diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,14368 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 29.885057471264368,
+  "global_step": 20800,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8e-06,
+      "loss": 24.5479,
+      "step": 10
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0799999999999998e-05,
+      "loss": 24.3336,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.68e-05,
+      "loss": 23.7501,
+      "step": 30
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2.28e-05,
+      "loss": 22.2394,
+      "step": 40
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2.88e-05,
+      "loss": 22.1365,
+      "step": 50
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 3.42e-05,
+      "loss": 20.337,
+      "step": 60
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 4.02e-05,
+      "loss": 19.8122,
+      "step": 70
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 4.62e-05,
+      "loss": 19.0623,
+      "step": 80
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 5.2199999999999995e-05,
+      "loss": 16.9603,
+      "step": 90
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 5.82e-05,
+      "loss": 15.8856,
+      "step": 100
+    },
+    {
+      "epoch": 0.14,
+      "eval_loss": 15.38314151763916,
+      "eval_runtime": 39.1757,
+      "eval_samples_per_second": 11.665,
+      "eval_steps_per_second": 1.481,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.419999999999999e-05,
+      "loss": 14.1248,
+      "step": 110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 7.02e-05,
+      "loss": 12.634,
+      "step": 120
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 7.62e-05,
+      "loss": 10.9827,
+      "step": 130
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.22e-05,
+      "loss": 9.6156,
+      "step": 140
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 8.819999999999999e-05,
+      "loss": 8.3307,
+      "step": 150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.419999999999999e-05,
+      "loss": 7.1401,
+      "step": 160
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.0001002,
+      "loss": 6.2121,
+      "step": 170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 0.00010619999999999998,
+      "loss": 5.5431,
+      "step": 180
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00011219999999999999,
+      "loss": 5.1547,
+      "step": 190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.0001182,
+      "loss": 4.9694,
+      "step": 200
+    },
+    {
+      "epoch": 0.29,
+      "eval_loss": 4.939667701721191,
+      "eval_runtime": 39.3067,
+      "eval_samples_per_second": 11.627,
+      "eval_steps_per_second": 1.476,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 0.00012419999999999998,
+      "loss": 4.8874,
+      "step": 210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0001302,
+      "loss": 4.8633,
+      "step": 220
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.0001362,
+      "loss": 4.8666,
+      "step": 230
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.0001422,
+      "loss": 4.8345,
+      "step": 240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001482,
+      "loss": 4.8414,
+      "step": 250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00015419999999999998,
+      "loss": 4.8116,
+      "step": 260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.0001602,
+      "loss": 4.7888,
+      "step": 270
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.0001662,
+      "loss": 4.7886,
+      "step": 280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00017219999999999998,
+      "loss": 4.7816,
+      "step": 290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00017819999999999997,
+      "loss": 4.8183,
+      "step": 300
+    },
+    {
+      "epoch": 0.43,
+      "eval_loss": 4.77665901184082,
+      "eval_runtime": 39.6162,
+      "eval_samples_per_second": 11.536,
+      "eval_steps_per_second": 1.464,
+      "eval_wer": 1.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00018419999999999998,
+      "loss": 4.7999,
+      "step": 310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.0001902,
+      "loss": 4.7751,
+      "step": 320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.0001962,
+      "loss": 4.7608,
+      "step": 330
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0002022,
+      "loss": 4.743,
+      "step": 340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00020819999999999996,
+      "loss": 4.7291,
+      "step": 350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.00021419999999999998,
+      "loss": 4.7118,
+      "step": 360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00022019999999999999,
+      "loss": 4.7286,
+      "step": 370
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.00022619999999999997,
+      "loss": 4.715,
+      "step": 380
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.00023219999999999998,
+      "loss": 4.709,
+      "step": 390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.0002382,
+      "loss": 4.683,
+      "step": 400
+    },
+    {
+      "epoch": 0.57,
+      "eval_loss": 4.713740348815918,
+      "eval_runtime": 39.3947,
+      "eval_samples_per_second": 11.601,
+      "eval_steps_per_second": 1.472,
+      "eval_wer": 1.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00024419999999999997,
+      "loss": 4.7096,
+      "step": 410
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.00025019999999999996,
+      "loss": 4.7024,
+      "step": 420
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002562,
+      "loss": 4.7074,
+      "step": 430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0002622,
+      "loss": 4.7047,
+      "step": 440
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00026819999999999996,
+      "loss": 4.6906,
+      "step": 450
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002742,
+      "loss": 4.6973,
+      "step": 460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.0002802,
+      "loss": 4.6773,
+      "step": 470
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00028619999999999996,
+      "loss": 4.6905,
+      "step": 480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00029219999999999995,
+      "loss": 4.6664,
+      "step": 490
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.0002982,
+      "loss": 4.6742,
+      "step": 500
+    },
+    {
+      "epoch": 0.72,
+      "eval_loss": 4.695952415466309,
+      "eval_runtime": 39.4463,
+      "eval_samples_per_second": 11.585,
+      "eval_steps_per_second": 1.47,
+      "eval_wer": 1.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002998969578017664,
+      "loss": 4.7007,
+      "step": 510
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002997497546614327,
+      "loss": 4.6779,
+      "step": 520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002996025515210991,
+      "loss": 4.6749,
+      "step": 530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00029945534838076544,
+      "loss": 4.6655,
+      "step": 540
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002993081452404318,
+      "loss": 4.6543,
+      "step": 550
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002991609421000981,
+      "loss": 4.6716,
+      "step": 560
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00029901373895976446,
+      "loss": 4.6725,
+      "step": 570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.00029886653581943077,
+      "loss": 4.6846,
+      "step": 580
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00029871933267909713,
+      "loss": 4.6431,
+      "step": 590
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00029857212953876344,
+      "loss": 4.6485,
+      "step": 600
+    },
+    {
+      "epoch": 0.86,
+      "eval_loss": 4.668909072875977,
+      "eval_runtime": 39.3362,
+      "eval_samples_per_second": 11.618,
+      "eval_steps_per_second": 1.474,
+      "eval_wer": 1.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002984249263984298,
+      "loss": 4.6752,
+      "step": 610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00029827772325809616,
+      "loss": 4.6491,
+      "step": 620
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.00029813052011776246,
+      "loss": 4.6515,
+      "step": 630
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002979833169774288,
+      "loss": 4.6491,
+      "step": 640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.0002978361138370952,
+      "loss": 4.6027,
+      "step": 650
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002976889106967615,
+      "loss": 4.6736,
+      "step": 660
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029754170755642785,
+      "loss": 4.6461,
+      "step": 670
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029739450441609416,
+      "loss": 4.6184,
+      "step": 680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002972473012757605,
+      "loss": 4.6395,
+      "step": 690
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002971000981354269,
+      "loss": 4.6174,
+      "step": 700
+    },
+    {
+      "epoch": 1.01,
+      "eval_loss": 4.663065433502197,
+      "eval_runtime": 39.4396,
+      "eval_samples_per_second": 11.587,
+      "eval_steps_per_second": 1.471,
+      "eval_wer": 0.9998494202680319,
+      "step": 700
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002969528949950932,
+      "loss": 4.6355,
+      "step": 710
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029680569185475955,
+      "loss": 4.611,
+      "step": 720
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002966584887144259,
+      "loss": 4.6402,
+      "step": 730
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.0002965112855740922,
+      "loss": 4.6334,
+      "step": 740
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002963640824337586,
+      "loss": 4.6247,
+      "step": 750
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.0002962168792934249,
+      "loss": 4.6173,
+      "step": 760
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00029606967615309124,
+      "loss": 4.5934,
+      "step": 770
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00029592247301275755,
+      "loss": 4.6392,
+      "step": 780
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.0002957752698724239,
+      "loss": 4.5958,
+      "step": 790
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00029562806673209027,
+      "loss": 4.5931,
+      "step": 800
+    },
+    {
+      "epoch": 1.15,
+      "eval_loss": 4.590127944946289,
+      "eval_runtime": 39.464,
+      "eval_samples_per_second": 11.58,
+      "eval_steps_per_second": 1.47,
+      "eval_wer": 0.9968378256286704,
+      "step": 800
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029548086359175663,
+      "loss": 4.5797,
+      "step": 810
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00029533366045142293,
+      "loss": 4.5442,
+      "step": 820
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0002951864573110893,
+      "loss": 4.5553,
+      "step": 830
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002950392541707556,
+      "loss": 4.5279,
+      "step": 840
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00029489205103042196,
+      "loss": 4.5188,
+      "step": 850
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00029474484789008827,
+      "loss": 4.5404,
+      "step": 860
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00029459764474975463,
+      "loss": 4.5379,
+      "step": 870
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.000294450441609421,
+      "loss": 4.4347,
+      "step": 880
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002943032384690873,
+      "loss": 4.4093,
+      "step": 890
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.00029415603532875366,
+      "loss": 4.3427,
+      "step": 900
+    },
+    {
+      "epoch": 1.29,
+      "eval_loss": 4.386187553405762,
+      "eval_runtime": 39.5926,
+      "eval_samples_per_second": 11.543,
+      "eval_steps_per_second": 1.465,
+      "eval_wer": 1.0,
+      "step": 900
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.00029400883218842,
+      "loss": 4.28,
+      "step": 910
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.0002938616290480863,
+      "loss": 4.1754,
+      "step": 920
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.0002937144259077527,
+      "loss": 4.0524,
+      "step": 930
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.000293567222767419,
+      "loss": 3.8665,
+      "step": 940
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00029342001962708535,
+      "loss": 3.786,
+      "step": 950
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002932728164867517,
+      "loss": 3.4787,
+      "step": 960
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.000293125613346418,
+      "loss": 3.2242,
+      "step": 970
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.0002929784102060844,
+      "loss": 2.9959,
+      "step": 980
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00029283120706575074,
+      "loss": 2.8204,
+      "step": 990
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00029268400392541704,
+      "loss": 2.6594,
+      "step": 1000
+    },
+    {
+      "epoch": 1.44,
+      "eval_loss": 2.4058356285095215,
+      "eval_runtime": 39.9286,
+      "eval_samples_per_second": 11.445,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.9799728956482457,
+      "step": 1000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002925368007850834,
+      "loss": 2.527,
+      "step": 1010
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0002923895976447497,
+      "loss": 2.371,
+      "step": 1020
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.00029224239450441607,
+      "loss": 2.271,
+      "step": 1030
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002920951913640824,
+      "loss": 2.185,
+      "step": 1040
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00029194798822374874,
+      "loss": 2.167,
+      "step": 1050
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002918007850834151,
+      "loss": 2.0004,
+      "step": 1060
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00029165358194308146,
+      "loss": 1.9067,
+      "step": 1070
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.00029150637880274777,
+      "loss": 1.8721,
+      "step": 1080
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.0002913591756624141,
+      "loss": 1.7606,
+      "step": 1090
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00029121197252208043,
+      "loss": 1.729,
+      "step": 1100
+    },
+    {
+      "epoch": 1.58,
+      "eval_loss": 1.5358420610427856,
+      "eval_runtime": 39.5025,
+      "eval_samples_per_second": 11.569,
+      "eval_steps_per_second": 1.468,
+      "eval_wer": 0.8920343321788887,
+      "step": 1100
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002910647693817468,
+      "loss": 1.621,
+      "step": 1110
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002909175662414131,
+      "loss": 1.6394,
+      "step": 1120
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00029077036310107946,
+      "loss": 1.626,
+      "step": 1130
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0002906231599607458,
+      "loss": 1.5561,
+      "step": 1140
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0002904759568204122,
+      "loss": 1.5221,
+      "step": 1150
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.0002903287536800785,
+      "loss": 1.4047,
+      "step": 1160
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00029018155053974485,
+      "loss": 1.4521,
+      "step": 1170
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.00029003434739941115,
+      "loss": 1.3526,
+      "step": 1180
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028988714425907746,
+      "loss": 1.2765,
+      "step": 1190
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0002897399411187438,
+      "loss": 1.3998,
+      "step": 1200
+    },
+    {
+      "epoch": 1.72,
+      "eval_loss": 1.1598572731018066,
+      "eval_runtime": 39.5369,
+      "eval_samples_per_second": 11.559,
+      "eval_steps_per_second": 1.467,
+      "eval_wer": 0.8021382321939466,
+      "step": 1200
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0002895927379784102,
+      "loss": 1.2611,
+      "step": 1210
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00028944553483807654,
+      "loss": 1.2765,
+      "step": 1220
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.00028929833169774285,
+      "loss": 1.1803,
+      "step": 1230
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0002891511285574092,
+      "loss": 1.2348,
+      "step": 1240
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.00028900392541707557,
+      "loss": 1.1927,
+      "step": 1250
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0002888567222767419,
+      "loss": 1.141,
+      "step": 1260
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002887095191364082,
+      "loss": 1.2261,
+      "step": 1270
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00028856231599607454,
+      "loss": 1.1062,
+      "step": 1280
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002884151128557409,
+      "loss": 1.095,
+      "step": 1290
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.00028826790971540726,
+      "loss": 1.1236,
+      "step": 1300
+    },
+    {
+      "epoch": 1.87,
+      "eval_loss": 0.9830673336982727,
+      "eval_runtime": 39.4178,
+      "eval_samples_per_second": 11.594,
+      "eval_steps_per_second": 1.471,
+      "eval_wer": 0.7566631531395874,
+      "step": 1300
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.00028812070657507357,
+      "loss": 1.0961,
+      "step": 1310
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.00028797350343473993,
+      "loss": 1.09,
+      "step": 1320
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.0002878263002944063,
+      "loss": 1.0263,
+      "step": 1330
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002876790971540726,
+      "loss": 1.0797,
+      "step": 1340
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002875318940137389,
+      "loss": 1.0631,
+      "step": 1350
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00028738469087340526,
+      "loss": 0.9842,
+      "step": 1360
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0002872374877330716,
+      "loss": 1.129,
+      "step": 1370
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00028709028459273793,
+      "loss": 0.9745,
+      "step": 1380
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002869430814524043,
+      "loss": 0.9876,
+      "step": 1390
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.00028679587831207065,
+      "loss": 1.0142,
+      "step": 1400
+    },
+    {
+      "epoch": 2.01,
+      "eval_loss": 0.8522623181343079,
+      "eval_runtime": 39.9987,
+      "eval_samples_per_second": 11.425,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.6712844451136877,
+      "step": 1400
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.000286648675171737,
+      "loss": 0.9384,
+      "step": 1410
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0002865014720314033,
+      "loss": 0.9115,
+      "step": 1420
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.0002863542688910696,
+      "loss": 0.9601,
+      "step": 1430
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.000286207065750736,
+      "loss": 0.9199,
+      "step": 1440
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.00028605986261040234,
+      "loss": 0.9359,
+      "step": 1450
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00028591265947006865,
+      "loss": 0.9353,
+      "step": 1460
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.000285765456329735,
+      "loss": 0.8925,
+      "step": 1470
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.00028561825318940137,
+      "loss": 0.9429,
+      "step": 1480
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0002854710500490677,
+      "loss": 0.8772,
+      "step": 1490
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.00028532384690873404,
+      "loss": 0.9132,
+      "step": 1500
+    },
+    {
+      "epoch": 2.16,
+      "eval_loss": 0.802708089351654,
+      "eval_runtime": 39.7319,
+      "eval_samples_per_second": 11.502,
+      "eval_steps_per_second": 1.46,
+      "eval_wer": 0.6539677759373589,
+      "step": 1500
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00028517664376840035,
+      "loss": 0.8947,
+      "step": 1510
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0002850294406280667,
+      "loss": 0.8716,
+      "step": 1520
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.000284882237487733,
+      "loss": 0.8453,
+      "step": 1530
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00028473503434739937,
+      "loss": 0.8724,
+      "step": 1540
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00028458783120706573,
+      "loss": 0.835,
+      "step": 1550
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0002844406280667321,
+      "loss": 0.8617,
+      "step": 1560
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0002842934249263984,
+      "loss": 0.8362,
+      "step": 1570
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00028414622178606476,
+      "loss": 0.7738,
+      "step": 1580
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00028399901864573107,
+      "loss": 0.8412,
+      "step": 1590
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002838518155053974,
+      "loss": 0.8157,
+      "step": 1600
+    },
+    {
+      "epoch": 2.3,
+      "eval_loss": 0.7228513956069946,
+      "eval_runtime": 39.6806,
+      "eval_samples_per_second": 11.517,
+      "eval_steps_per_second": 1.462,
+      "eval_wer": 0.6179792199969885,
+      "step": 1600
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.00028370461236506373,
+      "loss": 0.7675,
+      "step": 1610
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002835574092247301,
+      "loss": 0.8323,
+      "step": 1620
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.00028341020608439645,
+      "loss": 0.784,
+      "step": 1630
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00028326300294406276,
+      "loss": 0.782,
+      "step": 1640
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002831157998037291,
+      "loss": 0.7919,
+      "step": 1650
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.0002829685966633955,
+      "loss": 0.7782,
+      "step": 1660
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.0002828213935230618,
+      "loss": 0.8246,
+      "step": 1670
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.00028267419038272815,
+      "loss": 0.7631,
+      "step": 1680
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.00028252698724239445,
+      "loss": 0.7991,
+      "step": 1690
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0002823797841020608,
+      "loss": 0.8154,
+      "step": 1700
+    },
+    {
+      "epoch": 2.44,
+      "eval_loss": 0.6884952783584595,
+      "eval_runtime": 39.9781,
+      "eval_samples_per_second": 11.431,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.6060834211715103,
+      "step": 1700
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002822325809617272,
+      "loss": 0.7625,
+      "step": 1710
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0002820853778213935,
+      "loss": 0.7888,
+      "step": 1720
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.00028193817468105984,
+      "loss": 0.7091,
+      "step": 1730
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0002817909715407262,
+      "loss": 0.7846,
+      "step": 1740
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0002816437684003925,
+      "loss": 0.7651,
+      "step": 1750
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00028149656526005887,
+      "loss": 0.6977,
+      "step": 1760
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0002813493621197252,
+      "loss": 0.7321,
+      "step": 1770
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00028120215897939154,
+      "loss": 0.6753,
+      "step": 1780
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00028105495583905784,
+      "loss": 0.7269,
+      "step": 1790
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0002809077526987242,
+      "loss": 0.7424,
+      "step": 1800
+    },
+    {
+      "epoch": 2.59,
+      "eval_loss": 0.6070171594619751,
+      "eval_runtime": 39.745,
+      "eval_samples_per_second": 11.498,
+      "eval_steps_per_second": 1.459,
+      "eval_wer": 0.5691913868393315,
+      "step": 1800
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00028076054955839056,
+      "loss": 0.6781,
+      "step": 1810
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0002806133464180569,
+      "loss": 0.7446,
+      "step": 1820
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.00028046614327772323,
+      "loss": 0.7286,
+      "step": 1830
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.0002803189401373896,
+      "loss": 0.707,
+      "step": 1840
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0002801717369970559,
+      "loss": 0.7032,
+      "step": 1850
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00028002453385672226,
+      "loss": 0.7005,
+      "step": 1860
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.00027987733071638856,
+      "loss": 0.6833,
+      "step": 1870
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002797301275760549,
+      "loss": 0.6812,
+      "step": 1880
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0002795829244357213,
+      "loss": 0.6614,
+      "step": 1890
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.00027943572129538765,
+      "loss": 0.6949,
+      "step": 1900
+    },
+    {
+      "epoch": 2.73,
+      "eval_loss": 0.5801286697387695,
+      "eval_runtime": 39.6689,
+      "eval_samples_per_second": 11.52,
+      "eval_steps_per_second": 1.462,
+      "eval_wer": 0.5460021081162475,
+      "step": 1900
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00027928851815505395,
+      "loss": 0.676,
+      "step": 1910
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0002791413150147203,
+      "loss": 0.7185,
+      "step": 1920
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0002789941118743866,
+      "loss": 0.6175,
+      "step": 1930
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.000278846908734053,
+      "loss": 0.6519,
+      "step": 1940
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.0002786997055937193,
+      "loss": 0.662,
+      "step": 1950
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.00027855250245338565,
+      "loss": 0.6317,
+      "step": 1960
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000278405299313052,
+      "loss": 0.6532,
+      "step": 1970
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0002782580961727183,
+      "loss": 0.6097,
+      "step": 1980
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002781108930323847,
+      "loss": 0.6441,
+      "step": 1990
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.00027796368989205103,
+      "loss": 0.6748,
+      "step": 2000
+    },
+    {
+      "epoch": 2.87,
+      "eval_loss": 0.5619427561759949,
+      "eval_runtime": 39.9278,
+      "eval_samples_per_second": 11.446,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.5387742809817798,
+      "step": 2000
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00027781648675171734,
+      "loss": 0.6749,
+      "step": 2010
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0002776692836113837,
+      "loss": 0.6225,
+      "step": 2020
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00027752208047105,
+      "loss": 0.592,
+      "step": 2030
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.00027737487733071637,
+      "loss": 0.6007,
+      "step": 2040
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.00027722767419038273,
+      "loss": 0.6488,
+      "step": 2050
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00027708047105004903,
+      "loss": 0.6334,
+      "step": 2060
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0002769332679097154,
+      "loss": 0.6266,
+      "step": 2070
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.00027678606476938175,
+      "loss": 0.5994,
+      "step": 2080
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00027663886162904806,
+      "loss": 0.6151,
+      "step": 2090
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0002764916584887144,
+      "loss": 0.5569,
+      "step": 2100
+    },
+    {
+      "epoch": 3.02,
+      "eval_loss": 0.5206429958343506,
+      "eval_runtime": 39.8569,
+      "eval_samples_per_second": 11.466,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.5130251468152387,
+      "step": 2100
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00027634445534838073,
+      "loss": 0.5417,
+      "step": 2110
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.0002761972522080471,
+      "loss": 0.6448,
+      "step": 2120
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002760500490677134,
+      "loss": 0.5474,
+      "step": 2130
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00027590284592737976,
+      "loss": 0.5912,
+      "step": 2140
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0002757556427870461,
+      "loss": 0.5482,
+      "step": 2150
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002756084396467125,
+      "loss": 0.5619,
+      "step": 2160
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.0002754612365063788,
+      "loss": 0.581,
+      "step": 2170
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.00027531403336604514,
+      "loss": 0.5275,
+      "step": 2180
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00027516683022571145,
+      "loss": 0.5497,
+      "step": 2190
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002750196270853778,
+      "loss": 0.5584,
+      "step": 2200
+    },
+    {
+      "epoch": 3.16,
+      "eval_loss": 0.4875911772251129,
+      "eval_runtime": 39.9849,
+      "eval_samples_per_second": 11.429,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.49028760728805904,
+      "step": 2200
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002748724239450441,
+      "loss": 0.5276,
+      "step": 2210
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002747252208047105,
+      "loss": 0.5582,
+      "step": 2220
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00027457801766437684,
+      "loss": 0.5294,
+      "step": 2230
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.00027443081452404314,
+      "loss": 0.5978,
+      "step": 2240
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002742836113837095,
+      "loss": 0.5331,
+      "step": 2250
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00027413640824337586,
+      "loss": 0.5714,
+      "step": 2260
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.00027398920510304217,
+      "loss": 0.549,
+      "step": 2270
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002738420019627085,
+      "loss": 0.5706,
+      "step": 2280
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00027369479882237484,
+      "loss": 0.5235,
+      "step": 2290
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002735475956820412,
+      "loss": 0.5333,
+      "step": 2300
+    },
+    {
+      "epoch": 3.3,
+      "eval_loss": 0.46249067783355713,
+      "eval_runtime": 40.1021,
+      "eval_samples_per_second": 11.396,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.4661948501731667,
+      "step": 2300
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00027340039254170756,
+      "loss": 0.5333,
+      "step": 2310
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00027325318940137386,
+      "loss": 0.578,
+      "step": 2320
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.0002731059862610402,
+      "loss": 0.5243,
+      "step": 2330
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0002729587831207066,
+      "loss": 0.5668,
+      "step": 2340
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002728115799803729,
+      "loss": 0.5112,
+      "step": 2350
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002726643768400392,
+      "loss": 0.5039,
+      "step": 2360
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00027251717369970556,
+      "loss": 0.5342,
+      "step": 2370
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.0002723699705593719,
+      "loss": 0.4879,
+      "step": 2380
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002722227674190382,
+      "loss": 0.5277,
+      "step": 2390
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0002720755642787046,
+      "loss": 0.5277,
+      "step": 2400
+    },
+    {
+      "epoch": 3.45,
+      "eval_loss": 0.4557473361492157,
+      "eval_runtime": 39.6153,
+      "eval_samples_per_second": 11.536,
+      "eval_steps_per_second": 1.464,
+      "eval_wer": 0.4631832555338051,
+      "step": 2400
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00027192836113837095,
+      "loss": 0.5052,
+      "step": 2410
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002717811579980373,
+      "loss": 0.5292,
+      "step": 2420
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.0002716339548577036,
+      "loss": 0.5448,
+      "step": 2430
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002714867517173699,
+      "loss": 0.5089,
+      "step": 2440
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002713395485770363,
+      "loss": 0.5381,
+      "step": 2450
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00027119234543670264,
+      "loss": 0.5119,
+      "step": 2460
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.00027104514229636895,
+      "loss": 0.5093,
+      "step": 2470
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002708979391560353,
+      "loss": 0.4931,
+      "step": 2480
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.00027075073601570167,
+      "loss": 0.4823,
+      "step": 2490
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.000270603532875368,
+      "loss": 0.479,
+      "step": 2500
+    },
+    {
+      "epoch": 3.59,
+      "eval_loss": 0.42780354619026184,
+      "eval_runtime": 39.8004,
+      "eval_samples_per_second": 11.482,
+      "eval_steps_per_second": 1.457,
+      "eval_wer": 0.44421020930582744,
+      "step": 2500
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00027045632973503433,
+      "loss": 0.4703,
+      "step": 2510
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00027030912659470064,
+      "loss": 0.4812,
+      "step": 2520
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.000270161923454367,
+      "loss": 0.5042,
+      "step": 2530
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.0002700147203140333,
+      "loss": 0.486,
+      "step": 2540
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.00026986751717369967,
+      "loss": 0.4923,
+      "step": 2550
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.00026972031403336603,
+      "loss": 0.4595,
+      "step": 2560
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002695731108930324,
+      "loss": 0.6039,
+      "step": 2570
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002694259077526987,
+      "loss": 0.4577,
+      "step": 2580
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.00026927870461236506,
+      "loss": 0.4911,
+      "step": 2590
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00026913150147203136,
+      "loss": 0.5368,
+      "step": 2600
+    },
+    {
+      "epoch": 3.74,
+      "eval_loss": 0.43260544538497925,
+      "eval_runtime": 39.7779,
+      "eval_samples_per_second": 11.489,
+      "eval_steps_per_second": 1.458,
+      "eval_wer": 0.44677006474928477,
+      "step": 2600
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0002689842983316977,
+      "loss": 0.4625,
+      "step": 2610
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.00026883709519136403,
+      "loss": 0.5655,
+      "step": 2620
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.0002686898920510304,
+      "loss": 0.4575,
+      "step": 2630
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.00026854268891069675,
+      "loss": 0.5072,
+      "step": 2640
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00026839548577036306,
+      "loss": 0.4799,
+      "step": 2650
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.0002682482826300294,
+      "loss": 0.4849,
+      "step": 2660
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002681010794896958,
+      "loss": 0.5101,
+      "step": 2670
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.0002679538763493621,
+      "loss": 0.4985,
+      "step": 2680
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.00026780667320902844,
+      "loss": 0.4664,
+      "step": 2690
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00026765947006869475,
+      "loss": 0.5113,
+      "step": 2700
+    },
+    {
+      "epoch": 3.88,
+      "eval_loss": 0.40216904878616333,
+      "eval_runtime": 40.3356,
+      "eval_samples_per_second": 11.33,
+      "eval_steps_per_second": 1.438,
+      "eval_wer": 0.4338202078000301,
+      "step": 2700
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0002675122669283611,
+      "loss": 0.4684,
+      "step": 2710
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.00026736506378802747,
+      "loss": 0.4687,
+      "step": 2720
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002672178606476938,
+      "loss": 0.4609,
+      "step": 2730
