diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,9784 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 19.9992034411343,
+  "global_step": 15680,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7499999999999997e-06,
+      "loss": 10.6608,
+      "step": 10
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1249999999999999e-05,
+      "loss": 8.0703,
+      "step": 20
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.875e-05,
+      "loss": 5.9157,
+      "step": 30
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 2.6249999999999998e-05,
+      "loss": 5.2418,
+      "step": 40
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 3.375e-05,
+      "loss": 4.7367,
+      "step": 50
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.125e-05,
+      "loss": 5.3521,
+      "step": 60
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 4.875e-05,
+      "loss": 4.515,
+      "step": 70
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 5.625e-05,
+      "loss": 4.3019,
+      "step": 80
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 6.374999999999999e-05,
+      "loss": 4.0842,
+      "step": 90
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 7.125e-05,
+      "loss": 3.7084,
+      "step": 100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 7.874999999999999e-05,
+      "loss": 4.4597,
+      "step": 110
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 8.624999999999998e-05,
+      "loss": 3.8854,
+      "step": 120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.374999999999999e-05,
+      "loss": 3.7205,
+      "step": 130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 0.00010125,
+      "loss": 3.6268,
+      "step": 140
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.00010874999999999999,
+      "loss": 3.2092,
+      "step": 150
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 0.00011624999999999999,
+      "loss": 4.0974,
+      "step": 160
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.00012374999999999997,
+      "loss": 3.591,
+      "step": 170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 0.00013125,
+      "loss": 3.5278,
+      "step": 180
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 0.00013874999999999998,
+      "loss": 3.4416,
+      "step": 190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 0.00014624999999999998,
+      "loss": 2.9812,
+      "step": 200
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00015374999999999997,
+      "loss": 3.8608,
+      "step": 210
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 0.00016125,
+      "loss": 3.414,
+      "step": 220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 0.00016874999999999998,
+      "loss": 3.3173,
+      "step": 230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 0.00017625,
+      "loss": 3.2286,
+      "step": 240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.00018375,
+      "loss": 2.6632,
+      "step": 250
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 0.00019124999999999996,
+      "loss": 3.6559,
+      "step": 260
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 0.00019874999999999998,
+      "loss": 3.2761,
+      "step": 270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.00020624999999999997,
+      "loss": 3.1426,
+      "step": 280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.00021375,
+      "loss": 3.1081,
+      "step": 290
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.00022124999999999998,
+      "loss": 2.4845,
+      "step": 300
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00022874999999999998,
+      "loss": 3.6627,
+      "step": 310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.00023624999999999997,
+      "loss": 3.2377,
+      "step": 320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.00024375,
+      "loss": 3.1296,
+      "step": 330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00025125,
+      "loss": 3.0432,
+      "step": 340
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.00025875,
+      "loss": 2.3069,
+      "step": 350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00026624999999999994,
+      "loss": 3.5711,
+      "step": 360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00027374999999999996,
+      "loss": 3.1114,
+      "step": 370
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.00028125,
+      "loss": 3.0522,
+      "step": 380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00028875,
+      "loss": 2.8813,
+      "step": 390
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.00029624999999999996,
+      "loss": 2.2943,
+      "step": 400
+    },
+    {
+      "epoch": 0.51,
+      "eval_loss": 4.436714172363281,
+      "eval_runtime": 912.2312,
+      "eval_samples_per_second": 6.833,
+      "eval_steps_per_second": 0.855,
+      "eval_wer": 0.9745079087749299,
+      "step": 400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000299901832460733,
+      "loss": 3.813,
+      "step": 410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00029970549738219895,
+      "loss": 3.1101,
+      "step": 420
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.0002995091623036649,
+      "loss": 2.9516,
+      "step": 430
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0002993128272251309,
+      "loss": 2.8798,
+      "step": 440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00029911649214659685,
+      "loss": 2.0947,
+      "step": 450
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.0002989201570680628,
+      "loss": 3.7556,
+      "step": 460
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0002987238219895288,
+      "loss": 3.1256,
+      "step": 470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00029852748691099474,
+      "loss": 3.032,
+      "step": 480
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.0002983311518324607,
+      "loss": 2.8627,
+      "step": 490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.0002981348167539267,
+      "loss": 2.346,
+      "step": 500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.00029793848167539264,
+      "loss": 3.678,
+      "step": 510
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.0002977421465968586,
+      "loss": 3.0202,
+      "step": 520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00029754581151832457,
+      "loss": 2.9468,
+      "step": 530
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00029734947643979054,
+      "loss": 2.8079,
+      "step": 540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0002971531413612565,
+      "loss": 2.2192,
+      "step": 550
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002969568062827225,
+      "loss": 3.4516,
+      "step": 560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.0002967604712041885,
+      "loss": 2.9412,
+      "step": 570
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.00029656413612565445,
+      "loss": 2.9339,
+      "step": 580
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002963678010471204,
+      "loss": 2.7984,
+      "step": 590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.00029617146596858633,
+      "loss": 2.1576,
+      "step": 600
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002959751308900523,
+      "loss": 3.4613,
+      "step": 610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.0002957787958115183,
+      "loss": 2.9413,
+      "step": 620
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.0002955824607329843,
+      "loss": 2.871,
+      "step": 630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00029538612565445024,
+      "loss": 2.6864,
+      "step": 640
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002951897905759162,
+      "loss": 2.138,
+      "step": 650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.0002949934554973822,
+      "loss": 3.3475,
+      "step": 660
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.00029479712041884814,
+      "loss": 2.9137,
+      "step": 670
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.0002946007853403141,
+      "loss": 2.7982,
+      "step": 680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.00029440445026178007,
+      "loss": 2.6471,
+      "step": 690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00029420811518324604,
+      "loss": 2.1628,
+      "step": 700
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.000294011780104712,
+      "loss": 3.7613,
+      "step": 710
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.000293815445026178,
+      "loss": 3.2122,
+      "step": 720
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00029361910994764393,
+      "loss": 3.0674,
+      "step": 730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.0002934227748691099,
+      "loss": 2.6641,
+      "step": 740
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00029322643979057587,
+      "loss": 2.2099,
+      "step": 750
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.00029303010471204183,
+      "loss": 3.2629,
+      "step": 760
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.00029283376963350785,
+      "loss": 2.8116,
+      "step": 770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002926374345549738,
+      "loss": 2.587,
+      "step": 780
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002924410994764398,
+      "loss": 2.981,
+      "step": 790
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.00029224476439790575,
+      "loss": 2.8816,
+      "step": 800
+    },
+    {
+      "epoch": 1.02,
+      "eval_loss": 2.810335874557495,
+      "eval_runtime": 1355.6131,
+      "eval_samples_per_second": 4.598,
+      "eval_steps_per_second": 0.575,
+      "eval_wer": 1.2576067359342764,
+      "step": 800
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.0002920484293193717,
+      "loss": 2.7035,
+      "step": 810
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.0002918520942408377,
+      "loss": 2.5633,
+      "step": 820
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029165575916230364,
+      "loss": 2.3199,
+      "step": 830
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0002914594240837696,
+      "loss": 2.8772,
+      "step": 840
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0002912630890052356,
+      "loss": 2.8404,
+      "step": 850
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.00029106675392670154,
+      "loss": 2.7054,
+      "step": 860
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0002908704188481675,
+      "loss": 2.5925,
+      "step": 870
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00029067408376963347,
+      "loss": 2.2584,
+      "step": 880
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00029047774869109944,
+      "loss": 2.8465,
+      "step": 890
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.0002902814136125654,
+      "loss": 2.8842,
+      "step": 900
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00029008507853403137,
+      "loss": 2.685,
+      "step": 910
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00028988874345549733,
+      "loss": 2.5287,
+      "step": 920
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.00028969240837696335,
+      "loss": 2.2153,
+      "step": 930
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0002894960732984293,
+      "loss": 2.6404,
+      "step": 940
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.0002892997382198953,
+      "loss": 2.82,
+      "step": 950
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00028910340314136125,
+      "loss": 2.6917,
+      "step": 960
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.00028890706806282716,
+      "loss": 2.495,
+      "step": 970
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.0002887107329842932,
+      "loss": 2.1993,
+      "step": 980
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.00028851439790575915,
+      "loss": 2.6478,
+      "step": 990
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002883180628272251,
+      "loss": 2.8197,
+      "step": 1000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0002881217277486911,
+      "loss": 2.7265,
+      "step": 1010
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.00028792539267015704,
+      "loss": 2.5507,
+      "step": 1020
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.000287729057591623,
+      "loss": 2.2095,
+      "step": 1030
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.000287532722513089,
+      "loss": 2.8337,
+      "step": 1040
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00028733638743455494,
+      "loss": 2.8985,
+      "step": 1050
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0002871400523560209,
+      "loss": 2.6719,
+      "step": 1060
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00028694371727748687,
+      "loss": 2.657,
+      "step": 1070
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.0002867473821989529,
+      "loss": 2.282,
+      "step": 1080
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00028655104712041886,
+      "loss": 2.6275,
+      "step": 1090
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.00028635471204188477,
+      "loss": 2.7491,
+      "step": 1100
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.00028615837696335073,
+      "loss": 2.6163,
+      "step": 1110
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0002859620418848167,
+      "loss": 2.5673,
+      "step": 1120
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.0002857657068062827,
+      "loss": 2.2049,
+      "step": 1130
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0002855693717277487,
+      "loss": 2.6092,
+      "step": 1140
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.00028537303664921465,
+      "loss": 2.7417,
+      "step": 1150
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.0002851767015706806,
+      "loss": 2.6391,
+      "step": 1160
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0002849803664921466,
+      "loss": 2.5289,
+      "step": 1170
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00028478403141361255,
+      "loss": 2.169,
+      "step": 1180
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0002845876963350785,
+      "loss": 2.6139,
+      "step": 1190
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.0002843913612565445,
+      "loss": 2.7842,
+      "step": 1200
+    },
+    {
+      "epoch": 1.53,
+      "eval_loss": 2.683229684829712,
+      "eval_runtime": 1423.409,
+      "eval_samples_per_second": 4.379,
+      "eval_steps_per_second": 0.548,
+      "eval_wer": 1.1330458541662793,
+      "step": 1200
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.00028419502617801044,
+      "loss": 2.6498,
+      "step": 1210
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0002839986910994764,
+      "loss": 2.4721,
+      "step": 1220
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00028380235602094237,
+      "loss": 2.1417,
+      "step": 1230
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.00028360602094240834,
+      "loss": 2.6908,
+      "step": 1240
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.0002834096858638743,
+      "loss": 2.7411,
+      "step": 1250
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00028321335078534027,
+      "loss": 2.6083,
+      "step": 1260
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.00028301701570680623,
+      "loss": 2.4592,
+      "step": 1270
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0002828206806282722,
+      "loss": 2.141,
+      "step": 1280
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.0002826243455497382,
+      "loss": 2.6857,
+      "step": 1290
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002824280104712042,
+      "loss": 2.7469,
+      "step": 1300
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00028223167539267015,
+      "loss": 2.6124,
+      "step": 1310
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0002820353403141361,
+      "loss": 2.4334,
+      "step": 1320
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0002818390052356021,
+      "loss": 2.1501,
+      "step": 1330
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00028164267015706805,
+      "loss": 2.5496,
+      "step": 1340
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.000281446335078534,
+      "loss": 2.707,
+      "step": 1350
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00028125,
+      "loss": 2.5717,
+      "step": 1360
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00028105366492146594,
+      "loss": 2.4321,
+      "step": 1370
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0002808573298429319,
+      "loss": 2.1794,
+      "step": 1380
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002806609947643979,
+      "loss": 2.5534,
+      "step": 1390
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.00028046465968586384,
+      "loss": 2.6953,
+      "step": 1400
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.0002802683246073298,
+      "loss": 2.5469,
+      "step": 1410
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.00028007198952879577,
+      "loss": 2.4954,
+      "step": 1420
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00027987565445026174,
+      "loss": 2.099,
+      "step": 1430
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.00027967931937172776,
+      "loss": 2.4482,
+      "step": 1440
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0002794829842931937,
+      "loss": 2.6436,
+      "step": 1450
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0002792866492146597,
+      "loss": 2.577,
+      "step": 1460
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002790903141361256,
+      "loss": 2.4556,
+      "step": 1470
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.00027889397905759156,
+      "loss": 2.1424,
+      "step": 1480
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0002786976439790576,
+      "loss": 2.5765,
+      "step": 1490
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.00027850130890052355,
+      "loss": 2.7068,
+      "step": 1500
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0002783049738219895,
+      "loss": 2.5316,
+      "step": 1510
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.0002781086387434555,
+      "loss": 2.4406,
+      "step": 1520
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.00027791230366492145,
+      "loss": 2.1381,
+      "step": 1530
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.0002777159685863874,
+      "loss": 2.5644,
+      "step": 1540
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.0002775196335078534,
+      "loss": 2.6976,
+      "step": 1550
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.00027732329842931934,
+      "loss": 2.4955,
+      "step": 1560
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0002771269633507853,
+      "loss": 2.2316,
+      "step": 1570
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0002769306282722513,
+      "loss": 2.6237,
+      "step": 1580
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00027673429319371724,
+      "loss": 2.4281,
+      "step": 1590
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.00027653795811518326,
+      "loss": 2.313,
+      "step": 1600
+    },
+    {
+      "epoch": 2.04,
+      "eval_loss": 2.667757511138916,
+      "eval_runtime": 1343.7695,
+      "eval_samples_per_second": 4.638,
+      "eval_steps_per_second": 0.58,
+      "eval_wer": 1.134774446571625,
+      "step": 1600
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.00027634162303664917,
+      "loss": 2.0772,
+      "step": 1610
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.00027614528795811514,
+      "loss": 1.9423,
+      "step": 1620
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0002759489528795811,
+      "loss": 2.8466,
+      "step": 1630
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.00027575261780104707,
+      "loss": 2.4786,
+      "step": 1640
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.0002755562827225131,
+      "loss": 2.3059,
+      "step": 1650
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.00027535994764397905,
+      "loss": 2.1039,
+      "step": 1660
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.000275163612565445,
+      "loss": 1.9146,
+      "step": 1670
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.000274967277486911,
+      "loss": 2.7422,
+      "step": 1680
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.00027477094240837695,
+      "loss": 2.4815,
+      "step": 1690
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0002745746073298429,
+      "loss": 2.3244,
+      "step": 1700
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0002743782722513089,
+      "loss": 2.1384,
+      "step": 1710
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.00027418193717277485,
+      "loss": 1.8722,
+      "step": 1720
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.0002739856020942408,
+      "loss": 2.7037,
+      "step": 1730
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0002737892670157068,
+      "loss": 2.4667,
+      "step": 1740
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.00027359293193717274,
+      "loss": 2.3463,
+      "step": 1750
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0002733965968586387,
+      "loss": 2.0723,
+      "step": 1760
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.0002732002617801047,
+      "loss": 1.8559,
+      "step": 1770
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.00027300392670157064,
+      "loss": 2.6737,
+      "step": 1780
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.0002728075916230366,
+      "loss": 2.4785,
+      "step": 1790
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002726112565445026,
+      "loss": 2.3714,
+      "step": 1800
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0002724149214659686,
+      "loss": 2.1554,
+      "step": 1810
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.00027221858638743456,
+      "loss": 1.9632,
+      "step": 1820
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0002720222513089005,
+      "loss": 2.6765,
+      "step": 1830
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00027182591623036643,
+      "loss": 2.4835,
+      "step": 1840
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0002716295811518324,
+      "loss": 2.3542,
+      "step": 1850
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.0002714332460732984,
+      "loss": 2.1265,
+      "step": 1860
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0002712369109947644,
+      "loss": 1.9591,
+      "step": 1870
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.00027104057591623035,
+      "loss": 2.7187,
+      "step": 1880
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0002708442408376963,
+      "loss": 2.4609,
+      "step": 1890
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0002706479057591623,
+      "loss": 2.3543,
+      "step": 1900
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.00027045157068062824,
+      "loss": 2.0617,
+      "step": 1910
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0002702552356020942,
+      "loss": 1.9237,
+      "step": 1920
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.0002700589005235602,
+      "loss": 2.6788,
+      "step": 1930
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.00026986256544502614,
+      "loss": 2.4417,
+      "step": 1940
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0002696662303664921,
+      "loss": 2.2983,
+      "step": 1950
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0002694698952879581,
+      "loss": 2.085,
+      "step": 1960
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.0002692735602094241,
+      "loss": 1.845,
+      "step": 1970
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00026907722513089,
+      "loss": 2.6836,
+      "step": 1980
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.00026888089005235597,
+      "loss": 2.4245,
+      "step": 1990
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.00026868455497382193,
+      "loss": 2.2897,
+      "step": 2000
+    },
+    {
+      "epoch": 2.55,
+      "eval_loss": 2.574383020401001,
+      "eval_runtime": 1426.3895,
+      "eval_samples_per_second": 4.37,
+      "eval_steps_per_second": 0.547,
+      "eval_wer": 1.2101726733703835,
+      "step": 2000
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.00026848821989528795,
+      "loss": 2.0934,
+      "step": 2010
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0002682918848167539,
+      "loss": 1.8096,
+      "step": 2020
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0002680955497382199,
+      "loss": 2.6541,
+      "step": 2030
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00026789921465968585,
+      "loss": 2.415,
+      "step": 2040
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0002677028795811518,
