval_wer": 1.1315072381493045,
+      "step": 775
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.362880886426592e-05,
+      "loss": 2.5889,
+      "step": 776
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.358725761772853e-05,
+      "loss": 2.6307,
+      "step": 777
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.354570637119114e-05,
+      "loss": 2.6655,
+      "step": 778
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 4.350415512465373e-05,
+      "loss": 2.5951,
+      "step": 779
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.346260387811634e-05,
+      "loss": 2.6733,
+      "step": 780
+    },
+    {
+      "epoch": 2.14,
+      "eval_loss": 2.707202196121216,
+      "eval_runtime": 141.7678,
+      "eval_samples_per_second": 34.161,
+      "eval_steps_per_second": 4.275,
+      "eval_wer": 1.1144479137099064,
+      "step": 780
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.342105263157895e-05,
+      "loss": 2.7627,
+      "step": 781
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 4.337950138504155e-05,
+      "loss": 2.7326,
+      "step": 782
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.333795013850415e-05,
+      "loss": 2.71,
+      "step": 783
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.329639889196675e-05,
+      "loss": 2.7234,
+      "step": 784
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.325484764542936e-05,
+      "loss": 2.6998,
+      "step": 785
+    },
+    {
+      "epoch": 2.15,
+      "eval_loss": 2.7131922245025635,
+      "eval_runtime": 140.6616,
+      "eval_samples_per_second": 34.43,
+      "eval_steps_per_second": 4.308,
+      "eval_wer": 1.0205790519443656,
+      "step": 785
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 4.321329639889197e-05,
+      "loss": 2.7412,
+      "step": 786
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.3171745152354564e-05,
+      "loss": 2.7296,
+      "step": 787
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.313019390581717e-05,
+      "loss": 2.7296,
+      "step": 788
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.3088642659279774e-05,
+      "loss": 2.6516,
+      "step": 789
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 4.304709141274238e-05,
+      "loss": 2.796,
+      "step": 790
+    },
+    {
+      "epoch": 2.16,
+      "eval_loss": 2.707557439804077,
+      "eval_runtime": 139.1991,
+      "eval_samples_per_second": 34.792,
+      "eval_steps_per_second": 4.353,
+      "eval_wer": 1.1261708770933863,
+      "step": 790
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.300554016620498e-05,
+      "loss": 2.7491,
+      "step": 791
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.2963988919667586e-05,
+      "loss": 2.668,
+      "step": 792
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 4.2922437673130194e-05,
+      "loss": 2.6521,
+      "step": 793
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.2880886426592795e-05,
+      "loss": 2.7049,
+      "step": 794
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.28393351800554e-05,
+      "loss": 2.6881,
+      "step": 795
+    },
+    {
+      "epoch": 2.18,
+      "eval_loss": 2.69525408744812,
+      "eval_runtime": 141.7773,
+      "eval_samples_per_second": 34.159,
+      "eval_steps_per_second": 4.274,
+      "eval_wer": 1.0841044564291797,
+      "step": 795
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.2797783933518e-05,
+      "loss": 2.7087,
+      "step": 796
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 4.275623268698061e-05,
+      "loss": 2.6626,
+      "step": 797
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.2714681440443215e-05,
+      "loss": 2.7188,
+      "step": 798
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.267313019390581e-05,
+      "loss": 2.6864,
+      "step": 799
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.263157894736842e-05,
+      "loss": 2.7382,
+      "step": 800
+    },
+    {
+      "epoch": 2.19,
+      "eval_loss": 2.660480499267578,
+      "eval_runtime": 139.2218,
+      "eval_samples_per_second": 34.786,
+      "eval_steps_per_second": 4.353,
+      "eval_wer": 1.1234175418677264,
+      "step": 800
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 4.259002770083102e-05,
+      "loss": 2.6756,
+      "step": 801
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.254847645429363e-05,
+      "loss": 2.7063,
+      "step": 802
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.250692520775623e-05,
+      "loss": 2.6239,
+      "step": 803
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 4.246537396121883e-05,
+      "loss": 2.7096,
+      "step": 804
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.242382271468144e-05,
+      "loss": 2.5814,
+      "step": 805
+    },
+    {
+      "epoch": 2.21,
+      "eval_loss": 2.681447744369507,
+      "eval_runtime": 148.4391,
+      "eval_samples_per_second": 32.626,
+      "eval_steps_per_second": 4.082,
+      "eval_wer": 1.1864604030655692,
+      "step": 805
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.238227146814404e-05,
+      "loss": 2.6484,
+      "step": 806
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.234072022160664e-05,
+      "loss": 2.6799,
+      "step": 807
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 4.229916897506925e-05,
+      "loss": 2.632,
+      "step": 808
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.225761772853185e-05,
+      "loss": 2.6419,
+      "step": 809
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.221606648199446e-05,
+      "loss": 2.6695,
+      "step": 810
+    },
+    {
+      "epoch": 2.22,
+      "eval_loss": 2.6531379222869873,
+      "eval_runtime": 139.9128,
+      "eval_samples_per_second": 34.614,
+      "eval_steps_per_second": 4.331,
+      "eval_wer": 1.0984956003406188,
+      "step": 810
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.2174515235457055e-05,
+      "loss": 2.6403,
+      "step": 811
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 4.213296398891966e-05,
+      "loss": 2.6355,
+      "step": 812
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.209141274238227e-05,
+      "loss": 2.5871,
+      "step": 813
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.204986149584487e-05,
+      "loss": 2.7353,
+      "step": 814
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 4.2008310249307475e-05,
+      "loss": 2.6415,
+      "step": 815
+    },
+    {
+      "epoch": 2.23,
+      "eval_loss": 2.6589951515197754,
+      "eval_runtime": 140.0789,
+      "eval_samples_per_second": 34.573,
+      "eval_steps_per_second": 4.326,
+      "eval_wer": 1.0804144195288106,
+      "step": 815
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.1966759002770076e-05,
+      "loss": 2.6633,
+      "step": 816
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.1925207756232684e-05,
+      "loss": 2.6116,
+      "step": 817
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.188365650969529e-05,
+      "loss": 2.6993,
+      "step": 818
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 4.184210526315789e-05,
+      "loss": 2.568,
+      "step": 819
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.1800554016620496e-05,
+      "loss": 2.646,
+      "step": 820
+    },
+    {
+      "epoch": 2.25,
+      "eval_loss": 2.6513803005218506,
+      "eval_runtime": 139.1779,
+      "eval_samples_per_second": 34.797,
+      "eval_steps_per_second": 4.354,
+      "eval_wer": 1.0853250070962248,
+      "step": 820
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.17590027700831e-05,
+      "loss": 2.6249,
+      "step": 821
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.1717451523545706e-05,
+      "loss": 2.5463,
+      "step": 822
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.16759002770083e-05,
+      "loss": 2.6901,
+      "step": 823
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.163434903047091e-05,
+      "loss": 2.6175,
+      "step": 824
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.159279778393352e-05,
+      "loss": 2.6028,
+      "step": 825
+    },
+    {
+      "epoch": 2.26,
+      "eval_loss": 2.672335624694824,
+      "eval_runtime": 140.0857,
+      "eval_samples_per_second": 34.572,
+      "eval_steps_per_second": 4.326,
+      "eval_wer": 1.1410729491910303,
+      "step": 825
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.155124653739612e-05,
+      "loss": 2.6191,
+      "step": 826
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.150969529085872e-05,
+      "loss": 2.5994,
+      "step": 827
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.146814404432132e-05,
+      "loss": 2.7541,
+      "step": 828
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.142659279778393e-05,
+      "loss": 2.7348,
+      "step": 829
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.138504155124654e-05,
+      "loss": 2.6429,
+      "step": 830
+    },
+    {
+      "epoch": 2.27,
+      "eval_loss": 2.672929286956787,
+      "eval_runtime": 144.0755,
+      "eval_samples_per_second": 33.614,
+      "eval_steps_per_second": 4.206,
+      "eval_wer": 1.0394833948339484,
+      "step": 830
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.134349030470913e-05,
+      "loss": 2.787,
+      "step": 831
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.130193905817174e-05,
+      "loss": 2.7276,
+      "step": 832
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.126038781163435e-05,
+      "loss": 2.7426,
+      "step": 833
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.121883656509695e-05,
+      "loss": 2.7152,
+      "step": 834
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.117728531855955e-05,
+      "loss": 2.6736,
+      "step": 835
+    },
+    {
+      "epoch": 2.29,
+      "eval_loss": 2.7038564682006836,
+      "eval_runtime": 139.6762,
+      "eval_samples_per_second": 34.673,
+      "eval_steps_per_second": 4.339,
+      "eval_wer": 1.035537893840477,
+      "step": 835
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.1135734072022154e-05,
+      "loss": 2.6877,
+      "step": 836
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.109418282548476e-05,
+      "loss": 2.6905,
+      "step": 837
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.105263157894737e-05,
+      "loss": 2.7637,
+      "step": 838
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.1011080332409965e-05,
+      "loss": 2.7069,
+      "step": 839
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.0969529085872574e-05,
+      "loss": 2.6959,
+      "step": 840
+    },
+    {
+      "epoch": 2.3,
+      "eval_loss": 2.651028633117676,
+      "eval_runtime": 140.53,
+      "eval_samples_per_second": 34.462,
+      "eval_steps_per_second": 4.312,
+      "eval_wer": 1.0414135679818337,
+      "step": 840
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.0927977839335175e-05,
+      "loss": 2.6842,
+      "step": 841
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.088642659279778e-05,
+      "loss": 2.6516,
+      "step": 842
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.084487534626038e-05,
+      "loss": 2.6443,
+      "step": 843
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.0803324099722986e-05,
+      "loss": 2.6402,
+      "step": 844
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.0761772853185595e-05,
+      "loss": 2.6426,
+      "step": 845
+    },
+    {
+      "epoch": 2.31,
+      "eval_loss": 2.6659910678863525,
+      "eval_runtime": 139.5783,
+      "eval_samples_per_second": 34.697,
+      "eval_steps_per_second": 4.342,
+      "eval_wer": 1.1591257451036048,
+      "step": 845
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.0720221606648196e-05,
+      "loss": 2.6714,
+      "step": 846
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.06786703601108e-05,
+      "loss": 2.7276,
+      "step": 847
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.06371191135734e-05,
+      "loss": 2.6044,
+      "step": 848
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.059556786703601e-05,
+      "loss": 2.6401,
+      "step": 849
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.0554016620498616e-05,
+      "loss": 2.7152,
+      "step": 850
+    },
+    {
+      "epoch": 2.33,
+      "eval_loss": 2.636141538619995,
+      "eval_runtime": 140.7341,
+      "eval_samples_per_second": 34.412,
+      "eval_steps_per_second": 4.306,
+      "eval_wer": 1.027646891853534,
+      "step": 850
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.051246537396121e-05,
+      "loss": 2.6399,
+      "step": 851
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.047091412742382e-05,
+      "loss": 2.6995,
+      "step": 852
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.042936288088642e-05,
+      "loss": 2.6531,
+      "step": 853
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.038781163434903e-05,
+      "loss": 2.5376,
+      "step": 854
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.034626038781163e-05,
+      "loss": 2.7148,
+      "step": 855
+    },
+    {
+      "epoch": 2.34,
+      "eval_loss": 2.672334671020508,
+      "eval_runtime": 139.4164,
+      "eval_samples_per_second": 34.738,
+      "eval_steps_per_second": 4.347,
+      "eval_wer": 1.2461254612546127,
+      "step": 855
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.030470914127423e-05,
+      "loss": 2.6914,
+      "step": 856
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.026315789473684e-05,
+      "loss": 2.6138,
+      "step": 857
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.022160664819944e-05,
+      "loss": 2.6221,
+      "step": 858
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.018005540166204e-05,
+      "loss": 2.5736,
+      "step": 859
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.013850415512465e-05,
+      "loss": 2.6336,
+      "step": 860
+    },
+    {
+      "epoch": 2.36,
+      "eval_loss": 2.633176326751709,
+      "eval_runtime": 140.2482,
+      "eval_samples_per_second": 34.532,
+      "eval_steps_per_second": 4.321,
+      "eval_wer": 1.0309963099630997,
+      "step": 860
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.009695290858725e-05,
+      "loss": 2.5826,
+      "step": 861
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.005540166204986e-05,
+      "loss": 2.606,
+      "step": 862
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.0013850415512456e-05,
+      "loss": 2.5936,
+      "step": 863
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.9972299168975064e-05,
+      "loss": 2.6391,
+      "step": 864
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.993074792243767e-05,
+      "loss": 2.665,
+      "step": 865
+    },
+    {
+      "epoch": 2.37,
+      "eval_loss": 2.6365127563476562,
+      "eval_runtime": 140.65,
+      "eval_samples_per_second": 34.433,
+      "eval_steps_per_second": 4.309,
+      "eval_wer": 1.1311950042577348,
+      "step": 865
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.9889196675900274e-05,
+      "loss": 2.5865,
+      "step": 866
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 3.9847645429362875e-05,
+      "loss": 2.633,
+      "step": 867
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.980609418282548e-05,
+      "loss": 2.5912,
+      "step": 868
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.9764542936288085e-05,
+      "loss": 2.5546,
+      "step": 869
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 3.9722991689750694e-05,
+      "loss": 2.5607,
+      "step": 870
+    },
+    {
+      "epoch": 2.38,
+      "eval_loss": 2.6344053745269775,
+      "eval_runtime": 140.5167,
+      "eval_samples_per_second": 34.466,
+      "eval_steps_per_second": 4.313,
+      "eval_wer": 1.1301447629860915,
+      "step": 870
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.9681440443213295e-05,
+      "loss": 2.582,
+      "step": 871
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.96398891966759e-05,
+      "loss": 2.59,
+      "step": 872
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.95983379501385e-05,
+      "loss": 2.6797,
+      "step": 873
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 3.9556786703601106e-05,
+      "loss": 2.6385,
+      "step": 874
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9515235457063715e-05,
+      "loss": 2.5614,
+      "step": 875
+    },
+    {
+      "epoch": 2.4,
+      "eval_loss": 2.643657922744751,
+      "eval_runtime": 140.4419,
+      "eval_samples_per_second": 34.484,
+      "eval_steps_per_second": 4.315,
+      "eval_wer": 1.1513198978143628,
+      "step": 875
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.947368421052631e-05,
+      "loss": 2.6793,
+      "step": 876
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.943213296398892e-05,
+      "loss": 2.6164,
+      "step": 877
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.939058171745152e-05,
+      "loss": 2.5675,
+      "step": 878
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.934903047091413e-05,
+      "loss": 2.6355,
+      "step": 879
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.930747922437672e-05,
+      "loss": 2.4899,
+      "step": 880
+    },
+    {
+      "epoch": 2.41,
+      "eval_loss": 2.641817092895508,
+      "eval_runtime": 140.5523,
+      "eval_samples_per_second": 34.457,
+      "eval_steps_per_second": 4.312,
+      "eval_wer": 1.1532216860630145,
+      "step": 880
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.926592797783933e-05,
+      "loss": 2.7238,
+      "step": 881
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.922437673130194e-05,
+      "loss": 2.7609,
+      "step": 882
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.918282548476454e-05,
+      "loss": 2.6853,
+      "step": 883
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.914127423822714e-05,
+      "loss": 2.7105,
+      "step": 884
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.909972299168975e-05,
+      "loss": 2.6794,
+      "step": 885
+    },
+    {
+      "epoch": 2.42,
+      "eval_loss": 2.6402528285980225,
+      "eval_runtime": 140.4059,
+      "eval_samples_per_second": 34.493,
+      "eval_steps_per_second": 4.316,
+      "eval_wer": 1.0271643485665627,
+      "step": 885
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.905817174515235e-05,
+      "loss": 2.6737,
+      "step": 886
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.901662049861496e-05,
+      "loss": 2.6492,
+      "step": 887
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8975069252077555e-05,
+      "loss": 2.6932,
+      "step": 888
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.893351800554016e-05,
+      "loss": 2.6776,
+      "step": 889
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.889196675900277e-05,
+      "loss": 2.6814,
+      "step": 890
+    },
+    {
+      "epoch": 2.44,
+      "eval_loss": 2.6419880390167236,
+      "eval_runtime": 140.6412,
+      "eval_samples_per_second": 34.435,
+      "eval_steps_per_second": 4.309,
+      "eval_wer": 1.132330400227079,
+      "step": 890
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.885041551246537e-05,
+      "loss": 2.6805,
+      "step": 891
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.8808864265927974e-05,
+      "loss": 2.6938,
+      "step": 892
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.8767313019390576e-05,
+      "loss": 2.7019,
+      "step": 893
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.8725761772853184e-05,
+      "loss": 2.6607,
+      "step": 894
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.868421052631579e-05,
+      "loss": 2.6614,
+      "step": 895
+    },
+    {
+      "epoch": 2.45,
+      "eval_loss": 2.618271827697754,
+      "eval_runtime": 141.942,
+      "eval_samples_per_second": 34.12,
+      "eval_steps_per_second": 4.269,
+      "eval_wer": 1.0524836786829406,
+      "step": 895
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.864265927977839e-05,
+      "loss": 2.6357,
+      "step": 896
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.8601108033240996e-05,
+      "loss": 2.6423,
+      "step": 897
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.85595567867036e-05,
+      "loss": 2.663,
+      "step": 898
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.8518005540166205e-05,
+      "loss": 2.6235,
+      "step": 899
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.84764542936288e-05,
+      "loss": 2.6629,
+      "step": 900
+    },
+    {
+      "epoch": 2.47,
+      "eval_loss": 2.6413512229919434,
+      "eval_runtime": 139.4501,
+      "eval_samples_per_second": 34.729,
+      "eval_steps_per_second": 4.346,
+      "eval_wer": 1.1568549531649162,
+      "step": 900
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.843490304709141e-05,
+      "loss": 2.6615,
+      "step": 901
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.839335180055402e-05,
+      "loss": 2.5994,
+      "step": 902
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.835180055401662e-05,
+      "loss": 2.6987,
+      "step": 903
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.831024930747922e-05,
+      "loss": 2.6827,
+      "step": 904
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.826869806094182e-05,
+      "loss": 2.6166,
+      "step": 905
+    },
+    {
+      "epoch": 2.48,
+      "eval_loss": 2.61674427986145,
+      "eval_runtime": 148.9037,
+      "eval_samples_per_second": 32.524,
+      "eval_steps_per_second": 4.07,
+      "eval_wer": 1.0265398807834232,
+      "step": 905
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.822714681440443e-05,
+      "loss": 2.6852,
+      "step": 906
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.818559556786704e-05,
+      "loss": 2.6355,
+      "step": 907
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.814404432132963e-05,
+      "loss": 2.608,
+      "step": 908
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.810249307479224e-05,
