diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,14314 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 14.999946469675072,
+  "global_step": 140100,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "eval_loss": 3.625196695327759,
+      "eval_runtime": 48.6485,
+      "eval_samples_per_second": 41.666,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.02,
+      "eval_loss": 3.190624713897705,
+      "eval_runtime": 48.8326,
+      "eval_samples_per_second": 41.509,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.03,
+      "eval_loss": 3.108981132507324,
+      "eval_runtime": 48.8211,
+      "eval_samples_per_second": 41.519,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 1.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.04,
+      "eval_loss": 1.8795998096466064,
+      "eval_runtime": 48.7588,
+      "eval_samples_per_second": 41.572,
+      "eval_steps_per_second": 5.209,
+      "eval_wer": 0.9954865499187578,
+      "step": 400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.00029759999999999997,
+      "loss": 6.2575,
+      "step": 500
+    },
+    {
+      "epoch": 0.05,
+      "eval_loss": 1.3514989614486694,
+      "eval_runtime": 48.9732,
+      "eval_samples_per_second": 41.39,
+      "eval_steps_per_second": 5.187,
+      "eval_wer": 0.9057591623036649,
+      "step": 500
+    },
+    {
+      "epoch": 0.06,
+      "eval_loss": 1.1209492683410645,
+      "eval_runtime": 48.2472,
+      "eval_samples_per_second": 42.013,
+      "eval_steps_per_second": 5.265,
+      "eval_wer": 0.8328218089907926,
+      "step": 600
+    },
+    {
+      "epoch": 0.07,
+      "eval_loss": 1.1404383182525635,
+      "eval_runtime": 48.9274,
+      "eval_samples_per_second": 41.429,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.8309261599566709,
+      "step": 700
+    },
+    {
+      "epoch": 0.09,
+      "eval_loss": 1.0599348545074463,
+      "eval_runtime": 48.4506,
+      "eval_samples_per_second": 41.836,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.8021303484383463,
+      "step": 800
+    },
+    {
+      "epoch": 0.1,
+      "eval_loss": 0.9901070594787598,
+      "eval_runtime": 48.6803,
+      "eval_samples_per_second": 41.639,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.8335439610037912,
+      "step": 900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0002989340974212034,
+      "loss": 0.7737,
+      "step": 1000
+    },
+    {
+      "epoch": 0.11,
+      "eval_loss": 0.884602963924408,
+      "eval_runtime": 48.4241,
+      "eval_samples_per_second": 41.859,
+      "eval_steps_per_second": 5.245,
+      "eval_wer": 0.7400252753204549,
+      "step": 1000
+    },
+    {
+      "epoch": 0.12,
+      "eval_loss": 0.9971189498901367,
+      "eval_runtime": 48.5998,
+      "eval_samples_per_second": 41.708,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.7820003610760065,
+      "step": 1100
+    },
+    {
+      "epoch": 0.13,
+      "eval_loss": 0.8665233850479126,
+      "eval_runtime": 48.3232,
+      "eval_samples_per_second": 41.947,
+      "eval_steps_per_second": 5.256,
+      "eval_wer": 0.7123126918216285,
+      "step": 1200
+    },
+    {
+      "epoch": 0.14,
+      "eval_loss": 0.8489500284194946,
+      "eval_runtime": 48.8631,
+      "eval_samples_per_second": 41.483,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.736595053258711,
+      "step": 1300
+    },
+    {
+      "epoch": 0.15,
+      "eval_loss": 0.8249963521957397,
+      "eval_runtime": 48.6019,
+      "eval_samples_per_second": 41.706,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.6764758981765662,
+      "step": 1400
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.00029785959885386817,
+      "loss": 0.6183,
+      "step": 1500
+    },
+    {
+      "epoch": 0.16,
+      "eval_loss": 0.8290594220161438,
+      "eval_runtime": 48.6721,
+      "eval_samples_per_second": 41.646,
+      "eval_steps_per_second": 5.219,
+      "eval_wer": 0.6965156165372811,
+      "step": 1500
+    },
+    {
+      "epoch": 0.17,
+      "eval_loss": 0.7945746779441833,
+      "eval_runtime": 48.366,
+      "eval_samples_per_second": 41.91,
+      "eval_steps_per_second": 5.252,
+      "eval_wer": 0.682253114280556,
+      "step": 1600
+    },
+    {
+      "epoch": 0.18,
+      "eval_loss": 0.8239201903343201,
+      "eval_runtime": 49.119,
+      "eval_samples_per_second": 41.267,
+      "eval_steps_per_second": 5.171,
+      "eval_wer": 0.6893843654089186,
+      "step": 1700
+    },
+    {
+      "epoch": 0.19,
+      "eval_loss": 0.828230082988739,
+      "eval_runtime": 48.821,
+      "eval_samples_per_second": 41.519,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.6796353132334356,
+      "step": 1800
+    },
+    {
+      "epoch": 0.2,
+      "eval_loss": 0.764546275138855,
+      "eval_runtime": 49.1104,
+      "eval_samples_per_second": 41.274,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.6518324607329843,
+      "step": 1900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 0.0002967851002865329,
+      "loss": 0.561,
+      "step": 2000
+    },
+    {
+      "epoch": 0.21,
+      "eval_loss": 0.7530480623245239,
+      "eval_runtime": 48.5997,
+      "eval_samples_per_second": 41.708,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.6366672684600109,
+      "step": 2000
+    },
+    {
+      "epoch": 0.22,
+      "eval_loss": 0.7295656800270081,
+      "eval_runtime": 48.6615,
+      "eval_samples_per_second": 41.655,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.617710778118794,
+      "step": 2100
+    },
+    {
+      "epoch": 0.24,
+      "eval_loss": 0.7527036070823669,
+      "eval_runtime": 48.4283,
+      "eval_samples_per_second": 41.856,
+      "eval_steps_per_second": 5.245,
+      "eval_wer": 0.6498465426972377,
+      "step": 2200
+    },
+    {
+      "epoch": 0.25,
+      "eval_loss": 0.7209727764129639,
+      "eval_runtime": 48.8209,
+      "eval_samples_per_second": 41.519,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.6316122043690197,
+      "step": 2300
+    },
+    {
+      "epoch": 0.26,
+      "eval_loss": 0.7938166856765747,
+      "eval_runtime": 48.2216,
+      "eval_samples_per_second": 42.035,
+      "eval_steps_per_second": 5.267,
+      "eval_wer": 0.6756634771619426,
+      "step": 2400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 0.00029571060171919766,
+      "loss": 0.5402,
+      "step": 2500
+    },
+    {
+      "epoch": 0.27,
+      "eval_loss": 0.7485235929489136,
+      "eval_runtime": 48.3609,
+      "eval_samples_per_second": 41.914,
+      "eval_steps_per_second": 5.252,
+      "eval_wer": 0.6372088824697599,
+      "step": 2500
+    },
+    {
+      "epoch": 0.28,
+      "eval_loss": 0.7146492004394531,
+      "eval_runtime": 48.3522,
+      "eval_samples_per_second": 41.922,
+      "eval_steps_per_second": 5.253,
+      "eval_wer": 0.6132875970391768,
+      "step": 2600
+    },
+    {
+      "epoch": 0.29,
+      "eval_loss": 0.7308168411254883,
+      "eval_runtime": 48.4029,
+      "eval_samples_per_second": 41.878,
+      "eval_steps_per_second": 5.248,
+      "eval_wer": 0.6625744719263404,
+      "step": 2700
+    },
+    {
+      "epoch": 0.3,
+      "eval_loss": 0.7078224420547485,
+      "eval_runtime": 48.7403,
+      "eval_samples_per_second": 41.588,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.594872720707709,
+      "step": 2800
+    },
+    {
+      "epoch": 0.31,
+      "eval_loss": 0.7679122090339661,
+      "eval_runtime": 48.4081,
+      "eval_samples_per_second": 41.873,
+      "eval_steps_per_second": 5.247,
+      "eval_wer": 0.6372991514713847,
+      "step": 2900
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 0.0002946361031518624,
+      "loss": 0.5303,
+      "step": 3000
+    },
+    {
+      "epoch": 0.32,
+      "eval_loss": 0.72626131772995,
+      "eval_runtime": 48.7452,
+      "eval_samples_per_second": 41.584,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.6502076187037371,
+      "step": 3000
+    },
+    {
+      "epoch": 0.33,
+      "eval_loss": 0.6612817049026489,
+      "eval_runtime": 48.9995,
+      "eval_samples_per_second": 41.368,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.584582054522477,
+      "step": 3100
+    },
+    {
+      "epoch": 0.34,
+      "eval_loss": 0.6783962249755859,
+      "eval_runtime": 48.6023,
+      "eval_samples_per_second": 41.706,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.578263224408738,
+      "step": 3200
+    },
+    {
+      "epoch": 0.35,
+      "eval_loss": 0.6908129453659058,
+      "eval_runtime": 48.5922,
+      "eval_samples_per_second": 41.715,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.5833182884997292,
+      "step": 3300
+    },
+    {
+      "epoch": 0.36,
+      "eval_loss": 0.6595033407211304,
+      "eval_runtime": 48.473,
+      "eval_samples_per_second": 41.817,
+      "eval_steps_per_second": 5.24,
+      "eval_wer": 0.5825961364867305,
+      "step": 3400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0002935616045845272,
+      "loss": 0.503,
+      "step": 3500
+    },
+    {
+      "epoch": 0.37,
+      "eval_loss": 0.6716800928115845,
+      "eval_runtime": 48.6209,
+      "eval_samples_per_second": 41.69,
+      "eval_steps_per_second": 5.224,
+      "eval_wer": 0.5937894926882109,
+      "step": 3500
+    },
+    {
+      "epoch": 0.39,
+      "eval_loss": 0.6937839984893799,
+      "eval_runtime": 48.9232,
+      "eval_samples_per_second": 41.432,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.5790756454233617,
+      "step": 3600
+    },
+    {
+      "epoch": 0.4,
+      "eval_loss": 0.667679488658905,
+      "eval_runtime": 48.3286,
+      "eval_samples_per_second": 41.942,
+      "eval_steps_per_second": 5.256,
+      "eval_wer": 0.6051633868929409,
+      "step": 3700
+    },
+    {
+      "epoch": 0.41,
+      "eval_loss": 0.65436190366745,
+      "eval_runtime": 48.7685,
+      "eval_samples_per_second": 41.564,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.555425166997653,
+      "step": 3800
+    },
+    {
+      "epoch": 0.42,
+      "eval_loss": 0.6514427661895752,
+      "eval_runtime": 48.532,
+      "eval_samples_per_second": 41.766,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.5727568153096226,
+      "step": 3900
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00029248710601719195,
+      "loss": 0.4959,
+      "step": 4000
+    },
+    {
+      "epoch": 0.43,
+      "eval_loss": 0.6847020983695984,
+      "eval_runtime": 48.5364,
+      "eval_samples_per_second": 41.763,
+      "eval_steps_per_second": 5.233,
+      "eval_wer": 0.6187940061382922,
+      "step": 4000
+    },
+    {
+      "epoch": 0.44,
+      "eval_loss": 0.6626438498497009,
+      "eval_runtime": 48.3485,
+      "eval_samples_per_second": 41.925,
+      "eval_steps_per_second": 5.254,
+      "eval_wer": 0.5869290485647228,
+      "step": 4100
+    },
+    {
+      "epoch": 0.45,
+      "eval_loss": 0.6669931411743164,
+      "eval_runtime": 48.5641,
+      "eval_samples_per_second": 41.739,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.5699584762592526,
+      "step": 4200
+    },
+    {
+      "epoch": 0.46,
+      "eval_loss": 0.6595851182937622,
+      "eval_runtime": 48.6164,
+      "eval_samples_per_second": 41.694,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.584582054522477,
+      "step": 4300
+    },
+    {
+      "epoch": 0.47,
+      "eval_loss": 0.6522601246833801,
+      "eval_runtime": 48.5548,
+      "eval_samples_per_second": 41.747,
+      "eval_steps_per_second": 5.231,
+      "eval_wer": 0.5467593428416682,
+      "step": 4400
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.0002914126074498567,
+      "loss": 0.4824,
+      "step": 4500
+    },
+    {
+      "epoch": 0.48,
+      "eval_loss": 0.6391969919204712,
+      "eval_runtime": 48.4793,
+      "eval_samples_per_second": 41.812,
+      "eval_steps_per_second": 5.239,
+      "eval_wer": 0.5687849792381297,
+      "step": 4500
+    },
+    {
+      "epoch": 0.49,
+      "eval_loss": 0.6561304926872253,
+      "eval_runtime": 48.7756,
+      "eval_samples_per_second": 41.558,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.5686947102365048,
+      "step": 4600
+    },
+    {
+      "epoch": 0.5,
+      "eval_loss": 0.6697142124176025,
+      "eval_runtime": 48.9375,
+      "eval_samples_per_second": 41.42,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.581693446470482,
+      "step": 4700
+    },
+    {
+      "epoch": 0.51,
+      "eval_loss": 0.6348330974578857,
+      "eval_runtime": 49.0519,
+      "eval_samples_per_second": 41.324,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.5608413070951436,
+      "step": 4800
+    },
+    {
+      "epoch": 0.52,
+      "eval_loss": 0.656074047088623,
+      "eval_runtime": 48.3379,
+      "eval_samples_per_second": 41.934,
+      "eval_steps_per_second": 5.255,
+      "eval_wer": 0.5600288860805199,
+      "step": 4900
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.00029033810888252145,
+      "loss": 0.4714,
+      "step": 5000
+    },
+    {
+      "epoch": 0.54,
+      "eval_loss": 0.652241051197052,
+      "eval_runtime": 48.5527,
+      "eval_samples_per_second": 41.748,
+      "eval_steps_per_second": 5.231,
+      "eval_wer": 0.6180718541252934,
+      "step": 5000
+    },
+    {
+      "epoch": 0.55,
+      "eval_loss": 0.6857942938804626,
+      "eval_runtime": 48.4567,
+      "eval_samples_per_second": 41.831,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.5920743816573388,
+      "step": 5100
+    },
+    {
+      "epoch": 0.56,
+      "eval_loss": 0.6706011891365051,
+      "eval_runtime": 48.9061,
+      "eval_samples_per_second": 41.447,
+      "eval_steps_per_second": 5.194,
+      "eval_wer": 0.5497382198952879,
+      "step": 5200
+    },
+    {
+      "epoch": 0.57,
+      "eval_loss": 0.7122880220413208,
+      "eval_runtime": 49.0884,
+      "eval_samples_per_second": 41.293,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.5768189203827405,
+      "step": 5300
+    },
+    {
+      "epoch": 0.58,
+      "eval_loss": 0.659943699836731,
+      "eval_runtime": 49.0325,
+      "eval_samples_per_second": 41.34,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.6100379129806824,
+      "step": 5400
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.00028926361031518625,
+      "loss": 0.471,
+      "step": 5500
+    },
+    {
+      "epoch": 0.59,
+      "eval_loss": 0.6421207785606384,
+      "eval_runtime": 48.992,
+      "eval_samples_per_second": 41.374,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.5625564181260155,
+      "step": 5500
+    },
+    {
+      "epoch": 0.6,
+      "eval_loss": 0.6395331621170044,
+      "eval_runtime": 48.7511,
+      "eval_samples_per_second": 41.579,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.5752843473551182,
+      "step": 5600
+    },
+    {
+      "epoch": 0.61,
+      "eval_loss": 0.6788127422332764,
+      "eval_runtime": 48.7567,
+      "eval_samples_per_second": 41.574,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.5481133778660408,
+      "step": 5700
+    },
+    {
+      "epoch": 0.62,
+      "eval_loss": 0.6385824084281921,
+      "eval_runtime": 48.9331,
+      "eval_samples_per_second": 41.424,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.5516338689294097,
+      "step": 5800
+    },
+    {
+      "epoch": 0.63,
+      "eval_loss": 0.6694206595420837,
+      "eval_runtime": 48.8182,
+      "eval_samples_per_second": 41.521,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.5912619606427153,
+      "step": 5900
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.000288189111747851,
+      "loss": 0.4707,
+      "step": 6000
+    },
+    {
+      "epoch": 0.64,
+      "eval_loss": 0.6250674724578857,
+      "eval_runtime": 48.8318,
+      "eval_samples_per_second": 41.51,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.5698682072576278,
+      "step": 6000
+    },
+    {
+      "epoch": 0.65,
+      "eval_loss": 0.6243389844894409,
+      "eval_runtime": 48.6735,
+      "eval_samples_per_second": 41.645,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.5566889330204008,
+      "step": 6100
+    },
+    {
+      "epoch": 0.66,
+      "eval_loss": 0.6645160913467407,
+      "eval_runtime": 48.8686,
+      "eval_samples_per_second": 41.479,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.5629174941325149,
+      "step": 6200
+    },
+    {
+      "epoch": 0.67,
+      "eval_loss": 0.6295756101608276,
+      "eval_runtime": 49.1221,
+      "eval_samples_per_second": 41.265,
+      "eval_steps_per_second": 5.171,
+      "eval_wer": 0.5895468496118433,
+      "step": 6300
+    },
+    {
+      "epoch": 0.69,
+      "eval_loss": 0.6077612042427063,
+      "eval_runtime": 48.8377,
+      "eval_samples_per_second": 41.505,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.518324607329843,
+      "step": 6400
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00028711461318051574,
+      "loss": 0.4632,
+      "step": 6500
+    },
+    {
+      "epoch": 0.7,
+      "eval_loss": 0.6269996762275696,
+      "eval_runtime": 48.7596,
+      "eval_samples_per_second": 41.571,
+      "eval_steps_per_second": 5.209,
+      "eval_wer": 0.5619245351146416,
+      "step": 6500
+    },
+    {
+      "epoch": 0.71,
+      "eval_loss": 0.6050315499305725,
+      "eval_runtime": 49.3099,
+      "eval_samples_per_second": 41.107,
+      "eval_steps_per_second": 5.151,
+      "eval_wer": 0.5335800686044412,
+      "step": 6600
+    },
+    {
+      "epoch": 0.72,
+      "eval_loss": 0.6185252666473389,
+      "eval_runtime": 49.0917,
+      "eval_samples_per_second": 41.29,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.5448636938075465,
+      "step": 6700
+    },
+    {
+      "epoch": 0.73,
+      "eval_loss": 0.6281481385231018,
+      "eval_runtime": 48.533,
+      "eval_samples_per_second": 41.765,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.5645423361617621,
+      "step": 6800
+    },
+    {
+      "epoch": 0.74,
+      "eval_loss": 0.5876803398132324,
+      "eval_runtime": 48.7707,
+      "eval_samples_per_second": 41.562,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.5083950171511104,
+      "step": 6900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.0002860401146131805,
+      "loss": 0.4514,
+      "step": 7000
+    },
+    {
+      "epoch": 0.75,
+      "eval_loss": 0.6199076175689697,
+      "eval_runtime": 49.1573,
+      "eval_samples_per_second": 41.235,
+      "eval_steps_per_second": 5.167,
+      "eval_wer": 0.5402599747246796,
+      "step": 7000
+    },
+    {
+      "epoch": 0.76,
+      "eval_loss": 0.6293303966522217,
+      "eval_runtime": 48.9241,
+      "eval_samples_per_second": 41.432,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.5275320454955769,
+      "step": 7100
+    },
+    {
+      "epoch": 0.77,
+      "eval_loss": 0.6289793848991394,
+      "eval_runtime": 49.1249,
+      "eval_samples_per_second": 41.262,
+      "eval_steps_per_second": 5.17,
+      "eval_wer": 0.5446831558042968,
+      "step": 7200
+    },
+    {
+      "epoch": 0.78,
+      "eval_loss": 0.613017737865448,
+      "eval_runtime": 48.9639,
+      "eval_samples_per_second": 41.398,
+      "eval_steps_per_second": 5.187,
+      "eval_wer": 0.5372810976710598,
+      "step": 7300
+    },
+    {
+      "epoch": 0.79,
+      "eval_loss": 0.6137964129447937,
+      "eval_runtime": 49.0353,
+      "eval_samples_per_second": 41.338,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.5285250045134501,
+      "step": 7400
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00028496561604584523,
+      "loss": 0.4457,
+      "step": 7500
+    },
+    {
+      "epoch": 0.8,
+      "eval_loss": 0.6040118336677551,
+      "eval_runtime": 48.8487,
+      "eval_samples_per_second": 41.495,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.5259072034663297,
+      "step": 7500
+    },
+    {
+      "epoch": 0.81,
+      "eval_loss": 0.622002363204956,
+      "eval_runtime": 48.8408,
+      "eval_samples_per_second": 41.502,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.56860444123488,
+      "step": 7600
+    },
+    {
+      "epoch": 0.82,
+      "eval_loss": 0.5914533734321594,
+      "eval_runtime": 48.6432,
+      "eval_samples_per_second": 41.671,
+      "eval_steps_per_second": 5.222,
+      "eval_wer": 0.5164289582957212,
+      "step": 7700
+    },
+    {
+      "epoch": 0.84,
+      "eval_loss": 0.6269855499267578,
+      "eval_runtime": 48.6798,
+      "eval_samples_per_second": 41.639,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.5288860805199495,
+      "step": 7800
+    },
+    {
+      "epoch": 0.85,
+      "eval_loss": 0.6224400997161865,
+      "eval_runtime": 48.9678,
+      "eval_samples_per_second": 41.395,
+      "eval_steps_per_second": 5.187,
+      "eval_wer": 0.55145333092616,
+      "step": 7900
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00028389111747851003,
+      "loss": 0.4458,
+      "step": 8000
+    },
+    {
+      "epoch": 0.86,
+      "eval_loss": 0.6161192059516907,
+      "eval_runtime": 48.8802,
+      "eval_samples_per_second": 41.469,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.5323163025816935,
+      "step": 8000
+    },
+    {
+      "epoch": 0.87,
+      "eval_loss": 0.5826652646064758,
+      "eval_runtime": 48.8638,
+      "eval_samples_per_second": 41.483,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.5121863152193536,
+      "step": 8100
+    },
+    {
+      "epoch": 0.88,
+      "eval_loss": 0.6067497730255127,
+      "eval_runtime": 48.7237,
+      "eval_samples_per_second": 41.602,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.5202202563639646,
+      "step": 8200
+    },
+    {
+      "epoch": 0.89,
+      "eval_loss": 0.6086713671684265,
+      "eval_runtime": 49.0772,
+      "eval_samples_per_second": 41.302,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.5192272973460913,
+      "step": 8300
+    },
+    {
+      "epoch": 0.9,
+      "eval_loss": 0.6859478950500488,
+      "eval_runtime": 48.6001,
+      "eval_samples_per_second": 41.708,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.5796172594331107,
+      "step": 8400
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.0002828187679083094,
+      "loss": 0.4409,
+      "step": 8500
+    },
+    {
+      "epoch": 0.91,
+      "eval_loss": 0.6179741621017456,
+      "eval_runtime": 48.6105,
+      "eval_samples_per_second": 41.699,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.5130890052356021,
+      "step": 8500
+    },
+    {
+      "epoch": 0.92,
+      "eval_loss": 0.5945226550102234,
+      "eval_runtime": 49.073,
+      "eval_samples_per_second": 41.306,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.49476439790575916,
+      "step": 8600
+    },
+    {
+      "epoch": 0.93,
+      "eval_loss": 0.5967476963996887,
+      "eval_runtime": 49.0045,
+      "eval_samples_per_second": 41.364,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.553168441957032,
+      "step": 8700
+    },
+    {
+      "epoch": 0.94,
+      "eval_loss": 0.5770399570465088,
+      "eval_runtime": 48.7028,
+      "eval_samples_per_second": 41.62,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.4961184329301318,
+      "step": 8800
+    },
+    {
+      "epoch": 0.95,
+      "eval_loss": 0.5808792114257812,
+      "eval_runtime": 48.8469,
+      "eval_samples_per_second": 41.497,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.5203105253655894,
+      "step": 8900
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00028174426934097417,
+      "loss": 0.4305,
+      "step": 9000
+    },
+    {
+      "epoch": 0.96,
+      "eval_loss": 0.5804657936096191,
+      "eval_runtime": 49.9542,
+      "eval_samples_per_second": 40.577,
+      "eval_steps_per_second": 5.085,
+      "eval_wer": 0.5038815670698682,
+      "step": 9000
+    },
+    {
+      "epoch": 0.97,
+      "eval_loss": 0.5873450040817261,
+      "eval_runtime": 48.5365,
+      "eval_samples_per_second": 41.762,
+      "eval_steps_per_second": 5.233,
+      "eval_wer": 0.5187759523379671,
+      "step": 9100
+    },
+    {
+      "epoch": 0.98,
+      "eval_loss": 0.6276754140853882,
+      "eval_runtime": 48.6742,
+      "eval_samples_per_second": 41.644,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.5516338689294097,
+      "step": 9200
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 0.5726681351661682,
+      "eval_runtime": 48.9081,
+      "eval_samples_per_second": 41.445,
+      "eval_steps_per_second": 5.193,
+      "eval_wer": 0.5052356020942408,
+      "step": 9300
+    },
+    {
+      "epoch": 1.01,
+      "eval_loss": 0.5858115553855896,
+      "eval_runtime": 48.4967,
+      "eval_samples_per_second": 41.797,
+      "eval_steps_per_second": 5.237,
+      "eval_wer": 0.5122765842209785,
+      "step": 9400
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002806719197707736,
+      "loss": 0.4264,
+      "step": 9500
+    },
+    {
+      "epoch": 1.02,
+      "eval_loss": 0.5691950917243958,
+      "eval_runtime": 48.3307,
+      "eval_samples_per_second": 41.94,
+      "eval_steps_per_second": 5.255,
+      "eval_wer": 0.49684058494313055,
+      "step": 9500
+    },
+    {
+      "epoch": 1.03,
+      "eval_loss": 0.5954357981681824,
+      "eval_runtime": 48.3796,
+      "eval_samples_per_second": 41.898,
+      "eval_steps_per_second": 5.25,
+      "eval_wer": 0.5117349702112295,
+      "step": 9600
+    },
+    {
+      "epoch": 1.04,
+      "eval_loss": 0.5903880000114441,
+      "eval_runtime": 48.2992,
+      "eval_samples_per_second": 41.968,
+      "eval_steps_per_second": 5.259,
+      "eval_wer": 0.5075825961364867,
+      "step": 9700
+    },
+    {
+      "epoch": 1.05,
+      "eval_loss": 0.6046377420425415,
+      "eval_runtime": 48.5859,
+      "eval_samples_per_second": 41.72,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.5101101281819823,
+      "step": 9800
+    },
+    {
+      "epoch": 1.06,
+      "eval_loss": 0.5616235136985779,
+      "eval_runtime": 48.5046,
+      "eval_samples_per_second": 41.79,
+      "eval_steps_per_second": 5.237,
+      "eval_wer": 0.492597941866763,
+      "step": 9900
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.00027959742120343837,
+      "loss": 0.4176,
+      "step": 10000
+    },
+    {
+      "epoch": 1.07,
+      "eval_loss": 0.5971259474754333,
+      "eval_runtime": 48.5096,
+      "eval_samples_per_second": 41.786,
+      "eval_steps_per_second": 5.236,
+      "eval_wer": 0.5368297526629355,
+      "step": 10000
+    },
+    {
+      "epoch": 1.08,
+      "eval_loss": 0.5706018805503845,
+      "eval_runtime": 48.1561,
+      "eval_samples_per_second": 42.092,
+      "eval_steps_per_second": 5.275,
+      "eval_wer": 0.4940422458927604,
+      "step": 10100
+    },
+    {
+      "epoch": 1.09,
+      "eval_loss": 0.5612200498580933,
+      "eval_runtime": 48.0575,
+      "eval_samples_per_second": 42.179,
+      "eval_steps_per_second": 5.285,
+      "eval_wer": 0.5032496840584944,
+      "step": 10200
+    },
+    {
+      "epoch": 1.1,
+      "eval_loss": 0.5671827793121338,
+      "eval_runtime": 48.4862,
+      "eval_samples_per_second": 41.806,
+      "eval_steps_per_second": 5.239,
+      "eval_wer": 0.4944033218992598,
+      "step": 10300
+    },
+    {
+      "epoch": 1.11,
+      "eval_loss": 0.5915140509605408,
+      "eval_runtime": 48.186,
+      "eval_samples_per_second": 42.066,
+      "eval_steps_per_second": 5.271,
+      "eval_wer": 0.5218450983932118,
+      "step": 10400
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.00027852292263610317,
+      "loss": 0.4033,
+      "step": 10500
+    },
+    {
+      "epoch": 1.12,
+      "eval_loss": 0.5705661177635193,
+      "eval_runtime": 48.5793,
+      "eval_samples_per_second": 41.726,
+      "eval_steps_per_second": 5.229,
+      "eval_wer": 0.505145333092616,
+      "step": 10500
+    },
+    {
+      "epoch": 1.13,
+      "eval_loss": 0.56607985496521,
+      "eval_runtime": 48.4918,
+      "eval_samples_per_second": 41.801,
+      "eval_steps_per_second": 5.238,
+      "eval_wer": 0.4934103628813865,
+      "step": 10600
+    },
+    {
+      "epoch": 1.15,
+      "eval_loss": 0.5724338889122009,
+      "eval_runtime": 48.2921,
+      "eval_samples_per_second": 41.974,
+      "eval_steps_per_second": 5.26,
+      "eval_wer": 0.49025094782451706,
+      "step": 10700
+    },
+    {
+      "epoch": 1.16,
+      "eval_loss": 0.5792076587677002,
+      "eval_runtime": 48.2324,
+      "eval_samples_per_second": 42.026,
+      "eval_steps_per_second": 5.266,
+      "eval_wer": 0.49395197689113557,
+      "step": 10800
+    },
+    {
+      "epoch": 1.17,
+      "eval_loss": 0.5744290351867676,
+      "eval_runtime": 48.5673,
+      "eval_samples_per_second": 41.736,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.49106336883914065,
+      "step": 10900
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00027744842406876786,
+      "loss": 0.392,
+      "step": 11000
+    },
+    {
+      "epoch": 1.18,
+      "eval_loss": 0.5767287015914917,
+      "eval_runtime": 48.5934,
+      "eval_samples_per_second": 41.714,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.5161581512908467,
+      "step": 11000
+    },
+    {
+      "epoch": 1.19,
+      "eval_loss": 0.5588058829307556,
+      "eval_runtime": 48.5001,
+      "eval_samples_per_second": 41.794,
+      "eval_steps_per_second": 5.237,
+      "eval_wer": 0.4834807727026539,
+      "step": 11100
+    },
+    {
+      "epoch": 1.2,
+      "eval_loss": 0.5609337687492371,
+      "eval_runtime": 48.6478,
+      "eval_samples_per_second": 41.667,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.4922368658602636,
+      "step": 11200
+    },
+    {
+      "epoch": 1.21,
+      "eval_loss": 0.5889888405799866,
+      "eval_runtime": 48.4613,
+      "eval_samples_per_second": 41.827,
+      "eval_steps_per_second": 5.241,
+      "eval_wer": 0.49142444484564,
+      "step": 11300
+    },
+    {
+      "epoch": 1.22,
+      "eval_loss": 0.5525108575820923,
+      "eval_runtime": 48.6323,
+      "eval_samples_per_second": 41.68,
+      "eval_steps_per_second": 5.223,
+      "eval_wer": 0.489709333814768,
+      "step": 11400
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.0002763739255014326,
+      "loss": 0.387,
+      "step": 11500
+    },
+    {
+      "epoch": 1.23,
+      "eval_loss": 0.5704483389854431,
+      "eval_runtime": 48.2001,
+      "eval_samples_per_second": 42.054,
+      "eval_steps_per_second": 5.27,
+      "eval_wer": 0.5050550640909911,
+      "step": 11500
+    },
+    {
+      "epoch": 1.24,
+      "eval_loss": 0.5538840889930725,
+      "eval_runtime": 48.397,
+      "eval_samples_per_second": 41.883,
+      "eval_steps_per_second": 5.248,
+      "eval_wer": 0.5013540350243726,
+      "step": 11600
+    },
+    {
+      "epoch": 1.25,
+      "eval_loss": 0.5473034381866455,
+      "eval_runtime": 48.5532,
+      "eval_samples_per_second": 41.748,
+      "eval_steps_per_second": 5.231,
+      "eval_wer": 0.48817476078714567,
+      "step": 11700
+    },
+    {
+      "epoch": 1.26,
+      "eval_loss": 0.5661569237709045,
+      "eval_runtime": 48.3301,
+      "eval_samples_per_second": 41.941,
+      "eval_steps_per_second": 5.256,
+      "eval_wer": 0.5003610760064994,
+      "step": 11800
+    },
+    {
+      "epoch": 1.27,
+      "eval_loss": 0.5784680247306824,
+      "eval_runtime": 48.4292,
+      "eval_samples_per_second": 41.855,
+      "eval_steps_per_second": 5.245,
+      "eval_wer": 0.5220256363964615,
+      "step": 11900
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.0002752994269340974,
+      "loss": 0.3956,
+      "step": 12000
+    },
+    {
+      "epoch": 1.28,
+      "eval_loss": 0.5989814400672913,
+      "eval_runtime": 48.3584,
+      "eval_samples_per_second": 41.916,
+      "eval_steps_per_second": 5.252,
+      "eval_wer": 0.5113738942047301,
+      "step": 12000
+    },
+    {
+      "epoch": 1.3,
+      "eval_loss": 0.549746036529541,
+      "eval_runtime": 48.5897,
+      "eval_samples_per_second": 41.717,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.4895287958115183,
+      "step": 12100
+    },
+    {
+      "epoch": 1.31,
+      "eval_loss": 0.5537735819816589,
+      "eval_runtime": 48.5304,
+      "eval_samples_per_second": 41.768,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.4895287958115183,
+      "step": 12200
+    },
+    {
+      "epoch": 1.32,
+      "eval_loss": 0.5651851296424866,
+      "eval_runtime": 48.2022,
+      "eval_samples_per_second": 42.052,
+      "eval_steps_per_second": 5.269,
+      "eval_wer": 0.4913341758440152,
+      "step": 12300
+    },
+    {
+      "epoch": 1.33,
+      "eval_loss": 0.5682100057601929,
+      "eval_runtime": 48.2298,
+      "eval_samples_per_second": 42.028,
+      "eval_steps_per_second": 5.266,
+      "eval_wer": 0.5128181982307276,
+      "step": 12400
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.00027422492836676215,
+      "loss": 0.4043,
+      "step": 12500
+    },
+    {
+      "epoch": 1.34,
+      "eval_loss": 0.5829596519470215,
+      "eval_runtime": 48.3024,
+      "eval_samples_per_second": 41.965,
+      "eval_steps_per_second": 5.259,
+      "eval_wer": 0.49990973099837516,
+      "step": 12500
+    },
+    {
+      "epoch": 1.35,
+      "eval_loss": 0.5685592889785767,
+      "eval_runtime": 48.5496,
+      "eval_samples_per_second": 41.751,
+      "eval_steps_per_second": 5.232,
+      "eval_wer": 0.4864596497562737,
+      "step": 12600
+    },
+    {
+      "epoch": 1.36,
+      "eval_loss": 0.5688390135765076,
+      "eval_runtime": 48.1792,
+      "eval_samples_per_second": 42.072,
+      "eval_steps_per_second": 5.272,
+      "eval_wer": 0.49368116988626104,
+      "step": 12700
+    },
+    {
+      "epoch": 1.37,
+      "eval_loss": 0.575319230556488,
+      "eval_runtime": 48.6663,
+      "eval_samples_per_second": 41.651,
+      "eval_steps_per_second": 5.219,
+      "eval_wer": 0.503430222061744,
+      "step": 12800
+    },
+    {
+      "epoch": 1.38,
+      "eval_loss": 0.5897743701934814,
+      "eval_runtime": 48.6109,
+      "eval_samples_per_second": 41.698,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.48654991875789855,
+      "step": 12900
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.0002731504297994269,
+      "loss": 0.3997,
+      "step": 13000
+    },
+    {
+      "epoch": 1.39,
+      "eval_loss": 0.5723254680633545,
+      "eval_runtime": 48.325,
+      "eval_samples_per_second": 41.945,
+      "eval_steps_per_second": 5.256,
+      "eval_wer": 0.4962989709333815,
+      "step": 13000
+    },
+    {
+      "epoch": 1.4,
+      "eval_loss": 0.5766825675964355,
+      "eval_runtime": 48.386,
+      "eval_samples_per_second": 41.892,
+      "eval_steps_per_second": 5.249,
+      "eval_wer": 0.4985556959740025,
+      "step": 13100
+    },
+    {
+      "epoch": 1.41,
+      "eval_loss": 0.5959895849227905,
+      "eval_runtime": 48.351,
+      "eval_samples_per_second": 41.923,
+      "eval_steps_per_second": 5.253,
+      "eval_wer": 0.5083950171511104,
+      "step": 13200
+    },
+    {
+      "epoch": 1.42,
+      "eval_loss": 0.5859235525131226,
+      "eval_runtime": 49.246,
+      "eval_samples_per_second": 41.161,
+      "eval_steps_per_second": 5.158,
+      "eval_wer": 0.5095685141722333,
+      "step": 13300
+    },
+    {
+      "epoch": 1.43,
+      "eval_loss": 0.5490578413009644,
+      "eval_runtime": 48.6297,
+      "eval_samples_per_second": 41.682,
+      "eval_steps_per_second": 5.223,
+      "eval_wer": 0.4784257086116628,
+      "step": 13400
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.00027207593123209164,
+      "loss": 0.3997,
+      "step": 13500
+    },
+    {
+      "epoch": 1.45,
+      "eval_loss": 0.5636021494865417,
+      "eval_runtime": 48.4175,
+      "eval_samples_per_second": 41.865,
+      "eval_steps_per_second": 5.246,
+      "eval_wer": 0.5048745260877414,
+      "step": 13500
+    },
+    {
+      "epoch": 1.46,
+      "eval_loss": 0.5667489767074585,
+      "eval_runtime": 48.7553,
+      "eval_samples_per_second": 41.575,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.470843112475176,
+      "step": 13600
+    },
+    {
+      "epoch": 1.47,
+      "eval_loss": 0.5757395029067993,
+      "eval_runtime": 48.5863,
+      "eval_samples_per_second": 41.72,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.4861888427513992,
+      "step": 13700
+    },
+    {
+      "epoch": 1.48,
+      "eval_loss": 0.5444319844245911,
+      "eval_runtime": 48.6276,
+      "eval_samples_per_second": 41.684,
+      "eval_steps_per_second": 5.223,
+      "eval_wer": 0.48158512366853223,
+      "step": 13800
+    },
+    {
+      "epoch": 1.49,
+      "eval_loss": 0.5556561350822449,
+      "eval_runtime": 48.6386,
+      "eval_samples_per_second": 41.675,
+      "eval_steps_per_second": 5.222,
+      "eval_wer": 0.4792381296262863,
+      "step": 13900
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00027100143266475644,
+      "loss": 0.3954,
+      "step": 14000
+    },
+    {
+      "epoch": 1.5,
+      "eval_loss": 0.5437420010566711,
+      "eval_runtime": 48.3246,
+      "eval_samples_per_second": 41.946,
+      "eval_steps_per_second": 5.256,
+      "eval_wer": 0.48095324065715833,
+      "step": 14000
+    },
+    {
+      "epoch": 1.51,
+      "eval_loss": 0.5488930940628052,
+      "eval_runtime": 48.1775,
+      "eval_samples_per_second": 42.074,
+      "eval_steps_per_second": 5.272,
+      "eval_wer": 0.46741289041343204,
+      "step": 14100
+    },
+    {
+      "epoch": 1.52,
+      "eval_loss": 0.541516900062561,
+      "eval_runtime": 48.22,
+      "eval_samples_per_second": 42.037,
+      "eval_steps_per_second": 5.268,
+      "eval_wer": 0.46741289041343204,
+      "step": 14200
+    },
+    {
+      "epoch": 1.53,
+      "eval_loss": 0.5481301546096802,
+      "eval_runtime": 49.0132,
+      "eval_samples_per_second": 41.356,
+      "eval_steps_per_second": 5.182,
+      "eval_wer": 0.4901606788228922,
+      "step": 14300
+    },
+    {
+      "epoch": 1.54,
+      "eval_loss": 0.547429621219635,
+      "eval_runtime": 48.9783,
+      "eval_samples_per_second": 41.386,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.4763495215742914,
+      "step": 14400
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.0002699269340974212,
+      "loss": 0.3814,
+      "step": 14500
+    },
+    {
+      "epoch": 1.55,
+      "eval_loss": 0.5588225722312927,
+      "eval_runtime": 48.3905,
+      "eval_samples_per_second": 41.888,
+      "eval_steps_per_second": 5.249,
+      "eval_wer": 0.4730998375157971,
+      "step": 14500
+    },
+    {
+      "epoch": 1.56,
+      "eval_loss": 0.5745720267295837,
+      "eval_runtime": 48.5865,
+      "eval_samples_per_second": 41.719,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.48203646867665645,
+      "step": 14600
+    },
+    {
+      "epoch": 1.57,
+      "eval_loss": 0.5675551891326904,
+      "eval_runtime": 48.3999,
+      "eval_samples_per_second": 41.88,
+      "eval_steps_per_second": 5.248,
+      "eval_wer": 0.48835529879039535,
+      "step": 14700
+    },
+    {
+      "epoch": 1.58,
+      "eval_loss": 0.5494562387466431,
+      "eval_runtime": 48.3296,
+      "eval_samples_per_second": 41.941,
+      "eval_steps_per_second": 5.256,
+      "eval_wer": 0.47111391948005055,
+      "step": 14800
+    },
+    {
+      "epoch": 1.6,
+      "eval_loss": 0.556481659412384,
+      "eval_runtime": 48.4994,
+      "eval_samples_per_second": 41.794,
+      "eval_steps_per_second": 5.237,
+      "eval_wer": 0.4782451706084131,
+      "step": 14900
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.0002688545845272206,
+      "loss": 0.3877,
+      "step": 15000
+    },
+    {
+      "epoch": 1.61,
+      "eval_loss": 0.5670508742332458,
+      "eval_runtime": 48.5206,
+      "eval_samples_per_second": 41.776,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.5135403502437264,
+      "step": 15000
+    },
+    {
+      "epoch": 1.62,
+      "eval_loss": 0.5511665344238281,
+      "eval_runtime": 48.5022,
+      "eval_samples_per_second": 41.792,
+      "eval_steps_per_second": 5.237,
+      "eval_wer": 0.4868207257627731,
+      "step": 15100
+    },
+    {
+      "epoch": 1.63,
+      "eval_loss": 0.5683439373970032,
+      "eval_runtime": 48.7471,
+      "eval_samples_per_second": 41.582,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.4649756273695613,
+      "step": 15200
+    },
+    {
+      "epoch": 1.64,
+      "eval_loss": 0.5426750183105469,
+      "eval_runtime": 48.8663,
+      "eval_samples_per_second": 41.481,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.47174580249142445,
+      "step": 15300
+    },
+    {
+      "epoch": 1.65,
+      "eval_loss": 0.5519290566444397,
+      "eval_runtime": 48.7168,
+      "eval_samples_per_second": 41.608,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.4650658963711861,
+      "step": 15400
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.0002677800859598854,
+      "loss": 0.387,
+      "step": 15500
+    },
+    {
+      "epoch": 1.66,
+      "eval_loss": 0.532715916633606,
+      "eval_runtime": 48.8241,
+      "eval_samples_per_second": 41.516,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.44556779202022023,
+      "step": 15500
+    },
+    {
+      "epoch": 1.67,
+      "eval_loss": 0.5371445417404175,
+      "eval_runtime": 48.9886,
+      "eval_samples_per_second": 41.377,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.4673226214118072,
+      "step": 15600
+    },
+    {
+      "epoch": 1.68,
+      "eval_loss": 0.5337085127830505,
+      "eval_runtime": 48.4513,
+      "eval_samples_per_second": 41.836,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.47048203646867665,
+      "step": 15700
+    },
+    {
+      "epoch": 1.69,
+      "eval_loss": 0.5606207847595215,
+      "eval_runtime": 48.5238,
+      "eval_samples_per_second": 41.773,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.4991875789853764,
+      "step": 15800
+    },
+    {
+      "epoch": 1.7,
+      "eval_loss": 0.5253767967224121,
+      "eval_runtime": 48.1837,
+      "eval_samples_per_second": 42.068,
+      "eval_steps_per_second": 5.271,
+      "eval_wer": 0.46127459830294276,
+      "step": 15900
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00026670558739255013,
+      "loss": 0.3877,
+      "step": 16000
+    },
+    {
+      "epoch": 1.71,
+      "eval_loss": 0.5618729591369629,
+      "eval_runtime": 48.4128,
+      "eval_samples_per_second": 41.869,
+      "eval_steps_per_second": 5.247,
+      "eval_wer": 0.48817476078714567,
+      "step": 16000
+    },
+    {
+      "epoch": 1.72,
+      "eval_loss": 0.5212122797966003,
+      "eval_runtime": 48.591,
+      "eval_samples_per_second": 41.716,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.4560389962087019,
+      "step": 16100
+    },
+    {
+      "epoch": 1.73,
+      "eval_loss": 0.536945641040802,
+      "eval_runtime": 48.7143,
+      "eval_samples_per_second": 41.61,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.4695793464524282,
+      "step": 16200
+    },
+    {
+      "epoch": 1.75,
+      "eval_loss": 0.5391582250595093,
+      "eval_runtime": 48.284,
+      "eval_samples_per_second": 41.981,
+      "eval_steps_per_second": 5.261,
+      "eval_wer": 0.46768369741830657,
+      "step": 16300
+    },
+    {
+      "epoch": 1.76,
+      "eval_loss": 0.535341739654541,
+      "eval_runtime": 48.3278,
+      "eval_samples_per_second": 41.943,
+      "eval_steps_per_second": 5.256,
+      "eval_wer": 0.4768008665824156,
+      "step": 16400
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0002656310888252149,
+      "loss": 0.3739,
+      "step": 16500
+    },
+    {
+      "epoch": 1.77,
+      "eval_loss": 0.5435373187065125,
+      "eval_runtime": 48.2949,
+      "eval_samples_per_second": 41.971,
+      "eval_steps_per_second": 5.259,
+      "eval_wer": 0.47770355659866404,
+      "step": 16500
+    },
+    {
+      "epoch": 1.78,
+      "eval_loss": 0.5342755317687988,
+      "eval_runtime": 48.4704,
+      "eval_samples_per_second": 41.819,
+      "eval_steps_per_second": 5.24,
+      "eval_wer": 0.4884455677920202,
+      "step": 16600
+    },
+    {
+      "epoch": 1.79,
+      "eval_loss": 0.5309295058250427,
+      "eval_runtime": 48.3519,
+      "eval_samples_per_second": 41.922,
+      "eval_steps_per_second": 5.253,
+      "eval_wer": 0.4942227838960101,
+      "step": 16700
+    },
+    {
+      "epoch": 1.8,
+      "eval_loss": 0.5372947454452515,
+      "eval_runtime": 48.1794,
+      "eval_samples_per_second": 42.072,
+      "eval_steps_per_second": 5.272,
+      "eval_wer": 0.4727387615092977,
+      "step": 16800
+    },
+    {
+      "epoch": 1.81,
+      "eval_loss": 0.5549677610397339,
+      "eval_runtime": 48.1603,
+      "eval_samples_per_second": 42.089,
+      "eval_steps_per_second": 5.274,
+      "eval_wer": 0.468586387434555,
+      "step": 16900
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.0002645565902578796,
+      "loss": 0.3884,
+      "step": 17000
+    },
+    {
+      "epoch": 1.82,
+      "eval_loss": 0.5485758185386658,
+      "eval_runtime": 48.2605,
+      "eval_samples_per_second": 42.001,
+      "eval_steps_per_second": 5.263,
+      "eval_wer": 0.4825780826864055,
+      "step": 17000
+    },
+    {
+      "epoch": 1.83,
+      "eval_loss": 0.5507860779762268,
+      "eval_runtime": 48.2489,
+      "eval_samples_per_second": 42.011,
+      "eval_steps_per_second": 5.264,
+      "eval_wer": 0.4861888427513992,
+      "step": 17100
+    },
+    {
+      "epoch": 1.84,
+      "eval_loss": 0.5423179864883423,
+      "eval_runtime": 48.238,
+      "eval_samples_per_second": 42.021,
+      "eval_steps_per_second": 5.266,
+      "eval_wer": 0.48546669073840043,
+      "step": 17200
+    },
+    {
+      "epoch": 1.85,
+      "eval_loss": 0.5478358864784241,
+      "eval_runtime": 48.2249,
+      "eval_samples_per_second": 42.032,
+      "eval_steps_per_second": 5.267,
+      "eval_wer": 0.47300956851417225,
+      "step": 17300
+    },
+    {
+      "epoch": 1.86,
+      "eval_loss": 0.5437774062156677,
+      "eval_runtime": 48.3875,
+      "eval_samples_per_second": 41.891,
+      "eval_steps_per_second": 5.249,
+      "eval_wer": 0.4937714388878859,
+      "step": 17400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0002634820916905444,
+      "loss": 0.3842,
+      "step": 17500
+    },
+    {
+      "epoch": 1.87,
+      "eval_loss": 0.5570672750473022,
+      "eval_runtime": 48.3857,
+      "eval_samples_per_second": 41.893,
+      "eval_steps_per_second": 5.249,
+      "eval_wer": 0.4817656616717819,
+      "step": 17500
+    },
+    {
+      "epoch": 1.88,
+      "eval_loss": 0.5401641130447388,
+      "eval_runtime": 48.625,
+      "eval_samples_per_second": 41.686,
+      "eval_steps_per_second": 5.224,
+      "eval_wer": 0.4752662935547933,
+      "step": 17600
+    },
+    {
+      "epoch": 1.9,
+      "eval_loss": 0.5679250955581665,
+      "eval_runtime": 48.1534,
+      "eval_samples_per_second": 42.095,
+      "eval_steps_per_second": 5.275,
+      "eval_wer": 0.48266835168803035,
+      "step": 17700
+    },
+    {
+      "epoch": 1.91,
+      "eval_loss": 0.5385044813156128,