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00026707065750736014,
+      "loss": 0.4715,
+      "step": 2740
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002669234543670265,
+      "loss": 0.5112,
+      "step": 2750
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.0002667762512266928,
+      "loss": 0.4614,
+      "step": 2760
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00026662904808635917,
+      "loss": 0.4613,
+      "step": 2770
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.00026648184494602547,
+      "loss": 0.4496,
+      "step": 2780
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.00026633464180569183,
+      "loss": 0.4614,
+      "step": 2790
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00026618743866535814,
+      "loss": 0.4377,
+      "step": 2800
+    },
+    {
+      "epoch": 4.02,
+      "eval_loss": 0.3904813528060913,
+      "eval_runtime": 40.1396,
+      "eval_samples_per_second": 11.385,
+      "eval_steps_per_second": 1.445,
+      "eval_wer": 0.42824875771721127,
+      "step": 2800
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002660402355250245,
+      "loss": 0.4191,
+      "step": 2810
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.00026589303238469086,
+      "loss": 0.3959,
+      "step": 2820
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002657458292443572,
+      "loss": 0.4222,
+      "step": 2830
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.0002655986261040235,
+      "loss": 0.3959,
+      "step": 2840
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0002654514229636899,
+      "loss": 0.3709,
+      "step": 2850
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.0002653042198233562,
+      "loss": 0.4196,
+      "step": 2860
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.00026515701668302255,
+      "loss": 0.4266,
+      "step": 2870
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.00026500981354268886,
+      "loss": 0.4069,
+      "step": 2880
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.0002648626104023552,
+      "loss": 0.4123,
+      "step": 2890
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.0002647154072620216,
+      "loss": 0.3715,
+      "step": 2900
+    },
+    {
+      "epoch": 4.17,
+      "eval_loss": 0.3695952892303467,
+      "eval_runtime": 39.8629,
+      "eval_samples_per_second": 11.464,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.40581237765396777,
+      "step": 2900
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00026456820412168794,
+      "loss": 0.4446,
+      "step": 2910
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.00026442100098135425,
+      "loss": 0.4482,
+      "step": 2920
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.0002642737978410206,
+      "loss": 0.4032,
+      "step": 2930
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.0002641265947006869,
+      "loss": 0.425,
+      "step": 2940
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.0002639793915603533,
+      "loss": 0.3987,
+      "step": 2950
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.0002638321884200196,
+      "loss": 0.3933,
+      "step": 2960
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.00026368498527968594,
+      "loss": 0.4381,
+      "step": 2970
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.0002635377821393523,
+      "loss": 0.3662,
+      "step": 2980
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.0002633905789990186,
+      "loss": 0.4008,
+      "step": 2990
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00026324337585868497,
+      "loss": 0.3764,
+      "step": 3000
+    },
+    {
+      "epoch": 4.31,
+      "eval_loss": 0.36124685406684875,
+      "eval_runtime": 39.8744,
+      "eval_samples_per_second": 11.461,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.3918084625809366,
+      "step": 3000
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.00026309617271835133,
+      "loss": 0.4272,
+      "step": 3010
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00026294896957801764,
+      "loss": 0.4066,
+      "step": 3020
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.000262801766437684,
+      "loss": 0.3517,
+      "step": 3030
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.0002626545632973503,
+      "loss": 0.4108,
+      "step": 3040
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.00026250736015701666,
+      "loss": 0.3775,
+      "step": 3050
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.000262360157016683,
+      "loss": 0.3643,
+      "step": 3060
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.00026221295387634933,
+      "loss": 0.395,
+      "step": 3070
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.0002620657507360157,
+      "loss": 0.3699,
+      "step": 3080
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00026191854759568205,
+      "loss": 0.3657,
+      "step": 3090
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00026177134445534836,
+      "loss": 0.3915,
+      "step": 3100
+    },
+    {
+      "epoch": 4.45,
+      "eval_loss": 0.3556678593158722,
+      "eval_runtime": 39.9374,
+      "eval_samples_per_second": 11.443,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.384731215178437,
+      "step": 3100
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.00026162414131501466,
+      "loss": 0.3968,
+      "step": 3110
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.000261476938174681,
+      "loss": 0.4099,
+      "step": 3120
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0002613297350343474,
+      "loss": 0.422,
+      "step": 3130
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002611825318940137,
+      "loss": 0.4122,
+      "step": 3140
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00026103532875368005,
+      "loss": 0.3962,
+      "step": 3150
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.0002608881256133464,
+      "loss": 0.3721,
+      "step": 3160
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00026074092247301277,
+      "loss": 0.468,
+      "step": 3170
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.0002605937193326791,
+      "loss": 0.3766,
+      "step": 3180
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.0002604465161923454,
+      "loss": 0.3768,
+      "step": 3190
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00026029931305201175,
+      "loss": 0.369,
+      "step": 3200
+    },
+    {
+      "epoch": 4.6,
+      "eval_loss": 0.36044731736183167,
+      "eval_runtime": 39.912,
+      "eval_samples_per_second": 11.45,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.39286252070471317,
+      "step": 3200
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.0002601521099116781,
+      "loss": 0.4042,
+      "step": 3210
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.0002600049067713444,
+      "loss": 0.3628,
+      "step": 3220
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00025985770363101077,
+      "loss": 0.3977,
+      "step": 3230
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00025971050049067713,
+      "loss": 0.4106,
+      "step": 3240
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00025956329735034344,
+      "loss": 0.3944,
+      "step": 3250
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.0002594160942100098,
+      "loss": 0.3964,
+      "step": 3260
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.0002592688910696761,
+      "loss": 0.3937,
+      "step": 3270
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00025912168792934247,
+      "loss": 0.33,
+      "step": 3280
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.00025897448478900877,
+      "loss": 0.4144,
+      "step": 3290
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00025882728164867513,
+      "loss": 0.3644,
+      "step": 3300
+    },
+    {
+      "epoch": 4.74,
+      "eval_loss": 0.3271957039833069,
+      "eval_runtime": 39.9356,
+      "eval_samples_per_second": 11.443,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.3604878783315766,
+      "step": 3300
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.0002586800785083415,
+      "loss": 0.3605,
+      "step": 3310
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.00025853287536800785,
+      "loss": 0.3926,
+      "step": 3320
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00025838567222767416,
+      "loss": 0.3565,
+      "step": 3330
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.0002582384690873405,
+      "loss": 0.3856,
+      "step": 3340
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00025809126594700683,
+      "loss": 0.3295,
+      "step": 3350
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.0002579440628066732,
+      "loss": 0.3327,
+      "step": 3360
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.0002577968596663395,
+      "loss": 0.374,
+      "step": 3370
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00025764965652600585,
+      "loss": 0.3291,
+      "step": 3380
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.0002575024533856722,
+      "loss": 0.3902,
+      "step": 3390
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.0002573552502453385,
+      "loss": 0.3959,
+      "step": 3400
+    },
+    {
+      "epoch": 4.89,
+      "eval_loss": 0.3158166706562042,
+      "eval_runtime": 40.0972,
+      "eval_samples_per_second": 11.397,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.36681222707423583,
+      "step": 3400
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.0002572080471050049,
+      "loss": 0.3636,
+      "step": 3410
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00025706084396467124,
+      "loss": 0.3581,
+      "step": 3420
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00025691364082433755,
+      "loss": 0.3505,
+      "step": 3430
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0002567664376840039,
+      "loss": 0.3967,
+      "step": 3440
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.0002566192345436702,
+      "loss": 0.3328,
+      "step": 3450
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.0002564720314033366,
+      "loss": 0.3625,
+      "step": 3460
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00025632482826300294,
+      "loss": 0.386,
+      "step": 3470
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00025617762512266924,
+      "loss": 0.3102,
+      "step": 3480
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002560304219823356,
+      "loss": 0.3945,
+      "step": 3490
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00025588321884200196,
+      "loss": 0.3307,
+      "step": 3500
+    },
+    {
+      "epoch": 5.03,
+      "eval_loss": 0.32223132252693176,
+      "eval_runtime": 40.0169,
+      "eval_samples_per_second": 11.42,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.35702454449631077,
+      "step": 3500
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00025573601570166827,
+      "loss": 0.3412,
+      "step": 3510
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.00025558881256133463,
+      "loss": 0.3243,
+      "step": 3520
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.00025544160942100094,
+      "loss": 0.3513,
+      "step": 3530
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0002552944062806673,
+      "loss": 0.3336,
+      "step": 3540
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0002551472031403336,
+      "loss": 0.318,
+      "step": 3550
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00025499999999999996,
+      "loss": 0.3662,
+      "step": 3560
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.0002548527968596663,
+      "loss": 0.3265,
+      "step": 3570
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002547055937193327,
+      "loss": 0.2784,
+      "step": 3580
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.000254558390578999,
+      "loss": 0.3218,
+      "step": 3590
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00025441118743866535,
+      "loss": 0.3283,
+      "step": 3600
+    },
+    {
+      "epoch": 5.17,
+      "eval_loss": 0.3063613176345825,
+      "eval_runtime": 40.0819,
+      "eval_samples_per_second": 11.402,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.3576268634241831,
+      "step": 3600
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00025426398429833166,
+      "loss": 0.3153,
+      "step": 3610
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.000254116781157998,
+      "loss": 0.2926,
+      "step": 3620
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002539695780176643,
+      "loss": 0.3065,
+      "step": 3630
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.0002538223748773307,
+      "loss": 0.3656,
+      "step": 3640
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00025367517173699705,
+      "loss": 0.2764,
+      "step": 3650
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.0002535279685966634,
+      "loss": 0.3197,
+      "step": 3660
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0002533807654563297,
+      "loss": 0.3113,
+      "step": 3670
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.0002532335623159961,
+      "loss": 0.3154,
+      "step": 3680
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.0002530863591756624,
+      "loss": 0.3771,
+      "step": 3690
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00025293915603532874,
+      "loss": 0.2934,
+      "step": 3700
+    },
+    {
+      "epoch": 5.32,
+      "eval_loss": 0.30499157309532166,
+      "eval_runtime": 39.856,
+      "eval_samples_per_second": 11.466,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.34829092004216233,
+      "step": 3700
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00025279195289499505,
+      "loss": 0.3419,
+      "step": 3710
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.0002526447497546614,
+      "loss": 0.3312,
+      "step": 3720
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00025249754661432777,
+      "loss": 0.2881,
+      "step": 3730
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.0002523503434739941,
+      "loss": 0.3223,
+      "step": 3740
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.00025220314033366043,
+      "loss": 0.3118,
+      "step": 3750
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.0002520559371933268,
+      "loss": 0.3139,
+      "step": 3760
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0002519087340529931,
+      "loss": 0.3198,
+      "step": 3770
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00025176153091265946,
+      "loss": 0.2908,
+      "step": 3780
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00025161432777232577,
+      "loss": 0.3446,
+      "step": 3790
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00025146712463199213,
+      "loss": 0.3015,
+      "step": 3800
+    },
+    {
+      "epoch": 5.46,
+      "eval_loss": 0.28972870111465454,
+      "eval_runtime": 40.1556,
+      "eval_samples_per_second": 11.381,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.3416654118355669,
+      "step": 3800
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0002513199214916585,
+      "loss": 0.3488,
+      "step": 3810
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.0002511727183513248,
+      "loss": 0.2873,
+      "step": 3820
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.00025102551521099116,
+      "loss": 0.2951,
+      "step": 3830
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.0002508783120706575,
+      "loss": 0.3305,
+      "step": 3840
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.0002507311089303238,
+      "loss": 0.3062,
+      "step": 3850
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0002505839057899902,
+      "loss": 0.3912,
+      "step": 3860
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.0002504367026496565,
+      "loss": 0.2884,
+      "step": 3870
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00025028949950932285,
+      "loss": 0.3128,
+      "step": 3880
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00025014229636898916,
+      "loss": 0.3343,
+      "step": 3890
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.0002499950932286555,
+      "loss": 0.2929,
+      "step": 3900
+    },
+    {
+      "epoch": 5.6,
+      "eval_loss": 0.28876641392707825,
+      "eval_runtime": 40.338,
+      "eval_samples_per_second": 11.329,
+      "eval_steps_per_second": 1.438,
+      "eval_wer": 0.31862671284445115,
+      "step": 3900
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.0002498478900883219,
+      "loss": 0.3341,
+      "step": 3910
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.00024970068694798824,
+      "loss": 0.3196,
+      "step": 3920
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00024955348380765454,
+      "loss": 0.3025,
+      "step": 3930
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0002494062806673209,
+      "loss": 0.3231,
+      "step": 3940
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.0002492590775269872,
+      "loss": 0.2989,
+      "step": 3950
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00024911187438665357,
+      "loss": 0.304,
+      "step": 3960
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0002489646712463199,
+      "loss": 0.2816,
+      "step": 3970
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.00024881746810598624,
+      "loss": 0.3096,
+      "step": 3980
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.0002486702649656526,
+      "loss": 0.3,
+      "step": 3990
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002485230618253189,
+      "loss": 0.2815,
+      "step": 4000
+    },
+    {
+      "epoch": 5.75,
+      "eval_loss": 0.28788086771965027,
+      "eval_runtime": 40.0114,
+      "eval_samples_per_second": 11.422,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.31666917632886615,
+      "step": 4000
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00024837585868498526,
+      "loss": 0.3083,
+      "step": 4010
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.0002482286555446516,
+      "loss": 0.2857,
+      "step": 4020
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00024808145240431793,
+      "loss": 0.3066,
+      "step": 4030
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00024793424926398424,
+      "loss": 0.2842,
+      "step": 4040
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.0002477870461236506,
+      "loss": 0.2664,
+      "step": 4050
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00024763984298331696,
+      "loss": 0.3455,
+      "step": 4060
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.0002474926398429833,
+      "loss": 0.2824,
+      "step": 4070
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.0002473454367026496,
+      "loss": 0.2934,
+      "step": 4080
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.000247198233562316,
+      "loss": 0.3343,
+      "step": 4090
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00024705103042198235,
+      "loss": 0.2778,
+      "step": 4100
+    },
+    {
+      "epoch": 5.89,
+      "eval_loss": 0.2641572952270508,
+      "eval_runtime": 39.8509,
+      "eval_samples_per_second": 11.468,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.3136575816895046,
+      "step": 4100
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.00024690382728164865,
+      "loss": 0.2685,
+      "step": 4110
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.00024675662414131496,
+      "loss": 0.3127,
+      "step": 4120
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.0002466094210009813,
+      "loss": 0.2618,
+      "step": 4130
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.0002464622178606477,
+      "loss": 0.3303,
+      "step": 4140
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.000246315014720314,
+      "loss": 0.2635,
+      "step": 4150
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00024616781157998035,
+      "loss": 0.3033,
+      "step": 4160
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002460206084396467,
+      "loss": 0.2614,
+      "step": 4170
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.00024587340529931307,
+      "loss": 0.2824,
+      "step": 4180
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.0002457262021589794,
+      "loss": 0.247,
+      "step": 4190
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0002455789990186457,
+      "loss": 0.2822,
+      "step": 4200
+    },
+    {
+      "epoch": 6.03,
+      "eval_loss": 0.2591544985771179,
+      "eval_runtime": 40.0116,
+      "eval_samples_per_second": 11.422,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.29920192742056917,
+      "step": 4200
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00024543179587831204,
+      "loss": 0.2866,
+      "step": 4210
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.0002452845927379784,
+      "loss": 0.24,
+      "step": 4220
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.0002451373895976447,
+      "loss": 0.249,
+      "step": 4230
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.00024499018645731107,
+      "loss": 0.2542,
+      "step": 4240
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00024484298331697743,
+      "loss": 0.2672,
+      "step": 4250
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.0002446957801766438,
+      "loss": 0.3306,
+      "step": 4260
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.0002445485770363101,
+      "loss": 0.2413,
+      "step": 4270
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0002444013738959764,
+      "loss": 0.2675,
+      "step": 4280
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00024425417075564276,
+      "loss": 0.2492,
+      "step": 4290
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.0002441069676153091,
+      "loss": 0.2343,
+      "step": 4300
+    },
+    {
+      "epoch": 6.18,
+      "eval_loss": 0.2688565254211426,
+      "eval_runtime": 39.6355,
+      "eval_samples_per_second": 11.53,
+      "eval_steps_per_second": 1.463,
+      "eval_wer": 0.294835115193495,
+      "step": 4300
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00024395976447497543,
+      "loss": 0.3075,
+      "step": 4310
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0002438125613346418,
+      "loss": 0.2398,
+      "step": 4320
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00024366535819430812,
+      "loss": 0.2782,
+      "step": 4330
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.00024351815505397446,
+      "loss": 0.2511,
+      "step": 4340
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.0002433709519136408,
+      "loss": 0.234,
+      "step": 4350
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00024322374877330715,
+      "loss": 0.2972,
+      "step": 4360
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.00024307654563297348,
+      "loss": 0.2553,
+      "step": 4370
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00024292934249263982,
+      "loss": 0.2697,
+      "step": 4380
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.00024278213935230615,
+      "loss": 0.2303,
+      "step": 4390
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0002426349362119725,
+      "loss": 0.2556,
+      "step": 4400
+    },
+    {
+      "epoch": 6.32,
+      "eval_loss": 0.2687687575817108,
+      "eval_runtime": 40.0355,
+      "eval_samples_per_second": 11.415,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.28896250564674,
+      "step": 4400
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00024248773307163884,
+      "loss": 0.2581,
+      "step": 4410
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.00024234052993130518,
+      "loss": 0.2807,
+      "step": 4420
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.0002421933267909715,
+      "loss": 0.2925,
+      "step": 4430
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00024204612365063787,
+      "loss": 0.2417,
+      "step": 4440
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00024189892051030418,
+      "loss": 0.2539,
+      "step": 4450
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.00024175171736997054,
+      "loss": 0.2756,
+      "step": 4460
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00024160451422963687,
+      "loss": 0.223,
+      "step": 4470
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00024145731108930323,
+      "loss": 0.2717,
+      "step": 4480
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00024131010794896954,
+      "loss": 0.2731,
+      "step": 4490
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.0002411629048086359,
+      "loss": 0.2437,
+      "step": 4500
+    },
+    {
+      "epoch": 6.47,
+      "eval_loss": 0.2555336654186249,
+      "eval_runtime": 40.1329,
+      "eval_samples_per_second": 11.387,
+      "eval_steps_per_second": 1.445,
+      "eval_wer": 0.28760728805902724,
+      "step": 4500
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00024101570166830223,
+      "loss": 0.2746,
+      "step": 4510
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.0002408684985279686,
+      "loss": 0.2506,
+      "step": 4520
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.0002407212953876349,
+      "loss": 0.2604,
+      "step": 4530
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.00024057409224730126,
+      "loss": 0.2634,
+      "step": 4540
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.0002404268891069676,
+      "loss": 0.2561,
+      "step": 4550
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.00024027968596663395,
+      "loss": 0.2648,
+      "step": 4560
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00024013248282630026,
+      "loss": 0.2446,
+      "step": 4570
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00023998527968596662,
+      "loss": 0.2685,
+      "step": 4580
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.00023983807654563295,
+      "loss": 0.2434,
+      "step": 4590
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.0002396908734052993,
+      "loss": 0.237,
+      "step": 4600
+    },
+    {
+      "epoch": 6.61,
+      "eval_loss": 0.2497735321521759,
+      "eval_runtime": 40.0633,
+      "eval_samples_per_second": 11.407,
+      "eval_steps_per_second": 1.448,
+      "eval_wer": 0.28233699744014457,
+      "step": 4600
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00023954367026496562,
+      "loss": 0.2712,
+      "step": 4610
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00023939646712463198,
+      "loss": 0.2464,
+      "step": 4620
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.00023924926398429831,
+      "loss": 0.2903,
+      "step": 4630
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00023910206084396465,
+      "loss": 0.2472,
+      "step": 4640
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.00023895485770363098,
+      "loss": 0.2602,
+      "step": 4650
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00023880765456329734,
+      "loss": 0.2728,
+      "step": 4660
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00023866045142296368,
+      "loss": 0.2636,
+      "step": 4670
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00023851324828263,
+      "loss": 0.2515,
+      "step": 4680
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.00023836604514229634,
+      "loss": 0.2602,
+      "step": 4690
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.0002382188420019627,
+      "loss": 0.2375,
+      "step": 4700
+    },
+    {
+      "epoch": 6.75,
+      "eval_loss": 0.2609308660030365,
+      "eval_runtime": 39.7988,
+      "eval_samples_per_second": 11.483,
+      "eval_steps_per_second": 1.457,
+      "eval_wer": 0.29363047733775033,
+      "step": 4700
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00023807163886162904,
+      "loss": 0.2602,
+      "step": 4710
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 0.00023792443572129537,
+      "loss": 0.2384,
+      "step": 4720
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0002377772325809617,
+      "loss": 0.2466,
+      "step": 4730
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00023763002944062806,
+      "loss": 0.2674,
+      "step": 4740
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00023748282630029437,
+      "loss": 0.2497,
+      "step": 4750
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.00023733562315996073,
+      "loss": 0.2771,
+      "step": 4760
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00023718842001962706,
+      "loss": 0.2276,
+      "step": 4770
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00023704121687929342,
+      "loss": 0.2459,
+      "step": 4780
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 0.00023689401373895973,
+      "loss": 0.2425,
+      "step": 4790
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.0002367468105986261,
+      "loss": 0.2361,
+      "step": 4800
+    },
+    {
+      "epoch": 6.9,
+      "eval_loss": 0.2400684505701065,
+      "eval_runtime": 39.986,
+      "eval_samples_per_second": 11.429,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.28037946092455956,
+      "step": 4800
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00023659960745829242,
+      "loss": 0.2473,
+      "step": 4810
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.00023645240431795878,
+      "loss": 0.2144,
+      "step": 4820
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.0002363052011776251,
+      "loss": 0.2641,
+      "step": 4830
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00023615799803729145,
+      "loss": 0.2559,
+      "step": 4840
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 0.00023601079489695778,
+      "loss": 0.2404,
+      "step": 4850
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.0002358635917566241,
+      "loss": 0.2533,
+      "step": 4860
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.00023571638861629045,
+      "loss": 0.2301,
+      "step": 4870
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.0002355691854759568,
+      "loss": 0.2553,
+      "step": 4880
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00023542198233562315,
+      "loss": 0.2102,
+      "step": 4890
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00023527477919528945,
+      "loss": 0.2266,
+      "step": 4900
+    },
+    {
+      "epoch": 7.04,
+      "eval_loss": 0.2387649267911911,
+      "eval_runtime": 40.0231,
+      "eval_samples_per_second": 11.418,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.2708929378105707,
+      "step": 4900
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0002351275760549558,
+      "loss": 0.2141,
+      "step": 4910
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00023498037291462217,
+      "loss": 0.2192,
+      "step": 4920
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.0002348331697742885,
+      "loss": 0.2375,
+      "step": 4930
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.0002346859666339548,
+      "loss": 0.2104,
+      "step": 4940
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 0.00023453876349362117,
+      "loss": 0.2054,
+      "step": 4950
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.00023439156035328753,
+      "loss": 0.2081,
+      "step": 4960
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00023424435721295387,
+      "loss": 0.2141,
+      "step": 4970
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00023409715407262017,
+      "loss": 0.2106,
+      "step": 4980
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00023394995093228653,
+      "loss": 0.1952,
+      "step": 4990
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.0002338027477919529,
+      "loss": 0.2299,
+      "step": 5000
+    },
+    {
+      "epoch": 7.18,
+      "eval_loss": 0.24284440279006958,
+      "eval_runtime": 39.8933,
+      "eval_samples_per_second": 11.456,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.2772172865532299,
+      "step": 5000
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 0.0002336555446516192,
+      "loss": 0.2312,
+      "step": 5010
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.00023350834151128553,
+      "loss": 0.2173,
+      "step": 5020
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0002333611383709519,
+      "loss": 0.2383,
+      "step": 5030
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00023321393523061825,
+      "loss": 0.2071,
+      "step": 5040
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00023306673209028456,
+      "loss": 0.2402,
+      "step": 5050
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.0002329195289499509,
+      "loss": 0.2559,
+      "step": 5060
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00023277232580961725,
+      "loss": 0.2094,
+      "step": 5070
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.00023262512266928362,
+      "loss": 0.2216,
+      "step": 5080
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00023247791952894992,
+      "loss": 0.2184,
+      "step": 5090
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00023233071638861625,
+      "loss": 0.2322,
+      "step": 5100
+    },
+    {
+      "epoch": 7.33,
+      "eval_loss": 0.24409331381320953,
+      "eval_runtime": 40.158,
+      "eval_samples_per_second": 11.38,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.27450685137780456,
+      "step": 5100
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 0.00023218351324828262,
+      "loss": 0.2169,
+      "step": 5110
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00023203631010794898,
+      "loss": 0.243,
+      "step": 5120
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.00023188910696761528,
+      "loss": 0.2499,
+      "step": 5130
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00023174190382728162,
+      "loss": 0.2151,
+      "step": 5140
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00023159470068694798,
+      "loss": 0.2384,
+      "step": 5150
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.00023144749754661428,
+      "loss": 0.2193,
+      "step": 5160
+    },
+    {
+      "epoch": 7.43,
+      "learning_rate": 0.00023130029440628064,
+      "loss": 0.214,
+      "step": 5170
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00023115309126594698,
+      "loss": 0.2394,
+      "step": 5180
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.00023100588812561334,
+      "loss": 0.1871,
+      "step": 5190
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00023085868498527964,
+      "loss": 0.214,
+      "step": 5200
+    },
+    {
+      "epoch": 7.47,
+      "eval_loss": 0.24094292521476746,
+      "eval_runtime": 40.0199,
+      "eval_samples_per_second": 11.419,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.2722481553982834,
+      "step": 5200
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.000230711481844946,
+      "loss": 0.2096,
+      "step": 5210
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.00023056427870461234,
+      "loss": 0.2079,
+      "step": 5220
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.0002304170755642787,
+      "loss": 0.233,
+      "step": 5230
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.000230269872423945,
+      "loss": 0.244,
+      "step": 5240
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00023012266928361136,
+      "loss": 0.2502,
+      "step": 5250
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.0002299754661432777,
+      "loss": 0.2079,
+      "step": 5260
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 0.00022982826300294406,
+      "loss": 0.2067,
+      "step": 5270
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00022968105986261036,
+      "loss": 0.2274,
+      "step": 5280
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00022953385672227672,
+      "loss": 0.2019,
+      "step": 5290
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.00022938665358194306,
+      "loss": 0.2223,
+      "step": 5300
+    },
+    {
+      "epoch": 7.61,
+      "eval_loss": 0.24156498908996582,
+      "eval_runtime": 39.8819,
+      "eval_samples_per_second": 11.459,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.26923656075892183,
+      "step": 5300
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0002292394504416094,
+      "loss": 0.2297,
+      "step": 5310
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00022909224730127573,
+      "loss": 0.1978,
+      "step": 5320
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00022894504416094209,
+      "loss": 0.2437,
+      "step": 5330
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00022879784102060842,
+      "loss": 0.2047,
+      "step": 5340
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00022865063788027475,
+      "loss": 0.2329,
+      "step": 5350
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00022850343473994109,
+      "loss": 0.2225,
+      "step": 5360
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00022835623159960745,
+      "loss": 0.2198,
+      "step": 5370
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00022820902845927378,
+      "loss": 0.2289,
+      "step": 5380
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.0002280618253189401,
+      "loss": 0.2167,
+      "step": 5390
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00022791462217860645,
+      "loss": 0.2201,
+      "step": 5400
+    },
+    {
+      "epoch": 7.76,
+      "eval_loss": 0.22141778469085693,
+      "eval_runtime": 39.8574,
+      "eval_samples_per_second": 11.466,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.2546303267580184,
+      "step": 5400
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.0002277674190382728,
