diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,12238 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 23.28,
+  "global_step": 87300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0001,
+      "loss": 16.3893,
+      "step": 100
+    },
+    {
+      "epoch": 0.03,
+      "eval_loss": 8.154696464538574,
+      "eval_runtime": 18.876,
+      "eval_samples_per_second": 4.715,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 9.999111032091743e-05,
+      "loss": 3.6998,
+      "step": 200
+    },
+    {
+      "epoch": 0.05,
+      "eval_loss": 3.48256254196167,
+      "eval_runtime": 19.0512,
+      "eval_samples_per_second": 4.672,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.998222064183483e-05,
+      "loss": 3.0156,
+      "step": 300
+    },
+    {
+      "epoch": 0.08,
+      "eval_loss": 3.4081289768218994,
+      "eval_runtime": 19.2561,
+      "eval_samples_per_second": 4.622,
+      "eval_wer": 1.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.997333096275225e-05,
+      "loss": 3.0049,
+      "step": 400
+    },
+    {
+      "epoch": 0.11,
+      "eval_loss": 3.4253759384155273,
+      "eval_runtime": 18.7811,
+      "eval_samples_per_second": 4.739,
+      "eval_wer": 1.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.996444128366966e-05,
+      "loss": 2.9946,
+      "step": 500
+    },
+    {
+      "epoch": 0.13,
+      "eval_loss": 3.3150370121002197,
+      "eval_runtime": 18.7872,
+      "eval_samples_per_second": 4.737,
+      "eval_wer": 1.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.995555160458708e-05,
+      "loss": 2.9508,
+      "step": 600
+    },
+    {
+      "epoch": 0.16,
+      "eval_loss": 3.2138397693634033,
+      "eval_runtime": 18.8338,
+      "eval_samples_per_second": 4.726,
+      "eval_wer": 1.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.994666192550449e-05,
+      "loss": 2.9476,
+      "step": 700
+    },
+    {
+      "epoch": 0.19,
+      "eval_loss": 3.145754814147949,
+      "eval_runtime": 18.7407,
+      "eval_samples_per_second": 4.749,
+      "eval_wer": 1.0,
+      "step": 700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.993777224642191e-05,
+      "loss": 2.9281,
+      "step": 800
+    },
+    {
+      "epoch": 0.21,
+      "eval_loss": 3.0304245948791504,
+      "eval_runtime": 18.9776,
+      "eval_samples_per_second": 4.69,
+      "eval_wer": 1.0,
+      "step": 800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.992888256733932e-05,
+      "loss": 2.7794,
+      "step": 900
+    },
+    {
+      "epoch": 0.24,
+      "eval_loss": 2.3334481716156006,
+      "eval_runtime": 18.8427,
+      "eval_samples_per_second": 4.723,
+      "eval_wer": 0.9905020352781547,
+      "step": 900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.991999288825674e-05,
+      "loss": 1.5845,
+      "step": 1000
+    },
+    {
+      "epoch": 0.27,
+      "eval_loss": 1.2006635665893555,
+      "eval_runtime": 19.0563,
+      "eval_samples_per_second": 4.67,
+      "eval_wer": 0.7639077340569878,
+      "step": 1000
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.991110320917416e-05,
+      "loss": 0.951,
+      "step": 1100
+    },
+    {
+      "epoch": 0.29,
+      "eval_loss": 0.9567620158195496,
+      "eval_runtime": 18.9557,
+      "eval_samples_per_second": 4.695,
+      "eval_wer": 0.6797829036635007,
+      "step": 1100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.990221353009156e-05,
+      "loss": 0.7728,
+      "step": 1200
+    },
+    {
+      "epoch": 0.32,
+      "eval_loss": 0.7444379329681396,
+      "eval_runtime": 18.779,
+      "eval_samples_per_second": 4.739,
+      "eval_wer": 0.6200814111261872,
+      "step": 1200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.989332385100899e-05,
+      "loss": 0.6914,
+      "step": 1300
+    },
+    {
+      "epoch": 0.35,
+      "eval_loss": 0.6914311647415161,
+      "eval_runtime": 19.0416,
+      "eval_samples_per_second": 4.674,
+      "eval_wer": 0.5739484396200815,
+      "step": 1300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.98844341719264e-05,
+      "loss": 0.6038,
+      "step": 1400
+    },
+    {
+      "epoch": 0.37,
+      "eval_loss": 0.6906067132949829,
+      "eval_runtime": 18.9276,
+      "eval_samples_per_second": 4.702,
+      "eval_wer": 0.5549525101763908,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.987554449284381e-05,
+      "loss": 0.5876,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4,
+      "eval_loss": 0.6231486201286316,
+      "eval_runtime": 18.7367,
+      "eval_samples_per_second": 4.75,
+      "eval_wer": 0.5332428765264586,
+      "step": 1500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.986665481376123e-05,
+      "loss": 0.5595,
+      "step": 1600
+    },
+    {
+      "epoch": 0.43,
+      "eval_loss": 0.5640144944190979,
+      "eval_runtime": 18.7452,
+      "eval_samples_per_second": 4.748,
+      "eval_wer": 0.5183175033921302,
+      "step": 1600
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.985776513467864e-05,
+      "loss": 0.5351,
+      "step": 1700
+    },
+    {
+      "epoch": 0.45,
+      "eval_loss": 0.5782244801521301,
+      "eval_runtime": 19.084,
+      "eval_samples_per_second": 4.664,
+      "eval_wer": 0.5223880597014925,
+      "step": 1700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.984887545559606e-05,
+      "loss": 0.5108,
+      "step": 1800
+    },
+    {
+      "epoch": 0.48,
+      "eval_loss": 0.569290816783905,
+      "eval_runtime": 19.0559,
+      "eval_samples_per_second": 4.67,
+      "eval_wer": 0.4966078697421981,
+      "step": 1800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.983998577651347e-05,
+      "loss": 0.4791,
+      "step": 1900
+    },
+    {
+      "epoch": 0.51,
+      "eval_loss": 0.5879985690116882,
+      "eval_runtime": 18.7275,
+      "eval_samples_per_second": 4.752,
+      "eval_wer": 0.5074626865671642,
+      "step": 1900
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.983109609743089e-05,
+      "loss": 0.4807,
+      "step": 2000
+    },
+    {
+      "epoch": 0.53,
+      "eval_loss": 0.5059410333633423,
+      "eval_runtime": 18.9221,
+      "eval_samples_per_second": 4.703,
+      "eval_wer": 0.4654002713704206,
+      "step": 2000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.982220641834831e-05,
+      "loss": 0.4752,
+      "step": 2100
+    },
+    {
+      "epoch": 0.56,
+      "eval_loss": 0.5032157301902771,
+      "eval_runtime": 19.1547,
+      "eval_samples_per_second": 4.646,
+      "eval_wer": 0.4694708276797829,
+      "step": 2100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.981331673926572e-05,
+      "loss": 0.4403,
+      "step": 2200
+    },
+    {
+      "epoch": 0.59,
+      "eval_loss": 0.4845319986343384,
+      "eval_runtime": 18.8285,
+      "eval_samples_per_second": 4.727,
+      "eval_wer": 0.4816824966078697,
+      "step": 2200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.980442706018314e-05,
+      "loss": 0.4498,
+      "step": 2300
+    },
+    {
+      "epoch": 0.61,
+      "eval_loss": 0.5130582451820374,
+      "eval_runtime": 18.6616,
+      "eval_samples_per_second": 4.769,
+      "eval_wer": 0.4626865671641791,
+      "step": 2300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.979553738110054e-05,
+      "loss": 0.4184,
+      "step": 2400
+    },
+    {
+      "epoch": 0.64,
+      "eval_loss": 0.5423906445503235,
+      "eval_runtime": 18.8876,
+      "eval_samples_per_second": 4.712,
+      "eval_wer": 0.4694708276797829,
+      "step": 2400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.978664770201797e-05,
+      "loss": 0.4174,
+      "step": 2500
+    },
+    {
+      "epoch": 0.67,
+      "eval_loss": 0.5088437795639038,
+      "eval_runtime": 18.9805,
+      "eval_samples_per_second": 4.689,
+      "eval_wer": 0.47761194029850745,
+      "step": 2500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.977775802293537e-05,
+      "loss": 0.4283,
+      "step": 2600
+    },
+    {
+      "epoch": 0.69,
+      "eval_loss": 0.4854317009449005,
+      "eval_runtime": 18.8551,
+      "eval_samples_per_second": 4.72,
+      "eval_wer": 0.451831750339213,
+      "step": 2600
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.976886834385279e-05,
+      "loss": 0.391,
+      "step": 2700
+    },
+    {
+      "epoch": 0.72,
+      "eval_loss": 0.5065292716026306,
+      "eval_runtime": 18.9612,
+      "eval_samples_per_second": 4.694,
+      "eval_wer": 0.4450474898236092,
+      "step": 2700
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.97599786647702e-05,
+      "loss": 0.4326,
+      "step": 2800
+    },
+    {
+      "epoch": 0.75,
+      "eval_loss": 0.43501290678977966,
+      "eval_runtime": 18.9522,
+      "eval_samples_per_second": 4.696,
+      "eval_wer": 0.46404341926729986,
+      "step": 2800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.975108898568762e-05,
+      "loss": 0.4054,
+      "step": 2900
+    },
+    {
+      "epoch": 0.77,
+      "eval_loss": 0.4695347845554352,
+      "eval_runtime": 18.8926,
+      "eval_samples_per_second": 4.711,
+      "eval_wer": 0.451831750339213,
+      "step": 2900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.974219930660503e-05,
+      "loss": 0.406,
+      "step": 3000
+    },
+    {
+      "epoch": 0.8,
+      "eval_loss": 0.5084555745124817,
+      "eval_runtime": 18.6863,
+      "eval_samples_per_second": 4.763,
+      "eval_wer": 0.4396200814111262,
+      "step": 3000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.973330962752245e-05,
+      "loss": 0.4113,
+      "step": 3100
+    },
+    {
+      "epoch": 0.83,
+      "eval_loss": 0.4934608042240143,
+      "eval_runtime": 19.0684,
+      "eval_samples_per_second": 4.667,
+      "eval_wer": 0.4396200814111262,
+      "step": 3100
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.972441994843986e-05,
+      "loss": 0.4023,
+      "step": 3200
+    },
+    {
+      "epoch": 0.85,
+      "eval_loss": 0.48704320192337036,
+      "eval_runtime": 19.0045,
+      "eval_samples_per_second": 4.683,
+      "eval_wer": 0.44776119402985076,
+      "step": 3200
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.971553026935728e-05,
+      "loss": 0.4152,
+      "step": 3300
+    },
+    {
+      "epoch": 0.88,
+      "eval_loss": 0.4661681354045868,
+      "eval_runtime": 18.6667,
+      "eval_samples_per_second": 4.768,
+      "eval_wer": 0.43419267299864317,
+      "step": 3300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.97066405902747e-05,
+      "loss": 0.3951,
+      "step": 3400
+    },
+    {
+      "epoch": 0.91,
+      "eval_loss": 0.4628346562385559,
+      "eval_runtime": 18.6945,
+      "eval_samples_per_second": 4.761,
+      "eval_wer": 0.4260515603799186,
+      "step": 3400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.969775091119212e-05,
+      "loss": 0.3739,
+      "step": 3500
+    },
+    {
+      "epoch": 0.93,
+      "eval_loss": 0.4508801996707916,
+      "eval_runtime": 18.9995,
+      "eval_samples_per_second": 4.684,
+      "eval_wer": 0.45047489823609227,
+      "step": 3500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.968886123210952e-05,
+      "loss": 0.3966,
+      "step": 3600
+    },
+    {
+      "epoch": 0.96,
+      "eval_loss": 0.5112784504890442,
+      "eval_runtime": 18.9324,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.44640434192673,
+      "step": 3600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.967997155302695e-05,
+      "loss": 0.3738,
+      "step": 3700
+    },
+    {
+      "epoch": 0.99,
+      "eval_loss": 0.5186980962753296,
+      "eval_runtime": 18.9294,
+      "eval_samples_per_second": 4.702,
+      "eval_wer": 0.42740841248303935,
+      "step": 3700
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 9.967108187394437e-05,
+      "loss": 0.35,
+      "step": 3800
+    },
+    {
+      "epoch": 1.01,
+      "eval_loss": 0.47684764862060547,
+      "eval_runtime": 18.784,
+      "eval_samples_per_second": 4.738,
+      "eval_wer": 0.42469470827679784,
+      "step": 3800
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.966219219486177e-05,
+      "loss": 0.3292,
+      "step": 3900
+    },
+    {
+      "epoch": 1.04,
+      "eval_loss": 0.44702979922294617,
+      "eval_runtime": 18.7696,
+      "eval_samples_per_second": 4.742,
+      "eval_wer": 0.42740841248303935,
+      "step": 3900
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.96533025157792e-05,
+      "loss": 0.3317,
+      "step": 4000
+    },
+    {
+      "epoch": 1.07,
+      "eval_loss": 0.44879746437072754,
+      "eval_runtime": 18.8524,
+      "eval_samples_per_second": 4.721,
+      "eval_wer": 0.4314789687924016,
+      "step": 4000
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.96444128366966e-05,
+      "loss": 0.3247,
+      "step": 4100
+    },
+    {
+      "epoch": 1.09,
+      "eval_loss": 0.45640674233436584,
+      "eval_runtime": 18.9416,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.43826322930800543,
+      "step": 4100
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.963552315761402e-05,
+      "loss": 0.3492,
+      "step": 4200
+    },
+    {
+      "epoch": 1.12,
+      "eval_loss": 0.4597824215888977,
+      "eval_runtime": 19.0336,
+      "eval_samples_per_second": 4.676,
+      "eval_wer": 0.4355495251017639,
+      "step": 4200
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 9.962663347853143e-05,
+      "loss": 0.3328,
+      "step": 4300
+    },
+    {
+      "epoch": 1.15,
+      "eval_loss": 0.4559178352355957,
+      "eval_runtime": 18.879,
+      "eval_samples_per_second": 4.714,
+      "eval_wer": 0.4260515603799186,
+      "step": 4300
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 9.961774379944885e-05,
+      "loss": 0.3091,
+      "step": 4400
+    },
+    {
+      "epoch": 1.17,
+      "eval_loss": 0.4449477791786194,
+      "eval_runtime": 18.8429,
+      "eval_samples_per_second": 4.723,
+      "eval_wer": 0.4491180461329715,
+      "step": 4400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 9.960885412036626e-05,
+      "loss": 0.3347,
+      "step": 4500
+    },
+    {
+      "epoch": 1.2,
+      "eval_loss": 0.4380190968513489,
+      "eval_runtime": 19.1851,
+      "eval_samples_per_second": 4.639,
+      "eval_wer": 0.41383989145183175,
+      "step": 4500
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 9.959996444128368e-05,
+      "loss": 0.3351,
+      "step": 4600
+    },
+    {
+      "epoch": 1.23,
+      "eval_loss": 0.42470598220825195,
+      "eval_runtime": 18.8771,
+      "eval_samples_per_second": 4.715,
+      "eval_wer": 0.4151967435549525,
+      "step": 4600
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 9.959107476220108e-05,
+      "loss": 0.337,
+      "step": 4700
+    },
+    {
+      "epoch": 1.25,
+      "eval_loss": 0.43264126777648926,
+      "eval_runtime": 18.7349,
+      "eval_samples_per_second": 4.751,
+      "eval_wer": 0.41926729986431477,
+      "step": 4700
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 9.95821850831185e-05,
+      "loss": 0.3261,
+      "step": 4800
+    },
+    {
+      "epoch": 1.28,
+      "eval_loss": 0.4580553472042084,
+      "eval_runtime": 19.162,
+      "eval_samples_per_second": 4.645,
+      "eval_wer": 0.42469470827679784,
+      "step": 4800
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 9.957329540403591e-05,
+      "loss": 0.3096,
+      "step": 4900
+    },
+    {
+      "epoch": 1.31,
+      "eval_loss": 0.4603330194950104,
+      "eval_runtime": 18.99,
+      "eval_samples_per_second": 4.687,
+      "eval_wer": 0.42740841248303935,
+      "step": 4900
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 9.956440572495333e-05,
+      "loss": 0.3087,
+      "step": 5000
+    },
+    {
+      "epoch": 1.33,
+      "eval_loss": 0.4239232838153839,
+      "eval_runtime": 18.9567,
+      "eval_samples_per_second": 4.695,
+      "eval_wer": 0.4260515603799186,
+      "step": 5000
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.955551604587074e-05,
+      "loss": 0.3302,
+      "step": 5100
+    },
+    {
+      "epoch": 1.36,
+      "eval_loss": 0.4312976002693176,
+      "eval_runtime": 18.9538,
+      "eval_samples_per_second": 4.696,
+      "eval_wer": 0.43283582089552236,
+      "step": 5100
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.954662636678816e-05,
+      "loss": 0.322,
+      "step": 5200
+    },
+    {
+      "epoch": 1.39,
+      "eval_loss": 0.4891665279865265,
+      "eval_runtime": 19.1223,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.41112618724559025,
+      "step": 5200
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 9.953773668770557e-05,
+      "loss": 0.3138,
+      "step": 5300
+    },
+    {
+      "epoch": 1.41,
+      "eval_loss": 0.4485594928264618,
+      "eval_runtime": 18.7069,
+      "eval_samples_per_second": 4.758,
+      "eval_wer": 0.4151967435549525,
+      "step": 5300
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 9.952884700862299e-05,
+      "loss": 0.3132,
+      "step": 5400
+    },
+    {
+      "epoch": 1.44,
+      "eval_loss": 0.44786882400512695,
+      "eval_runtime": 19.0411,
+      "eval_samples_per_second": 4.674,
+      "eval_wer": 0.4369063772048847,
+      "step": 5400
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.951995732954041e-05,
+      "loss": 0.3007,
+      "step": 5500
+    },
+    {
+      "epoch": 1.47,
+      "eval_loss": 0.45796269178390503,
+      "eval_runtime": 18.8766,
+      "eval_samples_per_second": 4.715,
+      "eval_wer": 0.4097693351424695,
+      "step": 5500
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 9.951106765045782e-05,
+      "loss": 0.3047,
+      "step": 5600
+    },
+    {
+      "epoch": 1.49,
+      "eval_loss": 0.4453733265399933,
+      "eval_runtime": 18.9395,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.4233378561736771,
+      "step": 5600
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.950217797137524e-05,
+      "loss": 0.3053,
+      "step": 5700
+    },
+    {
+      "epoch": 1.52,
+      "eval_loss": 0.4589693248271942,
+      "eval_runtime": 19.0887,
+      "eval_samples_per_second": 4.662,
+      "eval_wer": 0.4206241519674355,
+      "step": 5700
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.949328829229266e-05,
+      "loss": 0.318,
+      "step": 5800
+    },
+    {
+      "epoch": 1.55,
+      "eval_loss": 0.4058464467525482,
+      "eval_runtime": 18.9303,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.40705563093622793,
+      "step": 5800
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.948439861321008e-05,
+      "loss": 0.3217,
+      "step": 5900
+    },
+    {
+      "epoch": 1.57,
+      "eval_loss": 0.4143780469894409,
+      "eval_runtime": 19.0228,
+      "eval_samples_per_second": 4.679,
+      "eval_wer": 0.41112618724559025,
+      "step": 5900
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.947550893412749e-05,
+      "loss": 0.3104,
+      "step": 6000
+    },
+    {
+      "epoch": 1.6,
+      "eval_loss": 0.4316040575504303,
+      "eval_runtime": 18.7382,
+      "eval_samples_per_second": 4.75,
+      "eval_wer": 0.3989145183175034,
+      "step": 6000
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.94666192550449e-05,
+      "loss": 0.3007,
+      "step": 6100
+    },
+    {
+      "epoch": 1.63,
+      "eval_loss": 0.4299749732017517,
+      "eval_runtime": 18.9977,
+      "eval_samples_per_second": 4.685,
+      "eval_wer": 0.39755766621438265,
+      "step": 6100
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.945772957596231e-05,
+      "loss": 0.3119,
+      "step": 6200
+    },
+    {
+      "epoch": 1.65,
+      "eval_loss": 0.41399529576301575,
+      "eval_runtime": 19.0146,
+      "eval_samples_per_second": 4.681,
+      "eval_wer": 0.3853459972862958,
+      "step": 6200
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 9.944883989687973e-05,
+      "loss": 0.3092,
+      "step": 6300
+    },
+    {
+      "epoch": 1.68,
+      "eval_loss": 0.4074103832244873,
+      "eval_runtime": 18.9146,
+      "eval_samples_per_second": 4.705,
+      "eval_wer": 0.39620081411126185,
+      "step": 6300
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 9.943995021779714e-05,
+      "loss": 0.2953,
+      "step": 6400
+    },
+    {
+      "epoch": 1.71,
+      "eval_loss": 0.4460729658603668,
+      "eval_runtime": 18.9295,
+      "eval_samples_per_second": 4.702,
+      "eval_wer": 0.40298507462686567,
+      "step": 6400
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 9.943106053871456e-05,
+      "loss": 0.3215,
+      "step": 6500
+    },
+    {
+      "epoch": 1.73,
+      "eval_loss": 0.40716394782066345,
+      "eval_runtime": 19.1028,
+      "eval_samples_per_second": 4.659,
+      "eval_wer": 0.40705563093622793,
+      "step": 6500
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 9.942217085963197e-05,
+      "loss": 0.2985,
+      "step": 6600
+    },
+    {
+      "epoch": 1.76,
+      "eval_loss": 0.4205399751663208,
+      "eval_runtime": 18.8795,
+      "eval_samples_per_second": 4.714,
+      "eval_wer": 0.3880597014925373,
+      "step": 6600
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 9.941328118054939e-05,
+      "loss": 0.2902,
+      "step": 6700
+    },
+    {
+      "epoch": 1.79,
+      "eval_loss": 0.4227641224861145,
+      "eval_runtime": 18.929,
+      "eval_samples_per_second": 4.702,
+      "eval_wer": 0.39348710990502034,
+      "step": 6700
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.94043915014668e-05,
+      "loss": 0.2962,
+      "step": 6800
+    },
+    {
+      "epoch": 1.81,
+      "eval_loss": 0.4051659107208252,
+      "eval_runtime": 18.9694,
+      "eval_samples_per_second": 4.692,
+      "eval_wer": 0.4084124830393487,
+      "step": 6800
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 9.939550182238422e-05,
+      "loss": 0.2962,
+      "step": 6900
+    },
+    {
+      "epoch": 1.84,
+      "eval_loss": 0.4632880389690399,
+      "eval_runtime": 19.0719,
+      "eval_samples_per_second": 4.667,
+      "eval_wer": 0.4206241519674355,
+      "step": 6900
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 9.938661214330162e-05,
+      "loss": 0.321,
+      "step": 7000
+    },
+    {
+      "epoch": 1.87,
+      "eval_loss": 0.41146379709243774,
+      "eval_runtime": 19.2638,
+      "eval_samples_per_second": 4.62,
+      "eval_wer": 0.40298507462686567,
+      "step": 7000
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 9.937772246421905e-05,
+      "loss": 0.2962,
+      "step": 7100
+    },
+    {
+      "epoch": 1.89,
+      "eval_loss": 0.42060184478759766,
+      "eval_runtime": 19.0464,
+      "eval_samples_per_second": 4.673,
+      "eval_wer": 0.41383989145183175,
+      "step": 7100
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.936883278513645e-05,
+      "loss": 0.2865,
+      "step": 7200
+    },
+    {
+      "epoch": 1.92,
+      "eval_loss": 0.4277098476886749,
+      "eval_runtime": 19.195,
+      "eval_samples_per_second": 4.637,
+      "eval_wer": 0.40027137042062416,
+      "step": 7200
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 9.935994310605387e-05,
+      "loss": 0.285,
+      "step": 7300
+    },
+    {
+      "epoch": 1.95,
+      "eval_loss": 0.3912534713745117,
+      "eval_runtime": 18.8937,
+      "eval_samples_per_second": 4.711,
+      "eval_wer": 0.40298507462686567,
+      "step": 7300
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.935105342697128e-05,
+      "loss": 0.3124,
+      "step": 7400
+    },
+    {
+      "epoch": 1.97,
+      "eval_loss": 0.41216832399368286,
+      "eval_runtime": 19.4044,
+      "eval_samples_per_second": 4.587,
+      "eval_wer": 0.39348710990502034,
+      "step": 7400
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.93421637478887e-05,
+      "loss": 0.2811,
+      "step": 7500
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.4155157208442688,
+      "eval_runtime": 19.4461,
+      "eval_samples_per_second": 4.577,
+      "eval_wer": 0.40027137042062416,
+      "step": 7500
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.933327406880612e-05,
+      "loss": 0.2578,
+      "step": 7600
+    },
+    {
+      "epoch": 2.03,
+      "eval_loss": 0.38838207721710205,
+      "eval_runtime": 18.8382,
+      "eval_samples_per_second": 4.724,
+      "eval_wer": 0.39348710990502034,
+      "step": 7600
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 9.932438438972353e-05,
+      "loss": 0.2469,
+      "step": 7700
+    },
+    {
+      "epoch": 2.05,
+      "eval_loss": 0.3911269009113312,
+      "eval_runtime": 18.8861,
+      "eval_samples_per_second": 4.712,
+      "eval_wer": 0.3894165535956581,
+      "step": 7700
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.931549471064095e-05,
+      "loss": 0.2719,
+      "step": 7800
+    },
+    {
+      "epoch": 2.08,
+      "eval_loss": 0.3859439492225647,
+      "eval_runtime": 18.8223,
+      "eval_samples_per_second": 4.728,
+      "eval_wer": 0.39620081411126185,
+      "step": 7800
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.930660503155837e-05,
+      "loss": 0.2634,
+      "step": 7900
+    },
+    {
+      "epoch": 2.11,
+      "eval_loss": 0.40456128120422363,
+      "eval_runtime": 18.9193,
+      "eval_samples_per_second": 4.704,
+      "eval_wer": 0.3948439620081411,
+      "step": 7900
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 9.929771535247579e-05,
+      "loss": 0.2564,
+      "step": 8000
+    },
+    {
+      "epoch": 2.13,
+      "eval_loss": 0.4597094655036926,
+      "eval_runtime": 19.0139,
+      "eval_samples_per_second": 4.681,
+      "eval_wer": 0.40705563093622793,
+      "step": 8000
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 9.92888256733932e-05,
+      "loss": 0.2758,
+      "step": 8100
+    },
+    {
+      "epoch": 2.16,
+      "eval_loss": 0.4559546411037445,
+      "eval_runtime": 18.8509,
+      "eval_samples_per_second": 4.721,
+      "eval_wer": 0.4043419267299864,
+      "step": 8100
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 9.927993599431062e-05,
+      "loss": 0.2437,
+      "step": 8200
+    },
+    {
+      "epoch": 2.19,
+      "eval_loss": 0.472124844789505,
+      "eval_runtime": 19.1348,
+      "eval_samples_per_second": 4.651,
+      "eval_wer": 0.4043419267299864,
+      "step": 8200
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 9.927104631522803e-05,
+      "loss": 0.2463,
+      "step": 8300
+    },
+    {
+      "epoch": 2.21,
+      "eval_loss": 0.4547001123428345,
+      "eval_runtime": 18.8736,
+      "eval_samples_per_second": 4.716,
+      "eval_wer": 0.41655359565807326,
+      "step": 8300
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 9.926215663614545e-05,
+      "loss": 0.2698,
+      "step": 8400
+    },
+    {
+      "epoch": 2.24,
+      "eval_loss": 0.44032570719718933,
+      "eval_runtime": 18.9332,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.3921302578018996,
+      "step": 8400
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 9.925326695706285e-05,
+      "loss": 0.2615,
+      "step": 8500
+    },
+    {
+      "epoch": 2.27,
+      "eval_loss": 0.42494475841522217,
+      "eval_runtime": 18.8931,
+      "eval_samples_per_second": 4.711,
+      "eval_wer": 0.4056987788331072,
+      "step": 8500
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 9.924437727798027e-05,
+      "loss": 0.2563,
+      "step": 8600
+    },
+    {
+      "epoch": 2.29,
+      "eval_loss": 0.44931527972221375,
+      "eval_runtime": 18.8781,
+      "eval_samples_per_second": 4.714,
+      "eval_wer": 0.3989145183175034,
+      "step": 8600
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 9.923548759889768e-05,
+      "loss": 0.2516,
+      "step": 8700
+    },
+    {
+      "epoch": 2.32,
+      "eval_loss": 0.48540976643562317,
+      "eval_runtime": 18.8321,
+      "eval_samples_per_second": 4.726,
+      "eval_wer": 0.3921302578018996,
+      "step": 8700
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 9.92265979198151e-05,
+      "loss": 0.289,
+      "step": 8800
+    },
+    {
+      "epoch": 2.35,
+      "eval_loss": 0.47304674983024597,
+      "eval_runtime": 18.7854,
+      "eval_samples_per_second": 4.738,
+      "eval_wer": 0.39755766621438265,
+      "step": 8800
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 9.921770824073251e-05,
+      "loss": 0.2456,
+      "step": 8900
+    },
+    {
+      "epoch": 2.37,
+      "eval_loss": 0.49813735485076904,
+      "eval_runtime": 18.7937,
+      "eval_samples_per_second": 4.736,
+      "eval_wer": 0.4016282225237449,
+      "step": 8900
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 9.920881856164993e-05,
+      "loss": 0.2582,
+      "step": 9000
+    },
+    {
+      "epoch": 2.4,
+      "eval_loss": 0.47900137305259705,
+      "eval_runtime": 18.6913,
+      "eval_samples_per_second": 4.762,
+      "eval_wer": 0.4097693351424695,
+      "step": 9000
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 9.919992888256734e-05,
+      "loss": 0.255,
+      "step": 9100
+    },
+    {
+      "epoch": 2.43,
+      "eval_loss": 0.4543055295944214,
+      "eval_runtime": 18.8262,
+      "eval_samples_per_second": 4.727,
+      "eval_wer": 0.4056987788331072,
+      "step": 9100
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 9.919103920348476e-05,
+      "loss": 0.2513,
+      "step": 9200
+    },
+    {
+      "epoch": 2.45,
+      "eval_loss": 0.4764789938926697,
+      "eval_runtime": 18.9195,
+      "eval_samples_per_second": 4.704,
+      "eval_wer": 0.40027137042062416,
+      "step": 9200
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 9.918214952440218e-05,
+      "loss": 0.247,
+      "step": 9300
+    },
+    {
+      "epoch": 2.48,
+      "eval_loss": 0.4753105938434601,
+      "eval_runtime": 18.7974,
+      "eval_samples_per_second": 4.735,
+      "eval_wer": 0.39348710990502034,
+      "step": 9300
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 9.917325984531959e-05,
+      "loss": 0.2704,
+      "step": 9400
+    },
+    {
+      "epoch": 2.51,
+      "eval_loss": 0.44719693064689636,
+      "eval_runtime": 18.9517,
+      "eval_samples_per_second": 4.696,
+      "eval_wer": 0.39077340569877883,
+      "step": 9400
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 9.9164370166237e-05,
+      "loss": 0.2601,
+      "step": 9500
+    },
+    {
+      "epoch": 2.53,
+      "eval_loss": 0.43241894245147705,
+      "eval_runtime": 19.3423,
+      "eval_samples_per_second": 4.601,
+      "eval_wer": 0.4043419267299864,
+      "step": 9500
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 9.915548048715441e-05,
+      "loss": 0.2453,
+      "step": 9600
+    },
+    {
+      "epoch": 2.56,
+      "eval_loss": 0.42892536520957947,
+      "eval_runtime": 18.8648,
+      "eval_samples_per_second": 4.718,
+      "eval_wer": 0.3948439620081411,
+      "step": 9600
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 9.914659080807183e-05,
+      "loss": 0.2382,
+      "step": 9700
+    },
+    {
+      "epoch": 2.59,
+      "eval_loss": 0.42483776807785034,
+      "eval_runtime": 18.8169,
+      "eval_samples_per_second": 4.73,
+      "eval_wer": 0.40298507462686567,
+      "step": 9700
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 9.913770112898924e-05,
+      "loss": 0.2505,
+      "step": 9800
+    },
+    {
+      "epoch": 2.61,
+      "eval_loss": 0.3920654058456421,
+      "eval_runtime": 18.8682,
+      "eval_samples_per_second": 4.717,
+      "eval_wer": 0.3948439620081411,
+      "step": 9800
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 9.912881144990666e-05,
+      "loss": 0.254,
+      "step": 9900
+    },
+    {
+      "epoch": 2.64,
+      "eval_loss": 0.4071875214576721,
+      "eval_runtime": 19.105,
+      "eval_samples_per_second": 4.658,
+      "eval_wer": 0.417910447761194,
+      "step": 9900
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 9.911992177082408e-05,
+      "loss": 0.2502,
+      "step": 10000
+    },
+    {
+      "epoch": 2.67,
+      "eval_loss": 0.44064927101135254,
+      "eval_runtime": 18.7382,
+      "eval_samples_per_second": 4.75,
+      "eval_wer": 0.40027137042062416,
+      "step": 10000
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 9.911103209174149e-05,
+      "loss": 0.2591,
+      "step": 10100
+    },
+    {
+      "epoch": 2.69,
+      "eval_loss": 0.3873971998691559,
+      "eval_runtime": 18.6681,
+      "eval_samples_per_second": 4.767,
+      "eval_wer": 0.3989145183175034,
+      "step": 10100
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 9.910214241265891e-05,
+      "loss": 0.2538,
+      "step": 10200
+    },
+    {
+      "epoch": 2.72,
+      "eval_loss": 0.3714084029197693,
+      "eval_runtime": 19.1713,
+      "eval_samples_per_second": 4.642,
+      "eval_wer": 0.3812754409769335,
+      "step": 10200
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 9.909325273357633e-05,
+      "loss": 0.2367,
+      "step": 10300
+    },
+    {
+      "epoch": 2.75,
+      "eval_loss": 0.38263094425201416,
+      "eval_runtime": 19.1515,
+      "eval_samples_per_second": 4.647,
+      "eval_wer": 0.4056987788331072,
+      "step": 10300
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 9.908436305449374e-05,
+      "loss": 0.2253,
+      "step": 10400
+    },
+    {
+      "epoch": 2.77,
+      "eval_loss": 0.39084652066230774,
+      "eval_runtime": 18.9766,
+      "eval_samples_per_second": 4.69,
+      "eval_wer": 0.40027137042062416,
+      "step": 10400
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 9.907547337541116e-05,
+      "loss": 0.2557,
+      "step": 10500
+    },
+    {
+      "epoch": 2.8,
+      "eval_loss": 0.386184960603714,
+      "eval_runtime": 19.0577,
+      "eval_samples_per_second": 4.67,
+      "eval_wer": 0.4016282225237449,
+      "step": 10500
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 9.906658369632857e-05,
+      "loss": 0.2623,
+      "step": 10600
+    },
+    {
+      "epoch": 2.83,
+      "eval_loss": 0.41050976514816284,
+      "eval_runtime": 19.2149,
+      "eval_samples_per_second": 4.632,
+      "eval_wer": 0.39620081411126185,
+      "step": 10600
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 9.905769401724599e-05,
+      "loss": 0.247,
+      "step": 10700
+    },
+    {
+      "epoch": 2.85,
+      "eval_loss": 0.39522433280944824,
+      "eval_runtime": 19.0014,
+      "eval_samples_per_second": 4.684,
+      "eval_wer": 0.4260515603799186,
+      "step": 10700
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 9.904880433816339e-05,
+      "loss": 0.2707,
+      "step": 10800
+    },
+    {
+      "epoch": 2.88,
+      "eval_loss": 0.40348905324935913,
+      "eval_runtime": 18.7692,
+      "eval_samples_per_second": 4.742,
+      "eval_wer": 0.3989145183175034,
+      "step": 10800