+      "eval_runtime": 49.0923,
+      "eval_samples_per_second": 41.29,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.46416320635493774,
+      "step": 17800
+    },
+    {
+      "epoch": 1.92,
+      "eval_loss": 0.5519374012947083,
+      "eval_runtime": 48.6773,
+      "eval_samples_per_second": 41.642,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.4942227838960101,
+      "step": 17900
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00026240759312320917,
+      "loss": 0.3953,
+      "step": 18000
+    },
+    {
+      "epoch": 1.93,
+      "eval_loss": 0.5559102892875671,
+      "eval_runtime": 48.5742,
+      "eval_samples_per_second": 41.73,
+      "eval_steps_per_second": 5.229,
+      "eval_wer": 0.47454414154179453,
+      "step": 18000
+    },
+    {
+      "epoch": 1.94,
+      "eval_loss": 0.5656890869140625,
+      "eval_runtime": 48.6283,
+      "eval_samples_per_second": 41.684,
+      "eval_steps_per_second": 5.223,
+      "eval_wer": 0.4962989709333815,
+      "step": 18100
+    },
+    {
+      "epoch": 1.95,
+      "eval_loss": 0.5296332836151123,
+      "eval_runtime": 48.245,
+      "eval_samples_per_second": 42.015,
+      "eval_steps_per_second": 5.265,
+      "eval_wer": 0.46416320635493774,
+      "step": 18200
+    },
+    {
+      "epoch": 1.96,
+      "eval_loss": 0.5529346466064453,
+      "eval_runtime": 48.1639,
+      "eval_samples_per_second": 42.085,
+      "eval_steps_per_second": 5.274,
+      "eval_wer": 0.4907022928326413,
+      "step": 18300
+    },
+    {
+      "epoch": 1.97,
+      "eval_loss": 0.5379639267921448,
+      "eval_runtime": 48.3399,
+      "eval_samples_per_second": 41.932,
+      "eval_steps_per_second": 5.254,
+      "eval_wer": 0.4536017331648312,
+      "step": 18400
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.00026133524355300856,
+      "loss": 0.3745,
+      "step": 18500
+    },
+    {
+      "epoch": 1.98,
+      "eval_loss": 0.5275648832321167,
+      "eval_runtime": 48.1451,
+      "eval_samples_per_second": 42.102,
+      "eval_steps_per_second": 5.276,
+      "eval_wer": 0.4677739664199314,
+      "step": 18500
+    },
+    {
+      "epoch": 1.99,
+      "eval_loss": 0.5544295310974121,
+      "eval_runtime": 48.5316,
+      "eval_samples_per_second": 41.767,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.4853764217367756,
+      "step": 18600
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.5195306539535522,
+      "eval_runtime": 48.7173,
+      "eval_samples_per_second": 41.607,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.45351146416320637,
+      "step": 18700
+    },
+    {
+      "epoch": 2.01,
+      "eval_loss": 0.5165307521820068,
+      "eval_runtime": 49.0261,
+      "eval_samples_per_second": 41.345,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.46353132334356384,
+      "step": 18800
+    },
+    {
+      "epoch": 2.02,
+      "eval_loss": 0.5062076449394226,
+      "eval_runtime": 49.107,
+      "eval_samples_per_second": 41.277,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.4431305289763495,
+      "step": 18900
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00026026074498567336,
+      "loss": 0.3538,
+      "step": 19000
+    },
+    {
+      "epoch": 2.03,
+      "eval_loss": 0.5254501104354858,
+      "eval_runtime": 55.2719,
+      "eval_samples_per_second": 36.673,
+      "eval_steps_per_second": 4.595,
+      "eval_wer": 0.4508936631160859,
+      "step": 19000
+    },
+    {
+      "epoch": 2.04,
+      "eval_loss": 0.5124872326850891,
+      "eval_runtime": 48.7773,
+      "eval_samples_per_second": 41.556,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.45116447012096045,
+      "step": 19100
+    },
+    {
+      "epoch": 2.06,
+      "eval_loss": 0.510496973991394,
+      "eval_runtime": 49.0354,
+      "eval_samples_per_second": 41.337,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.45035204910633686,
+      "step": 19200
+    },
+    {
+      "epoch": 2.07,
+      "eval_loss": 0.5000460147857666,
+      "eval_runtime": 48.7939,
+      "eval_samples_per_second": 41.542,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.44899801408196427,
+      "step": 19300
+    },
+    {
+      "epoch": 2.08,
+      "eval_loss": 0.5149863362312317,
+      "eval_runtime": 48.7295,
+      "eval_samples_per_second": 41.597,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.45197689113558404,
+      "step": 19400
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0002591862464183381,
+      "loss": 0.356,
+      "step": 19500
+    },
+    {
+      "epoch": 2.09,
+      "eval_loss": 0.5052781701087952,
+      "eval_runtime": 48.7825,
+      "eval_samples_per_second": 41.552,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.43825600288860805,
+      "step": 19500
+    },
+    {
+      "epoch": 2.1,
+      "eval_loss": 0.5084646344184875,
+      "eval_runtime": 48.8304,
+      "eval_samples_per_second": 41.511,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.44168622495035204,
+      "step": 19600
+    },
+    {
+      "epoch": 2.11,
+      "eval_loss": 0.5229296088218689,
+      "eval_runtime": 49.9067,
+      "eval_samples_per_second": 40.616,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.44899801408196427,
+      "step": 19700
+    },
+    {
+      "epoch": 2.12,
+      "eval_loss": 0.5325890183448792,
+      "eval_runtime": 49.0157,
+      "eval_samples_per_second": 41.354,
+      "eval_steps_per_second": 5.182,
+      "eval_wer": 0.44917855208521396,
+      "step": 19800
+    },
+    {
+      "epoch": 2.13,
+      "eval_loss": 0.5138561725616455,
+      "eval_runtime": 48.4503,
+      "eval_samples_per_second": 41.837,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.4490882830835891,
+      "step": 19900
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.00025811174785100285,
+      "loss": 0.3474,
+      "step": 20000
+    },
+    {
+      "epoch": 2.14,
+      "eval_loss": 0.5133703351020813,
+      "eval_runtime": 48.4371,
+      "eval_samples_per_second": 41.848,
+      "eval_steps_per_second": 5.244,
+      "eval_wer": 0.43843654089185774,
+      "step": 20000
+    },
+    {
+      "epoch": 2.15,
+      "eval_loss": 0.5498203635215759,
+      "eval_runtime": 49.0522,
+      "eval_samples_per_second": 41.323,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.46064271529156886,
+      "step": 20100
+    },
+    {
+      "epoch": 2.16,
+      "eval_loss": 0.532422661781311,
+      "eval_runtime": 49.1671,
+      "eval_samples_per_second": 41.227,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.4539628091713306,
+      "step": 20200
+    },
+    {
+      "epoch": 2.17,
+      "eval_loss": 0.5338008999824524,
+      "eval_runtime": 48.8328,
+      "eval_samples_per_second": 41.509,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.4547752301859541,
+      "step": 20300
+    },
+    {
+      "epoch": 2.18,
+      "eval_loss": 0.5075809955596924,
+      "eval_runtime": 49.1135,
+      "eval_samples_per_second": 41.272,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.4424986459649756,
+      "step": 20400
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.0002570393982808023,
+      "loss": 0.345,
+      "step": 20500
+    },
+    {
+      "epoch": 2.19,
+      "eval_loss": 0.5252628326416016,
+      "eval_runtime": 48.651,
+      "eval_samples_per_second": 41.664,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.4549557681892038,
+      "step": 20500
+    },
+    {
+      "epoch": 2.21,
+      "eval_loss": 0.5125060677528381,
+      "eval_runtime": 48.9394,
+      "eval_samples_per_second": 41.419,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.4618162123126918,
+      "step": 20600
+    },
+    {
+      "epoch": 2.22,
+      "eval_loss": 0.5170891284942627,
+      "eval_runtime": 48.4981,
+      "eval_samples_per_second": 41.795,
+      "eval_steps_per_second": 5.237,
+      "eval_wer": 0.44872720707708974,
+      "step": 20700
+    },
+    {
+      "epoch": 2.23,
+      "eval_loss": 0.5232036113739014,
+      "eval_runtime": 50.0089,
+      "eval_samples_per_second": 40.533,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.4463802130348438,
+      "step": 20800
+    },
+    {
+      "epoch": 2.24,
+      "eval_loss": 0.5297756195068359,
+      "eval_runtime": 49.4494,
+      "eval_samples_per_second": 40.991,
+      "eval_steps_per_second": 5.137,
+      "eval_wer": 0.45883733525907205,
+      "step": 20900
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.000255964899713467,
+      "loss": 0.341,
+      "step": 21000
+    },
+    {
+      "epoch": 2.25,
+      "eval_loss": 0.5342200994491577,
+      "eval_runtime": 48.7244,
+      "eval_samples_per_second": 41.601,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.45757356923632425,
+      "step": 21000
+    },
+    {
+      "epoch": 2.26,
+      "eval_loss": 0.5514694452285767,
+      "eval_runtime": 48.4433,
+      "eval_samples_per_second": 41.843,
+      "eval_steps_per_second": 5.243,
+      "eval_wer": 0.4677739664199314,
+      "step": 21100
+    },
+    {
+      "epoch": 2.27,
+      "eval_loss": 0.5041313767433167,
+      "eval_runtime": 48.8996,
+      "eval_samples_per_second": 41.452,
+      "eval_steps_per_second": 5.194,
+      "eval_wer": 0.44953962809171333,
+      "step": 21200
+    },
+    {
+      "epoch": 2.28,
+      "eval_loss": 0.5169302821159363,
+      "eval_runtime": 48.8222,
+      "eval_samples_per_second": 41.518,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.44728290305109225,
+      "step": 21300
+    },
+    {
+      "epoch": 2.29,
+      "eval_loss": 0.522724449634552,
+      "eval_runtime": 48.6369,
+      "eval_samples_per_second": 41.676,
+      "eval_steps_per_second": 5.222,
+      "eval_wer": 0.44935909008846364,
+      "step": 21400
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0002548904011461318,
+      "loss": 0.354,
+      "step": 21500
+    },
+    {
+      "epoch": 2.3,
+      "eval_loss": 0.5213791131973267,
+      "eval_runtime": 48.5544,
+      "eval_samples_per_second": 41.747,
+      "eval_steps_per_second": 5.231,
+      "eval_wer": 0.44583859902509476,
+      "step": 21500
+    },
+    {
+      "epoch": 2.31,
+      "eval_loss": 0.5303418636322021,
+      "eval_runtime": 49.529,
+      "eval_samples_per_second": 40.925,
+      "eval_steps_per_second": 5.128,
+      "eval_wer": 0.4587470662574472,
+      "step": 21600
+    },
+    {
+      "epoch": 2.32,
+      "eval_loss": 0.5236622095108032,
+      "eval_runtime": 48.9492,
+      "eval_samples_per_second": 41.41,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.45974002527532043,
+      "step": 21700
+    },
+    {
+      "epoch": 2.33,
+      "eval_loss": 0.5067057013511658,
+      "eval_runtime": 49.3403,
+      "eval_samples_per_second": 41.082,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.44601913702834445,
+      "step": 21800
+    },
+    {
+      "epoch": 2.34,
+      "eval_loss": 0.5117440819740295,
+      "eval_runtime": 48.9443,
+      "eval_samples_per_second": 41.414,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.4560389962087019,
+      "step": 21900
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.00025381590257879654,
+      "loss": 0.3333,
+      "step": 22000
+    },
+    {
+      "epoch": 2.36,
+      "eval_loss": 0.5104334950447083,
+      "eval_runtime": 49.2332,
+      "eval_samples_per_second": 41.171,
+      "eval_steps_per_second": 5.159,
+      "eval_wer": 0.43590900884636213,
+      "step": 22000
+    },
+    {
+      "epoch": 2.37,
+      "eval_loss": 0.5325857996940613,
+      "eval_runtime": 48.7968,
+      "eval_samples_per_second": 41.54,
+      "eval_steps_per_second": 5.205,
+      "eval_wer": 0.46786423542155625,
+      "step": 22100
+    },
+    {
+      "epoch": 2.38,
+      "eval_loss": 0.509769082069397,
+      "eval_runtime": 49.5392,
+      "eval_samples_per_second": 40.917,
+      "eval_steps_per_second": 5.127,
+      "eval_wer": 0.45098393211771076,
+      "step": 22200
+    },
+    {
+      "epoch": 2.39,
+      "eval_loss": 0.5043946504592896,
+      "eval_runtime": 49.0673,
+      "eval_samples_per_second": 41.311,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.44448456400072217,
+      "step": 22300
+    },
+    {
+      "epoch": 2.4,
+      "eval_loss": 0.5219407677650452,
+      "eval_runtime": 49.3521,
+      "eval_samples_per_second": 41.072,
+      "eval_steps_per_second": 5.147,
+      "eval_wer": 0.4489077450803394,
+      "step": 22400
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.0002527414040114613,
+      "loss": 0.3514,
+      "step": 22500
+    },
+    {
+      "epoch": 2.41,
+      "eval_loss": 0.4986798167228699,
+      "eval_runtime": 48.8881,
+      "eval_samples_per_second": 41.462,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.4433110669795992,
+      "step": 22500
+    },
+    {
+      "epoch": 2.42,
+      "eval_loss": 0.5008650422096252,
+      "eval_runtime": 48.9214,
+      "eval_samples_per_second": 41.434,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.4338328218089908,
+      "step": 22600
+    },
+    {
+      "epoch": 2.43,
+      "eval_loss": 0.5252286195755005,
+      "eval_runtime": 48.9441,
+      "eval_samples_per_second": 41.415,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.44439429499909733,
+      "step": 22700
+    },
+    {
+      "epoch": 2.44,
+      "eval_loss": 0.4861030876636505,
+      "eval_runtime": 49.1509,
+      "eval_samples_per_second": 41.24,
+      "eval_steps_per_second": 5.168,
+      "eval_wer": 0.42688210868387794,
+      "step": 22800
+    },
+    {
+      "epoch": 2.45,
+      "eval_loss": 0.5156984925270081,
+      "eval_runtime": 48.9934,
+      "eval_samples_per_second": 41.373,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.44213756995847625,
+      "step": 22900
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.00025166690544412603,
+      "loss": 0.3444,
+      "step": 23000
+    },
+    {
+      "epoch": 2.46,
+      "eval_loss": 0.5277285575866699,
+      "eval_runtime": 49.0393,
+      "eval_samples_per_second": 41.334,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.44258891496660047,
+      "step": 23000
+    },
+    {
+      "epoch": 2.47,
+      "eval_loss": 0.5213038325309753,
+      "eval_runtime": 48.8044,
+      "eval_samples_per_second": 41.533,
+      "eval_steps_per_second": 5.204,
+      "eval_wer": 0.43780465788048384,
+      "step": 23100
+    },
+    {
+      "epoch": 2.48,
+      "eval_loss": 0.5172164440155029,
+      "eval_runtime": 48.9505,
+      "eval_samples_per_second": 41.409,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.4481855930673407,
+      "step": 23200
+    },
+    {
+      "epoch": 2.49,
+      "eval_loss": 0.51420658826828,
+      "eval_runtime": 49.0092,
+      "eval_samples_per_second": 41.36,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.43762411987723415,
+      "step": 23300
+    },
+    {
+      "epoch": 2.51,
+      "eval_loss": 0.5044341087341309,
+      "eval_runtime": 48.7967,
+      "eval_samples_per_second": 41.54,
+      "eval_steps_per_second": 5.205,
+      "eval_wer": 0.4230908106156346,
+      "step": 23400
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0002505924068767908,
+      "loss": 0.3536,
+      "step": 23500
+    },
+    {
+      "epoch": 2.52,
+      "eval_loss": 0.5267765522003174,
+      "eval_runtime": 49.0223,
+      "eval_samples_per_second": 41.349,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.44962989709333817,
+      "step": 23500
+    },
+    {
+      "epoch": 2.53,
+      "eval_loss": 0.5175958871841431,
+      "eval_runtime": 48.8831,
+      "eval_samples_per_second": 41.466,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.432569055786243,
+      "step": 23600
+    },
+    {
+      "epoch": 2.54,
+      "eval_loss": 0.5031694173812866,
+      "eval_runtime": 48.7866,
+      "eval_samples_per_second": 41.548,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.42959017873262323,
+      "step": 23700
+    },
+    {
+      "epoch": 2.55,
+      "eval_loss": 0.5210540890693665,
+      "eval_runtime": 49.2301,
+      "eval_samples_per_second": 41.174,
+      "eval_steps_per_second": 5.159,
+      "eval_wer": 0.44601913702834445,
+      "step": 23800
+    },
+    {
+      "epoch": 2.56,
+      "eval_loss": 0.5092755556106567,
+      "eval_runtime": 48.5206,
+      "eval_samples_per_second": 41.776,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.4378949268821087,
+      "step": 23900
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.00024952005730659023,
+      "loss": 0.337,
+      "step": 24000
+    },
+    {
+      "epoch": 2.57,
+      "eval_loss": 0.49896007776260376,
+      "eval_runtime": 48.9304,
+      "eval_samples_per_second": 41.426,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.4311247517602455,
+      "step": 24000
+    },
+    {
+      "epoch": 2.58,
+      "eval_loss": 0.49622592329978943,
+      "eval_runtime": 48.829,
+      "eval_samples_per_second": 41.512,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.43293013179274237,
+      "step": 24100
+    },
+    {
+      "epoch": 2.59,
+      "eval_loss": 0.5032628178596497,
+      "eval_runtime": 48.9877,
+      "eval_samples_per_second": 41.378,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.4288680267196245,
+      "step": 24200
+    },
+    {
+      "epoch": 2.6,
+      "eval_loss": 0.5260281562805176,
+      "eval_runtime": 48.83,
+      "eval_samples_per_second": 41.511,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.4534211951615815,
+      "step": 24300
+    },
+    {
+      "epoch": 2.61,
+      "eval_loss": 0.5309225916862488,
+      "eval_runtime": 48.8251,
+      "eval_samples_per_second": 41.516,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.4441234879942228,
+      "step": 24400
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.000248445558739255,
+      "loss": 0.3393,
+      "step": 24500
+    },
+    {
+      "epoch": 2.62,
+      "eval_loss": 0.5131687521934509,
+      "eval_runtime": 48.6954,
+      "eval_samples_per_second": 41.626,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.43455497382198954,
+      "step": 24500
+    },
+    {
+      "epoch": 2.63,
+      "eval_loss": 0.5188684463500977,
+      "eval_runtime": 48.6759,
+      "eval_samples_per_second": 41.643,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.42327134861888427,
+      "step": 24600
+    },
+    {
+      "epoch": 2.64,
+      "eval_loss": 0.5074316263198853,
+      "eval_runtime": 48.5643,
+      "eval_samples_per_second": 41.738,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.432569055786243,
+      "step": 24700
+    },
+    {
+      "epoch": 2.66,
+      "eval_loss": 0.5110915899276733,
+      "eval_runtime": 48.9356,
+      "eval_samples_per_second": 41.422,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.4254378046578805,
+      "step": 24800
+    },
+    {
+      "epoch": 2.67,
+      "eval_loss": 0.4932540953159332,
+      "eval_runtime": 49.84,
+      "eval_samples_per_second": 40.67,
+      "eval_steps_per_second": 5.096,
+      "eval_wer": 0.4254378046578805,
+      "step": 24900
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.0002473710601719197,
+      "loss": 0.3334,
+      "step": 25000
+    },
+    {
+      "epoch": 2.68,
+      "eval_loss": 0.5045706629753113,
+      "eval_runtime": 48.7805,
+      "eval_samples_per_second": 41.553,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.44069326593247876,
+      "step": 25000
+    },
+    {
+      "epoch": 2.69,
+      "eval_loss": 0.500957190990448,
+      "eval_runtime": 48.5639,
+      "eval_samples_per_second": 41.739,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.4404224589276043,
+      "step": 25100
+    },
+    {
+      "epoch": 2.7,
+      "eval_loss": 0.5044623017311096,
+      "eval_runtime": 48.5838,
+      "eval_samples_per_second": 41.722,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.42363242462538364,
+      "step": 25200
+    },
+    {
+      "epoch": 2.71,
+      "eval_loss": 0.4938253164291382,
+      "eval_runtime": 49.0033,
+      "eval_samples_per_second": 41.365,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.43049286874887166,
+      "step": 25300
+    },
+    {
+      "epoch": 2.72,
+      "eval_loss": 0.5020812153816223,
+      "eval_runtime": 48.9392,
+      "eval_samples_per_second": 41.419,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.4383462718902329,
+      "step": 25400
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0002462965616045845,
+      "loss": 0.3366,
+      "step": 25500
+    },
+    {
+      "epoch": 2.73,
+      "eval_loss": 0.4953358471393585,
+      "eval_runtime": 48.8378,
+      "eval_samples_per_second": 41.505,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.42020220256363966,
+      "step": 25500
+    },
+    {
+      "epoch": 2.74,
+      "eval_loss": 0.4984549880027771,
+      "eval_runtime": 48.6616,
+      "eval_samples_per_second": 41.655,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.4338328218089908,
+      "step": 25600
+    },
+    {
+      "epoch": 2.75,
+      "eval_loss": 0.4765419065952301,
+      "eval_runtime": 48.8911,
+      "eval_samples_per_second": 41.46,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.4161400974905218,
+      "step": 25700
+    },
+    {
+      "epoch": 2.76,
+      "eval_loss": 0.4873102903366089,
+      "eval_runtime": 49.0181,
+      "eval_samples_per_second": 41.352,
+      "eval_steps_per_second": 5.182,
+      "eval_wer": 0.42922910272612386,
+      "step": 25800
+    },
+    {
+      "epoch": 2.77,
+      "eval_loss": 0.49981385469436646,
+      "eval_runtime": 48.989,
+      "eval_samples_per_second": 41.377,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.41893843654089186,
+      "step": 25900
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0002452242120343839,
+      "loss": 0.3359,
+      "step": 26000
+    },
+    {
+      "epoch": 2.78,
+      "eval_loss": 0.49909472465515137,
+      "eval_runtime": 49.0099,
+      "eval_samples_per_second": 41.359,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.4248059216465066,
+      "step": 26000
+    },
+    {
+      "epoch": 2.79,
+      "eval_loss": 0.5011939406394958,
+      "eval_runtime": 49.1474,
+      "eval_samples_per_second": 41.243,
+      "eval_steps_per_second": 5.168,
+      "eval_wer": 0.43067340675212135,
+      "step": 26100
+    },
+    {
+      "epoch": 2.81,
+      "eval_loss": 0.5080650448799133,
+      "eval_runtime": 48.8435,
+      "eval_samples_per_second": 41.5,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.4151471384726485,
+      "step": 26200
+    },
+    {
+      "epoch": 2.82,
+      "eval_loss": 0.4996890127658844,
+      "eval_runtime": 48.7317,
+      "eval_samples_per_second": 41.595,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.43049286874887166,
+      "step": 26300
+    },
+    {
+      "epoch": 2.83,
+      "eval_loss": 0.49694108963012695,
+      "eval_runtime": 49.1204,
+      "eval_samples_per_second": 41.266,
+      "eval_steps_per_second": 5.171,
+      "eval_wer": 0.43022206174399713,
+      "step": 26400
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0002441497134670487,
+      "loss": 0.3396,
+      "step": 26500
+    },
+    {
+      "epoch": 2.84,
+      "eval_loss": 0.478350967168808,
+      "eval_runtime": 48.8451,
+      "eval_samples_per_second": 41.499,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.4270626466871276,
+      "step": 26500
+    },
+    {
+      "epoch": 2.85,
+      "eval_loss": 0.4804205298423767,
+      "eval_runtime": 48.7331,
+      "eval_samples_per_second": 41.594,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.414876331467774,
+      "step": 26600
+    },
+    {
+      "epoch": 2.86,
+      "eval_loss": 0.48998621106147766,
+      "eval_runtime": 49.5501,
+      "eval_samples_per_second": 40.908,
+      "eval_steps_per_second": 5.126,
+      "eval_wer": 0.4192092435457664,
+      "step": 26700
+    },
+    {
+      "epoch": 2.87,
+      "eval_loss": 0.5043953061103821,
+      "eval_runtime": 49.2804,
+      "eval_samples_per_second": 41.132,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.43247878678461815,
+      "step": 26800
+    },
+    {
+      "epoch": 2.88,
+      "eval_loss": 0.4934612512588501,
+      "eval_runtime": 48.9475,
+      "eval_samples_per_second": 41.412,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.43762411987723415,
+      "step": 26900
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.00024307521489971343,
+      "loss": 0.3356,
+      "step": 27000
+    },
+    {
+      "epoch": 2.89,
+      "eval_loss": 0.5006960034370422,
+      "eval_runtime": 48.7024,
+      "eval_samples_per_second": 41.62,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.42688210868387794,
+      "step": 27000
+    },
+    {
+      "epoch": 2.9,
+      "eval_loss": 0.48872122168540955,
+      "eval_runtime": 48.7611,
+      "eval_samples_per_second": 41.57,
+      "eval_steps_per_second": 5.209,
+      "eval_wer": 0.4177649395197689,
+      "step": 27100
+    },
+    {
+      "epoch": 2.91,
+      "eval_loss": 0.47697293758392334,
+      "eval_runtime": 48.9744,
+      "eval_samples_per_second": 41.389,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.4169525185051453,
+      "step": 27200
+    },
+    {
+      "epoch": 2.92,
+      "eval_loss": 0.48469677567481995,
+      "eval_runtime": 48.5931,
+      "eval_samples_per_second": 41.714,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.4166817115002708,
+      "step": 27300
+    },
+    {
+      "epoch": 2.93,
+      "eval_loss": 0.4861487150192261,
+      "eval_runtime": 49.0078,
+      "eval_samples_per_second": 41.361,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.4138833724499007,
+      "step": 27400
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.00024200071633237818,
+      "loss": 0.3395,
+      "step": 27500
+    },
+    {
+      "epoch": 2.94,
+      "eval_loss": 0.4975493848323822,
+      "eval_runtime": 48.6597,
+      "eval_samples_per_second": 41.657,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.429138833724499,
+      "step": 27500
+    },
+    {
+      "epoch": 2.95,
+      "eval_loss": 0.5055610537528992,
+      "eval_runtime": 51.3239,
+      "eval_samples_per_second": 39.494,
+      "eval_steps_per_second": 4.949,
+      "eval_wer": 0.44710236504784256,
+      "step": 27600
+    },
+    {
+      "epoch": 2.97,
+      "eval_loss": 0.5111228823661804,
+      "eval_runtime": 50.7513,
+      "eval_samples_per_second": 39.94,
+      "eval_steps_per_second": 5.005,
+      "eval_wer": 0.4375338508756093,
+      "step": 27700
+    },
+    {
+      "epoch": 2.98,
+      "eval_loss": 0.5326820611953735,
+      "eval_runtime": 48.8673,
+      "eval_samples_per_second": 41.48,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.4576638382379491,
+      "step": 27800
+    },
+    {
+      "epoch": 2.99,
+      "eval_loss": 0.506744384765625,
+      "eval_runtime": 49.0908,
+      "eval_samples_per_second": 41.291,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.43933923090810617,
+      "step": 27900
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.00024092836676217763,
+      "loss": 0.3332,
+      "step": 28000
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 0.48983654379844666,
+      "eval_runtime": 48.9912,
+      "eval_samples_per_second": 41.375,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.41875789853764217,
+      "step": 28000
+    },
+    {
+      "epoch": 3.01,
+      "eval_loss": 0.4790142774581909,
+      "eval_runtime": 48.9473,
+      "eval_samples_per_second": 41.412,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.40927965336703376,
+      "step": 28100
+    },
+    {
+      "epoch": 3.02,
+      "eval_loss": 0.48284170031547546,
+      "eval_runtime": 49.4292,
+      "eval_samples_per_second": 41.008,
+      "eval_steps_per_second": 5.139,
+      "eval_wer": 0.42020220256363966,
+      "step": 28200
+    },
+    {
+      "epoch": 3.03,
+      "eval_loss": 0.48360347747802734,
+      "eval_runtime": 48.8613,
+      "eval_samples_per_second": 41.485,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.41460552446289944,
+      "step": 28300
+    },
+    {
+      "epoch": 3.04,
+      "eval_loss": 0.49010956287384033,
+      "eval_runtime": 49.2751,
+      "eval_samples_per_second": 41.136,
+      "eval_steps_per_second": 5.155,
+      "eval_wer": 0.4241740386351327,
+      "step": 28400
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.00023985386819484238,
+      "loss": 0.2984,
+      "step": 28500
+    },
+    {
+      "epoch": 3.05,
+      "eval_loss": 0.47723203897476196,
+      "eval_runtime": 49.0961,
+      "eval_samples_per_second": 41.286,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.41180718541252936,
+      "step": 28500
+    },
+    {
+      "epoch": 3.06,
+      "eval_loss": 0.5054998397827148,
+      "eval_runtime": 49.4055,
+      "eval_samples_per_second": 41.028,
+      "eval_steps_per_second": 5.141,
+      "eval_wer": 0.4212854305831378,
+      "step": 28600
+    },
+    {
+      "epoch": 3.07,
+      "eval_loss": 0.4910781979560852,
+      "eval_runtime": 49.1346,
+      "eval_samples_per_second": 41.254,
+      "eval_steps_per_second": 5.169,
+      "eval_wer": 0.4100018053800325,
+      "step": 28700
+    },
+    {
+      "epoch": 3.08,
+      "eval_loss": 0.47373151779174805,
+      "eval_runtime": 49.1679,
+      "eval_samples_per_second": 41.226,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.4087380393572847,
+      "step": 28800
+    },
+    {
+      "epoch": 3.09,
+      "eval_loss": 0.49302175641059875,
+      "eval_runtime": 48.9528,
+      "eval_samples_per_second": 41.407,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.4216465065896371,
+      "step": 28900
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00023877936962750715,
+      "loss": 0.3056,
+      "step": 29000
+    },
+    {
+      "epoch": 3.1,
+      "eval_loss": 0.47357192635536194,
+      "eval_runtime": 49.4511,
+      "eval_samples_per_second": 40.99,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.41090449539628093,
+      "step": 29000
+    },
+    {
+      "epoch": 3.12,
+      "eval_loss": 0.4862995743751526,
+      "eval_runtime": 49.3674,
+      "eval_samples_per_second": 41.059,
+      "eval_steps_per_second": 5.145,
+      "eval_wer": 0.4058494313052898,
+      "step": 29100
+    },
+    {
+      "epoch": 3.13,
+      "eval_loss": 0.47840672731399536,
+      "eval_runtime": 50.0693,
+      "eval_samples_per_second": 40.484,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.4183968225311428,
+      "step": 29200
+    },
+    {
+      "epoch": 3.14,
+      "eval_loss": 0.49233099818229675,
+      "eval_runtime": 49.5351,
+      "eval_samples_per_second": 40.921,
+      "eval_steps_per_second": 5.128,
+      "eval_wer": 0.423993500631883,
+      "step": 29300
+    },
+    {
+      "epoch": 3.15,
+      "eval_loss": 0.4845847189426422,
+      "eval_runtime": 49.6836,
+      "eval_samples_per_second": 40.798,
+      "eval_steps_per_second": 5.112,
+      "eval_wer": 0.42263946560751037,
+      "step": 29400
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0002377048710601719,
+      "loss": 0.2995,
+      "step": 29500
+    },
+    {
+      "epoch": 3.16,
+      "eval_loss": 0.4829016923904419,
+      "eval_runtime": 48.9864,
+      "eval_samples_per_second": 41.379,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.408557501354035,
+      "step": 29500
+    },
+    {
+      "epoch": 3.17,
+      "eval_loss": 0.4933793544769287,
+      "eval_runtime": 49.2945,
+      "eval_samples_per_second": 41.12,
+      "eval_steps_per_second": 5.153,
+      "eval_wer": 0.423993500631883,
+      "step": 29600
+    },
+    {
+      "epoch": 3.18,
+      "eval_loss": 0.48932012915611267,
+      "eval_runtime": 49.1056,
+      "eval_samples_per_second": 41.278,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.41523740747427335,
+      "step": 29700
+    },
+    {
+      "epoch": 3.19,
+      "eval_loss": 0.47299858927726746,
+      "eval_runtime": 48.9953,
+      "eval_samples_per_second": 41.371,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.4227297346091352,
+      "step": 29800
+    },
+    {
+      "epoch": 3.2,
+      "eval_loss": 0.5027060508728027,
+      "eval_runtime": 50.4849,
+      "eval_samples_per_second": 40.151,
+      "eval_steps_per_second": 5.031,
+      "eval_wer": 0.4330204007943672,
+      "step": 29900
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.00023663037249283664,
+      "loss": 0.2926,
+      "step": 30000
+    },
+    {
+      "epoch": 3.21,
+      "eval_loss": 0.4902673363685608,
+      "eval_runtime": 107.9623,
+      "eval_samples_per_second": 18.775,
+      "eval_steps_per_second": 2.353,
+      "eval_wer": 0.41117530240115546,
+      "step": 30000
+    },
+    {
+      "epoch": 3.22,
+      "eval_loss": 0.49610400199890137,
+      "eval_runtime": 50.4886,
+      "eval_samples_per_second": 40.148,
+      "eval_steps_per_second": 5.031,
+      "eval_wer": 0.41568875248239756,
+      "step": 30100
+    },
+    {
+      "epoch": 3.23,
+      "eval_loss": 0.4980136454105377,
+      "eval_runtime": 51.2384,
+      "eval_samples_per_second": 39.56,
+      "eval_steps_per_second": 4.957,
+      "eval_wer": 0.42688210868387794,
+      "step": 30200
+    },
+    {
+      "epoch": 3.24,
+      "eval_loss": 0.48962903022766113,
+      "eval_runtime": 50.3961,
+      "eval_samples_per_second": 40.221,
+      "eval_steps_per_second": 5.04,
+      "eval_wer": 0.4126196064271529,
+      "step": 30300
+    },
+    {
+      "epoch": 3.25,
+      "eval_loss": 0.47261691093444824,
+      "eval_runtime": 50.6222,
+      "eval_samples_per_second": 40.042,
+      "eval_steps_per_second": 5.018,
+      "eval_wer": 0.40621050731178915,
+      "step": 30400
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.00023555587392550141,
+      "loss": 0.301,
+      "step": 30500
+    },
+    {
+      "epoch": 3.27,
+      "eval_loss": 0.47325438261032104,
+      "eval_runtime": 50.4878,
+      "eval_samples_per_second": 40.148,
+      "eval_steps_per_second": 5.031,
+      "eval_wer": 0.39853764217367754,
+      "step": 30500
+    },
+    {
+      "epoch": 3.28,
+      "eval_loss": 0.47721266746520996,
+      "eval_runtime": 50.584,
+      "eval_samples_per_second": 40.072,
+      "eval_steps_per_second": 5.021,
+      "eval_wer": 0.4046759342841668,
+      "step": 30600
+    },
+    {
+      "epoch": 3.29,
+      "eval_loss": 0.48059606552124023,
+      "eval_runtime": 50.9656,
+      "eval_samples_per_second": 39.772,
+      "eval_steps_per_second": 4.984,
+      "eval_wer": 0.40819642534753564,
+      "step": 30700
+    },
+    {
+      "epoch": 3.3,
+      "eval_loss": 0.4683433771133423,
+      "eval_runtime": 50.2652,
+      "eval_samples_per_second": 40.326,
+      "eval_steps_per_second": 5.053,
+      "eval_wer": 0.40106517421917315,
+      "step": 30800
+    },
+    {
+      "epoch": 3.31,
+      "eval_loss": 0.4774733781814575,
+      "eval_runtime": 50.7544,
+      "eval_samples_per_second": 39.937,
+      "eval_steps_per_second": 5.004,
+      "eval_wer": 0.4079256183426611,
+      "step": 30900
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00023448352435530084,
+      "loss": 0.2933,
+      "step": 31000
+    },
+    {
+      "epoch": 3.32,
+      "eval_loss": 0.47292542457580566,
+      "eval_runtime": 49.9791,
+      "eval_samples_per_second": 40.557,
+      "eval_steps_per_second": 5.082,
+      "eval_wer": 0.4082866943491605,
+      "step": 31000
+    },
+    {
+      "epoch": 3.33,
+      "eval_loss": 0.4628405272960663,
+      "eval_runtime": 49.7388,
+      "eval_samples_per_second": 40.753,
+      "eval_steps_per_second": 5.107,
+      "eval_wer": 0.4016067882289222,
+      "step": 31100
+    },
+    {
+      "epoch": 3.34,
+      "eval_loss": 0.475334107875824,
+      "eval_runtime": 49.8991,
+      "eval_samples_per_second": 40.622,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.4192092435457664,
+      "step": 31200
+    },
+    {
+      "epoch": 3.35,
+      "eval_loss": 0.46871381998062134,
+      "eval_runtime": 50.0989,
+      "eval_samples_per_second": 40.46,
+      "eval_steps_per_second": 5.07,
+      "eval_wer": 0.41848709153276764,
+      "step": 31300
+    },
+    {
+      "epoch": 3.36,
+      "eval_loss": 0.4805503487586975,
+      "eval_runtime": 50.1378,
+      "eval_samples_per_second": 40.429,
+      "eval_steps_per_second": 5.066,
+      "eval_wer": 0.4106336883914064,
+      "step": 31400
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.0002334090257879656,
+      "loss": 0.2957,
+      "step": 31500
+    },
+    {
+      "epoch": 3.37,
+      "eval_loss": 0.48885583877563477,
+      "eval_runtime": 49.7773,
+      "eval_samples_per_second": 40.721,
+      "eval_steps_per_second": 5.103,
+      "eval_wer": 0.423993500631883,
+      "step": 31500
+    },
+    {
+      "epoch": 3.38,
+      "eval_loss": 0.48816147446632385,
+      "eval_runtime": 49.8971,
+      "eval_samples_per_second": 40.624,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.4182162845278931,
+      "step": 31600
+    },
+    {
+      "epoch": 3.39,
+      "eval_loss": 0.4797750413417816,
+      "eval_runtime": 49.755,
+      "eval_samples_per_second": 40.74,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.4162303664921466,
+      "step": 31700
+    },
+    {
+      "epoch": 3.4,
+      "eval_loss": 0.47179311513900757,
+      "eval_runtime": 49.9097,
+      "eval_samples_per_second": 40.613,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.4108142263946561,
+      "step": 31800
+    },
+    {
+      "epoch": 3.42,
+      "eval_loss": 0.4685176610946655,
+      "eval_runtime": 49.6497,
+      "eval_samples_per_second": 40.826,
+      "eval_steps_per_second": 5.116,
+      "eval_wer": 0.41009207438165735,
+      "step": 31900
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.00023233452722063035,
+      "loss": 0.3039,
+      "step": 32000
+    },
+    {
+      "epoch": 3.43,
+      "eval_loss": 0.4815961420536041,
+      "eval_runtime": 49.8792,
+      "eval_samples_per_second": 40.638,
+      "eval_steps_per_second": 5.092,
+      "eval_wer": 0.418848167539267,
+      "step": 32000
+    },
+    {
+      "epoch": 3.44,
+      "eval_loss": 0.48744821548461914,
+      "eval_runtime": 49.9648,
+      "eval_samples_per_second": 40.569,
+      "eval_steps_per_second": 5.084,
+      "eval_wer": 0.4138833724499007,
+      "step": 32100
+    },
+    {
+      "epoch": 3.45,
+      "eval_loss": 0.4899212718009949,
+      "eval_runtime": 50.0433,
+      "eval_samples_per_second": 40.505,
+      "eval_steps_per_second": 5.076,
+      "eval_wer": 0.41153637840765483,
+      "step": 32200
+    },
+    {
+      "epoch": 3.46,
+      "eval_loss": 0.4851779043674469,
+      "eval_runtime": 49.6408,
+      "eval_samples_per_second": 40.833,
+      "eval_steps_per_second": 5.117,
+      "eval_wer": 0.4180357465246434,
+      "step": 32300
+    },
+    {
+      "epoch": 3.47,
+      "eval_loss": 0.5073689222335815,
+      "eval_runtime": 50.1621,
+      "eval_samples_per_second": 40.409,
+      "eval_steps_per_second": 5.064,
+      "eval_wer": 0.4128904134320274,
+      "step": 32400
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0002312600286532951,
+      "loss": 0.3006,
+      "step": 32500
+    },
+    {
+      "epoch": 3.48,
+      "eval_loss": 0.4837365448474884,
+      "eval_runtime": 49.6611,
+      "eval_samples_per_second": 40.817,
+      "eval_steps_per_second": 5.115,
+      "eval_wer": 0.40756454233616174,
+      "step": 32500
+    },
+    {
+      "epoch": 3.49,
+      "eval_loss": 0.49273020029067993,
+      "eval_runtime": 50.0235,
+      "eval_samples_per_second": 40.521,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.4098212673767828,
+      "step": 32600
+    },
+    {
+      "epoch": 3.5,
+      "eval_loss": 0.49989503622055054,
+      "eval_runtime": 50.1352,
+      "eval_samples_per_second": 40.431,
+      "eval_steps_per_second": 5.066,
+      "eval_wer": 0.41722332551001984,
+      "step": 32700
+    },
+    {
+      "epoch": 3.51,
+      "eval_loss": 0.47733768820762634,
+      "eval_runtime": 50.118,
+      "eval_samples_per_second": 40.445,
+      "eval_steps_per_second": 5.068,
+      "eval_wer": 0.41938978154901607,
+      "step": 32800
+    },
+    {
+      "epoch": 3.52,
+      "eval_loss": 0.4858909845352173,
+      "eval_runtime": 49.9497,
+      "eval_samples_per_second": 40.581,
+      "eval_steps_per_second": 5.085,
+      "eval_wer": 0.40575916230366493,
+      "step": 32900
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00023018553008595987,
+      "loss": 0.3089,
+      "step": 33000
+    },
+    {
+      "epoch": 3.53,
+      "eval_loss": 0.47827592492103577,
+      "eval_runtime": 50.8869,
+      "eval_samples_per_second": 39.833,
+      "eval_steps_per_second": 4.991,
+      "eval_wer": 0.4103628813865319,
+      "step": 33000
+    },
+    {
+      "epoch": 3.54,
+      "eval_loss": 0.46217525005340576,
+      "eval_runtime": 49.9647,
+      "eval_samples_per_second": 40.569,
+      "eval_steps_per_second": 5.084,
+      "eval_wer": 0.4019678642354216,
+      "step": 33100
+    },
+    {
+      "epoch": 3.55,
+      "eval_loss": 0.4839652180671692,
+      "eval_runtime": 50.4824,
+      "eval_samples_per_second": 40.153,
+      "eval_steps_per_second": 5.031,
+      "eval_wer": 0.4064813143166637,
+      "step": 33200
+    },
+    {
+      "epoch": 3.57,
+      "eval_loss": 0.47563862800598145,
+      "eval_runtime": 50.0519,
+      "eval_samples_per_second": 40.498,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.42408376963350786,
+      "step": 33300
+    },
+    {
+      "epoch": 3.58,
+      "eval_loss": 0.4830974340438843,
+      "eval_runtime": 50.2796,
+      "eval_samples_per_second": 40.315,
+      "eval_steps_per_second": 5.052,
+      "eval_wer": 0.41704278750677015,
+      "step": 33400
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.00022911103151862462,
+      "loss": 0.3061,
+      "step": 33500
+    },
+    {
+      "epoch": 3.59,
+      "eval_loss": 0.47941431403160095,
+      "eval_runtime": 49.7111,
+      "eval_samples_per_second": 40.776,
+      "eval_steps_per_second": 5.11,
+      "eval_wer": 0.4067521213215382,
+      "step": 33500
+    },
+    {
+      "epoch": 3.6,
+      "eval_loss": 0.47304731607437134,
+      "eval_runtime": 50.0694,
+      "eval_samples_per_second": 40.484,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.40368297526629354,
+      "step": 33600
+    },
+    {
+      "epoch": 3.61,
+      "eval_loss": 0.4808448851108551,
+      "eval_runtime": 49.8945,
+      "eval_samples_per_second": 40.626,
+      "eval_steps_per_second": 5.091,
+      "eval_wer": 0.41379310344827586,
+      "step": 33700
+    },
+    {
+      "epoch": 3.62,
+      "eval_loss": 0.4924474060535431,
+      "eval_runtime": 50.3079,
+      "eval_samples_per_second": 40.292,
+      "eval_steps_per_second": 5.049,
+      "eval_wer": 0.4248059216465066,
+      "step": 33800
+    },
+    {
+      "epoch": 3.63,
+      "eval_loss": 0.4748767614364624,
+      "eval_runtime": 49.9339,
+      "eval_samples_per_second": 40.594,
+      "eval_steps_per_second": 5.087,
+      "eval_wer": 0.41117530240115546,
+      "step": 33900
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.0002280365329512894,
+      "loss": 0.3047,
+      "step": 34000
+    },
+    {
+      "epoch": 3.64,
+      "eval_loss": 0.49240487813949585,
+      "eval_runtime": 50.046,
+      "eval_samples_per_second": 40.503,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.432569055786243,
+      "step": 34000
+    },
+    {
+      "epoch": 3.65,
+      "eval_loss": 0.4745273292064667,
+      "eval_runtime": 49.8993,
+      "eval_samples_per_second": 40.622,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.4103628813865319,
+      "step": 34100
+    },
+    {
+      "epoch": 3.66,
+      "eval_loss": 0.4759696125984192,
+      "eval_runtime": 50.0617,
+      "eval_samples_per_second": 40.49,
+      "eval_steps_per_second": 5.074,
+      "eval_wer": 0.4122585304206535,
+      "step": 34200
+    },
+    {
+      "epoch": 3.67,
+      "eval_loss": 0.47877970337867737,
+      "eval_runtime": 49.9144,
+      "eval_samples_per_second": 40.61,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.4065715833182885,
+      "step": 34300
+    },
+    {
+      "epoch": 3.68,
+      "eval_loss": 0.4626765549182892,
+      "eval_runtime": 50.4559,
+      "eval_samples_per_second": 40.174,
+      "eval_steps_per_second": 5.034,
+      "eval_wer": 0.4157790214840224,
+      "step": 34400
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.00022696203438395414,
+      "loss": 0.3042,
+      "step": 34500
+    },
+    {
+      "epoch": 3.69,
+      "eval_loss": 0.4974145293235779,
+      "eval_runtime": 49.7345,
+      "eval_samples_per_second": 40.756,
+      "eval_steps_per_second": 5.107,
+      "eval_wer": 0.4130709514352771,
+      "step": 34500
+    },
+    {
+      "epoch": 3.7,
+      "eval_loss": 0.4593214988708496,
+      "eval_runtime": 49.9902,
+      "eval_samples_per_second": 40.548,
+      "eval_steps_per_second": 5.081,
+      "eval_wer": 0.406300776313414,
+      "step": 34600
+    },
+    {
+      "epoch": 3.72,
+      "eval_loss": 0.4549331068992615,
+      "eval_runtime": 49.7975,
+      "eval_samples_per_second": 40.705,
+      "eval_steps_per_second": 5.101,
+      "eval_wer": 0.3927604260696877,
+      "step": 34700
+    },
+    {
+      "epoch": 3.73,
+      "eval_loss": 0.4690437614917755,
+      "eval_runtime": 50.7669,
+      "eval_samples_per_second": 39.928,
+      "eval_steps_per_second": 5.003,
+      "eval_wer": 0.3897815490160679,
+      "step": 34800
+    },
+    {
+      "epoch": 3.74,
+      "eval_loss": 0.4559744596481323,
+      "eval_runtime": 49.4008,
+      "eval_samples_per_second": 41.032,
+      "eval_steps_per_second": 5.142,
+      "eval_wer": 0.4007040982126738,
+      "step": 34900
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.0002258875358166189,
+      "loss": 0.2963,
+      "step": 35000
+    },
+    {
+      "epoch": 3.75,
+      "eval_loss": 0.46059346199035645,
+      "eval_runtime": 49.637,
+      "eval_samples_per_second": 40.836,
+      "eval_steps_per_second": 5.117,
+      "eval_wer": 0.39591984112655715,
+      "step": 35000
+    },
+    {
+      "epoch": 3.76,
+      "eval_loss": 0.4762277901172638,
+      "eval_runtime": 50.0278,
+      "eval_samples_per_second": 40.517,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.4056688933020401,
+      "step": 35100
+    },
+    {
+      "epoch": 3.77,
+      "eval_loss": 0.4749699831008911,
+      "eval_runtime": 49.7739,
+      "eval_samples_per_second": 40.724,
+      "eval_steps_per_second": 5.103,
+      "eval_wer": 0.403412168261419,
+      "step": 35200
+    },
+    {
+      "epoch": 3.78,
+      "eval_loss": 0.47716274857521057,
+      "eval_runtime": 49.6995,
+      "eval_samples_per_second": 40.785,
+      "eval_steps_per_second": 5.111,
+      "eval_wer": 0.41144610940603,
+      "step": 35300
+    },
+    {
+      "epoch": 3.79,
+      "eval_loss": 0.4668501615524292,
+      "eval_runtime": 49.7899,
+      "eval_samples_per_second": 40.711,
+      "eval_steps_per_second": 5.101,
+      "eval_wer": 0.3995306011915508,
+      "step": 35400
