diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,3025 +1,9385 @@
 {
-  "best_metric": 0.6712434887886047,
-  "best_model_checkpoint": "ai-light-dance_drums_ft_pretrain_wav2vec2-base-new-v6-1/checkpoint-2590",
-  "epoch": 99.99290780141844,
-  "global_step": 3500,
+  "best_metric": 0.5660640597343445,
+  "best_model_checkpoint": "ai-light-dance_drums_ft_pretrain_wav2vec2-base-new-v6-1/checkpoint-3525",
+  "epoch": 100.0,
+  "global_step": 14100,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "epoch": 0.28,
-      "learning_rate": 6.666666666666667e-05,
-      "loss": 55.7655,
+      "epoch": 0.07,
+      "learning_rate": 9e-05,
+      "loss": 0.419,
       "step": 10
     },
     {
-      "epoch": 0.57,
-      "learning_rate": 0.0001866666666666667,
-      "loss": 58.6473,
+      "epoch": 0.14,
+      "learning_rate": 9.994322214336409e-05,
+      "loss": 0.3042,
       "step": 20
     },
     {
-      "epoch": 0.85,
-      "learning_rate": 0.00032,
-      "loss": 17.2739,
+      "epoch": 0.21,
+      "learning_rate": 9.98722498225692e-05,
+      "loss": 0.4124,
       "step": 30
     },
     {
-      "epoch": 0.99,
-      "eval_loss": 2.9041128158569336,
-      "eval_runtime": 8.8924,
-      "eval_samples_per_second": 7.085,
-      "eval_steps_per_second": 1.799,
-      "eval_wer": 1.0,
-      "step": 35
-    },
-    {
-      "epoch": 1.14,
-      "learning_rate": 0.00039953890489913545,
-      "loss": 2.8972,
+      "epoch": 0.28,
+      "learning_rate": 9.980127750177432e-05,
+      "loss": 0.5157,
       "step": 40
     },
     {
-      "epoch": 1.43,
-      "learning_rate": 0.0003983861671469741,
-      "loss": 2.082,
+      "epoch": 0.35,
+      "learning_rate": 9.973030518097943e-05,
+      "loss": 0.275,
       "step": 50
     },
     {
-      "epoch": 1.71,
-      "learning_rate": 0.0003972334293948127,
-      "loss": 2.1557,
+      "epoch": 0.43,
+      "learning_rate": 9.965933286018453e-05,
+      "loss": 0.362,
       "step": 60
     },
     {
-      "epoch": 1.99,
-      "learning_rate": 0.00039608069164265134,
-      "loss": 1.8424,
-      "step": 70
-    },
-    {
-      "epoch": 1.99,
-      "eval_loss": 3.505493402481079,
-      "eval_runtime": 9.0402,
-      "eval_samples_per_second": 6.969,
-      "eval_steps_per_second": 1.77,
-      "eval_wer": 1.0,
+      "epoch": 0.5,
+      "learning_rate": 9.958836053938965e-05,
+      "loss": 0.401,
       "step": 70
     },
     {
-      "epoch": 2.28,
-      "learning_rate": 0.0003949279538904899,
-      "loss": 1.8014,
+      "epoch": 0.57,
+      "learning_rate": 9.952448545067424e-05,
+      "loss": 0.4658,
       "step": 80
     },
     {
-      "epoch": 2.57,
-      "learning_rate": 0.00039377521613832855,
-      "loss": 1.6721,
+      "epoch": 0.64,
+      "learning_rate": 9.946061036195884e-05,
+      "loss": 0.3456,
       "step": 90
     },
     {
-      "epoch": 2.85,
-      "learning_rate": 0.0003926224783861672,
-      "loss": 1.7092,
+      "epoch": 0.71,
+      "learning_rate": 9.938963804116395e-05,
+      "loss": 0.4761,
       "step": 100
     },
     {
-      "epoch": 2.99,
-      "eval_loss": 2.0045998096466064,
-      "eval_runtime": 9.0436,
-      "eval_samples_per_second": 6.966,
-      "eval_steps_per_second": 1.769,
-      "eval_wer": 1.0,
-      "step": 105
-    },
-    {
-      "epoch": 3.14,
-      "learning_rate": 0.00039146974063400576,
-      "loss": 1.6035,
+      "epoch": 0.78,
+      "learning_rate": 9.931866572036906e-05,
+      "loss": 0.9702,
       "step": 110
     },
     {
-      "epoch": 3.43,
-      "learning_rate": 0.0003903170028818444,
-      "loss": 1.6196,
+      "epoch": 0.85,
+      "learning_rate": 9.924769339957418e-05,
+      "loss": 0.3642,
       "step": 120
     },
     {
-      "epoch": 3.71,
-      "learning_rate": 0.000389164265129683,
-      "loss": 1.587,
+      "epoch": 0.92,
+      "learning_rate": 9.917672107877927e-05,
+      "loss": 0.342,
       "step": 130
     },
     {
-      "epoch": 3.99,
-      "learning_rate": 0.00038801152737752165,
-      "loss": 1.5022,
+      "epoch": 0.99,
+      "learning_rate": 9.91057487579844e-05,
+      "loss": 0.2764,
       "step": 140
     },
     {
-      "epoch": 3.99,
-      "eval_loss": 1.9661929607391357,
-      "eval_runtime": 9.0602,
-      "eval_samples_per_second": 6.954,
-      "eval_steps_per_second": 1.766,
-      "eval_wer": 0.9675242513707296,
-      "step": 140
+      "epoch": 1.0,
+      "eval_loss": 0.7125303149223328,
+      "eval_runtime": 8.1602,
+      "eval_samples_per_second": 7.72,
+      "eval_steps_per_second": 3.921,
+      "eval_wer": 0.3618726275832982,
+      "step": 141
     },
     {
-      "epoch": 4.28,
-      "learning_rate": 0.00038685878962536023,
-      "loss": 1.486,
+      "epoch": 1.06,
+      "learning_rate": 9.90347764371895e-05,
+      "loss": 0.3607,
       "step": 150
     },
     {
-      "epoch": 4.57,
-      "learning_rate": 0.00038570605187319886,
-      "loss": 1.382,
+      "epoch": 1.13,
+      "learning_rate": 9.896380411639461e-05,
+      "loss": 0.2474,
       "step": 160
     },
     {
-      "epoch": 4.85,
-      "learning_rate": 0.0003845533141210375,
-      "loss": 1.2964,
+      "epoch": 1.21,
+      "learning_rate": 9.889283179559971e-05,
+      "loss": 0.3964,
       "step": 170
     },
     {
-      "epoch": 4.99,
-      "eval_loss": 1.9017366170883179,
-      "eval_runtime": 8.7445,
-      "eval_samples_per_second": 7.204,
-      "eval_steps_per_second": 1.83,
-      "eval_wer": 0.5504006748207507,
-      "step": 175
-    },
-    {
-      "epoch": 5.14,
-      "learning_rate": 0.0003834005763688761,
-      "loss": 1.3537,
+      "epoch": 1.28,
+      "learning_rate": 9.882185947480484e-05,
+      "loss": 0.4184,
       "step": 180
     },
     {
-      "epoch": 5.43,
-      "learning_rate": 0.0003822478386167147,
-      "loss": 1.1033,
+      "epoch": 1.35,
+      "learning_rate": 9.875088715400994e-05,
+      "loss": 0.4364,
       "step": 190
     },
     {
-      "epoch": 5.71,
-      "learning_rate": 0.00038109510086455333,
-      "loss": 1.2447,
+      "epoch": 1.42,
+      "learning_rate": 9.867991483321505e-05,
+      "loss": 0.2955,
       "step": 200
     },
     {
-      "epoch": 5.99,
-      "learning_rate": 0.00037994236311239196,
-      "loss": 1.1235,
-      "step": 210
-    },
-    {
-      "epoch": 5.99,
-      "eval_loss": 1.9874823093414307,
-      "eval_runtime": 8.6787,
-      "eval_samples_per_second": 7.259,
-      "eval_steps_per_second": 1.844,
-      "eval_wer": 0.4643610291016449,
+      "epoch": 1.49,
+      "learning_rate": 9.860894251242016e-05,
+      "loss": 0.607,
       "step": 210
     },
     {
-      "epoch": 6.28,
-      "learning_rate": 0.00037878962536023054,
-      "loss": 1.1092,
+      "epoch": 1.56,
+      "learning_rate": 9.853797019162528e-05,
+      "loss": 0.4059,
       "step": 220
     },
     {
-      "epoch": 6.57,
-      "learning_rate": 0.00037763688760806917,
-      "loss": 1.0707,
+      "epoch": 1.63,
+      "learning_rate": 9.846699787083037e-05,
+      "loss": 0.3457,
       "step": 230
     },
     {
-      "epoch": 6.85,
-      "learning_rate": 0.0003764841498559078,
-      "loss": 1.1056,
+      "epoch": 1.7,
+      "learning_rate": 9.839602555003549e-05,
+      "loss": 0.3669,
       "step": 240
     },
     {
-      "epoch": 6.99,
-      "eval_loss": 1.8791475296020508,
-      "eval_runtime": 8.7777,
-      "eval_samples_per_second": 7.177,
-      "eval_steps_per_second": 1.823,
-      "eval_wer": 0.47617039223956137,
-      "step": 245
-    },
-    {
-      "epoch": 7.14,
-      "learning_rate": 0.00037533141210374643,
-      "loss": 1.0885,
+      "epoch": 1.77,
+      "learning_rate": 9.83250532292406e-05,
+      "loss": 0.4576,
       "step": 250
     },
     {
-      "epoch": 7.43,
-      "learning_rate": 0.000374178674351585,
-      "loss": 0.9235,
+      "epoch": 1.84,
+      "learning_rate": 9.825408090844571e-05,
+      "loss": 0.38,
       "step": 260
     },
     {
-      "epoch": 7.71,
-      "learning_rate": 0.00037302593659942364,
-      "loss": 1.0039,
+      "epoch": 1.91,
+      "learning_rate": 9.818310858765081e-05,
+      "loss": 0.4483,
       "step": 270
     },
     {
-      "epoch": 7.99,
-      "learning_rate": 0.0003718731988472623,
-      "loss": 0.8907,
+      "epoch": 1.99,
+      "learning_rate": 9.811213626685594e-05,
+      "loss": 0.5415,
       "step": 280
     },
     {
-      "epoch": 7.99,
-      "eval_loss": 1.4811413288116455,
-      "eval_runtime": 8.7992,
-      "eval_samples_per_second": 7.16,
-      "eval_steps_per_second": 1.818,
-      "eval_wer": 0.467313369886124,
-      "step": 280
+      "epoch": 2.0,
+      "eval_loss": 0.7251520156860352,
+      "eval_runtime": 8.5869,
+      "eval_samples_per_second": 7.337,
+      "eval_steps_per_second": 3.727,
+      "eval_wer": 0.36819907212146774,
+      "step": 282
     },
     {
-      "epoch": 8.28,
-      "learning_rate": 0.0003707204610951009,
-      "loss": 1.0557,
+      "epoch": 2.06,
+      "learning_rate": 9.804116394606104e-05,
+      "loss": 0.3413,
       "step": 290
     },
     {
-      "epoch": 8.57,
-      "learning_rate": 0.0003695677233429395,
-      "loss": 0.9808,
+      "epoch": 2.13,
+      "learning_rate": 9.797019162526615e-05,
+      "loss": 0.3201,
       "step": 300
     },
     {
-      "epoch": 8.85,
-      "learning_rate": 0.0003684149855907781,
-      "loss": 0.8605,
+      "epoch": 2.2,
+      "learning_rate": 9.789921930447126e-05,
+      "loss": 0.3184,
       "step": 310
     },
     {
-      "epoch": 8.99,
-      "eval_loss": 1.9113553762435913,
-      "eval_runtime": 8.737,
-      "eval_samples_per_second": 7.211,
-      "eval_steps_per_second": 1.831,
-      "eval_wer": 0.4479122733024041,
-      "step": 315
-    },
-    {
-      "epoch": 9.14,
-      "learning_rate": 0.00036726224783861674,
-      "loss": 0.8891,
+      "epoch": 2.27,
+      "learning_rate": 9.782824698367638e-05,
+      "loss": 1.1036,
       "step": 320
     },
     {
-      "epoch": 9.43,
-      "learning_rate": 0.0003661095100864553,
-      "loss": 0.9639,
+      "epoch": 2.34,
+      "learning_rate": 9.775727466288149e-05,
+      "loss": 0.3773,
       "step": 330
     },
     {
-      "epoch": 9.71,
-      "learning_rate": 0.00036495677233429395,
-      "loss": 0.8316,
+      "epoch": 2.41,
+      "learning_rate": 9.768630234208659e-05,
+      "loss": 0.3455,
       "step": 340
     },
     {
-      "epoch": 9.99,
-      "learning_rate": 0.0003638040345821326,
-      "loss": 0.8498,
-      "step": 350
-    },
-    {
-      "epoch": 9.99,
-      "eval_loss": 1.210748553276062,
-      "eval_runtime": 8.7375,
-      "eval_samples_per_second": 7.21,
-      "eval_steps_per_second": 1.831,
-      "eval_wer": 0.47279628848587096,
+      "epoch": 2.48,
+      "learning_rate": 9.76153300212917e-05,
+      "loss": 0.43,
       "step": 350
     },
     {
-      "epoch": 10.28,
-      "learning_rate": 0.0003626512968299712,
-      "loss": 0.9453,
+      "epoch": 2.55,
+      "learning_rate": 9.754435770049681e-05,
+      "loss": 0.4461,
       "step": 360
     },
     {
-      "epoch": 10.57,
-      "learning_rate": 0.0003614985590778098,
-      "loss": 0.8499,
+      "epoch": 2.62,
+      "learning_rate": 9.747338537970193e-05,
+      "loss": 0.2987,
       "step": 370
     },
     {
-      "epoch": 10.85,
-      "learning_rate": 0.0003603458213256484,
-      "loss": 0.7205,
+      "epoch": 2.7,
+      "learning_rate": 9.740241305890702e-05,
+      "loss": 0.4987,
       "step": 380
     },
     {
-      "epoch": 10.99,
-      "eval_loss": 1.574381947517395,
-      "eval_runtime": 8.7195,
-      "eval_samples_per_second": 7.225,
-      "eval_steps_per_second": 1.835,
-      "eval_wer": 0.45255166596372837,
-      "step": 385
-    },
-    {
-      "epoch": 11.14,
-      "learning_rate": 0.00035919308357348705,
-      "loss": 1.0183,
+      "epoch": 2.77,
+      "learning_rate": 9.733144073811215e-05,
+      "loss": 0.4053,
       "step": 390
     },
     {
-      "epoch": 11.43,
-      "learning_rate": 0.0003580403458213257,
-      "loss": 0.7708,
+      "epoch": 2.84,
+      "learning_rate": 9.726046841731725e-05,
+      "loss": 0.3217,
       "step": 400
     },
     {
-      "epoch": 11.71,
-      "learning_rate": 0.00035688760806916426,
-      "loss": 0.8779,
+      "epoch": 2.91,
+      "learning_rate": 9.718949609652236e-05,
+      "loss": 0.3183,
       "step": 410
     },
     {
-      "epoch": 11.99,
-      "learning_rate": 0.0003557348703170029,
-      "loss": 0.8417,
+      "epoch": 2.98,
+      "learning_rate": 9.711852377572746e-05,
+      "loss": 0.3324,
       "step": 420
     },
     {
-      "epoch": 11.99,
-      "eval_loss": 1.468916654586792,
-      "eval_runtime": 8.7774,
-      "eval_samples_per_second": 7.178,
-      "eval_steps_per_second": 1.823,
-      "eval_wer": 0.453395191902151,
-      "step": 420
+      "epoch": 3.0,
+      "eval_loss": 0.677878201007843,
+      "eval_runtime": 8.4841,
+      "eval_samples_per_second": 7.426,
+      "eval_steps_per_second": 3.772,
+      "eval_wer": 0.3728384647827921,
+      "step": 423
     },
     {
-      "epoch": 12.28,
-      "learning_rate": 0.0003545821325648415,
-      "loss": 0.8197,
+      "epoch": 3.05,
+      "learning_rate": 9.704755145493259e-05,
+      "loss": 0.432,
       "step": 430
     },
     {
-      "epoch": 12.57,
-      "learning_rate": 0.0003534293948126801,
-      "loss": 0.7148,
+      "epoch": 3.12,
+      "learning_rate": 9.697657913413769e-05,
+      "loss": 0.2925,
       "step": 440
     },
     {
-      "epoch": 12.85,
-      "learning_rate": 0.00035227665706051873,
-      "loss": 0.7734,
+      "epoch": 3.19,
+      "learning_rate": 9.69056068133428e-05,
+      "loss": 0.3407,
       "step": 450
     },
     {
-      "epoch": 12.99,
-      "eval_loss": 1.3531099557876587,
-      "eval_runtime": 8.7009,
-      "eval_samples_per_second": 7.241,
-      "eval_steps_per_second": 1.839,
-      "eval_wer": 0.4550822437789962,
-      "step": 455
-    },
-    {
-      "epoch": 13.14,
-      "learning_rate": 0.00035112391930835736,
-      "loss": 0.7908,
+      "epoch": 3.26,
+      "learning_rate": 9.683463449254791e-05,
+      "loss": 0.4149,
       "step": 460
     },
     {
-      "epoch": 13.43,
-      "learning_rate": 0.000349971181556196,
-      "loss": 0.7157,
+      "epoch": 3.33,
+      "learning_rate": 9.676366217175303e-05,
+      "loss": 0.3774,
       "step": 470
     },
     {
-      "epoch": 13.71,
-      "learning_rate": 0.00034881844380403457,
-      "loss": 0.7873,
+      "epoch": 3.4,
+      "learning_rate": 9.669268985095812e-05,
+      "loss": 0.3061,
       "step": 480
     },
     {
-      "epoch": 13.99,
-      "learning_rate": 0.00034766570605187326,
-      "loss": 0.7762,
-      "step": 490
-    },
-    {
-      "epoch": 13.99,
-      "eval_loss": 1.2924444675445557,
-      "eval_runtime": 8.7984,
-      "eval_samples_per_second": 7.16,
-      "eval_steps_per_second": 1.819,
-      "eval_wer": 0.4664698439477014,
+      "epoch": 3.48,
+      "learning_rate": 9.662171753016324e-05,
+      "loss": 0.3265,
       "step": 490
     },
     {
-      "epoch": 14.28,
-      "learning_rate": 0.00034651296829971183,
-      "loss": 0.8195,
+      "epoch": 3.55,
+      "learning_rate": 9.655074520936835e-05,
+      "loss": 0.3311,
       "step": 500
     },
     {
-      "epoch": 14.57,
-      "learning_rate": 0.00034536023054755047,
-      "loss": 0.7457,
+      "epoch": 3.62,
+      "learning_rate": 9.647977288857346e-05,
+      "loss": 0.2818,
       "step": 510
     },
     {
-      "epoch": 14.85,
-      "learning_rate": 0.00034420749279538904,
-      "loss": 0.6812,
+      "epoch": 3.69,
+      "learning_rate": 9.640880056777856e-05,
+      "loss": 0.3543,
       "step": 520
     },
     {
-      "epoch": 14.99,
-      "eval_loss": 1.0827070474624634,
-      "eval_runtime": 8.7923,
-      "eval_samples_per_second": 7.165,
-      "eval_steps_per_second": 1.82,
-      "eval_wer": 0.40995360607338677,
-      "step": 525
-    },
-    {
-      "epoch": 15.14,
-      "learning_rate": 0.0003430547550432277,
-      "loss": 0.7591,
+      "epoch": 3.76,
+      "learning_rate": 9.633782824698369e-05,
+      "loss": 0.5701,
       "step": 530
     },
     {
-      "epoch": 15.43,
-      "learning_rate": 0.0003419020172910663,
-      "loss": 0.7642,
+      "epoch": 3.83,
+      "learning_rate": 9.626685592618879e-05,
+      "loss": 0.3405,
       "step": 540
     },
     {
-      "epoch": 15.71,
-      "learning_rate": 0.0003407492795389049,
-      "loss": 1.27,
+      "epoch": 3.9,
+      "learning_rate": 9.61958836053939e-05,
+      "loss": 0.3205,
       "step": 550
     },
     {
-      "epoch": 15.99,
-      "learning_rate": 0.0003395965417867435,
-      "loss": 0.7245,
+      "epoch": 3.97,
+      "learning_rate": 9.6124911284599e-05,
+      "loss": 0.4244,
       "step": 560
     },
     {
-      "epoch": 15.99,
-      "eval_loss": 1.4069534540176392,
-      "eval_runtime": 8.8121,
-      "eval_samples_per_second": 7.149,
-      "eval_steps_per_second": 1.816,
-      "eval_wer": 0.43525938422606497,
-      "step": 560
+      "epoch": 4.0,
+      "eval_loss": 0.7402982115745544,
+      "eval_runtime": 8.6329,
+      "eval_samples_per_second": 7.298,
+      "eval_steps_per_second": 3.707,
+      "eval_wer": 0.37368199072121466,
+      "step": 564
     },
     {
-      "epoch": 16.28,
-      "learning_rate": 0.00033844380403458214,
-      "loss": 0.7721,
+      "epoch": 4.04,
+      "learning_rate": 9.605393896380413e-05,
+      "loss": 0.3539,
       "step": 570
     },
     {
-      "epoch": 16.57,
-      "learning_rate": 0.0003372910662824208,
-      "loss": 0.6945,
+      "epoch": 4.11,
+      "learning_rate": 9.598296664300923e-05,
+      "loss": 0.2606,
       "step": 580
     },
     {
-      "epoch": 16.85,
-      "learning_rate": 0.00033613832853025935,
-      "loss": 0.6508,
+      "epoch": 4.18,
+      "learning_rate": 9.591199432221434e-05,
+      "loss": 0.3101,
       "step": 590
     },
     {
-      "epoch": 16.99,
-      "eval_loss": 1.052022099494934,
-      "eval_runtime": 8.7064,
-      "eval_samples_per_second": 7.236,
-      "eval_steps_per_second": 1.838,
-      "eval_wer": 0.4086883171657528,
-      "step": 595
-    },
-    {
-      "epoch": 17.14,
-      "learning_rate": 0.00033498559077809804,
-      "loss": 0.735,
+      "epoch": 4.26,
+      "learning_rate": 9.584102200141945e-05,
+      "loss": 0.3915,
       "step": 600
     },
     {
-      "epoch": 17.43,
-      "learning_rate": 0.0003338328530259366,
-      "loss": 0.6526,
+      "epoch": 4.33,
+      "learning_rate": 9.577004968062456e-05,
+      "loss": 0.3793,
       "step": 610
     },
     {
-      "epoch": 17.71,
-      "learning_rate": 0.00033268011527377525,
-      "loss": 0.6955,
+      "epoch": 4.4,
+      "learning_rate": 9.569907735982968e-05,
+      "loss": 0.2797,
       "step": 620
     },
     {
-      "epoch": 17.99,
-      "learning_rate": 0.0003315273775216138,
-      "loss": 0.7144,
-      "step": 630
-    },
-    {
-      "epoch": 17.99,
-      "eval_loss": 1.0728662014007568,
-      "eval_runtime": 8.8004,
-      "eval_samples_per_second": 7.159,
-      "eval_steps_per_second": 1.818,
-      "eval_wer": 0.42091944327288067,
+      "epoch": 4.47,
+      "learning_rate": 9.563520227111426e-05,
+      "loss": 0.4823,
       "step": 630
     },
     {
-      "epoch": 18.28,
-      "learning_rate": 0.0003303746397694525,
-      "loss": 0.6136,
+      "epoch": 4.54,
+      "learning_rate": 9.556422995031939e-05,
+      "loss": 0.4453,
       "step": 640
     },
     {
-      "epoch": 18.57,
-      "learning_rate": 0.0003292219020172911,
-      "loss": 0.6545,
+      "epoch": 4.61,
+      "learning_rate": 9.549325762952449e-05,
+      "loss": 0.3081,
       "step": 650
     },
     {
-      "epoch": 18.85,
-      "learning_rate": 0.00032806916426512966,
-      "loss": 0.6566,
+      "epoch": 4.68,
+      "learning_rate": 9.54222853087296e-05,
+      "loss": 0.3722,
       "step": 660
     },
     {
-      "epoch": 18.99,
-      "eval_loss": 1.1671946048736572,
-      "eval_runtime": 8.7903,
-      "eval_samples_per_second": 7.167,
-      "eval_steps_per_second": 1.82,
-      "eval_wer": 0.4053142134120624,
-      "step": 665
-    },
-    {
-      "epoch": 19.14,
-      "learning_rate": 0.00032691642651296835,
-      "loss": 0.7112,
+      "epoch": 4.75,
+      "learning_rate": 9.535131298793471e-05,
+      "loss": 0.4097,
       "step": 670
     },
     {
-      "epoch": 19.43,
-      "learning_rate": 0.0003257636887608069,
-      "loss": 0.6853,
+      "epoch": 4.82,
+      "learning_rate": 9.528034066713982e-05,
+      "loss": 0.3396,
       "step": 680
     },
     {
-      "epoch": 19.71,
-      "learning_rate": 0.00032461095100864556,
-      "loss": 0.7222,
+      "epoch": 4.89,
+      "learning_rate": 9.520936834634492e-05,
+      "loss": 0.3178,
       "step": 690
     },
     {
-      "epoch": 19.99,
-      "learning_rate": 0.00032345821325648413,
-      "loss": 0.5802,
+      "epoch": 4.96,
+      "learning_rate": 9.513839602555004e-05,
+      "loss": 0.5234,
       "step": 700
     },
     {
-      "epoch": 19.99,
-      "eval_loss": 1.0128839015960693,
-      "eval_runtime": 8.823,
-      "eval_samples_per_second": 7.14,
-      "eval_steps_per_second": 1.813,
-      "eval_wer": 0.4015183466891607,
-      "step": 700
+      "epoch": 5.0,
+      "eval_loss": 0.8086081743240356,
+      "eval_runtime": 8.3374,
+      "eval_samples_per_second": 7.556,
+      "eval_steps_per_second": 3.838,
+      "eval_wer": 0.3534373681990721,
+      "step": 705
     },
     {
-      "epoch": 20.28,
-      "learning_rate": 0.0003223054755043228,
-      "loss": 0.6989,
+      "epoch": 5.04,
+      "learning_rate": 9.506742370475515e-05,
+      "loss": 0.5366,
       "step": 710
     },
     {
-      "epoch": 20.57,
-      "learning_rate": 0.0003211527377521614,
-      "loss": 0.6103,
+      "epoch": 5.11,
+      "learning_rate": 9.499645138396026e-05,
+      "loss": 0.4794,
       "step": 720
     },
     {
-      "epoch": 20.85,
-      "learning_rate": 0.00032,
-      "loss": 0.5924,
+      "epoch": 5.18,
+      "learning_rate": 9.492547906316536e-05,
+      "loss": 0.3289,
       "step": 730
     },
     {
-      "epoch": 20.99,
-      "eval_loss": 1.0761842727661133,
-      "eval_runtime": 8.6827,
-      "eval_samples_per_second": 7.256,
-      "eval_steps_per_second": 1.843,
-      "eval_wer": 0.40067482075073807,
-      "step": 735
-    },
-    {
-      "epoch": 21.14,
-      "learning_rate": 0.0003188472622478386,
-      "loss": 0.641,
+      "epoch": 5.25,
+      "learning_rate": 9.485450674237047e-05,
+      "loss": 0.4384,
       "step": 740
     },
     {
-      "epoch": 21.43,
-      "learning_rate": 0.0003176945244956773,
-      "loss": 0.7465,
+      "epoch": 5.32,
+      "learning_rate": 9.478353442157559e-05,
+      "loss": 0.3467,
       "step": 750
     },
     {
-      "epoch": 21.71,
-      "learning_rate": 0.00031654178674351587,
-      "loss": 0.6215,
+      "epoch": 5.39,
+      "learning_rate": 9.47125621007807e-05,
+      "loss": 0.2536,
       "step": 760
     },
     {
-      "epoch": 21.99,
-      "learning_rate": 0.00031538904899135444,
-      "loss": 0.7051,
-      "step": 770
-    },
-    {
-      "epoch": 21.99,
-      "eval_loss": 1.025338053703308,
-      "eval_runtime": 8.7155,
-      "eval_samples_per_second": 7.229,
-      "eval_steps_per_second": 1.836,
-      "eval_wer": 0.4027836355967946,
+      "epoch": 5.46,
+      "learning_rate": 9.464158977998581e-05,
+      "loss": 0.3686,
       "step": 770
     },
     {
-      "epoch": 22.28,
-      "learning_rate": 0.00031423631123919313,
-      "loss": 0.6773,
+      "epoch": 5.53,
+      "learning_rate": 9.457061745919093e-05,
+      "loss": 0.3635,
       "step": 780
     },
     {
-      "epoch": 22.57,
-      "learning_rate": 0.0003130835734870317,
-      "loss": 0.6137,
+      "epoch": 5.6,
+      "learning_rate": 9.449964513839604e-05,
+      "loss": 0.3513,
       "step": 790
     },
     {
-      "epoch": 22.85,
-      "learning_rate": 0.00031193083573487034,
-      "loss": 0.5669,
+      "epoch": 5.67,
+      "learning_rate": 9.442867281760114e-05,
+      "loss": 0.341,
       "step": 800
     },
     {
-      "epoch": 22.99,
-      "eval_loss": 1.0526078939437866,
-      "eval_runtime": 8.675,
-      "eval_samples_per_second": 7.262,
-      "eval_steps_per_second": 1.844,
-      "eval_wer": 0.41881062842682415,
-      "step": 805
-    },
-    {
-      "epoch": 23.14,
-      "learning_rate": 0.0003107780979827089,
-      "loss": 0.5976,
+      "epoch": 5.74,
+      "learning_rate": 9.435770049680625e-05,
+      "loss": 0.4465,
       "step": 810
     },
     {
-      "epoch": 23.43,
-      "learning_rate": 0.0003096253602305476,
-      "loss": 0.6745,
+      "epoch": 5.82,
+      "learning_rate": 9.428672817601136e-05,
+      "loss": 0.4103,
       "step": 820
     },
     {
-      "epoch": 23.71,
-      "learning_rate": 0.0003084726224783862,
-      "loss": 0.7289,
+      "epoch": 5.89,
+      "learning_rate": 9.421575585521648e-05,
+      "loss": 0.2147,
       "step": 830
     },
     {
-      "epoch": 23.99,
-      "learning_rate": 0.0003073198847262248,
-      "loss": 0.6209,
+      "epoch": 5.96,
+      "learning_rate": 9.414478353442157e-05,
+      "loss": 0.3339,
       "step": 840
     },
     {
-      "epoch": 23.99,
-      "eval_loss": 1.0177326202392578,
-      "eval_runtime": 8.832,
-      "eval_samples_per_second": 7.133,
-      "eval_steps_per_second": 1.812,
-      "eval_wer": 0.4213412062420919,
-      "step": 840
+      "epoch": 6.0,
+      "eval_loss": 0.718730092048645,
+      "eval_runtime": 8.4856,
+      "eval_samples_per_second": 7.424,
+      "eval_steps_per_second": 3.771,
+      "eval_wer": 0.3618726275832982,
+      "step": 846
     },
     {
-      "epoch": 24.28,
-      "learning_rate": 0.00030616714697406344,
-      "loss": 0.6383,
+      "epoch": 6.03,
+      "learning_rate": 9.40738112136267e-05,
+      "loss": 0.4226,
       "step": 850
     },
     {
-      "epoch": 24.57,
-      "learning_rate": 0.00030501440922190207,
-      "loss": 0.5804,
+      "epoch": 6.1,
+      "learning_rate": 9.40028388928318e-05,
+      "loss": 0.3365,
       "step": 860
     },
     {
-      "epoch": 24.85,
-      "learning_rate": 0.00030386167146974065,
-      "loss": 0.635,
+      "epoch": 6.17,
+      "learning_rate": 9.393186657203691e-05,
+      "loss": 0.3646,
       "step": 870
     },
     {
-      "epoch": 24.99,
-      "eval_loss": 0.929938554763794,
-      "eval_runtime": 8.8255,
-      "eval_samples_per_second": 7.138,
-      "eval_steps_per_second": 1.813,
-      "eval_wer": 0.401940109658372,
-      "step": 875
-    },
-    {
-      "epoch": 25.14,
-      "learning_rate": 0.0003027089337175792,
-      "loss": 0.74,
+      "epoch": 6.24,
+      "learning_rate": 9.386089425124201e-05,
+      "loss": 0.394,
       "step": 880
     },
     {
-      "epoch": 25.43,
-      "learning_rate": 0.0003015561959654179,
-      "loss": 0.6235,
+      "epoch": 6.31,
+      "learning_rate": 9.378992193044714e-05,
+      "loss": 0.3678,
       "step": 890
     },
     {
-      "epoch": 25.71,
-      "learning_rate": 0.0003004034582132565,
-      "loss": 0.6219,
+      "epoch": 6.38,
+      "learning_rate": 9.371894960965224e-05,
+      "loss": 0.2709,
       "step": 900
     },
     {
-      "epoch": 25.99,
-      "learning_rate": 0.0002992507204610951,
-      "loss": 0.5914,
-      "step": 910
-    },
-    {
-      "epoch": 25.99,
-      "eval_loss": 1.0058447122573853,
-      "eval_runtime": 8.6702,
-      "eval_samples_per_second": 7.266,
-      "eval_steps_per_second": 1.845,
-      "eval_wer": 0.4141712357654998,
+      "epoch": 6.45,
+      "learning_rate": 9.364797728885735e-05,
+      "loss": 0.3355,
       "step": 910
     },
     {
-      "epoch": 26.28,
-      "learning_rate": 0.00029809798270893375,
-      "loss": 0.6127,
+      "epoch": 6.52,
+      "learning_rate": 9.357700496806246e-05,
+      "loss": 0.3858,
       "step": 920
     },
     {
-      "epoch": 26.57,
-      "learning_rate": 0.0002969452449567724,
-      "loss": 0.5335,
+      "epoch": 6.6,
+      "learning_rate": 9.350603264726758e-05,
+      "loss": 0.3261,
       "step": 930
     },
     {
-      "epoch": 26.85,
-      "learning_rate": 0.00029579250720461096,
-      "loss": 0.5983,
+      "epoch": 6.67,
+      "learning_rate": 9.343506032647267e-05,
+      "loss": 0.3413,
       "step": 940
     },
     {
-      "epoch": 26.99,
-      "eval_loss": 0.9719786643981934,
-      "eval_runtime": 8.6673,
-      "eval_samples_per_second": 7.269,
-      "eval_steps_per_second": 1.846,
-      "eval_wer": 0.4141712357654998,
-      "step": 945
-    },
-    {
-      "epoch": 27.14,
-      "learning_rate": 0.0002946397694524496,
-      "loss": 0.559,
+      "epoch": 6.74,
+      "learning_rate": 9.336408800567779e-05,
+      "loss": 0.4866,
       "step": 950
     },
     {
-      "epoch": 27.43,
-      "learning_rate": 0.0002934870317002882,
-      "loss": 0.6208,
+      "epoch": 6.81,
+      "learning_rate": 9.32931156848829e-05,
+      "loss": 0.391,
       "step": 960
     },
     {
-      "epoch": 27.71,
-      "learning_rate": 0.00029233429394812685,
-      "loss": 0.5678,
+      "epoch": 6.88,
+      "learning_rate": 9.322214336408801e-05,
+      "loss": 0.3667,
       "step": 970
     },
     {
-      "epoch": 27.99,
-      "learning_rate": 0.0002911815561959654,
-      "loss": 0.5631,
+      "epoch": 6.95,
+      "learning_rate": 9.315117104329311e-05,
+      "loss": 0.5016,
       "step": 980
     },
     {
-      "epoch": 27.99,
-      "eval_loss": 0.8982506990432739,
-      "eval_runtime": 8.7362,
-      "eval_samples_per_second": 7.211,
-      "eval_steps_per_second": 1.831,
-      "eval_wer": 0.4027836355967946,
-      "step": 980
+      "epoch": 7.0,
+      "eval_loss": 0.8582101464271545,
+      "eval_runtime": 8.4772,
+      "eval_samples_per_second": 7.432,
+      "eval_steps_per_second": 3.775,
+      "eval_wer": 0.360185575706453,
+      "step": 987
     },
     {
-      "epoch": 28.28,
-      "learning_rate": 0.000290028818443804,
-      "loss": 0.5503,
+      "epoch": 7.02,
+      "learning_rate": 9.308019872249824e-05,
+      "loss": 0.3826,
       "step": 990
     },
     {
-      "epoch": 28.57,
-      "learning_rate": 0.0002888760806916427,
-      "loss": 0.5238,
+      "epoch": 7.09,
+      "learning_rate": 9.300922640170334e-05,
+      "loss": 0.3774,
       "step": 1000
     },
     {
-      "epoch": 28.85,
-      "learning_rate": 0.00028772334293948127,
-      "loss": 0.552,
+      "epoch": 7.16,
+      "learning_rate": 9.293825408090845e-05,
+      "loss": 0.4231,
       "step": 1010
     },
     {
-      "epoch": 28.99,
-      "eval_loss": 0.914803683757782,
-      "eval_runtime": 8.7166,
-      "eval_samples_per_second": 7.228,
-      "eval_steps_per_second": 1.836,
-      "eval_wer": 0.41838886545761284,
-      "step": 1015
-    },
-    {
-      "epoch": 29.14,
-      "learning_rate": 0.0002865706051873199,
-      "loss": 0.5816,
+      "epoch": 7.23,
+      "learning_rate": 9.286728176011355e-05,
+      "loss": 1.3863,
       "step": 1020
     },
     {
-      "epoch": 29.43,
-      "learning_rate": 0.00028541786743515853,
-      "loss": 0.5295,
+      "epoch": 7.3,
+      "learning_rate": 9.279630943931868e-05,
+      "loss": 0.3564,
       "step": 1030
     },
     {
-      "epoch": 29.71,
-      "learning_rate": 0.00028426512968299716,
-      "loss": 0.5742,
+      "epoch": 7.38,
+      "learning_rate": 9.272533711852377e-05,
+      "loss": 0.2481,
       "step": 1040
     },
     {
-      "epoch": 29.99,
-      "learning_rate": 0.00028311239193083574,
-      "loss": 0.5213,
+      "epoch": 7.45,
+      "learning_rate": 9.265436479772889e-05,
+      "loss": 0.391,
       "step": 1050
     },
     {
-      "epoch": 29.99,
-      "eval_loss": 1.0816500186920166,
-      "eval_runtime": 8.9248,
-      "eval_samples_per_second": 7.059,
-      "eval_steps_per_second": 1.793,
-      "eval_wer": 0.4141712357654998,
-      "step": 1050
-    },
-    {
-      "epoch": 30.28,
-      "learning_rate": 0.00028195965417867437,
-      "loss": 0.681,
+      "epoch": 7.52,
+      "learning_rate": 9.2583392476934e-05,
+      "loss": 0.4603,
       "step": 1060
     },
     {
-      "epoch": 30.57,
-      "learning_rate": 0.000280806916426513,
-      "loss": 0.5139,
+      "epoch": 7.59,
+      "learning_rate": 9.251242015613911e-05,
+      "loss": 0.3406,
       "step": 1070
     },
     {
-      "epoch": 30.85,
-      "learning_rate": 0.00027965417867435163,
-      "loss": 0.5387,
+      "epoch": 7.66,