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 9.903991465908081e-05,
+      "loss": 0.2436,
+      "step": 10900
+    },
+    {
+      "epoch": 2.91,
+      "eval_loss": 0.3818841576576233,
+      "eval_runtime": 19.2548,
+      "eval_samples_per_second": 4.622,
+      "eval_wer": 0.39755766621438265,
+      "step": 10900
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 9.903102497999822e-05,
+      "loss": 0.2592,
+      "step": 11000
+    },
+    {
+      "epoch": 2.93,
+      "eval_loss": 0.40791064500808716,
+      "eval_runtime": 18.991,
+      "eval_samples_per_second": 4.686,
+      "eval_wer": 0.39755766621438265,
+      "step": 11000
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 9.902213530091564e-05,
+      "loss": 0.2524,
+      "step": 11100
+    },
+    {
+      "epoch": 2.96,
+      "eval_loss": 0.4026681184768677,
+      "eval_runtime": 18.9403,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.38670284938941657,
+      "step": 11100
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 9.901324562183306e-05,
+      "loss": 0.2389,
+      "step": 11200
+    },
+    {
+      "epoch": 2.99,
+      "eval_loss": 0.38911232352256775,
+      "eval_runtime": 18.8555,
+      "eval_samples_per_second": 4.72,
+      "eval_wer": 0.3989145183175034,
+      "step": 11200
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 9.900435594275047e-05,
+      "loss": 0.2321,
+      "step": 11300
+    },
+    {
+      "epoch": 3.01,
+      "eval_loss": 0.3908725678920746,
+      "eval_runtime": 18.8446,
+      "eval_samples_per_second": 4.723,
+      "eval_wer": 0.4056987788331072,
+      "step": 11300
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 9.899546626366789e-05,
+      "loss": 0.2331,
+      "step": 11400
+    },
+    {
+      "epoch": 3.04,
+      "eval_loss": 0.3997111916542053,
+      "eval_runtime": 19.0613,
+      "eval_samples_per_second": 4.669,
+      "eval_wer": 0.3812754409769335,
+      "step": 11400
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 9.89865765845853e-05,
+      "loss": 0.209,
+      "step": 11500
+    },
+    {
+      "epoch": 3.07,
+      "eval_loss": 0.4227985143661499,
+      "eval_runtime": 19.0526,
+      "eval_samples_per_second": 4.671,
+      "eval_wer": 0.3948439620081411,
+      "step": 11500
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 9.897768690550272e-05,
+      "loss": 0.2155,
+      "step": 11600
+    },
+    {
+      "epoch": 3.09,
+      "eval_loss": 0.420188307762146,
+      "eval_runtime": 19.1299,
+      "eval_samples_per_second": 4.652,
+      "eval_wer": 0.39348710990502034,
+      "step": 11600
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 9.896879722642013e-05,
+      "loss": 0.2228,
+      "step": 11700
+    },
+    {
+      "epoch": 3.12,
+      "eval_loss": 0.4401314854621887,
+      "eval_runtime": 18.9789,
+      "eval_samples_per_second": 4.689,
+      "eval_wer": 0.38670284938941657,
+      "step": 11700
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 9.895990754733755e-05,
+      "loss": 0.2232,
+      "step": 11800
+    },
+    {
+      "epoch": 3.15,
+      "eval_loss": 0.46247828006744385,
+      "eval_runtime": 19.3603,
+      "eval_samples_per_second": 4.597,
+      "eval_wer": 0.38263229308005425,
+      "step": 11800
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 9.895101786825495e-05,
+      "loss": 0.2214,
+      "step": 11900
+    },
+    {
+      "epoch": 3.17,
+      "eval_loss": 0.4425117075443268,
+      "eval_runtime": 18.9946,
+      "eval_samples_per_second": 4.686,
+      "eval_wer": 0.3812754409769335,
+      "step": 11900
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 9.894212818917237e-05,
+      "loss": 0.2174,
+      "step": 12000
+    },
+    {
+      "epoch": 3.2,
+      "eval_loss": 0.4542634189128876,
+      "eval_runtime": 19.0943,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.3989145183175034,
+      "step": 12000
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 9.89332385100898e-05,
+      "loss": 0.2299,
+      "step": 12100
+    },
+    {
+      "epoch": 3.23,
+      "eval_loss": 0.4233208894729614,
+      "eval_runtime": 19.0266,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.39755766621438265,
+      "step": 12100
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 9.89243488310072e-05,
+      "loss": 0.2386,
+      "step": 12200
+    },
+    {
+      "epoch": 3.25,
+      "eval_loss": 0.4648773670196533,
+      "eval_runtime": 18.9972,
+      "eval_samples_per_second": 4.685,
+      "eval_wer": 0.3853459972862958,
+      "step": 12200
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 9.891545915192462e-05,
+      "loss": 0.2059,
+      "step": 12300
+    },
+    {
+      "epoch": 3.28,
+      "eval_loss": 0.43632611632347107,
+      "eval_runtime": 19.1176,
+      "eval_samples_per_second": 4.655,
+      "eval_wer": 0.3989145183175034,
+      "step": 12300
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 9.890656947284204e-05,
+      "loss": 0.2188,
+      "step": 12400
+    },
+    {
+      "epoch": 3.31,
+      "eval_loss": 0.4346342086791992,
+      "eval_runtime": 19.0243,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.4043419267299864,
+      "step": 12400
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 9.889767979375945e-05,
+      "loss": 0.2188,
+      "step": 12500
+    },
+    {
+      "epoch": 3.33,
+      "eval_loss": 0.425716370344162,
+      "eval_runtime": 19.1207,
+      "eval_samples_per_second": 4.655,
+      "eval_wer": 0.3853459972862958,
+      "step": 12500
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 9.888879011467687e-05,
+      "loss": 0.2199,
+      "step": 12600
+    },
+    {
+      "epoch": 3.36,
+      "eval_loss": 0.43285807967185974,
+      "eval_runtime": 18.852,
+      "eval_samples_per_second": 4.721,
+      "eval_wer": 0.3921302578018996,
+      "step": 12600
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 9.887990043559428e-05,
+      "loss": 0.2127,
+      "step": 12700
+    },
+    {
+      "epoch": 3.39,
+      "eval_loss": 0.45121872425079346,
+      "eval_runtime": 18.7657,
+      "eval_samples_per_second": 4.743,
+      "eval_wer": 0.4056987788331072,
+      "step": 12700
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 9.88710107565117e-05,
+      "loss": 0.217,
+      "step": 12800
+    },
+    {
+      "epoch": 3.41,
+      "eval_loss": 0.4849798083305359,
+      "eval_runtime": 18.7794,
+      "eval_samples_per_second": 4.739,
+      "eval_wer": 0.3921302578018996,
+      "step": 12800
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 9.88621210774291e-05,
+      "loss": 0.2398,
+      "step": 12900
+    },
+    {
+      "epoch": 3.44,
+      "eval_loss": 0.5062362551689148,
+      "eval_runtime": 18.7298,
+      "eval_samples_per_second": 4.752,
+      "eval_wer": 0.37177747625508817,
+      "step": 12900
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 9.885323139834653e-05,
+      "loss": 0.221,
+      "step": 13000
+    },
+    {
+      "epoch": 3.47,
+      "eval_loss": 0.47657549381256104,
+      "eval_runtime": 19.0148,
+      "eval_samples_per_second": 4.681,
+      "eval_wer": 0.373134328358209,
+      "step": 13000
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 9.884434171926395e-05,
+      "loss": 0.2212,
+      "step": 13100
+    },
+    {
+      "epoch": 3.49,
+      "eval_loss": 0.44042253494262695,
+      "eval_runtime": 19.0374,
+      "eval_samples_per_second": 4.675,
+      "eval_wer": 0.39077340569877883,
+      "step": 13100
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 9.883545204018135e-05,
+      "loss": 0.2179,
+      "step": 13200
+    },
+    {
+      "epoch": 3.52,
+      "eval_loss": 0.4313114583492279,
+      "eval_runtime": 19.1165,
+      "eval_samples_per_second": 4.656,
+      "eval_wer": 0.39348710990502034,
+      "step": 13200
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 9.882656236109877e-05,
+      "loss": 0.2293,
+      "step": 13300
+    },
+    {
+      "epoch": 3.55,
+      "eval_loss": 0.45125553011894226,
+      "eval_runtime": 19.0854,
+      "eval_samples_per_second": 4.663,
+      "eval_wer": 0.383989145183175,
+      "step": 13300
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 9.881767268201618e-05,
+      "loss": 0.2077,
+      "step": 13400
+    },
+    {
+      "epoch": 3.57,
+      "eval_loss": 0.4451759457588196,
+      "eval_runtime": 19.0728,
+      "eval_samples_per_second": 4.666,
+      "eval_wer": 0.4097693351424695,
+      "step": 13400
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 9.88087830029336e-05,
+      "loss": 0.2271,
+      "step": 13500
+    },
+    {
+      "epoch": 3.6,
+      "eval_loss": 0.4819214642047882,
+      "eval_runtime": 18.9467,
+      "eval_samples_per_second": 4.697,
+      "eval_wer": 0.4084124830393487,
+      "step": 13500
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 9.879989332385101e-05,
+      "loss": 0.2209,
+      "step": 13600
+    },
+    {
+      "epoch": 3.63,
+      "eval_loss": 0.45108529925346375,
+      "eval_runtime": 19.2092,
+      "eval_samples_per_second": 4.633,
+      "eval_wer": 0.40298507462686567,
+      "step": 13600
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 9.879100364476843e-05,
+      "loss": 0.224,
+      "step": 13700
+    },
+    {
+      "epoch": 3.65,
+      "eval_loss": 0.44665461778640747,
+      "eval_runtime": 19.0304,
+      "eval_samples_per_second": 4.677,
+      "eval_wer": 0.3853459972862958,
+      "step": 13700
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 9.878211396568584e-05,
+      "loss": 0.2182,
+      "step": 13800
+    },
+    {
+      "epoch": 3.68,
+      "eval_loss": 0.3846975266933441,
+      "eval_runtime": 18.9861,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.3758480325644505,
+      "step": 13800
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 9.877322428660326e-05,
+      "loss": 0.233,
+      "step": 13900
+    },
+    {
+      "epoch": 3.71,
+      "eval_loss": 0.3762931823730469,
+      "eval_runtime": 18.9347,
+      "eval_samples_per_second": 4.7,
+      "eval_wer": 0.373134328358209,
+      "step": 13900
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 9.876433460752066e-05,
+      "loss": 0.212,
+      "step": 14000
+    },
+    {
+      "epoch": 3.73,
+      "eval_loss": 0.37193208932876587,
+      "eval_runtime": 19.0798,
+      "eval_samples_per_second": 4.665,
+      "eval_wer": 0.3853459972862958,
+      "step": 14000
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 9.875544492843809e-05,
+      "loss": 0.231,
+      "step": 14100
+    },
+    {
+      "epoch": 3.76,
+      "eval_loss": 0.3909999132156372,
+      "eval_runtime": 18.6352,
+      "eval_samples_per_second": 4.776,
+      "eval_wer": 0.3880597014925373,
+      "step": 14100
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 9.874655524935549e-05,
+      "loss": 0.2181,
+      "step": 14200
+    },
+    {
+      "epoch": 3.79,
+      "eval_loss": 0.39621177315711975,
+      "eval_runtime": 18.9455,
+      "eval_samples_per_second": 4.698,
+      "eval_wer": 0.3704206241519674,
+      "step": 14200
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 9.873766557027291e-05,
+      "loss": 0.2018,
+      "step": 14300
+    },
+    {
+      "epoch": 3.81,
+      "eval_loss": 0.41908976435661316,
+      "eval_runtime": 18.8525,
+      "eval_samples_per_second": 4.721,
+      "eval_wer": 0.3880597014925373,
+      "step": 14300
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 9.872877589119033e-05,
+      "loss": 0.2267,
+      "step": 14400
+    },
+    {
+      "epoch": 3.84,
+      "eval_loss": 0.39545124769210815,
+      "eval_runtime": 19.1242,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.4043419267299864,
+      "step": 14400
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 9.871988621210775e-05,
+      "loss": 0.2248,
+      "step": 14500
+    },
+    {
+      "epoch": 3.87,
+      "eval_loss": 0.40026775002479553,
+      "eval_runtime": 18.8697,
+      "eval_samples_per_second": 4.717,
+      "eval_wer": 0.3880597014925373,
+      "step": 14500
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 9.871099653302516e-05,
+      "loss": 0.2394,
+      "step": 14600
+    },
+    {
+      "epoch": 3.89,
+      "eval_loss": 0.40464088320732117,
+      "eval_runtime": 19.3048,
+      "eval_samples_per_second": 4.61,
+      "eval_wer": 0.3948439620081411,
+      "step": 14600
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 9.870210685394258e-05,
+      "loss": 0.2041,
+      "step": 14700
+    },
+    {
+      "epoch": 3.92,
+      "eval_loss": 0.4450311064720154,
+      "eval_runtime": 18.9653,
+      "eval_samples_per_second": 4.693,
+      "eval_wer": 0.39620081411126185,
+      "step": 14700
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 9.869321717486e-05,
+      "loss": 0.2128,
+      "step": 14800
+    },
+    {
+      "epoch": 3.95,
+      "eval_loss": 0.4340137541294098,
+      "eval_runtime": 18.8543,
+      "eval_samples_per_second": 4.72,
+      "eval_wer": 0.3921302578018996,
+      "step": 14800
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 9.868432749577741e-05,
+      "loss": 0.247,
+      "step": 14900
+    },
+    {
+      "epoch": 3.97,
+      "eval_loss": 0.4467424154281616,
+      "eval_runtime": 18.9813,
+      "eval_samples_per_second": 4.689,
+      "eval_wer": 0.39348710990502034,
+      "step": 14900
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 9.867543781669483e-05,
+      "loss": 0.2065,
+      "step": 15000
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 0.4233141243457794,
+      "eval_runtime": 19.0909,
+      "eval_samples_per_second": 4.662,
+      "eval_wer": 0.3880597014925373,
+      "step": 15000
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 9.866654813761224e-05,
+      "loss": 0.1929,
+      "step": 15100
+    },
+    {
+      "epoch": 4.03,
+      "eval_loss": 0.4598129689693451,
+      "eval_runtime": 19.2258,
+      "eval_samples_per_second": 4.629,
+      "eval_wer": 0.38670284938941657,
+      "step": 15100
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 9.865765845852966e-05,
+      "loss": 0.1897,
+      "step": 15200
+    },
+    {
+      "epoch": 4.05,
+      "eval_loss": 0.48485830426216125,
+      "eval_runtime": 19.2033,
+      "eval_samples_per_second": 4.635,
+      "eval_wer": 0.3704206241519674,
+      "step": 15200
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 9.864876877944707e-05,
+      "loss": 0.1974,
+      "step": 15300
+    },
+    {
+      "epoch": 4.08,
+      "eval_loss": 0.465469092130661,
+      "eval_runtime": 18.9331,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.3880597014925373,
+      "step": 15300
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 9.863987910036449e-05,
+      "loss": 0.2077,
+      "step": 15400
+    },
+    {
+      "epoch": 4.11,
+      "eval_loss": 0.42971137166023254,
+      "eval_runtime": 19.061,
+      "eval_samples_per_second": 4.669,
+      "eval_wer": 0.383989145183175,
+      "step": 15400
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 9.86309894212819e-05,
+      "loss": 0.2157,
+      "step": 15500
+    },
+    {
+      "epoch": 4.13,
+      "eval_loss": 0.49831441044807434,
+      "eval_runtime": 18.956,
+      "eval_samples_per_second": 4.695,
+      "eval_wer": 0.3853459972862958,
+      "step": 15500
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 9.862209974219931e-05,
+      "loss": 0.2083,
+      "step": 15600
+    },
+    {
+      "epoch": 4.16,
+      "eval_loss": 0.44665566086769104,
+      "eval_runtime": 18.9129,
+      "eval_samples_per_second": 4.706,
+      "eval_wer": 0.4016282225237449,
+      "step": 15600
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 9.861321006311672e-05,
+      "loss": 0.1823,
+      "step": 15700
+    },
+    {
+      "epoch": 4.19,
+      "eval_loss": 0.49660953879356384,
+      "eval_runtime": 18.799,
+      "eval_samples_per_second": 4.734,
+      "eval_wer": 0.39077340569877883,
+      "step": 15700
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 9.860432038403414e-05,
+      "loss": 0.1879,
+      "step": 15800
+    },
+    {
+      "epoch": 4.21,
+      "eval_loss": 0.44513949751853943,
+      "eval_runtime": 18.9952,
+      "eval_samples_per_second": 4.685,
+      "eval_wer": 0.39755766621438265,
+      "step": 15800
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 9.859543070495155e-05,
+      "loss": 0.1976,
+      "step": 15900
+    },
+    {
+      "epoch": 4.24,
+      "eval_loss": 0.526491641998291,
+      "eval_runtime": 18.8119,
+      "eval_samples_per_second": 4.731,
+      "eval_wer": 0.39755766621438265,
+      "step": 15900
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 9.858654102586897e-05,
+      "loss": 0.1972,
+      "step": 16000
+    },
+    {
+      "epoch": 4.27,
+      "eval_loss": 0.47604671120643616,
+      "eval_runtime": 18.7759,
+      "eval_samples_per_second": 4.74,
+      "eval_wer": 0.4016282225237449,
+      "step": 16000
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 9.857765134678638e-05,
+      "loss": 0.1918,
+      "step": 16100
+    },
+    {
+      "epoch": 4.29,
+      "eval_loss": 0.48783737421035767,
+      "eval_runtime": 18.9324,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.4084124830393487,
+      "step": 16100
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 9.85687616677038e-05,
+      "loss": 0.2054,
+      "step": 16200
+    },
+    {
+      "epoch": 4.32,
+      "eval_loss": 0.45347392559051514,
+      "eval_runtime": 18.9595,
+      "eval_samples_per_second": 4.694,
+      "eval_wer": 0.3948439620081411,
+      "step": 16200
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 9.85598719886212e-05,
+      "loss": 0.2004,
+      "step": 16300
+    },
+    {
+      "epoch": 4.35,
+      "eval_loss": 0.46713659167289734,
+      "eval_runtime": 18.8846,
+      "eval_samples_per_second": 4.713,
+      "eval_wer": 0.3880597014925373,
+      "step": 16300
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 9.855098230953863e-05,
+      "loss": 0.1946,
+      "step": 16400
+    },
+    {
+      "epoch": 4.37,
+      "eval_loss": 0.4721645414829254,
+      "eval_runtime": 18.7894,
+      "eval_samples_per_second": 4.737,
+      "eval_wer": 0.3894165535956581,
+      "step": 16400
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 9.854209263045605e-05,
+      "loss": 0.1887,
+      "step": 16500
+    },
+    {
+      "epoch": 4.4,
+      "eval_loss": 0.47203660011291504,
+      "eval_runtime": 18.9104,
+      "eval_samples_per_second": 4.706,
+      "eval_wer": 0.3853459972862958,
+      "step": 16500
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 9.853320295137345e-05,
+      "loss": 0.1855,
+      "step": 16600
+    },
+    {
+      "epoch": 4.43,
+      "eval_loss": 0.4632361829280853,
+      "eval_runtime": 18.7781,
+      "eval_samples_per_second": 4.74,
+      "eval_wer": 0.38263229308005425,
+      "step": 16600
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 9.852431327229087e-05,
+      "loss": 0.2013,
+      "step": 16700
+    },
+    {
+      "epoch": 4.45,
+      "eval_loss": 0.490334153175354,
+      "eval_runtime": 18.7724,
+      "eval_samples_per_second": 4.741,
+      "eval_wer": 0.3880597014925373,
+      "step": 16700
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 9.85154235932083e-05,
+      "loss": 0.1935,
+      "step": 16800
+    },
+    {
+      "epoch": 4.48,
+      "eval_loss": 0.4184252619743347,
+      "eval_runtime": 18.9885,
+      "eval_samples_per_second": 4.687,
+      "eval_wer": 0.378561736770692,
+      "step": 16800
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 9.850653391412572e-05,
+      "loss": 0.1958,
+      "step": 16900
+    },
+    {
+      "epoch": 4.51,
+      "eval_loss": 0.4693225026130676,
+      "eval_runtime": 18.7677,
+      "eval_samples_per_second": 4.742,
+      "eval_wer": 0.383989145183175,
+      "step": 16900
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 9.849764423504312e-05,
+      "loss": 0.2093,
+      "step": 17000
+    },
+    {
+      "epoch": 4.53,
+      "eval_loss": 0.40667590498924255,
+      "eval_runtime": 18.7231,
+      "eval_samples_per_second": 4.753,
+      "eval_wer": 0.39755766621438265,
+      "step": 17000
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 9.848875455596054e-05,
+      "loss": 0.1918,
+      "step": 17100
+    },
+    {
+      "epoch": 4.56,
+      "eval_loss": 0.4672726094722748,
+      "eval_runtime": 18.8205,
+      "eval_samples_per_second": 4.729,
+      "eval_wer": 0.373134328358209,
+      "step": 17100
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 9.847986487687795e-05,
+      "loss": 0.1962,
+      "step": 17200
+    },
+    {
+      "epoch": 4.59,
+      "eval_loss": 0.4968659579753876,
+      "eval_runtime": 18.8945,
+      "eval_samples_per_second": 4.71,
+      "eval_wer": 0.3921302578018996,
+      "step": 17200
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 9.847097519779537e-05,
+      "loss": 0.1966,
+      "step": 17300
+    },
+    {
+      "epoch": 4.61,
+      "eval_loss": 0.44510090351104736,
+      "eval_runtime": 18.829,
+      "eval_samples_per_second": 4.727,
+      "eval_wer": 0.378561736770692,
+      "step": 17300
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 9.846208551871278e-05,
+      "loss": 0.204,
+      "step": 17400
+    },
+    {
+      "epoch": 4.64,
+      "eval_loss": 0.47826775908470154,
+      "eval_runtime": 18.6704,
+      "eval_samples_per_second": 4.767,
+      "eval_wer": 0.3894165535956581,
+      "step": 17400
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 9.84531958396302e-05,
+      "loss": 0.2062,
+      "step": 17500
+    },
+    {
+      "epoch": 4.67,
+      "eval_loss": 0.49247556924819946,
+      "eval_runtime": 18.8407,
+      "eval_samples_per_second": 4.724,
+      "eval_wer": 0.37991858887381275,
+      "step": 17500
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 9.84443061605476e-05,
+      "loss": 0.2184,
+      "step": 17600
+    },
+    {
+      "epoch": 4.69,
+      "eval_loss": 0.39191409945487976,
+      "eval_runtime": 18.9316,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.37991858887381275,
+      "step": 17600
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 9.843541648146503e-05,
+      "loss": 0.1794,
+      "step": 17700
+    },
+    {
+      "epoch": 4.72,
+      "eval_loss": 0.4386317729949951,
+      "eval_runtime": 18.9393,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.378561736770692,
+      "step": 17700
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 9.842652680238243e-05,
+      "loss": 0.2079,
+      "step": 17800
+    },
+    {
+      "epoch": 4.75,
+      "eval_loss": 0.4679855704307556,
+      "eval_runtime": 18.9374,
+      "eval_samples_per_second": 4.7,
+      "eval_wer": 0.39620081411126185,
+      "step": 17800
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 9.841763712329985e-05,
+      "loss": 0.1873,
+      "step": 17900
+    },
+    {
+      "epoch": 4.77,
+      "eval_loss": 0.3962444067001343,
+      "eval_runtime": 19.0239,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.373134328358209,
+      "step": 17900
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 9.840874744421726e-05,
+      "loss": 0.1983,
+      "step": 18000
+    },
+    {
+      "epoch": 4.8,
+      "eval_loss": 0.4402616620063782,
+      "eval_runtime": 18.8476,
+      "eval_samples_per_second": 4.722,
+      "eval_wer": 0.3880597014925373,
+      "step": 18000
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 9.839985776513468e-05,
+      "loss": 0.2073,
+      "step": 18100
+    },
+    {
+      "epoch": 4.83,
+      "eval_loss": 0.4626671075820923,
+      "eval_runtime": 18.902,
+      "eval_samples_per_second": 4.709,
+      "eval_wer": 0.3921302578018996,
+      "step": 18100
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 9.839096808605209e-05,
+      "loss": 0.197,
+      "step": 18200
+    },
+    {
+      "epoch": 4.85,
+      "eval_loss": 0.4363816976547241,
+      "eval_runtime": 18.7853,
+      "eval_samples_per_second": 4.738,
+      "eval_wer": 0.40027137042062416,
+      "step": 18200
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 9.838207840696951e-05,
+      "loss": 0.1914,
+      "step": 18300
+    },
+    {
+      "epoch": 4.88,
+      "eval_loss": 0.4775172770023346,
+      "eval_runtime": 18.9008,
+      "eval_samples_per_second": 4.709,
+      "eval_wer": 0.4056987788331072,
+      "step": 18300
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 9.837318872788693e-05,
+      "loss": 0.2045,
+      "step": 18400
+    },
+    {
+      "epoch": 4.91,
+      "eval_loss": 0.4852813482284546,
+      "eval_runtime": 18.9611,
+      "eval_samples_per_second": 4.694,
+      "eval_wer": 0.3894165535956581,
+      "step": 18400
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 9.836429904880434e-05,
+      "loss": 0.2032,
+      "step": 18500
+    },
+    {
+      "epoch": 4.93,
+      "eval_loss": 0.4517078697681427,
+      "eval_runtime": 19.0005,
+      "eval_samples_per_second": 4.684,
+      "eval_wer": 0.4016282225237449,
+      "step": 18500
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 9.835540936972176e-05,
+      "loss": 0.2019,
+      "step": 18600
+    },
+    {
+      "epoch": 4.96,
+      "eval_loss": 0.45572879910469055,
+      "eval_runtime": 19.0507,
+      "eval_samples_per_second": 4.672,
+      "eval_wer": 0.3989145183175034,
+      "step": 18600
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 9.834651969063917e-05,
+      "loss": 0.1906,
+      "step": 18700
+    },
+    {
+      "epoch": 4.99,
+      "eval_loss": 0.44796085357666016,
+      "eval_runtime": 18.984,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.3921302578018996,
+      "step": 18700
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 9.833763001155659e-05,
+      "loss": 0.1978,
+      "step": 18800
+    },
+    {
+      "epoch": 5.01,
+      "eval_loss": 0.4583187699317932,
+      "eval_runtime": 18.8739,
+      "eval_samples_per_second": 4.716,
+      "eval_wer": 0.37449118046132973,
+      "step": 18800
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 9.8328740332474e-05,
+      "loss": 0.1697,
+      "step": 18900
+    },
+    {
+      "epoch": 5.04,
+      "eval_loss": 0.48077714443206787,
+      "eval_runtime": 18.862,
+      "eval_samples_per_second": 4.718,
+      "eval_wer": 0.3894165535956581,
+      "step": 18900
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 9.831985065339143e-05,
+      "loss": 0.1757,
+      "step": 19000
+    },
+    {
+      "epoch": 5.07,
+      "eval_loss": 0.4542636573314667,
+      "eval_runtime": 19.1516,
+      "eval_samples_per_second": 4.647,
+      "eval_wer": 0.3758480325644505,
+      "step": 19000
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 9.831096097430883e-05,
+      "loss": 0.184,
+      "step": 19100
+    },
+    {
+      "epoch": 5.09,
+      "eval_loss": 0.4621862471103668,
+      "eval_runtime": 18.9225,
+      "eval_samples_per_second": 4.703,
+      "eval_wer": 0.37991858887381275,
+      "step": 19100
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 9.830207129522626e-05,
+      "loss": 0.1759,
+      "step": 19200
+    },
+    {
+      "epoch": 5.12,
+      "eval_loss": 0.5149765610694885,
+      "eval_runtime": 18.8482,
+      "eval_samples_per_second": 4.722,
+      "eval_wer": 0.3921302578018996,
+      "step": 19200
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 9.829318161614366e-05,
+      "loss": 0.1835,
+      "step": 19300
+    },
+    {
+      "epoch": 5.15,
+      "eval_loss": 0.5005657076835632,
+      "eval_runtime": 18.8498,
+      "eval_samples_per_second": 4.722,
+      "eval_wer": 0.37177747625508817,
+      "step": 19300
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 9.828429193706108e-05,
+      "loss": 0.1743,
+      "step": 19400
+    },
+    {
+      "epoch": 5.17,
+      "eval_loss": 0.5164467096328735,
+      "eval_runtime": 18.9088,
+      "eval_samples_per_second": 4.707,
+      "eval_wer": 0.3812754409769335,
+      "step": 19400
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 9.827540225797849e-05,
+      "loss": 0.1823,
+      "step": 19500
+    },
+    {
+      "epoch": 5.2,
+      "eval_loss": 0.5010002851486206,
+      "eval_runtime": 18.9784,
+      "eval_samples_per_second": 4.69,
+      "eval_wer": 0.39077340569877883,
+      "step": 19500
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 9.826651257889591e-05,
+      "loss": 0.1923,
+      "step": 19600
+    },
+    {
+      "epoch": 5.23,
+      "eval_loss": 0.43779751658439636,
+      "eval_runtime": 18.8458,
+      "eval_samples_per_second": 4.723,
+      "eval_wer": 0.39348710990502034,
+      "step": 19600
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 9.825762289981332e-05,
+      "loss": 0.178,
+      "step": 19700
+    },
+    {
+      "epoch": 5.25,
+      "eval_loss": 0.4191744923591614,
+      "eval_runtime": 19.1842,
+      "eval_samples_per_second": 4.639,
+      "eval_wer": 0.3758480325644505,
+      "step": 19700
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 9.824873322073074e-05,
+      "loss": 0.1784,
+      "step": 19800
+    },
+    {
+      "epoch": 5.28,
+      "eval_loss": 0.47597554326057434,
+      "eval_runtime": 18.7659,
+      "eval_samples_per_second": 4.743,
+      "eval_wer": 0.39348710990502034,
+      "step": 19800
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 9.823984354164815e-05,
+      "loss": 0.175,
+      "step": 19900
+    },
+    {
+      "epoch": 5.31,
+      "eval_loss": 0.39441633224487305,
+      "eval_runtime": 19.1297,
+      "eval_samples_per_second": 4.652,
+      "eval_wer": 0.378561736770692,
+      "step": 19900
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 9.823095386256557e-05,
+      "loss": 0.1788,
+      "step": 20000
+    },
+    {
+      "epoch": 5.33,
+      "eval_loss": 0.3814501464366913,
+      "eval_runtime": 18.9356,
+      "eval_samples_per_second": 4.7,
+      "eval_wer": 0.373134328358209,
+      "step": 20000
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 9.822206418348297e-05,
+      "loss": 0.1902,
+      "step": 20100
+    },
+    {
+      "epoch": 5.36,
+      "eval_loss": 0.3846692740917206,
+      "eval_runtime": 18.9756,
+      "eval_samples_per_second": 4.69,
+      "eval_wer": 0.37720488466757124,
+      "step": 20100
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 9.82131745044004e-05,
+      "loss": 0.1653,
+      "step": 20200
+    },
+    {
+      "epoch": 5.39,
+      "eval_loss": 0.4306989908218384,
+      "eval_runtime": 18.821,
+      "eval_samples_per_second": 4.729,
+      "eval_wer": 0.37449118046132973,
+      "step": 20200
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 9.820428482531781e-05,
+      "loss": 0.1924,
+      "step": 20300
+    },
+    {
+      "epoch": 5.41,
+      "eval_loss": 0.3729378879070282,
+      "eval_runtime": 18.9831,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.38670284938941657,
+      "step": 20300
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 9.819539514623522e-05,
+      "loss": 0.1787,
+      "step": 20400
+    },
+    {
+      "epoch": 5.44,
+      "eval_loss": 0.4352020025253296,
+      "eval_runtime": 19.1388,
+      "eval_samples_per_second": 4.65,
+      "eval_wer": 0.40705563093622793,
+      "step": 20400
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 9.818650546715264e-05,
+      "loss": 0.1883,
+      "step": 20500
+    },
+    {
+      "epoch": 5.47,
+      "eval_loss": 0.42689651250839233,
+      "eval_runtime": 18.9631,
+      "eval_samples_per_second": 4.693,
+      "eval_wer": 0.383989145183175,
+      "step": 20500
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 9.817761578807005e-05,
+      "loss": 0.1849,
+      "step": 20600
+    },
+    {
+      "epoch": 5.49,
+      "eval_loss": 0.4585600793361664,
+      "eval_runtime": 18.8897,
+      "eval_samples_per_second": 4.712,
+      "eval_wer": 0.39620081411126185,
+      "step": 20600
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 9.816872610898747e-05,
+      "loss": 0.1753,
+      "step": 20700
+    },
+    {
+      "epoch": 5.52,
+      "eval_loss": 0.38371387124061584,
+      "eval_runtime": 19.0346,
+      "eval_samples_per_second": 4.676,
+      "eval_wer": 0.36906377204884666,
+      "step": 20700
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 9.815983642990488e-05,
+      "loss": 0.1706,
+      "step": 20800
+    },
+    {
+      "epoch": 5.55,
+      "eval_loss": 0.40636616945266724,
+      "eval_runtime": 18.8791,
+      "eval_samples_per_second": 4.714,
+      "eval_wer": 0.3948439620081411,
+      "step": 20800
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 9.81509467508223e-05,
+      "loss": 0.1887,
+      "step": 20900
+    },
+    {
+      "epoch": 5.57,
+      "eval_loss": 0.4221135675907135,
+      "eval_runtime": 19.0591,
+      "eval_samples_per_second": 4.67,
+      "eval_wer": 0.3853459972862958,
+      "step": 20900
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 9.814205707173972e-05,
+      "loss": 0.1888,
+      "step": 21000
+    },
+    {
+      "epoch": 5.6,
+      "eval_loss": 0.4211312532424927,
+      "eval_runtime": 18.9281,
+      "eval_samples_per_second": 4.702,
+      "eval_wer": 0.3894165535956581,
+      "step": 21000
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 9.813316739265713e-05,
+      "loss": 0.1836,
+      "step": 21100
+    },
+    {
+      "epoch": 5.63,
+      "eval_loss": 0.40096545219421387,
+      "eval_runtime": 19.1784,
+      "eval_samples_per_second": 4.641,
+      "eval_wer": 0.40705563093622793,
+      "step": 21100
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 9.812427771357455e-05,
+      "loss": 0.1769,
+      "step": 21200
+    },
+    {
+      "epoch": 5.65,
+      "eval_loss": 0.45822688937187195,
+      "eval_runtime": 18.9665,
+      "eval_samples_per_second": 4.692,
+      "eval_wer": 0.38263229308005425,
+      "step": 21200
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 9.811538803449197e-05,
+      "loss": 0.1765,
+      "step": 21300
+    },
+    {
+      "epoch": 5.68,
+      "eval_loss": 0.3974924087524414,
+      "eval_runtime": 18.9732,
+      "eval_samples_per_second": 4.691,
+      "eval_wer": 0.4056987788331072,
+      "step": 21300
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 9.810649835540937e-05,
+      "loss": 0.1898,
+      "step": 21400
+    },
+    {
+      "epoch": 5.71,
+      "eval_loss": 0.47260570526123047,
+      "eval_runtime": 18.8966,
+      "eval_samples_per_second": 4.71,
+      "eval_wer": 0.3921302578018996,
+      "step": 21400
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 9.80976086763268e-05,
+      "loss": 0.172,
+      "step": 21500
+    },
+    {
+      "epoch": 5.73,
+      "eval_loss": 0.47722673416137695,
+      "eval_runtime": 19.1058,