+      "loss": 2.6037,
+      "step": 909
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.806094182825484e-05,
+      "loss": 2.6374,
+      "step": 910
+    },
+    {
+      "epoch": 2.49,
+      "eval_loss": 2.629850387573242,
+      "eval_runtime": 140.4885,
+      "eval_samples_per_second": 34.473,
+      "eval_steps_per_second": 4.314,
+      "eval_wer": 1.1720408742548964,
+      "step": 910
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.801939058171745e-05,
+      "loss": 2.5998,
+      "step": 911
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.797783933518005e-05,
+      "loss": 2.6625,
+      "step": 912
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.7936288088642654e-05,
+      "loss": 2.6031,
+      "step": 913
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.789473684210526e-05,
+      "loss": 2.5866,
+      "step": 914
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.785318559556787e-05,
+      "loss": 2.6035,
+      "step": 915
+    },
+    {
+      "epoch": 2.51,
+      "eval_loss": 2.6139261722564697,
+      "eval_runtime": 140.2461,
+      "eval_samples_per_second": 34.532,
+      "eval_steps_per_second": 4.321,
+      "eval_wer": 1.1565427192733466,
+      "step": 915
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.7811634349030465e-05,
+      "loss": 2.6247,
+      "step": 916
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.777008310249307e-05,
+      "loss": 2.6632,
+      "step": 917
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.7728531855955675e-05,
+      "loss": 2.6268,
+      "step": 918
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.768698060941828e-05,
+      "loss": 2.5891,
+      "step": 919
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.764542936288088e-05,
+      "loss": 2.595,
+      "step": 920
+    },
+    {
+      "epoch": 2.52,
+      "eval_loss": 2.6126370429992676,
+      "eval_runtime": 140.6689,
+      "eval_samples_per_second": 34.428,
+      "eval_steps_per_second": 4.308,
+      "eval_wer": 1.0556627873971047,
+      "step": 920
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.7603878116343486e-05,
+      "loss": 2.5504,
+      "step": 921
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.7562326869806094e-05,
+      "loss": 2.5722,
+      "step": 922
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.7520775623268696e-05,
+      "loss": 2.6546,
+      "step": 923
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.74792243767313e-05,
+      "loss": 2.5729,
+      "step": 924
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.74376731301939e-05,
+      "loss": 2.6416,
+      "step": 925
+    },
+    {
+      "epoch": 2.53,
+      "eval_loss": 2.619037389755249,
+      "eval_runtime": 141.7338,
+      "eval_samples_per_second": 34.17,
+      "eval_steps_per_second": 4.276,
+      "eval_wer": 1.0414419528810672,
+      "step": 925
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.739612188365651e-05,
+      "loss": 2.5694,
+      "step": 926
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.735457063711911e-05,
+      "loss": 2.7109,
+      "step": 927
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.731301939058172e-05,
+      "loss": 2.54,
+      "step": 928
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.727146814404432e-05,
+      "loss": 2.6116,
+      "step": 929
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.722991689750692e-05,
+      "loss": 2.6785,
+      "step": 930
+    },
+    {
+      "epoch": 2.55,
+      "eval_loss": 2.635225772857666,
+      "eval_runtime": 140.3161,
+      "eval_samples_per_second": 34.515,
+      "eval_steps_per_second": 4.319,
+      "eval_wer": 1.0288958274198126,
+      "step": 930
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.718836565096952e-05,
+      "loss": 2.695,
+      "step": 931
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.714681440443213e-05,
+      "loss": 2.6851,
+      "step": 932
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.710526315789473e-05,
+      "loss": 2.6712,
+      "step": 933
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.706371191135734e-05,
+      "loss": 2.7343,
+      "step": 934
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.702216066481994e-05,
+      "loss": 2.6986,
+      "step": 935
+    },
+    {
+      "epoch": 2.56,
+      "eval_loss": 2.6267940998077393,
+      "eval_runtime": 148.1702,
+      "eval_samples_per_second": 32.685,
+      "eval_steps_per_second": 4.09,
+      "eval_wer": 1.0076923076923077,
+      "step": 935
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 3.698060941828254e-05,
+      "loss": 2.6477,
+      "step": 936
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.693905817174515e-05,
+      "loss": 2.6396,
+      "step": 937
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.689750692520775e-05,
+      "loss": 2.6473,
+      "step": 938
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.6855955678670354e-05,
+      "loss": 2.6596,
+      "step": 939
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 3.681440443213296e-05,
+      "loss": 2.6145,
+      "step": 940
+    },
+    {
+      "epoch": 2.57,
+      "eval_loss": 2.6166110038757324,
+      "eval_runtime": 139.5921,
+      "eval_samples_per_second": 34.694,
+      "eval_steps_per_second": 4.341,
+      "eval_wer": 1.0444507521998296,
+      "step": 940
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.6772853185595564e-05,
+      "loss": 2.6345,
+      "step": 941
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.673130193905817e-05,
+      "loss": 2.6359,
+      "step": 942
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 3.6689750692520774e-05,
+      "loss": 2.6898,
+      "step": 943
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.6648199445983375e-05,
+      "loss": 2.6554,
+      "step": 944
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.660664819944598e-05,
+      "loss": 2.6961,
+      "step": 945
+    },
+    {
+      "epoch": 2.59,
+      "eval_loss": 2.614225387573242,
+      "eval_runtime": 141.1144,
+      "eval_samples_per_second": 34.32,
+      "eval_steps_per_second": 4.294,
+      "eval_wer": 1.0185353391995458,
+      "step": 945
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.6565096952908585e-05,
+      "loss": 2.622,
+      "step": 946
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 3.6523545706371187e-05,
+      "loss": 2.6716,
+      "step": 947
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.6481994459833795e-05,
+      "loss": 2.7486,
+      "step": 948
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.6440443213296396e-05,
+      "loss": 2.6455,
+      "step": 949
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.6398891966759e-05,
+      "loss": 2.6852,
+      "step": 950
+    },
+    {
+      "epoch": 2.6,
+      "eval_loss": 2.607232093811035,
+      "eval_runtime": 149.4774,
+      "eval_samples_per_second": 32.4,
+      "eval_steps_per_second": 4.054,
+      "eval_wer": 1.012233891569685,
+      "step": 950
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 3.63573407202216e-05,
+      "loss": 2.68,
+      "step": 951
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.631578947368421e-05,
+      "loss": 2.6479,
+      "step": 952
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.627423822714681e-05,
+      "loss": 2.6228,
+      "step": 953
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 3.623268698060942e-05,
+      "loss": 2.6673,
+      "step": 954
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.619113573407202e-05,
+      "loss": 2.5792,
+      "step": 955
+    },
+    {
+      "epoch": 2.62,
+      "eval_loss": 2.6078407764434814,
+      "eval_runtime": 143.081,
+      "eval_samples_per_second": 33.848,
+      "eval_steps_per_second": 4.235,
+      "eval_wer": 1.116491626454726,
+      "step": 955
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.614958448753462e-05,
+      "loss": 2.6426,
+      "step": 956
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.610803324099722e-05,
+      "loss": 2.6061,
+      "step": 957
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 3.606648199445983e-05,
+      "loss": 2.6111,
+      "step": 958
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.602493074792243e-05,
+      "loss": 2.6043,
+      "step": 959
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.598337950138504e-05,
+      "loss": 2.6118,
+      "step": 960
+    },
+    {
+      "epoch": 2.63,
+      "eval_loss": 2.6176657676696777,
+      "eval_runtime": 150.2175,
+      "eval_samples_per_second": 32.24,
+      "eval_steps_per_second": 4.034,
+      "eval_wer": 1.1209764405336362,
+      "step": 960
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.594182825484764e-05,
+      "loss": 2.6354,
+      "step": 961
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 3.590027700831024e-05,
+      "loss": 2.6413,
+      "step": 962
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.585872576177285e-05,
+      "loss": 2.5868,
+      "step": 963
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.581717451523545e-05,
+      "loss": 2.7232,
+      "step": 964
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 3.5775623268698055e-05,
+      "loss": 2.5472,
+      "step": 965
+    },
+    {
+      "epoch": 2.64,
+      "eval_loss": 2.6125783920288086,
+      "eval_runtime": 148.6283,
+      "eval_samples_per_second": 32.585,
+      "eval_steps_per_second": 4.077,
+      "eval_wer": 1.0044280442804427,
+      "step": 965
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.573407202216066e-05,
+      "loss": 2.6403,
+      "step": 966
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.5692520775623264e-05,
+      "loss": 2.635,
+      "step": 967
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.565096952908587e-05,
+      "loss": 2.5512,
+      "step": 968
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 3.5609418282548474e-05,
+      "loss": 2.5858,
+      "step": 969
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.5567867036011076e-05,
+      "loss": 2.577,
+      "step": 970
+    },
+    {
+      "epoch": 2.66,
+      "eval_loss": 2.6051268577575684,
+      "eval_runtime": 152.7302,
+      "eval_samples_per_second": 31.71,
+      "eval_steps_per_second": 3.968,
+      "eval_wer": 1.088135112120352,
+      "step": 970
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.552631578947368e-05,
+      "loss": 2.54,
+      "step": 971
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.5484764542936285e-05,
+      "loss": 2.542,
+      "step": 972
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 3.544321329639889e-05,
+      "loss": 2.5901,
+      "step": 973
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.5401662049861495e-05,
+      "loss": 2.6001,
+      "step": 974
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.53601108033241e-05,
+      "loss": 2.5602,
+      "step": 975
+    },
+    {
+      "epoch": 2.67,
+      "eval_loss": 2.5991711616516113,
+      "eval_runtime": 139.8925,
+      "eval_samples_per_second": 34.619,
+      "eval_steps_per_second": 4.332,
+      "eval_wer": 1.017797331819472,
+      "step": 975
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 3.53185595567867e-05,
+      "loss": 2.5948,
+      "step": 976
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.5277008310249307e-05,
+      "loss": 2.5473,
+      "step": 977
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.523545706371191e-05,
+      "loss": 2.5476,
+      "step": 978
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.5193905817174516e-05,
+      "loss": 2.5644,
+      "step": 979
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 3.515235457063712e-05,
+      "loss": 2.695,
+      "step": 980
+    },
+    {
+      "epoch": 2.68,
+      "eval_loss": 2.602264404296875,
+      "eval_runtime": 151.0197,
+      "eval_samples_per_second": 32.069,
+      "eval_steps_per_second": 4.013,
+      "eval_wer": 1.0247516321317058,
+      "step": 980
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.511080332409972e-05,
+      "loss": 2.7191,
+      "step": 981
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.506925207756232e-05,
+      "loss": 2.639,
+      "step": 982
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.502770083102493e-05,
+      "loss": 2.7155,
+      "step": 983
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 3.498614958448753e-05,
+      "loss": 2.6456,
+      "step": 984
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.494459833795014e-05,
+      "loss": 2.7017,
+      "step": 985
+    },
+    {
+      "epoch": 2.7,
+      "eval_loss": 2.6189544200897217,
+      "eval_runtime": 139.912,
+      "eval_samples_per_second": 34.615,
+      "eval_steps_per_second": 4.331,
+      "eval_wer": 1.004059040590406,
+      "step": 985
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.490304709141274e-05,
+      "loss": 2.6867,
+      "step": 986
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 3.486149584487534e-05,
+      "loss": 2.7123,
+      "step": 987
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.4819944598337944e-05,
+      "loss": 2.7224,
+      "step": 988
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.477839335180055e-05,
+      "loss": 2.6807,
+      "step": 989
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.4736842105263153e-05,
+      "loss": 2.6327,
+      "step": 990
+    },
+    {
+      "epoch": 2.71,
+      "eval_loss": 2.60239839553833,
+      "eval_runtime": 154.9675,
+      "eval_samples_per_second": 31.252,
+      "eval_steps_per_second": 3.91,
+      "eval_wer": 1.014249219415271,
+      "step": 990
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 3.469529085872576e-05,
+      "loss": 2.6733,
+      "step": 991
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.465373961218836e-05,
+      "loss": 2.6564,
+      "step": 992
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.461218836565097e-05,
+      "loss": 2.6688,
+      "step": 993
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 3.457063711911357e-05,
+      "loss": 2.6269,
+      "step": 994
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.4529085872576175e-05,
+      "loss": 2.6193,
+      "step": 995
+    },
+    {
+      "epoch": 2.73,
+      "eval_loss": 2.5897159576416016,
+      "eval_runtime": 141.0123,
+      "eval_samples_per_second": 34.345,
+      "eval_steps_per_second": 4.297,
+      "eval_wer": 1.0147885325007096,
+      "step": 995
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.4487534626038776e-05,
+      "loss": 2.585,
+      "step": 996
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.4445983379501384e-05,
+      "loss": 2.6455,
+      "step": 997
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 3.4404432132963986e-05,
+      "loss": 2.6029,
+      "step": 998
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.4362880886426594e-05,
+      "loss": 2.6363,
+      "step": 999
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.4321329639889196e-05,
+      "loss": 2.5939,
+      "step": 1000
+    },
+    {
+      "epoch": 2.74,
+      "eval_loss": 2.5900378227233887,
+      "eval_runtime": 142.8356,
+      "eval_samples_per_second": 33.906,
+      "eval_steps_per_second": 4.243,
+      "eval_wer": 1.0328980982117513,
+      "step": 1000
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.42797783933518e-05,
+      "loss": 2.617,
+      "step": 1001
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 3.42382271468144e-05,
+      "loss": 2.607,
+      "step": 1002
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.419667590027701e-05,
+      "loss": 2.6708,
+      "step": 1003
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.415512465373961e-05,
+      "loss": 2.5678,
+      "step": 1004
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 3.411357340720222e-05,
+      "loss": 2.5477,
+      "step": 1005
+    },
+    {
+      "epoch": 2.75,
+      "eval_loss": 2.597130060195923,
+      "eval_runtime": 140.8229,
+      "eval_samples_per_second": 34.391,
+      "eval_steps_per_second": 4.303,
+      "eval_wer": 1.0338347998864603,
+      "step": 1005
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.407202216066482e-05,
+      "loss": 2.6318,
+      "step": 1006
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.403047091412742e-05,
+      "loss": 2.6538,
+      "step": 1007
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.398891966759002e-05,
+      "loss": 2.5786,
+      "step": 1008
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 3.394736842105263e-05,
+      "loss": 2.5831,
+      "step": 1009
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.390581717451523e-05,
+      "loss": 2.6089,
+      "step": 1010
+    },
+    {
+      "epoch": 2.77,
+      "eval_loss": 2.596896171569824,
+      "eval_runtime": 141.6502,
+      "eval_samples_per_second": 34.19,
+      "eval_steps_per_second": 4.278,
+      "eval_wer": 1.006358217428328,
+      "step": 1010
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.386426592797784e-05,
+      "loss": 2.6291,
+      "step": 1011
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.382271468144044e-05,
+      "loss": 2.6366,
+      "step": 1012
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 3.378116343490304e-05,
+      "loss": 2.5969,
+      "step": 1013
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.3739612188365644e-05,
+      "loss": 2.6686,
+      "step": 1014
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.369806094182825e-05,
+      "loss": 2.5625,
+      "step": 1015
+    },
+    {
+      "epoch": 2.78,
+      "eval_loss": 2.589853048324585,
+      "eval_runtime": 142.4939,
+      "eval_samples_per_second": 33.987,
+      "eval_steps_per_second": 4.253,
+      "eval_wer": 1.0648027249503265,
+      "step": 1015
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 3.3656509695290854e-05,
+      "loss": 2.6448,
+      "step": 1016
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.361495844875346e-05,
+      "loss": 2.5491,
+      "step": 1017
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.3573407202216064e-05,
+      "loss": 2.6089,
+      "step": 1018
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.353185595567867e-05,
+      "loss": 2.5364,
+      "step": 1019
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 3.3490304709141273e-05,
+      "loss": 2.5745,
+      "step": 1020
+    },
+    {
+      "epoch": 2.79,
+      "eval_loss": 2.586090326309204,
+      "eval_runtime": 140.1471,
+      "eval_samples_per_second": 34.557,
+      "eval_steps_per_second": 4.324,
+      "eval_wer": 1.0627022424070394,
+      "step": 1020
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.3448753462603875e-05,
+      "loss": 2.5555,
+      "step": 1021
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.3407202216066476e-05,
+      "loss": 2.6456,
+      "step": 1022
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.3365650969529085e-05,
+      "loss": 2.6165,
+      "step": 1023
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 3.3324099722991686e-05,
+      "loss": 2.6059,
+      "step": 1024
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.3282548476454295e-05,
+      "loss": 2.5702,
+      "step": 1025
+    },
+    {
+      "epoch": 2.81,
+      "eval_loss": 2.5922608375549316,
+      "eval_runtime": 142.7304,
+      "eval_samples_per_second": 33.931,
+      "eval_steps_per_second": 4.246,
+      "eval_wer": 1.052597218279875,
+      "step": 1025
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.3240997229916896e-05,
+      "loss": 2.7317,
+      "step": 1026
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 3.31994459833795e-05,
+      "loss": 2.6421,
+      "step": 1027
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.31578947368421e-05,
+      "loss": 2.6052,
+      "step": 1028
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.311634349030471e-05,
+      "loss": 2.6021,
+      "step": 1029
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.307479224376731e-05,
+      "loss": 2.645,
+      "step": 1030
+    },
+    {
+      "epoch": 2.82,
+      "eval_loss": 2.6053032875061035,
+      "eval_runtime": 146.1281,
+      "eval_samples_per_second": 33.142,
+      "eval_steps_per_second": 4.147,
+      "eval_wer": 1.0198694294635253,
+      "step": 1030
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 3.303324099722992e-05,
+      "loss": 2.7207,
+      "step": 1031
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.299168975069252e-05,
+      "loss": 2.7174,
+      "step": 1032
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.295013850415512e-05,
+      "loss": 2.7229,
+      "step": 1033
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.290858725761772e-05,
+      "loss": 2.7535,
+      "step": 1034
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 3.286703601108033e-05,
+      "loss": 2.6869,
+      "step": 1035
+    },
+    {
+      "epoch": 2.83,
+      "eval_loss": 2.6226749420166016,
+      "eval_runtime": 139.922,
+      "eval_samples_per_second": 34.612,
+      "eval_steps_per_second": 4.331,
+      "eval_wer": 1.001078626170877,
+      "step": 1035
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.282548476454293e-05,
+      "loss": 2.7362,
+      "step": 1036
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.278393351800554e-05,
+      "loss": 2.6669,
+      "step": 1037
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 3.274238227146814e-05,
+      "loss": 2.7309,
+      "step": 1038
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.270083102493074e-05,
+      "loss": 2.6516,
+      "step": 1039
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.2659279778393344e-05,
+      "loss": 2.6678,
+      "step": 1040
+    },
+    {
+      "epoch": 2.85,
+      "eval_loss": 2.6094069480895996,
+      "eval_runtime": 140.0132,
+      "eval_samples_per_second": 34.59,
+      "eval_steps_per_second": 4.328,
+      "eval_wer": 1.017882486517173,