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.00022481518624641833,
+      "loss": 0.3012,
+      "step": 35500
+    },
+    {
+      "epoch": 3.8,
+      "eval_loss": 0.47086089849472046,
+      "eval_runtime": 49.7381,
+      "eval_samples_per_second": 40.753,
+      "eval_steps_per_second": 5.107,
+      "eval_wer": 0.40900884636215923,
+      "step": 35500
+    },
+    {
+      "epoch": 3.81,
+      "eval_loss": 0.47223708033561707,
+      "eval_runtime": 49.8147,
+      "eval_samples_per_second": 40.691,
+      "eval_steps_per_second": 5.099,
+      "eval_wer": 0.41234879942227837,
+      "step": 35600
+    },
+    {
+      "epoch": 3.82,
+      "eval_loss": 0.49128904938697815,
+      "eval_runtime": 50.3648,
+      "eval_samples_per_second": 40.246,
+      "eval_steps_per_second": 5.043,
+      "eval_wer": 0.41650117349702115,
+      "step": 35700
+    },
+    {
+      "epoch": 3.83,
+      "eval_loss": 0.48138895630836487,
+      "eval_runtime": 51.0291,
+      "eval_samples_per_second": 39.722,
+      "eval_steps_per_second": 4.978,
+      "eval_wer": 0.406300776313414,
+      "step": 35800
+    },
+    {
+      "epoch": 3.84,
+      "eval_loss": 0.4868621230125427,
+      "eval_runtime": 49.5664,
+      "eval_samples_per_second": 40.895,
+      "eval_steps_per_second": 5.124,
+      "eval_wer": 0.417133056508395,
+      "step": 35900
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.00022374068767908305,
+      "loss": 0.3015,
+      "step": 36000
+    },
+    {
+      "epoch": 3.85,
+      "eval_loss": 0.4791094660758972,
+      "eval_runtime": 50.1246,
+      "eval_samples_per_second": 40.439,
+      "eval_steps_per_second": 5.067,
+      "eval_wer": 0.4059397003069146,
+      "step": 36000
+    },
+    {
+      "epoch": 3.87,
+      "eval_loss": 0.45353808999061584,
+      "eval_runtime": 49.889,
+      "eval_samples_per_second": 40.63,
+      "eval_steps_per_second": 5.091,
+      "eval_wer": 0.3976349521574291,
+      "step": 36100
+    },
+    {
+      "epoch": 3.88,
+      "eval_loss": 0.4706454575061798,
+      "eval_runtime": 50.016,
+      "eval_samples_per_second": 40.527,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.40088463621592346,
+      "step": 36200
+    },
+    {
+      "epoch": 3.89,
+      "eval_loss": 0.4678768217563629,
+      "eval_runtime": 49.8077,
+      "eval_samples_per_second": 40.697,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.40124571222242283,
+      "step": 36300
+    },
+    {
+      "epoch": 3.9,
+      "eval_loss": 0.4735689163208008,
+      "eval_runtime": 50.0515,
+      "eval_samples_per_second": 40.498,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.40964072937353313,
+      "step": 36400
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.00022266618911174783,
+      "loss": 0.2965,
+      "step": 36500
+    },
+    {
+      "epoch": 3.91,
+      "eval_loss": 0.4755512475967407,
+      "eval_runtime": 49.592,
+      "eval_samples_per_second": 40.874,
+      "eval_steps_per_second": 5.122,
+      "eval_wer": 0.4106336883914064,
+      "step": 36500
+    },
+    {
+      "epoch": 3.92,
+      "eval_loss": 0.4669349789619446,
+      "eval_runtime": 50.1764,
+      "eval_samples_per_second": 40.397,
+      "eval_steps_per_second": 5.062,
+      "eval_wer": 0.40846723235241017,
+      "step": 36600
+    },
+    {
+      "epoch": 3.93,
+      "eval_loss": 0.47956836223602295,
+      "eval_runtime": 49.6711,
+      "eval_samples_per_second": 40.808,
+      "eval_steps_per_second": 5.114,
+      "eval_wer": 0.40539808629716556,
+      "step": 36700
+    },
+    {
+      "epoch": 3.94,
+      "eval_loss": 0.4583274722099304,
+      "eval_runtime": 49.5945,
+      "eval_samples_per_second": 40.871,
+      "eval_steps_per_second": 5.122,
+      "eval_wer": 0.39321177107781186,
+      "step": 36800
+    },
+    {
+      "epoch": 3.95,
+      "eval_loss": 0.4429765045642853,
+      "eval_runtime": 49.6156,
+      "eval_samples_per_second": 40.854,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.3969128001444304,
+      "step": 36900
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.00022159169054441257,
+      "loss": 0.2993,
+      "step": 37000
+    },
+    {
+      "epoch": 3.96,
+      "eval_loss": 0.45601701736450195,
+      "eval_runtime": 50.1949,
+      "eval_samples_per_second": 40.383,
+      "eval_steps_per_second": 5.06,
+      "eval_wer": 0.39140639104531505,
+      "step": 37000
+    },
+    {
+      "epoch": 3.97,
+      "eval_loss": 0.4738839864730835,
+      "eval_runtime": 49.7611,
+      "eval_samples_per_second": 40.735,
+      "eval_steps_per_second": 5.104,
+      "eval_wer": 0.4001624842029247,
+      "step": 37100
+    },
+    {
+      "epoch": 3.98,
+      "eval_loss": 0.45979049801826477,
+      "eval_runtime": 50.036,
+      "eval_samples_per_second": 40.511,
+      "eval_steps_per_second": 5.076,
+      "eval_wer": 0.39122585304206536,
+      "step": 37200
+    },
+    {
+      "epoch": 3.99,
+      "eval_loss": 0.46073099970817566,
+      "eval_runtime": 49.9116,
+      "eval_samples_per_second": 40.612,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.3906842390323163,
+      "step": 37300
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 0.4708978235721588,
+      "eval_runtime": 50.1501,
+      "eval_samples_per_second": 40.419,
+      "eval_steps_per_second": 5.065,
+      "eval_wer": 0.3986279111753024,
+      "step": 37400
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.000220519340974212,
+      "loss": 0.2886,
+      "step": 37500
+    },
+    {
+      "epoch": 4.01,
+      "eval_loss": 0.46422716975212097,
+      "eval_runtime": 49.9196,
+      "eval_samples_per_second": 40.605,
+      "eval_steps_per_second": 5.088,
+      "eval_wer": 0.40666185231991336,
+      "step": 37500
+    },
+    {
+      "epoch": 4.03,
+      "eval_loss": 0.4683905839920044,
+      "eval_runtime": 50.3714,
+      "eval_samples_per_second": 40.241,
+      "eval_steps_per_second": 5.043,
+      "eval_wer": 0.3984473731720527,
+      "step": 37600
+    },
+    {
+      "epoch": 4.04,
+      "eval_loss": 0.4689880311489105,
+      "eval_runtime": 50.2882,
+      "eval_samples_per_second": 40.308,
+      "eval_steps_per_second": 5.051,
+      "eval_wer": 0.39790575916230364,
+      "step": 37700
+    },
+    {
+      "epoch": 4.05,
+      "eval_loss": 0.47223958373069763,
+      "eval_runtime": 50.6582,
+      "eval_samples_per_second": 40.013,
+      "eval_steps_per_second": 5.014,
+      "eval_wer": 0.3979960281639285,
+      "step": 37800
+    },
+    {
+      "epoch": 4.06,
+      "eval_loss": 0.47339245676994324,
+      "eval_runtime": 50.5606,
+      "eval_samples_per_second": 40.091,
+      "eval_steps_per_second": 5.024,
+      "eval_wer": 0.39267015706806285,
+      "step": 37900
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.00021944484240687677,
+      "loss": 0.2534,
+      "step": 38000
+    },
+    {
+      "epoch": 4.07,
+      "eval_loss": 0.4724343717098236,
+      "eval_runtime": 51.0309,
+      "eval_samples_per_second": 39.721,
+      "eval_steps_per_second": 4.977,
+      "eval_wer": 0.39880844917855207,
+      "step": 38000
+    },
+    {
+      "epoch": 4.08,
+      "eval_loss": 0.4664517343044281,
+      "eval_runtime": 51.1346,
+      "eval_samples_per_second": 39.641,
+      "eval_steps_per_second": 4.967,
+      "eval_wer": 0.3986279111753024,
+      "step": 38100
+    },
+    {
+      "epoch": 4.09,
+      "eval_loss": 0.4658842086791992,
+      "eval_runtime": 50.3243,
+      "eval_samples_per_second": 40.279,
+      "eval_steps_per_second": 5.047,
+      "eval_wer": 0.4035927062646687,
+      "step": 38200
+    },
+    {
+      "epoch": 4.1,
+      "eval_loss": 0.469433456659317,
+      "eval_runtime": 49.619,
+      "eval_samples_per_second": 40.851,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.3951976891135584,
+      "step": 38300
+    },
+    {
+      "epoch": 4.11,
+      "eval_loss": 0.47190093994140625,
+      "eval_runtime": 49.7683,
+      "eval_samples_per_second": 40.729,
+      "eval_steps_per_second": 5.104,
+      "eval_wer": 0.38905939700306913,
+      "step": 38400
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.0002183703438395415,
+      "loss": 0.2596,
+      "step": 38500
+    },
+    {
+      "epoch": 4.12,
+      "eval_loss": 0.46866583824157715,
+      "eval_runtime": 50.0111,
+      "eval_samples_per_second": 40.531,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.399440332189926,
+      "step": 38500
+    },
+    {
+      "epoch": 4.13,
+      "eval_loss": 0.47052884101867676,
+      "eval_runtime": 49.7745,
+      "eval_samples_per_second": 40.724,
+      "eval_steps_per_second": 5.103,
+      "eval_wer": 0.39032316302581693,
+      "step": 38600
+    },
+    {
+      "epoch": 4.14,
+      "eval_loss": 0.460104376077652,
+      "eval_runtime": 49.9336,
+      "eval_samples_per_second": 40.594,
+      "eval_steps_per_second": 5.087,
+      "eval_wer": 0.3974544141541795,
+      "step": 38700
+    },
+    {
+      "epoch": 4.15,
+      "eval_loss": 0.4666154384613037,
+      "eval_runtime": 50.032,
+      "eval_samples_per_second": 40.514,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.3970933381476801,
+      "step": 38800
+    },
+    {
+      "epoch": 4.16,
+      "eval_loss": 0.4772312641143799,
+      "eval_runtime": 49.9053,
+      "eval_samples_per_second": 40.617,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.3892399350063188,
+      "step": 38900
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.00021729799426934096,
+      "loss": 0.2643,
+      "step": 39000
+    },
+    {
+      "epoch": 4.18,
+      "eval_loss": 0.48102715611457825,
+      "eval_runtime": 49.7263,
+      "eval_samples_per_second": 40.763,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.4071131973280376,
+      "step": 39000
+    },
+    {
+      "epoch": 4.19,
+      "eval_loss": 0.49802666902542114,
+      "eval_runtime": 49.6686,
+      "eval_samples_per_second": 40.81,
+      "eval_steps_per_second": 5.114,
+      "eval_wer": 0.4166817115002708,
+      "step": 39100
+    },
+    {
+      "epoch": 4.2,
+      "eval_loss": 0.4657323956489563,
+      "eval_runtime": 49.7618,
+      "eval_samples_per_second": 40.734,
+      "eval_steps_per_second": 5.104,
+      "eval_wer": 0.39962087019317566,
+      "step": 39200
+    },
+    {
+      "epoch": 4.21,
+      "eval_loss": 0.48691385984420776,
+      "eval_runtime": 49.7124,
+      "eval_samples_per_second": 40.775,
+      "eval_steps_per_second": 5.109,
+      "eval_wer": 0.4001624842029247,
+      "step": 39300
+    },
+    {
+      "epoch": 4.22,
+      "eval_loss": 0.46561890840530396,
+      "eval_runtime": 49.9565,
+      "eval_samples_per_second": 40.575,
+      "eval_steps_per_second": 5.084,
+      "eval_wer": 0.3913161220436902,
+      "step": 39400
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.0002162234957020057,
+      "loss": 0.265,
+      "step": 39500
+    },
+    {
+      "epoch": 4.23,
+      "eval_loss": 0.47198420763015747,
+      "eval_runtime": 49.854,
+      "eval_samples_per_second": 40.659,
+      "eval_steps_per_second": 5.095,
+      "eval_wer": 0.3947463441054342,
+      "step": 39500
+    },
+    {
+      "epoch": 4.24,
+      "eval_loss": 0.4710632562637329,
+      "eval_runtime": 50.0216,
+      "eval_samples_per_second": 40.523,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.39700306914605527,
+      "step": 39600
+    },
+    {
+      "epoch": 4.25,
+      "eval_loss": 0.46885937452316284,
+      "eval_runtime": 50.8629,
+      "eval_samples_per_second": 39.852,
+      "eval_steps_per_second": 4.994,
+      "eval_wer": 0.3933020400794367,
+      "step": 39700
+    },
+    {
+      "epoch": 4.26,
+      "eval_loss": 0.47284629940986633,
+      "eval_runtime": 50.0447,
+      "eval_samples_per_second": 40.504,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.40169705723054705,
+      "step": 39800
+    },
+    {
+      "epoch": 4.27,
+      "eval_loss": 0.46733906865119934,
+      "eval_runtime": 49.6974,
+      "eval_samples_per_second": 40.787,
+      "eval_steps_per_second": 5.111,
+      "eval_wer": 0.3847264849250767,
+      "step": 39900
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.00021514899713467045,
+      "loss": 0.2644,
+      "step": 40000
+    },
+    {
+      "epoch": 4.28,
+      "eval_loss": 0.46359318494796753,
+      "eval_runtime": 49.5988,
+      "eval_samples_per_second": 40.868,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.396010110128182,
+      "step": 40000
+    },
+    {
+      "epoch": 4.29,
+      "eval_loss": 0.46991944313049316,
+      "eval_runtime": 50.3998,
+      "eval_samples_per_second": 40.218,
+      "eval_steps_per_second": 5.04,
+      "eval_wer": 0.38644159595594874,
+      "step": 40100
+    },
+    {
+      "epoch": 4.3,
+      "eval_loss": 0.4580465257167816,
+      "eval_runtime": 49.4799,
+      "eval_samples_per_second": 40.966,
+      "eval_steps_per_second": 5.133,
+      "eval_wer": 0.387434554973822,
+      "step": 40200
+    },
+    {
+      "epoch": 4.31,
+      "eval_loss": 0.47626662254333496,
+      "eval_runtime": 49.6505,
+      "eval_samples_per_second": 40.825,
+      "eval_steps_per_second": 5.116,
+      "eval_wer": 0.39510742011193356,
+      "step": 40300
+    },
+    {
+      "epoch": 4.33,
+      "eval_loss": 0.47517192363739014,
+      "eval_runtime": 49.3981,
+      "eval_samples_per_second": 41.034,
+      "eval_steps_per_second": 5.142,
+      "eval_wer": 0.4140639104531504,
+      "step": 40400
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.00021407449856733523,
+      "loss": 0.2633,
+      "step": 40500
+    },
+    {
+      "epoch": 4.34,
+      "eval_loss": 0.49175700545310974,
+      "eval_runtime": 49.6012,
+      "eval_samples_per_second": 40.866,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.39935006318830113,
+      "step": 40500
+    },
+    {
+      "epoch": 4.35,
+      "eval_loss": 0.4782505929470062,
+      "eval_runtime": 49.7601,
+      "eval_samples_per_second": 40.735,
+      "eval_steps_per_second": 5.104,
+      "eval_wer": 0.4025997472467954,
+      "step": 40600
+    },
+    {
+      "epoch": 4.36,
+      "eval_loss": 0.4738580286502838,
+      "eval_runtime": 50.9085,
+      "eval_samples_per_second": 39.817,
+      "eval_steps_per_second": 4.989,
+      "eval_wer": 0.403412168261419,
+      "step": 40700
+    },
+    {
+      "epoch": 4.37,
+      "eval_loss": 0.47499799728393555,
+      "eval_runtime": 49.9805,
+      "eval_samples_per_second": 40.556,
+      "eval_steps_per_second": 5.082,
+      "eval_wer": 0.39998194619967503,
+      "step": 40800
+    },
+    {
+      "epoch": 4.38,
+      "eval_loss": 0.46082910895347595,
+      "eval_runtime": 49.827,
+      "eval_samples_per_second": 40.681,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.39429499909731,
+      "step": 40900
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.00021299999999999997,
+      "loss": 0.2679,
+      "step": 41000
+    },
+    {
+      "epoch": 4.39,
+      "eval_loss": 0.46146294474601746,
+      "eval_runtime": 49.5199,
+      "eval_samples_per_second": 40.933,
+      "eval_steps_per_second": 5.129,
+      "eval_wer": 0.389149666004694,
+      "step": 41000
+    },
+    {
+      "epoch": 4.4,
+      "eval_loss": 0.47296470403671265,
+      "eval_runtime": 49.5896,
+      "eval_samples_per_second": 40.875,
+      "eval_steps_per_second": 5.122,
+      "eval_wer": 0.3984473731720527,
+      "step": 41100
+    },
+    {
+      "epoch": 4.41,
+      "eval_loss": 0.47276759147644043,
+      "eval_runtime": 49.8499,
+      "eval_samples_per_second": 40.662,
+      "eval_steps_per_second": 5.095,
+      "eval_wer": 0.40106517421917315,
+      "step": 41200
+    },
+    {
+      "epoch": 4.42,
+      "eval_loss": 0.46751782298088074,
+      "eval_runtime": 49.7874,
+      "eval_samples_per_second": 40.713,
+      "eval_steps_per_second": 5.102,
+      "eval_wer": 0.39321177107781186,
+      "step": 41300
+    },
+    {
+      "epoch": 4.43,
+      "eval_loss": 0.46617743372917175,
+      "eval_runtime": 50.1545,
+      "eval_samples_per_second": 40.415,
+      "eval_steps_per_second": 5.064,
+      "eval_wer": 0.39285069507131254,
+      "step": 41400
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.00021192550143266475,
+      "loss": 0.2682,
+      "step": 41500
+    },
+    {
+      "epoch": 4.44,
+      "eval_loss": 0.44897398352622986,
+      "eval_runtime": 50.9668,
+      "eval_samples_per_second": 39.771,
+      "eval_steps_per_second": 4.984,
+      "eval_wer": 0.38373352590720344,
+      "step": 41500
+    },
+    {
+      "epoch": 4.45,
+      "eval_loss": 0.46106940507888794,
+      "eval_runtime": 49.9064,
+      "eval_samples_per_second": 40.616,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.3838237949088283,
+      "step": 41600
+    },
+    {
+      "epoch": 4.46,
+      "eval_loss": 0.46049293875694275,
+      "eval_runtime": 49.6932,
+      "eval_samples_per_second": 40.79,
+      "eval_steps_per_second": 5.111,
+      "eval_wer": 0.39447553710055966,
+      "step": 41700
+    },
+    {
+      "epoch": 4.48,
+      "eval_loss": 0.4730290472507477,
+      "eval_runtime": 49.4559,
+      "eval_samples_per_second": 40.986,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.39384365408918576,
+      "step": 41800
+    },
+    {
+      "epoch": 4.49,
+      "eval_loss": 0.4566921293735504,
+      "eval_runtime": 49.6481,
+      "eval_samples_per_second": 40.827,
+      "eval_steps_per_second": 5.116,
+      "eval_wer": 0.387434554973822,
+      "step": 41900
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.0002108510028653295,
+      "loss": 0.2658,
+      "step": 42000
+    },
+    {
+      "epoch": 4.5,
+      "eval_loss": 0.47149142622947693,
+      "eval_runtime": 49.8048,
+      "eval_samples_per_second": 40.699,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.38689294096407295,
+      "step": 42000
+    },
+    {
+      "epoch": 4.51,
+      "eval_loss": 0.45138201117515564,
+      "eval_runtime": 50.5648,
+      "eval_samples_per_second": 40.087,
+      "eval_steps_per_second": 5.023,
+      "eval_wer": 0.3832821808990793,
+      "step": 42100
+    },
+    {
+      "epoch": 4.52,
+      "eval_loss": 0.46020662784576416,
+      "eval_runtime": 49.6896,
+      "eval_samples_per_second": 40.793,
+      "eval_steps_per_second": 5.112,
+      "eval_wer": 0.3897815490160679,
+      "step": 42200
+    },
+    {
+      "epoch": 4.53,
+      "eval_loss": 0.4846310615539551,
+      "eval_runtime": 49.7205,
+      "eval_samples_per_second": 40.768,
+      "eval_steps_per_second": 5.109,
+      "eval_wer": 0.4022386712402961,
+      "step": 42300
+    },
+    {
+      "epoch": 4.54,
+      "eval_loss": 0.44736534357070923,
+      "eval_runtime": 49.9456,
+      "eval_samples_per_second": 40.584,
+      "eval_steps_per_second": 5.086,
+      "eval_wer": 0.3810254558584582,
+      "step": 42400
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.00020977650429799427,
+      "loss": 0.2676,
+      "step": 42500
+    },
+    {
+      "epoch": 4.55,
+      "eval_loss": 0.4513036608695984,
+      "eval_runtime": 49.7727,
+      "eval_samples_per_second": 40.725,
+      "eval_steps_per_second": 5.103,
+      "eval_wer": 0.3820184148763315,
+      "step": 42500
+    },
+    {
+      "epoch": 4.56,
+      "eval_loss": 0.4588390588760376,
+      "eval_runtime": 49.7619,
+      "eval_samples_per_second": 40.734,
+      "eval_steps_per_second": 5.104,
+      "eval_wer": 0.3927604260696877,
+      "step": 42600
+    },
+    {
+      "epoch": 4.57,
+      "eval_loss": 0.4601156413555145,
+      "eval_runtime": 49.7259,
+      "eval_samples_per_second": 40.763,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.3894204730095685,
+      "step": 42700
+    },
+    {
+      "epoch": 4.58,
+      "eval_loss": 0.45155754685401917,
+      "eval_runtime": 49.7075,
+      "eval_samples_per_second": 40.779,
+      "eval_steps_per_second": 5.11,
+      "eval_wer": 0.37922007582596134,
+      "step": 42800
+    },
+    {
+      "epoch": 4.59,
+      "eval_loss": 0.4481757581233978,
+      "eval_runtime": 49.8315,
+      "eval_samples_per_second": 40.677,
+      "eval_steps_per_second": 5.097,
+      "eval_wer": 0.38481675392670156,
+      "step": 42900
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.000208702005730659,
+      "loss": 0.2693,
+      "step": 43000
+    },
+    {
+      "epoch": 4.6,
+      "eval_loss": 0.4695383310317993,
+      "eval_runtime": 50.2493,
+      "eval_samples_per_second": 40.339,
+      "eval_steps_per_second": 5.055,
+      "eval_wer": 0.4007943672142986,
+      "step": 43000
+    },
+    {
+      "epoch": 4.61,
+      "eval_loss": 0.45803362131118774,
+      "eval_runtime": 50.1652,
+      "eval_samples_per_second": 40.407,
+      "eval_steps_per_second": 5.063,
+      "eval_wer": 0.38707347896732264,
+      "step": 43100
+    },
+    {
+      "epoch": 4.63,
+      "eval_loss": 0.44191282987594604,
+      "eval_runtime": 49.9048,
+      "eval_samples_per_second": 40.617,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.38571944394295,
+      "step": 43200
+    },
+    {
+      "epoch": 4.64,
+      "eval_loss": 0.45340240001678467,
+      "eval_runtime": 51.1795,
+      "eval_samples_per_second": 39.606,
+      "eval_steps_per_second": 4.963,
+      "eval_wer": 0.3795811518324607,
+      "step": 43300
+    },
+    {
+      "epoch": 4.65,
+      "eval_loss": 0.4532068073749542,
+      "eval_runtime": 49.8489,
+      "eval_samples_per_second": 40.663,
+      "eval_steps_per_second": 5.095,
+      "eval_wer": 0.38562917494132515,
+      "step": 43400
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.00020762750716332376,
+      "loss": 0.2641,
+      "step": 43500
+    },
+    {
+      "epoch": 4.66,
+      "eval_loss": 0.442064106464386,
+      "eval_runtime": 49.9503,
+      "eval_samples_per_second": 40.58,
+      "eval_steps_per_second": 5.085,
+      "eval_wer": 0.38093518685683336,
+      "step": 43500
+    },
+    {
+      "epoch": 4.67,
+      "eval_loss": 0.44000962376594543,
+      "eval_runtime": 50.1474,
+      "eval_samples_per_second": 40.421,
+      "eval_steps_per_second": 5.065,
+      "eval_wer": 0.38436540891857734,
+      "step": 43600
+    },
+    {
+      "epoch": 4.68,
+      "eval_loss": 0.45147180557250977,
+      "eval_runtime": 49.9827,
+      "eval_samples_per_second": 40.554,
+      "eval_steps_per_second": 5.082,
+      "eval_wer": 0.3832821808990793,
+      "step": 43700
+    },
+    {
+      "epoch": 4.69,
+      "eval_loss": 0.4462360143661499,
+      "eval_runtime": 50.1109,
+      "eval_samples_per_second": 40.45,
+      "eval_steps_per_second": 5.069,
+      "eval_wer": 0.3807546488535837,
+      "step": 43800
+    },
+    {
+      "epoch": 4.7,
+      "eval_loss": 0.47406265139579773,
+      "eval_runtime": 50.0121,
+      "eval_samples_per_second": 40.53,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.392579888066438,
+      "step": 43900
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.0002065551575931232,
+      "loss": 0.2626,
+      "step": 44000
+    },
+    {
+      "epoch": 4.71,
+      "eval_loss": 0.4542163014411926,
+      "eval_runtime": 50.5956,
+      "eval_samples_per_second": 40.063,
+      "eval_steps_per_second": 5.02,
+      "eval_wer": 0.393121502076187,
+      "step": 44000
+    },
+    {
+      "epoch": 4.72,
+      "eval_loss": 0.4554780125617981,
+      "eval_runtime": 49.9867,
+      "eval_samples_per_second": 40.551,
+      "eval_steps_per_second": 5.081,
+      "eval_wer": 0.38851778299332007,
+      "step": 44100
+    },
+    {
+      "epoch": 4.73,
+      "eval_loss": 0.45047158002853394,
+      "eval_runtime": 50.0094,
+      "eval_samples_per_second": 40.532,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.38454594692182703,
+      "step": 44200
+    },
+    {
+      "epoch": 4.74,
+      "eval_loss": 0.45926421880722046,
+      "eval_runtime": 49.6573,
+      "eval_samples_per_second": 40.82,
+      "eval_steps_per_second": 5.115,
+      "eval_wer": 0.38707347896732264,
+      "step": 44300
+    },
+    {
+      "epoch": 4.75,
+      "eval_loss": 0.43586495518684387,
+      "eval_runtime": 50.2993,
+      "eval_samples_per_second": 40.299,
+      "eval_steps_per_second": 5.05,
+      "eval_wer": 0.38301137389420475,
+      "step": 44400
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.00020548280802292263,
+      "loss": 0.2648,
+      "step": 44500
+    },
+    {
+      "epoch": 4.76,
+      "eval_loss": 0.43868881464004517,
+      "eval_runtime": 49.9728,
+      "eval_samples_per_second": 40.562,
+      "eval_steps_per_second": 5.083,
+      "eval_wer": 0.3736233977252212,
+      "step": 44500
+    },
+    {
+      "epoch": 4.78,
+      "eval_loss": 0.45290684700012207,
+      "eval_runtime": 49.8631,
+      "eval_samples_per_second": 40.651,
+      "eval_steps_per_second": 5.094,
+      "eval_wer": 0.38066437985195883,
+      "step": 44600
+    },
+    {
+      "epoch": 4.79,
+      "eval_loss": 0.45659151673316956,
+      "eval_runtime": 50.0989,
+      "eval_samples_per_second": 40.46,
+      "eval_steps_per_second": 5.07,
+      "eval_wer": 0.38373352590720344,
+      "step": 44700
+    },
+    {
+      "epoch": 4.8,
+      "eval_loss": 0.45571401715278625,
+      "eval_runtime": 50.5772,
+      "eval_samples_per_second": 40.077,
+      "eval_steps_per_second": 5.022,
+      "eval_wer": 0.40666185231991336,
+      "step": 44800
+    },
+    {
+      "epoch": 4.81,
+      "eval_loss": 0.4609202742576599,
+      "eval_runtime": 50.0904,
+      "eval_samples_per_second": 40.467,
+      "eval_steps_per_second": 5.071,
+      "eval_wer": 0.38517782993320093,
+      "step": 44900
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.00020440830945558735,
+      "loss": 0.2603,
+      "step": 45000
+    },
+    {
+      "epoch": 4.82,
+      "eval_loss": 0.46665722131729126,
+      "eval_runtime": 49.8766,
+      "eval_samples_per_second": 40.64,
+      "eval_steps_per_second": 5.093,
+      "eval_wer": 0.4005235602094241,
+      "step": 45000
+    },
+    {
+      "epoch": 4.83,
+      "eval_loss": 0.46655377745628357,
+      "eval_runtime": 49.8776,
+      "eval_samples_per_second": 40.639,
+      "eval_steps_per_second": 5.092,
+      "eval_wer": 0.3836432569055786,
+      "step": 45100
+    },
+    {
+      "epoch": 4.84,
+      "eval_loss": 0.47749215364456177,
+      "eval_runtime": 49.9,
+      "eval_samples_per_second": 40.621,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.3945658061021845,
+      "step": 45200
+    },
+    {
+      "epoch": 4.85,
+      "eval_loss": 0.4701036810874939,
+      "eval_runtime": 50.0063,
+      "eval_samples_per_second": 40.535,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.39248961906481317,
+      "step": 45300
+    },
+    {
+      "epoch": 4.86,
+      "eval_loss": 0.4578703045845032,
+      "eval_runtime": 50.0708,
+      "eval_samples_per_second": 40.483,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.38887885899981944,
+      "step": 45400
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.00020333381088825215,
+      "loss": 0.2626,
+      "step": 45500
+    },
+    {
+      "epoch": 4.87,
+      "eval_loss": 0.4515564739704132,
+      "eval_runtime": 49.915,
+      "eval_samples_per_second": 40.609,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.38842751399169523,
+      "step": 45500
+    },
+    {
+      "epoch": 4.88,
+      "eval_loss": 0.46054086089134216,
+      "eval_runtime": 49.7949,
+      "eval_samples_per_second": 40.707,
+      "eval_steps_per_second": 5.101,
+      "eval_wer": 0.3877956309803214,
+      "step": 45600
+    },
+    {
+      "epoch": 4.89,
+      "eval_loss": 0.4576418101787567,
+      "eval_runtime": 50.3057,
+      "eval_samples_per_second": 40.294,
+      "eval_steps_per_second": 5.049,
+      "eval_wer": 0.3802130348438346,
+      "step": 45700
+    },
+    {
+      "epoch": 4.9,
+      "eval_loss": 0.45533105731010437,
+      "eval_runtime": 49.7254,
+      "eval_samples_per_second": 40.764,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.37804657880483844,
+      "step": 45800
+    },
+    {
+      "epoch": 4.91,
+      "eval_loss": 0.4335707724094391,
+      "eval_runtime": 50.3097,
+      "eval_samples_per_second": 40.29,
+      "eval_steps_per_second": 5.049,
+      "eval_wer": 0.3752482397544683,
+      "step": 45900
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.00020225931232091687,
+      "loss": 0.2602,
+      "step": 46000
+    },
+    {
+      "epoch": 4.93,
+      "eval_loss": 0.44188374280929565,
+      "eval_runtime": 49.9873,
+      "eval_samples_per_second": 40.55,
+      "eval_steps_per_second": 5.081,
+      "eval_wer": 0.3880664379851959,
+      "step": 46000
+    },
+    {
+      "epoch": 4.94,
+      "eval_loss": 0.46014997363090515,
+      "eval_runtime": 50.4086,
+      "eval_samples_per_second": 40.211,
+      "eval_steps_per_second": 5.039,
+      "eval_wer": 0.3842751399169525,
+      "step": 46100
+    },
+    {
+      "epoch": 4.95,
+      "eval_loss": 0.44370460510253906,
+      "eval_runtime": 49.8698,
+      "eval_samples_per_second": 40.646,
+      "eval_steps_per_second": 5.093,
+      "eval_wer": 0.3956490341216826,
+      "step": 46200
+    },
+    {
+      "epoch": 4.96,
+      "eval_loss": 0.4524065852165222,
+      "eval_runtime": 50.044,
+      "eval_samples_per_second": 40.504,
+      "eval_steps_per_second": 5.076,
+      "eval_wer": 0.38436540891857734,
+      "step": 46300
+    },
+    {
+      "epoch": 4.97,
+      "eval_loss": 0.4709448218345642,
+      "eval_runtime": 49.8637,
+      "eval_samples_per_second": 40.651,
+      "eval_steps_per_second": 5.094,
+      "eval_wer": 0.4031413612565445,
+      "step": 46400
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.00020118481375358164,
+      "loss": 0.2609,
+      "step": 46500
+    },
+    {
+      "epoch": 4.98,
+      "eval_loss": 0.45001140236854553,
+      "eval_runtime": 50.2152,
+      "eval_samples_per_second": 40.366,
+      "eval_steps_per_second": 5.058,
+      "eval_wer": 0.3871637479689475,
+      "step": 46500
+    },
+    {
+      "epoch": 4.99,
+      "eval_loss": 0.43656888604164124,
+      "eval_runtime": 49.6836,
+      "eval_samples_per_second": 40.798,
+      "eval_steps_per_second": 5.112,
+      "eval_wer": 0.3846362159234519,
+      "step": 46600
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 0.46530061960220337,
+      "eval_runtime": 49.7235,
+      "eval_samples_per_second": 40.765,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.38842751399169523,
+      "step": 46700
+    },
+    {
+      "epoch": 5.01,
+      "eval_loss": 0.4602296054363251,
+      "eval_runtime": 50.5077,
+      "eval_samples_per_second": 40.133,
+      "eval_steps_per_second": 5.029,
+      "eval_wer": 0.39321177107781186,
+      "step": 46800
+    },
+    {
+      "epoch": 5.02,
+      "eval_loss": 0.4668322801589966,
+      "eval_runtime": 50.0245,
+      "eval_samples_per_second": 40.52,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.3853583679364506,
+      "step": 46900
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.00020011031518624639,
+      "loss": 0.2472,
+      "step": 47000
+    },
+    {
+      "epoch": 5.03,
+      "eval_loss": 0.4615883529186249,
+      "eval_runtime": 50.0103,
+      "eval_samples_per_second": 40.532,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.389149666004694,
+      "step": 47000
+    },
+    {
+      "epoch": 5.04,
+      "eval_loss": 0.4542787969112396,
+      "eval_runtime": 50.2585,
+      "eval_samples_per_second": 40.331,
+      "eval_steps_per_second": 5.054,
+      "eval_wer": 0.38355298790395376,
+      "step": 47100
+    },
+    {
+      "epoch": 5.05,
+      "eval_loss": 0.4526177644729614,
+      "eval_runtime": 50.0483,
+      "eval_samples_per_second": 40.501,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.38219895287958117,
+      "step": 47200
+    },
+    {
+      "epoch": 5.06,
+      "eval_loss": 0.45387881994247437,
+      "eval_runtime": 50.0828,
+      "eval_samples_per_second": 40.473,
+      "eval_steps_per_second": 5.072,
+      "eval_wer": 0.37407474273334534,
+      "step": 47300
+    },
+    {
+      "epoch": 5.07,
+      "eval_loss": 0.4776493310928345,
+      "eval_runtime": 49.8454,
+      "eval_samples_per_second": 40.666,
+      "eval_steps_per_second": 5.096,
+      "eval_wer": 0.3818378768730818,
+      "step": 47400
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.0001990379656160458,
+      "loss": 0.2278,
+      "step": 47500
+    },
+    {
+      "epoch": 5.09,
+      "eval_loss": 0.47709885239601135,
+      "eval_runtime": 51.1106,
+      "eval_samples_per_second": 39.659,
+      "eval_steps_per_second": 4.97,
+      "eval_wer": 0.37940061382921103,
+      "step": 47500
+    },
+    {
+      "epoch": 5.1,
+      "eval_loss": 0.46624356508255005,
+      "eval_runtime": 49.8637,
+      "eval_samples_per_second": 40.651,
+      "eval_steps_per_second": 5.094,
+      "eval_wer": 0.3831016428958296,
+      "step": 47600
+    },
+    {
+      "epoch": 5.11,
+      "eval_loss": 0.45577922463417053,
+      "eval_runtime": 49.9202,
+      "eval_samples_per_second": 40.605,
+      "eval_steps_per_second": 5.088,
+      "eval_wer": 0.4032316302581693,
+      "step": 47700
+    },
+    {
+      "epoch": 5.12,
+      "eval_loss": 0.4903823733329773,
+      "eval_runtime": 50.3451,
+      "eval_samples_per_second": 40.262,
+      "eval_steps_per_second": 5.045,
+      "eval_wer": 0.3917674670518144,
+      "step": 47800
+    },
+    {
+      "epoch": 5.13,
+      "eval_loss": 0.4764585494995117,
+      "eval_runtime": 50.0404,
+      "eval_samples_per_second": 40.507,
+      "eval_steps_per_second": 5.076,
+      "eval_wer": 0.3889691280014443,
+      "step": 47900
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.00019796346704871058,
+      "loss": 0.2311,
+      "step": 48000
+    },
+    {
+      "epoch": 5.14,
+      "eval_loss": 0.4673701822757721,
+      "eval_runtime": 49.7968,
+      "eval_samples_per_second": 40.705,
+      "eval_steps_per_second": 5.101,
+      "eval_wer": 0.38824697598844554,
+      "step": 48000
+    },
+    {
+      "epoch": 5.15,
+      "eval_loss": 0.460916131734848,
+      "eval_runtime": 49.6923,
+      "eval_samples_per_second": 40.791,
+      "eval_steps_per_second": 5.111,
+      "eval_wer": 0.39465607510380935,
+      "step": 48100
+    },
+    {
+      "epoch": 5.16,
+      "eval_loss": 0.4587761163711548,
+      "eval_runtime": 49.8661,
+      "eval_samples_per_second": 40.649,
+      "eval_steps_per_second": 5.094,
+      "eval_wer": 0.38373352590720344,
+      "step": 48200
+    },
+    {
+      "epoch": 5.17,
+      "eval_loss": 0.48272979259490967,
+      "eval_runtime": 49.3238,
+      "eval_samples_per_second": 41.096,
+      "eval_steps_per_second": 5.15,
+      "eval_wer": 0.3844556779202022,
+      "step": 48300
+    },
+    {
+      "epoch": 5.18,
+      "eval_loss": 0.4711352288722992,
+      "eval_runtime": 49.1004,
+      "eval_samples_per_second": 41.283,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.38391406391045313,
+      "step": 48400
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.00019689111747851,
+      "loss": 0.229,
+      "step": 48500
+    },
+    {
+      "epoch": 5.19,
+      "eval_loss": 0.45826002955436707,
+      "eval_runtime": 48.9232,
+      "eval_samples_per_second": 41.432,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.3872540169705723,
+      "step": 48500
+    },
+    {
+      "epoch": 5.2,
+      "eval_loss": 0.47996604442596436,
+      "eval_runtime": 48.7886,
+      "eval_samples_per_second": 41.547,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.38580971294457483,
+      "step": 48600
+    },
+    {
+      "epoch": 5.21,
+      "eval_loss": 0.46112576127052307,
+      "eval_runtime": 48.9997,
+      "eval_samples_per_second": 41.368,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.38003249684058493,
+      "step": 48700
+    },
+    {
+      "epoch": 5.22,
+      "eval_loss": 0.45038747787475586,
+      "eval_runtime": 48.9685,
+      "eval_samples_per_second": 41.394,
+      "eval_steps_per_second": 5.187,
+      "eval_wer": 0.38887885899981944,
+      "step": 48800
+    },
+    {
+      "epoch": 5.24,
+      "eval_loss": 0.45691677927970886,
+      "eval_runtime": 49.0298,
+      "eval_samples_per_second": 41.342,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.3760606607690919,
+      "step": 48900
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.00019581661891117478,
+      "loss": 0.2313,
+      "step": 49000
+    },
+    {
+      "epoch": 5.25,
+      "eval_loss": 0.47323185205459595,
+      "eval_runtime": 48.7503,
+      "eval_samples_per_second": 41.579,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.3914966600469399,
+      "step": 49000
+    },
+    {
+      "epoch": 5.26,
+      "eval_loss": 0.4727950692176819,
+      "eval_runtime": 49.3693,
+      "eval_samples_per_second": 41.058,
+      "eval_steps_per_second": 5.145,
+      "eval_wer": 0.38319191189745444,
+      "step": 49100
+    },
+    {
+      "epoch": 5.27,
+      "eval_loss": 0.4667152464389801,
+      "eval_runtime": 48.9142,
+      "eval_samples_per_second": 41.44,
+      "eval_steps_per_second": 5.193,
+      "eval_wer": 0.3814768008665824,
+      "step": 49200
+    },
+    {
+      "epoch": 5.28,
+      "eval_loss": 0.49122875928878784,
+      "eval_runtime": 49.9108,
+      "eval_samples_per_second": 40.612,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.38562917494132515,
+      "step": 49300
+    },
+    {
+      "epoch": 5.29,
+      "eval_loss": 0.47898349165916443,
+      "eval_runtime": 48.5465,
+      "eval_samples_per_second": 41.754,
+      "eval_steps_per_second": 5.232,
+      "eval_wer": 0.3945658061021845,
+      "step": 49400
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.00019474212034383952,
+      "loss": 0.2266,
+      "step": 49500
+    },
+    {
+      "epoch": 5.3,
+      "eval_loss": 0.4596710801124573,
+      "eval_runtime": 49.2944,
+      "eval_samples_per_second": 41.12,
+      "eval_steps_per_second": 5.153,
+      "eval_wer": 0.3763314677739664,
+      "step": 49500
+    },
+    {
+      "epoch": 5.31,
+      "eval_loss": 0.45802178978919983,
+      "eval_runtime": 49.1967,
+      "eval_samples_per_second": 41.202,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3777757717999639,
+      "step": 49600
+    },
+    {
+      "epoch": 5.32,
+      "eval_loss": 0.4439472556114197,
+      "eval_runtime": 48.6894,
+      "eval_samples_per_second": 41.631,
+      "eval_steps_per_second": 5.217,
+      "eval_wer": 0.37208882469759885,
+      "step": 49700
+    },
+    {
+      "epoch": 5.33,
+      "eval_loss": 0.46113863587379456,
+      "eval_runtime": 48.9061,
+      "eval_samples_per_second": 41.447,
+      "eval_steps_per_second": 5.194,
+      "eval_wer": 0.37037371366672683,
+      "step": 49800
+    },
+    {
+      "epoch": 5.34,
+      "eval_loss": 0.45986300706863403,
+      "eval_runtime": 49.0073,
+      "eval_samples_per_second": 41.361,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.3768730817837155,
+      "step": 49900
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.00019366762177650427,
+      "loss": 0.235,
+      "step": 50000
+    },
+    {
+      "epoch": 5.35,
+      "eval_loss": 0.45431894063949585,
+      "eval_runtime": 49.0294,
+      "eval_samples_per_second": 41.343,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.3808449178552085,
+      "step": 50000
+    },
+    {
+      "epoch": 5.36,
+      "eval_loss": 0.45549678802490234,
+      "eval_runtime": 49.0356,
+      "eval_samples_per_second": 41.337,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.3773244267918397,
+      "step": 50100
+    },
+    {
+      "epoch": 5.37,
+      "eval_loss": 0.45248955488204956,
+      "eval_runtime": 48.9938,
+      "eval_samples_per_second": 41.373,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.3814768008665824,
+      "step": 50200
+    },
+    {
+      "epoch": 5.39,
+      "eval_loss": 0.45570600032806396,
+      "eval_runtime": 48.9527,
+      "eval_samples_per_second": 41.407,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.3813865318649576,
+      "step": 50300
+    },
+    {
+      "epoch": 5.4,
+      "eval_loss": 0.46039825677871704,
+      "eval_runtime": 48.6359,
+      "eval_samples_per_second": 41.677,
+      "eval_steps_per_second": 5.222,
+      "eval_wer": 0.375428777757718,
+      "step": 50400
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.00019259312320916904,
+      "loss": 0.2299,
+      "step": 50500
+    },
+    {
+      "epoch": 5.41,
+      "eval_loss": 0.4657667875289917,
+      "eval_runtime": 49.1397,
+      "eval_samples_per_second": 41.25,
+      "eval_steps_per_second": 5.169,
+      "eval_wer": 0.3769633507853403,
+      "step": 50500
+    },
+    {
+      "epoch": 5.42,
+      "eval_loss": 0.4657803177833557,
+      "eval_runtime": 49.0879,
+      "eval_samples_per_second": 41.293,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.38842751399169523,
+      "step": 50600
+    },
+    {
+      "epoch": 5.43,
+      "eval_loss": 0.4701368510723114,
+      "eval_runtime": 48.9419,
+      "eval_samples_per_second": 41.416,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.3919480050550641,
+      "step": 50700
+    },
+    {
+      "epoch": 5.44,
+      "eval_loss": 0.4495311677455902,
+      "eval_runtime": 48.8139,
+      "eval_samples_per_second": 41.525,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.3818378768730818,
+      "step": 50800
+    },
+    {
+      "epoch": 5.45,
+      "eval_loss": 0.47025540471076965,
+      "eval_runtime": 48.7211,
+      "eval_samples_per_second": 41.604,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.3886080519949449,
+      "step": 50900
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.0001915186246418338,
+      "loss": 0.2307,
+      "step": 51000
+    },
+    {
+      "epoch": 5.46,
+      "eval_loss": 0.4395087659358978,
+      "eval_runtime": 48.5097,
+      "eval_samples_per_second": 41.785,
+      "eval_steps_per_second": 5.236,
+      "eval_wer": 0.37425528073659503,
+      "step": 51000
+    },
+    {
+      "epoch": 5.47,
+      "eval_loss": 0.44871771335601807,
+      "eval_runtime": 48.9604,
+      "eval_samples_per_second": 41.401,
+      "eval_steps_per_second": 5.188,
+      "eval_wer": 0.3750677017512186,
+      "step": 51100
+    },
+    {
+      "epoch": 5.48,
+      "eval_loss": 0.4355129301548004,
+      "eval_runtime": 49.0556,
+      "eval_samples_per_second": 41.32,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.3732623217187218,
+      "step": 51200
+    },
+    {
+      "epoch": 5.49,
+      "eval_loss": 0.4622361958026886,
+      "eval_runtime": 48.6307,
+      "eval_samples_per_second": 41.681,
+      "eval_steps_per_second": 5.223,
+      "eval_wer": 0.38111572486008305,
+      "step": 51300
+    },
+    {
+      "epoch": 5.5,
+      "eval_loss": 0.44426313042640686,
+      "eval_runtime": 49.0933,
+      "eval_samples_per_second": 41.289,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.3801227658422098,
+      "step": 51400
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.00019044412607449856,
+      "loss": 0.2383,
+      "step": 51500
+    },
+    {
+      "epoch": 5.51,
+      "eval_loss": 0.4410970211029053,
+      "eval_runtime": 48.6889,
+      "eval_samples_per_second": 41.632,
+      "eval_steps_per_second": 5.217,
+      "eval_wer": 0.3743455497382199,
+      "step": 51500
+    },
+    {
+      "epoch": 5.52,
+      "eval_loss": 0.4437633454799652,
+      "eval_runtime": 49.0805,
+      "eval_samples_per_second": 41.3,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3777757717999639,
+      "step": 51600
+    },
+    {
+      "epoch": 5.54,
+      "eval_loss": 0.45590972900390625,
+      "eval_runtime": 48.7252,
+      "eval_samples_per_second": 41.601,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.3784076548113378,
+      "step": 51700
+    },
+    {
+      "epoch": 5.55,
+      "eval_loss": 0.43089374899864197,
+      "eval_runtime": 49.0544,
+      "eval_samples_per_second": 41.322,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.3655894565806102,
+      "step": 51800
+    },
+    {
+      "epoch": 5.56,
+      "eval_loss": 0.4454806447029114,
+      "eval_runtime": 50.091,
+      "eval_samples_per_second": 40.466,
+      "eval_steps_per_second": 5.071,
+      "eval_wer": 0.3660408015887344,
+      "step": 51900
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.0001893696275071633,
+      "loss": 0.23,
+      "step": 52000
+    },
+    {
+      "epoch": 5.57,
+      "eval_loss": 0.4435708224773407,
+      "eval_runtime": 48.8139,
+      "eval_samples_per_second": 41.525,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.3598122404766203,
+      "step": 52000
+    },
+    {
+      "epoch": 5.58,
+      "eval_loss": 0.43444132804870605,
+      "eval_runtime": 48.9776,
+      "eval_samples_per_second": 41.386,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.3684780646326052,
+      "step": 52100
+    },
+    {
+      "epoch": 5.59,
+      "eval_loss": 0.42824628949165344,
+      "eval_runtime": 49.2816,
+      "eval_samples_per_second": 41.131,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.36901967864235424,
+      "step": 52200
+    },
+    {
+      "epoch": 5.6,
+      "eval_loss": 0.44636350870132446,
+      "eval_runtime": 49.2151,
+      "eval_samples_per_second": 41.187,
+      "eval_steps_per_second": 5.161,
+      "eval_wer": 0.38003249684058493,
+      "step": 52300
+    },
+    {
+      "epoch": 5.61,
+      "eval_loss": 0.4458146393299103,
+      "eval_runtime": 48.8414,
+      "eval_samples_per_second": 41.502,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.390864777035566,
+      "step": 52400
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.00018829512893982808,
+      "loss": 0.2305,
+      "step": 52500
+    },
+    {
+      "epoch": 5.62,
+      "eval_loss": 0.448255330324173,
+      "eval_runtime": 49.3278,
+      "eval_samples_per_second": 41.092,
+      "eval_steps_per_second": 5.149,
+      "eval_wer": 0.3756093157609677,
+      "step": 52500
+    },
+    {
+      "epoch": 5.63,
+      "eval_loss": 0.4547184705734253,
+      "eval_runtime": 49.758,
+      "eval_samples_per_second": 40.737,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.37849792381296266,
+      "step": 52600
+    },
+    {
+      "epoch": 5.64,
+      "eval_loss": 0.46712735295295715,
+      "eval_runtime": 48.8295,
+      "eval_samples_per_second": 41.512,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.3820184148763315,
+      "step": 52700
+    },
+    {
+      "epoch": 5.65,
+      "eval_loss": 0.4449251890182495,
+      "eval_runtime": 51.0114,
+      "eval_samples_per_second": 39.736,
+      "eval_steps_per_second": 4.979,
+      "eval_wer": 0.3657699945838599,
+      "step": 52800
+    },
+    {
+      "epoch": 5.66,
+      "eval_loss": 0.45962706208229065,
+      "eval_runtime": 48.8671,
+      "eval_samples_per_second": 41.48,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.37163747968947464,
+      "step": 52900
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.0001872227793696275,
+      "loss": 0.2237,
+      "step": 53000
+    },
+    {
+      "epoch": 5.67,