+      "loss": 2.2685,
+      "step": 2050
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0002675065445026178,
+      "loss": 2.0645,
+      "step": 2060
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00026731020942408375,
+      "loss": 1.8799,
+      "step": 2070
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0002671138743455497,
+      "loss": 2.6958,
+      "step": 2080
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.0002669175392670157,
+      "loss": 2.4073,
+      "step": 2090
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.00026672120418848164,
+      "loss": 2.2642,
+      "step": 2100
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.0002665248691099476,
+      "loss": 2.0429,
+      "step": 2110
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0002663285340314136,
+      "loss": 1.9115,
+      "step": 2120
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.00026613219895287954,
+      "loss": 2.6661,
+      "step": 2130
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002659358638743455,
+      "loss": 2.4516,
+      "step": 2140
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.00026573952879581147,
+      "loss": 2.2485,
+      "step": 2150
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.00026554319371727744,
+      "loss": 2.0369,
+      "step": 2160
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.00026534685863874346,
+      "loss": 1.7468,
+      "step": 2170
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0002651505235602094,
+      "loss": 2.5992,
+      "step": 2180
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0002649541884816754,
+      "loss": 2.365,
+      "step": 2190
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.00026475785340314135,
+      "loss": 2.2485,
+      "step": 2200
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.00026456151832460726,
+      "loss": 1.9479,
+      "step": 2210
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0002643651832460733,
+      "loss": 1.8616,
+      "step": 2220
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.00026416884816753925,
+      "loss": 2.6966,
+      "step": 2230
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0002639725130890052,
+      "loss": 2.3569,
+      "step": 2240
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.0002637761780104712,
+      "loss": 2.1948,
+      "step": 2250
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.00026357984293193715,
+      "loss": 2.0404,
+      "step": 2260
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0002633835078534031,
+      "loss": 1.6561,
+      "step": 2270
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0002631871727748691,
+      "loss": 2.6196,
+      "step": 2280
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00026299083769633504,
+      "loss": 2.373,
+      "step": 2290
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.000262794502617801,
+      "loss": 2.2689,
+      "step": 2300
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.000262598167539267,
+      "loss": 2.0032,
+      "step": 2310
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.000262401832460733,
+      "loss": 1.8228,
+      "step": 2320
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.00026220549738219896,
+      "loss": 2.5379,
+      "step": 2330
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.0002620091623036649,
+      "loss": 2.2594,
+      "step": 2340
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00026181282722513084,
+      "loss": 1.8865,
+      "step": 2350
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0002616164921465968,
+      "loss": 2.5077,
+      "step": 2360
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0002614201570680628,
+      "loss": 2.3351,
+      "step": 2370
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.0002612238219895288,
+      "loss": 2.1007,
+      "step": 2380
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00026102748691099475,
+      "loss": 1.9247,
+      "step": 2390
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0002608311518324607,
+      "loss": 1.3821,
+      "step": 2400
+    },
+    {
+      "epoch": 3.06,
+      "eval_loss": 2.9908056259155273,
+      "eval_runtime": 1158.3432,
+      "eval_samples_per_second": 5.381,
+      "eval_steps_per_second": 0.673,
+      "eval_wer": 0.9377149123622237,
+      "step": 2400
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0002606348167539267,
+      "loss": 2.4299,
+      "step": 2410
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.00026043848167539265,
+      "loss": 2.3084,
+      "step": 2420
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.0002602421465968586,
+      "loss": 2.1145,
+      "step": 2430
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0002600458115183246,
+      "loss": 1.8714,
+      "step": 2440
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.00025984947643979055,
+      "loss": 1.4162,
+      "step": 2450
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.0002596531413612565,
+      "loss": 2.4488,
+      "step": 2460
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.00025945680628272253,
+      "loss": 2.3654,
+      "step": 2470
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002592604712041885,
+      "loss": 2.0721,
+      "step": 2480
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0002590641361256544,
+      "loss": 1.8091,
+      "step": 2490
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0002588678010471204,
+      "loss": 1.3244,
+      "step": 2500
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.00025867146596858634,
+      "loss": 2.4118,
+      "step": 2510
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0002584751308900523,
+      "loss": 2.2881,
+      "step": 2520
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0002582787958115183,
+      "loss": 2.0995,
+      "step": 2530
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0002580824607329843,
+      "loss": 1.8236,
+      "step": 2540
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.00025788612565445025,
+      "loss": 1.3496,
+      "step": 2550
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0002576897905759162,
+      "loss": 2.4261,
+      "step": 2560
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.0002574934554973822,
+      "loss": 2.2874,
+      "step": 2570
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.00025729712041884815,
+      "loss": 2.0784,
+      "step": 2580
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.0002571007853403141,
+      "loss": 1.8425,
+      "step": 2590
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0002569044502617801,
+      "loss": 1.3747,
+      "step": 2600
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.00025670811518324605,
+      "loss": 2.465,
+      "step": 2610
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.000256511780104712,
+      "loss": 2.2774,
+      "step": 2620
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.000256315445026178,
+      "loss": 2.051,
+      "step": 2630
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.00025611910994764394,
+      "loss": 1.8113,
+      "step": 2640
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0002559227748691099,
+      "loss": 1.2219,
+      "step": 2650
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.0002557264397905759,
+      "loss": 2.4022,
+      "step": 2660
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.00025553010471204184,
+      "loss": 2.2788,
+      "step": 2670
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00025533376963350786,
+      "loss": 2.0516,
+      "step": 2680
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0002551374345549738,
+      "loss": 1.7858,
+      "step": 2690
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.0002549410994764398,
+      "loss": 1.2485,
+      "step": 2700
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.00025474476439790576,
+      "loss": 2.4865,
+      "step": 2710
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.00025454842931937167,
+      "loss": 2.2058,
+      "step": 2720
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002543520942408377,
+      "loss": 1.9552,
+      "step": 2730
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00025415575916230365,
+      "loss": 1.688,
+      "step": 2740
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.0002539594240837696,
+      "loss": 1.2323,
+      "step": 2750
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0002537630890052356,
+      "loss": 2.3544,
+      "step": 2760
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00025356675392670155,
+      "loss": 2.1961,
+      "step": 2770
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.0002533704188481675,
+      "loss": 1.9179,
+      "step": 2780
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0002531740837696335,
+      "loss": 1.5599,
+      "step": 2790
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.00025297774869109945,
+      "loss": 1.074,
+      "step": 2800
+    },
+    {
+      "epoch": 3.57,
+      "eval_loss": 2.6648833751678467,
+      "eval_runtime": 1202.5402,
+      "eval_samples_per_second": 5.183,
+      "eval_steps_per_second": 0.649,
+      "eval_wer": 0.8966469024739318,
+      "step": 2800
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.0002527814136125654,
+      "loss": 2.2922,
+      "step": 2810
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.0002525850785340314,
+      "loss": 2.0673,
+      "step": 2820
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.00025238874345549734,
+      "loss": 1.7602,
+      "step": 2830
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.00025219240837696336,
+      "loss": 1.4873,
+      "step": 2840
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.00025199607329842933,
+      "loss": 1.0685,
+      "step": 2850
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.00025179973821989524,
+      "loss": 2.3202,
+      "step": 2860
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.0002516034031413612,
+      "loss": 1.9736,
+      "step": 2870
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.00025140706806282717,
+      "loss": 1.6341,
+      "step": 2880
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.0002512107329842932,
+      "loss": 1.3519,
+      "step": 2890
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.00025101439790575916,
+      "loss": 0.9903,
+      "step": 2900
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.0002508180628272251,
+      "loss": 2.4131,
+      "step": 2910
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.0002506217277486911,
+      "loss": 1.9066,
+      "step": 2920
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.00025042539267015705,
+      "loss": 1.5151,
+      "step": 2930
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.000250229057591623,
+      "loss": 1.1935,
+      "step": 2940
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.000250032722513089,
+      "loss": 0.996,
+      "step": 2950
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.00024983638743455495,
+      "loss": 2.1993,
+      "step": 2960
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.0002496400523560209,
+      "loss": 1.7478,
+      "step": 2970
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.0002494437172774869,
+      "loss": 1.3926,
+      "step": 2980
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.00024924738219895285,
+      "loss": 1.1185,
+      "step": 2990
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.0002490510471204188,
+      "loss": 0.9758,
+      "step": 3000
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.0002488547120418848,
+      "loss": 2.0239,
+      "step": 3010
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00024865837696335074,
+      "loss": 1.5411,
+      "step": 3020
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.0002484620418848167,
+      "loss": 1.2329,
+      "step": 3030
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.00024826570680628273,
+      "loss": 1.0219,
+      "step": 3040
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.0002480693717277487,
+      "loss": 0.8315,
+      "step": 3050
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.00024787303664921466,
+      "loss": 2.0886,
+      "step": 3060
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.0002476767015706806,
+      "loss": 1.4358,
+      "step": 3070
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.0002474803664921466,
+      "loss": 1.1161,
+      "step": 3080
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.00024728403141361256,
+      "loss": 0.9732,
+      "step": 3090
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.0002470876963350785,
+      "loss": 0.8213,
+      "step": 3100
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.0002468913612565445,
+      "loss": 1.9164,
+      "step": 3110
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.00024669502617801045,
+      "loss": 1.332,
+      "step": 3120
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.0002464986910994764,
+      "loss": 0.9794,
+      "step": 3130
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.0002463023560209424,
+      "loss": 1.4493,
+      "step": 3140
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.00024610602094240835,
+      "loss": 1.3535,
+      "step": 3150
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.0002459096858638743,
+      "loss": 0.9229,
+      "step": 3160
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.0002457133507853403,
+      "loss": 0.7988,
+      "step": 3170
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.00024551701570680624,
+      "loss": 0.6798,
+      "step": 3180
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.0002453206806282722,
+      "loss": 1.301,
+      "step": 3190
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.00024512434554973823,
+      "loss": 1.3643,
+      "step": 3200
+    },
+    {
+      "epoch": 4.08,
+      "eval_loss": 1.0063679218292236,
+      "eval_runtime": 1365.76,
+      "eval_samples_per_second": 4.564,
+      "eval_steps_per_second": 0.571,
+      "eval_wer": 0.7236296723109236,
+      "step": 3200
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.0002449280104712042,
+      "loss": 0.8936,
+      "step": 3210
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.00024473167539267016,
+      "loss": 0.7426,
+      "step": 3220
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.00024453534031413607,
+      "loss": 0.6584,
+      "step": 3230
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.00024433900523560204,
+      "loss": 1.3437,
+      "step": 3240
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.00024414267015706806,
+      "loss": 1.3082,
+      "step": 3250
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.00024394633507853402,
+      "loss": 0.8755,
+      "step": 3260
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.00024375,
+      "loss": 0.735,
+      "step": 3270
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00024355366492146595,
+      "loss": 0.6546,
+      "step": 3280
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.0002433573298429319,
+      "loss": 1.2138,
+      "step": 3290
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.00024316099476439789,
+      "loss": 1.1497,
+      "step": 3300
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.00024296465968586385,
+      "loss": 0.8393,
+      "step": 3310
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.00024276832460732982,
+      "loss": 0.6995,
+      "step": 3320
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.00024257198952879578,
+      "loss": 0.6347,
+      "step": 3330
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.00024237565445026175,
+      "loss": 1.2088,
+      "step": 3340
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.00024217931937172774,
+      "loss": 1.0903,
+      "step": 3350
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.0002419829842931937,
+      "loss": 0.7817,
+      "step": 3360
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.00024178664921465967,
+      "loss": 0.7012,
+      "step": 3370
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.00024159031413612564,
+      "loss": 0.6103,
+      "step": 3380
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.0002413939790575916,
+      "loss": 1.1574,
+      "step": 3390
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.0002411976439790576,
+      "loss": 1.0844,
+      "step": 3400
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.00024100130890052356,
+      "loss": 0.7571,
+      "step": 3410
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.0002408049738219895,
+      "loss": 0.6706,
+      "step": 3420
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.00024060863874345546,
+      "loss": 0.6273,
+      "step": 3430
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00024041230366492143,
+      "loss": 1.0405,
+      "step": 3440
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.0002402159685863874,
+      "loss": 0.9718,
+      "step": 3450
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.0002400196335078534,
+      "loss": 0.7145,
+      "step": 3460
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.00023982329842931935,
+      "loss": 0.6564,
+      "step": 3470
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00023962696335078532,
+      "loss": 0.5634,
+      "step": 3480
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.00023943062827225128,
+      "loss": 1.0246,
+      "step": 3490
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.00023923429319371725,
+      "loss": 0.9484,
+      "step": 3500
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.00023903795811518324,
+      "loss": 0.6733,
+      "step": 3510
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.0002388416230366492,
+      "loss": 0.6366,
+      "step": 3520
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.00023864528795811517,
+      "loss": 0.5755,
+      "step": 3530
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.0002384489528795811,
+      "loss": 1.0351,
+      "step": 3540
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.00023825261780104708,
+      "loss": 0.8986,
+      "step": 3550
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.00023805628272251307,
+      "loss": 0.6612,
+      "step": 3560
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00023785994764397904,
+      "loss": 0.5871,
+      "step": 3570
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.000237663612565445,
+      "loss": 0.5601,
+      "step": 3580
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.00023746727748691097,
+      "loss": 1.0413,
+      "step": 3590
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.00023727094240837693,
+      "loss": 0.8286,
+      "step": 3600
+    },
+    {
+      "epoch": 4.59,
+      "eval_loss": 0.6339302062988281,
+      "eval_runtime": 1363.8954,
+      "eval_samples_per_second": 4.57,
+      "eval_steps_per_second": 0.572,
+      "eval_wer": 0.5453523168714336,
+      "step": 3600
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.00023707460732984292,
+      "loss": 0.6376,
+      "step": 3610
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.0002368782722513089,
+      "loss": 0.6132,
+      "step": 3620
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.00023668193717277486,
+      "loss": 0.5198,
+      "step": 3630
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.00023648560209424082,
+      "loss": 0.9745,
+      "step": 3640
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.0002362892670157068,
+      "loss": 0.8313,
+      "step": 3650
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.00023609293193717278,
+      "loss": 0.6251,
+      "step": 3660
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.00023589659685863872,
+      "loss": 0.5662,
+      "step": 3670
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.00023570026178010468,
+      "loss": 0.5428,
+      "step": 3680
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.00023550392670157065,
+      "loss": 0.9495,
+      "step": 3690
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.00023530759162303661,
+      "loss": 0.7888,
+      "step": 3700
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.0002351112565445026,
+      "loss": 0.6159,
+      "step": 3710
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.00023491492146596857,
+      "loss": 0.5395,
+      "step": 3720
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00023471858638743454,
+      "loss": 0.5349,
+      "step": 3730
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.0002345222513089005,
+      "loss": 0.9063,
+      "step": 3740
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.00023432591623036647,
+      "loss": 0.7716,
+      "step": 3750
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.00023412958115183243,
+      "loss": 0.5782,
+      "step": 3760
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.00023393324607329843,
+      "loss": 0.5188,
+      "step": 3770
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.0002337369109947644,
+      "loss": 0.559,
+      "step": 3780
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.00023354057591623033,
+      "loss": 0.8986,
+      "step": 3790
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.0002333442408376963,
+      "loss": 0.7888,
+      "step": 3800
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.00023314790575916226,
+      "loss": 0.5603,
+      "step": 3810
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00023295157068062825,
+      "loss": 0.5168,
+      "step": 3820
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.00023275523560209422,
+      "loss": 0.5285,
+      "step": 3830
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.00023255890052356019,
+      "loss": 0.8083,
+      "step": 3840
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.00023236256544502615,
+      "loss": 0.7496,
+      "step": 3850
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.00023216623036649212,
+      "loss": 0.5935,
+      "step": 3860
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.0002319698952879581,
+      "loss": 0.5104,
+      "step": 3870
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.00023177356020942408,
+      "loss": 0.4716,
+      "step": 3880
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.00023157722513089004,
+      "loss": 0.7965,
+      "step": 3890
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.000231380890052356,
+      "loss": 0.6845,
+      "step": 3900
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.00023118455497382194,
+      "loss": 0.5182,
+      "step": 3910
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.00023098821989528796,
+      "loss": 0.4957,
+      "step": 3920
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.0002307918848167539,
+      "loss": 0.7714,
+      "step": 3930
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00023059554973821987,
+      "loss": 0.4629,
+      "step": 3940
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.00023039921465968583,
+      "loss": 0.4137,
+      "step": 3950
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.0002302028795811518,
+      "loss": 0.3444,
+      "step": 3960
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.0002300065445026178,
+      "loss": 0.4056,
+      "step": 3970
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.00022981020942408376,
+      "loss": 0.8155,
+      "step": 3980
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.00022961387434554972,
+      "loss": 0.461,
+      "step": 3990
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.0002294175392670157,
+      "loss": 0.3872,
+      "step": 4000
+    },
+    {
+      "epoch": 5.1,
+      "eval_loss": 0.5169897079467773,
+      "eval_runtime": 1364.33,
+      "eval_samples_per_second": 4.569,
+      "eval_steps_per_second": 0.572,
+      "eval_wer": 0.47175703053846585,
+      "step": 4000
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.00022922120418848165,
+      "loss": 0.3563,
+      "step": 4010
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.00022902486910994765,
+      "loss": 0.4178,
+      "step": 4020
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.0002288285340314136,
+      "loss": 0.7408,
+      "step": 4030
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.00022863219895287955,
+      "loss": 0.4519,
+      "step": 4040
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.00022843586387434552,
+      "loss": 0.3947,
+      "step": 4050
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.00022823952879581148,
+      "loss": 0.3745,
+      "step": 4060
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00022804319371727745,
+      "loss": 0.4124,
+      "step": 4070
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.00022784685863874344,