+      "learning_rate": 9.244144783534423e-05,
+      "loss": 0.2895,
       "step": 1080
     },
     {
-      "epoch": 30.99,
-      "eval_loss": 0.9432059526443481,
-      "eval_runtime": 8.8063,
-      "eval_samples_per_second": 7.154,
-      "eval_steps_per_second": 1.817,
-      "eval_wer": 0.41881062842682415,
-      "step": 1085
-    },
-    {
-      "epoch": 31.14,
-      "learning_rate": 0.0002785014409221902,
-      "loss": 0.5541,
+      "epoch": 7.73,
+      "learning_rate": 9.237047551454932e-05,
+      "loss": 0.3409,
       "step": 1090
     },
     {
-      "epoch": 31.43,
-      "learning_rate": 0.00027734870317002884,
-      "loss": 0.6006,
+      "epoch": 7.8,
+      "learning_rate": 9.229950319375444e-05,
+      "loss": 0.4088,
       "step": 1100
     },
     {
-      "epoch": 31.71,
-      "learning_rate": 0.00027619596541786747,
-      "loss": 0.6104,
+      "epoch": 7.87,
+      "learning_rate": 9.222853087295955e-05,
+      "loss": 0.333,
       "step": 1110
     },
     {
-      "epoch": 31.99,
-      "learning_rate": 0.00027504322766570605,
-      "loss": 0.5276,
+      "epoch": 7.94,
+      "learning_rate": 9.215755855216466e-05,
+      "loss": 0.3376,
       "step": 1120
     },
     {
-      "epoch": 31.99,
-      "eval_loss": 1.1206682920455933,
-      "eval_runtime": 8.7507,
-      "eval_samples_per_second": 7.199,
-      "eval_steps_per_second": 1.828,
-      "eval_wer": 0.40067482075073807,
-      "step": 1120
+      "epoch": 8.0,
+      "eval_loss": 0.880143940448761,
+      "eval_runtime": 8.4956,
+      "eval_samples_per_second": 7.416,
+      "eval_steps_per_second": 3.767,
+      "eval_wer": 0.3673555461830451,
+      "step": 1128
     },
     {
-      "epoch": 32.28,
-      "learning_rate": 0.0002738904899135447,
-      "loss": 0.511,
+      "epoch": 8.01,
+      "learning_rate": 9.208658623136976e-05,
+      "loss": 0.531,
       "step": 1130
     },
     {
-      "epoch": 32.57,
-      "learning_rate": 0.0002727377521613833,
-      "loss": 0.4948,
+      "epoch": 8.09,
+      "learning_rate": 9.201561391057489e-05,
+      "loss": 0.2671,
       "step": 1140
     },
     {
-      "epoch": 32.85,
-      "learning_rate": 0.00027158501440922194,
-      "loss": 0.5778,
+      "epoch": 8.16,
+      "learning_rate": 9.194464158977999e-05,
+      "loss": 0.3075,
       "step": 1150
     },
     {
-      "epoch": 32.99,
-      "eval_loss": 0.9253532886505127,
-      "eval_runtime": 8.8188,
-      "eval_samples_per_second": 7.144,
-      "eval_steps_per_second": 1.814,
-      "eval_wer": 0.4150147617039224,
-      "step": 1155
-    },
-    {
-      "epoch": 33.14,
-      "learning_rate": 0.0002704322766570605,
-      "loss": 0.551,
+      "epoch": 8.23,
+      "learning_rate": 9.18736692689851e-05,
+      "loss": 0.4256,
       "step": 1160
     },
     {
-      "epoch": 33.43,
-      "learning_rate": 0.00026927953890489915,
-      "loss": 0.5115,
+      "epoch": 8.3,
+      "learning_rate": 9.180269694819021e-05,
+      "loss": 0.3646,
       "step": 1170
     },
     {
-      "epoch": 33.71,
-      "learning_rate": 0.0002681268011527378,
-      "loss": 0.6795,
+      "epoch": 8.37,
+      "learning_rate": 9.173172462739533e-05,
+      "loss": 0.3085,
       "step": 1180
     },
     {
-      "epoch": 33.99,
-      "learning_rate": 0.0002669740634005764,
-      "loss": 0.5001,
+      "epoch": 8.44,
+      "learning_rate": 9.166075230660042e-05,
+      "loss": 0.3801,
       "step": 1190
     },
     {
-      "epoch": 33.99,
-      "eval_loss": 1.0393145084381104,
-      "eval_runtime": 8.783,
-      "eval_samples_per_second": 7.173,
-      "eval_steps_per_second": 1.822,
-      "eval_wer": 0.4192323913960354,
-      "step": 1190
-    },
-    {
-      "epoch": 34.28,
-      "learning_rate": 0.000265821325648415,
-      "loss": 0.5146,
+      "epoch": 8.51,
+      "learning_rate": 9.158977998580554e-05,
+      "loss": 0.36,
       "step": 1200
     },
     {
-      "epoch": 34.57,
-      "learning_rate": 0.0002646685878962536,
-      "loss": 0.5165,
+      "epoch": 8.58,
+      "learning_rate": 9.151880766501065e-05,
+      "loss": 0.4412,
       "step": 1210
     },
     {
-      "epoch": 34.85,
-      "learning_rate": 0.00026351585014409225,
-      "loss": 0.5329,
+      "epoch": 8.65,
+      "learning_rate": 9.144783534421576e-05,
+      "loss": 0.4121,
       "step": 1220
     },
     {
-      "epoch": 34.99,
-      "eval_loss": 0.9108927249908447,
-      "eval_runtime": 8.8479,
-      "eval_samples_per_second": 7.12,
-      "eval_steps_per_second": 1.808,
-      "eval_wer": 0.39645719105862504,
-      "step": 1225
-    },
-    {
-      "epoch": 35.14,
-      "learning_rate": 0.0002623631123919308,
-      "loss": 0.5703,
+      "epoch": 8.72,
+      "learning_rate": 9.137686302342086e-05,
+      "loss": 0.4252,
       "step": 1230
     },
     {
-      "epoch": 35.43,
-      "learning_rate": 0.00026121037463976946,
-      "loss": 0.5296,
+      "epoch": 8.79,
+      "learning_rate": 9.130589070262599e-05,
+      "loss": 0.4037,
       "step": 1240
     },
     {
-      "epoch": 35.71,
-      "learning_rate": 0.0002600576368876081,
-      "loss": 0.5212,
+      "epoch": 8.87,
+      "learning_rate": 9.123491838183109e-05,
+      "loss": 0.3078,
       "step": 1250
     },
     {
-      "epoch": 35.99,
-      "learning_rate": 0.0002589048991354467,
-      "loss": 0.5168,
+      "epoch": 8.94,
+      "learning_rate": 9.11639460610362e-05,
+      "loss": 0.3507,
       "step": 1260
     },
     {
-      "epoch": 35.99,
-      "eval_loss": 0.8983320593833923,
-      "eval_runtime": 8.8493,
-      "eval_samples_per_second": 7.119,
-      "eval_steps_per_second": 1.808,
-      "eval_wer": 0.429776465626318,
-      "step": 1260
+      "epoch": 9.0,
+      "eval_loss": 0.8524307012557983,
+      "eval_runtime": 8.4497,
+      "eval_samples_per_second": 7.456,
+      "eval_steps_per_second": 3.787,
+      "eval_wer": 0.35596794601433995,
+      "step": 1269
     },
     {
-      "epoch": 36.28,
-      "learning_rate": 0.0002577521613832853,
-      "loss": 0.528,
+      "epoch": 9.01,
+      "learning_rate": 9.10929737402413e-05,
+      "loss": 0.3534,
       "step": 1270
     },
     {
-      "epoch": 36.57,
-      "learning_rate": 0.00025659942363112393,
-      "loss": 0.4487,
+      "epoch": 9.08,
+      "learning_rate": 9.102200141944643e-05,
+      "loss": 0.3294,
       "step": 1280
     },
     {
-      "epoch": 36.85,
-      "learning_rate": 0.00025544668587896256,
-      "loss": 0.4918,
+      "epoch": 9.15,
+      "learning_rate": 9.095102909865152e-05,
+      "loss": 0.454,
       "step": 1290
     },
     {
-      "epoch": 36.99,
-      "eval_loss": 0.8411929607391357,
-      "eval_runtime": 8.8153,
-      "eval_samples_per_second": 7.147,
-      "eval_steps_per_second": 1.815,
-      "eval_wer": 0.4086883171657528,
-      "step": 1295
-    },
-    {
-      "epoch": 37.14,
-      "learning_rate": 0.0002542939481268012,
-      "loss": 0.5363,
+      "epoch": 9.22,
+      "learning_rate": 9.088005677785664e-05,
+      "loss": 0.4106,
       "step": 1300
     },
     {
-      "epoch": 37.43,
-      "learning_rate": 0.00025314121037463977,
-      "loss": 0.4954,
+      "epoch": 9.29,
+      "learning_rate": 9.080908445706175e-05,
+      "loss": 0.5058,
       "step": 1310
     },
     {
-      "epoch": 37.71,
-      "learning_rate": 0.0002519884726224784,
-      "loss": 0.4878,
+      "epoch": 9.36,
+      "learning_rate": 9.073811213626686e-05,
+      "loss": 0.2653,
       "step": 1320
     },
     {
-      "epoch": 37.99,
-      "learning_rate": 0.00025083573487031703,
-      "loss": 0.5651,
-      "step": 1330
-    },
-    {
-      "epoch": 37.99,
-      "eval_loss": 0.8560371398925781,
-      "eval_runtime": 9.0193,
-      "eval_samples_per_second": 6.985,
-      "eval_steps_per_second": 1.774,
-      "eval_wer": 0.42176296921130324,
+      "epoch": 9.43,
+      "learning_rate": 9.066713981547196e-05,
+      "loss": 0.352,
       "step": 1330
     },
     {
-      "epoch": 38.28,
-      "learning_rate": 0.0002496829971181556,
-      "loss": 0.5888,
+      "epoch": 9.5,
+      "learning_rate": 9.059616749467708e-05,
+      "loss": 0.3728,
       "step": 1340
     },
     {
-      "epoch": 38.57,
-      "learning_rate": 0.00024853025936599424,
-      "loss": 0.485,
+      "epoch": 9.57,
+      "learning_rate": 9.052519517388219e-05,
+      "loss": 0.4493,
       "step": 1350
     },
     {
-      "epoch": 38.85,
-      "learning_rate": 0.00024737752161383287,
-      "loss": 0.438,
+      "epoch": 9.65,
+      "learning_rate": 9.04542228530873e-05,
+      "loss": 0.3085,
       "step": 1360
     },
     {
-      "epoch": 38.99,
-      "eval_loss": 0.8555735349655151,
-      "eval_runtime": 8.8578,
-      "eval_samples_per_second": 7.112,
-      "eval_steps_per_second": 1.806,
-      "eval_wer": 0.4171235765499789,
-      "step": 1365
-    },
-    {
-      "epoch": 39.14,
-      "learning_rate": 0.0002462247838616715,
-      "loss": 0.5556,
+      "epoch": 9.72,
+      "learning_rate": 9.038325053229241e-05,
+      "loss": 0.3595,
       "step": 1370
     },
     {
-      "epoch": 39.43,
-      "learning_rate": 0.0002450720461095101,
-      "loss": 0.4795,
+      "epoch": 9.79,
+      "learning_rate": 9.031227821149753e-05,
+      "loss": 0.5616,
       "step": 1380
     },
     {
-      "epoch": 39.71,
-      "learning_rate": 0.00024391930835734874,
-      "loss": 0.4785,
+      "epoch": 9.86,
+      "learning_rate": 9.024130589070264e-05,
+      "loss": 0.2889,
       "step": 1390
     },
     {
-      "epoch": 39.99,
-      "learning_rate": 0.00024276657060518734,
-      "loss": 0.4808,
+      "epoch": 9.93,
+      "learning_rate": 9.017033356990774e-05,
+      "loss": 0.2847,
       "step": 1400
     },
     {
-      "epoch": 39.99,
-      "eval_loss": 0.8319878578186035,
-      "eval_runtime": 8.7692,
-      "eval_samples_per_second": 7.184,
-      "eval_steps_per_second": 1.825,
-      "eval_wer": 0.4175453395191902,
-      "step": 1400
+      "epoch": 10.0,
+      "learning_rate": 9.009936124911285e-05,
+      "loss": 0.4844,
+      "step": 1410
     },
     {
-      "epoch": 40.28,
-      "learning_rate": 0.00024161383285302597,
-      "loss": 0.5304,
+      "epoch": 10.0,
+      "eval_loss": 0.7152120471000671,
+      "eval_runtime": 8.3589,
+      "eval_samples_per_second": 7.537,
+      "eval_steps_per_second": 3.828,
+      "eval_wer": 0.36482496836777734,
       "step": 1410
     },
     {
-      "epoch": 40.57,
-      "learning_rate": 0.00024046109510086458,
-      "loss": 0.4896,
+      "epoch": 10.07,
+      "learning_rate": 9.002838892831796e-05,
+      "loss": 0.5095,
       "step": 1420
     },
     {
-      "epoch": 40.85,
-      "learning_rate": 0.00023930835734870315,
-      "loss": 0.5372,
+      "epoch": 10.14,
+      "learning_rate": 8.995741660752308e-05,
+      "loss": 0.3538,
       "step": 1430
     },
     {
-      "epoch": 40.99,
-      "eval_loss": 0.9745492339134216,
-      "eval_runtime": 8.7927,
-      "eval_samples_per_second": 7.165,
-      "eval_steps_per_second": 1.82,
-      "eval_wer": 0.39561366512020246,
-      "step": 1435
-    },
-    {
-      "epoch": 41.14,
-      "learning_rate": 0.0002381556195965418,
-      "loss": 0.4811,
+      "epoch": 10.21,
+      "learning_rate": 8.988644428672818e-05,
+      "loss": 0.5236,
       "step": 1440
     },
     {
-      "epoch": 41.43,
-      "learning_rate": 0.0002370028818443804,
-      "loss": 0.4714,
+      "epoch": 10.28,
+      "learning_rate": 8.98154719659333e-05,
+      "loss": 0.5021,
       "step": 1450
     },
     {
-      "epoch": 41.71,
-      "learning_rate": 0.00023585014409221905,
-      "loss": 0.5127,
+      "epoch": 10.35,
+      "learning_rate": 8.97444996451384e-05,
+      "loss": 0.3653,
       "step": 1460
     },
     {
-      "epoch": 41.99,
-      "learning_rate": 0.00023469740634005762,
-      "loss": 0.4814,
-      "step": 1470
-    },
-    {
-      "epoch": 41.99,
-      "eval_loss": 0.8032535910606384,
-      "eval_runtime": 8.8256,
-      "eval_samples_per_second": 7.138,
-      "eval_steps_per_second": 1.813,
-      "eval_wer": 0.4120624209194433,
+      "epoch": 10.43,
+      "learning_rate": 8.967352732434351e-05,
+      "loss": 0.3093,
       "step": 1470
     },
     {
-      "epoch": 42.28,
-      "learning_rate": 0.00023354466858789628,
-      "loss": 0.4302,
+      "epoch": 10.5,
+      "learning_rate": 8.960255500354861e-05,
+      "loss": 0.4504,
       "step": 1480
     },
     {
-      "epoch": 42.57,
-      "learning_rate": 0.00023239193083573489,
-      "loss": 0.4579,
+      "epoch": 10.57,
+      "learning_rate": 8.953158268275374e-05,
+      "loss": 0.4039,
       "step": 1490
     },
     {
-      "epoch": 42.85,
-      "learning_rate": 0.00023123919308357352,
-      "loss": 0.4416,
+      "epoch": 10.64,
+      "learning_rate": 8.946061036195884e-05,
+      "loss": 0.1843,
       "step": 1500
     },
     {
-      "epoch": 42.99,
-      "eval_loss": 0.8194773197174072,
-      "eval_runtime": 8.7558,
-      "eval_samples_per_second": 7.195,
-      "eval_steps_per_second": 1.827,
-      "eval_wer": 0.39898776887389287,
-      "step": 1505
-    },
-    {
-      "epoch": 43.14,
-      "learning_rate": 0.00023008645533141212,
-      "loss": 0.4628,
+      "epoch": 10.71,
+      "learning_rate": 8.938963804116395e-05,
+      "loss": 0.3346,
       "step": 1510
     },
     {
-      "epoch": 43.43,
-      "learning_rate": 0.00022893371757925075,
-      "loss": 0.4765,
+      "epoch": 10.78,
+      "learning_rate": 8.931866572036906e-05,
+      "loss": 0.3941,
       "step": 1520
     },
     {
-      "epoch": 43.71,
-      "learning_rate": 0.00022778097982708936,
-      "loss": 0.4504,
+      "epoch": 10.85,
+      "learning_rate": 8.924769339957418e-05,
+      "loss": 0.3102,
       "step": 1530
     },
     {
-      "epoch": 43.99,
-      "learning_rate": 0.00022662824207492793,
-      "loss": 0.4958,
+      "epoch": 10.92,
+      "learning_rate": 8.917672107877928e-05,
+      "loss": 0.4267,
       "step": 1540
     },
     {
-      "epoch": 43.99,
-      "eval_loss": 0.8263903856277466,
-      "eval_runtime": 8.775,
-      "eval_samples_per_second": 7.179,
-      "eval_steps_per_second": 1.823,
-      "eval_wer": 0.39561366512020246,
-      "step": 1540
+      "epoch": 10.99,
+      "learning_rate": 8.910574875798439e-05,
+      "loss": 0.4282,
+      "step": 1550
     },
     {
-      "epoch": 44.28,
-      "learning_rate": 0.0002254755043227666,
-      "loss": 0.4835,
-      "step": 1550
+      "epoch": 11.0,
+      "eval_loss": 0.6718541383743286,
+      "eval_runtime": 8.4294,
+      "eval_samples_per_second": 7.474,
+      "eval_steps_per_second": 3.796,
+      "eval_wer": 0.3475326866301139,
+      "step": 1551
     },
     {
-      "epoch": 44.57,
-      "learning_rate": 0.00022432276657060517,
-      "loss": 0.4188,
+      "epoch": 11.06,
+      "learning_rate": 8.90347764371895e-05,
+      "loss": 0.3585,
       "step": 1560
     },
     {
-      "epoch": 44.85,
-      "learning_rate": 0.00022317002881844383,
-      "loss": 0.4665,
+      "epoch": 11.13,
+      "learning_rate": 8.896380411639461e-05,
+      "loss": 0.7773,
       "step": 1570
     },
     {
-      "epoch": 44.99,
-      "eval_loss": 0.8172122240066528,
-      "eval_runtime": 8.7583,
-      "eval_samples_per_second": 7.193,
-      "eval_steps_per_second": 1.827,
-      "eval_wer": 0.4070012652889076,
-      "step": 1575
-    },
-    {
-      "epoch": 45.14,
-      "learning_rate": 0.00022201729106628243,
-      "loss": 0.4708,
+      "epoch": 11.21,
+      "learning_rate": 8.889283179559971e-05,
+      "loss": 0.4016,
       "step": 1580
     },
     {
-      "epoch": 45.43,
-      "learning_rate": 0.00022086455331412106,
-      "loss": 0.4321,
+      "epoch": 11.28,
+      "learning_rate": 8.882185947480483e-05,
+      "loss": 0.3903,
       "step": 1590
     },
     {
-      "epoch": 45.71,
-      "learning_rate": 0.00021971181556195967,
-      "loss": 0.5132,
+      "epoch": 11.35,
+      "learning_rate": 8.875088715400994e-05,
+      "loss": 0.3357,
       "step": 1600
     },
     {
-      "epoch": 45.99,
-      "learning_rate": 0.0002185590778097983,
-      "loss": 0.4196,
-      "step": 1610
-    },
-    {
-      "epoch": 45.99,
-      "eval_loss": 0.7970888018608093,
-      "eval_runtime": 8.8547,
-      "eval_samples_per_second": 7.115,
-      "eval_steps_per_second": 1.807,
-      "eval_wer": 0.39519190215099115,
+      "epoch": 11.42,
+      "learning_rate": 8.867991483321505e-05,
+      "loss": 0.3951,
       "step": 1610
     },
     {
-      "epoch": 46.28,
-      "learning_rate": 0.0002174063400576369,
-      "loss": 0.5437,
+      "epoch": 11.49,
+      "learning_rate": 8.860894251242015e-05,
+      "loss": 0.7644,
       "step": 1620
     },
     {
-      "epoch": 46.57,
-      "learning_rate": 0.00021625360230547553,
-      "loss": 0.4058,
+      "epoch": 11.56,
+      "learning_rate": 8.853797019162528e-05,
+      "loss": 0.3324,
       "step": 1630
     },
     {
-      "epoch": 46.85,
-      "learning_rate": 0.00021510086455331414,
-      "loss": 0.4088,
+      "epoch": 11.63,
+      "learning_rate": 8.846699787083038e-05,
+      "loss": 0.3739,
       "step": 1640
     },
     {
-      "epoch": 46.99,
-      "eval_loss": 0.7417133450508118,
-      "eval_runtime": 8.7997,
-      "eval_samples_per_second": 7.159,
-      "eval_steps_per_second": 1.818,
-      "eval_wer": 0.38802193167439897,
-      "step": 1645
-    },
-    {
-      "epoch": 47.14,
-      "learning_rate": 0.0002139481268011527,
-      "loss": 0.4692,
+      "epoch": 11.7,
+      "learning_rate": 8.839602555003549e-05,
+      "loss": 0.2898,
       "step": 1650
     },
     {
-      "epoch": 47.43,
-      "learning_rate": 0.00021279538904899137,
-      "loss": 0.4082,
+      "epoch": 11.77,
+      "learning_rate": 8.83250532292406e-05,
+      "loss": 0.3933,
       "step": 1660
     },
     {
-      "epoch": 47.71,
-      "learning_rate": 0.00021164265129682998,
-      "loss": 0.4934,
+      "epoch": 11.84,
+      "learning_rate": 8.825408090844571e-05,
+      "loss": 0.3054,
       "step": 1670
     },
     {
-      "epoch": 47.99,
-      "learning_rate": 0.0002104899135446686,
-      "loss": 0.4308,
+      "epoch": 11.91,
+      "learning_rate": 8.818310858765083e-05,
+      "loss": 0.8342,
       "step": 1680
     },
     {
-      "epoch": 47.99,
-      "eval_loss": 0.7806001901626587,
-      "eval_runtime": 8.8422,
-      "eval_samples_per_second": 7.125,
-      "eval_steps_per_second": 1.81,
-      "eval_wer": 0.39308308730493463,
-      "step": 1680
+      "epoch": 11.99,
+      "learning_rate": 8.811213626685593e-05,
+      "loss": 0.4398,
+      "step": 1690
     },
     {
-      "epoch": 48.28,
-      "learning_rate": 0.0002093371757925072,
-      "loss": 0.4835,
-      "step": 1690
+      "epoch": 12.0,
+      "eval_loss": 0.7130092978477478,
+      "eval_runtime": 8.4904,
+      "eval_samples_per_second": 7.42,
+      "eval_steps_per_second": 3.769,
+      "eval_wer": 0.36862083509067906,
+      "step": 1692
     },
     {
-      "epoch": 48.57,
-      "learning_rate": 0.00020818443804034584,
-      "loss": 0.4237,
+      "epoch": 12.06,
+      "learning_rate": 8.804116394606104e-05,
+      "loss": 0.3842,
       "step": 1700
     },
     {
-      "epoch": 48.85,
-      "learning_rate": 0.00020703170028818445,
-      "loss": 0.4173,
+      "epoch": 12.13,
+      "learning_rate": 8.797019162526615e-05,
+      "loss": 0.2242,
       "step": 1710
     },
     {
-      "epoch": 48.99,
-      "eval_loss": 0.7380473613739014,
-      "eval_runtime": 8.9386,
-      "eval_samples_per_second": 7.048,
-      "eval_steps_per_second": 1.79,
-      "eval_wer": 0.392239561366512,
-      "step": 1715
-    },
-    {
-      "epoch": 49.14,
-      "learning_rate": 0.00020587896253602308,
-      "loss": 0.6076,
+      "epoch": 12.2,
+      "learning_rate": 8.789921930447126e-05,
+      "loss": 0.3456,
       "step": 1720
     },
     {
-      "epoch": 49.43,
-      "learning_rate": 0.00020472622478386168,
-      "loss": 0.4505,
+      "epoch": 12.27,
+      "learning_rate": 8.782824698367636e-05,
+      "loss": 0.5004,
       "step": 1730
     },
     {
-      "epoch": 49.71,
-      "learning_rate": 0.0002035734870317003,
-      "loss": 0.4394,
+      "epoch": 12.34,
+      "learning_rate": 8.775727466288149e-05,
+      "loss": 0.3666,
       "step": 1740
     },
     {
-      "epoch": 49.99,
-      "learning_rate": 0.00020242074927953892,
-      "loss": 0.4653,
-      "step": 1750
-    },
-    {
-      "epoch": 49.99,
-      "eval_loss": 0.8961731195449829,
-      "eval_runtime": 8.8706,
-      "eval_samples_per_second": 7.102,
-      "eval_steps_per_second": 1.804,
-      "eval_wer": 0.4027836355967946,
+      "epoch": 12.41,
+      "learning_rate": 8.768630234208659e-05,
+      "loss": 0.3736,
       "step": 1750
     },
     {
-      "epoch": 50.28,
-      "learning_rate": 0.00020126801152737752,
-      "loss": 0.4578,
+      "epoch": 12.48,
+      "learning_rate": 8.76153300212917e-05,
+      "loss": 0.3188,
       "step": 1760
     },
     {
-      "epoch": 50.57,
-      "learning_rate": 0.00020011527377521615,
-      "loss": 1.0308,
+      "epoch": 12.55,
+      "learning_rate": 8.754435770049681e-05,
+      "loss": 0.3954,
       "step": 1770
     },
     {
-      "epoch": 50.85,
-      "learning_rate": 0.00019896253602305476,
-      "loss": 0.4406,
+      "epoch": 12.62,
+      "learning_rate": 8.747338537970193e-05,
+      "loss": 0.3274,
       "step": 1780
     },
     {
-      "epoch": 50.99,
-      "eval_loss": 0.7790281176567078,
-      "eval_runtime": 8.8636,
-      "eval_samples_per_second": 7.108,
-      "eval_steps_per_second": 1.805,
-      "eval_wer": 0.39350485027414595,
-      "step": 1785
-    },
-    {
-      "epoch": 51.14,
-      "learning_rate": 0.0001978097982708934,
-      "loss": 0.5174,
+      "epoch": 12.7,
+      "learning_rate": 8.740241305890703e-05,
+      "loss": 0.3119,
       "step": 1790
     },
     {
-      "epoch": 51.43,
-      "learning_rate": 0.00019665706051873202,
-      "loss": 0.489,
+      "epoch": 12.77,
+      "learning_rate": 8.733144073811214e-05,
+      "loss": 0.3678,
       "step": 1800
     },
     {
-      "epoch": 51.71,
-      "learning_rate": 0.00019550432276657062,
-      "loss": 0.6726,
+      "epoch": 12.84,
+      "learning_rate": 8.726046841731725e-05,
+      "loss": 0.3254,
       "step": 1810
     },
     {
-      "epoch": 51.99,
-      "learning_rate": 0.00019435158501440923,
-      "loss": 0.4664,
+      "epoch": 12.91,
+      "learning_rate": 8.718949609652236e-05,
+      "loss": 0.3593,
       "step": 1820
     },
     {
-      "epoch": 51.99,
-      "eval_loss": 0.917263925075531,
-      "eval_runtime": 8.8897,
-      "eval_samples_per_second": 7.087,
-      "eval_steps_per_second": 1.8,
-      "eval_wer": 0.3892872205820329,
-      "step": 1820
+      "epoch": 12.98,
+      "learning_rate": 8.711852377572746e-05,
+      "loss": 0.331,
+      "step": 1830
     },
     {
-      "epoch": 52.28,
-      "learning_rate": 0.00019319884726224783,
-      "loss": 0.4466,
-      "step": 1830
+      "epoch": 13.0,
+      "eval_loss": 0.6425178050994873,
+      "eval_runtime": 8.4315,
+      "eval_samples_per_second": 7.472,
+      "eval_steps_per_second": 3.795,
+      "eval_wer": 0.3627161535217208,
+      "step": 1833
     },
     {
-      "epoch": 52.57,
-      "learning_rate": 0.00019204610951008646,
-      "loss": 0.4465,
+      "epoch": 13.05,
+      "learning_rate": 8.704755145493259e-05,
+      "loss": 0.3567,
       "step": 1840
     },
     {
-      "epoch": 52.85,
-      "learning_rate": 0.00019089337175792507,
-      "loss": 0.4486,
+      "epoch": 13.12,
+      "learning_rate": 8.697657913413769e-05,
+      "loss": 0.2779,
       "step": 1850
     },
     {
-      "epoch": 52.99,
-      "eval_loss": 0.8234918713569641,
-      "eval_runtime": 8.8605,
-      "eval_samples_per_second": 7.11,
-      "eval_steps_per_second": 1.806,
-      "eval_wer": 0.392239561366512,
-      "step": 1855
-    },
-    {
-      "epoch": 53.14,
-      "learning_rate": 0.0001897406340057637,
-      "loss": 0.4336,
+      "epoch": 13.19,
+      "learning_rate": 8.69056068133428e-05,
+      "loss": 0.2866,
       "step": 1860
     },
     {
-      "epoch": 53.43,
-      "learning_rate": 0.0001885878962536023,
-      "loss": 0.4376,
+      "epoch": 13.26,
+      "learning_rate": 8.68346344925479e-05,
+      "loss": 0.3417,
       "step": 1870
     },
     {
-      "epoch": 53.71,
-      "learning_rate": 0.00018743515850144093,
-      "loss": 0.4706,
+      "epoch": 13.33,
+      "learning_rate": 8.676366217175303e-05,
+      "loss": 0.3396,
       "step": 1880
     },
     {
-      "epoch": 53.99,
-      "learning_rate": 0.00018628242074927956,
-      "loss": 0.4137,
-      "step": 1890
-    },
-    {
-      "epoch": 53.99,
-      "eval_loss": 0.8031628727912903,
-      "eval_runtime": 8.8634,
-      "eval_samples_per_second": 7.108,
-      "eval_steps_per_second": 1.805,
-      "eval_wer": 0.3926613243357233,
+      "epoch": 13.4,
+      "learning_rate": 8.669268985095813e-05,
+      "loss": 0.3115,
       "step": 1890
     },
     {
-      "epoch": 54.28,
-      "learning_rate": 0.00018512968299711817,
-      "loss": 0.4758,
+      "epoch": 13.48,
+      "learning_rate": 8.662171753016324e-05,
+      "loss": 0.399,
       "step": 1900
     },
     {
-      "epoch": 54.57,
-      "learning_rate": 0.0001839769452449568,
-      "loss": 0.3621,
+      "epoch": 13.55,
+      "learning_rate": 8.655074520936835e-05,
+      "loss": 0.4087,
       "step": 1910
     },
     {
-      "epoch": 54.85,
-      "learning_rate": 0.0001828242074927954,
-      "loss": 0.4402,
+      "epoch": 13.62,
+      "learning_rate": 8.647977288857346e-05,
+      "loss": 0.381,
       "step": 1920
     },
     {
-      "epoch": 54.99,
-      "eval_loss": 0.7658433318138123,
-      "eval_runtime": 8.8656,
-      "eval_samples_per_second": 7.106,
-      "eval_steps_per_second": 1.805,
-      "eval_wer": 0.38296077604386336,
-      "step": 1925
-    },
-    {
-      "epoch": 55.14,
-      "learning_rate": 0.000181671469740634,
-      "loss": 0.4614,
+      "epoch": 13.69,
+      "learning_rate": 8.640880056777856e-05,
+      "loss": 0.3189,
       "step": 1930
     },
     {
-      "epoch": 55.43,
-      "learning_rate": 0.0001805187319884726,
-      "loss": 0.4183,
+      "epoch": 13.76,
+      "learning_rate": 8.633782824698368e-05,
+      "loss": 0.516,
       "step": 1940
     },
     {
-      "epoch": 55.71,
-      "learning_rate": 0.00017936599423631124,
-      "loss": 0.4065,
+      "epoch": 13.83,
+      "learning_rate": 8.626685592618879e-05,
+      "loss": 0.3476,
       "step": 1950
     },
     {
-      "epoch": 55.99,
-      "learning_rate": 0.00017821325648414985,
-      "loss": 0.4101,
+      "epoch": 13.9,
+      "learning_rate": 8.61958836053939e-05,
+      "loss": 0.301,
       "step": 1960
     },
     {
-      "epoch": 55.99,
-      "eval_loss": 0.8620884418487549,
-      "eval_runtime": 8.9342,
-      "eval_samples_per_second": 7.052,
-      "eval_steps_per_second": 1.791,
-      "eval_wer": 0.3994095318431042,
-      "step": 1960
+      "epoch": 13.97,
+      "learning_rate": 8.612491128459901e-05,
+      "loss": 0.4488,
+      "step": 1970
     },
     {
-      "epoch": 56.28,
-      "learning_rate": 0.00017706051873198848,
-      "loss": 0.5536,
-      "step": 1970
+      "epoch": 14.0,
+      "eval_loss": 0.6483317017555237,
+      "eval_runtime": 8.4054,
+      "eval_samples_per_second": 7.495,
+      "eval_steps_per_second": 3.807,
+      "eval_wer": 0.36482496836777734,
+      "step": 1974
     },
     {
-      "epoch": 56.57,
-      "learning_rate": 0.0001759077809798271,
-      "loss": 0.4167,
+      "epoch": 14.04,
+      "learning_rate": 8.605393896380413e-05,
+      "loss": 0.3696,
       "step": 1980
     },
     {
-      "epoch": 56.85,
-      "learning_rate": 0.0001747550432276657,
-      "loss": 0.5239,
+      "epoch": 14.11,
+      "learning_rate": 8.598296664300924e-05,
+      "loss": 0.2732,
       "step": 1990
     },
     {
-      "epoch": 56.99,
-      "eval_loss": 0.7903460264205933,
-      "eval_runtime": 8.918,
-      "eval_samples_per_second": 7.064,
-      "eval_steps_per_second": 1.794,
-      "eval_wer": 0.39561366512020246,
-      "step": 1995
-    },
-    {
-      "epoch": 57.14,
-      "learning_rate": 0.00017360230547550434,
-      "loss": 0.6093,
+      "epoch": 14.18,
+      "learning_rate": 8.591199432221434e-05,
+      "loss": 0.3822,
       "step": 2000
     },
     {
-      "epoch": 57.43,
-      "learning_rate": 0.00017244956772334295,
-      "loss": 0.4394,
+      "epoch": 14.26,
+      "learning_rate": 8.584102200141945e-05,
+      "loss": 0.4847,
       "step": 2010
     },
     {
-      "epoch": 57.71,
-      "learning_rate": 0.00017129682997118158,
-      "loss": 0.4504,
+      "epoch": 14.33,
+      "learning_rate": 8.577004968062456e-05,
+      "loss": 0.355,
       "step": 2020
     },
     {
-      "epoch": 57.99,
-      "learning_rate": 0.00017014409221902018,
-      "loss": 0.4151,
-      "step": 2030
-    },
-    {
-      "epoch": 57.99,
-      "eval_loss": 0.7849175333976746,
-      "eval_runtime": 8.9241,
-      "eval_samples_per_second": 7.06,
-      "eval_steps_per_second": 1.793,
-      "eval_wer": 0.3871784057359764,
+      "epoch": 14.4,
+      "learning_rate": 8.569907735982968e-05,
+      "loss": 0.2866,
       "step": 2030
     },
     {
-      "epoch": 58.28,
-      "learning_rate": 0.0001689913544668588,
-      "loss": 0.463,
+      "epoch": 14.47,
+      "learning_rate": 8.562810503903478e-05,
+      "loss": 0.4715,
       "step": 2040
     },
     {
-      "epoch": 58.57,
-      "learning_rate": 0.00016783861671469742,
-      "loss": 0.4195,
+      "epoch": 14.54,
+      "learning_rate": 8.555713271823989e-05,
+      "loss": 0.4262,
       "step": 2050
     },
     {
-      "epoch": 58.85,
-      "learning_rate": 0.00016668587896253602,
-      "loss": 0.4766,
+      "epoch": 14.61,
+      "learning_rate": 8.5486160397445e-05,
+      "loss": 0.2553,
       "step": 2060
     },
     {
-      "epoch": 58.99,
-      "eval_loss": 0.8306421041488647,
-      "eval_runtime": 9.0122,
-      "eval_samples_per_second": 6.99,
-      "eval_steps_per_second": 1.775,
-      "eval_wer": 0.39181779839730074,
-      "step": 2065
-    },
-    {
-      "epoch": 59.14,
-      "learning_rate": 0.00016553314121037465,
-      "loss": 0.4292,
+      "epoch": 14.68,
+      "learning_rate": 8.541518807665011e-05,
+      "loss": 0.2963,
       "step": 2070
     },
     {
-      "epoch": 59.43,
-      "learning_rate": 0.00016438040345821326,
-      "loss": 0.4278,
+      "epoch": 14.75,
+      "learning_rate": 8.534421575585521e-05,
+      "loss": 0.3588,
       "step": 2080
     },
     {
-      "epoch": 59.71,
-      "learning_rate": 0.0001632276657060519,
-      "loss": 0.4586,
+      "epoch": 14.82,
+      "learning_rate": 8.527324343506034e-05,
+      "loss": 0.338,
       "step": 2090
     },
     {
-      "epoch": 59.99,
-      "learning_rate": 0.0001620749279538905,
-      "loss": 0.4882,
+      "epoch": 14.89,
+      "learning_rate": 8.520227111426544e-05,
+      "loss": 0.2616,
       "step": 2100
     },
     {
-      "epoch": 59.99,
-      "eval_loss": 0.8134064674377441,
-      "eval_runtime": 8.8415,
-      "eval_samples_per_second": 7.125,
-      "eval_steps_per_second": 1.81,
-      "eval_wer": 0.3926613243357233,
-      "step": 2100
+      "epoch": 14.96,
+      "learning_rate": 8.513129879347055e-05,
+      "loss": 0.3876,
+      "step": 2110
     },
     {
-      "epoch": 60.28,
-      "learning_rate": 0.00016092219020172912,
-      "loss": 0.3925,
-      "step": 2110
+      "epoch": 15.0,
+      "eval_loss": 0.637477457523346,
+      "eval_runtime": 8.3988,
+      "eval_samples_per_second": 7.501,
+      "eval_steps_per_second": 3.81,
+      "eval_wer": 0.3509067903838043,
+      "step": 2115
     },
     {
-      "epoch": 60.57,
-      "learning_rate": 0.00015976945244956773,
-      "loss": 0.4318,
+      "epoch": 15.04,
+      "learning_rate": 8.506032647267565e-05,
+      "loss": 0.3431,
       "step": 2120
     },
     {
-      "epoch": 60.85,
-      "learning_rate": 0.00015861671469740636,
-      "loss": 0.4583,
+      "epoch": 15.11,
+      "learning_rate": 8.498935415188078e-05,
+      "loss": 0.3285,
       "step": 2130
     },
     {
-      "epoch": 60.99,
-      "eval_loss": 0.9527338743209839,
-      "eval_runtime": 8.8501,
-      "eval_samples_per_second": 7.119,
-      "eval_steps_per_second": 1.808,
-      "eval_wer": 0.3850695908899199,
-      "step": 2135
-    },
-    {
-      "epoch": 61.14,