+      "eval_loss": 0.4399178624153137,
+      "eval_runtime": 48.9657,
+      "eval_samples_per_second": 41.396,
+      "eval_steps_per_second": 5.187,
+      "eval_wer": 0.36694349160498285,
+      "step": 53000
+    },
+    {
+      "epoch": 5.69,
+      "eval_loss": 0.44098395109176636,
+      "eval_runtime": 50.1368,
+      "eval_samples_per_second": 40.429,
+      "eval_steps_per_second": 5.066,
+      "eval_wer": 0.37190828669434917,
+      "step": 53100
+    },
+    {
+      "epoch": 5.7,
+      "eval_loss": 0.45740246772766113,
+      "eval_runtime": 48.7603,
+      "eval_samples_per_second": 41.571,
+      "eval_steps_per_second": 5.209,
+      "eval_wer": 0.3618884275139917,
+      "step": 53200
+    },
+    {
+      "epoch": 5.71,
+      "eval_loss": 0.44431784749031067,
+      "eval_runtime": 48.9893,
+      "eval_samples_per_second": 41.376,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.36901967864235424,
+      "step": 53300
+    },
+    {
+      "epoch": 5.72,
+      "eval_loss": 0.4380861818790436,
+      "eval_runtime": 49.01,
+      "eval_samples_per_second": 41.359,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.36775591261960644,
+      "step": 53400
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.00018614828080229225,
+      "loss": 0.2337,
+      "step": 53500
+    },
+    {
+      "epoch": 5.73,
+      "eval_loss": 0.44900110363960266,
+      "eval_runtime": 49.3359,
+      "eval_samples_per_second": 41.086,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.3687488716374797,
+      "step": 53500
+    },
+    {
+      "epoch": 5.74,
+      "eval_loss": 0.4427066743373871,
+      "eval_runtime": 49.2278,
+      "eval_samples_per_second": 41.176,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.37515797075284346,
+      "step": 53600
+    },
+    {
+      "epoch": 5.75,
+      "eval_loss": 0.4423009157180786,
+      "eval_runtime": 49.2643,
+      "eval_samples_per_second": 41.145,
+      "eval_steps_per_second": 5.156,
+      "eval_wer": 0.38580971294457483,
+      "step": 53700
+    },
+    {
+      "epoch": 5.76,
+      "eval_loss": 0.4701789617538452,
+      "eval_runtime": 49.1364,
+      "eval_samples_per_second": 41.253,
+      "eval_steps_per_second": 5.169,
+      "eval_wer": 0.3824697598844557,
+      "step": 53800
+    },
+    {
+      "epoch": 5.77,
+      "eval_loss": 0.4724409580230713,
+      "eval_runtime": 49.3384,
+      "eval_samples_per_second": 41.084,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.38003249684058493,
+      "step": 53900
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.00018507378223495702,
+      "loss": 0.23,
+      "step": 54000
+    },
+    {
+      "epoch": 5.78,
+      "eval_loss": 0.44759470224380493,
+      "eval_runtime": 49.0243,
+      "eval_samples_per_second": 41.347,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.3826502978877054,
+      "step": 54000
+    },
+    {
+      "epoch": 5.79,
+      "eval_loss": 0.4508274793624878,
+      "eval_runtime": 48.7547,
+      "eval_samples_per_second": 41.575,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.39185773605343927,
+      "step": 54100
+    },
+    {
+      "epoch": 5.8,
+      "eval_loss": 0.45642468333244324,
+      "eval_runtime": 49.1232,
+      "eval_samples_per_second": 41.264,
+      "eval_steps_per_second": 5.171,
+      "eval_wer": 0.37876873081783713,
+      "step": 54200
+    },
+    {
+      "epoch": 5.81,
+      "eval_loss": 0.46016448736190796,
+      "eval_runtime": 48.8959,
+      "eval_samples_per_second": 41.455,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.3887885899981946,
+      "step": 54300
+    },
+    {
+      "epoch": 5.82,
+      "eval_loss": 0.4537946581840515,
+      "eval_runtime": 49.0323,
+      "eval_samples_per_second": 41.34,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.37317205271709697,
+      "step": 54400
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.00018399928366762177,
+      "loss": 0.2334,
+      "step": 54500
+    },
+    {
+      "epoch": 5.84,
+      "eval_loss": 0.4499867558479309,
+      "eval_runtime": 48.5461,
+      "eval_samples_per_second": 41.754,
+      "eval_steps_per_second": 5.232,
+      "eval_wer": 0.3807546488535837,
+      "step": 54500
+    },
+    {
+      "epoch": 5.85,
+      "eval_loss": 0.44750362634658813,
+      "eval_runtime": 48.962,
+      "eval_samples_per_second": 41.399,
+      "eval_steps_per_second": 5.188,
+      "eval_wer": 0.3704639826683517,
+      "step": 54600
+    },
+    {
+      "epoch": 5.86,
+      "eval_loss": 0.4414827823638916,
+      "eval_runtime": 48.6484,
+      "eval_samples_per_second": 41.666,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.37723415779021485,
+      "step": 54700
+    },
+    {
+      "epoch": 5.87,
+      "eval_loss": 0.4515135586261749,
+      "eval_runtime": 48.8475,
+      "eval_samples_per_second": 41.497,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.37705361978696517,
+      "step": 54800
+    },
+    {
+      "epoch": 5.88,
+      "eval_loss": 0.44104623794555664,
+      "eval_runtime": 48.6499,
+      "eval_samples_per_second": 41.665,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.3676656436179816,
+      "step": 54900
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.00018292693409742116,
+      "loss": 0.2259,
+      "step": 55000
+    },
+    {
+      "epoch": 5.89,
+      "eval_loss": 0.45551472902297974,
+      "eval_runtime": 48.9097,
+      "eval_samples_per_second": 41.444,
+      "eval_steps_per_second": 5.193,
+      "eval_wer": 0.37019317566347715,
+      "step": 55000
+    },
+    {
+      "epoch": 5.9,
+      "eval_loss": 0.4508730471134186,
+      "eval_runtime": 48.4437,
+      "eval_samples_per_second": 41.842,
+      "eval_steps_per_second": 5.243,
+      "eval_wer": 0.3894204730095685,
+      "step": 55100
+    },
+    {
+      "epoch": 5.91,
+      "eval_loss": 0.4472365379333496,
+      "eval_runtime": 48.759,
+      "eval_samples_per_second": 41.572,
+      "eval_steps_per_second": 5.209,
+      "eval_wer": 0.3692002166456039,
+      "step": 55200
+    },
+    {
+      "epoch": 5.92,
+      "eval_loss": 0.44380319118499756,
+      "eval_runtime": 48.6197,
+      "eval_samples_per_second": 41.691,
+      "eval_steps_per_second": 5.224,
+      "eval_wer": 0.375428777757718,
+      "step": 55300
+    },
+    {
+      "epoch": 5.93,
+      "eval_loss": 0.4398874044418335,
+      "eval_runtime": 48.7782,
+      "eval_samples_per_second": 41.555,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.3698320996569778,
+      "step": 55400
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.00018185243553008596,
+      "loss": 0.2289,
+      "step": 55500
+    },
+    {
+      "epoch": 5.94,
+      "eval_loss": 0.44961434602737427,
+      "eval_runtime": 50.0268,
+      "eval_samples_per_second": 40.518,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.37533850875609315,
+      "step": 55500
+    },
+    {
+      "epoch": 5.95,
+      "eval_loss": 0.4505833685398102,
+      "eval_runtime": 51.0999,
+      "eval_samples_per_second": 39.667,
+      "eval_steps_per_second": 4.971,
+      "eval_wer": 0.3752482397544683,
+      "step": 55600
+    },
+    {
+      "epoch": 5.96,
+      "eval_loss": 0.448152095079422,
+      "eval_runtime": 49.8088,
+      "eval_samples_per_second": 40.696,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.37660227477884095,
+      "step": 55700
+    },
+    {
+      "epoch": 5.97,
+      "eval_loss": 0.4415203630924225,
+      "eval_runtime": 50.6483,
+      "eval_samples_per_second": 40.021,
+      "eval_steps_per_second": 5.015,
+      "eval_wer": 0.37723415779021485,
+      "step": 55800
+    },
+    {
+      "epoch": 5.98,
+      "eval_loss": 0.44465985894203186,
+      "eval_runtime": 49.654,
+      "eval_samples_per_second": 40.822,
+      "eval_steps_per_second": 5.115,
+      "eval_wer": 0.3749774327495938,
+      "step": 55900
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.00018078008595988536,
+      "loss": 0.2281,
+      "step": 56000
+    },
+    {
+      "epoch": 6.0,
+      "eval_loss": 0.45663803815841675,
+      "eval_runtime": 49.5628,
+      "eval_samples_per_second": 40.898,
+      "eval_steps_per_second": 5.125,
+      "eval_wer": 0.38418487091532766,
+      "step": 56000
+    },
+    {
+      "epoch": 6.01,
+      "eval_loss": 0.4693831205368042,
+      "eval_runtime": 50.0809,
+      "eval_samples_per_second": 40.475,
+      "eval_steps_per_second": 5.072,
+      "eval_wer": 0.37741469579346454,
+      "step": 56100
+    },
+    {
+      "epoch": 6.02,
+      "eval_loss": 0.4454186260700226,
+      "eval_runtime": 50.0614,
+      "eval_samples_per_second": 40.49,
+      "eval_steps_per_second": 5.074,
+      "eval_wer": 0.37876873081783713,
+      "step": 56200
+    },
+    {
+      "epoch": 6.03,
+      "eval_loss": 0.467598557472229,
+      "eval_runtime": 50.2364,
+      "eval_samples_per_second": 40.349,
+      "eval_steps_per_second": 5.056,
+      "eval_wer": 0.3718180176927243,
+      "step": 56300
+    },
+    {
+      "epoch": 6.04,
+      "eval_loss": 0.46499374508857727,
+      "eval_runtime": 50.1973,
+      "eval_samples_per_second": 40.381,
+      "eval_steps_per_second": 5.06,
+      "eval_wer": 0.3750677017512186,
+      "step": 56400
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.00017970558739255016,
+      "loss": 0.1979,
+      "step": 56500
+    },
+    {
+      "epoch": 6.05,
+      "eval_loss": 0.4601166844367981,
+      "eval_runtime": 49.9764,
+      "eval_samples_per_second": 40.559,
+      "eval_steps_per_second": 5.082,
+      "eval_wer": 0.3765120057772161,
+      "step": 56500
+    },
+    {
+      "epoch": 6.06,
+      "eval_loss": 0.4647436738014221,
+      "eval_runtime": 50.0745,
+      "eval_samples_per_second": 40.48,
+      "eval_steps_per_second": 5.072,
+      "eval_wer": 0.384004332912078,
+      "step": 56600
+    },
+    {
+      "epoch": 6.07,
+      "eval_loss": 0.4781748950481415,
+      "eval_runtime": 50.0517,
+      "eval_samples_per_second": 40.498,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.3756093157609677,
+      "step": 56700
+    },
+    {
+      "epoch": 6.08,
+      "eval_loss": 0.4708513617515564,
+      "eval_runtime": 49.8736,
+      "eval_samples_per_second": 40.643,
+      "eval_steps_per_second": 5.093,
+      "eval_wer": 0.3736233977252212,
+      "step": 56800
+    },
+    {
+      "epoch": 6.09,
+      "eval_loss": 0.4707356095314026,
+      "eval_runtime": 50.2175,
+      "eval_samples_per_second": 40.364,
+      "eval_steps_per_second": 5.058,
+      "eval_wer": 0.37335259072034666,
+      "step": 56900
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.00017863108882521487,
+      "loss": 0.1923,
+      "step": 57000
+    },
+    {
+      "epoch": 6.1,
+      "eval_loss": 0.4704015254974365,
+      "eval_runtime": 50.2985,
+      "eval_samples_per_second": 40.299,
+      "eval_steps_per_second": 5.05,
+      "eval_wer": 0.3750677017512186,
+      "step": 57000
+    },
+    {
+      "epoch": 6.11,
+      "eval_loss": 0.4542071223258972,
+      "eval_runtime": 49.8684,
+      "eval_samples_per_second": 40.647,
+      "eval_steps_per_second": 5.093,
+      "eval_wer": 0.37208882469759885,
+      "step": 57100
+    },
+    {
+      "epoch": 6.12,
+      "eval_loss": 0.45417317748069763,
+      "eval_runtime": 51.0162,
+      "eval_samples_per_second": 39.732,
+      "eval_steps_per_second": 4.979,
+      "eval_wer": 0.37353312872359634,
+      "step": 57200
+    },
+    {
+      "epoch": 6.13,
+      "eval_loss": 0.4587162435054779,
+      "eval_runtime": 49.7221,
+      "eval_samples_per_second": 40.767,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.3803935728470843,
+      "step": 57300
+    },
+    {
+      "epoch": 6.15,
+      "eval_loss": 0.4428020715713501,
+      "eval_runtime": 50.0675,
+      "eval_samples_per_second": 40.485,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.3687488716374797,
+      "step": 57400
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.00017755659025787962,
+      "loss": 0.2012,
+      "step": 57500
+    },
+    {
+      "epoch": 6.16,
+      "eval_loss": 0.4456498920917511,
+      "eval_runtime": 50.0097,
+      "eval_samples_per_second": 40.532,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.3747968947463441,
+      "step": 57500
+    },
+    {
+      "epoch": 6.17,
+      "eval_loss": 0.45776355266571045,
+      "eval_runtime": 50.0851,
+      "eval_samples_per_second": 40.471,
+      "eval_steps_per_second": 5.071,
+      "eval_wer": 0.37615092977071674,
+      "step": 57600
+    },
+    {
+      "epoch": 6.18,
+      "eval_loss": 0.4699217975139618,
+      "eval_runtime": 50.0917,
+      "eval_samples_per_second": 40.466,
+      "eval_steps_per_second": 5.071,
+      "eval_wer": 0.3721790936992237,
+      "step": 57700
+    },
+    {
+      "epoch": 6.19,
+      "eval_loss": 0.44986045360565186,
+      "eval_runtime": 50.1466,
+      "eval_samples_per_second": 40.422,
+      "eval_steps_per_second": 5.065,
+      "eval_wer": 0.3756093157609677,
+      "step": 57800
+    },
+    {
+      "epoch": 6.2,
+      "eval_loss": 0.46329542994499207,
+      "eval_runtime": 49.6477,
+      "eval_samples_per_second": 40.828,
+      "eval_steps_per_second": 5.116,
+      "eval_wer": 0.36802671962448097,
+      "step": 57900
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.0001764820916905444,
+      "loss": 0.1951,
+      "step": 58000
+    },
+    {
+      "epoch": 6.21,
+      "eval_loss": 0.45481452345848083,
+      "eval_runtime": 50.2015,
+      "eval_samples_per_second": 40.377,
+      "eval_steps_per_second": 5.06,
+      "eval_wer": 0.3711861346813504,
+      "step": 58000
+    },
+    {
+      "epoch": 6.22,
+      "eval_loss": 0.4520038366317749,
+      "eval_runtime": 50.0272,
+      "eval_samples_per_second": 40.518,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.3758801227658422,
+      "step": 58100
+    },
+    {
+      "epoch": 6.23,
+      "eval_loss": 0.4458475410938263,
+      "eval_runtime": 50.0331,
+      "eval_samples_per_second": 40.513,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.36161762050911717,
+      "step": 58200
+    },
+    {
+      "epoch": 6.24,
+      "eval_loss": 0.43071892857551575,
+      "eval_runtime": 50.4559,
+      "eval_samples_per_second": 40.174,
+      "eval_steps_per_second": 5.034,
+      "eval_wer": 0.36369380754648856,
+      "step": 58300
+    },
+    {
+      "epoch": 6.25,
+      "eval_loss": 0.45458078384399414,
+      "eval_runtime": 50.2084,
+      "eval_samples_per_second": 40.372,
+      "eval_steps_per_second": 5.059,
+      "eval_wer": 0.3620689655172414,
+      "step": 58400
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.00017540759312320914,
+      "loss": 0.1967,
+      "step": 58500
+    },
+    {
+      "epoch": 6.26,
+      "eval_loss": 0.4458640217781067,
+      "eval_runtime": 50.4951,
+      "eval_samples_per_second": 40.143,
+      "eval_steps_per_second": 5.03,
+      "eval_wer": 0.3623397725221159,
+      "step": 58500
+    },
+    {
+      "epoch": 6.27,
+      "eval_loss": 0.45346349477767944,
+      "eval_runtime": 50.6118,
+      "eval_samples_per_second": 40.05,
+      "eval_steps_per_second": 5.019,
+      "eval_wer": 0.36901967864235424,
+      "step": 58600
+    },
+    {
+      "epoch": 6.28,
+      "eval_loss": 0.4573749899864197,
+      "eval_runtime": 50.4308,
+      "eval_samples_per_second": 40.194,
+      "eval_steps_per_second": 5.037,
+      "eval_wer": 0.37714388878859,
+      "step": 58700
+    },
+    {
+      "epoch": 6.3,
+      "eval_loss": 0.44925975799560547,
+      "eval_runtime": 50.8786,
+      "eval_samples_per_second": 39.84,
+      "eval_steps_per_second": 4.992,
+      "eval_wer": 0.3744358187398447,
+      "step": 58800
+    },
+    {
+      "epoch": 6.31,
+      "eval_loss": 0.44936132431030273,
+      "eval_runtime": 50.0403,
+      "eval_samples_per_second": 40.507,
+      "eval_steps_per_second": 5.076,
+      "eval_wer": 0.3768730817837155,
+      "step": 58900
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.0001743330945558739,
+      "loss": 0.1998,
+      "step": 59000
+    },
+    {
+      "epoch": 6.32,
+      "eval_loss": 0.45291775465011597,
+      "eval_runtime": 50.0663,
+      "eval_samples_per_second": 40.486,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.36441595955948725,
+      "step": 59000
+    },
+    {
+      "epoch": 6.33,
+      "eval_loss": 0.44164779782295227,
+      "eval_runtime": 50.3619,
+      "eval_samples_per_second": 40.249,
+      "eval_steps_per_second": 5.043,
+      "eval_wer": 0.3662213395919841,
+      "step": 59100
+    },
+    {
+      "epoch": 6.34,
+      "eval_loss": 0.44675391912460327,
+      "eval_runtime": 50.2818,
+      "eval_samples_per_second": 40.313,
+      "eval_steps_per_second": 5.052,
+      "eval_wer": 0.37849792381296266,
+      "step": 59200
+    },
+    {
+      "epoch": 6.35,
+      "eval_loss": 0.43772903084754944,
+      "eval_runtime": 51.8248,
+      "eval_samples_per_second": 39.113,
+      "eval_steps_per_second": 4.901,
+      "eval_wer": 0.3664018775952338,
+      "step": 59300
+    },
+    {
+      "epoch": 6.36,
+      "eval_loss": 0.46470198035240173,
+      "eval_runtime": 49.603,
+      "eval_samples_per_second": 40.865,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.37551904675934283,
+      "step": 59400
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.00017325859598853866,
+      "loss": 0.2009,
+      "step": 59500
+    },
+    {
+      "epoch": 6.37,
+      "eval_loss": 0.46995046734809875,
+      "eval_runtime": 49.547,
+      "eval_samples_per_second": 40.911,
+      "eval_steps_per_second": 5.126,
+      "eval_wer": 0.38237949088283085,
+      "step": 59500
+    },
+    {
+      "epoch": 6.38,
+      "eval_loss": 0.448755145072937,
+      "eval_runtime": 49.6968,
+      "eval_samples_per_second": 40.787,
+      "eval_steps_per_second": 5.111,
+      "eval_wer": 0.3684780646326052,
+      "step": 59600
+    },
+    {
+      "epoch": 6.39,
+      "eval_loss": 0.4649437665939331,
+      "eval_runtime": 49.5562,
+      "eval_samples_per_second": 40.903,
+      "eval_steps_per_second": 5.125,
+      "eval_wer": 0.3803935728470843,
+      "step": 59700
+    },
+    {
+      "epoch": 6.4,
+      "eval_loss": 0.4388504922389984,
+      "eval_runtime": 49.9037,
+      "eval_samples_per_second": 40.618,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.3689294096407294,
+      "step": 59800
+    },
+    {
+      "epoch": 6.41,
+      "eval_loss": 0.4456236958503723,
+      "eval_runtime": 49.5129,
+      "eval_samples_per_second": 40.939,
+      "eval_steps_per_second": 5.13,
+      "eval_wer": 0.353132334356382,
+      "step": 59900
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.00017218409742120343,
+      "loss": 0.2007,
+      "step": 60000
+    },
+    {
+      "epoch": 6.42,
+      "eval_loss": 0.45720431208610535,
+      "eval_runtime": 49.5962,
+      "eval_samples_per_second": 40.87,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.3657699945838599,
+      "step": 60000
+    },
+    {
+      "epoch": 6.43,
+      "eval_loss": 0.44642525911331177,
+      "eval_runtime": 49.5448,
+      "eval_samples_per_second": 40.912,
+      "eval_steps_per_second": 5.127,
+      "eval_wer": 0.366853222603358,
+      "step": 60100
+    },
+    {
+      "epoch": 6.45,
+      "eval_loss": 0.46656882762908936,
+      "eval_runtime": 49.5603,
+      "eval_samples_per_second": 40.9,
+      "eval_steps_per_second": 5.125,
+      "eval_wer": 0.3710958656797256,
+      "step": 60200
+    },
+    {
+      "epoch": 6.46,
+      "eval_loss": 0.4399396777153015,
+      "eval_runtime": 49.6905,
+      "eval_samples_per_second": 40.793,
+      "eval_steps_per_second": 5.112,
+      "eval_wer": 0.3659505325871096,
+      "step": 60300
+    },
+    {
+      "epoch": 6.47,
+      "eval_loss": 0.44454124569892883,
+      "eval_runtime": 49.5472,
+      "eval_samples_per_second": 40.91,
+      "eval_steps_per_second": 5.126,
+      "eval_wer": 0.36306192453511466,
+      "step": 60400
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.00017111174785100285,
+      "loss": 0.2005,
+      "step": 60500
+    },
+    {
+      "epoch": 6.48,
+      "eval_loss": 0.4450042247772217,
+      "eval_runtime": 49.7013,
+      "eval_samples_per_second": 40.784,
+      "eval_steps_per_second": 5.111,
+      "eval_wer": 0.3620689655172414,
+      "step": 60500
+    },
+    {
+      "epoch": 6.49,
+      "eval_loss": 0.4346451163291931,
+      "eval_runtime": 49.7485,
+      "eval_samples_per_second": 40.745,
+      "eval_steps_per_second": 5.106,
+      "eval_wer": 0.35710417042787507,
+      "step": 60600
+    },
+    {
+      "epoch": 6.5,
+      "eval_loss": 0.43582871556282043,
+      "eval_runtime": 49.5116,
+      "eval_samples_per_second": 40.94,
+      "eval_steps_per_second": 5.13,
+      "eval_wer": 0.35809712944574834,
+      "step": 60700
+    },
+    {
+      "epoch": 6.51,
+      "eval_loss": 0.43437883257865906,
+      "eval_runtime": 49.4394,
+      "eval_samples_per_second": 41.0,
+      "eval_steps_per_second": 5.138,
+      "eval_wer": 0.36459649756273693,
+      "step": 60800
+    },
+    {
+      "epoch": 6.52,
+      "eval_loss": 0.4377478063106537,
+      "eval_runtime": 49.4072,
+      "eval_samples_per_second": 41.026,
+      "eval_steps_per_second": 5.141,
+      "eval_wer": 0.3620689655172414,
+      "step": 60900
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.0001700372492836676,
+      "loss": 0.2038,
+      "step": 61000
+    },
+    {
+      "epoch": 6.53,
+      "eval_loss": 0.42619436979293823,
+      "eval_runtime": 49.2219,
+      "eval_samples_per_second": 41.181,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.3570139014262502,
+      "step": 61000
+    },
+    {
+      "epoch": 6.54,
+      "eval_loss": 0.4269231855869293,
+      "eval_runtime": 49.1941,
+      "eval_samples_per_second": 41.204,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3614370825058675,
+      "step": 61100
+    },
+    {
+      "epoch": 6.55,
+      "eval_loss": 0.42971500754356384,
+      "eval_runtime": 49.2417,
+      "eval_samples_per_second": 41.164,
+      "eval_steps_per_second": 5.158,
+      "eval_wer": 0.35918035746524646,
+      "step": 61200
+    },
+    {
+      "epoch": 6.56,
+      "eval_loss": 0.4432744085788727,
+      "eval_runtime": 49.2842,
+      "eval_samples_per_second": 41.129,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.36820725762773066,
+      "step": 61300
+    },
+    {
+      "epoch": 6.57,
+      "eval_loss": 0.44737276434898376,
+      "eval_runtime": 49.0666,
+      "eval_samples_per_second": 41.311,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.36441595955948725,
+      "step": 61400
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.00016896275071633237,
+      "loss": 0.199,
+      "step": 61500
+    },
+    {
+      "epoch": 6.58,
+      "eval_loss": 0.44639796018600464,
+      "eval_runtime": 49.2396,
+      "eval_samples_per_second": 41.166,
+      "eval_steps_per_second": 5.158,
+      "eval_wer": 0.3678461816212313,
+      "step": 61500
+    },
+    {
+      "epoch": 6.6,
+      "eval_loss": 0.43974536657333374,
+      "eval_runtime": 50.0219,
+      "eval_samples_per_second": 40.522,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.35620148041162664,
+      "step": 61600
+    },
+    {
+      "epoch": 6.61,
+      "eval_loss": 0.4415305256843567,
+      "eval_runtime": 49.2213,
+      "eval_samples_per_second": 41.181,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.36116627550099295,
+      "step": 61700
+    },
+    {
+      "epoch": 6.62,
+      "eval_loss": 0.4362453520298004,
+      "eval_runtime": 49.2548,
+      "eval_samples_per_second": 41.153,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.36008304748149483,
+      "step": 61800
+    },
+    {
+      "epoch": 6.63,
+      "eval_loss": 0.44418972730636597,
+      "eval_runtime": 49.1447,
+      "eval_samples_per_second": 41.246,
+      "eval_steps_per_second": 5.168,
+      "eval_wer": 0.3623397725221159,
+      "step": 61900
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.00016788825214899712,
+      "loss": 0.1995,
+      "step": 62000
+    },
+    {
+      "epoch": 6.64,
+      "eval_loss": 0.45579132437705994,
+      "eval_runtime": 49.0853,
+      "eval_samples_per_second": 41.295,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3662213395919841,
+      "step": 62000
+    },
+    {
+      "epoch": 6.65,
+      "eval_loss": 0.44767463207244873,
+      "eval_runtime": 49.0065,
+      "eval_samples_per_second": 41.362,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.3646867665643618,
+      "step": 62100
+    },
+    {
+      "epoch": 6.66,
+      "eval_loss": 0.4542370140552521,
+      "eval_runtime": 49.7999,
+      "eval_samples_per_second": 40.703,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.3699223686586026,
+      "step": 62200
+    },
+    {
+      "epoch": 6.67,
+      "eval_loss": 0.4411347508430481,
+      "eval_runtime": 49.1039,
+      "eval_samples_per_second": 41.28,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.36324246253836434,
+      "step": 62300
+    },
+    {
+      "epoch": 6.68,
+      "eval_loss": 0.44079679250717163,
+      "eval_runtime": 48.9326,
+      "eval_samples_per_second": 41.424,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.3657699945838599,
+      "step": 62400
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.00016681590257879654,
+      "loss": 0.2014,
+      "step": 62500
+    },
+    {
+      "epoch": 6.69,
+      "eval_loss": 0.4426039755344391,
+      "eval_runtime": 50.1246,
+      "eval_samples_per_second": 40.439,
+      "eval_steps_per_second": 5.067,
+      "eval_wer": 0.36910994764397903,
+      "step": 62500
+    },
+    {
+      "epoch": 6.7,
+      "eval_loss": 0.42458194494247437,
+      "eval_runtime": 49.2003,
+      "eval_samples_per_second": 41.199,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3645062285611121,
+      "step": 62600
+    },
+    {
+      "epoch": 6.71,
+      "eval_loss": 0.4466419219970703,
+      "eval_runtime": 49.1925,
+      "eval_samples_per_second": 41.205,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.36757537461635675,
+      "step": 62700
+    },
+    {
+      "epoch": 6.72,
+      "eval_loss": 0.4493255913257599,
+      "eval_runtime": 49.3043,
+      "eval_samples_per_second": 41.112,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.356562556418126,
+      "step": 62800
+    },
+    {
+      "epoch": 6.73,
+      "eval_loss": 0.43360260128974915,
+      "eval_runtime": 49.2395,
+      "eval_samples_per_second": 41.166,
+      "eval_steps_per_second": 5.158,
+      "eval_wer": 0.3620689655172414,
+      "step": 62900
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.00016574140401146131,
+      "loss": 0.2015,
+      "step": 63000
+    },
+    {
+      "epoch": 6.75,
+      "eval_loss": 0.4367372393608093,
+      "eval_runtime": 49.3894,
+      "eval_samples_per_second": 41.041,
+      "eval_steps_per_second": 5.143,
+      "eval_wer": 0.36035385448636936,
+      "step": 63000
+    },
+    {
+      "epoch": 6.76,
+      "eval_loss": 0.4424467086791992,
+      "eval_runtime": 50.5332,
+      "eval_samples_per_second": 40.112,
+      "eval_steps_per_second": 5.026,
+      "eval_wer": 0.375428777757718,
+      "step": 63100
+    },
+    {
+      "epoch": 6.77,
+      "eval_loss": 0.46792829036712646,
+      "eval_runtime": 49.3356,
+      "eval_samples_per_second": 41.086,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.3732623217187218,
+      "step": 63200
+    },
+    {
+      "epoch": 6.78,
+      "eval_loss": 0.4482695758342743,
+      "eval_runtime": 49.3364,
+      "eval_samples_per_second": 41.085,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.37515797075284346,
+      "step": 63300
+    },
+    {
+      "epoch": 6.79,
+      "eval_loss": 0.47460225224494934,
+      "eval_runtime": 49.3305,
+      "eval_samples_per_second": 41.09,
+      "eval_steps_per_second": 5.149,
+      "eval_wer": 0.38219895287958117,
+      "step": 63400
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 0.00016466690544412606,
+      "loss": 0.2048,
+      "step": 63500
+    },
+    {
+      "epoch": 6.8,
+      "eval_loss": 0.4340318739414215,
+      "eval_runtime": 49.2,
+      "eval_samples_per_second": 41.199,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3730817837154721,
+      "step": 63500
+    },
+    {
+      "epoch": 6.81,
+      "eval_loss": 0.4346422851085663,
+      "eval_runtime": 49.239,
+      "eval_samples_per_second": 41.167,
+      "eval_steps_per_second": 5.159,
+      "eval_wer": 0.36306192453511466,
+      "step": 63600
+    },
+    {
+      "epoch": 6.82,
+      "eval_loss": 0.452511727809906,
+      "eval_runtime": 49.0618,
+      "eval_samples_per_second": 41.315,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.36802671962448097,
+      "step": 63700
+    },
+    {
+      "epoch": 6.83,
+      "eval_loss": 0.4359569549560547,
+      "eval_runtime": 49.4157,
+      "eval_samples_per_second": 41.019,
+      "eval_steps_per_second": 5.14,
+      "eval_wer": 0.36405488355298793,
+      "step": 63800
+    },
+    {
+      "epoch": 6.84,
+      "eval_loss": 0.42985743284225464,
+      "eval_runtime": 49.1072,
+      "eval_samples_per_second": 41.277,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.3557501354035024,
+      "step": 63900
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 0.00016359240687679083,
+      "loss": 0.2017,
+      "step": 64000
+    },
+    {
+      "epoch": 6.85,
+      "eval_loss": 0.43702712655067444,
+      "eval_runtime": 49.7634,
+      "eval_samples_per_second": 40.733,
+      "eval_steps_per_second": 5.104,
+      "eval_wer": 0.3533128723596317,
+      "step": 64000
+    },
+    {
+      "epoch": 6.86,
+      "eval_loss": 0.4292781352996826,
+      "eval_runtime": 49.4279,
+      "eval_samples_per_second": 41.009,
+      "eval_steps_per_second": 5.139,
+      "eval_wer": 0.361707889510742,
+      "step": 64100
+    },
+    {
+      "epoch": 6.87,
+      "eval_loss": 0.443094402551651,
+      "eval_runtime": 49.4116,
+      "eval_samples_per_second": 41.023,
+      "eval_steps_per_second": 5.14,
+      "eval_wer": 0.3660408015887344,
+      "step": 64200
+    },
+    {
+      "epoch": 6.88,
+      "eval_loss": 0.4362151026725769,
+      "eval_runtime": 48.7406,
+      "eval_samples_per_second": 41.588,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.36883914063910456,
+      "step": 64300
+    },
+    {
+      "epoch": 6.9,
+      "eval_loss": 0.4507463276386261,
+      "eval_runtime": 49.1068,
+      "eval_samples_per_second": 41.277,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.3647770355659866,
+      "step": 64400
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 0.00016251790830945558,
+      "loss": 0.2045,
+      "step": 64500
+    },
+    {
+      "epoch": 6.91,
+      "eval_loss": 0.44388145208358765,
+      "eval_runtime": 48.9274,
+      "eval_samples_per_second": 41.429,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.3612565445026178,
+      "step": 64500
+    },
+    {
+      "epoch": 6.92,
+      "eval_loss": 0.4249202013015747,
+      "eval_runtime": 50.6467,
+      "eval_samples_per_second": 40.022,
+      "eval_steps_per_second": 5.015,
+      "eval_wer": 0.3493410362881387,
+      "step": 64600
+    },
+    {
+      "epoch": 6.93,
+      "eval_loss": 0.436200350522995,
+      "eval_runtime": 48.8605,
+      "eval_samples_per_second": 41.485,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.36116627550099295,
+      "step": 64700
+    },
+    {
+      "epoch": 6.94,
+      "eval_loss": 0.4335877299308777,
+      "eval_runtime": 49.1047,
+      "eval_samples_per_second": 41.279,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.3584582054522477,
+      "step": 64800
+    },
+    {
+      "epoch": 6.95,
+      "eval_loss": 0.4387299120426178,
+      "eval_runtime": 49.1708,
+      "eval_samples_per_second": 41.224,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.35683336342300054,
+      "step": 64900
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 0.000161445558739255,
+      "loss": 0.1977,
+      "step": 65000
+    },
+    {
+      "epoch": 6.96,
+      "eval_loss": 0.43134012818336487,
+      "eval_runtime": 49.5932,
+      "eval_samples_per_second": 40.873,
+      "eval_steps_per_second": 5.122,
+      "eval_wer": 0.35421556237588014,
+      "step": 65000
+    },
+    {
+      "epoch": 6.97,
+      "eval_loss": 0.42868292331695557,
+      "eval_runtime": 49.0964,
+      "eval_samples_per_second": 41.286,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.35520852139375336,
+      "step": 65100
+    },
+    {
+      "epoch": 6.98,
+      "eval_loss": 0.4371795654296875,
+      "eval_runtime": 49.0695,
+      "eval_samples_per_second": 41.309,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.3586387434554974,
+      "step": 65200
+    },
+    {
+      "epoch": 6.99,
+      "eval_loss": 0.43784716725349426,
+      "eval_runtime": 48.9834,
+      "eval_samples_per_second": 41.381,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.36288138653186497,
+      "step": 65300
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 0.4518013298511505,
+      "eval_runtime": 49.3743,
+      "eval_samples_per_second": 41.054,
+      "eval_steps_per_second": 5.144,
+      "eval_wer": 0.3639646145513631,
+      "step": 65400
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 0.00016037106017191977,
+      "loss": 0.1971,
+      "step": 65500
+    },
+    {
+      "epoch": 7.01,
+      "eval_loss": 0.44795140624046326,
+      "eval_runtime": 48.9935,
+      "eval_samples_per_second": 41.373,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.3556598664018776,
+      "step": 65500
+    },
+    {
+      "epoch": 7.02,
+      "eval_loss": 0.452963650226593,
+      "eval_runtime": 49.4173,
+      "eval_samples_per_second": 41.018,
+      "eval_steps_per_second": 5.14,
+      "eval_wer": 0.35602094240837695,
+      "step": 65600
+    },
+    {
+      "epoch": 7.03,
+      "eval_loss": 0.4580971300601959,
+      "eval_runtime": 50.065,
+      "eval_samples_per_second": 40.487,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.3581873984473732,
+      "step": 65700
+    },
+    {
+      "epoch": 7.04,
+      "eval_loss": 0.44924312829971313,
+      "eval_runtime": 49.1382,
+      "eval_samples_per_second": 41.251,
+      "eval_steps_per_second": 5.169,
+      "eval_wer": 0.354305831377505,
+      "step": 65800
+    },
+    {
+      "epoch": 7.06,
+      "eval_loss": 0.44480279088020325,
+      "eval_runtime": 49.2997,
+      "eval_samples_per_second": 41.116,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.3608051994944936,
+      "step": 65900
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 0.0001592965616045845,
+      "loss": 0.1672,
+      "step": 66000
+    },
+    {
+      "epoch": 7.07,
+      "eval_loss": 0.4469369351863861,
+      "eval_runtime": 49.2087,
+      "eval_samples_per_second": 41.192,
+      "eval_steps_per_second": 5.162,
+      "eval_wer": 0.354305831377505,
+      "step": 66000
+    },
+    {
+      "epoch": 7.08,
+      "eval_loss": 0.42615774273872375,
+      "eval_runtime": 49.8244,
+      "eval_samples_per_second": 40.683,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.3487994222783896,
+      "step": 66100
+    },
+    {
+      "epoch": 7.09,
+      "eval_loss": 0.4289032220840454,
+      "eval_runtime": 49.6551,
+      "eval_samples_per_second": 40.822,
+      "eval_steps_per_second": 5.115,
+      "eval_wer": 0.3570139014262502,
+      "step": 66200
+    },
+    {
+      "epoch": 7.1,
+      "eval_loss": 0.4455362558364868,
+      "eval_runtime": 49.8614,
+      "eval_samples_per_second": 40.653,
+      "eval_steps_per_second": 5.094,
+      "eval_wer": 0.3544863693807547,
+      "step": 66300
+    },
+    {
+      "epoch": 7.11,
+      "eval_loss": 0.4449104070663452,
+      "eval_runtime": 49.8819,
+      "eval_samples_per_second": 40.636,
+      "eval_steps_per_second": 5.092,
+      "eval_wer": 0.3562917494132515,
+      "step": 66400
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 0.0001582220630372493,
+      "loss": 0.169,
+      "step": 66500
+    },
+    {
+      "epoch": 7.12,
+      "eval_loss": 0.4554521143436432,
+      "eval_runtime": 49.8243,
+      "eval_samples_per_second": 40.683,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.35647228741650117,
+      "step": 66500
+    },
+    {
+      "epoch": 7.13,
+      "eval_loss": 0.443204402923584,
+      "eval_runtime": 50.0734,
+      "eval_samples_per_second": 40.481,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.3655894565806102,
+      "step": 66600
+    },
+    {
+      "epoch": 7.14,
+      "eval_loss": 0.4398520588874817,
+      "eval_runtime": 49.6006,
+      "eval_samples_per_second": 40.866,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.36098573749774326,
+      "step": 66700
+    },
+    {
+      "epoch": 7.15,
+      "eval_loss": 0.4383464753627777,
+      "eval_runtime": 49.6217,
+      "eval_samples_per_second": 40.849,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.35538905939700305,
+      "step": 66800
+    },
+    {
+      "epoch": 7.16,
+      "eval_loss": 0.43758177757263184,
+      "eval_runtime": 50.6011,
+      "eval_samples_per_second": 40.058,
+      "eval_steps_per_second": 5.02,
+      "eval_wer": 0.35358367936450624,
+      "step": 66900
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 0.0001571497134670487,
+      "loss": 0.1724,
+      "step": 67000
+    },
+    {
+      "epoch": 7.17,
+      "eval_loss": 0.4382694363594055,
+      "eval_runtime": 49.755,
+      "eval_samples_per_second": 40.74,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.3571944394294999,
+      "step": 67000
+    },
+    {
+      "epoch": 7.18,
+      "eval_loss": 0.4452250301837921,
+      "eval_runtime": 49.6952,
+      "eval_samples_per_second": 40.789,
+      "eval_steps_per_second": 5.111,
+      "eval_wer": 0.3534934103628814,
+      "step": 67100
+    },
+    {
+      "epoch": 7.19,
+      "eval_loss": 0.4610207974910736,
+      "eval_runtime": 50.0238,
+      "eval_samples_per_second": 40.521,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.36676295360173317,
+      "step": 67200
+    },
+    {
+      "epoch": 7.21,
+      "eval_loss": 0.45337799191474915,
+      "eval_runtime": 49.5531,
+      "eval_samples_per_second": 40.906,
+      "eval_steps_per_second": 5.126,
+      "eval_wer": 0.3545766383823795,
+      "step": 67300
+    },
+    {
+      "epoch": 7.22,
+      "eval_loss": 0.450591504573822,
+      "eval_runtime": 49.4218,
+      "eval_samples_per_second": 41.014,
+      "eval_steps_per_second": 5.139,
+      "eval_wer": 0.3604441234879942,
+      "step": 67400
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 0.0001560773638968481,
+      "loss": 0.1729,
+      "step": 67500
+    },
+    {
+      "epoch": 7.23,
+      "eval_loss": 0.4463426470756531,
+      "eval_runtime": 49.5524,
+      "eval_samples_per_second": 40.906,
+      "eval_steps_per_second": 5.126,
+      "eval_wer": 0.35069507131251126,
+      "step": 67500
+    },
+    {
+      "epoch": 7.24,
+      "eval_loss": 0.44396212697029114,
+      "eval_runtime": 49.687,
+      "eval_samples_per_second": 40.795,
+      "eval_steps_per_second": 5.112,
+      "eval_wer": 0.3629716555334898,
+      "step": 67600
+    },
+    {
+      "epoch": 7.25,
+      "eval_loss": 0.4361093044281006,
+      "eval_runtime": 50.0356,
+      "eval_samples_per_second": 40.511,
+      "eval_steps_per_second": 5.076,
+      "eval_wer": 0.3550279833905037,
+      "step": 67700
+    },
+    {
+      "epoch": 7.26,
+      "eval_loss": 0.4397481679916382,
+      "eval_runtime": 49.5204,
+      "eval_samples_per_second": 40.933,
+      "eval_steps_per_second": 5.129,
+      "eval_wer": 0.3643256905578624,
+      "step": 67800
+    },
+    {
+      "epoch": 7.27,
+      "eval_loss": 0.43276941776275635,
+      "eval_runtime": 49.9383,
+      "eval_samples_per_second": 40.59,
+      "eval_steps_per_second": 5.086,
+      "eval_wer": 0.3547571763856292,
+      "step": 67900
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 0.00015500286532951288,
+      "loss": 0.1736,
+      "step": 68000
+    },
+    {
+      "epoch": 7.28,
+      "eval_loss": 0.45456382632255554,
+      "eval_runtime": 49.2784,
+      "eval_samples_per_second": 41.134,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.3614370825058675,
+      "step": 68000
+    },
+    {
+      "epoch": 7.29,
+      "eval_loss": 0.45059072971343994,
+      "eval_runtime": 49.8297,
+      "eval_samples_per_second": 40.679,
+      "eval_steps_per_second": 5.097,
+      "eval_wer": 0.35584040440512726,
+      "step": 68100
+    },
+    {
+      "epoch": 7.3,
+      "eval_loss": 0.4360811114311218,
+      "eval_runtime": 49.6076,
+      "eval_samples_per_second": 40.861,
+      "eval_steps_per_second": 5.12,
+      "eval_wer": 0.35132695432388517,
+      "step": 68200
+    },
+    {
+      "epoch": 7.31,
+      "eval_loss": 0.4222584366798401,
+      "eval_runtime": 49.813,
+      "eval_samples_per_second": 40.692,
+      "eval_steps_per_second": 5.099,
+      "eval_wer": 0.3499729192995126,
+      "step": 68300
+    },
+    {
+      "epoch": 7.32,
+      "eval_loss": 0.4474307596683502,
+      "eval_runtime": 49.5259,
+      "eval_samples_per_second": 40.928,
+      "eval_steps_per_second": 5.129,
+      "eval_wer": 0.34970211229463805,
+      "step": 68400
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 0.00015392836676217763,
+      "loss": 0.1733,
+      "step": 68500
+    },
+    {
+      "epoch": 7.33,
+      "eval_loss": 0.4303099513053894,
+      "eval_runtime": 49.5599,
+      "eval_samples_per_second": 40.9,
+      "eval_steps_per_second": 5.125,
+      "eval_wer": 0.35493771438887883,
+      "step": 68500
+    },
+    {
+      "epoch": 7.34,
+      "eval_loss": 0.42649605870246887,
+      "eval_runtime": 49.6335,
+      "eval_samples_per_second": 40.839,
+      "eval_steps_per_second": 5.118,
+      "eval_wer": 0.3483480772702654,
+      "step": 68600
+    },
+    {
+      "epoch": 7.36,
+      "eval_loss": 0.43393298983573914,
+      "eval_runtime": 50.0205,
+      "eval_samples_per_second": 40.523,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.3557501354035024,
+      "step": 68700
+    },
+    {
+      "epoch": 7.37,
+      "eval_loss": 0.4266127347946167,
+      "eval_runtime": 50.1516,
+      "eval_samples_per_second": 40.417,
+      "eval_steps_per_second": 5.065,
+      "eval_wer": 0.34907022928326414,
+      "step": 68800
+    },
+    {
+      "epoch": 7.38,
+      "eval_loss": 0.44230374693870544,
+      "eval_runtime": 49.8261,
+      "eval_samples_per_second": 40.681,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.35647228741650117,
+      "step": 68900
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 0.00015285386819484238,
+      "loss": 0.1764,
+      "step": 69000
+    },
+    {
+      "epoch": 7.39,
+      "eval_loss": 0.44104790687561035,
+      "eval_runtime": 49.6079,
+      "eval_samples_per_second": 40.86,
+      "eval_steps_per_second": 5.12,
+      "eval_wer": 0.35538905939700305,
+      "step": 69000
+    },
+    {
+      "epoch": 7.4,
+      "eval_loss": 0.44824397563934326,
+      "eval_runtime": 49.4864,
+      "eval_samples_per_second": 40.961,
+      "eval_steps_per_second": 5.133,
+      "eval_wer": 0.370283444665102,
+      "step": 69100
+    },
+    {
+      "epoch": 7.41,
+      "eval_loss": 0.4479742646217346,
+      "eval_runtime": 49.5287,
+      "eval_samples_per_second": 40.926,
+      "eval_steps_per_second": 5.128,
+      "eval_wer": 0.36405488355298793,
+      "step": 69200
+    },
+    {
+      "epoch": 7.42,
+      "eval_loss": 0.4361330270767212,
+      "eval_runtime": 49.8379,
+      "eval_samples_per_second": 40.672,
+      "eval_steps_per_second": 5.097,
+      "eval_wer": 0.3499729192995126,
+      "step": 69300
+    },
+    {
+      "epoch": 7.43,
+      "eval_loss": 0.4399217963218689,
+      "eval_runtime": 50.0303,
+      "eval_samples_per_second": 40.515,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.36324246253836434,
+      "step": 69400
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 0.00015177936962750715,
+      "loss": 0.1711,
+      "step": 69500
+    },
+    {
+      "epoch": 7.44,
+      "eval_loss": 0.43825072050094604,
+      "eval_runtime": 50.1866,
+      "eval_samples_per_second": 40.389,
+      "eval_steps_per_second": 5.061,
+      "eval_wer": 0.3590900884636216,
+      "step": 69500
+    },
+    {
+      "epoch": 7.45,
+      "eval_loss": 0.45228099822998047,
+      "eval_runtime": 49.884,
+      "eval_samples_per_second": 40.634,
+      "eval_steps_per_second": 5.092,
+      "eval_wer": 0.3636035385448637,
+      "step": 69600
+    },
+    {
+      "epoch": 7.46,
+      "eval_loss": 0.438789963722229,
+      "eval_runtime": 49.7132,
+      "eval_samples_per_second": 40.774,
+      "eval_steps_per_second": 5.109,
+      "eval_wer": 0.35024372630438705,
+      "step": 69700
+    },
+    {
+      "epoch": 7.47,
+      "eval_loss": 0.430462509393692,
+      "eval_runtime": 49.7445,
+      "eval_samples_per_second": 40.748,
+      "eval_steps_per_second": 5.106,
+      "eval_wer": 0.35647228741650117,
+      "step": 69800
+    },
+    {
+      "epoch": 7.48,
+      "eval_loss": 0.4289742410182953,
+      "eval_runtime": 49.8969,
+      "eval_samples_per_second": 40.624,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.35376421736775593,
+      "step": 69900
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 0.0001507048710601719,
+      "loss": 0.1748,
+      "step": 70000
+    },
+    {
+      "epoch": 7.49,
+      "eval_loss": 0.435860276222229,
+      "eval_runtime": 49.6706,
+      "eval_samples_per_second": 40.809,
+      "eval_steps_per_second": 5.114,
+      "eval_wer": 0.3511464163206355,
+      "step": 70000
+    },
+    {
+      "epoch": 7.51,
+      "eval_loss": 0.4314829707145691,
+      "eval_runtime": 49.5429,
+      "eval_samples_per_second": 40.914,
+      "eval_steps_per_second": 5.127,
+      "eval_wer": 0.3460010832280195,
+      "step": 70100
+    },
+    {
+      "epoch": 7.52,
+      "eval_loss": 0.42675673961639404,
+      "eval_runtime": 49.8146,
+      "eval_samples_per_second": 40.691,
+      "eval_steps_per_second": 5.099,
+      "eval_wer": 0.3554793283986279,
+      "step": 70200
+    },
+    {
+      "epoch": 7.53,
+      "eval_loss": 0.42668795585632324,
+      "eval_runtime": 49.5577,
+      "eval_samples_per_second": 40.902,
+      "eval_steps_per_second": 5.125,
+      "eval_wer": 0.34554973821989526,
+      "step": 70300
+    },
+    {
+      "epoch": 7.54,
+      "eval_loss": 0.43592557311058044,
+      "eval_runtime": 49.9069,
+      "eval_samples_per_second": 40.616,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.35168803033038454,
+      "step": 70400
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 0.00014963037249283667,
+      "loss": 0.1739,
+      "step": 70500
+    },
+    {
+      "epoch": 7.55,
+      "eval_loss": 0.4299277067184448,
+      "eval_runtime": 49.4302,
+      "eval_samples_per_second": 41.007,
+      "eval_steps_per_second": 5.139,
+      "eval_wer": 0.34907022928326414,