+      "step": 1040
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.261772853185595e-05,
+      "loss": 2.6565,
+      "step": 1041
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 3.2576177285318554e-05,
+      "loss": 2.7457,
+      "step": 1042
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.253462603878116e-05,
+      "loss": 2.6513,
+      "step": 1043
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.2493074792243764e-05,
+      "loss": 2.6587,
+      "step": 1044
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.245152354570637e-05,
+      "loss": 2.6787,
+      "step": 1045
+    },
+    {
+      "epoch": 2.86,
+      "eval_loss": 2.5977933406829834,
+      "eval_runtime": 141.7965,
+      "eval_samples_per_second": 34.155,
+      "eval_steps_per_second": 4.274,
+      "eval_wer": 1.0028101050241272,
+      "step": 1045
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 3.2409972299168974e-05,
+      "loss": 2.6396,
+      "step": 1046
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.2368421052631575e-05,
+      "loss": 2.6629,
+      "step": 1047
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.232686980609418e-05,
+      "loss": 2.6321,
+      "step": 1048
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 3.2285318559556785e-05,
+      "loss": 2.6517,
+      "step": 1049
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.224376731301939e-05,
+      "loss": 2.6246,
+      "step": 1050
+    },
+    {
+      "epoch": 2.88,
+      "eval_loss": 2.596486806869507,
+      "eval_runtime": 141.192,
+      "eval_samples_per_second": 34.301,
+      "eval_steps_per_second": 4.292,
+      "eval_wer": 1.0092818620493897,
+      "step": 1050
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.2202216066481995e-05,
+      "loss": 2.6658,
+      "step": 1051
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.2160664819944597e-05,
+      "loss": 2.603,
+      "step": 1052
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 3.21191135734072e-05,
+      "loss": 2.6856,
+      "step": 1053
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.20775623268698e-05,
+      "loss": 2.6517,
+      "step": 1054
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.203601108033241e-05,
+      "loss": 2.5676,
+      "step": 1055
+    },
+    {
+      "epoch": 2.89,
+      "eval_loss": 2.592679500579834,
+      "eval_runtime": 145.0407,
+      "eval_samples_per_second": 33.391,
+      "eval_steps_per_second": 4.178,
+      "eval_wer": 1.062730627306273,
+      "step": 1055
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.199445983379501e-05,
+      "loss": 2.5968,
+      "step": 1056
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 3.195290858725762e-05,
+      "loss": 2.6432,
+      "step": 1057
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.191135734072022e-05,
+      "loss": 2.6205,
+      "step": 1058
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.186980609418282e-05,
+      "loss": 2.5305,
+      "step": 1059
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 3.182825484764542e-05,
+      "loss": 2.6773,
+      "step": 1060
+    },
+    {
+      "epoch": 2.9,
+      "eval_loss": 2.5907163619995117,
+      "eval_runtime": 156.3634,
+      "eval_samples_per_second": 30.973,
+      "eval_steps_per_second": 3.876,
+      "eval_wer": 1.0817485097927901,
+      "step": 1060
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.178670360110803e-05,
+      "loss": 2.648,
+      "step": 1061
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.174515235457063e-05,
+      "loss": 2.5314,
+      "step": 1062
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.170360110803324e-05,
+      "loss": 2.6579,
+      "step": 1063
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 3.166204986149584e-05,
+      "loss": 2.5713,
+      "step": 1064
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.162049861495844e-05,
+      "loss": 2.6114,
+      "step": 1065
+    },
+    {
+      "epoch": 2.92,
+      "eval_loss": 2.593247413635254,
+      "eval_runtime": 153.9779,
+      "eval_samples_per_second": 31.453,
+      "eval_steps_per_second": 3.936,
+      "eval_wer": 1.101305705364746,
+      "step": 1065
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1578947368421045e-05,
+      "loss": 2.6003,
+      "step": 1066
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.153739612188365e-05,
+      "loss": 2.5824,
+      "step": 1067
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 3.1495844875346255e-05,
+      "loss": 2.6219,
+      "step": 1068
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.145429362880886e-05,
+      "loss": 2.6234,
+      "step": 1069
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.1412742382271464e-05,
+      "loss": 2.6227,
+      "step": 1070
+    },
+    {
+      "epoch": 2.93,
+      "eval_loss": 2.584027051925659,
+      "eval_runtime": 139.6475,
+      "eval_samples_per_second": 34.68,
+      "eval_steps_per_second": 4.339,
+      "eval_wer": 1.0401930173147886,
+      "step": 1070
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 3.137119113573407e-05,
+      "loss": 2.6051,
+      "step": 1071
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.1329639889196674e-05,
+      "loss": 2.5372,
+      "step": 1072
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.1288088642659276e-05,
+      "loss": 2.5539,
+      "step": 1073
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.124653739612188e-05,
+      "loss": 2.5365,
+      "step": 1074
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 3.1204986149584486e-05,
+      "loss": 2.594,
+      "step": 1075
+    },
+    {
+      "epoch": 2.94,
+      "eval_loss": 2.59973406791687,
+      "eval_runtime": 144.6623,
+      "eval_samples_per_second": 33.478,
+      "eval_steps_per_second": 4.189,
+      "eval_wer": 1.1371274481975588,
+      "step": 1075
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.116343490304709e-05,
+      "loss": 2.5861,
+      "step": 1076
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.1121883656509695e-05,
+      "loss": 2.6044,
+      "step": 1077
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.10803324099723e-05,
+      "loss": 2.5448,
+      "step": 1078
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 3.10387811634349e-05,
+      "loss": 2.6662,
+      "step": 1079
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.09972299168975e-05,
+      "loss": 2.751,
+      "step": 1080
+    },
+    {
+      "epoch": 2.96,
+      "eval_loss": 2.590857982635498,
+      "eval_runtime": 144.6161,
+      "eval_samples_per_second": 33.489,
+      "eval_steps_per_second": 4.19,
+      "eval_wer": 1.0972182798751065,
+      "step": 1080
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.095567867036011e-05,
+      "loss": 2.7415,
+      "step": 1081
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 3.091412742382271e-05,
+      "loss": 2.6923,
+      "step": 1082
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.087257617728532e-05,
+      "loss": 2.6413,
+      "step": 1083
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.083102493074792e-05,
+      "loss": 2.6784,
+      "step": 1084
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.078947368421052e-05,
+      "loss": 2.6366,
+      "step": 1085
+    },
+    {
+      "epoch": 2.97,
+      "eval_loss": 2.6080873012542725,
+      "eval_runtime": 143.8483,
+      "eval_samples_per_second": 33.667,
+      "eval_steps_per_second": 4.213,
+      "eval_wer": 1.0597502128867442,
+      "step": 1085
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 3.074792243767313e-05,
+      "loss": 2.6346,
+      "step": 1086
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.070637119113573e-05,
+      "loss": 2.5989,
+      "step": 1087
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.066481994459834e-05,
+      "loss": 2.6227,
+      "step": 1088
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.062326869806094e-05,
+      "loss": 2.5712,
+      "step": 1089
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 3.058171745152354e-05,
+      "loss": 2.577,
+      "step": 1090
+    },
+    {
+      "epoch": 2.98,
+      "eval_loss": 2.591460943222046,
+      "eval_runtime": 146.3232,
+      "eval_samples_per_second": 33.098,
+      "eval_steps_per_second": 4.142,
+      "eval_wer": 1.0410161793925632,
+      "step": 1090
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.0540166204986144e-05,
+      "loss": 2.6296,
+      "step": 1091
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.0498614958448752e-05,
+      "loss": 2.6557,
+      "step": 1092
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 3.0457063711911354e-05,
+      "loss": 2.5853,
+      "step": 1093
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.0415512465373962e-05,
+      "loss": 2.5998,
+      "step": 1094
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.0373961218836563e-05,
+      "loss": 2.579,
+      "step": 1095
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 2.5952680110931396,
+      "eval_runtime": 141.2401,
+      "eval_samples_per_second": 34.289,
+      "eval_steps_per_second": 4.291,
+      "eval_wer": 1.1432585864320182,
+      "step": 1095
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 3.0332409972299168e-05,
+      "loss": 4.0855,
+      "step": 1096
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.029085872576177e-05,
+      "loss": 2.656,
+      "step": 1097
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.0249307479224375e-05,
+      "loss": 2.6641,
+      "step": 1098
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.0207756232686976e-05,
+      "loss": 2.7159,
+      "step": 1099
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 3.0166204986149585e-05,
+      "loss": 2.6706,
+      "step": 1100
+    },
+    {
+      "epoch": 3.01,
+      "eval_loss": 2.591264486312866,
+      "eval_runtime": 140.5773,
+      "eval_samples_per_second": 34.451,
+      "eval_steps_per_second": 4.311,
+      "eval_wer": 1.0456429179676412,
+      "step": 1100
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.0124653739612186e-05,
+      "loss": 2.6311,
+      "step": 1101
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.008310249307479e-05,
+      "loss": 2.7027,
+      "step": 1102
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 3.0041551246537392e-05,
+      "loss": 2.6313,
+      "step": 1103
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 2.5907,
+      "step": 1104
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.9958448753462602e-05,
+      "loss": 2.6161,
+      "step": 1105
+    },
+    {
+      "epoch": 3.03,
+      "eval_loss": 2.60790753364563,
+      "eval_runtime": 142.8987,
+      "eval_samples_per_second": 33.891,
+      "eval_steps_per_second": 4.241,
+      "eval_wer": 1.100879931876242,
+      "step": 1105
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.9916897506925207e-05,
+      "loss": 2.6878,
+      "step": 1106
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 2.987534626038781e-05,
+      "loss": 2.6162,
+      "step": 1107
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.9833795013850414e-05,
+      "loss": 2.6136,
+      "step": 1108
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.9792243767313015e-05,
+      "loss": 2.613,
+      "step": 1109
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.9750692520775623e-05,
+      "loss": 2.6397,
+      "step": 1110
+    },
+    {
+      "epoch": 3.04,
+      "eval_loss": 2.5951175689697266,
+      "eval_runtime": 140.1696,
+      "eval_samples_per_second": 34.551,
+      "eval_steps_per_second": 4.323,
+      "eval_wer": 1.1771217712177122,
+      "step": 1110
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 2.9709141274238225e-05,
+      "loss": 2.6065,
+      "step": 1111
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.966759002770083e-05,
+      "loss": 2.6346,
+      "step": 1112
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.962603878116343e-05,
+      "loss": 2.6567,
+      "step": 1113
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.9584487534626036e-05,
+      "loss": 2.6403,
+      "step": 1114
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 2.9542936288088638e-05,
+      "loss": 2.6246,
+      "step": 1115
+    },
+    {
+      "epoch": 3.05,
+      "eval_loss": 2.5729894638061523,
+      "eval_runtime": 142.0088,
+      "eval_samples_per_second": 34.104,
+      "eval_steps_per_second": 4.267,
+      "eval_wer": 1.0298892988929889,
+      "step": 1115
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.9501385041551246e-05,
+      "loss": 2.6297,
+      "step": 1116
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.9459833795013848e-05,
+      "loss": 2.5378,
+      "step": 1117
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 2.9418282548476452e-05,
+      "loss": 2.5485,
+      "step": 1118
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.9376731301939054e-05,
+      "loss": 2.6042,
+      "step": 1119
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.9335180055401662e-05,
+      "loss": 2.5637,
+      "step": 1120
+    },
+    {
+      "epoch": 3.07,
+      "eval_loss": 2.5621588230133057,
+      "eval_runtime": 157.6164,
+      "eval_samples_per_second": 30.727,
+      "eval_steps_per_second": 3.845,
+      "eval_wer": 1.0847573091115525,
+      "step": 1120
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.9293628808864264e-05,
+      "loss": 2.542,
+      "step": 1121
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 2.925207756232687e-05,
+      "loss": 2.5961,
+      "step": 1122
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.921052631578947e-05,
+      "loss": 2.5422,
+      "step": 1123
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.9168975069252075e-05,
+      "loss": 2.563,
+      "step": 1124
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.9127423822714677e-05,
+      "loss": 2.5692,
+      "step": 1125
+    },
+    {
+      "epoch": 3.08,
+      "eval_loss": 2.556077241897583,
+      "eval_runtime": 141.5277,
+      "eval_samples_per_second": 34.219,
+      "eval_steps_per_second": 4.282,
+      "eval_wer": 1.1471757025262561,
+      "step": 1125
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 2.9085872576177285e-05,
+      "loss": 2.6206,
+      "step": 1126
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.9044321329639886e-05,
+      "loss": 2.6016,
+      "step": 1127
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.900277008310249e-05,
+      "loss": 2.5876,
+      "step": 1128
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 2.8961218836565093e-05,
+      "loss": 2.5195,
+      "step": 1129
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.8919667590027698e-05,
+      "loss": 2.5948,
+      "step": 1130
+    },
+    {
+      "epoch": 3.1,
+      "eval_loss": 2.556828260421753,
+      "eval_runtime": 140.7914,
+      "eval_samples_per_second": 34.398,
+      "eval_steps_per_second": 4.304,
+      "eval_wer": 1.0801589554357083,
+      "step": 1130
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.8878116343490303e-05,
+      "loss": 2.5581,
+      "step": 1131
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.8836565096952908e-05,
+      "loss": 2.5012,
+      "step": 1132
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 2.879501385041551e-05,
+      "loss": 2.5066,
+      "step": 1133
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.8753462603878114e-05,
+      "loss": 2.5888,
+      "step": 1134
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.8711911357340716e-05,
+      "loss": 2.5372,
+      "step": 1135
+    },
+    {
+      "epoch": 3.11,
+      "eval_loss": 2.5638327598571777,
+      "eval_runtime": 159.3627,
+      "eval_samples_per_second": 30.39,
+      "eval_steps_per_second": 3.803,
+      "eval_wer": 1.126114107294919,
+      "step": 1135
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.8670360110803324e-05,
+      "loss": 2.5483,
+      "step": 1136
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 2.8628808864265925e-05,
+      "loss": 2.5979,
+      "step": 1137
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.858725761772853e-05,
+      "loss": 2.5599,
+      "step": 1138
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.8545706371191132e-05,
+      "loss": 2.5663,
+      "step": 1139
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 2.8504155124653737e-05,
+      "loss": 2.4995,
+      "step": 1140
+    },
+    {
+      "epoch": 3.12,
+      "eval_loss": 2.572706937789917,
+      "eval_runtime": 141.419,
+      "eval_samples_per_second": 34.246,
+      "eval_steps_per_second": 4.285,
+      "eval_wer": 1.139511779733182,
+      "step": 1140
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.8462603878116338e-05,
+      "loss": 2.548,
+      "step": 1141
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.8421052631578946e-05,
+      "loss": 2.5133,
+      "step": 1142
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.8379501385041548e-05,
+      "loss": 2.5249,
+      "step": 1143
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 2.8337950138504153e-05,
+      "loss": 2.578,
+      "step": 1144
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.8296398891966754e-05,
+      "loss": 2.6304,
+      "step": 1145
+    },
+    {
+      "epoch": 3.14,
+      "eval_loss": 2.5671443939208984,
+      "eval_runtime": 140.4914,
+      "eval_samples_per_second": 34.472,
+      "eval_steps_per_second": 4.313,
+      "eval_wer": 1.0259437978995174,
+      "step": 1145
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.8254847645429363e-05,
+      "loss": 2.7281,
+      "step": 1146
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 2.8213296398891964e-05,
+      "loss": 2.6481,
+      "step": 1147
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.817174515235457e-05,
+      "loss": 2.5677,
+      "step": 1148
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.813019390581717e-05,
+      "loss": 2.6792,
+      "step": 1149
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.8088642659279776e-05,
+      "loss": 2.6395,
+      "step": 1150
+    },
+    {
+      "epoch": 3.15,
+      "eval_loss": 2.5778486728668213,
+      "eval_runtime": 140.0801,
+      "eval_samples_per_second": 34.573,
+      "eval_steps_per_second": 4.326,
+      "eval_wer": 1.021203519727505,
+      "step": 1150
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 2.8047091412742377e-05,
+      "loss": 2.6219,
+      "step": 1151
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.8005540166204985e-05,
+      "loss": 2.6536,
+      "step": 1152
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.7963988919667587e-05,
+      "loss": 2.6565,
+      "step": 1153
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.7922437673130192e-05,
+      "loss": 2.5832,
+      "step": 1154
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 2.7880886426592793e-05,
+      "loss": 2.6127,
+      "step": 1155
+    },
+    {
+      "epoch": 3.16,
+      "eval_loss": 2.560915470123291,
+      "eval_runtime": 150.0724,
+      "eval_samples_per_second": 32.271,
+      "eval_steps_per_second": 4.038,
+      "eval_wer": 1.0456996877661084,
+      "step": 1155
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.7839335180055398e-05,
+      "loss": 2.6248,
+      "step": 1156
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.7797783933518003e-05,
+      "loss": 2.5818,
+      "step": 1157
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 2.7756232686980608e-05,
+      "loss": 2.5946,
+      "step": 1158
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.771468144044321e-05,
+      "loss": 2.6157,
+      "step": 1159
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.7673130193905814e-05,
+      "loss": 2.5919,
+      "step": 1160
+    },
+    {
+      "epoch": 3.18,
+      "eval_loss": 2.5603954792022705,
+      "eval_runtime": 139.9789,
+      "eval_samples_per_second": 34.598,
+      "eval_steps_per_second": 4.329,
+      "eval_wer": 1.0901788248651718,
+      "step": 1160
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.7631578947368416e-05,
+      "loss": 2.5974,
+      "step": 1161
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 2.7590027700831024e-05,
+      "loss": 2.5705,
+      "step": 1162
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.7548476454293626e-05,
+      "loss": 2.6249,
+      "step": 1163
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.750692520775623e-05,
+      "loss": 2.6387,
+      "step": 1164
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.7465373961218832e-05,
+      "loss": 2.6111,
+      "step": 1165
+    },
+    {
+      "epoch": 3.19,
+      "eval_loss": 2.546295166015625,
+      "eval_runtime": 140.1701,
+      "eval_samples_per_second": 34.551,
+      "eval_steps_per_second": 4.323,
+      "eval_wer": 1.001362475163213,
+      "step": 1165
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 2.7423822714681437e-05,
+      "loss": 2.6139,
+      "step": 1166
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.738227146814404e-05,
+      "loss": 2.6438,
+      "step": 1167
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.7340720221606647e-05,
+      "loss": 2.6285,
+      "step": 1168
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 2.729916897506925e-05,
+      "loss": 2.599,
+      "step": 1169
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.7257617728531853e-05,
+      "loss": 2.5971,
+      "step": 1170
+    },
+    {
+      "epoch": 3.21,
+      "eval_loss": 2.5429396629333496,
+      "eval_runtime": 139.5793,
+      "eval_samples_per_second": 34.697,
+      "eval_steps_per_second": 4.342,
+      "eval_wer": 1.0021572523417541,
+      "step": 1170
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.7216066481994455e-05,
+      "loss": 2.5423,
+      "step": 1171
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.7174515235457063e-05,
+      "loss": 2.6009,