-      "learning_rate": 0.00015746397694524496,
-      "loss": 0.6257,
+      "epoch": 15.18,
+      "learning_rate": 8.491838183108588e-05,
+      "loss": 0.2948,
       "step": 2140
     },
     {
-      "epoch": 61.43,
-      "learning_rate": 0.0001563112391930836,
-      "loss": 0.3906,
+      "epoch": 15.25,
+      "learning_rate": 8.484740951029099e-05,
+      "loss": 0.3892,
       "step": 2150
     },
     {
-      "epoch": 61.71,
-      "learning_rate": 0.0001551585014409222,
-      "loss": 0.468,
+      "epoch": 15.32,
+      "learning_rate": 8.47764371894961e-05,
+      "loss": 0.3617,
       "step": 2160
     },
     {
-      "epoch": 61.99,
-      "learning_rate": 0.0001540057636887608,
-      "loss": 0.4284,
-      "step": 2170
-    },
-    {
-      "epoch": 61.99,
-      "eval_loss": 0.9743154048919678,
-      "eval_runtime": 8.9267,
-      "eval_samples_per_second": 7.057,
-      "eval_steps_per_second": 1.792,
-      "eval_wer": 0.3998312948123155,
+      "epoch": 15.39,
+      "learning_rate": 8.470546486870121e-05,
+      "loss": 0.2205,
       "step": 2170
     },
     {
-      "epoch": 62.28,
-      "learning_rate": 0.00015285302593659943,
-      "loss": 0.3827,
+      "epoch": 15.46,
+      "learning_rate": 8.463449254790631e-05,
+      "loss": 0.3829,
       "step": 2180
     },
     {
-      "epoch": 62.57,
-      "learning_rate": 0.00015170028818443804,
-      "loss": 0.3974,
+      "epoch": 15.53,
+      "learning_rate": 8.456352022711143e-05,
+      "loss": 0.4073,
       "step": 2190
     },
     {
-      "epoch": 62.85,
-      "learning_rate": 0.00015054755043227667,
-      "loss": 0.46,
+      "epoch": 15.6,
+      "learning_rate": 8.449254790631654e-05,
+      "loss": 0.4901,
       "step": 2200
     },
     {
-      "epoch": 62.99,
-      "eval_loss": 0.7807419300079346,
-      "eval_runtime": 8.9409,
-      "eval_samples_per_second": 7.046,
-      "eval_steps_per_second": 1.79,
-      "eval_wer": 0.38296077604386336,
-      "step": 2205
-    },
-    {
-      "epoch": 63.14,
-      "learning_rate": 0.00014939481268011527,
-      "loss": 0.4637,
+      "epoch": 15.67,
+      "learning_rate": 8.442157558552165e-05,
+      "loss": 0.3071,
       "step": 2210
     },
     {
-      "epoch": 63.43,
-      "learning_rate": 0.0001482420749279539,
-      "loss": 0.3948,
+      "epoch": 15.74,
+      "learning_rate": 8.435060326472675e-05,
+      "loss": 0.3896,
       "step": 2220
     },
     {
-      "epoch": 63.71,
-      "learning_rate": 0.0001470893371757925,
-      "loss": 0.4112,
+      "epoch": 15.82,
+      "learning_rate": 8.427963094393188e-05,
+      "loss": 0.3653,
       "step": 2230
     },
     {
-      "epoch": 63.99,
-      "learning_rate": 0.00014593659942363114,
-      "loss": 0.4039,
+      "epoch": 15.89,
+      "learning_rate": 8.420865862313698e-05,
+      "loss": 0.3004,
       "step": 2240
     },
     {
-      "epoch": 63.99,
-      "eval_loss": 0.8864094614982605,
-      "eval_runtime": 8.7598,
-      "eval_samples_per_second": 7.192,
-      "eval_steps_per_second": 1.827,
-      "eval_wer": 0.3884436946436103,
-      "step": 2240
+      "epoch": 15.96,
+      "learning_rate": 8.413768630234209e-05,
+      "loss": 0.3361,
+      "step": 2250
     },
     {
-      "epoch": 64.28,
-      "learning_rate": 0.00014478386167146974,
-      "loss": 0.3935,
-      "step": 2250
+      "epoch": 16.0,
+      "eval_loss": 0.6791129112243652,
+      "eval_runtime": 8.4384,
+      "eval_samples_per_second": 7.466,
+      "eval_steps_per_second": 3.792,
+      "eval_wer": 0.37030788696752426,
+      "step": 2256
     },
     {
-      "epoch": 64.57,
-      "learning_rate": 0.00014363112391930838,
-      "loss": 0.3677,
+      "epoch": 16.03,
+      "learning_rate": 8.40667139815472e-05,
+      "loss": 0.3807,
       "step": 2260
     },
     {
-      "epoch": 64.85,
-      "learning_rate": 0.00014247838616714698,
-      "loss": 0.3868,
+      "epoch": 16.1,
+      "learning_rate": 8.399574166075231e-05,
+      "loss": 0.3632,
       "step": 2270
     },
     {
-      "epoch": 64.99,
-      "eval_loss": 0.7304245233535767,
-      "eval_runtime": 8.9023,
-      "eval_samples_per_second": 7.077,
-      "eval_steps_per_second": 1.797,
-      "eval_wer": 0.3816954871362294,
-      "step": 2275
-    },
-    {
-      "epoch": 65.14,
-      "learning_rate": 0.00014132564841498558,
-      "loss": 0.4321,
+      "epoch": 16.17,
+      "learning_rate": 8.392476933995743e-05,
+      "loss": 0.3268,
       "step": 2280
     },
     {
-      "epoch": 65.43,
-      "learning_rate": 0.00014017291066282421,
-      "loss": 0.3726,
+      "epoch": 16.24,
+      "learning_rate": 8.385379701916253e-05,
+      "loss": 0.4717,
       "step": 2290
     },
     {
-      "epoch": 65.71,
-      "learning_rate": 0.00013902017291066282,
-      "loss": 0.428,
+      "epoch": 16.31,
+      "learning_rate": 8.378282469836764e-05,
+      "loss": 0.3313,
       "step": 2300
     },
     {
-      "epoch": 65.99,
-      "learning_rate": 0.00013786743515850145,
-      "loss": 0.3934,
-      "step": 2310
-    },
-    {
-      "epoch": 65.99,
-      "eval_loss": 0.875798761844635,
-      "eval_runtime": 8.8052,
-      "eval_samples_per_second": 7.155,
-      "eval_steps_per_second": 1.817,
-      "eval_wer": 0.38464782792070856,
+      "epoch": 16.38,
+      "learning_rate": 8.371185237757275e-05,
+      "loss": 0.2431,
       "step": 2310
     },
     {
-      "epoch": 66.28,
-      "learning_rate": 0.00013671469740634005,
-      "loss": 0.3963,
+      "epoch": 16.45,
+      "learning_rate": 8.364088005677786e-05,
+      "loss": 0.2565,
       "step": 2320
     },
     {
-      "epoch": 66.57,
-      "learning_rate": 0.00013556195965417869,
-      "loss": 0.3708,
+      "epoch": 16.52,
+      "learning_rate": 8.356990773598296e-05,
+      "loss": 0.3185,
       "step": 2330
     },
     {
-      "epoch": 66.85,
-      "learning_rate": 0.0001344092219020173,
-      "loss": 0.3776,
+      "epoch": 16.6,
+      "learning_rate": 8.349893541518809e-05,
+      "loss": 0.4059,
       "step": 2340
     },
     {
-      "epoch": 66.99,
-      "eval_loss": 0.8156087398529053,
-      "eval_runtime": 8.9243,
-      "eval_samples_per_second": 7.059,
-      "eval_steps_per_second": 1.793,
-      "eval_wer": 0.3762125685364825,
-      "step": 2345
-    },
-    {
-      "epoch": 67.14,
-      "learning_rate": 0.00013325648414985592,
-      "loss": 0.4032,
+      "epoch": 16.67,
+      "learning_rate": 8.342796309439319e-05,
+      "loss": 0.2951,
       "step": 2350
     },
     {
-      "epoch": 67.43,
-      "learning_rate": 0.00013210374639769455,
-      "loss": 0.4047,
+      "epoch": 16.74,
+      "learning_rate": 8.33569907735983e-05,
+      "loss": 0.4013,
       "step": 2360
     },
     {
-      "epoch": 67.71,
-      "learning_rate": 0.00013095100864553316,
-      "loss": 0.4063,
+      "epoch": 16.81,
+      "learning_rate": 8.328601845280341e-05,
+      "loss": 0.3153,
       "step": 2370
     },
     {
-      "epoch": 67.99,
-      "learning_rate": 0.00012979827089337176,
-      "loss": 0.3499,
+      "epoch": 16.88,
+      "learning_rate": 8.321504613200853e-05,
+      "loss": 0.2732,
       "step": 2380
     },
     {
-      "epoch": 67.99,
-      "eval_loss": 0.814339816570282,
-      "eval_runtime": 8.9099,
-      "eval_samples_per_second": 7.071,
-      "eval_steps_per_second": 1.796,
-      "eval_wer": 0.3888654576128216,
-      "step": 2380
+      "epoch": 16.95,
+      "learning_rate": 8.314407381121363e-05,
+      "loss": 0.344,
+      "step": 2390
     },
     {
-      "epoch": 68.28,
-      "learning_rate": 0.00012864553314121036,
-      "loss": 0.4262,
-      "step": 2390
+      "epoch": 17.0,
+      "eval_loss": 0.727933943271637,
+      "eval_runtime": 8.3629,
+      "eval_samples_per_second": 7.533,
+      "eval_steps_per_second": 3.826,
+      "eval_wer": 0.3551244200759173,
+      "step": 2397
     },
     {
-      "epoch": 68.57,
-      "learning_rate": 0.000127492795389049,
-      "loss": 0.4213,
+      "epoch": 17.02,
+      "learning_rate": 8.307310149041874e-05,
+      "loss": 0.4237,
       "step": 2400
     },
     {
-      "epoch": 68.85,
-      "learning_rate": 0.0001263400576368876,
-      "loss": 0.4055,
+      "epoch": 17.09,
+      "learning_rate": 8.300212916962385e-05,
+      "loss": 0.2894,
       "step": 2410
     },
     {
-      "epoch": 68.99,
-      "eval_loss": 0.7502649426460266,
-      "eval_runtime": 8.8806,
-      "eval_samples_per_second": 7.094,
-      "eval_steps_per_second": 1.802,
-      "eval_wer": 0.3795866722901729,
-      "step": 2415
-    },
-    {
-      "epoch": 69.14,
-      "learning_rate": 0.00012518731988472623,
-      "loss": 0.4393,
+      "epoch": 17.16,
+      "learning_rate": 8.293115684882896e-05,
+      "loss": 0.3385,
       "step": 2420
     },
     {
-      "epoch": 69.43,
-      "learning_rate": 0.00012403458213256483,
-      "loss": 0.3695,
+      "epoch": 17.23,
+      "learning_rate": 8.286018452803406e-05,
+      "loss": 0.3934,
       "step": 2430
     },
     {
-      "epoch": 69.71,
-      "learning_rate": 0.00012288184438040347,
-      "loss": 0.4452,
+      "epoch": 17.3,
+      "learning_rate": 8.278921220723918e-05,
+      "loss": 0.3708,
       "step": 2440
     },
     {
-      "epoch": 69.99,
-      "learning_rate": 0.00012172910662824208,
-      "loss": 0.3505,
+      "epoch": 17.38,
+      "learning_rate": 8.271823988644429e-05,
+      "loss": 0.2129,
       "step": 2450
     },
     {
-      "epoch": 69.99,
-      "eval_loss": 0.7138158082962036,
-      "eval_runtime": 8.8451,
-      "eval_samples_per_second": 7.123,
-      "eval_steps_per_second": 1.809,
-      "eval_wer": 0.38043019822859553,
-      "step": 2450
+      "epoch": 17.45,
+      "learning_rate": 8.26472675656494e-05,
+      "loss": 0.3097,
+      "step": 2460
     },
     {
-      "epoch": 70.28,
-      "learning_rate": 0.0001205763688760807,
-      "loss": 0.4222,
-      "step": 2460
-    },
-    {
-      "epoch": 70.57,
-      "learning_rate": 0.00011942363112391932,
-      "loss": 0.3144,
+      "epoch": 17.52,
+      "learning_rate": 8.25762952448545e-05,
+      "loss": 0.3606,
       "step": 2470
     },
     {
-      "epoch": 70.85,
-      "learning_rate": 0.00011827089337175794,
-      "loss": 0.3755,
+      "epoch": 17.59,
+      "learning_rate": 8.250532292405963e-05,
+      "loss": 0.2913,
       "step": 2480
     },
     {
-      "epoch": 70.99,
-      "eval_loss": 0.8071521520614624,
-      "eval_runtime": 8.8022,
-      "eval_samples_per_second": 7.157,
-      "eval_steps_per_second": 1.818,
-      "eval_wer": 0.3800084352593842,
-      "step": 2485
-    },
-    {
-      "epoch": 71.14,
-      "learning_rate": 0.00011711815561959654,
-      "loss": 0.3778,
+      "epoch": 17.66,
+      "learning_rate": 8.243435060326473e-05,
+      "loss": 0.3282,
       "step": 2490
     },
     {
-      "epoch": 71.43,
-      "learning_rate": 0.00011596541786743516,
-      "loss": 0.3831,
+      "epoch": 17.73,
+      "learning_rate": 8.236337828246984e-05,
+      "loss": 0.3118,
       "step": 2500
     },
     {
-      "epoch": 71.71,
-      "learning_rate": 0.00011481268011527378,
-      "loss": 0.3552,
+      "epoch": 17.8,
+      "learning_rate": 8.229240596167494e-05,
+      "loss": 0.3547,
       "step": 2510
     },
     {
-      "epoch": 71.99,
-      "learning_rate": 0.0001136599423631124,
-      "loss": 0.3594,
+      "epoch": 17.87,
+      "learning_rate": 8.222143364088006e-05,
+      "loss": 0.3776,
       "step": 2520
     },
     {
-      "epoch": 71.99,
-      "eval_loss": 0.7692469358444214,
-      "eval_runtime": 8.7672,
-      "eval_samples_per_second": 7.186,
-      "eval_steps_per_second": 1.825,
-      "eval_wer": 0.3850695908899199,
-      "step": 2520
+      "epoch": 17.94,
+      "learning_rate": 8.215046132008516e-05,
+      "loss": 0.3198,
+      "step": 2530
     },
     {
-      "epoch": 72.28,
-      "learning_rate": 0.00011250720461095101,
-      "loss": 0.4549,
-      "step": 2530
+      "epoch": 18.0,
+      "eval_loss": 0.6801475882530212,
+      "eval_runtime": 8.3766,
+      "eval_samples_per_second": 7.521,
+      "eval_steps_per_second": 3.82,
+      "eval_wer": 0.3509067903838043,
+      "step": 2538
     },
     {
-      "epoch": 72.57,
-      "learning_rate": 0.00011135446685878963,
-      "loss": 0.3636,
+      "epoch": 18.01,
+      "learning_rate": 8.207948899929028e-05,
+      "loss": 0.362,
       "step": 2540
     },
     {
-      "epoch": 72.85,
-      "learning_rate": 0.00011020172910662825,
-      "loss": 0.3167,
+      "epoch": 18.09,
+      "learning_rate": 8.200851667849539e-05,
+      "loss": 0.3886,
       "step": 2550
     },
     {
-      "epoch": 72.99,
-      "eval_loss": 0.6994994282722473,
-      "eval_runtime": 8.766,
-      "eval_samples_per_second": 7.187,
-      "eval_steps_per_second": 1.825,
-      "eval_wer": 0.3745255166596373,
-      "step": 2555
-    },
-    {
-      "epoch": 73.14,
-      "learning_rate": 0.00010904899135446686,
-      "loss": 0.3651,
+      "epoch": 18.16,
+      "learning_rate": 8.19375443577005e-05,
+      "loss": 0.3027,
       "step": 2560
     },
     {
-      "epoch": 73.43,
-      "learning_rate": 0.00010789625360230548,
-      "loss": 0.3751,
+      "epoch": 18.23,
+      "learning_rate": 8.186657203690562e-05,
+      "loss": 0.392,
       "step": 2570
     },
     {
-      "epoch": 73.71,
-      "learning_rate": 0.0001067435158501441,
-      "loss": 0.4196,
+      "epoch": 18.3,
+      "learning_rate": 8.179559971611071e-05,
+      "loss": 0.2925,
       "step": 2580
     },
     {
-      "epoch": 73.99,
-      "learning_rate": 0.00010559077809798272,
-      "loss": 0.3915,
-      "step": 2590
-    },
-    {
-      "epoch": 73.99,
-      "eval_loss": 0.6712434887886047,
-      "eval_runtime": 8.8181,
-      "eval_samples_per_second": 7.144,
-      "eval_steps_per_second": 1.814,
-      "eval_wer": 0.3762125685364825,
+      "epoch": 18.37,
+      "learning_rate": 8.172462739531583e-05,
+      "loss": 0.2887,
       "step": 2590
     },
     {
-      "epoch": 74.28,
-      "learning_rate": 0.00010443804034582132,
-      "loss": 0.3952,
+      "epoch": 18.44,
+      "learning_rate": 8.165365507452094e-05,
+      "loss": 0.2679,
       "step": 2600
     },
     {
-      "epoch": 74.57,
-      "learning_rate": 0.00010328530259365994,
-      "loss": 0.3115,
+      "epoch": 18.51,
+      "learning_rate": 8.158268275372605e-05,
+      "loss": 0.3519,
       "step": 2610
     },
     {
-      "epoch": 74.85,
-      "learning_rate": 0.00010213256484149856,
-      "loss": 0.3741,
+      "epoch": 18.58,
+      "learning_rate": 8.151171043293117e-05,
+      "loss": 0.3358,
       "step": 2620
     },
     {
-      "epoch": 74.99,
-      "eval_loss": 0.7139262557029724,
-      "eval_runtime": 8.8329,
-      "eval_samples_per_second": 7.132,
-      "eval_steps_per_second": 1.811,
-      "eval_wer": 0.3800084352593842,
-      "step": 2625
-    },
-    {
-      "epoch": 75.14,
-      "learning_rate": 0.00010097982708933717,
-      "loss": 0.3542,
+      "epoch": 18.65,
+      "learning_rate": 8.144073811213628e-05,
+      "loss": 0.2472,
       "step": 2630
     },
     {
-      "epoch": 75.43,
-      "learning_rate": 9.982708933717579e-05,
-      "loss": 0.4682,
+      "epoch": 18.72,
+      "learning_rate": 8.136976579134138e-05,
+      "loss": 0.4711,
       "step": 2640
     },
     {
-      "epoch": 75.71,
-      "learning_rate": 9.867435158501441e-05,
-      "loss": 0.3378,
+      "epoch": 18.79,
+      "learning_rate": 8.129879347054649e-05,
+      "loss": 0.399,
       "step": 2650
     },
     {
-      "epoch": 75.99,
-      "learning_rate": 9.752161383285303e-05,
-      "loss": 0.3708,
+      "epoch": 18.87,
+      "learning_rate": 8.12278211497516e-05,
+      "loss": 0.2654,
       "step": 2660
     },
     {
-      "epoch": 75.99,
-      "eval_loss": 0.7064934968948364,
-      "eval_runtime": 8.8273,
-      "eval_samples_per_second": 7.137,
-      "eval_steps_per_second": 1.813,
-      "eval_wer": 0.3833825390130747,
-      "step": 2660
+      "epoch": 18.94,
+      "learning_rate": 8.115684882895672e-05,
+      "loss": 0.2753,
+      "step": 2670
     },
     {
-      "epoch": 76.28,
-      "learning_rate": 9.636887608069164e-05,
-      "loss": 0.4179,
-      "step": 2670
+      "epoch": 19.0,
+      "eval_loss": 0.6238527894020081,
+      "eval_runtime": 8.3979,
+      "eval_samples_per_second": 7.502,
+      "eval_steps_per_second": 3.81,
+      "eval_wer": 0.3509067903838043,
+      "step": 2679
     },
     {
-      "epoch": 76.57,
-      "learning_rate": 9.521613832853028e-05,
-      "loss": 0.3543,
+      "epoch": 19.01,
+      "learning_rate": 8.108587650816181e-05,
+      "loss": 0.5353,
       "step": 2680
     },
     {
-      "epoch": 76.85,
-      "learning_rate": 9.406340057636888e-05,
+      "epoch": 19.08,
+      "learning_rate": 8.101490418736694e-05,
       "loss": 0.3731,
       "step": 2690
     },
     {
-      "epoch": 76.99,
-      "eval_loss": 0.7315667867660522,
-      "eval_runtime": 8.8516,
-      "eval_samples_per_second": 7.117,
-      "eval_steps_per_second": 1.808,
-      "eval_wer": 0.37536904259805987,
-      "step": 2695
-    },
-    {
-      "epoch": 77.14,
-      "learning_rate": 9.29106628242075e-05,
-      "loss": 0.4854,
+      "epoch": 19.15,
+      "learning_rate": 8.094393186657204e-05,
+      "loss": 0.2425,
       "step": 2700
     },
     {
-      "epoch": 77.43,
-      "learning_rate": 9.175792507204611e-05,
-      "loss": 0.3493,
+      "epoch": 19.22,
+      "learning_rate": 8.087295954577715e-05,
+      "loss": 0.3249,
       "step": 2710
     },
     {
-      "epoch": 77.71,
-      "learning_rate": 9.060518731988473e-05,
-      "loss": 0.3559,
+      "epoch": 19.29,
+      "learning_rate": 8.080198722498225e-05,
+      "loss": 0.4575,
       "step": 2720
     },
     {
-      "epoch": 77.99,
-      "learning_rate": 8.945244956772335e-05,
-      "loss": 0.3785,
-      "step": 2730
-    },
-    {
-      "epoch": 77.99,
-      "eval_loss": 0.7071495056152344,
-      "eval_runtime": 8.8836,
-      "eval_samples_per_second": 7.092,
-      "eval_steps_per_second": 1.801,
-      "eval_wer": 0.3757908055672712,
+      "epoch": 19.36,
+      "learning_rate": 8.073101490418738e-05,
+      "loss": 0.2898,
       "step": 2730
     },
     {
-      "epoch": 78.28,
-      "learning_rate": 8.829971181556195e-05,
-      "loss": 0.3751,
+      "epoch": 19.43,
+      "learning_rate": 8.066004258339248e-05,
+      "loss": 0.3392,
       "step": 2740
     },
     {
-      "epoch": 78.57,
-      "learning_rate": 8.714697406340057e-05,
-      "loss": 0.3622,
+      "epoch": 19.5,
+      "learning_rate": 8.058907026259759e-05,
+      "loss": 0.4448,
       "step": 2750
     },
     {
-      "epoch": 78.85,
-      "learning_rate": 8.599423631123919e-05,
-      "loss": 0.3466,
+      "epoch": 19.57,
+      "learning_rate": 8.05180979418027e-05,
+      "loss": 0.2676,
       "step": 2760
     },
     {
-      "epoch": 78.99,
-      "eval_loss": 0.7361500263214111,
-      "eval_runtime": 9.0867,
-      "eval_samples_per_second": 6.933,
-      "eval_steps_per_second": 1.761,
-      "eval_wer": 0.3833825390130747,
-      "step": 2765
-    },
-    {
-      "epoch": 79.14,
-      "learning_rate": 8.484149855907782e-05,
-      "loss": 0.3678,
+      "epoch": 19.65,
+      "learning_rate": 8.044712562100782e-05,
+      "loss": 0.321,
       "step": 2770
     },
     {
-      "epoch": 79.43,
-      "learning_rate": 8.368876080691644e-05,
-      "loss": 0.3829,
+      "epoch": 19.72,
+      "learning_rate": 8.037615330021291e-05,
+      "loss": 0.3333,
       "step": 2780
     },
     {
-      "epoch": 79.71,
-      "learning_rate": 8.253602305475506e-05,
-      "loss": 0.4725,
+      "epoch": 19.79,
+      "learning_rate": 8.030518097941803e-05,
+      "loss": 0.3355,
       "step": 2790
     },
     {
-      "epoch": 79.99,
-      "learning_rate": 8.138328530259366e-05,
-      "loss": 0.3505,
-      "step": 2800
-    },
-    {
-      "epoch": 79.99,
-      "eval_loss": 0.6964561343193054,
-      "eval_runtime": 8.8201,
-      "eval_samples_per_second": 7.143,
-      "eval_steps_per_second": 1.814,
-      "eval_wer": 0.3800084352593842,
+      "epoch": 19.86,
+      "learning_rate": 8.023420865862314e-05,
+      "loss": 0.3245,
       "step": 2800
     },
     {
-      "epoch": 80.28,
-      "learning_rate": 8.023054755043228e-05,
-      "loss": 0.3439,
+      "epoch": 19.93,
+      "learning_rate": 8.016323633782825e-05,
+      "loss": 0.2878,
       "step": 2810
     },
     {
-      "epoch": 80.57,
-      "learning_rate": 7.90778097982709e-05,
-      "loss": 0.3645,
+      "epoch": 20.0,
+      "learning_rate": 8.009226401703335e-05,
+      "loss": 0.2962,
       "step": 2820
     },
     {
-      "epoch": 80.85,
-      "learning_rate": 7.792507204610951e-05,
-      "loss": 0.4003,
-      "step": 2830
+      "epoch": 20.0,
+      "eval_loss": 0.7418911457061768,
+      "eval_runtime": 8.4497,
+      "eval_samples_per_second": 7.456,
+      "eval_steps_per_second": 3.787,
+      "eval_wer": 0.3441585828764234,
+      "step": 2820
     },
     {
-      "epoch": 80.99,
-      "eval_loss": 0.752064049243927,
-      "eval_runtime": 8.7801,
-      "eval_samples_per_second": 7.175,
-      "eval_steps_per_second": 1.822,
-      "eval_wer": 0.3766343315056938,
-      "step": 2835
+      "epoch": 20.07,
+      "learning_rate": 8.002129169623848e-05,
+      "loss": 0.3513,
+      "step": 2830
     },
     {
-      "epoch": 81.14,
-      "learning_rate": 7.677233429394813e-05,
-      "loss": 0.4152,
+      "epoch": 20.14,
+      "learning_rate": 7.995031937544358e-05,
+      "loss": 0.2747,
       "step": 2840
     },
     {
-      "epoch": 81.43,
-      "learning_rate": 7.561959654178675e-05,
-      "loss": 0.3652,
+      "epoch": 20.21,
+      "learning_rate": 7.987934705464869e-05,
+      "loss": 0.3844,
       "step": 2850
     },
     {
-      "epoch": 81.71,
-      "learning_rate": 7.446685878962537e-05,
-      "loss": 0.3254,
+      "epoch": 20.28,
+      "learning_rate": 7.98083747338538e-05,
+      "loss": 0.465,
       "step": 2860
     },
     {
-      "epoch": 81.99,
-      "learning_rate": 7.331412103746398e-05,
-      "loss": 0.3723,
-      "step": 2870
-    },
-    {
-      "epoch": 81.99,
-      "eval_loss": 0.7616933584213257,
-      "eval_runtime": 8.8771,
-      "eval_samples_per_second": 7.097,
-      "eval_steps_per_second": 1.802,
-      "eval_wer": 0.3749472796288486,
+      "epoch": 20.35,
+      "learning_rate": 7.973740241305892e-05,
+      "loss": 0.3068,
       "step": 2870
     },
     {
-      "epoch": 82.28,
-      "learning_rate": 7.21613832853026e-05,
-      "loss": 0.3791,
+      "epoch": 20.43,
+      "learning_rate": 7.966643009226403e-05,
+      "loss": 0.2795,
       "step": 2880
     },
     {
-      "epoch": 82.57,
-      "learning_rate": 7.100864553314122e-05,
-      "loss": 0.3127,
+      "epoch": 20.5,
+      "learning_rate": 7.959545777146913e-05,
+      "loss": 0.4048,
       "step": 2890
     },
     {
-      "epoch": 82.85,
-      "learning_rate": 6.985590778097984e-05,
-      "loss": 0.4029,
+      "epoch": 20.57,
+      "learning_rate": 7.952448545067424e-05,
+      "loss": 0.3131,
       "step": 2900
     },
     {
-      "epoch": 82.99,
-      "eval_loss": 0.7659004926681519,
-      "eval_runtime": 8.8596,
-      "eval_samples_per_second": 7.111,
-      "eval_steps_per_second": 1.806,
-      "eval_wer": 0.38127372416701816,
-      "step": 2905
-    },
-    {
-      "epoch": 83.14,
-      "learning_rate": 6.870317002881844e-05,
-      "loss": 0.3742,
+      "epoch": 20.64,
+      "learning_rate": 7.945351312987935e-05,
+      "loss": 0.2241,
       "step": 2910
     },
     {
-      "epoch": 83.43,
-      "learning_rate": 6.755043227665706e-05,
-      "loss": 0.3072,
+      "epoch": 20.71,
+      "learning_rate": 7.938254080908447e-05,
+      "loss": 0.3629,
       "step": 2920
     },
     {
-      "epoch": 83.71,
-      "learning_rate": 6.639769452449568e-05,
-      "loss": 0.3836,
+      "epoch": 20.78,
+      "learning_rate": 7.931156848828956e-05,
+      "loss": 0.4955,
       "step": 2930
     },
     {
-      "epoch": 83.99,
-      "learning_rate": 6.52449567723343e-05,
-      "loss": 0.3478,
-      "step": 2940
-    },
-    {
-      "epoch": 83.99,
-      "eval_loss": 0.707669198513031,
-      "eval_runtime": 8.9849,
-      "eval_samples_per_second": 7.012,
-      "eval_steps_per_second": 1.781,
-      "eval_wer": 0.3833825390130747,
+      "epoch": 20.85,
+      "learning_rate": 7.924059616749469e-05,
+      "loss": 0.3154,
       "step": 2940
     },
     {
-      "epoch": 84.28,
-      "learning_rate": 6.409221902017291e-05,
-      "loss": 0.3467,
+      "epoch": 20.92,
+      "learning_rate": 7.916962384669979e-05,
+      "loss": 0.2936,
       "step": 2950
     },
     {
-      "epoch": 84.57,
-      "learning_rate": 6.293948126801153e-05,
-      "loss": 0.324,
+      "epoch": 20.99,
+      "learning_rate": 7.90986515259049e-05,
+      "loss": 0.7503,
       "step": 2960
     },
     {
-      "epoch": 84.85,
-      "learning_rate": 6.178674351585015e-05,
-      "loss": 0.3363,
-      "step": 2970
+      "epoch": 21.0,
+      "eval_loss": 0.7278538942337036,
+      "eval_runtime": 8.392,
+      "eval_samples_per_second": 7.507,
+      "eval_steps_per_second": 3.813,
+      "eval_wer": 0.3500632644453817,
+      "step": 2961
     },
     {
-      "epoch": 84.99,
-      "eval_loss": 0.7333377599716187,
-      "eval_runtime": 8.8448,
-      "eval_samples_per_second": 7.123,
-      "eval_steps_per_second": 1.809,
-      "eval_wer": 0.3787431463517503,
-      "step": 2975
+      "epoch": 21.06,
+      "learning_rate": 7.902767920511e-05,
+      "loss": 0.3804,
+      "step": 2970
     },
     {
-      "epoch": 85.14,
-      "learning_rate": 6.0634005763688764e-05,
-      "loss": 0.343,
+      "epoch": 21.13,
+      "learning_rate": 7.895670688431513e-05,
+      "loss": 0.2244,
       "step": 2980
     },
     {
-      "epoch": 85.43,
-      "learning_rate": 5.948126801152738e-05,
-      "loss": 0.3371,
+      "epoch": 21.21,
+      "learning_rate": 7.888573456352023e-05,
+      "loss": 0.315,
       "step": 2990
     },
     {
-      "epoch": 85.71,
-      "learning_rate": 5.8328530259366e-05,
-      "loss": 0.3989,
+      "epoch": 21.28,
+      "learning_rate": 7.881476224272534e-05,
+      "loss": 0.4814,
       "step": 3000
     },
     {
-      "epoch": 85.99,
-      "learning_rate": 5.717579250720462e-05,
-      "loss": 0.4228,
-      "step": 3010
-    },
-    {
-      "epoch": 85.99,
-      "eval_loss": 0.7196153998374939,
-      "eval_runtime": 8.9305,
-      "eval_samples_per_second": 7.054,
-      "eval_steps_per_second": 1.792,
-      "eval_wer": 0.3745255166596373,
+      "epoch": 21.35,
+      "learning_rate": 7.874378992193045e-05,
+      "loss": 0.3235,
       "step": 3010
     },
     {
-      "epoch": 86.28,
-      "learning_rate": 5.602305475504323e-05,
-      "loss": 0.3931,
+      "epoch": 21.42,
+      "learning_rate": 7.867281760113557e-05,
+      "loss": 0.3563,
       "step": 3020
     },
     {
-      "epoch": 86.57,
-      "learning_rate": 5.4870317002881845e-05,
-      "loss": 0.3359,
+      "epoch": 21.49,
+      "learning_rate": 7.860184528034066e-05,
+      "loss": 0.3318,
       "step": 3030
     },
     {
-      "epoch": 86.85,
-      "learning_rate": 5.371757925072046e-05,
-      "loss": 0.3823,
+      "epoch": 21.56,
+      "learning_rate": 7.853087295954578e-05,
+      "loss": 0.361,
       "step": 3040
     },
     {
-      "epoch": 86.99,
-      "eval_loss": 0.7194885015487671,
-      "eval_runtime": 8.8773,
-      "eval_samples_per_second": 7.097,
-      "eval_steps_per_second": 1.802,
-      "eval_wer": 0.37536904259805987,
-      "step": 3045
-    },
-    {
-      "epoch": 87.14,
-      "learning_rate": 5.256484149855908e-05,
-      "loss": 0.3635,
+      "epoch": 21.63,
+      "learning_rate": 7.845990063875089e-05,
+      "loss": 0.2439,
       "step": 3050
     },
     {
-      "epoch": 87.43,
-      "learning_rate": 5.14121037463977e-05,
-      "loss": 0.3576,
+      "epoch": 21.7,
+      "learning_rate": 7.8388928317956e-05,
+      "loss": 0.3815,
       "step": 3060
     },
     {
-      "epoch": 87.71,
-      "learning_rate": 5.0259365994236316e-05,
-      "loss": 0.4654,
+      "epoch": 21.77,
+      "learning_rate": 7.83179559971611e-05,
+      "loss": 0.5302,
       "step": 3070
     },
     {
-      "epoch": 87.99,
-      "learning_rate": 4.9106628242074933e-05,
-      "loss": 0.3574,
-      "step": 3080
-    },
-    {
-      "epoch": 87.99,
-      "eval_loss": 0.713660478591919,
-      "eval_runtime": 8.8845,
-      "eval_samples_per_second": 7.091,
-      "eval_steps_per_second": 1.801,
-      "eval_wer": 0.3795866722901729,
+      "epoch": 21.84,
+      "learning_rate": 7.824698367636623e-05,
+      "loss": 0.2964,
       "step": 3080
     },
     {
-      "epoch": 88.28,
-      "learning_rate": 4.7953890489913544e-05,
-      "loss": 0.3677,
+      "epoch": 21.91,
+      "learning_rate": 7.817601135557133e-05,
+      "loss": 0.2972,
       "step": 3090
     },
     {
-      "epoch": 88.57,
-      "learning_rate": 4.680115273775216e-05,
-      "loss": 0.3065,
+      "epoch": 21.99,
+      "learning_rate": 7.810503903477644e-05,
+      "loss": 0.4013,
       "step": 3100
     },
     {
-      "epoch": 88.85,
-      "learning_rate": 4.564841498559078e-05,
-      "loss": 0.3371,
-      "step": 3110
+      "epoch": 22.0,
+      "eval_loss": 0.6899057030677795,
+      "eval_runtime": 8.432,
+      "eval_samples_per_second": 7.472,
+      "eval_steps_per_second": 3.795,
+      "eval_wer": 0.37916490932096164,
+      "step": 3102
     },
     {
-      "epoch": 88.99,
-      "eval_loss": 0.7164372801780701,
-      "eval_runtime": 8.865,
-      "eval_samples_per_second": 7.107,
-      "eval_steps_per_second": 1.805,
-      "eval_wer": 0.3762125685364825,
-      "step": 3115
+      "epoch": 22.06,
+      "learning_rate": 7.803406671398154e-05,
+      "loss": 0.3204,
+      "step": 3110
     },
     {
-      "epoch": 89.14,
-      "learning_rate": 4.44956772334294e-05,
-      "loss": 0.6018,
+      "epoch": 22.13,
+      "learning_rate": 7.796309439318667e-05,
+      "loss": 0.2901,
       "step": 3120
     },
     {
-      "epoch": 89.43,
-      "learning_rate": 4.3342939481268015e-05,
-      "loss": 0.3404,
+      "epoch": 22.2,
+      "learning_rate": 7.789212207239177e-05,
+      "loss": 0.3243,
       "step": 3130
     },
     {
-      "epoch": 89.71,
-      "learning_rate": 4.219020172910663e-05,
-      "loss": 0.3659,
+      "epoch": 22.27,
+      "learning_rate": 7.782114975159688e-05,
+      "loss": 0.3352,
       "step": 3140
     },
     {
-      "epoch": 89.99,
-      "learning_rate": 4.103746397694524e-05,
-      "loss": 0.3548,
-      "step": 3150
-    },
-    {
-      "epoch": 89.99,
-      "eval_loss": 0.7766362428665161,
-      "eval_runtime": 8.874,
-      "eval_samples_per_second": 7.099,
-      "eval_steps_per_second": 1.803,
-      "eval_wer": 0.37916490932096164,
+      "epoch": 22.34,
+      "learning_rate": 7.775017743080199e-05,
+      "loss": 0.3619,
       "step": 3150
     },
     {
-      "epoch": 90.28,
-      "learning_rate": 3.988472622478386e-05,
-      "loss": 0.3393,
+      "epoch": 22.41,
+      "learning_rate": 7.76792051100071e-05,
+      "loss": 0.2761,
       "step": 3160
     },
     {
-      "epoch": 90.57,
-      "learning_rate": 3.8731988472622485e-05,
-      "loss": 0.2986,
+      "epoch": 22.48,
+      "learning_rate": 7.760823278921222e-05,
+      "loss": 0.308,
       "step": 3170
     },
     {
-      "epoch": 90.85,
-      "learning_rate": 3.7579250720461096e-05,
-      "loss": 0.4042,
+      "epoch": 22.55,
+      "learning_rate": 7.753726046841732e-05,
+      "loss": 0.4181,
       "step": 3180
     },
     {
-      "epoch": 90.99,
-      "eval_loss": 0.7587747573852539,
-      "eval_runtime": 8.7904,
-      "eval_samples_per_second": 7.167,
-      "eval_steps_per_second": 1.82,
-      "eval_wer": 0.3766343315056938,
-      "step": 3185
-    },
-    {
-      "epoch": 91.14,
-      "learning_rate": 3.6426512968299714e-05,
-      "loss": 0.3223,
+      "epoch": 22.62,
+      "learning_rate": 7.746628814762243e-05,
+      "loss": 0.2802,
       "step": 3190
     },
     {
-      "epoch": 91.43,
-      "learning_rate": 3.527377521613833e-05,
-      "loss": 0.5178,
+      "epoch": 22.7,
+      "learning_rate": 7.739531582682754e-05,
+      "loss": 0.3316,
       "step": 3200
     },
     {
-      "epoch": 91.71,
-      "learning_rate": 3.412103746397695e-05,
-      "loss": 0.3521,
+      "epoch": 22.77,