+      "step": 70500
+    },
+    {
+      "epoch": 7.56,
+      "eval_loss": 0.4423278570175171,
+      "eval_runtime": 50.5024,
+      "eval_samples_per_second": 40.137,
+      "eval_steps_per_second": 5.029,
+      "eval_wer": 0.3409460191370283,
+      "step": 70600
+    },
+    {
+      "epoch": 7.57,
+      "eval_loss": 0.425134539604187,
+      "eval_runtime": 49.993,
+      "eval_samples_per_second": 40.546,
+      "eval_steps_per_second": 5.081,
+      "eval_wer": 0.34202924715652644,
+      "step": 70700
+    },
+    {
+      "epoch": 7.58,
+      "eval_loss": 0.4300227463245392,
+      "eval_runtime": 49.8429,
+      "eval_samples_per_second": 40.668,
+      "eval_steps_per_second": 5.096,
+      "eval_wer": 0.34139736414515254,
+      "step": 70800
+    },
+    {
+      "epoch": 7.59,
+      "eval_loss": 0.43489590287208557,
+      "eval_runtime": 49.9603,
+      "eval_samples_per_second": 40.572,
+      "eval_steps_per_second": 5.084,
+      "eval_wer": 0.3422097851597761,
+      "step": 70900
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 0.0001485558739255014,
+      "loss": 0.1763,
+      "step": 71000
+    },
+    {
+      "epoch": 7.6,
+      "eval_loss": 0.432796835899353,
+      "eval_runtime": 49.7256,
+      "eval_samples_per_second": 40.764,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.34184870915327675,
+      "step": 71000
+    },
+    {
+      "epoch": 7.61,
+      "eval_loss": 0.4312654137611389,
+      "eval_runtime": 49.7843,
+      "eval_samples_per_second": 40.716,
+      "eval_steps_per_second": 5.102,
+      "eval_wer": 0.34518866221339595,
+      "step": 71100
+    },
+    {
+      "epoch": 7.62,
+      "eval_loss": 0.42399337887763977,
+      "eval_runtime": 49.7353,
+      "eval_samples_per_second": 40.756,
+      "eval_steps_per_second": 5.107,
+      "eval_wer": 0.35340314136125656,
+      "step": 71200
+    },
+    {
+      "epoch": 7.63,
+      "eval_loss": 0.42743441462516785,
+      "eval_runtime": 49.616,
+      "eval_samples_per_second": 40.854,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.34744538725401697,
+      "step": 71300
+    },
+    {
+      "epoch": 7.64,
+      "eval_loss": 0.43036043643951416,
+      "eval_runtime": 49.8989,
+      "eval_samples_per_second": 40.622,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.3467232352410182,
+      "step": 71400
+    },
+    {
+      "epoch": 7.66,
+      "learning_rate": 0.00014748137535816616,
+      "loss": 0.171,
+      "step": 71500
+    },
+    {
+      "epoch": 7.66,
+      "eval_loss": 0.43307775259017944,
+      "eval_runtime": 49.8783,
+      "eval_samples_per_second": 40.639,
+      "eval_steps_per_second": 5.092,
+      "eval_wer": 0.3509658783173858,
+      "step": 71500
+    },
+    {
+      "epoch": 7.67,
+      "eval_loss": 0.42630788683891296,
+      "eval_runtime": 49.6005,
+      "eval_samples_per_second": 40.866,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.34780646326051634,
+      "step": 71600
+    },
+    {
+      "epoch": 7.68,
+      "eval_loss": 0.4301213026046753,
+      "eval_runtime": 49.8474,
+      "eval_samples_per_second": 40.664,
+      "eval_steps_per_second": 5.096,
+      "eval_wer": 0.34473731720527173,
+      "step": 71700
+    },
+    {
+      "epoch": 7.69,
+      "eval_loss": 0.40457987785339355,
+      "eval_runtime": 49.7536,
+      "eval_samples_per_second": 40.741,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.34518866221339595,
+      "step": 71800
+    },
+    {
+      "epoch": 7.7,
+      "eval_loss": 0.42995375394821167,
+      "eval_runtime": 50.5859,
+      "eval_samples_per_second": 40.07,
+      "eval_steps_per_second": 5.021,
+      "eval_wer": 0.35277125834988265,
+      "step": 71900
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 0.00014641117478510026,
+      "loss": 0.1792,
+      "step": 72000
+    },
+    {
+      "epoch": 7.71,
+      "eval_loss": 0.42525142431259155,
+      "eval_runtime": 49.9158,
+      "eval_samples_per_second": 40.608,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.349160498284889,
+      "step": 72000
+    },
+    {
+      "epoch": 7.72,
+      "eval_loss": 0.4295995533466339,
+      "eval_runtime": 50.2083,
+      "eval_samples_per_second": 40.372,
+      "eval_steps_per_second": 5.059,
+      "eval_wer": 0.34907022928326414,
+      "step": 72100
+    },
+    {
+      "epoch": 7.73,
+      "eval_loss": 0.4117787778377533,
+      "eval_runtime": 50.1254,
+      "eval_samples_per_second": 40.439,
+      "eval_steps_per_second": 5.067,
+      "eval_wer": 0.3450983932117711,
+      "step": 72200
+    },
+    {
+      "epoch": 7.74,
+      "eval_loss": 0.4347759485244751,
+      "eval_runtime": 49.7417,
+      "eval_samples_per_second": 40.751,
+      "eval_steps_per_second": 5.106,
+      "eval_wer": 0.3345369200216646,
+      "step": 72300
+    },
+    {
+      "epoch": 7.75,
+      "eval_loss": 0.4282897412776947,
+      "eval_runtime": 49.9268,
+      "eval_samples_per_second": 40.599,
+      "eval_steps_per_second": 5.087,
+      "eval_wer": 0.34473731720527173,
+      "step": 72400
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 0.00014533667621776503,
+      "loss": 0.1801,
+      "step": 72500
+    },
+    {
+      "epoch": 7.76,
+      "eval_loss": 0.423183798789978,
+      "eval_runtime": 50.0273,
+      "eval_samples_per_second": 40.518,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.3449178552085214,
+      "step": 72500
+    },
+    {
+      "epoch": 7.77,
+      "eval_loss": 0.4490908980369568,
+      "eval_runtime": 49.8341,
+      "eval_samples_per_second": 40.675,
+      "eval_steps_per_second": 5.097,
+      "eval_wer": 0.34861888427513993,
+      "step": 72600
+    },
+    {
+      "epoch": 7.78,
+      "eval_loss": 0.42611071467399597,
+      "eval_runtime": 49.9659,
+      "eval_samples_per_second": 40.568,
+      "eval_steps_per_second": 5.083,
+      "eval_wer": 0.33426611301679005,
+      "step": 72700
+    },
+    {
+      "epoch": 7.79,
+      "eval_loss": 0.43815183639526367,
+      "eval_runtime": 49.9365,
+      "eval_samples_per_second": 40.592,
+      "eval_steps_per_second": 5.086,
+      "eval_wer": 0.34554973821989526,
+      "step": 72800
+    },
+    {
+      "epoch": 7.81,
+      "eval_loss": 0.43009382486343384,
+      "eval_runtime": 49.5744,
+      "eval_samples_per_second": 40.888,
+      "eval_steps_per_second": 5.124,
+      "eval_wer": 0.3414876331467774,
+      "step": 72900
+    },
+    {
+      "epoch": 7.82,
+      "learning_rate": 0.00014426432664756445,
+      "loss": 0.1731,
+      "step": 73000
+    },
+    {
+      "epoch": 7.82,
+      "eval_loss": 0.42360547184944153,
+      "eval_runtime": 50.0488,
+      "eval_samples_per_second": 40.5,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.3438346271890233,
+      "step": 73000
+    },
+    {
+      "epoch": 7.83,
+      "eval_loss": 0.42565011978149414,
+      "eval_runtime": 49.9316,
+      "eval_samples_per_second": 40.596,
+      "eval_steps_per_second": 5.087,
+      "eval_wer": 0.3419389781549016,
+      "step": 73100
+    },
+    {
+      "epoch": 7.84,
+      "eval_loss": 0.4367884397506714,
+      "eval_runtime": 49.8253,
+      "eval_samples_per_second": 40.682,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.34103628813865317,
+      "step": 73200
+    },
+    {
+      "epoch": 7.85,
+      "eval_loss": 0.4206501841545105,
+      "eval_runtime": 49.6241,
+      "eval_samples_per_second": 40.847,
+      "eval_steps_per_second": 5.118,
+      "eval_wer": 0.3397725221159054,
+      "step": 73300
+    },
+    {
+      "epoch": 7.86,
+      "eval_loss": 0.4118278920650482,
+      "eval_runtime": 49.7109,
+      "eval_samples_per_second": 40.776,
+      "eval_steps_per_second": 5.11,
+      "eval_wer": 0.3417584401516519,
+      "step": 73400
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 0.00014318982808022923,
+      "loss": 0.1748,
+      "step": 73500
+    },
+    {
+      "epoch": 7.87,
+      "eval_loss": 0.4357281029224396,
+      "eval_runtime": 49.9503,
+      "eval_samples_per_second": 40.58,
+      "eval_steps_per_second": 5.085,
+      "eval_wer": 0.34293193717277487,
+      "step": 73500
+    },
+    {
+      "epoch": 7.88,
+      "eval_loss": 0.4276559054851532,
+      "eval_runtime": 49.4295,
+      "eval_samples_per_second": 41.008,
+      "eval_steps_per_second": 5.139,
+      "eval_wer": 0.34518866221339595,
+      "step": 73600
+    },
+    {
+      "epoch": 7.89,
+      "eval_loss": 0.4172598719596863,
+      "eval_runtime": 51.6349,
+      "eval_samples_per_second": 39.256,
+      "eval_steps_per_second": 4.919,
+      "eval_wer": 0.34762592525726665,
+      "step": 73700
+    },
+    {
+      "epoch": 7.9,
+      "eval_loss": 0.41912218928337097,
+      "eval_runtime": 49.7354,
+      "eval_samples_per_second": 40.756,
+      "eval_steps_per_second": 5.107,
+      "eval_wer": 0.34780646326051634,
+      "step": 73800
+    },
+    {
+      "epoch": 7.91,
+      "eval_loss": 0.4197118282318115,
+      "eval_runtime": 49.7715,
+      "eval_samples_per_second": 40.726,
+      "eval_steps_per_second": 5.103,
+      "eval_wer": 0.34573027622314495,
+      "step": 73900
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 0.00014211532951289397,
+      "loss": 0.1745,
+      "step": 74000
+    },
+    {
+      "epoch": 7.92,
+      "eval_loss": 0.4197350740432739,
+      "eval_runtime": 50.0166,
+      "eval_samples_per_second": 40.527,
+      "eval_steps_per_second": 5.078,
+      "eval_wer": 0.34356382018414877,
+      "step": 74000
+    },
+    {
+      "epoch": 7.93,
+      "eval_loss": 0.4252673089504242,
+      "eval_runtime": 49.9578,
+      "eval_samples_per_second": 40.574,
+      "eval_steps_per_second": 5.084,
+      "eval_wer": 0.3512366853222603,
+      "step": 74100
+    },
+    {
+      "epoch": 7.94,
+      "eval_loss": 0.4216529130935669,
+      "eval_runtime": 49.7253,
+      "eval_samples_per_second": 40.764,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.346271890232894,
+      "step": 74200
+    },
+    {
+      "epoch": 7.95,
+      "eval_loss": 0.4305253326892853,
+      "eval_runtime": 50.2727,
+      "eval_samples_per_second": 40.32,
+      "eval_steps_per_second": 5.052,
+      "eval_wer": 0.3472648492507673,
+      "step": 74300
+    },
+    {
+      "epoch": 7.97,
+      "eval_loss": 0.42150747776031494,
+      "eval_runtime": 49.8232,
+      "eval_samples_per_second": 40.684,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.35069507131251126,
+      "step": 74400
+    },
+    {
+      "epoch": 7.98,
+      "learning_rate": 0.00014104083094555872,
+      "loss": 0.1743,
+      "step": 74500
+    },
+    {
+      "epoch": 7.98,
+      "eval_loss": 0.41268059611320496,
+      "eval_runtime": 50.0689,
+      "eval_samples_per_second": 40.484,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.34076548113377864,
+      "step": 74500
+    },
+    {
+      "epoch": 7.99,
+      "eval_loss": 0.419066846370697,
+      "eval_runtime": 49.8664,
+      "eval_samples_per_second": 40.649,
+      "eval_steps_per_second": 5.094,
+      "eval_wer": 0.34681350424264307,
+      "step": 74600
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": 0.43805456161499023,
+      "eval_runtime": 49.856,
+      "eval_samples_per_second": 40.657,
+      "eval_steps_per_second": 5.095,
+      "eval_wer": 0.34907022928326414,
+      "step": 74700
+    },
+    {
+      "epoch": 8.01,
+      "eval_loss": 0.45096319913864136,
+      "eval_runtime": 50.647,
+      "eval_samples_per_second": 40.022,
+      "eval_steps_per_second": 5.015,
+      "eval_wer": 0.3477161942588915,
+      "step": 74800
+    },
+    {
+      "epoch": 8.02,
+      "eval_loss": 0.4481862485408783,
+      "eval_runtime": 50.0491,
+      "eval_samples_per_second": 40.5,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.3470843112475176,
+      "step": 74900
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 0.0001399663323782235,
+      "loss": 0.1588,
+      "step": 75000
+    },
+    {
+      "epoch": 8.03,
+      "eval_loss": 0.4470798671245575,
+      "eval_runtime": 50.1478,
+      "eval_samples_per_second": 40.421,
+      "eval_steps_per_second": 5.065,
+      "eval_wer": 0.3430222061743997,
+      "step": 75000
+    },
+    {
+      "epoch": 8.04,
+      "eval_loss": 0.4295661449432373,
+      "eval_runtime": 49.994,
+      "eval_samples_per_second": 40.545,
+      "eval_steps_per_second": 5.081,
+      "eval_wer": 0.3393211771077812,
+      "step": 75100
+    },
+    {
+      "epoch": 8.05,
+      "eval_loss": 0.4479615390300751,
+      "eval_runtime": 50.1363,
+      "eval_samples_per_second": 40.43,
+      "eval_steps_per_second": 5.066,
+      "eval_wer": 0.3397725221159054,
+      "step": 75200
+    },
+    {
+      "epoch": 8.06,
+      "eval_loss": 0.43022599816322327,
+      "eval_runtime": 49.9468,
+      "eval_samples_per_second": 40.583,
+      "eval_steps_per_second": 5.085,
+      "eval_wer": 0.34518866221339595,
+      "step": 75300
+    },
+    {
+      "epoch": 8.07,
+      "eval_loss": 0.44102075695991516,
+      "eval_runtime": 49.7515,
+      "eval_samples_per_second": 40.742,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.34311247517602456,
+      "step": 75400
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 0.00013889183381088824,
+      "loss": 0.144,
+      "step": 75500
+    },
+    {
+      "epoch": 8.08,
+      "eval_loss": 0.426250696182251,
+      "eval_runtime": 49.7571,
+      "eval_samples_per_second": 40.738,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.34554973821989526,
+      "step": 75500
+    },
+    {
+      "epoch": 8.09,
+      "eval_loss": 0.4522950053215027,
+      "eval_runtime": 50.1232,
+      "eval_samples_per_second": 40.44,
+      "eval_steps_per_second": 5.068,
+      "eval_wer": 0.34952157429138836,
+      "step": 75600
+    },
+    {
+      "epoch": 8.1,
+      "eval_loss": 0.4454551041126251,
+      "eval_runtime": 50.0404,
+      "eval_samples_per_second": 40.507,
+      "eval_steps_per_second": 5.076,
+      "eval_wer": 0.3511464163206355,
+      "step": 75700
+    },
+    {
+      "epoch": 8.12,
+      "eval_loss": 0.43785789608955383,
+      "eval_runtime": 50.0138,
+      "eval_samples_per_second": 40.529,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.3444665102003972,
+      "step": 75800
+    },
+    {
+      "epoch": 8.13,
+      "eval_loss": 0.4417654275894165,
+      "eval_runtime": 49.6684,
+      "eval_samples_per_second": 40.811,
+      "eval_steps_per_second": 5.114,
+      "eval_wer": 0.341126557140278,
+      "step": 75900
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 0.00013781733524355298,
+      "loss": 0.1483,
+      "step": 76000
+    },
+    {
+      "epoch": 8.14,
+      "eval_loss": 0.4490506649017334,
+      "eval_runtime": 49.753,
+      "eval_samples_per_second": 40.741,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.346271890232894,
+      "step": 76000
+    },
+    {
+      "epoch": 8.15,
+      "eval_loss": 0.43855246901512146,
+      "eval_runtime": 49.593,
+      "eval_samples_per_second": 40.873,
+      "eval_steps_per_second": 5.122,
+      "eval_wer": 0.3467232352410182,
+      "step": 76100
+    },
+    {
+      "epoch": 8.16,
+      "eval_loss": 0.43268758058547974,
+      "eval_runtime": 49.787,
+      "eval_samples_per_second": 40.713,
+      "eval_steps_per_second": 5.102,
+      "eval_wer": 0.3524101823433833,
+      "step": 76200
+    },
+    {
+      "epoch": 8.17,
+      "eval_loss": 0.4359542727470398,
+      "eval_runtime": 49.6772,
+      "eval_samples_per_second": 40.803,
+      "eval_steps_per_second": 5.113,
+      "eval_wer": 0.36134681350424264,
+      "step": 76300
+    },
+    {
+      "epoch": 8.18,
+      "eval_loss": 0.4352158308029175,
+      "eval_runtime": 49.6241,
+      "eval_samples_per_second": 40.847,
+      "eval_steps_per_second": 5.118,
+      "eval_wer": 0.3497923812962629,
+      "step": 76400
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 0.00013674498567335243,
+      "loss": 0.1541,
+      "step": 76500
+    },
+    {
+      "epoch": 8.19,
+      "eval_loss": 0.43757787346839905,
+      "eval_runtime": 49.5023,
+      "eval_samples_per_second": 40.948,
+      "eval_steps_per_second": 5.131,
+      "eval_wer": 0.34139736414515254,
+      "step": 76500
+    },
+    {
+      "epoch": 8.2,
+      "eval_loss": 0.4408476650714874,
+      "eval_runtime": 49.9235,
+      "eval_samples_per_second": 40.602,
+      "eval_steps_per_second": 5.088,
+      "eval_wer": 0.34636215923451885,
+      "step": 76600
+    },
+    {
+      "epoch": 8.21,
+      "eval_loss": 0.44147056341171265,
+      "eval_runtime": 50.6402,
+      "eval_samples_per_second": 40.027,
+      "eval_steps_per_second": 5.016,
+      "eval_wer": 0.3444665102003972,
+      "step": 76700
+    },
+    {
+      "epoch": 8.22,
+      "eval_loss": 0.44549474120140076,
+      "eval_runtime": 49.8586,
+      "eval_samples_per_second": 40.655,
+      "eval_steps_per_second": 5.094,
+      "eval_wer": 0.3481675392670157,
+      "step": 76800
+    },
+    {
+      "epoch": 8.23,
+      "eval_loss": 0.45420777797698975,
+      "eval_runtime": 50.7469,
+      "eval_samples_per_second": 39.943,
+      "eval_steps_per_second": 5.005,
+      "eval_wer": 0.3414876331467774,
+      "step": 76900
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 0.00013567048710601718,
+      "loss": 0.1479,
+      "step": 77000
+    },
+    {
+      "epoch": 8.24,
+      "eval_loss": 0.44616568088531494,
+      "eval_runtime": 49.6192,
+      "eval_samples_per_second": 40.851,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.3425708611662755,
+      "step": 77000
+    },
+    {
+      "epoch": 8.25,
+      "eval_loss": 0.4459743797779083,
+      "eval_runtime": 49.6132,
+      "eval_samples_per_second": 40.856,
+      "eval_steps_per_second": 5.12,
+      "eval_wer": 0.3413070951435277,
+      "step": 77100
+    },
+    {
+      "epoch": 8.27,
+      "eval_loss": 0.4433516561985016,
+      "eval_runtime": 49.9376,
+      "eval_samples_per_second": 40.591,
+      "eval_steps_per_second": 5.086,
+      "eval_wer": 0.33751579707528434,
+      "step": 77200
+    },
+    {
+      "epoch": 8.28,
+      "eval_loss": 0.43969422578811646,
+      "eval_runtime": 50.4883,
+      "eval_samples_per_second": 40.148,
+      "eval_steps_per_second": 5.031,
+      "eval_wer": 0.3472648492507673,
+      "step": 77300
+    },
+    {
+      "epoch": 8.29,
+      "eval_loss": 0.4378765821456909,
+      "eval_runtime": 49.5816,
+      "eval_samples_per_second": 40.882,
+      "eval_steps_per_second": 5.123,
+      "eval_wer": 0.34843834627189024,
+      "step": 77400
+    },
+    {
+      "epoch": 8.3,
+      "learning_rate": 0.00013459598853868192,
+      "loss": 0.1479,
+      "step": 77500
+    },
+    {
+      "epoch": 8.3,
+      "eval_loss": 0.44405272603034973,
+      "eval_runtime": 50.0477,
+      "eval_samples_per_second": 40.501,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.3494313052897635,
+      "step": 77500
+    },
+    {
+      "epoch": 8.31,
+      "eval_loss": 0.430147647857666,
+      "eval_runtime": 49.5341,
+      "eval_samples_per_second": 40.921,
+      "eval_steps_per_second": 5.128,
+      "eval_wer": 0.3466329662393934,
+      "step": 77600
+    },
+    {
+      "epoch": 8.32,
+      "eval_loss": 0.44200408458709717,
+      "eval_runtime": 49.7936,
+      "eval_samples_per_second": 40.708,
+      "eval_steps_per_second": 5.101,
+      "eval_wer": 0.3473551182523921,
+      "step": 77700
+    },
+    {
+      "epoch": 8.33,
+      "eval_loss": 0.4519636034965515,
+      "eval_runtime": 49.663,
+      "eval_samples_per_second": 40.815,
+      "eval_steps_per_second": 5.114,
+      "eval_wer": 0.35890955046037193,
+      "step": 77800
+    },
+    {
+      "epoch": 8.34,
+      "eval_loss": 0.4283126890659332,
+      "eval_runtime": 49.7811,
+      "eval_samples_per_second": 40.718,
+      "eval_steps_per_second": 5.102,
+      "eval_wer": 0.3481675392670157,
+      "step": 77900
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 0.0001335214899713467,
+      "loss": 0.1531,
+      "step": 78000
+    },
+    {
+      "epoch": 8.35,
+      "eval_loss": 0.4324764311313629,
+      "eval_runtime": 49.9512,
+      "eval_samples_per_second": 40.58,
+      "eval_steps_per_second": 5.085,
+      "eval_wer": 0.34455677920202205,
+      "step": 78000
+    },
+    {
+      "epoch": 8.36,
+      "eval_loss": 0.4379710555076599,
+      "eval_runtime": 50.5917,
+      "eval_samples_per_second": 40.066,
+      "eval_steps_per_second": 5.021,
+      "eval_wer": 0.3469037732442679,
+      "step": 78100
+    },
+    {
+      "epoch": 8.37,
+      "eval_loss": 0.44632431864738464,
+      "eval_runtime": 49.7303,
+      "eval_samples_per_second": 40.76,
+      "eval_steps_per_second": 5.108,
+      "eval_wer": 0.3503339953060119,
+      "step": 78200
+    },
+    {
+      "epoch": 8.38,
+      "eval_loss": 0.44785556197166443,
+      "eval_runtime": 50.0294,
+      "eval_samples_per_second": 40.516,
+      "eval_steps_per_second": 5.077,
+      "eval_wer": 0.34988265029788773,
+      "step": 78300
+    },
+    {
+      "epoch": 8.39,
+      "eval_loss": 0.4477352499961853,
+      "eval_runtime": 50.9014,
+      "eval_samples_per_second": 39.822,
+      "eval_steps_per_second": 4.99,
+      "eval_wer": 0.3528615273515075,
+      "step": 78400
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 0.00013244699140401144,
+      "loss": 0.1507,
+      "step": 78500
+    },
+    {
+      "epoch": 8.4,
+      "eval_loss": 0.47092047333717346,
+      "eval_runtime": 50.1089,
+      "eval_samples_per_second": 40.452,
+      "eval_steps_per_second": 5.069,
+      "eval_wer": 0.3551182523921285,
+      "step": 78500
+    },
+    {
+      "epoch": 8.42,
+      "eval_loss": 0.45331060886383057,
+      "eval_runtime": 49.7737,
+      "eval_samples_per_second": 40.724,
+      "eval_steps_per_second": 5.103,
+      "eval_wer": 0.353132334356382,
+      "step": 78600
+    },
+    {
+      "epoch": 8.43,
+      "eval_loss": 0.450659841299057,
+      "eval_runtime": 49.9373,
+      "eval_samples_per_second": 40.591,
+      "eval_steps_per_second": 5.086,
+      "eval_wer": 0.3522296443401336,
+      "step": 78700
+    },
+    {
+      "epoch": 8.44,
+      "eval_loss": 0.45615944266319275,
+      "eval_runtime": 49.8778,
+      "eval_samples_per_second": 40.639,
+      "eval_steps_per_second": 5.092,
+      "eval_wer": 0.358277667448998,
+      "step": 78800
+    },
+    {
+      "epoch": 8.45,
+      "eval_loss": 0.442095547914505,
+      "eval_runtime": 49.5311,
+      "eval_samples_per_second": 40.924,
+      "eval_steps_per_second": 5.128,
+      "eval_wer": 0.35773605343924897,
+      "step": 78900
+    },
+    {
+      "epoch": 8.46,
+      "learning_rate": 0.00013137249283667622,
+      "loss": 0.1545,
+      "step": 79000
+    },
+    {
+      "epoch": 8.46,
+      "eval_loss": 0.4484567642211914,
+      "eval_runtime": 49.8883,
+      "eval_samples_per_second": 40.631,
+      "eval_steps_per_second": 5.091,
+      "eval_wer": 0.35466690738400436,
+      "step": 79000
+    },
+    {
+      "epoch": 8.47,
+      "eval_loss": 0.4389214813709259,
+      "eval_runtime": 49.6752,
+      "eval_samples_per_second": 40.805,
+      "eval_steps_per_second": 5.113,
+      "eval_wer": 0.34654269723776854,
+      "step": 79100
+    },
+    {
+      "epoch": 8.48,
+      "eval_loss": 0.43973833322525024,
+      "eval_runtime": 49.8321,
+      "eval_samples_per_second": 40.677,
+      "eval_steps_per_second": 5.097,
+      "eval_wer": 0.3501534573027622,
+      "step": 79200
+    },
+    {
+      "epoch": 8.49,
+      "eval_loss": 0.44029539823532104,
+      "eval_runtime": 50.1573,
+      "eval_samples_per_second": 40.413,
+      "eval_steps_per_second": 5.064,
+      "eval_wer": 0.3470843112475176,
+      "step": 79300
+    },
+    {
+      "epoch": 8.5,
+      "eval_loss": 0.4393804967403412,
+      "eval_runtime": 49.826,
+      "eval_samples_per_second": 40.682,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.3481675392670157,
+      "step": 79400
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 0.00013029799426934096,
+      "loss": 0.153,
+      "step": 79500
+    },
+    {
+      "epoch": 8.51,
+      "eval_loss": 0.43932780623435974,
+      "eval_runtime": 50.1463,
+      "eval_samples_per_second": 40.422,
+      "eval_steps_per_second": 5.065,
+      "eval_wer": 0.3473551182523921,
+      "step": 79500
+    },
+    {
+      "epoch": 8.52,
+      "eval_loss": 0.434330016374588,
+      "eval_runtime": 49.9012,
+      "eval_samples_per_second": 40.62,
+      "eval_steps_per_second": 5.09,
+      "eval_wer": 0.34952157429138836,
+      "step": 79600
+    },
+    {
+      "epoch": 8.53,
+      "eval_loss": 0.4395470917224884,
+      "eval_runtime": 49.8862,
+      "eval_samples_per_second": 40.633,
+      "eval_steps_per_second": 5.092,
+      "eval_wer": 0.35385448636938077,
+      "step": 79700
+    },
+    {
+      "epoch": 8.54,
+      "eval_loss": 0.44967034459114075,
+      "eval_runtime": 49.7096,
+      "eval_samples_per_second": 40.777,
+      "eval_steps_per_second": 5.11,
+      "eval_wer": 0.3534934103628814,
+      "step": 79800
+    },
+    {
+      "epoch": 8.55,
+      "eval_loss": 0.44434839487075806,
+      "eval_runtime": 49.5821,
+      "eval_samples_per_second": 40.882,
+      "eval_steps_per_second": 5.123,
+      "eval_wer": 0.35403502437263046,
+      "step": 79900
+    },
+    {
+      "epoch": 8.57,
+      "learning_rate": 0.00012922349570200573,
+      "loss": 0.1558,
+      "step": 80000
+    },
+    {
+      "epoch": 8.57,
+      "eval_loss": 0.4494616985321045,
+      "eval_runtime": 49.8291,
+      "eval_samples_per_second": 40.679,
+      "eval_steps_per_second": 5.097,
+      "eval_wer": 0.35538905939700305,
+      "step": 80000
+    },
+    {
+      "epoch": 8.58,
+      "eval_loss": 0.4387182295322418,
+      "eval_runtime": 50.0487,
+      "eval_samples_per_second": 40.501,
+      "eval_steps_per_second": 5.075,
+      "eval_wer": 0.3460010832280195,
+      "step": 80100
+    },
+    {
+      "epoch": 8.59,
+      "eval_loss": 0.4377879202365875,
+      "eval_runtime": 49.7994,
+      "eval_samples_per_second": 40.703,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.3520491063368839,
+      "step": 80200
+    },
+    {
+      "epoch": 8.6,
+      "eval_loss": 0.44459307193756104,
+      "eval_runtime": 50.011,
+      "eval_samples_per_second": 40.531,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.3526809893482578,
+      "step": 80300
+    },
+    {
+      "epoch": 8.61,
+      "eval_loss": 0.451284795999527,
+      "eval_runtime": 49.4538,
+      "eval_samples_per_second": 40.988,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.3507853403141361,
+      "step": 80400
+    },
+    {
+      "epoch": 8.62,
+      "learning_rate": 0.00012814899713467048,
+      "loss": 0.1527,
+      "step": 80500
+    },
+    {
+      "epoch": 8.62,
+      "eval_loss": 0.43961384892463684,
+      "eval_runtime": 49.3742,
+      "eval_samples_per_second": 41.054,
+      "eval_steps_per_second": 5.144,
+      "eval_wer": 0.3536739483661311,
+      "step": 80500
+    },
+    {
+      "epoch": 8.63,
+      "eval_loss": 0.4405277967453003,
+      "eval_runtime": 49.7186,
+      "eval_samples_per_second": 40.769,
+      "eval_steps_per_second": 5.109,
+      "eval_wer": 0.35069507131251126,
+      "step": 80600
+    },
+    {
+      "epoch": 8.64,
+      "eval_loss": 0.4397623538970947,
+      "eval_runtime": 49.7906,
+      "eval_samples_per_second": 40.71,
+      "eval_steps_per_second": 5.101,
+      "eval_wer": 0.34500812421014626,
+      "step": 80700
+    },
+    {
+      "epoch": 8.65,
+      "eval_loss": 0.44584351778030396,
+      "eval_runtime": 49.9839,
+      "eval_samples_per_second": 40.553,
+      "eval_steps_per_second": 5.082,
+      "eval_wer": 0.3507853403141361,
+      "step": 80800
+    },
+    {
+      "epoch": 8.66,
+      "eval_loss": 0.43796154856681824,
+      "eval_runtime": 50.8439,
+      "eval_samples_per_second": 39.867,
+      "eval_steps_per_second": 4.996,
+      "eval_wer": 0.3464524282361437,
+      "step": 80900
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 0.00012707449856733523,
+      "loss": 0.1522,
+      "step": 81000
+    },
+    {
+      "epoch": 8.67,
+      "eval_loss": 0.4372806251049042,
+      "eval_runtime": 49.486,
+      "eval_samples_per_second": 40.961,
+      "eval_steps_per_second": 5.133,
+      "eval_wer": 0.3481675392670157,
+      "step": 81000
+    },
+    {
+      "epoch": 8.68,
+      "eval_loss": 0.4363256096839905,
+      "eval_runtime": 49.4566,
+      "eval_samples_per_second": 40.985,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.34103628813865317,
+      "step": 81100
+    },
+    {
+      "epoch": 8.69,
+      "eval_loss": 0.4289993941783905,
+      "eval_runtime": 49.6427,
+      "eval_samples_per_second": 40.832,
+      "eval_steps_per_second": 5.117,
+      "eval_wer": 0.34473731720527173,
+      "step": 81200
+    },
+    {
+      "epoch": 8.7,
+      "eval_loss": 0.4408806264400482,
+      "eval_runtime": 50.5803,
+      "eval_samples_per_second": 40.075,
+      "eval_steps_per_second": 5.022,
+      "eval_wer": 0.35150749232713485,
+      "step": 81300
+    },
+    {
+      "epoch": 8.72,
+      "eval_loss": 0.43628886342048645,
+      "eval_runtime": 49.5854,
+      "eval_samples_per_second": 40.879,
+      "eval_steps_per_second": 5.122,
+      "eval_wer": 0.34329301317927424,
+      "step": 81400
+    },
+    {
+      "epoch": 8.73,
+      "learning_rate": 0.00012599999999999997,
+      "loss": 0.1502,
+      "step": 81500
+    },
+    {
+      "epoch": 8.73,
+      "eval_loss": 0.4313265383243561,
+      "eval_runtime": 50.648,
+      "eval_samples_per_second": 40.021,
+      "eval_steps_per_second": 5.015,
+      "eval_wer": 0.34293193717277487,
+      "step": 81500
+    },
+    {
+      "epoch": 8.74,
+      "eval_loss": 0.4263302981853485,
+      "eval_runtime": 49.6108,
+      "eval_samples_per_second": 40.858,
+      "eval_steps_per_second": 5.12,
+      "eval_wer": 0.3450983932117711,
+      "step": 81600
+    },
+    {
+      "epoch": 8.75,
+      "eval_loss": 0.42967385053634644,
+      "eval_runtime": 49.8154,
+      "eval_samples_per_second": 40.69,
+      "eval_steps_per_second": 5.099,
+      "eval_wer": 0.34518866221339595,
+      "step": 81700
+    },
+    {
+      "epoch": 8.76,
+      "eval_loss": 0.4448642432689667,
+      "eval_runtime": 49.7694,
+      "eval_samples_per_second": 40.728,
+      "eval_steps_per_second": 5.104,
+      "eval_wer": 0.341126557140278,
+      "step": 81800
+    },
+    {
+      "epoch": 8.77,
+      "eval_loss": 0.4465280771255493,
+      "eval_runtime": 49.2797,
+      "eval_samples_per_second": 41.133,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.34554973821989526,
+      "step": 81900
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 0.00012492550143266475,
+      "loss": 0.151,
+      "step": 82000
+    },
+    {
+      "epoch": 8.78,
+      "eval_loss": 0.42739376425743103,
+      "eval_runtime": 49.4978,
+      "eval_samples_per_second": 40.951,
+      "eval_steps_per_second": 5.132,
+      "eval_wer": 0.34248059216465065,
+      "step": 82000
+    },
+    {
+      "epoch": 8.79,
+      "eval_loss": 0.45250403881073,
+      "eval_runtime": 49.9239,
+      "eval_samples_per_second": 40.602,
+      "eval_steps_per_second": 5.088,
+      "eval_wer": 0.35322260335800687,
+      "step": 82100
+    },
+    {
+      "epoch": 8.8,
+      "eval_loss": 0.4282127618789673,
+      "eval_runtime": 49.2,
+      "eval_samples_per_second": 41.199,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.35024372630438705,
+      "step": 82200
+    },
+    {
+      "epoch": 8.81,
+      "eval_loss": 0.418887734413147,
+      "eval_runtime": 49.616,
+      "eval_samples_per_second": 40.854,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.35069507131251126,
+      "step": 82300
+    },
+    {
+      "epoch": 8.82,
+      "eval_loss": 0.4379144608974457,
+      "eval_runtime": 49.3537,
+      "eval_samples_per_second": 41.071,
+      "eval_steps_per_second": 5.147,
+      "eval_wer": 0.3450983932117711,
+      "step": 82400
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 0.00012385315186246417,
+      "loss": 0.1529,
+      "step": 82500
+    },
+    {
+      "epoch": 8.83,
+      "eval_loss": 0.43779709935188293,
+      "eval_runtime": 49.8201,
+      "eval_samples_per_second": 40.686,
+      "eval_steps_per_second": 5.098,
+      "eval_wer": 0.3419389781549016,
+      "step": 82500
+    },
+    {
+      "epoch": 8.84,
+      "eval_loss": 0.42828336358070374,
+      "eval_runtime": 49.4645,
+      "eval_samples_per_second": 40.979,
+      "eval_steps_per_second": 5.135,
+      "eval_wer": 0.33923090810615636,
+      "step": 82600
+    },
+    {
+      "epoch": 8.85,
+      "eval_loss": 0.4358876943588257,
+      "eval_runtime": 49.5058,
+      "eval_samples_per_second": 40.945,
+      "eval_steps_per_second": 5.131,
+      "eval_wer": 0.33986279111753026,
+      "step": 82700
+    },
+    {
+      "epoch": 8.87,
+      "eval_loss": 0.4307969808578491,
+      "eval_runtime": 49.606,
+      "eval_samples_per_second": 40.862,
+      "eval_steps_per_second": 5.12,
+      "eval_wer": 0.3358006860444123,
+      "step": 82800
+    },
+    {
+      "epoch": 8.88,
+      "eval_loss": 0.42962419986724854,
+      "eval_runtime": 50.0738,
+      "eval_samples_per_second": 40.48,
+      "eval_steps_per_second": 5.073,
+      "eval_wer": 0.33345369200216646,
+      "step": 82900
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 0.00012277865329512894,
+      "loss": 0.151,
+      "step": 83000
+    },
+    {
+      "epoch": 8.89,
+      "eval_loss": 0.4386581778526306,
+      "eval_runtime": 49.6144,
+      "eval_samples_per_second": 40.855,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.3372449900704098,
+      "step": 83000
+    },
+    {
+      "epoch": 8.9,
+      "eval_loss": 0.4334877133369446,
+      "eval_runtime": 49.5844,
+      "eval_samples_per_second": 40.88,
+      "eval_steps_per_second": 5.123,
+      "eval_wer": 0.34202924715652644,
+      "step": 83100
+    },
+    {
+      "epoch": 8.91,
+      "eval_loss": 0.4329209625720978,
+      "eval_runtime": 49.5046,
+      "eval_samples_per_second": 40.946,
+      "eval_steps_per_second": 5.131,
+      "eval_wer": 0.3374255280736595,
+      "step": 83200
+    },
+    {
+      "epoch": 8.92,
+      "eval_loss": 0.43530502915382385,
+      "eval_runtime": 49.6901,
+      "eval_samples_per_second": 40.793,
+      "eval_steps_per_second": 5.112,
+      "eval_wer": 0.3404044051272793,
+      "step": 83300
+    },
+    {
+      "epoch": 8.93,
+      "eval_loss": 0.43844395875930786,
+      "eval_runtime": 49.5686,
+      "eval_samples_per_second": 40.893,
+      "eval_steps_per_second": 5.124,
+      "eval_wer": 0.34473731720527173,
+      "step": 83400
+    },
+    {
+      "epoch": 8.94,
+      "learning_rate": 0.00012170415472779367,
+      "loss": 0.1522,
+      "step": 83500
+    },
+    {
+      "epoch": 8.94,
+      "eval_loss": 0.4444495141506195,
+      "eval_runtime": 49.7157,
+      "eval_samples_per_second": 40.772,
+      "eval_steps_per_second": 5.109,
+      "eval_wer": 0.3352590720346633,
+      "step": 83500
+    },
+    {
+      "epoch": 8.95,
+      "eval_loss": 0.4412956237792969,
+      "eval_runtime": 49.7084,
+      "eval_samples_per_second": 40.778,
+      "eval_steps_per_second": 5.11,
+      "eval_wer": 0.34807727026539087,
+      "step": 83600
+    },
+    {
+      "epoch": 8.96,
+      "eval_loss": 0.42466339468955994,
+      "eval_runtime": 49.4542,
+      "eval_samples_per_second": 40.987,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.3473551182523921,
+      "step": 83700
+    },
+    {
+      "epoch": 8.97,
+      "eval_loss": 0.41966524720191956,
+      "eval_runtime": 49.9194,
+      "eval_samples_per_second": 40.605,
+      "eval_steps_per_second": 5.088,
+      "eval_wer": 0.33859902509478246,
+      "step": 83800
+    },
+    {
+      "epoch": 8.98,
+      "eval_loss": 0.42158830165863037,
+      "eval_runtime": 49.7701,
+      "eval_samples_per_second": 40.727,
+      "eval_steps_per_second": 5.103,
+      "eval_wer": 0.33841848709153277,
+      "step": 83900
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 0.00012062965616045843,
+      "loss": 0.1511,
+      "step": 84000
+    },
+    {
+      "epoch": 8.99,
+      "eval_loss": 0.41589921712875366,
+      "eval_runtime": 50.0082,
+      "eval_samples_per_second": 40.533,
+      "eval_steps_per_second": 5.079,
+      "eval_wer": 0.33959198411265573,
+      "step": 84000
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": 0.4212520122528076,
+      "eval_runtime": 49.9688,
+      "eval_samples_per_second": 40.565,
+      "eval_steps_per_second": 5.083,
+      "eval_wer": 0.3415779021484022,
+      "step": 84100
+    },
+    {
+      "epoch": 9.01,
+      "eval_loss": 0.43991902470588684,
+      "eval_runtime": 50.2667,
+      "eval_samples_per_second": 40.325,
+      "eval_steps_per_second": 5.053,
+      "eval_wer": 0.3378768730817837,
+      "step": 84200
+    },
+    {
+      "epoch": 9.03,
+      "eval_loss": 0.43180111050605774,
+      "eval_runtime": 50.1864,
+      "eval_samples_per_second": 40.389,
+      "eval_steps_per_second": 5.061,
+      "eval_wer": 0.3436540891857736,
+      "step": 84300
+    },
+    {
+      "epoch": 9.04,
+      "eval_loss": 0.43560150265693665,
+      "eval_runtime": 50.6404,
+      "eval_samples_per_second": 40.027,
+      "eval_steps_per_second": 5.016,
+      "eval_wer": 0.3370644520671601,
+      "step": 84400
+    },
+    {
+      "epoch": 9.05,
+      "learning_rate": 0.00011955515759312319,
+      "loss": 0.1336,
+      "step": 84500
+    },
+    {
+      "epoch": 9.05,
+      "eval_loss": 0.44032755494117737,
+      "eval_runtime": 49.9078,
+      "eval_samples_per_second": 40.615,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.33733525907203465,
+      "step": 84500
+    },
+    {
+      "epoch": 9.06,
+      "eval_loss": 0.4544905722141266,
+      "eval_runtime": 48.9473,
+      "eval_samples_per_second": 41.412,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.33814768008665824,
+      "step": 84600
+    },
+    {
+      "epoch": 9.07,
+      "eval_loss": 0.4313288629055023,
+      "eval_runtime": 49.8429,
+      "eval_samples_per_second": 40.668,
+      "eval_steps_per_second": 5.096,
+      "eval_wer": 0.3330926159956671,
+      "step": 84700
+    },
+    {
+      "epoch": 9.08,
+      "eval_loss": 0.425656259059906,
+      "eval_runtime": 49.1736,
+      "eval_samples_per_second": 41.221,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.335981224047662,
+      "step": 84800
+    },
+    {
+      "epoch": 9.09,
+      "eval_loss": 0.4285389482975006,
+      "eval_runtime": 49.1078,
+      "eval_samples_per_second": 41.277,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.33715472106878497,
+      "step": 84900
+    },
+    {
+      "epoch": 9.1,
+      "learning_rate": 0.00011848065902578795,
+      "loss": 0.1315,
+      "step": 85000
+    },
+    {
+      "epoch": 9.1,
+      "eval_loss": 0.43780598044395447,
+      "eval_runtime": 49.1411,
+      "eval_samples_per_second": 41.249,
+      "eval_steps_per_second": 5.169,
+      "eval_wer": 0.33318288499729193,
+      "step": 85000
+    },
+    {
+      "epoch": 9.11,
+      "eval_loss": 0.4351807236671448,
+      "eval_runtime": 48.8456,
+      "eval_samples_per_second": 41.498,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.3282180899079256,
+      "step": 85100
+    },
+    {
+      "epoch": 9.12,
+      "eval_loss": 0.4359615743160248,
+      "eval_runtime": 50.6269,
+      "eval_samples_per_second": 40.038,
+      "eval_steps_per_second": 5.017,
+      "eval_wer": 0.3339050370102907,
+      "step": 85200
+    },
+    {
+      "epoch": 9.13,
+      "eval_loss": 0.4403829574584961,
+      "eval_runtime": 48.958,
+      "eval_samples_per_second": 41.403,
+      "eval_steps_per_second": 5.188,
+      "eval_wer": 0.33652283805741107,
+      "step": 85300
+    },
+    {
+      "epoch": 9.14,
+      "eval_loss": 0.4345133602619171,
+      "eval_runtime": 49.1877,
+      "eval_samples_per_second": 41.21,
+      "eval_steps_per_second": 5.164,
+      "eval_wer": 0.3356201480411627,
+      "step": 85400
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 0.00011740616045845271,
+      "loss": 0.1272,
+      "step": 85500
+    },
+    {
+      "epoch": 9.15,
+      "eval_loss": 0.4467580020427704,
+      "eval_runtime": 48.8216,
+      "eval_samples_per_second": 41.519,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.33751579707528434,
+      "step": 85500
+    },
+    {
+      "epoch": 9.16,
+      "eval_loss": 0.43308141827583313,
+      "eval_runtime": 49.2077,
+      "eval_samples_per_second": 41.193,
+      "eval_steps_per_second": 5.162,
+      "eval_wer": 0.3363423000541614,
+      "step": 85600
+    },
+    {
+      "epoch": 9.18,
+      "eval_loss": 0.4329800605773926,
+      "eval_runtime": 50.5519,
+      "eval_samples_per_second": 40.097,
+      "eval_steps_per_second": 5.025,
+      "eval_wer": 0.3309261599566709,
+      "step": 85700
+    },
+    {
+      "epoch": 9.19,
+      "eval_loss": 0.4423808753490448,
+      "eval_runtime": 48.8833,
+      "eval_samples_per_second": 41.466,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.3301137389420473,
+      "step": 85800
+    },
+    {
+      "epoch": 9.2,
+      "eval_loss": 0.4520210027694702,
+      "eval_runtime": 48.6979,
+      "eval_samples_per_second": 41.624,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.33264127098754287,
+      "step": 85900
+    },
+    {
+      "epoch": 9.21,
+      "learning_rate": 0.00011633166189111747,
+      "loss": 0.1289,
+      "step": 86000
+    },
+    {
+      "epoch": 9.21,
+      "eval_loss": 0.44208621978759766,
+      "eval_runtime": 49.4655,
+      "eval_samples_per_second": 40.978,
+      "eval_steps_per_second": 5.135,
+      "eval_wer": 0.332551001985918,
+      "step": 86000
+    },
+    {
+      "epoch": 9.22,
+      "eval_loss": 0.44799816608428955,
+      "eval_runtime": 49.0094,
+      "eval_samples_per_second": 41.359,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.33345369200216646,
+      "step": 86100
+    },
+    {
+      "epoch": 9.23,
+      "eval_loss": 0.4351326525211334,
+      "eval_runtime": 49.1297,
+      "eval_samples_per_second": 41.258,
+      "eval_steps_per_second": 5.17,
+      "eval_wer": 0.33796714208340856,
+      "step": 86200
+    },
+    {
+      "epoch": 9.24,
+      "eval_loss": 0.43502259254455566,
+      "eval_runtime": 49.1339,
+      "eval_samples_per_second": 41.255,
+      "eval_steps_per_second": 5.17,
+      "eval_wer": 0.34266113016790034,
+      "step": 86300
+    },
+    {
+      "epoch": 9.25,
+      "eval_loss": 0.43621236085891724,
+      "eval_runtime": 49.6078,
+      "eval_samples_per_second": 40.86,
+      "eval_steps_per_second": 5.12,
+      "eval_wer": 0.33200938797616897,
+      "step": 86400
+    },
+    {
+      "epoch": 9.26,
+      "learning_rate": 0.00011525716332378223,
+      "loss": 0.1333,
+      "step": 86500
+    },
+    {
+      "epoch": 9.26,
+      "eval_loss": 0.42599332332611084,
+      "eval_runtime": 49.1903,
+      "eval_samples_per_second": 41.207,
+      "eval_steps_per_second": 5.164,
+      "eval_wer": 0.3341758440151652,
+      "step": 86500
+    },
+    {
+      "epoch": 9.27,
+      "eval_loss": 0.4356515407562256,
+      "eval_runtime": 48.8406,
+      "eval_samples_per_second": 41.502,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.335981224047662,
+      "step": 86600
+    },
+    {
+      "epoch": 9.28,
+      "eval_loss": 0.45051419734954834,
+      "eval_runtime": 49.1958,
+      "eval_samples_per_second": 41.203,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3372449900704098,
+      "step": 86700
+    },
+    {
+      "epoch": 9.29,
+      "eval_loss": 0.4341755509376526,
+      "eval_runtime": 49.4875,
+      "eval_samples_per_second": 40.96,
+      "eval_steps_per_second": 5.133,
+      "eval_wer": 0.33589095504603717,
+      "step": 86800
+    },
+    {
+      "epoch": 9.3,
+      "eval_loss": 0.4295257329940796,
+      "eval_runtime": 49.2323,
+      "eval_samples_per_second": 41.172,
+      "eval_steps_per_second": 5.159,
+      "eval_wer": 0.33670337606066075,
+      "step": 86900
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 0.00011418266475644698,
+      "loss": 0.1318,
+      "step": 87000
+    },
+    {
+      "epoch": 9.31,
+      "eval_loss": 0.43195515871047974,
+      "eval_runtime": 50.3285,
+      "eval_samples_per_second": 40.275,
+      "eval_steps_per_second": 5.047,
+      "eval_wer": 0.3335439610037913,
+      "step": 87000
+    },
+    {
+      "epoch": 9.33,
+      "eval_loss": 0.4331757128238678,
+      "eval_runtime": 49.2032,
+      "eval_samples_per_second": 41.197,
+      "eval_steps_per_second": 5.162,
+      "eval_wer": 0.33444665102003973,
+      "step": 87100
+    },
+    {
+      "epoch": 9.34,
+      "eval_loss": 0.43729162216186523,
+      "eval_runtime": 49.2547,
+      "eval_samples_per_second": 41.153,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.33300234699404224,
+      "step": 87200
+    },
+    {
+      "epoch": 9.35,
+      "eval_loss": 0.44904646277427673,
+      "eval_runtime": 49.1768,
+      "eval_samples_per_second": 41.219,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.33155804296804475,
+      "step": 87300
+    },
+    {
+      "epoch": 9.36,
+      "eval_loss": 0.41877713799476624,
+      "eval_runtime": 49.0605,
+      "eval_samples_per_second": 41.316,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.34293193717277487,
+      "step": 87400
+    },
+    {
+      "epoch": 9.37,
+      "learning_rate": 0.00011311031518624641,
+      "loss": 0.1275,
+      "step": 87500
+    },
+    {
+      "epoch": 9.37,
+      "eval_loss": 0.45020759105682373,
+      "eval_runtime": 48.7704,
+      "eval_samples_per_second": 41.562,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.33832821808990793,
+      "step": 87500
+    },
+    {
+      "epoch": 9.38,
+      "eval_loss": 0.44634371995925903,
+      "eval_runtime": 49.2783,
+      "eval_samples_per_second": 41.134,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.3386892940964073,
+      "step": 87600
+    },
+    {
+      "epoch": 9.39,
+      "eval_loss": 0.4385134279727936,
+      "eval_runtime": 49.1376,
+      "eval_samples_per_second": 41.251,
+      "eval_steps_per_second": 5.169,
+      "eval_wer": 0.33083589095504606,
+      "step": 87700
+    },
+    {
+      "epoch": 9.4,
+      "eval_loss": 0.446433424949646,
+      "eval_runtime": 48.9791,
+      "eval_samples_per_second": 41.385,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.34139736414515254,
+      "step": 87800
+    },
+    {
+      "epoch": 9.41,
+      "eval_loss": 0.4562661945819855,
+      "eval_runtime": 49.0763,
+      "eval_samples_per_second": 41.303,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.3404946741289041,
+      "step": 87900
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 0.00011203581661891117,
+      "loss": 0.1331,
+      "step": 88000
+    },
+    {
+      "epoch": 9.42,
+      "eval_loss": 0.4285636842250824,
+      "eval_runtime": 49.0786,