+      "loss": 0.7667,
+      "step": 4080
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.0002276505235602094,
+      "loss": 0.4667,
+      "step": 4090
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.00022745418848167537,
+      "loss": 0.3947,
+      "step": 4100
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.00022725785340314134,
+      "loss": 0.35,
+      "step": 4110
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.0002270615183246073,
+      "loss": 0.4057,
+      "step": 4120
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.0002268651832460733,
+      "loss": 0.7254,
+      "step": 4130
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.00022666884816753926,
+      "loss": 0.4605,
+      "step": 4140
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.00022647251308900523,
+      "loss": 0.389,
+      "step": 4150
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.00022627617801047116,
+      "loss": 0.3588,
+      "step": 4160
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.00022607984293193713,
+      "loss": 0.4358,
+      "step": 4170
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.00022588350785340312,
+      "loss": 0.7098,
+      "step": 4180
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.0002256871727748691,
+      "loss": 0.4503,
+      "step": 4190
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.00022549083769633505,
+      "loss": 0.3745,
+      "step": 4200
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.00022529450261780102,
+      "loss": 0.3538,
+      "step": 4210
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.00022509816753926698,
+      "loss": 0.3951,
+      "step": 4220
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.00022490183246073298,
+      "loss": 0.6996,
+      "step": 4230
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00022470549738219894,
+      "loss": 0.4202,
+      "step": 4240
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.0002245091623036649,
+      "loss": 0.3801,
+      "step": 4250
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.00022431282722513087,
+      "loss": 0.353,
+      "step": 4260
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.00022411649214659684,
+      "loss": 0.3701,
+      "step": 4270
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.00022392015706806283,
+      "loss": 0.6898,
+      "step": 4280
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.0002237238219895288,
+      "loss": 0.4247,
+      "step": 4290
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.00022352748691099474,
+      "loss": 0.3809,
+      "step": 4300
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.0002233311518324607,
+      "loss": 0.3563,
+      "step": 4310
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00022313481675392667,
+      "loss": 0.3691,
+      "step": 4320
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.00022293848167539266,
+      "loss": 0.6588,
+      "step": 4330
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.00022274214659685862,
+      "loss": 0.4361,
+      "step": 4340
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.0002225458115183246,
+      "loss": 0.3538,
+      "step": 4350
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.00022234947643979056,
+      "loss": 0.3586,
+      "step": 4360
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.00022215314136125652,
+      "loss": 0.4101,
+      "step": 4370
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.00022195680628272251,
+      "loss": 0.6699,
+      "step": 4380
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.00022176047120418848,
+      "loss": 0.423,
+      "step": 4390
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.00022156413612565444,
+      "loss": 0.3654,
+      "step": 4400
+    },
+    {
+      "epoch": 5.61,
+      "eval_loss": 0.43857690691947937,
+      "eval_runtime": 1365.733,
+      "eval_samples_per_second": 4.564,
+      "eval_steps_per_second": 0.571,
+      "eval_wer": 0.4419620453151428,
+      "step": 4400
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.0002213678010471204,
+      "loss": 0.3372,
+      "step": 4410
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.00022117146596858635,
+      "loss": 0.3896,
+      "step": 4420
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.00022097513089005231,
+      "loss": 0.6146,
+      "step": 4430
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.0002207787958115183,
+      "loss": 0.4175,
+      "step": 4440
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.00022058246073298427,
+      "loss": 0.3842,
+      "step": 4450
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.00022038612565445024,
+      "loss": 0.3201,
+      "step": 4460
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.0002201897905759162,
+      "loss": 0.4245,
+      "step": 4470
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.00021999345549738217,
+      "loss": 0.597,
+      "step": 4480
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00021979712041884816,
+      "loss": 0.4227,
+      "step": 4490
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.00021960078534031413,
+      "loss": 0.3707,
+      "step": 4500
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.0002194044502617801,
+      "loss": 0.335,
+      "step": 4510
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.00021920811518324606,
+      "loss": 0.3694,
+      "step": 4520
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.000219011780104712,
+      "loss": 0.6092,
+      "step": 4530
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.00021881544502617802,
+      "loss": 0.432,
+      "step": 4540
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.00021861910994764395,
+      "loss": 0.3613,
+      "step": 4550
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.00021842277486910992,
+      "loss": 0.3221,
+      "step": 4560
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.00021822643979057589,
+      "loss": 0.3863,
+      "step": 4570
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00021803010471204185,
+      "loss": 0.6141,
+      "step": 4580
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.00021783376963350784,
+      "loss": 0.4157,
+      "step": 4590
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.0002176374345549738,
+      "loss": 0.3407,
+      "step": 4600
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.00021744109947643977,
+      "loss": 0.3217,
+      "step": 4610
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00021724476439790574,
+      "loss": 0.3651,
+      "step": 4620
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.0002170484293193717,
+      "loss": 0.5891,
+      "step": 4630
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.0002168520942408377,
+      "loss": 0.4029,
+      "step": 4640
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.00021665575916230366,
+      "loss": 0.3595,
+      "step": 4650
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00021645942408376963,
+      "loss": 0.3529,
+      "step": 4660
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.00021626308900523557,
+      "loss": 0.4116,
+      "step": 4670
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.00021606675392670153,
+      "loss": 0.5429,
+      "step": 4680
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.00021587041884816753,
+      "loss": 0.3666,
+      "step": 4690
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.0002156740837696335,
+      "loss": 0.3353,
+      "step": 4700
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.00021547774869109946,
+      "loss": 0.5384,
+      "step": 4710
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.00021528141361256542,
+      "loss": 0.3432,
+      "step": 4720
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.0002150850785340314,
+      "loss": 0.2736,
+      "step": 4730
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00021488874345549735,
+      "loss": 0.2313,
+      "step": 4740
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.00021469240837696335,
+      "loss": 0.2382,
+      "step": 4750
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.0002144960732984293,
+      "loss": 0.4981,
+      "step": 4760
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.00021429973821989528,
+      "loss": 0.3455,
+      "step": 4770
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00021410340314136124,
+      "loss": 0.2643,
+      "step": 4780
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.00021390706806282718,
+      "loss": 0.254,
+      "step": 4790
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.00021371073298429317,
+      "loss": 0.2672,
+      "step": 4800
+    },
+    {
+      "epoch": 6.12,
+      "eval_loss": 0.5185807347297668,
+      "eval_runtime": 1365.146,
+      "eval_samples_per_second": 4.566,
+      "eval_steps_per_second": 0.571,
+      "eval_wer": 0.46785375736510476,
+      "step": 4800
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.00021351439790575914,
+      "loss": 0.5044,
+      "step": 4810
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.0002133180628272251,
+      "loss": 0.3535,
+      "step": 4820
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00021312172774869107,
+      "loss": 0.2837,
+      "step": 4830
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.00021292539267015704,
+      "loss": 0.2281,
+      "step": 4840
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.00021272905759162303,
+      "loss": 0.2728,
+      "step": 4850
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.000212532722513089,
+      "loss": 0.4953,
+      "step": 4860
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.00021233638743455496,
+      "loss": 0.3822,
+      "step": 4870
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.00021214005235602092,
+      "loss": 0.2864,
+      "step": 4880
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.0002119437172774869,
+      "loss": 0.2488,
+      "step": 4890
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.00021174738219895288,
+      "loss": 0.2436,
+      "step": 4900
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00021155104712041885,
+      "loss": 0.4638,
+      "step": 4910
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.0002113547120418848,
+      "loss": 0.3644,
+      "step": 4920
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.00021115837696335075,
+      "loss": 0.2702,
+      "step": 4930
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.00021096204188481672,
+      "loss": 0.2489,
+      "step": 4940
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.0002107657068062827,
+      "loss": 0.2662,
+      "step": 4950
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.00021056937172774868,
+      "loss": 0.4687,
+      "step": 4960
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.00021037303664921464,
+      "loss": 0.3702,
+      "step": 4970
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.0002101767015706806,
+      "loss": 0.2909,
+      "step": 4980
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.00020998036649214657,
+      "loss": 0.2428,
+      "step": 4990
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.00020978403141361257,
+      "loss": 0.2843,
+      "step": 5000
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.00020958769633507853,
+      "loss": 0.4471,
+      "step": 5010
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.0002093913612565445,
+      "loss": 0.3476,
+      "step": 5020
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00020919502617801046,
+      "loss": 0.3004,
+      "step": 5030
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.0002089986910994764,
+      "loss": 0.2578,
+      "step": 5040
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.00020880235602094237,
+      "loss": 0.2586,
+      "step": 5050
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.00020860602094240836,
+      "loss": 0.4379,
+      "step": 5060
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.00020840968586387432,
+      "loss": 0.3609,
+      "step": 5070
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.0002082133507853403,
+      "loss": 0.2871,
+      "step": 5080
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.00020801701570680626,
+      "loss": 0.2576,
+      "step": 5090
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.00020782068062827222,
+      "loss": 0.2512,
+      "step": 5100
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.0002076243455497382,
+      "loss": 0.4695,
+      "step": 5110
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.00020742801047120418,
+      "loss": 0.3397,
+      "step": 5120
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.00020723167539267014,
+      "loss": 0.2798,
+      "step": 5130
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.0002070353403141361,
+      "loss": 0.2453,
+      "step": 5140
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.00020683900523560208,
+      "loss": 0.2512,
+      "step": 5150
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00020664267015706807,
+      "loss": 0.5053,
+      "step": 5160
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.000206446335078534,
+      "loss": 0.3528,
+      "step": 5170
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.00020624999999999997,
+      "loss": 0.2963,
+      "step": 5180
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.00020605366492146594,
+      "loss": 0.263,
+      "step": 5190
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.0002058573298429319,
+      "loss": 0.2519,
+      "step": 5200
+    },
+    {
+      "epoch": 6.63,
+      "eval_loss": 0.4238142669200897,
+      "eval_runtime": 1364.3081,
+      "eval_samples_per_second": 4.569,
+      "eval_steps_per_second": 0.572,
+      "eval_wer": 0.4177245776100816,
+      "step": 5200
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.0002056609947643979,
+      "loss": 0.4852,
+      "step": 5210
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.00020546465968586386,
+      "loss": 0.3287,
+      "step": 5220
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.00020526832460732983,
+      "loss": 0.277,
+      "step": 5230
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.0002050719895287958,
+      "loss": 0.2373,
+      "step": 5240
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.00020487565445026176,
+      "loss": 0.2603,
+      "step": 5250
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.00020467931937172775,
+      "loss": 0.4455,
+      "step": 5260
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.00020448298429319372,
+      "loss": 0.34,
+      "step": 5270
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.00020428664921465968,
+      "loss": 0.2875,
+      "step": 5280
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00020409031413612562,
+      "loss": 0.2555,
+      "step": 5290
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.00020389397905759159,
+      "loss": 0.2654,
+      "step": 5300
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.00020369764397905758,
+      "loss": 0.4317,
+      "step": 5310
+    },
+    {
+      "epoch": 6.79,
+      "learning_rate": 0.00020350130890052354,
+      "loss": 0.3373,
+      "step": 5320
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.0002033049738219895,
+      "loss": 0.2725,
+      "step": 5330
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 0.00020310863874345547,
+      "loss": 0.2231,
+      "step": 5340
+    },
+    {
+      "epoch": 6.82,
+      "learning_rate": 0.00020291230366492144,
+      "loss": 0.2662,
+      "step": 5350
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 0.0002027159685863874,
+      "loss": 0.4472,
+      "step": 5360
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.0002025196335078534,
+      "loss": 0.3278,
+      "step": 5370
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 0.00020232329842931936,
+      "loss": 0.2765,
+      "step": 5380
+    },
+    {
+      "epoch": 6.87,
+      "learning_rate": 0.00020212696335078533,
+      "loss": 0.2356,
+      "step": 5390
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 0.0002019306282722513,
+      "loss": 0.2468,
+      "step": 5400
+    },
+    {
+      "epoch": 6.9,
+      "learning_rate": 0.00020173429319371723,
+      "loss": 0.4314,
+      "step": 5410
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00020153795811518325,
+      "loss": 0.3245,
+      "step": 5420
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 0.0002013416230366492,
+      "loss": 0.289,
+      "step": 5430
+    },
+    {
+      "epoch": 6.94,
+      "learning_rate": 0.00020114528795811516,
+      "loss": 0.232,
+      "step": 5440
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 0.00020094895287958112,
+      "loss": 0.2622,
+      "step": 5450
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.0002007526178010471,
+      "loss": 0.4571,
+      "step": 5460
+    },
+    {
+      "epoch": 6.98,
+      "learning_rate": 0.00020055628272251308,
+      "loss": 0.3164,
+      "step": 5470
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 0.00020035994764397905,
+      "loss": 0.2764,
+      "step": 5480
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 0.000200163612565445,
+      "loss": 0.3875,
+      "step": 5490
+    },
+    {
+      "epoch": 7.02,
+      "learning_rate": 0.00019996727748691098,
+      "loss": 0.325,
+      "step": 5500
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 0.00019977094240837694,
+      "loss": 0.2272,
+      "step": 5510
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 0.00019957460732984293,
+      "loss": 0.1944,
+      "step": 5520
+    },
+    {
+      "epoch": 7.05,
+      "learning_rate": 0.0001993782722513089,
+      "loss": 0.1683,
+      "step": 5530
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.00019918193717277487,
+      "loss": 0.3496,
+      "step": 5540
+    },
+    {
+      "epoch": 7.08,
+      "learning_rate": 0.0001990052356020942,
+      "loss": 0.3397,
+      "step": 5550
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 0.00019880890052356018,
+      "loss": 0.2654,
+      "step": 5560
+    },
+    {
+      "epoch": 7.1,
+      "learning_rate": 0.00019861256544502617,
+      "loss": 0.192,
+      "step": 5570
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.00019841623036649214,
+      "loss": 0.1725,
+      "step": 5580
+    },
+    {
+      "epoch": 7.13,
+      "learning_rate": 0.0001982198952879581,
+      "loss": 0.3103,
+      "step": 5590
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 0.00019802356020942407,
+      "loss": 0.3293,
+      "step": 5600
+    },
+    {
+      "epoch": 7.14,
+      "eval_loss": 0.3583916127681732,
+      "eval_runtime": 1369.6624,
+      "eval_samples_per_second": 4.551,
+      "eval_steps_per_second": 0.569,
+      "eval_wer": 0.3969535882232672,
+      "step": 5600
+    },
+    {
+      "epoch": 7.16,
+      "learning_rate": 0.00019782722513089003,
+      "loss": 0.2435,
+      "step": 5610
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.00019763089005235603,
+      "loss": 0.205,
+      "step": 5620
+    },
+    {
+      "epoch": 7.18,
+      "learning_rate": 0.000197434554973822,
+      "loss": 0.1868,
+      "step": 5630
+    },
+    {
+      "epoch": 7.19,
+      "learning_rate": 0.00019723821989528793,
+      "loss": 0.2964,
+      "step": 5640
+    },
+    {
+      "epoch": 7.21,
+      "learning_rate": 0.0001970418848167539,
+      "loss": 0.3549,
+      "step": 5650
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 0.00019684554973821986,
+      "loss": 0.2544,
+      "step": 5660
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.00019664921465968585,
+      "loss": 0.2109,
+      "step": 5670
+    },
+    {
+      "epoch": 7.24,
+      "learning_rate": 0.00019645287958115182,
+      "loss": 0.1782,
+      "step": 5680
+    },
+    {
+      "epoch": 7.26,
+      "learning_rate": 0.00019625654450261778,
+      "loss": 0.3182,
+      "step": 5690
+    },
+    {
+      "epoch": 7.27,
+      "learning_rate": 0.00019606020942408375,
+      "loss": 0.3435,
+      "step": 5700
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00019586387434554972,
+      "loss": 0.2444,
+      "step": 5710
+    },
+    {
+      "epoch": 7.3,
+      "learning_rate": 0.0001956675392670157,
+      "loss": 0.1897,
+      "step": 5720
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 0.00019547120418848167,
+      "loss": 0.2053,
+      "step": 5730
+    },
+    {
+      "epoch": 7.32,
+      "learning_rate": 0.00019527486910994764,
+      "loss": 0.3107,
+      "step": 5740
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.0001950785340314136,
+      "loss": 0.3381,
+      "step": 5750
+    },
+    {
+      "epoch": 7.35,
+      "learning_rate": 0.00019488219895287954,
+      "loss": 0.2586,
+      "step": 5760
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 0.00019468586387434554,
+      "loss": 0.2043,
+      "step": 5770
+    },
+    {
+      "epoch": 7.37,
+      "learning_rate": 0.0001944895287958115,
+      "loss": 0.1843,
+      "step": 5780
+    },
+    {
+      "epoch": 7.38,
+      "learning_rate": 0.00019429319371727747,
+      "loss": 0.3085,
+      "step": 5790
+    },
+    {
+      "epoch": 7.4,
+      "learning_rate": 0.00019409685863874343,
+      "loss": 0.3402,
+      "step": 5800
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 0.0001939005235602094,
+      "loss": 0.2324,
+      "step": 5810
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 0.00019370418848167536,
+      "loss": 0.1846,
+      "step": 5820
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00019350785340314136,
+      "loss": 0.179,
+      "step": 5830
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 0.00019331151832460732,
+      "loss": 0.3396,
+      "step": 5840
+    },
+    {
+      "epoch": 7.46,
+      "learning_rate": 0.0001931151832460733,
+      "loss": 0.3293,
+      "step": 5850
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 0.00019291884816753925,
+      "loss": 0.2491,
+      "step": 5860
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.0001927225130890052,
+      "loss": 0.2153,
+      "step": 5870
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 0.0001925261780104712,
+      "loss": 0.1753,
+      "step": 5880
+    },
+    {
+      "epoch": 7.51,
+      "learning_rate": 0.00019232984293193715,
+      "loss": 0.3342,
+      "step": 5890
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 0.00019213350785340311,
+      "loss": 0.3418,
+      "step": 5900
+    },
+    {
+      "epoch": 7.54,
+      "learning_rate": 0.00019193717277486908,
+      "loss": 0.2302,
+      "step": 5910
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00019174083769633505,
+      "loss": 0.1994,
+      "step": 5920
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 0.00019154450261780104,
+      "loss": 0.1872,
+      "step": 5930
+    },
+    {
+      "epoch": 7.58,
+      "learning_rate": 0.000191348167539267,
+      "loss": 0.3014,
+      "step": 5940
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 0.00019115183246073297,
+      "loss": 0.3302,
+      "step": 5950
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.00019095549738219893,
+      "loss": 0.2249,
+      "step": 5960
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 0.0001907591623036649,
+      "loss": 0.1994,
+      "step": 5970
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 0.0001905628272251309,
+      "loss": 0.1859,
+      "step": 5980
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 0.00019036649214659686,
+      "loss": 0.3089,
+      "step": 5990
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 0.00019017015706806282,
+      "loss": 0.314,
+      "step": 6000
+    },
+    {
+      "epoch": 7.65,
+      "eval_loss": 0.3325476050376892,
+      "eval_runtime": 1361.7679,
+      "eval_samples_per_second": 4.577,
+      "eval_steps_per_second": 0.573,
+      "eval_wer": 0.39112655898589244,
+      "step": 6000
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 0.00018997382198952876,
+      "loss": 0.2404,
+      "step": 6010
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 0.00018977748691099473,
+      "loss": 0.1933,
+      "step": 6020
+    },
+    {
+      "epoch": 7.69,
+      "learning_rate": 0.00018958115183246072,
+      "loss": 0.1847,
+      "step": 6030
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 0.00018938481675392669,
+      "loss": 0.3643,
+      "step": 6040
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 0.00018918848167539265,
+      "loss": 0.3283,
+      "step": 6050
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 0.00018899214659685862,
+      "loss": 0.2328,
+      "step": 6060
+    },
+    {
+      "epoch": 7.74,
+      "learning_rate": 0.00018879581151832458,