+      "learning_rate": 7.732434350603265e-05,
+      "loss": 0.4193,
       "step": 3210
     },
     {
-      "epoch": 91.99,
-      "learning_rate": 3.296829971181557e-05,
-      "loss": 0.3989,
-      "step": 3220
-    },
-    {
-      "epoch": 91.99,
-      "eval_loss": 0.7310720086097717,
-      "eval_runtime": 8.8614,
-      "eval_samples_per_second": 7.109,
-      "eval_steps_per_second": 1.806,
-      "eval_wer": 0.3774778574441164,
+      "epoch": 22.84,
+      "learning_rate": 7.725337118523777e-05,
+      "loss": 0.2907,
       "step": 3220
     },
     {
-      "epoch": 92.28,
-      "learning_rate": 3.181556195965418e-05,
-      "loss": 0.329,
+      "epoch": 22.91,
+      "learning_rate": 7.718239886444288e-05,
+      "loss": 0.2809,
       "step": 3230
     },
     {
-      "epoch": 92.57,
-      "learning_rate": 3.0662824207492795e-05,
-      "loss": 0.3594,
+      "epoch": 22.98,
+      "learning_rate": 7.711142654364798e-05,
+      "loss": 0.5134,
       "step": 3240
     },
     {
-      "epoch": 92.85,
-      "learning_rate": 2.9510086455331416e-05,
-      "loss": 0.3625,
-      "step": 3250
+      "epoch": 23.0,
+      "eval_loss": 0.6571967005729675,
+      "eval_runtime": 8.5678,
+      "eval_samples_per_second": 7.353,
+      "eval_steps_per_second": 3.735,
+      "eval_wer": 0.3787431463517503,
+      "step": 3243
     },
     {
-      "epoch": 92.99,
-      "eval_loss": 0.7475365400314331,
-      "eval_runtime": 8.8406,
-      "eval_samples_per_second": 7.126,
-      "eval_steps_per_second": 1.81,
-      "eval_wer": 0.3745255166596373,
-      "step": 3255
+      "epoch": 23.05,
+      "learning_rate": 7.704045422285309e-05,
+      "loss": 0.2606,
+      "step": 3250
     },
     {
-      "epoch": 93.14,
-      "learning_rate": 2.8357348703170027e-05,
-      "loss": 0.3361,
+      "epoch": 23.12,
+      "learning_rate": 7.69694819020582e-05,
+      "loss": 0.2414,
       "step": 3260
     },
     {
-      "epoch": 93.43,
-      "learning_rate": 2.7204610951008648e-05,
-      "loss": 0.3521,
+      "epoch": 23.19,
+      "learning_rate": 7.689850958126332e-05,
+      "loss": 0.2561,
       "step": 3270
     },
     {
-      "epoch": 93.71,
-      "learning_rate": 2.6051873198847266e-05,
-      "loss": 0.3261,
+      "epoch": 23.26,
+      "learning_rate": 7.682753726046842e-05,
+      "loss": 0.3898,
       "step": 3280
     },
     {
-      "epoch": 93.99,
-      "learning_rate": 2.489913544668588e-05,
-      "loss": 0.3036,
-      "step": 3290
-    },
-    {
-      "epoch": 93.99,
-      "eval_loss": 0.7138022184371948,
-      "eval_runtime": 8.8395,
-      "eval_samples_per_second": 7.127,
-      "eval_steps_per_second": 1.81,
-      "eval_wer": 0.37157317587515815,
+      "epoch": 23.33,
+      "learning_rate": 7.675656493967354e-05,
+      "loss": 0.2709,
       "step": 3290
     },
     {
-      "epoch": 94.28,
-      "learning_rate": 2.3746397694524498e-05,
-      "loss": 0.3412,
+      "epoch": 23.4,
+      "learning_rate": 7.668559261887864e-05,
+      "loss": 0.385,
       "step": 3300
     },
     {
-      "epoch": 94.57,
-      "learning_rate": 2.2593659942363112e-05,
-      "loss": 0.3451,
+      "epoch": 23.48,
+      "learning_rate": 7.661462029808375e-05,
+      "loss": 0.3725,
       "step": 3310
     },
     {
-      "epoch": 94.85,
-      "learning_rate": 2.144092219020173e-05,
-      "loss": 0.5157,
+      "epoch": 23.55,
+      "learning_rate": 7.654364797728885e-05,
+      "loss": 0.4937,
       "step": 3320
     },
     {
-      "epoch": 94.99,
-      "eval_loss": 0.7245524525642395,
-      "eval_runtime": 8.8889,
-      "eval_samples_per_second": 7.087,
-      "eval_steps_per_second": 1.8,
-      "eval_wer": 0.3787431463517503,
-      "step": 3325
-    },
-    {
-      "epoch": 95.14,
-      "learning_rate": 2.0288184438040347e-05,
-      "loss": 0.3089,
+      "epoch": 23.62,
+      "learning_rate": 7.647267565649398e-05,
+      "loss": 0.2654,
       "step": 3330
     },
     {
-      "epoch": 95.43,
-      "learning_rate": 1.9135446685878965e-05,
-      "loss": 0.3115,
+      "epoch": 23.69,
+      "learning_rate": 7.640170333569908e-05,
+      "loss": 0.2891,
       "step": 3340
     },
     {
-      "epoch": 95.71,
-      "learning_rate": 1.798270893371758e-05,
-      "loss": 0.3742,
+      "epoch": 23.76,
+      "learning_rate": 7.633073101490419e-05,
+      "loss": 0.3545,
       "step": 3350
     },
     {
-      "epoch": 95.99,
-      "learning_rate": 1.6829971181556197e-05,
-      "loss": 0.4072,
+      "epoch": 23.83,
+      "learning_rate": 7.625975869410929e-05,
+      "loss": 0.3639,
       "step": 3360
     },
     {
-      "epoch": 95.99,
-      "eval_loss": 0.7321768999099731,
-      "eval_runtime": 8.9645,
-      "eval_samples_per_second": 7.028,
-      "eval_steps_per_second": 1.785,
-      "eval_wer": 0.3762125685364825,
-      "step": 3360
-    },
-    {
-      "epoch": 96.28,
-      "learning_rate": 1.5677233429394814e-05,
-      "loss": 0.3496,
+      "epoch": 23.9,
+      "learning_rate": 7.618878637331442e-05,
+      "loss": 0.2874,
       "step": 3370
     },
     {
-      "epoch": 96.57,
-      "learning_rate": 1.452449567723343e-05,
-      "loss": 0.4089,
+      "epoch": 23.97,
+      "learning_rate": 7.611781405251952e-05,
+      "loss": 0.3144,
       "step": 3380
     },
     {
-      "epoch": 96.85,
-      "learning_rate": 1.3371757925072045e-05,
-      "loss": 0.3406,
-      "step": 3390
+      "epoch": 24.0,
+      "eval_loss": 0.5881998538970947,
+      "eval_runtime": 8.4771,
+      "eval_samples_per_second": 7.432,
+      "eval_steps_per_second": 3.775,
+      "eval_wer": 0.35428089413749475,
+      "step": 3384
     },
     {
-      "epoch": 96.99,
-      "eval_loss": 0.7133801579475403,
-      "eval_runtime": 8.9109,
-      "eval_samples_per_second": 7.07,
-      "eval_steps_per_second": 1.796,
-      "eval_wer": 0.3770560944749051,
-      "step": 3395
+      "epoch": 24.04,
+      "learning_rate": 7.604684173172463e-05,
+      "loss": 0.3707,
+      "step": 3390
     },
     {
-      "epoch": 97.14,
-      "learning_rate": 1.2219020172910664e-05,
-      "loss": 0.3305,
+      "epoch": 24.11,
+      "learning_rate": 7.597586941092974e-05,
+      "loss": 0.2284,
       "step": 3400
     },
     {
-      "epoch": 97.43,
-      "learning_rate": 1.106628242074928e-05,
-      "loss": 0.3521,
+      "epoch": 24.18,
+      "learning_rate": 7.590489709013485e-05,
+      "loss": 0.4784,
       "step": 3410
     },
     {
-      "epoch": 97.71,
-      "learning_rate": 9.913544668587897e-06,
-      "loss": 0.3405,
+      "epoch": 24.26,
+      "learning_rate": 7.583392476933995e-05,
+      "loss": 0.3584,
       "step": 3420
     },
     {
-      "epoch": 97.99,
-      "learning_rate": 8.760806916426513e-06,
-      "loss": 0.2987,
-      "step": 3430
-    },
-    {
-      "epoch": 97.99,
-      "eval_loss": 0.69514000415802,
-      "eval_runtime": 8.9709,
-      "eval_samples_per_second": 7.023,
-      "eval_steps_per_second": 1.784,
-      "eval_wer": 0.37536904259805987,
+      "epoch": 24.33,
+      "learning_rate": 7.576295244854507e-05,
+      "loss": 0.3331,
       "step": 3430
     },
     {
-      "epoch": 98.28,
-      "learning_rate": 7.60806916426513e-06,
-      "loss": 0.3394,
+      "epoch": 24.4,
+      "learning_rate": 7.569198012775018e-05,
+      "loss": 0.2717,
       "step": 3440
     },
     {
-      "epoch": 98.57,
-      "learning_rate": 6.455331412103746e-06,
-      "loss": 0.3566,
+      "epoch": 24.47,
+      "learning_rate": 7.562100780695529e-05,
+      "loss": 0.3537,
       "step": 3450
     },
     {
-      "epoch": 98.85,
-      "learning_rate": 5.302593659942364e-06,
-      "loss": 0.3355,
+      "epoch": 24.54,
+      "learning_rate": 7.55500354861604e-05,
+      "loss": 0.3631,
       "step": 3460
     },
     {
-      "epoch": 98.99,
-      "eval_loss": 0.7004749178886414,
-      "eval_runtime": 8.9802,
-      "eval_samples_per_second": 7.015,
-      "eval_steps_per_second": 1.782,
-      "eval_wer": 0.3766343315056938,
-      "step": 3465
-    },
-    {
-      "epoch": 99.14,
-      "learning_rate": 4.14985590778098e-06,
-      "loss": 0.3434,
+      "epoch": 24.61,
+      "learning_rate": 7.547906316536552e-05,
+      "loss": 0.283,
       "step": 3470
     },
     {
-      "epoch": 99.43,
-      "learning_rate": 2.9971181556195965e-06,
-      "loss": 0.3309,
+      "epoch": 24.68,
+      "learning_rate": 7.540809084457063e-05,
+      "loss": 0.258,
       "step": 3480
     },
     {
-      "epoch": 99.71,
-      "learning_rate": 1.8443804034582135e-06,
-      "loss": 0.369,
+      "epoch": 24.75,
+      "learning_rate": 7.533711852377573e-05,
+      "loss": 0.3421,
       "step": 3490
     },
     {
-      "epoch": 99.99,
-      "learning_rate": 6.9164265129683e-07,
-      "loss": 0.341,
+      "epoch": 24.82,
+      "learning_rate": 7.526614620298084e-05,
+      "loss": 0.3671,
       "step": 3500
     },
     {
-      "epoch": 99.99,
-      "eval_loss": 0.6989557147026062,
-      "eval_runtime": 8.9073,
-      "eval_samples_per_second": 7.073,
-      "eval_steps_per_second": 1.796,
-      "eval_wer": 0.3770560944749051,
-      "step": 3500
+      "epoch": 24.89,
+      "learning_rate": 7.519517388218595e-05,
+      "loss": 0.3133,
+      "step": 3510
+    },
+    {
+      "epoch": 24.96,
+      "learning_rate": 7.512420156139107e-05,
+      "loss": 0.3534,
+      "step": 3520
+    },
+    {
+      "epoch": 25.0,
+      "eval_loss": 0.5660640597343445,
+      "eval_runtime": 8.4653,
+      "eval_samples_per_second": 7.442,
+      "eval_steps_per_second": 3.78,
+      "eval_wer": 0.34162800506115565,
+      "step": 3525
+    },
+    {
+      "epoch": 25.04,
+      "learning_rate": 7.505322924059617e-05,
+      "loss": 0.282,
+      "step": 3530
+    },
+    {
+      "epoch": 25.11,
+      "learning_rate": 7.498225691980129e-05,
+      "loss": 0.3512,
+      "step": 3540
+    },
+    {
+      "epoch": 25.18,
+      "learning_rate": 7.491128459900639e-05,
+      "loss": 0.3079,
+      "step": 3550
+    },
+    {
+      "epoch": 25.25,
+      "learning_rate": 7.48403122782115e-05,
+      "loss": 0.2728,
+      "step": 3560
+    },
+    {
+      "epoch": 25.32,
+      "learning_rate": 7.47693399574166e-05,
+      "loss": 0.3536,
+      "step": 3570
+    },
+    {
+      "epoch": 25.39,
+      "learning_rate": 7.469836763662173e-05,
+      "loss": 0.3129,
+      "step": 3580
+    },
+    {
+      "epoch": 25.46,
+      "learning_rate": 7.462739531582683e-05,
+      "loss": 0.3502,
+      "step": 3590
+    },
+    {
+      "epoch": 25.53,
+      "learning_rate": 7.455642299503194e-05,
+      "loss": 0.3017,
+      "step": 3600
+    },
+    {
+      "epoch": 25.6,
+      "learning_rate": 7.448545067423705e-05,
+      "loss": 0.3009,
+      "step": 3610
+    },
+    {
+      "epoch": 25.67,
+      "learning_rate": 7.441447835344217e-05,
+      "loss": 0.2406,
+      "step": 3620
+    },
+    {
+      "epoch": 25.74,
+      "learning_rate": 7.434350603264727e-05,
+      "loss": 1.3185,
+      "step": 3630
+    },
+    {
+      "epoch": 25.82,
+      "learning_rate": 7.427253371185238e-05,
+      "loss": 0.4062,
+      "step": 3640
+    },
+    {
+      "epoch": 25.89,
+      "learning_rate": 7.420156139105749e-05,
+      "loss": 0.2571,
+      "step": 3650
+    },
+    {
+      "epoch": 25.96,
+      "learning_rate": 7.41305890702626e-05,
+      "loss": 0.2555,
+      "step": 3660
+    },
+    {
+      "epoch": 26.0,
+      "eval_loss": 0.5976953506469727,
+      "eval_runtime": 8.5028,
+      "eval_samples_per_second": 7.409,
+      "eval_steps_per_second": 3.763,
+      "eval_wer": 0.35892028679881904,
+      "step": 3666
+    },
+    {
+      "epoch": 26.03,
+      "learning_rate": 7.40596167494677e-05,
+      "loss": 0.3319,
+      "step": 3670
+    },
+    {
+      "epoch": 26.1,
+      "learning_rate": 7.398864442867283e-05,
+      "loss": 0.3536,
+      "step": 3680
+    },
+    {
+      "epoch": 26.17,
+      "learning_rate": 7.391767210787793e-05,
+      "loss": 0.2111,
+      "step": 3690
+    },
+    {
+      "epoch": 26.24,
+      "learning_rate": 7.384669978708304e-05,
+      "loss": 0.3994,
+      "step": 3700
+    },
+    {
+      "epoch": 26.31,
+      "learning_rate": 7.377572746628814e-05,
+      "loss": 0.5157,
+      "step": 3710
+    },
+    {
+      "epoch": 26.38,
+      "learning_rate": 7.370475514549327e-05,
+      "loss": 0.2034,
+      "step": 3720
+    },
+    {
+      "epoch": 26.45,
+      "learning_rate": 7.363378282469837e-05,
+      "loss": 0.3259,
+      "step": 3730
+    },
+    {
+      "epoch": 26.52,
+      "learning_rate": 7.356281050390348e-05,
+      "loss": 0.358,
+      "step": 3740
+    },
+    {
+      "epoch": 26.6,
+      "learning_rate": 7.349183818310859e-05,
+      "loss": 0.2461,
+      "step": 3750
+    },
+    {
+      "epoch": 26.67,
+      "learning_rate": 7.34208658623137e-05,
+      "loss": 0.3318,
+      "step": 3760
+    },
+    {
+      "epoch": 26.74,
+      "learning_rate": 7.334989354151882e-05,
+      "loss": 0.3374,
+      "step": 3770
+    },
+    {
+      "epoch": 26.81,
+      "learning_rate": 7.327892122072392e-05,
+      "loss": 0.263,
+      "step": 3780
+    },
+    {
+      "epoch": 26.88,
+      "learning_rate": 7.320794889992903e-05,
+      "loss": 0.2467,
+      "step": 3790
+    },
+    {
+      "epoch": 26.95,
+      "learning_rate": 7.313697657913414e-05,
+      "loss": 0.3524,
+      "step": 3800
+    },
+    {
+      "epoch": 27.0,
+      "eval_loss": 0.595256507396698,
+      "eval_runtime": 8.4556,
+      "eval_samples_per_second": 7.451,
+      "eval_steps_per_second": 3.784,
+      "eval_wer": 0.3584985238296078,
+      "step": 3807
+    },
+    {
+      "epoch": 27.02,
+      "learning_rate": 7.306600425833925e-05,
+      "loss": 0.3722,
+      "step": 3810
+    },
+    {
+      "epoch": 27.09,
+      "learning_rate": 7.299503193754435e-05,
+      "loss": 0.2954,
+      "step": 3820
+    },
+    {
+      "epoch": 27.16,
+      "learning_rate": 7.292405961674948e-05,
+      "loss": 0.2942,
+      "step": 3830
+    },
+    {
+      "epoch": 27.23,
+      "learning_rate": 7.285308729595458e-05,
+      "loss": 0.3869,
+      "step": 3840
+    },
+    {
+      "epoch": 27.3,
+      "learning_rate": 7.278211497515969e-05,
+      "loss": 0.3699,
+      "step": 3850
+    },
+    {
+      "epoch": 27.38,
+      "learning_rate": 7.27111426543648e-05,
+      "loss": 0.209,
+      "step": 3860
+    },
+    {
+      "epoch": 27.45,
+      "learning_rate": 7.264017033356992e-05,
+      "loss": 0.325,
+      "step": 3870
+    },
+    {
+      "epoch": 27.52,
+      "learning_rate": 7.256919801277502e-05,
+      "loss": 0.3722,
+      "step": 3880
+    },
+    {
+      "epoch": 27.59,
+      "learning_rate": 7.249822569198013e-05,
+      "loss": 0.3216,
+      "step": 3890
+    },
+    {
+      "epoch": 27.66,
+      "learning_rate": 7.242725337118524e-05,
+      "loss": 0.2944,
+      "step": 3900
+    },
+    {
+      "epoch": 27.73,
+      "learning_rate": 7.235628105039035e-05,
+      "loss": 0.3565,
+      "step": 3910
+    },
+    {
+      "epoch": 27.8,
+      "learning_rate": 7.228530872959545e-05,
+      "loss": 0.4623,
+      "step": 3920
+    },
+    {
+      "epoch": 27.87,
+      "learning_rate": 7.221433640880058e-05,
+      "loss": 0.3105,
+      "step": 3930
+    },
+    {
+      "epoch": 27.94,
+      "learning_rate": 7.214336408800568e-05,
+      "loss": 0.314,
+      "step": 3940
+    },
+    {
+      "epoch": 28.0,
+      "eval_loss": 0.6359146237373352,
+      "eval_runtime": 8.3572,
+      "eval_samples_per_second": 7.538,
+      "eval_steps_per_second": 3.829,
+      "eval_wer": 0.35934204976803036,
+      "step": 3948
+    },
+    {
+      "epoch": 28.01,
+      "learning_rate": 7.207239176721079e-05,
+      "loss": 0.3712,
+      "step": 3950
+    },
+    {
+      "epoch": 28.09,
+      "learning_rate": 7.200141944641589e-05,
+      "loss": 0.3517,
+      "step": 3960
+    },
+    {
+      "epoch": 28.16,
+      "learning_rate": 7.193044712562102e-05,
+      "loss": 0.2737,
+      "step": 3970
+    },
+    {
+      "epoch": 28.23,
+      "learning_rate": 7.185947480482612e-05,
+      "loss": 0.3002,
+      "step": 3980
+    },
+    {
+      "epoch": 28.3,
+      "learning_rate": 7.178850248403123e-05,
+      "loss": 0.3866,
+      "step": 3990
+    },
+    {
+      "epoch": 28.37,
+      "learning_rate": 7.171753016323634e-05,
+      "loss": 0.2567,
+      "step": 4000
+    },
+    {
+      "epoch": 28.44,
+      "learning_rate": 7.164655784244145e-05,
+      "loss": 0.3106,
+      "step": 4010
+    },
+    {
+      "epoch": 28.51,
+      "learning_rate": 7.157558552164655e-05,
+      "loss": 0.3031,
+      "step": 4020
+    },
+    {
+      "epoch": 28.58,
+      "learning_rate": 7.150461320085167e-05,
+      "loss": 0.2936,
+      "step": 4030
+    },
+    {
+      "epoch": 28.65,
+      "learning_rate": 7.143364088005678e-05,
+      "loss": 0.3149,
+      "step": 4040
+    },
+    {
+      "epoch": 28.72,
+      "learning_rate": 7.136266855926189e-05,
+      "loss": 0.3877,
+      "step": 4050
+    },
+    {
+      "epoch": 28.79,
+      "learning_rate": 7.1291696238467e-05,
+      "loss": 0.3484,
+      "step": 4060
+    },
+    {
+      "epoch": 28.87,
+      "learning_rate": 7.122072391767212e-05,
+      "loss": 0.2438,
+      "step": 4070
+    },
+    {
+      "epoch": 28.94,
+      "learning_rate": 7.114975159687722e-05,
+      "loss": 0.2565,
+      "step": 4080
+    },
+    {
+      "epoch": 29.0,
+      "eval_loss": 0.619179904460907,
+      "eval_runtime": 8.4385,
+      "eval_samples_per_second": 7.466,
+      "eval_steps_per_second": 3.792,
+      "eval_wer": 0.3614508646140869,
+      "step": 4089
+    },
+    {
+      "epoch": 29.01,
+      "learning_rate": 7.107877927608233e-05,
+      "loss": 0.2524,
+      "step": 4090
+    },
+    {
+      "epoch": 29.08,
+      "learning_rate": 7.100780695528744e-05,
+      "loss": 0.28,
+      "step": 4100
+    },
+    {
+      "epoch": 29.15,
+      "learning_rate": 7.093683463449255e-05,
+      "loss": 0.2847,
+      "step": 4110
+    },
+    {
+      "epoch": 29.22,
+      "learning_rate": 7.086586231369767e-05,
+      "loss": 0.2755,
+      "step": 4120
+    },
+    {
+      "epoch": 29.29,
+      "learning_rate": 7.079488999290277e-05,
+      "loss": 0.4346,
+      "step": 4130
+    },
+    {
+      "epoch": 29.36,
+      "learning_rate": 7.072391767210789e-05,
+      "loss": 0.2375,
+      "step": 4140
+    },
+    {
+      "epoch": 29.43,
+      "learning_rate": 7.065294535131299e-05,
+      "loss": 0.3072,
+      "step": 4150
+    },
+    {
+      "epoch": 29.5,
+      "learning_rate": 7.05819730305181e-05,
+      "loss": 1.0986,
+      "step": 4160
+    },
+    {
+      "epoch": 29.57,
+      "learning_rate": 7.05110007097232e-05,
+      "loss": 0.3082,
+      "step": 4170
+    },
+    {
+      "epoch": 29.65,
+      "learning_rate": 7.044002838892833e-05,
+      "loss": 0.3417,
+      "step": 4180
+    },
+    {
+      "epoch": 29.72,
+      "learning_rate": 7.036905606813343e-05,
+      "loss": 0.3263,
+      "step": 4190
+    },
+    {
+      "epoch": 29.79,
+      "learning_rate": 7.029808374733854e-05,
+      "loss": 0.2604,
+      "step": 4200
+    },
+    {
+      "epoch": 29.86,
+      "learning_rate": 7.022711142654364e-05,
+      "loss": 0.2549,
+      "step": 4210
+    },
+    {
+      "epoch": 29.93,
+      "learning_rate": 7.015613910574877e-05,
+      "loss": 0.2332,
+      "step": 4220
+    },
+    {
+      "epoch": 30.0,
+      "learning_rate": 7.008516678495387e-05,
+      "loss": 0.5023,
+      "step": 4230
+    },
+    {
+      "epoch": 30.0,
+      "eval_loss": 0.6229135394096375,
+      "eval_runtime": 8.3954,
+      "eval_samples_per_second": 7.504,
+      "eval_steps_per_second": 3.812,
+      "eval_wer": 0.33783213833825393,
+      "step": 4230
+    },
+    {
+      "epoch": 30.07,
+      "learning_rate": 7.001419446415898e-05,
+      "loss": 0.3438,
+      "step": 4240
+    },
+    {
+      "epoch": 30.14,
+      "learning_rate": 6.994322214336409e-05,
+      "loss": 0.2485,
+      "step": 4250
+    },
+    {
+      "epoch": 30.21,
+      "learning_rate": 6.98722498225692e-05,
+      "loss": 0.3658,
+      "step": 4260
+    },
+    {
+      "epoch": 30.28,
+      "learning_rate": 6.98012775017743e-05,
+      "loss": 0.3625,
+      "step": 4270
+    },
+    {
+      "epoch": 30.35,
+      "learning_rate": 6.973030518097942e-05,
+      "loss": 0.2155,
+      "step": 4280
+    },
+    {
+      "epoch": 30.43,
+      "learning_rate": 6.965933286018453e-05,
+      "loss": 0.2392,
+      "step": 4290
+    },
+    {
+      "epoch": 30.5,
+      "learning_rate": 6.958836053938964e-05,
+      "loss": 0.4407,
+      "step": 4300
+    },
+    {
+      "epoch": 30.57,
+      "learning_rate": 6.951738821859474e-05,
+      "loss": 0.3791,
+      "step": 4310
+    },
+    {
+      "epoch": 30.64,
+      "learning_rate": 6.944641589779987e-05,
+      "loss": 0.28,
+      "step": 4320
+    },
+    {
+      "epoch": 30.71,
+      "learning_rate": 6.937544357700497e-05,
+      "loss": 0.3249,
+      "step": 4330
+    },
+    {
+      "epoch": 30.78,
+      "learning_rate": 6.930447125621008e-05,
+      "loss": 0.4017,
+      "step": 4340
+    },
+    {
+      "epoch": 30.85,
+      "learning_rate": 6.923349893541519e-05,
+      "loss": 0.2568,
+      "step": 4350
+    },
+    {
+      "epoch": 30.92,
+      "learning_rate": 6.91625266146203e-05,
+      "loss": 0.2721,
+      "step": 4360
+    },
+    {
+      "epoch": 30.99,
+      "learning_rate": 6.909155429382542e-05,
+      "loss": 0.3025,
+      "step": 4370
+    },
+    {
+      "epoch": 31.0,
+      "eval_loss": 0.6002037525177002,
+      "eval_runtime": 8.385,
+      "eval_samples_per_second": 7.513,
+      "eval_steps_per_second": 3.816,
+      "eval_wer": 0.3441585828764234,
+      "step": 4371
+    },
+    {
+      "epoch": 31.06,
+      "learning_rate": 6.902058197303052e-05,
+      "loss": 0.3279,
+      "step": 4380
+    },
+    {
+      "epoch": 31.13,
+      "learning_rate": 6.894960965223563e-05,
+      "loss": 0.2087,
+      "step": 4390
+    },
+    {
+      "epoch": 31.21,
+      "learning_rate": 6.887863733144074e-05,
+      "loss": 0.5199,
+      "step": 4400
+    },
+    {
+      "epoch": 31.28,
+      "learning_rate": 6.880766501064586e-05,
+      "loss": 0.3573,
+      "step": 4410
+    },
+    {
+      "epoch": 31.35,
+      "learning_rate": 6.873669268985095e-05,
+      "loss": 0.2642,
+      "step": 4420
+    },
+    {
+      "epoch": 31.42,
+      "learning_rate": 6.866572036905608e-05,
+      "loss": 0.2947,
+      "step": 4430
+    },
+    {
+      "epoch": 31.49,
+      "learning_rate": 6.859474804826118e-05,
+      "loss": 0.3337,
+      "step": 4440
+    },
+    {
+      "epoch": 31.56,
+      "learning_rate": 6.852377572746629e-05,
+      "loss": 0.3989,
+      "step": 4450
+    },
+    {
+      "epoch": 31.63,
+      "learning_rate": 6.84528034066714e-05,
+      "loss": 0.2785,
+      "step": 4460
+    },
+    {
+      "epoch": 31.7,
+      "learning_rate": 6.838183108587652e-05,
+      "loss": 0.3,
+      "step": 4470
+    },
+    {
+      "epoch": 31.77,
+      "learning_rate": 6.831085876508162e-05,
+      "loss": 0.3592,
+      "step": 4480
+    },
+    {
+      "epoch": 31.84,
+      "learning_rate": 6.823988644428673e-05,
+      "loss": 0.2559,
+      "step": 4490
+    },
+    {
+      "epoch": 31.91,
+      "learning_rate": 6.816891412349184e-05,
+      "loss": 0.2456,
+      "step": 4500
+    },
+    {
+      "epoch": 31.99,
+      "learning_rate": 6.809794180269696e-05,
+      "loss": 0.3329,
+      "step": 4510
+    },
+    {
+      "epoch": 32.0,
+      "eval_loss": 0.6234570145606995,
+      "eval_runtime": 8.4148,
+      "eval_samples_per_second": 7.487,
+      "eval_steps_per_second": 3.803,
+      "eval_wer": 0.3513285533530156,
+      "step": 4512
+    },
+    {
+      "epoch": 32.06,
+      "learning_rate": 6.802696948190205e-05,
+      "loss": 0.2698,
+      "step": 4520
+    },
+    {
+      "epoch": 32.13,
+      "learning_rate": 6.795599716110718e-05,
+      "loss": 0.2643,
+      "step": 4530
+    },
+    {
+      "epoch": 32.2,
+      "learning_rate": 6.788502484031228e-05,
+      "loss": 0.4749,
+      "step": 4540
+    },
+    {
+      "epoch": 32.27,
+      "learning_rate": 6.781405251951739e-05,
+      "loss": 0.3372,
+      "step": 4550
+    },
+    {
+      "epoch": 32.34,
+      "learning_rate": 6.774308019872249e-05,
+      "loss": 0.3082,
+      "step": 4560
+    },
+    {
+      "epoch": 32.41,
+      "learning_rate": 6.767210787792762e-05,
+      "loss": 0.243,
+      "step": 4570
+    },
+    {
+      "epoch": 32.48,
+      "learning_rate": 6.760113555713272e-05,
+      "loss": 0.4117,
+      "step": 4580
+    },
+    {
+      "epoch": 32.55,
+      "learning_rate": 6.753016323633783e-05,
+      "loss": 0.3074,
+      "step": 4590
+    },
+    {
+      "epoch": 32.62,
+      "learning_rate": 6.745919091554294e-05,
+      "loss": 0.3157,
+      "step": 4600
+    },
+    {
+      "epoch": 32.7,
+      "learning_rate": 6.738821859474806e-05,
+      "loss": 0.2848,
+      "step": 4610
+    },
+    {
+      "epoch": 32.77,
+      "learning_rate": 6.731724627395315e-05,
+      "loss": 0.3737,
+      "step": 4620
+    },
+    {
+      "epoch": 32.84,
+      "learning_rate": 6.724627395315827e-05,
+      "loss": 0.3029,
+      "step": 4630
+    },
+    {
+      "epoch": 32.91,
+      "learning_rate": 6.717530163236338e-05,
+      "loss": 0.2747,
+      "step": 4640
+    },
+    {
+      "epoch": 32.98,
+      "learning_rate": 6.710432931156849e-05,
+      "loss": 0.3744,
+      "step": 4650
+    },
+    {
+      "epoch": 33.0,
+      "eval_loss": 0.5781772136688232,
+      "eval_runtime": 8.6325,
+      "eval_samples_per_second": 7.298,
+      "eval_steps_per_second": 3.707,
+      "eval_wer": 0.34162800506115565,
+      "step": 4653
+    },
+    {
+      "epoch": 33.05,
+      "learning_rate": 6.70333569907736e-05,
+      "loss": 0.4845,
+      "step": 4660
+    },
+    {
+      "epoch": 33.12,
+      "learning_rate": 6.69623846699787e-05,
+      "loss": 0.2426,
+      "step": 4670
+    },
+    {
+      "epoch": 33.19,
+      "learning_rate": 6.689141234918382e-05,
+      "loss": 0.3018,
+      "step": 4680
+    },
+    {
+      "epoch": 33.26,
+      "learning_rate": 6.682044002838893e-05,
+      "loss": 0.2633,
+      "step": 4690
+    },
+    {
+      "epoch": 33.33,
+      "learning_rate": 6.674946770759404e-05,
+      "loss": 0.3021,
+      "step": 4700
+    },
+    {
+      "epoch": 33.4,
+      "learning_rate": 6.667849538679916e-05,
+      "loss": 0.2766,
+      "step": 4710
+    },
+    {
+      "epoch": 33.48,
+      "learning_rate": 6.660752306600427e-05,
+      "loss": 0.2843,
+      "step": 4720
+    },
+    {
+      "epoch": 33.55,
+      "learning_rate": 6.653655074520937e-05,
+      "loss": 0.3754,
+      "step": 4730
+    },
+    {
+      "epoch": 33.62,
+      "learning_rate": 6.646557842441448e-05,
+      "loss": 0.3603,
+      "step": 4740
+    },
+    {
+      "epoch": 33.69,
+      "learning_rate": 6.639460610361959e-05,
+      "loss": 0.3315,
+      "step": 4750
+    },
+    {
+      "epoch": 33.76,
+      "learning_rate": 6.63236337828247e-05,
+      "loss": 0.3918,
+      "step": 4760
+    },
+    {
+      "epoch": 33.83,
+      "learning_rate": 6.62526614620298e-05,
+      "loss": 0.3538,
+      "step": 4770
+    },
+    {
+      "epoch": 33.9,
+      "learning_rate": 6.618168914123493e-05,
+      "loss": 0.2728,
+      "step": 4780
+    },
+    {
+      "epoch": 33.97,
+      "learning_rate": 6.611071682044003e-05,
+      "loss": 0.2899,
+      "step": 4790
+    },
+    {
+      "epoch": 34.0,
+      "eval_loss": 0.5834619402885437,
+      "eval_runtime": 8.5973,
+      "eval_samples_per_second": 7.328,
+      "eval_steps_per_second": 3.722,
+      "eval_wer": 0.3336145086461409,
+      "step": 4794
+    },
+    {
+      "epoch": 34.04,
+      "learning_rate": 6.603974449964514e-05,
+      "loss": 0.2843,
+      "step": 4800
+    },
+    {
+      "epoch": 34.11,
+      "learning_rate": 6.596877217885024e-05,
+      "loss": 0.2314,
+      "step": 4810
+    },
+    {
+      "epoch": 34.18,
+      "learning_rate": 6.589779985805537e-05,
+      "loss": 0.3179,
+      "step": 4820
+    },
+    {
+      "epoch": 34.26,
+      "learning_rate": 6.582682753726047e-05,
+      "loss": 0.3155,
+      "step": 4830
+    },
+    {
+      "epoch": 34.33,
+      "learning_rate": 6.575585521646558e-05,
+      "loss": 0.3545,
+      "step": 4840
+    },
+    {
+      "epoch": 34.4,
+      "learning_rate": 6.568488289567069e-05,
+      "loss": 0.2801,
+      "step": 4850
+    },
+    {
+      "epoch": 34.47,
+      "learning_rate": 6.56139105748758e-05,
+      "loss": 0.2635,
+      "step": 4860
+    },
+    {
+      "epoch": 34.54,
+      "learning_rate": 6.55429382540809e-05,
+      "loss": 0.3329,
+      "step": 4870
+    },
+    {
+      "epoch": 34.61,
+      "learning_rate": 6.547196593328602e-05,
+      "loss": 0.4071,
+      "step": 4880
+    },
+    {
+      "epoch": 34.68,
+      "learning_rate": 6.540099361249113e-05,
+      "loss": 0.2843,
+      "step": 4890
+    },
+    {
+      "epoch": 34.75,
+      "learning_rate": 6.533002129169624e-05,
+      "loss": 0.3218,
+      "step": 4900
+    },
+    {
+      "epoch": 34.82,
+      "learning_rate": 6.525904897090134e-05,
+      "loss": 0.3113,
+      "step": 4910
+    },
+    {
+      "epoch": 34.89,
+      "learning_rate": 6.518807665010647e-05,
+      "loss": 0.2122,
+      "step": 4920
+    },
+    {
+      "epoch": 34.96,
+      "learning_rate": 6.511710432931157e-05,
+      "loss": 0.306,
+      "step": 4930
+    },
+    {
+      "epoch": 35.0,
+      "eval_loss": 0.606121838092804,
+      "eval_runtime": 8.6182,
+      "eval_samples_per_second": 7.31,
+      "eval_steps_per_second": 3.713,
+      "eval_wer": 0.3496415014761704,
+      "step": 4935
+    },
+    {
+      "epoch": 35.04,
+      "learning_rate": 6.504613200851668e-05,
+      "loss": 0.2875,
+      "step": 4940
+    },
+    {
+      "epoch": 35.11,
+      "learning_rate": 6.49751596877218e-05,
+      "loss": 0.2454,
+      "step": 4950
+    },
+    {
+      "epoch": 35.18,
+      "learning_rate": 6.49041873669269e-05,
+      "loss": 0.3266,
+      "step": 4960
+    },
+    {
+      "epoch": 35.25,
+      "learning_rate": 6.483321504613202e-05,
+      "loss": 0.4325,
+      "step": 4970
+    },
+    {
+      "epoch": 35.32,
+      "learning_rate": 6.476224272533712e-05,
+      "loss": 0.3243,
+      "step": 4980
+    },
+    {
+      "epoch": 35.39,
+      "learning_rate": 6.469836763662173e-05,
+      "loss": 0.3504,
+      "step": 4990
+    },
+    {
+      "epoch": 35.46,
+      "learning_rate": 6.462739531582683e-05,
+      "loss": 0.3304,
+      "step": 5000
+    },
+    {
+      "epoch": 35.53,
+      "learning_rate": 6.456352022711143e-05,
+      "loss": 0.5036,
+      "step": 5010
+    },
+    {
+      "epoch": 35.6,
+      "learning_rate": 6.449254790631654e-05,
+      "loss": 0.2183,
+      "step": 5020
+    },
+    {
+      "epoch": 35.67,
+      "learning_rate": 6.442157558552165e-05,
+      "loss": 0.2698,
+      "step": 5030
+    },
+    {
+      "epoch": 35.74,
+      "learning_rate": 6.435060326472677e-05,
+      "loss": 0.377,
+      "step": 5040
+    },
+    {
+      "epoch": 35.82,
+      "learning_rate": 6.427963094393187e-05,
+      "loss": 0.3307,
+      "step": 5050
+    },
+    {
+      "epoch": 35.89,
+      "learning_rate": 6.420865862313698e-05,
+      "loss": 0.3117,
+      "step": 5060
+    },
+    {
+      "epoch": 35.96,
+      "learning_rate": 6.413768630234209e-05,
+      "loss": 0.2519,
+      "step": 5070
+    },
+    {
+      "epoch": 36.0,
+      "eval_loss": 0.5957750678062439,
+      "eval_runtime": 8.7524,
+      "eval_samples_per_second": 7.198,
+      "eval_steps_per_second": 3.656,
+      "eval_wer": 0.3652467313369886,
+      "step": 5076
+    },
+    {
+      "epoch": 36.03,
+      "learning_rate": 6.40667139815472e-05,
+      "loss": 0.2093,
+      "step": 5080
+    },
+    {
+      "epoch": 36.1,
+      "learning_rate": 6.39957416607523e-05,
+      "loss": 0.3378,
+      "step": 5090
+    },
+    {
+      "epoch": 36.17,
+      "learning_rate": 6.392476933995743e-05,
+      "loss": 0.2608,
+      "step": 5100
+    },
+    {
+      "epoch": 36.24,
+      "learning_rate": 6.385379701916253e-05,
+      "loss": 0.4409,
+      "step": 5110
+    },
+    {
+      "epoch": 36.31,
+      "learning_rate": 6.378282469836764e-05,
+      "loss": 0.3885,
+      "step": 5120
+    },
+    {
+      "epoch": 36.38,
+      "learning_rate": 6.371185237757274e-05,