+      "loss": 0.2066,
+      "step": 5410
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00022762021589793914,
+      "loss": 0.2034,
+      "step": 5420
+    },
+    {
+      "epoch": 7.8,
+      "learning_rate": 0.00022747301275760547,
+      "loss": 0.2038,
+      "step": 5430
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.0002273258096172718,
+      "loss": 0.2034,
+      "step": 5440
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00022717860647693817,
+      "loss": 0.2328,
+      "step": 5450
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00022703140333660447,
+      "loss": 0.1861,
+      "step": 5460
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.00022688420019627083,
+      "loss": 0.2146,
+      "step": 5470
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00022673699705593717,
+      "loss": 0.2463,
+      "step": 5480
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00022658979391560353,
+      "loss": 0.1953,
+      "step": 5490
+    },
+    {
+      "epoch": 7.9,
+      "learning_rate": 0.00022644259077526983,
+      "loss": 0.2225,
+      "step": 5500
+    },
+    {
+      "epoch": 7.9,
+      "eval_loss": 0.22256095707416534,
+      "eval_runtime": 39.882,
+      "eval_samples_per_second": 11.459,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.2653214877277518,
+      "step": 5500
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.0002262953876349362,
+      "loss": 0.2189,
+      "step": 5510
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00022614818449460253,
+      "loss": 0.1998,
+      "step": 5520
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.0002260009813542689,
+      "loss": 0.2587,
+      "step": 5530
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.0002258537782139352,
+      "loss": 0.2029,
+      "step": 5540
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00022570657507360156,
+      "loss": 0.1974,
+      "step": 5550
+    },
+    {
+      "epoch": 7.99,
+      "learning_rate": 0.0002255593719332679,
+      "loss": 0.2219,
+      "step": 5560
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00022541216879293425,
+      "loss": 0.2126,
+      "step": 5570
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.00022526496565260056,
+      "loss": 0.2219,
+      "step": 5580
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.00022511776251226692,
+      "loss": 0.1896,
+      "step": 5590
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00022497055937193325,
+      "loss": 0.2159,
+      "step": 5600
+    },
+    {
+      "epoch": 8.05,
+      "eval_loss": 0.21107521653175354,
+      "eval_runtime": 40.0872,
+      "eval_samples_per_second": 11.4,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.24318626712844452,
+      "step": 5600
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.00022482335623159958,
+      "loss": 0.1805,
+      "step": 5610
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00022467615309126592,
+      "loss": 0.1778,
+      "step": 5620
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00022452894995093228,
+      "loss": 0.1797,
+      "step": 5630
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.0002243817468105986,
+      "loss": 0.1686,
+      "step": 5640
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00022423454367026494,
+      "loss": 0.1895,
+      "step": 5650
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 0.00022408734052993128,
+      "loss": 0.1648,
+      "step": 5660
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00022394013738959764,
+      "loss": 0.2194,
+      "step": 5670
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00022379293424926397,
+      "loss": 0.1901,
+      "step": 5680
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.0002236457311089303,
+      "loss": 0.1913,
+      "step": 5690
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00022349852796859664,
+      "loss": 0.2291,
+      "step": 5700
+    },
+    {
+      "epoch": 8.19,
+      "eval_loss": 0.20670226216316223,
+      "eval_runtime": 40.0404,
+      "eval_samples_per_second": 11.413,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.24680018069567836,
+      "step": 5700
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.000223351324828263,
+      "loss": 0.1716,
+      "step": 5710
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 0.00022320412168792933,
+      "loss": 0.1894,
+      "step": 5720
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.00022305691854759567,
+      "loss": 0.2024,
+      "step": 5730
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.000222909715407262,
+      "loss": 0.1778,
+      "step": 5740
+    },
+    {
+      "epoch": 8.26,
+      "learning_rate": 0.00022276251226692836,
+      "loss": 0.2244,
+      "step": 5750
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00022261530912659467,
+      "loss": 0.2153,
+      "step": 5760
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00022246810598626103,
+      "loss": 0.2085,
+      "step": 5770
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00022232090284592736,
+      "loss": 0.1727,
+      "step": 5780
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00022217369970559372,
+      "loss": 0.1769,
+      "step": 5790
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00022202649656526003,
+      "loss": 0.2067,
+      "step": 5800
+    },
+    {
+      "epoch": 8.33,
+      "eval_loss": 0.2133297473192215,
+      "eval_runtime": 40.2177,
+      "eval_samples_per_second": 11.363,
+      "eval_steps_per_second": 1.442,
+      "eval_wer": 0.23701249811775335,
+      "step": 5800
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.00022187929342492639,
+      "loss": 0.1915,
+      "step": 5810
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 0.00022173209028459272,
+      "loss": 0.1666,
+      "step": 5820
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00022158488714425908,
+      "loss": 0.2038,
+      "step": 5830
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.0002214376840039254,
+      "loss": 0.1724,
+      "step": 5840
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00022129048086359175,
+      "loss": 0.2457,
+      "step": 5850
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00022114327772325808,
+      "loss": 0.1776,
+      "step": 5860
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.00022099607458292444,
+      "loss": 0.1952,
+      "step": 5870
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 0.00022084887144259075,
+      "loss": 0.2253,
+      "step": 5880
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.0002207016683022571,
+      "loss": 0.1772,
+      "step": 5890
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00022055446516192344,
+      "loss": 0.1882,
+      "step": 5900
+    },
+    {
+      "epoch": 8.48,
+      "eval_loss": 0.21073894202709198,
+      "eval_runtime": 39.9763,
+      "eval_samples_per_second": 11.432,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.23520554133413643,
+      "step": 5900
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00022040726202158975,
+      "loss": 0.1848,
+      "step": 5910
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0002202600588812561,
+      "loss": 0.1788,
+      "step": 5920
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00022011285574092247,
+      "loss": 0.2083,
+      "step": 5930
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.0002199656526005888,
+      "loss": 0.206,
+      "step": 5940
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.0002198184494602551,
+      "loss": 0.2169,
+      "step": 5950
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.00021967124631992147,
+      "loss": 0.1949,
+      "step": 5960
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00021952404317958783,
+      "loss": 0.1824,
+      "step": 5970
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 0.00021937684003925416,
+      "loss": 0.2049,
+      "step": 5980
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00021922963689892047,
+      "loss": 0.166,
+      "step": 5990
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00021908243375858683,
+      "loss": 0.2055,
+      "step": 6000
+    },
+    {
+      "epoch": 8.62,
+      "eval_loss": 0.21278780698776245,
+      "eval_runtime": 39.9574,
+      "eval_samples_per_second": 11.437,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.2550820659539226,
+      "step": 6000
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 0.0002189352306182532,
+      "loss": 0.1874,
+      "step": 6010
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00021878802747791952,
+      "loss": 0.1799,
+      "step": 6020
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00021864082433758583,
+      "loss": 0.2084,
+      "step": 6030
+    },
+    {
+      "epoch": 8.68,
+      "learning_rate": 0.0002184936211972522,
+      "loss": 0.207,
+      "step": 6040
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.00021834641805691855,
+      "loss": 0.215,
+      "step": 6050
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.00021819921491658486,
+      "loss": 0.1749,
+      "step": 6060
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.0002180520117762512,
+      "loss": 0.1796,
+      "step": 6070
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00021790480863591755,
+      "loss": 0.1734,
+      "step": 6080
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.0002177576054955839,
+      "loss": 0.1894,
+      "step": 6090
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.00021761040235525022,
+      "loss": 0.2144,
+      "step": 6100
+    },
+    {
+      "epoch": 8.76,
+      "eval_loss": 0.20413599908351898,
+      "eval_runtime": 40.0035,
+      "eval_samples_per_second": 11.424,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.2302364101791899,
+      "step": 6100
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00021746319921491655,
+      "loss": 0.2061,
+      "step": 6110
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.0002173159960745829,
+      "loss": 0.1862,
+      "step": 6120
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00021716879293424927,
+      "loss": 0.2183,
+      "step": 6130
+    },
+    {
+      "epoch": 8.82,
+      "learning_rate": 0.00021702158979391558,
+      "loss": 0.1819,
+      "step": 6140
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0002168743866535819,
+      "loss": 0.2072,
+      "step": 6150
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00021672718351324827,
+      "loss": 0.1876,
+      "step": 6160
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.0002165799803729146,
+      "loss": 0.2077,
+      "step": 6170
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.00021643277723258094,
+      "loss": 0.2064,
+      "step": 6180
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00021628557409224727,
+      "loss": 0.1796,
+      "step": 6190
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 0.00021613837095191363,
+      "loss": 0.1948,
+      "step": 6200
+    },
+    {
+      "epoch": 8.91,
+      "eval_loss": 0.207631453871727,
+      "eval_runtime": 39.8141,
+      "eval_samples_per_second": 11.478,
+      "eval_steps_per_second": 1.457,
+      "eval_wer": 0.2347538021382322,
+      "step": 6200
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.00021599116781157994,
+      "loss": 0.1759,
+      "step": 6210
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.0002158439646712463,
+      "loss": 0.2199,
+      "step": 6220
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.00021569676153091263,
+      "loss": 0.1782,
+      "step": 6230
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.000215549558390579,
+      "loss": 0.1646,
+      "step": 6240
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.0002154023552502453,
+      "loss": 0.203,
+      "step": 6250
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00021525515210991166,
+      "loss": 0.185,
+      "step": 6260
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.000215107948969578,
+      "loss": 0.1783,
+      "step": 6270
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00021496074582924435,
+      "loss": 0.163,
+      "step": 6280
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.00021481354268891066,
+      "loss": 0.1734,
+      "step": 6290
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00021466633954857702,
+      "loss": 0.176,
+      "step": 6300
+    },
+    {
+      "epoch": 9.05,
+      "eval_loss": 0.20396728813648224,
+      "eval_runtime": 39.8795,
+      "eval_samples_per_second": 11.46,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.22933293178738143,
+      "step": 6300
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00021451913640824335,
+      "loss": 0.1499,
+      "step": 6310
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00021437193326790971,
+      "loss": 0.1693,
+      "step": 6320
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00021422473012757602,
+      "loss": 0.1554,
+      "step": 6330
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00021407752698724238,
+      "loss": 0.158,
+      "step": 6340
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00021393032384690871,
+      "loss": 0.1957,
+      "step": 6350
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 0.00021378312070657505,
+      "loss": 0.1862,
+      "step": 6360
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00021363591756624138,
+      "loss": 0.1581,
+      "step": 6370
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00021348871442590774,
+      "loss": 0.133,
+      "step": 6380
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00021334151128557408,
+      "loss": 0.1534,
+      "step": 6390
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.0002131943081452404,
+      "loss": 0.1892,
+      "step": 6400
+    },
+    {
+      "epoch": 9.2,
+      "eval_loss": 0.1975768804550171,
+      "eval_runtime": 39.8911,
+      "eval_samples_per_second": 11.456,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.22466496009637102,
+      "step": 6400
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00021304710500490674,
+      "loss": 0.146,
+      "step": 6410
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.0002128999018645731,
+      "loss": 0.1676,
+      "step": 6420
+    },
+    {
+      "epoch": 9.24,
+      "learning_rate": 0.00021275269872423944,
+      "loss": 0.135,
+      "step": 6430
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.00021260549558390577,
+      "loss": 0.1776,
+      "step": 6440
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.0002124582924435721,
+      "loss": 0.155,
+      "step": 6450
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 0.00021231108930323846,
+      "loss": 0.1667,
+      "step": 6460
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.0002121638861629048,
+      "loss": 0.1514,
+      "step": 6470
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00021201668302257113,
+      "loss": 0.1718,
+      "step": 6480
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00021186947988223746,
+      "loss": 0.1912,
+      "step": 6490
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00021172227674190382,
+      "loss": 0.1645,
+      "step": 6500
+    },
+    {
+      "epoch": 9.34,
+      "eval_loss": 0.19538117945194244,
+      "eval_runtime": 40.169,
+      "eval_samples_per_second": 11.377,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.21547959644631834,
+      "step": 6500
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00021157507360157013,
+      "loss": 0.1581,
+      "step": 6510
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.0002114278704612365,
+      "loss": 0.1781,
+      "step": 6520
+    },
+    {
+      "epoch": 9.38,
+      "learning_rate": 0.00021128066732090282,
+      "loss": 0.1437,
+      "step": 6530
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00021113346418056918,
+      "loss": 0.1538,
+      "step": 6540
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0002109862610402355,
+      "loss": 0.1945,
+      "step": 6550
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00021083905789990185,
+      "loss": 0.1629,
+      "step": 6560
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.00021069185475956818,
+      "loss": 0.1905,
+      "step": 6570
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00021054465161923455,
+      "loss": 0.1688,
+      "step": 6580
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 0.00021039744847890085,
+      "loss": 0.1678,
+      "step": 6590
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.0002102502453385672,
+      "loss": 0.1723,
+      "step": 6600
+    },
+    {
+      "epoch": 9.48,
+      "eval_loss": 0.19737230241298676,
+      "eval_runtime": 40.207,
+      "eval_samples_per_second": 11.366,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.2215027857250414,
+      "step": 6600
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00021010304219823355,
+      "loss": 0.1664,
+      "step": 6610
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.0002099558390578999,
+      "loss": 0.1783,
+      "step": 6620
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.0002098086359175662,
+      "loss": 0.1776,
+      "step": 6630
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00020966143277723257,
+      "loss": 0.1511,
+      "step": 6640
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.0002095142296368989,
+      "loss": 0.1905,
+      "step": 6650
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.00020936702649656524,
+      "loss": 0.1478,
+      "step": 6660
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00020921982335623157,
+      "loss": 0.1802,
+      "step": 6670
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.00020907262021589793,
+      "loss": 0.1668,
+      "step": 6680
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 0.00020892541707556427,
+      "loss": 0.1529,
+      "step": 6690
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.0002087782139352306,
+      "loss": 0.1685,
+      "step": 6700
+    },
+    {
+      "epoch": 9.63,
+      "eval_loss": 0.20146189630031586,
+      "eval_runtime": 39.9776,
+      "eval_samples_per_second": 11.431,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.21773829242583947,
+      "step": 6700
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00020863101079489693,
+      "loss": 0.1339,
+      "step": 6710
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.0002084838076545633,
+      "loss": 0.175,
+      "step": 6720
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00020833660451422963,
+      "loss": 0.1607,
+      "step": 6730
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00020818940137389596,
+      "loss": 0.1571,
+      "step": 6740
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 0.0002080421982335623,
+      "loss": 0.1678,
+      "step": 6750
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00020789499509322865,
+      "loss": 0.1585,
+      "step": 6760
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.000207747791952895,
+      "loss": 0.2009,
+      "step": 6770
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00020760058881256132,
+      "loss": 0.1597,
+      "step": 6780
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00020745338567222765,
+      "loss": 0.1763,
+      "step": 6790
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.00020730618253189402,
+      "loss": 0.1617,
+      "step": 6800
+    },
+    {
+      "epoch": 9.77,
+      "eval_loss": 0.20486493408679962,
+      "eval_runtime": 39.9706,
+      "eval_samples_per_second": 11.433,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.22195452492094564,
+      "step": 6800
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00020715897939156032,
+      "loss": 0.1613,
+      "step": 6810
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00020701177625122666,
+      "loss": 0.1911,
+      "step": 6820
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.00020686457311089302,
+      "loss": 0.1849,
+      "step": 6830
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00020671736997055938,
+      "loss": 0.1719,
+      "step": 6840
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 0.00020657016683022568,
+      "loss": 0.1817,
+      "step": 6850
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00020642296368989202,
+      "loss": 0.1575,
+      "step": 6860
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00020627576054955838,
+      "loss": 0.1655,
+      "step": 6870
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 0.00020612855740922474,
+      "loss": 0.1773,
+      "step": 6880
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00020598135426889104,
+      "loss": 0.1749,
+      "step": 6890
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00020583415112855738,
+      "loss": 0.1669,
+      "step": 6900
+    },
+    {
+      "epoch": 9.91,
+      "eval_loss": 0.19124703109264374,
+      "eval_runtime": 39.8375,
+      "eval_samples_per_second": 11.472,
+      "eval_steps_per_second": 1.456,
+      "eval_wer": 0.21547959644631834,
+      "step": 6900
+    },
+    {
+      "epoch": 9.93,
+      "learning_rate": 0.00020568694798822374,
+      "loss": 0.1504,
+      "step": 6910
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0002055397448478901,
+      "loss": 0.1713,
+      "step": 6920
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.0002053925417075564,
+      "loss": 0.1773,
+      "step": 6930
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.00020524533856722274,
+      "loss": 0.158,
+      "step": 6940
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.0002050981354268891,
+      "loss": 0.1632,
+      "step": 6950
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.0002049509322865554,
+      "loss": 0.1622,
+      "step": 6960
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00020480372914622176,
+      "loss": 0.1693,
+      "step": 6970
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.0002046565260058881,
+      "loss": 0.1505,
+      "step": 6980
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00020450932286555446,
+      "loss": 0.1477,
+      "step": 6990
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.00020436211972522076,
+      "loss": 0.1504,
+      "step": 7000
+    },
+    {
+      "epoch": 10.06,
+      "eval_loss": 0.1917961686849594,
+      "eval_runtime": 40.0723,
+      "eval_samples_per_second": 11.404,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.21547959644631834,
+      "step": 7000
+    },
+    {
+      "epoch": 10.07,
+      "learning_rate": 0.00020421491658488713,
+      "loss": 0.121,
+      "step": 7010
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.00020406771344455346,
+      "loss": 0.1654,
+      "step": 7020
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00020392051030421982,
+      "loss": 0.1567,
+      "step": 7030
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00020377330716388613,
+      "loss": 0.1748,
+      "step": 7040
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00020362610402355249,
+      "loss": 0.1355,
+      "step": 7050
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.00020347890088321882,
+      "loss": 0.1484,
+      "step": 7060
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 0.00020333169774288518,
+      "loss": 0.1646,
+      "step": 7070
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00020318449460255149,
+      "loss": 0.1906,
+      "step": 7080
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00020303729146221785,
+      "loss": 0.1629,
+      "step": 7090
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00020289008832188418,
+      "loss": 0.1715,
+      "step": 7100
+    },
+    {
+      "epoch": 10.2,
+      "eval_loss": 0.20631931722164154,
+      "eval_runtime": 40.224,
+      "eval_samples_per_second": 11.361,
+      "eval_steps_per_second": 1.442,
+      "eval_wer": 0.2124680018069568,
+      "step": 7100
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.0002027428851815505,
+      "loss": 0.1555,
+      "step": 7110
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00020259568204121685,
+      "loss": 0.1834,
+      "step": 7120
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.0002024484789008832,
+      "loss": 0.1437,
+      "step": 7130
+    },
+    {
+      "epoch": 10.26,
+      "learning_rate": 0.00020230127576054954,
+      "loss": 0.1608,
+      "step": 7140
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00020215407262021587,
+      "loss": 0.159,
+      "step": 7150
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0002020068694798822,
+      "loss": 0.1512,
+      "step": 7160
+    },
+    {
+      "epoch": 10.3,
+      "learning_rate": 0.00020185966633954857,
+      "loss": 0.1701,
+      "step": 7170
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.0002017124631992149,
+      "loss": 0.1324,
+      "step": 7180
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.00020156526005888123,
+      "loss": 0.1629,
+      "step": 7190
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.00020141805691854757,
+      "loss": 0.1604,
+      "step": 7200
+    },
+    {
+      "epoch": 10.34,
+      "eval_loss": 0.1930931955575943,
+      "eval_runtime": 40.1705,
+      "eval_samples_per_second": 11.377,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.21261858153892485,
+      "step": 7200
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00020127085377821393,
+      "loss": 0.1234,
+      "step": 7210
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.00020112365063788026,
+      "loss": 0.1739,
+      "step": 7220
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 0.0002009764474975466,
+      "loss": 0.1482,
+      "step": 7230
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00020082924435721293,
+      "loss": 0.1717,
+      "step": 7240
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.0002006820412168793,
+      "loss": 0.1697,
+      "step": 7250
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.0002005348380765456,
+      "loss": 0.1609,
+      "step": 7260
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00020038763493621196,
+      "loss": 0.1583,
+      "step": 7270
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.0002002404317958783,
+      "loss": 0.1319,
+      "step": 7280
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.00020009322865554465,
+      "loss": 0.1548,
+      "step": 7290
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 0.00019994602551521096,
+      "loss": 0.1296,
+      "step": 7300
+    },
+    {
+      "epoch": 10.49,
+      "eval_loss": 0.1908556967973709,
+      "eval_runtime": 39.9017,
+      "eval_samples_per_second": 11.453,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.21261858153892485,
+      "step": 7300
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.00019979882237487732,
+      "loss": 0.1352,
+      "step": 7310
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00019965161923454365,
+      "loss": 0.1664,
+      "step": 7320
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 0.00019950441609421,
+      "loss": 0.1202,
+      "step": 7330
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00019935721295387632,
+      "loss": 0.1642,
+      "step": 7340
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00019921000981354268,
+      "loss": 0.1413,
+      "step": 7350
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.000199062806673209,
+      "loss": 0.1233,
+      "step": 7360
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.00019891560353287537,
+      "loss": 0.1356,
+      "step": 7370
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00019876840039254168,
+      "loss": 0.1423,
+      "step": 7380
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.00019862119725220804,
+      "loss": 0.1568,
+      "step": 7390
+    },
+    {
+      "epoch": 10.63,
+      "learning_rate": 0.00019847399411187437,
+      "loss": 0.1405,
+      "step": 7400
+    },
+    {
+      "epoch": 10.63,
+      "eval_loss": 0.19520747661590576,
+      "eval_runtime": 39.9029,
+      "eval_samples_per_second": 11.453,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.21066104502333985,
+      "step": 7400
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.0001983267909715407,
+      "loss": 0.1478,
+      "step": 7410
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.00019817958783120704,
+      "loss": 0.1633,
+      "step": 7420
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.0001980323846908734,
+      "loss": 0.1438,
+      "step": 7430
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00019788518155053973,
+      "loss": 0.1668,
+      "step": 7440
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00019773797841020607,
+      "loss": 0.1472,
+      "step": 7450
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 0.0001975907752698724,
+      "loss": 0.1323,
+      "step": 7460
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00019744357212953876,
+      "loss": 0.1607,
+      "step": 7470
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.0001972963689892051,
+      "loss": 0.1649,
+      "step": 7480
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00019714916584887143,
+      "loss": 0.1469,
+      "step": 7490
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00019700196270853776,
+      "loss": 0.1423,
+      "step": 7500
+    },
+    {
+      "epoch": 10.78,
+      "eval_loss": 0.20175614953041077,
+      "eval_runtime": 39.9105,
+      "eval_samples_per_second": 11.451,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.2096069868995633,
+      "step": 7500
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00019685475956820412,
+      "loss": 0.1369,
+      "step": 7510
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00019670755642787043,
+      "loss": 0.1513,
+      "step": 7520
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.0001965603532875368,
+      "loss": 0.1426,
+      "step": 7530
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.00019641315014720312,
+      "loss": 0.182,
+      "step": 7540
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00019626594700686948,
+      "loss": 0.1548,
+      "step": 7550
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 0.0001961187438665358,
+      "loss": 0.1208,
+      "step": 7560
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00019597154072620215,
+      "loss": 0.1596,
+      "step": 7570
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00019582433758586848,
+      "loss": 0.1369,
+      "step": 7580
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 0.00019567713444553484,
+      "loss": 0.1583,
+      "step": 7590
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00019552993130520115,
+      "loss": 0.1434,
+      "step": 7600
+    },
+    {
+      "epoch": 10.92,
+      "eval_loss": 0.18903155624866486,
+      "eval_runtime": 40.197,
+      "eval_samples_per_second": 11.369,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.2064448125282337,
+      "step": 7600
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.0001953827281648675,
+      "loss": 0.1513,
+      "step": 7610
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 0.00019523552502453384,
+      "loss": 0.1791,
+      "step": 7620
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0001950883218842002,
+      "loss": 0.1272,
+      "step": 7630
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.0001949411187438665,
+      "loss": 0.1441,
+      "step": 7640
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00019479391560353287,
+      "loss": 0.1205,
+      "step": 7650
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.0001946467124631992,
+      "loss": 0.151,
+      "step": 7660
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.0001944995093228655,
+      "loss": 0.136,
+      "step": 7670
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.00019435230618253187,
+      "loss": 0.1342,
+      "step": 7680
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00019420510304219823,
+      "loss": 0.1474,
+      "step": 7690
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00019405789990186456,
+      "loss": 0.1439,
+      "step": 7700
+    },
+    {
+      "epoch": 11.06,
+      "eval_loss": 0.1940617859363556,
+      "eval_runtime": 39.8635,
+      "eval_samples_per_second": 11.464,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.20268031922903176,
+      "step": 7700
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00019391069676153087,
+      "loss": 0.1328,
+      "step": 7710
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 0.00019376349362119723,
+      "loss": 0.1136,
+      "step": 7720
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.0001936162904808636,
+      "loss": 0.1119,
+      "step": 7730
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00019346908734052992,
+      "loss": 0.1645,
+      "step": 7740
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00019332188420019623,
+      "loss": 0.1351,
+      "step": 7750
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0001931746810598626,
+      "loss": 0.1427,
+      "step": 7760
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.00019302747791952895,
+      "loss": 0.1949,
+      "step": 7770
+    },
+    {
+      "epoch": 11.18,
+      "learning_rate": 0.00019288027477919528,
+      "loss": 0.1581,
+      "step": 7780
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.0001927330716388616,
+      "loss": 0.1372,
+      "step": 7790
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00019258586849852795,
+      "loss": 0.1415,
+      "step": 7800
+    },
+    {
+      "epoch": 11.21,
+      "eval_loss": 0.18760690093040466,
+      "eval_runtime": 40.0158,
+      "eval_samples_per_second": 11.42,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.20268031922903176,
+      "step": 7800
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.0001924386653581943,
+      "loss": 0.1523,
+      "step": 7810
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00019229146221786062,
+      "loss": 0.1295,
+      "step": 7820
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00019214425907752695,
+      "loss": 0.1288,
+      "step": 7830
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.0001919970559371933,
+      "loss": 0.1567,
+      "step": 7840
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.00019184985279685967,
+      "loss": 0.1315,
+      "step": 7850
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00019170264965652598,
+      "loss": 0.153,
+      "step": 7860
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0001915554465161923,
+      "loss": 0.1413,
+      "step": 7870
+    },
+    {
+      "epoch": 11.32,
+      "learning_rate": 0.00019140824337585867,
+      "loss": 0.1203,
+      "step": 7880
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00019126104023552503,
+      "loss": 0.1459,
+      "step": 7890
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.00019111383709519134,
+      "loss": 0.1289,
+      "step": 7900
+    },
+    {
+      "epoch": 11.35,
+      "eval_loss": 0.18372580409049988,
+      "eval_runtime": 40.1863,
+      "eval_samples_per_second": 11.372,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.20629423279626563,
+      "step": 7900
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.00019096663395485767,
+      "loss": 0.138,
+      "step": 7910
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00019081943081452403,
+      "loss": 0.1593,
+      "step": 7920
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.0001906722276741904,
+      "loss": 0.1145,
+      "step": 7930
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 0.0001905250245338567,
+      "loss": 0.1582,
+      "step": 7940
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00019037782139352303,
+      "loss": 0.1597,
+      "step": 7950
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.0001902306182531894,
+      "loss": 0.1344,
+      "step": 7960
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.0001900834151128557,
+      "loss": 0.1387,
+      "step": 7970
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00018993621197252206,
+      "loss": 0.1321,
+      "step": 7980
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00018980372914622178,
+      "loss": 0.1783,
+      "step": 7990
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00018965652600588809,
+      "loss": 0.1268,
+      "step": 8000
+    },
+    {
+      "epoch": 11.49,
+      "eval_loss": 0.17705726623535156,
+      "eval_runtime": 39.9982,
+      "eval_samples_per_second": 11.426,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.20358379762084022,
+      "step": 8000
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 0.00018950932286555445,
+      "loss": 0.1468,
+      "step": 8010
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.0001893621197252208,
+      "loss": 0.1278,
+      "step": 8020
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00018921491658488714,
+      "loss": 0.1432,
+      "step": 8030
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 0.00018906771344455345,
+      "loss": 0.1383,
+      "step": 8040
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.0001889205103042198,
+      "loss": 0.1267,
+      "step": 8050
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00018877330716388617,
+      "loss": 0.1315,
+      "step": 8060
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00018862610402355247,
+      "loss": 0.1618,
+      "step": 8070
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.0001884789008832188,