+      "eval_samples_per_second": 4.658,
+      "eval_wer": 0.4016282225237449,
+      "step": 21500
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 9.80887189972442e-05,
+      "loss": 0.1781,
+      "step": 21600
+    },
+    {
+      "epoch": 5.76,
+      "eval_loss": 0.5133776664733887,
+      "eval_runtime": 19.0038,
+      "eval_samples_per_second": 4.683,
+      "eval_wer": 0.3948439620081411,
+      "step": 21600
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 9.807982931816162e-05,
+      "loss": 0.1878,
+      "step": 21700
+    },
+    {
+      "epoch": 5.79,
+      "eval_loss": 0.4732833504676819,
+      "eval_runtime": 19.0195,
+      "eval_samples_per_second": 4.679,
+      "eval_wer": 0.37991858887381275,
+      "step": 21700
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 9.807093963907903e-05,
+      "loss": 0.1745,
+      "step": 21800
+    },
+    {
+      "epoch": 5.81,
+      "eval_loss": 0.5174434185028076,
+      "eval_runtime": 19.1422,
+      "eval_samples_per_second": 4.649,
+      "eval_wer": 0.39348710990502034,
+      "step": 21800
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 9.806204995999645e-05,
+      "loss": 0.1708,
+      "step": 21900
+    },
+    {
+      "epoch": 5.84,
+      "eval_loss": 0.5051631927490234,
+      "eval_runtime": 18.9207,
+      "eval_samples_per_second": 4.704,
+      "eval_wer": 0.3880597014925373,
+      "step": 21900
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 9.805316028091386e-05,
+      "loss": 0.1859,
+      "step": 22000
+    },
+    {
+      "epoch": 5.87,
+      "eval_loss": 0.45525211095809937,
+      "eval_runtime": 18.9528,
+      "eval_samples_per_second": 4.696,
+      "eval_wer": 0.37991858887381275,
+      "step": 22000
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 9.804427060183128e-05,
+      "loss": 0.1831,
+      "step": 22100
+    },
+    {
+      "epoch": 5.89,
+      "eval_loss": 0.4619895815849304,
+      "eval_runtime": 18.9216,
+      "eval_samples_per_second": 4.704,
+      "eval_wer": 0.39755766621438265,
+      "step": 22100
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 9.80353809227487e-05,
+      "loss": 0.1954,
+      "step": 22200
+    },
+    {
+      "epoch": 5.92,
+      "eval_loss": 0.4404244124889374,
+      "eval_runtime": 19.252,
+      "eval_samples_per_second": 4.623,
+      "eval_wer": 0.3853459972862958,
+      "step": 22200
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 9.80264912436661e-05,
+      "loss": 0.1835,
+      "step": 22300
+    },
+    {
+      "epoch": 5.95,
+      "eval_loss": 0.43536993861198425,
+      "eval_runtime": 19.2551,
+      "eval_samples_per_second": 4.622,
+      "eval_wer": 0.3853459972862958,
+      "step": 22300
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 9.801760156458353e-05,
+      "loss": 0.1811,
+      "step": 22400
+    },
+    {
+      "epoch": 5.97,
+      "eval_loss": 0.49131497740745544,
+      "eval_runtime": 19.2506,
+      "eval_samples_per_second": 4.623,
+      "eval_wer": 0.39077340569877883,
+      "step": 22400
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 9.800871188550093e-05,
+      "loss": 0.1727,
+      "step": 22500
+    },
+    {
+      "epoch": 6.0,
+      "eval_loss": 0.44274598360061646,
+      "eval_runtime": 19.7324,
+      "eval_samples_per_second": 4.51,
+      "eval_wer": 0.39348710990502034,
+      "step": 22500
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 9.799982220641835e-05,
+      "loss": 0.1557,
+      "step": 22600
+    },
+    {
+      "epoch": 6.03,
+      "eval_loss": 0.5067463517189026,
+      "eval_runtime": 19.6903,
+      "eval_samples_per_second": 4.52,
+      "eval_wer": 0.39348710990502034,
+      "step": 22600
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 9.799093252733576e-05,
+      "loss": 0.1676,
+      "step": 22700
+    },
+    {
+      "epoch": 6.05,
+      "eval_loss": 0.43477436900138855,
+      "eval_runtime": 19.1478,
+      "eval_samples_per_second": 4.648,
+      "eval_wer": 0.37449118046132973,
+      "step": 22700
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 9.798204284825318e-05,
+      "loss": 0.1672,
+      "step": 22800
+    },
+    {
+      "epoch": 6.08,
+      "eval_loss": 0.4951562285423279,
+      "eval_runtime": 19.1813,
+      "eval_samples_per_second": 4.64,
+      "eval_wer": 0.4151967435549525,
+      "step": 22800
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 9.797315316917059e-05,
+      "loss": 0.1737,
+      "step": 22900
+    },
+    {
+      "epoch": 6.11,
+      "eval_loss": 0.5281330943107605,
+      "eval_runtime": 19.4497,
+      "eval_samples_per_second": 4.576,
+      "eval_wer": 0.3948439620081411,
+      "step": 22900
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 9.796426349008801e-05,
+      "loss": 0.1544,
+      "step": 23000
+    },
+    {
+      "epoch": 6.13,
+      "eval_loss": 0.5187743306159973,
+      "eval_runtime": 18.9388,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.3989145183175034,
+      "step": 23000
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 9.795537381100542e-05,
+      "loss": 0.1733,
+      "step": 23100
+    },
+    {
+      "epoch": 6.16,
+      "eval_loss": 0.45824718475341797,
+      "eval_runtime": 19.2709,
+      "eval_samples_per_second": 4.618,
+      "eval_wer": 0.39620081411126185,
+      "step": 23100
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 9.794648413192284e-05,
+      "loss": 0.1715,
+      "step": 23200
+    },
+    {
+      "epoch": 6.19,
+      "eval_loss": 0.4227101504802704,
+      "eval_runtime": 19.2061,
+      "eval_samples_per_second": 4.634,
+      "eval_wer": 0.3853459972862958,
+      "step": 23200
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 9.793759445284026e-05,
+      "loss": 0.1758,
+      "step": 23300
+    },
+    {
+      "epoch": 6.21,
+      "eval_loss": 0.5284545421600342,
+      "eval_runtime": 19.1393,
+      "eval_samples_per_second": 4.65,
+      "eval_wer": 0.39077340569877883,
+      "step": 23300
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 9.792870477375768e-05,
+      "loss": 0.1679,
+      "step": 23400
+    },
+    {
+      "epoch": 6.24,
+      "eval_loss": 0.431808739900589,
+      "eval_runtime": 19.2888,
+      "eval_samples_per_second": 4.614,
+      "eval_wer": 0.3853459972862958,
+      "step": 23400
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 9.791981509467509e-05,
+      "loss": 0.163,
+      "step": 23500
+    },
+    {
+      "epoch": 6.27,
+      "eval_loss": 0.4140615165233612,
+      "eval_runtime": 19.0526,
+      "eval_samples_per_second": 4.671,
+      "eval_wer": 0.39077340569877883,
+      "step": 23500
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 9.791092541559251e-05,
+      "loss": 0.1591,
+      "step": 23600
+    },
+    {
+      "epoch": 6.29,
+      "eval_loss": 0.44218435883522034,
+      "eval_runtime": 19.0588,
+      "eval_samples_per_second": 4.67,
+      "eval_wer": 0.36906377204884666,
+      "step": 23600
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 9.790203573650991e-05,
+      "loss": 0.1653,
+      "step": 23700
+    },
+    {
+      "epoch": 6.32,
+      "eval_loss": 0.4993055462837219,
+      "eval_runtime": 19.3253,
+      "eval_samples_per_second": 4.605,
+      "eval_wer": 0.39755766621438265,
+      "step": 23700
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 9.789314605742734e-05,
+      "loss": 0.1569,
+      "step": 23800
+    },
+    {
+      "epoch": 6.35,
+      "eval_loss": 0.4586738348007202,
+      "eval_runtime": 19.0305,
+      "eval_samples_per_second": 4.677,
+      "eval_wer": 0.39755766621438265,
+      "step": 23800
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 9.788425637834476e-05,
+      "loss": 0.1603,
+      "step": 23900
+    },
+    {
+      "epoch": 6.37,
+      "eval_loss": 0.4952552318572998,
+      "eval_runtime": 19.0943,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.38670284938941657,
+      "step": 23900
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 9.787536669926216e-05,
+      "loss": 0.1682,
+      "step": 24000
+    },
+    {
+      "epoch": 6.4,
+      "eval_loss": 0.4896041750907898,
+      "eval_runtime": 19.3288,
+      "eval_samples_per_second": 4.605,
+      "eval_wer": 0.39077340569877883,
+      "step": 24000
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 9.786647702017958e-05,
+      "loss": 0.16,
+      "step": 24100
+    },
+    {
+      "epoch": 6.43,
+      "eval_loss": 0.5756000280380249,
+      "eval_runtime": 18.8148,
+      "eval_samples_per_second": 4.73,
+      "eval_wer": 0.3853459972862958,
+      "step": 24100
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 9.785758734109699e-05,
+      "loss": 0.1608,
+      "step": 24200
+    },
+    {
+      "epoch": 6.45,
+      "eval_loss": 0.4900316298007965,
+      "eval_runtime": 19.1765,
+      "eval_samples_per_second": 4.641,
+      "eval_wer": 0.37720488466757124,
+      "step": 24200
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 9.784869766201441e-05,
+      "loss": 0.1742,
+      "step": 24300
+    },
+    {
+      "epoch": 6.48,
+      "eval_loss": 0.4866500198841095,
+      "eval_runtime": 18.8311,
+      "eval_samples_per_second": 4.726,
+      "eval_wer": 0.39077340569877883,
+      "step": 24300
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 9.783980798293182e-05,
+      "loss": 0.1705,
+      "step": 24400
+    },
+    {
+      "epoch": 6.51,
+      "eval_loss": 0.4403529763221741,
+      "eval_runtime": 19.127,
+      "eval_samples_per_second": 4.653,
+      "eval_wer": 0.37720488466757124,
+      "step": 24400
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 9.783091830384924e-05,
+      "loss": 0.1717,
+      "step": 24500
+    },
+    {
+      "epoch": 6.53,
+      "eval_loss": 0.46126067638397217,
+      "eval_runtime": 18.9436,
+      "eval_samples_per_second": 4.698,
+      "eval_wer": 0.37720488466757124,
+      "step": 24500
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 9.782202862476665e-05,
+      "loss": 0.1657,
+      "step": 24600
+    },
+    {
+      "epoch": 6.56,
+      "eval_loss": 0.4159884452819824,
+      "eval_runtime": 19.0307,
+      "eval_samples_per_second": 4.677,
+      "eval_wer": 0.383989145183175,
+      "step": 24600
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 9.781313894568407e-05,
+      "loss": 0.1733,
+      "step": 24700
+    },
+    {
+      "epoch": 6.59,
+      "eval_loss": 0.4532783031463623,
+      "eval_runtime": 19.2877,
+      "eval_samples_per_second": 4.614,
+      "eval_wer": 0.378561736770692,
+      "step": 24700
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 9.780424926660147e-05,
+      "loss": 0.1808,
+      "step": 24800
+    },
+    {
+      "epoch": 6.61,
+      "eval_loss": 0.5106078386306763,
+      "eval_runtime": 19.1695,
+      "eval_samples_per_second": 4.643,
+      "eval_wer": 0.37720488466757124,
+      "step": 24800
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 9.77953595875189e-05,
+      "loss": 0.1672,
+      "step": 24900
+    },
+    {
+      "epoch": 6.64,
+      "eval_loss": 0.5461359620094299,
+      "eval_runtime": 19.1261,
+      "eval_samples_per_second": 4.653,
+      "eval_wer": 0.383989145183175,
+      "step": 24900
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.77864699084363e-05,
+      "loss": 0.1596,
+      "step": 25000
+    },
+    {
+      "epoch": 6.67,
+      "eval_loss": 0.5272344946861267,
+      "eval_runtime": 19.6123,
+      "eval_samples_per_second": 4.538,
+      "eval_wer": 0.37449118046132973,
+      "step": 25000
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 9.777758022935372e-05,
+      "loss": 0.1635,
+      "step": 25100
+    },
+    {
+      "epoch": 6.69,
+      "eval_loss": 0.5185968279838562,
+      "eval_runtime": 19.4571,
+      "eval_samples_per_second": 4.574,
+      "eval_wer": 0.37720488466757124,
+      "step": 25100
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.776869055027113e-05,
+      "loss": 0.1869,
+      "step": 25200
+    },
+    {
+      "epoch": 6.72,
+      "eval_loss": 0.4779977798461914,
+      "eval_runtime": 19.3154,
+      "eval_samples_per_second": 4.608,
+      "eval_wer": 0.37449118046132973,
+      "step": 25200
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.775980087118855e-05,
+      "loss": 0.1867,
+      "step": 25300
+    },
+    {
+      "epoch": 6.75,
+      "eval_loss": 0.49571913480758667,
+      "eval_runtime": 18.9183,
+      "eval_samples_per_second": 4.704,
+      "eval_wer": 0.3894165535956581,
+      "step": 25300
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 9.775091119210597e-05,
+      "loss": 0.1658,
+      "step": 25400
+    },
+    {
+      "epoch": 6.77,
+      "eval_loss": 0.5112097859382629,
+      "eval_runtime": 18.9492,
+      "eval_samples_per_second": 4.697,
+      "eval_wer": 0.3853459972862958,
+      "step": 25400
+    },
+    {
+      "epoch": 6.8,
+      "learning_rate": 9.774202151302339e-05,
+      "loss": 0.152,
+      "step": 25500
+    },
+    {
+      "epoch": 6.8,
+      "eval_loss": 0.4930531680583954,
+      "eval_runtime": 19.1582,
+      "eval_samples_per_second": 4.646,
+      "eval_wer": 0.3853459972862958,
+      "step": 25500
+    },
+    {
+      "epoch": 6.83,
+      "learning_rate": 9.77331318339408e-05,
+      "loss": 0.1644,
+      "step": 25600
+    },
+    {
+      "epoch": 6.83,
+      "eval_loss": 0.43873706459999084,
+      "eval_runtime": 19.1193,
+      "eval_samples_per_second": 4.655,
+      "eval_wer": 0.37177747625508817,
+      "step": 25600
+    },
+    {
+      "epoch": 6.85,
+      "learning_rate": 9.772424215485822e-05,
+      "loss": 0.1629,
+      "step": 25700
+    },
+    {
+      "epoch": 6.85,
+      "eval_loss": 0.4649239778518677,
+      "eval_runtime": 18.8975,
+      "eval_samples_per_second": 4.71,
+      "eval_wer": 0.37720488466757124,
+      "step": 25700
+    },
+    {
+      "epoch": 6.88,
+      "learning_rate": 9.771535247577564e-05,
+      "loss": 0.179,
+      "step": 25800
+    },
+    {
+      "epoch": 6.88,
+      "eval_loss": 0.4687182903289795,
+      "eval_runtime": 18.8458,
+      "eval_samples_per_second": 4.723,
+      "eval_wer": 0.39620081411126185,
+      "step": 25800
+    },
+    {
+      "epoch": 6.91,
+      "learning_rate": 9.770646279669305e-05,
+      "loss": 0.1577,
+      "step": 25900
+    },
+    {
+      "epoch": 6.91,
+      "eval_loss": 0.4671737253665924,
+      "eval_runtime": 19.0069,
+      "eval_samples_per_second": 4.682,
+      "eval_wer": 0.37991858887381275,
+      "step": 25900
+    },
+    {
+      "epoch": 6.93,
+      "learning_rate": 9.769757311761047e-05,
+      "loss": 0.1758,
+      "step": 26000
+    },
+    {
+      "epoch": 6.93,
+      "eval_loss": 0.45876505970954895,
+      "eval_runtime": 18.9976,
+      "eval_samples_per_second": 4.685,
+      "eval_wer": 0.383989145183175,
+      "step": 26000
+    },
+    {
+      "epoch": 6.96,
+      "learning_rate": 9.768868343852787e-05,
+      "loss": 0.1748,
+      "step": 26100
+    },
+    {
+      "epoch": 6.96,
+      "eval_loss": 0.4998129904270172,
+      "eval_runtime": 19.1262,
+      "eval_samples_per_second": 4.653,
+      "eval_wer": 0.4056987788331072,
+      "step": 26100
+    },
+    {
+      "epoch": 6.99,
+      "learning_rate": 9.76797937594453e-05,
+      "loss": 0.1767,
+      "step": 26200
+    },
+    {
+      "epoch": 6.99,
+      "eval_loss": 0.49574416875839233,
+      "eval_runtime": 18.7605,
+      "eval_samples_per_second": 4.744,
+      "eval_wer": 0.40298507462686567,
+      "step": 26200
+    },
+    {
+      "epoch": 7.01,
+      "learning_rate": 9.76709040803627e-05,
+      "loss": 0.1648,
+      "step": 26300
+    },
+    {
+      "epoch": 7.01,
+      "eval_loss": 0.4767066538333893,
+      "eval_runtime": 18.9848,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.3853459972862958,
+      "step": 26300
+    },
+    {
+      "epoch": 7.04,
+      "learning_rate": 9.766201440128012e-05,
+      "loss": 0.15,
+      "step": 26400
+    },
+    {
+      "epoch": 7.04,
+      "eval_loss": 0.5086139440536499,
+      "eval_runtime": 18.9119,
+      "eval_samples_per_second": 4.706,
+      "eval_wer": 0.39077340569877883,
+      "step": 26400
+    },
+    {
+      "epoch": 7.07,
+      "learning_rate": 9.765312472219753e-05,
+      "loss": 0.1457,
+      "step": 26500
+    },
+    {
+      "epoch": 7.07,
+      "eval_loss": 0.6223266124725342,
+      "eval_runtime": 18.9032,
+      "eval_samples_per_second": 4.708,
+      "eval_wer": 0.37449118046132973,
+      "step": 26500
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 9.764423504311495e-05,
+      "loss": 0.1625,
+      "step": 26600
+    },
+    {
+      "epoch": 7.09,
+      "eval_loss": 0.5288943648338318,
+      "eval_runtime": 18.9249,
+      "eval_samples_per_second": 4.703,
+      "eval_wer": 0.3921302578018996,
+      "step": 26600
+    },
+    {
+      "epoch": 7.12,
+      "learning_rate": 9.763534536403236e-05,
+      "loss": 0.1522,
+      "step": 26700
+    },
+    {
+      "epoch": 7.12,
+      "eval_loss": 0.504027783870697,
+      "eval_runtime": 18.9327,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.4056987788331072,
+      "step": 26700
+    },
+    {
+      "epoch": 7.15,
+      "learning_rate": 9.762645568494978e-05,
+      "loss": 0.1483,
+      "step": 26800
+    },
+    {
+      "epoch": 7.15,
+      "eval_loss": 0.4940141439437866,
+      "eval_runtime": 19.0342,
+      "eval_samples_per_second": 4.676,
+      "eval_wer": 0.40027137042062416,
+      "step": 26800
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 9.761756600586719e-05,
+      "loss": 0.1594,
+      "step": 26900
+    },
+    {
+      "epoch": 7.17,
+      "eval_loss": 0.4934203624725342,
+      "eval_runtime": 18.8191,
+      "eval_samples_per_second": 4.729,
+      "eval_wer": 0.3948439620081411,
+      "step": 26900
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 9.76086763267846e-05,
+      "loss": 0.1465,
+      "step": 27000
+    },
+    {
+      "epoch": 7.2,
+      "eval_loss": 0.4947309195995331,
+      "eval_runtime": 19.0463,
+      "eval_samples_per_second": 4.673,
+      "eval_wer": 0.3989145183175034,
+      "step": 27000
+    },
+    {
+      "epoch": 7.23,
+      "learning_rate": 9.759978664770201e-05,
+      "loss": 0.146,
+      "step": 27100
+    },
+    {
+      "epoch": 7.23,
+      "eval_loss": 0.4862762689590454,
+      "eval_runtime": 19.1467,
+      "eval_samples_per_second": 4.648,
+      "eval_wer": 0.38670284938941657,
+      "step": 27100
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 9.759089696861943e-05,
+      "loss": 0.1601,
+      "step": 27200
+    },
+    {
+      "epoch": 7.25,
+      "eval_loss": 0.44045352935791016,
+      "eval_runtime": 18.9651,
+      "eval_samples_per_second": 4.693,
+      "eval_wer": 0.3677069199457259,
+      "step": 27200
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 9.758200728953684e-05,
+      "loss": 0.1439,
+      "step": 27300
+    },
+    {
+      "epoch": 7.28,
+      "eval_loss": 0.4298516511917114,
+      "eval_runtime": 19.1246,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.378561736770692,
+      "step": 27300
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 9.757311761045426e-05,
+      "loss": 0.1584,
+      "step": 27400
+    },
+    {
+      "epoch": 7.31,
+      "eval_loss": 0.44751495122909546,
+      "eval_runtime": 19.1338,
+      "eval_samples_per_second": 4.651,
+      "eval_wer": 0.3677069199457259,
+      "step": 27400
+    },
+    {
+      "epoch": 7.33,
+      "learning_rate": 9.756422793137168e-05,
+      "loss": 0.1671,
+      "step": 27500
+    },
+    {
+      "epoch": 7.33,
+      "eval_loss": 0.4183838367462158,
+      "eval_runtime": 19.0601,
+      "eval_samples_per_second": 4.669,
+      "eval_wer": 0.3649932157394844,
+      "step": 27500
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 9.755533825228909e-05,
+      "loss": 0.1745,
+      "step": 27600
+    },
+    {
+      "epoch": 7.36,
+      "eval_loss": 0.39715662598609924,
+      "eval_runtime": 18.8454,
+      "eval_samples_per_second": 4.723,
+      "eval_wer": 0.36906377204884666,
+      "step": 27600
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 9.754644857320651e-05,
+      "loss": 0.1438,
+      "step": 27700
+    },
+    {
+      "epoch": 7.39,
+      "eval_loss": 0.4280313551425934,
+      "eval_runtime": 19.2376,
+      "eval_samples_per_second": 4.626,
+      "eval_wer": 0.37720488466757124,
+      "step": 27700
+    },
+    {
+      "epoch": 7.41,
+      "learning_rate": 9.753755889412393e-05,
+      "loss": 0.1468,
+      "step": 27800
+    },
+    {
+      "epoch": 7.41,
+      "eval_loss": 0.5265226364135742,
+      "eval_runtime": 19.1478,
+      "eval_samples_per_second": 4.648,
+      "eval_wer": 0.37177747625508817,
+      "step": 27800
+    },
+    {
+      "epoch": 7.44,
+      "learning_rate": 9.752866921504135e-05,
+      "loss": 0.1523,
+      "step": 27900
+    },
+    {
+      "epoch": 7.44,
+      "eval_loss": 0.5133256316184998,
+      "eval_runtime": 19.0216,
+      "eval_samples_per_second": 4.679,
+      "eval_wer": 0.383989145183175,
+      "step": 27900
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 9.751977953595876e-05,
+      "loss": 0.1527,
+      "step": 28000
+    },
+    {
+      "epoch": 7.47,
+      "eval_loss": 0.4403773248195648,
+      "eval_runtime": 19.0958,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.3758480325644505,
+      "step": 28000
+    },
+    {
+      "epoch": 7.49,
+      "learning_rate": 9.751088985687618e-05,
+      "loss": 0.165,
+      "step": 28100
+    },
+    {
+      "epoch": 7.49,
+      "eval_loss": 0.5208707451820374,
+      "eval_runtime": 19.031,
+      "eval_samples_per_second": 4.677,
+      "eval_wer": 0.3622795115332429,
+      "step": 28100
+    },
+    {
+      "epoch": 7.52,
+      "learning_rate": 9.750200017779359e-05,
+      "loss": 0.1632,
+      "step": 28200
+    },
+    {
+      "epoch": 7.52,
+      "eval_loss": 0.4097565710544586,
+      "eval_runtime": 18.9804,
+      "eval_samples_per_second": 4.689,
+      "eval_wer": 0.36635006784260515,
+      "step": 28200
+    },
+    {
+      "epoch": 7.55,
+      "learning_rate": 9.749311049871101e-05,
+      "loss": 0.1528,
+      "step": 28300
+    },
+    {
+      "epoch": 7.55,
+      "eval_loss": 0.4614786207675934,
+      "eval_runtime": 19.1726,
+      "eval_samples_per_second": 4.642,
+      "eval_wer": 0.378561736770692,
+      "step": 28300
+    },
+    {
+      "epoch": 7.57,
+      "learning_rate": 9.748422081962841e-05,
+      "loss": 0.164,
+      "step": 28400
+    },
+    {
+      "epoch": 7.57,
+      "eval_loss": 0.4744202196598053,
+      "eval_runtime": 18.8568,
+      "eval_samples_per_second": 4.72,
+      "eval_wer": 0.3812754409769335,
+      "step": 28400
+    },
+    {
+      "epoch": 7.6,
+      "learning_rate": 9.747533114054584e-05,
+      "loss": 0.1705,
+      "step": 28500
+    },
+    {
+      "epoch": 7.6,
+      "eval_loss": 0.5456199049949646,
+      "eval_runtime": 18.974,
+      "eval_samples_per_second": 4.691,
+      "eval_wer": 0.37449118046132973,
+      "step": 28500
+    },
+    {
+      "epoch": 7.63,
+      "learning_rate": 9.746644146146324e-05,
+      "loss": 0.1585,
+      "step": 28600
+    },
+    {
+      "epoch": 7.63,
+      "eval_loss": 0.5013868808746338,
+      "eval_runtime": 18.9998,
+      "eval_samples_per_second": 4.684,
+      "eval_wer": 0.3758480325644505,
+      "step": 28600
+    },
+    {
+      "epoch": 7.65,
+      "learning_rate": 9.745755178238066e-05,
+      "loss": 0.144,
+      "step": 28700
+    },
+    {
+      "epoch": 7.65,
+      "eval_loss": 0.5348823070526123,
+      "eval_runtime": 19.3804,
+      "eval_samples_per_second": 4.592,
+      "eval_wer": 0.3812754409769335,
+      "step": 28700
+    },
+    {
+      "epoch": 7.68,
+      "learning_rate": 9.744866210329807e-05,
+      "loss": 0.1697,
+      "step": 28800
+    },
+    {
+      "epoch": 7.68,
+      "eval_loss": 0.5246402621269226,
+      "eval_runtime": 18.9415,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.3649932157394844,
+      "step": 28800
+    },
+    {
+      "epoch": 7.71,
+      "learning_rate": 9.743977242421549e-05,
+      "loss": 0.1523,
+      "step": 28900
+    },
+    {
+      "epoch": 7.71,
+      "eval_loss": 0.5185402035713196,
+      "eval_runtime": 19.1314,
+      "eval_samples_per_second": 4.652,
+      "eval_wer": 0.3568521031207598,
+      "step": 28900
+    },
+    {
+      "epoch": 7.73,
+      "learning_rate": 9.74308827451329e-05,
+      "loss": 0.147,
+      "step": 29000
+    },
+    {
+      "epoch": 7.73,
+      "eval_loss": 0.4967338442802429,
+      "eval_runtime": 19.1765,
+      "eval_samples_per_second": 4.641,
+      "eval_wer": 0.37177747625508817,
+      "step": 29000
+    },
+    {
+      "epoch": 7.76,
+      "learning_rate": 9.742199306605032e-05,
+      "loss": 0.1457,
+      "step": 29100
+    },
+    {
+      "epoch": 7.76,
+      "eval_loss": 0.5241490006446838,
+      "eval_runtime": 19.6732,
+      "eval_samples_per_second": 4.524,
+      "eval_wer": 0.3677069199457259,
+      "step": 29100
+    },
+    {
+      "epoch": 7.79,
+      "learning_rate": 9.741310338696773e-05,
+      "loss": 0.1616,
+      "step": 29200
+    },
+    {
+      "epoch": 7.79,
+      "eval_loss": 0.49654269218444824,
+      "eval_runtime": 19.0618,
+      "eval_samples_per_second": 4.669,
+      "eval_wer": 0.3677069199457259,
+      "step": 29200
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 9.740421370788515e-05,
+      "loss": 0.1564,
+      "step": 29300
+    },
+    {
+      "epoch": 7.81,
+      "eval_loss": 0.4634263217449188,
+      "eval_runtime": 19.0562,
+      "eval_samples_per_second": 4.67,
+      "eval_wer": 0.34328358208955223,
+      "step": 29300
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 9.739532402880257e-05,
+      "loss": 0.1621,
+      "step": 29400
+    },
+    {
+      "epoch": 7.84,
+      "eval_loss": 0.4548163115978241,
+      "eval_runtime": 19.1231,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.35956580732700133,
+      "step": 29400
+    },
+    {
+      "epoch": 7.87,
+      "learning_rate": 9.738643434971997e-05,
+      "loss": 0.15,
+      "step": 29500
+    },
+    {
+      "epoch": 7.87,
+      "eval_loss": 0.48281076550483704,
+      "eval_runtime": 18.9524,
+      "eval_samples_per_second": 4.696,
+      "eval_wer": 0.3677069199457259,
+      "step": 29500
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 9.73775446706374e-05,
+      "loss": 0.1548,
+      "step": 29600
+    },
+    {
+      "epoch": 7.89,
+      "eval_loss": 0.48112940788269043,
+      "eval_runtime": 19.1999,
+      "eval_samples_per_second": 4.635,
+      "eval_wer": 0.35956580732700133,
+      "step": 29600
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 9.73686549915548e-05,
+      "loss": 0.1523,
+      "step": 29700
+    },
+    {
+      "epoch": 7.92,
+      "eval_loss": 0.44887080788612366,
+      "eval_runtime": 19.1733,
+      "eval_samples_per_second": 4.642,
+      "eval_wer": 0.36092265943012214,
+      "step": 29700
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 9.735976531247222e-05,
+      "loss": 0.1647,
+      "step": 29800
+    },
+    {
+      "epoch": 7.95,
+      "eval_loss": 0.4797247052192688,
+      "eval_runtime": 19.189,
+      "eval_samples_per_second": 4.638,
+      "eval_wer": 0.36092265943012214,
+      "step": 29800
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 9.735087563338964e-05,
+      "loss": 0.1684,
+      "step": 29900
+    },
+    {
+      "epoch": 7.97,
+      "eval_loss": 0.48400458693504333,
+      "eval_runtime": 19.1425,
+      "eval_samples_per_second": 4.649,
+      "eval_wer": 0.36635006784260515,
+      "step": 29900
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 9.734198595430706e-05,
+      "loss": 0.1563,
+      "step": 30000
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": 0.4427766799926758,
+      "eval_runtime": 19.215,
+      "eval_samples_per_second": 4.632,
+      "eval_wer": 0.36635006784260515,
+      "step": 30000
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 9.733309627522447e-05,
+      "loss": 0.14,
+      "step": 30100
+    },
+    {
+      "epoch": 8.03,
+      "eval_loss": 0.45060795545578003,
+      "eval_runtime": 19.1737,
+      "eval_samples_per_second": 4.642,
+      "eval_wer": 0.36635006784260515,
+      "step": 30100
+    },
+    {
+      "epoch": 8.05,
+      "learning_rate": 9.732420659614189e-05,
+      "loss": 0.1343,
+      "step": 30200
+    },
+    {
+      "epoch": 8.05,
+      "eval_loss": 0.46462002396583557,
+      "eval_runtime": 19.1567,
+      "eval_samples_per_second": 4.646,
+      "eval_wer": 0.36635006784260515,
+      "step": 30200
+    },
+    {
+      "epoch": 8.08,
+      "learning_rate": 9.73153169170593e-05,
+      "loss": 0.1285,
+      "step": 30300
+    },
+    {
+      "epoch": 8.08,
+      "eval_loss": 0.5208112001419067,
+      "eval_runtime": 19.4158,
+      "eval_samples_per_second": 4.584,
+      "eval_wer": 0.36363636363636365,
+      "step": 30300
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 9.730642723797672e-05,
+      "loss": 0.1365,
+      "step": 30400
+    },
+    {
+      "epoch": 8.11,
+      "eval_loss": 0.45026347041130066,
+      "eval_runtime": 19.0875,
+      "eval_samples_per_second": 4.663,
+      "eval_wer": 0.3487109905020353,
+      "step": 30400
+    },
+    {
+      "epoch": 8.13,
+      "learning_rate": 9.729753755889413e-05,
+      "loss": 0.1365,
+      "step": 30500
+    },
+    {
+      "epoch": 8.13,
+      "eval_loss": 0.48427531123161316,
+      "eval_runtime": 19.1077,
+      "eval_samples_per_second": 4.658,
+      "eval_wer": 0.36906377204884666,
+      "step": 30500
+    },
+    {
+      "epoch": 8.16,
+      "learning_rate": 9.728864787981155e-05,
+      "loss": 0.1342,
+      "step": 30600
+    },
+    {
+      "epoch": 8.16,
+      "eval_loss": 0.47319698333740234,
+      "eval_runtime": 19.1238,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.36635006784260515,
+      "step": 30600
+    },
+    {
+      "epoch": 8.19,
+      "learning_rate": 9.727975820072895e-05,
+      "loss": 0.152,
+      "step": 30700
+    },
+    {
+      "epoch": 8.19,
+      "eval_loss": 0.43120163679122925,
+      "eval_runtime": 19.1531,
+      "eval_samples_per_second": 4.647,
+      "eval_wer": 0.3677069199457259,
+      "step": 30700
+    },
+    {
+      "epoch": 8.21,
+      "learning_rate": 9.727086852164638e-05,
+      "loss": 0.1581,
+      "step": 30800
+    },
+    {
+      "epoch": 8.21,
+      "eval_loss": 0.43442800641059875,
+      "eval_runtime": 19.2568,
+      "eval_samples_per_second": 4.622,
+      "eval_wer": 0.36092265943012214,
+      "step": 30800
+    },
+    {
+      "epoch": 8.24,
+      "learning_rate": 9.726197884256378e-05,
+      "loss": 0.1497,
+      "step": 30900
+    },
+    {
+      "epoch": 8.24,
+      "eval_loss": 0.459628164768219,
+      "eval_runtime": 19.0439,
+      "eval_samples_per_second": 4.673,
+      "eval_wer": 0.37449118046132973,
+      "step": 30900
+    },
+    {
+      "epoch": 8.27,
+      "learning_rate": 9.72530891634812e-05,
+      "loss": 0.1407,
+      "step": 31000
+    },
+    {
+      "epoch": 8.27,
+      "eval_loss": 0.4642084836959839,
+      "eval_runtime": 19.1127,
+      "eval_samples_per_second": 4.657,
+      "eval_wer": 0.38263229308005425,
+      "step": 31000
+    },
+    {
+      "epoch": 8.29,
+      "learning_rate": 9.724419948439861e-05,
+      "loss": 0.1394,
+      "step": 31100
+    },
+    {
+      "epoch": 8.29,
+      "eval_loss": 0.44429096579551697,
+      "eval_runtime": 19.2032,
+      "eval_samples_per_second": 4.635,
+      "eval_wer": 0.3649932157394844,
+      "step": 31100
+    },
+    {
+      "epoch": 8.32,
+      "learning_rate": 9.723530980531603e-05,
+      "loss": 0.1575,
+      "step": 31200
+    },
+    {
+      "epoch": 8.32,
+      "eval_loss": 0.47867223620414734,
+      "eval_runtime": 18.8918,
+      "eval_samples_per_second": 4.711,
+      "eval_wer": 0.3677069199457259,
+      "step": 31200
+    },
+    {
+      "epoch": 8.35,
+      "learning_rate": 9.722642012623345e-05,
+      "loss": 0.1602,
+      "step": 31300
+    },
+    {
+      "epoch": 8.35,
+      "eval_loss": 0.4566819369792938,
+      "eval_runtime": 19.0551,
+      "eval_samples_per_second": 4.671,
+      "eval_wer": 0.3649932157394844,
+      "step": 31300
+    },
+    {
+      "epoch": 8.37,
+      "learning_rate": 9.721753044715086e-05,
+      "loss": 0.1561,
+      "step": 31400
+    },
+    {
+      "epoch": 8.37,
+      "eval_loss": 0.4551659822463989,
+      "eval_runtime": 19.0889,
+      "eval_samples_per_second": 4.662,
+      "eval_wer": 0.3704206241519674,
+      "step": 31400
+    },
+    {
+      "epoch": 8.4,
+      "learning_rate": 9.720864076806828e-05,
+      "loss": 0.1389,
+      "step": 31500
+    },
+    {
+      "epoch": 8.4,
+      "eval_loss": 0.5054765343666077,
+      "eval_runtime": 19.1734,
+      "eval_samples_per_second": 4.642,
+      "eval_wer": 0.39077340569877883,
+      "step": 31500
+    },
+    {
+      "epoch": 8.43,
+      "learning_rate": 9.719975108898569e-05,
+      "loss": 0.1446,
+      "step": 31600
+    },
+    {
+      "epoch": 8.43,
+      "eval_loss": 0.5110938549041748,
+      "eval_runtime": 19.1594,
+      "eval_samples_per_second": 4.645,
+      "eval_wer": 0.37449118046132973,
+      "step": 31600
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 9.719086140990311e-05,
+      "loss": 0.1395,
+      "step": 31700
+    },
+    {
+      "epoch": 8.45,
+      "eval_loss": 0.4636899530887604,
+      "eval_runtime": 19.0384,
+      "eval_samples_per_second": 4.675,
+      "eval_wer": 0.3622795115332429,
+      "step": 31700
+    },
+    {
+      "epoch": 8.48,
+      "learning_rate": 9.718197173082051e-05,
+      "loss": 0.1375,
+      "step": 31800
+    },
+    {
+      "epoch": 8.48,
+      "eval_loss": 0.5089759230613708,
+      "eval_runtime": 19.2182,
+      "eval_samples_per_second": 4.631,
+      "eval_wer": 0.3649932157394844,
+      "step": 31800
+    },
+    {
+      "epoch": 8.51,
+      "learning_rate": 9.717308205173794e-05,
+      "loss": 0.1587,
+      "step": 31900
+    },
+    {
+      "epoch": 8.51,
+      "eval_loss": 0.5211026072502136,
+      "eval_runtime": 19.2116,
+      "eval_samples_per_second": 4.633,
+      "eval_wer": 0.37720488466757124,
+      "step": 31900
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 9.716419237265536e-05,
+      "loss": 0.1445,
+      "step": 32000
+    },
+    {
+      "epoch": 8.53,
+      "eval_loss": 0.5122462511062622,
+      "eval_runtime": 18.6805,
+      "eval_samples_per_second": 4.764,
+      "eval_wer": 0.37720488466757124,
+      "step": 32000
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 9.715530269357276e-05,