+      "eval_samples_per_second": 41.301,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3374255280736595,
+      "step": 88000
+    },
+    {
+      "epoch": 9.43,
+      "eval_loss": 0.4389163553714752,
+      "eval_runtime": 49.1661,
+      "eval_samples_per_second": 41.228,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.3351688030330385,
+      "step": 88100
+    },
+    {
+      "epoch": 9.44,
+      "eval_loss": 0.43011942505836487,
+      "eval_runtime": 49.0545,
+      "eval_samples_per_second": 41.321,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.3339953060119155,
+      "step": 88200
+    },
+    {
+      "epoch": 9.45,
+      "eval_loss": 0.4416744112968445,
+      "eval_runtime": 49.0788,
+      "eval_samples_per_second": 41.301,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.33733525907203465,
+      "step": 88300
+    },
+    {
+      "epoch": 9.46,
+      "eval_loss": 0.44503673911094666,
+      "eval_runtime": 49.266,
+      "eval_samples_per_second": 41.144,
+      "eval_steps_per_second": 5.156,
+      "eval_wer": 0.34248059216465065,
+      "step": 88400
+    },
+    {
+      "epoch": 9.48,
+      "learning_rate": 0.00011096346704871058,
+      "loss": 0.1266,
+      "step": 88500
+    },
+    {
+      "epoch": 9.48,
+      "eval_loss": 0.44555312395095825,
+      "eval_runtime": 49.2181,
+      "eval_samples_per_second": 41.184,
+      "eval_steps_per_second": 5.161,
+      "eval_wer": 0.3450983932117711,
+      "step": 88500
+    },
+    {
+      "epoch": 9.49,
+      "eval_loss": 0.45165377855300903,
+      "eval_runtime": 49.095,
+      "eval_samples_per_second": 41.287,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.3403141361256545,
+      "step": 88600
+    },
+    {
+      "epoch": 9.5,
+      "eval_loss": 0.44465458393096924,
+      "eval_runtime": 48.9889,
+      "eval_samples_per_second": 41.377,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.3419389781549016,
+      "step": 88700
+    },
+    {
+      "epoch": 9.51,
+      "eval_loss": 0.4486336410045624,
+      "eval_runtime": 49.0554,
+      "eval_samples_per_second": 41.321,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.3427513991695252,
+      "step": 88800
+    },
+    {
+      "epoch": 9.52,
+      "eval_loss": 0.4590870141983032,
+      "eval_runtime": 48.9673,
+      "eval_samples_per_second": 41.395,
+      "eval_steps_per_second": 5.187,
+      "eval_wer": 0.341126557140278,
+      "step": 88900
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 0.00010988896848137534,
+      "loss": 0.1316,
+      "step": 89000
+    },
+    {
+      "epoch": 9.53,
+      "eval_loss": 0.44813600182533264,
+      "eval_runtime": 49.5192,
+      "eval_samples_per_second": 40.934,
+      "eval_steps_per_second": 5.129,
+      "eval_wer": 0.3386892940964073,
+      "step": 89000
+    },
+    {
+      "epoch": 9.54,
+      "eval_loss": 0.43075862526893616,
+      "eval_runtime": 49.2146,
+      "eval_samples_per_second": 41.187,
+      "eval_steps_per_second": 5.161,
+      "eval_wer": 0.33489799602816395,
+      "step": 89100
+    },
+    {
+      "epoch": 9.55,
+      "eval_loss": 0.44113072752952576,
+      "eval_runtime": 49.2342,
+      "eval_samples_per_second": 41.171,
+      "eval_steps_per_second": 5.159,
+      "eval_wer": 0.3404946741289041,
+      "step": 89200
+    },
+    {
+      "epoch": 9.56,
+      "eval_loss": 0.43779146671295166,
+      "eval_runtime": 48.7095,
+      "eval_samples_per_second": 41.614,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.33896010110128183,
+      "step": 89300
+    },
+    {
+      "epoch": 9.57,
+      "eval_loss": 0.44476673007011414,
+      "eval_runtime": 49.0767,
+      "eval_samples_per_second": 41.303,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.33652283805741107,
+      "step": 89400
+    },
+    {
+      "epoch": 9.58,
+      "learning_rate": 0.0001088144699140401,
+      "loss": 0.1325,
+      "step": 89500
+    },
+    {
+      "epoch": 9.58,
+      "eval_loss": 0.45753398537635803,
+      "eval_runtime": 50.1245,
+      "eval_samples_per_second": 40.439,
+      "eval_steps_per_second": 5.067,
+      "eval_wer": 0.3415779021484022,
+      "step": 89500
+    },
+    {
+      "epoch": 9.59,
+      "eval_loss": 0.46084266901016235,
+      "eval_runtime": 48.8264,
+      "eval_samples_per_second": 41.514,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.3422097851597761,
+      "step": 89600
+    },
+    {
+      "epoch": 9.6,
+      "eval_loss": 0.4396139681339264,
+      "eval_runtime": 48.7221,
+      "eval_samples_per_second": 41.603,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.3349882650297888,
+      "step": 89700
+    },
+    {
+      "epoch": 9.61,
+      "eval_loss": 0.43795400857925415,
+      "eval_runtime": 49.7416,
+      "eval_samples_per_second": 40.751,
+      "eval_steps_per_second": 5.106,
+      "eval_wer": 0.3397725221159054,
+      "step": 89800
+    },
+    {
+      "epoch": 9.63,
+      "eval_loss": 0.43372565507888794,
+      "eval_runtime": 49.3047,
+      "eval_samples_per_second": 41.112,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.33877956309803214,
+      "step": 89900
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 0.00010773997134670486,
+      "loss": 0.1324,
+      "step": 90000
+    },
+    {
+      "epoch": 9.64,
+      "eval_loss": 0.4376404285430908,
+      "eval_runtime": 49.2265,
+      "eval_samples_per_second": 41.177,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.33877956309803214,
+      "step": 90000
+    },
+    {
+      "epoch": 9.65,
+      "eval_loss": 0.418549120426178,
+      "eval_runtime": 49.0407,
+      "eval_samples_per_second": 41.333,
+      "eval_steps_per_second": 5.179,
+      "eval_wer": 0.33796714208340856,
+      "step": 90100
+    },
+    {
+      "epoch": 9.66,
+      "eval_loss": 0.43939289450645447,
+      "eval_runtime": 49.1046,
+      "eval_samples_per_second": 41.279,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.33841848709153277,
+      "step": 90200
+    },
+    {
+      "epoch": 9.67,
+      "eval_loss": 0.4472477436065674,
+      "eval_runtime": 48.6772,
+      "eval_samples_per_second": 41.642,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.34004332912077995,
+      "step": 90300
+    },
+    {
+      "epoch": 9.68,
+      "eval_loss": 0.45231983065605164,
+      "eval_runtime": 49.342,
+      "eval_samples_per_second": 41.081,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.33896010110128183,
+      "step": 90400
+    },
+    {
+      "epoch": 9.69,
+      "learning_rate": 0.00010666762177650428,
+      "loss": 0.1361,
+      "step": 90500
+    },
+    {
+      "epoch": 9.69,
+      "eval_loss": 0.4465869069099426,
+      "eval_runtime": 49.0952,
+      "eval_samples_per_second": 41.287,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.338869832099657,
+      "step": 90500
+    },
+    {
+      "epoch": 9.7,
+      "eval_loss": 0.44142165780067444,
+      "eval_runtime": 48.9328,
+      "eval_samples_per_second": 41.424,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.33832821808990793,
+      "step": 90600
+    },
+    {
+      "epoch": 9.71,
+      "eval_loss": 0.42880019545555115,
+      "eval_runtime": 49.1913,
+      "eval_samples_per_second": 41.206,
+      "eval_steps_per_second": 5.164,
+      "eval_wer": 0.3348077270265391,
+      "step": 90700
+    },
+    {
+      "epoch": 9.72,
+      "eval_loss": 0.4444648027420044,
+      "eval_runtime": 49.0743,
+      "eval_samples_per_second": 41.305,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.3374255280736595,
+      "step": 90800
+    },
+    {
+      "epoch": 9.73,
+      "eval_loss": 0.4251817762851715,
+      "eval_runtime": 49.2092,
+      "eval_samples_per_second": 41.191,
+      "eval_steps_per_second": 5.162,
+      "eval_wer": 0.33218992597941865,
+      "step": 90900
+    },
+    {
+      "epoch": 9.74,
+      "learning_rate": 0.00010559312320916904,
+      "loss": 0.1353,
+      "step": 91000
+    },
+    {
+      "epoch": 9.74,
+      "eval_loss": 0.4311528205871582,
+      "eval_runtime": 49.3374,
+      "eval_samples_per_second": 41.084,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.33381476800866583,
+      "step": 91000
+    },
+    {
+      "epoch": 9.75,
+      "eval_loss": 0.4325965344905853,
+      "eval_runtime": 49.4186,
+      "eval_samples_per_second": 41.017,
+      "eval_steps_per_second": 5.14,
+      "eval_wer": 0.3319191189745441,
+      "step": 91100
+    },
+    {
+      "epoch": 9.76,
+      "eval_loss": 0.42121651768684387,
+      "eval_runtime": 50.098,
+      "eval_samples_per_second": 40.461,
+      "eval_steps_per_second": 5.07,
+      "eval_wer": 0.3399530601191551,
+      "step": 91200
+    },
+    {
+      "epoch": 9.78,
+      "eval_loss": 0.4190988838672638,
+      "eval_runtime": 48.9423,
+      "eval_samples_per_second": 41.416,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.3374255280736595,
+      "step": 91300
+    },
+    {
+      "epoch": 9.79,
+      "eval_loss": 0.4398752450942993,
+      "eval_runtime": 49.0663,
+      "eval_samples_per_second": 41.311,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.33318288499729193,
+      "step": 91400
+    },
+    {
+      "epoch": 9.8,
+      "learning_rate": 0.0001045186246418338,
+      "loss": 0.1308,
+      "step": 91500
+    },
+    {
+      "epoch": 9.8,
+      "eval_loss": 0.43398961424827576,
+      "eval_runtime": 49.3702,
+      "eval_samples_per_second": 41.057,
+      "eval_steps_per_second": 5.145,
+      "eval_wer": 0.33489799602816395,
+      "step": 91500
+    },
+    {
+      "epoch": 9.81,
+      "eval_loss": 0.4279971718788147,
+      "eval_runtime": 49.3652,
+      "eval_samples_per_second": 41.061,
+      "eval_steps_per_second": 5.145,
+      "eval_wer": 0.3378768730817837,
+      "step": 91600
+    },
+    {
+      "epoch": 9.82,
+      "eval_loss": 0.44193169474601746,
+      "eval_runtime": 49.0009,
+      "eval_samples_per_second": 41.367,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.3376060660769092,
+      "step": 91700
+    },
+    {
+      "epoch": 9.83,
+      "eval_loss": 0.43088042736053467,
+      "eval_runtime": 49.3626,
+      "eval_samples_per_second": 41.063,
+      "eval_steps_per_second": 5.146,
+      "eval_wer": 0.33327315399891677,
+      "step": 91800
+    },
+    {
+      "epoch": 9.84,
+      "eval_loss": 0.42743775248527527,
+      "eval_runtime": 49.588,
+      "eval_samples_per_second": 40.877,
+      "eval_steps_per_second": 5.122,
+      "eval_wer": 0.3351688030330385,
+      "step": 91900
+    },
+    {
+      "epoch": 9.85,
+      "learning_rate": 0.00010344627507163324,
+      "loss": 0.1321,
+      "step": 92000
+    },
+    {
+      "epoch": 9.85,
+      "eval_loss": 0.41469934582710266,
+      "eval_runtime": 49.1545,
+      "eval_samples_per_second": 41.237,
+      "eval_steps_per_second": 5.167,
+      "eval_wer": 0.333724499007041,
+      "step": 92000
+    },
+    {
+      "epoch": 9.86,
+      "eval_loss": 0.4252135455608368,
+      "eval_runtime": 49.2985,
+      "eval_samples_per_second": 41.117,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.3316483119696696,
+      "step": 92100
+    },
+    {
+      "epoch": 9.87,
+      "eval_loss": 0.43777018785476685,
+      "eval_runtime": 49.1225,
+      "eval_samples_per_second": 41.264,
+      "eval_steps_per_second": 5.171,
+      "eval_wer": 0.33814768008665824,
+      "step": 92200
+    },
+    {
+      "epoch": 9.88,
+      "eval_loss": 0.4264971911907196,
+      "eval_runtime": 49.522,
+      "eval_samples_per_second": 40.931,
+      "eval_steps_per_second": 5.129,
+      "eval_wer": 0.33552987903953785,
+      "step": 92300
+    },
+    {
+      "epoch": 9.89,
+      "eval_loss": 0.4247148036956787,
+      "eval_runtime": 49.2205,
+      "eval_samples_per_second": 41.182,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.3330926159956671,
+      "step": 92400
+    },
+    {
+      "epoch": 9.9,
+      "learning_rate": 0.000102371776504298,
+      "loss": 0.1358,
+      "step": 92500
+    },
+    {
+      "epoch": 9.9,
+      "eval_loss": 0.4099043309688568,
+      "eval_runtime": 49.0334,
+      "eval_samples_per_second": 41.339,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.3378768730817837,
+      "step": 92500
+    },
+    {
+      "epoch": 9.91,
+      "eval_loss": 0.414156436920166,
+      "eval_runtime": 49.0837,
+      "eval_samples_per_second": 41.297,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3356201480411627,
+      "step": 92600
+    },
+    {
+      "epoch": 9.93,
+      "eval_loss": 0.4220016598701477,
+      "eval_runtime": 49.194,
+      "eval_samples_per_second": 41.204,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.33318288499729193,
+      "step": 92700
+    },
+    {
+      "epoch": 9.94,
+      "eval_loss": 0.42191407084465027,
+      "eval_runtime": 50.2805,
+      "eval_samples_per_second": 40.314,
+      "eval_steps_per_second": 5.052,
+      "eval_wer": 0.33688391406391044,
+      "step": 92800
+    },
+    {
+      "epoch": 9.95,
+      "eval_loss": 0.4178451597690582,
+      "eval_runtime": 49.1736,
+      "eval_samples_per_second": 41.221,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.33318288499729193,
+      "step": 92900
+    },
+    {
+      "epoch": 9.96,
+      "learning_rate": 0.00010129727793696274,
+      "loss": 0.1331,
+      "step": 93000
+    },
+    {
+      "epoch": 9.96,
+      "eval_loss": 0.43051832914352417,
+      "eval_runtime": 49.1956,
+      "eval_samples_per_second": 41.203,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3352590720346633,
+      "step": 93000
+    },
+    {
+      "epoch": 9.97,
+      "eval_loss": 0.43244585394859314,
+      "eval_runtime": 49.2837,
+      "eval_samples_per_second": 41.129,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.3306553529517964,
+      "step": 93100
+    },
+    {
+      "epoch": 9.98,
+      "eval_loss": 0.4315347671508789,
+      "eval_runtime": 49.2581,
+      "eval_samples_per_second": 41.151,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.33444665102003973,
+      "step": 93200
+    },
+    {
+      "epoch": 9.99,
+      "eval_loss": 0.4212498664855957,
+      "eval_runtime": 49.1731,
+      "eval_samples_per_second": 41.222,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.33137750496479507,
+      "step": 93300
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 0.42032095789909363,
+      "eval_runtime": 49.4382,
+      "eval_samples_per_second": 41.001,
+      "eval_steps_per_second": 5.138,
+      "eval_wer": 0.33318288499729193,
+      "step": 93400
+    },
+    {
+      "epoch": 10.01,
+      "learning_rate": 0.00010022277936962749,
+      "loss": 0.1304,
+      "step": 93500
+    },
+    {
+      "epoch": 10.01,
+      "eval_loss": 0.4423731863498688,
+      "eval_runtime": 50.0952,
+      "eval_samples_per_second": 40.463,
+      "eval_steps_per_second": 5.07,
+      "eval_wer": 0.33507853403141363,
+      "step": 93500
+    },
+    {
+      "epoch": 10.02,
+      "eval_loss": 0.44737526774406433,
+      "eval_runtime": 49.7473,
+      "eval_samples_per_second": 40.746,
+      "eval_steps_per_second": 5.106,
+      "eval_wer": 0.33408557501354036,
+      "step": 93600
+    },
+    {
+      "epoch": 10.03,
+      "eval_loss": 0.4465775191783905,
+      "eval_runtime": 49.5611,
+      "eval_samples_per_second": 40.899,
+      "eval_steps_per_second": 5.125,
+      "eval_wer": 0.33778660408015887,
+      "step": 93700
+    },
+    {
+      "epoch": 10.04,
+      "eval_loss": 0.43884560465812683,
+      "eval_runtime": 49.0695,
+      "eval_samples_per_second": 41.309,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.3327315399891677,
+      "step": 93800
+    },
+    {
+      "epoch": 10.05,
+      "eval_loss": 0.431158185005188,
+      "eval_runtime": 49.6041,
+      "eval_samples_per_second": 40.864,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.335981224047662,
+      "step": 93900
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 9.914828080229225e-05,
+      "loss": 0.1152,
+      "step": 94000
+    },
+    {
+      "epoch": 10.06,
+      "eval_loss": 0.44709327816963196,
+      "eval_runtime": 49.482,
+      "eval_samples_per_second": 40.964,
+      "eval_steps_per_second": 5.133,
+      "eval_wer": 0.3306553529517964,
+      "step": 94000
+    },
+    {
+      "epoch": 10.07,
+      "eval_loss": 0.4472164809703827,
+      "eval_runtime": 49.2263,
+      "eval_samples_per_second": 41.177,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.33155804296804475,
+      "step": 94100
+    },
+    {
+      "epoch": 10.09,
+      "eval_loss": 0.4462222158908844,
+      "eval_runtime": 50.6279,
+      "eval_samples_per_second": 40.037,
+      "eval_steps_per_second": 5.017,
+      "eval_wer": 0.33237046398266834,
+      "step": 94200
+    },
+    {
+      "epoch": 10.1,
+      "eval_loss": 0.4382689297199249,
+      "eval_runtime": 49.1505,
+      "eval_samples_per_second": 41.241,
+      "eval_steps_per_second": 5.168,
+      "eval_wer": 0.3343563820184149,
+      "step": 94300
+    },
+    {
+      "epoch": 10.11,
+      "eval_loss": 0.46711546182632446,
+      "eval_runtime": 49.598,
+      "eval_samples_per_second": 40.869,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.33652283805741107,
+      "step": 94400
+    },
+    {
+      "epoch": 10.12,
+      "learning_rate": 9.8073782234957e-05,
+      "loss": 0.1097,
+      "step": 94500
+    },
+    {
+      "epoch": 10.12,
+      "eval_loss": 0.4596339464187622,
+      "eval_runtime": 49.2966,
+      "eval_samples_per_second": 41.118,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.3306553529517964,
+      "step": 94500
+    },
+    {
+      "epoch": 10.13,
+      "eval_loss": 0.4517350196838379,
+      "eval_runtime": 49.4575,
+      "eval_samples_per_second": 40.985,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.3382379490882831,
+      "step": 94600
+    },
+    {
+      "epoch": 10.14,
+      "eval_loss": 0.4285356402397156,
+      "eval_runtime": 49.5466,
+      "eval_samples_per_second": 40.911,
+      "eval_steps_per_second": 5.126,
+      "eval_wer": 0.33796714208340856,
+      "step": 94700
+    },
+    {
+      "epoch": 10.15,
+      "eval_loss": 0.46282461285591125,
+      "eval_runtime": 49.2102,
+      "eval_samples_per_second": 41.191,
+      "eval_steps_per_second": 5.162,
+      "eval_wer": 0.3363423000541614,
+      "step": 94800
+    },
+    {
+      "epoch": 10.16,
+      "eval_loss": 0.4478081464767456,
+      "eval_runtime": 49.3065,
+      "eval_samples_per_second": 41.11,
+      "eval_steps_per_second": 5.151,
+      "eval_wer": 0.33652283805741107,
+      "step": 94900
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 9.699928366762177e-05,
+      "loss": 0.1153,
+      "step": 95000
+    },
+    {
+      "epoch": 10.17,
+      "eval_loss": 0.44635722041130066,
+      "eval_runtime": 49.4158,
+      "eval_samples_per_second": 41.019,
+      "eval_steps_per_second": 5.14,
+      "eval_wer": 0.3346271890232894,
+      "step": 95000
+    },
+    {
+      "epoch": 10.18,
+      "eval_loss": 0.44317421317100525,
+      "eval_runtime": 49.2548,
+      "eval_samples_per_second": 41.153,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.33923090810615636,
+      "step": 95100
+    },
+    {
+      "epoch": 10.19,
+      "eval_loss": 0.4326357841491699,
+      "eval_runtime": 49.4047,
+      "eval_samples_per_second": 41.029,
+      "eval_steps_per_second": 5.141,
+      "eval_wer": 0.33300234699404224,
+      "step": 95200
+    },
+    {
+      "epoch": 10.2,
+      "eval_loss": 0.4479786455631256,
+      "eval_runtime": 49.1321,
+      "eval_samples_per_second": 41.256,
+      "eval_steps_per_second": 5.17,
+      "eval_wer": 0.3327315399891677,
+      "step": 95300
+    },
+    {
+      "epoch": 10.21,
+      "eval_loss": 0.44355833530426025,
+      "eval_runtime": 50.8799,
+      "eval_samples_per_second": 39.839,
+      "eval_steps_per_second": 4.992,
+      "eval_wer": 0.3259613648673046,
+      "step": 95400
+    },
+    {
+      "epoch": 10.22,
+      "learning_rate": 9.592693409742119e-05,
+      "loss": 0.1149,
+      "step": 95500
+    },
+    {
+      "epoch": 10.22,
+      "eval_loss": 0.4549073874950409,
+      "eval_runtime": 49.3035,
+      "eval_samples_per_second": 41.113,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.33110669795992054,
+      "step": 95500
+    },
+    {
+      "epoch": 10.24,
+      "eval_loss": 0.4572867751121521,
+      "eval_runtime": 49.5772,
+      "eval_samples_per_second": 40.886,
+      "eval_steps_per_second": 5.123,
+      "eval_wer": 0.33534934103628816,
+      "step": 95600
+    },
+    {
+      "epoch": 10.25,
+      "eval_loss": 0.4373367428779602,
+      "eval_runtime": 49.5738,
+      "eval_samples_per_second": 40.889,
+      "eval_steps_per_second": 5.124,
+      "eval_wer": 0.33688391406391044,
+      "step": 95700
+    },
+    {
+      "epoch": 10.26,
+      "eval_loss": 0.4459189176559448,
+      "eval_runtime": 49.4912,
+      "eval_samples_per_second": 40.957,
+      "eval_steps_per_second": 5.132,
+      "eval_wer": 0.3358006860444123,
+      "step": 95800
+    },
+    {
+      "epoch": 10.27,
+      "eval_loss": 0.42883867025375366,
+      "eval_runtime": 49.3495,
+      "eval_samples_per_second": 41.074,
+      "eval_steps_per_second": 5.147,
+      "eval_wer": 0.32704459288680265,
+      "step": 95900
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 9.485243553008595e-05,
+      "loss": 0.1169,
+      "step": 96000
+    },
+    {
+      "epoch": 10.28,
+      "eval_loss": 0.44739025831222534,
+      "eval_runtime": 49.7992,
+      "eval_samples_per_second": 40.703,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.33300234699404224,
+      "step": 96000
+    },
+    {
+      "epoch": 10.29,
+      "eval_loss": 0.4523959755897522,
+      "eval_runtime": 49.381,
+      "eval_samples_per_second": 41.048,
+      "eval_steps_per_second": 5.144,
+      "eval_wer": 0.3298429319371728,
+      "step": 96100
+    },
+    {
+      "epoch": 10.3,
+      "eval_loss": 0.4517040550708771,
+      "eval_runtime": 49.4501,
+      "eval_samples_per_second": 40.991,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.3257808268640549,
+      "step": 96200
+    },
+    {
+      "epoch": 10.31,
+      "eval_loss": 0.43664810061454773,
+      "eval_runtime": 49.3021,
+      "eval_samples_per_second": 41.114,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.3288499729192995,
+      "step": 96300
+    },
+    {
+      "epoch": 10.32,
+      "eval_loss": 0.4574427008628845,
+      "eval_runtime": 49.5018,
+      "eval_samples_per_second": 40.948,
+      "eval_steps_per_second": 5.131,
+      "eval_wer": 0.33237046398266834,
+      "step": 96400
+    },
+    {
+      "epoch": 10.33,
+      "learning_rate": 9.37779369627507e-05,
+      "loss": 0.1137,
+      "step": 96500
+    },
+    {
+      "epoch": 10.33,
+      "eval_loss": 0.4506693482398987,
+      "eval_runtime": 49.6177,
+      "eval_samples_per_second": 40.852,
+      "eval_steps_per_second": 5.119,
+      "eval_wer": 0.33426611301679005,
+      "step": 96500
+    },
+    {
+      "epoch": 10.34,
+      "eval_loss": 0.44143351912498474,
+      "eval_runtime": 49.2006,
+      "eval_samples_per_second": 41.199,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3301137389420473,
+      "step": 96600
+    },
+    {
+      "epoch": 10.35,
+      "eval_loss": 0.45244306325912476,
+      "eval_runtime": 49.2654,
+      "eval_samples_per_second": 41.144,
+      "eval_steps_per_second": 5.156,
+      "eval_wer": 0.3366131070590359,
+      "step": 96700
+    },
+    {
+      "epoch": 10.36,
+      "eval_loss": 0.45626798272132874,
+      "eval_runtime": 50.7297,
+      "eval_samples_per_second": 39.957,
+      "eval_steps_per_second": 5.007,
+      "eval_wer": 0.34347355118252393,
+      "step": 96800
+    },
+    {
+      "epoch": 10.37,
+      "eval_loss": 0.4315440356731415,
+      "eval_runtime": 49.4934,
+      "eval_samples_per_second": 40.955,
+      "eval_steps_per_second": 5.132,
+      "eval_wer": 0.33751579707528434,
+      "step": 96900
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 9.270343839541547e-05,
+      "loss": 0.1162,
+      "step": 97000
+    },
+    {
+      "epoch": 10.39,
+      "eval_loss": 0.44286635518074036,
+      "eval_runtime": 49.454,
+      "eval_samples_per_second": 40.988,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.33652283805741107,
+      "step": 97000
+    },
+    {
+      "epoch": 10.4,
+      "eval_loss": 0.4488806128501892,
+      "eval_runtime": 49.716,
+      "eval_samples_per_second": 40.772,
+      "eval_steps_per_second": 5.109,
+      "eval_wer": 0.33796714208340856,
+      "step": 97100
+    },
+    {
+      "epoch": 10.41,
+      "eval_loss": 0.4351952075958252,
+      "eval_runtime": 49.4834,
+      "eval_samples_per_second": 40.963,
+      "eval_steps_per_second": 5.133,
+      "eval_wer": 0.33571041704278753,
+      "step": 97200
+    },
+    {
+      "epoch": 10.42,
+      "eval_loss": 0.4390479624271393,
+      "eval_runtime": 49.1335,
+      "eval_samples_per_second": 41.255,
+      "eval_steps_per_second": 5.17,
+      "eval_wer": 0.3319191189745441,
+      "step": 97300
+    },
+    {
+      "epoch": 10.43,
+      "eval_loss": 0.4569990336894989,
+      "eval_runtime": 49.1672,
+      "eval_samples_per_second": 41.227,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.330294276945297,
+      "step": 97400
+    },
+    {
+      "epoch": 10.44,
+      "learning_rate": 9.162893982808023e-05,
+      "loss": 0.1151,
+      "step": 97500
+    },
+    {
+      "epoch": 10.44,
+      "eval_loss": 0.4691784083843231,
+      "eval_runtime": 49.324,
+      "eval_samples_per_second": 41.096,
+      "eval_steps_per_second": 5.15,
+      "eval_wer": 0.3343563820184149,
+      "step": 97500
+    },
+    {
+      "epoch": 10.45,
+      "eval_loss": 0.46045053005218506,
+      "eval_runtime": 49.3284,
+      "eval_samples_per_second": 41.092,
+      "eval_steps_per_second": 5.149,
+      "eval_wer": 0.33318288499729193,
+      "step": 97600
+    },
+    {
+      "epoch": 10.46,
+      "eval_loss": 0.4457150101661682,
+      "eval_runtime": 48.9643,
+      "eval_samples_per_second": 41.398,
+      "eval_steps_per_second": 5.187,
+      "eval_wer": 0.32379490882830836,
+      "step": 97700
+    },
+    {
+      "epoch": 10.47,
+      "eval_loss": 0.4298110902309418,
+      "eval_runtime": 49.1962,
+      "eval_samples_per_second": 41.202,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.33038454594692185,
+      "step": 97800
+    },
+    {
+      "epoch": 10.48,
+      "eval_loss": 0.4619194269180298,
+      "eval_runtime": 49.4405,
+      "eval_samples_per_second": 40.999,
+      "eval_steps_per_second": 5.137,
+      "eval_wer": 0.327405668893302,
+      "step": 97900
+    },
+    {
+      "epoch": 10.49,
+      "learning_rate": 9.055444126074499e-05,
+      "loss": 0.1105,
+      "step": 98000
+    },
+    {
+      "epoch": 10.49,
+      "eval_loss": 0.4361836612224579,
+      "eval_runtime": 49.0712,
+      "eval_samples_per_second": 41.307,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.32442679183968226,
+      "step": 98000
+    },
+    {
+      "epoch": 10.5,
+      "eval_loss": 0.45676031708717346,
+      "eval_runtime": 49.0451,
+      "eval_samples_per_second": 41.329,
+      "eval_steps_per_second": 5.179,
+      "eval_wer": 0.32894024192092436,
+      "step": 98100
+    },
+    {
+      "epoch": 10.51,
+      "eval_loss": 0.4522392153739929,
+      "eval_runtime": 49.3352,
+      "eval_samples_per_second": 41.086,
+      "eval_steps_per_second": 5.148,
+      "eval_wer": 0.33363423000541614,
+      "step": 98200
+    },
+    {
+      "epoch": 10.52,
+      "eval_loss": 0.43021589517593384,
+      "eval_runtime": 49.3191,
+      "eval_samples_per_second": 41.1,
+      "eval_steps_per_second": 5.15,
+      "eval_wer": 0.32569055786243006,
+      "step": 98300
+    },
+    {
+      "epoch": 10.54,
+      "eval_loss": 0.4505043625831604,
+      "eval_runtime": 49.4141,
+      "eval_samples_per_second": 41.021,
+      "eval_steps_per_second": 5.14,
+      "eval_wer": 0.32379490882830836,
+      "step": 98400
+    },
+    {
+      "epoch": 10.55,
+      "learning_rate": 8.947994269340974e-05,
+      "loss": 0.1164,
+      "step": 98500
+    },
+    {
+      "epoch": 10.55,
+      "eval_loss": 0.4430020749568939,
+      "eval_runtime": 49.2412,
+      "eval_samples_per_second": 41.165,
+      "eval_steps_per_second": 5.158,
+      "eval_wer": 0.3301137389420473,
+      "step": 98500
+    },
+    {
+      "epoch": 10.56,
+      "eval_loss": 0.4575265944004059,
+      "eval_runtime": 48.9818,
+      "eval_samples_per_second": 41.383,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.32830835890955046,
+      "step": 98600
+    },
+    {
+      "epoch": 10.57,
+      "eval_loss": 0.4447467029094696,
+      "eval_runtime": 49.3508,
+      "eval_samples_per_second": 41.073,
+      "eval_steps_per_second": 5.147,
+      "eval_wer": 0.32767647589817656,
+      "step": 98700
+    },
+    {
+      "epoch": 10.58,
+      "eval_loss": 0.4400414824485779,
+      "eval_runtime": 49.1998,
+      "eval_samples_per_second": 41.199,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3301137389420473,
+      "step": 98800
+    },
+    {
+      "epoch": 10.59,
+      "eval_loss": 0.44269418716430664,
+      "eval_runtime": 49.2271,
+      "eval_samples_per_second": 41.176,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.3287597039176747,
+      "step": 98900
+    },
+    {
+      "epoch": 10.6,
+      "learning_rate": 8.84054441260745e-05,
+      "loss": 0.1113,
+      "step": 99000
+    },
+    {
+      "epoch": 10.6,
+      "eval_loss": 0.4538458585739136,
+      "eval_runtime": 49.0089,
+      "eval_samples_per_second": 41.36,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.32478786784618163,
+      "step": 99000
+    },
+    {
+      "epoch": 10.61,
+      "eval_loss": 0.4518674612045288,
+      "eval_runtime": 49.3026,
+      "eval_samples_per_second": 41.113,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.3298429319371728,
+      "step": 99100
+    },
+    {
+      "epoch": 10.62,
+      "eval_loss": 0.4290253520011902,
+      "eval_runtime": 49.1748,
+      "eval_samples_per_second": 41.22,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.3248781368478065,
+      "step": 99200
+    },
+    {
+      "epoch": 10.63,
+      "eval_loss": 0.45013627409935,
+      "eval_runtime": 49.3453,
+      "eval_samples_per_second": 41.078,
+      "eval_steps_per_second": 5.147,
+      "eval_wer": 0.3219895287958115,
+      "step": 99300
+    },
+    {
+      "epoch": 10.64,
+      "eval_loss": 0.4410213530063629,
+      "eval_runtime": 49.1693,
+      "eval_samples_per_second": 41.225,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.3218089907925618,
+      "step": 99400
+    },
+    {
+      "epoch": 10.65,
+      "learning_rate": 8.733309455587391e-05,
+      "loss": 0.1159,
+      "step": 99500
+    },
+    {
+      "epoch": 10.65,
+      "eval_loss": 0.447797566652298,
+      "eval_runtime": 49.0342,
+      "eval_samples_per_second": 41.338,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.3210868387795631,
+      "step": 99500
+    },
+    {
+      "epoch": 10.66,
+      "eval_loss": 0.44618627429008484,
+      "eval_runtime": 48.7054,
+      "eval_samples_per_second": 41.618,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.3249684058494313,
+      "step": 99600
+    },
+    {
+      "epoch": 10.67,
+      "eval_loss": 0.45431581139564514,
+      "eval_runtime": 48.7336,
+      "eval_samples_per_second": 41.594,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.33020400794367216,
+      "step": 99700
+    },
+    {
+      "epoch": 10.69,
+      "eval_loss": 0.4462348520755768,
+      "eval_runtime": 48.6987,
+      "eval_samples_per_second": 41.623,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.3301137389420473,
+      "step": 99800
+    },
+    {
+      "epoch": 10.7,
+      "eval_loss": 0.44683581590652466,
+      "eval_runtime": 49.026,
+      "eval_samples_per_second": 41.345,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.32289221881205993,
+      "step": 99900
+    },
+    {
+      "epoch": 10.71,
+      "learning_rate": 8.625859598853867e-05,
+      "loss": 0.1161,
+      "step": 100000
+    },
+    {
+      "epoch": 10.71,
+      "eval_loss": 0.45149165391921997,
+      "eval_runtime": 49.1288,
+      "eval_samples_per_second": 41.259,
+      "eval_steps_per_second": 5.17,
+      "eval_wer": 0.3240657158331829,
+      "step": 100000
+    },
+    {
+      "epoch": 10.72,
+      "eval_loss": 0.44036853313446045,
+      "eval_runtime": 49.9413,
+      "eval_samples_per_second": 40.588,
+      "eval_steps_per_second": 5.086,
+      "eval_wer": 0.3275862068965517,
+      "step": 100100
+    },
+    {
+      "epoch": 10.73,
+      "eval_loss": 0.4439077079296112,
+      "eval_runtime": 48.7005,
+      "eval_samples_per_second": 41.622,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.3221700667990612,
+      "step": 100200
+    },
+    {
+      "epoch": 10.74,
+      "eval_loss": 0.43919479846954346,
+      "eval_runtime": 48.8991,
+      "eval_samples_per_second": 41.453,
+      "eval_steps_per_second": 5.194,
+      "eval_wer": 0.32569055786243006,
+      "step": 100300
+    },
+    {
+      "epoch": 10.75,
+      "eval_loss": 0.4475514590740204,
+      "eval_runtime": 49.0257,
+      "eval_samples_per_second": 41.346,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.33137750496479507,
+      "step": 100400
+    },
+    {
+      "epoch": 10.76,
+      "learning_rate": 8.518409742120343e-05,
+      "loss": 0.1199,
+      "step": 100500
+    },
+    {
+      "epoch": 10.76,
+      "eval_loss": 0.44927603006362915,
+      "eval_runtime": 49.1482,
+      "eval_samples_per_second": 41.243,
+      "eval_steps_per_second": 5.168,
+      "eval_wer": 0.32704459288680265,
+      "step": 100500
+    },
+    {
+      "epoch": 10.77,
+      "eval_loss": 0.44615232944488525,
+      "eval_runtime": 48.9489,
+      "eval_samples_per_second": 41.411,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.32235060480231087,
+      "step": 100600
+    },
+    {
+      "epoch": 10.78,
+      "eval_loss": 0.44666746258735657,
+      "eval_runtime": 48.7214,
+      "eval_samples_per_second": 41.604,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.33110669795992054,
+      "step": 100700
+    },
+    {
+      "epoch": 10.79,
+      "eval_loss": 0.4198402762413025,
+      "eval_runtime": 49.5384,
+      "eval_samples_per_second": 40.918,
+      "eval_steps_per_second": 5.127,
+      "eval_wer": 0.3228019498104351,
+      "step": 100800
+    },
+    {
+      "epoch": 10.8,
+      "eval_loss": 0.4348670244216919,
+      "eval_runtime": 48.746,
+      "eval_samples_per_second": 41.583,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.32253114280556056,
+      "step": 100900
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 8.410959885386819e-05,
+      "loss": 0.1146,
+      "step": 101000
+    },
+    {
+      "epoch": 10.81,
+      "eval_loss": 0.4371398985385895,
+      "eval_runtime": 49.8683,
+      "eval_samples_per_second": 40.647,
+      "eval_steps_per_second": 5.093,
+      "eval_wer": 0.32722513089005234,
+      "step": 101000
+    },
+    {
+      "epoch": 10.82,
+      "eval_loss": 0.4525171220302582,
+      "eval_runtime": 49.2008,
+      "eval_samples_per_second": 41.199,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3209965697779383,
+      "step": 101100
+    },
+    {
+      "epoch": 10.84,
+      "eval_loss": 0.4292885959148407,
+      "eval_runtime": 48.606,
+      "eval_samples_per_second": 41.703,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.32189925979418665,
+      "step": 101200
+    },
+    {
+      "epoch": 10.85,
+      "eval_loss": 0.4238409698009491,
+      "eval_runtime": 48.6683,
+      "eval_samples_per_second": 41.649,
+      "eval_steps_per_second": 5.219,
+      "eval_wer": 0.3216284527893121,
+      "step": 101300
+    },
+    {
+      "epoch": 10.86,
+      "eval_loss": 0.43767470121383667,
+      "eval_runtime": 48.6531,
+      "eval_samples_per_second": 41.662,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.32523921285430585,
+      "step": 101400
+    },
+    {
+      "epoch": 10.87,
+      "learning_rate": 8.303510028653295e-05,
+      "loss": 0.118,
+      "step": 101500
+    },
+    {
+      "epoch": 10.87,
+      "eval_loss": 0.43711698055267334,
+      "eval_runtime": 48.5847,
+      "eval_samples_per_second": 41.721,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.3208160317746886,
+      "step": 101500
+    },
+    {
+      "epoch": 10.88,
+      "eval_loss": 0.4216325879096985,
+      "eval_runtime": 48.5704,
+      "eval_samples_per_second": 41.733,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.31738580971294456,
+      "step": 101600
+    },
+    {
+      "epoch": 10.89,
+      "eval_loss": 0.43119776248931885,
+      "eval_runtime": 48.5795,
+      "eval_samples_per_second": 41.725,
+      "eval_steps_per_second": 5.229,
+      "eval_wer": 0.3189203827405669,
+      "step": 101700
+    },
+    {
+      "epoch": 10.9,
+      "eval_loss": 0.4317311942577362,
+      "eval_runtime": 48.6163,
+      "eval_samples_per_second": 41.694,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.3203646867665644,
+      "step": 101800
+    },
+    {
+      "epoch": 10.91,
+      "eval_loss": 0.43034303188323975,
+      "eval_runtime": 48.6584,
+      "eval_samples_per_second": 41.658,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.32352410182343383,
+      "step": 101900
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 8.19606017191977e-05,
+      "loss": 0.114,
+      "step": 102000
+    },
+    {
+      "epoch": 10.92,
+      "eval_loss": 0.44159621000289917,
+      "eval_runtime": 49.7514,
+      "eval_samples_per_second": 40.743,
+      "eval_steps_per_second": 5.105,
+      "eval_wer": 0.31576096768369744,
+      "step": 102000
+    },
+    {
+      "epoch": 10.93,
+      "eval_loss": 0.4239681363105774,
+      "eval_runtime": 48.5431,
+      "eval_samples_per_second": 41.757,
+      "eval_steps_per_second": 5.232,
+      "eval_wer": 0.31946199675031595,
+      "step": 102100
+    },
+    {
+      "epoch": 10.94,
+      "eval_loss": 0.4340362548828125,
+      "eval_runtime": 49.0823,
+      "eval_samples_per_second": 41.298,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.31494854666907385,
+      "step": 102200
+    },
+    {
+      "epoch": 10.95,
+      "eval_loss": 0.431062251329422,
+      "eval_runtime": 48.4884,
+      "eval_samples_per_second": 41.804,
+      "eval_steps_per_second": 5.238,
+      "eval_wer": 0.3215381837876873,
+      "step": 102300
+    },
+    {
+      "epoch": 10.96,
+      "eval_loss": 0.4261479377746582,
+      "eval_runtime": 49.008,
+      "eval_samples_per_second": 41.361,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.32379490882830836,
+      "step": 102400
+    },
+    {
+      "epoch": 10.97,
+      "learning_rate": 8.088610315186246e-05,
+      "loss": 0.1152,
+      "step": 102500
+    },
+    {
+      "epoch": 10.97,
+      "eval_loss": 0.42632901668548584,
+      "eval_runtime": 49.1051,
+      "eval_samples_per_second": 41.279,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.3206354937714389,
+      "step": 102500
+    },
+    {
+      "epoch": 10.98,
+      "eval_loss": 0.4325226843357086,
+      "eval_runtime": 49.3257,
+      "eval_samples_per_second": 41.094,
+      "eval_steps_per_second": 5.149,
+      "eval_wer": 0.3293915869290486,
+      "step": 102600
+    },
+    {
+      "epoch": 11.0,
+      "eval_loss": 0.4327019155025482,
+      "eval_runtime": 49.2829,
+      "eval_samples_per_second": 41.13,
+      "eval_steps_per_second": 5.154,
+      "eval_wer": 0.3187398447373172,
+      "step": 102700
+    },
+    {
+      "epoch": 11.01,
+      "eval_loss": 0.44230592250823975,
+      "eval_runtime": 49.246,
+      "eval_samples_per_second": 41.161,
+      "eval_steps_per_second": 5.158,
+      "eval_wer": 0.31946199675031595,
+      "step": 102800
+    },
+    {
+      "epoch": 11.02,
+      "eval_loss": 0.43406760692596436,
+      "eval_runtime": 49.0206,
+      "eval_samples_per_second": 41.35,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.32767647589817656,
+      "step": 102900
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 7.981160458452722e-05,
+      "loss": 0.1084,
+      "step": 103000
+    },
+    {
+      "epoch": 11.03,
+      "eval_loss": 0.42317888140678406,
+      "eval_runtime": 49.1704,
+      "eval_samples_per_second": 41.224,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.3243365228380574,
+      "step": 103000
+    },
+    {
+      "epoch": 11.04,
+      "eval_loss": 0.4355052411556244,
+      "eval_runtime": 49.0712,
+      "eval_samples_per_second": 41.307,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.31837876873081783,
+      "step": 103100
+    },
+    {
+      "epoch": 11.05,
+      "eval_loss": 0.43737396597862244,
+      "eval_runtime": 49.4194,
+      "eval_samples_per_second": 41.016,
+      "eval_steps_per_second": 5.14,
+      "eval_wer": 0.327405668893302,
+      "step": 103200
+    },
+    {
+      "epoch": 11.06,
+      "eval_loss": 0.44839176535606384,
+      "eval_runtime": 49.5435,
+      "eval_samples_per_second": 40.914,
+      "eval_steps_per_second": 5.127,
+      "eval_wer": 0.3304748149485467,
+      "step": 103300
+    },
+    {
+      "epoch": 11.07,
+      "eval_loss": 0.44234499335289,
+      "eval_runtime": 48.8359,
+      "eval_samples_per_second": 41.506,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.3226214118071854,
+      "step": 103400
+    },
+    {
+      "epoch": 11.08,
+      "learning_rate": 7.873925501432665e-05,
+      "loss": 0.1003,
+      "step": 103500
+    },
+    {
+      "epoch": 11.08,
+      "eval_loss": 0.45182204246520996,
+      "eval_runtime": 49.1484,
+      "eval_samples_per_second": 41.242,
+      "eval_steps_per_second": 5.168,
+      "eval_wer": 0.3224408738039357,
+      "step": 103500
+    },
+    {
+      "epoch": 11.09,
+      "eval_loss": 0.45178499817848206,
+      "eval_runtime": 48.7083,
+      "eval_samples_per_second": 41.615,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.3243365228380574,
+      "step": 103600
+    },
+    {
+      "epoch": 11.1,
+      "eval_loss": 0.4281817376613617,
+      "eval_runtime": 48.8171,
+      "eval_samples_per_second": 41.522,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.32072576277306375,
+      "step": 103700
+    },
+    {
+      "epoch": 11.11,
+      "eval_loss": 0.44184988737106323,
+      "eval_runtime": 48.9326,
+      "eval_samples_per_second": 41.424,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.3219895287958115,
+      "step": 103800
+    },
+    {
+      "epoch": 11.12,
+      "eval_loss": 0.4411134123802185,
+      "eval_runtime": 48.7339,
+      "eval_samples_per_second": 41.593,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.3216284527893121,
+      "step": 103900
+    },
+    {
+      "epoch": 11.13,
+      "learning_rate": 7.766475644699141e-05,
+      "loss": 0.1009,
+      "step": 104000
+    },
+    {
+      "epoch": 11.13,
+      "eval_loss": 0.44739264249801636,
+      "eval_runtime": 49.071,
+      "eval_samples_per_second": 41.307,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.32379490882830836,
+      "step": 104000
+    },
+    {
+      "epoch": 11.15,
+      "eval_loss": 0.4405987858772278,
+      "eval_runtime": 48.9357,
+      "eval_samples_per_second": 41.422,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.3245170608413071,
+      "step": 104100
+    },
+    {
+      "epoch": 11.16,
+      "eval_loss": 0.43838658928871155,
+      "eval_runtime": 48.6654,
+      "eval_samples_per_second": 41.652,
+      "eval_steps_per_second": 5.219,
+      "eval_wer": 0.32415598483480773,
+      "step": 104200
+    },
+    {
+      "epoch": 11.17,
+      "eval_loss": 0.470173716545105,
+      "eval_runtime": 48.9255,
+      "eval_samples_per_second": 41.43,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.3265029788770536,
+      "step": 104300
+    },
+    {
+      "epoch": 11.18,
+      "eval_loss": 0.4611179828643799,
+      "eval_runtime": 48.9633,
+      "eval_samples_per_second": 41.398,
+      "eval_steps_per_second": 5.188,
+      "eval_wer": 0.32659324787867844,
+      "step": 104400
+    },
+    {
+      "epoch": 11.19,
+      "learning_rate": 7.659025787965614e-05,
+      "loss": 0.0992,
+      "step": 104500
+    },
+    {
+      "epoch": 11.19,
+      "eval_loss": 0.44245555996894836,
+      "eval_runtime": 48.853,
+      "eval_samples_per_second": 41.492,
+      "eval_steps_per_second": 5.199,
+      "eval_wer": 0.3210868387795631,
+      "step": 104500
+    },
+    {
+      "epoch": 11.2,
+      "eval_loss": 0.45753201842308044,
+      "eval_runtime": 48.7428,
+      "eval_samples_per_second": 41.586,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.3221700667990612,
+      "step": 104600
+    },
+    {
+      "epoch": 11.21,
+      "eval_loss": 0.4448811411857605,
+      "eval_runtime": 48.4957,
+      "eval_samples_per_second": 41.797,
+      "eval_steps_per_second": 5.238,
+      "eval_wer": 0.3208160317746886,
+      "step": 104700
+    },
+    {
+      "epoch": 11.22,
+      "eval_loss": 0.47151175141334534,
+      "eval_runtime": 48.9938,
+      "eval_samples_per_second": 41.373,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.3208160317746886,
+      "step": 104800
+    },
+    {
+      "epoch": 11.23,
+      "eval_loss": 0.4468552768230438,
+      "eval_runtime": 48.8885,
+      "eval_samples_per_second": 41.462,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.322260335800686,
+      "step": 104900
+    },
+    {
+      "epoch": 11.24,
+      "learning_rate": 7.55157593123209e-05,
+      "loss": 0.1021,
+      "step": 105000
+    },
+    {
+      "epoch": 11.24,
+      "eval_loss": 0.4536341726779938,
+      "eval_runtime": 48.4312,
+      "eval_samples_per_second": 41.853,
+      "eval_steps_per_second": 5.245,
+      "eval_wer": 0.32253114280556056,
+      "step": 105000
+    },
+    {
+      "epoch": 11.25,
+      "eval_loss": 0.4629049599170685,
+      "eval_runtime": 48.4787,
+      "eval_samples_per_second": 41.812,
+      "eval_steps_per_second": 5.239,
+      "eval_wer": 0.323433832821809,
+      "step": 105100
+    },
+    {
+      "epoch": 11.26,
+      "eval_loss": 0.4549843966960907,
+      "eval_runtime": 48.5414,
+      "eval_samples_per_second": 41.758,
+      "eval_steps_per_second": 5.233,
+      "eval_wer": 0.32054522476981406,
+      "step": 105200
+    },
+    {
+      "epoch": 11.27,
+      "eval_loss": 0.4598093032836914,
+      "eval_runtime": 48.7765,
+      "eval_samples_per_second": 41.557,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.3212673767828128,
+      "step": 105300
+    },
+    {
+      "epoch": 11.28,
+      "eval_loss": 0.45224353671073914,
+      "eval_runtime": 48.7101,
+      "eval_samples_per_second": 41.614,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.3179274237226936,
+      "step": 105400
+    },
+    {
+      "epoch": 11.3,
+      "learning_rate": 7.444126074498566e-05,
+      "loss": 0.1021,
+      "step": 105500
+    },
+    {
+      "epoch": 11.3,