+      "loss": 0.2015,
+      "step": 6070
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 0.00018859947643979058,
+      "loss": 0.172,
+      "step": 6080
+    },
+    {
+      "epoch": 7.77,
+      "learning_rate": 0.00018840314136125654,
+      "loss": 0.2959,
+      "step": 6090
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 0.0001882068062827225,
+      "loss": 0.333,
+      "step": 6100
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 0.00018801047120418847,
+      "loss": 0.2308,
+      "step": 6110
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 0.00018781413612565444,
+      "loss": 0.2256,
+      "step": 6120
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00018761780104712038,
+      "loss": 0.1825,
+      "step": 6130
+    },
+    {
+      "epoch": 7.83,
+      "learning_rate": 0.00018742146596858637,
+      "loss": 0.2924,
+      "step": 6140
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 0.00018722513089005233,
+      "loss": 0.3146,
+      "step": 6150
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 0.0001870287958115183,
+      "loss": 0.2384,
+      "step": 6160
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00018683246073298426,
+      "loss": 0.1956,
+      "step": 6170
+    },
+    {
+      "epoch": 7.88,
+      "learning_rate": 0.00018663612565445023,
+      "loss": 0.1781,
+      "step": 6180
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 0.00018643979057591622,
+      "loss": 0.3478,
+      "step": 6190
+    },
+    {
+      "epoch": 7.91,
+      "learning_rate": 0.0001862434554973822,
+      "loss": 0.3168,
+      "step": 6200
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00018604712041884815,
+      "loss": 0.2203,
+      "step": 6210
+    },
+    {
+      "epoch": 7.93,
+      "learning_rate": 0.00018585078534031412,
+      "loss": 0.1985,
+      "step": 6220
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 0.00018565445026178009,
+      "loss": 0.1816,
+      "step": 6230
+    },
+    {
+      "epoch": 7.96,
+      "learning_rate": 0.00018545811518324608,
+      "loss": 0.3022,
+      "step": 6240
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 0.00018526178010471204,
+      "loss": 0.2997,
+      "step": 6250
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00018506544502617798,
+      "loss": 0.1941,
+      "step": 6260
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 0.00018486910994764395,
+      "loss": 0.2014,
+      "step": 6270
+    },
+    {
+      "epoch": 8.01,
+      "learning_rate": 0.0001846727748691099,
+      "loss": 0.3263,
+      "step": 6280
+    },
+    {
+      "epoch": 8.02,
+      "learning_rate": 0.0001844764397905759,
+      "loss": 0.2091,
+      "step": 6290
+    },
+    {
+      "epoch": 8.04,
+      "learning_rate": 0.00018428010471204187,
+      "loss": 0.1689,
+      "step": 6300
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 0.00018408376963350784,
+      "loss": 0.139,
+      "step": 6310
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 0.0001838874345549738,
+      "loss": 0.1635,
+      "step": 6320
+    },
+    {
+      "epoch": 8.07,
+      "learning_rate": 0.00018369109947643977,
+      "loss": 0.3436,
+      "step": 6330
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 0.00018349476439790576,
+      "loss": 0.2094,
+      "step": 6340
+    },
+    {
+      "epoch": 8.1,
+      "learning_rate": 0.00018329842931937173,
+      "loss": 0.1653,
+      "step": 6350
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 0.0001831020942408377,
+      "loss": 0.1412,
+      "step": 6360
+    },
+    {
+      "epoch": 8.12,
+      "learning_rate": 0.00018290575916230366,
+      "loss": 0.1905,
+      "step": 6370
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.0001827094240837696,
+      "loss": 0.3324,
+      "step": 6380
+    },
+    {
+      "epoch": 8.15,
+      "learning_rate": 0.00018251308900523561,
+      "loss": 0.2082,
+      "step": 6390
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 0.00018231675392670155,
+      "loss": 0.1698,
+      "step": 6400
+    },
+    {
+      "epoch": 8.16,
+      "eval_loss": 0.34114429354667664,
+      "eval_runtime": 1398.1775,
+      "eval_samples_per_second": 4.458,
+      "eval_steps_per_second": 0.558,
+      "eval_wer": 0.38549469340718573,
+      "step": 6400
+    },
+    {
+      "epoch": 8.18,
+      "learning_rate": 0.00018212041884816752,
+      "loss": 0.1344,
+      "step": 6410
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00018192408376963348,
+      "loss": 0.1596,
+      "step": 6420
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 0.00018172774869109945,
+      "loss": 0.3183,
+      "step": 6430
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 0.00018153141361256542,
+      "loss": 0.2245,
+      "step": 6440
+    },
+    {
+      "epoch": 8.23,
+      "learning_rate": 0.0001813350785340314,
+      "loss": 0.1684,
+      "step": 6450
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00018113874345549737,
+      "loss": 0.1487,
+      "step": 6460
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 0.00018094240837696334,
+      "loss": 0.1856,
+      "step": 6470
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 0.0001807460732984293,
+      "loss": 0.3307,
+      "step": 6480
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 0.00018054973821989527,
+      "loss": 0.2087,
+      "step": 6490
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 0.00018035340314136126,
+      "loss": 0.1641,
+      "step": 6500
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.0001801570680628272,
+      "loss": 0.1669,
+      "step": 6510
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 0.00017996073298429317,
+      "loss": 0.186,
+      "step": 6520
+    },
+    {
+      "epoch": 8.33,
+      "learning_rate": 0.00017976439790575913,
+      "loss": 0.3674,
+      "step": 6530
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 0.0001795680628272251,
+      "loss": 0.2171,
+      "step": 6540
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.0001793717277486911,
+      "loss": 0.1897,
+      "step": 6550
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 0.00017917539267015706,
+      "loss": 0.1413,
+      "step": 6560
+    },
+    {
+      "epoch": 8.38,
+      "learning_rate": 0.00017897905759162302,
+      "loss": 0.1873,
+      "step": 6570
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 0.000178782722513089,
+      "loss": 0.3289,
+      "step": 6580
+    },
+    {
+      "epoch": 8.41,
+      "learning_rate": 0.00017858638743455495,
+      "loss": 0.2077,
+      "step": 6590
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 0.00017839005235602094,
+      "loss": 0.1808,
+      "step": 6600
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 0.0001781937172774869,
+      "loss": 0.1488,
+      "step": 6610
+    },
+    {
+      "epoch": 8.44,
+      "learning_rate": 0.00017799738219895288,
+      "loss": 0.1871,
+      "step": 6620
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00017780104712041881,
+      "loss": 0.3656,
+      "step": 6630
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 0.00017760471204188478,
+      "loss": 0.2395,
+      "step": 6640
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 0.00017740837696335077,
+      "loss": 0.1906,
+      "step": 6650
+    },
+    {
+      "epoch": 8.49,
+      "learning_rate": 0.00017721204188481674,
+      "loss": 0.1456,
+      "step": 6660
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.0001770157068062827,
+      "loss": 0.1786,
+      "step": 6670
+    },
+    {
+      "epoch": 8.52,
+      "learning_rate": 0.00017681937172774867,
+      "loss": 0.368,
+      "step": 6680
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 0.00017662303664921463,
+      "loss": 0.208,
+      "step": 6690
+    },
+    {
+      "epoch": 8.55,
+      "learning_rate": 0.00017642670157068063,
+      "loss": 0.164,
+      "step": 6700
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 0.0001762303664921466,
+      "loss": 0.1565,
+      "step": 6710
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00017603403141361256,
+      "loss": 0.1861,
+      "step": 6720
+    },
+    {
+      "epoch": 8.58,
+      "learning_rate": 0.00017583769633507852,
+      "loss": 0.3166,
+      "step": 6730
+    },
+    {
+      "epoch": 8.6,
+      "learning_rate": 0.0001756413612565445,
+      "loss": 0.2314,
+      "step": 6740
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 0.00017544502617801043,
+      "loss": 0.1798,
+      "step": 6750
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00017524869109947645,
+      "loss": 0.1489,
+      "step": 6760
+    },
+    {
+      "epoch": 8.63,
+      "learning_rate": 0.00017505235602094239,
+      "loss": 0.171,
+      "step": 6770
+    },
+    {
+      "epoch": 8.65,
+      "learning_rate": 0.00017485602094240835,
+      "loss": 0.3342,
+      "step": 6780
+    },
+    {
+      "epoch": 8.66,
+      "learning_rate": 0.00017465968586387432,
+      "loss": 0.2143,
+      "step": 6790
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00017446335078534028,
+      "loss": 0.1682,
+      "step": 6800
+    },
+    {
+      "epoch": 8.67,
+      "eval_loss": 0.32393980026245117,
+      "eval_runtime": 1364.7581,
+      "eval_samples_per_second": 4.567,
+      "eval_steps_per_second": 0.572,
+      "eval_wer": 0.3800858720098139,
+      "step": 6800
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 0.00017426701570680627,
+      "loss": 0.1393,
+      "step": 6810
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 0.00017407068062827224,
+      "loss": 0.1638,
+      "step": 6820
+    },
+    {
+      "epoch": 8.71,
+      "learning_rate": 0.0001738743455497382,
+      "loss": 0.3465,
+      "step": 6830
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 0.00017367801047120417,
+      "loss": 0.2254,
+      "step": 6840
+    },
+    {
+      "epoch": 8.74,
+      "learning_rate": 0.00017348167539267014,
+      "loss": 0.1721,
+      "step": 6850
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 0.00017328534031413613,
+      "loss": 0.1507,
+      "step": 6860
+    },
+    {
+      "epoch": 8.76,
+      "learning_rate": 0.0001730890052356021,
+      "loss": 0.1939,
+      "step": 6870
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 0.00017289267015706806,
+      "loss": 0.3321,
+      "step": 6880
+    },
+    {
+      "epoch": 8.79,
+      "learning_rate": 0.000172696335078534,
+      "loss": 0.2132,
+      "step": 6890
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 0.00017249999999999996,
+      "loss": 0.1872,
+      "step": 6900
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 0.00017230366492146596,
+      "loss": 0.1524,
+      "step": 6910
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00017210732984293192,
+      "loss": 0.1913,
+      "step": 6920
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 0.0001719109947643979,
+      "loss": 0.322,
+      "step": 6930
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 0.00017171465968586385,
+      "loss": 0.2124,
+      "step": 6940
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 0.00017151832460732982,
+      "loss": 0.18,
+      "step": 6950
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 0.0001713219895287958,
+      "loss": 0.133,
+      "step": 6960
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00017112565445026178,
+      "loss": 0.1686,
+      "step": 6970
+    },
+    {
+      "epoch": 8.9,
+      "learning_rate": 0.00017092931937172774,
+      "loss": 0.328,
+      "step": 6980
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 0.0001707329842931937,
+      "loss": 0.2151,
+      "step": 6990
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 0.00017053664921465965,
+      "loss": 0.1738,
+      "step": 7000
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00017034031413612567,
+      "loss": 0.1505,
+      "step": 7010
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 0.0001701439790575916,
+      "loss": 0.1733,
+      "step": 7020
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 0.00016994764397905757,
+      "loss": 0.3218,
+      "step": 7030
+    },
+    {
+      "epoch": 8.98,
+      "learning_rate": 0.00016975130890052354,
+      "loss": 0.1946,
+      "step": 7040
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.0001695549738219895,
+      "loss": 0.1476,
+      "step": 7050
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 0.00016935863874345547,
+      "loss": 0.267,
+      "step": 7060
+    },
+    {
+      "epoch": 9.02,
+      "learning_rate": 0.00016916230366492146,
+      "loss": 0.2029,
+      "step": 7070
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 0.00016896596858638743,
+      "loss": 0.1506,
+      "step": 7080
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 0.0001687696335078534,
+      "loss": 0.1325,
+      "step": 7090
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 0.00016857329842931936,
+      "loss": 0.1385,
+      "step": 7100
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 0.00016837696335078532,
+      "loss": 0.2153,
+      "step": 7110
+    },
+    {
+      "epoch": 9.08,
+      "learning_rate": 0.00016818062827225131,
+      "loss": 0.2034,
+      "step": 7120
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 0.00016798429319371728,
+      "loss": 0.1632,
+      "step": 7130
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 0.00016778795811518322,
+      "loss": 0.1309,
+      "step": 7140
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 0.00016759162303664918,
+      "loss": 0.1267,
+      "step": 7150
+    },
+    {
+      "epoch": 9.13,
+      "learning_rate": 0.00016739528795811515,
+      "loss": 0.2547,
+      "step": 7160
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00016719895287958114,
+      "loss": 0.2026,
+      "step": 7170
+    },
+    {
+      "epoch": 9.16,
+      "learning_rate": 0.0001670026178010471,
+      "loss": 0.147,
+      "step": 7180
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 0.00016680628272251307,
+      "loss": 0.1161,
+      "step": 7190
+    },
+    {
+      "epoch": 9.18,
+      "learning_rate": 0.00016660994764397904,
+      "loss": 0.1325,
+      "step": 7200
+    },
+    {
+      "epoch": 9.18,
+      "eval_loss": 0.34735119342803955,
+      "eval_runtime": 1369.1036,
+      "eval_samples_per_second": 4.553,
+      "eval_steps_per_second": 0.57,
+      "eval_wer": 0.38318990353339155,
+      "step": 7200
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 0.000166413612565445,
+      "loss": 0.2826,
+      "step": 7210
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.000166217277486911,
+      "loss": 0.2062,
+      "step": 7220
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 0.00016602094240837696,
+      "loss": 0.1644,
+      "step": 7230
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 0.00016582460732984293,
+      "loss": 0.143,
+      "step": 7240
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 0.0001656282722513089,
+      "loss": 0.1266,
+      "step": 7250
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00016543193717277483,
+      "loss": 0.2596,
+      "step": 7260
+    },
+    {
+      "epoch": 9.27,
+      "learning_rate": 0.00016523560209424082,
+      "loss": 0.2224,
+      "step": 7270
+    },
+    {
+      "epoch": 9.29,
+      "learning_rate": 0.0001650392670157068,
+      "loss": 0.1756,
+      "step": 7280
+    },
+    {
+      "epoch": 9.3,
+      "learning_rate": 0.00016484293193717276,
+      "loss": 0.1324,
+      "step": 7290
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00016464659685863872,
+      "loss": 0.1286,
+      "step": 7300
+    },
+    {
+      "epoch": 9.32,
+      "learning_rate": 0.00016445026178010469,
+      "loss": 0.2256,
+      "step": 7310
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 0.00016425392670157068,
+      "loss": 0.1991,
+      "step": 7320
+    },
+    {
+      "epoch": 9.35,
+      "learning_rate": 0.00016405759162303664,
+      "loss": 0.1505,
+      "step": 7330
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 0.0001638612565445026,
+      "loss": 0.1245,
+      "step": 7340
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00016366492146596858,
+      "loss": 0.1287,
+      "step": 7350
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 0.00016346858638743454,
+      "loss": 0.2492,
+      "step": 7360
+    },
+    {
+      "epoch": 9.4,
+      "learning_rate": 0.00016327225130890053,
+      "loss": 0.2291,
+      "step": 7370
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 0.0001630759162303665,
+      "loss": 0.1614,
+      "step": 7380
+    },
+    {
+      "epoch": 9.43,
+      "learning_rate": 0.00016287958115183244,
+      "loss": 0.1247,
+      "step": 7390
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 0.0001626832460732984,
+      "loss": 0.1307,
+      "step": 7400
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 0.00016248691099476437,
+      "loss": 0.2721,
+      "step": 7410
+    },
+    {
+      "epoch": 9.46,
+      "learning_rate": 0.00016229057591623033,
+      "loss": 0.2251,
+      "step": 7420
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00016209424083769633,
+      "loss": 0.1712,
+      "step": 7430
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 0.0001618979057591623,
+      "loss": 0.1193,
+      "step": 7440
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 0.00016170157068062826,
+      "loss": 0.1309,
+      "step": 7450
+    },
+    {
+      "epoch": 9.51,
+      "learning_rate": 0.00016150523560209422,
+      "loss": 0.2234,
+      "step": 7460
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.0001613089005235602,
+      "loss": 0.2045,
+      "step": 7470
+    },
+    {
+      "epoch": 9.54,
+      "learning_rate": 0.00016111256544502618,
+      "loss": 0.1753,
+      "step": 7480
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 0.00016091623036649215,
+      "loss": 0.1531,
+      "step": 7490
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 0.0001607198952879581,
+      "loss": 0.1207,
+      "step": 7500
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.00016052356020942405,
+      "loss": 0.2653,
+      "step": 7510
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 0.00016032722513089002,
+      "loss": 0.1977,
+      "step": 7520
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 0.000160130890052356,
+      "loss": 0.1636,
+      "step": 7530
+    },
+    {
+      "epoch": 9.62,
+      "learning_rate": 0.00015993455497382197,
+      "loss": 0.1271,
+      "step": 7540
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 0.00015973821989528794,
+      "loss": 0.1216,
+      "step": 7550
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.0001595418848167539,
+      "loss": 0.2405,
+      "step": 7560
+    },
+    {
+      "epoch": 9.66,
+      "learning_rate": 0.00015934554973821987,
+      "loss": 0.2063,
+      "step": 7570
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 0.00015914921465968586,
+      "loss": 0.1432,
+      "step": 7580
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 0.00015895287958115183,
+      "loss": 0.1361,
+      "step": 7590
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.0001587565445026178,
+      "loss": 0.1577,
+      "step": 7600
+    },
+    {
+      "epoch": 9.69,
+      "eval_loss": 0.32892361283302307,
+      "eval_runtime": 1370.6266,
+      "eval_samples_per_second": 4.548,
+      "eval_steps_per_second": 0.569,
+      "eval_wer": 0.38386832958495193,
+      "step": 7600
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 0.00015856020942408376,
+      "loss": 0.2458,
+      "step": 7610
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 0.00015836387434554973,
+      "loss": 0.2081,
+      "step": 7620
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 0.00015816753926701572,
+      "loss": 0.1561,
+      "step": 7630
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00015797120418848168,
+      "loss": 0.121,
+      "step": 7640
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 0.00015777486910994762,
+      "loss": 0.1425,
+      "step": 7650
+    },
+    {
+      "epoch": 9.77,
+      "learning_rate": 0.0001575785340314136,
+      "loss": 0.2933,
+      "step": 7660
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 0.00015738219895287955,
+      "loss": 0.2143,
+      "step": 7670
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.00015718586387434555,
+      "loss": 0.1709,
+      "step": 7680
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 0.0001569895287958115,
+      "loss": 0.1308,
+      "step": 7690
+    },
+    {
+      "epoch": 9.82,
+      "learning_rate": 0.00015679319371727748,
+      "loss": 0.1145,
+      "step": 7700
+    },
+    {
+      "epoch": 9.83,
+      "learning_rate": 0.00015659685863874344,
+      "loss": 0.2629,
+      "step": 7710
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.0001564005235602094,
+      "loss": 0.2246,
+      "step": 7720
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 0.00015620418848167537,
+      "loss": 0.1456,
+      "step": 7730
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 0.00015600785340314137,
+      "loss": 0.1331,
+      "step": 7740
+    },
+    {
+      "epoch": 9.88,
+      "learning_rate": 0.00015581151832460733,
+      "loss": 0.1219,
+      "step": 7750
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.00015561518324607327,
+      "loss": 0.2413,
+      "step": 7760
+    },
+    {
+      "epoch": 9.91,
+      "learning_rate": 0.00015541884816753924,
+      "loss": 0.2153,
+      "step": 7770
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 0.0001552225130890052,
+      "loss": 0.1605,
+      "step": 7780
+    },
+    {
+      "epoch": 9.94,
+      "learning_rate": 0.0001550261780104712,
+      "loss": 0.1282,
+      "step": 7790
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 0.00015482984293193716,
+      "loss": 0.1246,
+      "step": 7800
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00015463350785340312,
+      "loss": 0.2416,
+      "step": 7810
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 0.0001544371727748691,
+      "loss": 0.1857,
+      "step": 7820
+    },
+    {
+      "epoch": 9.99,
+      "learning_rate": 0.00015424083769633506,
+      "loss": 0.1353,
+      "step": 7830
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 0.00015404450261780105,
+      "loss": 0.1872,
+      "step": 7840
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00015384816753926701,
+      "loss": 0.2413,
+      "step": 7850
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 0.00015365183246073298,
+      "loss": 0.1483,
+      "step": 7860
+    },
+    {
+      "epoch": 10.04,
+      "learning_rate": 0.00015345549738219894,
+      "loss": 0.108,
+      "step": 7870
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 0.00015325916230366488,
+      "loss": 0.1024,
+      "step": 7880
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 0.0001530628272251309,
+      "loss": 0.1305,
+      "step": 7890
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 0.00015286649214659684,
+      "loss": 0.2373,
+      "step": 7900
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 0.0001526701570680628,
+      "loss": 0.138,
+      "step": 7910
+    },
+    {
+      "epoch": 10.1,
+      "learning_rate": 0.00015247382198952877,
+      "loss": 0.1215,
+      "step": 7920
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 0.00015227748691099474,
+      "loss": 0.0984,
+      "step": 7930
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 0.00015208115183246073,
+      "loss": 0.153,
+      "step": 7940
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 0.0001518848167539267,
+      "loss": 0.2221,
+      "step": 7950
+    },
+    {
+      "epoch": 10.15,
+      "learning_rate": 0.00015168848167539266,
+      "loss": 0.1454,
+      "step": 7960
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 0.00015149214659685863,
+      "loss": 0.1387,
+      "step": 7970
+    },
+    {
+      "epoch": 10.18,
+      "learning_rate": 0.0001512958115183246,
+      "loss": 0.1012,
+      "step": 7980
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 0.00015109947643979059,
+      "loss": 0.1612,
+      "step": 7990
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 0.00015090314136125655,
+      "loss": 0.2259,
+      "step": 8000
+    },
+    {
+      "epoch": 10.2,
+      "eval_loss": 0.3182670474052429,
+      "eval_runtime": 1403.678,
+      "eval_samples_per_second": 4.44,
+      "eval_steps_per_second": 0.556,
+      "eval_wer": 0.3755971078604487,
+      "step": 8000
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 0.00015070680628272252,
+      "loss": 0.1496,
+      "step": 8010
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 0.00015051047120418845,
+      "loss": 0.1166,
+      "step": 8020