+      "loss": 0.1757,
+      "step": 5130
+    },
+    {
+      "epoch": 36.45,
+      "learning_rate": 6.364088005677787e-05,
+      "loss": 0.2864,
+      "step": 5140
+    },
+    {
+      "epoch": 36.52,
+      "learning_rate": 6.356990773598297e-05,
+      "loss": 0.3663,
+      "step": 5150
+    },
+    {
+      "epoch": 36.6,
+      "learning_rate": 6.349893541518808e-05,
+      "loss": 0.2438,
+      "step": 5160
+    },
+    {
+      "epoch": 36.67,
+      "learning_rate": 6.342796309439319e-05,
+      "loss": 0.3371,
+      "step": 5170
+    },
+    {
+      "epoch": 36.74,
+      "learning_rate": 6.33569907735983e-05,
+      "loss": 0.4315,
+      "step": 5180
+    },
+    {
+      "epoch": 36.81,
+      "learning_rate": 6.32860184528034e-05,
+      "loss": 0.3056,
+      "step": 5190
+    },
+    {
+      "epoch": 36.88,
+      "learning_rate": 6.321504613200852e-05,
+      "loss": 0.2268,
+      "step": 5200
+    },
+    {
+      "epoch": 36.95,
+      "learning_rate": 6.314407381121363e-05,
+      "loss": 0.3201,
+      "step": 5210
+    },
+    {
+      "epoch": 37.0,
+      "eval_loss": 0.5777791142463684,
+      "eval_runtime": 8.5751,
+      "eval_samples_per_second": 7.347,
+      "eval_steps_per_second": 3.732,
+      "eval_wer": 0.3652467313369886,
+      "step": 5217
+    },
+    {
+      "epoch": 37.02,
+      "learning_rate": 6.307310149041874e-05,
+      "loss": 0.3539,
+      "step": 5220
+    },
+    {
+      "epoch": 37.09,
+      "learning_rate": 6.300212916962384e-05,
+      "loss": 0.4534,
+      "step": 5230
+    },
+    {
+      "epoch": 37.16,
+      "learning_rate": 6.293115684882897e-05,
+      "loss": 0.3003,
+      "step": 5240
+    },
+    {
+      "epoch": 37.23,
+      "learning_rate": 6.286018452803407e-05,
+      "loss": 0.241,
+      "step": 5250
+    },
+    {
+      "epoch": 37.3,
+      "learning_rate": 6.278921220723918e-05,
+      "loss": 0.26,
+      "step": 5260
+    },
+    {
+      "epoch": 37.38,
+      "learning_rate": 6.271823988644429e-05,
+      "loss": 0.3437,
+      "step": 5270
+    },
+    {
+      "epoch": 37.45,
+      "learning_rate": 6.26472675656494e-05,
+      "loss": 0.254,
+      "step": 5280
+    },
+    {
+      "epoch": 37.52,
+      "learning_rate": 6.257629524485452e-05,
+      "loss": 0.4285,
+      "step": 5290
+    },
+    {
+      "epoch": 37.59,
+      "learning_rate": 6.250532292405962e-05,
+      "loss": 0.3217,
+      "step": 5300
+    },
+    {
+      "epoch": 37.66,
+      "learning_rate": 6.243435060326473e-05,
+      "loss": 0.2184,
+      "step": 5310
+    },
+    {
+      "epoch": 37.73,
+      "learning_rate": 6.236337828246984e-05,
+      "loss": 0.3274,
+      "step": 5320
+    },
+    {
+      "epoch": 37.8,
+      "learning_rate": 6.229240596167495e-05,
+      "loss": 0.3348,
+      "step": 5330
+    },
+    {
+      "epoch": 37.87,
+      "learning_rate": 6.222143364088005e-05,
+      "loss": 0.2953,
+      "step": 5340
+    },
+    {
+      "epoch": 37.94,
+      "learning_rate": 6.215046132008518e-05,
+      "loss": 0.3011,
+      "step": 5350
+    },
+    {
+      "epoch": 38.0,
+      "eval_loss": 0.6238256096839905,
+      "eval_runtime": 8.4953,
+      "eval_samples_per_second": 7.416,
+      "eval_steps_per_second": 3.767,
+      "eval_wer": 0.35892028679881904,
+      "step": 5358
+    },
+    {
+      "epoch": 38.01,
+      "learning_rate": 6.207948899929028e-05,
+      "loss": 0.338,
+      "step": 5360
+    },
+    {
+      "epoch": 38.09,
+      "learning_rate": 6.200851667849539e-05,
+      "loss": 0.3479,
+      "step": 5370
+    },
+    {
+      "epoch": 38.16,
+      "learning_rate": 6.193754435770049e-05,
+      "loss": 0.2699,
+      "step": 5380
+    },
+    {
+      "epoch": 38.23,
+      "learning_rate": 6.186657203690562e-05,
+      "loss": 0.5177,
+      "step": 5390
+    },
+    {
+      "epoch": 38.3,
+      "learning_rate": 6.179559971611072e-05,
+      "loss": 0.304,
+      "step": 5400
+    },
+    {
+      "epoch": 38.37,
+      "learning_rate": 6.172462739531583e-05,
+      "loss": 0.2442,
+      "step": 5410
+    },
+    {
+      "epoch": 38.44,
+      "learning_rate": 6.165365507452094e-05,
+      "loss": 0.2664,
+      "step": 5420
+    },
+    {
+      "epoch": 38.51,
+      "learning_rate": 6.158268275372605e-05,
+      "loss": 0.294,
+      "step": 5430
+    },
+    {
+      "epoch": 38.58,
+      "learning_rate": 6.151171043293115e-05,
+      "loss": 0.3588,
+      "step": 5440
+    },
+    {
+      "epoch": 38.65,
+      "learning_rate": 6.144073811213627e-05,
+      "loss": 0.2744,
+      "step": 5450
+    },
+    {
+      "epoch": 38.72,
+      "learning_rate": 6.136976579134138e-05,
+      "loss": 0.3519,
+      "step": 5460
+    },
+    {
+      "epoch": 38.79,
+      "learning_rate": 6.129879347054649e-05,
+      "loss": 0.5183,
+      "step": 5470
+    },
+    {
+      "epoch": 38.87,
+      "learning_rate": 6.122782114975159e-05,
+      "loss": 0.2305,
+      "step": 5480
+    },
+    {
+      "epoch": 38.94,
+      "learning_rate": 6.115684882895672e-05,
+      "loss": 0.2882,
+      "step": 5490
+    },
+    {
+      "epoch": 39.0,
+      "eval_loss": 0.6501014232635498,
+      "eval_runtime": 8.5383,
+      "eval_samples_per_second": 7.379,
+      "eval_steps_per_second": 3.748,
+      "eval_wer": 0.33614508646140867,
+      "step": 5499
+    },
+    {
+      "epoch": 39.01,
+      "learning_rate": 6.108587650816182e-05,
+      "loss": 0.2515,
+      "step": 5500
+    },
+    {
+      "epoch": 39.08,
+      "learning_rate": 6.101490418736693e-05,
+      "loss": 0.2852,
+      "step": 5510
+    },
+    {
+      "epoch": 39.15,
+      "learning_rate": 6.0943931866572035e-05,
+      "loss": 0.27,
+      "step": 5520
+    },
+    {
+      "epoch": 39.22,
+      "learning_rate": 6.0872959545777154e-05,
+      "loss": 0.2866,
+      "step": 5530
+    },
+    {
+      "epoch": 39.29,
+      "learning_rate": 6.080198722498226e-05,
+      "loss": 0.3485,
+      "step": 5540
+    },
+    {
+      "epoch": 39.36,
+      "learning_rate": 6.0731014904187366e-05,
+      "loss": 0.2235,
+      "step": 5550
+    },
+    {
+      "epoch": 39.43,
+      "learning_rate": 6.0660042583392486e-05,
+      "loss": 0.2411,
+      "step": 5560
+    },
+    {
+      "epoch": 39.5,
+      "learning_rate": 6.058907026259759e-05,
+      "loss": 0.3484,
+      "step": 5570
+    },
+    {
+      "epoch": 39.57,
+      "learning_rate": 6.05180979418027e-05,
+      "loss": 0.4276,
+      "step": 5580
+    },
+    {
+      "epoch": 39.65,
+      "learning_rate": 6.0447125621007804e-05,
+      "loss": 0.238,
+      "step": 5590
+    },
+    {
+      "epoch": 39.72,
+      "learning_rate": 6.037615330021292e-05,
+      "loss": 0.479,
+      "step": 5600
+    },
+    {
+      "epoch": 39.79,
+      "learning_rate": 6.030518097941803e-05,
+      "loss": 0.2841,
+      "step": 5610
+    },
+    {
+      "epoch": 39.86,
+      "learning_rate": 6.0234208658623135e-05,
+      "loss": 0.3854,
+      "step": 5620
+    },
+    {
+      "epoch": 39.93,
+      "learning_rate": 6.0163236337828254e-05,
+      "loss": 0.3052,
+      "step": 5630
+    },
+    {
+      "epoch": 40.0,
+      "learning_rate": 6.009226401703336e-05,
+      "loss": 0.2542,
+      "step": 5640
+    },
+    {
+      "epoch": 40.0,
+      "eval_loss": 0.6340907216072083,
+      "eval_runtime": 8.4944,
+      "eval_samples_per_second": 7.417,
+      "eval_steps_per_second": 3.767,
+      "eval_wer": 0.3487979755377478,
+      "step": 5640
+    },
+    {
+      "epoch": 40.07,
+      "learning_rate": 6.0021291696238466e-05,
+      "loss": 0.3474,
+      "step": 5650
+    },
+    {
+      "epoch": 40.14,
+      "learning_rate": 5.995031937544357e-05,
+      "loss": 0.2695,
+      "step": 5660
+    },
+    {
+      "epoch": 40.21,
+      "learning_rate": 5.987934705464869e-05,
+      "loss": 0.3707,
+      "step": 5670
+    },
+    {
+      "epoch": 40.28,
+      "learning_rate": 5.98083747338538e-05,
+      "loss": 0.28,
+      "step": 5680
+    },
+    {
+      "epoch": 40.35,
+      "learning_rate": 5.973740241305891e-05,
+      "loss": 0.2745,
+      "step": 5690
+    },
+    {
+      "epoch": 40.43,
+      "learning_rate": 5.966643009226402e-05,
+      "loss": 0.3151,
+      "step": 5700
+    },
+    {
+      "epoch": 40.5,
+      "learning_rate": 5.959545777146913e-05,
+      "loss": 0.4127,
+      "step": 5710
+    },
+    {
+      "epoch": 40.57,
+      "learning_rate": 5.952448545067424e-05,
+      "loss": 0.3172,
+      "step": 5720
+    },
+    {
+      "epoch": 40.64,
+      "learning_rate": 5.945351312987935e-05,
+      "loss": 0.249,
+      "step": 5730
+    },
+    {
+      "epoch": 40.71,
+      "learning_rate": 5.938254080908446e-05,
+      "loss": 0.2968,
+      "step": 5740
+    },
+    {
+      "epoch": 40.78,
+      "learning_rate": 5.931156848828957e-05,
+      "loss": 0.3314,
+      "step": 5750
+    },
+    {
+      "epoch": 40.85,
+      "learning_rate": 5.924059616749468e-05,
+      "loss": 0.2152,
+      "step": 5760
+    },
+    {
+      "epoch": 40.92,
+      "learning_rate": 5.91696238466998e-05,
+      "loss": 0.3094,
+      "step": 5770
+    },
+    {
+      "epoch": 40.99,
+      "learning_rate": 5.9098651525904905e-05,
+      "loss": 0.2717,
+      "step": 5780
+    },
+    {
+      "epoch": 41.0,
+      "eval_loss": 0.5890086889266968,
+      "eval_runtime": 8.5723,
+      "eval_samples_per_second": 7.349,
+      "eval_steps_per_second": 3.733,
+      "eval_wer": 0.3530156052298608,
+      "step": 5781
+    },
+    {
+      "epoch": 41.06,
+      "learning_rate": 5.902767920511001e-05,
+      "loss": 0.3601,
+      "step": 5790
+    },
+    {
+      "epoch": 41.13,
+      "learning_rate": 5.8956706884315117e-05,
+      "loss": 0.2123,
+      "step": 5800
+    },
+    {
+      "epoch": 41.21,
+      "learning_rate": 5.8885734563520236e-05,
+      "loss": 0.3188,
+      "step": 5810
+    },
+    {
+      "epoch": 41.28,
+      "learning_rate": 5.881476224272534e-05,
+      "loss": 1.1286,
+      "step": 5820
+    },
+    {
+      "epoch": 41.35,
+      "learning_rate": 5.874378992193045e-05,
+      "loss": 0.3037,
+      "step": 5830
+    },
+    {
+      "epoch": 41.42,
+      "learning_rate": 5.8672817601135554e-05,
+      "loss": 0.2433,
+      "step": 5840
+    },
+    {
+      "epoch": 41.49,
+      "learning_rate": 5.8601845280340673e-05,
+      "loss": 0.3291,
+      "step": 5850
+    },
+    {
+      "epoch": 41.56,
+      "learning_rate": 5.853087295954578e-05,
+      "loss": 0.2721,
+      "step": 5860
+    },
+    {
+      "epoch": 41.63,
+      "learning_rate": 5.8459900638750885e-05,
+      "loss": 0.228,
+      "step": 5870
+    },
+    {
+      "epoch": 41.7,
+      "learning_rate": 5.8388928317956005e-05,
+      "loss": 0.3515,
+      "step": 5880
+    },
+    {
+      "epoch": 41.77,
+      "learning_rate": 5.831795599716111e-05,
+      "loss": 0.3712,
+      "step": 5890
+    },
+    {
+      "epoch": 41.84,
+      "learning_rate": 5.824698367636622e-05,
+      "loss": 0.2264,
+      "step": 5900
+    },
+    {
+      "epoch": 41.91,
+      "learning_rate": 5.817601135557132e-05,
+      "loss": 0.3169,
+      "step": 5910
+    },
+    {
+      "epoch": 41.99,
+      "learning_rate": 5.810503903477644e-05,
+      "loss": 0.3197,
+      "step": 5920
+    },
+    {
+      "epoch": 42.0,
+      "eval_loss": 0.5877115726470947,
+      "eval_runtime": 8.5428,
+      "eval_samples_per_second": 7.375,
+      "eval_steps_per_second": 3.746,
+      "eval_wer": 0.34711092366090257,
+      "step": 5922
+    },
+    {
+      "epoch": 42.06,
+      "learning_rate": 5.803406671398155e-05,
+      "loss": 0.3204,
+      "step": 5930
+    },
+    {
+      "epoch": 42.13,
+      "learning_rate": 5.7963094393186654e-05,
+      "loss": 0.3614,
+      "step": 5940
+    },
+    {
+      "epoch": 42.2,
+      "learning_rate": 5.7892122072391774e-05,
+      "loss": 0.263,
+      "step": 5950
+    },
+    {
+      "epoch": 42.27,
+      "learning_rate": 5.782114975159688e-05,
+      "loss": 0.2805,
+      "step": 5960
+    },
+    {
+      "epoch": 42.34,
+      "learning_rate": 5.7750177430801985e-05,
+      "loss": 0.2712,
+      "step": 5970
+    },
+    {
+      "epoch": 42.41,
+      "learning_rate": 5.76792051100071e-05,
+      "loss": 0.2292,
+      "step": 5980
+    },
+    {
+      "epoch": 42.48,
+      "learning_rate": 5.760823278921221e-05,
+      "loss": 0.3057,
+      "step": 5990
+    },
+    {
+      "epoch": 42.55,
+      "learning_rate": 5.753726046841732e-05,
+      "loss": 0.3446,
+      "step": 6000
+    },
+    {
+      "epoch": 42.62,
+      "learning_rate": 5.746628814762243e-05,
+      "loss": 0.3518,
+      "step": 6010
+    },
+    {
+      "epoch": 42.7,
+      "learning_rate": 5.739531582682754e-05,
+      "loss": 0.3175,
+      "step": 6020
+    },
+    {
+      "epoch": 42.77,
+      "learning_rate": 5.7324343506032655e-05,
+      "loss": 0.4367,
+      "step": 6030
+    },
+    {
+      "epoch": 42.84,
+      "learning_rate": 5.725337118523776e-05,
+      "loss": 0.2073,
+      "step": 6040
+    },
+    {
+      "epoch": 42.91,
+      "learning_rate": 5.718239886444287e-05,
+      "loss": 0.2769,
+      "step": 6050
+    },
+    {
+      "epoch": 42.98,
+      "learning_rate": 5.7111426543647986e-05,
+      "loss": 0.2816,
+      "step": 6060
+    },
+    {
+      "epoch": 43.0,
+      "eval_loss": 0.6614282727241516,
+      "eval_runtime": 8.5505,
+      "eval_samples_per_second": 7.368,
+      "eval_steps_per_second": 3.742,
+      "eval_wer": 0.3420497680303669,
+      "step": 6063
+    },
+    {
+      "epoch": 43.05,
+      "learning_rate": 5.704045422285309e-05,
+      "loss": 0.3708,
+      "step": 6070
+    },
+    {
+      "epoch": 43.12,
+      "learning_rate": 5.69694819020582e-05,
+      "loss": 0.215,
+      "step": 6080
+    },
+    {
+      "epoch": 43.19,
+      "learning_rate": 5.689850958126332e-05,
+      "loss": 0.26,
+      "step": 6090
+    },
+    {
+      "epoch": 43.26,
+      "learning_rate": 5.6827537260468424e-05,
+      "loss": 0.2678,
+      "step": 6100
+    },
+    {
+      "epoch": 43.33,
+      "learning_rate": 5.675656493967353e-05,
+      "loss": 0.331,
+      "step": 6110
+    },
+    {
+      "epoch": 43.4,
+      "learning_rate": 5.6685592618878636e-05,
+      "loss": 0.221,
+      "step": 6120
+    },
+    {
+      "epoch": 43.48,
+      "learning_rate": 5.6614620298083755e-05,
+      "loss": 0.6329,
+      "step": 6130
+    },
+    {
+      "epoch": 43.55,
+      "learning_rate": 5.654364797728886e-05,
+      "loss": 0.3507,
+      "step": 6140
+    },
+    {
+      "epoch": 43.62,
+      "learning_rate": 5.647267565649397e-05,
+      "loss": 0.3738,
+      "step": 6150
+    },
+    {
+      "epoch": 43.69,
+      "learning_rate": 5.6401703335699087e-05,
+      "loss": 0.2949,
+      "step": 6160
+    },
+    {
+      "epoch": 43.76,
+      "learning_rate": 5.633073101490419e-05,
+      "loss": 0.3074,
+      "step": 6170
+    },
+    {
+      "epoch": 43.83,
+      "learning_rate": 5.62597586941093e-05,
+      "loss": 0.2972,
+      "step": 6180
+    },
+    {
+      "epoch": 43.9,
+      "learning_rate": 5.6188786373314404e-05,
+      "loss": 0.2435,
+      "step": 6190
+    },
+    {
+      "epoch": 43.97,
+      "learning_rate": 5.6117814052519524e-05,
+      "loss": 0.3301,
+      "step": 6200
+    },
+    {
+      "epoch": 44.0,
+      "eval_loss": 0.63343745470047,
+      "eval_runtime": 8.5803,
+      "eval_samples_per_second": 7.342,
+      "eval_steps_per_second": 3.729,
+      "eval_wer": 0.3475326866301139,
+      "step": 6204
+    },
+    {
+      "epoch": 44.04,
+      "learning_rate": 5.604684173172463e-05,
+      "loss": 0.2632,
+      "step": 6210
+    },
+    {
+      "epoch": 44.11,
+      "learning_rate": 5.5975869410929736e-05,
+      "loss": 0.2534,
+      "step": 6220
+    },
+    {
+      "epoch": 44.18,
+      "learning_rate": 5.590489709013484e-05,
+      "loss": 0.2968,
+      "step": 6230
+    },
+    {
+      "epoch": 44.26,
+      "learning_rate": 5.583392476933996e-05,
+      "loss": 0.2767,
+      "step": 6240
+    },
+    {
+      "epoch": 44.33,
+      "learning_rate": 5.576295244854507e-05,
+      "loss": 0.3589,
+      "step": 6250
+    },
+    {
+      "epoch": 44.4,
+      "learning_rate": 5.569198012775017e-05,
+      "loss": 0.2715,
+      "step": 6260
+    },
+    {
+      "epoch": 44.47,
+      "learning_rate": 5.562100780695529e-05,
+      "loss": 0.2852,
+      "step": 6270
+    },
+    {
+      "epoch": 44.54,
+      "learning_rate": 5.55500354861604e-05,
+      "loss": 0.3722,
+      "step": 6280
+    },
+    {
+      "epoch": 44.61,
+      "learning_rate": 5.5479063165365505e-05,
+      "loss": 0.2152,
+      "step": 6290
+    },
+    {
+      "epoch": 44.68,
+      "learning_rate": 5.540809084457062e-05,
+      "loss": 0.2592,
+      "step": 6300
+    },
+    {
+      "epoch": 44.75,
+      "learning_rate": 5.533711852377573e-05,
+      "loss": 0.2651,
+      "step": 6310
+    },
+    {
+      "epoch": 44.82,
+      "learning_rate": 5.526614620298084e-05,
+      "loss": 0.2841,
+      "step": 6320
+    },
+    {
+      "epoch": 44.89,
+      "learning_rate": 5.519517388218595e-05,
+      "loss": 0.1942,
+      "step": 6330
+    },
+    {
+      "epoch": 44.96,
+      "learning_rate": 5.512420156139106e-05,
+      "loss": 0.2466,
+      "step": 6340
+    },
+    {
+      "epoch": 45.0,
+      "eval_loss": 0.6663392782211304,
+      "eval_runtime": 8.5921,
+      "eval_samples_per_second": 7.332,
+      "eval_steps_per_second": 3.724,
+      "eval_wer": 0.34289329396878954,
+      "step": 6345
+    },
+    {
+      "epoch": 45.04,
+      "learning_rate": 5.5053229240596174e-05,
+      "loss": 0.3651,
+      "step": 6350
+    },
+    {
+      "epoch": 45.11,
+      "learning_rate": 5.498225691980128e-05,
+      "loss": 0.2512,
+      "step": 6360
+    },
+    {
+      "epoch": 45.18,
+      "learning_rate": 5.4911284599006386e-05,
+      "loss": 0.2905,
+      "step": 6370
+    },
+    {
+      "epoch": 45.25,
+      "learning_rate": 5.4840312278211506e-05,
+      "loss": 0.2855,
+      "step": 6380
+    },
+    {
+      "epoch": 45.32,
+      "learning_rate": 5.476933995741661e-05,
+      "loss": 0.2681,
+      "step": 6390
+    },
+    {
+      "epoch": 45.39,
+      "learning_rate": 5.469836763662172e-05,
+      "loss": 0.2958,
+      "step": 6400
+    },
+    {
+      "epoch": 45.46,
+      "learning_rate": 5.462739531582684e-05,
+      "loss": 0.3322,
+      "step": 6410
+    },
+    {
+      "epoch": 45.53,
+      "learning_rate": 5.455642299503194e-05,
+      "loss": 0.4016,
+      "step": 6420
+    },
+    {
+      "epoch": 45.6,
+      "learning_rate": 5.448545067423705e-05,
+      "loss": 0.204,
+      "step": 6430
+    },
+    {
+      "epoch": 45.67,
+      "learning_rate": 5.4414478353442155e-05,
+      "loss": 0.2698,
+      "step": 6440
+    },
+    {
+      "epoch": 45.74,
+      "learning_rate": 5.4343506032647274e-05,
+      "loss": 0.3043,
+      "step": 6450
+    },
+    {
+      "epoch": 45.82,
+      "learning_rate": 5.427253371185238e-05,
+      "loss": 0.3044,
+      "step": 6460
+    },
+    {
+      "epoch": 45.89,
+      "learning_rate": 5.4201561391057486e-05,
+      "loss": 0.1805,
+      "step": 6470
+    },
+    {
+      "epoch": 45.96,
+      "learning_rate": 5.4130589070262606e-05,
+      "loss": 0.2908,
+      "step": 6480
+    },
+    {
+      "epoch": 46.0,
+      "eval_loss": 0.594066321849823,
+      "eval_runtime": 8.5199,
+      "eval_samples_per_second": 7.394,
+      "eval_steps_per_second": 3.756,
+      "eval_wer": 0.3475326866301139,
+      "step": 6486
+    },
+    {
+      "epoch": 46.03,
+      "learning_rate": 5.405961674946771e-05,
+      "loss": 0.2973,
+      "step": 6490
+    },
+    {
+      "epoch": 46.1,
+      "learning_rate": 5.398864442867282e-05,
+      "loss": 0.3801,
+      "step": 6500
+    },
+    {
+      "epoch": 46.17,
+      "learning_rate": 5.3917672107877924e-05,
+      "loss": 0.3148,
+      "step": 6510
+    },
+    {
+      "epoch": 46.24,
+      "learning_rate": 5.384669978708304e-05,
+      "loss": 0.293,
+      "step": 6520
+    },
+    {
+      "epoch": 46.31,
+      "learning_rate": 5.377572746628815e-05,
+      "loss": 0.3326,
+      "step": 6530
+    },
+    {
+      "epoch": 46.38,
+      "learning_rate": 5.3704755145493255e-05,
+      "loss": 0.1631,
+      "step": 6540
+    },
+    {
+      "epoch": 46.45,
+      "learning_rate": 5.3633782824698374e-05,
+      "loss": 0.2645,
+      "step": 6550
+    },
+    {
+      "epoch": 46.52,
+      "learning_rate": 5.356281050390348e-05,
+      "loss": 0.3975,
+      "step": 6560
+    },
+    {
+      "epoch": 46.6,
+      "learning_rate": 5.3491838183108586e-05,
+      "loss": 0.2558,
+      "step": 6570
+    },
+    {
+      "epoch": 46.67,
+      "learning_rate": 5.34208658623137e-05,
+      "loss": 0.2651,
+      "step": 6580
+    },
+    {
+      "epoch": 46.74,
+      "learning_rate": 5.334989354151881e-05,
+      "loss": 0.388,
+      "step": 6590
+    },
+    {
+      "epoch": 46.81,
+      "learning_rate": 5.327892122072392e-05,
+      "loss": 0.3663,
+      "step": 6600
+    },
+    {
+      "epoch": 46.88,
+      "learning_rate": 5.320794889992903e-05,
+      "loss": 0.2184,
+      "step": 6610
+    },
+    {
+      "epoch": 46.95,
+      "learning_rate": 5.313697657913414e-05,
+      "loss": 0.2785,
+      "step": 6620
+    },
+    {
+      "epoch": 47.0,
+      "eval_loss": 0.6336867213249207,
+      "eval_runtime": 8.56,
+      "eval_samples_per_second": 7.36,
+      "eval_steps_per_second": 3.738,
+      "eval_wer": 0.3568114719527625,
+      "step": 6627
+    },
+    {
+      "epoch": 47.02,
+      "learning_rate": 5.306600425833925e-05,
+      "loss": 0.5427,
+      "step": 6630
+    },
+    {
+      "epoch": 47.09,
+      "learning_rate": 5.299503193754436e-05,
+      "loss": 0.3401,
+      "step": 6640
+    },
+    {
+      "epoch": 47.16,
+      "learning_rate": 5.292405961674947e-05,
+      "loss": 0.3042,
+      "step": 6650
+    },
+    {
+      "epoch": 47.23,
+      "learning_rate": 5.285308729595459e-05,
+      "loss": 0.2539,
+      "step": 6660
+    },
+    {
+      "epoch": 47.3,
+      "learning_rate": 5.278211497515969e-05,
+      "loss": 0.3101,
+      "step": 6670
+    },
+    {
+      "epoch": 47.38,
+      "learning_rate": 5.27111426543648e-05,
+      "loss": 0.1548,
+      "step": 6680
+    },
+    {
+      "epoch": 47.45,
+      "learning_rate": 5.2640170333569905e-05,
+      "loss": 0.2557,
+      "step": 6690
+    },
+    {
+      "epoch": 47.52,
+      "learning_rate": 5.2569198012775025e-05,
+      "loss": 0.526,
+      "step": 6700
+    },
+    {
+      "epoch": 47.59,
+      "learning_rate": 5.249822569198013e-05,
+      "loss": 0.2895,
+      "step": 6710
+    },
+    {
+      "epoch": 47.66,
+      "learning_rate": 5.242725337118524e-05,
+      "loss": 0.3268,
+      "step": 6720
+    },
+    {
+      "epoch": 47.73,
+      "learning_rate": 5.2356281050390356e-05,
+      "loss": 0.306,
+      "step": 6730
+    },
+    {
+      "epoch": 47.8,
+      "learning_rate": 5.228530872959546e-05,
+      "loss": 0.4037,
+      "step": 6740
+    },
+    {
+      "epoch": 47.87,
+      "learning_rate": 5.221433640880057e-05,
+      "loss": 0.2334,
+      "step": 6750
+    },
+    {
+      "epoch": 47.94,
+      "learning_rate": 5.2143364088005674e-05,
+      "loss": 0.2361,
+      "step": 6760
+    },
+    {
+      "epoch": 48.0,
+      "eval_loss": 0.5845135450363159,
+      "eval_runtime": 8.6098,
+      "eval_samples_per_second": 7.317,
+      "eval_steps_per_second": 3.717,
+      "eval_wer": 0.3399409531843104,
+      "step": 6768
+    },
+    {
+      "epoch": 48.01,
+      "learning_rate": 5.2072391767210793e-05,
+      "loss": 0.2627,
+      "step": 6770
+    },
+    {
+      "epoch": 48.09,
+      "learning_rate": 5.20014194464159e-05,
+      "loss": 0.2765,
+      "step": 6780
+    },
+    {
+      "epoch": 48.16,
+      "learning_rate": 5.1930447125621005e-05,
+      "loss": 0.2074,
+      "step": 6790
+    },
+    {
+      "epoch": 48.23,
+      "learning_rate": 5.1859474804826125e-05,
+      "loss": 0.3596,
+      "step": 6800
+    },
+    {
+      "epoch": 48.3,
+      "learning_rate": 5.178850248403123e-05,
+      "loss": 0.2663,
+      "step": 6810
+    },
+    {
+      "epoch": 48.37,
+      "learning_rate": 5.171753016323634e-05,
+      "loss": 0.2299,
+      "step": 6820
+    },
+    {
+      "epoch": 48.44,
+      "learning_rate": 5.164655784244144e-05,
+      "loss": 0.2753,
+      "step": 6830
+    },
+    {
+      "epoch": 48.51,
+      "learning_rate": 5.157558552164656e-05,
+      "loss": 0.2834,
+      "step": 6840
+    },
+    {
+      "epoch": 48.58,
+      "learning_rate": 5.150461320085167e-05,
+      "loss": 0.277,
+      "step": 6850
+    },
+    {
+      "epoch": 48.65,
+      "learning_rate": 5.1433640880056774e-05,
+      "loss": 0.2361,
+      "step": 6860
+    },
+    {
+      "epoch": 48.72,
+      "learning_rate": 5.1362668559261894e-05,
+      "loss": 0.3205,
+      "step": 6870
+    },
+    {
+      "epoch": 48.79,
+      "learning_rate": 5.1291696238467e-05,
+      "loss": 0.3748,
+      "step": 6880
+    },
+    {
+      "epoch": 48.87,
+      "learning_rate": 5.1220723917672106e-05,
+      "loss": 0.246,
+      "step": 6890
+    },
+    {
+      "epoch": 48.94,
+      "learning_rate": 5.114975159687722e-05,
+      "loss": 0.4729,
+      "step": 6900
+    },
+    {
+      "epoch": 49.0,
+      "eval_loss": 0.6465837359428406,
+      "eval_runtime": 8.6076,
+      "eval_samples_per_second": 7.319,
+      "eval_steps_per_second": 3.718,
+      "eval_wer": 0.3424715309995782,
+      "step": 6909
+    },
+    {
+      "epoch": 49.01,
+      "learning_rate": 5.107877927608233e-05,
+      "loss": 1.0998,
+      "step": 6910
+    },
+    {
+      "epoch": 49.08,
+      "learning_rate": 5.1007806955287444e-05,
+      "loss": 0.3247,
+      "step": 6920
+    },
+    {
+      "epoch": 49.15,
+      "learning_rate": 5.093683463449255e-05,
+      "loss": 0.2948,
+      "step": 6930
+    },
+    {
+      "epoch": 49.22,
+      "learning_rate": 5.086586231369766e-05,
+      "loss": 0.2801,
+      "step": 6940
+    },
+    {
+      "epoch": 49.29,
+      "learning_rate": 5.0794889992902775e-05,
+      "loss": 0.3145,
+      "step": 6950
+    },
+    {
+      "epoch": 49.36,
+      "learning_rate": 5.072391767210788e-05,
+      "loss": 0.2245,
+      "step": 6960
+    },
+    {
+      "epoch": 49.43,
+      "learning_rate": 5.065294535131299e-05,
+      "loss": 0.2317,
+      "step": 6970
+    },
+    {
+      "epoch": 49.5,
+      "learning_rate": 5.0581973030518107e-05,
+      "loss": 0.2945,
+      "step": 6980
+    },
+    {
+      "epoch": 49.57,
+      "learning_rate": 5.051100070972321e-05,
+      "loss": 0.3259,
+      "step": 6990
+    },
+    {
+      "epoch": 49.65,
+      "learning_rate": 5.044002838892832e-05,
+      "loss": 0.1846,
+      "step": 7000
+    },
+    {
+      "epoch": 49.72,
+      "learning_rate": 5.036905606813344e-05,
+      "loss": 0.2498,
+      "step": 7010
+    },
+    {
+      "epoch": 49.79,
+      "learning_rate": 5.0298083747338544e-05,
+      "loss": 0.292,
+      "step": 7020
+    },
+    {
+      "epoch": 49.86,
+      "learning_rate": 5.022711142654365e-05,
+      "loss": 0.2975,
+      "step": 7030
+    },
+    {
+      "epoch": 49.93,
+      "learning_rate": 5.0156139105748756e-05,
+      "loss": 0.2524,
+      "step": 7040
+    },
+    {
+      "epoch": 50.0,
+      "learning_rate": 5.0085166784953875e-05,
+      "loss": 0.5103,
+      "step": 7050
+    },
+    {
+      "epoch": 50.0,
+      "eval_loss": 0.71123868227005,
+      "eval_runtime": 8.4028,
+      "eval_samples_per_second": 7.497,
+      "eval_steps_per_second": 3.808,
+      "eval_wer": 0.34162800506115565,
+      "step": 7050
+    },
+    {
+      "epoch": 50.07,
+      "learning_rate": 5.001419446415898e-05,
+      "loss": 0.374,
+      "step": 7060
+    },
+    {
+      "epoch": 50.14,
+      "learning_rate": 4.994322214336409e-05,
+      "loss": 0.247,
+      "step": 7070
+    },
+    {
+      "epoch": 50.21,
+      "learning_rate": 4.987934705464869e-05,
+      "loss": 0.2219,
+      "step": 7080
+    },
+    {
+      "epoch": 50.28,
+      "learning_rate": 4.9808374733853805e-05,
+      "loss": 0.303,
+      "step": 7090
+    },
+    {
+      "epoch": 50.35,
+      "learning_rate": 4.973740241305891e-05,
+      "loss": 0.2254,
+      "step": 7100
+    },
+    {
+      "epoch": 50.43,
+      "learning_rate": 4.9666430092264024e-05,
+      "loss": 0.2708,
+      "step": 7110
+    },
+    {
+      "epoch": 50.5,
+      "learning_rate": 4.959545777146913e-05,
+      "loss": 0.3334,
+      "step": 7120
+    },
+    {
+      "epoch": 50.57,
+      "learning_rate": 4.952448545067424e-05,
+      "loss": 0.2839,
+      "step": 7130
+    },
+    {
+      "epoch": 50.64,
+      "learning_rate": 4.945351312987935e-05,
+      "loss": 0.2395,
+      "step": 7140
+    },
+    {
+      "epoch": 50.71,
+      "learning_rate": 4.938254080908446e-05,
+      "loss": 0.277,
+      "step": 7150
+    },
+    {
+      "epoch": 50.78,
+      "learning_rate": 4.931156848828957e-05,
+      "loss": 0.2924,
+      "step": 7160
+    },
+    {
+      "epoch": 50.85,
+      "learning_rate": 4.924059616749468e-05,
+      "loss": 0.2232,
+      "step": 7170
+    },
+    {
+      "epoch": 50.92,
+      "learning_rate": 4.916962384669979e-05,
+      "loss": 0.3027,
+      "step": 7180
+    },
+    {
+      "epoch": 50.99,
+      "learning_rate": 4.90986515259049e-05,
+      "loss": 0.2676,
+      "step": 7190
+    },
+    {
+      "epoch": 51.0,
+      "eval_loss": 0.6259974241256714,
+      "eval_runtime": 8.4619,
+      "eval_samples_per_second": 7.445,
+      "eval_steps_per_second": 3.782,
+      "eval_wer": 0.33066216786166175,
+      "step": 7191
+    },
+    {
+      "epoch": 51.06,
+      "learning_rate": 4.902767920511001e-05,
+      "loss": 0.3311,
+      "step": 7200
+    },
+    {
+      "epoch": 51.13,
+      "learning_rate": 4.895670688431512e-05,
+      "loss": 0.2613,
+      "step": 7210
+    },
+    {
+      "epoch": 51.21,
+      "learning_rate": 4.888573456352023e-05,
+      "loss": 0.2897,
+      "step": 7220
+    },
+    {
+      "epoch": 51.28,
+      "learning_rate": 4.8814762242725336e-05,
+      "loss": 0.2148,
+      "step": 7230
+    },
+    {
+      "epoch": 51.35,
+      "learning_rate": 4.874378992193045e-05,
+      "loss": 0.2615,
+      "step": 7240
+    },
+    {
+      "epoch": 51.42,
+      "learning_rate": 4.867281760113556e-05,
+      "loss": 0.3662,
+      "step": 7250
+    },
+    {
+      "epoch": 51.49,
+      "learning_rate": 4.860184528034067e-05,
+      "loss": 0.3403,
+      "step": 7260
+    },
+    {
+      "epoch": 51.56,
+      "learning_rate": 4.853087295954578e-05,
+      "loss": 0.2696,
+      "step": 7270
+    },
+    {
+      "epoch": 51.63,
+      "learning_rate": 4.8459900638750886e-05,
+      "loss": 0.1889,
+      "step": 7280
+    },
+    {
+      "epoch": 51.7,
+      "learning_rate": 4.8388928317956e-05,
+      "loss": 0.1904,
+      "step": 7290
+    },
+    {
+      "epoch": 51.77,
+      "learning_rate": 4.8317955997161105e-05,
+      "loss": 0.3629,
+      "step": 7300
+    },
+    {
+      "epoch": 51.84,
+      "learning_rate": 4.824698367636622e-05,
+      "loss": 0.2285,
+      "step": 7310
+    },
+    {
+      "epoch": 51.91,
+      "learning_rate": 4.817601135557133e-05,
+      "loss": 0.2943,
+      "step": 7320
+    },
+    {
+      "epoch": 51.99,
+      "learning_rate": 4.8105039034776436e-05,
+      "loss": 0.3533,
+      "step": 7330
+    },
+    {
+      "epoch": 52.0,
+      "eval_loss": 0.7327254414558411,
+      "eval_runtime": 8.4414,
+      "eval_samples_per_second": 7.463,
+      "eval_steps_per_second": 3.791,
+      "eval_wer": 0.34542387178405737,
+      "step": 7332
+    },
+    {
+      "epoch": 52.06,
+      "learning_rate": 4.803406671398155e-05,
+      "loss": 0.2908,
+      "step": 7340
+    },
+    {
+      "epoch": 52.13,
+      "learning_rate": 4.7963094393186655e-05,
+      "loss": 0.2752,
+      "step": 7350
+    },
+    {
+      "epoch": 52.2,
+      "learning_rate": 4.789212207239177e-05,
+      "loss": 0.2432,
+      "step": 7360
+    },
+    {
+      "epoch": 52.27,
+      "learning_rate": 4.782114975159688e-05,
+      "loss": 0.3063,
+      "step": 7370
+    },
+    {
+      "epoch": 52.34,
+      "learning_rate": 4.775017743080199e-05,
+      "loss": 0.2829,
+      "step": 7380
+    },
+    {
+      "epoch": 52.41,
+      "learning_rate": 4.76792051100071e-05,
+      "loss": 0.247,
+      "step": 7390