+      "eval_loss": 0.4658335745334625,
+      "eval_runtime": 48.593,
+      "eval_samples_per_second": 41.714,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.3210868387795631,
+      "step": 105500
+    },
+    {
+      "epoch": 11.31,
+      "eval_loss": 0.4664027690887451,
+      "eval_runtime": 49.0611,
+      "eval_samples_per_second": 41.316,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.3195522657519408,
+      "step": 105600
+    },
+    {
+      "epoch": 11.32,
+      "eval_loss": 0.47357499599456787,
+      "eval_runtime": 49.0629,
+      "eval_samples_per_second": 41.314,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.31774688571944393,
+      "step": 105700
+    },
+    {
+      "epoch": 11.33,
+      "eval_loss": 0.4587169885635376,
+      "eval_runtime": 49.4541,
+      "eval_samples_per_second": 40.987,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.31576096768369744,
+      "step": 105800
+    },
+    {
+      "epoch": 11.34,
+      "eval_loss": 0.45886215567588806,
+      "eval_runtime": 48.5172,
+      "eval_samples_per_second": 41.779,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.3193717277486911,
+      "step": 105900
+    },
+    {
+      "epoch": 11.35,
+      "learning_rate": 7.336676217765042e-05,
+      "loss": 0.1025,
+      "step": 106000
+    },
+    {
+      "epoch": 11.35,
+      "eval_loss": 0.4691576361656189,
+      "eval_runtime": 48.5278,
+      "eval_samples_per_second": 41.77,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.32135764578443765,
+      "step": 106000
+    },
+    {
+      "epoch": 11.36,
+      "eval_loss": 0.43815869092941284,
+      "eval_runtime": 48.5469,
+      "eval_samples_per_second": 41.753,
+      "eval_steps_per_second": 5.232,
+      "eval_wer": 0.3181079617259433,
+      "step": 106100
+    },
+    {
+      "epoch": 11.37,
+      "eval_loss": 0.4555544853210449,
+      "eval_runtime": 48.8057,
+      "eval_samples_per_second": 41.532,
+      "eval_steps_per_second": 5.204,
+      "eval_wer": 0.3184690377324427,
+      "step": 106200
+    },
+    {
+      "epoch": 11.38,
+      "eval_loss": 0.4445129334926605,
+      "eval_runtime": 49.8518,
+      "eval_samples_per_second": 40.661,
+      "eval_steps_per_second": 5.095,
+      "eval_wer": 0.3191009207438166,
+      "step": 106300
+    },
+    {
+      "epoch": 11.39,
+      "eval_loss": 0.4379484951496124,
+      "eval_runtime": 48.9128,
+      "eval_samples_per_second": 41.441,
+      "eval_steps_per_second": 5.193,
+      "eval_wer": 0.3163025816934465,
+      "step": 106400
+    },
+    {
+      "epoch": 11.4,
+      "learning_rate": 7.229226361031518e-05,
+      "loss": 0.104,
+      "step": 106500
+    },
+    {
+      "epoch": 11.4,
+      "eval_loss": 0.44537481665611267,
+      "eval_runtime": 48.5687,
+      "eval_samples_per_second": 41.735,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.3219895287958115,
+      "step": 106500
+    },
+    {
+      "epoch": 11.41,
+      "eval_loss": 0.44634634256362915,
+      "eval_runtime": 48.8665,
+      "eval_samples_per_second": 41.48,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.32009387976168985,
+      "step": 106600
+    },
+    {
+      "epoch": 11.42,
+      "eval_loss": 0.4550469219684601,
+      "eval_runtime": 48.5571,
+      "eval_samples_per_second": 41.745,
+      "eval_steps_per_second": 5.231,
+      "eval_wer": 0.3172955407113197,
+      "step": 106700
+    },
+    {
+      "epoch": 11.43,
+      "eval_loss": 0.44044601917266846,
+      "eval_runtime": 48.518,
+      "eval_samples_per_second": 41.778,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.31675392670157065,
+      "step": 106800
+    },
+    {
+      "epoch": 11.45,
+      "eval_loss": 0.45691826939582825,
+      "eval_runtime": 48.6409,
+      "eval_samples_per_second": 41.673,
+      "eval_steps_per_second": 5.222,
+      "eval_wer": 0.3170247337064452,
+      "step": 106900
+    },
+    {
+      "epoch": 11.46,
+      "learning_rate": 7.121776504297993e-05,
+      "loss": 0.1016,
+      "step": 107000
+    },
+    {
+      "epoch": 11.46,
+      "eval_loss": 0.45285138487815857,
+      "eval_runtime": 48.4798,
+      "eval_samples_per_second": 41.811,
+      "eval_steps_per_second": 5.239,
+      "eval_wer": 0.3168441957031955,
+      "step": 107000
+    },
+    {
+      "epoch": 11.47,
+      "eval_loss": 0.45870763063430786,
+      "eval_runtime": 49.1364,
+      "eval_samples_per_second": 41.252,
+      "eval_steps_per_second": 5.169,
+      "eval_wer": 0.3172955407113197,
+      "step": 107100
+    },
+    {
+      "epoch": 11.48,
+      "eval_loss": 0.4505434036254883,
+      "eval_runtime": 48.8484,
+      "eval_samples_per_second": 41.496,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.31720527170969487,
+      "step": 107200
+    },
+    {
+      "epoch": 11.49,
+      "eval_loss": 0.44892099499702454,
+      "eval_runtime": 48.8062,
+      "eval_samples_per_second": 41.532,
+      "eval_steps_per_second": 5.204,
+      "eval_wer": 0.3158512366853223,
+      "step": 107300
+    },
+    {
+      "epoch": 11.5,
+      "eval_loss": 0.4527936577796936,
+      "eval_runtime": 48.8297,
+      "eval_samples_per_second": 41.512,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.3129626286333273,
+      "step": 107400
+    },
+    {
+      "epoch": 11.51,
+      "learning_rate": 7.014326647564469e-05,
+      "loss": 0.1001,
+      "step": 107500
+    },
+    {
+      "epoch": 11.51,
+      "eval_loss": 0.4472792446613312,
+      "eval_runtime": 48.6292,
+      "eval_samples_per_second": 41.683,
+      "eval_steps_per_second": 5.223,
+      "eval_wer": 0.3181079617259433,
+      "step": 107500
+    },
+    {
+      "epoch": 11.52,
+      "eval_loss": 0.4434313476085663,
+      "eval_runtime": 48.7732,
+      "eval_samples_per_second": 41.56,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.31756634771619424,
+      "step": 107600
+    },
+    {
+      "epoch": 11.53,
+      "eval_loss": 0.45973971486091614,
+      "eval_runtime": 48.7123,
+      "eval_samples_per_second": 41.612,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.3185593067340675,
+      "step": 107700
+    },
+    {
+      "epoch": 11.54,
+      "eval_loss": 0.4351407289505005,
+      "eval_runtime": 48.8735,
+      "eval_samples_per_second": 41.474,
+      "eval_steps_per_second": 5.197,
+      "eval_wer": 0.3159415056869471,
+      "step": 107800
+    },
+    {
+      "epoch": 11.55,
+      "eval_loss": 0.44710567593574524,
+      "eval_runtime": 49.9625,
+      "eval_samples_per_second": 40.57,
+      "eval_steps_per_second": 5.084,
+      "eval_wer": 0.3184690377324427,
+      "step": 107900
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 6.907091690544412e-05,
+      "loss": 0.1005,
+      "step": 108000
+    },
+    {
+      "epoch": 11.56,
+      "eval_loss": 0.44572997093200684,
+      "eval_runtime": 48.7976,
+      "eval_samples_per_second": 41.539,
+      "eval_steps_per_second": 5.205,
+      "eval_wer": 0.3191009207438166,
+      "step": 108000
+    },
+    {
+      "epoch": 11.57,
+      "eval_loss": 0.45437440276145935,
+      "eval_runtime": 48.6844,
+      "eval_samples_per_second": 41.636,
+      "eval_steps_per_second": 5.217,
+      "eval_wer": 0.32930131792742373,
+      "step": 108100
+    },
+    {
+      "epoch": 11.58,
+      "eval_loss": 0.4436147212982178,
+      "eval_runtime": 49.2491,
+      "eval_samples_per_second": 41.158,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.32207979779743634,
+      "step": 108200
+    },
+    {
+      "epoch": 11.6,
+      "eval_loss": 0.4641793668270111,
+      "eval_runtime": 48.7793,
+      "eval_samples_per_second": 41.554,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.3269543238851778,
+      "step": 108300
+    },
+    {
+      "epoch": 11.61,
+      "eval_loss": 0.44737499952316284,
+      "eval_runtime": 48.6883,
+      "eval_samples_per_second": 41.632,
+      "eval_steps_per_second": 5.217,
+      "eval_wer": 0.32704459288680265,
+      "step": 108400
+    },
+    {
+      "epoch": 11.62,
+      "learning_rate": 6.799641833810888e-05,
+      "loss": 0.1031,
+      "step": 108500
+    },
+    {
+      "epoch": 11.62,
+      "eval_loss": 0.44577571749687195,
+      "eval_runtime": 48.8043,
+      "eval_samples_per_second": 41.533,
+      "eval_steps_per_second": 5.204,
+      "eval_wer": 0.31964253475356563,
+      "step": 108500
+    },
+    {
+      "epoch": 11.63,
+      "eval_loss": 0.4723000228404999,
+      "eval_runtime": 48.8424,
+      "eval_samples_per_second": 41.501,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.3204549557681892,
+      "step": 108600
+    },
+    {
+      "epoch": 11.64,
+      "eval_loss": 0.4507168233394623,
+      "eval_runtime": 48.9897,
+      "eval_samples_per_second": 41.376,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.3226214118071854,
+      "step": 108700
+    },
+    {
+      "epoch": 11.65,
+      "eval_loss": 0.44235825538635254,
+      "eval_runtime": 48.7448,
+      "eval_samples_per_second": 41.584,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.3212673767828128,
+      "step": 108800
+    },
+    {
+      "epoch": 11.66,
+      "eval_loss": 0.4511132836341858,
+      "eval_runtime": 48.7014,
+      "eval_samples_per_second": 41.621,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.3212673767828128,
+      "step": 108900
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 6.692191977077364e-05,
+      "loss": 0.1014,
+      "step": 109000
+    },
+    {
+      "epoch": 11.67,
+      "eval_loss": 0.4421851634979248,
+      "eval_runtime": 48.8914,
+      "eval_samples_per_second": 41.459,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.3204549557681892,
+      "step": 109000
+    },
+    {
+      "epoch": 11.68,
+      "eval_loss": 0.44982463121414185,
+      "eval_runtime": 48.4128,
+      "eval_samples_per_second": 41.869,
+      "eval_steps_per_second": 5.247,
+      "eval_wer": 0.31801769272431846,
+      "step": 109100
+    },
+    {
+      "epoch": 11.69,
+      "eval_loss": 0.43027353286743164,
+      "eval_runtime": 48.7237,
+      "eval_samples_per_second": 41.602,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.31666365769994587,
+      "step": 109200
+    },
+    {
+      "epoch": 11.7,
+      "eval_loss": 0.4483342170715332,
+      "eval_runtime": 48.6826,
+      "eval_samples_per_second": 41.637,
+      "eval_steps_per_second": 5.217,
+      "eval_wer": 0.3107961725943311,
+      "step": 109300
+    },
+    {
+      "epoch": 11.71,
+      "eval_loss": 0.4547644257545471,
+      "eval_runtime": 49.6041,
+      "eval_samples_per_second": 40.864,
+      "eval_steps_per_second": 5.121,
+      "eval_wer": 0.31693446470482034,
+      "step": 109400
+    },
+    {
+      "epoch": 11.72,
+      "learning_rate": 6.584742120343839e-05,
+      "loss": 0.0981,
+      "step": 109500
+    },
+    {
+      "epoch": 11.72,
+      "eval_loss": 0.4406187832355499,
+      "eval_runtime": 49.0859,
+      "eval_samples_per_second": 41.295,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3121502076187037,
+      "step": 109500
+    },
+    {
+      "epoch": 11.73,
+      "eval_loss": 0.42933210730552673,
+      "eval_runtime": 49.1781,
+      "eval_samples_per_second": 41.218,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.311428055605705,
+      "step": 109600
+    },
+    {
+      "epoch": 11.75,
+      "eval_loss": 0.4368530511856079,
+      "eval_runtime": 48.9575,
+      "eval_samples_per_second": 41.403,
+      "eval_steps_per_second": 5.188,
+      "eval_wer": 0.3158512366853223,
+      "step": 109700
+    },
+    {
+      "epoch": 11.76,
+      "eval_loss": 0.43642425537109375,
+      "eval_runtime": 48.9813,
+      "eval_samples_per_second": 41.383,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.31639285069507134,
+      "step": 109800
+    },
+    {
+      "epoch": 11.77,
+      "eval_loss": 0.4358440339565277,
+      "eval_runtime": 48.7711,
+      "eval_samples_per_second": 41.562,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.3189203827405669,
+      "step": 109900
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 6.477292263610315e-05,
+      "loss": 0.1023,
+      "step": 110000
+    },
+    {
+      "epoch": 11.78,
+      "eval_loss": 0.4281153976917267,
+      "eval_runtime": 48.845,
+      "eval_samples_per_second": 41.499,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.318288499729193,
+      "step": 110000
+    },
+    {
+      "epoch": 11.79,
+      "eval_loss": 0.44035249948501587,
+      "eval_runtime": 49.3957,
+      "eval_samples_per_second": 41.036,
+      "eval_steps_per_second": 5.142,
+      "eval_wer": 0.3158512366853223,
+      "step": 110100
+    },
+    {
+      "epoch": 11.8,
+      "eval_loss": 0.44708630442619324,
+      "eval_runtime": 48.6464,
+      "eval_samples_per_second": 41.668,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.31350424264307636,
+      "step": 110200
+    },
+    {
+      "epoch": 11.81,
+      "eval_loss": 0.449814110994339,
+      "eval_runtime": 48.8935,
+      "eval_samples_per_second": 41.457,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.32009387976168985,
+      "step": 110300
+    },
+    {
+      "epoch": 11.82,
+      "eval_loss": 0.45270591974258423,
+      "eval_runtime": 48.7747,
+      "eval_samples_per_second": 41.558,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.3161220436901968,
+      "step": 110400
+    },
+    {
+      "epoch": 11.83,
+      "learning_rate": 6.370057306590257e-05,
+      "loss": 0.0988,
+      "step": 110500
+    },
+    {
+      "epoch": 11.83,
+      "eval_loss": 0.4439750909805298,
+      "eval_runtime": 48.8822,
+      "eval_samples_per_second": 41.467,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.3172955407113197,
+      "step": 110500
+    },
+    {
+      "epoch": 11.84,
+      "eval_loss": 0.4356108605861664,
+      "eval_runtime": 48.8938,
+      "eval_samples_per_second": 41.457,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.3135945116447012,
+      "step": 110600
+    },
+    {
+      "epoch": 11.85,
+      "eval_loss": 0.43082764744758606,
+      "eval_runtime": 48.9279,
+      "eval_samples_per_second": 41.428,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.31350424264307636,
+      "step": 110700
+    },
+    {
+      "epoch": 11.86,
+      "eval_loss": 0.4294142723083496,
+      "eval_runtime": 49.0537,
+      "eval_samples_per_second": 41.322,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.3191911897454414,
+      "step": 110800
+    },
+    {
+      "epoch": 11.87,
+      "eval_loss": 0.4240707755088806,
+      "eval_runtime": 48.8428,
+      "eval_samples_per_second": 41.501,
+      "eval_steps_per_second": 5.2,
+      "eval_wer": 0.3168441957031955,
+      "step": 110900
+    },
+    {
+      "epoch": 11.88,
+      "learning_rate": 6.263037249283667e-05,
+      "loss": 0.1022,
+      "step": 111000
+    },
+    {
+      "epoch": 11.88,
+      "eval_loss": 0.44202566146850586,
+      "eval_runtime": 49.1216,
+      "eval_samples_per_second": 41.265,
+      "eval_steps_per_second": 5.171,
+      "eval_wer": 0.3156706986820726,
+      "step": 111000
+    },
+    {
+      "epoch": 11.9,
+      "eval_loss": 0.43126294016838074,
+      "eval_runtime": 49.1248,
+      "eval_samples_per_second": 41.262,
+      "eval_steps_per_second": 5.171,
+      "eval_wer": 0.3125112836252031,
+      "step": 111100
+    },
+    {
+      "epoch": 11.91,
+      "eval_loss": 0.42133891582489014,
+      "eval_runtime": 48.989,
+      "eval_samples_per_second": 41.377,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.31675392670157065,
+      "step": 111200
+    },
+    {
+      "epoch": 11.92,
+      "eval_loss": 0.4351927936077118,
+      "eval_runtime": 50.4732,
+      "eval_samples_per_second": 40.16,
+      "eval_steps_per_second": 5.032,
+      "eval_wer": 0.31350424264307636,
+      "step": 111300
+    },
+    {
+      "epoch": 11.93,
+      "eval_loss": 0.4296703040599823,
+      "eval_runtime": 49.0792,
+      "eval_samples_per_second": 41.301,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3116085936089547,
+      "step": 111400
+    },
+    {
+      "epoch": 11.94,
+      "learning_rate": 6.155587392550143e-05,
+      "loss": 0.1032,
+      "step": 111500
+    },
+    {
+      "epoch": 11.94,
+      "eval_loss": 0.42183753848075867,
+      "eval_runtime": 49.0367,
+      "eval_samples_per_second": 41.336,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.31368478064632604,
+      "step": 111500
+    },
+    {
+      "epoch": 11.95,
+      "eval_loss": 0.43339774012565613,
+      "eval_runtime": 48.4486,
+      "eval_samples_per_second": 41.838,
+      "eval_steps_per_second": 5.243,
+      "eval_wer": 0.3123307456219534,
+      "step": 111600
+    },
+    {
+      "epoch": 11.96,
+      "eval_loss": 0.4373456537723541,
+      "eval_runtime": 48.7637,
+      "eval_samples_per_second": 41.568,
+      "eval_steps_per_second": 5.209,
+      "eval_wer": 0.3174760787145694,
+      "step": 111700
+    },
+    {
+      "epoch": 11.97,
+      "eval_loss": 0.4298686385154724,
+      "eval_runtime": 49.5415,
+      "eval_samples_per_second": 40.915,
+      "eval_steps_per_second": 5.127,
+      "eval_wer": 0.31603177468857196,
+      "step": 111800
+    },
+    {
+      "epoch": 11.98,
+      "eval_loss": 0.432638943195343,
+      "eval_runtime": 48.7304,
+      "eval_samples_per_second": 41.596,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.3189203827405669,
+      "step": 111900
+    },
+    {
+      "epoch": 11.99,
+      "learning_rate": 6.0481375358166186e-05,
+      "loss": 0.0969,
+      "step": 112000
+    },
+    {
+      "epoch": 11.99,
+      "eval_loss": 0.4207654595375061,
+      "eval_runtime": 48.9546,
+      "eval_samples_per_second": 41.406,
+      "eval_steps_per_second": 5.188,
+      "eval_wer": 0.31864957573569236,
+      "step": 112000
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": 0.4385160207748413,
+      "eval_runtime": 48.8057,
+      "eval_samples_per_second": 41.532,
+      "eval_steps_per_second": 5.204,
+      "eval_wer": 0.31693446470482034,
+      "step": 112100
+    },
+    {
+      "epoch": 12.01,
+      "eval_loss": 0.4453379809856415,
+      "eval_runtime": 48.7148,
+      "eval_samples_per_second": 41.609,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.31558042968044775,
+      "step": 112200
+    },
+    {
+      "epoch": 12.02,
+      "eval_loss": 0.4596421718597412,
+      "eval_runtime": 48.8968,
+      "eval_samples_per_second": 41.455,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.3133237046398267,
+      "step": 112300
+    },
+    {
+      "epoch": 12.03,
+      "eval_loss": 0.45092177391052246,
+      "eval_runtime": 48.8687,
+      "eval_samples_per_second": 41.478,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.3092615995667088,
+      "step": 112400
+    },
+    {
+      "epoch": 12.04,
+      "learning_rate": 5.9406876790830945e-05,
+      "loss": 0.0901,
+      "step": 112500
+    },
+    {
+      "epoch": 12.04,
+      "eval_loss": 0.45350125432014465,
+      "eval_runtime": 48.6941,
+      "eval_samples_per_second": 41.627,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.3137750496479509,
+      "step": 112500
+    },
+    {
+      "epoch": 12.06,
+      "eval_loss": 0.4370967447757721,
+      "eval_runtime": 49.3215,
+      "eval_samples_per_second": 41.098,
+      "eval_steps_per_second": 5.15,
+      "eval_wer": 0.3144069326593248,
+      "step": 112600
+    },
+    {
+      "epoch": 12.07,
+      "eval_loss": 0.4498507082462311,
+      "eval_runtime": 49.0717,
+      "eval_samples_per_second": 41.307,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.31539989167719806,
+      "step": 112700
+    },
+    {
+      "epoch": 12.08,
+      "eval_loss": 0.46147477626800537,
+      "eval_runtime": 48.753,
+      "eval_samples_per_second": 41.577,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.3198230727568153,
+      "step": 112800
+    },
+    {
+      "epoch": 12.09,
+      "eval_loss": 0.45230022072792053,
+      "eval_runtime": 48.945,
+      "eval_samples_per_second": 41.414,
+      "eval_steps_per_second": 5.19,
+      "eval_wer": 0.3176566167178191,
+      "step": 112900
+    },
+    {
+      "epoch": 12.1,
+      "learning_rate": 5.833237822349569e-05,
+      "loss": 0.0889,
+      "step": 113000
+    },
+    {
+      "epoch": 12.1,
+      "eval_loss": 0.44123727083206177,
+      "eval_runtime": 49.3729,
+      "eval_samples_per_second": 41.055,
+      "eval_steps_per_second": 5.145,
+      "eval_wer": 0.3129626286333273,
+      "step": 113000
+    },
+    {
+      "epoch": 12.11,
+      "eval_loss": 0.4471343457698822,
+      "eval_runtime": 49.1556,
+      "eval_samples_per_second": 41.236,
+      "eval_steps_per_second": 5.167,
+      "eval_wer": 0.3181079617259433,
+      "step": 113100
+    },
+    {
+      "epoch": 12.12,
+      "eval_loss": 0.45298895239830017,
+      "eval_runtime": 49.0223,
+      "eval_samples_per_second": 41.349,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.31693446470482034,
+      "step": 113200
+    },
+    {
+      "epoch": 12.13,
+      "eval_loss": 0.467046320438385,
+      "eval_runtime": 49.039,
+      "eval_samples_per_second": 41.334,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.314858277667449,
+      "step": 113300
+    },
+    {
+      "epoch": 12.14,
+      "eval_loss": 0.4594302475452423,
+      "eval_runtime": 50.5981,
+      "eval_samples_per_second": 40.061,
+      "eval_steps_per_second": 5.02,
+      "eval_wer": 0.31413612565445026,
+      "step": 113400
+    },
+    {
+      "epoch": 12.15,
+      "learning_rate": 5.725787965616045e-05,
+      "loss": 0.0917,
+      "step": 113500
+    },
+    {
+      "epoch": 12.15,
+      "eval_loss": 0.4622795879840851,
+      "eval_runtime": 50.5474,
+      "eval_samples_per_second": 40.101,
+      "eval_steps_per_second": 5.025,
+      "eval_wer": 0.31269182162845277,
+      "step": 113500
+    },
+    {
+      "epoch": 12.16,
+      "eval_loss": 0.4459761083126068,
+      "eval_runtime": 50.9405,
+      "eval_samples_per_second": 39.792,
+      "eval_steps_per_second": 4.986,
+      "eval_wer": 0.3133237046398267,
+      "step": 113600
+    },
+    {
+      "epoch": 12.17,
+      "eval_loss": 0.45123758912086487,
+      "eval_runtime": 50.7689,
+      "eval_samples_per_second": 39.926,
+      "eval_steps_per_second": 5.003,
+      "eval_wer": 0.3191009207438166,
+      "step": 113700
+    },
+    {
+      "epoch": 12.18,
+      "eval_loss": 0.46812236309051514,
+      "eval_runtime": 50.5875,
+      "eval_samples_per_second": 40.069,
+      "eval_steps_per_second": 5.021,
+      "eval_wer": 0.3135945116447012,
+      "step": 113800
+    },
+    {
+      "epoch": 12.19,
+      "eval_loss": 0.45636752247810364,
+      "eval_runtime": 50.7928,
+      "eval_samples_per_second": 39.907,
+      "eval_steps_per_second": 5.001,
+      "eval_wer": 0.31287235963170246,
+      "step": 113900
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 5.618553008595988e-05,
+      "loss": 0.0906,
+      "step": 114000
+    },
+    {
+      "epoch": 12.21,
+      "eval_loss": 0.44822677969932556,
+      "eval_runtime": 50.8831,
+      "eval_samples_per_second": 39.836,
+      "eval_steps_per_second": 4.992,
+      "eval_wer": 0.3107059035927063,
+      "step": 114000
+    },
+    {
+      "epoch": 12.22,
+      "eval_loss": 0.4594801664352417,
+      "eval_runtime": 50.6192,
+      "eval_samples_per_second": 40.044,
+      "eval_steps_per_second": 5.018,
+      "eval_wer": 0.3133237046398267,
+      "step": 114100
+    },
+    {
+      "epoch": 12.23,
+      "eval_loss": 0.4509586989879608,
+      "eval_runtime": 50.466,
+      "eval_samples_per_second": 40.166,
+      "eval_steps_per_second": 5.033,
+      "eval_wer": 0.3117891316122044,
+      "step": 114200
+    },
+    {
+      "epoch": 12.24,
+      "eval_loss": 0.44721925258636475,
+      "eval_runtime": 50.6245,
+      "eval_samples_per_second": 40.04,
+      "eval_steps_per_second": 5.017,
+      "eval_wer": 0.31305289763495214,
+      "step": 114300
+    },
+    {
+      "epoch": 12.25,
+      "eval_loss": 0.44987425208091736,
+      "eval_runtime": 51.5037,
+      "eval_samples_per_second": 39.356,
+      "eval_steps_per_second": 4.932,
+      "eval_wer": 0.3129626286333273,
+      "step": 114400
+    },
+    {
+      "epoch": 12.26,
+      "learning_rate": 5.511103151862464e-05,
+      "loss": 0.0918,
+      "step": 114500
+    },
+    {
+      "epoch": 12.26,
+      "eval_loss": 0.45026734471321106,
+      "eval_runtime": 49.4953,
+      "eval_samples_per_second": 40.953,
+      "eval_steps_per_second": 5.132,
+      "eval_wer": 0.3137750496479509,
+      "step": 114500
+    },
+    {
+      "epoch": 12.27,
+      "eval_loss": 0.45180875062942505,
+      "eval_runtime": 49.496,
+      "eval_samples_per_second": 40.953,
+      "eval_steps_per_second": 5.132,
+      "eval_wer": 0.31350424264307636,
+      "step": 114600
+    },
+    {
+      "epoch": 12.28,
+      "eval_loss": 0.44931671023368835,
+      "eval_runtime": 49.0213,
+      "eval_samples_per_second": 41.349,
+      "eval_steps_per_second": 5.181,
+      "eval_wer": 0.311428055605705,
+      "step": 114700
+    },
+    {
+      "epoch": 12.29,
+      "eval_loss": 0.45739245414733887,
+      "eval_runtime": 49.1885,
+      "eval_samples_per_second": 41.209,
+      "eval_steps_per_second": 5.164,
+      "eval_wer": 0.3133237046398267,
+      "step": 114800
+    },
+    {
+      "epoch": 12.3,
+      "eval_loss": 0.46827754378318787,
+      "eval_runtime": 49.2072,
+      "eval_samples_per_second": 41.193,
+      "eval_steps_per_second": 5.162,
+      "eval_wer": 0.320003610760065,
+      "step": 114900
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 5.40365329512894e-05,
+      "loss": 0.0869,
+      "step": 115000
+    },
+    {
+      "epoch": 12.31,
+      "eval_loss": 0.4608277678489685,
+      "eval_runtime": 49.5789,
+      "eval_samples_per_second": 40.884,
+      "eval_steps_per_second": 5.123,
+      "eval_wer": 0.3164831196966962,
+      "step": 115000
+    },
+    {
+      "epoch": 12.32,
+      "eval_loss": 0.46175676584243774,
+      "eval_runtime": 49.9224,
+      "eval_samples_per_second": 40.603,
+      "eval_steps_per_second": 5.088,
+      "eval_wer": 0.318288499729193,
+      "step": 115100
+    },
+    {
+      "epoch": 12.33,
+      "eval_loss": 0.46893182396888733,
+      "eval_runtime": 49.2463,
+      "eval_samples_per_second": 41.16,
+      "eval_steps_per_second": 5.158,
+      "eval_wer": 0.3172955407113197,
+      "step": 115200
+    },
+    {
+      "epoch": 12.34,
+      "eval_loss": 0.4680798351764679,
+      "eval_runtime": 48.8639,
+      "eval_samples_per_second": 41.483,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.32235060480231087,
+      "step": 115300
+    },
+    {
+      "epoch": 12.36,
+      "eval_loss": 0.4576069712638855,
+      "eval_runtime": 49.1095,
+      "eval_samples_per_second": 41.275,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.3230727568153096,
+      "step": 115400
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 5.2962034383954145e-05,
+      "loss": 0.0885,
+      "step": 115500
+    },
+    {
+      "epoch": 12.37,
+      "eval_loss": 0.48311322927474976,
+      "eval_runtime": 48.7739,
+      "eval_samples_per_second": 41.559,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.31756634771619424,
+      "step": 115500
+    },
+    {
+      "epoch": 12.38,
+      "eval_loss": 0.4601800739765167,
+      "eval_runtime": 49.0417,
+      "eval_samples_per_second": 41.332,
+      "eval_steps_per_second": 5.179,
+      "eval_wer": 0.3181079617259433,
+      "step": 115600
+    },
+    {
+      "epoch": 12.39,
+      "eval_loss": 0.4493272006511688,
+      "eval_runtime": 49.2267,
+      "eval_samples_per_second": 41.177,
+      "eval_steps_per_second": 5.16,
+      "eval_wer": 0.31675392670157065,
+      "step": 115700
+    },
+    {
+      "epoch": 12.4,
+      "eval_loss": 0.45637446641921997,
+      "eval_runtime": 48.8932,
+      "eval_samples_per_second": 41.458,
+      "eval_steps_per_second": 5.195,
+      "eval_wer": 0.314858277667449,
+      "step": 115800
+    },
+    {
+      "epoch": 12.41,
+      "eval_loss": 0.45846298336982727,
+      "eval_runtime": 48.5992,
+      "eval_samples_per_second": 41.708,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.31576096768369744,
+      "step": 115900
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 5.1887535816618904e-05,
+      "loss": 0.091,
+      "step": 116000
+    },
+    {
+      "epoch": 12.42,
+      "eval_loss": 0.4713011384010315,
+      "eval_runtime": 48.7915,
+      "eval_samples_per_second": 41.544,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.31928145874706626,
+      "step": 116000
+    },
+    {
+      "epoch": 12.43,
+      "eval_loss": 0.45806244015693665,
+      "eval_runtime": 48.5172,
+      "eval_samples_per_second": 41.779,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.31386531864957573,
+      "step": 116100
+    },
+    {
+      "epoch": 12.44,
+      "eval_loss": 0.4636545777320862,
+      "eval_runtime": 48.7872,
+      "eval_samples_per_second": 41.548,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.313143166636577,
+      "step": 116200
+    },
+    {
+      "epoch": 12.45,
+      "eval_loss": 0.4571922719478607,
+      "eval_runtime": 48.9824,
+      "eval_samples_per_second": 41.382,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.31242101462357824,
+      "step": 116300
+    },
+    {
+      "epoch": 12.46,
+      "eval_loss": 0.44893747568130493,
+      "eval_runtime": 51.4376,
+      "eval_samples_per_second": 39.407,
+      "eval_steps_per_second": 4.938,
+      "eval_wer": 0.3163025816934465,
+      "step": 116400
+    },
+    {
+      "epoch": 12.47,
+      "learning_rate": 5.0813037249283664e-05,
+      "loss": 0.0886,
+      "step": 116500
+    },
+    {
+      "epoch": 12.47,
+      "eval_loss": 0.46794840693473816,
+      "eval_runtime": 49.1991,
+      "eval_samples_per_second": 41.2,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3159415056869471,
+      "step": 116500
+    },
+    {
+      "epoch": 12.48,
+      "eval_loss": 0.4712357819080353,
+      "eval_runtime": 49.2178,
+      "eval_samples_per_second": 41.184,
+      "eval_steps_per_second": 5.161,
+      "eval_wer": 0.31512908467232353,
+      "step": 116600
+    },
+    {
+      "epoch": 12.49,
+      "eval_loss": 0.47498226165771484,
+      "eval_runtime": 48.8815,
+      "eval_samples_per_second": 41.468,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.31864957573569236,
+      "step": 116700
+    },
+    {
+      "epoch": 12.51,
+      "eval_loss": 0.4672822654247284,
+      "eval_runtime": 48.7696,
+      "eval_samples_per_second": 41.563,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.31756634771619424,
+      "step": 116800
+    },
+    {
+      "epoch": 12.52,
+      "eval_loss": 0.4601333439350128,
+      "eval_runtime": 48.6974,
+      "eval_samples_per_second": 41.624,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.3113377866040802,
+      "step": 116900
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 4.9738538681948424e-05,
+      "loss": 0.0917,
+      "step": 117000
+    },
+    {
+      "epoch": 12.53,
+      "eval_loss": 0.43411996960639954,
+      "eval_runtime": 49.4938,
+      "eval_samples_per_second": 40.955,
+      "eval_steps_per_second": 5.132,
+      "eval_wer": 0.3125112836252031,
+      "step": 117000
+    },
+    {
+      "epoch": 12.54,
+      "eval_loss": 0.4462231397628784,
+      "eval_runtime": 48.7809,
+      "eval_samples_per_second": 41.553,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.30772702653908646,
+      "step": 117100
+    },
+    {
+      "epoch": 12.55,
+      "eval_loss": 0.45020830631256104,
+      "eval_runtime": 48.9828,
+      "eval_samples_per_second": 41.382,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.3098934825780827,
+      "step": 117200
+    },
+    {
+      "epoch": 12.56,
+      "eval_loss": 0.4482104480266571,
+      "eval_runtime": 49.0098,
+      "eval_samples_per_second": 41.359,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.3116085936089547,
+      "step": 117300
+    },
+    {
+      "epoch": 12.57,
+      "eval_loss": 0.4459335505962372,
+      "eval_runtime": 48.885,
+      "eval_samples_per_second": 41.465,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.313143166636577,
+      "step": 117400
+    },
+    {
+      "epoch": 12.58,
+      "learning_rate": 4.8664040114613176e-05,
+      "loss": 0.0881,
+      "step": 117500
+    },
+    {
+      "epoch": 12.58,
+      "eval_loss": 0.4464477598667145,
+      "eval_runtime": 48.7534,
+      "eval_samples_per_second": 41.577,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.3121502076187037,
+      "step": 117500
+    },
+    {
+      "epoch": 12.59,
+      "eval_loss": 0.4471108019351959,
+      "eval_runtime": 49.1827,
+      "eval_samples_per_second": 41.214,
+      "eval_steps_per_second": 5.164,
+      "eval_wer": 0.3125112836252031,
+      "step": 117600
+    },
+    {
+      "epoch": 12.6,
+      "eval_loss": 0.4318666458129883,
+      "eval_runtime": 48.6999,
+      "eval_samples_per_second": 41.622,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.31224047662032856,
+      "step": 117700
+    },
+    {
+      "epoch": 12.61,
+      "eval_loss": 0.4420875310897827,
+      "eval_runtime": 48.5883,
+      "eval_samples_per_second": 41.718,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.31025455858458206,
+      "step": 117800
+    },
+    {
+      "epoch": 12.62,
+      "eval_loss": 0.43261706829071045,
+      "eval_runtime": 48.5361,
+      "eval_samples_per_second": 41.763,
+      "eval_steps_per_second": 5.233,
+      "eval_wer": 0.3107961725943311,
+      "step": 117900
+    },
+    {
+      "epoch": 12.63,
+      "learning_rate": 4.758954154727793e-05,
+      "loss": 0.0913,
+      "step": 118000
+    },
+    {
+      "epoch": 12.63,
+      "eval_loss": 0.44137805700302124,
+      "eval_runtime": 49.0062,
+      "eval_samples_per_second": 41.362,
+      "eval_steps_per_second": 5.183,
+      "eval_wer": 0.3068243365228381,
+      "step": 118000
+    },
+    {
+      "epoch": 12.64,
+      "eval_loss": 0.4421420097351074,
+      "eval_runtime": 48.7431,
+      "eval_samples_per_second": 41.585,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.3082686405488355,
+      "step": 118100
+    },
+    {
+      "epoch": 12.66,
+      "eval_loss": 0.44493910670280457,
+      "eval_runtime": 48.7918,
+      "eval_samples_per_second": 41.544,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.31025455858458206,
+      "step": 118200
+    },
+    {
+      "epoch": 12.67,
+      "eval_loss": 0.43802663683891296,
+      "eval_runtime": 48.9254,
+      "eval_samples_per_second": 41.43,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.3127820906300776,
+      "step": 118300
+    },
+    {
+      "epoch": 12.68,
+      "eval_loss": 0.43897026777267456,
+      "eval_runtime": 48.7195,
+      "eval_samples_per_second": 41.606,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.3135945116447012,
+      "step": 118400
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 4.651504297994269e-05,
+      "loss": 0.0921,
+      "step": 118500
+    },
+    {
+      "epoch": 12.69,
+      "eval_loss": 0.4452337324619293,
+      "eval_runtime": 48.675,
+      "eval_samples_per_second": 41.644,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.31043509658783175,
+      "step": 118500
+    },
+    {
+      "epoch": 12.7,
+      "eval_loss": 0.437765896320343,
+      "eval_runtime": 48.8617,
+      "eval_samples_per_second": 41.484,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.3121502076187037,
+      "step": 118600
+    },
+    {
+      "epoch": 12.71,
+      "eval_loss": 0.44589871168136597,
+      "eval_runtime": 49.4576,
+      "eval_samples_per_second": 40.985,
+      "eval_steps_per_second": 5.136,
+      "eval_wer": 0.307997833543961,
+      "step": 118700
+    },
+    {
+      "epoch": 12.72,
+      "eval_loss": 0.43694981932640076,
+      "eval_runtime": 48.5647,
+      "eval_samples_per_second": 41.738,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.30510922549196606,
+      "step": 118800
+    },
+    {
+      "epoch": 12.73,
+      "eval_loss": 0.4473933279514313,
+      "eval_runtime": 48.8399,
+      "eval_samples_per_second": 41.503,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.3076367575374616,
+      "step": 118900
+    },
+    {
+      "epoch": 12.74,
+      "learning_rate": 4.544054441260745e-05,
+      "loss": 0.0886,
+      "step": 119000
+    },
+    {
+      "epoch": 12.74,
+      "eval_loss": 0.45080724358558655,
+      "eval_runtime": 48.7926,
+      "eval_samples_per_second": 41.543,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.3066437985195884,
+      "step": 119000
+    },
+    {
+      "epoch": 12.75,
+      "eval_loss": 0.445557177066803,
+      "eval_runtime": 49.0576,
+      "eval_samples_per_second": 41.319,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.309712944574833,
+      "step": 119100
+    },
+    {
+      "epoch": 12.76,
+      "eval_loss": 0.4503107964992523,
+      "eval_runtime": 49.0549,
+      "eval_samples_per_second": 41.321,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.3078172955407113,
+      "step": 119200
+    },
+    {
+      "epoch": 12.77,
+      "eval_loss": 0.4459659457206726,
+      "eval_runtime": 49.8084,
+      "eval_samples_per_second": 40.696,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.30808810254558583,
+      "step": 119300
+    },
+    {
+      "epoch": 12.78,
+      "eval_loss": 0.4403787851333618,
+      "eval_runtime": 48.9807,
+      "eval_samples_per_second": 41.384,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.307997833543961,
+      "step": 119400
+    },
+    {
+      "epoch": 12.79,
+      "learning_rate": 4.43660458452722e-05,
+      "loss": 0.0897,
+      "step": 119500
+    },
+    {
+      "epoch": 12.79,
+      "eval_loss": 0.43507125973701477,
+      "eval_runtime": 49.5648,
+      "eval_samples_per_second": 40.896,
+      "eval_steps_per_second": 5.125,
+      "eval_wer": 0.30998375157970753,
+      "step": 119500
+    },
+    {
+      "epoch": 12.81,
+      "eval_loss": 0.44462278485298157,
+      "eval_runtime": 48.5763,
+      "eval_samples_per_second": 41.728,
+      "eval_steps_per_second": 5.229,
+      "eval_wer": 0.311969669615454,
+      "step": 119600
+    },
+    {
+      "epoch": 12.82,
+      "eval_loss": 0.44070571660995483,
+      "eval_runtime": 49.2374,
+      "eval_samples_per_second": 41.168,
+      "eval_steps_per_second": 5.159,
+      "eval_wer": 0.30980321357645785,
+      "step": 119700
+    },
+    {
+      "epoch": 12.83,
+      "eval_loss": 0.4405609667301178,
+      "eval_runtime": 48.5684,
+      "eval_samples_per_second": 41.735,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.30835890955046036,
+      "step": 119800
+    },
+    {
+      "epoch": 12.84,
+      "eval_loss": 0.4491572678089142,
+      "eval_runtime": 49.0594,
+      "eval_samples_per_second": 41.317,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.30673406752121324,
+      "step": 119900
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 4.329369627507163e-05,
+      "loss": 0.09,
+      "step": 120000
+    },
+    {
+      "epoch": 12.85,
+      "eval_loss": 0.45461514592170715,
+      "eval_runtime": 49.0505,
+      "eval_samples_per_second": 41.325,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.30980321357645785,
+      "step": 120000
+    },
+    {
+      "epoch": 12.86,
+      "eval_loss": 0.45472848415374756,
+      "eval_runtime": 49.0834,
+      "eval_samples_per_second": 41.297,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3073659505325871,
+      "step": 120100
+    },
+    {
+      "epoch": 12.87,
+      "eval_loss": 0.45165643095970154,
+      "eval_runtime": 49.1082,
+      "eval_samples_per_second": 41.276,
+      "eval_steps_per_second": 5.172,
+      "eval_wer": 0.31106697959920565,
+      "step": 120200
+    },
+    {
+      "epoch": 12.88,
+      "eval_loss": 0.43198153376579285,
+      "eval_runtime": 49.0197,
+      "eval_samples_per_second": 41.351,
+      "eval_steps_per_second": 5.182,
+      "eval_wer": 0.30637299151471387,
+      "step": 120300
+    },
+    {
+      "epoch": 12.89,
+      "eval_loss": 0.4293752610683441,
+      "eval_runtime": 49.0833,
+      "eval_samples_per_second": 41.297,
+      "eval_steps_per_second": 5.175,
+      "eval_wer": 0.3071854125293374,
+      "step": 120400
+    },
+    {
+      "epoch": 12.9,
+      "learning_rate": 4.221919770773638e-05,
+      "loss": 0.0898,
+      "step": 120500
+    },
+    {
+      "epoch": 12.9,
+      "eval_loss": 0.44117698073387146,
+      "eval_runtime": 49.2326,
+      "eval_samples_per_second": 41.172,
+      "eval_steps_per_second": 5.159,
+      "eval_wer": 0.3050189564903412,
+      "step": 120500
+    },
+    {
+      "epoch": 12.91,
+      "eval_loss": 0.4254050552845001,
+      "eval_runtime": 49.3023,
+      "eval_samples_per_second": 41.114,
+      "eval_steps_per_second": 5.152,
+      "eval_wer": 0.3073659505325871,
+      "step": 120600
+    },
+    {
+      "epoch": 12.92,
+      "eval_loss": 0.4409028887748718,
+      "eval_runtime": 48.4446,
+      "eval_samples_per_second": 41.842,
+      "eval_steps_per_second": 5.243,
+      "eval_wer": 0.3070951435277126,
+      "step": 120700
+    },
+    {
+      "epoch": 12.93,
+      "eval_loss": 0.4362035095691681,
+      "eval_runtime": 49.258,
+      "eval_samples_per_second": 41.151,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.3070951435277126,
+      "step": 120800
+    },
+    {
+      "epoch": 12.94,
+      "eval_loss": 0.45788970589637756,
+      "eval_runtime": 49.1783,
+      "eval_samples_per_second": 41.217,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.30899079256183426,
+      "step": 120900
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 4.114469914040114e-05,
+      "loss": 0.0892,
+      "step": 121000
+    },
+    {
+      "epoch": 12.95,
+      "eval_loss": 0.44916099309921265,
+      "eval_runtime": 49.252,
+      "eval_samples_per_second": 41.156,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.30592164650658965,
+      "step": 121000
+    },
+    {
+      "epoch": 12.97,
+      "eval_loss": 0.44036194682121277,
+      "eval_runtime": 48.5298,
+      "eval_samples_per_second": 41.768,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.3105253655894566,
+      "step": 121100
+    },
+    {
+      "epoch": 12.98,
+      "eval_loss": 0.43648219108581543,
+      "eval_runtime": 48.6252,
+      "eval_samples_per_second": 41.686,
+      "eval_steps_per_second": 5.224,
+      "eval_wer": 0.3066437985195884,
+      "step": 121200
+    },
+    {
+      "epoch": 12.99,
+      "eval_loss": 0.43682196736335754,
+      "eval_runtime": 49.1606,
+      "eval_samples_per_second": 41.232,
+      "eval_steps_per_second": 5.167,
+      "eval_wer": 0.30483841848709153,
+      "step": 121300
+    },
+    {
+      "epoch": 13.0,
+      "eval_loss": 0.44096165895462036,
+      "eval_runtime": 49.9358,
+      "eval_samples_per_second": 40.592,
+      "eval_steps_per_second": 5.087,
+      "eval_wer": 0.3033038454594692,
+      "step": 121400
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 4.00702005730659e-05,
+      "loss": 0.085,
+      "step": 121500
+    },
+    {
+      "epoch": 13.01,
+      "eval_loss": 0.44503235816955566,
+      "eval_runtime": 48.8269,
+      "eval_samples_per_second": 41.514,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.30465788048384185,
+      "step": 121500
+    },
+    {
+      "epoch": 13.02,
+      "eval_loss": 0.4633066952228546,
+      "eval_runtime": 49.1922,
+      "eval_samples_per_second": 41.206,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.3013179274237227,
+      "step": 121600
+    },
+    {
+      "epoch": 13.03,
+      "eval_loss": 0.46001821756362915,
+      "eval_runtime": 49.1699,
+      "eval_samples_per_second": 41.224,
+      "eval_steps_per_second": 5.166,
+      "eval_wer": 0.3053800324968406,
+      "step": 121700
+    },
+    {
+      "epoch": 13.04,
+      "eval_loss": 0.4540691375732422,
+      "eval_runtime": 49.0697,
+      "eval_samples_per_second": 41.309,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.3047481494854667,
+      "step": 121800
+    },
+    {
+      "epoch": 13.05,
+      "eval_loss": 0.45461952686309814,
+      "eval_runtime": 49.096,
+      "eval_samples_per_second": 41.286,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.3058313775049648,
+      "step": 121900
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 3.8995702005730654e-05,
+      "loss": 0.0791,
+      "step": 122000
+    },
+    {
+      "epoch": 13.06,
+      "eval_loss": 0.4536155164241791,
+      "eval_runtime": 48.7723,
+      "eval_samples_per_second": 41.56,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.30447734248059216,
+      "step": 122000
+    },
+    {
+      "epoch": 13.07,
+      "eval_loss": 0.4589266777038574,
+      "eval_runtime": 49.0728,
+      "eval_samples_per_second": 41.306,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.30655352951796355,
+      "step": 122100
+    },
+    {
+      "epoch": 13.08,
+      "eval_loss": 0.4580624997615814,
+      "eval_runtime": 49.0335,
+      "eval_samples_per_second": 41.339,
+      "eval_steps_per_second": 5.18,
+      "eval_wer": 0.30574110850333996,
+      "step": 122200
+    },
+    {
+      "epoch": 13.09,
+      "eval_loss": 0.45460861921310425,
+      "eval_runtime": 48.9092,
+      "eval_samples_per_second": 41.444,
+      "eval_steps_per_second": 5.193,
+      "eval_wer": 0.30483841848709153,
+      "step": 122300
+    },
+    {
+      "epoch": 13.1,
+      "eval_loss": 0.46732285618782043,
+      "eval_runtime": 48.8035,
+      "eval_samples_per_second": 41.534,
+      "eval_steps_per_second": 5.205,
+      "eval_wer": 0.30059577541072396,
+      "step": 122400
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 3.792335243553008e-05,
+      "loss": 0.0789,
+      "step": 122500
+    },
+    {
+      "epoch": 13.12,
+      "eval_loss": 0.45507410168647766,
+      "eval_runtime": 48.977,
+      "eval_samples_per_second": 41.387,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.30185954143347177,
+      "step": 122500
+    },
+    {
+      "epoch": 13.13,
+      "eval_loss": 0.4466601014137268,
+      "eval_runtime": 48.675,
+      "eval_samples_per_second": 41.644,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.3024914244448456,
+      "step": 122600
+    },
+    {
+      "epoch": 13.14,
+      "eval_loss": 0.4593279957771301,
+      "eval_runtime": 49.0944,
+      "eval_samples_per_second": 41.288,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.3014984654269724,
+      "step": 122700
+    },
+    {
+      "epoch": 13.15,
+      "eval_loss": 0.4597959518432617,
+      "eval_runtime": 48.7201,
+      "eval_samples_per_second": 41.605,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.3036649214659686,
+      "step": 122800
+    },
+    {
+      "epoch": 13.16,
+      "eval_loss": 0.4531610608100891,
+      "eval_runtime": 49.6573,
+      "eval_samples_per_second": 40.82,
+      "eval_steps_per_second": 5.115,
+      "eval_wer": 0.3037551904675934,