+      "loss": 0.1414,
+      "step": 32100
+    },
+    {
+      "epoch": 8.56,
+      "eval_loss": 0.5316699147224426,
+      "eval_runtime": 18.7239,
+      "eval_samples_per_second": 4.753,
+      "eval_wer": 0.37449118046132973,
+      "step": 32100
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 9.714641301449018e-05,
+      "loss": 0.1483,
+      "step": 32200
+    },
+    {
+      "epoch": 8.59,
+      "eval_loss": 0.5292964577674866,
+      "eval_runtime": 19.0909,
+      "eval_samples_per_second": 4.662,
+      "eval_wer": 0.378561736770692,
+      "step": 32200
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 9.71375233354076e-05,
+      "loss": 0.1563,
+      "step": 32300
+    },
+    {
+      "epoch": 8.61,
+      "eval_loss": 0.5341648459434509,
+      "eval_runtime": 18.9523,
+      "eval_samples_per_second": 4.696,
+      "eval_wer": 0.36906377204884666,
+      "step": 32300
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 9.712863365632501e-05,
+      "loss": 0.1469,
+      "step": 32400
+    },
+    {
+      "epoch": 8.64,
+      "eval_loss": 0.4673183262348175,
+      "eval_runtime": 19.0738,
+      "eval_samples_per_second": 4.666,
+      "eval_wer": 0.36092265943012214,
+      "step": 32400
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 9.711974397724243e-05,
+      "loss": 0.137,
+      "step": 32500
+    },
+    {
+      "epoch": 8.67,
+      "eval_loss": 0.48212361335754395,
+      "eval_runtime": 19.1355,
+      "eval_samples_per_second": 4.651,
+      "eval_wer": 0.37449118046132973,
+      "step": 32500
+    },
+    {
+      "epoch": 8.69,
+      "learning_rate": 9.711085429815984e-05,
+      "loss": 0.1463,
+      "step": 32600
+    },
+    {
+      "epoch": 8.69,
+      "eval_loss": 0.4796565771102905,
+      "eval_runtime": 19.0044,
+      "eval_samples_per_second": 4.683,
+      "eval_wer": 0.3704206241519674,
+      "step": 32600
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 9.710196461907726e-05,
+      "loss": 0.1502,
+      "step": 32700
+    },
+    {
+      "epoch": 8.72,
+      "eval_loss": 0.5013092756271362,
+      "eval_runtime": 19.1019,
+      "eval_samples_per_second": 4.659,
+      "eval_wer": 0.37449118046132973,
+      "step": 32700
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 9.709307493999467e-05,
+      "loss": 0.1501,
+      "step": 32800
+    },
+    {
+      "epoch": 8.75,
+      "eval_loss": 0.4674939811229706,
+      "eval_runtime": 19.1309,
+      "eval_samples_per_second": 4.652,
+      "eval_wer": 0.3758480325644505,
+      "step": 32800
+    },
+    {
+      "epoch": 8.77,
+      "learning_rate": 9.708418526091209e-05,
+      "loss": 0.1436,
+      "step": 32900
+    },
+    {
+      "epoch": 8.77,
+      "eval_loss": 0.4998147487640381,
+      "eval_runtime": 19.3624,
+      "eval_samples_per_second": 4.597,
+      "eval_wer": 0.36635006784260515,
+      "step": 32900
+    },
+    {
+      "epoch": 8.8,
+      "learning_rate": 9.70752955818295e-05,
+      "loss": 0.1451,
+      "step": 33000
+    },
+    {
+      "epoch": 8.8,
+      "eval_loss": 0.4489138424396515,
+      "eval_runtime": 19.0423,
+      "eval_samples_per_second": 4.674,
+      "eval_wer": 0.3704206241519674,
+      "step": 33000
+    },
+    {
+      "epoch": 8.83,
+      "learning_rate": 9.706640590274692e-05,
+      "loss": 0.1488,
+      "step": 33100
+    },
+    {
+      "epoch": 8.83,
+      "eval_loss": 0.4716799557209015,
+      "eval_runtime": 18.9872,
+      "eval_samples_per_second": 4.687,
+      "eval_wer": 0.3649932157394844,
+      "step": 33100
+    },
+    {
+      "epoch": 8.85,
+      "learning_rate": 9.705751622366434e-05,
+      "loss": 0.1538,
+      "step": 33200
+    },
+    {
+      "epoch": 8.85,
+      "eval_loss": 0.46106600761413574,
+      "eval_runtime": 19.1462,
+      "eval_samples_per_second": 4.648,
+      "eval_wer": 0.3649932157394844,
+      "step": 33200
+    },
+    {
+      "epoch": 8.88,
+      "learning_rate": 9.704862654458174e-05,
+      "loss": 0.1504,
+      "step": 33300
+    },
+    {
+      "epoch": 8.88,
+      "eval_loss": 0.4125480353832245,
+      "eval_runtime": 19.0767,
+      "eval_samples_per_second": 4.665,
+      "eval_wer": 0.3677069199457259,
+      "step": 33300
+    },
+    {
+      "epoch": 8.91,
+      "learning_rate": 9.703973686549916e-05,
+      "loss": 0.1462,
+      "step": 33400
+    },
+    {
+      "epoch": 8.91,
+      "eval_loss": 0.44749605655670166,
+      "eval_runtime": 19.1081,
+      "eval_samples_per_second": 4.658,
+      "eval_wer": 0.3677069199457259,
+      "step": 33400
+    },
+    {
+      "epoch": 8.93,
+      "learning_rate": 9.703084718641657e-05,
+      "loss": 0.137,
+      "step": 33500
+    },
+    {
+      "epoch": 8.93,
+      "eval_loss": 0.46455812454223633,
+      "eval_runtime": 19.1584,
+      "eval_samples_per_second": 4.645,
+      "eval_wer": 0.3649932157394844,
+      "step": 33500
+    },
+    {
+      "epoch": 8.96,
+      "learning_rate": 9.702195750733399e-05,
+      "loss": 0.143,
+      "step": 33600
+    },
+    {
+      "epoch": 8.96,
+      "eval_loss": 0.4270174205303192,
+      "eval_runtime": 19.4937,
+      "eval_samples_per_second": 4.566,
+      "eval_wer": 0.373134328358209,
+      "step": 33600
+    },
+    {
+      "epoch": 8.99,
+      "learning_rate": 9.70130678282514e-05,
+      "loss": 0.1412,
+      "step": 33700
+    },
+    {
+      "epoch": 8.99,
+      "eval_loss": 0.47272413969039917,
+      "eval_runtime": 19.0926,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.3622795115332429,
+      "step": 33700
+    },
+    {
+      "epoch": 9.01,
+      "learning_rate": 9.700417814916882e-05,
+      "loss": 0.1513,
+      "step": 33800
+    },
+    {
+      "epoch": 9.01,
+      "eval_loss": 0.4483870565891266,
+      "eval_runtime": 19.1149,
+      "eval_samples_per_second": 4.656,
+      "eval_wer": 0.35278154681139756,
+      "step": 33800
+    },
+    {
+      "epoch": 9.04,
+      "learning_rate": 9.699528847008623e-05,
+      "loss": 0.1296,
+      "step": 33900
+    },
+    {
+      "epoch": 9.04,
+      "eval_loss": 0.45633700489997864,
+      "eval_runtime": 19.2465,
+      "eval_samples_per_second": 4.624,
+      "eval_wer": 0.3622795115332429,
+      "step": 33900
+    },
+    {
+      "epoch": 9.07,
+      "learning_rate": 9.698639879100365e-05,
+      "loss": 0.1317,
+      "step": 34000
+    },
+    {
+      "epoch": 9.07,
+      "eval_loss": 0.4496014714241028,
+      "eval_runtime": 18.9397,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.3677069199457259,
+      "step": 34000
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.697750911192105e-05,
+      "loss": 0.1348,
+      "step": 34100
+    },
+    {
+      "epoch": 9.09,
+      "eval_loss": 0.5121065378189087,
+      "eval_runtime": 18.8553,
+      "eval_samples_per_second": 4.72,
+      "eval_wer": 0.3649932157394844,
+      "step": 34100
+    },
+    {
+      "epoch": 9.12,
+      "learning_rate": 9.696861943283847e-05,
+      "loss": 0.1314,
+      "step": 34200
+    },
+    {
+      "epoch": 9.12,
+      "eval_loss": 0.4888863265514374,
+      "eval_runtime": 19.0624,
+      "eval_samples_per_second": 4.669,
+      "eval_wer": 0.36906377204884666,
+      "step": 34200
+    },
+    {
+      "epoch": 9.15,
+      "learning_rate": 9.69597297537559e-05,
+      "loss": 0.1424,
+      "step": 34300
+    },
+    {
+      "epoch": 9.15,
+      "eval_loss": 0.45483386516571045,
+      "eval_runtime": 19.1973,
+      "eval_samples_per_second": 4.636,
+      "eval_wer": 0.37991858887381275,
+      "step": 34300
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 9.695084007467332e-05,
+      "loss": 0.1251,
+      "step": 34400
+    },
+    {
+      "epoch": 9.17,
+      "eval_loss": 0.49185308814048767,
+      "eval_runtime": 18.9376,
+      "eval_samples_per_second": 4.7,
+      "eval_wer": 0.37991858887381275,
+      "step": 34400
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 9.694195039559072e-05,
+      "loss": 0.1346,
+      "step": 34500
+    },
+    {
+      "epoch": 9.2,
+      "eval_loss": 0.4568560719490051,
+      "eval_runtime": 19.5444,
+      "eval_samples_per_second": 4.554,
+      "eval_wer": 0.37991858887381275,
+      "step": 34500
+    },
+    {
+      "epoch": 9.23,
+      "learning_rate": 9.693306071650814e-05,
+      "loss": 0.1404,
+      "step": 34600
+    },
+    {
+      "epoch": 9.23,
+      "eval_loss": 0.48937907814979553,
+      "eval_runtime": 19.1171,
+      "eval_samples_per_second": 4.656,
+      "eval_wer": 0.37991858887381275,
+      "step": 34600
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 9.692417103742555e-05,
+      "loss": 0.1345,
+      "step": 34700
+    },
+    {
+      "epoch": 9.25,
+      "eval_loss": 0.4677034020423889,
+      "eval_runtime": 19.1666,
+      "eval_samples_per_second": 4.643,
+      "eval_wer": 0.36092265943012214,
+      "step": 34700
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 9.691528135834297e-05,
+      "loss": 0.1369,
+      "step": 34800
+    },
+    {
+      "epoch": 9.28,
+      "eval_loss": 0.5112490653991699,
+      "eval_runtime": 18.986,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.3704206241519674,
+      "step": 34800
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 9.690639167926039e-05,
+      "loss": 0.1402,
+      "step": 34900
+    },
+    {
+      "epoch": 9.31,
+      "eval_loss": 0.4531557261943817,
+      "eval_runtime": 19.0339,
+      "eval_samples_per_second": 4.676,
+      "eval_wer": 0.35006784260515605,
+      "step": 34900
+    },
+    {
+      "epoch": 9.33,
+      "learning_rate": 9.68975020001778e-05,
+      "loss": 0.1352,
+      "step": 35000
+    },
+    {
+      "epoch": 9.33,
+      "eval_loss": 0.4898829162120819,
+      "eval_runtime": 19.0066,
+      "eval_samples_per_second": 4.683,
+      "eval_wer": 0.3758480325644505,
+      "step": 35000
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 9.688861232109522e-05,
+      "loss": 0.1232,
+      "step": 35100
+    },
+    {
+      "epoch": 9.36,
+      "eval_loss": 0.4806523621082306,
+      "eval_runtime": 18.9953,
+      "eval_samples_per_second": 4.685,
+      "eval_wer": 0.36092265943012214,
+      "step": 35100
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 9.687972264201263e-05,
+      "loss": 0.1444,
+      "step": 35200
+    },
+    {
+      "epoch": 9.39,
+      "eval_loss": 0.47065097093582153,
+      "eval_runtime": 19.2965,
+      "eval_samples_per_second": 4.612,
+      "eval_wer": 0.37720488466757124,
+      "step": 35200
+    },
+    {
+      "epoch": 9.41,
+      "learning_rate": 9.687083296293005e-05,
+      "loss": 0.1312,
+      "step": 35300
+    },
+    {
+      "epoch": 9.41,
+      "eval_loss": 0.5507692098617554,
+      "eval_runtime": 19.4882,
+      "eval_samples_per_second": 4.567,
+      "eval_wer": 0.373134328358209,
+      "step": 35300
+    },
+    {
+      "epoch": 9.44,
+      "learning_rate": 9.686194328384746e-05,
+      "loss": 0.1279,
+      "step": 35400
+    },
+    {
+      "epoch": 9.44,
+      "eval_loss": 0.5002427697181702,
+      "eval_runtime": 19.2686,
+      "eval_samples_per_second": 4.619,
+      "eval_wer": 0.3677069199457259,
+      "step": 35400
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 9.685305360476488e-05,
+      "loss": 0.1363,
+      "step": 35500
+    },
+    {
+      "epoch": 9.47,
+      "eval_loss": 0.5028083920478821,
+      "eval_runtime": 19.3563,
+      "eval_samples_per_second": 4.598,
+      "eval_wer": 0.36635006784260515,
+      "step": 35500
+    },
+    {
+      "epoch": 9.49,
+      "learning_rate": 9.684416392568228e-05,
+      "loss": 0.1321,
+      "step": 35600
+    },
+    {
+      "epoch": 9.49,
+      "eval_loss": 0.5599233508110046,
+      "eval_runtime": 19.6796,
+      "eval_samples_per_second": 4.522,
+      "eval_wer": 0.37449118046132973,
+      "step": 35600
+    },
+    {
+      "epoch": 9.52,
+      "learning_rate": 9.68352742465997e-05,
+      "loss": 0.1271,
+      "step": 35700
+    },
+    {
+      "epoch": 9.52,
+      "eval_loss": 0.47436535358428955,
+      "eval_runtime": 19.446,
+      "eval_samples_per_second": 4.577,
+      "eval_wer": 0.3649932157394844,
+      "step": 35700
+    },
+    {
+      "epoch": 9.55,
+      "learning_rate": 9.682638456751711e-05,
+      "loss": 0.1391,
+      "step": 35800
+    },
+    {
+      "epoch": 9.55,
+      "eval_loss": 0.42880725860595703,
+      "eval_runtime": 19.1821,
+      "eval_samples_per_second": 4.64,
+      "eval_wer": 0.3582089552238806,
+      "step": 35800
+    },
+    {
+      "epoch": 9.57,
+      "learning_rate": 9.681749488843453e-05,
+      "loss": 0.1443,
+      "step": 35900
+    },
+    {
+      "epoch": 9.57,
+      "eval_loss": 0.4953589141368866,
+      "eval_runtime": 19.1926,
+      "eval_samples_per_second": 4.637,
+      "eval_wer": 0.37991858887381275,
+      "step": 35900
+    },
+    {
+      "epoch": 9.6,
+      "learning_rate": 9.680860520935194e-05,
+      "loss": 0.1251,
+      "step": 36000
+    },
+    {
+      "epoch": 9.6,
+      "eval_loss": 0.5396003723144531,
+      "eval_runtime": 19.4069,
+      "eval_samples_per_second": 4.586,
+      "eval_wer": 0.35549525101763907,
+      "step": 36000
+    },
+    {
+      "epoch": 9.63,
+      "learning_rate": 9.679971553026936e-05,
+      "loss": 0.1358,
+      "step": 36100
+    },
+    {
+      "epoch": 9.63,
+      "eval_loss": 0.5394467711448669,
+      "eval_runtime": 19.5331,
+      "eval_samples_per_second": 4.556,
+      "eval_wer": 0.378561736770692,
+      "step": 36100
+    },
+    {
+      "epoch": 9.65,
+      "learning_rate": 9.679082585118677e-05,
+      "loss": 0.1503,
+      "step": 36200
+    },
+    {
+      "epoch": 9.65,
+      "eval_loss": 0.500866174697876,
+      "eval_runtime": 19.222,
+      "eval_samples_per_second": 4.63,
+      "eval_wer": 0.35956580732700133,
+      "step": 36200
+    },
+    {
+      "epoch": 9.68,
+      "learning_rate": 9.678193617210419e-05,
+      "loss": 0.1372,
+      "step": 36300
+    },
+    {
+      "epoch": 9.68,
+      "eval_loss": 0.5512708425521851,
+      "eval_runtime": 19.3018,
+      "eval_samples_per_second": 4.611,
+      "eval_wer": 0.3677069199457259,
+      "step": 36300
+    },
+    {
+      "epoch": 9.71,
+      "learning_rate": 9.677304649302161e-05,
+      "loss": 0.1282,
+      "step": 36400
+    },
+    {
+      "epoch": 9.71,
+      "eval_loss": 0.6068547964096069,
+      "eval_runtime": 19.0233,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.3622795115332429,
+      "step": 36400
+    },
+    {
+      "epoch": 9.73,
+      "learning_rate": 9.676415681393903e-05,
+      "loss": 0.1332,
+      "step": 36500
+    },
+    {
+      "epoch": 9.73,
+      "eval_loss": 0.5263712406158447,
+      "eval_runtime": 19.2024,
+      "eval_samples_per_second": 4.635,
+      "eval_wer": 0.373134328358209,
+      "step": 36500
+    },
+    {
+      "epoch": 9.76,
+      "learning_rate": 9.675526713485644e-05,
+      "loss": 0.1278,
+      "step": 36600
+    },
+    {
+      "epoch": 9.76,
+      "eval_loss": 0.500022292137146,
+      "eval_runtime": 19.3329,
+      "eval_samples_per_second": 4.604,
+      "eval_wer": 0.3812754409769335,
+      "step": 36600
+    },
+    {
+      "epoch": 9.79,
+      "learning_rate": 9.674637745577386e-05,
+      "loss": 0.1464,
+      "step": 36700
+    },
+    {
+      "epoch": 9.79,
+      "eval_loss": 0.5074013471603394,
+      "eval_runtime": 19.9614,
+      "eval_samples_per_second": 4.459,
+      "eval_wer": 0.40705563093622793,
+      "step": 36700
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 9.673748777669128e-05,
+      "loss": 0.1463,
+      "step": 36800
+    },
+    {
+      "epoch": 9.81,
+      "eval_loss": 0.5642330050468445,
+      "eval_runtime": 19.6554,
+      "eval_samples_per_second": 4.528,
+      "eval_wer": 0.3921302578018996,
+      "step": 36800
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 9.672859809760868e-05,
+      "loss": 0.1447,
+      "step": 36900
+    },
+    {
+      "epoch": 9.84,
+      "eval_loss": 0.563325047492981,
+      "eval_runtime": 19.5728,
+      "eval_samples_per_second": 4.547,
+      "eval_wer": 0.3704206241519674,
+      "step": 36900
+    },
+    {
+      "epoch": 9.87,
+      "learning_rate": 9.67197084185261e-05,
+      "loss": 0.1477,
+      "step": 37000
+    },
+    {
+      "epoch": 9.87,
+      "eval_loss": 0.546058177947998,
+      "eval_runtime": 19.9056,
+      "eval_samples_per_second": 4.471,
+      "eval_wer": 0.36906377204884666,
+      "step": 37000
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 9.671081873944351e-05,
+      "loss": 0.1369,
+      "step": 37100
+    },
+    {
+      "epoch": 9.89,
+      "eval_loss": 0.5840484499931335,
+      "eval_runtime": 19.377,
+      "eval_samples_per_second": 4.593,
+      "eval_wer": 0.36635006784260515,
+      "step": 37100
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 9.670192906036093e-05,
+      "loss": 0.135,
+      "step": 37200
+    },
+    {
+      "epoch": 9.92,
+      "eval_loss": 0.5639474391937256,
+      "eval_runtime": 19.8886,
+      "eval_samples_per_second": 4.475,
+      "eval_wer": 0.36635006784260515,
+      "step": 37200
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 9.669303938127834e-05,
+      "loss": 0.1296,
+      "step": 37300
+    },
+    {
+      "epoch": 9.95,
+      "eval_loss": 0.6042764782905579,
+      "eval_runtime": 19.2824,
+      "eval_samples_per_second": 4.616,
+      "eval_wer": 0.3853459972862958,
+      "step": 37300
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.668414970219576e-05,
+      "loss": 0.1348,
+      "step": 37400
+    },
+    {
+      "epoch": 9.97,
+      "eval_loss": 0.5782718658447266,
+      "eval_runtime": 19.3843,
+      "eval_samples_per_second": 4.591,
+      "eval_wer": 0.38263229308005425,
+      "step": 37400
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.667526002311317e-05,
+      "loss": 0.1493,
+      "step": 37500
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 0.5749413371086121,
+      "eval_runtime": 19.5524,
+      "eval_samples_per_second": 4.552,
+      "eval_wer": 0.3853459972862958,
+      "step": 37500
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 9.666637034403059e-05,
+      "loss": 0.1202,
+      "step": 37600
+    },
+    {
+      "epoch": 10.03,
+      "eval_loss": 0.622056782245636,
+      "eval_runtime": 18.9859,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.383989145183175,
+      "step": 37600
+    },
+    {
+      "epoch": 10.05,
+      "learning_rate": 9.6657480664948e-05,
+      "loss": 0.1215,
+      "step": 37700
+    },
+    {
+      "epoch": 10.05,
+      "eval_loss": 0.540052056312561,
+      "eval_runtime": 19.5469,
+      "eval_samples_per_second": 4.553,
+      "eval_wer": 0.38670284938941657,
+      "step": 37700
+    },
+    {
+      "epoch": 10.08,
+      "learning_rate": 9.664859098586542e-05,
+      "loss": 0.1261,
+      "step": 37800
+    },
+    {
+      "epoch": 10.08,
+      "eval_loss": 0.5521669983863831,
+      "eval_runtime": 19.2945,
+      "eval_samples_per_second": 4.613,
+      "eval_wer": 0.3758480325644505,
+      "step": 37800
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 9.663970130678282e-05,
+      "loss": 0.1174,
+      "step": 37900
+    },
+    {
+      "epoch": 10.11,
+      "eval_loss": 0.5957828164100647,
+      "eval_runtime": 18.8411,
+      "eval_samples_per_second": 4.724,
+      "eval_wer": 0.37720488466757124,
+      "step": 37900
+    },
+    {
+      "epoch": 10.13,
+      "learning_rate": 9.663081162770024e-05,
+      "loss": 0.1265,
+      "step": 38000
+    },
+    {
+      "epoch": 10.13,
+      "eval_loss": 0.4838184416294098,
+      "eval_runtime": 18.8435,
+      "eval_samples_per_second": 4.723,
+      "eval_wer": 0.37991858887381275,
+      "step": 38000
+    },
+    {
+      "epoch": 10.16,
+      "learning_rate": 9.662192194861765e-05,
+      "loss": 0.1198,
+      "step": 38100
+    },
+    {
+      "epoch": 10.16,
+      "eval_loss": 0.5236405730247498,
+      "eval_runtime": 19.2289,
+      "eval_samples_per_second": 4.628,
+      "eval_wer": 0.3880597014925373,
+      "step": 38100
+    },
+    {
+      "epoch": 10.19,
+      "learning_rate": 9.661303226953507e-05,
+      "loss": 0.1206,
+      "step": 38200
+    },
+    {
+      "epoch": 10.19,
+      "eval_loss": 0.4768201410770416,
+      "eval_runtime": 19.7782,
+      "eval_samples_per_second": 4.5,
+      "eval_wer": 0.3812754409769335,
+      "step": 38200
+    },
+    {
+      "epoch": 10.21,
+      "learning_rate": 9.660414259045248e-05,
+      "loss": 0.1258,
+      "step": 38300
+    },
+    {
+      "epoch": 10.21,
+      "eval_loss": 0.508847713470459,
+      "eval_runtime": 19.5815,
+      "eval_samples_per_second": 4.545,
+      "eval_wer": 0.378561736770692,
+      "step": 38300
+    },
+    {
+      "epoch": 10.24,
+      "learning_rate": 9.65952529113699e-05,
+      "loss": 0.1459,
+      "step": 38400
+    },
+    {
+      "epoch": 10.24,
+      "eval_loss": 0.5468646883964539,
+      "eval_runtime": 18.9569,
+      "eval_samples_per_second": 4.695,
+      "eval_wer": 0.36635006784260515,
+      "step": 38400
+    },
+    {
+      "epoch": 10.27,
+      "learning_rate": 9.658636323228732e-05,
+      "loss": 0.1287,
+      "step": 38500
+    },
+    {
+      "epoch": 10.27,
+      "eval_loss": 0.476516455411911,
+      "eval_runtime": 19.4399,
+      "eval_samples_per_second": 4.578,
+      "eval_wer": 0.373134328358209,
+      "step": 38500
+    },
+    {
+      "epoch": 10.29,
+      "learning_rate": 9.657747355320473e-05,
+      "loss": 0.1311,
+      "step": 38600
+    },
+    {
+      "epoch": 10.29,
+      "eval_loss": 0.47492021322250366,
+      "eval_runtime": 18.9544,
+      "eval_samples_per_second": 4.695,
+      "eval_wer": 0.3622795115332429,
+      "step": 38600
+    },
+    {
+      "epoch": 10.32,
+      "learning_rate": 9.656858387412215e-05,
+      "loss": 0.126,
+      "step": 38700
+    },
+    {
+      "epoch": 10.32,
+      "eval_loss": 0.5082312822341919,
+      "eval_runtime": 18.9391,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.37720488466757124,
+      "step": 38700
+    },
+    {
+      "epoch": 10.35,
+      "learning_rate": 9.655969419503957e-05,
+      "loss": 0.1208,
+      "step": 38800
+    },
+    {
+      "epoch": 10.35,
+      "eval_loss": 0.47253718972206116,
+      "eval_runtime": 18.883,
+      "eval_samples_per_second": 4.713,
+      "eval_wer": 0.36363636363636365,
+      "step": 38800
+    },
+    {
+      "epoch": 10.37,
+      "learning_rate": 9.655080451595699e-05,
+      "loss": 0.1327,
+      "step": 38900
+    },
+    {
+      "epoch": 10.37,
+      "eval_loss": 0.4717899262905121,
+      "eval_runtime": 19.0233,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.378561736770692,
+      "step": 38900
+    },
+    {
+      "epoch": 10.4,
+      "learning_rate": 9.65419148368744e-05,
+      "loss": 0.1247,
+      "step": 39000
+    },
+    {
+      "epoch": 10.4,
+      "eval_loss": 0.533586859703064,
+      "eval_runtime": 18.7807,
+      "eval_samples_per_second": 4.739,
+      "eval_wer": 0.37991858887381275,
+      "step": 39000
+    },
+    {
+      "epoch": 10.43,
+      "learning_rate": 9.653302515779182e-05,
+      "loss": 0.1241,
+      "step": 39100
+    },
+    {
+      "epoch": 10.43,
+      "eval_loss": 0.5169569253921509,
+      "eval_runtime": 18.7542,
+      "eval_samples_per_second": 4.746,
+      "eval_wer": 0.3622795115332429,
+      "step": 39100
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 9.652413547870922e-05,
+      "loss": 0.1278,
+      "step": 39200
+    },
+    {
+      "epoch": 10.45,
+      "eval_loss": 0.49946340918540955,
+      "eval_runtime": 18.8803,
+      "eval_samples_per_second": 4.714,
+      "eval_wer": 0.3704206241519674,
+      "step": 39200
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 9.651524579962664e-05,
+      "loss": 0.1229,
+      "step": 39300
+    },
+    {
+      "epoch": 10.48,
+      "eval_loss": 0.5186386704444885,
+      "eval_runtime": 18.9282,
+      "eval_samples_per_second": 4.702,
+      "eval_wer": 0.36635006784260515,
+      "step": 39300
+    },
+    {
+      "epoch": 10.51,
+      "learning_rate": 9.650635612054405e-05,
+      "loss": 0.1329,
+      "step": 39400
+    },
+    {
+      "epoch": 10.51,
+      "eval_loss": 0.5104267001152039,
+      "eval_runtime": 18.7535,
+      "eval_samples_per_second": 4.746,
+      "eval_wer": 0.37720488466757124,
+      "step": 39400
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 9.649746644146147e-05,
+      "loss": 0.1156,
+      "step": 39500
+    },
+    {
+      "epoch": 10.53,
+      "eval_loss": 0.5080288052558899,
+      "eval_runtime": 18.8158,
+      "eval_samples_per_second": 4.73,
+      "eval_wer": 0.383989145183175,
+      "step": 39500
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 9.648857676237888e-05,
+      "loss": 0.1366,
+      "step": 39600
+    },
+    {
+      "epoch": 10.56,
+      "eval_loss": 0.5457862019538879,
+      "eval_runtime": 18.9381,
+      "eval_samples_per_second": 4.7,
+      "eval_wer": 0.37449118046132973,
+      "step": 39600
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 9.64796870832963e-05,
+      "loss": 0.124,
+      "step": 39700
+    },
+    {
+      "epoch": 10.59,
+      "eval_loss": 0.5074524283409119,
+      "eval_runtime": 19.2431,
+      "eval_samples_per_second": 4.625,
+      "eval_wer": 0.3704206241519674,
+      "step": 39700
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 9.647079740421371e-05,
+      "loss": 0.1405,
+      "step": 39800
+    },
+    {
+      "epoch": 10.61,
+      "eval_loss": 0.5374652147293091,
+      "eval_runtime": 19.3353,
+      "eval_samples_per_second": 4.603,
+      "eval_wer": 0.3758480325644505,
+      "step": 39800
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 9.646190772513113e-05,
+      "loss": 0.1432,
+      "step": 39900
+    },
+    {
+      "epoch": 10.64,
+      "eval_loss": 0.5069397687911987,
+      "eval_runtime": 18.9745,
+      "eval_samples_per_second": 4.691,
+      "eval_wer": 0.3704206241519674,
+      "step": 39900
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 9.645301804604854e-05,
+      "loss": 0.131,
+      "step": 40000
+    },
+    {
+      "epoch": 10.67,
+      "eval_loss": 0.507075846195221,
+      "eval_runtime": 18.9831,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.37720488466757124,
+      "step": 40000
+    },
+    {
+      "epoch": 10.69,
+      "learning_rate": 9.644412836696596e-05,
+      "loss": 0.1338,
+      "step": 40100
+    },
+    {
+      "epoch": 10.69,
+      "eval_loss": 0.5374857187271118,
+      "eval_runtime": 19.0282,
+      "eval_samples_per_second": 4.677,
+      "eval_wer": 0.34599728629579374,
+      "step": 40100
+    },
+    {
+      "epoch": 10.72,
+      "learning_rate": 9.643523868788336e-05,
+      "loss": 0.1504,
+      "step": 40200
+    },
+    {
+      "epoch": 10.72,
+      "eval_loss": 0.5982678532600403,
+      "eval_runtime": 18.8179,
+      "eval_samples_per_second": 4.73,
+      "eval_wer": 0.36906377204884666,
+      "step": 40200
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 9.642634900880078e-05,
+      "loss": 0.1239,
+      "step": 40300
+    },
+    {
+      "epoch": 10.75,
+      "eval_loss": 0.6135703325271606,
+      "eval_runtime": 18.9185,
+      "eval_samples_per_second": 4.704,
+      "eval_wer": 0.3704206241519674,
+      "step": 40300
+    },
+    {
+      "epoch": 10.77,
+      "learning_rate": 9.64174593297182e-05,
+      "loss": 0.1215,
+      "step": 40400
+    },
+    {
+      "epoch": 10.77,
+      "eval_loss": 0.5666403770446777,
+      "eval_runtime": 19.0473,
+      "eval_samples_per_second": 4.673,
+      "eval_wer": 0.35278154681139756,
+      "step": 40400
+    },
+    {
+      "epoch": 10.8,
+      "learning_rate": 9.640856965063561e-05,
+      "loss": 0.1183,
+      "step": 40500
+    },
+    {
+      "epoch": 10.8,
+      "eval_loss": 0.4917553663253784,
+      "eval_runtime": 18.779,
+      "eval_samples_per_second": 4.739,
+      "eval_wer": 0.37449118046132973,
+      "step": 40500
+    },
+    {
+      "epoch": 10.83,
+      "learning_rate": 9.639967997155303e-05,
+      "loss": 0.1303,
+      "step": 40600
+    },
+    {
+      "epoch": 10.83,
+      "eval_loss": 0.4532963037490845,
+      "eval_runtime": 19.0031,
+      "eval_samples_per_second": 4.683,
+      "eval_wer": 0.3568521031207598,
+      "step": 40600
+    },
+    {
+      "epoch": 10.85,
+      "learning_rate": 9.639079029247044e-05,
+      "loss": 0.1393,
+      "step": 40700
+    },
+    {
+      "epoch": 10.85,
+      "eval_loss": 0.4896491467952728,
+      "eval_runtime": 19.1512,
+      "eval_samples_per_second": 4.647,
+      "eval_wer": 0.36363636363636365,
+      "step": 40700
+    },
+    {
+      "epoch": 10.88,
+      "learning_rate": 9.638190061338786e-05,
+      "loss": 0.1386,
+      "step": 40800
+    },
+    {
+      "epoch": 10.88,
+      "eval_loss": 0.4572480022907257,
+      "eval_runtime": 19.033,
+      "eval_samples_per_second": 4.676,
+      "eval_wer": 0.37720488466757124,
+      "step": 40800
+    },
+    {
+      "epoch": 10.91,
+      "learning_rate": 9.637301093430528e-05,
+      "loss": 0.1367,
+      "step": 40900
+    },
+    {
+      "epoch": 10.91,
+      "eval_loss": 0.49036428332328796,
+      "eval_runtime": 18.9156,
+      "eval_samples_per_second": 4.705,
+      "eval_wer": 0.3758480325644505,
+      "step": 40900
+    },
+    {
+      "epoch": 10.93,
+      "learning_rate": 9.63641212552227e-05,
+      "loss": 0.1404,
+      "step": 41000
+    },
+    {
+      "epoch": 10.93,
+      "eval_loss": 0.49353957176208496,
+      "eval_runtime": 18.9922,
+      "eval_samples_per_second": 4.686,
+      "eval_wer": 0.3622795115332429,
+      "step": 41000
+    },
+    {
+      "epoch": 10.96,
+      "learning_rate": 9.635523157614011e-05,
+      "loss": 0.1207,
+      "step": 41100
+    },
+    {
+      "epoch": 10.96,
+      "eval_loss": 0.5062761902809143,
+      "eval_runtime": 18.94,
+      "eval_samples_per_second": 4.699,
+      "eval_wer": 0.37177747625508817,
+      "step": 41100
+    },
+    {
+      "epoch": 10.99,
+      "learning_rate": 9.634634189705753e-05,
+      "loss": 0.125,
+      "step": 41200
+    },
+    {
+      "epoch": 10.99,
+      "eval_loss": 0.4883849620819092,
+      "eval_runtime": 19.0294,
+      "eval_samples_per_second": 4.677,
+      "eval_wer": 0.383989145183175,
+      "step": 41200
+    },
+    {
+      "epoch": 11.01,
+      "learning_rate": 9.633745221797494e-05,
+      "loss": 0.1244,
+      "step": 41300
+    },
+    {
+      "epoch": 11.01,
+      "eval_loss": 0.5206499695777893,
+      "eval_runtime": 18.9844,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.378561736770692,
+      "step": 41300
+    },
+    {
+      "epoch": 11.04,
+      "learning_rate": 9.632856253889236e-05,
+      "loss": 0.1256,
+      "step": 41400
+    },
+    {
+      "epoch": 11.04,
+      "eval_loss": 0.4768509566783905,
+      "eval_runtime": 18.8631,
+      "eval_samples_per_second": 4.718,
+      "eval_wer": 0.36363636363636365,
+      "step": 41400
+    },
+    {
+      "epoch": 11.07,
+      "learning_rate": 9.631967285980976e-05,
+      "loss": 0.1189,
+      "step": 41500
+    },
+    {
+      "epoch": 11.07,
+      "eval_loss": 0.5374042391777039,
+      "eval_runtime": 18.9892,
+      "eval_samples_per_second": 4.687,
+      "eval_wer": 0.38263229308005425,
+      "step": 41500
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 9.631078318072718e-05,
+      "loss": 0.1179,
+      "step": 41600
+    },
+    {
+      "epoch": 11.09,
+      "eval_loss": 0.5065696239471436,
+      "eval_runtime": 19.1588,
+      "eval_samples_per_second": 4.645,
+      "eval_wer": 0.36635006784260515,
+      "step": 41600
+    },
+    {
+      "epoch": 11.12,
+      "learning_rate": 9.630189350164459e-05,
+      "loss": 0.1185,
+      "step": 41700
+    },
+    {
+      "epoch": 11.12,
+      "eval_loss": 0.47732967138290405,
+      "eval_runtime": 18.9008,
+      "eval_samples_per_second": 4.709,
+      "eval_wer": 0.36363636363636365,
+      "step": 41700
+    },
+    {
+      "epoch": 11.15,
+      "learning_rate": 9.629300382256201e-05,
+      "loss": 0.1219,
+      "step": 41800
+    },
+    {
+      "epoch": 11.15,
+      "eval_loss": 0.5960556864738464,
+      "eval_runtime": 19.3719,
+      "eval_samples_per_second": 4.594,
+      "eval_wer": 0.383989145183175,
+      "step": 41800
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 9.628411414347942e-05,
+      "loss": 0.1243,
+      "step": 41900
+    },
+    {
+      "epoch": 11.17,
+      "eval_loss": 0.5030884742736816,
+      "eval_runtime": 19.1628,
+      "eval_samples_per_second": 4.644,
+      "eval_wer": 0.3758480325644505,
+      "step": 41900
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 9.627522446439684e-05,
+      "loss": 0.1175,
+      "step": 42000
+    },
+    {
+      "epoch": 11.2,
+      "eval_loss": 0.4822719395160675,
+      "eval_runtime": 18.9648,
+      "eval_samples_per_second": 4.693,
+      "eval_wer": 0.3677069199457259,
+      "step": 42000
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 9.626633478531425e-05,
+      "loss": 0.1219,
+      "step": 42100
+    },
+    {
+      "epoch": 11.23,
+      "eval_loss": 0.4833529591560364,
+      "eval_runtime": 18.8526,
+      "eval_samples_per_second": 4.721,
+      "eval_wer": 0.378561736770692,
+      "step": 42100
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 9.625744510623167e-05,
+      "loss": 0.1201,
+      "step": 42200
+    },
+    {
+      "epoch": 11.25,
+      "eval_loss": 0.4882830083370209,
+      "eval_runtime": 18.5208,
+      "eval_samples_per_second": 4.805,
+      "eval_wer": 0.3622795115332429,
+      "step": 42200
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 9.624855542714909e-05,
+      "loss": 0.1243,
+      "step": 42300
+    },
+    {
+      "epoch": 11.28,
+      "eval_loss": 0.4852515161037445,
+      "eval_runtime": 18.5525,
+      "eval_samples_per_second": 4.797,
+      "eval_wer": 0.378561736770692,
+      "step": 42300
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 9.62396657480665e-05,
+      "loss": 0.1298,
+      "step": 42400
+    },
+    {
+      "epoch": 11.31,
+      "eval_loss": 0.5610016584396362,
+      "eval_runtime": 18.6812,
+      "eval_samples_per_second": 4.764,
+      "eval_wer": 0.378561736770692,
+      "step": 42400
+    },
+    {
+      "epoch": 11.33,
+      "learning_rate": 9.623077606898392e-05,
+      "loss": 0.1177,
+      "step": 42500
+    },
+    {
+      "epoch": 11.33,
+      "eval_loss": 0.4733968675136566,
+      "eval_runtime": 18.5726,
+      "eval_samples_per_second": 4.792,
+      "eval_wer": 0.3894165535956581,
+      "step": 42500
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 9.622188638990132e-05,