+    },
+    {
+      "epoch": 52.48,
+      "learning_rate": 4.760823278921221e-05,
+      "loss": 0.2481,
+      "step": 7400
+    },
+    {
+      "epoch": 52.55,
+      "learning_rate": 4.7537260468417325e-05,
+      "loss": 0.2989,
+      "step": 7410
+    },
+    {
+      "epoch": 52.62,
+      "learning_rate": 4.746628814762243e-05,
+      "loss": 0.2498,
+      "step": 7420
+    },
+    {
+      "epoch": 52.7,
+      "learning_rate": 4.739531582682754e-05,
+      "loss": 0.2779,
+      "step": 7430
+    },
+    {
+      "epoch": 52.77,
+      "learning_rate": 4.732434350603265e-05,
+      "loss": 0.3295,
+      "step": 7440
+    },
+    {
+      "epoch": 52.84,
+      "learning_rate": 4.725337118523776e-05,
+      "loss": 0.3394,
+      "step": 7450
+    },
+    {
+      "epoch": 52.91,
+      "learning_rate": 4.718239886444287e-05,
+      "loss": 0.254,
+      "step": 7460
+    },
+    {
+      "epoch": 52.98,
+      "learning_rate": 4.711142654364798e-05,
+      "loss": 0.3308,
+      "step": 7470
+    },
+    {
+      "epoch": 53.0,
+      "eval_loss": 0.7149801254272461,
+      "eval_runtime": 8.5139,
+      "eval_samples_per_second": 7.4,
+      "eval_steps_per_second": 3.759,
+      "eval_wer": 0.3277098270771826,
+      "step": 7473
+    },
+    {
+      "epoch": 53.05,
+      "learning_rate": 4.704045422285309e-05,
+      "loss": 0.2592,
+      "step": 7480
+    },
+    {
+      "epoch": 53.12,
+      "learning_rate": 4.69694819020582e-05,
+      "loss": 0.2273,
+      "step": 7490
+    },
+    {
+      "epoch": 53.19,
+      "learning_rate": 4.689850958126331e-05,
+      "loss": 0.2575,
+      "step": 7500
+    },
+    {
+      "epoch": 53.26,
+      "learning_rate": 4.682753726046842e-05,
+      "loss": 0.2796,
+      "step": 7510
+    },
+    {
+      "epoch": 53.33,
+      "learning_rate": 4.675656493967353e-05,
+      "loss": 0.3359,
+      "step": 7520
+    },
+    {
+      "epoch": 53.4,
+      "learning_rate": 4.6685592618878637e-05,
+      "loss": 0.2229,
+      "step": 7530
+    },
+    {
+      "epoch": 53.48,
+      "learning_rate": 4.661462029808375e-05,
+      "loss": 0.2945,
+      "step": 7540
+    },
+    {
+      "epoch": 53.55,
+      "learning_rate": 4.654364797728886e-05,
+      "loss": 0.3534,
+      "step": 7550
+    },
+    {
+      "epoch": 53.62,
+      "learning_rate": 4.647267565649397e-05,
+      "loss": 0.3587,
+      "step": 7560
+    },
+    {
+      "epoch": 53.69,
+      "learning_rate": 4.640170333569908e-05,
+      "loss": 0.2756,
+      "step": 7570
+    },
+    {
+      "epoch": 53.76,
+      "learning_rate": 4.633073101490419e-05,
+      "loss": 0.3332,
+      "step": 7580
+    },
+    {
+      "epoch": 53.83,
+      "learning_rate": 4.62597586941093e-05,
+      "loss": 0.3785,
+      "step": 7590
+    },
+    {
+      "epoch": 53.9,
+      "learning_rate": 4.6188786373314405e-05,
+      "loss": 0.1912,
+      "step": 7600
+    },
+    {
+      "epoch": 53.97,
+      "learning_rate": 4.611781405251952e-05,
+      "loss": 0.2617,
+      "step": 7610
+    },
+    {
+      "epoch": 54.0,
+      "eval_loss": 0.641165554523468,
+      "eval_runtime": 8.5457,
+      "eval_samples_per_second": 7.372,
+      "eval_steps_per_second": 3.745,
+      "eval_wer": 0.3390974272458878,
+      "step": 7614
+    },
+    {
+      "epoch": 54.04,
+      "learning_rate": 4.6046841731724624e-05,
+      "loss": 0.2767,
+      "step": 7620
+    },
+    {
+      "epoch": 54.11,
+      "learning_rate": 4.597586941092974e-05,
+      "loss": 0.2497,
+      "step": 7630
+    },
+    {
+      "epoch": 54.18,
+      "learning_rate": 4.590489709013485e-05,
+      "loss": 0.2864,
+      "step": 7640
+    },
+    {
+      "epoch": 54.26,
+      "learning_rate": 4.5833924769339955e-05,
+      "loss": 0.2516,
+      "step": 7650
+    },
+    {
+      "epoch": 54.33,
+      "learning_rate": 4.576295244854507e-05,
+      "loss": 0.3344,
+      "step": 7660
+    },
+    {
+      "epoch": 54.4,
+      "learning_rate": 4.569198012775018e-05,
+      "loss": 0.2157,
+      "step": 7670
+    },
+    {
+      "epoch": 54.47,
+      "learning_rate": 4.5621007806955294e-05,
+      "loss": 0.2786,
+      "step": 7680
+    },
+    {
+      "epoch": 54.54,
+      "learning_rate": 4.55500354861604e-05,
+      "loss": 0.2468,
+      "step": 7690
+    },
+    {
+      "epoch": 54.61,
+      "learning_rate": 4.547906316536551e-05,
+      "loss": 0.2466,
+      "step": 7700
+    },
+    {
+      "epoch": 54.68,
+      "learning_rate": 4.5408090844570625e-05,
+      "loss": 0.2902,
+      "step": 7710
+    },
+    {
+      "epoch": 54.75,
+      "learning_rate": 4.533711852377573e-05,
+      "loss": 0.3666,
+      "step": 7720
+    },
+    {
+      "epoch": 54.82,
+      "learning_rate": 4.5266146202980844e-05,
+      "loss": 0.3019,
+      "step": 7730
+    },
+    {
+      "epoch": 54.89,
+      "learning_rate": 4.519517388218595e-05,
+      "loss": 0.2002,
+      "step": 7740
+    },
+    {
+      "epoch": 54.96,
+      "learning_rate": 4.512420156139106e-05,
+      "loss": 0.2901,
+      "step": 7750
+    },
+    {
+      "epoch": 55.0,
+      "eval_loss": 0.6224503517150879,
+      "eval_runtime": 8.498,
+      "eval_samples_per_second": 7.413,
+      "eval_steps_per_second": 3.766,
+      "eval_wer": 0.3390974272458878,
+      "step": 7755
+    },
+    {
+      "epoch": 55.04,
+      "learning_rate": 4.505322924059617e-05,
+      "loss": 0.3045,
+      "step": 7760
+    },
+    {
+      "epoch": 55.11,
+      "learning_rate": 4.498225691980128e-05,
+      "loss": 0.284,
+      "step": 7770
+    },
+    {
+      "epoch": 55.18,
+      "learning_rate": 4.4911284599006394e-05,
+      "loss": 0.3475,
+      "step": 7780
+    },
+    {
+      "epoch": 55.25,
+      "learning_rate": 4.48403122782115e-05,
+      "loss": 0.3395,
+      "step": 7790
+    },
+    {
+      "epoch": 55.32,
+      "learning_rate": 4.476933995741661e-05,
+      "loss": 0.3347,
+      "step": 7800
+    },
+    {
+      "epoch": 55.39,
+      "learning_rate": 4.469836763662172e-05,
+      "loss": 0.245,
+      "step": 7810
+    },
+    {
+      "epoch": 55.46,
+      "learning_rate": 4.462739531582683e-05,
+      "loss": 0.3235,
+      "step": 7820
+    },
+    {
+      "epoch": 55.53,
+      "learning_rate": 4.455642299503194e-05,
+      "loss": 0.6167,
+      "step": 7830
+    },
+    {
+      "epoch": 55.6,
+      "learning_rate": 4.448545067423705e-05,
+      "loss": 0.2098,
+      "step": 7840
+    },
+    {
+      "epoch": 55.67,
+      "learning_rate": 4.4414478353442156e-05,
+      "loss": 0.1866,
+      "step": 7850
+    },
+    {
+      "epoch": 55.74,
+      "learning_rate": 4.434350603264727e-05,
+      "loss": 0.3688,
+      "step": 7860
+    },
+    {
+      "epoch": 55.82,
+      "learning_rate": 4.427253371185238e-05,
+      "loss": 0.2521,
+      "step": 7870
+    },
+    {
+      "epoch": 55.89,
+      "learning_rate": 4.420156139105749e-05,
+      "loss": 0.2041,
+      "step": 7880
+    },
+    {
+      "epoch": 55.96,
+      "learning_rate": 4.41305890702626e-05,
+      "loss": 0.2847,
+      "step": 7890
+    },
+    {
+      "epoch": 56.0,
+      "eval_loss": 0.7384737730026245,
+      "eval_runtime": 8.6447,
+      "eval_samples_per_second": 7.288,
+      "eval_steps_per_second": 3.702,
+      "eval_wer": 0.3390974272458878,
+      "step": 7896
+    },
+    {
+      "epoch": 56.03,
+      "learning_rate": 4.4059616749467706e-05,
+      "loss": 0.2371,
+      "step": 7900
+    },
+    {
+      "epoch": 56.1,
+      "learning_rate": 4.398864442867282e-05,
+      "loss": 0.2683,
+      "step": 7910
+    },
+    {
+      "epoch": 56.17,
+      "learning_rate": 4.3917672107877924e-05,
+      "loss": 0.2573,
+      "step": 7920
+    },
+    {
+      "epoch": 56.24,
+      "learning_rate": 4.384669978708304e-05,
+      "loss": 0.2967,
+      "step": 7930
+    },
+    {
+      "epoch": 56.31,
+      "learning_rate": 4.377572746628815e-05,
+      "loss": 0.3052,
+      "step": 7940
+    },
+    {
+      "epoch": 56.38,
+      "learning_rate": 4.3704755145493256e-05,
+      "loss": 0.2639,
+      "step": 7950
+    },
+    {
+      "epoch": 56.45,
+      "learning_rate": 4.363378282469837e-05,
+      "loss": 0.2826,
+      "step": 7960
+    },
+    {
+      "epoch": 56.52,
+      "learning_rate": 4.356281050390348e-05,
+      "loss": 0.3564,
+      "step": 7970
+    },
+    {
+      "epoch": 56.6,
+      "learning_rate": 4.3491838183108594e-05,
+      "loss": 0.23,
+      "step": 7980
+    },
+    {
+      "epoch": 56.67,
+      "learning_rate": 4.34208658623137e-05,
+      "loss": 0.2456,
+      "step": 7990
+    },
+    {
+      "epoch": 56.74,
+      "learning_rate": 4.334989354151881e-05,
+      "loss": 0.3744,
+      "step": 8000
+    },
+    {
+      "epoch": 56.81,
+      "learning_rate": 4.327892122072392e-05,
+      "loss": 0.3109,
+      "step": 8010
+    },
+    {
+      "epoch": 56.88,
+      "learning_rate": 4.320794889992903e-05,
+      "loss": 0.2497,
+      "step": 8020
+    },
+    {
+      "epoch": 56.95,
+      "learning_rate": 4.3136976579134144e-05,
+      "loss": 0.2621,
+      "step": 8030
+    },
+    {
+      "epoch": 57.0,
+      "eval_loss": 0.7240817546844482,
+      "eval_runtime": 8.5689,
+      "eval_samples_per_second": 7.352,
+      "eval_steps_per_second": 3.734,
+      "eval_wer": 0.3496415014761704,
+      "step": 8037
+    },
+    {
+      "epoch": 57.02,
+      "learning_rate": 4.306600425833925e-05,
+      "loss": 0.2675,
+      "step": 8040
+    },
+    {
+      "epoch": 57.09,
+      "learning_rate": 4.299503193754436e-05,
+      "loss": 0.3007,
+      "step": 8050
+    },
+    {
+      "epoch": 57.16,
+      "learning_rate": 4.292405961674947e-05,
+      "loss": 0.2588,
+      "step": 8060
+    },
+    {
+      "epoch": 57.23,
+      "learning_rate": 4.285308729595458e-05,
+      "loss": 0.3343,
+      "step": 8070
+    },
+    {
+      "epoch": 57.3,
+      "learning_rate": 4.278211497515969e-05,
+      "loss": 0.3248,
+      "step": 8080
+    },
+    {
+      "epoch": 57.38,
+      "learning_rate": 4.27111426543648e-05,
+      "loss": 0.1702,
+      "step": 8090
+    },
+    {
+      "epoch": 57.45,
+      "learning_rate": 4.264017033356991e-05,
+      "loss": 0.2126,
+      "step": 8100
+    },
+    {
+      "epoch": 57.52,
+      "learning_rate": 4.256919801277502e-05,
+      "loss": 0.2449,
+      "step": 8110
+    },
+    {
+      "epoch": 57.59,
+      "learning_rate": 4.249822569198013e-05,
+      "loss": 0.3149,
+      "step": 8120
+    },
+    {
+      "epoch": 57.66,
+      "learning_rate": 4.242725337118524e-05,
+      "loss": 0.249,
+      "step": 8130
+    },
+    {
+      "epoch": 57.73,
+      "learning_rate": 4.235628105039035e-05,
+      "loss": 0.3214,
+      "step": 8140
+    },
+    {
+      "epoch": 57.8,
+      "learning_rate": 4.2285308729595456e-05,
+      "loss": 0.2728,
+      "step": 8150
+    },
+    {
+      "epoch": 57.87,
+      "learning_rate": 4.221433640880057e-05,
+      "loss": 0.1911,
+      "step": 8160
+    },
+    {
+      "epoch": 57.94,
+      "learning_rate": 4.214336408800568e-05,
+      "loss": 0.2477,
+      "step": 8170
+    },
+    {
+      "epoch": 58.0,
+      "eval_loss": 0.6957076191902161,
+      "eval_runtime": 8.5146,
+      "eval_samples_per_second": 7.399,
+      "eval_steps_per_second": 3.758,
+      "eval_wer": 0.34289329396878954,
+      "step": 8178
+    },
+    {
+      "epoch": 58.01,
+      "learning_rate": 4.207239176721079e-05,
+      "loss": 0.2933,
+      "step": 8180
+    },
+    {
+      "epoch": 58.09,
+      "learning_rate": 4.20014194464159e-05,
+      "loss": 0.277,
+      "step": 8190
+    },
+    {
+      "epoch": 58.16,
+      "learning_rate": 4.1930447125621006e-05,
+      "loss": 0.2261,
+      "step": 8200
+    },
+    {
+      "epoch": 58.23,
+      "learning_rate": 4.185947480482612e-05,
+      "loss": 0.3048,
+      "step": 8210
+    },
+    {
+      "epoch": 58.3,
+      "learning_rate": 4.1788502484031225e-05,
+      "loss": 0.3011,
+      "step": 8220
+    },
+    {
+      "epoch": 58.37,
+      "learning_rate": 4.171753016323634e-05,
+      "loss": 0.2998,
+      "step": 8230
+    },
+    {
+      "epoch": 58.44,
+      "learning_rate": 4.164655784244145e-05,
+      "loss": 0.2122,
+      "step": 8240
+    },
+    {
+      "epoch": 58.51,
+      "learning_rate": 4.1575585521646556e-05,
+      "loss": 0.3195,
+      "step": 8250
+    },
+    {
+      "epoch": 58.58,
+      "learning_rate": 4.150461320085167e-05,
+      "loss": 0.2645,
+      "step": 8260
+    },
+    {
+      "epoch": 58.65,
+      "learning_rate": 4.143364088005678e-05,
+      "loss": 0.2668,
+      "step": 8270
+    },
+    {
+      "epoch": 58.72,
+      "learning_rate": 4.1362668559261895e-05,
+      "loss": 0.2873,
+      "step": 8280
+    },
+    {
+      "epoch": 58.79,
+      "learning_rate": 4.1291696238467e-05,
+      "loss": 0.3274,
+      "step": 8290
+    },
+    {
+      "epoch": 58.87,
+      "learning_rate": 4.122072391767211e-05,
+      "loss": 0.1785,
+      "step": 8300
+    },
+    {
+      "epoch": 58.94,
+      "learning_rate": 4.114975159687722e-05,
+      "loss": 0.3147,
+      "step": 8310
+    },
+    {
+      "epoch": 59.0,
+      "eval_loss": 0.6808367371559143,
+      "eval_runtime": 8.5164,
+      "eval_samples_per_second": 7.398,
+      "eval_steps_per_second": 3.757,
+      "eval_wer": 0.3424715309995782,
+      "step": 8319
+    },
+    {
+      "epoch": 59.01,
+      "learning_rate": 4.107877927608233e-05,
+      "loss": 0.7616,
+      "step": 8320
+    },
+    {
+      "epoch": 59.08,
+      "learning_rate": 4.1007806955287445e-05,
+      "loss": 0.3139,
+      "step": 8330
+    },
+    {
+      "epoch": 59.15,
+      "learning_rate": 4.093683463449255e-05,
+      "loss": 0.2172,
+      "step": 8340
+    },
+    {
+      "epoch": 59.22,
+      "learning_rate": 4.086586231369766e-05,
+      "loss": 0.2523,
+      "step": 8350
+    },
+    {
+      "epoch": 59.29,
+      "learning_rate": 4.079488999290277e-05,
+      "loss": 0.2458,
+      "step": 8360
+    },
+    {
+      "epoch": 59.36,
+      "learning_rate": 4.072391767210788e-05,
+      "loss": 0.2708,
+      "step": 8370
+    },
+    {
+      "epoch": 59.43,
+      "learning_rate": 4.065294535131299e-05,
+      "loss": 0.2279,
+      "step": 8380
+    },
+    {
+      "epoch": 59.5,
+      "learning_rate": 4.05819730305181e-05,
+      "loss": 0.2461,
+      "step": 8390
+    },
+    {
+      "epoch": 59.57,
+      "learning_rate": 4.051100070972321e-05,
+      "loss": 0.3414,
+      "step": 8400
+    },
+    {
+      "epoch": 59.65,
+      "learning_rate": 4.044002838892832e-05,
+      "loss": 0.2425,
+      "step": 8410
+    },
+    {
+      "epoch": 59.72,
+      "learning_rate": 4.036905606813343e-05,
+      "loss": 0.5483,
+      "step": 8420
+    },
+    {
+      "epoch": 59.79,
+      "learning_rate": 4.029808374733854e-05,
+      "loss": 0.4511,
+      "step": 8430
+    },
+    {
+      "epoch": 59.86,
+      "learning_rate": 4.022711142654365e-05,
+      "loss": 0.2809,
+      "step": 8440
+    },
+    {
+      "epoch": 59.93,
+      "learning_rate": 4.015613910574876e-05,
+      "loss": 0.2493,
+      "step": 8450
+    },
+    {
+      "epoch": 60.0,
+      "learning_rate": 4.008516678495387e-05,
+      "loss": 0.3761,
+      "step": 8460
+    },
+    {
+      "epoch": 60.0,
+      "eval_loss": 0.6709910035133362,
+      "eval_runtime": 8.5011,
+      "eval_samples_per_second": 7.411,
+      "eval_steps_per_second": 3.764,
+      "eval_wer": 0.34500210881484605,
+      "step": 8460
+    },
+    {
+      "epoch": 60.07,
+      "learning_rate": 4.0014194464158975e-05,
+      "loss": 0.2432,
+      "step": 8470
+    },
+    {
+      "epoch": 60.14,
+      "learning_rate": 3.994322214336409e-05,
+      "loss": 0.209,
+      "step": 8480
+    },
+    {
+      "epoch": 60.21,
+      "learning_rate": 3.98722498225692e-05,
+      "loss": 0.2951,
+      "step": 8490
+    },
+    {
+      "epoch": 60.28,
+      "learning_rate": 3.980127750177431e-05,
+      "loss": 0.2551,
+      "step": 8500
+    },
+    {
+      "epoch": 60.35,
+      "learning_rate": 3.973030518097942e-05,
+      "loss": 0.2393,
+      "step": 8510
+    },
+    {
+      "epoch": 60.43,
+      "learning_rate": 3.9659332860184525e-05,
+      "loss": 0.2293,
+      "step": 8520
+    },
+    {
+      "epoch": 60.5,
+      "learning_rate": 3.958836053938964e-05,
+      "loss": 0.4045,
+      "step": 8530
+    },
+    {
+      "epoch": 60.57,
+      "learning_rate": 3.951738821859475e-05,
+      "loss": 0.3139,
+      "step": 8540
+    },
+    {
+      "epoch": 60.64,
+      "learning_rate": 3.945351312987935e-05,
+      "loss": 0.9282,
+      "step": 8550
+    },
+    {
+      "epoch": 60.71,
+      "learning_rate": 3.938254080908446e-05,
+      "loss": 0.272,
+      "step": 8560
+    },
+    {
+      "epoch": 60.78,
+      "learning_rate": 3.931156848828957e-05,
+      "loss": 0.3775,
+      "step": 8570
+    },
+    {
+      "epoch": 60.85,
+      "learning_rate": 3.924059616749468e-05,
+      "loss": 0.304,
+      "step": 8580
+    },
+    {
+      "epoch": 60.92,
+      "learning_rate": 3.916962384669979e-05,
+      "loss": 0.2076,
+      "step": 8590
+    },
+    {
+      "epoch": 60.99,
+      "learning_rate": 3.90986515259049e-05,
+      "loss": 0.2609,
+      "step": 8600
+    },
+    {
+      "epoch": 61.0,
+      "eval_loss": 0.66289883852005,
+      "eval_runtime": 8.4612,
+      "eval_samples_per_second": 7.446,
+      "eval_steps_per_second": 3.782,
+      "eval_wer": 0.33445803458456347,
+      "step": 8601
+    },
+    {
+      "epoch": 61.06,
+      "learning_rate": 3.9027679205110006e-05,
+      "loss": 0.3749,
+      "step": 8610
+    },
+    {
+      "epoch": 61.13,
+      "learning_rate": 3.895670688431512e-05,
+      "loss": 0.2239,
+      "step": 8620
+    },
+    {
+      "epoch": 61.21,
+      "learning_rate": 3.8885734563520224e-05,
+      "loss": 0.2322,
+      "step": 8630
+    },
+    {
+      "epoch": 61.28,
+      "learning_rate": 3.881476224272534e-05,
+      "loss": 0.2893,
+      "step": 8640
+    },
+    {
+      "epoch": 61.35,
+      "learning_rate": 3.874378992193045e-05,
+      "loss": 0.2045,
+      "step": 8650
+    },
+    {
+      "epoch": 61.42,
+      "learning_rate": 3.8672817601135556e-05,
+      "loss": 0.2521,
+      "step": 8660
+    },
+    {
+      "epoch": 61.49,
+      "learning_rate": 3.860184528034067e-05,
+      "loss": 0.29,
+      "step": 8670
+    },
+    {
+      "epoch": 61.56,
+      "learning_rate": 3.8530872959545774e-05,
+      "loss": 0.2893,
+      "step": 8680
+    },
+    {
+      "epoch": 61.63,
+      "learning_rate": 3.845990063875089e-05,
+      "loss": 0.2351,
+      "step": 8690
+    },
+    {
+      "epoch": 61.7,
+      "learning_rate": 3.8388928317956e-05,
+      "loss": 0.2352,
+      "step": 8700
+    },
+    {
+      "epoch": 61.77,
+      "learning_rate": 3.8317955997161106e-05,
+      "loss": 0.3264,
+      "step": 8710
+    },
+    {
+      "epoch": 61.84,
+      "learning_rate": 3.824698367636622e-05,
+      "loss": 0.3034,
+      "step": 8720
+    },
+    {
+      "epoch": 61.91,
+      "learning_rate": 3.817601135557133e-05,
+      "loss": 0.2271,
+      "step": 8730
+    },
+    {
+      "epoch": 61.99,
+      "learning_rate": 3.8105039034776444e-05,
+      "loss": 0.388,
+      "step": 8740
+    },
+    {
+      "epoch": 62.0,
+      "eval_loss": 0.6687957644462585,
+      "eval_runtime": 8.5937,
+      "eval_samples_per_second": 7.331,
+      "eval_steps_per_second": 3.724,
+      "eval_wer": 0.34626739772247994,
+      "step": 8742
+    },
+    {
+      "epoch": 62.06,
+      "learning_rate": 3.803406671398155e-05,
+      "loss": 0.2523,
+      "step": 8750
+    },
+    {
+      "epoch": 62.13,
+      "learning_rate": 3.796309439318666e-05,
+      "loss": 0.1775,
+      "step": 8760
+    },
+    {
+      "epoch": 62.2,
+      "learning_rate": 3.789212207239177e-05,
+      "loss": 0.2654,
+      "step": 8770
+    },
+    {
+      "epoch": 62.27,
+      "learning_rate": 3.782114975159688e-05,
+      "loss": 0.35,
+      "step": 8780
+    },
+    {
+      "epoch": 62.34,
+      "learning_rate": 3.7750177430801994e-05,
+      "loss": 0.3156,
+      "step": 8790
+    },
+    {
+      "epoch": 62.41,
+      "learning_rate": 3.76792051100071e-05,
+      "loss": 0.3129,
+      "step": 8800
+    },
+    {
+      "epoch": 62.48,
+      "learning_rate": 3.760823278921221e-05,
+      "loss": 0.31,
+      "step": 8810
+    },
+    {
+      "epoch": 62.55,
+      "learning_rate": 3.753726046841732e-05,
+      "loss": 0.2929,
+      "step": 8820
+    },
+    {
+      "epoch": 62.62,
+      "learning_rate": 3.746628814762243e-05,
+      "loss": 0.2805,
+      "step": 8830
+    },
+    {
+      "epoch": 62.7,
+      "learning_rate": 3.739531582682754e-05,
+      "loss": 0.3599,
+      "step": 8840
+    },
+    {
+      "epoch": 62.77,
+      "learning_rate": 3.732434350603265e-05,
+      "loss": 0.3024,
+      "step": 8850
+    },
+    {
+      "epoch": 62.84,
+      "learning_rate": 3.7253371185237756e-05,
+      "loss": 0.2745,
+      "step": 8860
+    },
+    {
+      "epoch": 62.91,
+      "learning_rate": 3.718239886444287e-05,
+      "loss": 0.2187,
+      "step": 8870
+    },
+    {
+      "epoch": 62.98,
+      "learning_rate": 3.711142654364798e-05,
+      "loss": 0.3684,
+      "step": 8880
+    },
+    {
+      "epoch": 63.0,
+      "eval_loss": 0.7018281817436218,
+      "eval_runtime": 8.7278,
+      "eval_samples_per_second": 7.218,
+      "eval_steps_per_second": 3.666,
+      "eval_wer": 0.33403627161535215,
+      "step": 8883
+    },
+    {
+      "epoch": 63.05,
+      "learning_rate": 3.704045422285309e-05,
+      "loss": 0.2946,
+      "step": 8890
+    },
+    {
+      "epoch": 63.12,
+      "learning_rate": 3.69694819020582e-05,
+      "loss": 0.2203,
+      "step": 8900
+    },
+    {
+      "epoch": 63.19,
+      "learning_rate": 3.6898509581263306e-05,
+      "loss": 0.3315,
+      "step": 8910
+    },
+    {
+      "epoch": 63.26,
+      "learning_rate": 3.682753726046842e-05,
+      "loss": 0.2491,
+      "step": 8920
+    },
+    {
+      "epoch": 63.33,
+      "learning_rate": 3.6756564939673525e-05,
+      "loss": 0.284,
+      "step": 8930
+    },
+    {
+      "epoch": 63.4,
+      "learning_rate": 3.668559261887864e-05,
+      "loss": 0.2002,
+      "step": 8940
+    },
+    {
+      "epoch": 63.48,
+      "learning_rate": 3.661462029808375e-05,
+      "loss": 0.2953,
+      "step": 8950
+    },
+    {
+      "epoch": 63.55,
+      "learning_rate": 3.6543647977288856e-05,
+      "loss": 0.3717,
+      "step": 8960
+    },
+    {
+      "epoch": 63.62,
+      "learning_rate": 3.647267565649397e-05,
+      "loss": 0.3134,
+      "step": 8970
+    },
+    {
+      "epoch": 63.69,
+      "learning_rate": 3.6401703335699075e-05,
+      "loss": 0.4772,
+      "step": 8980
+    },
+    {
+      "epoch": 63.76,
+      "learning_rate": 3.633073101490419e-05,
+      "loss": 0.5724,
+      "step": 8990
+    },
+    {
+      "epoch": 63.83,
+      "learning_rate": 3.62597586941093e-05,
+      "loss": 0.2927,
+      "step": 9000
+    },
+    {
+      "epoch": 63.9,
+      "learning_rate": 3.6188786373314406e-05,
+      "loss": 0.1963,
+      "step": 9010
+    },
+    {
+      "epoch": 63.97,
+      "learning_rate": 3.611781405251952e-05,
+      "loss": 0.2494,
+      "step": 9020
+    },
+    {
+      "epoch": 64.0,
+      "eval_loss": 0.6611393094062805,
+      "eval_runtime": 8.8901,
+      "eval_samples_per_second": 7.087,
+      "eval_steps_per_second": 3.6,
+      "eval_wer": 0.3399409531843104,
+      "step": 9024
+    },
+    {
+      "epoch": 64.04,
+      "learning_rate": 3.604684173172463e-05,
+      "loss": 0.2768,
+      "step": 9030
+    },
+    {
+      "epoch": 64.11,
+      "learning_rate": 3.5975869410929744e-05,
+      "loss": 0.2305,
+      "step": 9040
+    },
+    {
+      "epoch": 64.18,
+      "learning_rate": 3.590489709013485e-05,
+      "loss": 0.2864,
+      "step": 9050
+    },
+    {
+      "epoch": 64.26,
+      "learning_rate": 3.583392476933996e-05,
+      "loss": 0.2588,
+      "step": 9060
+    },
+    {
+      "epoch": 64.33,
+      "learning_rate": 3.576295244854507e-05,
+      "loss": 0.3611,
+      "step": 9070
+    },
+    {
+      "epoch": 64.4,
+      "learning_rate": 3.569198012775018e-05,
+      "loss": 0.2813,
+      "step": 9080
+    },
+    {
+      "epoch": 64.47,
+      "learning_rate": 3.562100780695529e-05,
+      "loss": 0.2351,
+      "step": 9090
+    },
+    {
+      "epoch": 64.54,
+      "learning_rate": 3.55500354861604e-05,
+      "loss": 0.3237,
+      "step": 9100
+    },
+    {
+      "epoch": 64.61,
+      "learning_rate": 3.547906316536551e-05,
+      "loss": 0.2062,
+      "step": 9110
+    },
+    {
+      "epoch": 64.68,
+      "learning_rate": 3.540809084457062e-05,
+      "loss": 0.2735,
+      "step": 9120
+    },
+    {
+      "epoch": 64.75,
+      "learning_rate": 3.533711852377573e-05,
+      "loss": 0.2478,
+      "step": 9130
+    },
+    {
+      "epoch": 64.82,
+      "learning_rate": 3.526614620298084e-05,
+      "loss": 0.2994,
+      "step": 9140
+    },
+    {
+      "epoch": 64.89,
+      "learning_rate": 3.519517388218595e-05,
+      "loss": 0.2504,
+      "step": 9150
+    },
+    {
+      "epoch": 64.96,
+      "learning_rate": 3.5124201561391056e-05,
+      "loss": 0.2641,
+      "step": 9160
+    },
+    {
+      "epoch": 65.0,
+      "eval_loss": 0.6827792525291443,
+      "eval_runtime": 8.5165,
+      "eval_samples_per_second": 7.397,
+      "eval_steps_per_second": 3.757,
+      "eval_wer": 0.3399409531843104,
+      "step": 9165
+    },
+    {
+      "epoch": 65.04,
+      "learning_rate": 3.505322924059617e-05,
+      "loss": 0.3018,
+      "step": 9170
+    },
+    {
+      "epoch": 65.11,
+      "learning_rate": 3.498225691980128e-05,
+      "loss": 0.2247,
+      "step": 9180
+    },
+    {
+      "epoch": 65.18,
+      "learning_rate": 3.491128459900639e-05,
+      "loss": 0.4066,
+      "step": 9190
+    },
+    {
+      "epoch": 65.25,
+      "learning_rate": 3.48403122782115e-05,
+      "loss": 0.2373,
+      "step": 9200
+    },
+    {
+      "epoch": 65.32,
+      "learning_rate": 3.4769339957416607e-05,
+      "loss": 0.2791,
+      "step": 9210
+    },
+    {
+      "epoch": 65.39,
+      "learning_rate": 3.469836763662172e-05,
+      "loss": 0.2285,
+      "step": 9220
+    },
+    {
+      "epoch": 65.46,
+      "learning_rate": 3.4627395315826825e-05,
+      "loss": 0.4232,
+      "step": 9230
+    },
+    {
+      "epoch": 65.53,
+      "learning_rate": 3.455642299503194e-05,
+      "loss": 0.2646,
+      "step": 9240
+    },
+    {
+      "epoch": 65.6,
+      "learning_rate": 3.4485450674237044e-05,
+      "loss": 0.1887,
+      "step": 9250
+    },
+    {
+      "epoch": 65.67,
+      "learning_rate": 3.4414478353442157e-05,
+      "loss": 0.4363,
+      "step": 9260
+    },
+    {
+      "epoch": 65.74,
+      "learning_rate": 3.434350603264727e-05,
+      "loss": 0.4587,
+      "step": 9270
+    },
+    {
+      "epoch": 65.82,
+      "learning_rate": 3.4272533711852375e-05,
+      "loss": 0.3252,
+      "step": 9280
+    },
+    {
+      "epoch": 65.89,
+      "learning_rate": 3.420156139105749e-05,
+      "loss": 0.2025,
+      "step": 9290
+    },
+    {
+      "epoch": 65.96,
+      "learning_rate": 3.41305890702626e-05,
+      "loss": 0.2716,
+      "step": 9300
+    },
+    {
+      "epoch": 66.0,
+      "eval_loss": 0.6408915519714355,
+      "eval_runtime": 8.5558,
+      "eval_samples_per_second": 7.363,
+      "eval_steps_per_second": 3.74,
+      "eval_wer": 0.32939687895402786,
+      "step": 9306
+    },
+    {
+      "epoch": 66.03,
+      "learning_rate": 3.405961674946771e-05,
+      "loss": 0.2812,
+      "step": 9310
+    },
+    {
+      "epoch": 66.1,
+      "learning_rate": 3.398864442867282e-05,
+      "loss": 0.2285,
+      "step": 9320
+    },
+    {
+      "epoch": 66.17,
+      "learning_rate": 3.391767210787793e-05,
+      "loss": 0.3145,
+      "step": 9330
+    },
+    {
+      "epoch": 66.24,
+      "learning_rate": 3.3846699787083045e-05,
+      "loss": 0.2783,
+      "step": 9340
+    },
+    {
+      "epoch": 66.31,
+      "learning_rate": 3.377572746628815e-05,
+      "loss": 0.3086,
+      "step": 9350
+    },
+    {
+      "epoch": 66.38,
+      "learning_rate": 3.3704755145493264e-05,
+      "loss": 0.1965,
+      "step": 9360
+    },
+    {
+      "epoch": 66.45,
+      "learning_rate": 3.363378282469837e-05,
+      "loss": 0.2795,
+      "step": 9370
+    },
+    {
+      "epoch": 66.52,
+      "learning_rate": 3.356281050390348e-05,
+      "loss": 0.2213,
+      "step": 9380
+    },
+    {
+      "epoch": 66.6,
+      "learning_rate": 3.349183818310859e-05,
+      "loss": 0.2883,
+      "step": 9390
+    },
+    {
+      "epoch": 66.67,
+      "learning_rate": 3.34208658623137e-05,
+      "loss": 0.1711,
+      "step": 9400
+    },
+    {
+      "epoch": 66.74,
+      "learning_rate": 3.3349893541518814e-05,
+      "loss": 0.2907,
+      "step": 9410
+    },
+    {
+      "epoch": 66.81,
+      "learning_rate": 3.327892122072392e-05,
+      "loss": 0.2755,
+      "step": 9420
+    },
+    {
+      "epoch": 66.88,
+      "learning_rate": 3.320794889992903e-05,
+      "loss": 0.2137,
+      "step": 9430
+    },
+    {
+      "epoch": 66.95,
+      "learning_rate": 3.313697657913414e-05,
+      "loss": 0.2595,
+      "step": 9440
+    },
+    {
+      "epoch": 67.0,
+      "eval_loss": 0.6055729389190674,
+      "eval_runtime": 8.5087,
+      "eval_samples_per_second": 7.404,
+      "eval_steps_per_second": 3.761,
+      "eval_wer": 0.3230704344158583,
+      "step": 9447
+    },
+    {
+      "epoch": 67.02,
+      "learning_rate": 3.306600425833925e-05,
+      "loss": 0.3914,
+      "step": 9450
+    },
+    {
+      "epoch": 67.09,
+      "learning_rate": 3.299503193754436e-05,
+      "loss": 0.3529,
+      "step": 9460
+    },
+    {
+      "epoch": 67.16,
+      "learning_rate": 3.292405961674947e-05,
+      "loss": 0.2574,
+      "step": 9470
+    },
+    {
+      "epoch": 67.23,
+      "learning_rate": 3.2853087295954576e-05,
+      "loss": 0.3945,
+      "step": 9480
+    },
+    {
+      "epoch": 67.3,
+      "learning_rate": 3.278211497515969e-05,
+      "loss": 0.3274,
+      "step": 9490
+    },
+    {
+      "epoch": 67.38,
+      "learning_rate": 3.27111426543648e-05,
+      "loss": 0.2448,
+      "step": 9500
+    },
+    {
+      "epoch": 67.45,
+      "learning_rate": 3.264017033356991e-05,
+      "loss": 0.262,
+      "step": 9510
+    },
+    {
+      "epoch": 67.52,
+      "learning_rate": 3.256919801277502e-05,
+      "loss": 0.3499,
+      "step": 9520
+    },
+    {
+      "epoch": 67.59,
+      "learning_rate": 3.2498225691980126e-05,
+      "loss": 0.2505,
+      "step": 9530
+    },
+    {
+      "epoch": 67.66,
+      "learning_rate": 3.242725337118524e-05,
+      "loss": 0.2293,
+      "step": 9540
+    },
+    {
+      "epoch": 67.73,
+      "learning_rate": 3.2356281050390344e-05,
+      "loss": 0.3501,
+      "step": 9550
+    },
+    {
+      "epoch": 67.8,
+      "learning_rate": 3.228530872959546e-05,
+      "loss": 0.2544,
+      "step": 9560
+    },
+    {
+      "epoch": 67.87,
+      "learning_rate": 3.221433640880057e-05,
+      "loss": 0.4223,
+      "step": 9570
+    },
+    {
+      "epoch": 67.94,
+      "learning_rate": 3.2143364088005676e-05,
+      "loss": 0.2683,
+      "step": 9580
+    },
+    {
+      "epoch": 68.0,
+      "eval_loss": 0.6202967166900635,
+      "eval_runtime": 8.5252,
+      "eval_samples_per_second": 7.39,
+      "eval_steps_per_second": 3.754,
+      "eval_wer": 0.3331927456769296,
+      "step": 9588
+    },
+    {
+      "epoch": 68.01,
+      "learning_rate": 3.207239176721079e-05,
+      "loss": 0.2926,
+      "step": 9590
+    },
+    {
+      "epoch": 68.09,
+      "learning_rate": 3.20014194464159e-05,
+      "loss": 0.3014,
+      "step": 9600
+    },
+    {
+      "epoch": 68.16,
+      "learning_rate": 3.193044712562101e-05,
+      "loss": 0.1857,
+      "step": 9610
+    },
+    {
+      "epoch": 68.23,
+      "learning_rate": 3.185947480482612e-05,
+      "loss": 0.2324,
+      "step": 9620
+    },
+    {
+      "epoch": 68.3,
+      "learning_rate": 3.178850248403123e-05,
+      "loss": 0.2454,
+      "step": 9630
+    },
+    {
+      "epoch": 68.37,
+      "learning_rate": 3.1717530163236345e-05,
+      "loss": 0.186,
+      "step": 9640
+    },
+    {
+      "epoch": 68.44,
+      "learning_rate": 3.164655784244145e-05,
+      "loss": 0.3591,
+      "step": 9650
+    },
+    {
+      "epoch": 68.51,
+      "learning_rate": 3.1575585521646564e-05,