+      "loss": 0.144,
+      "step": 8080
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00018833169774288517,
+      "loss": 0.1267,
+      "step": 8090
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.00018818449460255153,
+      "loss": 0.1335,
+      "step": 8100
+    },
+    {
+      "epoch": 11.64,
+      "eval_loss": 0.1754847913980484,
+      "eval_runtime": 39.9481,
+      "eval_samples_per_second": 11.44,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.20087336244541484,
+      "step": 8100
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 0.00018803729146221783,
+      "loss": 0.1241,
+      "step": 8110
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00018789008832188417,
+      "loss": 0.1458,
+      "step": 8120
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.00018774288518155053,
+      "loss": 0.1269,
+      "step": 8130
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.0001875956820412169,
+      "loss": 0.1547,
+      "step": 8140
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.0001874484789008832,
+      "loss": 0.138,
+      "step": 8150
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.00018730127576054953,
+      "loss": 0.1513,
+      "step": 8160
+    },
+    {
+      "epoch": 11.74,
+      "learning_rate": 0.0001871540726202159,
+      "loss": 0.1523,
+      "step": 8170
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00018700686947988225,
+      "loss": 0.1463,
+      "step": 8180
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00018685966633954856,
+      "loss": 0.1541,
+      "step": 8190
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 0.0001867124631992149,
+      "loss": 0.1136,
+      "step": 8200
+    },
+    {
+      "epoch": 11.78,
+      "eval_loss": 0.19131112098693848,
+      "eval_runtime": 40.2824,
+      "eval_samples_per_second": 11.345,
+      "eval_steps_per_second": 1.44,
+      "eval_wer": 0.205089594940521,
+      "step": 8200
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.00018656526005888125,
+      "loss": 0.1186,
+      "step": 8210
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00018641805691854756,
+      "loss": 0.151,
+      "step": 8220
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00018627085377821392,
+      "loss": 0.1311,
+      "step": 8230
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00018612365063788025,
+      "loss": 0.1507,
+      "step": 8240
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0001859764474975466,
+      "loss": 0.1305,
+      "step": 8250
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00018582924435721292,
+      "loss": 0.1331,
+      "step": 8260
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 0.00018568204121687928,
+      "loss": 0.1288,
+      "step": 8270
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.0001855348380765456,
+      "loss": 0.1305,
+      "step": 8280
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00018538763493621197,
+      "loss": 0.1492,
+      "step": 8290
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.00018524043179587828,
+      "loss": 0.1321,
+      "step": 8300
+    },
+    {
+      "epoch": 11.93,
+      "eval_loss": 0.1786966621875763,
+      "eval_runtime": 39.7362,
+      "eval_samples_per_second": 11.501,
+      "eval_steps_per_second": 1.46,
+      "eval_wer": 0.201174521909351,
+      "step": 8300
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00018509322865554464,
+      "loss": 0.1387,
+      "step": 8310
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00018494602551521097,
+      "loss": 0.1183,
+      "step": 8320
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 0.00018479882237487733,
+      "loss": 0.1429,
+      "step": 8330
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00018465161923454364,
+      "loss": 0.142,
+      "step": 8340
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.00018450441609421,
+      "loss": 0.1511,
+      "step": 8350
+    },
+    {
+      "epoch": 12.01,
+      "learning_rate": 0.00018435721295387633,
+      "loss": 0.1594,
+      "step": 8360
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00018421000981354266,
+      "loss": 0.1273,
+      "step": 8370
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.000184062806673209,
+      "loss": 0.1324,
+      "step": 8380
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.00018391560353287536,
+      "loss": 0.1219,
+      "step": 8390
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.0001837684003925417,
+      "loss": 0.1088,
+      "step": 8400
+    },
+    {
+      "epoch": 12.07,
+      "eval_loss": 0.19158530235290527,
+      "eval_runtime": 39.7742,
+      "eval_samples_per_second": 11.49,
+      "eval_steps_per_second": 1.458,
+      "eval_wer": 0.19966872458967022,
+      "step": 8400
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00018362119725220803,
+      "loss": 0.1208,
+      "step": 8410
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00018347399411187436,
+      "loss": 0.1286,
+      "step": 8420
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 0.00018332679097154072,
+      "loss": 0.1394,
+      "step": 8430
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00018317958783120705,
+      "loss": 0.1601,
+      "step": 8440
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00018303238469087339,
+      "loss": 0.1211,
+      "step": 8450
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00018288518155053972,
+      "loss": 0.1519,
+      "step": 8460
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.00018273797841020608,
+      "loss": 0.1015,
+      "step": 8470
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.0001825907752698724,
+      "loss": 0.1214,
+      "step": 8480
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 0.00018244357212953875,
+      "loss": 0.1315,
+      "step": 8490
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00018229636898920508,
+      "loss": 0.1319,
+      "step": 8500
+    },
+    {
+      "epoch": 12.21,
+      "eval_loss": 0.178669735789299,
+      "eval_runtime": 39.4909,
+      "eval_samples_per_second": 11.572,
+      "eval_steps_per_second": 1.469,
+      "eval_wer": 0.20252973949706368,
+      "step": 8500
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00018214916584887144,
+      "loss": 0.1498,
+      "step": 8510
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00018200196270853775,
+      "loss": 0.1089,
+      "step": 8520
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.0001818547595682041,
+      "loss": 0.1212,
+      "step": 8530
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00018170755642787044,
+      "loss": 0.1483,
+      "step": 8540
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.0001815603532875368,
+      "loss": 0.1295,
+      "step": 8550
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.0001814131501472031,
+      "loss": 0.1221,
+      "step": 8560
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00018126594700686947,
+      "loss": 0.1058,
+      "step": 8570
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.0001811187438665358,
+      "loss": 0.1259,
+      "step": 8580
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 0.00018097154072620216,
+      "loss": 0.1267,
+      "step": 8590
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00018082433758586847,
+      "loss": 0.1274,
+      "step": 8600
+    },
+    {
+      "epoch": 12.36,
+      "eval_loss": 0.18227306008338928,
+      "eval_runtime": 40.0342,
+      "eval_samples_per_second": 11.415,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.2044872760126487,
+      "step": 8600
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00018067713444553483,
+      "loss": 0.1318,
+      "step": 8610
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 0.00018052993130520116,
+      "loss": 0.1286,
+      "step": 8620
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.0001803827281648675,
+      "loss": 0.1115,
+      "step": 8630
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00018023552502453383,
+      "loss": 0.1283,
+      "step": 8640
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 0.0001800883218842002,
+      "loss": 0.1297,
+      "step": 8650
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00017994111874386652,
+      "loss": 0.1287,
+      "step": 8660
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00017979391560353286,
+      "loss": 0.1334,
+      "step": 8670
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.0001796467124631992,
+      "loss": 0.1264,
+      "step": 8680
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00017949950932286555,
+      "loss": 0.1287,
+      "step": 8690
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00017935230618253188,
+      "loss": 0.1361,
+      "step": 8700
+    },
+    {
+      "epoch": 12.5,
+      "eval_loss": 0.1859862059354782,
+      "eval_runtime": 39.6559,
+      "eval_samples_per_second": 11.524,
+      "eval_steps_per_second": 1.463,
+      "eval_wer": 0.20749887065201025,
+      "step": 8700
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00017920510304219822,
+      "loss": 0.1718,
+      "step": 8710
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00017905789990186455,
+      "loss": 0.1127,
+      "step": 8720
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.0001789106967615309,
+      "loss": 0.1558,
+      "step": 8730
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00017876349362119724,
+      "loss": 0.1398,
+      "step": 8740
+    },
+    {
+      "epoch": 12.57,
+      "learning_rate": 0.00017861629048086358,
+      "loss": 0.1351,
+      "step": 8750
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.0001784690873405299,
+      "loss": 0.1291,
+      "step": 8760
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00017832188420019627,
+      "loss": 0.1339,
+      "step": 8770
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.00017817468105986258,
+      "loss": 0.152,
+      "step": 8780
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00017802747791952894,
+      "loss": 0.1313,
+      "step": 8790
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00017788027477919527,
+      "loss": 0.1541,
+      "step": 8800
+    },
+    {
+      "epoch": 12.64,
+      "eval_loss": 0.1849740892648697,
+      "eval_runtime": 39.6506,
+      "eval_samples_per_second": 11.526,
+      "eval_steps_per_second": 1.463,
+      "eval_wer": 0.20283089896099984,
+      "step": 8800
+    },
+    {
+      "epoch": 12.66,
+      "learning_rate": 0.00017773307163886163,
+      "loss": 0.1601,
+      "step": 8810
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00017758586849852794,
+      "loss": 0.123,
+      "step": 8820
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0001774386653581943,
+      "loss": 0.134,
+      "step": 8830
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00017729146221786063,
+      "loss": 0.1273,
+      "step": 8840
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.000177144259077527,
+      "loss": 0.1085,
+      "step": 8850
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.0001769970559371933,
+      "loss": 0.1438,
+      "step": 8860
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.00017684985279685966,
+      "loss": 0.1215,
+      "step": 8870
+    },
+    {
+      "epoch": 12.76,
+      "learning_rate": 0.000176702649656526,
+      "loss": 0.1301,
+      "step": 8880
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00017655544651619235,
+      "loss": 0.118,
+      "step": 8890
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00017640824337585866,
+      "loss": 0.0957,
+      "step": 8900
+    },
+    {
+      "epoch": 12.79,
+      "eval_loss": 0.1861564964056015,
+      "eval_runtime": 39.7711,
+      "eval_samples_per_second": 11.491,
+      "eval_steps_per_second": 1.458,
+      "eval_wer": 0.19469959343472368,
+      "step": 8900
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 0.00017626104023552502,
+      "loss": 0.1277,
+      "step": 8910
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.00017611383709519135,
+      "loss": 0.1311,
+      "step": 8920
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00017596663395485766,
+      "loss": 0.1545,
+      "step": 8930
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00017581943081452402,
+      "loss": 0.1182,
+      "step": 8940
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.00017567222767419038,
+      "loss": 0.12,
+      "step": 8950
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00017552502453385671,
+      "loss": 0.1413,
+      "step": 8960
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00017537782139352302,
+      "loss": 0.1131,
+      "step": 8970
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 0.00017523061825318938,
+      "loss": 0.124,
+      "step": 8980
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00017508341511285574,
+      "loss": 0.122,
+      "step": 8990
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.00017493621197252208,
+      "loss": 0.1139,
+      "step": 9000
+    },
+    {
+      "epoch": 12.93,
+      "eval_loss": 0.17861497402191162,
+      "eval_runtime": 39.712,
+      "eval_samples_per_second": 11.508,
+      "eval_steps_per_second": 1.461,
+      "eval_wer": 0.18310495407318175,
+      "step": 9000
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00017478900883218838,
+      "loss": 0.1459,
+      "step": 9010
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00017464180569185474,
+      "loss": 0.1092,
+      "step": 9020
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.0001744946025515211,
+      "loss": 0.1329,
+      "step": 9030
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 0.00017434739941118744,
+      "loss": 0.1047,
+      "step": 9040
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00017420019627085374,
+      "loss": 0.1094,
+      "step": 9050
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.0001740529931305201,
+      "loss": 0.1347,
+      "step": 9060
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 0.00017390578999018646,
+      "loss": 0.1158,
+      "step": 9070
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00017375858684985277,
+      "loss": 0.1443,
+      "step": 9080
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0001736113837095191,
+      "loss": 0.1027,
+      "step": 9090
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00017346418056918546,
+      "loss": 0.107,
+      "step": 9100
+    },
+    {
+      "epoch": 13.07,
+      "eval_loss": 0.18377935886383057,
+      "eval_runtime": 39.9042,
+      "eval_samples_per_second": 11.452,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.1959042312904683,
+      "step": 9100
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00017331697742885182,
+      "loss": 0.1134,
+      "step": 9110
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00017316977428851813,
+      "loss": 0.1044,
+      "step": 9120
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00017302257114818446,
+      "loss": 0.1194,
+      "step": 9130
+    },
+    {
+      "epoch": 13.13,
+      "learning_rate": 0.00017287536800785082,
+      "loss": 0.1139,
+      "step": 9140
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00017272816486751718,
+      "loss": 0.1271,
+      "step": 9150
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.0001725809617271835,
+      "loss": 0.1207,
+      "step": 9160
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00017243375858684982,
+      "loss": 0.117,
+      "step": 9170
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00017228655544651618,
+      "loss": 0.1332,
+      "step": 9180
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00017213935230618255,
+      "loss": 0.1186,
+      "step": 9190
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 0.00017199214916584885,
+      "loss": 0.1103,
+      "step": 9200
+    },
+    {
+      "epoch": 13.22,
+      "eval_loss": 0.1748352199792862,
+      "eval_runtime": 39.9358,
+      "eval_samples_per_second": 11.443,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.18732118656828792,
+      "step": 9200
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00017184494602551518,
+      "loss": 0.1177,
+      "step": 9210
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00017169774288518155,
+      "loss": 0.1022,
+      "step": 9220
+    },
+    {
+      "epoch": 13.26,
+      "learning_rate": 0.00017155053974484785,
+      "loss": 0.1417,
+      "step": 9230
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.0001714033366045142,
+      "loss": 0.1326,
+      "step": 9240
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00017125613346418055,
+      "loss": 0.1207,
+      "step": 9250
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.0001711089303238469,
+      "loss": 0.1132,
+      "step": 9260
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.0001709617271835132,
+      "loss": 0.097,
+      "step": 9270
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00017081452404317957,
+      "loss": 0.1492,
+      "step": 9280
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.0001706673209028459,
+      "loss": 0.1272,
+      "step": 9290
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 0.00017052011776251227,
+      "loss": 0.1274,
+      "step": 9300
+    },
+    {
+      "epoch": 13.36,
+      "eval_loss": 0.17697954177856445,
+      "eval_runtime": 39.8577,
+      "eval_samples_per_second": 11.466,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.19771118807408522,
+      "step": 9300
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00017037291462217857,
+      "loss": 0.1162,
+      "step": 9310
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00017022571148184493,
+      "loss": 0.1054,
+      "step": 9320
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00017007850834151127,
+      "loss": 0.1115,
+      "step": 9330
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00016993130520117763,
+      "loss": 0.103,
+      "step": 9340
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00016978410206084393,
+      "loss": 0.1359,
+      "step": 9350
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 0.0001696368989205103,
+      "loss": 0.1019,
+      "step": 9360
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00016948969578017663,
+      "loss": 0.1107,
+      "step": 9370
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00016934249263984296,
+      "loss": 0.126,
+      "step": 9380
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.0001691952894995093,
+      "loss": 0.1239,
+      "step": 9390
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.00016904808635917565,
+      "loss": 0.1243,
+      "step": 9400
+    },
+    {
+      "epoch": 13.51,
+      "eval_loss": 0.17389988899230957,
+      "eval_runtime": 39.7378,
+      "eval_samples_per_second": 11.5,
+      "eval_steps_per_second": 1.46,
+      "eval_wer": 0.19349495557897908,
+      "step": 9400
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 0.000168900883218842,
+      "loss": 0.1237,
+      "step": 9410
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 0.00016875368007850832,
+      "loss": 0.1058,
+      "step": 9420
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 0.00016860647693817465,
+      "loss": 0.1313,
+      "step": 9430
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 0.00016845927379784102,
+      "loss": 0.1101,
+      "step": 9440
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 0.00016831207065750735,
+      "loss": 0.1158,
+      "step": 9450
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 0.00016816486751717368,
+      "loss": 0.1189,
+      "step": 9460
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 0.00016801766437684002,
+      "loss": 0.1178,
+      "step": 9470
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 0.00016787046123650638,
+      "loss": 0.1273,
+      "step": 9480
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 0.0001677232580961727,
+      "loss": 0.1302,
+      "step": 9490
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 0.00016757605495583904,
+      "loss": 0.0975,
+      "step": 9500
+    },
+    {
+      "epoch": 13.65,
+      "eval_loss": 0.17656183242797852,
+      "eval_runtime": 39.8815,
+      "eval_samples_per_second": 11.459,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.19816292726998946,
+      "step": 9500
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 0.00016742885181550538,
+      "loss": 0.1057,
+      "step": 9510
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 0.00016728164867517174,
+      "loss": 0.1182,
+      "step": 9520
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 0.00016713444553483804,
+      "loss": 0.111,
+      "step": 9530
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 0.0001669872423945044,
+      "loss": 0.128,
+      "step": 9540
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 0.00016684003925417074,
+      "loss": 0.095,
+      "step": 9550
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 0.0001666928361138371,
+      "loss": 0.1286,
+      "step": 9560
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 0.0001665456329735034,
+      "loss": 0.1028,
+      "step": 9570
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 0.00016639842983316976,
+      "loss": 0.1336,
+      "step": 9580
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 0.0001662512266928361,
+      "loss": 0.095,
+      "step": 9590
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 0.00016610402355250246,
+      "loss": 0.1093,
+      "step": 9600
+    },
+    {
+      "epoch": 13.79,
+      "eval_loss": 0.1793922334909439,
+      "eval_runtime": 40.0085,
+      "eval_samples_per_second": 11.423,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.1959042312904683,
+      "step": 9600
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 0.00016595682041216876,
+      "loss": 0.1257,
+      "step": 9610
+    },
+    {
+      "epoch": 13.82,
+      "learning_rate": 0.00016580961727183512,
+      "loss": 0.0955,
+      "step": 9620
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 0.00016566241413150146,
+      "loss": 0.1234,
+      "step": 9630
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 0.00016551521099116782,
+      "loss": 0.0953,
+      "step": 9640
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 0.00016536800785083412,
+      "loss": 0.1147,
+      "step": 9650
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 0.00016522080471050049,
+      "loss": 0.1232,
+      "step": 9660
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 0.00016507360157016682,
+      "loss": 0.1264,
+      "step": 9670
+    },
+    {
+      "epoch": 13.91,
+      "learning_rate": 0.00016492639842983315,
+      "loss": 0.1401,
+      "step": 9680
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 0.00016477919528949949,
+      "loss": 0.1071,
+      "step": 9690
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 0.00016463199214916585,
+      "loss": 0.1123,
+      "step": 9700
+    },
+    {
+      "epoch": 13.94,
+      "eval_loss": 0.16937404870986938,
+      "eval_runtime": 40.1681,
+      "eval_samples_per_second": 11.377,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.193344375847011,
+      "step": 9700
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 0.00016448478900883218,
+      "loss": 0.13,
+      "step": 9710
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 0.0001643375858684985,
+      "loss": 0.1109,
+      "step": 9720
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 0.00016419038272816485,
+      "loss": 0.1241,
+      "step": 9730
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 0.0001640431795878312,
+      "loss": 0.115,
+      "step": 9740
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 0.00016389597644749754,
+      "loss": 0.11,
+      "step": 9750
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 0.00016374877330716387,
+      "loss": 0.1115,
+      "step": 9760
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 0.0001636015701668302,
+      "loss": 0.0945,
+      "step": 9770
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 0.00016345436702649657,
+      "loss": 0.1122,
+      "step": 9780
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 0.0001633071638861629,
+      "loss": 0.0999,
+      "step": 9790
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 0.00016315996074582923,
+      "loss": 0.1401,
+      "step": 9800
+    },
+    {
+      "epoch": 14.08,
+      "eval_loss": 0.17904061079025269,
+      "eval_runtime": 39.9612,
+      "eval_samples_per_second": 11.436,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.19093510013552176,
+      "step": 9800
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 0.00016301275760549557,
+      "loss": 0.1179,
+      "step": 9810
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 0.00016286555446516193,
+      "loss": 0.1311,
+      "step": 9820
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 0.00016271835132482823,
+      "loss": 0.1111,
+      "step": 9830
+    },
+    {
+      "epoch": 14.14,
+      "learning_rate": 0.0001625711481844946,
+      "loss": 0.1112,
+      "step": 9840
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 0.00016242394504416093,
+      "loss": 0.1189,
+      "step": 9850
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 0.0001622767419038273,
+      "loss": 0.1143,
+      "step": 9860
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 0.0001621295387634936,
+      "loss": 0.1192,
+      "step": 9870
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 0.00016198233562315996,
+      "loss": 0.1224,
+      "step": 9880
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 0.0001618351324828263,
+      "loss": 0.1,
+      "step": 9890
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 0.00016168792934249265,
+      "loss": 0.1255,
+      "step": 9900
+    },
+    {
+      "epoch": 14.22,
+      "eval_loss": 0.17239724099636078,
+      "eval_runtime": 40.3402,
+      "eval_samples_per_second": 11.329,
+      "eval_steps_per_second": 1.438,
+      "eval_wer": 0.18822466496009638,
+      "step": 9900
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 0.00016154072620215896,
+      "loss": 0.0956,
+      "step": 9910
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 0.00016139352306182532,
+      "loss": 0.1031,
+      "step": 9920
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 0.00016124631992149165,
+      "loss": 0.1059,
+      "step": 9930
+    },
+    {
+      "epoch": 14.28,
+      "learning_rate": 0.000161099116781158,
+      "loss": 0.1004,
+      "step": 9940
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 0.00016095191364082432,
+      "loss": 0.1064,
+      "step": 9950
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 0.00016080471050049065,
+      "loss": 0.1167,
+      "step": 9960
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 0.000160657507360157,
+      "loss": 0.1056,
+      "step": 9970
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 0.00016051030421982332,
+      "loss": 0.128,
+      "step": 9980
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 0.00016036310107948968,
+      "loss": 0.1267,
+      "step": 9990
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 0.000160215897939156,
+      "loss": 0.1126,
+      "step": 10000
+    },
+    {
+      "epoch": 14.37,
+      "eval_loss": 0.1660313904285431,
+      "eval_runtime": 40.0416,
+      "eval_samples_per_second": 11.413,
+      "eval_steps_per_second": 1.448,
+      "eval_wer": 0.18882698388796867,
+      "step": 10000
+    },
+    {
+      "epoch": 14.38,
+      "learning_rate": 0.00016006869479882237,
+      "loss": 0.0972,
+      "step": 10010
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 0.00015992149165848868,
+      "loss": 0.0785,
+      "step": 10020
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 0.00015977428851815504,
+      "loss": 0.1162,
+      "step": 10030
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 0.00015962708537782137,
+      "loss": 0.1037,
+      "step": 10040
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 0.00015947988223748773,
+      "loss": 0.1261,
+      "step": 10050
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 0.00015933267909715404,
+      "loss": 0.1,
+      "step": 10060
+    },
+    {
+      "epoch": 14.47,
+      "learning_rate": 0.0001591854759568204,
+      "loss": 0.1175,
+      "step": 10070
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 0.00015903827281648673,
+      "loss": 0.1218,
+      "step": 10080
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 0.0001588910696761531,
+      "loss": 0.0957,
+      "step": 10090
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 0.0001587438665358194,
+      "loss": 0.1241,
+      "step": 10100
+    },
+    {
+      "epoch": 14.51,
+      "eval_loss": 0.16984562575817108,
+      "eval_runtime": 40.0158,
+      "eval_samples_per_second": 11.42,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.19048336093961754,
+      "step": 10100
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 0.00015859666339548576,
+      "loss": 0.0949,
+      "step": 10110
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 0.0001584494602551521,
+      "loss": 0.1353,
+      "step": 10120
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 0.00015830225711481843,
+      "loss": 0.1164,
+      "step": 10130
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 0.00015815505397448476,
+      "loss": 0.0954,
+      "step": 10140
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 0.00015800785083415112,
+      "loss": 0.1477,
+      "step": 10150
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 0.00015786064769381745,
+      "loss": 0.1094,
+      "step": 10160
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 0.0001577134445534838,
+      "loss": 0.0973,
+      "step": 10170
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 0.00015756624141315012,
+      "loss": 0.1124,
+      "step": 10180
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 0.00015741903827281648,
+      "loss": 0.1131,
+      "step": 10190
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 0.00015727183513248281,
+      "loss": 0.1147,
+      "step": 10200
+    },
+    {
+      "epoch": 14.66,
+      "eval_loss": 0.16483080387115479,
+      "eval_runtime": 39.9302,
+      "eval_samples_per_second": 11.445,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.1802439391657883,
+      "step": 10200
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 0.00015712463199214915,
+      "loss": 0.1111,
+      "step": 10210
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 0.00015697742885181548,
+      "loss": 0.1048,
+      "step": 10220
+    },
+    {
+      "epoch": 14.7,
+      "learning_rate": 0.00015683022571148184,
+      "loss": 0.1285,
+      "step": 10230
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 0.00015668302257114817,
+      "loss": 0.092,
+      "step": 10240
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 0.0001565358194308145,
+      "loss": 0.1086,
+      "step": 10250
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 0.00015638861629048084,
+      "loss": 0.1002,
+      "step": 10260
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 0.0001562414131501472,
+      "loss": 0.1148,
+      "step": 10270
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 0.0001560942100098135,
+      "loss": 0.1267,
+      "step": 10280
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 0.00015594700686947987,
+      "loss": 0.1025,
+      "step": 10290
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 0.0001557998037291462,
+      "loss": 0.115,
+      "step": 10300
+    },
+    {
+      "epoch": 14.8,
+      "eval_loss": 0.1692885160446167,
+      "eval_runtime": 40.0262,
+      "eval_samples_per_second": 11.418,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.18581538924860713,
+      "step": 10300
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 0.00015565260058881256,
+      "loss": 0.0927,
+      "step": 10310
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 0.00015550539744847887,
+      "loss": 0.1171,
+      "step": 10320
+    },
+    {
+      "epoch": 14.84,
+      "learning_rate": 0.00015535819430814523,
+      "loss": 0.1267,
+      "step": 10330
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 0.00015521099116781156,
+      "loss": 0.1005,
+      "step": 10340
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 0.00015506378802747792,
+      "loss": 0.117,
+      "step": 10350
+    },
+    {
+      "epoch": 14.89,
+      "learning_rate": 0.00015491658488714423,
+      "loss": 0.0975,
+      "step": 10360
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 0.0001547693817468106,
+      "loss": 0.1131,
+      "step": 10370
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 0.00015462217860647692,
+      "loss": 0.1209,
+      "step": 10380
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 0.00015447497546614328,
+      "loss": 0.105,
+      "step": 10390
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 0.0001543277723258096,
+      "loss": 0.1078,
+      "step": 10400
+    },
+    {
+      "epoch": 14.94,
+      "eval_loss": 0.16425766050815582,
+      "eval_runtime": 39.9939,
+      "eval_samples_per_second": 11.427,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.18385785273302213,
+      "step": 10400
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 0.00015418056918547595,
+      "loss": 0.1102,
+      "step": 10410
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 0.00015403336604514228,
+      "loss": 0.1,
+      "step": 10420
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 0.00015388616290480862,
+      "loss": 0.0944,
+      "step": 10430
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 0.00015373895976447495,
+      "loss": 0.1232,
+      "step": 10440
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 0.0001535917566241413,
+      "loss": 0.1017,
+      "step": 10450
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 0.00015344455348380764,
+      "loss": 0.0991,
+      "step": 10460
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 0.00015329735034347398,
+      "loss": 0.0986,
+      "step": 10470
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 0.0001531501472031403,
+      "loss": 0.1012,
+      "step": 10480
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 0.00015300294406280667,
+      "loss": 0.0992,
+      "step": 10490
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 0.000152855740922473,
+      "loss": 0.107,
+      "step": 10500
+    },
+    {
+      "epoch": 15.09,
+      "eval_loss": 0.169921875,
+      "eval_runtime": 39.8342,
+      "eval_samples_per_second": 11.473,
+      "eval_steps_per_second": 1.456,
+      "eval_wer": 0.17964162023791597,
+      "step": 10500
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 0.00015270853778213934,
+      "loss": 0.0901,
+      "step": 10510
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 0.00015256133464180567,
+      "loss": 0.1056,
+      "step": 10520
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 0.00015241413150147203,
+      "loss": 0.1044,
+      "step": 10530
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 0.00015226692836113837,
+      "loss": 0.0921,
+      "step": 10540
+    },
+    {
+      "epoch": 15.16,
+      "learning_rate": 0.0001521197252208047,
+      "loss": 0.1146,
+      "step": 10550
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 0.00015197252208047103,
+      "loss": 0.0826,
+      "step": 10560
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 0.0001518253189401374,
+      "loss": 0.1078,
+      "step": 10570
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 0.0001516781157998037,
+      "loss": 0.1036,
+      "step": 10580
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 0.00015153091265947006,
+      "loss": 0.0755,
+      "step": 10590
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 0.0001513837095191364,
+      "loss": 0.0966,
+      "step": 10600
+    },
+    {
+      "epoch": 15.23,
+      "eval_loss": 0.1691877841949463,
+      "eval_runtime": 40.0386,
+      "eval_samples_per_second": 11.414,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.1802439391657883,
+      "step": 10600
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 0.00015123650637880275,
+      "loss": 0.087,
+      "step": 10610
+    },
+    {
+      "epoch": 15.26,
+      "learning_rate": 0.00015108930323846906,
+      "loss": 0.1237,
+      "step": 10620
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 0.00015094210009813542,
+      "loss": 0.087,
+      "step": 10630
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 0.00015079489695780175,
+      "loss": 0.0968,
+      "step": 10640
+    },
+    {
+      "epoch": 15.3,
+      "learning_rate": 0.00015064769381746811,
+      "loss": 0.1188,
+      "step": 10650
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 0.00015050049067713442,
+      "loss": 0.0945,
+      "step": 10660
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 0.00015035328753680078,
+      "loss": 0.1007,
+      "step": 10670
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 0.00015020608439646711,
+      "loss": 0.0988,
+      "step": 10680
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 0.00015005888125613348,
+      "loss": 0.0876,
+      "step": 10690
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 0.00014991167811579978,
+      "loss": 0.0931,
+      "step": 10700
+    },
+    {
+      "epoch": 15.37,
+      "eval_loss": 0.16350314021110535,
+      "eval_runtime": 39.8693,