+      "step": 1172
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 2.7132963988919665e-05,
+      "loss": 2.614,
+      "step": 1173
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.709141274238227e-05,
+      "loss": 2.6471,
+      "step": 1174
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.704986149584487e-05,
+      "loss": 2.5887,
+      "step": 1175
+    },
+    {
+      "epoch": 3.22,
+      "eval_loss": 2.539423942565918,
+      "eval_runtime": 139.2321,
+      "eval_samples_per_second": 34.784,
+      "eval_steps_per_second": 4.352,
+      "eval_wer": 1.0411864887879647,
+      "step": 1175
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.7008310249307476e-05,
+      "loss": 2.5649,
+      "step": 1176
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 2.6966759002770077e-05,
+      "loss": 2.5507,
+      "step": 1177
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.6925207756232686e-05,
+      "loss": 2.5415,
+      "step": 1178
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.6883656509695287e-05,
+      "loss": 2.5508,
+      "step": 1179
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 2.6842105263157892e-05,
+      "loss": 2.5644,
+      "step": 1180
+    },
+    {
+      "epoch": 3.23,
+      "eval_loss": 2.534177780151367,
+      "eval_runtime": 144.3721,
+      "eval_samples_per_second": 33.545,
+      "eval_steps_per_second": 4.197,
+      "eval_wer": 1.04689185353392,
+      "step": 1180
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.6800554016620494e-05,
+      "loss": 2.5909,
+      "step": 1181
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.67590027700831e-05,
+      "loss": 2.6377,
+      "step": 1182
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.6717451523545704e-05,
+      "loss": 2.5593,
+      "step": 1183
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 2.667590027700831e-05,
+      "loss": 2.5651,
+      "step": 1184
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.663434903047091e-05,
+      "loss": 2.4805,
+      "step": 1185
+    },
+    {
+      "epoch": 3.25,
+      "eval_loss": 2.6065704822540283,
+      "eval_runtime": 138.8303,
+      "eval_samples_per_second": 34.884,
+      "eval_steps_per_second": 4.365,
+      "eval_wer": 1.2668180527959125,
+      "step": 1185
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.6592797783933515e-05,
+      "loss": 2.5671,
+      "step": 1186
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.6551246537396116e-05,
+      "loss": 2.606,
+      "step": 1187
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 2.6509695290858725e-05,
+      "loss": 2.5668,
+      "step": 1188
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.646814404432133e-05,
+      "loss": 2.6273,
+      "step": 1189
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.642659279778393e-05,
+      "loss": 2.5324,
+      "step": 1190
+    },
+    {
+      "epoch": 3.26,
+      "eval_loss": 2.5394909381866455,
+      "eval_runtime": 141.2379,
+      "eval_samples_per_second": 34.29,
+      "eval_steps_per_second": 4.291,
+      "eval_wer": 1.0234175418677263,
+      "step": 1190
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 2.6385041551246536e-05,
+      "loss": 2.4968,
+      "step": 1191
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.6343490304709138e-05,
+      "loss": 2.5731,
+      "step": 1192
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.6301939058171746e-05,
+      "loss": 2.5527,
+      "step": 1193
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.6260387811634347e-05,
+      "loss": 2.569,
+      "step": 1194
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 2.6218836565096952e-05,
+      "loss": 2.5491,
+      "step": 1195
+    },
+    {
+      "epoch": 3.27,
+      "eval_loss": 2.543052911758423,
+      "eval_runtime": 139.7762,
+      "eval_samples_per_second": 34.648,
+      "eval_steps_per_second": 4.336,
+      "eval_wer": 1.064405336361056,
+      "step": 1195
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.6177285318559554e-05,
+      "loss": 2.6687,
+      "step": 1196
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.613573407202216e-05,
+      "loss": 2.6242,
+      "step": 1197
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.6094182825484764e-05,
+      "loss": 2.659,
+      "step": 1198
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 2.605263157894737e-05,
+      "loss": 2.6644,
+      "step": 1199
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.601108033240997e-05,
+      "loss": 2.6302,
+      "step": 1200
+    },
+    {
+      "epoch": 3.29,
+      "eval_loss": 2.5557799339294434,
+      "eval_runtime": 141.0302,
+      "eval_samples_per_second": 34.34,
+      "eval_steps_per_second": 4.297,
+      "eval_wer": 1.0679534487652569,
+      "step": 1200
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.5969529085872575e-05,
+      "loss": 2.598,
+      "step": 1201
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 2.5927977839335176e-05,
+      "loss": 2.6236,
+      "step": 1202
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.5886426592797785e-05,
+      "loss": 2.6231,
+      "step": 1203
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.5844875346260386e-05,
+      "loss": 2.5897,
+      "step": 1204
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.580332409972299e-05,
+      "loss": 2.6139,
+      "step": 1205
+    },
+    {
+      "epoch": 3.3,
+      "eval_loss": 2.5711026191711426,
+      "eval_runtime": 140.275,
+      "eval_samples_per_second": 34.525,
+      "eval_steps_per_second": 4.32,
+      "eval_wer": 1.056514334374113,
+      "step": 1205
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 2.5761772853185593e-05,
+      "loss": 2.5854,
+      "step": 1206
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.5720221606648198e-05,
+      "loss": 2.6024,
+      "step": 1207
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.56786703601108e-05,
+      "loss": 2.5771,
+      "step": 1208
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.5637119113573407e-05,
+      "loss": 2.616,
+      "step": 1209
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 2.559556786703601e-05,
+      "loss": 2.5607,
+      "step": 1210
+    },
+    {
+      "epoch": 3.31,
+      "eval_loss": 2.563533306121826,
+      "eval_runtime": 140.7436,
+      "eval_samples_per_second": 34.41,
+      "eval_steps_per_second": 4.306,
+      "eval_wer": 1.041527107578768,
+      "step": 1210
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.5554016620498614e-05,
+      "loss": 2.545,
+      "step": 1211
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.5512465373961215e-05,
+      "loss": 2.5622,
+      "step": 1212
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 2.5470914127423824e-05,
+      "loss": 2.6103,
+      "step": 1213
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.5429362880886425e-05,
+      "loss": 2.606,
+      "step": 1214
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.538781163434903e-05,
+      "loss": 2.6535,
+      "step": 1215
+    },
+    {
+      "epoch": 3.33,
+      "eval_loss": 2.5505189895629883,
+      "eval_runtime": 139.3559,
+      "eval_samples_per_second": 34.753,
+      "eval_steps_per_second": 4.349,
+      "eval_wer": 1.0612546125461255,
+      "step": 1215
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.534626038781163e-05,
+      "loss": 2.6134,
+      "step": 1216
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 2.5304709141274236e-05,
+      "loss": 2.6193,
+      "step": 1217
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.5263157894736838e-05,
+      "loss": 2.6594,
+      "step": 1218
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.5221606648199446e-05,
+      "loss": 2.6274,
+      "step": 1219
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.5180055401662048e-05,
+      "loss": 2.6129,
+      "step": 1220
+    },
+    {
+      "epoch": 3.34,
+      "eval_loss": 2.540283679962158,
+      "eval_runtime": 152.2512,
+      "eval_samples_per_second": 31.809,
+      "eval_steps_per_second": 3.98,
+      "eval_wer": 1.0723531081464661,
+      "step": 1220
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 2.5138504155124653e-05,
+      "loss": 2.5915,
+      "step": 1221
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.5096952908587254e-05,
+      "loss": 2.5009,
+      "step": 1222
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.505540166204986e-05,
+      "loss": 2.5471,
+      "step": 1223
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 2.5013850415512464e-05,
+      "loss": 2.5608,
+      "step": 1224
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.497229916897507e-05,
+      "loss": 2.5157,
+      "step": 1225
+    },
+    {
+      "epoch": 3.36,
+      "eval_loss": 2.5294299125671387,
+      "eval_runtime": 157.5914,
+      "eval_samples_per_second": 30.731,
+      "eval_steps_per_second": 3.845,
+      "eval_wer": 1.0584728924212319,
+      "step": 1225
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.493074792243767e-05,
+      "loss": 2.5284,
+      "step": 1226
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.4889196675900275e-05,
+      "loss": 2.5658,
+      "step": 1227
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 2.4847645429362877e-05,
+      "loss": 2.6195,
+      "step": 1228
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.4806094182825485e-05,
+      "loss": 2.5549,
+      "step": 1229
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.4764542936288087e-05,
+      "loss": 2.551,
+      "step": 1230
+    },
+    {
+      "epoch": 3.37,
+      "eval_loss": 2.524221181869507,
+      "eval_runtime": 143.6048,
+      "eval_samples_per_second": 33.725,
+      "eval_steps_per_second": 4.22,
+      "eval_wer": 1.1599489071813796,
+      "step": 1230
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.472299168975069e-05,
+      "loss": 2.5739,
+      "step": 1231
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 2.4681440443213293e-05,
+      "loss": 2.4905,
+      "step": 1232
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.4639889196675898e-05,
+      "loss": 2.5197,
+      "step": 1233
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.45983379501385e-05,
+      "loss": 2.5322,
+      "step": 1234
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 2.4556786703601108e-05,
+      "loss": 2.5527,
+      "step": 1235
+    },
+    {
+      "epoch": 3.38,
+      "eval_loss": 2.547403335571289,
+      "eval_runtime": 140.1677,
+      "eval_samples_per_second": 34.551,
+      "eval_steps_per_second": 4.323,
+      "eval_wer": 1.2326710190178825,
+      "step": 1235
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.451523545706371e-05,
+      "loss": 2.4555,
+      "step": 1236
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.4473684210526314e-05,
+      "loss": 2.4782,
+      "step": 1237
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.4432132963988916e-05,
+      "loss": 2.53,
+      "step": 1238
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 2.4390581717451524e-05,
+      "loss": 2.5533,
+      "step": 1239
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.4349030470914126e-05,
+      "loss": 2.4964,
+      "step": 1240
+    },
+    {
+      "epoch": 3.4,
+      "eval_loss": 2.5244102478027344,
+      "eval_runtime": 156.6755,
+      "eval_samples_per_second": 30.911,
+      "eval_steps_per_second": 3.868,
+      "eval_wer": 1.0856940107862616,
+      "step": 1240
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.430747922437673e-05,
+      "loss": 2.586,
+      "step": 1241
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.4265927977839332e-05,
+      "loss": 2.4998,
+      "step": 1242
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 2.4224376731301937e-05,
+      "loss": 2.5844,
+      "step": 1243
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.418282548476454e-05,
+      "loss": 2.5179,
+      "step": 1244
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.4141274238227147e-05,
+      "loss": 2.5781,
+      "step": 1245
+    },
+    {
+      "epoch": 3.41,
+      "eval_loss": 2.5298705101013184,
+      "eval_runtime": 145.4438,
+      "eval_samples_per_second": 33.298,
+      "eval_steps_per_second": 4.167,
+      "eval_wer": 1.0469770082316208,
+      "step": 1245
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 2.4099722991689748e-05,
+      "loss": 2.6277,
+      "step": 1246
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.4058171745152353e-05,
+      "loss": 2.6374,
+      "step": 1247
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.4016620498614955e-05,
+      "loss": 2.6683,
+      "step": 1248
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.397506925207756e-05,
+      "loss": 2.6298,
+      "step": 1249
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 2.3933518005540164e-05,
+      "loss": 2.6143,
+      "step": 1250
+    },
+    {
+      "epoch": 3.42,
+      "eval_loss": 2.5312798023223877,
+      "eval_runtime": 141.132,
+      "eval_samples_per_second": 34.315,
+      "eval_steps_per_second": 4.294,
+      "eval_wer": 1.0019301731478854,
+      "step": 1250
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.389196675900277e-05,
+      "loss": 2.6277,
+      "step": 1251
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.385041551246537e-05,
+      "loss": 2.6025,
+      "step": 1252
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 2.3808864265927976e-05,
+      "loss": 2.6399,
+      "step": 1253
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.3767313019390577e-05,
+      "loss": 2.6078,
+      "step": 1254
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.3725761772853186e-05,
+      "loss": 2.6566,
+      "step": 1255
+    },
+    {
+      "epoch": 3.44,
+      "eval_loss": 2.5431058406829834,
+      "eval_runtime": 140.2683,
+      "eval_samples_per_second": 34.527,
+      "eval_steps_per_second": 4.32,
+      "eval_wer": 1.048765256883338,
+      "step": 1255
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.3684210526315787e-05,
+      "loss": 2.6106,
+      "step": 1256
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 2.3642659279778392e-05,
+      "loss": 2.6295,
+      "step": 1257
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.3601108033240993e-05,
+      "loss": 2.6731,
+      "step": 1258
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.35595567867036e-05,
+      "loss": 2.5986,
+      "step": 1259
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.35180055401662e-05,
+      "loss": 2.5373,
+      "step": 1260
+    },
+    {
+      "epoch": 3.45,
+      "eval_loss": 2.528088331222534,
+      "eval_runtime": 139.931,
+      "eval_samples_per_second": 34.61,
+      "eval_steps_per_second": 4.331,
+      "eval_wer": 1.0901220550667046,
+      "step": 1260
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 2.3476454293628808e-05,
+      "loss": 2.5876,
+      "step": 1261
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.343490304709141e-05,
+      "loss": 2.602,
+      "step": 1262
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.3393351800554015e-05,
+      "loss": 2.6114,
+      "step": 1263
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 2.3351800554016616e-05,
+      "loss": 2.5745,
+      "step": 1264
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.3310249307479224e-05,
+      "loss": 2.6597,
+      "step": 1265
+    },
+    {
+      "epoch": 3.47,
+      "eval_loss": 2.5299735069274902,
+      "eval_runtime": 139.3913,
+      "eval_samples_per_second": 34.744,
+      "eval_steps_per_second": 4.347,
+      "eval_wer": 1.0609707635537893,
+      "step": 1265
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.3268698060941826e-05,
+      "loss": 2.5861,
+      "step": 1266
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.322714681440443e-05,
+      "loss": 2.5883,
+      "step": 1267
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 2.3185595567867032e-05,
+      "loss": 2.5713,
+      "step": 1268
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.3144044321329637e-05,
+      "loss": 2.5819,
+      "step": 1269
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.310249307479224e-05,
+      "loss": 2.5457,
+      "step": 1270
+    },
+    {
+      "epoch": 3.48,
+      "eval_loss": 2.5130109786987305,
+      "eval_runtime": 143.3843,
+      "eval_samples_per_second": 33.776,
+      "eval_steps_per_second": 4.226,
+      "eval_wer": 1.0419528810672722,
+      "step": 1270
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.3060941828254847e-05,
+      "loss": 2.5728,
+      "step": 1271
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 2.301939058171745e-05,
+      "loss": 2.5323,
+      "step": 1272
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.2977839335180053e-05,
+      "loss": 2.5843,
+      "step": 1273
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.2936288088642655e-05,
+      "loss": 2.5127,
+      "step": 1274
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 2.289473684210526e-05,
+      "loss": 2.5632,
+      "step": 1275
+    },
+    {
+      "epoch": 3.49,
+      "eval_loss": 2.530580520629883,
+      "eval_runtime": 139.7498,
+      "eval_samples_per_second": 34.655,
+      "eval_steps_per_second": 4.336,
+      "eval_wer": 1.141810956571104,
+      "step": 1275
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.2853185595567865e-05,
+      "loss": 2.6214,
+      "step": 1276
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.281163434903047e-05,
+      "loss": 2.5222,
+      "step": 1277
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.277008310249307e-05,
+      "loss": 2.5339,
+      "step": 1278
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 2.2728531855955676e-05,
+      "loss": 2.5666,
+      "step": 1279
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.2686980609418278e-05,
+      "loss": 2.5267,
+      "step": 1280
+    },
+    {
+      "epoch": 3.51,
+      "eval_loss": 2.502101182937622,
+      "eval_runtime": 142.8366,
+      "eval_samples_per_second": 33.906,
+      "eval_steps_per_second": 4.243,
+      "eval_wer": 1.0293499858075503,
+      "step": 1280
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.2645429362880886e-05,
+      "loss": 2.55,
+      "step": 1281
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.2603878116343487e-05,
+      "loss": 2.5833,
+      "step": 1282
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 2.2562326869806092e-05,
+      "loss": 2.5213,
+      "step": 1283
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.2520775623268694e-05,
+      "loss": 2.5311,
+      "step": 1284
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.24792243767313e-05,
+      "loss": 2.507,
+      "step": 1285
+    },
+    {
+      "epoch": 3.52,
+      "eval_loss": 2.5013275146484375,
+      "eval_runtime": 141.6525,
+      "eval_samples_per_second": 34.189,
+      "eval_steps_per_second": 4.278,
+      "eval_wer": 1.0196423502696565,
+      "step": 1285
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 2.24376731301939e-05,
+      "loss": 2.4812,
+      "step": 1286
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.239612188365651e-05,
+      "loss": 2.5609,
+      "step": 1287
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.235457063711911e-05,
+      "loss": 2.5508,
+      "step": 1288
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.2313019390581715e-05,
+      "loss": 2.532,
+      "step": 1289
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 2.2271468144044317e-05,
+      "loss": 2.5713,
+      "step": 1290
+    },
+    {
+      "epoch": 3.53,
+      "eval_loss": 2.4978487491607666,
+      "eval_runtime": 148.1171,
+      "eval_samples_per_second": 32.697,
+      "eval_steps_per_second": 4.091,
+      "eval_wer": 1.066420664206642,
+      "step": 1290
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.2229916897506925e-05,
+      "loss": 2.5371,
+      "step": 1291
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.2188365650969526e-05,
+      "loss": 2.6019,
+      "step": 1292
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.214681440443213e-05,
+      "loss": 2.5275,
+      "step": 1293
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 2.2105263157894733e-05,
+      "loss": 2.5739,
+      "step": 1294
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.2063711911357338e-05,
+      "loss": 2.4783,
+      "step": 1295
+    },
+    {
+      "epoch": 3.55,
+      "eval_loss": 2.4957897663116455,
+      "eval_runtime": 141.4895,
+      "eval_samples_per_second": 34.229,
+      "eval_steps_per_second": 4.283,
+      "eval_wer": 1.052966221969912,
+      "step": 1295
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.2022160664819946e-05,
+      "loss": 2.6544,
+      "step": 1296
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 2.1980609418282547e-05,
+      "loss": 2.6567,
+      "step": 1297
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.1939058171745152e-05,
+      "loss": 2.6406,
+      "step": 1298
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.1897506925207754e-05,
+      "loss": 2.6023,
+      "step": 1299
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.185595567867036e-05,
+      "loss": 2.5874,
+      "step": 1300
+    },
+    {
+      "epoch": 3.56,
+      "eval_loss": 2.4968109130859375,
+      "eval_runtime": 139.379,
+      "eval_samples_per_second": 34.747,
+      "eval_steps_per_second": 4.348,
+      "eval_wer": 1.0059040590405903,
+      "step": 1300
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 2.181440443213296e-05,
+      "loss": 2.612,
+      "step": 1301
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.177285318559557e-05,
+      "loss": 2.5328,
+      "step": 1302
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.173130193905817e-05,
+      "loss": 2.5541,
+      "step": 1303
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.1689750692520775e-05,