+      "loss": 0.394,
+      "step": 9660
+    },
+    {
+      "epoch": 68.58,
+      "learning_rate": 3.150461320085167e-05,
+      "loss": 0.2601,
+      "step": 9670
+    },
+    {
+      "epoch": 68.65,
+      "learning_rate": 3.143364088005678e-05,
+      "loss": 0.2442,
+      "step": 9680
+    },
+    {
+      "epoch": 68.72,
+      "learning_rate": 3.136266855926189e-05,
+      "loss": 0.2,
+      "step": 9690
+    },
+    {
+      "epoch": 68.79,
+      "learning_rate": 3.1291696238467e-05,
+      "loss": 0.3524,
+      "step": 9700
+    },
+    {
+      "epoch": 68.87,
+      "learning_rate": 3.122072391767211e-05,
+      "loss": 0.2951,
+      "step": 9710
+    },
+    {
+      "epoch": 68.94,
+      "learning_rate": 3.114975159687722e-05,
+      "loss": 0.2571,
+      "step": 9720
+    },
+    {
+      "epoch": 69.0,
+      "eval_loss": 0.6483610272407532,
+      "eval_runtime": 8.5368,
+      "eval_samples_per_second": 7.38,
+      "eval_steps_per_second": 3.748,
+      "eval_wer": 0.3336145086461409,
+      "step": 9729
+    },
+    {
+      "epoch": 69.01,
+      "learning_rate": 3.107877927608233e-05,
+      "loss": 0.3195,
+      "step": 9730
+    },
+    {
+      "epoch": 69.08,
+      "learning_rate": 3.100780695528744e-05,
+      "loss": 0.2935,
+      "step": 9740
+    },
+    {
+      "epoch": 69.15,
+      "learning_rate": 3.093683463449255e-05,
+      "loss": 0.2653,
+      "step": 9750
+    },
+    {
+      "epoch": 69.22,
+      "learning_rate": 3.086586231369766e-05,
+      "loss": 0.223,
+      "step": 9760
+    },
+    {
+      "epoch": 69.29,
+      "learning_rate": 3.079488999290277e-05,
+      "loss": 0.2389,
+      "step": 9770
+    },
+    {
+      "epoch": 69.36,
+      "learning_rate": 3.0723917672107876e-05,
+      "loss": 0.2729,
+      "step": 9780
+    },
+    {
+      "epoch": 69.43,
+      "learning_rate": 3.065294535131299e-05,
+      "loss": 0.2366,
+      "step": 9790
+    },
+    {
+      "epoch": 69.5,
+      "learning_rate": 3.05819730305181e-05,
+      "loss": 0.2661,
+      "step": 9800
+    },
+    {
+      "epoch": 69.57,
+      "learning_rate": 3.0511000709723207e-05,
+      "loss": 0.3213,
+      "step": 9810
+    },
+    {
+      "epoch": 69.65,
+      "learning_rate": 3.044002838892832e-05,
+      "loss": 0.1436,
+      "step": 9820
+    },
+    {
+      "epoch": 69.72,
+      "learning_rate": 3.036905606813343e-05,
+      "loss": 0.3178,
+      "step": 9830
+    },
+    {
+      "epoch": 69.79,
+      "learning_rate": 3.029808374733854e-05,
+      "loss": 0.2802,
+      "step": 9840
+    },
+    {
+      "epoch": 69.86,
+      "learning_rate": 3.0227111426543648e-05,
+      "loss": 0.228,
+      "step": 9850
+    },
+    {
+      "epoch": 69.93,
+      "learning_rate": 3.015613910574876e-05,
+      "loss": 0.2093,
+      "step": 9860
+    },
+    {
+      "epoch": 70.0,
+      "learning_rate": 3.0085166784953874e-05,
+      "loss": 0.2593,
+      "step": 9870
+    },
+    {
+      "epoch": 70.0,
+      "eval_loss": 0.6596509218215942,
+      "eval_runtime": 8.5696,
+      "eval_samples_per_second": 7.352,
+      "eval_steps_per_second": 3.734,
+      "eval_wer": 0.32939687895402786,
+      "step": 9870
+    },
+    {
+      "epoch": 70.07,
+      "learning_rate": 3.001419446415898e-05,
+      "loss": 0.2956,
+      "step": 9880
+    },
+    {
+      "epoch": 70.14,
+      "learning_rate": 2.9943222143364092e-05,
+      "loss": 0.2681,
+      "step": 9890
+    },
+    {
+      "epoch": 70.21,
+      "learning_rate": 2.9872249822569198e-05,
+      "loss": 0.2387,
+      "step": 9900
+    },
+    {
+      "epoch": 70.28,
+      "learning_rate": 2.980127750177431e-05,
+      "loss": 0.2632,
+      "step": 9910
+    },
+    {
+      "epoch": 70.35,
+      "learning_rate": 2.9730305180979417e-05,
+      "loss": 0.1867,
+      "step": 9920
+    },
+    {
+      "epoch": 70.43,
+      "learning_rate": 2.965933286018453e-05,
+      "loss": 0.187,
+      "step": 9930
+    },
+    {
+      "epoch": 70.5,
+      "learning_rate": 2.9588360539389636e-05,
+      "loss": 0.2356,
+      "step": 9940
+    },
+    {
+      "epoch": 70.57,
+      "learning_rate": 2.951738821859475e-05,
+      "loss": 0.2665,
+      "step": 9950
+    },
+    {
+      "epoch": 70.64,
+      "learning_rate": 2.944641589779986e-05,
+      "loss": 0.1672,
+      "step": 9960
+    },
+    {
+      "epoch": 70.71,
+      "learning_rate": 2.9375443577004967e-05,
+      "loss": 0.2955,
+      "step": 9970
+    },
+    {
+      "epoch": 70.78,
+      "learning_rate": 2.930447125621008e-05,
+      "loss": 0.3505,
+      "step": 9980
+    },
+    {
+      "epoch": 70.85,
+      "learning_rate": 2.923349893541519e-05,
+      "loss": 0.3014,
+      "step": 9990
+    },
+    {
+      "epoch": 70.92,
+      "learning_rate": 2.9162526614620302e-05,
+      "loss": 0.3333,
+      "step": 10000
+    },
+    {
+      "epoch": 70.99,
+      "learning_rate": 2.9091554293825408e-05,
+      "loss": 0.229,
+      "step": 10010
+    },
+    {
+      "epoch": 71.0,
+      "eval_loss": 0.6353591084480286,
+      "eval_runtime": 8.6152,
+      "eval_samples_per_second": 7.313,
+      "eval_steps_per_second": 3.714,
+      "eval_wer": 0.32349219738506957,
+      "step": 10011
+    },
+    {
+      "epoch": 71.06,
+      "learning_rate": 2.902058197303052e-05,
+      "loss": 0.2533,
+      "step": 10020
+    },
+    {
+      "epoch": 71.13,
+      "learning_rate": 2.8949609652235633e-05,
+      "loss": 0.2492,
+      "step": 10030
+    },
+    {
+      "epoch": 71.21,
+      "learning_rate": 2.887863733144074e-05,
+      "loss": 0.2775,
+      "step": 10040
+    },
+    {
+      "epoch": 71.28,
+      "learning_rate": 2.8807665010645852e-05,
+      "loss": 0.2395,
+      "step": 10050
+    },
+    {
+      "epoch": 71.35,
+      "learning_rate": 2.8736692689850958e-05,
+      "loss": 0.1978,
+      "step": 10060
+    },
+    {
+      "epoch": 71.42,
+      "learning_rate": 2.866572036905607e-05,
+      "loss": 0.2921,
+      "step": 10070
+    },
+    {
+      "epoch": 71.49,
+      "learning_rate": 2.8594748048261176e-05,
+      "loss": 0.238,
+      "step": 10080
+    },
+    {
+      "epoch": 71.56,
+      "learning_rate": 2.852377572746629e-05,
+      "loss": 0.3778,
+      "step": 10090
+    },
+    {
+      "epoch": 71.63,
+      "learning_rate": 2.8452803406671402e-05,
+      "loss": 0.1989,
+      "step": 10100
+    },
+    {
+      "epoch": 71.7,
+      "learning_rate": 2.8381831085876508e-05,
+      "loss": 0.2499,
+      "step": 10110
+    },
+    {
+      "epoch": 71.77,
+      "learning_rate": 2.831085876508162e-05,
+      "loss": 0.1927,
+      "step": 10120
+    },
+    {
+      "epoch": 71.84,
+      "learning_rate": 2.823988644428673e-05,
+      "loss": 0.312,
+      "step": 10130
+    },
+    {
+      "epoch": 71.91,
+      "learning_rate": 2.816891412349184e-05,
+      "loss": 0.1921,
+      "step": 10140
+    },
+    {
+      "epoch": 71.99,
+      "learning_rate": 2.809794180269695e-05,
+      "loss": 0.281,
+      "step": 10150
+    },
+    {
+      "epoch": 72.0,
+      "eval_loss": 0.6397704482078552,
+      "eval_runtime": 8.6804,
+      "eval_samples_per_second": 7.258,
+      "eval_steps_per_second": 3.686,
+      "eval_wer": 0.32939687895402786,
+      "step": 10152
+    },
+    {
+      "epoch": 72.06,
+      "learning_rate": 2.802696948190206e-05,
+      "loss": 0.2701,
+      "step": 10160
+    },
+    {
+      "epoch": 72.13,
+      "learning_rate": 2.7955997161107167e-05,
+      "loss": 0.2616,
+      "step": 10170
+    },
+    {
+      "epoch": 72.2,
+      "learning_rate": 2.788502484031228e-05,
+      "loss": 0.2467,
+      "step": 10180
+    },
+    {
+      "epoch": 72.27,
+      "learning_rate": 2.7814052519517393e-05,
+      "loss": 0.2987,
+      "step": 10190
+    },
+    {
+      "epoch": 72.34,
+      "learning_rate": 2.77430801987225e-05,
+      "loss": 0.2641,
+      "step": 10200
+    },
+    {
+      "epoch": 72.41,
+      "learning_rate": 2.767210787792761e-05,
+      "loss": 0.2708,
+      "step": 10210
+    },
+    {
+      "epoch": 72.48,
+      "learning_rate": 2.7601135557132717e-05,
+      "loss": 0.2941,
+      "step": 10220
+    },
+    {
+      "epoch": 72.55,
+      "learning_rate": 2.753016323633783e-05,
+      "loss": 0.6741,
+      "step": 10230
+    },
+    {
+      "epoch": 72.62,
+      "learning_rate": 2.7459190915542936e-05,
+      "loss": 0.1653,
+      "step": 10240
+    },
+    {
+      "epoch": 72.7,
+      "learning_rate": 2.738821859474805e-05,
+      "loss": 0.244,
+      "step": 10250
+    },
+    {
+      "epoch": 72.77,
+      "learning_rate": 2.731724627395316e-05,
+      "loss": 0.2485,
+      "step": 10260
+    },
+    {
+      "epoch": 72.84,
+      "learning_rate": 2.7246273953158267e-05,
+      "loss": 0.1573,
+      "step": 10270
+    },
+    {
+      "epoch": 72.91,
+      "learning_rate": 2.717530163236338e-05,
+      "loss": 0.2292,
+      "step": 10280
+    },
+    {
+      "epoch": 72.98,
+      "learning_rate": 2.710432931156849e-05,
+      "loss": 0.3779,
+      "step": 10290
+    },
+    {
+      "epoch": 73.0,
+      "eval_loss": 0.6870745420455933,
+      "eval_runtime": 8.7299,
+      "eval_samples_per_second": 7.217,
+      "eval_steps_per_second": 3.666,
+      "eval_wer": 0.33445803458456347,
+      "step": 10293
+    },
+    {
+      "epoch": 73.05,
+      "learning_rate": 2.7033356990773602e-05,
+      "loss": 0.3413,
+      "step": 10300
+    },
+    {
+      "epoch": 73.12,
+      "learning_rate": 2.6962384669978708e-05,
+      "loss": 0.245,
+      "step": 10310
+    },
+    {
+      "epoch": 73.19,
+      "learning_rate": 2.689141234918382e-05,
+      "loss": 0.195,
+      "step": 10320
+    },
+    {
+      "epoch": 73.26,
+      "learning_rate": 2.6820440028388934e-05,
+      "loss": 0.6626,
+      "step": 10330
+    },
+    {
+      "epoch": 73.33,
+      "learning_rate": 2.674946770759404e-05,
+      "loss": 0.2569,
+      "step": 10340
+    },
+    {
+      "epoch": 73.4,
+      "learning_rate": 2.6678495386799152e-05,
+      "loss": 0.246,
+      "step": 10350
+    },
+    {
+      "epoch": 73.48,
+      "learning_rate": 2.6607523066004258e-05,
+      "loss": 0.8424,
+      "step": 10360
+    },
+    {
+      "epoch": 73.55,
+      "learning_rate": 2.653655074520937e-05,
+      "loss": 0.2843,
+      "step": 10370
+    },
+    {
+      "epoch": 73.62,
+      "learning_rate": 2.6465578424414477e-05,
+      "loss": 0.1892,
+      "step": 10380
+    },
+    {
+      "epoch": 73.69,
+      "learning_rate": 2.639460610361959e-05,
+      "loss": 0.2237,
+      "step": 10390
+    },
+    {
+      "epoch": 73.76,
+      "learning_rate": 2.6323633782824696e-05,
+      "loss": 0.2172,
+      "step": 10400
+    },
+    {
+      "epoch": 73.83,
+      "learning_rate": 2.625266146202981e-05,
+      "loss": 0.2791,
+      "step": 10410
+    },
+    {
+      "epoch": 73.9,
+      "learning_rate": 2.618168914123492e-05,
+      "loss": 0.2826,
+      "step": 10420
+    },
+    {
+      "epoch": 73.97,
+      "learning_rate": 2.611071682044003e-05,
+      "loss": 0.2998,
+      "step": 10430
+    },
+    {
+      "epoch": 74.0,
+      "eval_loss": 0.7328982949256897,
+      "eval_runtime": 8.546,
+      "eval_samples_per_second": 7.372,
+      "eval_steps_per_second": 3.744,
+      "eval_wer": 0.33234921973850695,
+      "step": 10434
+    },
+    {
+      "epoch": 74.04,
+      "learning_rate": 2.604684173172463e-05,
+      "loss": 0.851,
+      "step": 10440
+    },
+    {
+      "epoch": 74.11,
+      "learning_rate": 2.597586941092974e-05,
+      "loss": 0.2103,
+      "step": 10450
+    },
+    {
+      "epoch": 74.18,
+      "learning_rate": 2.590489709013485e-05,
+      "loss": 0.2126,
+      "step": 10460
+    },
+    {
+      "epoch": 74.26,
+      "learning_rate": 2.5833924769339957e-05,
+      "loss": 0.8986,
+      "step": 10470
+    },
+    {
+      "epoch": 74.33,
+      "learning_rate": 2.576295244854507e-05,
+      "loss": 0.2385,
+      "step": 10480
+    },
+    {
+      "epoch": 74.4,
+      "learning_rate": 2.5691980127750176e-05,
+      "loss": 0.2488,
+      "step": 10490
+    },
+    {
+      "epoch": 74.47,
+      "learning_rate": 2.562100780695529e-05,
+      "loss": 0.2289,
+      "step": 10500
+    },
+    {
+      "epoch": 74.54,
+      "learning_rate": 2.55500354861604e-05,
+      "loss": 0.821,
+      "step": 10510
+    },
+    {
+      "epoch": 74.61,
+      "learning_rate": 2.5479063165365507e-05,
+      "loss": 0.2269,
+      "step": 10520
+    },
+    {
+      "epoch": 74.68,
+      "learning_rate": 2.540809084457062e-05,
+      "loss": 0.2409,
+      "step": 10530
+    },
+    {
+      "epoch": 74.75,
+      "learning_rate": 2.5337118523775726e-05,
+      "loss": 0.3034,
+      "step": 10540
+    },
+    {
+      "epoch": 74.82,
+      "learning_rate": 2.526614620298084e-05,
+      "loss": 0.2625,
+      "step": 10550
+    },
+    {
+      "epoch": 74.89,
+      "learning_rate": 2.5195173882185945e-05,
+      "loss": 0.332,
+      "step": 10560
+    },
+    {
+      "epoch": 74.96,
+      "learning_rate": 2.5124201561391057e-05,
+      "loss": 0.2095,
+      "step": 10570
+    },
+    {
+      "epoch": 75.0,
+      "eval_loss": 0.7365297675132751,
+      "eval_runtime": 8.5567,
+      "eval_samples_per_second": 7.363,
+      "eval_steps_per_second": 3.74,
+      "eval_wer": 0.3239139603542809,
+      "step": 10575
+    },
+    {
+      "epoch": 75.04,
+      "learning_rate": 2.505322924059617e-05,
+      "loss": 0.2735,
+      "step": 10580
+    },
+    {
+      "epoch": 75.11,
+      "learning_rate": 2.498225691980128e-05,
+      "loss": 0.158,
+      "step": 10590
+    },
+    {
+      "epoch": 75.18,
+      "learning_rate": 2.491128459900639e-05,
+      "loss": 0.2581,
+      "step": 10600
+    },
+    {
+      "epoch": 75.25,
+      "learning_rate": 2.48403122782115e-05,
+      "loss": 0.2818,
+      "step": 10610
+    },
+    {
+      "epoch": 75.32,
+      "learning_rate": 2.476933995741661e-05,
+      "loss": 0.3283,
+      "step": 10620
+    },
+    {
+      "epoch": 75.39,
+      "learning_rate": 2.469836763662172e-05,
+      "loss": 0.2447,
+      "step": 10630
+    },
+    {
+      "epoch": 75.46,
+      "learning_rate": 2.462739531582683e-05,
+      "loss": 0.237,
+      "step": 10640
+    },
+    {
+      "epoch": 75.53,
+      "learning_rate": 2.455642299503194e-05,
+      "loss": 0.2796,
+      "step": 10650
+    },
+    {
+      "epoch": 75.6,
+      "learning_rate": 2.4485450674237048e-05,
+      "loss": 0.1742,
+      "step": 10660
+    },
+    {
+      "epoch": 75.67,
+      "learning_rate": 2.4414478353442157e-05,
+      "loss": 0.2076,
+      "step": 10670
+    },
+    {
+      "epoch": 75.74,
+      "learning_rate": 2.4343506032647267e-05,
+      "loss": 0.2157,
+      "step": 10680
+    },
+    {
+      "epoch": 75.82,
+      "learning_rate": 2.427253371185238e-05,
+      "loss": 0.2834,
+      "step": 10690
+    },
+    {
+      "epoch": 75.89,
+      "learning_rate": 2.420156139105749e-05,
+      "loss": 0.227,
+      "step": 10700
+    },
+    {
+      "epoch": 75.96,
+      "learning_rate": 2.4130589070262598e-05,
+      "loss": 0.247,
+      "step": 10710
+    },
+    {
+      "epoch": 76.0,
+      "eval_loss": 0.6383683681488037,
+      "eval_runtime": 8.5016,
+      "eval_samples_per_second": 7.41,
+      "eval_steps_per_second": 3.764,
+      "eval_wer": 0.32897511598481655,
+      "step": 10716
+    },
+    {
+      "epoch": 76.03,
+      "learning_rate": 2.4059616749467708e-05,
+      "loss": 0.2927,
+      "step": 10720
+    },
+    {
+      "epoch": 76.1,
+      "learning_rate": 2.3988644428672817e-05,
+      "loss": 0.2487,
+      "step": 10730
+    },
+    {
+      "epoch": 76.17,
+      "learning_rate": 2.391767210787793e-05,
+      "loss": 0.2622,
+      "step": 10740
+    },
+    {
+      "epoch": 76.24,
+      "learning_rate": 2.384669978708304e-05,
+      "loss": 0.3057,
+      "step": 10750
+    },
+    {
+      "epoch": 76.31,
+      "learning_rate": 2.3775727466288148e-05,
+      "loss": 0.3467,
+      "step": 10760
+    },
+    {
+      "epoch": 76.38,
+      "learning_rate": 2.370475514549326e-05,
+      "loss": 0.2154,
+      "step": 10770
+    },
+    {
+      "epoch": 76.45,
+      "learning_rate": 2.363378282469837e-05,
+      "loss": 0.2281,
+      "step": 10780
+    },
+    {
+      "epoch": 76.52,
+      "learning_rate": 2.356281050390348e-05,
+      "loss": 0.3083,
+      "step": 10790
+    },
+    {
+      "epoch": 76.6,
+      "learning_rate": 2.349183818310859e-05,
+      "loss": 0.201,
+      "step": 10800
+    },
+    {
+      "epoch": 76.67,
+      "learning_rate": 2.34208658623137e-05,
+      "loss": 0.2349,
+      "step": 10810
+    },
+    {
+      "epoch": 76.74,
+      "learning_rate": 2.3349893541518808e-05,
+      "loss": 0.2511,
+      "step": 10820
+    },
+    {
+      "epoch": 76.81,
+      "learning_rate": 2.3278921220723917e-05,
+      "loss": 0.2944,
+      "step": 10830
+    },
+    {
+      "epoch": 76.88,
+      "learning_rate": 2.320794889992903e-05,
+      "loss": 0.1829,
+      "step": 10840
+    },
+    {
+      "epoch": 76.95,
+      "learning_rate": 2.313697657913414e-05,
+      "loss": 0.2095,
+      "step": 10850
+    },
+    {
+      "epoch": 77.0,
+      "eval_loss": 0.6703174710273743,
+      "eval_runtime": 8.4959,
+      "eval_samples_per_second": 7.415,
+      "eval_steps_per_second": 3.767,
+      "eval_wer": 0.33445803458456347,
+      "step": 10857
+    },
+    {
+      "epoch": 77.02,
+      "learning_rate": 2.306600425833925e-05,
+      "loss": 0.2628,
+      "step": 10860
+    },
+    {
+      "epoch": 77.09,
+      "learning_rate": 2.2995031937544358e-05,
+      "loss": 0.2327,
+      "step": 10870
+    },
+    {
+      "epoch": 77.16,
+      "learning_rate": 2.2924059616749467e-05,
+      "loss": 0.3404,
+      "step": 10880
+    },
+    {
+      "epoch": 77.23,
+      "learning_rate": 2.285308729595458e-05,
+      "loss": 0.2323,
+      "step": 10890
+    },
+    {
+      "epoch": 77.3,
+      "learning_rate": 2.278211497515969e-05,
+      "loss": 0.217,
+      "step": 10900
+    },
+    {
+      "epoch": 77.38,
+      "learning_rate": 2.27111426543648e-05,
+      "loss": 0.1934,
+      "step": 10910
+    },
+    {
+      "epoch": 77.45,
+      "learning_rate": 2.264017033356991e-05,
+      "loss": 0.2541,
+      "step": 10920
+    },
+    {
+      "epoch": 77.52,
+      "learning_rate": 2.256919801277502e-05,
+      "loss": 0.2838,
+      "step": 10930
+    },
+    {
+      "epoch": 77.59,
+      "learning_rate": 2.249822569198013e-05,
+      "loss": 0.2788,
+      "step": 10940
+    },
+    {
+      "epoch": 77.66,
+      "learning_rate": 2.242725337118524e-05,
+      "loss": 0.2112,
+      "step": 10950
+    },
+    {
+      "epoch": 77.73,
+      "learning_rate": 2.235628105039035e-05,
+      "loss": 0.26,
+      "step": 10960
+    },
+    {
+      "epoch": 77.8,
+      "learning_rate": 2.2285308729595458e-05,
+      "loss": 0.2879,
+      "step": 10970
+    },
+    {
+      "epoch": 77.87,
+      "learning_rate": 2.2214336408800567e-05,
+      "loss": 0.1768,
+      "step": 10980
+    },
+    {
+      "epoch": 77.94,
+      "learning_rate": 2.2143364088005677e-05,
+      "loss": 0.2074,
+      "step": 10990
+    },
+    {
+      "epoch": 78.0,
+      "eval_loss": 0.6576946377754211,
+      "eval_runtime": 8.4681,
+      "eval_samples_per_second": 7.44,
+      "eval_steps_per_second": 3.779,
+      "eval_wer": 0.3424715309995782,
+      "step": 10998
+    },
+    {
+      "epoch": 78.01,
+      "learning_rate": 2.207239176721079e-05,
+      "loss": 0.2426,
+      "step": 11000
+    },
+    {
+      "epoch": 78.09,
+      "learning_rate": 2.20014194464159e-05,
+      "loss": 0.2299,
+      "step": 11010
+    },
+    {
+      "epoch": 78.16,
+      "learning_rate": 2.1930447125621008e-05,
+      "loss": 0.2093,
+      "step": 11020
+    },
+    {
+      "epoch": 78.23,
+      "learning_rate": 2.1859474804826117e-05,
+      "loss": 0.2821,
+      "step": 11030
+    },
+    {
+      "epoch": 78.3,
+      "learning_rate": 2.178850248403123e-05,
+      "loss": 0.2183,
+      "step": 11040
+    },
+    {
+      "epoch": 78.37,
+      "learning_rate": 2.171753016323634e-05,
+      "loss": 0.2704,
+      "step": 11050
+    },
+    {
+      "epoch": 78.44,
+      "learning_rate": 2.164655784244145e-05,
+      "loss": 0.1946,
+      "step": 11060
+    },
+    {
+      "epoch": 78.51,
+      "learning_rate": 2.157558552164656e-05,
+      "loss": 1.5477,
+      "step": 11070
+    },
+    {
+      "epoch": 78.58,
+      "learning_rate": 2.150461320085167e-05,
+      "loss": 0.332,
+      "step": 11080
+    },
+    {
+      "epoch": 78.65,
+      "learning_rate": 2.143364088005678e-05,
+      "loss": 0.329,
+      "step": 11090
+    },
+    {
+      "epoch": 78.72,
+      "learning_rate": 2.136266855926189e-05,
+      "loss": 0.2844,
+      "step": 11100
+    },
+    {
+      "epoch": 78.79,
+      "learning_rate": 2.1291696238467e-05,
+      "loss": 0.202,
+      "step": 11110
+    },
+    {
+      "epoch": 78.87,
+      "learning_rate": 2.1220723917672108e-05,
+      "loss": 0.2461,
+      "step": 11120
+    },
+    {
+      "epoch": 78.94,
+      "learning_rate": 2.1149751596877217e-05,
+      "loss": 0.2519,
+      "step": 11130
+    },
+    {
+      "epoch": 79.0,
+      "eval_loss": 0.6359320282936096,
+      "eval_runtime": 8.4615,
+      "eval_samples_per_second": 7.445,
+      "eval_steps_per_second": 3.782,
+      "eval_wer": 0.3369886123998313,
+      "step": 11139
+    },
+    {
+      "epoch": 79.01,
+      "learning_rate": 2.1078779276082327e-05,
+      "loss": 0.2678,
+      "step": 11140
+    },
+    {
+      "epoch": 79.08,
+      "learning_rate": 2.100780695528744e-05,
+      "loss": 0.3181,
+      "step": 11150
+    },
+    {
+      "epoch": 79.15,
+      "learning_rate": 2.093683463449255e-05,
+      "loss": 0.2068,
+      "step": 11160
+    },
+    {
+      "epoch": 79.22,
+      "learning_rate": 2.0865862313697658e-05,
+      "loss": 0.2488,
+      "step": 11170
+    },
+    {
+      "epoch": 79.29,
+      "learning_rate": 2.0794889992902768e-05,
+      "loss": 0.2542,
+      "step": 11180
+    },
+    {
+      "epoch": 79.36,
+      "learning_rate": 2.072391767210788e-05,
+      "loss": 0.19,
+      "step": 11190
+    },
+    {
+      "epoch": 79.43,
+      "learning_rate": 2.065294535131299e-05,
+      "loss": 0.2304,
+      "step": 11200
+    },
+    {
+      "epoch": 79.5,
+      "learning_rate": 2.05819730305181e-05,
+      "loss": 0.2657,
+      "step": 11210
+    },
+    {
+      "epoch": 79.57,
+      "learning_rate": 2.051100070972321e-05,
+      "loss": 0.301,
+      "step": 11220
+    },
+    {
+      "epoch": 79.65,
+      "learning_rate": 2.044002838892832e-05,
+      "loss": 0.2092,
+      "step": 11230
+    },
+    {
+      "epoch": 79.72,
+      "learning_rate": 2.036905606813343e-05,
+      "loss": 0.2297,
+      "step": 11240
+    },
+    {
+      "epoch": 79.79,
+      "learning_rate": 2.029808374733854e-05,
+      "loss": 0.2357,
+      "step": 11250
+    },
+    {
+      "epoch": 79.86,
+      "learning_rate": 2.022711142654365e-05,
+      "loss": 0.1763,
+      "step": 11260
+    },
+    {
+      "epoch": 79.93,
+      "learning_rate": 2.015613910574876e-05,
+      "loss": 0.3121,
+      "step": 11270
+    },
+    {
+      "epoch": 80.0,
+      "learning_rate": 2.0085166784953868e-05,
+      "loss": 0.2046,
+      "step": 11280
+    },
+    {
+      "epoch": 80.0,
+      "eval_loss": 0.6221576929092407,
+      "eval_runtime": 8.5217,
+      "eval_samples_per_second": 7.393,
+      "eval_steps_per_second": 3.755,
+      "eval_wer": 0.3256010122311261,
+      "step": 11280
+    },
+    {
+      "epoch": 80.07,
+      "learning_rate": 2.0014194464158977e-05,
+      "loss": 0.2205,
+      "step": 11290
+    },
+    {
+      "epoch": 80.14,
+      "learning_rate": 1.994322214336409e-05,
+      "loss": 0.1807,
+      "step": 11300
+    },
+    {
+      "epoch": 80.21,
+      "learning_rate": 1.98722498225692e-05,
+      "loss": 0.258,
+      "step": 11310
+    },
+    {
+      "epoch": 80.28,
+      "learning_rate": 1.980127750177431e-05,
+      "loss": 0.3637,
+      "step": 11320
+    },
+    {
+      "epoch": 80.35,
+      "learning_rate": 1.9730305180979418e-05,
+      "loss": 0.1809,
+      "step": 11330
+    },
+    {
+      "epoch": 80.43,
+      "learning_rate": 1.965933286018453e-05,
+      "loss": 0.2168,
+      "step": 11340
+    },
+    {
+      "epoch": 80.5,
+      "learning_rate": 1.958836053938964e-05,
+      "loss": 0.2207,
+      "step": 11350
+    },
+    {
+      "epoch": 80.57,
+      "learning_rate": 1.951738821859475e-05,
+      "loss": 0.3524,
+      "step": 11360
+    },
+    {
+      "epoch": 80.64,
+      "learning_rate": 1.944641589779986e-05,
+      "loss": 0.2245,
+      "step": 11370
+    },
+    {
+      "epoch": 80.71,
+      "learning_rate": 1.937544357700497e-05,
+      "loss": 0.216,
+      "step": 11380
+    },
+    {
+      "epoch": 80.78,
+      "learning_rate": 1.930447125621008e-05,
+      "loss": 0.3583,
+      "step": 11390
+    },
+    {
+      "epoch": 80.85,
+      "learning_rate": 1.923349893541519e-05,
+      "loss": 0.3116,
+      "step": 11400
+    },
+    {
+      "epoch": 80.92,
+      "learning_rate": 1.91625266146203e-05,
+      "loss": 0.2244,
+      "step": 11410
+    },
+    {
+      "epoch": 80.99,
+      "learning_rate": 1.909155429382541e-05,
+      "loss": 1.3195,
+      "step": 11420
+    },
+    {
+      "epoch": 81.0,
+      "eval_loss": 0.6126046776771545,
+      "eval_runtime": 8.5497,
+      "eval_samples_per_second": 7.369,
+      "eval_steps_per_second": 3.743,
+      "eval_wer": 0.33445803458456347,
+      "step": 11421
+    },
+    {
+      "epoch": 81.06,
+      "learning_rate": 1.9020581973030518e-05,
+      "loss": 0.3326,
+      "step": 11430
+    },
+    {
+      "epoch": 81.13,
+      "learning_rate": 1.8949609652235627e-05,
+      "loss": 0.2092,
+      "step": 11440
+    },
+    {
+      "epoch": 81.21,
+      "learning_rate": 1.8878637331440737e-05,
+      "loss": 0.2808,
+      "step": 11450
+    },
+    {
+      "epoch": 81.28,
+      "learning_rate": 1.880766501064585e-05,
+      "loss": 0.2458,
+      "step": 11460
+    },
+    {
+      "epoch": 81.35,
+      "learning_rate": 1.873669268985096e-05,
+      "loss": 0.2638,
+      "step": 11470
+    },
+    {
+      "epoch": 81.42,
+      "learning_rate": 1.8665720369056068e-05,
+      "loss": 0.1828,
+      "step": 11480
+    },
+    {
+      "epoch": 81.49,
+      "learning_rate": 1.859474804826118e-05,
+      "loss": 0.2367,
+      "step": 11490
+    },
+    {
+      "epoch": 81.56,
+      "learning_rate": 1.852377572746629e-05,
+      "loss": 0.265,
+      "step": 11500
+    },
+    {
+      "epoch": 81.63,
+      "learning_rate": 1.84528034066714e-05,
+      "loss": 0.1512,
+      "step": 11510
+    },
+    {
+      "epoch": 81.7,
+      "learning_rate": 1.838183108587651e-05,
+      "loss": 0.2238,
+      "step": 11520
+    },
+    {
+      "epoch": 81.77,
+      "learning_rate": 1.8310858765081618e-05,
+      "loss": 0.2214,
+      "step": 11530
+    },
+    {
+      "epoch": 81.84,
+      "learning_rate": 1.823988644428673e-05,
+      "loss": 0.2005,
+      "step": 11540
+    },
+    {
+      "epoch": 81.91,
+      "learning_rate": 1.816891412349184e-05,
+      "loss": 0.2559,
+      "step": 11550
+    },
+    {
+      "epoch": 81.99,
+      "learning_rate": 1.809794180269695e-05,
+      "loss": 0.2821,
+      "step": 11560
+    },
+    {
+      "epoch": 82.0,
+      "eval_loss": 0.6193426251411438,
+      "eval_runtime": 8.4485,
+      "eval_samples_per_second": 7.457,
+      "eval_steps_per_second": 3.788,
+      "eval_wer": 0.32939687895402786,
+      "step": 11562
+    },
+    {
+      "epoch": 82.06,
+      "learning_rate": 1.802696948190206e-05,
+      "loss": 0.2758,
+      "step": 11570
+    },
+    {
+      "epoch": 82.13,
+      "learning_rate": 1.7955997161107168e-05,
+      "loss": 0.2292,
+      "step": 11580
+    },
+    {
+      "epoch": 82.2,
+      "learning_rate": 1.7885024840312278e-05,
+      "loss": 0.2359,
+      "step": 11590
+    },
+    {
+      "epoch": 82.27,
+      "learning_rate": 1.7814052519517387e-05,
+      "loss": 0.2328,
+      "step": 11600
+    },
+    {
+      "epoch": 82.34,
+      "learning_rate": 1.77430801987225e-05,
+      "loss": 0.2812,
+      "step": 11610
+    },
+    {
+      "epoch": 82.41,
+      "learning_rate": 1.767210787792761e-05,
+      "loss": 0.1795,
+      "step": 11620
+    },
+    {
+      "epoch": 82.48,
+      "learning_rate": 1.7601135557132718e-05,
+      "loss": 0.2196,
+      "step": 11630
+    },
+    {
+      "epoch": 82.55,
+      "learning_rate": 1.753016323633783e-05,
+      "loss": 0.2557,
+      "step": 11640
+    },
+    {
+      "epoch": 82.62,
+      "learning_rate": 1.745919091554294e-05,
+      "loss": 0.2415,
+      "step": 11650
+    },
+    {
+      "epoch": 82.7,
+      "learning_rate": 1.738821859474805e-05,
+      "loss": 0.2675,
+      "step": 11660
+    },
+    {
+      "epoch": 82.77,
+      "learning_rate": 1.731724627395316e-05,
+      "loss": 0.2298,
+      "step": 11670
+    },
+    {
+      "epoch": 82.84,
+      "learning_rate": 1.724627395315827e-05,
+      "loss": 0.2678,
+      "step": 11680
+    },
+    {
+      "epoch": 82.91,
+      "learning_rate": 1.717530163236338e-05,
+      "loss": 0.267,
+      "step": 11690
+    },
+    {
+      "epoch": 82.98,
+      "learning_rate": 1.710432931156849e-05,
+      "loss": 0.3256,
+      "step": 11700
+    },
+    {
+      "epoch": 83.0,
+      "eval_loss": 0.6140345931053162,
+      "eval_runtime": 8.4261,
+      "eval_samples_per_second": 7.477,
+      "eval_steps_per_second": 3.798,
+      "eval_wer": 0.3336145086461409,
+      "step": 11703
+    },
+    {
+      "epoch": 83.05,
+      "learning_rate": 1.70333569907736e-05,
+      "loss": 0.2812,
+      "step": 11710
+    },
+    {
+      "epoch": 83.12,
+      "learning_rate": 1.696238466997871e-05,
+      "loss": 0.2222,
+      "step": 11720
+    },
+    {
+      "epoch": 83.19,
+      "learning_rate": 1.689141234918382e-05,
+      "loss": 0.2015,
+      "step": 11730
+    },
+    {
+      "epoch": 83.26,
+      "learning_rate": 1.6820440028388928e-05,
+      "loss": 0.2767,
+      "step": 11740
+    },
+    {
+      "epoch": 83.33,
+      "learning_rate": 1.6749467707594037e-05,
+      "loss": 0.191,
+      "step": 11750
+    },
+    {
+      "epoch": 83.4,
+      "learning_rate": 1.6678495386799146e-05,
+      "loss": 0.2005,
+      "step": 11760
+    },
+    {
+      "epoch": 83.48,
+      "learning_rate": 1.660752306600426e-05,
+      "loss": 0.286,
+      "step": 11770
+    },
+    {
+      "epoch": 83.55,
+      "learning_rate": 1.653655074520937e-05,
+      "loss": 0.3074,
+      "step": 11780
+    },
+    {
+      "epoch": 83.62,
+      "learning_rate": 1.6465578424414478e-05,
+      "loss": 0.2082,
+      "step": 11790
+    },
+    {
+      "epoch": 83.69,
+      "learning_rate": 1.639460610361959e-05,
+      "loss": 0.2358,
+      "step": 11800
+    },
+    {
+      "epoch": 83.76,
+      "learning_rate": 1.63236337828247e-05,
+      "loss": 0.2544,
+      "step": 11810
+    },
+    {
+      "epoch": 83.83,
+      "learning_rate": 1.625266146202981e-05,
+      "loss": 0.3836,
+      "step": 11820
+    },
+    {
+      "epoch": 83.9,
+      "learning_rate": 1.618168914123492e-05,
+      "loss": 0.2533,
+      "step": 11830
+    },
+    {
+      "epoch": 83.97,
+      "learning_rate": 1.611071682044003e-05,
+      "loss": 0.2743,
+      "step": 11840
+    },
+    {
+      "epoch": 84.0,
+      "eval_loss": 0.6204140186309814,
+      "eval_runtime": 8.4692,
+      "eval_samples_per_second": 7.439,
+      "eval_steps_per_second": 3.778,
+      "eval_wer": 0.32897511598481655,
+      "step": 11844
+    },
+    {
+      "epoch": 84.04,
+      "learning_rate": 1.603974449964514e-05,
+      "loss": 0.3008,
+      "step": 11850
+    },
+    {
+      "epoch": 84.11,
+      "learning_rate": 1.596877217885025e-05,
+      "loss": 0.2039,
+      "step": 11860
+    },
+    {
+      "epoch": 84.18,
+      "learning_rate": 1.589779985805536e-05,
+      "loss": 0.2392,
+      "step": 11870
+    },
+    {
+      "epoch": 84.26,
+      "learning_rate": 1.582682753726047e-05,
+      "loss": 0.2138,
+      "step": 11880
+    },
+    {
+      "epoch": 84.33,
+      "learning_rate": 1.5755855216465578e-05,
+      "loss": 0.42,
+      "step": 11890
+    },
+    {
+      "epoch": 84.4,
+      "learning_rate": 1.5684882895670687e-05,
+      "loss": 0.2008,
+      "step": 11900
+    },
+    {
+      "epoch": 84.47,