+      "loss": 0.1288,
+      "step": 42600
+    },
+    {
+      "epoch": 11.36,
+      "eval_loss": 0.5176745057106018,
+      "eval_runtime": 18.5425,
+      "eval_samples_per_second": 4.8,
+      "eval_wer": 0.383989145183175,
+      "step": 42600
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 9.621299671081874e-05,
+      "loss": 0.1301,
+      "step": 42700
+    },
+    {
+      "epoch": 11.39,
+      "eval_loss": 0.4754810035228729,
+      "eval_runtime": 18.488,
+      "eval_samples_per_second": 4.814,
+      "eval_wer": 0.3758480325644505,
+      "step": 42700
+    },
+    {
+      "epoch": 11.41,
+      "learning_rate": 9.620410703173615e-05,
+      "loss": 0.1168,
+      "step": 42800
+    },
+    {
+      "epoch": 11.41,
+      "eval_loss": 0.4704125225543976,
+      "eval_runtime": 18.4696,
+      "eval_samples_per_second": 4.819,
+      "eval_wer": 0.383989145183175,
+      "step": 42800
+    },
+    {
+      "epoch": 11.44,
+      "learning_rate": 9.619521735265357e-05,
+      "loss": 0.1262,
+      "step": 42900
+    },
+    {
+      "epoch": 11.44,
+      "eval_loss": 0.4788670837879181,
+      "eval_runtime": 18.4324,
+      "eval_samples_per_second": 4.828,
+      "eval_wer": 0.39348710990502034,
+      "step": 42900
+    },
+    {
+      "epoch": 11.47,
+      "learning_rate": 9.618632767357099e-05,
+      "loss": 0.1279,
+      "step": 43000
+    },
+    {
+      "epoch": 11.47,
+      "eval_loss": 0.4369634985923767,
+      "eval_runtime": 18.4074,
+      "eval_samples_per_second": 4.835,
+      "eval_wer": 0.3812754409769335,
+      "step": 43000
+    },
+    {
+      "epoch": 11.49,
+      "learning_rate": 9.61774379944884e-05,
+      "loss": 0.1138,
+      "step": 43100
+    },
+    {
+      "epoch": 11.49,
+      "eval_loss": 0.4786142110824585,
+      "eval_runtime": 18.5427,
+      "eval_samples_per_second": 4.8,
+      "eval_wer": 0.373134328358209,
+      "step": 43100
+    },
+    {
+      "epoch": 11.52,
+      "learning_rate": 9.616854831540582e-05,
+      "loss": 0.1364,
+      "step": 43200
+    },
+    {
+      "epoch": 11.52,
+      "eval_loss": 0.5559784173965454,
+      "eval_runtime": 18.5316,
+      "eval_samples_per_second": 4.803,
+      "eval_wer": 0.39348710990502034,
+      "step": 43200
+    },
+    {
+      "epoch": 11.55,
+      "learning_rate": 9.615965863632324e-05,
+      "loss": 0.1205,
+      "step": 43300
+    },
+    {
+      "epoch": 11.55,
+      "eval_loss": 0.508516252040863,
+      "eval_runtime": 18.3785,
+      "eval_samples_per_second": 4.843,
+      "eval_wer": 0.36092265943012214,
+      "step": 43300
+    },
+    {
+      "epoch": 11.57,
+      "learning_rate": 9.615076895724065e-05,
+      "loss": 0.1212,
+      "step": 43400
+    },
+    {
+      "epoch": 11.57,
+      "eval_loss": 0.5075653195381165,
+      "eval_runtime": 18.4252,
+      "eval_samples_per_second": 4.83,
+      "eval_wer": 0.37177747625508817,
+      "step": 43400
+    },
+    {
+      "epoch": 11.6,
+      "learning_rate": 9.614187927815807e-05,
+      "loss": 0.1122,
+      "step": 43500
+    },
+    {
+      "epoch": 11.6,
+      "eval_loss": 0.5215174555778503,
+      "eval_runtime": 18.5083,
+      "eval_samples_per_second": 4.809,
+      "eval_wer": 0.39077340569877883,
+      "step": 43500
+    },
+    {
+      "epoch": 11.63,
+      "learning_rate": 9.613298959907548e-05,
+      "loss": 0.1292,
+      "step": 43600
+    },
+    {
+      "epoch": 11.63,
+      "eval_loss": 0.5167407989501953,
+      "eval_runtime": 18.4917,
+      "eval_samples_per_second": 4.813,
+      "eval_wer": 0.3677069199457259,
+      "step": 43600
+    },
+    {
+      "epoch": 11.65,
+      "learning_rate": 9.61240999199929e-05,
+      "loss": 0.1189,
+      "step": 43700
+    },
+    {
+      "epoch": 11.65,
+      "eval_loss": 0.48751795291900635,
+      "eval_runtime": 18.5007,
+      "eval_samples_per_second": 4.811,
+      "eval_wer": 0.3677069199457259,
+      "step": 43700
+    },
+    {
+      "epoch": 11.68,
+      "learning_rate": 9.61152102409103e-05,
+      "loss": 0.1192,
+      "step": 43800
+    },
+    {
+      "epoch": 11.68,
+      "eval_loss": 0.504744291305542,
+      "eval_runtime": 18.4708,
+      "eval_samples_per_second": 4.818,
+      "eval_wer": 0.38263229308005425,
+      "step": 43800
+    },
+    {
+      "epoch": 11.71,
+      "learning_rate": 9.610632056182772e-05,
+      "loss": 0.1193,
+      "step": 43900
+    },
+    {
+      "epoch": 11.71,
+      "eval_loss": 0.495815634727478,
+      "eval_runtime": 18.5931,
+      "eval_samples_per_second": 4.787,
+      "eval_wer": 0.36092265943012214,
+      "step": 43900
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 9.609743088274515e-05,
+      "loss": 0.126,
+      "step": 44000
+    },
+    {
+      "epoch": 11.73,
+      "eval_loss": 0.48839473724365234,
+      "eval_runtime": 18.4904,
+      "eval_samples_per_second": 4.813,
+      "eval_wer": 0.36635006784260515,
+      "step": 44000
+    },
+    {
+      "epoch": 11.76,
+      "learning_rate": 9.608854120366255e-05,
+      "loss": 0.1226,
+      "step": 44100
+    },
+    {
+      "epoch": 11.76,
+      "eval_loss": 0.5896198749542236,
+      "eval_runtime": 18.4049,
+      "eval_samples_per_second": 4.836,
+      "eval_wer": 0.3853459972862958,
+      "step": 44100
+    },
+    {
+      "epoch": 11.79,
+      "learning_rate": 9.607965152457997e-05,
+      "loss": 0.1292,
+      "step": 44200
+    },
+    {
+      "epoch": 11.79,
+      "eval_loss": 0.515791654586792,
+      "eval_runtime": 18.393,
+      "eval_samples_per_second": 4.839,
+      "eval_wer": 0.3677069199457259,
+      "step": 44200
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 9.607076184549738e-05,
+      "loss": 0.1165,
+      "step": 44300
+    },
+    {
+      "epoch": 11.81,
+      "eval_loss": 0.4869723916053772,
+      "eval_runtime": 18.5779,
+      "eval_samples_per_second": 4.791,
+      "eval_wer": 0.37991858887381275,
+      "step": 44300
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 9.60618721664148e-05,
+      "loss": 0.1134,
+      "step": 44400
+    },
+    {
+      "epoch": 11.84,
+      "eval_loss": 0.49207982420921326,
+      "eval_runtime": 18.4293,
+      "eval_samples_per_second": 4.829,
+      "eval_wer": 0.37177747625508817,
+      "step": 44400
+    },
+    {
+      "epoch": 11.87,
+      "learning_rate": 9.605298248733221e-05,
+      "loss": 0.1287,
+      "step": 44500
+    },
+    {
+      "epoch": 11.87,
+      "eval_loss": 0.456536203622818,
+      "eval_runtime": 18.4931,
+      "eval_samples_per_second": 4.813,
+      "eval_wer": 0.36635006784260515,
+      "step": 44500
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 9.604409280824963e-05,
+      "loss": 0.1115,
+      "step": 44600
+    },
+    {
+      "epoch": 11.89,
+      "eval_loss": 0.5330535173416138,
+      "eval_runtime": 18.4444,
+      "eval_samples_per_second": 4.825,
+      "eval_wer": 0.3622795115332429,
+      "step": 44600
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 9.603520312916704e-05,
+      "loss": 0.1317,
+      "step": 44700
+    },
+    {
+      "epoch": 11.92,
+      "eval_loss": 0.5398668050765991,
+      "eval_runtime": 18.5795,
+      "eval_samples_per_second": 4.79,
+      "eval_wer": 0.3649932157394844,
+      "step": 44700
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 9.602631345008446e-05,
+      "loss": 0.1255,
+      "step": 44800
+    },
+    {
+      "epoch": 11.95,
+      "eval_loss": 0.5336928963661194,
+      "eval_runtime": 18.4701,
+      "eval_samples_per_second": 4.819,
+      "eval_wer": 0.3649932157394844,
+      "step": 44800
+    },
+    {
+      "epoch": 11.97,
+      "learning_rate": 9.601742377100186e-05,
+      "loss": 0.1223,
+      "step": 44900
+    },
+    {
+      "epoch": 11.97,
+      "eval_loss": 0.5405169129371643,
+      "eval_runtime": 18.4125,
+      "eval_samples_per_second": 4.834,
+      "eval_wer": 0.36906377204884666,
+      "step": 44900
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 9.600853409191928e-05,
+      "loss": 0.1289,
+      "step": 45000
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": 0.5292983651161194,
+      "eval_runtime": 18.5669,
+      "eval_samples_per_second": 4.793,
+      "eval_wer": 0.36635006784260515,
+      "step": 45000
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 9.599964441283669e-05,
+      "loss": 0.1065,
+      "step": 45100
+    },
+    {
+      "epoch": 12.03,
+      "eval_loss": 0.5058318972587585,
+      "eval_runtime": 18.3529,
+      "eval_samples_per_second": 4.849,
+      "eval_wer": 0.36363636363636365,
+      "step": 45100
+    },
+    {
+      "epoch": 12.05,
+      "learning_rate": 9.599075473375411e-05,
+      "loss": 0.1134,
+      "step": 45200
+    },
+    {
+      "epoch": 12.05,
+      "eval_loss": 0.500919759273529,
+      "eval_runtime": 18.3964,
+      "eval_samples_per_second": 4.838,
+      "eval_wer": 0.36363636363636365,
+      "step": 45200
+    },
+    {
+      "epoch": 12.08,
+      "learning_rate": 9.598186505467153e-05,
+      "loss": 0.1146,
+      "step": 45300
+    },
+    {
+      "epoch": 12.08,
+      "eval_loss": 0.5870064496994019,
+      "eval_runtime": 18.6943,
+      "eval_samples_per_second": 4.761,
+      "eval_wer": 0.37720488466757124,
+      "step": 45300
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 9.597297537558895e-05,
+      "loss": 0.1238,
+      "step": 45400
+    },
+    {
+      "epoch": 12.11,
+      "eval_loss": 0.5421254634857178,
+      "eval_runtime": 18.5668,
+      "eval_samples_per_second": 4.793,
+      "eval_wer": 0.36906377204884666,
+      "step": 45400
+    },
+    {
+      "epoch": 12.13,
+      "learning_rate": 9.596408569650636e-05,
+      "loss": 0.1172,
+      "step": 45500
+    },
+    {
+      "epoch": 12.13,
+      "eval_loss": 0.5242326855659485,
+      "eval_runtime": 18.4687,
+      "eval_samples_per_second": 4.819,
+      "eval_wer": 0.3622795115332429,
+      "step": 45500
+    },
+    {
+      "epoch": 12.16,
+      "learning_rate": 9.595519601742378e-05,
+      "loss": 0.1125,
+      "step": 45600
+    },
+    {
+      "epoch": 12.16,
+      "eval_loss": 0.515609085559845,
+      "eval_runtime": 18.4604,
+      "eval_samples_per_second": 4.821,
+      "eval_wer": 0.3758480325644505,
+      "step": 45600
+    },
+    {
+      "epoch": 12.19,
+      "learning_rate": 9.594630633834119e-05,
+      "loss": 0.1302,
+      "step": 45700
+    },
+    {
+      "epoch": 12.19,
+      "eval_loss": 0.4599016606807709,
+      "eval_runtime": 18.4783,
+      "eval_samples_per_second": 4.816,
+      "eval_wer": 0.37449118046132973,
+      "step": 45700
+    },
+    {
+      "epoch": 12.21,
+      "learning_rate": 9.593741665925861e-05,
+      "loss": 0.1162,
+      "step": 45800
+    },
+    {
+      "epoch": 12.21,
+      "eval_loss": 0.498314768075943,
+      "eval_runtime": 18.3794,
+      "eval_samples_per_second": 4.842,
+      "eval_wer": 0.378561736770692,
+      "step": 45800
+    },
+    {
+      "epoch": 12.24,
+      "learning_rate": 9.592852698017603e-05,
+      "loss": 0.1072,
+      "step": 45900
+    },
+    {
+      "epoch": 12.24,
+      "eval_loss": 0.5079747438430786,
+      "eval_runtime": 18.47,
+      "eval_samples_per_second": 4.819,
+      "eval_wer": 0.38263229308005425,
+      "step": 45900
+    },
+    {
+      "epoch": 12.27,
+      "learning_rate": 9.591963730109344e-05,
+      "loss": 0.1258,
+      "step": 46000
+    },
+    {
+      "epoch": 12.27,
+      "eval_loss": 0.5236510634422302,
+      "eval_runtime": 18.3961,
+      "eval_samples_per_second": 4.838,
+      "eval_wer": 0.3948439620081411,
+      "step": 46000
+    },
+    {
+      "epoch": 12.29,
+      "learning_rate": 9.591074762201086e-05,
+      "loss": 0.1152,
+      "step": 46100
+    },
+    {
+      "epoch": 12.29,
+      "eval_loss": 0.5025539398193359,
+      "eval_runtime": 18.4595,
+      "eval_samples_per_second": 4.821,
+      "eval_wer": 0.4043419267299864,
+      "step": 46100
+    },
+    {
+      "epoch": 12.32,
+      "learning_rate": 9.590185794292826e-05,
+      "loss": 0.1117,
+      "step": 46200
+    },
+    {
+      "epoch": 12.32,
+      "eval_loss": 0.5044400691986084,
+      "eval_runtime": 18.4222,
+      "eval_samples_per_second": 4.831,
+      "eval_wer": 0.3894165535956581,
+      "step": 46200
+    },
+    {
+      "epoch": 12.35,
+      "learning_rate": 9.589296826384568e-05,
+      "loss": 0.1348,
+      "step": 46300
+    },
+    {
+      "epoch": 12.35,
+      "eval_loss": 0.5156105160713196,
+      "eval_runtime": 18.3863,
+      "eval_samples_per_second": 4.841,
+      "eval_wer": 0.39348710990502034,
+      "step": 46300
+    },
+    {
+      "epoch": 12.37,
+      "learning_rate": 9.588407858476309e-05,
+      "loss": 0.1183,
+      "step": 46400
+    },
+    {
+      "epoch": 12.37,
+      "eval_loss": 0.534095823764801,
+      "eval_runtime": 18.4687,
+      "eval_samples_per_second": 4.819,
+      "eval_wer": 0.39755766621438265,
+      "step": 46400
+    },
+    {
+      "epoch": 12.4,
+      "learning_rate": 9.587518890568051e-05,
+      "loss": 0.1159,
+      "step": 46500
+    },
+    {
+      "epoch": 12.4,
+      "eval_loss": 0.5106559991836548,
+      "eval_runtime": 18.5073,
+      "eval_samples_per_second": 4.809,
+      "eval_wer": 0.39348710990502034,
+      "step": 46500
+    },
+    {
+      "epoch": 12.43,
+      "learning_rate": 9.586629922659792e-05,
+      "loss": 0.127,
+      "step": 46600
+    },
+    {
+      "epoch": 12.43,
+      "eval_loss": 0.51495361328125,
+      "eval_runtime": 18.4662,
+      "eval_samples_per_second": 4.82,
+      "eval_wer": 0.3880597014925373,
+      "step": 46600
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 9.585740954751534e-05,
+      "loss": 0.1142,
+      "step": 46700
+    },
+    {
+      "epoch": 12.45,
+      "eval_loss": 0.5183359384536743,
+      "eval_runtime": 18.3637,
+      "eval_samples_per_second": 4.847,
+      "eval_wer": 0.3853459972862958,
+      "step": 46700
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 9.584851986843275e-05,
+      "loss": 0.1228,
+      "step": 46800
+    },
+    {
+      "epoch": 12.48,
+      "eval_loss": 0.4911718964576721,
+      "eval_runtime": 18.3832,
+      "eval_samples_per_second": 4.841,
+      "eval_wer": 0.3894165535956581,
+      "step": 46800
+    },
+    {
+      "epoch": 12.51,
+      "learning_rate": 9.583963018935017e-05,
+      "loss": 0.114,
+      "step": 46900
+    },
+    {
+      "epoch": 12.51,
+      "eval_loss": 0.5739838480949402,
+      "eval_runtime": 18.3971,
+      "eval_samples_per_second": 4.838,
+      "eval_wer": 0.39077340569877883,
+      "step": 46900
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 9.583074051026758e-05,
+      "loss": 0.1131,
+      "step": 47000
+    },
+    {
+      "epoch": 12.53,
+      "eval_loss": 0.5165678262710571,
+      "eval_runtime": 18.5537,
+      "eval_samples_per_second": 4.797,
+      "eval_wer": 0.383989145183175,
+      "step": 47000
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 9.5821850831185e-05,
+      "loss": 0.1018,
+      "step": 47100
+    },
+    {
+      "epoch": 12.56,
+      "eval_loss": 0.5533799529075623,
+      "eval_runtime": 18.4738,
+      "eval_samples_per_second": 4.818,
+      "eval_wer": 0.378561736770692,
+      "step": 47100
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 9.58129611521024e-05,
+      "loss": 0.1257,
+      "step": 47200
+    },
+    {
+      "epoch": 12.59,
+      "eval_loss": 0.5154994130134583,
+      "eval_runtime": 18.5116,
+      "eval_samples_per_second": 4.808,
+      "eval_wer": 0.39077340569877883,
+      "step": 47200
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 9.580407147301982e-05,
+      "loss": 0.1197,
+      "step": 47300
+    },
+    {
+      "epoch": 12.61,
+      "eval_loss": 0.5019519925117493,
+      "eval_runtime": 18.5328,
+      "eval_samples_per_second": 4.802,
+      "eval_wer": 0.3880597014925373,
+      "step": 47300
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 9.579518179393724e-05,
+      "loss": 0.12,
+      "step": 47400
+    },
+    {
+      "epoch": 12.64,
+      "eval_loss": 0.523768424987793,
+      "eval_runtime": 18.3757,
+      "eval_samples_per_second": 4.843,
+      "eval_wer": 0.37177747625508817,
+      "step": 47400
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 9.578629211485467e-05,
+      "loss": 0.1125,
+      "step": 47500
+    },
+    {
+      "epoch": 12.67,
+      "eval_loss": 0.5521425604820251,
+      "eval_runtime": 18.6393,
+      "eval_samples_per_second": 4.775,
+      "eval_wer": 0.37449118046132973,
+      "step": 47500
+    },
+    {
+      "epoch": 12.69,
+      "learning_rate": 9.577740243577207e-05,
+      "loss": 0.1256,
+      "step": 47600
+    },
+    {
+      "epoch": 12.69,
+      "eval_loss": 0.5110790133476257,
+      "eval_runtime": 18.4608,
+      "eval_samples_per_second": 4.821,
+      "eval_wer": 0.373134328358209,
+      "step": 47600
+    },
+    {
+      "epoch": 12.72,
+      "learning_rate": 9.576851275668949e-05,
+      "loss": 0.1286,
+      "step": 47700
+    },
+    {
+      "epoch": 12.72,
+      "eval_loss": 0.5211021900177002,
+      "eval_runtime": 18.6193,
+      "eval_samples_per_second": 4.78,
+      "eval_wer": 0.3649932157394844,
+      "step": 47700
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 9.575962307760691e-05,
+      "loss": 0.1113,
+      "step": 47800
+    },
+    {
+      "epoch": 12.75,
+      "eval_loss": 0.5323519110679626,
+      "eval_runtime": 18.4729,
+      "eval_samples_per_second": 4.818,
+      "eval_wer": 0.3677069199457259,
+      "step": 47800
+    },
+    {
+      "epoch": 12.77,
+      "learning_rate": 9.575073339852432e-05,
+      "loss": 0.1178,
+      "step": 47900
+    },
+    {
+      "epoch": 12.77,
+      "eval_loss": 0.5314516425132751,
+      "eval_runtime": 18.4189,
+      "eval_samples_per_second": 4.832,
+      "eval_wer": 0.378561736770692,
+      "step": 47900
+    },
+    {
+      "epoch": 12.8,
+      "learning_rate": 9.574184371944174e-05,
+      "loss": 0.1331,
+      "step": 48000
+    },
+    {
+      "epoch": 12.8,
+      "eval_loss": 0.5547075271606445,
+      "eval_runtime": 18.4315,
+      "eval_samples_per_second": 4.829,
+      "eval_wer": 0.37720488466757124,
+      "step": 48000
+    },
+    {
+      "epoch": 12.83,
+      "learning_rate": 9.573295404035915e-05,
+      "loss": 0.1215,
+      "step": 48100
+    },
+    {
+      "epoch": 12.83,
+      "eval_loss": 0.5018399953842163,
+      "eval_runtime": 18.5824,
+      "eval_samples_per_second": 4.789,
+      "eval_wer": 0.373134328358209,
+      "step": 48100
+    },
+    {
+      "epoch": 12.85,
+      "learning_rate": 9.572406436127657e-05,
+      "loss": 0.1112,
+      "step": 48200
+    },
+    {
+      "epoch": 12.85,
+      "eval_loss": 0.5037311315536499,
+      "eval_runtime": 18.53,
+      "eval_samples_per_second": 4.803,
+      "eval_wer": 0.3853459972862958,
+      "step": 48200
+    },
+    {
+      "epoch": 12.88,
+      "learning_rate": 9.571517468219398e-05,
+      "loss": 0.1181,
+      "step": 48300
+    },
+    {
+      "epoch": 12.88,
+      "eval_loss": 0.47216159105300903,
+      "eval_runtime": 18.6715,
+      "eval_samples_per_second": 4.767,
+      "eval_wer": 0.3758480325644505,
+      "step": 48300
+    },
+    {
+      "epoch": 12.91,
+      "learning_rate": 9.57062850031114e-05,
+      "loss": 0.1338,
+      "step": 48400
+    },
+    {
+      "epoch": 12.91,
+      "eval_loss": 0.526368260383606,
+      "eval_runtime": 18.4819,
+      "eval_samples_per_second": 4.816,
+      "eval_wer": 0.37991858887381275,
+      "step": 48400
+    },
+    {
+      "epoch": 12.93,
+      "learning_rate": 9.56973953240288e-05,
+      "loss": 0.1072,
+      "step": 48500
+    },
+    {
+      "epoch": 12.93,
+      "eval_loss": 0.5151082873344421,
+      "eval_runtime": 18.5674,
+      "eval_samples_per_second": 4.793,
+      "eval_wer": 0.3758480325644505,
+      "step": 48500
+    },
+    {
+      "epoch": 12.96,
+      "learning_rate": 9.568850564494622e-05,
+      "loss": 0.1164,
+      "step": 48600
+    },
+    {
+      "epoch": 12.96,
+      "eval_loss": 0.4576261341571808,
+      "eval_runtime": 18.5424,
+      "eval_samples_per_second": 4.8,
+      "eval_wer": 0.37177747625508817,
+      "step": 48600
+    },
+    {
+      "epoch": 12.99,
+      "learning_rate": 9.567961596586363e-05,
+      "loss": 0.1118,
+      "step": 48700
+    },
+    {
+      "epoch": 12.99,
+      "eval_loss": 0.5205701589584351,
+      "eval_runtime": 18.4893,
+      "eval_samples_per_second": 4.814,
+      "eval_wer": 0.37177747625508817,
+      "step": 48700
+    },
+    {
+      "epoch": 13.01,
+      "learning_rate": 9.567072628678105e-05,
+      "loss": 0.1218,
+      "step": 48800
+    },
+    {
+      "epoch": 13.01,
+      "eval_loss": 0.5717171430587769,
+      "eval_runtime": 18.4815,
+      "eval_samples_per_second": 4.816,
+      "eval_wer": 0.38263229308005425,
+      "step": 48800
+    },
+    {
+      "epoch": 13.04,
+      "learning_rate": 9.566183660769846e-05,
+      "loss": 0.1069,
+      "step": 48900
+    },
+    {
+      "epoch": 13.04,
+      "eval_loss": 0.5778408050537109,
+      "eval_runtime": 18.8486,
+      "eval_samples_per_second": 4.722,
+      "eval_wer": 0.37177747625508817,
+      "step": 48900
+    },
+    {
+      "epoch": 13.07,
+      "learning_rate": 9.565294692861588e-05,
+      "loss": 0.1087,
+      "step": 49000
+    },
+    {
+      "epoch": 13.07,
+      "eval_loss": 0.5404648780822754,
+      "eval_runtime": 18.6906,
+      "eval_samples_per_second": 4.762,
+      "eval_wer": 0.37177747625508817,
+      "step": 49000
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 9.564405724953329e-05,
+      "loss": 0.112,
+      "step": 49100
+    },
+    {
+      "epoch": 13.09,
+      "eval_loss": 0.5402548909187317,
+      "eval_runtime": 18.5721,
+      "eval_samples_per_second": 4.792,
+      "eval_wer": 0.3812754409769335,
+      "step": 49100
+    },
+    {
+      "epoch": 13.12,
+      "learning_rate": 9.563516757045071e-05,
+      "loss": 0.1139,
+      "step": 49200
+    },
+    {
+      "epoch": 13.12,
+      "eval_loss": 0.5458917617797852,
+      "eval_runtime": 18.5042,
+      "eval_samples_per_second": 4.81,
+      "eval_wer": 0.37720488466757124,
+      "step": 49200
+    },
+    {
+      "epoch": 13.15,
+      "learning_rate": 9.562627789136812e-05,
+      "loss": 0.1199,
+      "step": 49300
+    },
+    {
+      "epoch": 13.15,
+      "eval_loss": 0.5493699312210083,
+      "eval_runtime": 18.6046,
+      "eval_samples_per_second": 4.784,
+      "eval_wer": 0.37720488466757124,
+      "step": 49300
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 9.561738821228554e-05,
+      "loss": 0.129,
+      "step": 49400
+    },
+    {
+      "epoch": 13.17,
+      "eval_loss": 0.5191054344177246,
+      "eval_runtime": 18.4874,
+      "eval_samples_per_second": 4.814,
+      "eval_wer": 0.383989145183175,
+      "step": 49400
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 9.560849853320296e-05,
+      "loss": 0.1014,
+      "step": 49500
+    },
+    {
+      "epoch": 13.2,
+      "eval_loss": 0.55241858959198,
+      "eval_runtime": 18.5397,
+      "eval_samples_per_second": 4.801,
+      "eval_wer": 0.36906377204884666,
+      "step": 49500
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 9.559960885412036e-05,
+      "loss": 0.109,
+      "step": 49600
+    },
+    {
+      "epoch": 13.23,
+      "eval_loss": 0.5140576958656311,
+      "eval_runtime": 18.5894,
+      "eval_samples_per_second": 4.788,
+      "eval_wer": 0.35549525101763907,
+      "step": 49600
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 9.559071917503778e-05,
+      "loss": 0.1094,
+      "step": 49700
+    },
+    {
+      "epoch": 13.25,
+      "eval_loss": 0.5314241051673889,
+      "eval_runtime": 18.4155,
+      "eval_samples_per_second": 4.833,
+      "eval_wer": 0.37449118046132973,
+      "step": 49700
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 9.55818294959552e-05,
+      "loss": 0.104,
+      "step": 49800
+    },
+    {
+      "epoch": 13.28,
+      "eval_loss": 0.48007091879844666,
+      "eval_runtime": 18.6129,
+      "eval_samples_per_second": 4.782,
+      "eval_wer": 0.35956580732700133,
+      "step": 49800
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 9.557293981687263e-05,
+      "loss": 0.1003,
+      "step": 49900
+    },
+    {
+      "epoch": 13.31,
+      "eval_loss": 0.5088266134262085,
+      "eval_runtime": 18.5868,
+      "eval_samples_per_second": 4.788,
+      "eval_wer": 0.373134328358209,
+      "step": 49900
+    },
+    {
+      "epoch": 13.33,
+      "learning_rate": 9.556405013779003e-05,
+      "loss": 0.1159,
+      "step": 50000
+    },
+    {
+      "epoch": 13.33,
+      "eval_loss": 0.5004997849464417,
+      "eval_runtime": 18.5816,
+      "eval_samples_per_second": 4.79,
+      "eval_wer": 0.37991858887381275,
+      "step": 50000
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 9.555516045870745e-05,
+      "loss": 0.1094,
+      "step": 50100
+    },
+    {
+      "epoch": 13.36,
+      "eval_loss": 0.49865999817848206,
+      "eval_runtime": 18.5192,
+      "eval_samples_per_second": 4.806,
+      "eval_wer": 0.378561736770692,
+      "step": 50100
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 9.554627077962486e-05,
+      "loss": 0.1172,
+      "step": 50200
+    },
+    {
+      "epoch": 13.39,
+      "eval_loss": 0.5248690843582153,
+      "eval_runtime": 18.4808,
+      "eval_samples_per_second": 4.816,
+      "eval_wer": 0.3812754409769335,
+      "step": 50200
+    },
+    {
+      "epoch": 13.41,
+      "learning_rate": 9.553738110054228e-05,
+      "loss": 0.109,
+      "step": 50300
+    },
+    {
+      "epoch": 13.41,
+      "eval_loss": 0.5493407249450684,
+      "eval_runtime": 18.4468,
+      "eval_samples_per_second": 4.825,
+      "eval_wer": 0.36635006784260515,
+      "step": 50300
+    },
+    {
+      "epoch": 13.44,
+      "learning_rate": 9.552849142145969e-05,
+      "loss": 0.1076,
+      "step": 50400
+    },
+    {
+      "epoch": 13.44,
+      "eval_loss": 0.5546450018882751,
+      "eval_runtime": 18.4445,
+      "eval_samples_per_second": 4.825,
+      "eval_wer": 0.3812754409769335,
+      "step": 50400
+    },
+    {
+      "epoch": 13.47,
+      "learning_rate": 9.551960174237711e-05,
+      "loss": 0.1121,
+      "step": 50500
+    },
+    {
+      "epoch": 13.47,
+      "eval_loss": 0.5394149422645569,
+      "eval_runtime": 18.5243,
+      "eval_samples_per_second": 4.805,
+      "eval_wer": 0.3704206241519674,
+      "step": 50500
+    },
+    {
+      "epoch": 13.49,
+      "learning_rate": 9.551071206329452e-05,
+      "loss": 0.1202,
+      "step": 50600
+    },
+    {
+      "epoch": 13.49,
+      "eval_loss": 0.5272220373153687,
+      "eval_runtime": 18.4684,
+      "eval_samples_per_second": 4.819,
+      "eval_wer": 0.3758480325644505,
+      "step": 50600
+    },
+    {
+      "epoch": 13.52,
+      "learning_rate": 9.550182238421194e-05,
+      "loss": 0.114,
+      "step": 50700
+    },
+    {
+      "epoch": 13.52,
+      "eval_loss": 0.5850356817245483,
+      "eval_runtime": 18.5674,
+      "eval_samples_per_second": 4.793,
+      "eval_wer": 0.37991858887381275,
+      "step": 50700
+    },
+    {
+      "epoch": 13.55,
+      "learning_rate": 9.549293270512934e-05,
+      "loss": 0.1095,
+      "step": 50800
+    },
+    {
+      "epoch": 13.55,
+      "eval_loss": 0.6063650846481323,
+      "eval_runtime": 18.6128,
+      "eval_samples_per_second": 4.782,
+      "eval_wer": 0.3812754409769335,
+      "step": 50800
+    },
+    {
+      "epoch": 13.57,
+      "learning_rate": 9.548404302604676e-05,
+      "loss": 0.1161,
+      "step": 50900
+    },
+    {
+      "epoch": 13.57,
+      "eval_loss": 0.5305390357971191,
+      "eval_runtime": 18.4774,
+      "eval_samples_per_second": 4.817,
+      "eval_wer": 0.37720488466757124,
+      "step": 50900
+    },
+    {
+      "epoch": 13.6,
+      "learning_rate": 9.547515334696417e-05,
+      "loss": 0.1026,
+      "step": 51000
+    },
+    {
+      "epoch": 13.6,
+      "eval_loss": 0.5353760719299316,
+      "eval_runtime": 18.5337,
+      "eval_samples_per_second": 4.802,
+      "eval_wer": 0.37720488466757124,
+      "step": 51000
+    },
+    {
+      "epoch": 13.63,
+      "learning_rate": 9.546626366788159e-05,
+      "loss": 0.1176,
+      "step": 51100
+    },
+    {
+      "epoch": 13.63,
+      "eval_loss": 0.5384781956672668,
+      "eval_runtime": 18.5716,
+      "eval_samples_per_second": 4.792,
+      "eval_wer": 0.39620081411126185,
+      "step": 51100
+    },
+    {
+      "epoch": 13.65,
+      "learning_rate": 9.5457373988799e-05,
+      "loss": 0.1045,
+      "step": 51200
+    },
+    {
+      "epoch": 13.65,
+      "eval_loss": 0.5240272283554077,
+      "eval_runtime": 18.9049,
+      "eval_samples_per_second": 4.708,
+      "eval_wer": 0.37449118046132973,
+      "step": 51200
+    },
+    {
+      "epoch": 13.68,
+      "learning_rate": 9.544848430971642e-05,
+      "loss": 0.1168,
+      "step": 51300
+    },
+    {
+      "epoch": 13.68,
+      "eval_loss": 0.5680223703384399,
+      "eval_runtime": 19.104,
+      "eval_samples_per_second": 4.659,
+      "eval_wer": 0.383989145183175,
+      "step": 51300
+    },
+    {
+      "epoch": 13.71,
+      "learning_rate": 9.543959463063384e-05,
+      "loss": 0.1193,
+      "step": 51400
+    },
+    {
+      "epoch": 13.71,
+      "eval_loss": 0.5521659255027771,
+      "eval_runtime": 19.1465,
+      "eval_samples_per_second": 4.648,
+      "eval_wer": 0.3921302578018996,
+      "step": 51400
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.543070495155125e-05,
+      "loss": 0.1054,
+      "step": 51500
+    },
+    {
+      "epoch": 13.73,
+      "eval_loss": 0.5446234345436096,
+      "eval_runtime": 18.8246,
+      "eval_samples_per_second": 4.728,
+      "eval_wer": 0.3894165535956581,
+      "step": 51500
+    },
+    {
+      "epoch": 13.76,
+      "learning_rate": 9.542181527246867e-05,
+      "loss": 0.1116,
+      "step": 51600
+    },
+    {
+      "epoch": 13.76,
+      "eval_loss": 0.5196542143821716,
+      "eval_runtime": 18.4922,
+      "eval_samples_per_second": 4.813,
+      "eval_wer": 0.37991858887381275,
+      "step": 51600
+    },
+    {
+      "epoch": 13.79,
+      "learning_rate": 9.541292559338608e-05,
+      "loss": 0.1174,
+      "step": 51700
+    },
+    {
+      "epoch": 13.79,
+      "eval_loss": 0.5170068144798279,
+      "eval_runtime": 18.5764,
+      "eval_samples_per_second": 4.791,
+      "eval_wer": 0.3758480325644505,
+      "step": 51700
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.54040359143035e-05,
+      "loss": 0.1052,
+      "step": 51800
+    },
+    {
+      "epoch": 13.81,
+      "eval_loss": 0.5080772638320923,
+      "eval_runtime": 18.6329,
+      "eval_samples_per_second": 4.777,
+      "eval_wer": 0.3812754409769335,
+      "step": 51800
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.539514623522092e-05,
+      "loss": 0.1086,
+      "step": 51900
+    },
+    {
+      "epoch": 13.84,
+      "eval_loss": 0.5389041900634766,
+      "eval_runtime": 18.5717,
+      "eval_samples_per_second": 4.792,
+      "eval_wer": 0.3853459972862958,
+      "step": 51900
+    },
+    {
+      "epoch": 13.87,
+      "learning_rate": 9.538625655613834e-05,
+      "loss": 0.11,
+      "step": 52000
+    },
+    {
+      "epoch": 13.87,
+      "eval_loss": 0.5823717713356018,
+      "eval_runtime": 18.5063,
+      "eval_samples_per_second": 4.809,
+      "eval_wer": 0.3880597014925373,
+      "step": 52000
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.537736687705575e-05,
+      "loss": 0.1068,
+      "step": 52100
+    },
+    {
+      "epoch": 13.89,
+      "eval_loss": 0.5642216205596924,
+      "eval_runtime": 18.6265,
+      "eval_samples_per_second": 4.778,
+      "eval_wer": 0.383989145183175,
+      "step": 52100
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.536847719797317e-05,
+      "loss": 0.1203,
+      "step": 52200
+    },
+    {
+      "epoch": 13.92,
+      "eval_loss": 0.5410022735595703,
+      "eval_runtime": 18.6087,
+      "eval_samples_per_second": 4.783,
+      "eval_wer": 0.38670284938941657,
+      "step": 52200
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.535958751889057e-05,
+      "loss": 0.1135,
+      "step": 52300
+    },
+    {
+      "epoch": 13.95,
+      "eval_loss": 0.566125750541687,
+      "eval_runtime": 18.7048,
+      "eval_samples_per_second": 4.758,
+      "eval_wer": 0.3758480325644505,
+      "step": 52300
+    },
+    {
+      "epoch": 13.97,
+      "learning_rate": 9.5350697839808e-05,
+      "loss": 0.1201,
+      "step": 52400
+    },
+    {
+      "epoch": 13.97,
+      "eval_loss": 0.540148138999939,
+      "eval_runtime": 18.5553,
+      "eval_samples_per_second": 4.796,
+      "eval_wer": 0.37720488466757124,
+      "step": 52400
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.53418081607254e-05,
+      "loss": 0.1099,
+      "step": 52500
+    },
+    {
+      "epoch": 14.0,
+      "eval_loss": 0.585115909576416,
+      "eval_runtime": 18.6409,
+      "eval_samples_per_second": 4.774,
+      "eval_wer": 0.38670284938941657,
+      "step": 52500
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.533291848164282e-05,
+      "loss": 0.1142,
+      "step": 52600
+    },
+    {
+      "epoch": 14.03,
+      "eval_loss": 0.613251268863678,
+      "eval_runtime": 18.6571,
+      "eval_samples_per_second": 4.77,
+      "eval_wer": 0.37991858887381275,
+      "step": 52600
+    },
+    {
+      "epoch": 14.05,
+      "learning_rate": 9.532402880256023e-05,
+      "loss": 0.1099,
+      "step": 52700
+    },
+    {
+      "epoch": 14.05,
+      "eval_loss": 0.652717649936676,
+      "eval_runtime": 18.6595,
+      "eval_samples_per_second": 4.77,
+      "eval_wer": 0.373134328358209,
+      "step": 52700
+    },
+    {
+      "epoch": 14.08,
+      "learning_rate": 9.531513912347765e-05,
+      "loss": 0.1132,
+      "step": 52800
+    },
+    {
+      "epoch": 14.08,
+      "eval_loss": 0.6298044323921204,
+      "eval_runtime": 18.7011,
+      "eval_samples_per_second": 4.759,
+      "eval_wer": 0.3704206241519674,
+      "step": 52800
+    },
+    {
+      "epoch": 14.11,
+      "learning_rate": 9.530624944439506e-05,
+      "loss": 0.1016,
+      "step": 52900
+    },
+    {
+      "epoch": 14.11,
+      "eval_loss": 0.6007134914398193,
+      "eval_runtime": 18.4161,
+      "eval_samples_per_second": 4.833,
+      "eval_wer": 0.36906377204884666,
+      "step": 52900
+    },
+    {
+      "epoch": 14.13,
+      "learning_rate": 9.529735976531248e-05,
+      "loss": 0.1061,
+      "step": 53000
+    },
+    {
+      "epoch": 14.13,
+      "eval_loss": 0.6146613955497742,
+      "eval_runtime": 18.5509,