+      "eval_samples_per_second": 11.462,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.17753350399036288,
+      "step": 10700
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 0.00014976447497546614,
+      "loss": 0.0825,
+      "step": 10710
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 0.00014961727183513248,
+      "loss": 0.1109,
+      "step": 10720
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 0.0001494700686947988,
+      "loss": 0.095,
+      "step": 10730
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 0.00014932286555446514,
+      "loss": 0.1073,
+      "step": 10740
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 0.0001491756624141315,
+      "loss": 0.111,
+      "step": 10750
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 0.00014902845927379784,
+      "loss": 0.0962,
+      "step": 10760
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 0.00014888125613346417,
+      "loss": 0.096,
+      "step": 10770
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 0.0001487340529931305,
+      "loss": 0.0717,
+      "step": 10780
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 0.00014858684985279686,
+      "loss": 0.0893,
+      "step": 10790
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 0.0001484396467124632,
+      "loss": 0.1068,
+      "step": 10800
+    },
+    {
+      "epoch": 15.52,
+      "eval_loss": 0.16544069349765778,
+      "eval_runtime": 39.8983,
+      "eval_samples_per_second": 11.454,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.17407017015509713,
+      "step": 10800
+    },
+    {
+      "epoch": 15.53,
+      "learning_rate": 0.00014829244357212953,
+      "loss": 0.0968,
+      "step": 10810
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 0.00014814524043179586,
+      "loss": 0.0907,
+      "step": 10820
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 0.00014799803729146222,
+      "loss": 0.0836,
+      "step": 10830
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 0.00014785083415112856,
+      "loss": 0.0866,
+      "step": 10840
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 0.00014770363101079486,
+      "loss": 0.1004,
+      "step": 10850
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 0.00014755642787046122,
+      "loss": 0.0972,
+      "step": 10860
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 0.00014740922473012756,
+      "loss": 0.1118,
+      "step": 10870
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 0.00014726202158979392,
+      "loss": 0.1021,
+      "step": 10880
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 0.00014711481844946022,
+      "loss": 0.0791,
+      "step": 10890
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 0.00014696761530912658,
+      "loss": 0.1224,
+      "step": 10900
+    },
+    {
+      "epoch": 15.66,
+      "eval_loss": 0.16307871043682098,
+      "eval_runtime": 40.0248,
+      "eval_samples_per_second": 11.418,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.1767806053305225,
+      "step": 10900
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 0.00014682041216879292,
+      "loss": 0.0825,
+      "step": 10910
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 0.00014667320902845928,
+      "loss": 0.0958,
+      "step": 10920
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 0.00014652600588812558,
+      "loss": 0.112,
+      "step": 10930
+    },
+    {
+      "epoch": 15.72,
+      "learning_rate": 0.00014637880274779195,
+      "loss": 0.0934,
+      "step": 10940
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 0.00014623159960745828,
+      "loss": 0.0917,
+      "step": 10950
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 0.00014608439646712464,
+      "loss": 0.0813,
+      "step": 10960
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 0.00014593719332679095,
+      "loss": 0.1391,
+      "step": 10970
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 0.0001457899901864573,
+      "loss": 0.1232,
+      "step": 10980
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 0.00014564278704612364,
+      "loss": 0.1132,
+      "step": 10990
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 0.00014549558390578997,
+      "loss": 0.0999,
+      "step": 11000
+    },
+    {
+      "epoch": 15.8,
+      "eval_loss": 0.15900734066963196,
+      "eval_runtime": 40.0131,
+      "eval_samples_per_second": 11.421,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.17663002559855442,
+      "step": 11000
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 0.0001453483807654563,
+      "loss": 0.079,
+      "step": 11010
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 0.00014520117762512264,
+      "loss": 0.102,
+      "step": 11020
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 0.000145053974484789,
+      "loss": 0.088,
+      "step": 11030
+    },
+    {
+      "epoch": 15.86,
+      "learning_rate": 0.00014490677134445533,
+      "loss": 0.0976,
+      "step": 11040
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 0.00014475956820412167,
+      "loss": 0.1118,
+      "step": 11050
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 0.000144612365063788,
+      "loss": 0.0847,
+      "step": 11060
+    },
+    {
+      "epoch": 15.91,
+      "learning_rate": 0.00014446516192345436,
+      "loss": 0.0962,
+      "step": 11070
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 0.0001443179587831207,
+      "loss": 0.1053,
+      "step": 11080
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 0.00014417075564278703,
+      "loss": 0.1131,
+      "step": 11090
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 0.00014402355250245336,
+      "loss": 0.0879,
+      "step": 11100
+    },
+    {
+      "epoch": 15.95,
+      "eval_loss": 0.16153839230537415,
+      "eval_runtime": 40.5046,
+      "eval_samples_per_second": 11.283,
+      "eval_steps_per_second": 1.432,
+      "eval_wer": 0.17196205390754404,
+      "step": 11100
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 0.00014387634936211972,
+      "loss": 0.1013,
+      "step": 11110
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 0.00014372914622178605,
+      "loss": 0.1083,
+      "step": 11120
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 0.0001435819430814524,
+      "loss": 0.1037,
+      "step": 11130
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 0.00014343473994111872,
+      "loss": 0.0954,
+      "step": 11140
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 0.00014328753680078506,
+      "loss": 0.0845,
+      "step": 11150
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 0.00014314033366045142,
+      "loss": 0.0827,
+      "step": 11160
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 0.00014299313052011775,
+      "loss": 0.1113,
+      "step": 11170
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 0.00014284592737978408,
+      "loss": 0.0934,
+      "step": 11180
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 0.00014269872423945042,
+      "loss": 0.0903,
+      "step": 11190
+    },
+    {
+      "epoch": 16.09,
+      "learning_rate": 0.00014255152109911678,
+      "loss": 0.0761,
+      "step": 11200
+    },
+    {
+      "epoch": 16.09,
+      "eval_loss": 0.16535413265228271,
+      "eval_runtime": 40.0395,
+      "eval_samples_per_second": 11.414,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.17497364854690559,
+      "step": 11200
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 0.0001424043179587831,
+      "loss": 0.0709,
+      "step": 11210
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 0.00014225711481844944,
+      "loss": 0.1006,
+      "step": 11220
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 0.00014210991167811578,
+      "loss": 0.1218,
+      "step": 11230
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 0.00014196270853778214,
+      "loss": 0.0994,
+      "step": 11240
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 0.00014181550539744847,
+      "loss": 0.0986,
+      "step": 11250
+    },
+    {
+      "epoch": 16.18,
+      "learning_rate": 0.0001416683022571148,
+      "loss": 0.0789,
+      "step": 11260
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 0.00014152109911678114,
+      "loss": 0.1026,
+      "step": 11270
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 0.0001413738959764475,
+      "loss": 0.0982,
+      "step": 11280
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 0.00014122669283611383,
+      "loss": 0.1112,
+      "step": 11290
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 0.00014107948969578016,
+      "loss": 0.0821,
+      "step": 11300
+    },
+    {
+      "epoch": 16.24,
+      "eval_loss": 0.15813738107681274,
+      "eval_runtime": 39.8582,
+      "eval_samples_per_second": 11.466,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.17452190935100134,
+      "step": 11300
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 0.0001409322865554465,
+      "loss": 0.0935,
+      "step": 11310
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 0.00014078508341511283,
+      "loss": 0.1029,
+      "step": 11320
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 0.0001406378802747792,
+      "loss": 0.1002,
+      "step": 11330
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 0.00014049067713444553,
+      "loss": 0.095,
+      "step": 11340
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 0.00014034347399411186,
+      "loss": 0.1028,
+      "step": 11350
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 0.0001401962708537782,
+      "loss": 0.079,
+      "step": 11360
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 0.00014004906771344455,
+      "loss": 0.0941,
+      "step": 11370
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 0.00013990186457311089,
+      "loss": 0.0839,
+      "step": 11380
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 0.00013975466143277722,
+      "loss": 0.0976,
+      "step": 11390
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 0.00013960745829244355,
+      "loss": 0.0984,
+      "step": 11400
+    },
+    {
+      "epoch": 16.38,
+      "eval_loss": 0.16124248504638672,
+      "eval_runtime": 40.0383,
+      "eval_samples_per_second": 11.414,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.16940219846408675,
+      "step": 11400
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 0.0001394602551521099,
+      "loss": 0.0889,
+      "step": 11410
+    },
+    {
+      "epoch": 16.41,
+      "learning_rate": 0.00013931305201177625,
+      "loss": 0.0951,
+      "step": 11420
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 0.00013916584887144258,
+      "loss": 0.075,
+      "step": 11430
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 0.0001390186457311089,
+      "loss": 0.1025,
+      "step": 11440
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 0.00013887144259077525,
+      "loss": 0.0838,
+      "step": 11450
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 0.0001387242394504416,
+      "loss": 0.093,
+      "step": 11460
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 0.00013857703631010794,
+      "loss": 0.1044,
+      "step": 11470
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 0.00013842983316977427,
+      "loss": 0.0944,
+      "step": 11480
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 0.0001382826300294406,
+      "loss": 0.0967,
+      "step": 11490
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 0.00013813542688910697,
+      "loss": 0.0815,
+      "step": 11500
+    },
+    {
+      "epoch": 16.52,
+      "eval_loss": 0.16426511108875275,
+      "eval_runtime": 39.87,
+      "eval_samples_per_second": 11.462,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.17361843095919288,
+      "step": 11500
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 0.0001379882237487733,
+      "loss": 0.1004,
+      "step": 11510
+    },
+    {
+      "epoch": 16.55,
+      "learning_rate": 0.00013784102060843963,
+      "loss": 0.0916,
+      "step": 11520
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 0.00013769381746810597,
+      "loss": 0.0947,
+      "step": 11530
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 0.00013754661432777233,
+      "loss": 0.0905,
+      "step": 11540
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 0.00013739941118743866,
+      "loss": 0.0864,
+      "step": 11550
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 0.000137252208047105,
+      "loss": 0.0852,
+      "step": 11560
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 0.00013710500490677133,
+      "loss": 0.1071,
+      "step": 11570
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 0.0001369578017664377,
+      "loss": 0.0939,
+      "step": 11580
+    },
+    {
+      "epoch": 16.65,
+      "learning_rate": 0.00013681059862610402,
+      "loss": 0.1154,
+      "step": 11590
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 0.00013666339548577036,
+      "loss": 0.0898,
+      "step": 11600
+    },
+    {
+      "epoch": 16.67,
+      "eval_loss": 0.16324785351753235,
+      "eval_runtime": 39.9344,
+      "eval_samples_per_second": 11.444,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.17723234452642675,
+      "step": 11600
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 0.0001365161923454367,
+      "loss": 0.0818,
+      "step": 11610
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 0.00013636898920510302,
+      "loss": 0.0951,
+      "step": 11620
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 0.00013622178606476938,
+      "loss": 0.1037,
+      "step": 11630
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 0.00013607458292443572,
+      "loss": 0.0855,
+      "step": 11640
+    },
+    {
+      "epoch": 16.74,
+      "learning_rate": 0.00013592737978410205,
+      "loss": 0.0795,
+      "step": 11650
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 0.00013578017664376838,
+      "loss": 0.0877,
+      "step": 11660
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 0.00013563297350343474,
+      "loss": 0.1126,
+      "step": 11670
+    },
+    {
+      "epoch": 16.78,
+      "learning_rate": 0.00013548577036310108,
+      "loss": 0.0809,
+      "step": 11680
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 0.0001353385672227674,
+      "loss": 0.1051,
+      "step": 11690
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 0.00013519136408243374,
+      "loss": 0.0958,
+      "step": 11700
+    },
+    {
+      "epoch": 16.81,
+      "eval_loss": 0.159054696559906,
+      "eval_runtime": 40.0796,
+      "eval_samples_per_second": 11.402,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.17467248908296942,
+      "step": 11700
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 0.0001350441609421001,
+      "loss": 0.0972,
+      "step": 11710
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 0.00013489695780176644,
+      "loss": 0.0908,
+      "step": 11720
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 0.00013474975466143274,
+      "loss": 0.1005,
+      "step": 11730
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 0.0001346025515210991,
+      "loss": 0.1101,
+      "step": 11740
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 0.00013445534838076544,
+      "loss": 0.1086,
+      "step": 11750
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 0.0001343081452404318,
+      "loss": 0.0864,
+      "step": 11760
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 0.0001341609421000981,
+      "loss": 0.1002,
+      "step": 11770
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 0.00013401373895976447,
+      "loss": 0.1145,
+      "step": 11780
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 0.0001338665358194308,
+      "loss": 0.0848,
+      "step": 11790
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 0.00013371933267909716,
+      "loss": 0.0986,
+      "step": 11800
+    },
+    {
+      "epoch": 16.95,
+      "eval_loss": 0.16650956869125366,
+      "eval_runtime": 40.0741,
+      "eval_samples_per_second": 11.404,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.17331727149525675,
+      "step": 11800
+    },
+    {
+      "epoch": 16.97,
+      "learning_rate": 0.00013357212953876347,
+      "loss": 0.0942,
+      "step": 11810
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 0.00013342492639842983,
+      "loss": 0.0868,
+      "step": 11820
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 0.00013327772325809616,
+      "loss": 0.0857,
+      "step": 11830
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 0.00013313052011776252,
+      "loss": 0.0911,
+      "step": 11840
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 0.00013298331697742883,
+      "loss": 0.0743,
+      "step": 11850
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 0.0001328361138370952,
+      "loss": 0.0727,
+      "step": 11860
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 0.00013268891069676152,
+      "loss": 0.0905,
+      "step": 11870
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 0.00013254170755642785,
+      "loss": 0.0724,
+      "step": 11880
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 0.0001323945044160942,
+      "loss": 0.0872,
+      "step": 11890
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 0.00013224730127576052,
+      "loss": 0.0843,
+      "step": 11900
+    },
+    {
+      "epoch": 17.1,
+      "eval_loss": 0.16180233657360077,
+      "eval_runtime": 40.2757,
+      "eval_samples_per_second": 11.347,
+      "eval_steps_per_second": 1.44,
+      "eval_wer": 0.17256437283541634,
+      "step": 11900
+    },
+    {
+      "epoch": 17.11,
+      "learning_rate": 0.00013210009813542688,
+      "loss": 0.1044,
+      "step": 11910
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 0.00013195289499509321,
+      "loss": 0.1075,
+      "step": 11920
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 0.00013180569185475955,
+      "loss": 0.0632,
+      "step": 11930
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 0.00013165848871442588,
+      "loss": 0.1181,
+      "step": 11940
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 0.00013151128557409224,
+      "loss": 0.0798,
+      "step": 11950
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 0.00013136408243375857,
+      "loss": 0.0976,
+      "step": 11960
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 0.0001312168792934249,
+      "loss": 0.1028,
+      "step": 11970
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 0.00013106967615309124,
+      "loss": 0.0819,
+      "step": 11980
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 0.0001309224730127576,
+      "loss": 0.0977,
+      "step": 11990
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 0.00013077526987242394,
+      "loss": 0.0888,
+      "step": 12000
+    },
+    {
+      "epoch": 17.24,
+      "eval_loss": 0.1597016453742981,
+      "eval_runtime": 40.0997,
+      "eval_samples_per_second": 11.397,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.16985393765999096,
+      "step": 12000
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 0.00013062806673209027,
+      "loss": 0.0903,
+      "step": 12010
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 0.0001304808635917566,
+      "loss": 0.0879,
+      "step": 12020
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 0.00013033366045142294,
+      "loss": 0.0902,
+      "step": 12030
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 0.0001301864573110893,
+      "loss": 0.0922,
+      "step": 12040
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 0.00013003925417075563,
+      "loss": 0.0705,
+      "step": 12050
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 0.00012989205103042196,
+      "loss": 0.0861,
+      "step": 12060
+    },
+    {
+      "epoch": 17.34,
+      "learning_rate": 0.0001297448478900883,
+      "loss": 0.0923,
+      "step": 12070
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 0.00012959764474975466,
+      "loss": 0.0776,
+      "step": 12080
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 0.000129450441609421,
+      "loss": 0.1092,
+      "step": 12090
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 0.00012930323846908732,
+      "loss": 0.0851,
+      "step": 12100
+    },
+    {
+      "epoch": 17.39,
+      "eval_loss": 0.15801562368869781,
+      "eval_runtime": 40.1138,
+      "eval_samples_per_second": 11.393,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.1704562565878633,
+      "step": 12100
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 0.00012915603532875366,
+      "loss": 0.0997,
+      "step": 12110
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 0.00012900883218842002,
+      "loss": 0.0931,
+      "step": 12120
+    },
+    {
+      "epoch": 17.43,
+      "learning_rate": 0.00012886162904808635,
+      "loss": 0.0745,
+      "step": 12130
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 0.00012871442590775268,
+      "loss": 0.0965,
+      "step": 12140
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 0.00012856722276741902,
+      "loss": 0.0909,
+      "step": 12150
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 0.00012842001962708538,
+      "loss": 0.1066,
+      "step": 12160
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 0.0001282728164867517,
+      "loss": 0.0924,
+      "step": 12170
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 0.00012812561334641804,
+      "loss": 0.0757,
+      "step": 12180
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 0.00012797841020608438,
+      "loss": 0.0849,
+      "step": 12190
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 0.0001278459273797841,
+      "loss": 0.0928,
+      "step": 12200
+    },
+    {
+      "epoch": 17.53,
+      "eval_loss": 0.16045060753822327,
+      "eval_runtime": 40.0909,
+      "eval_samples_per_second": 11.399,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.1695527781960548,
+      "step": 12200
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 0.00012769872423945043,
+      "loss": 0.1008,
+      "step": 12210
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 0.00012755152109911676,
+      "loss": 0.1008,
+      "step": 12220
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 0.0001274043179587831,
+      "loss": 0.0863,
+      "step": 12230
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 0.00012725711481844946,
+      "loss": 0.0996,
+      "step": 12240
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 0.0001271099116781158,
+      "loss": 0.067,
+      "step": 12250
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 0.00012696270853778212,
+      "loss": 0.0988,
+      "step": 12260
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 0.00012681550539744846,
+      "loss": 0.0883,
+      "step": 12270
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 0.0001266683022571148,
+      "loss": 0.0811,
+      "step": 12280
+    },
+    {
+      "epoch": 17.66,
+      "learning_rate": 0.00012652109911678115,
+      "loss": 0.0999,
+      "step": 12290
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 0.00012637389597644749,
+      "loss": 0.0886,
+      "step": 12300
+    },
+    {
+      "epoch": 17.67,
+      "eval_loss": 0.1619402915239334,
+      "eval_runtime": 40.3366,
+      "eval_samples_per_second": 11.33,
+      "eval_steps_per_second": 1.438,
+      "eval_wer": 0.16940219846408675,
+      "step": 12300
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 0.00012622669283611382,
+      "loss": 0.0936,
+      "step": 12310
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 0.00012607948969578015,
+      "loss": 0.1139,
+      "step": 12320
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 0.0001259322865554465,
+      "loss": 0.0851,
+      "step": 12330
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 0.00012578508341511285,
+      "loss": 0.1101,
+      "step": 12340
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 0.00012563788027477918,
+      "loss": 0.0792,
+      "step": 12350
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 0.0001254906771344455,
+      "loss": 0.0789,
+      "step": 12360
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 0.00012534347399411187,
+      "loss": 0.1043,
+      "step": 12370
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 0.0001251962708537782,
+      "loss": 0.0701,
+      "step": 12380
+    },
+    {
+      "epoch": 17.8,
+      "learning_rate": 0.00012504906771344454,
+      "loss": 0.0817,
+      "step": 12390
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 0.00012490186457311087,
+      "loss": 0.0931,
+      "step": 12400
+    },
+    {
+      "epoch": 17.82,
+      "eval_loss": 0.15292048454284668,
+      "eval_runtime": 40.0068,
+      "eval_samples_per_second": 11.423,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.16819756060834212,
+      "step": 12400
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 0.0001247546614327772,
+      "loss": 0.1021,
+      "step": 12410
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 0.00012460745829244357,
+      "loss": 0.0851,
+      "step": 12420
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 0.0001244602551521099,
+      "loss": 0.07,
+      "step": 12430
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 0.00012431305201177623,
+      "loss": 0.108,
+      "step": 12440
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 0.00012416584887144257,
+      "loss": 0.0826,
+      "step": 12450
+    },
+    {
+      "epoch": 17.9,
+      "learning_rate": 0.00012401864573110893,
+      "loss": 0.0834,
+      "step": 12460
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 0.00012387144259077526,
+      "loss": 0.082,
+      "step": 12470
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 0.0001237242394504416,
+      "loss": 0.0886,
+      "step": 12480
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 0.00012357703631010793,
+      "loss": 0.0873,
+      "step": 12490
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 0.0001234298331697743,
+      "loss": 0.0747,
+      "step": 12500
+    },
+    {
+      "epoch": 17.96,
+      "eval_loss": 0.14587688446044922,
+      "eval_runtime": 40.2304,
+      "eval_samples_per_second": 11.36,
+      "eval_steps_per_second": 1.442,
+      "eval_wer": 0.16277669025749134,
+      "step": 12500
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 0.00012328263002944062,
+      "loss": 0.0806,
+      "step": 12510
+    },
+    {
+      "epoch": 17.99,
+      "learning_rate": 0.00012313542688910696,
+      "loss": 0.0835,
+      "step": 12520
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 0.0001229882237487733,
+      "loss": 0.0926,
+      "step": 12530
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 0.00012284102060843965,
+      "loss": 0.0802,
+      "step": 12540
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 0.00012269381746810598,
+      "loss": 0.0666,
+      "step": 12550
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 0.00012254661432777232,
+      "loss": 0.0892,
+      "step": 12560
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 0.00012239941118743865,
+      "loss": 0.0741,
+      "step": 12570
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 0.00012225220804710498,
+      "loss": 0.0788,
+      "step": 12580
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 0.00012210500490677134,
+      "loss": 0.0796,
+      "step": 12590
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 0.00012195780176643766,
+      "loss": 0.1014,
+      "step": 12600
+    },
+    {
+      "epoch": 18.1,
+      "eval_loss": 0.15678279101848602,
+      "eval_runtime": 40.1685,
+      "eval_samples_per_second": 11.377,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.16608944436078904,
+      "step": 12600
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 0.00012181059862610401,
+      "loss": 0.0859,
+      "step": 12610
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 0.00012166339548577034,
+      "loss": 0.0854,
+      "step": 12620
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 0.00012151619234543669,
+      "loss": 0.078,
+      "step": 12630
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 0.00012136898920510302,
+      "loss": 0.0905,
+      "step": 12640
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 0.00012122178606476937,
+      "loss": 0.0968,
+      "step": 12650
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 0.0001210745829244357,
+      "loss": 0.0987,
+      "step": 12660
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 0.00012092737978410205,
+      "loss": 0.0743,
+      "step": 12670
+    },
+    {
+      "epoch": 18.22,
+      "learning_rate": 0.00012078017664376838,
+      "loss": 0.0799,
+      "step": 12680
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 0.00012063297350343472,
+      "loss": 0.087,
+      "step": 12690
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 0.00012048577036310106,
+      "loss": 0.0729,
+      "step": 12700
+    },
+    {
+      "epoch": 18.25,
+      "eval_loss": 0.1656101793050766,
+      "eval_runtime": 39.9566,
+      "eval_samples_per_second": 11.437,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.17120915524770366,
+      "step": 12700
+    },
+    {
+      "epoch": 18.26,
+      "learning_rate": 0.0001203385672227674,
+      "loss": 0.0898,
+      "step": 12710
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 0.00012019136408243375,
+      "loss": 0.0798,
+      "step": 12720
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 0.00012004416094210008,
+      "loss": 0.0952,
+      "step": 12730
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 0.00011989695780176643,
+      "loss": 0.0874,
+      "step": 12740
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 0.00011974975466143276,
+      "loss": 0.0713,
+      "step": 12750
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 0.0001196025515210991,
+      "loss": 0.0865,
+      "step": 12760
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 0.00011945534838076544,
+      "loss": 0.073,
+      "step": 12770
+    },
+    {
+      "epoch": 18.36,
+      "learning_rate": 0.00011930814524043179,
+      "loss": 0.0902,
+      "step": 12780
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 0.00011916094210009812,
+      "loss": 0.0766,
+      "step": 12790
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 0.00011901373895976447,
+      "loss": 0.0884,
+      "step": 12800
+    },
+    {
+      "epoch": 18.39,
+      "eval_loss": 0.15617236495018005,
+      "eval_runtime": 40.4371,
+      "eval_samples_per_second": 11.302,
+      "eval_steps_per_second": 1.434,
+      "eval_wer": 0.16789640114440596,
+      "step": 12800
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 0.0001188665358194308,
+      "loss": 0.0802,
+      "step": 12810
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 0.00011871933267909715,
+      "loss": 0.0701,
+      "step": 12820
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 0.00011857212953876348,
+      "loss": 0.0678,
+      "step": 12830
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 0.00011842492639842981,
+      "loss": 0.08,
+      "step": 12840
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 0.00011827772325809616,
+      "loss": 0.0661,
+      "step": 12850
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 0.0001181305201177625,
+      "loss": 0.0891,
+      "step": 12860
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 0.00011798331697742884,
+      "loss": 0.0792,
+      "step": 12870
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 0.00011783611383709517,
+      "loss": 0.0976,
+      "step": 12880
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 0.00011768891069676152,
+      "loss": 0.0936,
+      "step": 12890
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 0.00011754170755642785,
+      "loss": 0.0823,
+      "step": 12900
+    },
+    {
+      "epoch": 18.53,
+      "eval_loss": 0.15830214321613312,
+      "eval_runtime": 40.3114,
+      "eval_samples_per_second": 11.337,
+      "eval_steps_per_second": 1.439,
+      "eval_wer": 0.16970335792802288,
+      "step": 12900
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 0.0001173945044160942,
+      "loss": 0.0743,
+      "step": 12910
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 0.00011724730127576054,
+      "loss": 0.0868,
+      "step": 12920
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 0.00011710009813542688,
+      "loss": 0.0812,
+      "step": 12930
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 0.00011695289499509322,
+      "loss": 0.0652,
+      "step": 12940
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 0.00011680569185475956,
+      "loss": 0.0738,
+      "step": 12950
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 0.0001166584887144259,
+      "loss": 0.1027,
+      "step": 12960
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 0.00011651128557409224,
+      "loss": 0.0777,
+      "step": 12970
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 0.00011636408243375858,
+      "loss": 0.1243,
+      "step": 12980
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 0.00011621687929342491,
+      "loss": 0.092,
+      "step": 12990
+    },
+    {
+      "epoch": 18.68,
+      "learning_rate": 0.00011606967615309126,
+      "loss": 0.066,
+      "step": 13000
+    },
+    {
+      "epoch": 18.68,
+      "eval_loss": 0.1592216044664383,
+      "eval_runtime": 40.6807,
+      "eval_samples_per_second": 11.234,
+      "eval_steps_per_second": 1.426,
+      "eval_wer": 0.17166089444360788,
+      "step": 13000
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 0.00011592247301275759,
+      "loss": 0.08,
+      "step": 13010
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 0.00011577526987242394,
+      "loss": 0.0768,
+      "step": 13020
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 0.00011562806673209027,
+      "loss": 0.078,
+      "step": 13030
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 0.00011548086359175662,
+      "loss": 0.0931,
+      "step": 13040
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 0.00011533366045142295,
+      "loss": 0.0784,
+      "step": 13050
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 0.0001151864573110893,
+      "loss": 0.0902,
+      "step": 13060
+    },
+    {
+      "epoch": 18.78,
+      "learning_rate": 0.00011503925417075563,
+      "loss": 0.0885,
+      "step": 13070
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 0.00011489205103042198,
+      "loss": 0.0786,
+      "step": 13080
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 0.00011474484789008831,
+      "loss": 0.0763,
+      "step": 13090
+    },
+    {
+      "epoch": 18.82,
+      "learning_rate": 0.00011459764474975466,
+      "loss": 0.0849,
+      "step": 13100
+    },
+    {
+      "epoch": 18.82,
+      "eval_loss": 0.15664438903331757,
+      "eval_runtime": 40.0732,
+      "eval_samples_per_second": 11.404,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.16684234302062942,
+      "step": 13100
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 0.00011445044160942099,
+      "loss": 0.0843,
+      "step": 13110
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 0.00011430323846908734,
+      "loss": 0.0776,
+      "step": 13120
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 0.00011415603532875367,
+      "loss": 0.0597,
+      "step": 13130
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 0.00011400883218842,
+      "loss": 0.0803,
+      "step": 13140
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 0.00011386162904808635,
+      "loss": 0.0802,
+      "step": 13150
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 0.00011371442590775269,
+      "loss": 0.0719,
+      "step": 13160
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 0.00011356722276741903,
+      "loss": 0.0707,
+      "step": 13170
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 0.00011342001962708537,
+      "loss": 0.0741,
+      "step": 13180
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 0.00011327281648675171,
+      "loss": 0.0834,
+      "step": 13190
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 0.00011312561334641805,
+      "loss": 0.0706,
+      "step": 13200
+    },
+    {
+      "epoch": 18.97,
+      "eval_loss": 0.16251307725906372,
+      "eval_runtime": 39.9205,
+      "eval_samples_per_second": 11.448,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.16774582141243788,
+      "step": 13200
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 0.00011297841020608439,
+      "loss": 0.0606,
+      "step": 13210
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 0.00011283120706575073,
+      "loss": 0.0619,
+      "step": 13220
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 0.00011268400392541707,
+      "loss": 0.0671,
+      "step": 13230
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 0.00011253680078508341,
+      "loss": 0.0765,
+      "step": 13240
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 0.00011238959764474975,
+      "loss": 0.0798,
+      "step": 13250
+    },
+    {
+      "epoch": 19.05,
+      "learning_rate": 0.00011224239450441609,
+      "loss": 0.0822,
+      "step": 13260
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 0.00011209519136408243,
+      "loss": 0.0743,
+      "step": 13270
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 0.00011194798822374877,
+      "loss": 0.0842,
+      "step": 13280
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 0.0001118007850834151,
+      "loss": 0.08,
+      "step": 13290
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 0.00011165358194308145,
+      "loss": 0.0693,
+      "step": 13300
+    },
+    {