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 0.00015031413612565442,
+      "loss": 0.1027,
+      "step": 8030
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 0.00015011780104712039,
+      "loss": 0.1577,
+      "step": 8040
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 0.00014992146596858638,
+      "loss": 0.2267,
+      "step": 8050
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 0.00014972513089005234,
+      "loss": 0.1527,
+      "step": 8060
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 0.0001495287958115183,
+      "loss": 0.155,
+      "step": 8070
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 0.00014933246073298428,
+      "loss": 0.1066,
+      "step": 8080
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 0.00014913612565445024,
+      "loss": 0.1573,
+      "step": 8090
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 0.0001489397905759162,
+      "loss": 0.2327,
+      "step": 8100
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 0.0001487434554973822,
+      "loss": 0.1481,
+      "step": 8110
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 0.00014854712041884816,
+      "loss": 0.1226,
+      "step": 8120
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 0.0001483507853403141,
+      "loss": 0.1083,
+      "step": 8130
+    },
+    {
+      "epoch": 10.38,
+      "learning_rate": 0.0001481544502617801,
+      "loss": 0.1837,
+      "step": 8140
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 0.00014795811518324606,
+      "loss": 0.2375,
+      "step": 8150
+    },
+    {
+      "epoch": 10.41,
+      "learning_rate": 0.00014776178010471203,
+      "loss": 0.1662,
+      "step": 8160
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 0.000147565445026178,
+      "loss": 0.1381,
+      "step": 8170
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 0.00014736910994764396,
+      "loss": 0.1059,
+      "step": 8180
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 0.00014717277486910995,
+      "loss": 0.1407,
+      "step": 8190
+    },
+    {
+      "epoch": 10.46,
+      "learning_rate": 0.0001469764397905759,
+      "loss": 0.2118,
+      "step": 8200
+    },
+    {
+      "epoch": 10.47,
+      "learning_rate": 0.00014678010471204188,
+      "loss": 0.155,
+      "step": 8210
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 0.00014658376963350785,
+      "loss": 0.1282,
+      "step": 8220
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 0.0001463874345549738,
+      "loss": 0.105,
+      "step": 8230
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 0.00014619109947643978,
+      "loss": 0.1696,
+      "step": 8240
+    },
+    {
+      "epoch": 10.52,
+      "learning_rate": 0.00014599476439790574,
+      "loss": 0.2137,
+      "step": 8250
+    },
+    {
+      "epoch": 10.54,
+      "learning_rate": 0.0001457984293193717,
+      "loss": 0.1468,
+      "step": 8260
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 0.00014560209424083767,
+      "loss": 0.1105,
+      "step": 8270
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 0.00014540575916230364,
+      "loss": 0.1082,
+      "step": 8280
+    },
+    {
+      "epoch": 10.57,
+      "learning_rate": 0.00014520942408376963,
+      "loss": 0.1443,
+      "step": 8290
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 0.0001450130890052356,
+      "loss": 0.2309,
+      "step": 8300
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 0.00014481675392670156,
+      "loss": 0.1469,
+      "step": 8310
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 0.00014462041884816753,
+      "loss": 0.1148,
+      "step": 8320
+    },
+    {
+      "epoch": 10.62,
+      "learning_rate": 0.0001444240837696335,
+      "loss": 0.1147,
+      "step": 8330
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 0.00014422774869109946,
+      "loss": 0.1584,
+      "step": 8340
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 0.00014403141361256543,
+      "loss": 0.2482,
+      "step": 8350
+    },
+    {
+      "epoch": 10.66,
+      "learning_rate": 0.0001438547120418848,
+      "loss": 0.1481,
+      "step": 8360
+    },
+    {
+      "epoch": 10.68,
+      "learning_rate": 0.00014365837696335077,
+      "loss": 0.1099,
+      "step": 8370
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 0.00014346204188481673,
+      "loss": 0.1048,
+      "step": 8380
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 0.00014326570680628272,
+      "loss": 0.1322,
+      "step": 8390
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 0.0001430693717277487,
+      "loss": 0.2473,
+      "step": 8400
+    },
+    {
+      "epoch": 10.71,
+      "eval_loss": 0.3132154941558838,
+      "eval_runtime": 1383.8655,
+      "eval_samples_per_second": 4.504,
+      "eval_steps_per_second": 0.564,
+      "eval_wer": 0.365420717087043,
+      "step": 8400
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 0.00014287303664921463,
+      "loss": 0.1666,
+      "step": 8410
+    },
+    {
+      "epoch": 10.74,
+      "learning_rate": 0.00014267670157068062,
+      "loss": 0.1253,
+      "step": 8420
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 0.00014248036649214659,
+      "loss": 0.1079,
+      "step": 8430
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 0.00014228403141361255,
+      "loss": 0.1529,
+      "step": 8440
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 0.00014208769633507852,
+      "loss": 0.2307,
+      "step": 8450
+    },
+    {
+      "epoch": 10.79,
+      "learning_rate": 0.00014189136125654448,
+      "loss": 0.1581,
+      "step": 8460
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 0.00014169502617801047,
+      "loss": 0.1279,
+      "step": 8470
+    },
+    {
+      "epoch": 10.82,
+      "learning_rate": 0.0001414986910994764,
+      "loss": 0.1175,
+      "step": 8480
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 0.0001413023560209424,
+      "loss": 0.1331,
+      "step": 8490
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 0.00014110602094240837,
+      "loss": 0.2305,
+      "step": 8500
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 0.00014090968586387434,
+      "loss": 0.1484,
+      "step": 8510
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 0.0001407133507853403,
+      "loss": 0.1286,
+      "step": 8520
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 0.00014051701570680627,
+      "loss": 0.1073,
+      "step": 8530
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 0.00014032068062827226,
+      "loss": 0.1634,
+      "step": 8540
+    },
+    {
+      "epoch": 10.9,
+      "learning_rate": 0.0001401243455497382,
+      "loss": 0.2268,
+      "step": 8550
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 0.00013992801047120416,
+      "loss": 0.1565,
+      "step": 8560
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 0.00013973167539267016,
+      "loss": 0.1179,
+      "step": 8570
+    },
+    {
+      "epoch": 10.94,
+      "learning_rate": 0.00013953534031413612,
+      "loss": 0.1152,
+      "step": 8580
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 0.0001393390052356021,
+      "loss": 0.1438,
+      "step": 8590
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 0.00013914267015706805,
+      "loss": 0.206,
+      "step": 8600
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 0.00013894633507853402,
+      "loss": 0.1249,
+      "step": 8610
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 0.00013874999999999998,
+      "loss": 0.1218,
+      "step": 8620
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 0.00013855366492146595,
+      "loss": 0.234,
+      "step": 8630
+    },
+    {
+      "epoch": 11.02,
+      "learning_rate": 0.00013835732984293192,
+      "loss": 0.1624,
+      "step": 8640
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 0.0001381609947643979,
+      "loss": 0.1205,
+      "step": 8650
+    },
+    {
+      "epoch": 11.05,
+      "learning_rate": 0.00013796465968586385,
+      "loss": 0.085,
+      "step": 8660
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 0.00013776832460732984,
+      "loss": 0.092,
+      "step": 8670
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 0.0001375719895287958,
+      "loss": 0.1959,
+      "step": 8680
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 0.00013737565445026177,
+      "loss": 0.1396,
+      "step": 8690
+    },
+    {
+      "epoch": 11.1,
+      "learning_rate": 0.00013717931937172774,
+      "loss": 0.1236,
+      "step": 8700
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 0.0001369829842931937,
+      "loss": 0.0883,
+      "step": 8710
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 0.00013678664921465967,
+      "loss": 0.1172,
+      "step": 8720
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 0.00013659031413612563,
+      "loss": 0.2029,
+      "step": 8730
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 0.0001363939790575916,
+      "loss": 0.1558,
+      "step": 8740
+    },
+    {
+      "epoch": 11.16,
+      "learning_rate": 0.0001361976439790576,
+      "loss": 0.1264,
+      "step": 8750
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 0.00013600130890052356,
+      "loss": 0.1348,
+      "step": 8760
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 0.00013580497382198952,
+      "loss": 0.1041,
+      "step": 8770
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 0.0001356086387434555,
+      "loss": 0.1963,
+      "step": 8780
+    },
+    {
+      "epoch": 11.21,
+      "learning_rate": 0.00013541230366492145,
+      "loss": 0.1653,
+      "step": 8790
+    },
+    {
+      "epoch": 11.22,
+      "learning_rate": 0.00013521596858638742,
+      "loss": 0.1136,
+      "step": 8800
+    },
+    {
+      "epoch": 11.22,
+      "eval_loss": 0.32418695092201233,
+      "eval_runtime": 1389.9638,
+      "eval_samples_per_second": 4.484,
+      "eval_steps_per_second": 0.561,
+      "eval_wer": 0.3670192003866099,
+      "step": 8800
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 0.00013501963350785338,
+      "loss": 0.0768,
+      "step": 8810
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 0.00013482329842931935,
+      "loss": 0.1071,
+      "step": 8820
+    },
+    {
+      "epoch": 11.26,
+      "learning_rate": 0.00013462696335078534,
+      "loss": 0.227,
+      "step": 8830
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 0.0001344306282722513,
+      "loss": 0.1728,
+      "step": 8840
+    },
+    {
+      "epoch": 11.29,
+      "learning_rate": 0.00013423429319371727,
+      "loss": 0.1197,
+      "step": 8850
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 0.00013403795811518324,
+      "loss": 0.0868,
+      "step": 8860
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 0.0001338416230366492,
+      "loss": 0.1066,
+      "step": 8870
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 0.00013364528795811517,
+      "loss": 0.2165,
+      "step": 8880
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 0.00013344895287958113,
+      "loss": 0.156,
+      "step": 8890
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 0.0001332526178010471,
+      "loss": 0.1056,
+      "step": 8900
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 0.0001330562827225131,
+      "loss": 0.0933,
+      "step": 8910
+    },
+    {
+      "epoch": 11.38,
+      "learning_rate": 0.00013285994764397903,
+      "loss": 0.1038,
+      "step": 8920
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 0.00013266361256544502,
+      "loss": 0.1903,
+      "step": 8930
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 0.000132467277486911,
+      "loss": 0.1476,
+      "step": 8940
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 0.00013227094240837695,
+      "loss": 0.1059,
+      "step": 8950
+    },
+    {
+      "epoch": 11.43,
+      "learning_rate": 0.00013207460732984292,
+      "loss": 0.0914,
+      "step": 8960
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 0.00013187827225130889,
+      "loss": 0.0847,
+      "step": 8970
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 0.00013168193717277488,
+      "loss": 0.195,
+      "step": 8980
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 0.00013148560209424082,
+      "loss": 0.1543,
+      "step": 8990
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 0.00013128926701570678,
+      "loss": 0.1181,
+      "step": 9000
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 0.00013109293193717278,
+      "loss": 0.087,
+      "step": 9010
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 0.00013089659685863874,
+      "loss": 0.1129,
+      "step": 9020
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 0.00013070026178010468,
+      "loss": 0.2049,
+      "step": 9030
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 0.00013050392670157067,
+      "loss": 0.1522,
+      "step": 9040
+    },
+    {
+      "epoch": 11.54,
+      "learning_rate": 0.00013030759162303664,
+      "loss": 0.1233,
+      "step": 9050
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 0.0001301112565445026,
+      "loss": 0.0912,
+      "step": 9060
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 0.00012991492146596857,
+      "loss": 0.1359,
+      "step": 9070
+    },
+    {
+      "epoch": 11.58,
+      "learning_rate": 0.00012971858638743453,
+      "loss": 0.1982,
+      "step": 9080
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 0.00012952225130890053,
+      "loss": 0.1522,
+      "step": 9090
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 0.00012932591623036646,
+      "loss": 0.1168,
+      "step": 9100
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 0.00012912958115183246,
+      "loss": 0.1026,
+      "step": 9110
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 0.00012893324607329842,
+      "loss": 0.1052,
+      "step": 9120
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 0.0001287369109947644,
+      "loss": 0.1982,
+      "step": 9130
+    },
+    {
+      "epoch": 11.66,
+      "learning_rate": 0.00012854057591623035,
+      "loss": 0.1349,
+      "step": 9140
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 0.00012834424083769632,
+      "loss": 0.1233,
+      "step": 9150
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 0.0001281479057591623,
+      "loss": 0.0937,
+      "step": 9160
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 0.00012795157068062825,
+      "loss": 0.1107,
+      "step": 9170
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 0.00012775523560209422,
+      "loss": 0.2197,
+      "step": 9180
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 0.0001275589005235602,
+      "loss": 0.1378,
+      "step": 9190
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 0.00012736256544502617,
+      "loss": 0.108,
+      "step": 9200
+    },
+    {
+      "epoch": 11.73,
+      "eval_loss": 0.3200973868370056,
+      "eval_runtime": 1388.3182,
+      "eval_samples_per_second": 4.49,
+      "eval_steps_per_second": 0.562,
+      "eval_wer": 0.36013271128789426,
+      "step": 9200
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 0.00012716623036649214,
+      "loss": 0.0872,
+      "step": 9210
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 0.0001269698952879581,
+      "loss": 0.0889,
+      "step": 9220
+    },
+    {
+      "epoch": 11.77,
+      "learning_rate": 0.00012677356020942407,
+      "loss": 0.2129,
+      "step": 9230
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 0.00012657722513089004,
+      "loss": 0.155,
+      "step": 9240
+    },
+    {
+      "epoch": 11.8,
+      "learning_rate": 0.000126380890052356,
+      "loss": 0.1204,
+      "step": 9250
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 0.00012618455497382197,
+      "loss": 0.0844,
+      "step": 9260
+    },
+    {
+      "epoch": 11.82,
+      "learning_rate": 0.00012598821989528796,
+      "loss": 0.0904,
+      "step": 9270
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 0.00012579188481675393,
+      "loss": 0.2074,
+      "step": 9280
+    },
+    {
+      "epoch": 11.85,
+      "learning_rate": 0.0001255955497382199,
+      "loss": 0.141,
+      "step": 9290
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 0.00012539921465968586,
+      "loss": 0.1134,
+      "step": 9300
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 0.00012520287958115182,
+      "loss": 0.0848,
+      "step": 9310
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 0.0001250065445026178,
+      "loss": 0.0929,
+      "step": 9320
+    },
+    {
+      "epoch": 11.9,
+      "learning_rate": 0.00012481020942408375,
+      "loss": 0.2138,
+      "step": 9330
+    },
+    {
+      "epoch": 11.91,
+      "learning_rate": 0.00012461387434554972,
+      "loss": 0.1591,
+      "step": 9340
+    },
+    {
+      "epoch": 11.93,
+      "learning_rate": 0.0001244175392670157,
+      "loss": 0.1118,
+      "step": 9350
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 0.00012422120418848165,
+      "loss": 0.0923,
+      "step": 9360
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 0.00012402486910994764,
+      "loss": 0.0974,
+      "step": 9370
+    },
+    {
+      "epoch": 11.96,
+      "learning_rate": 0.0001238285340314136,
+      "loss": 0.203,
+      "step": 9380
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 0.00012363219895287957,
+      "loss": 0.1309,
+      "step": 9390
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 0.00012343586387434554,
+      "loss": 0.1082,
+      "step": 9400
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 0.0001232395287958115,
+      "loss": 0.1743,
+      "step": 9410
+    },
+    {
+      "epoch": 12.02,
+      "learning_rate": 0.00012304319371727747,
+      "loss": 0.21,
+      "step": 9420
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 0.00012284685863874344,
+      "loss": 0.3056,
+      "step": 9430
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 0.0001226505235602094,
+      "loss": 0.1672,
+      "step": 9440
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 0.0001224541884816754,
+      "loss": 0.1673,
+      "step": 9450
+    },
+    {
+      "epoch": 12.07,
+      "learning_rate": 0.00012225785340314136,
+      "loss": 0.2128,
+      "step": 9460
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 0.00012206151832460732,
+      "loss": 0.2115,
+      "step": 9470
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 0.00012186518324607329,
+      "loss": 0.1185,
+      "step": 9480
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 0.00012166884816753926,
+      "loss": 0.1009,
+      "step": 9490
+    },
+    {
+      "epoch": 12.12,
+      "learning_rate": 0.00012147251308900523,
+      "loss": 0.119,
+      "step": 9500
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 0.00012127617801047119,
+      "loss": 0.1616,
+      "step": 9510
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 0.00012107984293193715,
+      "loss": 0.2202,
+      "step": 9520
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 0.00012088350785340313,
+      "loss": 0.2025,
+      "step": 9530
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 0.0001206871727748691,
+      "loss": 0.1168,
+      "step": 9540
+    },
+    {
+      "epoch": 12.18,
+      "learning_rate": 0.00012049083769633508,
+      "loss": 0.1177,
+      "step": 9550
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 0.00012029450261780104,
+      "loss": 0.1575,
+      "step": 9560
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 0.00012009816753926699,
+      "loss": 0.1669,
+      "step": 9570
+    },
+    {
+      "epoch": 12.22,
+      "learning_rate": 0.00011990183246073297,
+      "loss": 0.1353,
+      "step": 9580
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 0.00011970549738219894,
+      "loss": 0.0869,
+      "step": 9590
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 0.00011950916230366492,
+      "loss": 0.0806,
+      "step": 9600
+    },
+    {
+      "epoch": 12.24,
+      "eval_loss": 0.3223347067832947,
+      "eval_runtime": 1383.3623,
+      "eval_samples_per_second": 4.506,
+      "eval_steps_per_second": 0.564,
+      "eval_wer": 0.36094124644523334,
+      "step": 9600
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 0.00011931282722513088,
+      "loss": 0.1702,
+      "step": 9610
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 0.00011911649214659685,
+      "loss": 0.1506,
+      "step": 9620
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 0.00011892015706806283,
+      "loss": 0.1315,
+      "step": 9630
+    },
+    {
+      "epoch": 12.3,
+      "learning_rate": 0.00011872382198952878,
+      "loss": 0.0871,
+      "step": 9640
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 0.00011852748691099476,
+      "loss": 0.0811,
+      "step": 9650
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 0.00011833115183246072,
+      "loss": 0.1449,
+      "step": 9660
+    },
+    {
+      "epoch": 12.33,
+      "learning_rate": 0.00011813481675392669,
+      "loss": 0.1488,
+      "step": 9670
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 0.00011793848167539267,
+      "loss": 0.0972,
+      "step": 9680
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 0.00011774214659685863,
+      "loss": 0.0929,
+      "step": 9690
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 0.00011754581151832459,
+      "loss": 0.0762,
+      "step": 9700
+    },
+    {
+      "epoch": 12.38,
+      "learning_rate": 0.00011734947643979056,
+      "loss": 0.1797,
+      "step": 9710
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 0.00011715314136125653,
+      "loss": 0.1486,
+      "step": 9720
+    },
+    {
+      "epoch": 12.41,
+      "learning_rate": 0.00011695680628272251,
+      "loss": 0.116,
+      "step": 9730
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 0.00011676047120418847,
+      "loss": 0.0983,
+      "step": 9740
+    },
+    {
+      "epoch": 12.44,
+      "learning_rate": 0.00011656413612565444,
+      "loss": 0.0728,
+      "step": 9750
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 0.0001163678010471204,
+      "loss": 0.1737,
+      "step": 9760
+    },
+    {
+      "epoch": 12.46,
+      "learning_rate": 0.00011617146596858637,
+      "loss": 0.1463,
+      "step": 9770
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 0.00011597513089005235,
+      "loss": 0.1184,
+      "step": 9780
+    },
+    {
+      "epoch": 12.49,
+      "learning_rate": 0.00011577879581151832,
+      "loss": 0.0799,
+      "step": 9790
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 0.00011558246073298428,
+      "loss": 0.0929,
+      "step": 9800
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 0.00011538612565445026,
+      "loss": 0.1463,
+      "step": 9810
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 0.00011518979057591621,
+      "loss": 0.1509,
+      "step": 9820
+    },
+    {
+      "epoch": 12.54,
+      "learning_rate": 0.00011499345549738218,
+      "loss": 0.1234,
+      "step": 9830
+    },
+    {
+      "epoch": 12.55,
+      "learning_rate": 0.00011479712041884816,
+      "loss": 0.0917,
+      "step": 9840
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 0.00011460078534031412,
+      "loss": 0.0729,
+      "step": 9850
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 0.0001144044502617801,
+      "loss": 0.1627,
+      "step": 9860
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 0.00011420811518324607,
+      "loss": 0.1522,
+      "step": 9870
+    },
+    {
+      "epoch": 12.6,
+      "learning_rate": 0.00011401178010471202,
+      "loss": 0.1176,
+      "step": 9880
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 0.000113815445026178,
+      "loss": 0.0855,
+      "step": 9890
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 0.00011361910994764396,
+      "loss": 0.0832,
+      "step": 9900
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 0.00011342277486910994,
+      "loss": 0.1292,
+      "step": 9910
+    },
+    {
+      "epoch": 12.65,
+      "learning_rate": 0.00011322643979057591,
+      "loss": 0.1349,
+      "step": 9920
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 0.00011303010471204187,
+      "loss": 0.1178,
+      "step": 9930
+    },
+    {
+      "epoch": 12.68,
+      "learning_rate": 0.00011283376963350785,
+      "loss": 0.096,
+      "step": 9940
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 0.0001126374345549738,
+      "loss": 0.076,
+      "step": 9950
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 0.00011244109947643978,
+      "loss": 0.1355,
+      "step": 9960
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 0.00011224476439790575,
+      "loss": 0.1383,
+      "step": 9970
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 0.00011204842931937171,
+      "loss": 0.1045,
+      "step": 9980
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 0.0001118520942408377,
+      "loss": 0.109,
+      "step": 9990
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 0.00011165575916230366,