+      "eval_samples_per_second": 4.798,
+      "eval_wer": 0.37991858887381275,
+      "step": 53000
+    },
+    {
+      "epoch": 14.16,
+      "learning_rate": 9.528847008622988e-05,
+      "loss": 0.1159,
+      "step": 53100
+    },
+    {
+      "epoch": 14.16,
+      "eval_loss": 0.6246035099029541,
+      "eval_runtime": 18.6652,
+      "eval_samples_per_second": 4.768,
+      "eval_wer": 0.37991858887381275,
+      "step": 53100
+    },
+    {
+      "epoch": 14.19,
+      "learning_rate": 9.52795804071473e-05,
+      "loss": 0.1035,
+      "step": 53200
+    },
+    {
+      "epoch": 14.19,
+      "eval_loss": 0.6481459736824036,
+      "eval_runtime": 18.6091,
+      "eval_samples_per_second": 4.783,
+      "eval_wer": 0.37177747625508817,
+      "step": 53200
+    },
+    {
+      "epoch": 14.21,
+      "learning_rate": 9.527069072806473e-05,
+      "loss": 0.1104,
+      "step": 53300
+    },
+    {
+      "epoch": 14.21,
+      "eval_loss": 0.6105273962020874,
+      "eval_runtime": 18.5461,
+      "eval_samples_per_second": 4.799,
+      "eval_wer": 0.37449118046132973,
+      "step": 53300
+    },
+    {
+      "epoch": 14.24,
+      "learning_rate": 9.526180104898213e-05,
+      "loss": 0.1002,
+      "step": 53400
+    },
+    {
+      "epoch": 14.24,
+      "eval_loss": 0.6429329514503479,
+      "eval_runtime": 18.6977,
+      "eval_samples_per_second": 4.76,
+      "eval_wer": 0.3812754409769335,
+      "step": 53400
+    },
+    {
+      "epoch": 14.27,
+      "learning_rate": 9.525291136989955e-05,
+      "loss": 0.1072,
+      "step": 53500
+    },
+    {
+      "epoch": 14.27,
+      "eval_loss": 0.6524696350097656,
+      "eval_runtime": 18.5585,
+      "eval_samples_per_second": 4.796,
+      "eval_wer": 0.37449118046132973,
+      "step": 53500
+    },
+    {
+      "epoch": 14.29,
+      "learning_rate": 9.524402169081696e-05,
+      "loss": 0.1137,
+      "step": 53600
+    },
+    {
+      "epoch": 14.29,
+      "eval_loss": 0.5868683457374573,
+      "eval_runtime": 18.5632,
+      "eval_samples_per_second": 4.794,
+      "eval_wer": 0.3853459972862958,
+      "step": 53600
+    },
+    {
+      "epoch": 14.32,
+      "learning_rate": 9.523513201173438e-05,
+      "loss": 0.1096,
+      "step": 53700
+    },
+    {
+      "epoch": 14.32,
+      "eval_loss": 0.6672039031982422,
+      "eval_runtime": 18.4483,
+      "eval_samples_per_second": 4.824,
+      "eval_wer": 0.378561736770692,
+      "step": 53700
+    },
+    {
+      "epoch": 14.35,
+      "learning_rate": 9.522624233265179e-05,
+      "loss": 0.1181,
+      "step": 53800
+    },
+    {
+      "epoch": 14.35,
+      "eval_loss": 0.5820255279541016,
+      "eval_runtime": 18.6255,
+      "eval_samples_per_second": 4.778,
+      "eval_wer": 0.373134328358209,
+      "step": 53800
+    },
+    {
+      "epoch": 14.37,
+      "learning_rate": 9.521735265356921e-05,
+      "loss": 0.1201,
+      "step": 53900
+    },
+    {
+      "epoch": 14.37,
+      "eval_loss": 0.6158517003059387,
+      "eval_runtime": 18.5497,
+      "eval_samples_per_second": 4.798,
+      "eval_wer": 0.3921302578018996,
+      "step": 53900
+    },
+    {
+      "epoch": 14.4,
+      "learning_rate": 9.520846297448663e-05,
+      "loss": 0.1024,
+      "step": 54000
+    },
+    {
+      "epoch": 14.4,
+      "eval_loss": 0.6067594289779663,
+      "eval_runtime": 18.6991,
+      "eval_samples_per_second": 4.76,
+      "eval_wer": 0.383989145183175,
+      "step": 54000
+    },
+    {
+      "epoch": 14.43,
+      "learning_rate": 9.519957329540404e-05,
+      "loss": 0.1121,
+      "step": 54100
+    },
+    {
+      "epoch": 14.43,
+      "eval_loss": 0.5615008473396301,
+      "eval_runtime": 18.611,
+      "eval_samples_per_second": 4.782,
+      "eval_wer": 0.37720488466757124,
+      "step": 54100
+    },
+    {
+      "epoch": 14.45,
+      "learning_rate": 9.519068361632146e-05,
+      "loss": 0.1093,
+      "step": 54200
+    },
+    {
+      "epoch": 14.45,
+      "eval_loss": 0.5931609272956848,
+      "eval_runtime": 18.6112,
+      "eval_samples_per_second": 4.782,
+      "eval_wer": 0.378561736770692,
+      "step": 54200
+    },
+    {
+      "epoch": 14.48,
+      "learning_rate": 9.518179393723888e-05,
+      "loss": 0.1018,
+      "step": 54300
+    },
+    {
+      "epoch": 14.48,
+      "eval_loss": 0.5316422581672668,
+      "eval_runtime": 18.6104,
+      "eval_samples_per_second": 4.782,
+      "eval_wer": 0.36092265943012214,
+      "step": 54300
+    },
+    {
+      "epoch": 14.51,
+      "learning_rate": 9.517290425815628e-05,
+      "loss": 0.1004,
+      "step": 54400
+    },
+    {
+      "epoch": 14.51,
+      "eval_loss": 0.5725070238113403,
+      "eval_runtime": 18.6075,
+      "eval_samples_per_second": 4.783,
+      "eval_wer": 0.3758480325644505,
+      "step": 54400
+    },
+    {
+      "epoch": 14.53,
+      "learning_rate": 9.51640145790737e-05,
+      "loss": 0.1041,
+      "step": 54500
+    },
+    {
+      "epoch": 14.53,
+      "eval_loss": 0.5380896925926208,
+      "eval_runtime": 18.5708,
+      "eval_samples_per_second": 4.792,
+      "eval_wer": 0.37449118046132973,
+      "step": 54500
+    },
+    {
+      "epoch": 14.56,
+      "learning_rate": 9.515512489999111e-05,
+      "loss": 0.0993,
+      "step": 54600
+    },
+    {
+      "epoch": 14.56,
+      "eval_loss": 0.6202245950698853,
+      "eval_runtime": 18.7672,
+      "eval_samples_per_second": 4.742,
+      "eval_wer": 0.3758480325644505,
+      "step": 54600
+    },
+    {
+      "epoch": 14.59,
+      "learning_rate": 9.514623522090853e-05,
+      "loss": 0.1005,
+      "step": 54700
+    },
+    {
+      "epoch": 14.59,
+      "eval_loss": 0.6070139408111572,
+      "eval_runtime": 18.5843,
+      "eval_samples_per_second": 4.789,
+      "eval_wer": 0.39348710990502034,
+      "step": 54700
+    },
+    {
+      "epoch": 14.61,
+      "learning_rate": 9.513734554182594e-05,
+      "loss": 0.109,
+      "step": 54800
+    },
+    {
+      "epoch": 14.61,
+      "eval_loss": 0.6114675998687744,
+      "eval_runtime": 18.5027,
+      "eval_samples_per_second": 4.81,
+      "eval_wer": 0.37991858887381275,
+      "step": 54800
+    },
+    {
+      "epoch": 14.64,
+      "learning_rate": 9.512845586274336e-05,
+      "loss": 0.103,
+      "step": 54900
+    },
+    {
+      "epoch": 14.64,
+      "eval_loss": 0.6103177666664124,
+      "eval_runtime": 18.6988,
+      "eval_samples_per_second": 4.76,
+      "eval_wer": 0.3649932157394844,
+      "step": 54900
+    },
+    {
+      "epoch": 14.67,
+      "learning_rate": 9.511956618366078e-05,
+      "loss": 0.1059,
+      "step": 55000
+    },
+    {
+      "epoch": 14.67,
+      "eval_loss": 0.5796723365783691,
+      "eval_runtime": 18.4807,
+      "eval_samples_per_second": 4.816,
+      "eval_wer": 0.3853459972862958,
+      "step": 55000
+    },
+    {
+      "epoch": 14.69,
+      "learning_rate": 9.511067650457819e-05,
+      "loss": 0.1124,
+      "step": 55100
+    },
+    {
+      "epoch": 14.69,
+      "eval_loss": 0.5705472230911255,
+      "eval_runtime": 18.5087,
+      "eval_samples_per_second": 4.809,
+      "eval_wer": 0.3812754409769335,
+      "step": 55100
+    },
+    {
+      "epoch": 14.72,
+      "learning_rate": 9.510178682549561e-05,
+      "loss": 0.1136,
+      "step": 55200
+    },
+    {
+      "epoch": 14.72,
+      "eval_loss": 0.5421719551086426,
+      "eval_runtime": 18.5482,
+      "eval_samples_per_second": 4.798,
+      "eval_wer": 0.37177747625508817,
+      "step": 55200
+    },
+    {
+      "epoch": 14.75,
+      "learning_rate": 9.509289714641302e-05,
+      "loss": 0.1182,
+      "step": 55300
+    },
+    {
+      "epoch": 14.75,
+      "eval_loss": 0.5531747937202454,
+      "eval_runtime": 18.5003,
+      "eval_samples_per_second": 4.811,
+      "eval_wer": 0.37177747625508817,
+      "step": 55300
+    },
+    {
+      "epoch": 14.77,
+      "learning_rate": 9.508400746733044e-05,
+      "loss": 0.1198,
+      "step": 55400
+    },
+    {
+      "epoch": 14.77,
+      "eval_loss": 0.5494636297225952,
+      "eval_runtime": 18.4878,
+      "eval_samples_per_second": 4.814,
+      "eval_wer": 0.3541383989145183,
+      "step": 55400
+    },
+    {
+      "epoch": 14.8,
+      "learning_rate": 9.507511778824784e-05,
+      "loss": 0.1012,
+      "step": 55500
+    },
+    {
+      "epoch": 14.8,
+      "eval_loss": 0.5396161675453186,
+      "eval_runtime": 18.4857,
+      "eval_samples_per_second": 4.815,
+      "eval_wer": 0.36906377204884666,
+      "step": 55500
+    },
+    {
+      "epoch": 14.83,
+      "learning_rate": 9.506622810916527e-05,
+      "loss": 0.1198,
+      "step": 55600
+    },
+    {
+      "epoch": 14.83,
+      "eval_loss": 0.5467861294746399,
+      "eval_runtime": 18.6011,
+      "eval_samples_per_second": 4.785,
+      "eval_wer": 0.37720488466757124,
+      "step": 55600
+    },
+    {
+      "epoch": 14.85,
+      "learning_rate": 9.505733843008267e-05,
+      "loss": 0.1157,
+      "step": 55700
+    },
+    {
+      "epoch": 14.85,
+      "eval_loss": 0.5428489446640015,
+      "eval_runtime": 18.4353,
+      "eval_samples_per_second": 4.828,
+      "eval_wer": 0.37449118046132973,
+      "step": 55700
+    },
+    {
+      "epoch": 14.88,
+      "learning_rate": 9.504844875100009e-05,
+      "loss": 0.1099,
+      "step": 55800
+    },
+    {
+      "epoch": 14.88,
+      "eval_loss": 0.5461256504058838,
+      "eval_runtime": 18.4698,
+      "eval_samples_per_second": 4.819,
+      "eval_wer": 0.36906377204884666,
+      "step": 55800
+    },
+    {
+      "epoch": 14.91,
+      "learning_rate": 9.50395590719175e-05,
+      "loss": 0.1067,
+      "step": 55900
+    },
+    {
+      "epoch": 14.91,
+      "eval_loss": 0.5520471930503845,
+      "eval_runtime": 18.5318,
+      "eval_samples_per_second": 4.803,
+      "eval_wer": 0.3677069199457259,
+      "step": 55900
+    },
+    {
+      "epoch": 14.93,
+      "learning_rate": 9.503066939283492e-05,
+      "loss": 0.1174,
+      "step": 56000
+    },
+    {
+      "epoch": 14.93,
+      "eval_loss": 0.5237467885017395,
+      "eval_runtime": 18.5887,
+      "eval_samples_per_second": 4.788,
+      "eval_wer": 0.35549525101763907,
+      "step": 56000
+    },
+    {
+      "epoch": 14.96,
+      "learning_rate": 9.502177971375233e-05,
+      "loss": 0.1134,
+      "step": 56100
+    },
+    {
+      "epoch": 14.96,
+      "eval_loss": 0.5126647353172302,
+      "eval_runtime": 18.6923,
+      "eval_samples_per_second": 4.761,
+      "eval_wer": 0.3582089552238806,
+      "step": 56100
+    },
+    {
+      "epoch": 14.99,
+      "learning_rate": 9.501289003466975e-05,
+      "loss": 0.0971,
+      "step": 56200
+    },
+    {
+      "epoch": 14.99,
+      "eval_loss": 0.5562706589698792,
+      "eval_runtime": 18.3948,
+      "eval_samples_per_second": 4.838,
+      "eval_wer": 0.3677069199457259,
+      "step": 56200
+    },
+    {
+      "epoch": 15.01,
+      "learning_rate": 9.500400035558717e-05,
+      "loss": 0.1002,
+      "step": 56300
+    },
+    {
+      "epoch": 15.01,
+      "eval_loss": 0.5384283065795898,
+      "eval_runtime": 18.4397,
+      "eval_samples_per_second": 4.827,
+      "eval_wer": 0.37991858887381275,
+      "step": 56300
+    },
+    {
+      "epoch": 15.04,
+      "learning_rate": 9.499511067650459e-05,
+      "loss": 0.1004,
+      "step": 56400
+    },
+    {
+      "epoch": 15.04,
+      "eval_loss": 0.5369305610656738,
+      "eval_runtime": 18.5761,
+      "eval_samples_per_second": 4.791,
+      "eval_wer": 0.378561736770692,
+      "step": 56400
+    },
+    {
+      "epoch": 15.07,
+      "learning_rate": 9.4986220997422e-05,
+      "loss": 0.1017,
+      "step": 56500
+    },
+    {
+      "epoch": 15.07,
+      "eval_loss": 0.5496396422386169,
+      "eval_runtime": 18.4354,
+      "eval_samples_per_second": 4.828,
+      "eval_wer": 0.383989145183175,
+      "step": 56500
+    },
+    {
+      "epoch": 15.09,
+      "learning_rate": 9.497733131833942e-05,
+      "loss": 0.1071,
+      "step": 56600
+    },
+    {
+      "epoch": 15.09,
+      "eval_loss": 0.5438072085380554,
+      "eval_runtime": 18.5147,
+      "eval_samples_per_second": 4.807,
+      "eval_wer": 0.383989145183175,
+      "step": 56600
+    },
+    {
+      "epoch": 15.12,
+      "learning_rate": 9.496844163925682e-05,
+      "loss": 0.1001,
+      "step": 56700
+    },
+    {
+      "epoch": 15.12,
+      "eval_loss": 0.5238698124885559,
+      "eval_runtime": 18.5712,
+      "eval_samples_per_second": 4.792,
+      "eval_wer": 0.36906377204884666,
+      "step": 56700
+    },
+    {
+      "epoch": 15.15,
+      "learning_rate": 9.495955196017425e-05,
+      "loss": 0.096,
+      "step": 56800
+    },
+    {
+      "epoch": 15.15,
+      "eval_loss": 0.5002960562705994,
+      "eval_runtime": 18.4504,
+      "eval_samples_per_second": 4.824,
+      "eval_wer": 0.3677069199457259,
+      "step": 56800
+    },
+    {
+      "epoch": 15.17,
+      "learning_rate": 9.495066228109167e-05,
+      "loss": 0.1072,
+      "step": 56900
+    },
+    {
+      "epoch": 15.17,
+      "eval_loss": 0.5401918888092041,
+      "eval_runtime": 18.5104,
+      "eval_samples_per_second": 4.808,
+      "eval_wer": 0.3758480325644505,
+      "step": 56900
+    },
+    {
+      "epoch": 15.2,
+      "learning_rate": 9.494177260200907e-05,
+      "loss": 0.0964,
+      "step": 57000
+    },
+    {
+      "epoch": 15.2,
+      "eval_loss": 0.5129409432411194,
+      "eval_runtime": 18.4919,
+      "eval_samples_per_second": 4.813,
+      "eval_wer": 0.37720488466757124,
+      "step": 57000
+    },
+    {
+      "epoch": 15.23,
+      "learning_rate": 9.49328829229265e-05,
+      "loss": 0.0909,
+      "step": 57100
+    },
+    {
+      "epoch": 15.23,
+      "eval_loss": 0.5551543235778809,
+      "eval_runtime": 18.4551,
+      "eval_samples_per_second": 4.823,
+      "eval_wer": 0.378561736770692,
+      "step": 57100
+    },
+    {
+      "epoch": 15.25,
+      "learning_rate": 9.49239932438439e-05,
+      "loss": 0.1091,
+      "step": 57200
+    },
+    {
+      "epoch": 15.25,
+      "eval_loss": 0.5453340411186218,
+      "eval_runtime": 18.4272,
+      "eval_samples_per_second": 4.83,
+      "eval_wer": 0.37449118046132973,
+      "step": 57200
+    },
+    {
+      "epoch": 15.28,
+      "learning_rate": 9.491510356476132e-05,
+      "loss": 0.1028,
+      "step": 57300
+    },
+    {
+      "epoch": 15.28,
+      "eval_loss": 0.5148200392723083,
+      "eval_runtime": 18.5061,
+      "eval_samples_per_second": 4.809,
+      "eval_wer": 0.378561736770692,
+      "step": 57300
+    },
+    {
+      "epoch": 15.31,
+      "learning_rate": 9.490621388567873e-05,
+      "loss": 0.1053,
+      "step": 57400
+    },
+    {
+      "epoch": 15.31,
+      "eval_loss": 0.5528640747070312,
+      "eval_runtime": 18.9833,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.383989145183175,
+      "step": 57400
+    },
+    {
+      "epoch": 15.33,
+      "learning_rate": 9.489732420659615e-05,
+      "loss": 0.0998,
+      "step": 57500
+    },
+    {
+      "epoch": 15.33,
+      "eval_loss": 0.5291978716850281,
+      "eval_runtime": 18.9768,
+      "eval_samples_per_second": 4.69,
+      "eval_wer": 0.37720488466757124,
+      "step": 57500
+    },
+    {
+      "epoch": 15.36,
+      "learning_rate": 9.488843452751356e-05,
+      "loss": 0.0978,
+      "step": 57600
+    },
+    {
+      "epoch": 15.36,
+      "eval_loss": 0.5440502166748047,
+      "eval_runtime": 18.9558,
+      "eval_samples_per_second": 4.695,
+      "eval_wer": 0.373134328358209,
+      "step": 57600
+    },
+    {
+      "epoch": 15.39,
+      "learning_rate": 9.487954484843098e-05,
+      "loss": 0.1029,
+      "step": 57700
+    },
+    {
+      "epoch": 15.39,
+      "eval_loss": 0.5121389031410217,
+      "eval_runtime": 18.8688,
+      "eval_samples_per_second": 4.717,
+      "eval_wer": 0.37177747625508817,
+      "step": 57700
+    },
+    {
+      "epoch": 15.41,
+      "learning_rate": 9.487065516934838e-05,
+      "loss": 0.1003,
+      "step": 57800
+    },
+    {
+      "epoch": 15.41,
+      "eval_loss": 0.5384892821311951,
+      "eval_runtime": 19.219,
+      "eval_samples_per_second": 4.631,
+      "eval_wer": 0.38263229308005425,
+      "step": 57800
+    },
+    {
+      "epoch": 15.44,
+      "learning_rate": 9.48617654902658e-05,
+      "loss": 0.0978,
+      "step": 57900
+    },
+    {
+      "epoch": 15.44,
+      "eval_loss": 0.515407919883728,
+      "eval_runtime": 19.1965,
+      "eval_samples_per_second": 4.636,
+      "eval_wer": 0.38263229308005425,
+      "step": 57900
+    },
+    {
+      "epoch": 15.47,
+      "learning_rate": 9.485287581118321e-05,
+      "loss": 0.104,
+      "step": 58000
+    },
+    {
+      "epoch": 15.47,
+      "eval_loss": 0.5049002766609192,
+      "eval_runtime": 19.4065,
+      "eval_samples_per_second": 4.586,
+      "eval_wer": 0.378561736770692,
+      "step": 58000
+    },
+    {
+      "epoch": 15.49,
+      "learning_rate": 9.484398613210063e-05,
+      "loss": 0.1087,
+      "step": 58100
+    },
+    {
+      "epoch": 15.49,
+      "eval_loss": 0.5138339996337891,
+      "eval_runtime": 19.2018,
+      "eval_samples_per_second": 4.635,
+      "eval_wer": 0.3704206241519674,
+      "step": 58100
+    },
+    {
+      "epoch": 15.52,
+      "learning_rate": 9.483509645301804e-05,
+      "loss": 0.1145,
+      "step": 58200
+    },
+    {
+      "epoch": 15.52,
+      "eval_loss": 0.513146698474884,
+      "eval_runtime": 18.877,
+      "eval_samples_per_second": 4.715,
+      "eval_wer": 0.3758480325644505,
+      "step": 58200
+    },
+    {
+      "epoch": 15.55,
+      "learning_rate": 9.482620677393546e-05,
+      "loss": 0.1099,
+      "step": 58300
+    },
+    {
+      "epoch": 15.55,
+      "eval_loss": 0.47510138154029846,
+      "eval_runtime": 18.9849,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.37991858887381275,
+      "step": 58300
+    },
+    {
+      "epoch": 15.57,
+      "learning_rate": 9.481731709485288e-05,
+      "loss": 0.0989,
+      "step": 58400
+    },
+    {
+      "epoch": 15.57,
+      "eval_loss": 0.5093278884887695,
+      "eval_runtime": 19.3031,
+      "eval_samples_per_second": 4.611,
+      "eval_wer": 0.38263229308005425,
+      "step": 58400
+    },
+    {
+      "epoch": 15.6,
+      "learning_rate": 9.48084274157703e-05,
+      "loss": 0.1048,
+      "step": 58500
+    },
+    {
+      "epoch": 15.6,
+      "eval_loss": 0.5314071178436279,
+      "eval_runtime": 19.02,
+      "eval_samples_per_second": 4.679,
+      "eval_wer": 0.3880597014925373,
+      "step": 58500
+    },
+    {
+      "epoch": 15.63,
+      "learning_rate": 9.479953773668771e-05,
+      "loss": 0.11,
+      "step": 58600
+    },
+    {
+      "epoch": 15.63,
+      "eval_loss": 0.47913604974746704,
+      "eval_runtime": 18.8904,
+      "eval_samples_per_second": 4.711,
+      "eval_wer": 0.3704206241519674,
+      "step": 58600
+    },
+    {
+      "epoch": 15.65,
+      "learning_rate": 9.479064805760513e-05,
+      "loss": 0.1073,
+      "step": 58700
+    },
+    {
+      "epoch": 15.65,
+      "eval_loss": 0.5027610063552856,
+      "eval_runtime": 19.4692,
+      "eval_samples_per_second": 4.571,
+      "eval_wer": 0.378561736770692,
+      "step": 58700
+    },
+    {
+      "epoch": 15.68,
+      "learning_rate": 9.478175837852255e-05,
+      "loss": 0.1082,
+      "step": 58800
+    },
+    {
+      "epoch": 15.68,
+      "eval_loss": 0.5003597140312195,
+      "eval_runtime": 19.0432,
+      "eval_samples_per_second": 4.674,
+      "eval_wer": 0.3812754409769335,
+      "step": 58800
+    },
+    {
+      "epoch": 15.71,
+      "learning_rate": 9.477286869943996e-05,
+      "loss": 0.0985,
+      "step": 58900
+    },
+    {
+      "epoch": 15.71,
+      "eval_loss": 0.5045387148857117,
+      "eval_runtime": 19.0943,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.38263229308005425,
+      "step": 58900
+    },
+    {
+      "epoch": 15.73,
+      "learning_rate": 9.476397902035738e-05,
+      "loss": 0.1059,
+      "step": 59000
+    },
+    {
+      "epoch": 15.73,
+      "eval_loss": 0.5741746425628662,
+      "eval_runtime": 19.1223,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.3812754409769335,
+      "step": 59000
+    },
+    {
+      "epoch": 15.76,
+      "learning_rate": 9.475508934127479e-05,
+      "loss": 0.1079,
+      "step": 59100
+    },
+    {
+      "epoch": 15.76,
+      "eval_loss": 0.5420674681663513,
+      "eval_runtime": 19.4574,
+      "eval_samples_per_second": 4.574,
+      "eval_wer": 0.37991858887381275,
+      "step": 59100
+    },
+    {
+      "epoch": 15.79,
+      "learning_rate": 9.47461996621922e-05,
+      "loss": 0.0996,
+      "step": 59200
+    },
+    {
+      "epoch": 15.79,
+      "eval_loss": 0.6000908017158508,
+      "eval_runtime": 19.1293,
+      "eval_samples_per_second": 4.653,
+      "eval_wer": 0.3894165535956581,
+      "step": 59200
+    },
+    {
+      "epoch": 15.81,
+      "learning_rate": 9.473730998310961e-05,
+      "loss": 0.1046,
+      "step": 59300
+    },
+    {
+      "epoch": 15.81,
+      "eval_loss": 0.5264856815338135,
+      "eval_runtime": 19.406,
+      "eval_samples_per_second": 4.586,
+      "eval_wer": 0.3704206241519674,
+      "step": 59300
+    },
+    {
+      "epoch": 15.84,
+      "learning_rate": 9.472842030402703e-05,
+      "loss": 0.1,
+      "step": 59400
+    },
+    {
+      "epoch": 15.84,
+      "eval_loss": 0.5273507237434387,
+      "eval_runtime": 19.0823,
+      "eval_samples_per_second": 4.664,
+      "eval_wer": 0.3812754409769335,
+      "step": 59400
+    },
+    {
+      "epoch": 15.87,
+      "learning_rate": 9.471953062494444e-05,
+      "loss": 0.0984,
+      "step": 59500
+    },
+    {
+      "epoch": 15.87,
+      "eval_loss": 0.5884492993354797,
+      "eval_runtime": 19.2995,
+      "eval_samples_per_second": 4.612,
+      "eval_wer": 0.3989145183175034,
+      "step": 59500
+    },
+    {
+      "epoch": 15.89,
+      "learning_rate": 9.471064094586186e-05,
+      "loss": 0.1054,
+      "step": 59600
+    },
+    {
+      "epoch": 15.89,
+      "eval_loss": 0.572980523109436,
+      "eval_runtime": 18.8918,
+      "eval_samples_per_second": 4.711,
+      "eval_wer": 0.3853459972862958,
+      "step": 59600
+    },
+    {
+      "epoch": 15.92,
+      "learning_rate": 9.470175126677927e-05,
+      "loss": 0.0989,
+      "step": 59700
+    },
+    {
+      "epoch": 15.92,
+      "eval_loss": 0.5334805250167847,
+      "eval_runtime": 18.875,
+      "eval_samples_per_second": 4.715,
+      "eval_wer": 0.3948439620081411,
+      "step": 59700
+    },
+    {
+      "epoch": 15.95,
+      "learning_rate": 9.469286158769669e-05,
+      "loss": 0.1027,
+      "step": 59800
+    },
+    {
+      "epoch": 15.95,
+      "eval_loss": 0.5464542508125305,
+      "eval_runtime": 19.2289,
+      "eval_samples_per_second": 4.628,
+      "eval_wer": 0.3812754409769335,
+      "step": 59800
+    },
+    {
+      "epoch": 15.97,
+      "learning_rate": 9.46839719086141e-05,
+      "loss": 0.0935,
+      "step": 59900
+    },
+    {
+      "epoch": 15.97,
+      "eval_loss": 0.5629993081092834,
+      "eval_runtime": 19.1368,
+      "eval_samples_per_second": 4.651,
+      "eval_wer": 0.3921302578018996,
+      "step": 59900
+    },
+    {
+      "epoch": 16.0,
+      "learning_rate": 9.467508222953152e-05,
+      "loss": 0.1046,
+      "step": 60000
+    },
+    {
+      "epoch": 16.0,
+      "eval_loss": 0.601607620716095,
+      "eval_runtime": 19.3531,
+      "eval_samples_per_second": 4.599,
+      "eval_wer": 0.37177747625508817,
+      "step": 60000
+    },
+    {
+      "epoch": 16.03,
+      "learning_rate": 9.466619255044892e-05,
+      "loss": 0.0912,
+      "step": 60100
+    },
+    {
+      "epoch": 16.03,
+      "eval_loss": 0.5614283084869385,
+      "eval_runtime": 18.8088,
+      "eval_samples_per_second": 4.732,
+      "eval_wer": 0.37177747625508817,
+      "step": 60100
+    },
+    {
+      "epoch": 16.05,
+      "learning_rate": 9.465730287136635e-05,
+      "loss": 0.0961,
+      "step": 60200
+    },
+    {
+      "epoch": 16.05,
+      "eval_loss": 0.6098468899726868,
+      "eval_runtime": 18.9534,
+      "eval_samples_per_second": 4.696,
+      "eval_wer": 0.3649932157394844,
+      "step": 60200
+    },
+    {
+      "epoch": 16.08,
+      "learning_rate": 9.464841319228375e-05,
+      "loss": 0.1026,
+      "step": 60300
+    },
+    {
+      "epoch": 16.08,
+      "eval_loss": 0.5291151404380798,
+      "eval_runtime": 18.9768,
+      "eval_samples_per_second": 4.69,
+      "eval_wer": 0.3622795115332429,
+      "step": 60300
+    },
+    {
+      "epoch": 16.11,
+      "learning_rate": 9.463952351320117e-05,
+      "loss": 0.0951,
+      "step": 60400
+    },
+    {
+      "epoch": 16.11,
+      "eval_loss": 0.5626860857009888,
+      "eval_runtime": 18.8401,
+      "eval_samples_per_second": 4.724,
+      "eval_wer": 0.37720488466757124,
+      "step": 60400
+    },
+    {
+      "epoch": 16.13,
+      "learning_rate": 9.46306338341186e-05,
+      "loss": 0.0966,
+      "step": 60500
+    },
+    {
+      "epoch": 16.13,
+      "eval_loss": 0.5558755993843079,
+      "eval_runtime": 18.8014,
+      "eval_samples_per_second": 4.734,
+      "eval_wer": 0.383989145183175,
+      "step": 60500
+    },
+    {
+      "epoch": 16.16,
+      "learning_rate": 9.4621744155036e-05,
+      "loss": 0.1018,
+      "step": 60600
+    },
+    {
+      "epoch": 16.16,
+      "eval_loss": 0.5478349328041077,
+      "eval_runtime": 18.6885,
+      "eval_samples_per_second": 4.762,
+      "eval_wer": 0.3853459972862958,
+      "step": 60600
+    },
+    {
+      "epoch": 16.19,
+      "learning_rate": 9.461285447595342e-05,
+      "loss": 0.0972,
+      "step": 60700
+    },
+    {
+      "epoch": 16.19,
+      "eval_loss": 0.5810014009475708,
+      "eval_runtime": 18.5053,
+      "eval_samples_per_second": 4.809,
+      "eval_wer": 0.3758480325644505,
+      "step": 60700
+    },
+    {
+      "epoch": 16.21,
+      "learning_rate": 9.460396479687084e-05,
+      "loss": 0.1012,
+      "step": 60800
+    },
+    {
+      "epoch": 16.21,
+      "eval_loss": 0.5996187329292297,
+      "eval_runtime": 18.5013,
+      "eval_samples_per_second": 4.81,
+      "eval_wer": 0.3812754409769335,
+      "step": 60800
+    },
+    {
+      "epoch": 16.24,
+      "learning_rate": 9.459507511778826e-05,
+      "loss": 0.1013,
+      "step": 60900
+    },
+    {
+      "epoch": 16.24,
+      "eval_loss": 0.6554964184761047,
+      "eval_runtime": 19.0172,
+      "eval_samples_per_second": 4.68,
+      "eval_wer": 0.37177747625508817,
+      "step": 60900
+    },
+    {
+      "epoch": 16.27,
+      "learning_rate": 9.458618543870567e-05,
+      "loss": 0.0965,
+      "step": 61000
+    },
+    {
+      "epoch": 16.27,
+      "eval_loss": 0.6729394197463989,
+      "eval_runtime": 18.8941,
+      "eval_samples_per_second": 4.71,
+      "eval_wer": 0.39077340569877883,
+      "step": 61000
+    },
+    {
+      "epoch": 16.29,
+      "learning_rate": 9.457729575962309e-05,
+      "loss": 0.0959,
+      "step": 61100
+    },
+    {
+      "epoch": 16.29,
+      "eval_loss": 0.6258922815322876,
+      "eval_runtime": 18.8064,
+      "eval_samples_per_second": 4.732,
+      "eval_wer": 0.4056987788331072,
+      "step": 61100
+    },
+    {
+      "epoch": 16.32,
+      "learning_rate": 9.45684060805405e-05,
+      "loss": 0.0994,
+      "step": 61200
+    },
+    {
+      "epoch": 16.32,
+      "eval_loss": 0.6229995489120483,
+      "eval_runtime": 19.2529,
+      "eval_samples_per_second": 4.623,
+      "eval_wer": 0.40027137042062416,
+      "step": 61200
+    },
+    {
+      "epoch": 16.35,
+      "learning_rate": 9.455951640145792e-05,
+      "loss": 0.0937,
+      "step": 61300
+    },
+    {
+      "epoch": 16.35,
+      "eval_loss": 0.6871793270111084,
+      "eval_runtime": 19.0858,
+      "eval_samples_per_second": 4.663,
+      "eval_wer": 0.3894165535956581,
+      "step": 61300
+    },
+    {
+      "epoch": 16.37,
+      "learning_rate": 9.455062672237533e-05,
+      "loss": 0.1002,
+      "step": 61400
+    },
+    {
+      "epoch": 16.37,
+      "eval_loss": 0.7389774322509766,
+      "eval_runtime": 19.0123,
+      "eval_samples_per_second": 4.681,
+      "eval_wer": 0.39755766621438265,
+      "step": 61400
+    },
+    {
+      "epoch": 16.4,
+      "learning_rate": 9.454173704329275e-05,
+      "loss": 0.1029,
+      "step": 61500
+    },
+    {
+      "epoch": 16.4,
+      "eval_loss": 0.6294413805007935,
+      "eval_runtime": 19.2685,
+      "eval_samples_per_second": 4.619,
+      "eval_wer": 0.3853459972862958,
+      "step": 61500
+    },
+    {
+      "epoch": 16.43,
+      "learning_rate": 9.453284736421015e-05,
+      "loss": 0.1017,
+      "step": 61600
+    },
+    {
+      "epoch": 16.43,
+      "eval_loss": 0.6244038939476013,
+      "eval_runtime": 18.9224,
+      "eval_samples_per_second": 4.703,
+      "eval_wer": 0.3853459972862958,
+      "step": 61600
+    },
+    {
+      "epoch": 16.45,
+      "learning_rate": 9.452395768512757e-05,
+      "loss": 0.1044,
+      "step": 61700
+    },
+    {
+      "epoch": 16.45,
+      "eval_loss": 0.597368061542511,
+      "eval_runtime": 19.055,
+      "eval_samples_per_second": 4.671,
+      "eval_wer": 0.3853459972862958,
+      "step": 61700
+    },
+    {
+      "epoch": 16.48,
+      "learning_rate": 9.451506800604498e-05,
+      "loss": 0.105,
+      "step": 61800
+    },
+    {
+      "epoch": 16.48,
+      "eval_loss": 0.6092307567596436,
+      "eval_runtime": 19.0862,
+      "eval_samples_per_second": 4.663,
+      "eval_wer": 0.37449118046132973,
+      "step": 61800
+    },
+    {
+      "epoch": 16.51,
+      "learning_rate": 9.45061783269624e-05,
+      "loss": 0.1043,
+      "step": 61900
+    },
+    {
+      "epoch": 16.51,
+      "eval_loss": 0.6135939955711365,
+      "eval_runtime": 19.0253,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.38263229308005425,
+      "step": 61900
+    },
+    {
+      "epoch": 16.53,
+      "learning_rate": 9.449728864787981e-05,
+      "loss": 0.095,
+      "step": 62000
+    },
+    {
+      "epoch": 16.53,
+      "eval_loss": 0.5883112549781799,
+      "eval_runtime": 19.0943,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.373134328358209,
+      "step": 62000
+    },
+    {
+      "epoch": 16.56,
+      "learning_rate": 9.448839896879723e-05,
+      "loss": 0.0996,
+      "step": 62100
+    },
+    {
+      "epoch": 16.56,
+      "eval_loss": 0.6011040210723877,
+      "eval_runtime": 18.9218,
+      "eval_samples_per_second": 4.704,
+      "eval_wer": 0.37177747625508817,
+      "step": 62100
+    },
+    {
+      "epoch": 16.59,
+      "learning_rate": 9.447950928971464e-05,
+      "loss": 0.1027,
+      "step": 62200
+    },
+    {
+      "epoch": 16.59,
+      "eval_loss": 0.5422033071517944,
+      "eval_runtime": 19.1261,
+      "eval_samples_per_second": 4.653,
+      "eval_wer": 0.36906377204884666,
+      "step": 62200
+    },
+    {
+      "epoch": 16.61,
+      "learning_rate": 9.447061961063206e-05,
+      "loss": 0.0962,
+      "step": 62300
+    },
+    {
+      "epoch": 16.61,
+      "eval_loss": 0.5859438180923462,
+      "eval_runtime": 19.3397,
+      "eval_samples_per_second": 4.602,
+      "eval_wer": 0.37177747625508817,
+      "step": 62300
+    },
+    {
+      "epoch": 16.64,
+      "learning_rate": 9.446172993154948e-05,
+      "loss": 0.0957,
+      "step": 62400
+    },
+    {
+      "epoch": 16.64,
+      "eval_loss": 0.6062609553337097,
+      "eval_runtime": 19.0238,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.37720488466757124,
+      "step": 62400
+    },
+    {
+      "epoch": 16.67,
+      "learning_rate": 9.445284025246689e-05,
+      "loss": 0.117,
+      "step": 62500
+    },
+    {
+      "epoch": 16.67,
+      "eval_loss": 0.5630375146865845,
+      "eval_runtime": 18.9707,
+      "eval_samples_per_second": 4.691,
+      "eval_wer": 0.37449118046132973,
+      "step": 62500
+    },
+    {
+      "epoch": 16.69,
+      "learning_rate": 9.44439505733843e-05,
+      "loss": 0.0918,
+      "step": 62600
+    },
+    {
+      "epoch": 16.69,
+      "eval_loss": 0.6094798445701599,
+      "eval_runtime": 19.2578,
+      "eval_samples_per_second": 4.622,
+      "eval_wer": 0.37991858887381275,
+      "step": 62600
+    },
+    {
+      "epoch": 16.72,
+      "learning_rate": 9.443506089430171e-05,
+      "loss": 0.0987,
+      "step": 62700
+    },
+    {
+      "epoch": 16.72,
+      "eval_loss": 0.600867509841919,
+      "eval_runtime": 19.2798,
+      "eval_samples_per_second": 4.616,
+      "eval_wer": 0.3649932157394844,
+      "step": 62700
+    },
+    {
+      "epoch": 16.75,
+      "learning_rate": 9.442617121521913e-05,
+      "loss": 0.0998,
+      "step": 62800
+    },
+    {
+      "epoch": 16.75,
+      "eval_loss": 0.5615139603614807,
+      "eval_runtime": 19.0948,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.3677069199457259,
+      "step": 62800
+    },
+    {
+      "epoch": 16.77,
+      "learning_rate": 9.441728153613655e-05,
+      "loss": 0.1087,
+      "step": 62900
+    },
+    {
+      "epoch": 16.77,
+      "eval_loss": 0.5773482918739319,
+      "eval_runtime": 19.4313,
+      "eval_samples_per_second": 4.58,
+      "eval_wer": 0.3812754409769335,
+      "step": 62900
+    },
+    {
+      "epoch": 16.8,
+      "learning_rate": 9.440839185705397e-05,
+      "loss": 0.1007,
+      "step": 63000
+    },
+    {
+      "epoch": 16.8,
+      "eval_loss": 0.5691789984703064,
+      "eval_runtime": 19.3363,
+      "eval_samples_per_second": 4.603,
+      "eval_wer": 0.37177747625508817,
+      "step": 63000
+    },
+    {
+      "epoch": 16.83,
+      "learning_rate": 9.439950217797138e-05,
+      "loss": 0.1061,
+      "step": 63100
+    },
+    {
+      "epoch": 16.83,
+      "eval_loss": 0.6077779531478882,
+      "eval_runtime": 18.8096,
+      "eval_samples_per_second": 4.732,
+      "eval_wer": 0.37449118046132973,
+      "step": 63100
+    },
+    {
+      "epoch": 16.85,
+      "learning_rate": 9.43906124988888e-05,
+      "loss": 0.0953,
+      "step": 63200
+    },
+    {
+      "epoch": 16.85,
+      "eval_loss": 0.5832227468490601,
+      "eval_runtime": 19.0223,
+      "eval_samples_per_second": 4.679,
+      "eval_wer": 0.3568521031207598,
+      "step": 63200
+    },
+    {
+      "epoch": 16.88,
+      "learning_rate": 9.438172281980621e-05,
+      "loss": 0.1012,
+      "step": 63300
+    },
+    {
+      "epoch": 16.88,
+      "eval_loss": 0.5437654852867126,
+      "eval_runtime": 19.0455,
+      "eval_samples_per_second": 4.673,
+      "eval_wer": 0.37449118046132973,
+      "step": 63300
+    },
+    {
+      "epoch": 16.91,
+      "learning_rate": 9.437283314072363e-05,
+      "loss": 0.0957,
+      "step": 63400
+    },
+    {
+      "epoch": 16.91,