+      "loss": 2.5773,
+      "step": 1304
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 2.1648199445983377e-05,
+      "loss": 2.5744,
+      "step": 1305
+    },
+    {
+      "epoch": 3.57,
+      "eval_loss": 2.507821798324585,
+      "eval_runtime": 141.4958,
+      "eval_samples_per_second": 34.227,
+      "eval_steps_per_second": 4.283,
+      "eval_wer": 1.0286687482259438,
+      "step": 1305
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.1606648199445985e-05,
+      "loss": 2.6139,
+      "step": 1306
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.1565096952908586e-05,
+      "loss": 2.5506,
+      "step": 1307
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 2.152354570637119e-05,
+      "loss": 2.6884,
+      "step": 1308
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.1481994459833793e-05,
+      "loss": 2.599,
+      "step": 1309
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.1440443213296398e-05,
+      "loss": 2.5701,
+      "step": 1310
+    },
+    {
+      "epoch": 3.59,
+      "eval_loss": 2.4970922470092773,
+      "eval_runtime": 139.9396,
+      "eval_samples_per_second": 34.608,
+      "eval_steps_per_second": 4.33,
+      "eval_wer": 1.0365881351121204,
+      "step": 1310
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.1398891966759e-05,
+      "loss": 2.547,
+      "step": 1311
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 2.1357340720221607e-05,
+      "loss": 2.581,
+      "step": 1312
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.131578947368421e-05,
+      "loss": 2.5355,
+      "step": 1313
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.1274238227146814e-05,
+      "loss": 2.5744,
+      "step": 1314
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.1232686980609415e-05,
+      "loss": 2.5366,
+      "step": 1315
+    },
+    {
+      "epoch": 3.6,
+      "eval_loss": 2.489727258682251,
+      "eval_runtime": 140.3326,
+      "eval_samples_per_second": 34.511,
+      "eval_steps_per_second": 4.318,
+      "eval_wer": 1.019103037184218,
+      "step": 1315
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 2.119113573407202e-05,
+      "loss": 2.5345,
+      "step": 1316
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.1149584487534625e-05,
+      "loss": 2.5958,
+      "step": 1317
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.110803324099723e-05,
+      "loss": 2.5088,
+      "step": 1318
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 2.106648199445983e-05,
+      "loss": 2.5148,
+      "step": 1319
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.1024930747922437e-05,
+      "loss": 2.5679,
+      "step": 1320
+    },
+    {
+      "epoch": 3.62,
+      "eval_loss": 2.4830410480499268,
+      "eval_runtime": 139.7073,
+      "eval_samples_per_second": 34.665,
+      "eval_steps_per_second": 4.338,
+      "eval_wer": 1.0223105307976157,
+      "step": 1320
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.0983379501385038e-05,
+      "loss": 2.5739,
+      "step": 1321
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.0941828254847646e-05,
+      "loss": 2.5549,
+      "step": 1322
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 2.0900277008310248e-05,
+      "loss": 2.5686,
+      "step": 1323
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.0858725761772853e-05,
+      "loss": 2.5492,
+      "step": 1324
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.0817174515235454e-05,
+      "loss": 2.5239,
+      "step": 1325
+    },
+    {
+      "epoch": 3.63,
+      "eval_loss": 2.4832916259765625,
+      "eval_runtime": 141.0566,
+      "eval_samples_per_second": 34.334,
+      "eval_steps_per_second": 4.296,
+      "eval_wer": 1.0783990916832245,
+      "step": 1325
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.077562326869806e-05,
+      "loss": 2.5057,
+      "step": 1326
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 2.073407202216066e-05,
+      "loss": 2.5153,
+      "step": 1327
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.069252077562327e-05,
+      "loss": 2.5305,
+      "step": 1328
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.065096952908587e-05,
+      "loss": 2.5195,
+      "step": 1329
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 2.0609418282548475e-05,
+      "loss": 2.5411,
+      "step": 1330
+    },
+    {
+      "epoch": 3.64,
+      "eval_loss": 2.4850964546203613,
+      "eval_runtime": 140.8565,
+      "eval_samples_per_second": 34.383,
+      "eval_steps_per_second": 4.302,
+      "eval_wer": 1.1521998296906046,
+      "step": 1330
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.0567867036011077e-05,
+      "loss": 2.5186,
+      "step": 1331
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.0526315789473685e-05,
+      "loss": 2.5188,
+      "step": 1332
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.0484764542936287e-05,
+      "loss": 2.47,
+      "step": 1333
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 2.044321329639889e-05,
+      "loss": 2.5474,
+      "step": 1334
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.0401662049861493e-05,
+      "loss": 2.5037,
+      "step": 1335
+    },
+    {
+      "epoch": 3.66,
+      "eval_loss": 2.47922945022583,
+      "eval_runtime": 145.8466,
+      "eval_samples_per_second": 33.206,
+      "eval_steps_per_second": 4.155,
+      "eval_wer": 1.0928186204938972,
+      "step": 1335
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.0360110803324098e-05,
+      "loss": 2.5169,
+      "step": 1336
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.03185595567867e-05,
+      "loss": 2.545,
+      "step": 1337
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 2.0277008310249308e-05,
+      "loss": 2.5419,
+      "step": 1338
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.023545706371191e-05,
+      "loss": 2.5811,
+      "step": 1339
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.0193905817174514e-05,
+      "loss": 2.5907,
+      "step": 1340
+    },
+    {
+      "epoch": 3.67,
+      "eval_loss": 2.474959135055542,
+      "eval_runtime": 139.5377,
+      "eval_samples_per_second": 34.707,
+      "eval_steps_per_second": 4.343,
+      "eval_wer": 1.018677263695714,
+      "step": 1340
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 2.0152354570637116e-05,
+      "loss": 2.5418,
+      "step": 1341
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.011080332409972e-05,
+      "loss": 2.4793,
+      "step": 1342
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.0069252077562326e-05,
+      "loss": 2.5392,
+      "step": 1343
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 2.002770083102493e-05,
+      "loss": 2.5988,
+      "step": 1344
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 1.9986149584487532e-05,
+      "loss": 2.5107,
+      "step": 1345
+    },
+    {
+      "epoch": 3.68,
+      "eval_loss": 2.4805235862731934,
+      "eval_runtime": 140.577,
+      "eval_samples_per_second": 34.451,
+      "eval_steps_per_second": 4.311,
+      "eval_wer": 1.0872551802441102,
+      "step": 1345
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.9944598337950137e-05,
+      "loss": 2.6263,
+      "step": 1346
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.990304709141274e-05,
+      "loss": 2.6014,
+      "step": 1347
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.9861495844875347e-05,
+      "loss": 2.6408,
+      "step": 1348
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 1.981994459833795e-05,
+      "loss": 2.6359,
+      "step": 1349
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.9778393351800553e-05,
+      "loss": 2.5908,
+      "step": 1350
+    },
+    {
+      "epoch": 3.7,
+      "eval_loss": 2.475250720977783,
+      "eval_runtime": 160.4663,
+      "eval_samples_per_second": 30.181,
+      "eval_steps_per_second": 3.776,
+      "eval_wer": 1.0097927902355948,
+      "step": 1350
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.9736842105263155e-05,
+      "loss": 2.6107,
+      "step": 1351
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 1.969529085872576e-05,
+      "loss": 2.5853,
+      "step": 1352
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.965373961218836e-05,
+      "loss": 2.5924,
+      "step": 1353
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.961218836565097e-05,
+      "loss": 2.6257,
+      "step": 1354
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.957063711911357e-05,
+      "loss": 2.6274,
+      "step": 1355
+    },
+    {
+      "epoch": 3.71,
+      "eval_loss": 2.476541042327881,
+      "eval_runtime": 152.0314,
+      "eval_samples_per_second": 31.855,
+      "eval_steps_per_second": 3.986,
+      "eval_wer": 1.0045415838773772,
+      "step": 1355
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 1.9529085872576176e-05,
+      "loss": 2.5623,
+      "step": 1356
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.9487534626038777e-05,
+      "loss": 2.5641,
+      "step": 1357
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.9445983379501386e-05,
+      "loss": 2.5586,
+      "step": 1358
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 1.9404432132963987e-05,
+      "loss": 2.5546,
+      "step": 1359
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.9362880886426592e-05,
+      "loss": 2.5708,
+      "step": 1360
+    },
+    {
+      "epoch": 3.73,
+      "eval_loss": 2.45967698097229,
+      "eval_runtime": 157.4234,
+      "eval_samples_per_second": 30.764,
+      "eval_steps_per_second": 3.849,
+      "eval_wer": 1.045586148169174,
+      "step": 1360
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.9321329639889194e-05,
+      "loss": 2.5934,
+      "step": 1361
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.92797783933518e-05,
+      "loss": 2.5776,
+      "step": 1362
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 1.92382271468144e-05,
+      "loss": 2.5552,
+      "step": 1363
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.919667590027701e-05,
+      "loss": 2.4912,
+      "step": 1364
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.915512465373961e-05,
+      "loss": 2.6039,
+      "step": 1365
+    },
+    {
+      "epoch": 3.74,
+      "eval_loss": 2.4503345489501953,
+      "eval_runtime": 141.5529,
+      "eval_samples_per_second": 34.213,
+      "eval_steps_per_second": 4.281,
+      "eval_wer": 1.048481407891002,
+      "step": 1365
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.9113573407202215e-05,
+      "loss": 2.5709,
+      "step": 1366
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 1.9072022160664816e-05,
+      "loss": 2.4991,
+      "step": 1367
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.903047091412742e-05,
+      "loss": 2.5346,
+      "step": 1368
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.8988919667590026e-05,
+      "loss": 2.6553,
+      "step": 1369
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 1.894736842105263e-05,
+      "loss": 2.5305,
+      "step": 1370
+    },
+    {
+      "epoch": 3.75,
+      "eval_loss": 2.4438858032226562,
+      "eval_runtime": 140.8099,
+      "eval_samples_per_second": 34.394,
+      "eval_steps_per_second": 4.304,
+      "eval_wer": 1.0125745103604882,
+      "step": 1370
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8905817174515233e-05,
+      "loss": 2.4997,
+      "step": 1371
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8864265927977837e-05,
+      "loss": 2.5448,
+      "step": 1372
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.882271468144044e-05,
+      "loss": 2.5074,
+      "step": 1373
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 1.8781163434903047e-05,
+      "loss": 2.4434,
+      "step": 1374
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.873961218836565e-05,
+      "loss": 2.4878,
+      "step": 1375
+    },
+    {
+      "epoch": 3.77,
+      "eval_loss": 2.4406561851501465,
+      "eval_runtime": 141.5104,
+      "eval_samples_per_second": 34.224,
+      "eval_steps_per_second": 4.282,
+      "eval_wer": 1.01620777746239,
+      "step": 1375
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.8698060941828254e-05,
+      "loss": 2.5243,
+      "step": 1376
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.865650969529086e-05,
+      "loss": 2.4966,
+      "step": 1377
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 1.861495844875346e-05,
+      "loss": 2.4777,
+      "step": 1378
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.8573407202216065e-05,
+      "loss": 2.5574,
+      "step": 1379
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.853185595567867e-05,
+      "loss": 2.5055,
+      "step": 1380
+    },
+    {
+      "epoch": 3.78,
+      "eval_loss": 2.4420855045318604,
+      "eval_runtime": 141.4709,
+      "eval_samples_per_second": 34.233,
+      "eval_steps_per_second": 4.284,
+      "eval_wer": 1.0605449900652852,
+      "step": 1380
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 1.849030470914127e-05,
+      "loss": 2.6108,
+      "step": 1381
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8448753462603876e-05,
+      "loss": 2.5129,
+      "step": 1382
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.840720221606648e-05,
+      "loss": 2.5103,
+      "step": 1383
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8365650969529086e-05,
+      "loss": 2.5454,
+      "step": 1384
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 1.8324099722991688e-05,
+      "loss": 2.5249,
+      "step": 1385
+    },
+    {
+      "epoch": 3.79,
+      "eval_loss": 2.449862003326416,
+      "eval_runtime": 142.3281,
+      "eval_samples_per_second": 34.027,
+      "eval_steps_per_second": 4.258,
+      "eval_wer": 1.1162929321600907,
+      "step": 1385
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8282548476454293e-05,
+      "loss": 2.5665,
+      "step": 1386
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8240997229916897e-05,
+      "loss": 2.5636,
+      "step": 1387
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.81994459833795e-05,
+      "loss": 2.5557,
+      "step": 1388
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 1.8157894736842104e-05,
+      "loss": 2.5131,
+      "step": 1389
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.811634349030471e-05,
+      "loss": 2.5508,
+      "step": 1390
+    },
+    {
+      "epoch": 3.81,
+      "eval_loss": 2.4653842449188232,
+      "eval_runtime": 144.7652,
+      "eval_samples_per_second": 33.454,
+      "eval_steps_per_second": 4.186,
+      "eval_wer": 1.1471757025262561,
+      "step": 1390
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.807479224376731e-05,
+      "loss": 2.5625,
+      "step": 1391
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 1.8033240997229915e-05,
+      "loss": 2.6128,
+      "step": 1392
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.799168975069252e-05,
+      "loss": 2.5363,
+      "step": 1393
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.795013850415512e-05,
+      "loss": 2.5673,
+      "step": 1394
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.7908587257617727e-05,
+      "loss": 2.5827,
+      "step": 1395
+    },
+    {
+      "epoch": 3.82,
+      "eval_loss": 2.4509823322296143,
+      "eval_runtime": 139.7662,
+      "eval_samples_per_second": 34.651,
+      "eval_steps_per_second": 4.336,
+      "eval_wer": 1.056145330684076,
+      "step": 1395
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 1.786703601108033e-05,
+      "loss": 2.6625,
+      "step": 1396
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.7825484764542936e-05,
+      "loss": 2.6369,
+      "step": 1397
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.7783933518005538e-05,
+      "loss": 2.6523,
+      "step": 1398
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.7742382271468143e-05,
+      "loss": 2.6161,
+      "step": 1399
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 1.7700831024930748e-05,
+      "loss": 2.6148,
+      "step": 1400
+    },
+    {
+      "epoch": 3.83,
+      "eval_loss": 2.4495983123779297,
+      "eval_runtime": 156.7735,
+      "eval_samples_per_second": 30.892,
+      "eval_steps_per_second": 3.865,
+      "eval_wer": 0.9997729208061311,
+      "step": 1400
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.765927977839335e-05,
+      "loss": 2.5951,
+      "step": 1401
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.7617728531855954e-05,
+      "loss": 2.6617,
+      "step": 1402
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 1.757617728531856e-05,
+      "loss": 2.5516,
+      "step": 1403
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.753462603878116e-05,
+      "loss": 2.6057,
+      "step": 1404
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.7493074792243765e-05,
+      "loss": 2.5763,
+      "step": 1405
+    },
+    {
+      "epoch": 3.85,
+      "eval_loss": 2.44173002243042,
+      "eval_runtime": 145.2665,
+      "eval_samples_per_second": 33.339,
+      "eval_steps_per_second": 4.172,
+      "eval_wer": 1.0066988362191314,
+      "step": 1405
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.745152354570637e-05,
+      "loss": 2.5929,
+      "step": 1406
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 1.7409972299168972e-05,
+      "loss": 2.5835,
+      "step": 1407
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.7368421052631577e-05,
+      "loss": 2.6092,
+      "step": 1408
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.732686980609418e-05,
+      "loss": 2.5397,
+      "step": 1409
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.7285318559556787e-05,
+      "loss": 2.6077,
+      "step": 1410
+    },
+    {
+      "epoch": 3.86,
+      "eval_loss": 2.4457879066467285,
+      "eval_runtime": 148.7387,
+      "eval_samples_per_second": 32.56,
+      "eval_steps_per_second": 4.074,
+      "eval_wer": 1.0682089128583594,
+      "step": 1410
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 1.7243767313019388e-05,
+      "loss": 2.6019,
+      "step": 1411
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.7202216066481993e-05,
+      "loss": 2.579,
+      "step": 1412
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.7202216066481993e-05,
+      "loss": 2.5071,
+      "step": 1413
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 1.7160664819944598e-05,
+      "loss": 2.5964,
+      "step": 1414
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.71191135734072e-05,
+      "loss": 2.5388,
+      "step": 1415
+    },
+    {
+      "epoch": 3.88,
+      "eval_loss": 2.4352152347564697,
+      "eval_runtime": 140.2204,
+      "eval_samples_per_second": 34.538,
+      "eval_steps_per_second": 4.322,
+      "eval_wer": 1.0820323587851264,
+      "step": 1415
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.7077562326869804e-05,
+      "loss": 2.6199,
+      "step": 1416
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.703601108033241e-05,
+      "loss": 2.5453,
+      "step": 1417
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 1.699445983379501e-05,
+      "loss": 2.5166,
+      "step": 1418
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.6952908587257616e-05,
+      "loss": 2.5409,
+      "step": 1419
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.691135734072022e-05,
+      "loss": 2.5235,
+      "step": 1420
+    },
+    {
+      "epoch": 3.89,
+      "eval_loss": 2.4277079105377197,
+      "eval_runtime": 140.7652,
+      "eval_samples_per_second": 34.405,
+      "eval_steps_per_second": 4.305,
+      "eval_wer": 1.078370706783991,
+      "step": 1420
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.6869806094182822e-05,
+      "loss": 2.5293,
+      "step": 1421
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.6828254847645427e-05,
+      "loss": 2.564,
+      "step": 1422
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.6786703601108032e-05,
+      "loss": 2.5232,
+      "step": 1423
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.6745152354570637e-05,
+      "loss": 2.5388,
+      "step": 1424
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 1.6703601108033238e-05,
+      "loss": 2.4996,
+      "step": 1425
+    },
+    {
+      "epoch": 3.9,
+      "eval_loss": 2.4244942665100098,
+      "eval_runtime": 139.4701,
+      "eval_samples_per_second": 34.724,
+      "eval_steps_per_second": 4.345,
+      "eval_wer": 1.0671302866874823,
+      "step": 1425
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.6662049861495843e-05,
+      "loss": 2.5179,
+      "step": 1426
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.6620498614958448e-05,
+      "loss": 2.4875,
+      "step": 1427
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.657894736842105e-05,
+      "loss": 2.5264,
+      "step": 1428
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.6537396121883654e-05,
+      "loss": 2.5416,
+      "step": 1429
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.649584487534626e-05,
+      "loss": 2.5601,
+      "step": 1430
+    },
+    {
+      "epoch": 3.92,
+      "eval_loss": 2.4201695919036865,
+      "eval_runtime": 139.5554,
+      "eval_samples_per_second": 34.703,
+      "eval_steps_per_second": 4.342,
+      "eval_wer": 1.0650298041441952,
+      "step": 1430
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.645429362880886e-05,
+      "loss": 2.465,
+      "step": 1431
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.6412742382271466e-05,
+      "loss": 2.5486,
+      "step": 1432
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 1.637119113573407e-05,
+      "loss": 2.6054,
+      "step": 1433
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.6329639889196672e-05,
+      "loss": 2.4953,
+      "step": 1434
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.6288088642659277e-05,
+      "loss": 2.5805,
+      "step": 1435
+    },
+    {
+      "epoch": 3.93,