+      "loss": 0.0896,
+      "step": 10000
+    },
+    {
+      "epoch": 12.75,
+      "eval_loss": 0.3228430449962616,
+      "eval_runtime": 1386.0119,
+      "eval_samples_per_second": 4.497,
+      "eval_steps_per_second": 0.563,
+      "eval_wer": 0.3584227058976599,
+      "step": 10000
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 0.00011145942408376961,
+      "loss": 0.1356,
+      "step": 10010
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 0.00011126308900523559,
+      "loss": 0.1624,
+      "step": 10020
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 0.00011106675392670156,
+      "loss": 0.1019,
+      "step": 10030
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 0.00011087041884816754,
+      "loss": 0.0846,
+      "step": 10040
+    },
+    {
+      "epoch": 12.82,
+      "learning_rate": 0.0001106740837696335,
+      "loss": 0.094,
+      "step": 10050
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 0.00011047774869109947,
+      "loss": 0.1393,
+      "step": 10060
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 0.00011028141361256545,
+      "loss": 0.1464,
+      "step": 10070
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 0.0001100850785340314,
+      "loss": 0.1014,
+      "step": 10080
+    },
+    {
+      "epoch": 12.87,
+      "learning_rate": 0.00010988874345549738,
+      "loss": 0.0803,
+      "step": 10090
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 0.00010969240837696334,
+      "loss": 0.0777,
+      "step": 10100
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 0.00010949607329842931,
+      "loss": 0.1408,
+      "step": 10110
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 0.00010929973821989529,
+      "loss": 0.1711,
+      "step": 10120
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 0.00010910340314136125,
+      "loss": 0.113,
+      "step": 10130
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 0.0001089070680628272,
+      "loss": 0.1114,
+      "step": 10140
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 0.00010871073298429318,
+      "loss": 0.1093,
+      "step": 10150
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 0.00010851439790575915,
+      "loss": 0.1327,
+      "step": 10160
+    },
+    {
+      "epoch": 12.97,
+      "learning_rate": 0.00010831806282722513,
+      "loss": 0.1347,
+      "step": 10170
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 0.00010812172774869109,
+      "loss": 0.1042,
+      "step": 10180
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 0.00010792539267015704,
+      "loss": 0.0947,
+      "step": 10190
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 0.00010772905759162302,
+      "loss": 0.16,
+      "step": 10200
+    },
+    {
+      "epoch": 13.02,
+      "learning_rate": 0.00010753272251308899,
+      "loss": 0.0962,
+      "step": 10210
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 0.00010733638743455497,
+      "loss": 0.0818,
+      "step": 10220
+    },
+    {
+      "epoch": 13.05,
+      "learning_rate": 0.00010714005235602093,
+      "loss": 0.0606,
+      "step": 10230
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 0.0001069437172774869,
+      "loss": 0.0851,
+      "step": 10240
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 0.00010674738219895288,
+      "loss": 0.1352,
+      "step": 10250
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 0.00010655104712041883,
+      "loss": 0.1302,
+      "step": 10260
+    },
+    {
+      "epoch": 13.1,
+      "learning_rate": 0.00010635471204188481,
+      "loss": 0.0773,
+      "step": 10270
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 0.00010615837696335078,
+      "loss": 0.0693,
+      "step": 10280
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 0.00010596204188481674,
+      "loss": 0.09,
+      "step": 10290
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 0.00010576570680628272,
+      "loss": 0.1407,
+      "step": 10300
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 0.00010556937172774869,
+      "loss": 0.0946,
+      "step": 10310
+    },
+    {
+      "epoch": 13.16,
+      "learning_rate": 0.00010537303664921464,
+      "loss": 0.1019,
+      "step": 10320
+    },
+    {
+      "epoch": 13.18,
+      "learning_rate": 0.00010517670157068062,
+      "loss": 0.0695,
+      "step": 10330
+    },
+    {
+      "epoch": 13.19,
+      "learning_rate": 0.00010498036649214658,
+      "loss": 0.0646,
+      "step": 10340
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 0.00010478403141361256,
+      "loss": 0.172,
+      "step": 10350
+    },
+    {
+      "epoch": 13.21,
+      "learning_rate": 0.00010458769633507853,
+      "loss": 0.124,
+      "step": 10360
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 0.00010439136125654449,
+      "loss": 0.0795,
+      "step": 10370
+    },
+    {
+      "epoch": 13.24,
+      "learning_rate": 0.00010419502617801047,
+      "loss": 0.0558,
+      "step": 10380
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 0.00010399869109947642,
+      "loss": 0.1036,
+      "step": 10390
+    },
+    {
+      "epoch": 13.27,
+      "learning_rate": 0.0001038023560209424,
+      "loss": 0.1642,
+      "step": 10400
+    },
+    {
+      "epoch": 13.27,
+      "eval_loss": 0.31404751539230347,
+      "eval_runtime": 1383.9286,
+      "eval_samples_per_second": 4.504,
+      "eval_steps_per_second": 0.564,
+      "eval_wer": 0.35470530287541124,
+      "step": 10400
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 0.00010360602094240837,
+      "loss": 0.0981,
+      "step": 10410
+    },
+    {
+      "epoch": 13.29,
+      "learning_rate": 0.00010340968586387433,
+      "loss": 0.0742,
+      "step": 10420
+    },
+    {
+      "epoch": 13.3,
+      "learning_rate": 0.00010321335078534031,
+      "loss": 0.0704,
+      "step": 10430
+    },
+    {
+      "epoch": 13.32,
+      "learning_rate": 0.00010301701570680628,
+      "loss": 0.0831,
+      "step": 10440
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 0.00010282068062827226,
+      "loss": 0.1788,
+      "step": 10450
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 0.00010262434554973821,
+      "loss": 0.0991,
+      "step": 10460
+    },
+    {
+      "epoch": 13.35,
+      "learning_rate": 0.00010242801047120417,
+      "loss": 0.0738,
+      "step": 10470
+    },
+    {
+      "epoch": 13.37,
+      "learning_rate": 0.00010223167539267015,
+      "loss": 0.0683,
+      "step": 10480
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 0.00010203534031413612,
+      "loss": 0.092,
+      "step": 10490
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 0.00010183900523560208,
+      "loss": 0.1525,
+      "step": 10500
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 0.00010164267015706806,
+      "loss": 0.0986,
+      "step": 10510
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 0.00010144633507853402,
+      "loss": 0.0983,
+      "step": 10520
+    },
+    {
+      "epoch": 13.43,
+      "learning_rate": 0.00010125,
+      "loss": 0.0717,
+      "step": 10530
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 0.00010105366492146596,
+      "loss": 0.1029,
+      "step": 10540
+    },
+    {
+      "epoch": 13.46,
+      "learning_rate": 0.00010085732984293193,
+      "loss": 0.1707,
+      "step": 10550
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 0.0001006609947643979,
+      "loss": 0.1133,
+      "step": 10560
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 0.00010046465968586386,
+      "loss": 0.0857,
+      "step": 10570
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 0.00010026832460732984,
+      "loss": 0.0829,
+      "step": 10580
+    },
+    {
+      "epoch": 13.51,
+      "learning_rate": 0.0001000719895287958,
+      "loss": 0.0756,
+      "step": 10590
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 9.987565445026177e-05,
+      "loss": 0.1354,
+      "step": 10600
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 9.967931937172775e-05,
+      "loss": 0.1232,
+      "step": 10610
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 9.948298429319371e-05,
+      "loss": 0.0701,
+      "step": 10620
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 9.928664921465966e-05,
+      "loss": 0.0669,
+      "step": 10630
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 9.909031413612564e-05,
+      "loss": 0.0764,
+      "step": 10640
+    },
+    {
+      "epoch": 13.58,
+      "learning_rate": 9.889397905759161e-05,
+      "loss": 0.1572,
+      "step": 10650
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 9.869764397905759e-05,
+      "loss": 0.1071,
+      "step": 10660
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 9.850130890052355e-05,
+      "loss": 0.0868,
+      "step": 10670
+    },
+    {
+      "epoch": 13.62,
+      "learning_rate": 9.830497382198952e-05,
+      "loss": 0.0591,
+      "step": 10680
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 9.81086387434555e-05,
+      "loss": 0.0771,
+      "step": 10690
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 9.791230366492145e-05,
+      "loss": 0.1548,
+      "step": 10700
+    },
+    {
+      "epoch": 13.66,
+      "learning_rate": 9.771596858638743e-05,
+      "loss": 0.1099,
+      "step": 10710
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.75196335078534e-05,
+      "loss": 0.0738,
+      "step": 10720
+    },
+    {
+      "epoch": 13.69,
+      "learning_rate": 9.732329842931936e-05,
+      "loss": 0.0612,
+      "step": 10730
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.712696335078534e-05,
+      "loss": 0.0785,
+      "step": 10740
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.69306282722513e-05,
+      "loss": 0.1613,
+      "step": 10750
+    },
+    {
+      "epoch": 13.72,
+      "learning_rate": 9.673429319371728e-05,
+      "loss": 0.1243,
+      "step": 10760
+    },
+    {
+      "epoch": 13.74,
+      "learning_rate": 9.653795811518323e-05,
+      "loss": 0.0786,
+      "step": 10770
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.63416230366492e-05,
+      "loss": 0.0768,
+      "step": 10780
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.614528795811518e-05,
+      "loss": 0.0827,
+      "step": 10790
+    },
+    {
+      "epoch": 13.77,
+      "learning_rate": 9.596858638743454e-05,
+      "loss": 0.1442,
+      "step": 10800
+    },
+    {
+      "epoch": 13.77,
+      "eval_loss": 0.32345089316368103,
+      "eval_runtime": 1390.3928,
+      "eval_samples_per_second": 4.483,
+      "eval_steps_per_second": 0.561,
+      "eval_wer": 0.3538781807029609,
+      "step": 10800
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.577225130890052e-05,
+      "loss": 0.0933,
+      "step": 10810
+    },
+    {
+      "epoch": 13.8,
+      "learning_rate": 9.557591623036648e-05,
+      "loss": 0.0823,
+      "step": 10820
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.537958115183245e-05,
+      "loss": 0.0629,
+      "step": 10830
+    },
+    {
+      "epoch": 13.83,
+      "learning_rate": 9.518324607329843e-05,
+      "loss": 0.0832,
+      "step": 10840
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.498691099476438e-05,
+      "loss": 0.1491,
+      "step": 10850
+    },
+    {
+      "epoch": 13.85,
+      "learning_rate": 9.479057591623036e-05,
+      "loss": 0.1078,
+      "step": 10860
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.459424083769633e-05,
+      "loss": 0.0803,
+      "step": 10870
+    },
+    {
+      "epoch": 13.88,
+      "learning_rate": 9.439790575916229e-05,
+      "loss": 0.0586,
+      "step": 10880
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.420157068062827e-05,
+      "loss": 0.0819,
+      "step": 10890
+    },
+    {
+      "epoch": 13.9,
+      "learning_rate": 9.400523560209424e-05,
+      "loss": 0.1389,
+      "step": 10900
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.380890052356019e-05,
+      "loss": 0.0924,
+      "step": 10910
+    },
+    {
+      "epoch": 13.93,
+      "learning_rate": 9.361256544502617e-05,
+      "loss": 0.0742,
+      "step": 10920
+    },
+    {
+      "epoch": 13.94,
+      "learning_rate": 9.341623036649213e-05,
+      "loss": 0.0649,
+      "step": 10930
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.321989528795811e-05,
+      "loss": 0.0649,
+      "step": 10940
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.302356020942408e-05,
+      "loss": 0.1502,
+      "step": 10950
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.282722513089004e-05,
+      "loss": 0.1182,
+      "step": 10960
+    },
+    {
+      "epoch": 13.99,
+      "learning_rate": 9.263089005235602e-05,
+      "loss": 0.0694,
+      "step": 10970
+    },
+    {
+      "epoch": 14.01,
+      "learning_rate": 9.243455497382197e-05,
+      "loss": 0.1236,
+      "step": 10980
+    },
+    {
+      "epoch": 14.02,
+      "learning_rate": 9.223821989528795e-05,
+      "loss": 0.1018,
+      "step": 10990
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.204188481675392e-05,
+      "loss": 0.0713,
+      "step": 11000
+    },
+    {
+      "epoch": 14.04,
+      "learning_rate": 9.184554973821988e-05,
+      "loss": 0.0553,
+      "step": 11010
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.164921465968586e-05,
+      "loss": 0.0649,
+      "step": 11020
+    },
+    {
+      "epoch": 14.07,
+      "learning_rate": 9.145287958115183e-05,
+      "loss": 0.1469,
+      "step": 11030
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.125654450261781e-05,
+      "loss": 0.097,
+      "step": 11040
+    },
+    {
+      "epoch": 14.09,
+      "learning_rate": 9.106020942408376e-05,
+      "loss": 0.0752,
+      "step": 11050
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.086387434554972e-05,
+      "loss": 0.0726,
+      "step": 11060
+    },
+    {
+      "epoch": 14.12,
+      "learning_rate": 9.06675392670157e-05,
+      "loss": 0.06,
+      "step": 11070
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.047120418848167e-05,
+      "loss": 0.1317,
+      "step": 11080
+    },
+    {
+      "epoch": 14.15,
+      "learning_rate": 9.027486910994763e-05,
+      "loss": 0.111,
+      "step": 11090
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.00785340314136e-05,
+      "loss": 0.0702,
+      "step": 11100
+    },
+    {
+      "epoch": 14.17,
+      "learning_rate": 8.988219895287957e-05,
+      "loss": 0.0717,
+      "step": 11110
+    },
+    {
+      "epoch": 14.18,
+      "learning_rate": 8.968586387434555e-05,
+      "loss": 0.0589,
+      "step": 11120
+    },
+    {
+      "epoch": 14.2,
+      "learning_rate": 8.948952879581151e-05,
+      "loss": 0.1209,
+      "step": 11130
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 8.929319371727748e-05,
+      "loss": 0.1063,
+      "step": 11140
+    },
+    {
+      "epoch": 14.22,
+      "learning_rate": 8.909685863874346e-05,
+      "loss": 0.0738,
+      "step": 11150
+    },
+    {
+      "epoch": 14.23,
+      "learning_rate": 8.890052356020941e-05,
+      "loss": 0.0619,
+      "step": 11160
+    },
+    {
+      "epoch": 14.25,
+      "learning_rate": 8.870418848167539e-05,
+      "loss": 0.0791,
+      "step": 11170
+    },
+    {
+      "epoch": 14.26,
+      "learning_rate": 8.850785340314135e-05,
+      "loss": 0.1179,
+      "step": 11180
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 8.831151832460732e-05,
+      "loss": 0.1212,
+      "step": 11190
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 8.81151832460733e-05,
+      "loss": 0.0802,
+      "step": 11200
+    },
+    {
+      "epoch": 14.29,
+      "eval_loss": 0.31754833459854126,
+      "eval_runtime": 1384.3252,
+      "eval_samples_per_second": 4.503,
+      "eval_steps_per_second": 0.563,
+      "eval_wer": 0.355300087358971,
+      "step": 11200
+    },
+    {
+      "epoch": 14.3,
+      "learning_rate": 8.791884816753926e-05,
+      "loss": 0.0919,
+      "step": 11210
+    },
+    {
+      "epoch": 14.31,
+      "learning_rate": 8.772251308900521e-05,
+      "loss": 0.0681,
+      "step": 11220
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 8.752617801047119e-05,
+      "loss": 0.117,
+      "step": 11230
+    },
+    {
+      "epoch": 14.34,
+      "learning_rate": 8.732984293193716e-05,
+      "loss": 0.0968,
+      "step": 11240
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 8.713350785340314e-05,
+      "loss": 0.0739,
+      "step": 11250
+    },
+    {
+      "epoch": 14.36,
+      "learning_rate": 8.69371727748691e-05,
+      "loss": 0.0634,
+      "step": 11260
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 8.674083769633507e-05,
+      "loss": 0.0747,
+      "step": 11270
+    },
+    {
+      "epoch": 14.39,
+      "learning_rate": 8.654450261780105e-05,
+      "loss": 0.1164,
+      "step": 11280
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 8.6348167539267e-05,
+      "loss": 0.1129,
+      "step": 11290
+    },
+    {
+      "epoch": 14.41,
+      "learning_rate": 8.615183246073298e-05,
+      "loss": 0.0753,
+      "step": 11300
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 8.595549738219894e-05,
+      "loss": 0.0701,
+      "step": 11310
+    },
+    {
+      "epoch": 14.44,
+      "learning_rate": 8.575916230366491e-05,
+      "loss": 0.052,
+      "step": 11320
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 8.556282722513089e-05,
+      "loss": 0.1002,
+      "step": 11330
+    },
+    {
+      "epoch": 14.46,
+      "learning_rate": 8.536649214659685e-05,
+      "loss": 0.0917,
+      "step": 11340
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 8.517015706806283e-05,
+      "loss": 0.0726,
+      "step": 11350
+    },
+    {
+      "epoch": 14.49,
+      "learning_rate": 8.497382198952879e-05,
+      "loss": 0.067,
+      "step": 11360
+    },
+    {
+      "epoch": 14.5,
+      "learning_rate": 8.477748691099475e-05,
+      "loss": 0.0645,
+      "step": 11370
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 8.458115183246073e-05,
+      "loss": 0.1364,
+      "step": 11380
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 8.43848167539267e-05,
+      "loss": 0.0968,
+      "step": 11390
+    },
+    {
+      "epoch": 14.54,
+      "learning_rate": 8.418848167539266e-05,
+      "loss": 0.0748,
+      "step": 11400
+    },
+    {
+      "epoch": 14.55,
+      "learning_rate": 8.399214659685864e-05,
+      "loss": 0.0618,
+      "step": 11410
+    },
+    {
+      "epoch": 14.57,
+      "learning_rate": 8.379581151832459e-05,
+      "loss": 0.0567,
+      "step": 11420
+    },
+    {
+      "epoch": 14.58,
+      "learning_rate": 8.359947643979057e-05,
+      "loss": 0.1112,
+      "step": 11430
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 8.340314136125654e-05,
+      "loss": 0.0968,
+      "step": 11440
+    },
+    {
+      "epoch": 14.6,
+      "learning_rate": 8.32068062827225e-05,
+      "loss": 0.0884,
+      "step": 11450
+    },
+    {
+      "epoch": 14.62,
+      "learning_rate": 8.301047120418848e-05,
+      "loss": 0.0634,
+      "step": 11460
+    },
+    {
+      "epoch": 14.63,
+      "learning_rate": 8.281413612565445e-05,
+      "loss": 0.0711,
+      "step": 11470
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 8.261780104712041e-05,
+      "loss": 0.1249,
+      "step": 11480
+    },
+    {
+      "epoch": 14.66,
+      "learning_rate": 8.242146596858638e-05,
+      "loss": 0.0923,
+      "step": 11490
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 8.222513089005234e-05,
+      "loss": 0.0704,
+      "step": 11500
+    },
+    {
+      "epoch": 14.68,
+      "learning_rate": 8.202879581151832e-05,
+      "loss": 0.056,
+      "step": 11510
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 8.183246073298429e-05,
+      "loss": 0.071,
+      "step": 11520
+    },
+    {
+      "epoch": 14.71,
+      "learning_rate": 8.163612565445027e-05,
+      "loss": 0.1171,
+      "step": 11530
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 8.143979057591622e-05,
+      "loss": 0.0949,
+      "step": 11540
+    },
+    {
+      "epoch": 14.73,
+      "learning_rate": 8.124345549738218e-05,
+      "loss": 0.0845,
+      "step": 11550
+    },
+    {
+      "epoch": 14.74,
+      "learning_rate": 8.104712041884816e-05,
+      "loss": 0.0607,
+      "step": 11560
+    },
+    {
+      "epoch": 14.76,
+      "learning_rate": 8.085078534031413e-05,
+      "loss": 0.0529,
+      "step": 11570
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 8.06544502617801e-05,
+      "loss": 0.1006,
+      "step": 11580
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 8.045811518324607e-05,
+      "loss": 0.1075,
+      "step": 11590
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 8.026178010471203e-05,
+      "loss": 0.0747,
+      "step": 11600
+    },
+    {
+      "epoch": 14.8,
+      "eval_loss": 0.3126291334629059,
+      "eval_runtime": 1382.4429,
+      "eval_samples_per_second": 4.509,
+      "eval_steps_per_second": 0.564,
+      "eval_wer": 0.35120165052694186,
+      "step": 11600
+    },
+    {
+      "epoch": 14.81,
+      "learning_rate": 8.0065445026178e-05,
+      "loss": 0.0557,
+      "step": 11610
+    },
+    {
+      "epoch": 14.82,
+      "learning_rate": 7.986910994764397e-05,
+      "loss": 0.0556,
+      "step": 11620
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 7.967277486910994e-05,
+      "loss": 0.1105,
+      "step": 11630
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 7.947643979057591e-05,
+      "loss": 0.1135,
+      "step": 11640
+    },
+    {
+      "epoch": 14.86,
+      "learning_rate": 7.928010471204188e-05,
+      "loss": 0.0807,
+      "step": 11650
+    },
+    {
+      "epoch": 14.87,
+      "learning_rate": 7.908376963350786e-05,
+      "loss": 0.0482,
+      "step": 11660
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 7.888743455497381e-05,
+      "loss": 0.0626,
+      "step": 11670
+    },
+    {
+      "epoch": 14.9,
+      "learning_rate": 7.869109947643978e-05,
+      "loss": 0.1127,
+      "step": 11680
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 7.849476439790576e-05,
+      "loss": 0.1054,
+      "step": 11690
+    },
+    {
+      "epoch": 14.92,
+      "learning_rate": 7.829842931937172e-05,
+      "loss": 0.066,
+      "step": 11700
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 7.810209424083769e-05,
+      "loss": 0.0679,
+      "step": 11710
+    },
+    {
+      "epoch": 14.95,
+      "learning_rate": 7.790575916230367e-05,
+      "loss": 0.0637,
+      "step": 11720
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 7.770942408376962e-05,
+      "loss": 0.1381,
+      "step": 11730
+    },
+    {
+      "epoch": 14.97,
+      "learning_rate": 7.75130890052356e-05,
+      "loss": 0.0937,
+      "step": 11740
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 7.731675392670156e-05,
+      "loss": 0.0717,
+      "step": 11750
+    },
+    {
+      "epoch": 15.0,
+      "learning_rate": 7.712041884816753e-05,
+      "loss": 0.095,
+      "step": 11760
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 7.692408376963351e-05,
+      "loss": 0.1343,
+      "step": 11770
+    },
+    {
+      "epoch": 15.03,
+      "learning_rate": 7.672774869109947e-05,
+      "loss": 0.0884,
+      "step": 11780
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 7.653141361256545e-05,
+      "loss": 0.0723,
+      "step": 11790
+    },
+    {
+      "epoch": 15.05,
+      "learning_rate": 7.63350785340314e-05,
+      "loss": 0.0467,
+      "step": 11800
+    },
+    {
+      "epoch": 15.06,
+      "learning_rate": 7.613874345549737e-05,
+      "loss": 0.0778,
+      "step": 11810
+    },
+    {
+      "epoch": 15.08,
+      "learning_rate": 7.594240837696335e-05,
+      "loss": 0.1112,
+      "step": 11820
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 7.574607329842931e-05,
+      "loss": 0.0688,
+      "step": 11830
+    },
+    {
+      "epoch": 15.1,
+      "learning_rate": 7.554973821989529e-05,
+      "loss": 0.0585,
+      "step": 11840
+    },
+    {
+      "epoch": 15.11,
+      "learning_rate": 7.535340314136126e-05,
+      "loss": 0.0498,
+      "step": 11850
+    },
+    {
+      "epoch": 15.13,
+      "learning_rate": 7.515706806282721e-05,
+      "loss": 0.0813,
+      "step": 11860
+    },
+    {
+      "epoch": 15.14,
+      "learning_rate": 7.496073298429319e-05,
+      "loss": 0.1015,
+      "step": 11870
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 7.476439790575915e-05,
+      "loss": 0.069,
+      "step": 11880
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 7.456806282722512e-05,
+      "loss": 0.0532,
+      "step": 11890
+    },
+    {
+      "epoch": 15.18,
+      "learning_rate": 7.43717277486911e-05,
+      "loss": 0.0622,
+      "step": 11900
+    },
+    {
+      "epoch": 15.19,
+      "learning_rate": 7.417539267015705e-05,
+      "loss": 0.0698,
+      "step": 11910
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 7.397905759162303e-05,