+      "learning_rate": 1.5613910574875797e-05,
+      "loss": 0.1873,
+      "step": 11910
+    },
+    {
+      "epoch": 84.54,
+      "learning_rate": 1.554293825408091e-05,
+      "loss": 0.2086,
+      "step": 11920
+    },
+    {
+      "epoch": 84.61,
+      "learning_rate": 1.547196593328602e-05,
+      "loss": 0.1595,
+      "step": 11930
+    },
+    {
+      "epoch": 84.68,
+      "learning_rate": 1.5400993612491128e-05,
+      "loss": 0.2545,
+      "step": 11940
+    },
+    {
+      "epoch": 84.75,
+      "learning_rate": 1.533002129169624e-05,
+      "loss": 0.2662,
+      "step": 11950
+    },
+    {
+      "epoch": 84.82,
+      "learning_rate": 1.525904897090135e-05,
+      "loss": 0.2684,
+      "step": 11960
+    },
+    {
+      "epoch": 84.89,
+      "learning_rate": 1.518807665010646e-05,
+      "loss": 0.2389,
+      "step": 11970
+    },
+    {
+      "epoch": 84.96,
+      "learning_rate": 1.5117104329311569e-05,
+      "loss": 0.2761,
+      "step": 11980
+    },
+    {
+      "epoch": 85.0,
+      "eval_loss": 0.6598784923553467,
+      "eval_runtime": 8.4691,
+      "eval_samples_per_second": 7.439,
+      "eval_steps_per_second": 3.778,
+      "eval_wer": 0.3251792492619148,
+      "step": 11985
+    },
+    {
+      "epoch": 85.04,
+      "learning_rate": 1.5046132008516678e-05,
+      "loss": 0.3308,
+      "step": 11990
+    },
+    {
+      "epoch": 85.11,
+      "learning_rate": 1.4975159687721791e-05,
+      "loss": 0.2093,
+      "step": 12000
+    },
+    {
+      "epoch": 85.18,
+      "learning_rate": 1.49041873669269e-05,
+      "loss": 0.243,
+      "step": 12010
+    },
+    {
+      "epoch": 85.25,
+      "learning_rate": 1.483321504613201e-05,
+      "loss": 0.2727,
+      "step": 12020
+    },
+    {
+      "epoch": 85.32,
+      "learning_rate": 1.4762242725337119e-05,
+      "loss": 0.1852,
+      "step": 12030
+    },
+    {
+      "epoch": 85.39,
+      "learning_rate": 1.4691270404542228e-05,
+      "loss": 0.7108,
+      "step": 12040
+    },
+    {
+      "epoch": 85.46,
+      "learning_rate": 1.462029808374734e-05,
+      "loss": 0.256,
+      "step": 12050
+    },
+    {
+      "epoch": 85.53,
+      "learning_rate": 1.4549325762952449e-05,
+      "loss": 0.2647,
+      "step": 12060
+    },
+    {
+      "epoch": 85.6,
+      "learning_rate": 1.4478353442157561e-05,
+      "loss": 0.2309,
+      "step": 12070
+    },
+    {
+      "epoch": 85.67,
+      "learning_rate": 1.440738112136267e-05,
+      "loss": 0.2282,
+      "step": 12080
+    },
+    {
+      "epoch": 85.74,
+      "learning_rate": 1.433640880056778e-05,
+      "loss": 0.2037,
+      "step": 12090
+    },
+    {
+      "epoch": 85.82,
+      "learning_rate": 1.426543647977289e-05,
+      "loss": 0.3535,
+      "step": 12100
+    },
+    {
+      "epoch": 85.89,
+      "learning_rate": 1.4194464158977999e-05,
+      "loss": 0.1825,
+      "step": 12110
+    },
+    {
+      "epoch": 85.96,
+      "learning_rate": 1.412349183818311e-05,
+      "loss": 0.224,
+      "step": 12120
+    },
+    {
+      "epoch": 86.0,
+      "eval_loss": 0.6580356955528259,
+      "eval_runtime": 8.5118,
+      "eval_samples_per_second": 7.401,
+      "eval_steps_per_second": 3.759,
+      "eval_wer": 0.32939687895402786,
+      "step": 12126
+    },
+    {
+      "epoch": 86.03,
+      "learning_rate": 1.4052519517388219e-05,
+      "loss": 0.3194,
+      "step": 12130
+    },
+    {
+      "epoch": 86.1,
+      "learning_rate": 1.3981547196593328e-05,
+      "loss": 0.2826,
+      "step": 12140
+    },
+    {
+      "epoch": 86.17,
+      "learning_rate": 1.3910574875798441e-05,
+      "loss": 0.2415,
+      "step": 12150
+    },
+    {
+      "epoch": 86.24,
+      "learning_rate": 1.383960255500355e-05,
+      "loss": 0.2334,
+      "step": 12160
+    },
+    {
+      "epoch": 86.31,
+      "learning_rate": 1.376863023420866e-05,
+      "loss": 0.3545,
+      "step": 12170
+    },
+    {
+      "epoch": 86.38,
+      "learning_rate": 1.3697657913413769e-05,
+      "loss": 0.1522,
+      "step": 12180
+    },
+    {
+      "epoch": 86.45,
+      "learning_rate": 1.3626685592618878e-05,
+      "loss": 0.2383,
+      "step": 12190
+    },
+    {
+      "epoch": 86.52,
+      "learning_rate": 1.355571327182399e-05,
+      "loss": 0.3099,
+      "step": 12200
+    },
+    {
+      "epoch": 86.6,
+      "learning_rate": 1.3484740951029099e-05,
+      "loss": 0.1998,
+      "step": 12210
+    },
+    {
+      "epoch": 86.67,
+      "learning_rate": 1.3413768630234208e-05,
+      "loss": 0.2428,
+      "step": 12220
+    },
+    {
+      "epoch": 86.74,
+      "learning_rate": 1.3342796309439321e-05,
+      "loss": 0.2939,
+      "step": 12230
+    },
+    {
+      "epoch": 86.81,
+      "learning_rate": 1.327182398864443e-05,
+      "loss": 0.3152,
+      "step": 12240
+    },
+    {
+      "epoch": 86.88,
+      "learning_rate": 1.320085166784954e-05,
+      "loss": 0.2086,
+      "step": 12250
+    },
+    {
+      "epoch": 86.95,
+      "learning_rate": 1.3129879347054649e-05,
+      "loss": 0.2106,
+      "step": 12260
+    },
+    {
+      "epoch": 87.0,
+      "eval_loss": 0.6298220753669739,
+      "eval_runtime": 8.4291,
+      "eval_samples_per_second": 7.474,
+      "eval_steps_per_second": 3.796,
+      "eval_wer": 0.32939687895402786,
+      "step": 12267
+    },
+    {
+      "epoch": 87.02,
+      "learning_rate": 1.3058907026259758e-05,
+      "loss": 0.2707,
+      "step": 12270
+    },
+    {
+      "epoch": 87.09,
+      "learning_rate": 1.298793470546487e-05,
+      "loss": 0.2684,
+      "step": 12280
+    },
+    {
+      "epoch": 87.16,
+      "learning_rate": 1.2916962384669979e-05,
+      "loss": 0.2236,
+      "step": 12290
+    },
+    {
+      "epoch": 87.23,
+      "learning_rate": 1.2845990063875088e-05,
+      "loss": 0.364,
+      "step": 12300
+    },
+    {
+      "epoch": 87.3,
+      "learning_rate": 1.27750177430802e-05,
+      "loss": 0.2222,
+      "step": 12310
+    },
+    {
+      "epoch": 87.38,
+      "learning_rate": 1.270404542228531e-05,
+      "loss": 0.2293,
+      "step": 12320
+    },
+    {
+      "epoch": 87.45,
+      "learning_rate": 1.263307310149042e-05,
+      "loss": 0.2128,
+      "step": 12330
+    },
+    {
+      "epoch": 87.52,
+      "learning_rate": 1.2562100780695529e-05,
+      "loss": 0.2613,
+      "step": 12340
+    },
+    {
+      "epoch": 87.59,
+      "learning_rate": 1.249112845990064e-05,
+      "loss": 0.2476,
+      "step": 12350
+    },
+    {
+      "epoch": 87.66,
+      "learning_rate": 1.242015613910575e-05,
+      "loss": 0.2292,
+      "step": 12360
+    },
+    {
+      "epoch": 87.73,
+      "learning_rate": 1.234918381831086e-05,
+      "loss": 0.1973,
+      "step": 12370
+    },
+    {
+      "epoch": 87.8,
+      "learning_rate": 1.227821149751597e-05,
+      "loss": 0.3723,
+      "step": 12380
+    },
+    {
+      "epoch": 87.87,
+      "learning_rate": 1.2207239176721079e-05,
+      "loss": 0.1422,
+      "step": 12390
+    },
+    {
+      "epoch": 87.94,
+      "learning_rate": 1.213626685592619e-05,
+      "loss": 0.2706,
+      "step": 12400
+    },
+    {
+      "epoch": 88.0,
+      "eval_loss": 0.6411434412002563,
+      "eval_runtime": 8.4362,
+      "eval_samples_per_second": 7.468,
+      "eval_steps_per_second": 3.793,
+      "eval_wer": 0.3281315900463939,
+      "step": 12408
+    },
+    {
+      "epoch": 88.01,
+      "learning_rate": 1.2065294535131299e-05,
+      "loss": 0.2999,
+      "step": 12410
+    },
+    {
+      "epoch": 88.09,
+      "learning_rate": 1.1994322214336408e-05,
+      "loss": 0.2703,
+      "step": 12420
+    },
+    {
+      "epoch": 88.16,
+      "learning_rate": 1.192334989354152e-05,
+      "loss": 0.2953,
+      "step": 12430
+    },
+    {
+      "epoch": 88.23,
+      "learning_rate": 1.185237757274663e-05,
+      "loss": 0.2437,
+      "step": 12440
+    },
+    {
+      "epoch": 88.3,
+      "learning_rate": 1.178140525195174e-05,
+      "loss": 0.2818,
+      "step": 12450
+    },
+    {
+      "epoch": 88.37,
+      "learning_rate": 1.171043293115685e-05,
+      "loss": 0.1275,
+      "step": 12460
+    },
+    {
+      "epoch": 88.44,
+      "learning_rate": 1.1639460610361959e-05,
+      "loss": 0.2159,
+      "step": 12470
+    },
+    {
+      "epoch": 88.51,
+      "learning_rate": 1.156848828956707e-05,
+      "loss": 0.2239,
+      "step": 12480
+    },
+    {
+      "epoch": 88.58,
+      "learning_rate": 1.1497515968772179e-05,
+      "loss": 0.2818,
+      "step": 12490
+    },
+    {
+      "epoch": 88.65,
+      "learning_rate": 1.142654364797729e-05,
+      "loss": 0.269,
+      "step": 12500
+    },
+    {
+      "epoch": 88.72,
+      "learning_rate": 1.13555713271824e-05,
+      "loss": 0.2666,
+      "step": 12510
+    },
+    {
+      "epoch": 88.79,
+      "learning_rate": 1.128459900638751e-05,
+      "loss": 0.2968,
+      "step": 12520
+    },
+    {
+      "epoch": 88.87,
+      "learning_rate": 1.121362668559262e-05,
+      "loss": 0.2042,
+      "step": 12530
+    },
+    {
+      "epoch": 88.94,
+      "learning_rate": 1.1142654364797729e-05,
+      "loss": 0.2523,
+      "step": 12540
+    },
+    {
+      "epoch": 89.0,
+      "eval_loss": 0.6242873072624207,
+      "eval_runtime": 8.4613,
+      "eval_samples_per_second": 7.446,
+      "eval_steps_per_second": 3.782,
+      "eval_wer": 0.3264445381695487,
+      "step": 12549
+    },
+    {
+      "epoch": 89.01,
+      "learning_rate": 1.1071682044002838e-05,
+      "loss": 0.2966,
+      "step": 12550
+    },
+    {
+      "epoch": 89.08,
+      "learning_rate": 1.100070972320795e-05,
+      "loss": 0.3343,
+      "step": 12560
+    },
+    {
+      "epoch": 89.15,
+      "learning_rate": 1.0929737402413059e-05,
+      "loss": 0.1739,
+      "step": 12570
+    },
+    {
+      "epoch": 89.22,
+      "learning_rate": 1.085876508161817e-05,
+      "loss": 0.2722,
+      "step": 12580
+    },
+    {
+      "epoch": 89.29,
+      "learning_rate": 1.078779276082328e-05,
+      "loss": 0.2263,
+      "step": 12590
+    },
+    {
+      "epoch": 89.36,
+      "learning_rate": 1.071682044002839e-05,
+      "loss": 0.2336,
+      "step": 12600
+    },
+    {
+      "epoch": 89.43,
+      "learning_rate": 1.06458481192335e-05,
+      "loss": 0.2454,
+      "step": 12610
+    },
+    {
+      "epoch": 89.5,
+      "learning_rate": 1.0574875798438609e-05,
+      "loss": 0.2966,
+      "step": 12620
+    },
+    {
+      "epoch": 89.57,
+      "learning_rate": 1.050390347764372e-05,
+      "loss": 0.2043,
+      "step": 12630
+    },
+    {
+      "epoch": 89.65,
+      "learning_rate": 1.0432931156848829e-05,
+      "loss": 0.2351,
+      "step": 12640
+    },
+    {
+      "epoch": 89.72,
+      "learning_rate": 1.036195883605394e-05,
+      "loss": 0.2607,
+      "step": 12650
+    },
+    {
+      "epoch": 89.79,
+      "learning_rate": 1.029098651525905e-05,
+      "loss": 0.2644,
+      "step": 12660
+    },
+    {
+      "epoch": 89.86,
+      "learning_rate": 1.022001419446416e-05,
+      "loss": 0.2912,
+      "step": 12670
+    },
+    {
+      "epoch": 89.93,
+      "learning_rate": 1.014904187366927e-05,
+      "loss": 0.2167,
+      "step": 12680
+    },
+    {
+      "epoch": 90.0,
+      "learning_rate": 1.007806955287438e-05,
+      "loss": 0.3635,
+      "step": 12690
+    },
+    {
+      "epoch": 90.0,
+      "eval_loss": 0.6296595931053162,
+      "eval_runtime": 8.5838,
+      "eval_samples_per_second": 7.339,
+      "eval_steps_per_second": 3.728,
+      "eval_wer": 0.32897511598481655,
+      "step": 12690
+    },
+    {
+      "epoch": 90.07,
+      "learning_rate": 1.0007097232079489e-05,
+      "loss": 0.2771,
+      "step": 12700
+    },
+    {
+      "epoch": 90.14,
+      "learning_rate": 9.9361249112846e-06,
+      "loss": 0.2042,
+      "step": 12710
+    },
+    {
+      "epoch": 90.21,
+      "learning_rate": 9.865152590489709e-06,
+      "loss": 0.2304,
+      "step": 12720
+    },
+    {
+      "epoch": 90.28,
+      "learning_rate": 9.79418026969482e-06,
+      "loss": 0.3333,
+      "step": 12730
+    },
+    {
+      "epoch": 90.35,
+      "learning_rate": 9.72320794889993e-06,
+      "loss": 0.2067,
+      "step": 12740
+    },
+    {
+      "epoch": 90.43,
+      "learning_rate": 9.65223562810504e-06,
+      "loss": 0.1792,
+      "step": 12750
+    },
+    {
+      "epoch": 90.5,
+      "learning_rate": 9.58126330731015e-06,
+      "loss": 0.2282,
+      "step": 12760
+    },
+    {
+      "epoch": 90.57,
+      "learning_rate": 9.510290986515259e-06,
+      "loss": 0.2722,
+      "step": 12770
+    },
+    {
+      "epoch": 90.64,
+      "learning_rate": 9.439318665720368e-06,
+      "loss": 0.1991,
+      "step": 12780
+    },
+    {
+      "epoch": 90.71,
+      "learning_rate": 9.36834634492548e-06,
+      "loss": 0.2546,
+      "step": 12790
+    },
+    {
+      "epoch": 90.78,
+      "learning_rate": 9.29737402413059e-06,
+      "loss": 0.236,
+      "step": 12800
+    },
+    {
+      "epoch": 90.85,
+      "learning_rate": 9.2264017033357e-06,
+      "loss": 0.237,
+      "step": 12810
+    },
+    {
+      "epoch": 90.92,
+      "learning_rate": 9.155429382540809e-06,
+      "loss": 0.3315,
+      "step": 12820
+    },
+    {
+      "epoch": 90.99,
+      "learning_rate": 9.08445706174592e-06,
+      "loss": 0.353,
+      "step": 12830
+    },
+    {
+      "epoch": 91.0,
+      "eval_loss": 0.6145309805870056,
+      "eval_runtime": 8.6029,
+      "eval_samples_per_second": 7.323,
+      "eval_steps_per_second": 3.72,
+      "eval_wer": 0.32349219738506957,
+      "step": 12831
+    },
+    {
+      "epoch": 91.06,
+      "learning_rate": 9.01348474095103e-06,
+      "loss": 0.2944,
+      "step": 12840
+    },
+    {
+      "epoch": 91.13,
+      "learning_rate": 8.942512420156139e-06,
+      "loss": 0.133,
+      "step": 12850
+    },
+    {
+      "epoch": 91.21,
+      "learning_rate": 8.87154009936125e-06,
+      "loss": 0.2895,
+      "step": 12860
+    },
+    {
+      "epoch": 91.28,
+      "learning_rate": 8.800567778566359e-06,
+      "loss": 0.3509,
+      "step": 12870
+    },
+    {
+      "epoch": 91.35,
+      "learning_rate": 8.72959545777147e-06,
+      "loss": 0.1967,
+      "step": 12880
+    },
+    {
+      "epoch": 91.42,
+      "learning_rate": 8.65862313697658e-06,
+      "loss": 0.2549,
+      "step": 12890
+    },
+    {
+      "epoch": 91.49,
+      "learning_rate": 8.58765081618169e-06,
+      "loss": 0.2454,
+      "step": 12900
+    },
+    {
+      "epoch": 91.56,
+      "learning_rate": 8.5166784953868e-06,
+      "loss": 0.3575,
+      "step": 12910
+    },
+    {
+      "epoch": 91.63,
+      "learning_rate": 8.44570617459191e-06,
+      "loss": 0.1647,
+      "step": 12920
+    },
+    {
+      "epoch": 91.7,
+      "learning_rate": 8.374733853797019e-06,
+      "loss": 0.2122,
+      "step": 12930
+    },
+    {
+      "epoch": 91.77,
+      "learning_rate": 8.30376153300213e-06,
+      "loss": 0.2393,
+      "step": 12940
+    },
+    {
+      "epoch": 91.84,
+      "learning_rate": 8.232789212207239e-06,
+      "loss": 0.2655,
+      "step": 12950
+    },
+    {
+      "epoch": 91.91,
+      "learning_rate": 8.16181689141235e-06,
+      "loss": 0.2559,
+      "step": 12960
+    },
+    {
+      "epoch": 91.99,
+      "learning_rate": 8.09084457061746e-06,
+      "loss": 0.2491,
+      "step": 12970
+    },
+    {
+      "epoch": 92.0,
+      "eval_loss": 0.6295854449272156,
+      "eval_runtime": 8.4367,
+      "eval_samples_per_second": 7.467,
+      "eval_steps_per_second": 3.793,
+      "eval_wer": 0.31969633066216785,
+      "step": 12972
+    },
+    {
+      "epoch": 92.06,
+      "learning_rate": 8.01987224982257e-06,
+      "loss": 0.3154,
+      "step": 12980
+    },
+    {
+      "epoch": 92.13,
+      "learning_rate": 7.94889992902768e-06,
+      "loss": 0.2801,
+      "step": 12990
+    },
+    {
+      "epoch": 92.2,
+      "learning_rate": 7.877927608232789e-06,
+      "loss": 0.2368,
+      "step": 13000
+    },
+    {
+      "epoch": 92.27,
+      "learning_rate": 7.806955287437898e-06,
+      "loss": 0.2279,
+      "step": 13010
+    },
+    {
+      "epoch": 92.34,
+      "learning_rate": 7.73598296664301e-06,
+      "loss": 0.1701,
+      "step": 13020
+    },
+    {
+      "epoch": 92.41,
+      "learning_rate": 7.66501064584812e-06,
+      "loss": 0.2079,
+      "step": 13030
+    },
+    {
+      "epoch": 92.48,
+      "learning_rate": 7.59403832505323e-06,
+      "loss": 0.2345,
+      "step": 13040
+    },
+    {
+      "epoch": 92.55,
+      "learning_rate": 7.523066004258339e-06,
+      "loss": 0.3595,
+      "step": 13050
+    },
+    {
+      "epoch": 92.62,
+      "learning_rate": 7.45209368346345e-06,
+      "loss": 0.2368,
+      "step": 13060
+    },
+    {
+      "epoch": 92.7,
+      "learning_rate": 7.3811213626685594e-06,
+      "loss": 0.2925,
+      "step": 13070
+    },
+    {
+      "epoch": 92.77,
+      "learning_rate": 7.31014904187367e-06,
+      "loss": 0.2867,
+      "step": 13080
+    },
+    {
+      "epoch": 92.84,
+      "learning_rate": 7.239176721078781e-06,
+      "loss": 0.2769,
+      "step": 13090
+    },
+    {
+      "epoch": 92.91,
+      "learning_rate": 7.16820440028389e-06,
+      "loss": 0.1505,
+      "step": 13100
+    },
+    {
+      "epoch": 92.98,
+      "learning_rate": 7.097232079488999e-06,
+      "loss": 0.1999,
+      "step": 13110
+    },
+    {
+      "epoch": 93.0,
+      "eval_loss": 0.632934033870697,
+      "eval_runtime": 8.4074,
+      "eval_samples_per_second": 7.493,
+      "eval_steps_per_second": 3.806,
+      "eval_wer": 0.3222269084774357,
+      "step": 13113
+    },
+    {
+      "epoch": 93.05,
+      "learning_rate": 7.0262597586941095e-06,
+      "loss": 0.2301,
+      "step": 13120
+    },
+    {
+      "epoch": 93.12,
+      "learning_rate": 6.9552874378992205e-06,
+      "loss": 0.2689,
+      "step": 13130
+    },
+    {
+      "epoch": 93.19,
+      "learning_rate": 6.88431511710433e-06,
+      "loss": 0.2938,
+      "step": 13140
+    },
+    {
+      "epoch": 93.26,
+      "learning_rate": 6.813342796309439e-06,
+      "loss": 0.2878,
+      "step": 13150
+    },
+    {
+      "epoch": 93.33,
+      "learning_rate": 6.742370475514549e-06,
+      "loss": 0.234,
+      "step": 13160
+    },
+    {
+      "epoch": 93.4,
+      "learning_rate": 6.6713981547196604e-06,
+      "loss": 0.2317,
+      "step": 13170
+    },
+    {
+      "epoch": 93.48,
+      "learning_rate": 6.60042583392477e-06,
+      "loss": 0.3073,
+      "step": 13180
+    },
+    {
+      "epoch": 93.55,
+      "learning_rate": 6.529453513129879e-06,
+      "loss": 0.405,
+      "step": 13190
+    },
+    {
+      "epoch": 93.62,
+      "learning_rate": 6.458481192334989e-06,
+      "loss": 0.1661,
+      "step": 13200
+    },
+    {
+      "epoch": 93.69,
+      "learning_rate": 6.3875088715401e-06,
+      "loss": 0.2454,
+      "step": 13210
+    },
+    {
+      "epoch": 93.76,
+      "learning_rate": 6.31653655074521e-06,
+      "loss": 0.2061,
+      "step": 13220
+    },
+    {
+      "epoch": 93.83,
+      "learning_rate": 6.24556422995032e-06,
+      "loss": 0.2289,
+      "step": 13230
+    },
+    {
+      "epoch": 93.9,
+      "learning_rate": 6.17459190915543e-06,
+      "loss": 0.166,
+      "step": 13240
+    },
+    {
+      "epoch": 93.97,
+      "learning_rate": 6.103619588360539e-06,
+      "loss": 0.2417,
+      "step": 13250
+    },
+    {
+      "epoch": 94.0,
+      "eval_loss": 0.620030403137207,
+      "eval_runtime": 8.4465,
+      "eval_samples_per_second": 7.459,
+      "eval_steps_per_second": 3.789,
+      "eval_wer": 0.3222269084774357,
+      "step": 13254
+    },
+    {
+      "epoch": 94.04,
+      "learning_rate": 6.0326472675656496e-06,
+      "loss": 0.2652,
+      "step": 13260
+    },
+    {
+      "epoch": 94.11,
+      "learning_rate": 5.96167494677076e-06,
+      "loss": 0.4055,
+      "step": 13270
+    },
+    {
+      "epoch": 94.18,
+      "learning_rate": 5.89070262597587e-06,
+      "loss": 0.2285,
+      "step": 13280
+    },
+    {
+      "epoch": 94.26,
+      "learning_rate": 5.819730305180979e-06,
+      "loss": 0.2193,
+      "step": 13290
+    },
+    {
+      "epoch": 94.33,
+      "learning_rate": 5.7487579843860894e-06,
+      "loss": 0.2314,
+      "step": 13300
+    },
+    {
+      "epoch": 94.4,
+      "learning_rate": 5.6777856635912e-06,
+      "loss": 0.2313,
+      "step": 13310
+    },
+    {
+      "epoch": 94.47,
+      "learning_rate": 5.60681334279631e-06,
+      "loss": 0.2997,
+      "step": 13320
+    },
+    {
+      "epoch": 94.54,
+      "learning_rate": 5.535841022001419e-06,
+      "loss": 0.2441,
+      "step": 13330
+    },
+    {
+      "epoch": 94.61,
+      "learning_rate": 5.464868701206529e-06,
+      "loss": 0.2478,
+      "step": 13340
+    },
+    {
+      "epoch": 94.68,
+      "learning_rate": 5.39389638041164e-06,
+      "loss": 0.2528,
+      "step": 13350
+    },
+    {
+      "epoch": 94.75,
+      "learning_rate": 5.32292405961675e-06,
+      "loss": 0.1681,
+      "step": 13360
+    },
+    {
+      "epoch": 94.82,
+      "learning_rate": 5.25195173882186e-06,
+      "loss": 0.276,
+      "step": 13370
+    },
+    {
+      "epoch": 94.89,
+      "learning_rate": 5.18097941802697e-06,
+      "loss": 0.1868,
+      "step": 13380
+    },
+    {
+      "epoch": 94.96,
+      "learning_rate": 5.11000709723208e-06,
+      "loss": 0.2397,
+      "step": 13390
+    },
+    {
+      "epoch": 95.0,
+      "eval_loss": 0.6136645078659058,
+      "eval_runtime": 8.4936,
+      "eval_samples_per_second": 7.417,
+      "eval_steps_per_second": 3.768,
+      "eval_wer": 0.32686630113876003,
+      "step": 13395
+    },
+    {
+      "epoch": 95.04,
+      "learning_rate": 5.03903477643719e-06,
+      "loss": 0.1763,
+      "step": 13400
+    },
+    {
+      "epoch": 95.11,
+      "learning_rate": 4.9680624556423e-06,
+      "loss": 0.242,
+      "step": 13410
+    },
+    {
+      "epoch": 95.18,
+      "learning_rate": 4.89709013484741e-06,
+      "loss": 0.1797,
+      "step": 13420
+    },
+    {
+      "epoch": 95.25,
+      "learning_rate": 4.82611781405252e-06,
+      "loss": 0.2003,
+      "step": 13430
+    },
+    {
+      "epoch": 95.32,
+      "learning_rate": 4.7551454932576295e-06,
+      "loss": 0.2972,
+      "step": 13440
+    },
+    {
+      "epoch": 95.39,
+      "learning_rate": 4.68417317246274e-06,
+      "loss": 0.2214,
+      "step": 13450
+    },
+    {
+      "epoch": 95.46,
+      "learning_rate": 4.61320085166785e-06,
+      "loss": 0.2666,
+      "step": 13460
+    },
+    {
+      "epoch": 95.53,
+      "learning_rate": 4.54222853087296e-06,
+      "loss": 0.264,
+      "step": 13470
+    },
+    {
+      "epoch": 95.6,
+      "learning_rate": 4.471256210078069e-06,
+      "loss": 0.2361,
+      "step": 13480
+    },
+    {
+      "epoch": 95.67,
+      "learning_rate": 4.4002838892831796e-06,
+      "loss": 0.2482,
+      "step": 13490
+    },
+    {
+      "epoch": 95.74,
+      "learning_rate": 4.32931156848829e-06,
+      "loss": 0.3212,
+      "step": 13500
+    },
+    {
+      "epoch": 95.82,
+      "learning_rate": 4.2583392476934e-06,
+      "loss": 0.3053,
+      "step": 13510
+    },
+    {
+      "epoch": 95.89,
+      "learning_rate": 4.187366926898509e-06,
+      "loss": 0.1523,
+      "step": 13520
+    },
+    {
+      "epoch": 95.96,
+      "learning_rate": 4.1163946061036195e-06,
+      "loss": 0.2275,
+      "step": 13530
+    },
+    {
+      "epoch": 96.0,
+      "eval_loss": 0.6236584186553955,
+      "eval_runtime": 8.4452,
+      "eval_samples_per_second": 7.46,
+      "eval_steps_per_second": 3.789,
+      "eval_wer": 0.3277098270771826,
+      "step": 13536
+    },
+    {
+      "epoch": 96.03,
+      "learning_rate": 4.04542228530873e-06,
+      "loss": 0.4982,
+      "step": 13540
+    },
+    {
+      "epoch": 96.1,
+      "learning_rate": 3.97444996451384e-06,
+      "loss": 0.2927,
+      "step": 13550
+    },
+    {
+      "epoch": 96.17,
+      "learning_rate": 3.903477643718949e-06,
+      "loss": 0.223,
+      "step": 13560
+    },
+    {
+      "epoch": 96.24,
+      "learning_rate": 3.83250532292406e-06,
+      "loss": 0.4071,
+      "step": 13570
+    },
+    {
+      "epoch": 96.31,
+      "learning_rate": 3.7615330021291695e-06,
+      "loss": 0.295,
+      "step": 13580
+    },
+    {
+      "epoch": 96.38,
+      "learning_rate": 3.6905606813342797e-06,
+      "loss": 0.2015,
+      "step": 13590
+    },
+    {
+      "epoch": 96.45,
+      "learning_rate": 3.6195883605393903e-06,
+      "loss": 0.2517,
+      "step": 13600
+    },
+    {
+      "epoch": 96.52,
+      "learning_rate": 3.5486160397444997e-06,
+      "loss": 0.2328,
+      "step": 13610
+    },
+    {
+      "epoch": 96.6,
+      "learning_rate": 3.4776437189496103e-06,
+      "loss": 0.1414,
+      "step": 13620
+    },
+    {
+      "epoch": 96.67,
+      "learning_rate": 3.4066713981547196e-06,
+      "loss": 0.1819,
+      "step": 13630
+    },
+    {
+      "epoch": 96.74,
+      "learning_rate": 3.3356990773598302e-06,
+      "loss": 0.2394,
+      "step": 13640
+    },
+    {
+      "epoch": 96.81,
+      "learning_rate": 3.2647267565649396e-06,
+      "loss": 0.2972,
+      "step": 13650
+    },
+    {
+      "epoch": 96.88,
+      "learning_rate": 3.19375443577005e-06,
+      "loss": 0.1811,
+      "step": 13660
+    },
+    {
+      "epoch": 96.95,
+      "learning_rate": 3.12278211497516e-06,
+      "loss": 0.207,
+      "step": 13670
+    },
+    {
+      "epoch": 97.0,
+      "eval_loss": 0.6229757070541382,
+      "eval_runtime": 8.4,
+      "eval_samples_per_second": 7.5,
+      "eval_steps_per_second": 3.81,
+      "eval_wer": 0.32349219738506957,
+      "step": 13677
+    },
+    {
+      "epoch": 97.02,
+      "learning_rate": 3.0518097941802697e-06,
+      "loss": 0.4195,
+      "step": 13680
+    },
+    {
+      "epoch": 97.09,
+      "learning_rate": 2.98083747338538e-06,
+      "loss": 0.3336,
+      "step": 13690
+    },
+    {
+      "epoch": 97.16,
+      "learning_rate": 2.9098651525904896e-06,
+      "loss": 0.2007,
+      "step": 13700
+    },
+    {
+      "epoch": 97.23,
+      "learning_rate": 2.8388928317956e-06,
+      "loss": 0.2853,
+      "step": 13710
+    },
+    {
+      "epoch": 97.3,
+      "learning_rate": 2.7679205110007096e-06,
+      "loss": 0.3197,
+      "step": 13720
+    },
+    {
+      "epoch": 97.38,
+      "learning_rate": 2.69694819020582e-06,
+      "loss": 0.1906,
+      "step": 13730
+    },
+    {
+      "epoch": 97.45,
+      "learning_rate": 2.62597586941093e-06,
+      "loss": 0.1704,
+      "step": 13740
+    },
+    {
+      "epoch": 97.52,
+      "learning_rate": 2.55500354861604e-06,
+      "loss": 0.2465,
+      "step": 13750
+    },
+    {
+      "epoch": 97.59,
+      "learning_rate": 2.48403122782115e-06,
+      "loss": 0.2812,
+      "step": 13760
+    },
+    {
+      "epoch": 97.66,
+      "learning_rate": 2.41305890702626e-06,
+      "loss": 0.2217,
+      "step": 13770
+    },
+    {
+      "epoch": 97.73,
+      "learning_rate": 2.34208658623137e-06,
+      "loss": 0.2223,
+      "step": 13780
+    },
+    {
+      "epoch": 97.8,
+      "learning_rate": 2.27111426543648e-06,
+      "loss": 0.2453,
+      "step": 13790
+    },
+    {
+      "epoch": 97.87,
+      "learning_rate": 2.2001419446415898e-06,
+      "loss": 0.2094,
+      "step": 13800
+    },
+    {
+      "epoch": 97.94,
+      "learning_rate": 2.1291696238467e-06,
+      "loss": 0.2704,
+      "step": 13810
+    },
+    {
+      "epoch": 98.0,
+      "eval_loss": 0.6239410638809204,
+      "eval_runtime": 8.4443,
+      "eval_samples_per_second": 7.461,
+      "eval_steps_per_second": 3.79,
+      "eval_wer": 0.3281315900463939,
+      "step": 13818
+    },
+    {
+      "epoch": 98.01,
+      "learning_rate": 2.0581973030518097e-06,
+      "loss": 0.2622,
+      "step": 13820
+    },
+    {
+      "epoch": 98.09,
+      "learning_rate": 1.98722498225692e-06,
+      "loss": 0.3057,
+      "step": 13830
+    },
+    {
+      "epoch": 98.16,
+      "learning_rate": 1.91625266146203e-06,
+      "loss": 0.2091,
+      "step": 13840
+    },
+    {
+      "epoch": 98.23,
+      "learning_rate": 1.8452803406671399e-06,
+      "loss": 0.2059,
+      "step": 13850
+    },
+    {
+      "epoch": 98.3,
+      "learning_rate": 1.7743080198722498e-06,
+      "loss": 0.3139,
+      "step": 13860
+    },
+    {
+      "epoch": 98.37,
+      "learning_rate": 1.7033356990773598e-06,
+      "loss": 0.2163,
+      "step": 13870
+    },
+    {
+      "epoch": 98.44,
+      "learning_rate": 1.6323633782824698e-06,
+      "loss": 0.2711,
+      "step": 13880
+    },
+    {
+      "epoch": 98.51,
+      "learning_rate": 1.56139105748758e-06,
+      "loss": 0.2914,
+      "step": 13890
+    },
+    {
+      "epoch": 98.58,
+      "learning_rate": 1.49041873669269e-06,
+      "loss": 0.2942,
+      "step": 13900
+    },
+    {
+      "epoch": 98.65,
+      "learning_rate": 1.4194464158978e-06,
+      "loss": 0.2146,
+      "step": 13910
+    },
+    {
+      "epoch": 98.72,
+      "learning_rate": 1.34847409510291e-06,
+      "loss": 0.2038,
+      "step": 13920
+    },
+    {
+      "epoch": 98.79,
+      "learning_rate": 1.27750177430802e-06,
+      "loss": 0.2526,
+      "step": 13930
+    },
+    {
+      "epoch": 98.87,
+      "learning_rate": 1.20652945351313e-06,
+      "loss": 0.1591,
+      "step": 13940
+    },
+    {
+      "epoch": 98.94,
+      "learning_rate": 1.13555713271824e-06,
+      "loss": 0.2119,
+      "step": 13950
+    },
+    {
+      "epoch": 99.0,
+      "eval_loss": 0.6223601698875427,
+      "eval_runtime": 8.4755,
+      "eval_samples_per_second": 7.433,
+      "eval_steps_per_second": 3.776,
+      "eval_wer": 0.3277098270771826,
+      "step": 13959
+    },
+    {
+      "epoch": 99.01,
+      "learning_rate": 1.06458481192335e-06,
+      "loss": 0.2305,
+      "step": 13960
+    },
+    {
+      "epoch": 99.08,
+      "learning_rate": 9.9361249112846e-07,
+      "loss": 0.3098,
+      "step": 13970
+    },
+    {
+      "epoch": 99.15,
+      "learning_rate": 9.226401703335699e-07,
+      "loss": 0.4093,
+      "step": 13980
+    },
+    {
+      "epoch": 99.22,
+      "learning_rate": 8.516678495386799e-07,
+      "loss": 0.1856,
+      "step": 13990
+    },
+    {
+      "epoch": 99.29,
+      "learning_rate": 7.8069552874379e-07,
+      "loss": 0.4261,
+      "step": 14000
+    },
+    {
+      "epoch": 99.36,
+      "learning_rate": 7.097232079489e-07,
+      "loss": 0.1957,
+      "step": 14010
+    },
+    {
+      "epoch": 99.43,
+      "learning_rate": 6.3875088715401e-07,
+      "loss": 0.2591,
+      "step": 14020
+    },
+    {
+      "epoch": 99.5,
+      "learning_rate": 5.6777856635912e-07,
+      "loss": 0.3213,
+      "step": 14030
+    },
+    {
+      "epoch": 99.57,
+      "learning_rate": 4.9680624556423e-07,
+      "loss": 0.2394,
+      "step": 14040
+    },
+    {
+      "epoch": 99.65,
+      "learning_rate": 4.2583392476933995e-07,
+      "loss": 0.2385,
+      "step": 14050
+    },
+    {
+      "epoch": 99.72,
+      "learning_rate": 3.5486160397445e-07,
+      "loss": 0.2478,
+      "step": 14060
+    },
+    {
+      "epoch": 99.79,
+      "learning_rate": 2.8388928317956e-07,
+      "loss": 0.2611,
+      "step": 14070
+    },
+    {
+      "epoch": 99.86,
+      "learning_rate": 2.1291696238466998e-07,
+      "loss": 0.2643,
+      "step": 14080
+    },
+    {
+      "epoch": 99.93,
+      "learning_rate": 1.4194464158978e-07,
+      "loss": 0.233,
+      "step": 14090
+    },
+    {
+      "epoch": 100.0,
+      "learning_rate": 7.097232079489e-08,
+      "loss": 0.2561,
+      "step": 14100
+    },
+    {
+      "epoch": 100.0,
+      "eval_loss": 0.6187006831169128,
+      "eval_runtime": 8.521,
+      "eval_samples_per_second": 7.393,
+      "eval_steps_per_second": 3.755,
+      "eval_wer": 0.32686630113876003,
+      "step": 14100
     },
     {
-      "epoch": 99.99,
-      "step": 3500,
-      "total_flos": 5.739635222597828e+18,
-      "train_loss": 0.9362126126289367,
-      "train_runtime": 11890.6212,
-      "train_samples_per_second": 4.735,
-      "train_steps_per_second": 0.294
+      "epoch": 100.0,
+      "step": 14100,
+      "total_flos": 5.648331054745915e+18,
+      "train_loss": 0.30988129563365424,
+      "train_runtime": 12865.5249,
+      "train_samples_per_second": 4.376,
+      "train_steps_per_second": 1.096
     }
   ],
-  "max_steps": 3500,
+  "max_steps": 14100,
   "num_train_epochs": 100,
-  "total_flos": 5.739635222597828e+18,
+  "total_flos": 5.648331054745915e+18,
   "trial_name": null,
   "trial_params": null
 }