+      "eval_loss": 0.5404393076896667,
+      "eval_runtime": 19.1102,
+      "eval_samples_per_second": 4.657,
+      "eval_wer": 0.3649932157394844,
+      "step": 63400
+    },
+    {
+      "epoch": 16.93,
+      "learning_rate": 9.436394346164104e-05,
+      "loss": 0.0994,
+      "step": 63500
+    },
+    {
+      "epoch": 16.93,
+      "eval_loss": 0.5660609006881714,
+      "eval_runtime": 19.015,
+      "eval_samples_per_second": 4.681,
+      "eval_wer": 0.3758480325644505,
+      "step": 63500
+    },
+    {
+      "epoch": 16.96,
+      "learning_rate": 9.435505378255846e-05,
+      "loss": 0.1027,
+      "step": 63600
+    },
+    {
+      "epoch": 16.96,
+      "eval_loss": 0.6406200528144836,
+      "eval_runtime": 19.1755,
+      "eval_samples_per_second": 4.641,
+      "eval_wer": 0.3704206241519674,
+      "step": 63600
+    },
+    {
+      "epoch": 16.99,
+      "learning_rate": 9.434616410347587e-05,
+      "loss": 0.1014,
+      "step": 63700
+    },
+    {
+      "epoch": 16.99,
+      "eval_loss": 0.5906551480293274,
+      "eval_runtime": 19.0934,
+      "eval_samples_per_second": 4.661,
+      "eval_wer": 0.37177747625508817,
+      "step": 63700
+    },
+    {
+      "epoch": 17.01,
+      "learning_rate": 9.433727442439329e-05,
+      "loss": 0.0917,
+      "step": 63800
+    },
+    {
+      "epoch": 17.01,
+      "eval_loss": 0.6214393377304077,
+      "eval_runtime": 19.0471,
+      "eval_samples_per_second": 4.673,
+      "eval_wer": 0.3812754409769335,
+      "step": 63800
+    },
+    {
+      "epoch": 17.04,
+      "learning_rate": 9.432838474531069e-05,
+      "loss": 0.0965,
+      "step": 63900
+    },
+    {
+      "epoch": 17.04,
+      "eval_loss": 0.5787700414657593,
+      "eval_runtime": 19.0236,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.37177747625508817,
+      "step": 63900
+    },
+    {
+      "epoch": 17.07,
+      "learning_rate": 9.431949506622811e-05,
+      "loss": 0.098,
+      "step": 64000
+    },
+    {
+      "epoch": 17.07,
+      "eval_loss": 0.5748379230499268,
+      "eval_runtime": 19.1238,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.37720488466757124,
+      "step": 64000
+    },
+    {
+      "epoch": 17.09,
+      "learning_rate": 9.431060538714552e-05,
+      "loss": 0.0962,
+      "step": 64100
+    },
+    {
+      "epoch": 17.09,
+      "eval_loss": 0.5793355703353882,
+      "eval_runtime": 18.9867,
+      "eval_samples_per_second": 4.687,
+      "eval_wer": 0.38263229308005425,
+      "step": 64100
+    },
+    {
+      "epoch": 17.12,
+      "learning_rate": 9.430171570806294e-05,
+      "loss": 0.1125,
+      "step": 64200
+    },
+    {
+      "epoch": 17.12,
+      "eval_loss": 0.5658828020095825,
+      "eval_runtime": 19.2044,
+      "eval_samples_per_second": 4.634,
+      "eval_wer": 0.3812754409769335,
+      "step": 64200
+    },
+    {
+      "epoch": 17.15,
+      "learning_rate": 9.429282602898036e-05,
+      "loss": 0.0919,
+      "step": 64300
+    },
+    {
+      "epoch": 17.15,
+      "eval_loss": 0.5571489334106445,
+      "eval_runtime": 19.4744,
+      "eval_samples_per_second": 4.57,
+      "eval_wer": 0.37991858887381275,
+      "step": 64300
+    },
+    {
+      "epoch": 17.17,
+      "learning_rate": 9.428393634989777e-05,
+      "loss": 0.0914,
+      "step": 64400
+    },
+    {
+      "epoch": 17.17,
+      "eval_loss": 0.5881439447402954,
+      "eval_runtime": 19.2909,
+      "eval_samples_per_second": 4.614,
+      "eval_wer": 0.3758480325644505,
+      "step": 64400
+    },
+    {
+      "epoch": 17.2,
+      "learning_rate": 9.427504667081519e-05,
+      "loss": 0.0993,
+      "step": 64500
+    },
+    {
+      "epoch": 17.2,
+      "eval_loss": 0.5764723420143127,
+      "eval_runtime": 19.1231,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.3758480325644505,
+      "step": 64500
+    },
+    {
+      "epoch": 17.23,
+      "learning_rate": 9.42661569917326e-05,
+      "loss": 0.0986,
+      "step": 64600
+    },
+    {
+      "epoch": 17.23,
+      "eval_loss": 0.5930325984954834,
+      "eval_runtime": 19.1842,
+      "eval_samples_per_second": 4.639,
+      "eval_wer": 0.39077340569877883,
+      "step": 64600
+    },
+    {
+      "epoch": 17.25,
+      "learning_rate": 9.425726731265002e-05,
+      "loss": 0.0995,
+      "step": 64700
+    },
+    {
+      "epoch": 17.25,
+      "eval_loss": 0.6214385628700256,
+      "eval_runtime": 19.0986,
+      "eval_samples_per_second": 4.66,
+      "eval_wer": 0.378561736770692,
+      "step": 64700
+    },
+    {
+      "epoch": 17.28,
+      "learning_rate": 9.424837763356742e-05,
+      "loss": 0.0975,
+      "step": 64800
+    },
+    {
+      "epoch": 17.28,
+      "eval_loss": 0.5780035257339478,
+      "eval_runtime": 19.1908,
+      "eval_samples_per_second": 4.638,
+      "eval_wer": 0.3812754409769335,
+      "step": 64800
+    },
+    {
+      "epoch": 17.31,
+      "learning_rate": 9.423948795448485e-05,
+      "loss": 0.0942,
+      "step": 64900
+    },
+    {
+      "epoch": 17.31,
+      "eval_loss": 0.5606986880302429,
+      "eval_runtime": 18.9049,
+      "eval_samples_per_second": 4.708,
+      "eval_wer": 0.3677069199457259,
+      "step": 64900
+    },
+    {
+      "epoch": 17.33,
+      "learning_rate": 9.423059827540227e-05,
+      "loss": 0.0932,
+      "step": 65000
+    },
+    {
+      "epoch": 17.33,
+      "eval_loss": 0.5660507678985596,
+      "eval_runtime": 19.116,
+      "eval_samples_per_second": 4.656,
+      "eval_wer": 0.3758480325644505,
+      "step": 65000
+    },
+    {
+      "epoch": 17.36,
+      "learning_rate": 9.422170859631967e-05,
+      "loss": 0.1085,
+      "step": 65100
+    },
+    {
+      "epoch": 17.36,
+      "eval_loss": 0.5427435636520386,
+      "eval_runtime": 19.097,
+      "eval_samples_per_second": 4.66,
+      "eval_wer": 0.36635006784260515,
+      "step": 65100
+    },
+    {
+      "epoch": 17.39,
+      "learning_rate": 9.42128189172371e-05,
+      "loss": 0.0943,
+      "step": 65200
+    },
+    {
+      "epoch": 17.39,
+      "eval_loss": 0.5481781959533691,
+      "eval_runtime": 19.0896,
+      "eval_samples_per_second": 4.662,
+      "eval_wer": 0.37449118046132973,
+      "step": 65200
+    },
+    {
+      "epoch": 17.41,
+      "learning_rate": 9.420392923815451e-05,
+      "loss": 0.0804,
+      "step": 65300
+    },
+    {
+      "epoch": 17.41,
+      "eval_loss": 0.6351197361946106,
+      "eval_runtime": 19.0622,
+      "eval_samples_per_second": 4.669,
+      "eval_wer": 0.36635006784260515,
+      "step": 65300
+    },
+    {
+      "epoch": 17.44,
+      "learning_rate": 9.419503955907192e-05,
+      "loss": 0.0988,
+      "step": 65400
+    },
+    {
+      "epoch": 17.44,
+      "eval_loss": 0.612962543964386,
+      "eval_runtime": 19.1497,
+      "eval_samples_per_second": 4.648,
+      "eval_wer": 0.3812754409769335,
+      "step": 65400
+    },
+    {
+      "epoch": 17.47,
+      "learning_rate": 9.418614987998934e-05,
+      "loss": 0.0912,
+      "step": 65500
+    },
+    {
+      "epoch": 17.47,
+      "eval_loss": 0.5913006067276001,
+      "eval_runtime": 19.135,
+      "eval_samples_per_second": 4.651,
+      "eval_wer": 0.378561736770692,
+      "step": 65500
+    },
+    {
+      "epoch": 17.49,
+      "learning_rate": 9.417726020090675e-05,
+      "loss": 0.1021,
+      "step": 65600
+    },
+    {
+      "epoch": 17.49,
+      "eval_loss": 0.6023339033126831,
+      "eval_runtime": 19.2593,
+      "eval_samples_per_second": 4.621,
+      "eval_wer": 0.40027137042062416,
+      "step": 65600
+    },
+    {
+      "epoch": 17.52,
+      "learning_rate": 9.416837052182417e-05,
+      "loss": 0.1073,
+      "step": 65700
+    },
+    {
+      "epoch": 17.52,
+      "eval_loss": 0.6285555362701416,
+      "eval_runtime": 19.4737,
+      "eval_samples_per_second": 4.57,
+      "eval_wer": 0.37991858887381275,
+      "step": 65700
+    },
+    {
+      "epoch": 17.55,
+      "learning_rate": 9.415948084274158e-05,
+      "loss": 0.1034,
+      "step": 65800
+    },
+    {
+      "epoch": 17.55,
+      "eval_loss": 0.5944674015045166,
+      "eval_runtime": 19.2042,
+      "eval_samples_per_second": 4.634,
+      "eval_wer": 0.373134328358209,
+      "step": 65800
+    },
+    {
+      "epoch": 17.57,
+      "learning_rate": 9.4150591163659e-05,
+      "loss": 0.0932,
+      "step": 65900
+    },
+    {
+      "epoch": 17.57,
+      "eval_loss": 0.6914730668067932,
+      "eval_runtime": 18.9736,
+      "eval_samples_per_second": 4.691,
+      "eval_wer": 0.38263229308005425,
+      "step": 65900
+    },
+    {
+      "epoch": 17.6,
+      "learning_rate": 9.414170148457642e-05,
+      "loss": 0.0986,
+      "step": 66000
+    },
+    {
+      "epoch": 17.6,
+      "eval_loss": 0.6024160981178284,
+      "eval_runtime": 19.0161,
+      "eval_samples_per_second": 4.68,
+      "eval_wer": 0.36092265943012214,
+      "step": 66000
+    },
+    {
+      "epoch": 17.63,
+      "learning_rate": 9.413281180549383e-05,
+      "loss": 0.0986,
+      "step": 66100
+    },
+    {
+      "epoch": 17.63,
+      "eval_loss": 0.6499736309051514,
+      "eval_runtime": 19.0981,
+      "eval_samples_per_second": 4.66,
+      "eval_wer": 0.3677069199457259,
+      "step": 66100
+    },
+    {
+      "epoch": 17.65,
+      "learning_rate": 9.412392212641125e-05,
+      "loss": 0.1021,
+      "step": 66200
+    },
+    {
+      "epoch": 17.65,
+      "eval_loss": 0.6112159490585327,
+      "eval_runtime": 19.2156,
+      "eval_samples_per_second": 4.632,
+      "eval_wer": 0.3812754409769335,
+      "step": 66200
+    },
+    {
+      "epoch": 17.68,
+      "learning_rate": 9.411503244732865e-05,
+      "loss": 0.0992,
+      "step": 66300
+    },
+    {
+      "epoch": 17.68,
+      "eval_loss": 0.6412532925605774,
+      "eval_runtime": 19.1259,
+      "eval_samples_per_second": 4.653,
+      "eval_wer": 0.37449118046132973,
+      "step": 66300
+    },
+    {
+      "epoch": 17.71,
+      "learning_rate": 9.410614276824607e-05,
+      "loss": 0.0927,
+      "step": 66400
+    },
+    {
+      "epoch": 17.71,
+      "eval_loss": 0.6437122225761414,
+      "eval_runtime": 19.2929,
+      "eval_samples_per_second": 4.613,
+      "eval_wer": 0.373134328358209,
+      "step": 66400
+    },
+    {
+      "epoch": 17.73,
+      "learning_rate": 9.409725308916348e-05,
+      "loss": 0.101,
+      "step": 66500
+    },
+    {
+      "epoch": 17.73,
+      "eval_loss": 0.5899493098258972,
+      "eval_runtime": 19.3697,
+      "eval_samples_per_second": 4.595,
+      "eval_wer": 0.373134328358209,
+      "step": 66500
+    },
+    {
+      "epoch": 17.76,
+      "learning_rate": 9.40883634100809e-05,
+      "loss": 0.0955,
+      "step": 66600
+    },
+    {
+      "epoch": 17.76,
+      "eval_loss": 0.5804283618927002,
+      "eval_runtime": 19.2234,
+      "eval_samples_per_second": 4.63,
+      "eval_wer": 0.3677069199457259,
+      "step": 66600
+    },
+    {
+      "epoch": 17.79,
+      "learning_rate": 9.407947373099831e-05,
+      "loss": 0.0897,
+      "step": 66700
+    },
+    {
+      "epoch": 17.79,
+      "eval_loss": 0.5792164206504822,
+      "eval_runtime": 19.3584,
+      "eval_samples_per_second": 4.597,
+      "eval_wer": 0.36092265943012214,
+      "step": 66700
+    },
+    {
+      "epoch": 17.81,
+      "learning_rate": 9.407058405191573e-05,
+      "loss": 0.1071,
+      "step": 66800
+    },
+    {
+      "epoch": 17.81,
+      "eval_loss": 0.5760857462882996,
+      "eval_runtime": 19.1199,
+      "eval_samples_per_second": 4.655,
+      "eval_wer": 0.36906377204884666,
+      "step": 66800
+    },
+    {
+      "epoch": 17.84,
+      "learning_rate": 9.406169437283314e-05,
+      "loss": 0.0983,
+      "step": 66900
+    },
+    {
+      "epoch": 17.84,
+      "eval_loss": 0.5856578946113586,
+      "eval_runtime": 19.1526,
+      "eval_samples_per_second": 4.647,
+      "eval_wer": 0.36363636363636365,
+      "step": 66900
+    },
+    {
+      "epoch": 17.87,
+      "learning_rate": 9.405280469375056e-05,
+      "loss": 0.089,
+      "step": 67000
+    },
+    {
+      "epoch": 17.87,
+      "eval_loss": 0.5810869932174683,
+      "eval_runtime": 19.1806,
+      "eval_samples_per_second": 4.64,
+      "eval_wer": 0.37177747625508817,
+      "step": 67000
+    },
+    {
+      "epoch": 17.89,
+      "learning_rate": 9.404391501466796e-05,
+      "loss": 0.1029,
+      "step": 67100
+    },
+    {
+      "epoch": 17.89,
+      "eval_loss": 0.5800544619560242,
+      "eval_runtime": 19.2236,
+      "eval_samples_per_second": 4.63,
+      "eval_wer": 0.3622795115332429,
+      "step": 67100
+    },
+    {
+      "epoch": 17.92,
+      "learning_rate": 9.403502533558539e-05,
+      "loss": 0.0915,
+      "step": 67200
+    },
+    {
+      "epoch": 17.92,
+      "eval_loss": 0.6390753984451294,
+      "eval_runtime": 19.2757,
+      "eval_samples_per_second": 4.617,
+      "eval_wer": 0.37720488466757124,
+      "step": 67200
+    },
+    {
+      "epoch": 17.95,
+      "learning_rate": 9.40261356565028e-05,
+      "loss": 0.093,
+      "step": 67300
+    },
+    {
+      "epoch": 17.95,
+      "eval_loss": 0.5873878002166748,
+      "eval_runtime": 19.0734,
+      "eval_samples_per_second": 4.666,
+      "eval_wer": 0.3622795115332429,
+      "step": 67300
+    },
+    {
+      "epoch": 17.97,
+      "learning_rate": 9.401724597742023e-05,
+      "loss": 0.0961,
+      "step": 67400
+    },
+    {
+      "epoch": 17.97,
+      "eval_loss": 0.6587142944335938,
+      "eval_runtime": 19.2431,
+      "eval_samples_per_second": 4.625,
+      "eval_wer": 0.37177747625508817,
+      "step": 67400
+    },
+    {
+      "epoch": 18.0,
+      "learning_rate": 9.400835629833763e-05,
+      "loss": 0.101,
+      "step": 67500
+    },
+    {
+      "epoch": 18.0,
+      "eval_loss": 0.6221853494644165,
+      "eval_runtime": 19.2864,
+      "eval_samples_per_second": 4.615,
+      "eval_wer": 0.37449118046132973,
+      "step": 67500
+    },
+    {
+      "epoch": 18.03,
+      "learning_rate": 9.399946661925505e-05,
+      "loss": 0.0845,
+      "step": 67600
+    },
+    {
+      "epoch": 18.03,
+      "eval_loss": 0.6697494983673096,
+      "eval_runtime": 19.3071,
+      "eval_samples_per_second": 4.61,
+      "eval_wer": 0.3649932157394844,
+      "step": 67600
+    },
+    {
+      "epoch": 18.05,
+      "learning_rate": 9.399057694017246e-05,
+      "loss": 0.092,
+      "step": 67700
+    },
+    {
+      "epoch": 18.05,
+      "eval_loss": 0.6346781253814697,
+      "eval_runtime": 19.6304,
+      "eval_samples_per_second": 4.534,
+      "eval_wer": 0.373134328358209,
+      "step": 67700
+    },
+    {
+      "epoch": 18.08,
+      "learning_rate": 9.398168726108988e-05,
+      "loss": 0.0865,
+      "step": 67800
+    },
+    {
+      "epoch": 18.08,
+      "eval_loss": 0.648298978805542,
+      "eval_runtime": 19.4446,
+      "eval_samples_per_second": 4.577,
+      "eval_wer": 0.37720488466757124,
+      "step": 67800
+    },
+    {
+      "epoch": 18.11,
+      "learning_rate": 9.39727975820073e-05,
+      "loss": 0.0982,
+      "step": 67900
+    },
+    {
+      "epoch": 18.11,
+      "eval_loss": 0.6014466285705566,
+      "eval_runtime": 19.2506,
+      "eval_samples_per_second": 4.623,
+      "eval_wer": 0.38263229308005425,
+      "step": 67900
+    },
+    {
+      "epoch": 18.13,
+      "learning_rate": 9.396390790292471e-05,
+      "loss": 0.0927,
+      "step": 68000
+    },
+    {
+      "epoch": 18.13,
+      "eval_loss": 0.6841628551483154,
+      "eval_runtime": 19.1038,
+      "eval_samples_per_second": 4.659,
+      "eval_wer": 0.35549525101763907,
+      "step": 68000
+    },
+    {
+      "epoch": 18.16,
+      "learning_rate": 9.395501822384213e-05,
+      "loss": 0.0899,
+      "step": 68100
+    },
+    {
+      "epoch": 18.16,
+      "eval_loss": 0.6094717979431152,
+      "eval_runtime": 19.3954,
+      "eval_samples_per_second": 4.589,
+      "eval_wer": 0.37177747625508817,
+      "step": 68100
+    },
+    {
+      "epoch": 18.19,
+      "learning_rate": 9.394612854475954e-05,
+      "loss": 0.0837,
+      "step": 68200
+    },
+    {
+      "epoch": 18.19,
+      "eval_loss": 0.5623308420181274,
+      "eval_runtime": 19.112,
+      "eval_samples_per_second": 4.657,
+      "eval_wer": 0.36906377204884666,
+      "step": 68200
+    },
+    {
+      "epoch": 18.21,
+      "learning_rate": 9.393723886567696e-05,
+      "loss": 0.0845,
+      "step": 68300
+    },
+    {
+      "epoch": 18.21,
+      "eval_loss": 0.6320222020149231,
+      "eval_runtime": 19.2195,
+      "eval_samples_per_second": 4.631,
+      "eval_wer": 0.3894165535956581,
+      "step": 68300
+    },
+    {
+      "epoch": 18.24,
+      "learning_rate": 9.392834918659437e-05,
+      "loss": 0.0903,
+      "step": 68400
+    },
+    {
+      "epoch": 18.24,
+      "eval_loss": 0.6020944118499756,
+      "eval_runtime": 18.9466,
+      "eval_samples_per_second": 4.697,
+      "eval_wer": 0.37449118046132973,
+      "step": 68400
+    },
+    {
+      "epoch": 18.27,
+      "learning_rate": 9.391945950751179e-05,
+      "loss": 0.0819,
+      "step": 68500
+    },
+    {
+      "epoch": 18.27,
+      "eval_loss": 0.6818872690200806,
+      "eval_runtime": 19.3994,
+      "eval_samples_per_second": 4.588,
+      "eval_wer": 0.378561736770692,
+      "step": 68500
+    },
+    {
+      "epoch": 18.29,
+      "learning_rate": 9.39105698284292e-05,
+      "loss": 0.0796,
+      "step": 68600
+    },
+    {
+      "epoch": 18.29,
+      "eval_loss": 0.7323005199432373,
+      "eval_runtime": 19.1244,
+      "eval_samples_per_second": 4.654,
+      "eval_wer": 0.378561736770692,
+      "step": 68600
+    },
+    {
+      "epoch": 18.32,
+      "learning_rate": 9.390168014934661e-05,
+      "loss": 0.1068,
+      "step": 68700
+    },
+    {
+      "epoch": 18.32,
+      "eval_loss": 0.67662513256073,
+      "eval_runtime": 19.4965,
+      "eval_samples_per_second": 4.565,
+      "eval_wer": 0.37449118046132973,
+      "step": 68700
+    },
+    {
+      "epoch": 18.35,
+      "learning_rate": 9.389279047026402e-05,
+      "loss": 0.0889,
+      "step": 68800
+    },
+    {
+      "epoch": 18.35,
+      "eval_loss": 0.6837645769119263,
+      "eval_runtime": 19.4823,
+      "eval_samples_per_second": 4.568,
+      "eval_wer": 0.36363636363636365,
+      "step": 68800
+    },
+    {
+      "epoch": 18.37,
+      "learning_rate": 9.388390079118144e-05,
+      "loss": 0.1016,
+      "step": 68900
+    },
+    {
+      "epoch": 18.37,
+      "eval_loss": 0.6688793301582336,
+      "eval_runtime": 19.1947,
+      "eval_samples_per_second": 4.637,
+      "eval_wer": 0.373134328358209,
+      "step": 68900
+    },
+    {
+      "epoch": 18.4,
+      "learning_rate": 9.387501111209885e-05,
+      "loss": 0.0906,
+      "step": 69000
+    },
+    {
+      "epoch": 18.4,
+      "eval_loss": 0.7079416513442993,
+      "eval_runtime": 19.217,
+      "eval_samples_per_second": 4.631,
+      "eval_wer": 0.3622795115332429,
+      "step": 69000
+    },
+    {
+      "epoch": 18.43,
+      "learning_rate": 9.386612143301627e-05,
+      "loss": 0.0896,
+      "step": 69100
+    },
+    {
+      "epoch": 18.43,
+      "eval_loss": 0.6737688183784485,
+      "eval_runtime": 19.0677,
+      "eval_samples_per_second": 4.668,
+      "eval_wer": 0.378561736770692,
+      "step": 69100
+    },
+    {
+      "epoch": 18.45,
+      "learning_rate": 9.385723175393368e-05,
+      "loss": 0.0907,
+      "step": 69200
+    },
+    {
+      "epoch": 18.45,
+      "eval_loss": 0.6276966333389282,
+      "eval_runtime": 19.1858,
+      "eval_samples_per_second": 4.639,
+      "eval_wer": 0.3853459972862958,
+      "step": 69200
+    },
+    {
+      "epoch": 18.48,
+      "learning_rate": 9.38483420748511e-05,
+      "loss": 0.0909,
+      "step": 69300
+    },
+    {
+      "epoch": 18.48,
+      "eval_loss": 0.6839456558227539,
+      "eval_runtime": 19.0866,
+      "eval_samples_per_second": 4.663,
+      "eval_wer": 0.383989145183175,
+      "step": 69300
+    },
+    {
+      "epoch": 18.51,
+      "learning_rate": 9.383945239576852e-05,
+      "loss": 0.0981,
+      "step": 69400
+    },
+    {
+      "epoch": 18.51,
+      "eval_loss": 0.6838169097900391,
+      "eval_runtime": 19.0242,
+      "eval_samples_per_second": 4.678,
+      "eval_wer": 0.3880597014925373,
+      "step": 69400
+    },
+    {
+      "epoch": 18.53,
+      "learning_rate": 9.383056271668594e-05,
+      "loss": 0.0935,
+      "step": 69500
+    },
+    {
+      "epoch": 18.53,
+      "eval_loss": 0.6975194215774536,
+      "eval_runtime": 19.5512,
+      "eval_samples_per_second": 4.552,
+      "eval_wer": 0.36906377204884666,
+      "step": 69500
+    },
+    {
+      "epoch": 18.56,
+      "learning_rate": 9.382167303760335e-05,
+      "loss": 0.0919,
+      "step": 69600
+    },
+    {
+      "epoch": 18.56,
+      "eval_loss": 0.6357317566871643,
+      "eval_runtime": 19.2594,
+      "eval_samples_per_second": 4.621,
+      "eval_wer": 0.37177747625508817,
+      "step": 69600
+    },
+    {
+      "epoch": 18.59,
+      "learning_rate": 9.381278335852077e-05,
+      "loss": 0.0915,
+      "step": 69700
+    },
+    {
+      "epoch": 18.59,
+      "eval_loss": 0.6521219611167908,
+      "eval_runtime": 19.1959,
+      "eval_samples_per_second": 4.636,
+      "eval_wer": 0.37177747625508817,
+      "step": 69700
+    },
+    {
+      "epoch": 18.61,
+      "learning_rate": 9.380389367943819e-05,
+      "loss": 0.0979,
+      "step": 69800
+    },
+    {
+      "epoch": 18.61,
+      "eval_loss": 0.6195235252380371,
+      "eval_runtime": 19.0592,
+      "eval_samples_per_second": 4.67,
+      "eval_wer": 0.3568521031207598,
+      "step": 69800
+    },
+    {
+      "epoch": 18.64,
+      "learning_rate": 9.37950040003556e-05,
+      "loss": 0.1043,
+      "step": 69900
+    },
+    {
+      "epoch": 18.64,
+      "eval_loss": 0.6711440086364746,
+      "eval_runtime": 19.2178,
+      "eval_samples_per_second": 4.631,
+      "eval_wer": 0.3582089552238806,
+      "step": 69900
+    },
+    {
+      "epoch": 18.67,
+      "learning_rate": 9.378611432127302e-05,
+      "loss": 0.0881,
+      "step": 70000
+    },
+    {
+      "epoch": 18.67,
+      "eval_loss": 0.6858513355255127,
+      "eval_runtime": 19.0654,
+      "eval_samples_per_second": 4.668,
+      "eval_wer": 0.37177747625508817,
+      "step": 70000
+    },
+    {
+      "epoch": 18.69,
+      "learning_rate": 9.377722464219042e-05,
+      "loss": 0.0903,
+      "step": 70100
+    },
+    {
+      "epoch": 18.69,
+      "eval_loss": 0.6942087411880493,
+      "eval_runtime": 19.3047,
+      "eval_samples_per_second": 4.61,
+      "eval_wer": 0.378561736770692,
+      "step": 70100
+    },
+    {
+      "epoch": 18.72,
+      "learning_rate": 9.376833496310784e-05,
+      "loss": 0.1022,
+      "step": 70200
+    },
+    {
+      "epoch": 18.72,
+      "eval_loss": 0.6641057729721069,
+      "eval_runtime": 19.2833,
+      "eval_samples_per_second": 4.615,
+      "eval_wer": 0.37449118046132973,
+      "step": 70200
+    },
+    {
+      "epoch": 18.75,
+      "learning_rate": 9.375944528402525e-05,
+      "loss": 0.0897,
+      "step": 70300
+    },
+    {
+      "epoch": 18.75,
+      "eval_loss": 0.6515685319900513,
+      "eval_runtime": 19.1116,
+      "eval_samples_per_second": 4.657,
+      "eval_wer": 0.37991858887381275,
+      "step": 70300
+    },
+    {
+      "epoch": 18.77,
+      "learning_rate": 9.375055560494267e-05,
+      "loss": 0.1011,
+      "step": 70400
+    },
+    {
+      "epoch": 18.77,
+      "eval_loss": 0.7358932495117188,
+      "eval_runtime": 19.077,
+      "eval_samples_per_second": 4.665,
+      "eval_wer": 0.37177747625508817,
+      "step": 70400
+    },
+    {
+      "epoch": 18.8,
+      "learning_rate": 9.374166592586008e-05,
+      "loss": 0.0928,
+      "step": 70500
+    },
+    {
+      "epoch": 18.8,
+      "eval_loss": 0.7245281338691711,
+      "eval_runtime": 19.1297,
+      "eval_samples_per_second": 4.652,
+      "eval_wer": 0.3704206241519674,
+      "step": 70500
+    },
+    {
+      "epoch": 18.83,
+      "learning_rate": 9.37327762467775e-05,
+      "loss": 0.0868,
+      "step": 70600
+    },
+    {
+      "epoch": 18.83,
+      "eval_loss": 0.7375543713569641,
+      "eval_runtime": 19.6091,
+      "eval_samples_per_second": 4.539,
+      "eval_wer": 0.38263229308005425,
+      "step": 70600
+    },
+    {
+      "epoch": 18.85,
+      "learning_rate": 9.37238865676949e-05,
+      "loss": 0.0868,
+      "step": 70700
+    },
+    {
+      "epoch": 18.85,
+      "eval_loss": 0.7681329846382141,
+      "eval_runtime": 19.1465,
+      "eval_samples_per_second": 4.648,
+      "eval_wer": 0.39755766621438265,
+      "step": 70700
+    },
+    {
+      "epoch": 18.88,
+      "learning_rate": 9.371499688861233e-05,
+      "loss": 0.0994,
+      "step": 70800
+    },
+    {
+      "epoch": 18.88,
+      "eval_loss": 0.69522625207901,
+      "eval_runtime": 19.0977,
+      "eval_samples_per_second": 4.66,
+      "eval_wer": 0.38263229308005425,
+      "step": 70800
+    },
+    {
+      "epoch": 18.91,
+      "learning_rate": 9.370610720952973e-05,
+      "loss": 0.1001,
+      "step": 70900
+    },
+    {
+      "epoch": 18.91,
+      "eval_loss": 0.6889650225639343,
+      "eval_runtime": 19.4849,
+      "eval_samples_per_second": 4.568,
+      "eval_wer": 0.37449118046132973,
+      "step": 70900
+    },
+    {
+      "epoch": 18.93,
+      "learning_rate": 9.369721753044715e-05,
+      "loss": 0.0892,
+      "step": 71000
+    },
+    {
+      "epoch": 18.93,
+      "eval_loss": 0.6742197871208191,
+      "eval_runtime": 19.3207,
+      "eval_samples_per_second": 4.606,
+      "eval_wer": 0.37991858887381275,
+      "step": 71000
+    },
+    {
+      "epoch": 18.96,
+      "learning_rate": 9.368832785136456e-05,
+      "loss": 0.0846,
+      "step": 71100
+    },
+    {
+      "epoch": 18.96,
+      "eval_loss": 0.7130357027053833,
+      "eval_runtime": 19.1415,
+      "eval_samples_per_second": 4.65,
+      "eval_wer": 0.3812754409769335,
+      "step": 71100
+    },
+    {
+      "epoch": 18.99,
+      "learning_rate": 9.367943817228198e-05,
+      "loss": 0.1046,
+      "step": 71200
+    },
+    {
+      "epoch": 18.99,
+      "eval_loss": 0.6631144881248474,
+      "eval_runtime": 19.2421,
+      "eval_samples_per_second": 4.625,
+      "eval_wer": 0.36906377204884666,
+      "step": 71200
+    },
+    {
+      "epoch": 19.01,
+      "learning_rate": 9.367054849319939e-05,
+      "loss": 0.0932,
+      "step": 71300
+    },
+    {
+      "epoch": 19.01,
+      "eval_loss": 0.6675189733505249,
+      "eval_runtime": 19.1744,
+      "eval_samples_per_second": 4.642,
+      "eval_wer": 0.373134328358209,
+      "step": 71300
+    },
+    {
+      "epoch": 19.04,
+      "learning_rate": 9.366165881411681e-05,
+      "loss": 0.0835,
+      "step": 71400
+    },
+    {
+      "epoch": 19.04,
+      "eval_loss": 0.6527154445648193,
+      "eval_runtime": 18.9998,
+      "eval_samples_per_second": 4.684,
+      "eval_wer": 0.37720488466757124,
+      "step": 71400
+    },
+    {
+      "epoch": 19.07,
+      "learning_rate": 9.365276913503423e-05,
+      "loss": 0.083,
+      "step": 71500
+    },
+    {
+      "epoch": 19.07,
+      "eval_loss": 0.6096330881118774,
+      "eval_runtime": 19.0633,
+      "eval_samples_per_second": 4.669,
+      "eval_wer": 0.3704206241519674,
+      "step": 71500
+    },
+    {
+      "epoch": 19.09,
+      "learning_rate": 9.364387945595164e-05,
+      "loss": 0.0911,
+      "step": 71600
+    },
+    {
+      "epoch": 19.09,
+      "eval_loss": 0.6261327266693115,
+      "eval_runtime": 19.155,
+      "eval_samples_per_second": 4.646,
+      "eval_wer": 0.38263229308005425,
+      "step": 71600
+    },
+    {
+      "epoch": 19.12,
+      "learning_rate": 9.363498977686906e-05,
+      "loss": 0.0859,
+      "step": 71700
+    },
+    {
+      "epoch": 19.12,
+      "eval_loss": 0.6865926384925842,
+      "eval_runtime": 19.2355,
+      "eval_samples_per_second": 4.627,
+      "eval_wer": 0.37720488466757124,
+      "step": 71700
+    },
+    {
+      "epoch": 19.15,
+      "learning_rate": 9.362610009778648e-05,
+      "loss": 0.0967,
+      "step": 71800
+    },
+    {
+      "epoch": 19.15,
+      "eval_loss": 0.6426628828048706,
+      "eval_runtime": 19.1929,
+      "eval_samples_per_second": 4.637,
+      "eval_wer": 0.37177747625508817,
+      "step": 71800
+    },
+    {
+      "epoch": 19.17,
+      "learning_rate": 9.36172104187039e-05,
+      "loss": 0.0799,
+      "step": 71900
+    },
+    {
+      "epoch": 19.17,
+      "eval_loss": 0.6814299821853638,
+      "eval_runtime": 19.1209,
+      "eval_samples_per_second": 4.655,
+      "eval_wer": 0.3921302578018996,
+      "step": 71900
+    },
+    {
+      "epoch": 19.2,
+      "learning_rate": 9.36083207396213e-05,
+      "loss": 0.082,
+      "step": 72000
+    },
+    {
+      "epoch": 19.2,
+      "eval_loss": 0.7020444273948669,
+      "eval_runtime": 19.1677,
+      "eval_samples_per_second": 4.643,
+      "eval_wer": 0.37720488466757124,
+      "step": 72000
+    },
+    {
+      "epoch": 19.23,
+      "learning_rate": 9.359943106053873e-05,
+      "loss": 0.0831,
+      "step": 72100
+    },
+    {
+      "epoch": 19.23,
+      "eval_loss": 0.673681378364563,
+      "eval_runtime": 19.5281,
+      "eval_samples_per_second": 4.558,
+      "eval_wer": 0.378561736770692,
+      "step": 72100
+    },
+    {
+      "epoch": 19.25,
+      "learning_rate": 9.359054138145613e-05,
+      "loss": 0.0884,
+      "step": 72200
+    },
+    {
+      "epoch": 19.25,
+      "eval_loss": 0.6359362006187439,
+      "eval_runtime": 19.1567,
+      "eval_samples_per_second": 4.646,
+      "eval_wer": 0.3853459972862958,
+      "step": 72200
+    },
+    {
+      "epoch": 19.28,
+      "learning_rate": 9.358165170237356e-05,
+      "loss": 0.0865,
+      "step": 72300
+    },
+    {
+      "epoch": 19.28,
+      "eval_loss": 0.6532277464866638,
+      "eval_runtime": 19.3623,
+      "eval_samples_per_second": 4.597,
+      "eval_wer": 0.3894165535956581,
+      "step": 72300
+    },
+    {
+      "epoch": 19.31,
+      "learning_rate": 9.357276202329096e-05,
+      "loss": 0.088,
+      "step": 72400
+    },
+    {
+      "epoch": 19.31,
+      "eval_loss": 0.6576082706451416,
+      "eval_runtime": 19.2282,
+      "eval_samples_per_second": 4.629,
+      "eval_wer": 0.3921302578018996,
+      "step": 72400
+    },
+    {
+      "epoch": 19.33,
+      "learning_rate": 9.356387234420838e-05,
+      "loss": 0.0896,
+      "step": 72500
+    },
+    {
+      "epoch": 19.33,
+      "eval_loss": 0.6283882856369019,
+      "eval_runtime": 19.8099,
+      "eval_samples_per_second": 4.493,
+      "eval_wer": 0.3989145183175034,
+      "step": 72500
+    },
+    {
+      "epoch": 19.36,
+      "learning_rate": 9.355498266512579e-05,
+      "loss": 0.0887,
+      "step": 72600
+    },
+    {
+      "epoch": 19.36,
+      "eval_loss": 0.6282336711883545,
+      "eval_runtime": 19.3166,
+      "eval_samples_per_second": 4.607,
+      "eval_wer": 0.3948439620081411,
+      "step": 72600
+    },
+    {
+      "epoch": 19.39,
+      "learning_rate": 9.354609298604321e-05,
+      "loss": 0.0859,
+      "step": 72700
+    },
+    {
+      "epoch": 19.39,
+      "eval_loss": 0.6422051191329956,
+      "eval_runtime": 19.1758,
+      "eval_samples_per_second": 4.641,
+      "eval_wer": 0.39077340569877883,
+      "step": 72700
+    },
+    {
+      "epoch": 19.41,
+      "learning_rate": 9.353720330696062e-05,
+      "loss": 0.0885,
+      "step": 72800
+    },
+    {
+      "epoch": 19.41,
+      "eval_loss": 0.621360719203949,
+      "eval_runtime": 19.3191,
+      "eval_samples_per_second": 4.607,
+      "eval_wer": 0.36906377204884666,
+      "step": 72800
+    },
+    {
+      "epoch": 19.44,
+      "learning_rate": 9.352831362787804e-05,
+      "loss": 0.0818,
+      "step": 72900
+    },
+    {
+      "epoch": 19.44,
+      "eval_loss": 0.6056548357009888,
+      "eval_runtime": 19.1978,
+      "eval_samples_per_second": 4.636,
+      "eval_wer": 0.3853459972862958,
+      "step": 72900
+    },
+    {
+      "epoch": 19.47,
+      "learning_rate": 9.351942394879545e-05,
+      "loss": 0.0923,
+      "step": 73000
+    },
+    {
+      "epoch": 19.47,
+      "eval_loss": 0.6582766175270081,
+      "eval_runtime": 19.4404,
+      "eval_samples_per_second": 4.578,
+      "eval_wer": 0.3948439620081411,
+      "step": 73000
+    },
+    {
+      "epoch": 19.49,
+      "learning_rate": 9.351053426971287e-05,
+      "loss": 0.0908,
+      "step": 73100
+    },
+    {
+      "epoch": 19.49,
+      "eval_loss": 0.6576439738273621,
+      "eval_runtime": 19.3463,
+      "eval_samples_per_second": 4.6,
+      "eval_wer": 0.39755766621438265,
+      "step": 73100
+    },
+    {
+      "epoch": 19.52,
+      "learning_rate": 9.350164459063027e-05,
+      "loss": 0.0917,
+      "step": 73200
+    },
+    {
+      "epoch": 19.52,
+      "eval_loss": 0.5941808819770813,
+      "eval_runtime": 19.2274,
+      "eval_samples_per_second": 4.629,
+      "eval_wer": 0.37720488466757124,
+      "step": 73200
+    },
+    {
+      "epoch": 19.55,
+      "learning_rate": 9.34927549115477e-05,
+      "loss": 0.0875,
+      "step": 73300
+    },
+    {
+      "epoch": 19.55,
+      "eval_loss": 0.6104390621185303,
+      "eval_runtime": 19.2515,
+      "eval_samples_per_second": 4.623,
+      "eval_wer": 0.37991858887381275,
+      "step": 73300
+    },
+    {
+      "epoch": 19.57,
+      "learning_rate": 9.348386523246511e-05,
+      "loss": 0.0784,
+      "step": 73400
+    },
+    {
+      "epoch": 19.57,
+      "eval_loss": 0.6506833434104919,
+      "eval_runtime": 19.4135,
+      "eval_samples_per_second": 4.584,
+      "eval_wer": 0.383989145183175,
+      "step": 73400
+    },
+    {
+      "epoch": 19.6,
+      "learning_rate": 9.347497555338252e-05,
+      "loss": 0.0964,
+      "step": 73500
+    },
+    {
+      "epoch": 19.6,
+      "eval_loss": 0.6429141163825989,
+      "eval_runtime": 19.1422,
+      "eval_samples_per_second": 4.649,
+      "eval_wer": 0.383989145183175,
+      "step": 73500
+    },
+    {
+      "epoch": 19.63,
+      "learning_rate": 9.346608587429994e-05,
+      "loss": 0.0888,
+      "step": 73600
+    },
+    {
+      "epoch": 19.63,
+      "eval_loss": 0.6477982401847839,
+      "eval_runtime": 19.0417,
+      "eval_samples_per_second": 4.674,
+      "eval_wer": 0.3853459972862958,
+      "step": 73600
+    },
+    {
+      "epoch": 19.65,
+      "learning_rate": 9.345719619521735e-05,
+      "loss": 0.0935,
+      "step": 73700
+    },
+    {
+      "epoch": 19.65,
+      "eval_loss": 0.6525067687034607,
+      "eval_runtime": 19.1616,
+      "eval_samples_per_second": 4.645,
+      "eval_wer": 0.4056987788331072,
+      "step": 73700
+    },
+    {
+      "epoch": 19.68,