+      "loss": 0.1128,
+      "step": 11920
+    },
+    {
+      "epoch": 15.22,
+      "learning_rate": 7.3782722513089e-05,
+      "loss": 0.0657,
+      "step": 11930
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 7.358638743455497e-05,
+      "loss": 0.0584,
+      "step": 11940
+    },
+    {
+      "epoch": 15.24,
+      "learning_rate": 7.339005235602094e-05,
+      "loss": 0.0561,
+      "step": 11950
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 7.31937172774869e-05,
+      "loss": 0.0761,
+      "step": 11960
+    },
+    {
+      "epoch": 15.27,
+      "learning_rate": 7.299738219895287e-05,
+      "loss": 0.1219,
+      "step": 11970
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 7.280104712041884e-05,
+      "loss": 0.0777,
+      "step": 11980
+    },
+    {
+      "epoch": 15.29,
+      "learning_rate": 7.260471204188482e-05,
+      "loss": 0.0567,
+      "step": 11990
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 7.240837696335078e-05,
+      "loss": 0.0488,
+      "step": 12000
+    },
+    {
+      "epoch": 15.31,
+      "eval_loss": 0.3291836977005005,
+      "eval_runtime": 1380.3865,
+      "eval_samples_per_second": 4.515,
+      "eval_steps_per_second": 0.565,
+      "eval_wer": 0.3524655675545064,
+      "step": 12000
+    },
+    {
+      "epoch": 15.32,
+      "learning_rate": 7.221204188481675e-05,
+      "loss": 0.0753,
+      "step": 12010
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 7.201570680628271e-05,
+      "loss": 0.1031,
+      "step": 12020
+    },
+    {
+      "epoch": 15.34,
+      "learning_rate": 7.181937172774869e-05,
+      "loss": 0.0711,
+      "step": 12030
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 7.162303664921466e-05,
+      "loss": 0.0655,
+      "step": 12040
+    },
+    {
+      "epoch": 15.37,
+      "learning_rate": 7.142670157068062e-05,
+      "loss": 0.0674,
+      "step": 12050
+    },
+    {
+      "epoch": 15.38,
+      "learning_rate": 7.123036649214659e-05,
+      "loss": 0.088,
+      "step": 12060
+    },
+    {
+      "epoch": 15.4,
+      "learning_rate": 7.103403141361255e-05,
+      "loss": 0.1097,
+      "step": 12070
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 7.083769633507853e-05,
+      "loss": 0.0689,
+      "step": 12080
+    },
+    {
+      "epoch": 15.42,
+      "learning_rate": 7.06413612565445e-05,
+      "loss": 0.0644,
+      "step": 12090
+    },
+    {
+      "epoch": 15.43,
+      "learning_rate": 7.044502617801046e-05,
+      "loss": 0.0761,
+      "step": 12100
+    },
+    {
+      "epoch": 15.45,
+      "learning_rate": 7.024869109947643e-05,
+      "loss": 0.0737,
+      "step": 12110
+    },
+    {
+      "epoch": 15.46,
+      "learning_rate": 7.005235602094241e-05,
+      "loss": 0.1212,
+      "step": 12120
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 6.985602094240837e-05,
+      "loss": 0.0697,
+      "step": 12130
+    },
+    {
+      "epoch": 15.48,
+      "learning_rate": 6.965968586387434e-05,
+      "loss": 0.0638,
+      "step": 12140
+    },
+    {
+      "epoch": 15.5,
+      "learning_rate": 6.94633507853403e-05,
+      "loss": 0.0528,
+      "step": 12150
+    },
+    {
+      "epoch": 15.51,
+      "learning_rate": 6.926701570680628e-05,
+      "loss": 0.0708,
+      "step": 12160
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 6.907068062827225e-05,
+      "loss": 0.1115,
+      "step": 12170
+    },
+    {
+      "epoch": 15.54,
+      "learning_rate": 6.887434554973822e-05,
+      "loss": 0.0655,
+      "step": 12180
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 6.867801047120418e-05,
+      "loss": 0.0758,
+      "step": 12190
+    },
+    {
+      "epoch": 15.56,
+      "learning_rate": 6.848167539267015e-05,
+      "loss": 0.0523,
+      "step": 12200
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 6.828534031413613e-05,
+      "loss": 0.0895,
+      "step": 12210
+    },
+    {
+      "epoch": 15.59,
+      "learning_rate": 6.808900523560209e-05,
+      "loss": 0.1052,
+      "step": 12220
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 6.789267015706806e-05,
+      "loss": 0.0912,
+      "step": 12230
+    },
+    {
+      "epoch": 15.61,
+      "learning_rate": 6.769633507853402e-05,
+      "loss": 0.0766,
+      "step": 12240
+    },
+    {
+      "epoch": 15.62,
+      "learning_rate": 6.75e-05,
+      "loss": 0.0611,
+      "step": 12250
+    },
+    {
+      "epoch": 15.64,
+      "learning_rate": 6.730366492146597e-05,
+      "loss": 0.0664,
+      "step": 12260
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 6.710732984293193e-05,
+      "loss": 0.1003,
+      "step": 12270
+    },
+    {
+      "epoch": 15.66,
+      "learning_rate": 6.69109947643979e-05,
+      "loss": 0.0887,
+      "step": 12280
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 6.671465968586386e-05,
+      "loss": 0.0607,
+      "step": 12290
+    },
+    {
+      "epoch": 15.69,
+      "learning_rate": 6.651832460732984e-05,
+      "loss": 0.0469,
+      "step": 12300
+    },
+    {
+      "epoch": 15.7,
+      "learning_rate": 6.632198952879581e-05,
+      "loss": 0.0721,
+      "step": 12310
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 6.612565445026177e-05,
+      "loss": 0.1212,
+      "step": 12320
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 6.592931937172774e-05,
+      "loss": 0.0689,
+      "step": 12330
+    },
+    {
+      "epoch": 15.74,
+      "learning_rate": 6.573298429319372e-05,
+      "loss": 0.0611,
+      "step": 12340
+    },
+    {
+      "epoch": 15.75,
+      "learning_rate": 6.553664921465968e-05,
+      "loss": 0.0441,
+      "step": 12350
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 6.534031413612565e-05,
+      "loss": 0.0695,
+      "step": 12360
+    },
+    {
+      "epoch": 15.78,
+      "learning_rate": 6.514397905759161e-05,
+      "loss": 0.1082,
+      "step": 12370
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 6.49476439790576e-05,
+      "loss": 0.0793,
+      "step": 12380
+    },
+    {
+      "epoch": 15.8,
+      "learning_rate": 6.475130890052356e-05,
+      "loss": 0.0611,
+      "step": 12390
+    },
+    {
+      "epoch": 15.82,
+      "learning_rate": 6.455497382198952e-05,
+      "loss": 0.0469,
+      "step": 12400
+    },
+    {
+      "epoch": 15.82,
+      "eval_loss": 0.32305529713630676,
+      "eval_runtime": 1376.8764,
+      "eval_samples_per_second": 4.527,
+      "eval_steps_per_second": 0.566,
+      "eval_wer": 0.35036523484693594,
+      "step": 12400
+    },
+    {
+      "epoch": 15.83,
+      "learning_rate": 6.435863874345549e-05,
+      "loss": 0.0671,
+      "step": 12410
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 6.416230366492146e-05,
+      "loss": 0.1084,
+      "step": 12420
+    },
+    {
+      "epoch": 15.85,
+      "learning_rate": 6.396596858638743e-05,
+      "loss": 0.0739,
+      "step": 12430
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 6.37696335078534e-05,
+      "loss": 0.0724,
+      "step": 12440
+    },
+    {
+      "epoch": 15.88,
+      "learning_rate": 6.357329842931937e-05,
+      "loss": 0.0714,
+      "step": 12450
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 6.337696335078533e-05,
+      "loss": 0.0995,
+      "step": 12460
+    },
+    {
+      "epoch": 15.9,
+      "learning_rate": 6.318062827225131e-05,
+      "loss": 0.1095,
+      "step": 12470
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 6.298429319371728e-05,
+      "loss": 0.0719,
+      "step": 12480
+    },
+    {
+      "epoch": 15.93,
+      "learning_rate": 6.278795811518324e-05,
+      "loss": 0.0615,
+      "step": 12490
+    },
+    {
+      "epoch": 15.94,
+      "learning_rate": 6.25916230366492e-05,
+      "loss": 0.0689,
+      "step": 12500
+    },
+    {
+      "epoch": 15.96,
+      "learning_rate": 6.239528795811517e-05,
+      "loss": 0.0874,
+      "step": 12510
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 6.219895287958115e-05,
+      "loss": 0.0918,
+      "step": 12520
+    },
+    {
+      "epoch": 15.98,
+      "learning_rate": 6.200261780104712e-05,
+      "loss": 0.0703,
+      "step": 12530
+    },
+    {
+      "epoch": 15.99,
+      "learning_rate": 6.180628272251308e-05,
+      "loss": 0.0531,
+      "step": 12540
+    },
+    {
+      "epoch": 16.01,
+      "learning_rate": 6.160994764397905e-05,
+      "loss": 0.0944,
+      "step": 12550
+    },
+    {
+      "epoch": 16.02,
+      "learning_rate": 6.141361256544503e-05,
+      "loss": 0.0799,
+      "step": 12560
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 6.121727748691099e-05,
+      "loss": 0.0541,
+      "step": 12570
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 6.102094240837696e-05,
+      "loss": 0.07,
+      "step": 12580
+    },
+    {
+      "epoch": 16.06,
+      "learning_rate": 6.082460732984292e-05,
+      "loss": 0.0602,
+      "step": 12590
+    },
+    {
+      "epoch": 16.07,
+      "learning_rate": 6.0628272251308896e-05,
+      "loss": 0.0927,
+      "step": 12600
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 6.043193717277487e-05,
+      "loss": 0.0806,
+      "step": 12610
+    },
+    {
+      "epoch": 16.1,
+      "learning_rate": 6.023560209424083e-05,
+      "loss": 0.0605,
+      "step": 12620
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 6.00392670157068e-05,
+      "loss": 0.0426,
+      "step": 12630
+    },
+    {
+      "epoch": 16.12,
+      "learning_rate": 5.984293193717277e-05,
+      "loss": 0.0585,
+      "step": 12640
+    },
+    {
+      "epoch": 16.14,
+      "learning_rate": 5.964659685863874e-05,
+      "loss": 0.1018,
+      "step": 12650
+    },
+    {
+      "epoch": 16.15,
+      "learning_rate": 5.945026178010471e-05,
+      "loss": 0.0636,
+      "step": 12660
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 5.9253926701570675e-05,
+      "loss": 0.0498,
+      "step": 12670
+    },
+    {
+      "epoch": 16.17,
+      "learning_rate": 5.905759162303664e-05,
+      "loss": 0.0659,
+      "step": 12680
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 5.886125654450261e-05,
+      "loss": 0.0471,
+      "step": 12690
+    },
+    {
+      "epoch": 16.2,
+      "learning_rate": 5.8664921465968585e-05,
+      "loss": 0.1036,
+      "step": 12700
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 5.8468586387434543e-05,
+      "loss": 0.0713,
+      "step": 12710
+    },
+    {
+      "epoch": 16.22,
+      "learning_rate": 5.8272251308900516e-05,
+      "loss": 0.0553,
+      "step": 12720
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 5.807591623036649e-05,
+      "loss": 0.0431,
+      "step": 12730
+    },
+    {
+      "epoch": 16.25,
+      "learning_rate": 5.787958115183246e-05,
+      "loss": 0.0559,
+      "step": 12740
+    },
+    {
+      "epoch": 16.26,
+      "learning_rate": 5.7683246073298426e-05,
+      "loss": 0.0906,
+      "step": 12750
+    },
+    {
+      "epoch": 16.28,
+      "learning_rate": 5.748691099476439e-05,
+      "loss": 0.0894,
+      "step": 12760
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 5.7290575916230364e-05,
+      "loss": 0.0671,
+      "step": 12770
+    },
+    {
+      "epoch": 16.3,
+      "learning_rate": 5.709424083769633e-05,
+      "loss": 0.0477,
+      "step": 12780
+    },
+    {
+      "epoch": 16.31,
+      "learning_rate": 5.68979057591623e-05,
+      "loss": 0.0642,
+      "step": 12790
+    },
+    {
+      "epoch": 16.33,
+      "learning_rate": 5.670157068062827e-05,
+      "loss": 0.1021,
+      "step": 12800
+    },
+    {
+      "epoch": 16.33,
+      "eval_loss": 0.3229643702507019,
+      "eval_runtime": 1382.9575,
+      "eval_samples_per_second": 4.507,
+      "eval_steps_per_second": 0.564,
+      "eval_wer": 0.3502351257411572,
+      "step": 12800
+    },
+    {
+      "epoch": 16.34,
+      "learning_rate": 5.650523560209423e-05,
+      "loss": 0.072,
+      "step": 12810
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 5.6308900523560205e-05,
+      "loss": 0.0676,
+      "step": 12820
+    },
+    {
+      "epoch": 16.36,
+      "learning_rate": 5.611256544502618e-05,
+      "loss": 0.0421,
+      "step": 12830
+    },
+    {
+      "epoch": 16.38,
+      "learning_rate": 5.591623036649214e-05,
+      "loss": 0.067,
+      "step": 12840
+    },
+    {
+      "epoch": 16.39,
+      "learning_rate": 5.571989528795811e-05,
+      "loss": 0.0945,
+      "step": 12850
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 5.552356020942408e-05,
+      "loss": 0.0791,
+      "step": 12860
+    },
+    {
+      "epoch": 16.42,
+      "learning_rate": 5.5327225130890046e-05,
+      "loss": 0.0621,
+      "step": 12870
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 5.513089005235602e-05,
+      "loss": 0.0425,
+      "step": 12880
+    },
+    {
+      "epoch": 16.44,
+      "learning_rate": 5.4934554973821984e-05,
+      "loss": 0.0467,
+      "step": 12890
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 5.473821989528795e-05,
+      "loss": 0.0963,
+      "step": 12900
+    },
+    {
+      "epoch": 16.47,
+      "learning_rate": 5.454188481675392e-05,
+      "loss": 0.0733,
+      "step": 12910
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 5.4345549738219894e-05,
+      "loss": 0.0639,
+      "step": 12920
+    },
+    {
+      "epoch": 16.49,
+      "learning_rate": 5.414921465968585e-05,
+      "loss": 0.0505,
+      "step": 12930
+    },
+    {
+      "epoch": 16.5,
+      "learning_rate": 5.3952879581151825e-05,
+      "loss": 0.0616,
+      "step": 12940
+    },
+    {
+      "epoch": 16.52,
+      "learning_rate": 5.37565445026178e-05,
+      "loss": 0.1069,
+      "step": 12950
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 5.356020942408377e-05,
+      "loss": 0.0694,
+      "step": 12960
+    },
+    {
+      "epoch": 16.54,
+      "learning_rate": 5.3363874345549735e-05,
+      "loss": 0.0547,
+      "step": 12970
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 5.31675392670157e-05,
+      "loss": 0.0542,
+      "step": 12980
+    },
+    {
+      "epoch": 16.57,
+      "learning_rate": 5.297120418848167e-05,
+      "loss": 0.0562,
+      "step": 12990
+    },
+    {
+      "epoch": 16.58,
+      "learning_rate": 5.277486910994764e-05,
+      "loss": 0.084,
+      "step": 13000
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 5.257853403141361e-05,
+      "loss": 0.0672,
+      "step": 13010
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 5.2382198952879576e-05,
+      "loss": 0.0684,
+      "step": 13020
+    },
+    {
+      "epoch": 16.62,
+      "learning_rate": 5.218586387434554e-05,
+      "loss": 0.0556,
+      "step": 13030
+    },
+    {
+      "epoch": 16.63,
+      "learning_rate": 5.1989528795811514e-05,
+      "loss": 0.052,
+      "step": 13040
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 5.1793193717277486e-05,
+      "loss": 0.1033,
+      "step": 13050
+    },
+    {
+      "epoch": 16.66,
+      "learning_rate": 5.159685863874345e-05,
+      "loss": 0.0679,
+      "step": 13060
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 5.140052356020942e-05,
+      "loss": 0.0587,
+      "step": 13070
+    },
+    {
+      "epoch": 16.68,
+      "learning_rate": 5.120418848167539e-05,
+      "loss": 0.0418,
+      "step": 13080
+    },
+    {
+      "epoch": 16.7,
+      "learning_rate": 5.1007853403141355e-05,
+      "loss": 0.0407,
+      "step": 13090
+    },
+    {
+      "epoch": 16.71,
+      "learning_rate": 5.081151832460733e-05,
+      "loss": 0.0853,
+      "step": 13100
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 5.061518324607329e-05,
+      "loss": 0.0689,
+      "step": 13110
+    },
+    {
+      "epoch": 16.73,
+      "learning_rate": 5.041884816753926e-05,
+      "loss": 0.0632,
+      "step": 13120
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 5.022251308900523e-05,
+      "loss": 0.0572,
+      "step": 13130
+    },
+    {
+      "epoch": 16.76,
+      "learning_rate": 5.00261780104712e-05,
+      "loss": 0.0504,
+      "step": 13140
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 4.9829842931937175e-05,
+      "loss": 0.1229,
+      "step": 13150
+    },
+    {
+      "epoch": 16.79,
+      "learning_rate": 4.9633507853403134e-05,
+      "loss": 0.0673,
+      "step": 13160
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 4.9437172774869106e-05,
+      "loss": 0.0524,
+      "step": 13170
+    },
+    {
+      "epoch": 16.81,
+      "learning_rate": 4.924083769633508e-05,
+      "loss": 0.0426,
+      "step": 13180
+    },
+    {
+      "epoch": 16.82,
+      "learning_rate": 4.9044502617801044e-05,
+      "loss": 0.0527,
+      "step": 13190
+    },
+    {
+      "epoch": 16.84,
+      "learning_rate": 4.884816753926701e-05,
+      "loss": 0.0841,
+      "step": 13200
+    },
+    {
+      "epoch": 16.84,
+      "eval_loss": 0.3348087668418884,
+      "eval_runtime": 1379.6677,
+      "eval_samples_per_second": 4.518,
+      "eval_steps_per_second": 0.565,
+      "eval_wer": 0.3512945856024981,
+      "step": 13200
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 4.8651832460732975e-05,
+      "loss": 0.064,
+      "step": 13210
+    },
+    {
+      "epoch": 16.86,
+      "learning_rate": 4.845549738219895e-05,
+      "loss": 0.0501,
+      "step": 13220
+    },
+    {
+      "epoch": 16.87,
+      "learning_rate": 4.825916230366492e-05,
+      "loss": 0.0433,
+      "step": 13230
+    },
+    {
+      "epoch": 16.89,
+      "learning_rate": 4.806282722513089e-05,
+      "loss": 0.054,
+      "step": 13240
+    },
+    {
+      "epoch": 16.9,
+      "learning_rate": 4.786649214659685e-05,
+      "loss": 0.0975,
+      "step": 13250
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 4.767015706806282e-05,
+      "loss": 0.0802,
+      "step": 13260
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 4.7473821989528795e-05,
+      "loss": 0.0502,
+      "step": 13270
+    },
+    {
+      "epoch": 16.94,
+      "learning_rate": 4.727748691099476e-05,
+      "loss": 0.0384,
+      "step": 13280
+    },
+    {
+      "epoch": 16.95,
+      "learning_rate": 4.7081151832460726e-05,
+      "loss": 0.0431,
+      "step": 13290
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 4.68848167539267e-05,
+      "loss": 0.0907,
+      "step": 13300
+    },
+    {
+      "epoch": 16.98,
+      "learning_rate": 4.6688481675392664e-05,
+      "loss": 0.0705,
+      "step": 13310
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 4.6492146596858636e-05,
+      "loss": 0.0407,
+      "step": 13320
+    },
+    {
+      "epoch": 17.0,
+      "learning_rate": 4.62958115183246e-05,
+      "loss": 0.0715,
+      "step": 13330
+    },
+    {
+      "epoch": 17.02,
+      "learning_rate": 4.609947643979057e-05,
+      "loss": 0.0755,
+      "step": 13340
+    },
+    {
+      "epoch": 17.03,
+      "learning_rate": 4.590314136125654e-05,
+      "loss": 0.0586,
+      "step": 13350
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 4.570680628272251e-05,
+      "loss": 0.0469,
+      "step": 13360
+    },
+    {
+      "epoch": 17.05,
+      "learning_rate": 4.5510471204188484e-05,
+      "loss": 0.0298,
+      "step": 13370
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 4.531413612565444e-05,
+      "loss": 0.0591,
+      "step": 13380
+    },
+    {
+      "epoch": 17.08,
+      "learning_rate": 4.5117801047120415e-05,
+      "loss": 0.0682,
+      "step": 13390
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 4.492146596858638e-05,
+      "loss": 0.0626,
+      "step": 13400
+    },
+    {
+      "epoch": 17.1,
+      "learning_rate": 4.472513089005235e-05,
+      "loss": 0.0419,
+      "step": 13410
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 4.452879581151832e-05,
+      "loss": 0.0621,
+      "step": 13420
+    },
+    {
+      "epoch": 17.13,
+      "learning_rate": 4.4332460732984284e-05,
+      "loss": 0.0721,
+      "step": 13430
+    },
+    {
+      "epoch": 17.14,
+      "learning_rate": 4.4136125654450257e-05,
+      "loss": 0.0748,
+      "step": 13440
+    },
+    {
+      "epoch": 17.16,
+      "learning_rate": 4.393979057591623e-05,
+      "loss": 0.05,
+      "step": 13450
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 4.37434554973822e-05,
+      "loss": 0.057,
+      "step": 13460
+    },
+    {
+      "epoch": 17.18,
+      "learning_rate": 4.354712041884816e-05,
+      "loss": 0.0429,
+      "step": 13470
+    },
+    {
+      "epoch": 17.19,
+      "learning_rate": 4.335078534031413e-05,
+      "loss": 0.0751,
+      "step": 13480
+    },
+    {
+      "epoch": 17.21,
+      "learning_rate": 4.3154450261780104e-05,
+      "loss": 0.0783,
+      "step": 13490
+    },
+    {
+      "epoch": 17.22,
+      "learning_rate": 4.295811518324607e-05,
+      "loss": 0.0601,
+      "step": 13500
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 4.2761780104712035e-05,
+      "loss": 0.0423,
+      "step": 13510
+    },
+    {
+      "epoch": 17.24,
+      "learning_rate": 4.256544502617801e-05,
+      "loss": 0.0315,
+      "step": 13520
+    },
+    {
+      "epoch": 17.26,
+      "learning_rate": 4.236910994764397e-05,
+      "loss": 0.074,
+      "step": 13530
+    },
+    {
+      "epoch": 17.27,
+      "learning_rate": 4.2172774869109946e-05,
+      "loss": 0.0653,
+      "step": 13540
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 4.197643979057592e-05,
+      "loss": 0.0498,
+      "step": 13550
+    },
+    {
+      "epoch": 17.3,
+      "learning_rate": 4.178010471204188e-05,
+      "loss": 0.0464,
+      "step": 13560
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 4.158376963350785e-05,
+      "loss": 0.0322,
+      "step": 13570
+    },
+    {
+      "epoch": 17.32,
+      "learning_rate": 4.138743455497382e-05,
+      "loss": 0.0735,
+      "step": 13580
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 4.119109947643979e-05,
+      "loss": 0.0773,
+      "step": 13590
+    },
+    {
+      "epoch": 17.35,
+      "learning_rate": 4.099476439790575e-05,
+      "loss": 0.0502,
+      "step": 13600
+    },
+    {
+      "epoch": 17.35,
+      "eval_loss": 0.3318431079387665,
+      "eval_runtime": 1380.125,
+      "eval_samples_per_second": 4.516,
+      "eval_steps_per_second": 0.565,
+      "eval_wer": 0.3485530008735897,
+      "step": 13600
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 4.0798429319371725e-05,
+      "loss": 0.0402,
+      "step": 13610
+    },
+    {
+      "epoch": 17.37,
+      "learning_rate": 4.060209424083769e-05,
+      "loss": 0.0313,
+      "step": 13620
+    },
+    {
+      "epoch": 17.38,
+      "learning_rate": 4.040575916230366e-05,
+      "loss": 0.0724,
+      "step": 13630
+    },
+    {
+      "epoch": 17.4,
+      "learning_rate": 4.0209424083769635e-05,
+      "loss": 0.086,
+      "step": 13640
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 4.001308900523559e-05,
+      "loss": 0.0625,
+      "step": 13650
+    },
+    {
+      "epoch": 17.42,
+      "learning_rate": 3.9816753926701566e-05,
+      "loss": 0.0549,
+      "step": 13660
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 3.962041884816754e-05,
+      "loss": 0.0335,
+      "step": 13670
+    },
+    {
+      "epoch": 17.45,
+      "learning_rate": 3.942408376963351e-05,
+      "loss": 0.0645,
+      "step": 13680
+    },
+    {
+      "epoch": 17.46,
+      "learning_rate": 3.922774869109947e-05,
+      "loss": 0.0662,
+      "step": 13690
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 3.903141361256544e-05,
+      "loss": 0.0528,
+      "step": 13700
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 3.8835078534031414e-05,
+      "loss": 0.0536,
+      "step": 13710
+    },
+    {
+      "epoch": 17.5,
+      "learning_rate": 3.863874345549738e-05,
+      "loss": 0.0409,
+      "step": 13720
+    },
+    {
+      "epoch": 17.51,
+      "learning_rate": 3.8442408376963345e-05,
+      "loss": 0.0661,
+      "step": 13730
+    },
+    {
+      "epoch": 17.53,
+      "learning_rate": 3.824607329842932e-05,
+      "loss": 0.0916,
+      "step": 13740
+    },
+    {
+      "epoch": 17.54,
+      "learning_rate": 3.804973821989528e-05,
+      "loss": 0.0476,
+      "step": 13750
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 3.7853403141361255e-05,
+      "loss": 0.0468,
+      "step": 13760
+    },
+    {
+      "epoch": 17.56,
+      "learning_rate": 3.765706806282723e-05,
+      "loss": 0.0413,
+      "step": 13770
+    },
+    {
+      "epoch": 17.58,
+      "learning_rate": 3.746073298429319e-05,
+      "loss": 0.0769,
+      "step": 13780
+    },
+    {
+      "epoch": 17.59,
+      "learning_rate": 3.726439790575916e-05,
+      "loss": 0.0663,
+      "step": 13790
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 3.706806282722513e-05,
+      "loss": 0.0608,
+      "step": 13800
+    },
+    {
+      "epoch": 17.61,
+      "learning_rate": 3.6871727748691096e-05,
+      "loss": 0.038,
+      "step": 13810
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 3.667539267015707e-05,