+      "epoch": 19.11,
+      "eval_loss": 0.16297851502895355,
+      "eval_runtime": 40.169,
+      "eval_samples_per_second": 11.377,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.16910103900015058,
+      "step": 13300
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 0.00011150637880274778,
+      "loss": 0.0847,
+      "step": 13310
+    },
+    {
+      "epoch": 19.14,
+      "learning_rate": 0.00011135917566241413,
+      "loss": 0.0775,
+      "step": 13320
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 0.00011121197252208046,
+      "loss": 0.0876,
+      "step": 13330
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 0.00011106476938174681,
+      "loss": 0.0912,
+      "step": 13340
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 0.00011091756624141314,
+      "loss": 0.0635,
+      "step": 13350
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 0.00011077036310107949,
+      "loss": 0.0762,
+      "step": 13360
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 0.00011062315996074582,
+      "loss": 0.0885,
+      "step": 13370
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 0.00011047595682041217,
+      "loss": 0.0821,
+      "step": 13380
+    },
+    {
+      "epoch": 19.24,
+      "learning_rate": 0.0001103287536800785,
+      "loss": 0.0731,
+      "step": 13390
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 0.00011018155053974485,
+      "loss": 0.0736,
+      "step": 13400
+    },
+    {
+      "epoch": 19.25,
+      "eval_loss": 0.16010314226150513,
+      "eval_runtime": 40.0705,
+      "eval_samples_per_second": 11.405,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.16714350248456558,
+      "step": 13400
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 0.00011003434739941118,
+      "loss": 0.0817,
+      "step": 13410
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 0.00010988714425907753,
+      "loss": 0.0842,
+      "step": 13420
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 0.00010973994111874386,
+      "loss": 0.0852,
+      "step": 13430
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 0.00010959273797841018,
+      "loss": 0.0867,
+      "step": 13440
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 0.00010944553483807654,
+      "loss": 0.0984,
+      "step": 13450
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 0.00010929833169774286,
+      "loss": 0.0867,
+      "step": 13460
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 0.00010915112855740922,
+      "loss": 0.0619,
+      "step": 13470
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 0.00010900392541707554,
+      "loss": 0.089,
+      "step": 13480
+    },
+    {
+      "epoch": 19.38,
+      "learning_rate": 0.0001088567222767419,
+      "loss": 0.0666,
+      "step": 13490
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 0.00010870951913640822,
+      "loss": 0.0877,
+      "step": 13500
+    },
+    {
+      "epoch": 19.4,
+      "eval_loss": 0.16647975146770477,
+      "eval_runtime": 39.9295,
+      "eval_samples_per_second": 11.445,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.1689504592681825,
+      "step": 13500
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 0.00010856231599607458,
+      "loss": 0.0905,
+      "step": 13510
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 0.0001084151128557409,
+      "loss": 0.0656,
+      "step": 13520
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 0.00010826790971540726,
+      "loss": 0.0834,
+      "step": 13530
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 0.00010812070657507358,
+      "loss": 0.0769,
+      "step": 13540
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 0.00010797350343473995,
+      "loss": 0.0981,
+      "step": 13550
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 0.00010782630029440627,
+      "loss": 0.1006,
+      "step": 13560
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 0.00010767909715407263,
+      "loss": 0.0693,
+      "step": 13570
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 0.00010753189401373895,
+      "loss": 0.0747,
+      "step": 13580
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 0.00010738469087340528,
+      "loss": 0.077,
+      "step": 13590
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 0.00010723748773307163,
+      "loss": 0.0735,
+      "step": 13600
+    },
+    {
+      "epoch": 19.54,
+      "eval_loss": 0.16026227176189423,
+      "eval_runtime": 39.8849,
+      "eval_samples_per_second": 11.458,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.1675952416804698,
+      "step": 13600
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 0.00010709028459273796,
+      "loss": 0.1046,
+      "step": 13610
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 0.0001069430814524043,
+      "loss": 0.0685,
+      "step": 13620
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 0.00010679587831207064,
+      "loss": 0.0856,
+      "step": 13630
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 0.00010664867517173699,
+      "loss": 0.0621,
+      "step": 13640
+    },
+    {
+      "epoch": 19.61,
+      "learning_rate": 0.00010650147203140332,
+      "loss": 0.0743,
+      "step": 13650
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 0.00010635426889106967,
+      "loss": 0.0899,
+      "step": 13660
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 0.000106207065750736,
+      "loss": 0.0761,
+      "step": 13670
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 0.00010605986261040235,
+      "loss": 0.0887,
+      "step": 13680
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 0.00010591265947006868,
+      "loss": 0.082,
+      "step": 13690
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 0.00010576545632973503,
+      "loss": 0.0804,
+      "step": 13700
+    },
+    {
+      "epoch": 19.68,
+      "eval_loss": 0.16200566291809082,
+      "eval_runtime": 40.0981,
+      "eval_samples_per_second": 11.397,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.17060683631983134,
+      "step": 13700
+    },
+    {
+      "epoch": 19.7,
+      "learning_rate": 0.00010561825318940136,
+      "loss": 0.08,
+      "step": 13710
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 0.00010547105004906771,
+      "loss": 0.0683,
+      "step": 13720
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 0.00010532384690873404,
+      "loss": 0.0769,
+      "step": 13730
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 0.00010517664376840037,
+      "loss": 0.0768,
+      "step": 13740
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 0.00010502944062806672,
+      "loss": 0.0666,
+      "step": 13750
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 0.00010488223748773305,
+      "loss": 0.0676,
+      "step": 13760
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 0.0001047350343473994,
+      "loss": 0.0745,
+      "step": 13770
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 0.00010458783120706574,
+      "loss": 0.0802,
+      "step": 13780
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 0.00010444062806673208,
+      "loss": 0.0733,
+      "step": 13790
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 0.00010429342492639842,
+      "loss": 0.0871,
+      "step": 13800
+    },
+    {
+      "epoch": 19.83,
+      "eval_loss": 0.1567193865776062,
+      "eval_runtime": 39.8991,
+      "eval_samples_per_second": 11.454,
+      "eval_steps_per_second": 1.454,
+      "eval_wer": 0.1644330673091402,
+      "step": 13800
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 0.00010414622178606476,
+      "loss": 0.0767,
+      "step": 13810
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 0.0001039990186457311,
+      "loss": 0.0782,
+      "step": 13820
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 0.00010385181550539744,
+      "loss": 0.077,
+      "step": 13830
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 0.00010370461236506378,
+      "loss": 0.0592,
+      "step": 13840
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 0.00010355740922473012,
+      "loss": 0.0768,
+      "step": 13850
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 0.00010341020608439646,
+      "loss": 0.0797,
+      "step": 13860
+    },
+    {
+      "epoch": 19.93,
+      "learning_rate": 0.0001032630029440628,
+      "loss": 0.0721,
+      "step": 13870
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 0.00010311579980372914,
+      "loss": 0.0699,
+      "step": 13880
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 0.00010296859666339547,
+      "loss": 0.0656,
+      "step": 13890
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 0.00010282139352306182,
+      "loss": 0.0822,
+      "step": 13900
+    },
+    {
+      "epoch": 19.97,
+      "eval_loss": 0.1556914746761322,
+      "eval_runtime": 40.153,
+      "eval_samples_per_second": 11.381,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.16578828489685288,
+      "step": 13900
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 0.00010267419038272815,
+      "loss": 0.0944,
+      "step": 13910
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 0.0001025269872423945,
+      "loss": 0.0612,
+      "step": 13920
+    },
+    {
+      "epoch": 20.01,
+      "learning_rate": 0.00010237978410206083,
+      "loss": 0.082,
+      "step": 13930
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 0.00010224730127576054,
+      "loss": 0.0691,
+      "step": 13940
+    },
+    {
+      "epoch": 20.04,
+      "learning_rate": 0.00010210009813542687,
+      "loss": 0.068,
+      "step": 13950
+    },
+    {
+      "epoch": 20.06,
+      "learning_rate": 0.00010195289499509322,
+      "loss": 0.0663,
+      "step": 13960
+    },
+    {
+      "epoch": 20.07,
+      "learning_rate": 0.00010180569185475955,
+      "loss": 0.0654,
+      "step": 13970
+    },
+    {
+      "epoch": 20.09,
+      "learning_rate": 0.0001016584887144259,
+      "loss": 0.091,
+      "step": 13980
+    },
+    {
+      "epoch": 20.1,
+      "learning_rate": 0.00010151128557409223,
+      "loss": 0.075,
+      "step": 13990
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 0.00010136408243375858,
+      "loss": 0.0784,
+      "step": 14000
+    },
+    {
+      "epoch": 20.11,
+      "eval_loss": 0.16241466999053955,
+      "eval_runtime": 39.9865,
+      "eval_samples_per_second": 11.429,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.16187321186568288,
+      "step": 14000
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 0.00010121687929342491,
+      "loss": 0.0876,
+      "step": 14010
+    },
+    {
+      "epoch": 20.14,
+      "learning_rate": 0.00010106967615309126,
+      "loss": 0.0809,
+      "step": 14020
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 0.00010092247301275759,
+      "loss": 0.0855,
+      "step": 14030
+    },
+    {
+      "epoch": 20.17,
+      "learning_rate": 0.00010077526987242394,
+      "loss": 0.0702,
+      "step": 14040
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 0.00010062806673209027,
+      "loss": 0.0846,
+      "step": 14050
+    },
+    {
+      "epoch": 20.2,
+      "learning_rate": 0.00010048086359175662,
+      "loss": 0.0724,
+      "step": 14060
+    },
+    {
+      "epoch": 20.22,
+      "learning_rate": 0.00010033366045142295,
+      "loss": 0.0664,
+      "step": 14070
+    },
+    {
+      "epoch": 20.23,
+      "learning_rate": 0.0001001864573110893,
+      "loss": 0.0825,
+      "step": 14080
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 0.00010003925417075563,
+      "loss": 0.0651,
+      "step": 14090
+    },
+    {
+      "epoch": 20.26,
+      "learning_rate": 9.989205103042197e-05,
+      "loss": 0.0822,
+      "step": 14100
+    },
+    {
+      "epoch": 20.26,
+      "eval_loss": 0.16037575900554657,
+      "eval_runtime": 39.914,
+      "eval_samples_per_second": 11.45,
+      "eval_steps_per_second": 1.453,
+      "eval_wer": 0.1617226321337148,
+      "step": 14100
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 9.974484789008831e-05,
+      "loss": 0.0743,
+      "step": 14110
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 9.959764474975465e-05,
+      "loss": 0.0709,
+      "step": 14120
+    },
+    {
+      "epoch": 20.3,
+      "learning_rate": 9.945044160942099e-05,
+      "loss": 0.068,
+      "step": 14130
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 9.930323846908733e-05,
+      "loss": 0.067,
+      "step": 14140
+    },
+    {
+      "epoch": 20.33,
+      "learning_rate": 9.915603532875367e-05,
+      "loss": 0.0749,
+      "step": 14150
+    },
+    {
+      "epoch": 20.34,
+      "learning_rate": 9.900883218842e-05,
+      "loss": 0.0676,
+      "step": 14160
+    },
+    {
+      "epoch": 20.36,
+      "learning_rate": 9.886162904808635e-05,
+      "loss": 0.0652,
+      "step": 14170
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 9.871442590775269e-05,
+      "loss": 0.0841,
+      "step": 14180
+    },
+    {
+      "epoch": 20.39,
+      "learning_rate": 9.856722276741903e-05,
+      "loss": 0.0779,
+      "step": 14190
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 9.842001962708537e-05,
+      "loss": 0.0694,
+      "step": 14200
+    },
+    {
+      "epoch": 20.4,
+      "eval_loss": 0.14927269518375397,
+      "eval_runtime": 40.1956,
+      "eval_samples_per_second": 11.369,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.16142147266977866,
+      "step": 14200
+    },
+    {
+      "epoch": 20.42,
+      "learning_rate": 9.827281648675171e-05,
+      "loss": 0.0796,
+      "step": 14210
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 9.812561334641805e-05,
+      "loss": 0.0779,
+      "step": 14220
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 9.79784102060844e-05,
+      "loss": 0.0788,
+      "step": 14230
+    },
+    {
+      "epoch": 20.46,
+      "learning_rate": 9.783120706575073e-05,
+      "loss": 0.0567,
+      "step": 14240
+    },
+    {
+      "epoch": 20.47,
+      "learning_rate": 9.768400392541706e-05,
+      "loss": 0.0715,
+      "step": 14250
+    },
+    {
+      "epoch": 20.49,
+      "learning_rate": 9.753680078508341e-05,
+      "loss": 0.0744,
+      "step": 14260
+    },
+    {
+      "epoch": 20.5,
+      "learning_rate": 9.738959764474974e-05,
+      "loss": 0.0641,
+      "step": 14270
+    },
+    {
+      "epoch": 20.52,
+      "learning_rate": 9.724239450441609e-05,
+      "loss": 0.0776,
+      "step": 14280
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 9.709519136408242e-05,
+      "loss": 0.0637,
+      "step": 14290
+    },
+    {
+      "epoch": 20.55,
+      "learning_rate": 9.694798822374877e-05,
+      "loss": 0.0775,
+      "step": 14300
+    },
+    {
+      "epoch": 20.55,
+      "eval_loss": 0.1541454941034317,
+      "eval_runtime": 39.9553,
+      "eval_samples_per_second": 11.438,
+      "eval_steps_per_second": 1.452,
+      "eval_wer": 0.16081915374190633,
+      "step": 14300
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 9.68007850834151e-05,
+      "loss": 0.0805,
+      "step": 14310
+    },
+    {
+      "epoch": 20.57,
+      "learning_rate": 9.665358194308145e-05,
+      "loss": 0.0722,
+      "step": 14320
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 9.650637880274778e-05,
+      "loss": 0.0682,
+      "step": 14330
+    },
+    {
+      "epoch": 20.6,
+      "learning_rate": 9.635917566241413e-05,
+      "loss": 0.0724,
+      "step": 14340
+    },
+    {
+      "epoch": 20.62,
+      "learning_rate": 9.621197252208046e-05,
+      "loss": 0.0834,
+      "step": 14350
+    },
+    {
+      "epoch": 20.63,
+      "learning_rate": 9.606476938174681e-05,
+      "loss": 0.0795,
+      "step": 14360
+    },
+    {
+      "epoch": 20.65,
+      "learning_rate": 9.591756624141314e-05,
+      "loss": 0.0874,
+      "step": 14370
+    },
+    {
+      "epoch": 20.66,
+      "learning_rate": 9.577036310107949e-05,
+      "loss": 0.0689,
+      "step": 14380
+    },
+    {
+      "epoch": 20.68,
+      "learning_rate": 9.562315996074582e-05,
+      "loss": 0.0669,
+      "step": 14390
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 9.547595682041216e-05,
+      "loss": 0.0747,
+      "step": 14400
+    },
+    {
+      "epoch": 20.69,
+      "eval_loss": 0.15958380699157715,
+      "eval_runtime": 40.1805,
+      "eval_samples_per_second": 11.374,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.1624755307935552,
+      "step": 14400
+    },
+    {
+      "epoch": 20.7,
+      "learning_rate": 9.53287536800785e-05,
+      "loss": 0.0737,
+      "step": 14410
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 9.518155053974484e-05,
+      "loss": 0.0717,
+      "step": 14420
+    },
+    {
+      "epoch": 20.73,
+      "learning_rate": 9.503434739941118e-05,
+      "loss": 0.0923,
+      "step": 14430
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 9.488714425907752e-05,
+      "loss": 0.0623,
+      "step": 14440
+    },
+    {
+      "epoch": 20.76,
+      "learning_rate": 9.473994111874386e-05,
+      "loss": 0.0658,
+      "step": 14450
+    },
+    {
+      "epoch": 20.78,
+      "learning_rate": 9.45927379784102e-05,
+      "loss": 0.0658,
+      "step": 14460
+    },
+    {
+      "epoch": 20.79,
+      "learning_rate": 9.444553483807654e-05,
+      "loss": 0.068,
+      "step": 14470
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 9.429833169774288e-05,
+      "loss": 0.0697,
+      "step": 14480
+    },
+    {
+      "epoch": 20.82,
+      "learning_rate": 9.415112855740923e-05,
+      "loss": 0.0618,
+      "step": 14490
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 9.400392541707556e-05,
+      "loss": 0.0701,
+      "step": 14500
+    },
+    {
+      "epoch": 20.83,
+      "eval_loss": 0.15933384001255035,
+      "eval_runtime": 40.0625,
+      "eval_samples_per_second": 11.407,
+      "eval_steps_per_second": 1.448,
+      "eval_wer": 0.16081915374190633,
+      "step": 14500
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 9.38567222767419e-05,
+      "loss": 0.0737,
+      "step": 14510
+    },
+    {
+      "epoch": 20.86,
+      "learning_rate": 9.370951913640824e-05,
+      "loss": 0.0772,
+      "step": 14520
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 9.356231599607459e-05,
+      "loss": 0.0846,
+      "step": 14530
+    },
+    {
+      "epoch": 20.89,
+      "learning_rate": 9.341511285574092e-05,
+      "loss": 0.0514,
+      "step": 14540
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 9.326790971540725e-05,
+      "loss": 0.0853,
+      "step": 14550
+    },
+    {
+      "epoch": 20.92,
+      "learning_rate": 9.31207065750736e-05,
+      "loss": 0.0715,
+      "step": 14560
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 9.297350343473993e-05,
+      "loss": 0.0635,
+      "step": 14570
+    },
+    {
+      "epoch": 20.95,
+      "learning_rate": 9.282630029440628e-05,
+      "loss": 0.0978,
+      "step": 14580
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 9.267909715407261e-05,
+      "loss": 0.0803,
+      "step": 14590
+    },
+    {
+      "epoch": 20.98,
+      "learning_rate": 9.253189401373896e-05,
+      "loss": 0.0698,
+      "step": 14600
+    },
+    {
+      "epoch": 20.98,
+      "eval_loss": 0.15506555140018463,
+      "eval_runtime": 40.1892,
+      "eval_samples_per_second": 11.371,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.16081915374190633,
+      "step": 14600
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 9.23846908734053e-05,
+      "loss": 0.0791,
+      "step": 14610
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 9.223748773307164e-05,
+      "loss": 0.0568,
+      "step": 14620
+    },
+    {
+      "epoch": 21.02,
+      "learning_rate": 9.209028459273797e-05,
+      "loss": 0.0696,
+      "step": 14630
+    },
+    {
+      "epoch": 21.03,
+      "learning_rate": 9.194308145240432e-05,
+      "loss": 0.082,
+      "step": 14640
+    },
+    {
+      "epoch": 21.05,
+      "learning_rate": 9.179587831207065e-05,
+      "loss": 0.0648,
+      "step": 14650
+    },
+    {
+      "epoch": 21.06,
+      "learning_rate": 9.1648675171737e-05,
+      "loss": 0.056,
+      "step": 14660
+    },
+    {
+      "epoch": 21.08,
+      "learning_rate": 9.150147203140332e-05,
+      "loss": 0.0694,
+      "step": 14670
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 9.135426889106968e-05,
+      "loss": 0.0662,
+      "step": 14680
+    },
+    {
+      "epoch": 21.11,
+      "learning_rate": 9.1207065750736e-05,
+      "loss": 0.0671,
+      "step": 14690
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 9.105986261040233e-05,
+      "loss": 0.0785,
+      "step": 14700
+    },
+    {
+      "epoch": 21.12,
+      "eval_loss": 0.15715524554252625,
+      "eval_runtime": 40.3412,
+      "eval_samples_per_second": 11.328,
+      "eval_steps_per_second": 1.438,
+      "eval_wer": 0.1636801686492998,
+      "step": 14700
+    },
+    {
+      "epoch": 21.14,
+      "learning_rate": 9.091265947006868e-05,
+      "loss": 0.0661,
+      "step": 14710
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 9.076545632973502e-05,
+      "loss": 0.0702,
+      "step": 14720
+    },
+    {
+      "epoch": 21.16,
+      "learning_rate": 9.061825318940136e-05,
+      "loss": 0.0671,
+      "step": 14730
+    },
+    {
+      "epoch": 21.18,
+      "learning_rate": 9.04710500490677e-05,
+      "loss": 0.0707,
+      "step": 14740
+    },
+    {
+      "epoch": 21.19,
+      "learning_rate": 9.032384690873404e-05,
+      "loss": 0.0559,
+      "step": 14750
+    },
+    {
+      "epoch": 21.21,
+      "learning_rate": 9.017664376840038e-05,
+      "loss": 0.068,
+      "step": 14760
+    },
+    {
+      "epoch": 21.22,
+      "learning_rate": 9.002944062806672e-05,
+      "loss": 0.0817,
+      "step": 14770
+    },
+    {
+      "epoch": 21.24,
+      "learning_rate": 8.988223748773306e-05,
+      "loss": 0.0792,
+      "step": 14780
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 8.97350343473994e-05,
+      "loss": 0.054,
+      "step": 14790
+    },
+    {
+      "epoch": 21.26,
+      "learning_rate": 8.958783120706574e-05,
+      "loss": 0.0975,
+      "step": 14800
+    },
+    {
+      "epoch": 21.26,
+      "eval_loss": 0.16008907556533813,
+      "eval_runtime": 40.1457,
+      "eval_samples_per_second": 11.384,
+      "eval_steps_per_second": 1.445,
+      "eval_wer": 0.16322842945339558,
+      "step": 14800
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 8.944062806673208e-05,
+      "loss": 0.0653,
+      "step": 14810
+    },
+    {
+      "epoch": 21.29,
+      "learning_rate": 8.929342492639842e-05,
+      "loss": 0.0675,
+      "step": 14820
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 8.914622178606476e-05,
+      "loss": 0.0676,
+      "step": 14830
+    },
+    {
+      "epoch": 21.32,
+      "learning_rate": 8.89990186457311e-05,
+      "loss": 0.0631,
+      "step": 14840
+    },
+    {
+      "epoch": 21.34,
+      "learning_rate": 8.885181550539743e-05,
+      "loss": 0.0625,
+      "step": 14850
+    },
+    {
+      "epoch": 21.35,
+      "learning_rate": 8.870461236506378e-05,
+      "loss": 0.071,
+      "step": 14860
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 8.855740922473011e-05,
+      "loss": 0.0684,
+      "step": 14870
+    },
+    {
+      "epoch": 21.38,
+      "learning_rate": 8.841020608439646e-05,
+      "loss": 0.0656,
+      "step": 14880
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 8.826300294406279e-05,
+      "loss": 0.0609,
+      "step": 14890
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 8.811579980372914e-05,
+      "loss": 0.0697,
+      "step": 14900
+    },
+    {
+      "epoch": 21.41,
+      "eval_loss": 0.15438880026340485,
+      "eval_runtime": 40.0849,
+      "eval_samples_per_second": 11.401,
+      "eval_steps_per_second": 1.447,
+      "eval_wer": 0.15931335642222558,
+      "step": 14900
+    },
+    {
+      "epoch": 21.42,
+      "learning_rate": 8.796859666339547e-05,
+      "loss": 0.0698,
+      "step": 14910
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 8.782139352306182e-05,
+      "loss": 0.0645,
+      "step": 14920
+    },
+    {
+      "epoch": 21.45,
+      "learning_rate": 8.767419038272815e-05,
+      "loss": 0.0664,
+      "step": 14930
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 8.75269872423945e-05,
+      "loss": 0.0682,
+      "step": 14940
+    },
+    {
+      "epoch": 21.48,
+      "learning_rate": 8.737978410206083e-05,
+      "loss": 0.0693,
+      "step": 14950
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 8.723258096172718e-05,
+      "loss": 0.0717,
+      "step": 14960
+    },
+    {
+      "epoch": 21.51,
+      "learning_rate": 8.708537782139351e-05,
+      "loss": 0.0818,
+      "step": 14970
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 8.693817468105985e-05,
+      "loss": 0.0579,
+      "step": 14980
+    },
+    {
+      "epoch": 21.54,
+      "learning_rate": 8.679097154072619e-05,
+      "loss": 0.0598,
+      "step": 14990
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 8.664376840039253e-05,
+      "loss": 0.0638,
+      "step": 15000
+    },
+    {
+      "epoch": 21.55,
+      "eval_loss": 0.14896854758262634,
+      "eval_runtime": 40.2901,
+      "eval_samples_per_second": 11.343,
+      "eval_steps_per_second": 1.44,
+      "eval_wer": 0.1572052401746725,
+      "step": 15000
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 8.649656526005887e-05,
+      "loss": 0.07,
+      "step": 15010
+    },
+    {
+      "epoch": 21.58,
+      "learning_rate": 8.63493621197252e-05,
+      "loss": 0.0689,
+      "step": 15020
+    },
+    {
+      "epoch": 21.59,
+      "learning_rate": 8.620215897939155e-05,
+      "loss": 0.064,
+      "step": 15030
+    },
+    {
+      "epoch": 21.61,
+      "learning_rate": 8.605495583905789e-05,
+      "loss": 0.0582,
+      "step": 15040
+    },
+    {
+      "epoch": 21.62,
+      "learning_rate": 8.590775269872423e-05,
+      "loss": 0.0626,
+      "step": 15050
+    },
+    {
+      "epoch": 21.64,
+      "learning_rate": 8.576054955839057e-05,
+      "loss": 0.0564,
+      "step": 15060
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 8.561334641805691e-05,
+      "loss": 0.0642,
+      "step": 15070
+    },
+    {
+      "epoch": 21.67,
+      "learning_rate": 8.546614327772325e-05,
+      "loss": 0.056,
+      "step": 15080
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 8.53189401373896e-05,
+      "loss": 0.064,
+      "step": 15090
+    },
+    {
+      "epoch": 21.7,
+      "learning_rate": 8.517173699705593e-05,
+      "loss": 0.0665,
+      "step": 15100
+    },
+    {
+      "epoch": 21.7,
+      "eval_loss": 0.15601423382759094,
+      "eval_runtime": 40.1857,
+      "eval_samples_per_second": 11.372,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.1558500225869598,
+      "step": 15100
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 8.502453385672227e-05,
+      "loss": 0.0687,
+      "step": 15110
+    },
+    {
+      "epoch": 21.72,
+      "learning_rate": 8.487733071638861e-05,
+      "loss": 0.0617,
+      "step": 15120
+    },
+    {
+      "epoch": 21.74,
+      "learning_rate": 8.473012757605494e-05,
+      "loss": 0.0851,
+      "step": 15130
+    },
+    {
+      "epoch": 21.75,
+      "learning_rate": 8.458292443572129e-05,
+      "loss": 0.0644,
+      "step": 15140
+    },
+    {
+      "epoch": 21.77,
+      "learning_rate": 8.443572129538762e-05,
+      "loss": 0.073,
+      "step": 15150
+    },
+    {
+      "epoch": 21.78,
+      "learning_rate": 8.428851815505397e-05,
+      "loss": 0.0676,
+      "step": 15160
+    },
+    {
+      "epoch": 21.8,
+      "learning_rate": 8.41413150147203e-05,
+      "loss": 0.0772,
+      "step": 15170
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 8.399411187438665e-05,
+      "loss": 0.0654,
+      "step": 15180
+    },
+    {
+      "epoch": 21.82,
+      "learning_rate": 8.384690873405298e-05,
+      "loss": 0.0832,
+      "step": 15190
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 8.369970559371933e-05,
+      "loss": 0.0764,
+      "step": 15200
+    },
+    {
+      "epoch": 21.84,
+      "eval_loss": 0.15133680403232574,
+      "eval_runtime": 40.219,
+      "eval_samples_per_second": 11.363,
+      "eval_steps_per_second": 1.442,
+      "eval_wer": 0.15795813883451287,
+      "step": 15200
+    },
+    {
+      "epoch": 21.85,
+      "learning_rate": 8.355250245338566e-05,
+      "loss": 0.0564,
+      "step": 15210
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 8.340529931305201e-05,
+      "loss": 0.0689,
+      "step": 15220
+    },
+    {
+      "epoch": 21.88,
+      "learning_rate": 8.325809617271834e-05,
+      "loss": 0.0747,
+      "step": 15230
+    },
+    {
+      "epoch": 21.9,
+      "learning_rate": 8.311089303238469e-05,
+      "loss": 0.0566,
+      "step": 15240
+    },
+    {
+      "epoch": 21.91,
+      "learning_rate": 8.296368989205102e-05,
+      "loss": 0.0712,
+      "step": 15250
+    },
+    {
+      "epoch": 21.93,
+      "learning_rate": 8.281648675171737e-05,
+      "loss": 0.0637,
+      "step": 15260
+    },
+    {
+      "epoch": 21.94,
+      "learning_rate": 8.26692836113837e-05,
+      "loss": 0.0638,
+      "step": 15270
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 8.252208047105004e-05,
+      "loss": 0.0679,
+      "step": 15280
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 8.237487733071638e-05,
+      "loss": 0.0616,
+      "step": 15290
+    },
+    {
+      "epoch": 21.98,
+      "learning_rate": 8.222767419038272e-05,
+      "loss": 0.0642,
+      "step": 15300
+    },
+    {
+      "epoch": 21.98,
+      "eval_loss": 0.149861678481102,
+      "eval_runtime": 40.2796,
+      "eval_samples_per_second": 11.346,
+      "eval_steps_per_second": 1.44,
+      "eval_wer": 0.1552477036590875,
+      "step": 15300
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 8.208047105004906e-05,
+      "loss": 0.0694,
+      "step": 15310
+    },
+    {
+      "epoch": 22.01,
+      "learning_rate": 8.19332679097154e-05,
+      "loss": 0.0605,
+      "step": 15320
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 8.178606476938174e-05,
+      "loss": 0.0637,
+      "step": 15330
+    },
+    {
+      "epoch": 22.04,
+      "learning_rate": 8.163886162904808e-05,
+      "loss": 0.0606,
+      "step": 15340
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 8.149165848871443e-05,
+      "loss": 0.0583,
+      "step": 15350
+    },
+    {
+      "epoch": 22.07,
+      "learning_rate": 8.134445534838076e-05,
+      "loss": 0.0483,
+      "step": 15360
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 8.11972522080471e-05,
+      "loss": 0.0672,
+      "step": 15370
+    },
+    {
+      "epoch": 22.1,
+      "learning_rate": 8.105004906771344e-05,
+      "loss": 0.0558,
+      "step": 15380
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 8.090284592737979e-05,
+      "loss": 0.0598,
+      "step": 15390
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 8.075564278704612e-05,
+      "loss": 0.0577,
+      "step": 15400
+    },
+    {
+      "epoch": 22.13,
+      "eval_loss": 0.15612779557704926,
+      "eval_runtime": 40.1084,
+      "eval_samples_per_second": 11.394,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.15871103749435325,
+      "step": 15400
+    },
+    {
+      "epoch": 22.14,
+      "learning_rate": 8.060843964671247e-05,
+      "loss": 0.0704,
+      "step": 15410
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 8.04612365063788e-05,
+      "loss": 0.0887,
+      "step": 15420
+    },
+    {
+      "epoch": 22.17,
+      "learning_rate": 8.031403336604513e-05,
+      "loss": 0.0569,
+      "step": 15430
+    },
+    {
+      "epoch": 22.18,
+      "learning_rate": 8.016683022571148e-05,
+      "loss": 0.0672,
+      "step": 15440
+    },
+    {
+      "epoch": 22.2,
+      "learning_rate": 8.001962708537781e-05,
+      "loss": 0.0557,
+      "step": 15450
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 7.987242394504416e-05,
+      "loss": 0.041,
+      "step": 15460
+    },
+    {
+      "epoch": 22.23,
+      "learning_rate": 7.97252208047105e-05,
+      "loss": 0.0759,
+      "step": 15470
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 7.957801766437684e-05,
+      "loss": 0.0464,
+      "step": 15480
+    },
+    {
+      "epoch": 22.26,
+      "learning_rate": 7.943081452404317e-05,
+      "loss": 0.0764,
+      "step": 15490
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 7.928361138370952e-05,
+      "loss": 0.0551,
+      "step": 15500
+    },
+    {
+      "epoch": 22.27,
+      "eval_loss": 0.1587643325328827,
+      "eval_runtime": 40.2333,
+      "eval_samples_per_second": 11.359,
+      "eval_steps_per_second": 1.442,
+      "eval_wer": 0.1570546604427044,
+      "step": 15500
+    },
+    {
+      "epoch": 22.28,
+      "learning_rate": 7.913640824337585e-05,
+      "loss": 0.0516,
+      "step": 15510
+    },
+    {
+      "epoch": 22.3,
+      "learning_rate": 7.89892051030422e-05,
+      "loss": 0.0619,
+      "step": 15520
+    },
+    {
+      "epoch": 22.31,
+      "learning_rate": 7.884200196270853e-05,
+      "loss": 0.0618,
+      "step": 15530
+    },
+    {
+      "epoch": 22.33,
+      "learning_rate": 7.869479882237488e-05,
+      "loss": 0.0654,
+      "step": 15540
+    },
+    {
+      "epoch": 22.34,
+      "learning_rate": 7.854759568204122e-05,
+      "loss": 0.0652,
+      "step": 15550
+    },
+    {
+      "epoch": 22.36,
+      "learning_rate": 7.840039254170756e-05,
+      "loss": 0.0614,
+      "step": 15560
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 7.82531894013739e-05,
+      "loss": 0.0753,
+      "step": 15570
+    },
+    {
+      "epoch": 22.39,
+      "learning_rate": 7.810598626104022e-05,
+      "loss": 0.0618,
+      "step": 15580
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 7.795878312070658e-05,
+      "loss": 0.0634,
+      "step": 15590
+    },
+    {
+      "epoch": 22.41,
+      "learning_rate": 7.78115799803729e-05,
+      "loss": 0.0707,
+      "step": 15600
+    },
+    {
+      "epoch": 22.41,
+      "eval_loss": 0.1521938592195511,
+      "eval_runtime": 40.3124,
+      "eval_samples_per_second": 11.336,
+      "eval_steps_per_second": 1.439,
+      "eval_wer": 0.15810871856648095,
+      "step": 15600
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 7.766437684003926e-05,
+      "loss": 0.0505,
+      "step": 15610
+    },
+    {
+      "epoch": 22.44,
+      "learning_rate": 7.751717369970558e-05,
+      "loss": 0.0701,
+      "step": 15620
+    },
+    {
+      "epoch": 22.46,
+      "learning_rate": 7.736997055937194e-05,
+      "loss": 0.058,
+      "step": 15630
+    },
+    {
+      "epoch": 22.47,
+      "learning_rate": 7.722276741903826e-05,
+      "loss": 0.0668,
+      "step": 15640
+    },
+    {
+      "epoch": 22.49,
+      "learning_rate": 7.707556427870462e-05,
+      "loss": 0.0538,
+      "step": 15650
+    },
+    {
+      "epoch": 22.5,
+      "learning_rate": 7.692836113837094e-05,
+      "loss": 0.0613,
+      "step": 15660
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 7.67811579980373e-05,
+      "loss": 0.0562,
+      "step": 15670
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 7.663395485770362e-05,
+      "loss": 0.0581,
+      "step": 15680
+    },
+    {
+      "epoch": 22.54,
+      "learning_rate": 7.648675171736998e-05,
+      "loss": 0.0557,
+      "step": 15690
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 7.63395485770363e-05,
+      "loss": 0.076,
+      "step": 15700
+    },
+    {
+      "epoch": 22.56,
+      "eval_loss": 0.14777785539627075,
+      "eval_runtime": 40.7818,
+      "eval_samples_per_second": 11.206,
+      "eval_steps_per_second": 1.422,
+      "eval_wer": 0.1596145158861617,
+      "step": 15700
+    },
+    {
+      "epoch": 22.57,
+      "learning_rate": 7.619234543670264e-05,
+      "loss": 0.0626,
+      "step": 15710
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 7.604514229636898e-05,
+      "loss": 0.069,
+      "step": 15720
+    },
+    {
+      "epoch": 22.6,
+      "learning_rate": 7.589793915603531e-05,
+      "loss": 0.0668,
+      "step": 15730
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 7.575073601570166e-05,
+      "loss": 0.0886,
+      "step": 15740
+    },
+    {
+      "epoch": 22.63,
+      "learning_rate": 7.560353287536799e-05,
+      "loss": 0.0619,
+      "step": 15750
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 7.547105004906771e-05,
+      "loss": 0.0481,
+      "step": 15760
+    },
+    {
+      "epoch": 22.66,
+      "learning_rate": 7.532384690873406e-05,
+      "loss": 0.0674,
+      "step": 15770
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 7.517664376840039e-05,
+      "loss": 0.073,
+      "step": 15780
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 7.502944062806674e-05,
+      "loss": 0.057,
+      "step": 15790
+    },
+    {
+      "epoch": 22.7,
+      "learning_rate": 7.488223748773307e-05,
+      "loss": 0.0802,
+      "step": 15800
+    },
+    {
+      "epoch": 22.7,
+      "eval_loss": 0.1453210413455963,
+      "eval_runtime": 40.4045,
+      "eval_samples_per_second": 11.311,
+      "eval_steps_per_second": 1.435,
+      "eval_wer": 0.15464538473121517,
+      "step": 15800
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 7.47350343473994e-05,
+      "loss": 0.063,
+      "step": 15810
+    },
+    {
+      "epoch": 22.73,
+      "learning_rate": 7.458783120706575e-05,
+      "loss": 0.0889,
+      "step": 15820
+    },
+    {
+      "epoch": 22.74,
+      "learning_rate": 7.444062806673208e-05,
+      "loss": 0.0682,
+      "step": 15830
+    },
+    {
+      "epoch": 22.76,
+      "learning_rate": 7.429342492639843e-05,