+      "eval_loss": 2.4199044704437256,
+      "eval_runtime": 144.5086,
+      "eval_samples_per_second": 33.514,
+      "eval_steps_per_second": 4.194,
+      "eval_wer": 1.0530229917683793,
+      "step": 1435
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.6246537396121882e-05,
+      "loss": 2.5311,
+      "step": 1436
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.6204986149584487e-05,
+      "loss": 2.5084,
+      "step": 1437
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.616343490304709e-05,
+      "loss": 2.505,
+      "step": 1438
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.6121883656509693e-05,
+      "loss": 2.4964,
+      "step": 1439
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 1.6080332409972298e-05,
+      "loss": 2.5841,
+      "step": 1440
+    },
+    {
+      "epoch": 3.94,
+      "eval_loss": 2.4227607250213623,
+      "eval_runtime": 159.6123,
+      "eval_samples_per_second": 30.342,
+      "eval_steps_per_second": 3.797,
+      "eval_wer": 1.079733181947204,
+      "step": 1440
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.60387811634349e-05,
+      "loss": 2.5065,
+      "step": 1441
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.5997229916897505e-05,
+      "loss": 2.5736,
+      "step": 1442
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.595567867036011e-05,
+      "loss": 2.5266,
+      "step": 1443
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 1.591412742382271e-05,
+      "loss": 2.5491,
+      "step": 1444
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.5872576177285316e-05,
+      "loss": 2.4877,
+      "step": 1445
+    },
+    {
+      "epoch": 3.96,
+      "eval_loss": 2.4284441471099854,
+      "eval_runtime": 150.4761,
+      "eval_samples_per_second": 32.185,
+      "eval_steps_per_second": 4.027,
+      "eval_wer": 1.115923928470054,
+      "step": 1445
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.583102493074792e-05,
+      "loss": 2.5998,
+      "step": 1446
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 1.5789473684210522e-05,
+      "loss": 2.6028,
+      "step": 1447
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.5747922437673127e-05,
+      "loss": 2.5767,
+      "step": 1448
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.5706371191135732e-05,
+      "loss": 2.5556,
+      "step": 1449
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.5664819944598337e-05,
+      "loss": 2.5542,
+      "step": 1450
+    },
+    {
+      "epoch": 3.97,
+      "eval_loss": 2.4189741611480713,
+      "eval_runtime": 141.4662,
+      "eval_samples_per_second": 34.234,
+      "eval_steps_per_second": 4.284,
+      "eval_wer": 1.057451036048822,
+      "step": 1450
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 1.562326869806094e-05,
+      "loss": 2.5359,
+      "step": 1451
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.5581717451523544e-05,
+      "loss": 2.5129,
+      "step": 1452
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.554016620498615e-05,
+      "loss": 2.555,
+      "step": 1453
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.549861495844875e-05,
+      "loss": 2.4999,
+      "step": 1454
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 1.5457063711911355e-05,
+      "loss": 2.5961,
+      "step": 1455
+    },
+    {
+      "epoch": 3.98,
+      "eval_loss": 2.416226625442505,
+      "eval_runtime": 139.8778,
+      "eval_samples_per_second": 34.623,
+      "eval_steps_per_second": 4.332,
+      "eval_wer": 1.0675844450752199,
+      "step": 1455
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.541551246537396e-05,
+      "loss": 2.4591,
+      "step": 1456
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.5373961218836565e-05,
+      "loss": 2.5745,
+      "step": 1457
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 1.533240997229917e-05,
+      "loss": 2.5951,
+      "step": 1458
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.529085872576177e-05,
+      "loss": 2.5031,
+      "step": 1459
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.5249307479224376e-05,
+      "loss": 2.495,
+      "step": 1460
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 2.4164938926696777,
+      "eval_runtime": 140.4955,
+      "eval_samples_per_second": 34.471,
+      "eval_steps_per_second": 4.313,
+      "eval_wer": 1.0820891285835936,
+      "step": 1460
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 1.5207756232686981e-05,
+      "loss": 3.9209,
+      "step": 1461
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.5166204986149584e-05,
+      "loss": 2.623,
+      "step": 1462
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.5124653739612187e-05,
+      "loss": 2.5774,
+      "step": 1463
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.5083102493074792e-05,
+      "loss": 2.585,
+      "step": 1464
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 1.5041551246537395e-05,
+      "loss": 2.6157,
+      "step": 1465
+    },
+    {
+      "epoch": 4.01,
+      "eval_loss": 2.4118785858154297,
+      "eval_runtime": 139.8842,
+      "eval_samples_per_second": 34.621,
+      "eval_steps_per_second": 4.332,
+      "eval_wer": 1.0116661935850129,
+      "step": 1465
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.4999999999999999e-05,
+      "loss": 2.5665,
+      "step": 1466
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.4958448753462604e-05,
+      "loss": 2.5997,
+      "step": 1467
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.4916897506925207e-05,
+      "loss": 2.5887,
+      "step": 1468
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 1.4875346260387812e-05,
+      "loss": 2.5462,
+      "step": 1469
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.4833795013850415e-05,
+      "loss": 2.5415,
+      "step": 1470
+    },
+    {
+      "epoch": 4.03,
+      "eval_loss": 2.4089488983154297,
+      "eval_runtime": 142.6522,
+      "eval_samples_per_second": 33.95,
+      "eval_steps_per_second": 4.248,
+      "eval_wer": 1.0110417258018733,
+      "step": 1470
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.4792243767313018e-05,
+      "loss": 2.5649,
+      "step": 1471
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 1.4750692520775623e-05,
+      "loss": 2.5501,
+      "step": 1472
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.4709141274238226e-05,
+      "loss": 2.4939,
+      "step": 1473
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.4667590027700831e-05,
+      "loss": 2.533,
+      "step": 1474
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.4626038781163434e-05,
+      "loss": 2.4916,
+      "step": 1475
+    },
+    {
+      "epoch": 4.04,
+      "eval_loss": 2.4032280445098877,
+      "eval_runtime": 142.8548,
+      "eval_samples_per_second": 33.902,
+      "eval_steps_per_second": 4.242,
+      "eval_wer": 1.049787113255748,
+      "step": 1475
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 1.4584487534626038e-05,
+      "loss": 2.521,
+      "step": 1476
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.4542936288088642e-05,
+      "loss": 2.5257,
+      "step": 1477
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.4501385041551246e-05,
+      "loss": 2.5064,
+      "step": 1478
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.4459833795013849e-05,
+      "loss": 2.5356,
+      "step": 1479
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 1.4418282548476454e-05,
+      "loss": 2.5445,
+      "step": 1480
+    },
+    {
+      "epoch": 4.05,
+      "eval_loss": 2.3997323513031006,
+      "eval_runtime": 141.29,
+      "eval_samples_per_second": 34.277,
+      "eval_steps_per_second": 4.289,
+      "eval_wer": 1.0429463525404485,
+      "step": 1480
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.4376731301939057e-05,
+      "loss": 2.5841,
+      "step": 1481
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.4335180055401662e-05,
+      "loss": 2.6174,
+      "step": 1482
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 1.4293628808864265e-05,
+      "loss": 2.486,
+      "step": 1483
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.4252077562326868e-05,
+      "loss": 2.6067,
+      "step": 1484
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.4210526315789473e-05,
+      "loss": 2.4941,
+      "step": 1485
+    },
+    {
+      "epoch": 4.07,
+      "eval_loss": 2.4007747173309326,
+      "eval_runtime": 149.6997,
+      "eval_samples_per_second": 32.351,
+      "eval_steps_per_second": 4.048,
+      "eval_wer": 1.0140789100198695,
+      "step": 1485
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.4168975069252076e-05,
+      "loss": 2.5266,
+      "step": 1486
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 1.4127423822714681e-05,
+      "loss": 2.4845,
+      "step": 1487
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.4085872576177285e-05,
+      "loss": 2.5147,
+      "step": 1488
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.4044321329639888e-05,
+      "loss": 2.5391,
+      "step": 1489
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.4002770083102493e-05,
+      "loss": 2.5113,
+      "step": 1490
+    },
+    {
+      "epoch": 4.08,
+      "eval_loss": 2.397458553314209,
+      "eval_runtime": 140.4549,
+      "eval_samples_per_second": 34.481,
+      "eval_steps_per_second": 4.315,
+      "eval_wer": 1.0357082032358784,
+      "step": 1490
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 1.3961218836565096e-05,
+      "loss": 2.4974,
+      "step": 1491
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.3919667590027699e-05,
+      "loss": 2.5029,
+      "step": 1492
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.3878116343490304e-05,
+      "loss": 2.5135,
+      "step": 1493
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 1.3836565096952907e-05,
+      "loss": 2.485,
+      "step": 1494
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.3795013850415512e-05,
+      "loss": 2.4707,
+      "step": 1495
+    },
+    {
+      "epoch": 4.1,
+      "eval_loss": 2.3937840461730957,
+      "eval_runtime": 140.9908,
+      "eval_samples_per_second": 34.35,
+      "eval_steps_per_second": 4.298,
+      "eval_wer": 1.0288390576213455,
+      "step": 1495
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.3753462603878115e-05,
+      "loss": 2.4739,
+      "step": 1496
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.3711911357340719e-05,
+      "loss": 2.5664,
+      "step": 1497
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 1.3670360110803323e-05,
+      "loss": 2.4567,
+      "step": 1498
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.3628808864265927e-05,
+      "loss": 2.4789,
+      "step": 1499
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.3587257617728532e-05,
+      "loss": 2.4952,
+      "step": 1500
+    },
+    {
+      "epoch": 4.11,
+      "eval_loss": 2.3910233974456787,
+      "eval_runtime": 143.266,
+      "eval_samples_per_second": 33.804,
+      "eval_steps_per_second": 4.23,
+      "eval_wer": 1.0299744535906898,
+      "step": 1500
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.3545706371191135e-05,
+      "loss": 2.4834,
+      "step": 1501
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 1.3504155124653738e-05,
+      "loss": 2.4824,
+      "step": 1502
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.3462603878116343e-05,
+      "loss": 2.4848,
+      "step": 1503
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.3421052631578946e-05,
+      "loss": 2.5534,
+      "step": 1504
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 1.337950138504155e-05,
+      "loss": 2.5017,
+      "step": 1505
+    },
+    {
+      "epoch": 4.12,
+      "eval_loss": 2.386096477508545,
+      "eval_runtime": 146.5274,
+      "eval_samples_per_second": 33.052,
+      "eval_steps_per_second": 4.136,
+      "eval_wer": 1.0812943514050526,
+      "step": 1505
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.3337950138504154e-05,
+      "loss": 2.527,
+      "step": 1506
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.3296398891966757e-05,
+      "loss": 2.5528,
+      "step": 1507
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.3254847645429362e-05,
+      "loss": 2.4482,
+      "step": 1508
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 1.3213296398891966e-05,
+      "loss": 2.5257,
+      "step": 1509
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.3171745152354569e-05,
+      "loss": 2.5566,
+      "step": 1510
+    },
+    {
+      "epoch": 4.14,
+      "eval_loss": 2.391861915588379,
+      "eval_runtime": 140.1338,
+      "eval_samples_per_second": 34.56,
+      "eval_steps_per_second": 4.324,
+      "eval_wer": 1.108174850979279,
+      "step": 1510
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.3130193905817174e-05,
+      "loss": 2.589,
+      "step": 1511
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 1.3088642659279777e-05,
+      "loss": 2.6453,
+      "step": 1512
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.3047091412742382e-05,
+      "loss": 2.6068,
+      "step": 1513
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.3005540166204985e-05,
+      "loss": 2.5793,
+      "step": 1514
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.2963988919667588e-05,
+      "loss": 2.5754,
+      "step": 1515
+    },
+    {
+      "epoch": 4.15,
+      "eval_loss": 2.3946664333343506,
+      "eval_runtime": 159.2413,
+      "eval_samples_per_second": 30.413,
+      "eval_steps_per_second": 3.806,
+      "eval_wer": 1.007380073800738,
+      "step": 1515
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 1.2922437673130193e-05,
+      "loss": 2.5381,
+      "step": 1516
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.2880886426592796e-05,
+      "loss": 2.534,
+      "step": 1517
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.28393351800554e-05,
+      "loss": 2.5971,
+      "step": 1518
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.2797783933518004e-05,
+      "loss": 2.5409,
+      "step": 1519
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 1.2756232686980608e-05,
+      "loss": 2.6138,
+      "step": 1520
+    },
+    {
+      "epoch": 4.16,
+      "eval_loss": 2.4039783477783203,
+      "eval_runtime": 158.3691,
+      "eval_samples_per_second": 30.58,
+      "eval_steps_per_second": 3.827,
+      "eval_wer": 0.9989213738291229,
+      "step": 1520
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.2714681440443213e-05,
+      "loss": 2.575,
+      "step": 1521
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.2673130193905816e-05,
+      "loss": 2.5858,
+      "step": 1522
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 1.2631578947368419e-05,
+      "loss": 2.5557,
+      "step": 1523
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.2590027700831024e-05,
+      "loss": 2.579,
+      "step": 1524
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.2548476454293627e-05,
+      "loss": 2.5024,
+      "step": 1525
+    },
+    {
+      "epoch": 4.18,
+      "eval_loss": 2.394920825958252,
+      "eval_runtime": 146.2577,
+      "eval_samples_per_second": 33.113,
+      "eval_steps_per_second": 4.143,
+      "eval_wer": 1.0038603462957707,
+      "step": 1525
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.2506925207756232e-05,
+      "loss": 2.5674,
+      "step": 1526
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 1.2465373961218835e-05,
+      "loss": 2.4979,
+      "step": 1527
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.2423822714681438e-05,
+      "loss": 2.5494,
+      "step": 1528
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.2382271468144043e-05,
+      "loss": 2.6069,
+      "step": 1529
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.2340720221606647e-05,
+      "loss": 2.5136,
+      "step": 1530
+    },
+    {
+      "epoch": 4.19,
+      "eval_loss": 2.399284839630127,
+      "eval_runtime": 145.446,
+      "eval_samples_per_second": 33.298,
+      "eval_steps_per_second": 4.166,
+      "eval_wer": 1.04964518875958,
+      "step": 1530
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 1.229916897506925e-05,
+      "loss": 2.5401,
+      "step": 1531
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.2257617728531855e-05,
+      "loss": 2.5613,
+      "step": 1532
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.2216066481994458e-05,
+      "loss": 2.5148,
+      "step": 1533
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 1.2174515235457063e-05,
+      "loss": 2.5706,
+      "step": 1534
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.2132963988919666e-05,
+      "loss": 2.5646,
+      "step": 1535
+    },
+    {
+      "epoch": 4.21,
+      "eval_loss": 2.3980846405029297,
+      "eval_runtime": 141.3145,
+      "eval_samples_per_second": 34.271,
+      "eval_steps_per_second": 4.288,
+      "eval_wer": 1.072864036332671,
+      "step": 1535
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.209141274238227e-05,
+      "loss": 2.5162,
+      "step": 1536
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.2049861495844874e-05,
+      "loss": 2.5069,
+      "step": 1537
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 1.2008310249307477e-05,
+      "loss": 2.4451,
+      "step": 1538
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.1966759002770082e-05,
+      "loss": 2.5444,
+      "step": 1539
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.1925207756232685e-05,
+      "loss": 2.4556,
+      "step": 1540
+    },
+    {
+      "epoch": 4.22,
+      "eval_loss": 2.395197868347168,
+      "eval_runtime": 141.3826,
+      "eval_samples_per_second": 34.255,
+      "eval_steps_per_second": 4.286,
+      "eval_wer": 1.0494464944649446,
+      "step": 1540
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.1883656509695289e-05,
+      "loss": 2.5325,
+      "step": 1541
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 1.1842105263157894e-05,
+      "loss": 2.4633,
+      "step": 1542
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.1800554016620497e-05,
+      "loss": 2.497,
+      "step": 1543
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.17590027700831e-05,
+      "loss": 2.5419,
+      "step": 1544
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 1.1717451523545705e-05,
+      "loss": 2.5774,
+      "step": 1545
+    },
+    {
+      "epoch": 4.23,
+      "eval_loss": 2.392441511154175,
+      "eval_runtime": 141.9107,
+      "eval_samples_per_second": 34.127,
+      "eval_steps_per_second": 4.27,
+      "eval_wer": 1.034516037468067,
+      "step": 1545
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.1675900277008308e-05,
+      "loss": 2.5979,
+      "step": 1546
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.1634349030470913e-05,
+      "loss": 2.4978,
+      "step": 1547
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.1592797783933516e-05,
+      "loss": 2.5588,
+      "step": 1548
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 1.155124653739612e-05,
+      "loss": 2.5519,
+      "step": 1549
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.1509695290858724e-05,
+      "loss": 2.5126,
+      "step": 1550
+    },
+    {
+      "epoch": 4.25,
+      "eval_loss": 2.3887672424316406,
+      "eval_runtime": 139.9506,
+      "eval_samples_per_second": 34.605,
+      "eval_steps_per_second": 4.33,
+      "eval_wer": 1.0305989213738291,
+      "step": 1550
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.1468144044321327e-05,
+      "loss": 2.4828,
+      "step": 1551
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.1426592797783932e-05,
+      "loss": 2.4047,
+      "step": 1552
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 1.1385041551246536e-05,
+      "loss": 2.5188,
+      "step": 1553
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.1343490304709139e-05,
+      "loss": 2.5216,
+      "step": 1554
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.1301939058171744e-05,
+      "loss": 2.4596,
+      "step": 1555
+    },
+    {
+      "epoch": 4.26,
+      "eval_loss": 2.39601469039917,
+      "eval_runtime": 140.8025,
+      "eval_samples_per_second": 34.396,
+      "eval_steps_per_second": 4.304,
+      "eval_wer": 1.0774623900085154,
+      "step": 1555
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 1.1260387811634347e-05,
+      "loss": 2.4707,
+      "step": 1556
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.121883656509695e-05,
+      "loss": 2.5016,
+      "step": 1557
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.1177285318559555e-05,
+      "loss": 2.4901,
+      "step": 1558
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.1135734072022158e-05,
+      "loss": 2.5112,
+      "step": 1559
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 1.1094182825484763e-05,
+      "loss": 2.521,
+      "step": 1560
+    },
+    {
+      "epoch": 4.27,
+      "eval_loss": 2.3977975845336914,
+      "eval_runtime": 143.994,
+      "eval_samples_per_second": 33.633,
+      "eval_steps_per_second": 4.209,
+      "eval_wer": 1.1024978711325575,
+      "step": 1560
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.1052631578947366e-05,
+      "loss": 2.5309,
+      "step": 1561
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.1011080332409973e-05,
+      "loss": 2.6178,
+      "step": 1562
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.0969529085872576e-05,
+      "loss": 2.5363,
+      "step": 1563
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 1.092797783933518e-05,
+      "loss": 2.5819,
+      "step": 1564
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.0886426592797784e-05,
+      "loss": 2.6304,
+      "step": 1565
+    },
+    {
+      "epoch": 4.29,
+      "eval_loss": 2.3885209560394287,
+      "eval_runtime": 140.181,
+      "eval_samples_per_second": 34.548,
+      "eval_steps_per_second": 4.323,
+      "eval_wer": 1.0433153562304853,
+      "step": 1565