+      "loss": 0.0463,
+      "step": 13820
+    },
+    {
+      "epoch": 17.64,
+      "learning_rate": 3.6479057591623034e-05,
+      "loss": 0.0601,
+      "step": 13830
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 3.6282722513089e-05,
+      "loss": 0.0892,
+      "step": 13840
+    },
+    {
+      "epoch": 17.67,
+      "learning_rate": 3.608638743455497e-05,
+      "loss": 0.0541,
+      "step": 13850
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 3.589005235602094e-05,
+      "loss": 0.0499,
+      "step": 13860
+    },
+    {
+      "epoch": 17.69,
+      "learning_rate": 3.569371727748691e-05,
+      "loss": 0.0371,
+      "step": 13870
+    },
+    {
+      "epoch": 17.7,
+      "learning_rate": 3.5497382198952875e-05,
+      "loss": 0.0642,
+      "step": 13880
+    },
+    {
+      "epoch": 17.72,
+      "learning_rate": 3.530104712041885e-05,
+      "loss": 0.0667,
+      "step": 13890
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 3.510471204188481e-05,
+      "loss": 0.0849,
+      "step": 13900
+    },
+    {
+      "epoch": 17.74,
+      "learning_rate": 3.4908376963350785e-05,
+      "loss": 0.0486,
+      "step": 13910
+    },
+    {
+      "epoch": 17.75,
+      "learning_rate": 3.471204188481675e-05,
+      "loss": 0.0579,
+      "step": 13920
+    },
+    {
+      "epoch": 17.77,
+      "learning_rate": 3.451570680628272e-05,
+      "loss": 0.082,
+      "step": 13930
+    },
+    {
+      "epoch": 17.78,
+      "learning_rate": 3.431937172774869e-05,
+      "loss": 0.0786,
+      "step": 13940
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 3.4123036649214654e-05,
+      "loss": 0.0612,
+      "step": 13950
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 3.3926701570680626e-05,
+      "loss": 0.05,
+      "step": 13960
+    },
+    {
+      "epoch": 17.82,
+      "learning_rate": 3.373036649214659e-05,
+      "loss": 0.0357,
+      "step": 13970
+    },
+    {
+      "epoch": 17.83,
+      "learning_rate": 3.3534031413612564e-05,
+      "loss": 0.0639,
+      "step": 13980
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 3.333769633507853e-05,
+      "loss": 0.0664,
+      "step": 13990
+    },
+    {
+      "epoch": 17.86,
+      "learning_rate": 3.31413612565445e-05,
+      "loss": 0.059,
+      "step": 14000
+    },
+    {
+      "epoch": 17.86,
+      "eval_loss": 0.3359043598175049,
+      "eval_runtime": 1379.5922,
+      "eval_samples_per_second": 4.518,
+      "eval_steps_per_second": 0.565,
+      "eval_wer": 0.34620174346201743,
+      "step": 14000
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 3.294502617801047e-05,
+      "loss": 0.0363,
+      "step": 14010
+    },
+    {
+      "epoch": 17.88,
+      "learning_rate": 3.274869109947644e-05,
+      "loss": 0.0359,
+      "step": 14020
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 3.2552356020942405e-05,
+      "loss": 0.0641,
+      "step": 14030
+    },
+    {
+      "epoch": 17.91,
+      "learning_rate": 3.235602094240838e-05,
+      "loss": 0.0712,
+      "step": 14040
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 3.215968586387434e-05,
+      "loss": 0.0545,
+      "step": 14050
+    },
+    {
+      "epoch": 17.93,
+      "learning_rate": 3.196335078534031e-05,
+      "loss": 0.042,
+      "step": 14060
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 3.176701570680628e-05,
+      "loss": 0.0392,
+      "step": 14070
+    },
+    {
+      "epoch": 17.96,
+      "learning_rate": 3.1570680628272246e-05,
+      "loss": 0.0655,
+      "step": 14080
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 3.137434554973822e-05,
+      "loss": 0.0547,
+      "step": 14090
+    },
+    {
+      "epoch": 17.98,
+      "learning_rate": 3.1178010471204184e-05,
+      "loss": 0.0488,
+      "step": 14100
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 3.0981675392670156e-05,
+      "loss": 0.0359,
+      "step": 14110
+    },
+    {
+      "epoch": 18.01,
+      "learning_rate": 3.078534031413612e-05,
+      "loss": 0.0851,
+      "step": 14120
+    },
+    {
+      "epoch": 18.02,
+      "learning_rate": 3.0589005235602094e-05,
+      "loss": 0.0517,
+      "step": 14130
+    },
+    {
+      "epoch": 18.04,
+      "learning_rate": 3.0392670157068063e-05,
+      "loss": 0.0342,
+      "step": 14140
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 3.019633507853403e-05,
+      "loss": 0.0317,
+      "step": 14150
+    },
+    {
+      "epoch": 18.06,
+      "learning_rate": 2.9999999999999997e-05,
+      "loss": 0.0384,
+      "step": 14160
+    },
+    {
+      "epoch": 18.07,
+      "learning_rate": 2.9803664921465966e-05,
+      "loss": 0.0871,
+      "step": 14170
+    },
+    {
+      "epoch": 18.09,
+      "learning_rate": 2.9607329842931935e-05,
+      "loss": 0.051,
+      "step": 14180
+    },
+    {
+      "epoch": 18.1,
+      "learning_rate": 2.94109947643979e-05,
+      "loss": 0.0429,
+      "step": 14190
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 2.9214659685863873e-05,
+      "loss": 0.0339,
+      "step": 14200
+    },
+    {
+      "epoch": 18.12,
+      "learning_rate": 2.901832460732984e-05,
+      "loss": 0.0473,
+      "step": 14210
+    },
+    {
+      "epoch": 18.14,
+      "learning_rate": 2.882198952879581e-05,
+      "loss": 0.0719,
+      "step": 14220
+    },
+    {
+      "epoch": 18.15,
+      "learning_rate": 2.8625654450261776e-05,
+      "loss": 0.056,
+      "step": 14230
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 2.8429319371727745e-05,
+      "loss": 0.0384,
+      "step": 14240
+    },
+    {
+      "epoch": 18.18,
+      "learning_rate": 2.8232984293193718e-05,
+      "loss": 0.0384,
+      "step": 14250
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 2.8036649214659683e-05,
+      "loss": 0.0729,
+      "step": 14260
+    },
+    {
+      "epoch": 18.2,
+      "learning_rate": 2.7840314136125652e-05,
+      "loss": 0.0885,
+      "step": 14270
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 2.764397905759162e-05,
+      "loss": 0.0604,
+      "step": 14280
+    },
+    {
+      "epoch": 18.23,
+      "learning_rate": 2.744764397905759e-05,
+      "loss": 0.0394,
+      "step": 14290
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 2.7251308900523555e-05,
+      "loss": 0.0334,
+      "step": 14300
+    },
+    {
+      "epoch": 18.25,
+      "learning_rate": 2.7054973821989528e-05,
+      "loss": 0.0574,
+      "step": 14310
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 2.6858638743455493e-05,
+      "loss": 0.0812,
+      "step": 14320
+    },
+    {
+      "epoch": 18.28,
+      "learning_rate": 2.6662303664921465e-05,
+      "loss": 0.0506,
+      "step": 14330
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 2.646596858638743e-05,
+      "loss": 0.056,
+      "step": 14340
+    },
+    {
+      "epoch": 18.3,
+      "learning_rate": 2.62696335078534e-05,
+      "loss": 0.0327,
+      "step": 14350
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 2.6073298429319372e-05,
+      "loss": 0.0279,
+      "step": 14360
+    },
+    {
+      "epoch": 18.33,
+      "learning_rate": 2.5876963350785338e-05,
+      "loss": 0.087,
+      "step": 14370
+    },
+    {
+      "epoch": 18.34,
+      "learning_rate": 2.5680628272251307e-05,
+      "loss": 0.0527,
+      "step": 14380
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 2.5484293193717275e-05,
+      "loss": 0.0558,
+      "step": 14390
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 2.5287958115183244e-05,
+      "loss": 0.0417,
+      "step": 14400
+    },
+    {
+      "epoch": 18.37,
+      "eval_loss": 0.3309715986251831,
+      "eval_runtime": 1383.5299,
+      "eval_samples_per_second": 4.505,
+      "eval_steps_per_second": 0.564,
+      "eval_wer": 0.3466571253322429,
+      "step": 14400
+    },
+    {
+      "epoch": 18.38,
+      "learning_rate": 2.509162303664921e-05,
+      "loss": 0.0357,
+      "step": 14410
+    },
+    {
+      "epoch": 18.39,
+      "learning_rate": 2.4895287958115182e-05,
+      "loss": 0.0757,
+      "step": 14420
+    },
+    {
+      "epoch": 18.41,
+      "learning_rate": 2.4698952879581148e-05,
+      "loss": 0.0532,
+      "step": 14430
+    },
+    {
+      "epoch": 18.42,
+      "learning_rate": 2.450261780104712e-05,
+      "loss": 0.0463,
+      "step": 14440
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 2.430628272251309e-05,
+      "loss": 0.0333,
+      "step": 14450
+    },
+    {
+      "epoch": 18.44,
+      "learning_rate": 2.4109947643979054e-05,
+      "loss": 0.0303,
+      "step": 14460
+    },
+    {
+      "epoch": 18.46,
+      "learning_rate": 2.3913612565445027e-05,
+      "loss": 0.0694,
+      "step": 14470
+    },
+    {
+      "epoch": 18.47,
+      "learning_rate": 2.3717277486910992e-05,
+      "loss": 0.058,
+      "step": 14480
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 2.352094240837696e-05,
+      "loss": 0.0557,
+      "step": 14490
+    },
+    {
+      "epoch": 18.49,
+      "learning_rate": 2.332460732984293e-05,
+      "loss": 0.0351,
+      "step": 14500
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 2.31282722513089e-05,
+      "loss": 0.0505,
+      "step": 14510
+    },
+    {
+      "epoch": 18.52,
+      "learning_rate": 2.2931937172774864e-05,
+      "loss": 0.069,
+      "step": 14520
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 2.2735602094240837e-05,
+      "loss": 0.0493,
+      "step": 14530
+    },
+    {
+      "epoch": 18.55,
+      "learning_rate": 2.2539267015706802e-05,
+      "loss": 0.0582,
+      "step": 14540
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 2.234293193717277e-05,
+      "loss": 0.0402,
+      "step": 14550
+    },
+    {
+      "epoch": 18.57,
+      "learning_rate": 2.2146596858638743e-05,
+      "loss": 0.0442,
+      "step": 14560
+    },
+    {
+      "epoch": 18.58,
+      "learning_rate": 2.195026178010471e-05,
+      "loss": 0.0752,
+      "step": 14570
+    },
+    {
+      "epoch": 18.6,
+      "learning_rate": 2.175392670157068e-05,
+      "loss": 0.0763,
+      "step": 14580
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 2.1557591623036647e-05,
+      "loss": 0.0611,
+      "step": 14590
+    },
+    {
+      "epoch": 18.62,
+      "learning_rate": 2.1361256544502616e-05,
+      "loss": 0.0361,
+      "step": 14600
+    },
+    {
+      "epoch": 18.63,
+      "learning_rate": 2.1164921465968585e-05,
+      "loss": 0.0538,
+      "step": 14610
+    },
+    {
+      "epoch": 18.65,
+      "learning_rate": 2.0968586387434553e-05,
+      "loss": 0.067,
+      "step": 14620
+    },
+    {
+      "epoch": 18.66,
+      "learning_rate": 2.077225130890052e-05,
+      "loss": 0.0505,
+      "step": 14630
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 2.057591623036649e-05,
+      "loss": 0.0431,
+      "step": 14640
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 2.037958115183246e-05,
+      "loss": 0.0322,
+      "step": 14650
+    },
+    {
+      "epoch": 18.7,
+      "learning_rate": 2.0183246073298426e-05,
+      "loss": 0.0663,
+      "step": 14660
+    },
+    {
+      "epoch": 18.71,
+      "learning_rate": 1.9986910994764398e-05,
+      "loss": 0.0648,
+      "step": 14670
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 1.9790575916230363e-05,
+      "loss": 0.0456,
+      "step": 14680
+    },
+    {
+      "epoch": 18.74,
+      "learning_rate": 1.9594240837696336e-05,
+      "loss": 0.0485,
+      "step": 14690
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 1.93979057591623e-05,
+      "loss": 0.0264,
+      "step": 14700
+    },
+    {
+      "epoch": 18.76,
+      "learning_rate": 1.920157068062827e-05,
+      "loss": 0.0401,
+      "step": 14710
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 1.900523560209424e-05,
+      "loss": 0.078,
+      "step": 14720
+    },
+    {
+      "epoch": 18.79,
+      "learning_rate": 1.8808900523560208e-05,
+      "loss": 0.0495,
+      "step": 14730
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 1.8612565445026177e-05,
+      "loss": 0.037,
+      "step": 14740
+    },
+    {
+      "epoch": 18.81,
+      "learning_rate": 1.8416230366492146e-05,
+      "loss": 0.0291,
+      "step": 14750
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 1.8219895287958115e-05,
+      "loss": 0.0296,
+      "step": 14760
+    },
+    {
+      "epoch": 18.84,
+      "learning_rate": 1.802356020942408e-05,
+      "loss": 0.0827,
+      "step": 14770
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 1.782722513089005e-05,
+      "loss": 0.0495,
+      "step": 14780
+    },
+    {
+      "epoch": 18.86,
+      "learning_rate": 1.7630890052356018e-05,
+      "loss": 0.0547,
+      "step": 14790
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 1.743455497382199e-05,
+      "loss": 0.0322,
+      "step": 14800
+    },
+    {
+      "epoch": 18.88,
+      "eval_loss": 0.33246901631355286,
+      "eval_runtime": 1384.5739,
+      "eval_samples_per_second": 4.502,
+      "eval_steps_per_second": 0.563,
+      "eval_wer": 0.3466571253322429,
+      "step": 14800
+    },
+    {
+      "epoch": 18.89,
+      "learning_rate": 1.7238219895287956e-05,
+      "loss": 0.0359,
+      "step": 14810
+    },
+    {
+      "epoch": 18.9,
+      "learning_rate": 1.7041884816753925e-05,
+      "loss": 0.0737,
+      "step": 14820
+    },
+    {
+      "epoch": 18.92,
+      "learning_rate": 1.6845549738219894e-05,
+      "loss": 0.0716,
+      "step": 14830
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 1.6649214659685863e-05,
+      "loss": 0.0439,
+      "step": 14840
+    },
+    {
+      "epoch": 18.94,
+      "learning_rate": 1.645287958115183e-05,
+      "loss": 0.0348,
+      "step": 14850
+    },
+    {
+      "epoch": 18.95,
+      "learning_rate": 1.62565445026178e-05,
+      "loss": 0.0363,
+      "step": 14860
+    },
+    {
+      "epoch": 18.97,
+      "learning_rate": 1.606020942408377e-05,
+      "loss": 0.0668,
+      "step": 14870
+    },
+    {
+      "epoch": 18.98,
+      "learning_rate": 1.5863874345549735e-05,
+      "loss": 0.0442,
+      "step": 14880
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 1.5667539267015704e-05,
+      "loss": 0.0271,
+      "step": 14890
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 1.5471204188481676e-05,
+      "loss": 0.0562,
+      "step": 14900
+    },
+    {
+      "epoch": 19.02,
+      "learning_rate": 1.5274869109947645e-05,
+      "loss": 0.049,
+      "step": 14910
+    },
+    {
+      "epoch": 19.03,
+      "learning_rate": 1.5078534031413612e-05,
+      "loss": 0.0392,
+      "step": 14920
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 1.488219895287958e-05,
+      "loss": 0.0389,
+      "step": 14930
+    },
+    {
+      "epoch": 19.06,
+      "learning_rate": 1.4685863874345548e-05,
+      "loss": 0.0294,
+      "step": 14940
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 1.4489528795811517e-05,
+      "loss": 0.0483,
+      "step": 14950
+    },
+    {
+      "epoch": 19.08,
+      "learning_rate": 1.4293193717277486e-05,
+      "loss": 0.0617,
+      "step": 14960
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 1.4096858638743453e-05,
+      "loss": 0.0411,
+      "step": 14970
+    },
+    {
+      "epoch": 19.11,
+      "learning_rate": 1.3900523560209422e-05,
+      "loss": 0.0427,
+      "step": 14980
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 1.3704188481675391e-05,
+      "loss": 0.035,
+      "step": 14990
+    },
+    {
+      "epoch": 19.13,
+      "learning_rate": 1.3507853403141362e-05,
+      "loss": 0.0619,
+      "step": 15000
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 1.3311518324607329e-05,
+      "loss": 0.05,
+      "step": 15010
+    },
+    {
+      "epoch": 19.16,
+      "learning_rate": 1.3115183246073298e-05,
+      "loss": 0.0498,
+      "step": 15020
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 1.2918848167539267e-05,
+      "loss": 0.0291,
+      "step": 15030
+    },
+    {
+      "epoch": 19.18,
+      "learning_rate": 1.2722513089005234e-05,
+      "loss": 0.0392,
+      "step": 15040
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 1.2526178010471203e-05,
+      "loss": 0.0592,
+      "step": 15050
+    },
+    {
+      "epoch": 19.21,
+      "learning_rate": 1.2329842931937172e-05,
+      "loss": 0.0498,
+      "step": 15060
+    },
+    {
+      "epoch": 19.22,
+      "learning_rate": 1.2133507853403139e-05,
+      "loss": 0.041,
+      "step": 15070
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 1.1937172774869108e-05,
+      "loss": 0.0377,
+      "step": 15080
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 1.1740837696335077e-05,
+      "loss": 0.0277,
+      "step": 15090
+    },
+    {
+      "epoch": 19.26,
+      "learning_rate": 1.1544502617801047e-05,
+      "loss": 0.054,
+      "step": 15100
+    },
+    {
+      "epoch": 19.27,
+      "learning_rate": 1.1348167539267016e-05,
+      "loss": 0.0504,
+      "step": 15110
+    },
+    {
+      "epoch": 19.29,
+      "learning_rate": 1.1151832460732983e-05,
+      "loss": 0.0477,
+      "step": 15120
+    },
+    {
+      "epoch": 19.3,
+      "learning_rate": 1.0955497382198952e-05,
+      "loss": 0.0389,
+      "step": 15130
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 1.0759162303664921e-05,
+      "loss": 0.0267,
+      "step": 15140
+    },
+    {
+      "epoch": 19.32,
+      "learning_rate": 1.0562827225130888e-05,
+      "loss": 0.0719,
+      "step": 15150
+    },
+    {
+      "epoch": 19.34,
+      "learning_rate": 1.0366492146596857e-05,
+      "loss": 0.0597,
+      "step": 15160
+    },
+    {
+      "epoch": 19.35,
+      "learning_rate": 1.0170157068062826e-05,
+      "loss": 0.067,
+      "step": 15170
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 9.973821989528794e-06,
+      "loss": 0.0342,
+      "step": 15180
+    },
+    {
+      "epoch": 19.37,
+      "learning_rate": 9.777486910994762e-06,
+      "loss": 0.0297,
+      "step": 15190
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.581151832460733e-06,
+      "loss": 0.0531,
+      "step": 15200
+    },
+    {
+      "epoch": 19.39,
+      "eval_loss": 0.33570629358291626,
+      "eval_runtime": 1380.1267,
+      "eval_samples_per_second": 4.516,
+      "eval_steps_per_second": 0.565,
+      "eval_wer": 0.3449006524042304,
+      "step": 15200
+    },
+    {
+      "epoch": 19.4,
+      "learning_rate": 9.384816753926702e-06,
+      "loss": 0.0562,
+      "step": 15210
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.188481675392669e-06,
+      "loss": 0.037,
+      "step": 15220
+    },
+    {
+      "epoch": 19.43,
+      "learning_rate": 8.992146596858638e-06,
+      "loss": 0.0304,
+      "step": 15230
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 8.795811518324607e-06,
+      "loss": 0.0308,
+      "step": 15240
+    },
+    {
+      "epoch": 19.45,
+      "learning_rate": 8.599476439790576e-06,
+      "loss": 0.0707,
+      "step": 15250
+    },
+    {
+      "epoch": 19.46,
+      "learning_rate": 8.403141361256543e-06,
+      "loss": 0.0502,
+      "step": 15260
+    },
+    {
+      "epoch": 19.48,
+      "learning_rate": 8.206806282722512e-06,
+      "loss": 0.0417,
+      "step": 15270
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 8.010471204188481e-06,
+      "loss": 0.0488,
+      "step": 15280
+    },
+    {
+      "epoch": 19.5,
+      "learning_rate": 7.81413612565445e-06,
+      "loss": 0.0249,
+      "step": 15290
+    },
+    {
+      "epoch": 19.51,
+      "learning_rate": 7.617801047120418e-06,
+      "loss": 0.0618,
+      "step": 15300
+    },
+    {
+      "epoch": 19.53,
+      "learning_rate": 7.421465968586387e-06,
+      "loss": 0.063,
+      "step": 15310
+    },
+    {
+      "epoch": 19.54,
+      "learning_rate": 7.225130890052355e-06,
+      "loss": 0.0409,
+      "step": 15320
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 7.0287958115183245e-06,
+      "loss": 0.0274,
+      "step": 15330
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 6.832460732984293e-06,
+      "loss": 0.0291,
+      "step": 15340
+    },
+    {
+      "epoch": 19.58,
+      "learning_rate": 6.6361256544502615e-06,
+      "loss": 0.0523,
+      "step": 15350
+    },
+    {
+      "epoch": 19.59,
+      "learning_rate": 6.4397905759162296e-06,
+      "loss": 0.0675,
+      "step": 15360
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 6.243455497382198e-06,
+      "loss": 0.0374,
+      "step": 15370
+    },
+    {
+      "epoch": 19.62,
+      "learning_rate": 6.047120418848167e-06,
+      "loss": 0.0463,
+      "step": 15380
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 5.8507853403141354e-06,
+      "loss": 0.0272,
+      "step": 15390
+    },
+    {
+      "epoch": 19.64,
+      "learning_rate": 5.654450261780104e-06,
+      "loss": 0.0591,
+      "step": 15400
+    },
+    {
+      "epoch": 19.66,
+      "learning_rate": 5.458115183246072e-06,
+      "loss": 0.0611,
+      "step": 15410
+    },
+    {
+      "epoch": 19.67,
+      "learning_rate": 5.261780104712041e-06,
+      "loss": 0.0428,
+      "step": 15420
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 5.06544502617801e-06,
+      "loss": 0.0491,
+      "step": 15430
+    },
+    {
+      "epoch": 19.69,
+      "learning_rate": 4.869109947643979e-06,
+      "loss": 0.0313,
+      "step": 15440
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 4.672774869109947e-06,
+      "loss": 0.0517,
+      "step": 15450
+    },
+    {
+      "epoch": 19.72,
+      "learning_rate": 4.476439790575916e-06,
+      "loss": 0.0454,
+      "step": 15460
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 4.280104712041884e-06,
+      "loss": 0.0368,
+      "step": 15470
+    },
+    {
+      "epoch": 19.74,
+      "learning_rate": 4.083769633507853e-06,
+      "loss": 0.0525,
+      "step": 15480
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 3.887434554973822e-06,
+      "loss": 0.0329,
+      "step": 15490
+    },
+    {
+      "epoch": 19.77,
+      "learning_rate": 3.69109947643979e-06,
+      "loss": 0.0615,
+      "step": 15500
+    },
+    {
+      "epoch": 19.78,
+      "learning_rate": 3.494764397905759e-06,
+      "loss": 0.0503,
+      "step": 15510
+    },
+    {
+      "epoch": 19.8,
+      "learning_rate": 3.2984293193717274e-06,
+      "loss": 0.0393,
+      "step": 15520
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 3.102094240837696e-06,
+      "loss": 0.0299,
+      "step": 15530
+    },
+    {
+      "epoch": 19.82,
+      "learning_rate": 2.9057591623036648e-06,
+      "loss": 0.026,
+      "step": 15540
+    },
+    {
+      "epoch": 19.83,
+      "learning_rate": 2.7094240837696333e-06,
+      "loss": 0.0628,
+      "step": 15550
+    },
+    {
+      "epoch": 19.85,
+      "learning_rate": 2.513089005235602e-06,
+      "loss": 0.049,
+      "step": 15560
+    },
+    {
+      "epoch": 19.86,
+      "learning_rate": 2.3167539267015706e-06,
+      "loss": 0.052,
+      "step": 15570
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 2.120418848167539e-06,
+      "loss": 0.0308,
+      "step": 15580
+    },
+    {
+      "epoch": 19.88,
+      "learning_rate": 1.9240837696335076e-06,
+      "loss": 0.0443,
+      "step": 15590
+    },
+    {
+      "epoch": 19.9,
+      "learning_rate": 1.7277486910994763e-06,
+      "loss": 0.0623,
+      "step": 15600
+    },
+    {
+      "epoch": 19.9,
+      "eval_loss": 0.3349722921848297,
+      "eval_runtime": 1389.7911,
+      "eval_samples_per_second": 4.485,
+      "eval_steps_per_second": 0.561,
+      "eval_wer": 0.3440921172468913,
+      "step": 15600
+    },
+    {
+      "epoch": 19.91,
+      "learning_rate": 1.531413612565445e-06,
+      "loss": 0.0683,
+      "step": 15610
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 1.3350785340314135e-06,
+      "loss": 0.0656,
+      "step": 15620
+    },
+    {
+      "epoch": 19.94,
+      "learning_rate": 1.1387434554973822e-06,
+      "loss": 0.0302,
+      "step": 15630
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.424083769633508e-07,
+      "loss": 0.0403,
+      "step": 15640
+    },
+    {
+      "epoch": 19.96,
+      "learning_rate": 7.460732984293192e-07,
+      "loss": 0.0627,
+      "step": 15650
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 5.497382198952879e-07,
+      "loss": 0.0544,
+      "step": 15660
+    },
+    {
+      "epoch": 19.99,
+      "learning_rate": 3.534031413612565e-07,
+      "loss": 0.0415,
+      "step": 15670
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 1.5706806282722514e-07,
+      "loss": 0.0394,
+      "step": 15680
+    },
+    {
+      "epoch": 20.0,
+      "step": 15680,
+      "total_flos": 0.0,
+      "train_loss": 0.6550551648090156,
+      "train_runtime": 151124.7681,
+      "train_samples_per_second": 6.645,
+      "train_steps_per_second": 0.104
+    }
+  ],
+  "max_steps": 15680,
+  "num_train_epochs": 20,
+  "total_flos": 0.0,
+  "trial_name": null,
+  "trial_params": null
+}