+      "loss": 0.0548,
+      "step": 15840
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 7.414622178606477e-05,
+      "loss": 0.0677,
+      "step": 15850
+    },
+    {
+      "epoch": 22.79,
+      "learning_rate": 7.399901864573111e-05,
+      "loss": 0.0705,
+      "step": 15860
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 7.385181550539743e-05,
+      "loss": 0.0634,
+      "step": 15870
+    },
+    {
+      "epoch": 22.82,
+      "learning_rate": 7.370461236506378e-05,
+      "loss": 0.0574,
+      "step": 15880
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 7.355740922473011e-05,
+      "loss": 0.0654,
+      "step": 15890
+    },
+    {
+      "epoch": 22.84,
+      "learning_rate": 7.341020608439646e-05,
+      "loss": 0.0593,
+      "step": 15900
+    },
+    {
+      "epoch": 22.84,
+      "eval_loss": 0.14600467681884766,
+      "eval_runtime": 40.4483,
+      "eval_samples_per_second": 11.298,
+      "eval_steps_per_second": 1.434,
+      "eval_wer": 0.1552477036590875,
+      "step": 15900
+    },
+    {
+      "epoch": 22.86,
+      "learning_rate": 7.326300294406279e-05,
+      "loss": 0.0698,
+      "step": 15910
+    },
+    {
+      "epoch": 22.87,
+      "learning_rate": 7.311579980372914e-05,
+      "loss": 0.0741,
+      "step": 15920
+    },
+    {
+      "epoch": 22.89,
+      "learning_rate": 7.296859666339547e-05,
+      "loss": 0.0498,
+      "step": 15930
+    },
+    {
+      "epoch": 22.9,
+      "learning_rate": 7.282139352306182e-05,
+      "loss": 0.0633,
+      "step": 15940
+    },
+    {
+      "epoch": 22.92,
+      "learning_rate": 7.267419038272815e-05,
+      "loss": 0.0705,
+      "step": 15950
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 7.25269872423945e-05,
+      "loss": 0.0648,
+      "step": 15960
+    },
+    {
+      "epoch": 22.95,
+      "learning_rate": 7.237978410206083e-05,
+      "loss": 0.0609,
+      "step": 15970
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 7.223258096172718e-05,
+      "loss": 0.06,
+      "step": 15980
+    },
+    {
+      "epoch": 22.97,
+      "learning_rate": 7.208537782139351e-05,
+      "loss": 0.0535,
+      "step": 15990
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 7.193817468105986e-05,
+      "loss": 0.0609,
+      "step": 16000
+    },
+    {
+      "epoch": 22.99,
+      "eval_loss": 0.14838065207004547,
+      "eval_runtime": 40.4523,
+      "eval_samples_per_second": 11.297,
+      "eval_steps_per_second": 1.434,
+      "eval_wer": 0.15675350097876825,
+      "step": 16000
+    },
+    {
+      "epoch": 23.0,
+      "learning_rate": 7.17909715407262e-05,
+      "loss": 0.0718,
+      "step": 16010
+    },
+    {
+      "epoch": 23.02,
+      "learning_rate": 7.164376840039253e-05,
+      "loss": 0.061,
+      "step": 16020
+    },
+    {
+      "epoch": 23.03,
+      "learning_rate": 7.149656526005887e-05,
+      "loss": 0.0591,
+      "step": 16030
+    },
+    {
+      "epoch": 23.05,
+      "learning_rate": 7.134936211972521e-05,
+      "loss": 0.0588,
+      "step": 16040
+    },
+    {
+      "epoch": 23.06,
+      "learning_rate": 7.120215897939155e-05,
+      "loss": 0.0599,
+      "step": 16050
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 7.105495583905789e-05,
+      "loss": 0.0639,
+      "step": 16060
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 7.090775269872424e-05,
+      "loss": 0.0617,
+      "step": 16070
+    },
+    {
+      "epoch": 23.1,
+      "learning_rate": 7.076054955839057e-05,
+      "loss": 0.0563,
+      "step": 16080
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 7.061334641805692e-05,
+      "loss": 0.0638,
+      "step": 16090
+    },
+    {
+      "epoch": 23.13,
+      "learning_rate": 7.046614327772325e-05,
+      "loss": 0.0587,
+      "step": 16100
+    },
+    {
+      "epoch": 23.13,
+      "eval_loss": 0.1451614946126938,
+      "eval_runtime": 40.4862,
+      "eval_samples_per_second": 11.288,
+      "eval_steps_per_second": 1.433,
+      "eval_wer": 0.15359132660743863,
+      "step": 16100
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 7.03189401373896e-05,
+      "loss": 0.069,
+      "step": 16110
+    },
+    {
+      "epoch": 23.16,
+      "learning_rate": 7.017173699705593e-05,
+      "loss": 0.0651,
+      "step": 16120
+    },
+    {
+      "epoch": 23.18,
+      "learning_rate": 7.002453385672228e-05,
+      "loss": 0.0551,
+      "step": 16130
+    },
+    {
+      "epoch": 23.19,
+      "learning_rate": 6.987733071638861e-05,
+      "loss": 0.064,
+      "step": 16140
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 6.973012757605496e-05,
+      "loss": 0.0604,
+      "step": 16150
+    },
+    {
+      "epoch": 23.22,
+      "learning_rate": 6.958292443572129e-05,
+      "loss": 0.0697,
+      "step": 16160
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 6.943572129538762e-05,
+      "loss": 0.0564,
+      "step": 16170
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 6.928851815505397e-05,
+      "loss": 0.0583,
+      "step": 16180
+    },
+    {
+      "epoch": 23.26,
+      "learning_rate": 6.91413150147203e-05,
+      "loss": 0.0566,
+      "step": 16190
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 6.899411187438665e-05,
+      "loss": 0.0636,
+      "step": 16200
+    },
+    {
+      "epoch": 23.28,
+      "eval_loss": 0.14748641848564148,
+      "eval_runtime": 40.5091,
+      "eval_samples_per_second": 11.281,
+      "eval_steps_per_second": 1.432,
+      "eval_wer": 0.15283842794759825,
+      "step": 16200
+    },
+    {
+      "epoch": 23.29,
+      "learning_rate": 6.884690873405298e-05,
+      "loss": 0.0613,
+      "step": 16210
+    },
+    {
+      "epoch": 23.3,
+      "learning_rate": 6.869970559371933e-05,
+      "loss": 0.0664,
+      "step": 16220
+    },
+    {
+      "epoch": 23.32,
+      "learning_rate": 6.855250245338566e-05,
+      "loss": 0.0562,
+      "step": 16230
+    },
+    {
+      "epoch": 23.33,
+      "learning_rate": 6.840529931305201e-05,
+      "loss": 0.0588,
+      "step": 16240
+    },
+    {
+      "epoch": 23.35,
+      "learning_rate": 6.825809617271834e-05,
+      "loss": 0.0571,
+      "step": 16250
+    },
+    {
+      "epoch": 23.36,
+      "learning_rate": 6.811089303238469e-05,
+      "loss": 0.0562,
+      "step": 16260
+    },
+    {
+      "epoch": 23.38,
+      "learning_rate": 6.796368989205102e-05,
+      "loss": 0.0643,
+      "step": 16270
+    },
+    {
+      "epoch": 23.39,
+      "learning_rate": 6.781648675171737e-05,
+      "loss": 0.0578,
+      "step": 16280
+    },
+    {
+      "epoch": 23.41,
+      "learning_rate": 6.76692836113837e-05,
+      "loss": 0.0546,
+      "step": 16290
+    },
+    {
+      "epoch": 23.42,
+      "learning_rate": 6.752208047105005e-05,
+      "loss": 0.0506,
+      "step": 16300
+    },
+    {
+      "epoch": 23.42,
+      "eval_loss": 0.1494196057319641,
+      "eval_runtime": 40.2779,
+      "eval_samples_per_second": 11.346,
+      "eval_steps_per_second": 1.44,
+      "eval_wer": 0.15449480499924712,
+      "step": 16300
+    },
+    {
+      "epoch": 23.43,
+      "learning_rate": 6.737487733071637e-05,
+      "loss": 0.0556,
+      "step": 16310
+    },
+    {
+      "epoch": 23.45,
+      "learning_rate": 6.722767419038272e-05,
+      "loss": 0.0671,
+      "step": 16320
+    },
+    {
+      "epoch": 23.46,
+      "learning_rate": 6.708047105004905e-05,
+      "loss": 0.0531,
+      "step": 16330
+    },
+    {
+      "epoch": 23.48,
+      "learning_rate": 6.69332679097154e-05,
+      "loss": 0.0664,
+      "step": 16340
+    },
+    {
+      "epoch": 23.49,
+      "learning_rate": 6.678606476938173e-05,
+      "loss": 0.0583,
+      "step": 16350
+    },
+    {
+      "epoch": 23.51,
+      "learning_rate": 6.663886162904808e-05,
+      "loss": 0.0605,
+      "step": 16360
+    },
+    {
+      "epoch": 23.52,
+      "learning_rate": 6.649165848871441e-05,
+      "loss": 0.0523,
+      "step": 16370
+    },
+    {
+      "epoch": 23.53,
+      "learning_rate": 6.634445534838076e-05,
+      "loss": 0.0456,
+      "step": 16380
+    },
+    {
+      "epoch": 23.55,
+      "learning_rate": 6.61972522080471e-05,
+      "loss": 0.0773,
+      "step": 16390
+    },
+    {
+      "epoch": 23.56,
+      "learning_rate": 6.605004906771344e-05,
+      "loss": 0.0585,
+      "step": 16400
+    },
+    {
+      "epoch": 23.56,
+      "eval_loss": 0.15359418094158173,
+      "eval_runtime": 40.8191,
+      "eval_samples_per_second": 11.196,
+      "eval_steps_per_second": 1.421,
+      "eval_wer": 0.16021683481403404,
+      "step": 16400
+    },
+    {
+      "epoch": 23.58,
+      "learning_rate": 6.590284592737977e-05,
+      "loss": 0.0512,
+      "step": 16410
+    },
+    {
+      "epoch": 23.59,
+      "learning_rate": 6.575564278704612e-05,
+      "loss": 0.0639,
+      "step": 16420
+    },
+    {
+      "epoch": 23.61,
+      "learning_rate": 6.560843964671245e-05,
+      "loss": 0.0606,
+      "step": 16430
+    },
+    {
+      "epoch": 23.62,
+      "learning_rate": 6.54612365063788e-05,
+      "loss": 0.0591,
+      "step": 16440
+    },
+    {
+      "epoch": 23.64,
+      "learning_rate": 6.531403336604513e-05,
+      "loss": 0.0458,
+      "step": 16450
+    },
+    {
+      "epoch": 23.65,
+      "learning_rate": 6.516683022571147e-05,
+      "loss": 0.0609,
+      "step": 16460
+    },
+    {
+      "epoch": 23.66,
+      "learning_rate": 6.501962708537781e-05,
+      "loss": 0.0537,
+      "step": 16470
+    },
+    {
+      "epoch": 23.68,
+      "learning_rate": 6.487242394504415e-05,
+      "loss": 0.0546,
+      "step": 16480
+    },
+    {
+      "epoch": 23.69,
+      "learning_rate": 6.47252208047105e-05,
+      "loss": 0.0694,
+      "step": 16490
+    },
+    {
+      "epoch": 23.71,
+      "learning_rate": 6.457801766437683e-05,
+      "loss": 0.0585,
+      "step": 16500
+    },
+    {
+      "epoch": 23.71,
+      "eval_loss": 0.14631050825119019,
+      "eval_runtime": 40.2888,
+      "eval_samples_per_second": 11.343,
+      "eval_steps_per_second": 1.44,
+      "eval_wer": 0.15735581990664058,
+      "step": 16500
+    },
+    {
+      "epoch": 23.72,
+      "learning_rate": 6.443081452404318e-05,
+      "loss": 0.0717,
+      "step": 16510
+    },
+    {
+      "epoch": 23.74,
+      "learning_rate": 6.428361138370951e-05,
+      "loss": 0.0656,
+      "step": 16520
+    },
+    {
+      "epoch": 23.75,
+      "learning_rate": 6.413640824337586e-05,
+      "loss": 0.0461,
+      "step": 16530
+    },
+    {
+      "epoch": 23.76,
+      "learning_rate": 6.398920510304219e-05,
+      "loss": 0.0743,
+      "step": 16540
+    },
+    {
+      "epoch": 23.78,
+      "learning_rate": 6.384200196270854e-05,
+      "loss": 0.0533,
+      "step": 16550
+    },
+    {
+      "epoch": 23.79,
+      "learning_rate": 6.369479882237487e-05,
+      "loss": 0.0672,
+      "step": 16560
+    },
+    {
+      "epoch": 23.81,
+      "learning_rate": 6.354759568204122e-05,
+      "loss": 0.0631,
+      "step": 16570
+    },
+    {
+      "epoch": 23.82,
+      "learning_rate": 6.340039254170755e-05,
+      "loss": 0.0507,
+      "step": 16580
+    },
+    {
+      "epoch": 23.84,
+      "learning_rate": 6.32531894013739e-05,
+      "loss": 0.071,
+      "step": 16590
+    },
+    {
+      "epoch": 23.85,
+      "learning_rate": 6.310598626104023e-05,
+      "loss": 0.0546,
+      "step": 16600
+    },
+    {
+      "epoch": 23.85,
+      "eval_loss": 0.14672046899795532,
+      "eval_runtime": 40.2017,
+      "eval_samples_per_second": 11.368,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.1550971239271194,
+      "step": 16600
+    },
+    {
+      "epoch": 23.86,
+      "learning_rate": 6.295878312070656e-05,
+      "loss": 0.049,
+      "step": 16610
+    },
+    {
+      "epoch": 23.88,
+      "learning_rate": 6.281157998037291e-05,
+      "loss": 0.0709,
+      "step": 16620
+    },
+    {
+      "epoch": 23.89,
+      "learning_rate": 6.266437684003924e-05,
+      "loss": 0.0539,
+      "step": 16630
+    },
+    {
+      "epoch": 23.91,
+      "learning_rate": 6.251717369970559e-05,
+      "loss": 0.0669,
+      "step": 16640
+    },
+    {
+      "epoch": 23.92,
+      "learning_rate": 6.236997055937192e-05,
+      "loss": 0.0532,
+      "step": 16650
+    },
+    {
+      "epoch": 23.94,
+      "learning_rate": 6.222276741903827e-05,
+      "loss": 0.047,
+      "step": 16660
+    },
+    {
+      "epoch": 23.95,
+      "learning_rate": 6.20755642787046e-05,
+      "loss": 0.0563,
+      "step": 16670
+    },
+    {
+      "epoch": 23.97,
+      "learning_rate": 6.192836113837095e-05,
+      "loss": 0.0506,
+      "step": 16680
+    },
+    {
+      "epoch": 23.98,
+      "learning_rate": 6.178115799803728e-05,
+      "loss": 0.0708,
+      "step": 16690
+    },
+    {
+      "epoch": 23.99,
+      "learning_rate": 6.163395485770363e-05,
+      "loss": 0.0816,
+      "step": 16700
+    },
+    {
+      "epoch": 23.99,
+      "eval_loss": 0.15094982087612152,
+      "eval_runtime": 40.1931,
+      "eval_samples_per_second": 11.37,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.15464538473121517,
+      "step": 16700
+    },
+    {
+      "epoch": 24.01,
+      "learning_rate": 6.148675171736997e-05,
+      "loss": 0.0557,
+      "step": 16710
+    },
+    {
+      "epoch": 24.02,
+      "learning_rate": 6.133954857703631e-05,
+      "loss": 0.0544,
+      "step": 16720
+    },
+    {
+      "epoch": 24.04,
+      "learning_rate": 6.119234543670265e-05,
+      "loss": 0.0681,
+      "step": 16730
+    },
+    {
+      "epoch": 24.05,
+      "learning_rate": 6.104514229636899e-05,
+      "loss": 0.0576,
+      "step": 16740
+    },
+    {
+      "epoch": 24.07,
+      "learning_rate": 6.0897939156035326e-05,
+      "loss": 0.0663,
+      "step": 16750
+    },
+    {
+      "epoch": 24.08,
+      "learning_rate": 6.075073601570166e-05,
+      "loss": 0.0617,
+      "step": 16760
+    },
+    {
+      "epoch": 24.09,
+      "learning_rate": 6.0603532875368e-05,
+      "loss": 0.05,
+      "step": 16770
+    },
+    {
+      "epoch": 24.11,
+      "learning_rate": 6.045632973503434e-05,
+      "loss": 0.0622,
+      "step": 16780
+    },
+    {
+      "epoch": 24.12,
+      "learning_rate": 6.030912659470068e-05,
+      "loss": 0.0681,
+      "step": 16790
+    },
+    {
+      "epoch": 24.14,
+      "learning_rate": 6.016192345436702e-05,
+      "loss": 0.0556,
+      "step": 16800
+    },
+    {
+      "epoch": 24.14,
+      "eval_loss": 0.15066751837730408,
+      "eval_runtime": 40.0599,
+      "eval_samples_per_second": 11.408,
+      "eval_steps_per_second": 1.448,
+      "eval_wer": 0.15600060231892787,
+      "step": 16800
+    },
+    {
+      "epoch": 24.15,
+      "learning_rate": 6.001472031403336e-05,
+      "loss": 0.0585,
+      "step": 16810
+    },
+    {
+      "epoch": 24.17,
+      "learning_rate": 5.98675171736997e-05,
+      "loss": 0.0546,
+      "step": 16820
+    },
+    {
+      "epoch": 24.18,
+      "learning_rate": 5.972031403336604e-05,
+      "loss": 0.0532,
+      "step": 16830
+    },
+    {
+      "epoch": 24.2,
+      "learning_rate": 5.957311089303238e-05,
+      "loss": 0.0664,
+      "step": 16840
+    },
+    {
+      "epoch": 24.21,
+      "learning_rate": 5.942590775269872e-05,
+      "loss": 0.0492,
+      "step": 16850
+    },
+    {
+      "epoch": 24.22,
+      "learning_rate": 5.927870461236506e-05,
+      "loss": 0.0616,
+      "step": 16860
+    },
+    {
+      "epoch": 24.24,
+      "learning_rate": 5.91315014720314e-05,
+      "loss": 0.0582,
+      "step": 16870
+    },
+    {
+      "epoch": 24.25,
+      "learning_rate": 5.898429833169774e-05,
+      "loss": 0.0613,
+      "step": 16880
+    },
+    {
+      "epoch": 24.27,
+      "learning_rate": 5.883709519136408e-05,
+      "loss": 0.0545,
+      "step": 16890
+    },
+    {
+      "epoch": 24.28,
+      "learning_rate": 5.8689892051030415e-05,
+      "loss": 0.056,
+      "step": 16900
+    },
+    {
+      "epoch": 24.28,
+      "eval_loss": 0.15041780471801758,
+      "eval_runtime": 40.1424,
+      "eval_samples_per_second": 11.384,
+      "eval_steps_per_second": 1.445,
+      "eval_wer": 0.16051799427797017,
+      "step": 16900
+    },
+    {
+      "epoch": 24.3,
+      "learning_rate": 5.8542688910696755e-05,
+      "loss": 0.0638,
+      "step": 16910
+    },
+    {
+      "epoch": 24.31,
+      "learning_rate": 5.8395485770363095e-05,
+      "loss": 0.0512,
+      "step": 16920
+    },
+    {
+      "epoch": 24.32,
+      "learning_rate": 5.8248282630029435e-05,
+      "loss": 0.0619,
+      "step": 16930
+    },
+    {
+      "epoch": 24.34,
+      "learning_rate": 5.8101079489695775e-05,
+      "loss": 0.0641,
+      "step": 16940
+    },
+    {
+      "epoch": 24.35,
+      "learning_rate": 5.7953876349362116e-05,
+      "loss": 0.0507,
+      "step": 16950
+    },
+    {
+      "epoch": 24.37,
+      "learning_rate": 5.7806673209028456e-05,
+      "loss": 0.0543,
+      "step": 16960
+    },
+    {
+      "epoch": 24.38,
+      "learning_rate": 5.7659470068694796e-05,
+      "loss": 0.0493,
+      "step": 16970
+    },
+    {
+      "epoch": 24.4,
+      "learning_rate": 5.7512266928361136e-05,
+      "loss": 0.0567,
+      "step": 16980
+    },
+    {
+      "epoch": 24.41,
+      "learning_rate": 5.7365063788027476e-05,
+      "loss": 0.0776,
+      "step": 16990
+    },
+    {
+      "epoch": 24.43,
+      "learning_rate": 5.7217860647693816e-05,
+      "loss": 0.0621,
+      "step": 17000
+    },
+    {
+      "epoch": 24.43,
+      "eval_loss": 0.14634354412555695,
+      "eval_runtime": 40.0304,
+      "eval_samples_per_second": 11.416,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.15825929829844904,
+      "step": 17000
+    },
+    {
+      "epoch": 24.44,
+      "learning_rate": 5.7070657507360157e-05,
+      "loss": 0.0731,
+      "step": 17010
+    },
+    {
+      "epoch": 24.45,
+      "learning_rate": 5.69234543670265e-05,
+      "loss": 0.0588,
+      "step": 17020
+    },
+    {
+      "epoch": 24.47,
+      "learning_rate": 5.677625122669284e-05,
+      "loss": 0.0555,
+      "step": 17030
+    },
+    {
+      "epoch": 24.48,
+      "learning_rate": 5.662904808635918e-05,
+      "loss": 0.0633,
+      "step": 17040
+    },
+    {
+      "epoch": 24.5,
+      "learning_rate": 5.6481844946025504e-05,
+      "loss": 0.0477,
+      "step": 17050
+    },
+    {
+      "epoch": 24.51,
+      "learning_rate": 5.6334641805691844e-05,
+      "loss": 0.0584,
+      "step": 17060
+    },
+    {
+      "epoch": 24.53,
+      "learning_rate": 5.6187438665358184e-05,
+      "loss": 0.0594,
+      "step": 17070
+    },
+    {
+      "epoch": 24.54,
+      "learning_rate": 5.6040235525024524e-05,
+      "loss": 0.0613,
+      "step": 17080
+    },
+    {
+      "epoch": 24.55,
+      "learning_rate": 5.5893032384690864e-05,
+      "loss": 0.0608,
+      "step": 17090
+    },
+    {
+      "epoch": 24.57,
+      "learning_rate": 5.5745829244357205e-05,
+      "loss": 0.0514,
+      "step": 17100
+    },
+    {
+      "epoch": 24.57,
+      "eval_loss": 0.14699435234069824,
+      "eval_runtime": 40.2333,
+      "eval_samples_per_second": 11.359,
+      "eval_steps_per_second": 1.442,
+      "eval_wer": 0.15856045776238517,
+      "step": 17100
+    },
+    {
+      "epoch": 24.58,
+      "learning_rate": 5.5598626104023545e-05,
+      "loss": 0.0718,
+      "step": 17110
+    },
+    {
+      "epoch": 24.6,
+      "learning_rate": 5.5451422963689885e-05,
+      "loss": 0.0498,
+      "step": 17120
+    },
+    {
+      "epoch": 24.61,
+      "learning_rate": 5.5304219823356225e-05,
+      "loss": 0.0534,
+      "step": 17130
+    },
+    {
+      "epoch": 24.63,
+      "learning_rate": 5.5157016683022565e-05,
+      "loss": 0.0606,
+      "step": 17140
+    },
+    {
+      "epoch": 24.64,
+      "learning_rate": 5.5009813542688905e-05,
+      "loss": 0.0532,
+      "step": 17150
+    },
+    {
+      "epoch": 24.66,
+      "learning_rate": 5.4862610402355246e-05,
+      "loss": 0.062,
+      "step": 17160
+    },
+    {
+      "epoch": 24.67,
+      "learning_rate": 5.4715407262021586e-05,
+      "loss": 0.0483,
+      "step": 17170
+    },
+    {
+      "epoch": 24.68,
+      "learning_rate": 5.4568204121687926e-05,
+      "loss": 0.0428,
+      "step": 17180
+    },
+    {
+      "epoch": 24.7,
+      "learning_rate": 5.4421000981354266e-05,
+      "loss": 0.0725,
+      "step": 17190
+    },
+    {
+      "epoch": 24.71,
+      "learning_rate": 5.42737978410206e-05,
+      "loss": 0.0487,
+      "step": 17200
+    },
+    {
+      "epoch": 24.71,
+      "eval_loss": 0.1460464745759964,
+      "eval_runtime": 40.1151,
+      "eval_samples_per_second": 11.392,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.1556994428549917,
+      "step": 17200
+    },
+    {
+      "epoch": 24.73,
+      "learning_rate": 5.412659470068694e-05,
+      "loss": 0.0566,
+      "step": 17210
+    },
+    {
+      "epoch": 24.74,
+      "learning_rate": 5.397939156035328e-05,
+      "loss": 0.056,
+      "step": 17220
+    },
+    {
+      "epoch": 24.76,
+      "learning_rate": 5.383218842001962e-05,
+      "loss": 0.0526,
+      "step": 17230
+    },
+    {
+      "epoch": 24.77,
+      "learning_rate": 5.368498527968596e-05,
+      "loss": 0.0623,
+      "step": 17240
+    },
+    {
+      "epoch": 24.78,
+      "learning_rate": 5.35377821393523e-05,
+      "loss": 0.0532,
+      "step": 17250
+    },
+    {
+      "epoch": 24.8,
+      "learning_rate": 5.339057899901864e-05,
+      "loss": 0.0671,
+      "step": 17260
+    },
+    {
+      "epoch": 24.81,
+      "learning_rate": 5.324337585868498e-05,
+      "loss": 0.0525,
+      "step": 17270
+    },
+    {
+      "epoch": 24.83,
+      "learning_rate": 5.309617271835132e-05,
+      "loss": 0.0492,
+      "step": 17280
+    },
+    {
+      "epoch": 24.84,
+      "learning_rate": 5.294896957801766e-05,
+      "loss": 0.0518,
+      "step": 17290
+    },
+    {
+      "epoch": 24.86,
+      "learning_rate": 5.2801766437684e-05,
+      "loss": 0.0474,
+      "step": 17300
+    },
+    {
+      "epoch": 24.86,
+      "eval_loss": 0.1450837403535843,
+      "eval_runtime": 40.0465,
+      "eval_samples_per_second": 11.412,
+      "eval_steps_per_second": 1.448,
+      "eval_wer": 0.15238668875169403,
+      "step": 17300
+    },
+    {
+      "epoch": 24.87,
+      "learning_rate": 5.265456329735034e-05,
+      "loss": 0.0572,
+      "step": 17310
+    },
+    {
+      "epoch": 24.89,
+      "learning_rate": 5.250736015701668e-05,
+      "loss": 0.0502,
+      "step": 17320
+    },
+    {
+      "epoch": 24.9,
+      "learning_rate": 5.236015701668302e-05,
+      "loss": 0.0592,
+      "step": 17330
+    },
+    {
+      "epoch": 24.91,
+      "learning_rate": 5.221295387634936e-05,
+      "loss": 0.0609,
+      "step": 17340
+    },
+    {
+      "epoch": 24.93,
+      "learning_rate": 5.2065750736015695e-05,
+      "loss": 0.0617,
+      "step": 17350
+    },
+    {
+      "epoch": 24.94,
+      "learning_rate": 5.1918547595682035e-05,
+      "loss": 0.0525,
+      "step": 17360
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 5.1771344455348375e-05,
+      "loss": 0.0649,
+      "step": 17370
+    },
+    {
+      "epoch": 24.97,
+      "learning_rate": 5.1624141315014716e-05,
+      "loss": 0.0487,
+      "step": 17380
+    },
+    {
+      "epoch": 24.99,
+      "learning_rate": 5.1476938174681056e-05,
+      "loss": 0.062,
+      "step": 17390
+    },
+    {
+      "epoch": 25.0,
+      "learning_rate": 5.1329735034347396e-05,
+      "loss": 0.0402,
+      "step": 17400
+    },
+    {
+      "epoch": 25.0,
+      "eval_loss": 0.14534904062747955,
+      "eval_runtime": 40.0375,
+      "eval_samples_per_second": 11.414,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.1513326306279175,
+      "step": 17400
+    },
+    {
+      "epoch": 25.01,
+      "learning_rate": 5.1182531894013736e-05,
+      "loss": 0.0566,
+      "step": 17410
+    },
+    {
+      "epoch": 25.03,
+      "learning_rate": 5.1035328753680076e-05,
+      "loss": 0.0545,
+      "step": 17420
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 5.0888125613346416e-05,
+      "loss": 0.0524,
+      "step": 17430
+    },
+    {
+      "epoch": 25.06,
+      "learning_rate": 5.074092247301276e-05,
+      "loss": 0.0482,
+      "step": 17440
+    },
+    {
+      "epoch": 25.07,
+      "learning_rate": 5.05937193326791e-05,
+      "loss": 0.0479,
+      "step": 17450
+    },
+    {
+      "epoch": 25.09,
+      "learning_rate": 5.044651619234544e-05,
+      "loss": 0.0574,
+      "step": 17460
+    },
+    {
+      "epoch": 25.1,
+      "learning_rate": 5.029931305201178e-05,
+      "loss": 0.0516,
+      "step": 17470
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 5.015210991167812e-05,
+      "loss": 0.0564,
+      "step": 17480
+    },
+    {
+      "epoch": 25.13,
+      "learning_rate": 5.000490677134446e-05,
+      "loss": 0.0478,
+      "step": 17490
+    },
+    {
+      "epoch": 25.14,
+      "learning_rate": 4.9857703631010784e-05,
+      "loss": 0.0754,
+      "step": 17500
+    },
+    {
+      "epoch": 25.14,
+      "eval_loss": 0.14766934514045715,
+      "eval_runtime": 39.973,
+      "eval_samples_per_second": 11.433,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.15148321035988557,
+      "step": 17500
+    },
+    {
+      "epoch": 25.16,
+      "learning_rate": 4.9710500490677124e-05,
+      "loss": 0.0675,
+      "step": 17510
+    },
+    {
+      "epoch": 25.17,
+      "learning_rate": 4.9563297350343464e-05,
+      "loss": 0.0486,
+      "step": 17520
+    },
+    {
+      "epoch": 25.19,
+      "learning_rate": 4.9416094210009805e-05,
+      "loss": 0.0603,
+      "step": 17530
+    },
+    {
+      "epoch": 25.2,
+      "learning_rate": 4.9268891069676145e-05,
+      "loss": 0.0655,
+      "step": 17540
+    },
+    {
+      "epoch": 25.22,
+      "learning_rate": 4.9121687929342485e-05,
+      "loss": 0.0523,
+      "step": 17550
+    },
+    {
+      "epoch": 25.23,
+      "learning_rate": 4.8974484789008825e-05,
+      "loss": 0.0565,
+      "step": 17560
+    },
+    {
+      "epoch": 25.24,
+      "learning_rate": 4.8827281648675165e-05,
+      "loss": 0.051,
+      "step": 17570
+    },
+    {
+      "epoch": 25.26,
+      "learning_rate": 4.8680078508341505e-05,
+      "loss": 0.0615,
+      "step": 17580
+    },
+    {
+      "epoch": 25.27,
+      "learning_rate": 4.8532875368007846e-05,
+      "loss": 0.0576,
+      "step": 17590
+    },
+    {
+      "epoch": 25.29,
+      "learning_rate": 4.8385672227674186e-05,
+      "loss": 0.0409,
+      "step": 17600
+    },
+    {
+      "epoch": 25.29,
+      "eval_loss": 0.14516086876392365,
+      "eval_runtime": 40.1057,
+      "eval_samples_per_second": 11.395,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.14907393464839633,
+      "step": 17600
+    },
+    {
+      "epoch": 25.3,
+      "learning_rate": 4.8238469087340526e-05,
+      "loss": 0.0673,
+      "step": 17610
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 4.8091265947006866e-05,
+      "loss": 0.0686,
+      "step": 17620
+    },
+    {
+      "epoch": 25.33,
+      "learning_rate": 4.7944062806673206e-05,
+      "loss": 0.0512,
+      "step": 17630
+    },
+    {
+      "epoch": 25.34,
+      "learning_rate": 4.779685966633954e-05,
+      "loss": 0.0556,
+      "step": 17640
+    },
+    {
+      "epoch": 25.36,
+      "learning_rate": 4.764965652600588e-05,
+      "loss": 0.0619,
+      "step": 17650
+    },
+    {
+      "epoch": 25.37,
+      "learning_rate": 4.750245338567222e-05,
+      "loss": 0.0516,
+      "step": 17660
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 4.735525024533856e-05,
+      "loss": 0.0491,
+      "step": 17670
+    },
+    {
+      "epoch": 25.4,
+      "learning_rate": 4.72080471050049e-05,
+      "loss": 0.0561,
+      "step": 17680
+    },
+    {
+      "epoch": 25.42,
+      "learning_rate": 4.706084396467124e-05,
+      "loss": 0.0609,
+      "step": 17690
+    },
+    {
+      "epoch": 25.43,
+      "learning_rate": 4.691364082433758e-05,
+      "loss": 0.0414,
+      "step": 17700
+    },
+    {
+      "epoch": 25.43,
+      "eval_loss": 0.1499926894903183,
+      "eval_runtime": 39.8722,
+      "eval_samples_per_second": 11.462,
+      "eval_steps_per_second": 1.455,
+      "eval_wer": 0.1511820508959494,
+      "step": 17700
+    },
+    {
+      "epoch": 25.45,
+      "learning_rate": 4.676643768400392e-05,
+      "loss": 0.0555,
+      "step": 17710
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 4.661923454367026e-05,
+      "loss": 0.0568,
+      "step": 17720
+    },
+    {
+      "epoch": 25.47,
+      "learning_rate": 4.64720314033366e-05,
+      "loss": 0.0551,
+      "step": 17730
+    },
+    {
+      "epoch": 25.49,
+      "learning_rate": 4.632482826300294e-05,
+      "loss": 0.0441,
+      "step": 17740
+    },
+    {
+      "epoch": 25.5,
+      "learning_rate": 4.617762512266928e-05,
+      "loss": 0.0469,
+      "step": 17750
+    },
+    {
+      "epoch": 25.52,
+      "learning_rate": 4.603042198233562e-05,
+      "loss": 0.0562,
+      "step": 17760
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 4.588321884200196e-05,
+      "loss": 0.0511,
+      "step": 17770
+    },
+    {
+      "epoch": 25.55,
+      "learning_rate": 4.57360157016683e-05,
+      "loss": 0.0456,
+      "step": 17780
+    },
+    {
+      "epoch": 25.56,
+      "learning_rate": 4.5588812561334635e-05,
+      "loss": 0.0419,
+      "step": 17790
+    },
+    {
+      "epoch": 25.57,
+      "learning_rate": 4.5441609421000976e-05,
+      "loss": 0.0596,
+      "step": 17800
+    },
+    {
+      "epoch": 25.57,
+      "eval_loss": 0.14854370057582855,
+      "eval_runtime": 40.2407,
+      "eval_samples_per_second": 11.357,
+      "eval_steps_per_second": 1.441,
+      "eval_wer": 0.1538924860713748,
+      "step": 17800
+    },
+    {
+      "epoch": 25.59,
+      "learning_rate": 4.5294406280667316e-05,
+      "loss": 0.0575,
+      "step": 17810
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 4.5147203140333656e-05,
+      "loss": 0.0565,
+      "step": 17820
+    },
+    {
+      "epoch": 25.62,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 0.0598,
+      "step": 17830
+    },
+    {
+      "epoch": 25.63,
+      "learning_rate": 4.4852796859666336e-05,
+      "loss": 0.0482,
+      "step": 17840
+    },
+    {
+      "epoch": 25.65,
+      "learning_rate": 4.4705593719332676e-05,
+      "loss": 0.0505,
+      "step": 17850
+    },
+    {
+      "epoch": 25.66,
+      "learning_rate": 4.4558390578999017e-05,
+      "loss": 0.0588,
+      "step": 17860
+    },
+    {
+      "epoch": 25.68,
+      "learning_rate": 4.441118743866536e-05,
+      "loss": 0.0481,
+      "step": 17870
+    },
+    {
+      "epoch": 25.69,
+      "learning_rate": 4.42639842983317e-05,
+      "loss": 0.05,
+      "step": 17880
+    },
+    {
+      "epoch": 25.7,
+      "learning_rate": 4.411678115799804e-05,
+      "loss": 0.0587,
+      "step": 17890
+    },
+    {
+      "epoch": 25.72,
+      "learning_rate": 4.396957801766438e-05,
+      "loss": 0.0441,
+      "step": 17900
+    },
+    {
+      "epoch": 25.72,
+      "eval_loss": 0.14327119290828705,
+      "eval_runtime": 39.9632,
+      "eval_samples_per_second": 11.436,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.15344074687547057,
+      "step": 17900
+    },
+    {
+      "epoch": 25.73,
+      "learning_rate": 4.382237487733072e-05,
+      "loss": 0.0595,
+      "step": 17910
+    },
+    {
+      "epoch": 25.75,
+      "learning_rate": 4.367517173699706e-05,
+      "loss": 0.0519,
+      "step": 17920
+    },
+    {
+      "epoch": 25.76,
+      "learning_rate": 4.35279685966634e-05,
+      "loss": 0.052,
+      "step": 17930
+    },
+    {
+      "epoch": 25.78,
+      "learning_rate": 4.3380765456329724e-05,
+      "loss": 0.0412,
+      "step": 17940
+    },
+    {
+      "epoch": 25.79,
+      "learning_rate": 4.3233562315996065e-05,
+      "loss": 0.0462,
+      "step": 17950
+    },
+    {
+      "epoch": 25.8,
+      "learning_rate": 4.3086359175662405e-05,
+      "loss": 0.0646,
+      "step": 17960
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 4.2939156035328745e-05,
+      "loss": 0.0572,
+      "step": 17970
+    },
+    {
+      "epoch": 25.83,
+      "learning_rate": 4.2791952894995085e-05,
+      "loss": 0.062,
+      "step": 17980
+    },
+    {
+      "epoch": 25.85,
+      "learning_rate": 4.2644749754661425e-05,
+      "loss": 0.0485,
+      "step": 17990
+    },
+    {
+      "epoch": 25.86,
+      "learning_rate": 4.2497546614327765e-05,
+      "loss": 0.0517,
+      "step": 18000
+    },
+    {
+      "epoch": 25.86,
+      "eval_loss": 0.1446276158094406,
+      "eval_runtime": 40.1161,
+      "eval_samples_per_second": 11.392,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.15238668875169403,
+      "step": 18000
+    },
+    {
+      "epoch": 25.88,
+      "learning_rate": 4.2350343473994106e-05,
+      "loss": 0.0556,
+      "step": 18010
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 4.2203140333660446e-05,
+      "loss": 0.0542,
+      "step": 18020
+    },
+    {
+      "epoch": 25.91,
+      "learning_rate": 4.2055937193326786e-05,
+      "loss": 0.0517,
+      "step": 18030
+    },
+    {
+      "epoch": 25.92,
+      "learning_rate": 4.1908734052993126e-05,
+      "loss": 0.0483,
+      "step": 18040
+    },
+    {
+      "epoch": 25.93,
+      "learning_rate": 4.1761530912659466e-05,
+      "loss": 0.0454,
+      "step": 18050
+    },
+    {
+      "epoch": 25.95,
+      "learning_rate": 4.1614327772325806e-05,
+      "loss": 0.0649,
+      "step": 18060
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 4.1467124631992147e-05,
+      "loss": 0.0449,
+      "step": 18070
+    },
+    {
+      "epoch": 25.98,
+      "learning_rate": 4.131992149165849e-05,
+      "loss": 0.0604,
+      "step": 18080
+    },
+    {
+      "epoch": 25.99,
+      "learning_rate": 4.117271835132482e-05,
+      "loss": 0.0486,
+      "step": 18090
+    },
+    {
+      "epoch": 26.01,
+      "learning_rate": 4.102551521099116e-05,
+      "loss": 0.0752,
+      "step": 18100
+    },
+    {
+      "epoch": 26.01,
+      "eval_loss": 0.14016593992710114,
+      "eval_runtime": 40.0408,
+      "eval_samples_per_second": 11.413,
+      "eval_steps_per_second": 1.449,
+      "eval_wer": 0.14967625357626863,
+      "step": 18100
+    },
+    {
+      "epoch": 26.02,
+      "learning_rate": 4.08783120706575e-05,
+      "loss": 0.0568,
+      "step": 18110
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 4.073110893032384e-05,
+      "loss": 0.0504,
+      "step": 18120
+    },
+    {
+      "epoch": 26.05,
+      "learning_rate": 4.058390578999018e-05,
+      "loss": 0.0536,
+      "step": 18130
+    },
+    {
+      "epoch": 26.06,
+      "learning_rate": 4.043670264965652e-05,
+      "loss": 0.0465,
+      "step": 18140
+    },
+    {
+      "epoch": 26.08,
+      "learning_rate": 4.028949950932286e-05,
+      "loss": 0.05,
+      "step": 18150
+    },
+    {
+      "epoch": 26.09,
+      "learning_rate": 4.01422963689892e-05,
+      "loss": 0.047,
+      "step": 18160
+    },
+    {
+      "epoch": 26.11,
+      "learning_rate": 3.999509322865554e-05,
+      "loss": 0.0545,
+      "step": 18170
+    },
+    {
+      "epoch": 26.12,
+      "learning_rate": 3.984789008832188e-05,
+      "loss": 0.0559,
+      "step": 18180
+    },
+    {
+      "epoch": 26.14,
+      "learning_rate": 3.970068694798822e-05,
+      "loss": 0.049,
+      "step": 18190
+    },
+    {
+      "epoch": 26.15,
+      "learning_rate": 3.955348380765456e-05,
+      "loss": 0.0543,
+      "step": 18200
+    },
+    {
+      "epoch": 26.15,
+      "eval_loss": 0.14435161650180817,
+      "eval_runtime": 40.7685,
+      "eval_samples_per_second": 11.21,
+      "eval_steps_per_second": 1.423,
+      "eval_wer": 0.1519349495557898,
+      "step": 18200
+    },
+    {
+      "epoch": 26.16,
+      "learning_rate": 3.94062806673209e-05,
+      "loss": 0.058,
+      "step": 18210
+    },
+    {
+      "epoch": 26.18,
+      "learning_rate": 3.925907752698724e-05,
+      "loss": 0.0459,
+      "step": 18220
+    },
+    {
+      "epoch": 26.19,
+      "learning_rate": 3.9111874386653576e-05,
+      "loss": 0.064,
+      "step": 18230
+    },
+    {
+      "epoch": 26.21,
+      "learning_rate": 3.8964671246319916e-05,
+      "loss": 0.0452,
+      "step": 18240
+    },
+    {
+      "epoch": 26.22,
+      "learning_rate": 3.8817468105986256e-05,
+      "loss": 0.0497,
+      "step": 18250
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 3.8670264965652596e-05,
+      "loss": 0.0505,
+      "step": 18260
+    },
+    {
+      "epoch": 26.25,
+      "learning_rate": 3.8523061825318936e-05,
+      "loss": 0.0551,
+      "step": 18270
+    },
+    {
+      "epoch": 26.26,
+      "learning_rate": 3.8375858684985276e-05,
+      "loss": 0.0565,
+      "step": 18280
+    },
+    {
+      "epoch": 26.28,
+      "learning_rate": 3.822865554465162e-05,
+      "loss": 0.0501,
+      "step": 18290
+    },
+    {
+      "epoch": 26.29,
+      "learning_rate": 3.808145240431796e-05,
+      "loss": 0.0556,
+      "step": 18300
+    },
+    {
+      "epoch": 26.29,
+      "eval_loss": 0.1450069099664688,
+      "eval_runtime": 40.1396,
+      "eval_samples_per_second": 11.385,
+      "eval_steps_per_second": 1.445,
+      "eval_wer": 0.14967625357626863,
+      "step": 18300
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 3.79342492639843e-05,
+      "loss": 0.0543,
+      "step": 18310
+    },
+    {
+      "epoch": 26.32,
+      "learning_rate": 3.778704612365064e-05,
+      "loss": 0.0547,
+      "step": 18320
+    },
+    {
+      "epoch": 26.34,
+      "learning_rate": 3.763984298331698e-05,