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.0844875346260388e-05,
+      "loss": 2.5473,
+      "step": 1566
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 1.0803324099722992e-05,
+      "loss": 2.5859,
+      "step": 1567
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.0761772853185596e-05,
+      "loss": 2.5521,
+      "step": 1568
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.0720221606648199e-05,
+      "loss": 2.5361,
+      "step": 1569
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.0678670360110804e-05,
+      "loss": 2.543,
+      "step": 1570
+    },
+    {
+      "epoch": 4.3,
+      "eval_loss": 2.38488507270813,
+      "eval_runtime": 140.222,
+      "eval_samples_per_second": 34.538,
+      "eval_steps_per_second": 4.322,
+      "eval_wer": 1.00723814930457,
+      "step": 1570
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 1.0637119113573407e-05,
+      "loss": 2.5263,
+      "step": 1571
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.059556786703601e-05,
+      "loss": 2.5785,
+      "step": 1572
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.0554016620498615e-05,
+      "loss": 2.5757,
+      "step": 1573
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.0512465373961218e-05,
+      "loss": 2.5426,
+      "step": 1574
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 1.0470914127423823e-05,
+      "loss": 2.5601,
+      "step": 1575
+    },
+    {
+      "epoch": 4.31,
+      "eval_loss": 2.385526180267334,
+      "eval_runtime": 146.2432,
+      "eval_samples_per_second": 33.116,
+      "eval_steps_per_second": 4.144,
+      "eval_wer": 1.0110133409026398,
+      "step": 1575
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.0429362880886426e-05,
+      "loss": 2.5142,
+      "step": 1576
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.038781163434903e-05,
+      "loss": 2.591,
+      "step": 1577
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 1.0346260387811635e-05,
+      "loss": 2.535,
+      "step": 1578
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.0304709141274238e-05,
+      "loss": 2.6025,
+      "step": 1579
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.0263157894736843e-05,
+      "loss": 2.6304,
+      "step": 1580
+    },
+    {
+      "epoch": 4.33,
+      "eval_loss": 2.38775634765625,
+      "eval_runtime": 140.41,
+      "eval_samples_per_second": 34.492,
+      "eval_steps_per_second": 4.316,
+      "eval_wer": 1.0368719841044565,
+      "step": 1580
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.0221606648199446e-05,
+      "loss": 2.528,
+      "step": 1581
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 1.0180055401662049e-05,
+      "loss": 2.5493,
+      "step": 1582
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.0138504155124654e-05,
+      "loss": 2.6068,
+      "step": 1583
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.0096952908587257e-05,
+      "loss": 2.5067,
+      "step": 1584
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.005540166204986e-05,
+      "loss": 2.4121,
+      "step": 1585
+    },
+    {
+      "epoch": 4.34,
+      "eval_loss": 2.378288745880127,
+      "eval_runtime": 140.9782,
+      "eval_samples_per_second": 34.353,
+      "eval_steps_per_second": 4.299,
+      "eval_wer": 1.0365881351121204,
+      "step": 1585
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 1.0013850415512465e-05,
+      "loss": 2.432,
+      "step": 1586
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 9.972299168975068e-06,
+      "loss": 2.5205,
+      "step": 1587
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 9.930747922437673e-06,
+      "loss": 2.5213,
+      "step": 1588
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 9.889196675900277e-06,
+      "loss": 2.5412,
+      "step": 1589
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.84764542936288e-06,
+      "loss": 2.4261,
+      "step": 1590
+    },
+    {
+      "epoch": 4.36,
+      "eval_loss": 2.374589204788208,
+      "eval_runtime": 144.6416,
+      "eval_samples_per_second": 33.483,
+      "eval_steps_per_second": 4.19,
+      "eval_wer": 1.0307408458699971,
+      "step": 1590
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.806094182825485e-06,
+      "loss": 2.5382,
+      "step": 1591
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.764542936288088e-06,
+      "loss": 2.5176,
+      "step": 1592
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.722991689750693e-06,
+      "loss": 2.5282,
+      "step": 1593
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.681440443213296e-06,
+      "loss": 2.5183,
+      "step": 1594
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.6398891966759e-06,
+      "loss": 2.5038,
+      "step": 1595
+    },
+    {
+      "epoch": 4.37,
+      "eval_loss": 2.378915548324585,
+      "eval_runtime": 153.6256,
+      "eval_samples_per_second": 31.525,
+      "eval_steps_per_second": 3.945,
+      "eval_wer": 1.061141072949191,
+      "step": 1595
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.598337950138504e-06,
+      "loss": 2.481,
+      "step": 1596
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.556786703601107e-06,
+      "loss": 2.5587,
+      "step": 1597
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 9.51523545706371e-06,
+      "loss": 2.4982,
+      "step": 1598
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 9.473684210526315e-06,
+      "loss": 2.4992,
+      "step": 1599
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 9.432132963988919e-06,
+      "loss": 2.5391,
+      "step": 1600
+    },
+    {
+      "epoch": 4.38,
+      "eval_loss": 2.384855270385742,
+      "eval_runtime": 142.7744,
+      "eval_samples_per_second": 33.921,
+      "eval_steps_per_second": 4.244,
+      "eval_wer": 1.0738291229066137,
+      "step": 1600
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.390581717451524e-06,
+      "loss": 2.587,
+      "step": 1601
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.349030470914127e-06,
+      "loss": 2.4398,
+      "step": 1602
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.30747922437673e-06,
+      "loss": 2.5555,
+      "step": 1603
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.265927977839335e-06,
+      "loss": 2.5544,
+      "step": 1604
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.224376731301938e-06,
+      "loss": 2.4341,
+      "step": 1605
+    },
+    {
+      "epoch": 4.4,
+      "eval_loss": 2.377945899963379,
+      "eval_runtime": 140.116,
+      "eval_samples_per_second": 34.564,
+      "eval_steps_per_second": 4.325,
+      "eval_wer": 1.057309111552654,
+      "step": 1605
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.182825484764543e-06,
+      "loss": 2.5333,
+      "step": 1606
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.141274238227146e-06,
+      "loss": 2.5191,
+      "step": 1607
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.09972299168975e-06,
+      "loss": 2.5297,
+      "step": 1608
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 9.058171745152354e-06,
+      "loss": 2.6354,
+      "step": 1609
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 9.016620498614958e-06,
+      "loss": 2.5306,
+      "step": 1610
+    },
+    {
+      "epoch": 4.41,
+      "eval_loss": 2.375149726867676,
+      "eval_runtime": 159.8052,
+      "eval_samples_per_second": 30.306,
+      "eval_steps_per_second": 3.792,
+      "eval_wer": 1.0459835367584445,
+      "step": 1610
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 8.97506925207756e-06,
+      "loss": 2.6055,
+      "step": 1611
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.933518005540166e-06,
+      "loss": 2.5649,
+      "step": 1612
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.891966759002769e-06,
+      "loss": 2.5777,
+      "step": 1613
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.850415512465374e-06,
+      "loss": 2.5772,
+      "step": 1614
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 8.808864265927977e-06,
+      "loss": 2.5818,
+      "step": 1615
+    },
+    {
+      "epoch": 4.42,
+      "eval_loss": 2.3743481636047363,
+      "eval_runtime": 140.0792,
+      "eval_samples_per_second": 34.573,
+      "eval_steps_per_second": 4.326,
+      "eval_wer": 1.0250922509225093,
+      "step": 1615
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 8.76731301939058e-06,
+      "loss": 2.5188,
+      "step": 1616
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 8.725761772853185e-06,
+      "loss": 2.5539,
+      "step": 1617
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 8.684210526315788e-06,
+      "loss": 2.552,
+      "step": 1618
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 8.642659279778393e-06,
+      "loss": 2.5555,
+      "step": 1619
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 8.601108033240996e-06,
+      "loss": 2.5531,
+      "step": 1620
+    },
+    {
+      "epoch": 4.44,
+      "eval_loss": 2.3723344802856445,
+      "eval_runtime": 140.8188,
+      "eval_samples_per_second": 34.392,
+      "eval_steps_per_second": 4.303,
+      "eval_wer": 1.0209196707351689,
+      "step": 1620
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 8.5595567867036e-06,
+      "loss": 2.5475,
+      "step": 1621
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 8.518005540166205e-06,
+      "loss": 2.6148,
+      "step": 1622
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.476454293628808e-06,
+      "loss": 2.6216,
+      "step": 1623
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.434903047091411e-06,
+      "loss": 2.4575,
+      "step": 1624
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.393351800554016e-06,
+      "loss": 2.51,
+      "step": 1625
+    },
+    {
+      "epoch": 4.45,
+      "eval_loss": 2.3754518032073975,
+      "eval_runtime": 146.2568,
+      "eval_samples_per_second": 33.113,
+      "eval_steps_per_second": 4.143,
+      "eval_wer": 1.031620777746239,
+      "step": 1625
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 8.351800554016619e-06,
+      "loss": 2.5257,
+      "step": 1626
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 8.310249307479224e-06,
+      "loss": 2.4693,
+      "step": 1627
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 8.268698060941827e-06,
+      "loss": 2.5622,
+      "step": 1628
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 8.22714681440443e-06,
+      "loss": 2.5541,
+      "step": 1629
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.185595567867035e-06,
+      "loss": 2.5788,
+      "step": 1630
+    },
+    {
+      "epoch": 4.47,
+      "eval_loss": 2.372516393661499,
+      "eval_runtime": 141.6738,
+      "eval_samples_per_second": 34.184,
+      "eval_steps_per_second": 4.277,
+      "eval_wer": 1.0396253193301164,
+      "step": 1630
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.144044321329639e-06,
+      "loss": 2.4269,
+      "step": 1631
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.102493074792243e-06,
+      "loss": 2.5667,
+      "step": 1632
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 8.060941828254847e-06,
+      "loss": 2.5306,
+      "step": 1633
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 8.01939058171745e-06,
+      "loss": 2.4822,
+      "step": 1634
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 7.977839335180055e-06,
+      "loss": 2.5701,
+      "step": 1635
+    },
+    {
+      "epoch": 4.48,
+      "eval_loss": 2.3663015365600586,
+      "eval_runtime": 146.6348,
+      "eval_samples_per_second": 33.028,
+      "eval_steps_per_second": 4.133,
+      "eval_wer": 1.029151291512915,
+      "step": 1635
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 7.936288088642658e-06,
+      "loss": 2.5131,
+      "step": 1636
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 7.894736842105261e-06,
+      "loss": 2.6295,
+      "step": 1637
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 7.853185595567866e-06,
+      "loss": 2.4992,
+      "step": 1638
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 7.81163434903047e-06,
+      "loss": 2.5239,
+      "step": 1639
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 7.770083102493074e-06,
+      "loss": 2.4194,
+      "step": 1640
+    },
+    {
+      "epoch": 4.49,
+      "eval_loss": 2.3641042709350586,
+      "eval_runtime": 141.8357,
+      "eval_samples_per_second": 34.145,
+      "eval_steps_per_second": 4.273,
+      "eval_wer": 1.0261424921941527,
+      "step": 1640
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 7.728531855955677e-06,
+      "loss": 2.5042,
+      "step": 1641
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 7.686980609418282e-06,
+      "loss": 2.5308,
+      "step": 1642
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 7.645429362880886e-06,
+      "loss": 2.481,
+      "step": 1643
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 7.6038781163434905e-06,
+      "loss": 2.4786,
+      "step": 1644
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 7.562326869806094e-06,
+      "loss": 2.5439,
+      "step": 1645
+    },
+    {
+      "epoch": 4.51,
+      "eval_loss": 2.362913131713867,
+      "eval_runtime": 157.9107,
+      "eval_samples_per_second": 30.669,
+      "eval_steps_per_second": 3.838,
+      "eval_wer": 1.0376383763837638,
+      "step": 1645
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 7.520775623268698e-06,
+      "loss": 2.4422,
+      "step": 1646
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 7.479224376731302e-06,
+      "loss": 2.5159,
+      "step": 1647
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 7.437673130193906e-06,
+      "loss": 2.4263,
+      "step": 1648
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 7.396121883656509e-06,
+      "loss": 2.5145,
+      "step": 1649
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 7.354570637119113e-06,
+      "loss": 2.4527,
+      "step": 1650
+    },
+    {
+      "epoch": 4.52,
+      "eval_loss": 2.3628642559051514,
+      "eval_runtime": 161.5128,
+      "eval_samples_per_second": 29.985,
+      "eval_steps_per_second": 3.752,
+      "eval_wer": 1.0562588702810105,
+      "step": 1650
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 7.313019390581717e-06,
+      "loss": 2.4551,
+      "step": 1651
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 7.271468144044321e-06,
+      "loss": 2.4952,
+      "step": 1652
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 7.2299168975069244e-06,
+      "loss": 2.5051,
+      "step": 1653
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 7.1883656509695285e-06,
+      "loss": 2.5854,
+      "step": 1654
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 7.1468144044321326e-06,
+      "loss": 2.5705,
+      "step": 1655
+    },
+    {
+      "epoch": 4.53,
+      "eval_loss": 2.365365743637085,
+      "eval_runtime": 159.5208,
+      "eval_samples_per_second": 30.36,
+      "eval_steps_per_second": 3.799,
+      "eval_wer": 1.07655407323304,
+      "step": 1655
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 7.105263157894737e-06,
+      "loss": 2.5623,
+      "step": 1656
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 7.063711911357341e-06,
+      "loss": 2.5451,
+      "step": 1657
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 7.022160664819944e-06,
+      "loss": 2.515,
+      "step": 1658
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 6.980609418282548e-06,
+      "loss": 2.5289,
+      "step": 1659
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 6.939058171745152e-06,
+      "loss": 2.4552,
+      "step": 1660
+    },
+    {
+      "epoch": 4.55,
+      "eval_loss": 2.370793342590332,
+      "eval_runtime": 142.4091,
+      "eval_samples_per_second": 34.008,
+      "eval_steps_per_second": 4.255,
+      "eval_wer": 1.0801589554357083,
+      "step": 1660
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 6.897506925207756e-06,
+      "loss": 2.5884,
+      "step": 1661
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 6.855955678670359e-06,
+      "loss": 2.5812,
+      "step": 1662
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 6.814404432132963e-06,
+      "loss": 2.5576,
+      "step": 1663
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 6.772853185595567e-06,
+      "loss": 2.5552,
+      "step": 1664
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 6.7313019390581714e-06,
+      "loss": 2.5657,
+      "step": 1665
+    },
+    {
+      "epoch": 4.56,
+      "eval_loss": 2.3637919425964355,
+      "eval_runtime": 140.9418,
+      "eval_samples_per_second": 34.362,
+      "eval_steps_per_second": 4.3,
+      "eval_wer": 1.0247516321317058,
+      "step": 1665
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 6.689750692520775e-06,
+      "loss": 2.4985,
+      "step": 1666
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 6.648199445983379e-06,
+      "loss": 2.5127,
+      "step": 1667
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 6.606648199445983e-06,
+      "loss": 2.5258,
+      "step": 1668
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 6.565096952908587e-06,
+      "loss": 2.5434,
+      "step": 1669
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 6.523545706371191e-06,
+      "loss": 2.5371,
+      "step": 1670
+    },
+    {
+      "epoch": 4.57,
+      "eval_loss": 2.363882303237915,
+      "eval_runtime": 143.5985,
+      "eval_samples_per_second": 33.726,
+      "eval_steps_per_second": 4.22,
+      "eval_wer": 1.005279591257451,
+      "step": 1670
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 6.481994459833794e-06,
+      "loss": 2.4981,
+      "step": 1671
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 6.440443213296398e-06,
+      "loss": 2.5191,
+      "step": 1672
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 6.398891966759002e-06,
+      "loss": 2.598,
+      "step": 1673
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 6.357340720221606e-06,
+      "loss": 2.5796,
+      "step": 1674
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 6.3157894736842095e-06,
+      "loss": 2.5365,
+      "step": 1675
+    },
+    {
+      "epoch": 4.59,
+      "eval_loss": 2.362563371658325,
+      "eval_runtime": 160.6129,
+      "eval_samples_per_second": 30.153,
+      "eval_steps_per_second": 3.773,
+      "eval_wer": 1.0071529946068691,
+      "step": 1675
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 6.2742382271468135e-06,
+      "loss": 2.5202,
+      "step": 1676
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 6.232686980609418e-06,
+      "loss": 2.528,
+      "step": 1677
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 6.191135734072022e-06,
+      "loss": 2.5311,
+      "step": 1678
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 6.149584487534625e-06,
+      "loss": 2.5319,
+      "step": 1679
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 6.108033240997229e-06,
+      "loss": 2.5383,
+      "step": 1680
+    },
+    {
+      "epoch": 4.6,
+      "eval_loss": 2.3584203720092773,
+      "eval_runtime": 141.9293,
+      "eval_samples_per_second": 34.123,
+      "eval_steps_per_second": 4.27,
+      "eval_wer": 1.0169741697416974,
+      "step": 1680
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 6.066481994459833e-06,
+      "loss": 2.5571,
+      "step": 1681
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 6.024930747922437e-06,
+      "loss": 2.5372,
+      "step": 1682
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.983379501385041e-06,
+      "loss": 2.5638,
+      "step": 1683
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 5.941828254847644e-06,
+      "loss": 2.5316,
+      "step": 1684
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.900277008310248e-06,
+      "loss": 2.546,
+      "step": 1685
+    },
+    {
+      "epoch": 4.62,
+      "eval_loss": 2.3573923110961914,
+      "eval_runtime": 143.3676,
+      "eval_samples_per_second": 33.78,
+      "eval_steps_per_second": 4.227,
+      "eval_wer": 1.0468634686346863,
+      "step": 1685
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.8587257617728524e-06,
+      "loss": 2.5585,
+      "step": 1686
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.8171745152354565e-06,
+      "loss": 2.5237,
+      "step": 1687
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 5.77562326869806e-06,
+      "loss": 2.5222,
+      "step": 1688
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.734072022160664e-06,
+      "loss": 2.45,
+      "step": 1689
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.692520775623268e-06,
+      "loss": 2.6006,
+      "step": 1690
+    },
+    {
+      "epoch": 4.63,
+      "eval_loss": 2.351736307144165,
+      "eval_runtime": 142.9547,
+      "eval_samples_per_second": 33.878,
+      "eval_steps_per_second": 4.239,
+      "eval_wer": 1.0508941243258587,
+      "step": 1690
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.650969529085872e-06,
+      "loss": 2.4509,
+      "step": 1691
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 5.609418282548475e-06,
+      "loss": 2.4681,
+      "step": 1692
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.567867036011079e-06,
+      "loss": 2.5073,
+      "step": 1693
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.526315789473683e-06,
+      "loss": 2.4741,
+      "step": 1694
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 5.484764542936288e-06,
+      "loss": 2.4894,
+      "step": 1695
+    },
+    {
+      "epoch": 4.64,
+      "eval_loss": 2.3488974571228027,
+      "eval_runtime": 143.0199,
+      "eval_samples_per_second": 33.862,
+      "eval_steps_per_second": 4.237,
+      "eval_wer": 1.0451887595799034,
+      "step": 1695
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.443213296398892e-06,
+      "loss": 2.5711,
+      "step": 1696
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.401662049861496e-06,