+      "step": 122900
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 3.685100286532951e-05,
+      "loss": 0.077,
+      "step": 123000
+    },
+    {
+      "epoch": 13.17,
+      "eval_loss": 0.46069207787513733,
+      "eval_runtime": 49.0496,
+      "eval_samples_per_second": 41.325,
+      "eval_steps_per_second": 5.178,
+      "eval_wer": 0.3014984654269724,
+      "step": 123000
+    },
+    {
+      "epoch": 13.18,
+      "eval_loss": 0.43852800130844116,
+      "eval_runtime": 49.9159,
+      "eval_samples_per_second": 40.608,
+      "eval_steps_per_second": 5.089,
+      "eval_wer": 0.3005055064090991,
+      "step": 123100
+    },
+    {
+      "epoch": 13.19,
+      "eval_loss": 0.4589592218399048,
+      "eval_runtime": 49.0752,
+      "eval_samples_per_second": 41.304,
+      "eval_steps_per_second": 5.176,
+      "eval_wer": 0.3041162664740928,
+      "step": 123200
+    },
+    {
+      "epoch": 13.2,
+      "eval_loss": 0.4358598589897156,
+      "eval_runtime": 48.8305,
+      "eval_samples_per_second": 41.511,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.30465788048384185,
+      "step": 123300
+    },
+    {
+      "epoch": 13.21,
+      "eval_loss": 0.4457867443561554,
+      "eval_runtime": 48.6567,
+      "eval_samples_per_second": 41.659,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.3039357284708431,
+      "step": 123400
+    },
+    {
+      "epoch": 13.22,
+      "learning_rate": 3.5776504297994264e-05,
+      "loss": 0.0771,
+      "step": 123500
+    },
+    {
+      "epoch": 13.22,
+      "eval_loss": 0.4506053924560547,
+      "eval_runtime": 48.6615,
+      "eval_samples_per_second": 41.655,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.30754648853583677,
+      "step": 123500
+    },
+    {
+      "epoch": 13.23,
+      "eval_loss": 0.4456702768802643,
+      "eval_runtime": 48.5857,
+      "eval_samples_per_second": 41.72,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.30790756454233614,
+      "step": 123600
+    },
+    {
+      "epoch": 13.24,
+      "eval_loss": 0.4448252022266388,
+      "eval_runtime": 48.8145,
+      "eval_samples_per_second": 41.525,
+      "eval_steps_per_second": 5.203,
+      "eval_wer": 0.30483841848709153,
+      "step": 123700
+    },
+    {
+      "epoch": 13.25,
+      "eval_loss": 0.4397769272327423,
+      "eval_runtime": 48.444,
+      "eval_samples_per_second": 41.842,
+      "eval_steps_per_second": 5.243,
+      "eval_wer": 0.30357465246434373,
+      "step": 123800
+    },
+    {
+      "epoch": 13.27,
+      "eval_loss": 0.451018363237381,
+      "eval_runtime": 48.6358,
+      "eval_samples_per_second": 41.677,
+      "eval_steps_per_second": 5.222,
+      "eval_wer": 0.30547030149846544,
+      "step": 123900
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 3.4702005730659024e-05,
+      "loss": 0.0804,
+      "step": 124000
+    },
+    {
+      "epoch": 13.28,
+      "eval_loss": 0.4506898522377014,
+      "eval_runtime": 48.6155,
+      "eval_samples_per_second": 41.695,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.30592164650658965,
+      "step": 124000
+    },
+    {
+      "epoch": 13.29,
+      "eval_loss": 0.4543741047382355,
+      "eval_runtime": 49.2524,
+      "eval_samples_per_second": 41.155,
+      "eval_steps_per_second": 5.157,
+      "eval_wer": 0.3076367575374616,
+      "step": 124100
+    },
+    {
+      "epoch": 13.3,
+      "eval_loss": 0.45340901613235474,
+      "eval_runtime": 48.7151,
+      "eval_samples_per_second": 41.609,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.30727568153096224,
+      "step": 124200
+    },
+    {
+      "epoch": 13.31,
+      "eval_loss": 0.4441240131855011,
+      "eval_runtime": 48.5073,
+      "eval_samples_per_second": 41.788,
+      "eval_steps_per_second": 5.236,
+      "eval_wer": 0.30610218450983934,
+      "step": 124300
+    },
+    {
+      "epoch": 13.32,
+      "eval_loss": 0.43914496898651123,
+      "eval_runtime": 48.7563,
+      "eval_samples_per_second": 41.574,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.30754648853583677,
+      "step": 124400
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 3.362750716332378e-05,
+      "loss": 0.0774,
+      "step": 124500
+    },
+    {
+      "epoch": 13.33,
+      "eval_loss": 0.45265689492225647,
+      "eval_runtime": 49.1972,
+      "eval_samples_per_second": 41.202,
+      "eval_steps_per_second": 5.163,
+      "eval_wer": 0.306282722513089,
+      "step": 124500
+    },
+    {
+      "epoch": 13.34,
+      "eval_loss": 0.46383553743362427,
+      "eval_runtime": 48.6998,
+      "eval_samples_per_second": 41.622,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.30574110850333996,
+      "step": 124600
+    },
+    {
+      "epoch": 13.35,
+      "eval_loss": 0.45413774251937866,
+      "eval_runtime": 48.7189,
+      "eval_samples_per_second": 41.606,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.30637299151471387,
+      "step": 124700
+    },
+    {
+      "epoch": 13.36,
+      "eval_loss": 0.46165329217910767,
+      "eval_runtime": 48.6222,
+      "eval_samples_per_second": 41.689,
+      "eval_steps_per_second": 5.224,
+      "eval_wer": 0.3078172955407113,
+      "step": 124800
+    },
+    {
+      "epoch": 13.37,
+      "eval_loss": 0.4584297239780426,
+      "eval_runtime": 48.9274,
+      "eval_samples_per_second": 41.429,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.3041162664740928,
+      "step": 124900
+    },
+    {
+      "epoch": 13.38,
+      "learning_rate": 3.2555157593123205e-05,
+      "loss": 0.0795,
+      "step": 125000
+    },
+    {
+      "epoch": 13.38,
+      "eval_loss": 0.4663410782814026,
+      "eval_runtime": 48.6633,
+      "eval_samples_per_second": 41.654,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.30321357645784436,
+      "step": 125000
+    },
+    {
+      "epoch": 13.39,
+      "eval_loss": 0.4545777440071106,
+      "eval_runtime": 49.7034,
+      "eval_samples_per_second": 40.782,
+      "eval_steps_per_second": 5.11,
+      "eval_wer": 0.3024914244448456,
+      "step": 125100
+    },
+    {
+      "epoch": 13.4,
+      "eval_loss": 0.4616332948207855,
+      "eval_runtime": 48.8306,
+      "eval_samples_per_second": 41.511,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.3021303484383463,
+      "step": 125200
+    },
+    {
+      "epoch": 13.42,
+      "eval_loss": 0.4602724611759186,
+      "eval_runtime": 48.7695,
+      "eval_samples_per_second": 41.563,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.30158873442859724,
+      "step": 125300
+    },
+    {
+      "epoch": 13.43,
+      "eval_loss": 0.4615834057331085,
+      "eval_runtime": 48.6683,
+      "eval_samples_per_second": 41.649,
+      "eval_steps_per_second": 5.219,
+      "eval_wer": 0.30402599747246795,
+      "step": 125400
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 3.1480659025787965e-05,
+      "loss": 0.0791,
+      "step": 125500
+    },
+    {
+      "epoch": 13.44,
+      "eval_loss": 0.45479077100753784,
+      "eval_runtime": 48.3791,
+      "eval_samples_per_second": 41.898,
+      "eval_steps_per_second": 5.25,
+      "eval_wer": 0.3021303484383463,
+      "step": 125500
+    },
+    {
+      "epoch": 13.45,
+      "eval_loss": 0.4559873640537262,
+      "eval_runtime": 48.8585,
+      "eval_samples_per_second": 41.487,
+      "eval_steps_per_second": 5.199,
+      "eval_wer": 0.3024914244448456,
+      "step": 125600
+    },
+    {
+      "epoch": 13.46,
+      "eval_loss": 0.45155441761016846,
+      "eval_runtime": 48.7439,
+      "eval_samples_per_second": 41.585,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.3036649214659686,
+      "step": 125700
+    },
+    {
+      "epoch": 13.47,
+      "eval_loss": 0.45003730058670044,
+      "eval_runtime": 48.8595,
+      "eval_samples_per_second": 41.486,
+      "eval_steps_per_second": 5.199,
+      "eval_wer": 0.3013179274237227,
+      "step": 125800
+    },
+    {
+      "epoch": 13.48,
+      "eval_loss": 0.4539685547351837,
+      "eval_runtime": 49.61,
+      "eval_samples_per_second": 40.859,
+      "eval_steps_per_second": 5.12,
+      "eval_wer": 0.3008665824155985,
+      "step": 125900
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 3.0406160458452718e-05,
+      "loss": 0.0776,
+      "step": 126000
+    },
+    {
+      "epoch": 13.49,
+      "eval_loss": 0.4581133723258972,
+      "eval_runtime": 48.613,
+      "eval_samples_per_second": 41.697,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.30258169344647046,
+      "step": 126000
+    },
+    {
+      "epoch": 13.5,
+      "eval_loss": 0.45983752608299255,
+      "eval_runtime": 48.7145,
+      "eval_samples_per_second": 41.61,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.30276223144972014,
+      "step": 126100
+    },
+    {
+      "epoch": 13.51,
+      "eval_loss": 0.45870548486709595,
+      "eval_runtime": 48.7095,
+      "eval_samples_per_second": 41.614,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.30384545946921826,
+      "step": 126200
+    },
+    {
+      "epoch": 13.52,
+      "eval_loss": 0.45144903659820557,
+      "eval_runtime": 48.8311,
+      "eval_samples_per_second": 41.51,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.30240115544322077,
+      "step": 126300
+    },
+    {
+      "epoch": 13.53,
+      "eval_loss": 0.4495235085487366,
+      "eval_runtime": 48.8596,
+      "eval_samples_per_second": 41.486,
+      "eval_steps_per_second": 5.199,
+      "eval_wer": 0.30357465246434373,
+      "step": 126400
+    },
+    {
+      "epoch": 13.54,
+      "learning_rate": 2.9331661891117477e-05,
+      "loss": 0.0793,
+      "step": 126500
+    },
+    {
+      "epoch": 13.54,
+      "eval_loss": 0.45561927556991577,
+      "eval_runtime": 49.1727,
+      "eval_samples_per_second": 41.222,
+      "eval_steps_per_second": 5.165,
+      "eval_wer": 0.30158873442859724,
+      "step": 126500
+    },
+    {
+      "epoch": 13.55,
+      "eval_loss": 0.46026507019996643,
+      "eval_runtime": 48.7412,
+      "eval_samples_per_second": 41.587,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.3024914244448456,
+      "step": 126600
+    },
+    {
+      "epoch": 13.57,
+      "eval_loss": 0.4496086835861206,
+      "eval_runtime": 48.782,
+      "eval_samples_per_second": 41.552,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.29951254739122585,
+      "step": 126700
+    },
+    {
+      "epoch": 13.58,
+      "eval_loss": 0.44833648204803467,
+      "eval_runtime": 48.9526,
+      "eval_samples_per_second": 41.407,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.29689474634410545,
+      "step": 126800
+    },
+    {
+      "epoch": 13.59,
+      "eval_loss": 0.4462050795555115,
+      "eval_runtime": 48.8347,
+      "eval_samples_per_second": 41.507,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.2979779743636035,
+      "step": 126900
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 2.8257163323782234e-05,
+      "loss": 0.0816,
+      "step": 127000
+    },
+    {
+      "epoch": 13.6,
+      "eval_loss": 0.4521482586860657,
+      "eval_runtime": 48.794,
+      "eval_samples_per_second": 41.542,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.2981585123668532,
+      "step": 127000
+    },
+    {
+      "epoch": 13.61,
+      "eval_loss": 0.4580208659172058,
+      "eval_runtime": 48.7962,
+      "eval_samples_per_second": 41.54,
+      "eval_steps_per_second": 5.205,
+      "eval_wer": 0.30185954143347177,
+      "step": 127100
+    },
+    {
+      "epoch": 13.62,
+      "eval_loss": 0.46691346168518066,
+      "eval_runtime": 48.9324,
+      "eval_samples_per_second": 41.424,
+      "eval_steps_per_second": 5.191,
+      "eval_wer": 0.3008665824155985,
+      "step": 127200
+    },
+    {
+      "epoch": 13.63,
+      "eval_loss": 0.4513174295425415,
+      "eval_runtime": 48.9146,
+      "eval_samples_per_second": 41.44,
+      "eval_steps_per_second": 5.193,
+      "eval_wer": 0.3016790034302221,
+      "step": 127300
+    },
+    {
+      "epoch": 13.64,
+      "eval_loss": 0.46024101972579956,
+      "eval_runtime": 48.6915,
+      "eval_samples_per_second": 41.629,
+      "eval_steps_per_second": 5.217,
+      "eval_wer": 0.3014984654269724,
+      "step": 127400
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 2.718266475644699e-05,
+      "loss": 0.0779,
+      "step": 127500
+    },
+    {
+      "epoch": 13.65,
+      "eval_loss": 0.45917895436286926,
+      "eval_runtime": 48.782,
+      "eval_samples_per_second": 41.552,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.2997833543961004,
+      "step": 127500
+    },
+    {
+      "epoch": 13.66,
+      "eval_loss": 0.47004109621047974,
+      "eval_runtime": 48.902,
+      "eval_samples_per_second": 41.45,
+      "eval_steps_per_second": 5.194,
+      "eval_wer": 0.29806824336522836,
+      "step": 127600
+    },
+    {
+      "epoch": 13.67,
+      "eval_loss": 0.4726985692977905,
+      "eval_runtime": 48.8717,
+      "eval_samples_per_second": 41.476,
+      "eval_steps_per_second": 5.197,
+      "eval_wer": 0.29779743636035383,
+      "step": 127700
+    },
+    {
+      "epoch": 13.68,
+      "eval_loss": 0.45995599031448364,
+      "eval_runtime": 49.2728,
+      "eval_samples_per_second": 41.138,
+      "eval_steps_per_second": 5.155,
+      "eval_wer": 0.2983390503701029,
+      "step": 127800
+    },
+    {
+      "epoch": 13.69,
+      "eval_loss": 0.44724515080451965,
+      "eval_runtime": 48.727,
+      "eval_samples_per_second": 41.599,
+      "eval_steps_per_second": 5.213,
+      "eval_wer": 0.29779743636035383,
+      "step": 127900
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 2.6108166189111746e-05,
+      "loss": 0.0779,
+      "step": 128000
+    },
+    {
+      "epoch": 13.7,
+      "eval_loss": 0.4483438730239868,
+      "eval_runtime": 48.7121,
+      "eval_samples_per_second": 41.612,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.29842931937172773,
+      "step": 128000
+    },
+    {
+      "epoch": 13.72,
+      "eval_loss": 0.45116451382637024,
+      "eval_runtime": 48.9634,
+      "eval_samples_per_second": 41.398,
+      "eval_steps_per_second": 5.188,
+      "eval_wer": 0.2968044773424806,
+      "step": 128100
+    },
+    {
+      "epoch": 13.73,
+      "eval_loss": 0.4549117386341095,
+      "eval_runtime": 48.796,
+      "eval_samples_per_second": 41.54,
+      "eval_steps_per_second": 5.205,
+      "eval_wer": 0.2987903953782271,
+      "step": 128200
+    },
+    {
+      "epoch": 13.74,
+      "eval_loss": 0.45755627751350403,
+      "eval_runtime": 49.8079,
+      "eval_samples_per_second": 40.696,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.2992417403863513,
+      "step": 128300
+    },
+    {
+      "epoch": 13.75,
+      "eval_loss": 0.4400174915790558,
+      "eval_runtime": 48.596,
+      "eval_samples_per_second": 41.711,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.2974363603538545,
+      "step": 128400
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 2.50336676217765e-05,
+      "loss": 0.0793,
+      "step": 128500
+    },
+    {
+      "epoch": 13.76,
+      "eval_loss": 0.4433426260948181,
+      "eval_runtime": 48.6648,
+      "eval_samples_per_second": 41.652,
+      "eval_steps_per_second": 5.219,
+      "eval_wer": 0.3008665824155985,
+      "step": 128500
+    },
+    {
+      "epoch": 13.77,
+      "eval_loss": 0.44558921456336975,
+      "eval_runtime": 48.7178,
+      "eval_samples_per_second": 41.607,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.29824878136847804,
+      "step": 128600
+    },
+    {
+      "epoch": 13.78,
+      "eval_loss": 0.4559650719165802,
+      "eval_runtime": 48.5852,
+      "eval_samples_per_second": 41.72,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.3019498104350966,
+      "step": 128700
+    },
+    {
+      "epoch": 13.79,
+      "eval_loss": 0.45514196157455444,
+      "eval_runtime": 48.8865,
+      "eval_samples_per_second": 41.463,
+      "eval_steps_per_second": 5.196,
+      "eval_wer": 0.30077631341397365,
+      "step": 128800
+    },
+    {
+      "epoch": 13.8,
+      "eval_loss": 0.45125389099121094,
+      "eval_runtime": 48.9777,
+      "eval_samples_per_second": 41.386,
+      "eval_steps_per_second": 5.186,
+      "eval_wer": 0.3006860444123488,
+      "step": 128900
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 2.3959169054441258e-05,
+      "loss": 0.0769,
+      "step": 129000
+    },
+    {
+      "epoch": 13.81,
+      "eval_loss": 0.45180532336235046,
+      "eval_runtime": 48.7049,
+      "eval_samples_per_second": 41.618,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.30077631341397365,
+      "step": 129000
+    },
+    {
+      "epoch": 13.82,
+      "eval_loss": 0.45669108629226685,
+      "eval_runtime": 48.6617,
+      "eval_samples_per_second": 41.655,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.29806824336522836,
+      "step": 129100
+    },
+    {
+      "epoch": 13.83,
+      "eval_loss": 0.4437228739261627,
+      "eval_runtime": 48.6473,
+      "eval_samples_per_second": 41.667,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.2985195883733526,
+      "step": 129200
+    },
+    {
+      "epoch": 13.84,
+      "eval_loss": 0.4423811733722687,
+      "eval_runtime": 49.8087,
+      "eval_samples_per_second": 40.696,
+      "eval_steps_per_second": 5.1,
+      "eval_wer": 0.2969850153457303,
+      "step": 129300
+    },
+    {
+      "epoch": 13.85,
+      "eval_loss": 0.44225621223449707,
+      "eval_runtime": 48.5133,
+      "eval_samples_per_second": 41.782,
+      "eval_steps_per_second": 5.236,
+      "eval_wer": 0.30095685141722334,
+      "step": 129400
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 2.2884670487106014e-05,
+      "loss": 0.0785,
+      "step": 129500
+    },
+    {
+      "epoch": 13.87,
+      "eval_loss": 0.44948408007621765,
+      "eval_runtime": 48.6646,
+      "eval_samples_per_second": 41.652,
+      "eval_steps_per_second": 5.219,
+      "eval_wer": 0.2998736233977252,
+      "step": 129500
+    },
+    {
+      "epoch": 13.88,
+      "eval_loss": 0.4482867419719696,
+      "eval_runtime": 48.676,
+      "eval_samples_per_second": 41.643,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.29752662935547936,
+      "step": 129600
+    },
+    {
+      "epoch": 13.89,
+      "eval_loss": 0.44850584864616394,
+      "eval_runtime": 48.5728,
+      "eval_samples_per_second": 41.731,
+      "eval_steps_per_second": 5.229,
+      "eval_wer": 0.2981585123668532,
+      "step": 129700
+    },
+    {
+      "epoch": 13.9,
+      "eval_loss": 0.4428721070289612,
+      "eval_runtime": 48.6015,
+      "eval_samples_per_second": 41.707,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.29716555334898,
+      "step": 129800
+    },
+    {
+      "epoch": 13.91,
+      "eval_loss": 0.4429783523082733,
+      "eval_runtime": 48.8683,
+      "eval_samples_per_second": 41.479,
+      "eval_steps_per_second": 5.198,
+      "eval_wer": 0.29581151832460734,
+      "step": 129900
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 2.181017191977077e-05,
+      "loss": 0.0792,
+      "step": 130000
+    },
+    {
+      "epoch": 13.92,
+      "eval_loss": 0.44945263862609863,
+      "eval_runtime": 48.6546,
+      "eval_samples_per_second": 41.661,
+      "eval_steps_per_second": 5.22,
+      "eval_wer": 0.2953601733164831,
+      "step": 130000
+    },
+    {
+      "epoch": 13.93,
+      "eval_loss": 0.4485260546207428,
+      "eval_runtime": 48.9845,
+      "eval_samples_per_second": 41.38,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.2947282903051092,
+      "step": 130100
+    },
+    {
+      "epoch": 13.94,
+      "eval_loss": 0.4395206570625305,
+      "eval_runtime": 49.0973,
+      "eval_samples_per_second": 41.285,
+      "eval_steps_per_second": 5.173,
+      "eval_wer": 0.29716555334898,
+      "step": 130200
+    },
+    {
+      "epoch": 13.95,
+      "eval_loss": 0.43786346912384033,
+      "eval_runtime": 48.7456,
+      "eval_samples_per_second": 41.583,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.29734609135222967,
+      "step": 130300
+    },
+    {
+      "epoch": 13.96,
+      "eval_loss": 0.44279754161834717,
+      "eval_runtime": 48.6537,
+      "eval_samples_per_second": 41.662,
+      "eval_steps_per_second": 5.221,
+      "eval_wer": 0.29888066437985195,
+      "step": 130400
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 2.0735673352435527e-05,
+      "loss": 0.0795,
+      "step": 130500
+    },
+    {
+      "epoch": 13.97,
+      "eval_loss": 0.43852099776268005,
+      "eval_runtime": 48.7541,
+      "eval_samples_per_second": 41.576,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.29996389239935006,
+      "step": 130500
+    },
+    {
+      "epoch": 13.98,
+      "eval_loss": 0.4489758014678955,
+      "eval_runtime": 48.8778,
+      "eval_samples_per_second": 41.471,
+      "eval_steps_per_second": 5.197,
+      "eval_wer": 0.2983390503701029,
+      "step": 130600
+    },
+    {
+      "epoch": 13.99,
+      "eval_loss": 0.45682230591773987,
+      "eval_runtime": 48.7069,
+      "eval_samples_per_second": 41.616,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.2969850153457303,
+      "step": 130700
+    },
+    {
+      "epoch": 14.0,
+      "eval_loss": 0.4481572210788727,
+      "eval_runtime": 48.9961,
+      "eval_samples_per_second": 41.371,
+      "eval_steps_per_second": 5.184,
+      "eval_wer": 0.29626286333273155,
+      "step": 130800
+    },
+    {
+      "epoch": 14.01,
+      "eval_loss": 0.44789794087409973,
+      "eval_runtime": 48.6027,
+      "eval_samples_per_second": 41.705,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.2961725943311067,
+      "step": 130900
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 1.9663323782234952e-05,
+      "loss": 0.075,
+      "step": 131000
+    },
+    {
+      "epoch": 14.03,
+      "eval_loss": 0.4565275013446808,
+      "eval_runtime": 48.781,
+      "eval_samples_per_second": 41.553,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.2968044773424806,
+      "step": 131000
+    },
+    {
+      "epoch": 14.04,
+      "eval_loss": 0.46226829290390015,
+      "eval_runtime": 48.8274,
+      "eval_samples_per_second": 41.514,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.2961725943311067,
+      "step": 131100
+    },
+    {
+      "epoch": 14.05,
+      "eval_loss": 0.4617317020893097,
+      "eval_runtime": 48.5503,
+      "eval_samples_per_second": 41.751,
+      "eval_steps_per_second": 5.232,
+      "eval_wer": 0.2965336703376061,
+      "step": 131200
+    },
+    {
+      "epoch": 14.06,
+      "eval_loss": 0.46868252754211426,
+      "eval_runtime": 49.7035,
+      "eval_samples_per_second": 40.782,
+      "eval_steps_per_second": 5.11,
+      "eval_wer": 0.2949088283083589,
+      "step": 131300
+    },
+    {
+      "epoch": 14.07,
+      "eval_loss": 0.47178399562835693,
+      "eval_runtime": 48.788,
+      "eval_samples_per_second": 41.547,
+      "eval_steps_per_second": 5.206,
+      "eval_wer": 0.29292291027261236,
+      "step": 131400
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 1.8590974212034384e-05,
+      "loss": 0.0709,
+      "step": 131500
+    },
+    {
+      "epoch": 14.08,
+      "eval_loss": 0.4720158278942108,
+      "eval_runtime": 48.3508,
+      "eval_samples_per_second": 41.923,
+      "eval_steps_per_second": 5.253,
+      "eval_wer": 0.2944574833002347,
+      "step": 131500
+    },
+    {
+      "epoch": 14.09,
+      "eval_loss": 0.46036496758461,
+      "eval_runtime": 48.6823,
+      "eval_samples_per_second": 41.637,
+      "eval_steps_per_second": 5.218,
+      "eval_wer": 0.2952699043148583,
+      "step": 131600
+    },
+    {
+      "epoch": 14.1,
+      "eval_loss": 0.46548134088516235,
+      "eval_runtime": 48.9917,
+      "eval_samples_per_second": 41.374,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.29545044231810796,
+      "step": 131700
+    },
+    {
+      "epoch": 14.11,
+      "eval_loss": 0.4694814682006836,
+      "eval_runtime": 48.632,
+      "eval_samples_per_second": 41.68,
+      "eval_steps_per_second": 5.223,
+      "eval_wer": 0.29581151832460734,
+      "step": 131800
+    },
+    {
+      "epoch": 14.12,
+      "eval_loss": 0.4666370153427124,
+      "eval_runtime": 48.4823,
+      "eval_samples_per_second": 41.809,
+      "eval_steps_per_second": 5.239,
+      "eval_wer": 0.29454775230185953,
+      "step": 131900
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 1.751647564469914e-05,
+      "loss": 0.0705,
+      "step": 132000
+    },
+    {
+      "epoch": 14.13,
+      "eval_loss": 0.460516095161438,
+      "eval_runtime": 48.7139,
+      "eval_samples_per_second": 41.61,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.2959017873262322,
+      "step": 132000
+    },
+    {
+      "epoch": 14.14,
+      "eval_loss": 0.4581490755081177,
+      "eval_runtime": 48.7404,
+      "eval_samples_per_second": 41.588,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.2947282903051092,
+      "step": 132100
+    },
+    {
+      "epoch": 14.15,
+      "eval_loss": 0.4597261846065521,
+      "eval_runtime": 48.813,
+      "eval_samples_per_second": 41.526,
+      "eval_steps_per_second": 5.204,
+      "eval_wer": 0.29481855930673406,
+      "step": 132200
+    },
+    {
+      "epoch": 14.16,
+      "eval_loss": 0.4612421691417694,
+      "eval_runtime": 48.6056,
+      "eval_samples_per_second": 41.703,
+      "eval_steps_per_second": 5.226,
+      "eval_wer": 0.294276945296985,
+      "step": 132300
+    },
+    {
+      "epoch": 14.18,
+      "eval_loss": 0.46105560660362244,
+      "eval_runtime": 48.4486,
+      "eval_samples_per_second": 41.838,
+      "eval_steps_per_second": 5.243,
+      "eval_wer": 0.2959017873262322,
+      "step": 132400
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 1.6441977077363896e-05,
+      "loss": 0.0727,
+      "step": 132500
+    },
+    {
+      "epoch": 14.19,
+      "eval_loss": 0.4569126069545746,
+      "eval_runtime": 48.7541,
+      "eval_samples_per_second": 41.576,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.29581151832460734,
+      "step": 132500
+    },
+    {
+      "epoch": 14.2,
+      "eval_loss": 0.4556213319301605,
+      "eval_runtime": 48.9259,
+      "eval_samples_per_second": 41.43,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.2950893663116086,
+      "step": 132600
+    },
+    {
+      "epoch": 14.21,
+      "eval_loss": 0.4596796929836273,
+      "eval_runtime": 48.837,
+      "eval_samples_per_second": 41.505,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.2955407113197328,
+      "step": 132700
+    },
+    {
+      "epoch": 14.22,
+      "eval_loss": 0.4472346305847168,
+      "eval_runtime": 48.874,
+      "eval_samples_per_second": 41.474,
+      "eval_steps_per_second": 5.197,
+      "eval_wer": 0.2934645242823614,
+      "step": 132800
+    },
+    {
+      "epoch": 14.23,
+      "eval_loss": 0.4573153853416443,
+      "eval_runtime": 48.9521,
+      "eval_samples_per_second": 41.408,
+      "eval_steps_per_second": 5.189,
+      "eval_wer": 0.294276945296985,
+      "step": 132900
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 1.5367478510028653e-05,
+      "loss": 0.0723,
+      "step": 133000
+    },
+    {
+      "epoch": 14.24,
+      "eval_loss": 0.4571533501148224,
+      "eval_runtime": 48.8128,
+      "eval_samples_per_second": 41.526,
+      "eval_steps_per_second": 5.204,
+      "eval_wer": 0.294276945296985,
+      "step": 133000
+    },
+    {
+      "epoch": 14.25,
+      "eval_loss": 0.45821666717529297,
+      "eval_runtime": 48.7097,
+      "eval_samples_per_second": 41.614,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.29563098032135765,
+      "step": 133100
+    },
+    {
+      "epoch": 14.26,
+      "eval_loss": 0.4598552882671356,
+      "eval_runtime": 48.9917,
+      "eval_samples_per_second": 41.374,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.2968044773424806,
+      "step": 133200
+    },
+    {
+      "epoch": 14.27,
+      "eval_loss": 0.46326953172683716,
+      "eval_runtime": 48.9832,
+      "eval_samples_per_second": 41.382,
+      "eval_steps_per_second": 5.185,
+      "eval_wer": 0.2961725943311067,
+      "step": 133300
+    },
+    {
+      "epoch": 14.28,
+      "eval_loss": 0.46043142676353455,
+      "eval_runtime": 48.7305,
+      "eval_samples_per_second": 41.596,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.29716555334898,
+      "step": 133400
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 1.429512893982808e-05,
+      "loss": 0.071,
+      "step": 133500
+    },
+    {
+      "epoch": 14.29,
+      "eval_loss": 0.45868054032325745,
+      "eval_runtime": 48.925,
+      "eval_samples_per_second": 41.431,
+      "eval_steps_per_second": 5.192,
+      "eval_wer": 0.29707528434735514,
+      "step": 133500
+    },
+    {
+      "epoch": 14.3,
+      "eval_loss": 0.4598270058631897,
+      "eval_runtime": 48.8352,
+      "eval_samples_per_second": 41.507,
+      "eval_steps_per_second": 5.201,
+      "eval_wer": 0.2972558223506048,
+      "step": 133600
+    },
+    {
+      "epoch": 14.31,
+      "eval_loss": 0.4579356610774994,
+      "eval_runtime": 48.7823,
+      "eval_samples_per_second": 41.552,
+      "eval_steps_per_second": 5.207,
+      "eval_wer": 0.2976168983571042,
+      "step": 133700
+    },
+    {
+      "epoch": 14.33,
+      "eval_loss": 0.4539426565170288,
+      "eval_runtime": 49.092,
+      "eval_samples_per_second": 41.29,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.29689474634410545,
+      "step": 133800
+    },
+    {
+      "epoch": 14.34,
+      "eval_loss": 0.46281492710113525,
+      "eval_runtime": 48.5974,
+      "eval_samples_per_second": 41.71,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.29608232532948187,
+      "step": 133900
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 1.3220630372492836e-05,
+      "loss": 0.0703,
+      "step": 134000
+    },
+    {
+      "epoch": 14.35,
+      "eval_loss": 0.46268972754478455,
+      "eval_runtime": 48.6114,
+      "eval_samples_per_second": 41.698,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.2974363603538545,
+      "step": 134000
+    },
+    {
+      "epoch": 14.36,
+      "eval_loss": 0.4611077606678009,
+      "eval_runtime": 48.7363,
+      "eval_samples_per_second": 41.591,
+      "eval_steps_per_second": 5.212,
+      "eval_wer": 0.2974363603538545,
+      "step": 134100
+    },
+    {
+      "epoch": 14.37,
+      "eval_loss": 0.4607299268245697,
+      "eval_runtime": 48.8318,
+      "eval_samples_per_second": 41.51,
+      "eval_steps_per_second": 5.202,
+      "eval_wer": 0.297707167358729,
+      "step": 134200
+    },
+    {
+      "epoch": 14.38,
+      "eval_loss": 0.46377691626548767,
+      "eval_runtime": 48.5242,
+      "eval_samples_per_second": 41.773,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.2983390503701029,
+      "step": 134300
+    },
+    {
+      "epoch": 14.39,
+      "eval_loss": 0.46281489729881287,
+      "eval_runtime": 48.4537,
+      "eval_samples_per_second": 41.834,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.29689474634410545,
+      "step": 134400
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 1.2146131805157592e-05,
+      "loss": 0.0736,
+      "step": 134500
+    },
+    {
+      "epoch": 14.4,
+      "eval_loss": 0.45427754521369934,
+      "eval_runtime": 49.0592,
+      "eval_samples_per_second": 41.317,
+      "eval_steps_per_second": 5.177,
+      "eval_wer": 0.2965336703376061,
+      "step": 134500
+    },
+    {
+      "epoch": 14.41,
+      "eval_loss": 0.4584720730781555,
+      "eval_runtime": 48.8757,
+      "eval_samples_per_second": 41.473,
+      "eval_steps_per_second": 5.197,
+      "eval_wer": 0.29626286333273155,
+      "step": 134600
+    },
+    {
+      "epoch": 14.42,
+      "eval_loss": 0.4636397361755371,
+      "eval_runtime": 48.5914,
+      "eval_samples_per_second": 41.715,
+      "eval_steps_per_second": 5.227,
+      "eval_wer": 0.29499909730998375,
+      "step": 134700
+    },
+    {
+      "epoch": 14.43,
+      "eval_loss": 0.463554710149765,
+      "eval_runtime": 48.5443,
+      "eval_samples_per_second": 41.756,
+      "eval_steps_per_second": 5.232,
+      "eval_wer": 0.2963531323343564,
+      "step": 134800
+    },
+    {
+      "epoch": 14.44,
+      "eval_loss": 0.4630184769630432,
+      "eval_runtime": 48.4957,
+      "eval_samples_per_second": 41.798,
+      "eval_steps_per_second": 5.238,
+      "eval_wer": 0.29581151832460734,
+      "step": 134900
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 1.107163323782235e-05,
+      "loss": 0.0715,
+      "step": 135000
+    },
+    {
+      "epoch": 14.45,
+      "eval_loss": 0.46107855439186096,
+      "eval_runtime": 48.3967,
+      "eval_samples_per_second": 41.883,
+      "eval_steps_per_second": 5.248,
+      "eval_wer": 0.2968044773424806,
+      "step": 135000
+    },
+    {
+      "epoch": 14.46,
+      "eval_loss": 0.4633425176143646,
+      "eval_runtime": 48.7522,
+      "eval_samples_per_second": 41.578,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.2966239393392309,
+      "step": 135100
+    },
+    {
+      "epoch": 14.48,
+      "eval_loss": 0.46640947461128235,
+      "eval_runtime": 48.7469,
+      "eval_samples_per_second": 41.582,
+      "eval_steps_per_second": 5.211,
+      "eval_wer": 0.2953601733164831,
+      "step": 135200
+    },
+    {
+      "epoch": 14.49,
+      "eval_loss": 0.46697285771369934,
+      "eval_runtime": 48.3158,
+      "eval_samples_per_second": 41.953,
+      "eval_steps_per_second": 5.257,
+      "eval_wer": 0.29454775230185953,
+      "step": 135300
+    },
+    {
+      "epoch": 14.5,
+      "eval_loss": 0.4637785255908966,
+      "eval_runtime": 48.5635,
+      "eval_samples_per_second": 41.739,
+      "eval_steps_per_second": 5.23,
+      "eval_wer": 0.29608232532948187,
+      "step": 135400
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 9.997134670487106e-06,
+      "loss": 0.073,
+      "step": 135500
+    },
+    {
+      "epoch": 14.51,
+      "eval_loss": 0.46345365047454834,
+      "eval_runtime": 48.4562,
+      "eval_samples_per_second": 41.832,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.2965336703376061,
+      "step": 135500
+    },
+    {
+      "epoch": 14.52,
+      "eval_loss": 0.4638654291629791,
+      "eval_runtime": 48.5254,
+      "eval_samples_per_second": 41.772,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.29563098032135765,
+      "step": 135600
+    },
+    {
+      "epoch": 14.53,
+      "eval_loss": 0.461653470993042,
+      "eval_runtime": 48.3998,
+      "eval_samples_per_second": 41.88,
+      "eval_steps_per_second": 5.248,
+      "eval_wer": 0.29481855930673406,
+      "step": 135700
+    },
+    {
+      "epoch": 14.54,
+      "eval_loss": 0.46086516976356506,
+      "eval_runtime": 48.3403,
+      "eval_samples_per_second": 41.932,
+      "eval_steps_per_second": 5.254,
+      "eval_wer": 0.29328398627911173,
+      "step": 135800
+    },
+    {
+      "epoch": 14.55,
+      "eval_loss": 0.4613838493824005,
+      "eval_runtime": 49.0954,
+      "eval_samples_per_second": 41.287,
+      "eval_steps_per_second": 5.174,
+      "eval_wer": 0.2947282903051092,
+      "step": 135900
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 8.926934097421202e-06,
+      "loss": 0.0717,
+      "step": 136000
+    },
+    {
+      "epoch": 14.56,
+      "eval_loss": 0.4566675126552582,
+      "eval_runtime": 48.7004,
+      "eval_samples_per_second": 41.622,
+      "eval_steps_per_second": 5.216,
+      "eval_wer": 0.29581151832460734,
+      "step": 136000
+    },
+    {
+      "epoch": 14.57,
+      "eval_loss": 0.4615228474140167,
+      "eval_runtime": 48.4673,
+      "eval_samples_per_second": 41.822,
+      "eval_steps_per_second": 5.241,
+      "eval_wer": 0.2933742552807366,
+      "step": 136100
+    },
+    {
+      "epoch": 14.58,
+      "eval_loss": 0.4605715572834015,
+      "eval_runtime": 48.7692,
+      "eval_samples_per_second": 41.563,
+      "eval_steps_per_second": 5.208,
+      "eval_wer": 0.29292291027261236,
+      "step": 136200
+    },
+    {
+      "epoch": 14.59,
+      "eval_loss": 0.4651778042316437,
+      "eval_runtime": 48.3037,
+      "eval_samples_per_second": 41.964,
+      "eval_steps_per_second": 5.258,
+      "eval_wer": 0.2933742552807366,
+      "step": 136300
+    },
+    {
+      "epoch": 14.6,
+      "eval_loss": 0.466376394033432,
+      "eval_runtime": 48.4015,
+      "eval_samples_per_second": 41.879,
+      "eval_steps_per_second": 5.248,
+      "eval_wer": 0.2933742552807366,
+      "step": 136400
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 7.852435530085958e-06,
+      "loss": 0.0717,
+      "step": 136500
+    },
+    {
+      "epoch": 14.61,
+      "eval_loss": 0.4657481908798218,
+      "eval_runtime": 49.5201,
+      "eval_samples_per_second": 40.933,
+      "eval_steps_per_second": 5.129,
+      "eval_wer": 0.2922910272612385,
+      "step": 136500
+    },
+    {
+      "epoch": 14.63,
+      "eval_loss": 0.46331438422203064,
+      "eval_runtime": 48.1829,
+      "eval_samples_per_second": 42.069,
+      "eval_steps_per_second": 5.272,
+      "eval_wer": 0.29310344827586204,
+      "step": 136600
+    },
+    {
+      "epoch": 14.64,
+      "eval_loss": 0.4623650014400482,
+      "eval_runtime": 48.4558,
+      "eval_samples_per_second": 41.832,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.294276945296985,
+      "step": 136700
+    },
+    {
+      "epoch": 14.65,
+      "eval_loss": 0.46153533458709717,
+      "eval_runtime": 48.2857,
+      "eval_samples_per_second": 41.979,
+      "eval_steps_per_second": 5.26,
+      "eval_wer": 0.2949088283083589,
+      "step": 136800
+    },
+    {
+      "epoch": 14.66,
+      "eval_loss": 0.46194565296173096,
+      "eval_runtime": 48.511,
+      "eval_samples_per_second": 41.784,
+      "eval_steps_per_second": 5.236,
+      "eval_wer": 0.2930131792742372,
+      "step": 136900
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 6.777936962750715e-06,
+      "loss": 0.0707,
+      "step": 137000
+    },
+    {
+      "epoch": 14.67,
+      "eval_loss": 0.4607996642589569,
+      "eval_runtime": 48.4004,
+      "eval_samples_per_second": 41.88,
+      "eval_steps_per_second": 5.248,
+      "eval_wer": 0.29355479328398626,
+      "step": 137000
+    },
+    {
+      "epoch": 14.68,
+      "eval_loss": 0.4615156948566437,
+      "eval_runtime": 48.1236,
+      "eval_samples_per_second": 42.121,
+      "eval_steps_per_second": 5.278,
+      "eval_wer": 0.2944574833002347,
+      "step": 137100
+    },
+    {
+      "epoch": 14.69,
+      "eval_loss": 0.4605298340320587,
+      "eval_runtime": 48.3679,
+      "eval_samples_per_second": 41.908,
+      "eval_steps_per_second": 5.251,
+      "eval_wer": 0.2940964072937353,
+      "step": 137200
+    },
+    {
+      "epoch": 14.7,
+      "eval_loss": 0.4598088562488556,
+      "eval_runtime": 48.3883,
+      "eval_samples_per_second": 41.89,
+      "eval_steps_per_second": 5.249,
+      "eval_wer": 0.29310344827586204,
+      "step": 137300
+    },
+    {
+      "epoch": 14.71,
+      "eval_loss": 0.4595940113067627,
+      "eval_runtime": 48.3856,
+      "eval_samples_per_second": 41.893,
+      "eval_steps_per_second": 5.249,
+      "eval_wer": 0.294276945296985,
+      "step": 137400
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 5.703438395415472e-06,
+      "loss": 0.0694,
+      "step": 137500
+    },
+    {
+      "epoch": 14.72,
+      "eval_loss": 0.4623918831348419,
+      "eval_runtime": 48.1674,
+      "eval_samples_per_second": 42.082,
+      "eval_steps_per_second": 5.273,
+      "eval_wer": 0.2927423722693627,
+      "step": 137500
+    },
+    {
+      "epoch": 14.73,
+      "eval_loss": 0.46141380071640015,
+      "eval_runtime": 48.7535,
+      "eval_samples_per_second": 41.576,
+      "eval_steps_per_second": 5.21,
+      "eval_wer": 0.29310344827586204,
+      "step": 137600
+    },
+    {
+      "epoch": 14.74,
+      "eval_loss": 0.4621369540691376,
+      "eval_runtime": 48.3733,
+      "eval_samples_per_second": 41.903,
+      "eval_steps_per_second": 5.251,
+      "eval_wer": 0.29238129626286335,
+      "step": 137700
+    },
+    {
+      "epoch": 14.75,
+      "eval_loss": 0.4588609039783478,
+      "eval_runtime": 48.512,
+      "eval_samples_per_second": 41.783,
+      "eval_steps_per_second": 5.236,
+      "eval_wer": 0.292020220256364,
+      "step": 137800
+    },
+    {
+      "epoch": 14.76,
+      "eval_loss": 0.4590286314487457,
+      "eval_runtime": 48.3051,
+      "eval_samples_per_second": 41.962,
+      "eval_steps_per_second": 5.258,
+      "eval_wer": 0.29256183426611304,
+      "step": 137900
+    },
+    {
+      "epoch": 14.78,
+      "learning_rate": 4.6310888252149e-06,
+      "loss": 0.0706,
+      "step": 138000
+    },
+    {
+      "epoch": 14.78,
+      "eval_loss": 0.4587528705596924,
+      "eval_runtime": 48.6184,
+      "eval_samples_per_second": 41.692,
+      "eval_steps_per_second": 5.224,
+      "eval_wer": 0.29310344827586204,
+      "step": 138000
+    },
+    {
+      "epoch": 14.79,
+      "eval_loss": 0.4583205282688141,
+      "eval_runtime": 48.382,
+      "eval_samples_per_second": 41.896,
+      "eval_steps_per_second": 5.25,
+      "eval_wer": 0.29283264127098757,
+      "step": 138100
+    },
+    {
+      "epoch": 14.8,
+      "eval_loss": 0.4552071690559387,
+      "eval_runtime": 48.5176,
+      "eval_samples_per_second": 41.779,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.2933742552807366,
+      "step": 138200
+    },
+    {
+      "epoch": 14.81,
+      "eval_loss": 0.45508986711502075,
+      "eval_runtime": 48.381,
+      "eval_samples_per_second": 41.897,
+      "eval_steps_per_second": 5.25,
+      "eval_wer": 0.2922910272612385,
+      "step": 138300
+    },
+    {
+      "epoch": 14.82,
+      "eval_loss": 0.4554976224899292,
+      "eval_runtime": 48.4911,
+      "eval_samples_per_second": 41.801,
+      "eval_steps_per_second": 5.238,
+      "eval_wer": 0.2927423722693627,
+      "step": 138400
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 3.5565902578796557e-06,
+      "loss": 0.0717,
+      "step": 138500
+    },
+    {
+      "epoch": 14.83,
+      "eval_loss": 0.45470017194747925,
+      "eval_runtime": 48.872,
+      "eval_samples_per_second": 41.476,
+      "eval_steps_per_second": 5.197,
+      "eval_wer": 0.2930131792742372,
+      "step": 138500
+    },
+    {
+      "epoch": 14.84,
+      "eval_loss": 0.45462867617607117,
+      "eval_runtime": 48.527,
+      "eval_samples_per_second": 41.771,
+      "eval_steps_per_second": 5.234,
+      "eval_wer": 0.2930131792742372,
+      "step": 138600
+    },
+    {
+      "epoch": 14.85,
+      "eval_loss": 0.45525601506233215,
+      "eval_runtime": 48.4224,
+      "eval_samples_per_second": 41.861,
+      "eval_steps_per_second": 5.246,
+      "eval_wer": 0.2933742552807366,
+      "step": 138700
+    },
+    {
+      "epoch": 14.86,
+      "eval_loss": 0.4553750157356262,
+      "eval_runtime": 49.4095,
+      "eval_samples_per_second": 41.025,
+      "eval_steps_per_second": 5.141,
+      "eval_wer": 0.29238129626286335,
+      "step": 138800
+    },
+    {
+      "epoch": 14.87,
+      "eval_loss": 0.4572896361351013,
+      "eval_runtime": 48.7188,
+      "eval_samples_per_second": 41.606,
+      "eval_steps_per_second": 5.214,
+      "eval_wer": 0.29238129626286335,
+      "step": 138900
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 2.4820916905444123e-06,
+      "loss": 0.0722,
+      "step": 139000
+    },
+    {
+      "epoch": 14.88,
+      "eval_loss": 0.4582221806049347,
+      "eval_runtime": 48.6094,
+      "eval_samples_per_second": 41.7,
+      "eval_steps_per_second": 5.225,
+      "eval_wer": 0.2927423722693627,
+      "step": 139000
+    },
+    {
+      "epoch": 14.89,
+      "eval_loss": 0.45864611864089966,
+      "eval_runtime": 48.4648,
+      "eval_samples_per_second": 41.824,
+      "eval_steps_per_second": 5.241,
+      "eval_wer": 0.29256183426611304,
+      "step": 139100
+    },
+    {
+      "epoch": 14.9,
+      "eval_loss": 0.4569910168647766,
+      "eval_runtime": 48.3017,
+      "eval_samples_per_second": 41.965,
+      "eval_steps_per_second": 5.259,
+      "eval_wer": 0.29256183426611304,
+      "step": 139200
+    },
+    {
+      "epoch": 14.91,
+      "eval_loss": 0.45708757638931274,
+      "eval_runtime": 49.3455,
+      "eval_samples_per_second": 41.078,
+      "eval_steps_per_second": 5.147,
+      "eval_wer": 0.2922910272612385,
+      "step": 139300
+    },
+    {
+      "epoch": 14.93,
+      "eval_loss": 0.45643165707588196,
+      "eval_runtime": 48.5858,
+      "eval_samples_per_second": 41.72,
+      "eval_steps_per_second": 5.228,
+      "eval_wer": 0.2924715652644882,
+      "step": 139400
+    },
+    {
+      "epoch": 14.94,
+      "learning_rate": 1.4075931232091689e-06,
+      "loss": 0.0698,
+      "step": 139500
+    },
+    {
+      "epoch": 14.94,
+      "eval_loss": 0.45725566148757935,
+      "eval_runtime": 48.4887,
+      "eval_samples_per_second": 41.804,
+      "eval_steps_per_second": 5.238,
+      "eval_wer": 0.2927423722693627,
+      "step": 139500
+    },
+    {
+      "epoch": 14.95,
+      "eval_loss": 0.457427054643631,
+      "eval_runtime": 48.4081,
+      "eval_samples_per_second": 41.873,
+      "eval_steps_per_second": 5.247,
+      "eval_wer": 0.2927423722693627,
+      "step": 139600
+    },
+    {
+      "epoch": 14.96,
+      "eval_loss": 0.4572872817516327,
+      "eval_runtime": 48.5718,
+      "eval_samples_per_second": 41.732,
+      "eval_steps_per_second": 5.229,
+      "eval_wer": 0.2926521032677379,
+      "step": 139700
+    },
+    {
+      "epoch": 14.97,
+      "eval_loss": 0.45759284496307373,
+      "eval_runtime": 48.8708,
+      "eval_samples_per_second": 41.477,
+      "eval_steps_per_second": 5.197,
+      "eval_wer": 0.2921104892579888,
+      "step": 139800
+    },
+    {
+      "epoch": 14.98,
+      "eval_loss": 0.45777422189712524,
+      "eval_runtime": 48.4515,
+      "eval_samples_per_second": 41.836,
+      "eval_steps_per_second": 5.242,
+      "eval_wer": 0.2922910272612385,
+      "step": 139900
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 3.330945558739255e-07,
+      "loss": 0.0705,
+      "step": 140000
+    },
+    {
+      "epoch": 14.99,
+      "eval_loss": 0.45787951350212097,
+      "eval_runtime": 48.5181,
+      "eval_samples_per_second": 41.778,
+      "eval_steps_per_second": 5.235,
+      "eval_wer": 0.29283264127098757,
+      "step": 140000
+    },
+    {
+      "epoch": 15.0,
+      "eval_loss": 0.4578445553779602,
+      "eval_runtime": 48.7019,
+      "eval_samples_per_second": 41.621,
+      "eval_steps_per_second": 5.215,
+      "eval_wer": 0.2926521032677379,
+      "step": 140100
+    },
+    {
+      "epoch": 15.0,
+      "step": 140100,
+      "total_flos": 7.232425156722584e+20,
+      "train_loss": 0.23010640207313793,
+      "train_runtime": 407183.7095,
+      "train_samples_per_second": 11.011,
+      "train_steps_per_second": 0.344
+    }
+  ],
+  "max_steps": 140100,
+  "num_train_epochs": 15,
+  "total_flos": 7.232425156722584e+20,
+  "trial_name": null,
+  "trial_params": null
+}