+      "loss": 0.0608,
+      "step": 18330
+    },
+    {
+      "epoch": 26.35,
+      "learning_rate": 3.749263984298331e-05,
+      "loss": 0.0458,
+      "step": 18340
+    },
+    {
+      "epoch": 26.36,
+      "learning_rate": 3.734543670264965e-05,
+      "loss": 0.0481,
+      "step": 18350
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 3.719823356231599e-05,
+      "loss": 0.0601,
+      "step": 18360
+    },
+    {
+      "epoch": 26.39,
+      "learning_rate": 3.705103042198233e-05,
+      "loss": 0.0451,
+      "step": 18370
+    },
+    {
+      "epoch": 26.41,
+      "learning_rate": 3.690382728164867e-05,
+      "loss": 0.0623,
+      "step": 18380
+    },
+    {
+      "epoch": 26.42,
+      "learning_rate": 3.675662414131501e-05,
+      "loss": 0.0396,
+      "step": 18390
+    },
+    {
+      "epoch": 26.44,
+      "learning_rate": 3.660942100098135e-05,
+      "loss": 0.0514,
+      "step": 18400
+    },
+    {
+      "epoch": 26.44,
+      "eval_loss": 0.14546437561511993,
+      "eval_runtime": 40.3801,
+      "eval_samples_per_second": 11.317,
+      "eval_steps_per_second": 1.436,
+      "eval_wer": 0.15042915223610903,
+      "step": 18400
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 3.646221786064769e-05,
+      "loss": 0.065,
+      "step": 18410
+    },
+    {
+      "epoch": 26.47,
+      "learning_rate": 3.6315014720314025e-05,
+      "loss": 0.0446,
+      "step": 18420
+    },
+    {
+      "epoch": 26.48,
+      "learning_rate": 3.6167811579980365e-05,
+      "loss": 0.0568,
+      "step": 18430
+    },
+    {
+      "epoch": 26.49,
+      "learning_rate": 3.6020608439646706e-05,
+      "loss": 0.0467,
+      "step": 18440
+    },
+    {
+      "epoch": 26.51,
+      "learning_rate": 3.5873405299313046e-05,
+      "loss": 0.0515,
+      "step": 18450
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 3.5726202158979386e-05,
+      "loss": 0.0528,
+      "step": 18460
+    },
+    {
+      "epoch": 26.54,
+      "learning_rate": 3.5578999018645726e-05,
+      "loss": 0.0518,
+      "step": 18470
+    },
+    {
+      "epoch": 26.55,
+      "learning_rate": 3.5431795878312066e-05,
+      "loss": 0.0506,
+      "step": 18480
+    },
+    {
+      "epoch": 26.57,
+      "learning_rate": 3.5284592737978406e-05,
+      "loss": 0.0501,
+      "step": 18490
+    },
+    {
+      "epoch": 26.58,
+      "learning_rate": 3.5137389597644747e-05,
+      "loss": 0.0424,
+      "step": 18500
+    },
+    {
+      "epoch": 26.58,
+      "eval_loss": 0.14528827369213104,
+      "eval_runtime": 40.49,
+      "eval_samples_per_second": 11.287,
+      "eval_steps_per_second": 1.432,
+      "eval_wer": 0.15057973196807709,
+      "step": 18500
+    },
+    {
+      "epoch": 26.59,
+      "learning_rate": 3.499018645731109e-05,
+      "loss": 0.0556,
+      "step": 18510
+    },
+    {
+      "epoch": 26.61,
+      "learning_rate": 3.484298331697743e-05,
+      "loss": 0.0624,
+      "step": 18520
+    },
+    {
+      "epoch": 26.62,
+      "learning_rate": 3.469578017664377e-05,
+      "loss": 0.0551,
+      "step": 18530
+    },
+    {
+      "epoch": 26.64,
+      "learning_rate": 3.454857703631011e-05,
+      "loss": 0.0548,
+      "step": 18540
+    },
+    {
+      "epoch": 26.65,
+      "learning_rate": 3.440137389597645e-05,
+      "loss": 0.0501,
+      "step": 18550
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 3.425417075564279e-05,
+      "loss": 0.0622,
+      "step": 18560
+    },
+    {
+      "epoch": 26.68,
+      "learning_rate": 3.410696761530912e-05,
+      "loss": 0.0475,
+      "step": 18570
+    },
+    {
+      "epoch": 26.7,
+      "learning_rate": 3.395976447497546e-05,
+      "loss": 0.0534,
+      "step": 18580
+    },
+    {
+      "epoch": 26.71,
+      "learning_rate": 3.38125613346418e-05,
+      "loss": 0.0405,
+      "step": 18590
+    },
+    {
+      "epoch": 26.72,
+      "learning_rate": 3.366535819430814e-05,
+      "loss": 0.0689,
+      "step": 18600
+    },
+    {
+      "epoch": 26.72,
+      "eval_loss": 0.1431579440832138,
+      "eval_runtime": 39.9989,
+      "eval_samples_per_second": 11.425,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.15088089143201325,
+      "step": 18600
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 3.351815505397448e-05,
+      "loss": 0.0541,
+      "step": 18610
+    },
+    {
+      "epoch": 26.75,
+      "learning_rate": 3.337095191364082e-05,
+      "loss": 0.0604,
+      "step": 18620
+    },
+    {
+      "epoch": 26.77,
+      "learning_rate": 3.322374877330716e-05,
+      "loss": 0.0546,
+      "step": 18630
+    },
+    {
+      "epoch": 26.78,
+      "learning_rate": 3.30765456329735e-05,
+      "loss": 0.0476,
+      "step": 18640
+    },
+    {
+      "epoch": 26.8,
+      "learning_rate": 3.2929342492639836e-05,
+      "loss": 0.0532,
+      "step": 18650
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 3.2782139352306176e-05,
+      "loss": 0.0361,
+      "step": 18660
+    },
+    {
+      "epoch": 26.82,
+      "learning_rate": 3.2634936211972516e-05,
+      "loss": 0.0366,
+      "step": 18670
+    },
+    {
+      "epoch": 26.84,
+      "learning_rate": 3.2487733071638856e-05,
+      "loss": 0.0511,
+      "step": 18680
+    },
+    {
+      "epoch": 26.85,
+      "learning_rate": 3.2340529931305196e-05,
+      "loss": 0.0484,
+      "step": 18690
+    },
+    {
+      "epoch": 26.87,
+      "learning_rate": 3.2193326790971536e-05,
+      "loss": 0.0467,
+      "step": 18700
+    },
+    {
+      "epoch": 26.87,
+      "eval_loss": 0.1416984647512436,
+      "eval_runtime": 40.7855,
+      "eval_samples_per_second": 11.205,
+      "eval_steps_per_second": 1.422,
+      "eval_wer": 0.15088089143201325,
+      "step": 18700
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 3.2046123650637877e-05,
+      "loss": 0.0566,
+      "step": 18710
+    },
+    {
+      "epoch": 26.9,
+      "learning_rate": 3.189892051030422e-05,
+      "loss": 0.0387,
+      "step": 18720
+    },
+    {
+      "epoch": 26.91,
+      "learning_rate": 3.175171736997056e-05,
+      "loss": 0.0504,
+      "step": 18730
+    },
+    {
+      "epoch": 26.93,
+      "learning_rate": 3.16045142296369e-05,
+      "loss": 0.0447,
+      "step": 18740
+    },
+    {
+      "epoch": 26.94,
+      "learning_rate": 3.145731108930324e-05,
+      "loss": 0.0454,
+      "step": 18750
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 3.131010794896958e-05,
+      "loss": 0.0518,
+      "step": 18760
+    },
+    {
+      "epoch": 26.97,
+      "learning_rate": 3.116290480863592e-05,
+      "loss": 0.0476,
+      "step": 18770
+    },
+    {
+      "epoch": 26.98,
+      "learning_rate": 3.101570166830226e-05,
+      "loss": 0.0463,
+      "step": 18780
+    },
+    {
+      "epoch": 27.0,
+      "learning_rate": 3.086849852796859e-05,
+      "loss": 0.0527,
+      "step": 18790
+    },
+    {
+      "epoch": 27.01,
+      "learning_rate": 3.072129538763493e-05,
+      "loss": 0.0409,
+      "step": 18800
+    },
+    {
+      "epoch": 27.01,
+      "eval_loss": 0.14151474833488464,
+      "eval_runtime": 40.0113,
+      "eval_samples_per_second": 11.422,
+      "eval_steps_per_second": 1.45,
+      "eval_wer": 0.15298900767956633,
+      "step": 18800
+    },
+    {
+      "epoch": 27.03,
+      "learning_rate": 3.057409224730127e-05,
+      "loss": 0.047,
+      "step": 18810
+    },
+    {
+      "epoch": 27.04,
+      "learning_rate": 3.042688910696761e-05,
+      "loss": 0.0531,
+      "step": 18820
+    },
+    {
+      "epoch": 27.05,
+      "learning_rate": 3.0279685966633952e-05,
+      "loss": 0.0498,
+      "step": 18830
+    },
+    {
+      "epoch": 27.07,
+      "learning_rate": 3.0132482826300292e-05,
+      "loss": 0.0562,
+      "step": 18840
+    },
+    {
+      "epoch": 27.08,
+      "learning_rate": 2.9985279685966632e-05,
+      "loss": 0.0657,
+      "step": 18850
+    },
+    {
+      "epoch": 27.1,
+      "learning_rate": 2.983807654563297e-05,
+      "loss": 0.0474,
+      "step": 18860
+    },
+    {
+      "epoch": 27.11,
+      "learning_rate": 2.969087340529931e-05,
+      "loss": 0.0656,
+      "step": 18870
+    },
+    {
+      "epoch": 27.13,
+      "learning_rate": 2.954367026496565e-05,
+      "loss": 0.0483,
+      "step": 18880
+    },
+    {
+      "epoch": 27.14,
+      "learning_rate": 2.939646712463199e-05,
+      "loss": 0.0482,
+      "step": 18890
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 2.924926398429833e-05,
+      "loss": 0.0612,
+      "step": 18900
+    },
+    {
+      "epoch": 27.16,
+      "eval_loss": 0.1420021653175354,
+      "eval_runtime": 40.3238,
+      "eval_samples_per_second": 11.333,
+      "eval_steps_per_second": 1.438,
+      "eval_wer": 0.14952567384430054,
+      "step": 18900
+    },
+    {
+      "epoch": 27.17,
+      "learning_rate": 2.910206084396467e-05,
+      "loss": 0.0621,
+      "step": 18910
+    },
+    {
+      "epoch": 27.18,
+      "learning_rate": 2.895485770363101e-05,
+      "loss": 0.0583,
+      "step": 18920
+    },
+    {
+      "epoch": 27.2,
+      "learning_rate": 2.880765456329735e-05,
+      "loss": 0.0569,
+      "step": 18930
+    },
+    {
+      "epoch": 27.21,
+      "learning_rate": 2.8660451422963687e-05,
+      "loss": 0.0529,
+      "step": 18940
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 2.8513248282630027e-05,
+      "loss": 0.07,
+      "step": 18950
+    },
+    {
+      "epoch": 27.24,
+      "learning_rate": 2.8366045142296367e-05,
+      "loss": 0.0585,
+      "step": 18960
+    },
+    {
+      "epoch": 27.26,
+      "learning_rate": 2.8218842001962707e-05,
+      "loss": 0.054,
+      "step": 18970
+    },
+    {
+      "epoch": 27.27,
+      "learning_rate": 2.8071638861629047e-05,
+      "loss": 0.0523,
+      "step": 18980
+    },
+    {
+      "epoch": 27.28,
+      "learning_rate": 2.7924435721295388e-05,
+      "loss": 0.0416,
+      "step": 18990
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 2.7777232580961728e-05,
+      "loss": 0.0501,
+      "step": 19000
+    },
+    {
+      "epoch": 27.3,
+      "eval_loss": 0.14232105016708374,
+      "eval_runtime": 40.1869,
+      "eval_samples_per_second": 11.372,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.14892335491642825,
+      "step": 19000
+    },
+    {
+      "epoch": 27.31,
+      "learning_rate": 2.763002944062806e-05,
+      "loss": 0.0531,
+      "step": 19010
+    },
+    {
+      "epoch": 27.33,
+      "learning_rate": 2.74828263002944e-05,
+      "loss": 0.0529,
+      "step": 19020
+    },
+    {
+      "epoch": 27.34,
+      "learning_rate": 2.733562315996074e-05,
+      "loss": 0.0519,
+      "step": 19030
+    },
+    {
+      "epoch": 27.36,
+      "learning_rate": 2.718842001962708e-05,
+      "loss": 0.0456,
+      "step": 19040
+    },
+    {
+      "epoch": 27.37,
+      "learning_rate": 2.7041216879293422e-05,
+      "loss": 0.0546,
+      "step": 19050
+    },
+    {
+      "epoch": 27.39,
+      "learning_rate": 2.6894013738959762e-05,
+      "loss": 0.0488,
+      "step": 19060
+    },
+    {
+      "epoch": 27.4,
+      "learning_rate": 2.6746810598626102e-05,
+      "loss": 0.052,
+      "step": 19070
+    },
+    {
+      "epoch": 27.41,
+      "learning_rate": 2.659960745829244e-05,
+      "loss": 0.0537,
+      "step": 19080
+    },
+    {
+      "epoch": 27.43,
+      "learning_rate": 2.645240431795878e-05,
+      "loss": 0.0474,
+      "step": 19090
+    },
+    {
+      "epoch": 27.44,
+      "learning_rate": 2.630520117762512e-05,
+      "loss": 0.0579,
+      "step": 19100
+    },
+    {
+      "epoch": 27.44,
+      "eval_loss": 0.14417009055614471,
+      "eval_runtime": 40.3442,
+      "eval_samples_per_second": 11.328,
+      "eval_steps_per_second": 1.438,
+      "eval_wer": 0.15238668875169403,
+      "step": 19100
+    },
+    {
+      "epoch": 27.46,
+      "learning_rate": 2.615799803729146e-05,
+      "loss": 0.0403,
+      "step": 19110
+    },
+    {
+      "epoch": 27.47,
+      "learning_rate": 2.60107948969578e-05,
+      "loss": 0.0537,
+      "step": 19120
+    },
+    {
+      "epoch": 27.49,
+      "learning_rate": 2.586359175662414e-05,
+      "loss": 0.0529,
+      "step": 19130
+    },
+    {
+      "epoch": 27.5,
+      "learning_rate": 2.571638861629048e-05,
+      "loss": 0.0466,
+      "step": 19140
+    },
+    {
+      "epoch": 27.51,
+      "learning_rate": 2.556918547595682e-05,
+      "loss": 0.0552,
+      "step": 19150
+    },
+    {
+      "epoch": 27.53,
+      "learning_rate": 2.5421982335623157e-05,
+      "loss": 0.0525,
+      "step": 19160
+    },
+    {
+      "epoch": 27.54,
+      "learning_rate": 2.5274779195289497e-05,
+      "loss": 0.0508,
+      "step": 19170
+    },
+    {
+      "epoch": 27.56,
+      "learning_rate": 2.5127576054955837e-05,
+      "loss": 0.047,
+      "step": 19180
+    },
+    {
+      "epoch": 27.57,
+      "learning_rate": 2.4980372914622177e-05,
+      "loss": 0.0459,
+      "step": 19190
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 2.4833169774288518e-05,
+      "loss": 0.0449,
+      "step": 19200
+    },
+    {
+      "epoch": 27.59,
+      "eval_loss": 0.142539843916893,
+      "eval_runtime": 40.2027,
+      "eval_samples_per_second": 11.367,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.1519349495557898,
+      "step": 19200
+    },
+    {
+      "epoch": 27.6,
+      "learning_rate": 2.4685966633954858e-05,
+      "loss": 0.0526,
+      "step": 19210
+    },
+    {
+      "epoch": 27.61,
+      "learning_rate": 2.4538763493621198e-05,
+      "loss": 0.0453,
+      "step": 19220
+    },
+    {
+      "epoch": 27.63,
+      "learning_rate": 2.439156035328753e-05,
+      "loss": 0.0568,
+      "step": 19230
+    },
+    {
+      "epoch": 27.64,
+      "learning_rate": 2.424435721295387e-05,
+      "loss": 0.0441,
+      "step": 19240
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 2.409715407262021e-05,
+      "loss": 0.0516,
+      "step": 19250
+    },
+    {
+      "epoch": 27.67,
+      "learning_rate": 2.3949950932286552e-05,
+      "loss": 0.0448,
+      "step": 19260
+    },
+    {
+      "epoch": 27.69,
+      "learning_rate": 2.3802747791952892e-05,
+      "loss": 0.0411,
+      "step": 19270
+    },
+    {
+      "epoch": 27.7,
+      "learning_rate": 2.3655544651619232e-05,
+      "loss": 0.0495,
+      "step": 19280
+    },
+    {
+      "epoch": 27.72,
+      "learning_rate": 2.3508341511285572e-05,
+      "loss": 0.0483,
+      "step": 19290
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 2.336113837095191e-05,
+      "loss": 0.0615,
+      "step": 19300
+    },
+    {
+      "epoch": 27.73,
+      "eval_loss": 0.14353878796100616,
+      "eval_runtime": 40.2924,
+      "eval_samples_per_second": 11.342,
+      "eval_steps_per_second": 1.439,
+      "eval_wer": 0.15148321035988557,
+      "step": 19300
+    },
+    {
+      "epoch": 27.74,
+      "learning_rate": 2.321393523061825e-05,
+      "loss": 0.0478,
+      "step": 19310
+    },
+    {
+      "epoch": 27.76,
+      "learning_rate": 2.306673209028459e-05,
+      "loss": 0.0515,
+      "step": 19320
+    },
+    {
+      "epoch": 27.77,
+      "learning_rate": 2.291952894995093e-05,
+      "loss": 0.0488,
+      "step": 19330
+    },
+    {
+      "epoch": 27.79,
+      "learning_rate": 2.277232580961727e-05,
+      "loss": 0.0482,
+      "step": 19340
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 2.262512266928361e-05,
+      "loss": 0.0503,
+      "step": 19350
+    },
+    {
+      "epoch": 27.82,
+      "learning_rate": 2.247791952894995e-05,
+      "loss": 0.0388,
+      "step": 19360
+    },
+    {
+      "epoch": 27.83,
+      "learning_rate": 2.233071638861629e-05,
+      "loss": 0.0654,
+      "step": 19370
+    },
+    {
+      "epoch": 27.84,
+      "learning_rate": 2.2183513248282627e-05,
+      "loss": 0.0439,
+      "step": 19380
+    },
+    {
+      "epoch": 27.86,
+      "learning_rate": 2.2036310107948967e-05,
+      "loss": 0.0409,
+      "step": 19390
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 2.1889106967615307e-05,
+      "loss": 0.063,
+      "step": 19400
+    },
+    {
+      "epoch": 27.87,
+      "eval_loss": 0.14183540642261505,
+      "eval_runtime": 40.2851,
+      "eval_samples_per_second": 11.344,
+      "eval_steps_per_second": 1.44,
+      "eval_wer": 0.1511820508959494,
+      "step": 19400
+    },
+    {
+      "epoch": 27.89,
+      "learning_rate": 2.1741903827281648e-05,
+      "loss": 0.0444,
+      "step": 19410
+    },
+    {
+      "epoch": 27.9,
+      "learning_rate": 2.1594700686947988e-05,
+      "loss": 0.0494,
+      "step": 19420
+    },
+    {
+      "epoch": 27.92,
+      "learning_rate": 2.1447497546614328e-05,
+      "loss": 0.0464,
+      "step": 19430
+    },
+    {
+      "epoch": 27.93,
+      "learning_rate": 2.1300294406280668e-05,
+      "loss": 0.0425,
+      "step": 19440
+    },
+    {
+      "epoch": 27.95,
+      "learning_rate": 2.1153091265947e-05,
+      "loss": 0.0519,
+      "step": 19450
+    },
+    {
+      "epoch": 27.96,
+      "learning_rate": 2.100588812561334e-05,
+      "loss": 0.0466,
+      "step": 19460
+    },
+    {
+      "epoch": 27.97,
+      "learning_rate": 2.0858684985279682e-05,
+      "loss": 0.0479,
+      "step": 19470
+    },
+    {
+      "epoch": 27.99,
+      "learning_rate": 2.0711481844946022e-05,
+      "loss": 0.0433,
+      "step": 19480
+    },
+    {
+      "epoch": 28.0,
+      "learning_rate": 2.0564278704612362e-05,
+      "loss": 0.0472,
+      "step": 19490
+    },
+    {
+      "epoch": 28.02,
+      "learning_rate": 2.0417075564278702e-05,
+      "loss": 0.0422,
+      "step": 19500
+    },
+    {
+      "epoch": 28.02,
+      "eval_loss": 0.14156275987625122,
+      "eval_runtime": 40.1542,
+      "eval_samples_per_second": 11.381,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.14817045625658787,
+      "step": 19500
+    },
+    {
+      "epoch": 28.03,
+      "learning_rate": 2.0269872423945042e-05,
+      "loss": 0.0457,
+      "step": 19510
+    },
+    {
+      "epoch": 28.05,
+      "learning_rate": 2.0122669283611383e-05,
+      "loss": 0.051,
+      "step": 19520
+    },
+    {
+      "epoch": 28.06,
+      "learning_rate": 1.997546614327772e-05,
+      "loss": 0.0448,
+      "step": 19530
+    },
+    {
+      "epoch": 28.07,
+      "learning_rate": 1.982826300294406e-05,
+      "loss": 0.0419,
+      "step": 19540
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 1.96810598626104e-05,
+      "loss": 0.0488,
+      "step": 19550
+    },
+    {
+      "epoch": 28.1,
+      "learning_rate": 1.953385672227674e-05,
+      "loss": 0.0524,
+      "step": 19560
+    },
+    {
+      "epoch": 28.12,
+      "learning_rate": 1.938665358194308e-05,
+      "loss": 0.0404,
+      "step": 19570
+    },
+    {
+      "epoch": 28.13,
+      "learning_rate": 1.923945044160942e-05,
+      "loss": 0.0431,
+      "step": 19580
+    },
+    {
+      "epoch": 28.15,
+      "learning_rate": 1.909224730127576e-05,
+      "loss": 0.0428,
+      "step": 19590
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 1.8945044160942097e-05,
+      "loss": 0.0468,
+      "step": 19600
+    },
+    {
+      "epoch": 28.16,
+      "eval_loss": 0.14033865928649902,
+      "eval_runtime": 40.1669,
+      "eval_samples_per_second": 11.378,
+      "eval_steps_per_second": 1.444,
+      "eval_wer": 0.14952567384430054,
+      "step": 19600
+    },
+    {
+      "epoch": 28.18,
+      "learning_rate": 1.8797841020608437e-05,
+      "loss": 0.0554,
+      "step": 19610
+    },
+    {
+      "epoch": 28.19,
+      "learning_rate": 1.8650637880274777e-05,
+      "loss": 0.057,
+      "step": 19620
+    },
+    {
+      "epoch": 28.2,
+      "learning_rate": 1.8503434739941118e-05,
+      "loss": 0.0387,
+      "step": 19630
+    },
+    {
+      "epoch": 28.22,
+      "learning_rate": 1.8356231599607458e-05,
+      "loss": 0.0496,
+      "step": 19640
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 1.8209028459273798e-05,
+      "loss": 0.0547,
+      "step": 19650
+    },
+    {
+      "epoch": 28.25,
+      "learning_rate": 1.8061825318940135e-05,
+      "loss": 0.0383,
+      "step": 19660
+    },
+    {
+      "epoch": 28.26,
+      "learning_rate": 1.7914622178606475e-05,
+      "loss": 0.0445,
+      "step": 19670
+    },
+    {
+      "epoch": 28.28,
+      "learning_rate": 1.7767419038272815e-05,
+      "loss": 0.0391,
+      "step": 19680
+    },
+    {
+      "epoch": 28.29,
+      "learning_rate": 1.7620215897939155e-05,
+      "loss": 0.052,
+      "step": 19690
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 1.7473012757605495e-05,
+      "loss": 0.0421,
+      "step": 19700
+    },
+    {
+      "epoch": 28.3,
+      "eval_loss": 0.14019441604614258,
+      "eval_runtime": 40.1804,
+      "eval_samples_per_second": 11.374,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.1511820508959494,
+      "step": 19700
+    },
+    {
+      "epoch": 28.32,
+      "learning_rate": 1.7325809617271832e-05,
+      "loss": 0.0461,
+      "step": 19710
+    },
+    {
+      "epoch": 28.33,
+      "learning_rate": 1.7178606476938172e-05,
+      "loss": 0.0495,
+      "step": 19720
+    },
+    {
+      "epoch": 28.35,
+      "learning_rate": 1.7031403336604513e-05,
+      "loss": 0.0406,
+      "step": 19730
+    },
+    {
+      "epoch": 28.36,
+      "learning_rate": 1.6884200196270853e-05,
+      "loss": 0.0612,
+      "step": 19740
+    },
+    {
+      "epoch": 28.38,
+      "learning_rate": 1.6736997055937193e-05,
+      "loss": 0.044,
+      "step": 19750
+    },
+    {
+      "epoch": 28.39,
+      "learning_rate": 1.6589793915603533e-05,
+      "loss": 0.0439,
+      "step": 19760
+    },
+    {
+      "epoch": 28.41,
+      "learning_rate": 1.644259077526987e-05,
+      "loss": 0.0503,
+      "step": 19770
+    },
+    {
+      "epoch": 28.42,
+      "learning_rate": 1.629538763493621e-05,
+      "loss": 0.04,
+      "step": 19780
+    },
+    {
+      "epoch": 28.43,
+      "learning_rate": 1.614818449460255e-05,
+      "loss": 0.035,
+      "step": 19790
+    },
+    {
+      "epoch": 28.45,
+      "learning_rate": 1.600098135426889e-05,
+      "loss": 0.0446,
+      "step": 19800
+    },
+    {
+      "epoch": 28.45,
+      "eval_loss": 0.1425383985042572,
+      "eval_runtime": 40.1802,
+      "eval_samples_per_second": 11.374,
+      "eval_steps_per_second": 1.443,
+      "eval_wer": 0.1493750941123325,
+      "step": 19800
+    },
+    {
+      "epoch": 28.46,
+      "learning_rate": 1.585377821393523e-05,
+      "loss": 0.0487,
+      "step": 19810
+    },
+    {
+      "epoch": 28.48,
+      "learning_rate": 1.570657507360157e-05,
+      "loss": 0.066,
+      "step": 19820
+    },
+    {
+      "epoch": 28.49,
+      "learning_rate": 1.555937193326791e-05,
+      "loss": 0.0541,
+      "step": 19830
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 1.5412168792934248e-05,
+      "loss": 0.047,
+      "step": 19840
+    },
+    {
+      "epoch": 28.52,
+      "learning_rate": 1.5264965652600588e-05,
+      "loss": 0.0415,
+      "step": 19850
+    },
+    {
+      "epoch": 28.53,
+      "learning_rate": 1.5117762512266928e-05,
+      "loss": 0.0401,
+      "step": 19860
+    },
+    {
+      "epoch": 28.55,
+      "learning_rate": 1.4970559371933268e-05,
+      "loss": 0.0457,
+      "step": 19870
+    },
+    {
+      "epoch": 28.56,
+      "learning_rate": 1.4823356231599605e-05,
+      "loss": 0.0445,
+      "step": 19880
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 1.4676153091265945e-05,
+      "loss": 0.0419,
+      "step": 19890
+    },
+    {
+      "epoch": 28.59,
+      "learning_rate": 1.4528949950932285e-05,
+      "loss": 0.048,
+      "step": 19900
+    },
+    {
+      "epoch": 28.59,
+      "eval_loss": 0.14261671900749207,
+      "eval_runtime": 39.9856,
+      "eval_samples_per_second": 11.429,
+      "eval_steps_per_second": 1.451,
+      "eval_wer": 0.1474175575967475,
+      "step": 19900
+    },
+    {
+      "epoch": 28.61,
+      "learning_rate": 1.4381746810598624e-05,
+      "loss": 0.043,
+      "step": 19910
+    },
+    {
+      "epoch": 28.62,
+      "learning_rate": 1.4234543670264964e-05,
+      "loss": 0.0526,
+      "step": 19920
+    },
+    {
+      "epoch": 28.64,
+      "learning_rate": 1.4087340529931304e-05,
+      "loss": 0.0433,
+      "step": 19930
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 1.3940137389597644e-05,
+      "loss": 0.0523,
+      "step": 19940
+    },
+    {
+      "epoch": 28.66,
+      "learning_rate": 1.3792934249263983e-05,
+      "loss": 0.0477,
+      "step": 19950
+    },
+    {
+      "epoch": 28.68,
+      "learning_rate": 1.3645731108930323e-05,
+      "loss": 0.0438,
+      "step": 19960
+    },
+    {
+      "epoch": 28.69,
+      "learning_rate": 1.3498527968596663e-05,
+      "loss": 0.0634,
+      "step": 19970
+    },
+    {
+      "epoch": 28.71,
+      "learning_rate": 1.3351324828263003e-05,
+      "loss": 0.0488,
+      "step": 19980
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 1.320412168792934e-05,
+      "loss": 0.0561,
+      "step": 19990
+    },
+    {
+      "epoch": 28.74,
+      "learning_rate": 1.305691854759568e-05,
+      "loss": 0.0429,
+      "step": 20000
+    },
+    {
+      "epoch": 28.74,
+      "eval_loss": 0.14055244624614716,
+      "eval_runtime": 40.1463,
+      "eval_samples_per_second": 11.383,
+      "eval_steps_per_second": 1.445,
+      "eval_wer": 0.146514079204939,
+      "step": 20000
+    },
+    {
+      "epoch": 28.75,
+      "learning_rate": 1.290971540726202e-05,
+      "loss": 0.0455,
+      "step": 20010
+    },
+    {
+      "epoch": 28.76,
+      "learning_rate": 1.2762512266928359e-05,
+      "loss": 0.0559,
+      "step": 20020
+    },
+    {
+      "epoch": 28.78,
+      "learning_rate": 1.2615309126594699e-05,
+      "loss": 0.0384,
+      "step": 20030
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 1.2468105986261039e-05,
+      "loss": 0.0537,
+      "step": 20040
+    },
+    {
+      "epoch": 28.81,
+      "learning_rate": 1.232090284592738e-05,
+      "loss": 0.0477,
+      "step": 20050
+    },
+    {
+      "epoch": 28.82,
+      "learning_rate": 1.2173699705593718e-05,
+      "loss": 0.0438,
+      "step": 20060
+    },
+    {
+      "epoch": 28.84,
+      "learning_rate": 1.2026496565260058e-05,
+      "loss": 0.0424,
+      "step": 20070
+    },
+    {
+      "epoch": 28.85,
+      "learning_rate": 1.1879293424926398e-05,
+      "loss": 0.0658,
+      "step": 20080
+    },
+    {
+      "epoch": 28.86,
+      "learning_rate": 1.1732090284592738e-05,
+      "loss": 0.0496,
+      "step": 20090
+    },
+    {
+      "epoch": 28.88,
+      "learning_rate": 1.1584887144259077e-05,
+      "loss": 0.0457,
+      "step": 20100
+    },
+    {
+      "epoch": 28.88,
+      "eval_loss": 0.14235013723373413,
+      "eval_runtime": 40.2402,
+      "eval_samples_per_second": 11.357,
+      "eval_steps_per_second": 1.441,
+      "eval_wer": 0.14606234000903479,
+      "step": 20100
+    },
+    {
+      "epoch": 28.89,
+      "learning_rate": 1.1437684003925415e-05,
+      "loss": 0.0438,
+      "step": 20110
+    },
+    {
+      "epoch": 28.91,
+      "learning_rate": 1.1290480863591755e-05,
+      "loss": 0.0575,
+      "step": 20120
+    },
+    {
+      "epoch": 28.92,
+      "learning_rate": 1.1143277723258095e-05,
+      "loss": 0.048,
+      "step": 20130
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 1.0996074582924434e-05,
+      "loss": 0.0448,
+      "step": 20140
+    },
+    {
+      "epoch": 28.95,
+      "learning_rate": 1.0848871442590774e-05,
+      "loss": 0.0463,
+      "step": 20150
+    },
+    {
+      "epoch": 28.97,
+      "learning_rate": 1.0701668302257114e-05,
+      "loss": 0.0513,
+      "step": 20160
+    },
+    {
+      "epoch": 28.98,
+      "learning_rate": 1.0554465161923453e-05,
+      "loss": 0.0438,
+      "step": 20170
+    },
+    {
+      "epoch": 28.99,
+      "learning_rate": 1.0407262021589793e-05,
+      "loss": 0.0452,
+      "step": 20180
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 1.0260058881256133e-05,
+      "loss": 0.0487,
+      "step": 20190
+    },
+    {
+      "epoch": 29.02,
+      "learning_rate": 1.0112855740922473e-05,
+      "loss": 0.0469,
+      "step": 20200
+    },
+    {
+      "epoch": 29.02,
+      "eval_loss": 0.14129261672496796,
+      "eval_runtime": 40.3493,
+      "eval_samples_per_second": 11.326,
+      "eval_steps_per_second": 1.437,
+      "eval_wer": 0.14561060081313054,
+      "step": 20200
+    },
+    {
+      "epoch": 29.04,
+      "learning_rate": 9.965652600588812e-06,
+      "loss": 0.0488,
+      "step": 20210
+    },
+    {
+      "epoch": 29.05,
+      "learning_rate": 9.818449460255152e-06,
+      "loss": 0.0448,
+      "step": 20220
+    },
+    {
+      "epoch": 29.07,
+      "learning_rate": 9.671246319921492e-06,
+      "loss": 0.0451,
+      "step": 20230
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 9.52404317958783e-06,
+      "loss": 0.0551,
+      "step": 20240
+    },
+    {
+      "epoch": 29.09,
+      "learning_rate": 9.376840039254169e-06,
+      "loss": 0.0475,
+      "step": 20250
+    },
+    {
+      "epoch": 29.11,
+      "learning_rate": 9.22963689892051e-06,
+      "loss": 0.0424,
+      "step": 20260
+    },
+    {
+      "epoch": 29.12,
+      "learning_rate": 9.08243375858685e-06,
+      "loss": 0.0427,
+      "step": 20270
+    },
+    {
+      "epoch": 29.14,
+      "learning_rate": 8.93523061825319e-06,
+      "loss": 0.0487,
+      "step": 20280
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 8.788027477919528e-06,
+      "loss": 0.0553,
+      "step": 20290
+    },
+    {
+      "epoch": 29.17,
+      "learning_rate": 8.640824337585868e-06,
+      "loss": 0.0543,
+      "step": 20300
+    },
+    {
+      "epoch": 29.17,
+      "eval_loss": 0.14143255352973938,
+      "eval_runtime": 40.1054,
+      "eval_samples_per_second": 11.395,
+      "eval_steps_per_second": 1.446,
+      "eval_wer": 0.146514079204939,
+      "step": 20300
+    },
+    {
+      "epoch": 29.18,
+      "learning_rate": 8.493621197252207e-06,
+      "loss": 0.0515,
+      "step": 20310
+    },
+    {
+      "epoch": 29.2,
+      "learning_rate": 8.346418056918547e-06,
+      "loss": 0.0423,
+      "step": 20320
+    },
+    {
+      "epoch": 29.21,
+      "learning_rate": 8.199214916584887e-06,
+      "loss": 0.0382,
+      "step": 20330
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 8.052011776251227e-06,
+      "loss": 0.0436,
+      "step": 20340
+    },
+    {
+      "epoch": 29.24,
+      "learning_rate": 7.904808635917566e-06,
+      "loss": 0.0447,
+      "step": 20350
+    },
+    {
+      "epoch": 29.25,
+      "learning_rate": 7.757605495583906e-06,
+      "loss": 0.0386,
+      "step": 20360
+    },
+    {
+      "epoch": 29.27,
+      "learning_rate": 7.610402355250245e-06,
+      "loss": 0.0537,
+      "step": 20370
+    },
+    {
+      "epoch": 29.28,
+      "learning_rate": 7.4631992149165835e-06,
+      "loss": 0.0408,
+      "step": 20380
+    },
+    {
+      "epoch": 29.3,
+      "learning_rate": 7.315996074582924e-06,
+      "loss": 0.041,
+      "step": 20390
+    },
+    {
+      "epoch": 29.31,
+      "learning_rate": 7.168792934249263e-06,
+      "loss": 0.0416,
+      "step": 20400
+    },
+    {
+      "epoch": 29.31,
+      "eval_loss": 0.14171102643013,
+      "eval_runtime": 40.0654,
+      "eval_samples_per_second": 11.406,
+      "eval_steps_per_second": 1.448,
+      "eval_wer": 0.14681523866887516,
+      "step": 20400
+    },
+    {
+      "epoch": 29.32,
+      "learning_rate": 7.021589793915603e-06,
+      "loss": 0.0445,
+      "step": 20410
+    },
+    {
+      "epoch": 29.34,
+      "learning_rate": 6.8743866535819425e-06,
+      "loss": 0.0501,
+      "step": 20420
+    },
+    {
+      "epoch": 29.35,
+      "learning_rate": 6.741903827281648e-06,
+      "loss": 0.0454,
+      "step": 20430
+    },
+    {
+      "epoch": 29.37,
+      "learning_rate": 6.594700686947987e-06,
+      "loss": 0.0383,
+      "step": 20440
+    },
+    {
+      "epoch": 29.38,
+      "learning_rate": 6.4474975466143275e-06,
+      "loss": 0.0403,
+      "step": 20450
+    },
+    {
+      "epoch": 29.4,
+      "learning_rate": 6.300294406280667e-06,
+      "loss": 0.0387,
+      "step": 20460
+    },
+    {
+      "epoch": 29.41,
+      "learning_rate": 6.153091265947007e-06,
+      "loss": 0.055,
+      "step": 20470
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 6.0058881256133454e-06,
+      "loss": 0.0377,
+      "step": 20480
+    },
+    {
+      "epoch": 29.44,
+      "learning_rate": 5.858684985279686e-06,
+      "loss": 0.0539,
+      "step": 20490
+    },
+    {
+      "epoch": 29.45,
+      "learning_rate": 5.711481844946025e-06,
+      "loss": 0.0491,
+      "step": 20500
+    },
+    {
+      "epoch": 29.45,
+      "eval_loss": 0.1411919742822647,
+      "eval_runtime": 40.3501,
+      "eval_samples_per_second": 11.326,
+      "eval_steps_per_second": 1.437,
+      "eval_wer": 0.14711639813281133,
+      "step": 20500
+    },
+    {
+      "epoch": 29.47,
+      "learning_rate": 5.564278704612364e-06,
+      "loss": 0.0521,
+      "step": 20510
+    },
+    {
+      "epoch": 29.48,
+      "learning_rate": 5.417075564278704e-06,
+      "loss": 0.0384,
+      "step": 20520
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 5.269872423945043e-06,
+      "loss": 0.0496,
+      "step": 20530
+    },
+    {
+      "epoch": 29.51,
+      "learning_rate": 5.122669283611383e-06,
+      "loss": 0.0565,
+      "step": 20540
+    },
+    {
+      "epoch": 29.53,
+      "learning_rate": 4.975466143277722e-06,
+      "loss": 0.0496,
+      "step": 20550
+    },
+    {
+      "epoch": 29.54,
+      "learning_rate": 4.8282630029440625e-06,
+      "loss": 0.0423,
+      "step": 20560
+    },
+    {
+      "epoch": 29.55,
+      "learning_rate": 4.681059862610402e-06,
+      "loss": 0.0462,
+      "step": 20570
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 4.533856722276741e-06,
+      "loss": 0.0471,
+      "step": 20580
+    },
+    {
+      "epoch": 29.58,
+      "learning_rate": 4.386653581943081e-06,
+      "loss": 0.0482,
+      "step": 20590
+    },
+    {
+      "epoch": 29.6,
+      "learning_rate": 4.239450441609421e-06,
+      "loss": 0.0441,
+      "step": 20600
+    },
+    {
+      "epoch": 29.6,
+      "eval_loss": 0.14090383052825928,
+      "eval_runtime": 40.3963,
+      "eval_samples_per_second": 11.313,
+      "eval_steps_per_second": 1.436,
+      "eval_wer": 0.14711639813281133,
+      "step": 20600
+    },
+    {
+      "epoch": 29.61,
+      "learning_rate": 4.09224730127576e-06,
+      "loss": 0.05,
+      "step": 20610
+    },
+    {
+      "epoch": 29.63,
+      "learning_rate": 3.9450441609421e-06,
+      "loss": 0.0457,
+      "step": 20620
+    },
+    {
+      "epoch": 29.64,
+      "learning_rate": 3.7978410206084395e-06,
+      "loss": 0.0535,
+      "step": 20630
+    },
+    {
+      "epoch": 29.66,
+      "learning_rate": 3.6506378802747788e-06,
+      "loss": 0.0479,
+      "step": 20640
+    },
+    {
+      "epoch": 29.67,
+      "learning_rate": 3.5034347399411185e-06,
+      "loss": 0.0386,
+      "step": 20650
+    },
+    {
+      "epoch": 29.68,
+      "learning_rate": 3.3562315996074583e-06,
+      "loss": 0.0382,
+      "step": 20660
+    },
+    {
+      "epoch": 29.7,
+      "learning_rate": 3.2090284592737976e-06,
+      "loss": 0.0424,
+      "step": 20670
+    },
+    {
+      "epoch": 29.71,
+      "learning_rate": 3.0618253189401373e-06,
+      "loss": 0.0628,
+      "step": 20680
+    },
+    {
+      "epoch": 29.73,
+      "learning_rate": 2.9146221786064766e-06,
+      "loss": 0.0585,
+      "step": 20690
+    },
+    {
+      "epoch": 29.74,
+      "learning_rate": 2.767419038272816e-06,
+      "loss": 0.0542,
+      "step": 20700
+    },
+    {
+      "epoch": 29.74,
+      "eval_loss": 0.14069251716136932,
+      "eval_runtime": 40.3961,
+      "eval_samples_per_second": 11.313,
+      "eval_steps_per_second": 1.436,
+      "eval_wer": 0.14696581840084325,
+      "step": 20700
+    },
+    {
+      "epoch": 29.76,
+      "learning_rate": 2.6202158979391557e-06,
+      "loss": 0.0418,
+      "step": 20710
+    },
+    {
+      "epoch": 29.77,
+      "learning_rate": 2.4730127576054954e-06,
+      "loss": 0.0415,
+      "step": 20720
+    },
+    {
+      "epoch": 29.78,
+      "learning_rate": 2.3258096172718348e-06,
+      "loss": 0.0373,
+      "step": 20730
+    },
+    {
+      "epoch": 29.8,
+      "learning_rate": 2.1786064769381745e-06,
+      "loss": 0.0381,
+      "step": 20740
+    },
+    {
+      "epoch": 29.81,
+      "learning_rate": 2.0314033366045143e-06,
+      "loss": 0.0445,
+      "step": 20750
+    },
+    {
+      "epoch": 29.83,
+      "learning_rate": 1.8842001962708538e-06,
+      "loss": 0.0482,
+      "step": 20760
+    },
+    {
+      "epoch": 29.84,
+      "learning_rate": 1.7369970559371933e-06,
+      "loss": 0.0463,
+      "step": 20770
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 1.5897939156035326e-06,
+      "loss": 0.0402,
+      "step": 20780
+    },
+    {
+      "epoch": 29.87,
+      "learning_rate": 1.4425907752698722e-06,
+      "loss": 0.0503,
+      "step": 20790
+    },
+    {
+      "epoch": 29.89,
+      "learning_rate": 1.295387634936212e-06,
+      "loss": 0.0471,
+      "step": 20800
+    },
+    {
+      "epoch": 29.89,
+      "eval_loss": 0.1404000222682953,
+      "eval_runtime": 41.0194,
+      "eval_samples_per_second": 11.141,
+      "eval_steps_per_second": 1.414,
+      "eval_wer": 0.14711639813281133,
+      "step": 20800
+    }
+  ],
+  "max_steps": 20880,
+  "num_train_epochs": 30,
+  "total_flos": 1.7201005515836912e+20,
+  "trial_name": null,
+  "trial_params": null
+}