+      "loss": 2.4544,
+      "step": 1697
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.360110803324099e-06,
+      "loss": 2.5378,
+      "step": 1698
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 5.3185595567867035e-06,
+      "loss": 2.4544,
+      "step": 1699
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.2770083102493075e-06,
+      "loss": 2.4732,
+      "step": 1700
+    },
+    {
+      "epoch": 4.66,
+      "eval_loss": 2.348914861679077,
+      "eval_runtime": 153.5563,
+      "eval_samples_per_second": 31.539,
+      "eval_steps_per_second": 3.946,
+      "eval_wer": 1.0585864320181664,
+      "step": 1700
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.235457063711912e-06,
+      "loss": 2.4552,
+      "step": 1701
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.193905817174515e-06,
+      "loss": 2.4363,
+      "step": 1702
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 5.152354570637119e-06,
+      "loss": 2.4902,
+      "step": 1703
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.110803324099723e-06,
+      "loss": 2.5714,
+      "step": 1704
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.069252077562327e-06,
+      "loss": 2.4933,
+      "step": 1705
+    },
+    {
+      "epoch": 4.67,
+      "eval_loss": 2.3500583171844482,
+      "eval_runtime": 142.338,
+      "eval_samples_per_second": 34.025,
+      "eval_steps_per_second": 4.257,
+      "eval_wer": 1.069401078626171,
+      "step": 1705
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 5.02770083102493e-06,
+      "loss": 2.4341,
+      "step": 1706
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.986149584487534e-06,
+      "loss": 2.47,
+      "step": 1707
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.944598337950138e-06,
+      "loss": 2.4823,
+      "step": 1708
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.903047091412742e-06,
+      "loss": 2.4995,
+      "step": 1709
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 4.861495844875346e-06,
+      "loss": 2.4784,
+      "step": 1710
+    },
+    {
+      "epoch": 4.68,
+      "eval_loss": 2.347200632095337,
+      "eval_runtime": 140.0835,
+      "eval_samples_per_second": 34.572,
+      "eval_steps_per_second": 4.326,
+      "eval_wer": 1.0647175702526257,
+      "step": 1710
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.81994459833795e-06,
+      "loss": 2.5469,
+      "step": 1711
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.778393351800554e-06,
+      "loss": 2.5781,
+      "step": 1712
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.736842105263158e-06,
+      "loss": 2.5392,
+      "step": 1713
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 4.695290858725762e-06,
+      "loss": 2.5729,
+      "step": 1714
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.653739612188365e-06,
+      "loss": 2.5349,
+      "step": 1715
+    },
+    {
+      "epoch": 4.7,
+      "eval_loss": 2.3419370651245117,
+      "eval_runtime": 145.3075,
+      "eval_samples_per_second": 33.329,
+      "eval_steps_per_second": 4.17,
+      "eval_wer": 1.0299176837922226,
+      "step": 1715
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.612188365650969e-06,
+      "loss": 2.5522,
+      "step": 1716
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 4.570637119113573e-06,
+      "loss": 2.5657,
+      "step": 1717
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.529085872576177e-06,
+      "loss": 2.5509,
+      "step": 1718
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.48753462603878e-06,
+      "loss": 2.5198,
+      "step": 1719
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.4459833795013845e-06,
+      "loss": 2.553,
+      "step": 1720
+    },
+    {
+      "epoch": 4.71,
+      "eval_loss": 2.3420019149780273,
+      "eval_runtime": 143.8669,
+      "eval_samples_per_second": 33.663,
+      "eval_steps_per_second": 4.212,
+      "eval_wer": 1.0114674992903776,
+      "step": 1720
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 4.4044321329639885e-06,
+      "loss": 2.4339,
+      "step": 1721
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.3628808864265926e-06,
+      "loss": 2.5382,
+      "step": 1722
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.321329639889197e-06,
+      "loss": 2.5401,
+      "step": 1723
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 4.2797783933518e-06,
+      "loss": 2.5485,
+      "step": 1724
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.238227146814404e-06,
+      "loss": 2.5035,
+      "step": 1725
+    },
+    {
+      "epoch": 4.73,
+      "eval_loss": 2.3414924144744873,
+      "eval_runtime": 150.386,
+      "eval_samples_per_second": 32.204,
+      "eval_steps_per_second": 4.03,
+      "eval_wer": 1.0116945784842464,
+      "step": 1725
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.196675900277008e-06,
+      "loss": 2.5424,
+      "step": 1726
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.155124653739612e-06,
+      "loss": 2.547,
+      "step": 1727
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 4.113573407202215e-06,
+      "loss": 2.5482,
+      "step": 1728
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 4.072022160664819e-06,
+      "loss": 2.5535,
+      "step": 1729
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 4.030470914127423e-06,
+      "loss": 2.561,
+      "step": 1730
+    },
+    {
+      "epoch": 4.74,
+      "eval_loss": 2.34179949760437,
+      "eval_runtime": 147.2309,
+      "eval_samples_per_second": 32.894,
+      "eval_steps_per_second": 4.116,
+      "eval_wer": 1.0241555492478003,
+      "step": 1730
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.988919667590027e-06,
+      "loss": 2.5441,
+      "step": 1731
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 3.947368421052631e-06,
+      "loss": 2.5058,
+      "step": 1732
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.905817174515235e-06,
+      "loss": 2.4681,
+      "step": 1733
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.864265927977839e-06,
+      "loss": 2.5119,
+      "step": 1734
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 3.822714681440443e-06,
+      "loss": 2.4773,
+      "step": 1735
+    },
+    {
+      "epoch": 4.75,
+      "eval_loss": 2.3419573307037354,
+      "eval_runtime": 141.8367,
+      "eval_samples_per_second": 34.145,
+      "eval_steps_per_second": 4.273,
+      "eval_wer": 1.0325007096224808,
+      "step": 1735
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.781163434903047e-06,
+      "loss": 2.4355,
+      "step": 1736
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.739612188365651e-06,
+      "loss": 2.5295,
+      "step": 1737
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.6980609418282545e-06,
+      "loss": 2.4749,
+      "step": 1738
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 3.6565096952908586e-06,
+      "loss": 2.44,
+      "step": 1739
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.6149584487534622e-06,
+      "loss": 2.4691,
+      "step": 1740
+    },
+    {
+      "epoch": 4.77,
+      "eval_loss": 2.342160701751709,
+      "eval_runtime": 140.6734,
+      "eval_samples_per_second": 34.427,
+      "eval_steps_per_second": 4.308,
+      "eval_wer": 1.039426625035481,
+      "step": 1740
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.5734072022160663e-06,
+      "loss": 2.511,
+      "step": 1741
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.5318559556786703e-06,
+      "loss": 2.4891,
+      "step": 1742
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 3.490304709141274e-06,
+      "loss": 2.4933,
+      "step": 1743
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.448753462603878e-06,
+      "loss": 2.552,
+      "step": 1744
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.4072022160664817e-06,
+      "loss": 2.4959,
+      "step": 1745
+    },
+    {
+      "epoch": 4.78,
+      "eval_loss": 2.3405394554138184,
+      "eval_runtime": 140.8425,
+      "eval_samples_per_second": 34.386,
+      "eval_steps_per_second": 4.303,
+      "eval_wer": 1.041754186772637,
+      "step": 1745
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 3.3656509695290857e-06,
+      "loss": 2.4645,
+      "step": 1746
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.3240997229916894e-06,
+      "loss": 2.4843,
+      "step": 1747
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.2825484764542934e-06,
+      "loss": 2.4717,
+      "step": 1748
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.240997229916897e-06,
+      "loss": 2.4301,
+      "step": 1749
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 3.199445983379501e-06,
+      "loss": 2.4928,
+      "step": 1750
+    },
+    {
+      "epoch": 4.79,
+      "eval_loss": 2.3394219875335693,
+      "eval_runtime": 141.3184,
+      "eval_samples_per_second": 34.27,
+      "eval_steps_per_second": 4.288,
+      "eval_wer": 1.0449049105875674,
+      "step": 1750
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.1578947368421047e-06,
+      "loss": 2.4662,
+      "step": 1751
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.116343490304709e-06,
+      "loss": 2.5861,
+      "step": 1752
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.0747922437673124e-06,
+      "loss": 2.4717,
+      "step": 1753
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 3.0332409972299165e-06,
+      "loss": 2.4837,
+      "step": 1754
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.9916897506925205e-06,
+      "loss": 2.5058,
+      "step": 1755
+    },
+    {
+      "epoch": 4.81,
+      "eval_loss": 2.3391995429992676,
+      "eval_runtime": 139.9027,
+      "eval_samples_per_second": 34.617,
+      "eval_steps_per_second": 4.332,
+      "eval_wer": 1.0489071813795061,
+      "step": 1755
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.950138504155124e-06,
+      "loss": 2.436,
+      "step": 1756
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 2.9085872576177282e-06,
+      "loss": 2.4739,
+      "step": 1757
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.867036011080332e-06,
+      "loss": 2.4543,
+      "step": 1758
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.825484764542936e-06,
+      "loss": 2.6843,
+      "step": 1759
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.7839335180055396e-06,
+      "loss": 2.5193,
+      "step": 1760
+    },
+    {
+      "epoch": 4.82,
+      "eval_loss": 2.3390493392944336,
+      "eval_runtime": 151.2803,
+      "eval_samples_per_second": 32.013,
+      "eval_steps_per_second": 4.006,
+      "eval_wer": 1.0506102753335225,
+      "step": 1760
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 2.742382271468144e-06,
+      "loss": 2.5566,
+      "step": 1761
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.700831024930748e-06,
+      "loss": 2.5617,
+      "step": 1762
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.6592797783933517e-06,
+      "loss": 2.5593,
+      "step": 1763
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.617728531855956e-06,
+      "loss": 2.5633,
+      "step": 1764
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 2.5761772853185594e-06,
+      "loss": 2.5369,
+      "step": 1765
+    },
+    {
+      "epoch": 4.83,
+      "eval_loss": 2.339158058166504,
+      "eval_runtime": 155.5013,
+      "eval_samples_per_second": 31.144,
+      "eval_steps_per_second": 3.897,
+      "eval_wer": 1.0383763837638376,
+      "step": 1765
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.5346260387811635e-06,
+      "loss": 2.521,
+      "step": 1766
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.493074792243767e-06,
+      "loss": 2.5764,
+      "step": 1767
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 2.451523545706371e-06,
+      "loss": 2.5323,
+      "step": 1768
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.409972299168975e-06,
+      "loss": 2.5517,
+      "step": 1769
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.368421052631579e-06,
+      "loss": 2.4843,
+      "step": 1770
+    },
+    {
+      "epoch": 4.85,
+      "eval_loss": 2.3397862911224365,
+      "eval_runtime": 140.1076,
+      "eval_samples_per_second": 34.566,
+      "eval_steps_per_second": 4.325,
+      "eval_wer": 1.023587851263128,
+      "step": 1770
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.3268698060941825e-06,
+      "loss": 2.5745,
+      "step": 1771
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 2.2853185595567866e-06,
+      "loss": 2.5806,
+      "step": 1772
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.24376731301939e-06,
+      "loss": 2.5602,
+      "step": 1773
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.2022160664819943e-06,
+      "loss": 2.5206,
+      "step": 1774
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.1606648199445983e-06,
+      "loss": 2.5074,
+      "step": 1775
+    },
+    {
+      "epoch": 4.86,
+      "eval_loss": 2.3399529457092285,
+      "eval_runtime": 151.4274,
+      "eval_samples_per_second": 31.982,
+      "eval_steps_per_second": 4.002,
+      "eval_wer": 1.0149588418961113,
+      "step": 1775
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 2.119113573407202e-06,
+      "loss": 2.4752,
+      "step": 1776
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.077562326869806e-06,
+      "loss": 2.5011,
+      "step": 1777
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 2.0360110803324096e-06,
+      "loss": 2.525,
+      "step": 1778
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 1.9944598337950137e-06,
+      "loss": 2.5246,
+      "step": 1779
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.9529085872576173e-06,
+      "loss": 2.4941,
+      "step": 1780
+    },
+    {
+      "epoch": 4.88,
+      "eval_loss": 2.3385510444641113,
+      "eval_runtime": 145.0293,
+      "eval_samples_per_second": 33.393,
+      "eval_steps_per_second": 4.178,
+      "eval_wer": 1.0149872267953448,
+      "step": 1780
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.9113573407202214e-06,
+      "loss": 2.4665,
+      "step": 1781
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.8698060941828254e-06,
+      "loss": 2.5632,
+      "step": 1782
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 1.8282548476454293e-06,
+      "loss": 2.473,
+      "step": 1783
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.7867036011080331e-06,
+      "loss": 2.4518,
+      "step": 1784
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.745152354570637e-06,
+      "loss": 2.4352,
+      "step": 1785
+    },
+    {
+      "epoch": 4.89,
+      "eval_loss": 2.3370206356048584,
+      "eval_runtime": 141.7712,
+      "eval_samples_per_second": 34.161,
+      "eval_steps_per_second": 4.274,
+      "eval_wer": 1.0172296338348,
+      "step": 1785
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.7036011080332408e-06,
+      "loss": 2.4735,
+      "step": 1786
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 1.6620498614958447e-06,
+      "loss": 2.483,
+      "step": 1787
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.6204986149584485e-06,
+      "loss": 2.4563,
+      "step": 1788
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5789473684210524e-06,
+      "loss": 2.509,
+      "step": 1789
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 1.5373961218836562e-06,
+      "loss": 2.4372,
+      "step": 1790
+    },
+    {
+      "epoch": 4.9,
+      "eval_loss": 2.336162567138672,
+      "eval_runtime": 140.7255,
+      "eval_samples_per_second": 34.415,
+      "eval_steps_per_second": 4.306,
+      "eval_wer": 1.0208061311382344,
+      "step": 1790
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.4958448753462603e-06,
+      "loss": 2.4871,
+      "step": 1791
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.4542936288088641e-06,
+      "loss": 2.4704,
+      "step": 1792
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.412742382271468e-06,
+      "loss": 2.4653,
+      "step": 1793
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 1.371191135734072e-06,
+      "loss": 2.4765,
+      "step": 1794
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.3296398891966759e-06,
+      "loss": 2.4855,
+      "step": 1795
+    },
+    {
+      "epoch": 4.92,
+      "eval_loss": 2.3357536792755127,
+      "eval_runtime": 147.1147,
+      "eval_samples_per_second": 32.92,
+      "eval_steps_per_second": 4.119,
+      "eval_wer": 1.0238149304569968,
+      "step": 1795
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.2880886426592797e-06,
+      "loss": 2.4723,
+      "step": 1796
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.2465373961218836e-06,
+      "loss": 2.4209,
+      "step": 1797
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 1.2049861495844874e-06,
+      "loss": 2.431,
+      "step": 1798
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.1634349030470913e-06,
+      "loss": 2.4744,
+      "step": 1799
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.121883656509695e-06,
+      "loss": 2.4516,
+      "step": 1800
+    },
+    {
+      "epoch": 4.93,
+      "eval_loss": 2.335530996322632,
+      "eval_runtime": 144.2723,
+      "eval_samples_per_second": 33.568,
+      "eval_steps_per_second": 4.2,
+      "eval_wer": 1.0276185069543002,
+      "step": 1800
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 1.0803324099722992e-06,
+      "loss": 2.4579,
+      "step": 1801
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 1.038781163434903e-06,
+      "loss": 2.4824,
+      "step": 1802
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 9.972299168975068e-07,
+      "loss": 2.5019,
+      "step": 1803
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 9.556786703601107e-07,
+      "loss": 2.4765,
+      "step": 1804
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 9.141274238227146e-07,
+      "loss": 2.5281,
+      "step": 1805
+    },
+    {
+      "epoch": 4.94,
+      "eval_loss": 2.3355886936187744,
+      "eval_runtime": 139.944,
+      "eval_samples_per_second": 34.607,
+      "eval_steps_per_second": 4.33,
+      "eval_wer": 1.0312233891569684,
+      "step": 1805
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 8.725761772853185e-07,
+      "loss": 2.5515,
+      "step": 1806
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 8.310249307479223e-07,
+      "loss": 2.5382,
+      "step": 1807
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 7.894736842105262e-07,
+      "loss": 2.4093,
+      "step": 1808
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 7.479224376731301e-07,
+      "loss": 2.5102,
+      "step": 1809
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 7.06371191135734e-07,
+      "loss": 2.5519,
+      "step": 1810
+    },
+    {
+      "epoch": 4.96,
+      "eval_loss": 2.335160255432129,
+      "eval_runtime": 142.6698,
+      "eval_samples_per_second": 33.946,
+      "eval_steps_per_second": 4.248,
+      "eval_wer": 1.0318194720408742,
+      "step": 1810
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 6.648199445983379e-07,
+      "loss": 2.6029,
+      "step": 1811
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 6.232686980609418e-07,
+      "loss": 2.5347,
+      "step": 1812
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.817174515235456e-07,
+      "loss": 2.4606,
+      "step": 1813
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 5.401662049861496e-07,
+      "loss": 2.4955,
+      "step": 1814
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.986149584487534e-07,
+      "loss": 2.4641,
+      "step": 1815
+    },
+    {
+      "epoch": 4.97,
+      "eval_loss": 2.3349101543426514,
+      "eval_runtime": 152.5469,
+      "eval_samples_per_second": 31.748,
+      "eval_steps_per_second": 3.973,
+      "eval_wer": 1.0294067556060176,
+      "step": 1815
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 4.570637119113573e-07,
+      "loss": 2.5616,
+      "step": 1816
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 4.1551246537396117e-07,
+      "loss": 2.4761,
+      "step": 1817
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.7396121883656507e-07,
+      "loss": 2.5163,
+      "step": 1818
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 3.3240997229916897e-07,
+      "loss": 2.4638,
+      "step": 1819
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 2.908587257617728e-07,
+      "loss": 2.4515,
+      "step": 1820
+    },
+    {
+      "epoch": 4.98,
+      "eval_loss": 2.3347949981689453,
+      "eval_runtime": 141.2141,
+      "eval_samples_per_second": 34.295,
+      "eval_steps_per_second": 4.291,
+      "eval_wer": 1.028356514334374,
+      "step": 1820
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.493074792243767e-07,
+      "loss": 2.4388,
+      "step": 1821
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 2.0775623268698058e-07,
+      "loss": 2.4616,
+      "step": 1822
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 1.6620498614958448e-07,
+      "loss": 2.5311,
+      "step": 1823
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 1.2465373961218836e-07,
+      "loss": 2.4381,
+      "step": 1824
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 8.310249307479224e-08,
+      "loss": 2.553,
+      "step": 1825
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 2.3347036838531494,
+      "eval_runtime": 140.7491,
+      "eval_samples_per_second": 34.409,
+      "eval_steps_per_second": 4.306,
+      "eval_wer": 1.028583593528243,
+      "step": 1825
+    },
+    {
+      "epoch": 5.0,
+      "step": 1825,
+      "total_flos": 5.433166118134149e+18,
+      "train_loss": 2.8323297275909005,
+      "train_runtime": 64225.9806,
+      "train_samples_per_second": 0.91,
+      "train_steps_per_second": 0.028
+    }
+  ],
+  "max_steps": 1825,
+  "num_train_epochs": 5,
+  "total_flos": 5.433166118134149e+18,
+  "trial_name": null,
+  "trial_params": null
+}