+      "learning_rate": 9.344830651613477e-05,
+      "loss": 0.0896,
+      "step": 73800
+    },
+    {
+      "epoch": 19.68,
+      "eval_loss": 0.6392335295677185,
+      "eval_runtime": 19.1754,
+      "eval_samples_per_second": 4.641,
+      "eval_wer": 0.40298507462686567,
+      "step": 73800
+    },
+    {
+      "epoch": 19.71,
+      "learning_rate": 9.343941683705219e-05,
+      "loss": 0.088,
+      "step": 73900
+    },
+    {
+      "epoch": 19.71,
+      "eval_loss": 0.7739197611808777,
+      "eval_runtime": 19.1066,
+      "eval_samples_per_second": 4.658,
+      "eval_wer": 0.39077340569877883,
+      "step": 73900
+    },
+    {
+      "epoch": 19.73,
+      "learning_rate": 9.343052715796961e-05,
+      "loss": 0.0857,
+      "step": 74000
+    },
+    {
+      "epoch": 19.73,
+      "eval_loss": 0.6631598472595215,
+      "eval_runtime": 19.3706,
+      "eval_samples_per_second": 4.595,
+      "eval_wer": 0.38263229308005425,
+      "step": 74000
+    },
+    {
+      "epoch": 19.76,
+      "learning_rate": 9.342163747888702e-05,
+      "loss": 0.0934,
+      "step": 74100
+    },
+    {
+      "epoch": 19.76,
+      "eval_loss": 0.7168214917182922,
+      "eval_runtime": 19.3402,
+      "eval_samples_per_second": 4.602,
+      "eval_wer": 0.40705563093622793,
+      "step": 74100
+    },
+    {
+      "epoch": 19.79,
+      "learning_rate": 9.341274779980444e-05,
+      "loss": 0.0946,
+      "step": 74200
+    },
+    {
+      "epoch": 19.79,
+      "eval_loss": 0.6808632016181946,
+      "eval_runtime": 19.328,
+      "eval_samples_per_second": 4.605,
+      "eval_wer": 0.39077340569877883,
+      "step": 74200
+    },
+    {
+      "epoch": 19.81,
+      "learning_rate": 9.340385812072185e-05,
+      "loss": 0.0941,
+      "step": 74300
+    },
+    {
+      "epoch": 19.81,
+      "eval_loss": 0.6209776997566223,
+      "eval_runtime": 18.9844,
+      "eval_samples_per_second": 4.688,
+      "eval_wer": 0.38263229308005425,
+      "step": 74300
+    },
+    {
+      "epoch": 19.84,
+      "learning_rate": 9.339496844163927e-05,
+      "loss": 0.0972,
+      "step": 74400
+    },
+    {
+      "epoch": 19.84,
+      "eval_loss": 0.5869155526161194,
+      "eval_runtime": 19.5296,
+      "eval_samples_per_second": 4.557,
+      "eval_wer": 0.37991858887381275,
+      "step": 74400
+    },
+    {
+      "epoch": 19.87,
+      "learning_rate": 9.338607876255667e-05,
+      "loss": 0.0883,
+      "step": 74500
+    },
+    {
+      "epoch": 19.87,
+      "eval_loss": 0.6090362071990967,
+      "eval_runtime": 18.8546,
+      "eval_samples_per_second": 4.72,
+      "eval_wer": 0.3853459972862958,
+      "step": 74500
+    },
+    {
+      "epoch": 19.89,
+      "learning_rate": 9.33771890834741e-05,
+      "loss": 0.0951,
+      "step": 74600
+    },
+    {
+      "epoch": 19.89,
+      "eval_loss": 0.6196016669273376,
+      "eval_runtime": 18.9308,
+      "eval_samples_per_second": 4.701,
+      "eval_wer": 0.383989145183175,
+      "step": 74600
+    },
+    {
+      "epoch": 19.92,
+      "learning_rate": 9.33682994043915e-05,
+      "loss": 0.0971,
+      "step": 74700
+    },
+    {
+      "epoch": 19.92,
+      "eval_loss": 0.5604320168495178,
+      "eval_runtime": 19.1072,
+      "eval_samples_per_second": 4.658,
+      "eval_wer": 0.39620081411126185,
+      "step": 74700
+    },
+    {
+      "epoch": 19.95,
+      "learning_rate": 9.335940972530892e-05,
+      "loss": 0.0964,
+      "step": 74800
+    },
+    {
+      "epoch": 19.95,
+      "eval_loss": 0.5709301233291626,
+      "eval_runtime": 19.5052,
+      "eval_samples_per_second": 4.563,
+      "eval_wer": 0.38670284938941657,
+      "step": 74800
+    },
+    {
+      "epoch": 19.97,
+      "learning_rate": 9.335052004622633e-05,
+      "loss": 0.0984,
+      "step": 74900
+    },
+    {
+      "epoch": 19.97,
+      "eval_loss": 0.6479110717773438,
+      "eval_runtime": 18.9659,
+      "eval_samples_per_second": 4.693,
+      "eval_wer": 0.373134328358209,
+      "step": 74900
+    },
+    {
+      "epoch": 20.0,
+      "learning_rate": 9.334163036714375e-05,
+      "loss": 0.0874,
+      "step": 75000
+    },
+    {
+      "epoch": 20.0,
+      "eval_loss": 0.6282306909561157,
+      "eval_runtime": 18.9596,
+      "eval_samples_per_second": 4.694,
+      "eval_wer": 0.3853459972862958,
+      "step": 75000
+    },
+    {
+      "epoch": 20.03,
+      "learning_rate": 9.333274068806117e-05,
+      "loss": 0.0852,
+      "step": 75100
+    },
+    {
+      "epoch": 20.03,
+      "eval_loss": 0.5502228736877441,
+      "eval_runtime": 19.1592,
+      "eval_samples_per_second": 4.645,
+      "eval_wer": 0.38670284938941657,
+      "step": 75100
+    },
+    {
+      "epoch": 20.05,
+      "learning_rate": 9.332385100897858e-05,
+      "loss": 0.0806,
+      "step": 75200
+    },
+    {
+      "epoch": 20.05,
+      "eval_loss": 0.561530351638794,
+      "eval_runtime": 18.8221,
+      "eval_samples_per_second": 4.728,
+      "eval_wer": 0.3853459972862958,
+      "step": 75200
+    },
+    {
+      "epoch": 20.08,
+      "learning_rate": 9.3314961329896e-05,
+      "loss": 0.0807,
+      "step": 75300
+    },
+    {
+      "epoch": 20.08,
+      "eval_loss": 0.5761949419975281,
+      "eval_runtime": 19.3196,
+      "eval_samples_per_second": 4.607,
+      "eval_wer": 0.38263229308005425,
+      "step": 75300
+    },
+    {
+      "epoch": 20.11,
+      "learning_rate": 9.33060716508134e-05,
+      "loss": 0.0893,
+      "step": 75400
+    },
+    {
+      "epoch": 20.11,
+      "eval_loss": 0.5602518320083618,
+      "eval_runtime": 19.139,
+      "eval_samples_per_second": 4.65,
+      "eval_wer": 0.378561736770692,
+      "step": 75400
+    },
+    {
+      "epoch": 20.13,
+      "learning_rate": 9.329718197173083e-05,
+      "loss": 0.0887,
+      "step": 75500
+    },
+    {
+      "epoch": 20.13,
+      "eval_loss": 0.5833535194396973,
+      "eval_runtime": 19.1829,
+      "eval_samples_per_second": 4.64,
+      "eval_wer": 0.38263229308005425,
+      "step": 75500
+    },
+    {
+      "epoch": 20.16,
+      "learning_rate": 9.328829229264823e-05,
+      "loss": 0.0828,
+      "step": 75600
+    },
+    {
+      "epoch": 20.16,
+      "eval_loss": 0.6176598072052002,
+      "eval_runtime": 19.1408,
+      "eval_samples_per_second": 4.65,
+      "eval_wer": 0.36906377204884666,
+      "step": 75600
+    },
+    {
+      "epoch": 20.19,
+      "learning_rate": 9.327940261356565e-05,
+      "loss": 0.0851,
+      "step": 75700
+    },
+    {
+      "epoch": 20.19,
+      "eval_loss": 0.610307514667511,
+      "eval_runtime": 19.2991,
+      "eval_samples_per_second": 4.612,
+      "eval_wer": 0.373134328358209,
+      "step": 75700
+    },
+    {
+      "epoch": 20.21,
+      "learning_rate": 9.327051293448306e-05,
+      "loss": 0.0941,
+      "step": 75800
+    },
+    {
+      "epoch": 20.21,
+      "eval_loss": 0.6506506204605103,
+      "eval_runtime": 19.2155,
+      "eval_samples_per_second": 4.632,
+      "eval_wer": 0.378561736770692,
+      "step": 75800
+    },
+    {
+      "epoch": 20.24,
+      "learning_rate": 9.326162325540048e-05,
+      "loss": 0.0872,
+      "step": 75900
+    },
+    {
+      "epoch": 20.24,
+      "eval_loss": 0.608135461807251,
+      "eval_runtime": 19.1423,
+      "eval_samples_per_second": 4.649,
+      "eval_wer": 0.36906377204884666,
+      "step": 75900
+    },
+    {
+      "epoch": 20.27,
+      "learning_rate": 9.32527335763179e-05,
+      "loss": 0.0857,
+      "step": 76000
+    },
+    {
+      "epoch": 20.27,
+      "eval_loss": 0.710723876953125,
+      "eval_runtime": 19.2607,
+      "eval_samples_per_second": 4.621,
+      "eval_wer": 0.36635006784260515,
+      "step": 76000
+    },
+    {
+      "epoch": 20.29,
+      "learning_rate": 9.324384389723531e-05,
+      "loss": 0.0902,
+      "step": 76100
+    },
+    {
+      "epoch": 20.29,
+      "eval_loss": 0.6882886290550232,
+      "eval_runtime": 19.1457,
+      "eval_samples_per_second": 4.649,
+      "eval_wer": 0.3812754409769335,
+      "step": 76100
+    },
+    {
+      "epoch": 20.32,
+      "learning_rate": 9.323495421815273e-05,
+      "loss": 0.0903,
+      "step": 76200
+    },
+    {
+      "epoch": 20.32,
+      "eval_loss": 0.6835417747497559,
+      "eval_runtime": 19.1938,
+      "eval_samples_per_second": 4.637,
+      "eval_wer": 0.37720488466757124,
+      "step": 76200
+    },
+    {
+      "epoch": 20.35,
+      "learning_rate": 9.322606453907015e-05,
+      "loss": 0.0902,
+      "step": 76300
+    },
+    {
+      "epoch": 20.35,
+      "eval_loss": 0.6166879534721375,
+      "eval_runtime": 19.2325,
+      "eval_samples_per_second": 4.628,
+      "eval_wer": 0.373134328358209,
+      "step": 76300
+    },
+    {
+      "epoch": 20.37,
+      "learning_rate": 9.321717485998756e-05,
+      "loss": 0.0901,
+      "step": 76400
+    },
+    {
+      "epoch": 20.37,
+      "eval_loss": 0.7256922125816345,
+      "eval_runtime": 19.1718,
+      "eval_samples_per_second": 4.642,
+      "eval_wer": 0.37449118046132973,
+      "step": 76400
+    },
+    {
+      "epoch": 20.4,
+      "learning_rate": 9.320828518090498e-05,
+      "loss": 0.0853,
+      "step": 76500
+    },
+    {
+      "epoch": 20.4,
+      "eval_loss": 0.6770715713500977,
+      "eval_runtime": 19.3126,
+      "eval_samples_per_second": 4.608,
+      "eval_wer": 0.3812754409769335,
+      "step": 76500
+    },
+    {
+      "epoch": 20.43,
+      "learning_rate": 9.319939550182239e-05,
+      "loss": 0.0889,
+      "step": 76600
+    },
+    {
+      "epoch": 20.43,
+      "eval_loss": 0.6909158229827881,
+      "eval_runtime": 19.6021,
+      "eval_samples_per_second": 4.54,
+      "eval_wer": 0.38263229308005425,
+      "step": 76600
+    },
+    {
+      "epoch": 20.45,
+      "learning_rate": 9.319050582273981e-05,
+      "loss": 0.0822,
+      "step": 76700
+    },
+    {
+      "epoch": 20.45,
+      "eval_loss": 0.6846440434455872,
+      "eval_runtime": 19.3407,
+      "eval_samples_per_second": 4.602,
+      "eval_wer": 0.39348710990502034,
+      "step": 76700
+    },
+    {
+      "epoch": 20.48,
+      "learning_rate": 9.318161614365721e-05,
+      "loss": 0.0816,
+      "step": 76800
+    },
+    {
+      "epoch": 20.48,
+      "eval_loss": 0.6890459656715393,
+      "eval_runtime": 19.6124,
+      "eval_samples_per_second": 4.538,
+      "eval_wer": 0.3880597014925373,
+      "step": 76800
+    },
+    {
+      "epoch": 20.51,
+      "learning_rate": 9.317272646457463e-05,
+      "loss": 0.079,
+      "step": 76900
+    },
+    {
+      "epoch": 20.51,
+      "eval_loss": 0.7313602566719055,
+      "eval_runtime": 18.8506,
+      "eval_samples_per_second": 4.721,
+      "eval_wer": 0.3880597014925373,
+      "step": 76900
+    },
+    {
+      "epoch": 20.53,
+      "learning_rate": 9.316383678549206e-05,
+      "loss": 0.0803,
+      "step": 77000
+    },
+    {
+      "epoch": 20.53,
+      "eval_loss": 0.704619824886322,
+      "eval_runtime": 18.701,
+      "eval_samples_per_second": 4.759,
+      "eval_wer": 0.37177747625508817,
+      "step": 77000
+    },
+    {
+      "epoch": 20.56,
+      "learning_rate": 9.315494710640946e-05,
+      "loss": 0.0898,
+      "step": 77100
+    },
+    {
+      "epoch": 20.56,
+      "eval_loss": 0.6564654111862183,
+      "eval_runtime": 18.8369,
+      "eval_samples_per_second": 4.725,
+      "eval_wer": 0.3894165535956581,
+      "step": 77100
+    },
+    {
+      "epoch": 20.59,
+      "learning_rate": 9.314605742732688e-05,
+      "loss": 0.0917,
+      "step": 77200
+    },
+    {
+      "epoch": 20.59,
+      "eval_loss": 0.7199671268463135,
+      "eval_runtime": 18.8162,
+      "eval_samples_per_second": 4.73,
+      "eval_wer": 0.37991858887381275,
+      "step": 77200
+    },
+    {
+      "epoch": 20.61,
+      "learning_rate": 9.313716774824429e-05,
+      "loss": 0.0791,
+      "step": 77300
+    },
+    {
+      "epoch": 20.61,
+      "eval_loss": 0.6633087396621704,
+      "eval_runtime": 18.7345,
+      "eval_samples_per_second": 4.751,
+      "eval_wer": 0.373134328358209,
+      "step": 77300
+    },
+    {
+      "epoch": 20.64,
+      "learning_rate": 9.312827806916171e-05,
+      "loss": 0.0928,
+      "step": 77400
+    },
+    {
+      "epoch": 20.64,
+      "eval_loss": 0.6368780136108398,
+      "eval_runtime": 18.7701,
+      "eval_samples_per_second": 4.742,
+      "eval_wer": 0.36635006784260515,
+      "step": 77400
+    },
+    {
+      "epoch": 20.67,
+      "learning_rate": 9.311938839007912e-05,
+      "loss": 0.0892,
+      "step": 77500
+    },
+    {
+      "epoch": 20.67,
+      "eval_loss": 0.7035632133483887,
+      "eval_runtime": 18.977,
+      "eval_samples_per_second": 4.69,
+      "eval_wer": 0.3812754409769335,
+      "step": 77500
+    },
+    {
+      "epoch": 20.69,
+      "learning_rate": 9.311049871099654e-05,
+      "loss": 0.0776,
+      "step": 77600
+    },
+    {
+      "epoch": 20.69,
+      "eval_loss": 0.6616564393043518,
+      "eval_runtime": 18.991,
+      "eval_samples_per_second": 4.686,
+      "eval_wer": 0.37720488466757124,
+      "step": 77600
+    },
+    {
+      "epoch": 20.72,
+      "learning_rate": 9.310160903191395e-05,
+      "loss": 0.0865,
+      "step": 77700
+    },
+    {
+      "epoch": 20.72,
+      "eval_loss": 0.7506115436553955,
+      "eval_runtime": 18.8769,
+      "eval_samples_per_second": 4.715,
+      "eval_wer": 0.3758480325644505,
+      "step": 77700
+    },
+    {
+      "epoch": 20.75,
+      "learning_rate": 9.309271935283137e-05,
+      "loss": 0.0926,
+      "step": 77800
+    },
+    {
+      "epoch": 20.75,
+      "eval_loss": 0.6775010824203491,
+      "eval_runtime": 19.0093,
+      "eval_samples_per_second": 4.682,
+      "eval_wer": 0.378561736770692,
+      "step": 77800
+    },
+    {
+      "epoch": 20.77,
+      "learning_rate": 9.308382967374877e-05,
+      "loss": 0.0825,
+      "step": 77900
+    },
+    {
+      "epoch": 20.77,
+      "eval_loss": 0.660775363445282,
+      "eval_runtime": 18.876,
+      "eval_samples_per_second": 4.715,
+      "eval_wer": 0.37449118046132973,
+      "step": 77900
+    },
+    {
+      "epoch": 20.8,
+      "learning_rate": 9.30749399946662e-05,
+      "loss": 0.0866,
+      "step": 78000
+    },
+    {
+      "epoch": 20.8,
+      "eval_loss": 0.6021457314491272,
+      "eval_runtime": 18.6363,
+      "eval_samples_per_second": 4.776,
+      "eval_wer": 0.383989145183175,
+      "step": 78000
+    },
+    {
+      "epoch": 20.83,
+      "learning_rate": 9.30660503155836e-05,
+      "loss": 0.0962,
+      "step": 78100
+    },
+    {
+      "epoch": 20.83,
+      "eval_loss": 0.6218739748001099,
+      "eval_runtime": 18.656,
+      "eval_samples_per_second": 4.771,
+      "eval_wer": 0.3758480325644505,
+      "step": 78100
+    },
+    {
+      "epoch": 20.85,
+      "learning_rate": 9.305716063650102e-05,
+      "loss": 0.0951,
+      "step": 78200
+    },
+    {
+      "epoch": 20.85,
+      "eval_loss": 0.6719256043434143,
+      "eval_runtime": 18.7202,
+      "eval_samples_per_second": 4.754,
+      "eval_wer": 0.383989145183175,
+      "step": 78200
+    },
+    {
+      "epoch": 20.88,
+      "learning_rate": 9.304827095741844e-05,
+      "loss": 0.0806,
+      "step": 78300
+    },
+    {
+      "epoch": 20.88,
+      "eval_loss": 0.6920618414878845,
+      "eval_runtime": 18.8869,
+      "eval_samples_per_second": 4.712,
+      "eval_wer": 0.37177747625508817,
+      "step": 78300
+    },
+    {
+      "epoch": 20.91,
+      "learning_rate": 9.303938127833586e-05,
+      "loss": 0.0896,
+      "step": 78400
+    },
+    {
+      "epoch": 20.91,
+      "eval_loss": 0.7108328342437744,
+      "eval_runtime": 18.5663,
+      "eval_samples_per_second": 4.794,
+      "eval_wer": 0.383989145183175,
+      "step": 78400
+    },
+    {
+      "epoch": 20.93,
+      "learning_rate": 9.303049159925327e-05,
+      "loss": 0.0932,
+      "step": 78500
+    },
+    {
+      "epoch": 20.93,
+      "eval_loss": 0.7304340600967407,
+      "eval_runtime": 18.6578,
+      "eval_samples_per_second": 4.77,
+      "eval_wer": 0.36906377204884666,
+      "step": 78500
+    },
+    {
+      "epoch": 20.96,
+      "learning_rate": 9.302160192017069e-05,
+      "loss": 0.0869,
+      "step": 78600
+    },
+    {
+      "epoch": 20.96,
+      "eval_loss": 0.6971027255058289,
+      "eval_runtime": 18.645,
+      "eval_samples_per_second": 4.773,
+      "eval_wer": 0.3649932157394844,
+      "step": 78600
+    },
+    {
+      "epoch": 20.99,
+      "learning_rate": 9.30127122410881e-05,
+      "loss": 0.0993,
+      "step": 78700
+    },
+    {
+      "epoch": 20.99,
+      "eval_loss": 0.7250904440879822,
+      "eval_runtime": 18.5993,
+      "eval_samples_per_second": 4.785,
+      "eval_wer": 0.37991858887381275,
+      "step": 78700
+    },
+    {
+      "epoch": 21.01,
+      "learning_rate": 9.300382256200552e-05,
+      "loss": 0.0898,
+      "step": 78800
+    },
+    {
+      "epoch": 21.01,
+      "eval_loss": 0.6961299777030945,
+      "eval_runtime": 18.7464,
+      "eval_samples_per_second": 4.748,
+      "eval_wer": 0.37720488466757124,
+      "step": 78800
+    },
+    {
+      "epoch": 21.04,
+      "learning_rate": 9.299493288292294e-05,
+      "loss": 0.0823,
+      "step": 78900
+    },
+    {
+      "epoch": 21.04,
+      "eval_loss": 0.7197515368461609,
+      "eval_runtime": 18.7633,
+      "eval_samples_per_second": 4.743,
+      "eval_wer": 0.39077340569877883,
+      "step": 78900
+    },
+    {
+      "epoch": 21.07,
+      "learning_rate": 9.298604320384035e-05,
+      "loss": 0.0858,
+      "step": 79000
+    },
+    {
+      "epoch": 21.07,
+      "eval_loss": 0.7113422155380249,
+      "eval_runtime": 18.6224,
+      "eval_samples_per_second": 4.779,
+      "eval_wer": 0.3758480325644505,
+      "step": 79000
+    },
+    {
+      "epoch": 21.09,
+      "learning_rate": 9.297715352475777e-05,
+      "loss": 0.0843,
+      "step": 79100
+    },
+    {
+      "epoch": 21.09,
+      "eval_loss": 0.7641347646713257,
+      "eval_runtime": 18.7632,
+      "eval_samples_per_second": 4.743,
+      "eval_wer": 0.38263229308005425,
+      "step": 79100
+    },
+    {
+      "epoch": 21.12,
+      "learning_rate": 9.296826384567517e-05,
+      "loss": 0.0783,
+      "step": 79200
+    },
+    {
+      "epoch": 21.12,
+      "eval_loss": 0.7237738370895386,
+      "eval_runtime": 18.7137,
+      "eval_samples_per_second": 4.756,
+      "eval_wer": 0.3812754409769335,
+      "step": 79200
+    },
+    {
+      "epoch": 21.15,
+      "learning_rate": 9.29593741665926e-05,
+      "loss": 0.0932,
+      "step": 79300
+    },
+    {
+      "epoch": 21.15,
+      "eval_loss": 0.690632700920105,
+      "eval_runtime": 18.6167,
+      "eval_samples_per_second": 4.781,
+      "eval_wer": 0.3812754409769335,
+      "step": 79300
+    },
+    {
+      "epoch": 21.17,
+      "learning_rate": 9.295048448751e-05,
+      "loss": 0.0813,
+      "step": 79400
+    },
+    {
+      "epoch": 21.17,
+      "eval_loss": 0.7066590189933777,
+      "eval_runtime": 18.5705,
+      "eval_samples_per_second": 4.793,
+      "eval_wer": 0.3812754409769335,
+      "step": 79400
+    },
+    {
+      "epoch": 21.2,
+      "learning_rate": 9.294159480842742e-05,
+      "loss": 0.0852,
+      "step": 79500
+    },
+    {
+      "epoch": 21.2,
+      "eval_loss": 0.7364128232002258,
+      "eval_runtime": 18.5328,
+      "eval_samples_per_second": 4.802,
+      "eval_wer": 0.3880597014925373,
+      "step": 79500
+    },
+    {
+      "epoch": 21.23,
+      "learning_rate": 9.293270512934483e-05,
+      "loss": 0.0812,
+      "step": 79600
+    },
+    {
+      "epoch": 21.23,
+      "eval_loss": 0.7462046146392822,
+      "eval_runtime": 18.7006,
+      "eval_samples_per_second": 4.759,
+      "eval_wer": 0.37991858887381275,
+      "step": 79600
+    },
+    {
+      "epoch": 21.25,
+      "learning_rate": 9.292381545026225e-05,
+      "loss": 0.0859,
+      "step": 79700
+    },
+    {
+      "epoch": 21.25,
+      "eval_loss": 0.7259342074394226,
+      "eval_runtime": 18.8564,
+      "eval_samples_per_second": 4.72,
+      "eval_wer": 0.3880597014925373,
+      "step": 79700
+    },
+    {
+      "epoch": 21.28,
+      "learning_rate": 9.291492577117966e-05,
+      "loss": 0.0799,
+      "step": 79800
+    },
+    {
+      "epoch": 21.28,
+      "eval_loss": 0.7363275289535522,
+      "eval_runtime": 18.6509,
+      "eval_samples_per_second": 4.772,
+      "eval_wer": 0.383989145183175,
+      "step": 79800
+    },
+    {
+      "epoch": 21.31,
+      "learning_rate": 9.290603609209708e-05,
+      "loss": 0.0738,
+      "step": 79900
+    },
+    {
+      "epoch": 21.31,
+      "eval_loss": 0.7374128699302673,
+      "eval_runtime": 18.6603,
+      "eval_samples_per_second": 4.769,
+      "eval_wer": 0.38263229308005425,
+      "step": 79900
+    },
+    {
+      "epoch": 21.33,
+      "learning_rate": 9.289714641301449e-05,
+      "loss": 0.0845,
+      "step": 80000
+    },
+    {
+      "epoch": 21.33,
+      "eval_loss": 0.7262102365493774,
+      "eval_runtime": 18.5367,
+      "eval_samples_per_second": 4.801,
+      "eval_wer": 0.3853459972862958,
+      "step": 80000
+    },
+    {
+      "epoch": 21.36,
+      "learning_rate": 9.28882567339319e-05,
+      "loss": 0.0857,
+      "step": 80100
+    },
+    {
+      "epoch": 21.36,
+      "eval_loss": 0.6921221017837524,
+      "eval_runtime": 18.5898,
+      "eval_samples_per_second": 4.788,
+      "eval_wer": 0.3812754409769335,
+      "step": 80100
+    },
+    {
+      "epoch": 21.39,
+      "learning_rate": 9.287936705484931e-05,
+      "loss": 0.0868,
+      "step": 80200
+    },
+    {
+      "epoch": 21.39,
+      "eval_loss": 0.6997746825218201,
+      "eval_runtime": 18.6265,
+      "eval_samples_per_second": 4.778,
+      "eval_wer": 0.378561736770692,
+      "step": 80200
+    },
+    {
+      "epoch": 21.41,
+      "learning_rate": 9.287047737576673e-05,
+      "loss": 0.0806,
+      "step": 80300
+    },
+    {
+      "epoch": 21.41,
+      "eval_loss": 0.7043644189834595,
+      "eval_runtime": 18.5897,
+      "eval_samples_per_second": 4.788,
+      "eval_wer": 0.39077340569877883,
+      "step": 80300
+    },
+    {
+      "epoch": 21.44,
+      "learning_rate": 9.286158769668416e-05,
+      "loss": 0.0773,
+      "step": 80400
+    },
+    {
+      "epoch": 21.44,
+      "eval_loss": 0.6518604159355164,
+      "eval_runtime": 18.5303,
+      "eval_samples_per_second": 4.803,
+      "eval_wer": 0.37720488466757124,
+      "step": 80400
+    },
+    {
+      "epoch": 21.47,
+      "learning_rate": 9.285269801760158e-05,
+      "loss": 0.0866,
+      "step": 80500
+    },
+    {
+      "epoch": 21.47,
+      "eval_loss": 0.6679775714874268,
+      "eval_runtime": 18.5768,
+      "eval_samples_per_second": 4.791,
+      "eval_wer": 0.3812754409769335,
+      "step": 80500
+    },
+    {
+      "epoch": 21.49,
+      "learning_rate": 9.284380833851898e-05,
+      "loss": 0.0829,
+      "step": 80600
+    },
+    {
+      "epoch": 21.49,
+      "eval_loss": 0.6407601237297058,
+      "eval_runtime": 18.488,
+      "eval_samples_per_second": 4.814,
+      "eval_wer": 0.38263229308005425,
+      "step": 80600
+    },
+    {
+      "epoch": 21.52,
+      "learning_rate": 9.28349186594364e-05,
+      "loss": 0.0937,
+      "step": 80700
+    },
+    {
+      "epoch": 21.52,
+      "eval_loss": 0.6457295417785645,
+      "eval_runtime": 18.5871,
+      "eval_samples_per_second": 4.788,
+      "eval_wer": 0.40027137042062416,
+      "step": 80700
+    },
+    {
+      "epoch": 21.55,
+      "learning_rate": 9.282602898035382e-05,
+      "loss": 0.0822,
+      "step": 80800
+    },
+    {
+      "epoch": 21.55,
+      "eval_loss": 0.6667584776878357,
+      "eval_runtime": 18.667,
+      "eval_samples_per_second": 4.768,
+      "eval_wer": 0.38263229308005425,
+      "step": 80800
+    },
+    {
+      "epoch": 21.57,
+      "learning_rate": 9.281713930127123e-05,
+      "loss": 0.0838,
+      "step": 80900
+    },
+    {
+      "epoch": 21.57,
+      "eval_loss": 0.688599705696106,
+      "eval_runtime": 18.7488,
+      "eval_samples_per_second": 4.747,
+      "eval_wer": 0.37720488466757124,
+      "step": 80900
+    },
+    {
+      "epoch": 21.6,
+      "learning_rate": 9.280824962218865e-05,
+      "loss": 0.0904,
+      "step": 81000
+    },
+    {
+      "epoch": 21.6,
+      "eval_loss": 0.6245527863502502,
+      "eval_runtime": 18.5514,
+      "eval_samples_per_second": 4.797,
+      "eval_wer": 0.3853459972862958,
+      "step": 81000
+    },
+    {
+      "epoch": 21.63,
+      "learning_rate": 9.279935994310606e-05,
+      "loss": 0.0846,
+      "step": 81100
+    },
+    {
+      "epoch": 21.63,
+      "eval_loss": 0.6749595999717712,
+      "eval_runtime": 18.636,
+      "eval_samples_per_second": 4.776,
+      "eval_wer": 0.37720488466757124,
+      "step": 81100
+    },
+    {
+      "epoch": 21.65,
+      "learning_rate": 9.279047026402348e-05,
+      "loss": 0.0886,
+      "step": 81200
+    },
+    {
+      "epoch": 21.65,
+      "eval_loss": 0.6561444401741028,
+      "eval_runtime": 18.6763,
+      "eval_samples_per_second": 4.765,
+      "eval_wer": 0.37991858887381275,
+      "step": 81200
+    },
+    {
+      "epoch": 21.68,
+      "learning_rate": 9.278158058494089e-05,
+      "loss": 0.0811,
+      "step": 81300
+    },
+    {
+      "epoch": 21.68,
+      "eval_loss": 0.641154944896698,
+      "eval_runtime": 18.6518,
+      "eval_samples_per_second": 4.772,
+      "eval_wer": 0.39077340569877883,
+      "step": 81300
+    },
+    {
+      "epoch": 21.71,
+      "learning_rate": 9.277269090585831e-05,
+      "loss": 0.0841,
+      "step": 81400
+    },
+    {
+      "epoch": 21.71,
+      "eval_loss": 0.5872231125831604,
+      "eval_runtime": 18.7494,
+      "eval_samples_per_second": 4.747,
+      "eval_wer": 0.3921302578018996,
+      "step": 81400
+    },
+    {
+      "epoch": 21.73,
+      "learning_rate": 9.276380122677571e-05,
+      "loss": 0.0768,
+      "step": 81500
+    },
+    {
+      "epoch": 21.73,
+      "eval_loss": 0.5869088172912598,
+      "eval_runtime": 18.6444,
+      "eval_samples_per_second": 4.774,
+      "eval_wer": 0.36635006784260515,
+      "step": 81500
+    },
+    {
+      "epoch": 21.76,
+      "learning_rate": 9.275491154769314e-05,
+      "loss": 0.0883,
+      "step": 81600
+    },
+    {
+      "epoch": 21.76,
+      "eval_loss": 0.6907068490982056,
+      "eval_runtime": 18.5577,
+      "eval_samples_per_second": 4.796,
+      "eval_wer": 0.378561736770692,
+      "step": 81600
+    },
+    {
+      "epoch": 21.79,
+      "learning_rate": 9.274602186861054e-05,
+      "loss": 0.0772,
+      "step": 81700
+    },
+    {
+      "epoch": 21.79,
+      "eval_loss": 0.6918976306915283,
+      "eval_runtime": 18.6434,
+      "eval_samples_per_second": 4.774,
+      "eval_wer": 0.37720488466757124,
+      "step": 81700
+    },
+    {
+      "epoch": 21.81,
+      "learning_rate": 9.273713218952796e-05,
+      "loss": 0.085,
+      "step": 81800
+    },
+    {
+      "epoch": 21.81,
+      "eval_loss": 0.7292914390563965,
+      "eval_runtime": 18.8083,
+      "eval_samples_per_second": 4.732,
+      "eval_wer": 0.39077340569877883,
+      "step": 81800
+    },
+    {
+      "epoch": 21.84,
+      "learning_rate": 9.272824251044537e-05,
+      "loss": 0.0842,
+      "step": 81900
+    },
+    {
+      "epoch": 21.84,
+      "eval_loss": 0.7436701059341431,
+      "eval_runtime": 18.5721,
+      "eval_samples_per_second": 4.792,
+      "eval_wer": 0.3880597014925373,
+      "step": 81900
+    },
+    {
+      "epoch": 21.87,
+      "learning_rate": 9.271935283136279e-05,
+      "loss": 0.0897,
+      "step": 82000
+    },
+    {
+      "epoch": 21.87,
+      "eval_loss": 0.6747327446937561,
+      "eval_runtime": 18.6027,
+      "eval_samples_per_second": 4.784,
+      "eval_wer": 0.3880597014925373,
+      "step": 82000
+    },
+    {
+      "epoch": 21.89,
+      "learning_rate": 9.27104631522802e-05,
+      "loss": 0.0856,
+      "step": 82100
+    },
+    {
+      "epoch": 21.89,
+      "eval_loss": 0.6943639516830444,
+      "eval_runtime": 18.5437,
+      "eval_samples_per_second": 4.799,
+      "eval_wer": 0.37449118046132973,
+      "step": 82100
+    },
+    {
+      "epoch": 21.92,
+      "learning_rate": 9.270157347319762e-05,
+      "loss": 0.0866,
+      "step": 82200
+    },
+    {
+      "epoch": 21.92,
+      "eval_loss": 0.6735579371452332,
+      "eval_runtime": 18.5475,
+      "eval_samples_per_second": 4.798,
+      "eval_wer": 0.3894165535956581,
+      "step": 82200
+    },
+    {
+      "epoch": 21.95,
+      "learning_rate": 9.269268379411503e-05,
+      "loss": 0.0743,
+      "step": 82300
+    },
+    {
+      "epoch": 21.95,
+      "eval_loss": 0.6760044693946838,
+      "eval_runtime": 18.6158,
+      "eval_samples_per_second": 4.781,
+      "eval_wer": 0.3853459972862958,
+      "step": 82300
+    },
+    {
+      "epoch": 21.97,
+      "learning_rate": 9.268379411503245e-05,
+      "loss": 0.0745,
+      "step": 82400
+    },
+    {
+      "epoch": 21.97,
+      "eval_loss": 0.6835464239120483,
+      "eval_runtime": 18.8119,
+      "eval_samples_per_second": 4.731,
+      "eval_wer": 0.37991858887381275,
+      "step": 82400
+    },
+    {
+      "epoch": 22.0,
+      "learning_rate": 9.267490443594987e-05,
+      "loss": 0.092,
+      "step": 82500
+    },
+    {
+      "epoch": 22.0,
+      "eval_loss": 0.6800022125244141,
+      "eval_runtime": 18.7917,
+      "eval_samples_per_second": 4.736,
+      "eval_wer": 0.36906377204884666,
+      "step": 82500
+    },
+    {
+      "epoch": 22.03,
+      "learning_rate": 9.266601475686727e-05,
+      "loss": 0.085,
+      "step": 82600
+    },
+    {
+      "epoch": 22.03,
+      "eval_loss": 0.8254941701889038,
+      "eval_runtime": 18.766,
+      "eval_samples_per_second": 4.743,
+      "eval_wer": 0.3758480325644505,
+      "step": 82600
+    },
+    {
+      "epoch": 22.05,
+      "learning_rate": 9.26571250777847e-05,
+      "loss": 0.0811,
+      "step": 82700
+    },
+    {
+      "epoch": 22.05,
+      "eval_loss": 0.6936663389205933,
+      "eval_runtime": 18.6661,
+      "eval_samples_per_second": 4.768,
+      "eval_wer": 0.36906377204884666,
+      "step": 82700
+    },
+    {
+      "epoch": 22.08,
+      "learning_rate": 9.264823539870212e-05,
+      "loss": 0.0793,
+      "step": 82800
+    },
+    {
+      "epoch": 22.08,
+      "eval_loss": 0.6859510540962219,
+      "eval_runtime": 18.9605,
+      "eval_samples_per_second": 4.694,
+      "eval_wer": 0.383989145183175,
+      "step": 82800
+    },
+    {
+      "epoch": 22.11,
+      "learning_rate": 9.263934571961954e-05,
+      "loss": 0.0809,
+      "step": 82900
+    },
+    {
+      "epoch": 22.11,
+      "eval_loss": 0.7462514638900757,
+      "eval_runtime": 18.5081,
+      "eval_samples_per_second": 4.809,
+      "eval_wer": 0.3812754409769335,
+      "step": 82900
+    },
+    {
+      "epoch": 22.13,
+      "learning_rate": 9.263045604053694e-05,
+      "loss": 0.0754,
+      "step": 83000
+    },
+    {
+      "epoch": 22.13,
+      "eval_loss": 0.7102154493331909,
+      "eval_runtime": 18.7407,
+      "eval_samples_per_second": 4.749,
+      "eval_wer": 0.378561736770692,
+      "step": 83000
+    },
+    {
+      "epoch": 22.16,
+      "learning_rate": 9.262156636145436e-05,
+      "loss": 0.0797,
+      "step": 83100
+    },
+    {
+      "epoch": 22.16,
+      "eval_loss": 0.7404577136039734,
+      "eval_runtime": 18.7167,
+      "eval_samples_per_second": 4.755,
+      "eval_wer": 0.39077340569877883,
+      "step": 83100
+    },
+    {
+      "epoch": 22.19,
+      "learning_rate": 9.261267668237177e-05,
+      "loss": 0.0779,
+      "step": 83200
+    },
+    {
+      "epoch": 22.19,
+      "eval_loss": 0.7697685956954956,
+      "eval_runtime": 18.595,
+      "eval_samples_per_second": 4.786,
+      "eval_wer": 0.3853459972862958,
+      "step": 83200
+    },
+    {
+      "epoch": 22.21,
+      "learning_rate": 9.260378700328919e-05,
+      "loss": 0.0742,
+      "step": 83300
+    },
+    {
+      "epoch": 22.21,
+      "eval_loss": 0.7799381613731384,
+      "eval_runtime": 18.6827,
+      "eval_samples_per_second": 4.764,
+      "eval_wer": 0.38263229308005425,
+      "step": 83300
+    },
+    {
+      "epoch": 22.24,
+      "learning_rate": 9.25948973242066e-05,
+      "loss": 0.0846,
+      "step": 83400
+    },
+    {
+      "epoch": 22.24,
+      "eval_loss": 0.7063450217247009,
+      "eval_runtime": 18.7388,
+      "eval_samples_per_second": 4.75,
+      "eval_wer": 0.38263229308005425,
+      "step": 83400
+    },
+    {
+      "epoch": 22.27,
+      "learning_rate": 9.258600764512402e-05,
+      "loss": 0.0834,
+      "step": 83500
+    },
+    {
+      "epoch": 22.27,
+      "eval_loss": 0.7352509498596191,
+      "eval_runtime": 18.617,
+      "eval_samples_per_second": 4.781,
+      "eval_wer": 0.378561736770692,
+      "step": 83500
+    },
+    {
+      "epoch": 22.29,
+      "learning_rate": 9.257711796604143e-05,
+      "loss": 0.0921,
+      "step": 83600
+    },
+    {
+      "epoch": 22.29,
+      "eval_loss": 0.7542688846588135,
+      "eval_runtime": 18.5997,
+      "eval_samples_per_second": 4.785,
+      "eval_wer": 0.38670284938941657,
+      "step": 83600
+    },
+    {
+      "epoch": 22.32,
+      "learning_rate": 9.256822828695885e-05,
+      "loss": 0.0814,
+      "step": 83700
+    },
+    {
+      "epoch": 22.32,
+      "eval_loss": 0.7032181620597839,
+      "eval_runtime": 18.6162,
+      "eval_samples_per_second": 4.781,
+      "eval_wer": 0.38670284938941657,
+      "step": 83700
+    },
+    {
+      "epoch": 22.35,
+      "learning_rate": 9.255933860787625e-05,
+      "loss": 0.0849,
+      "step": 83800
+    },
+    {
+      "epoch": 22.35,
+      "eval_loss": 0.7397233843803406,
+      "eval_runtime": 18.8613,
+      "eval_samples_per_second": 4.719,
+      "eval_wer": 0.3880597014925373,
+      "step": 83800
+    },
+    {
+      "epoch": 22.37,
+      "learning_rate": 9.255044892879368e-05,
+      "loss": 0.0792,
+      "step": 83900
+    },
+    {
+      "epoch": 22.37,
+      "eval_loss": 0.7147604823112488,
+      "eval_runtime": 18.6986,
+      "eval_samples_per_second": 4.76,
+      "eval_wer": 0.3758480325644505,
+      "step": 83900
+    },
+    {
+      "epoch": 22.4,
+      "learning_rate": 9.254155924971108e-05,
+      "loss": 0.0788,
+      "step": 84000
+    },
+    {
+      "epoch": 22.4,
+      "eval_loss": 0.7215421795845032,
+      "eval_runtime": 18.5868,
+      "eval_samples_per_second": 4.788,
+      "eval_wer": 0.3758480325644505,
+      "step": 84000
+    },
+    {
+      "epoch": 22.43,
+      "learning_rate": 9.25326695706285e-05,
+      "loss": 0.0823,
+      "step": 84100
+    },
+    {
+      "epoch": 22.43,
+      "eval_loss": 0.7105488777160645,
+      "eval_runtime": 18.5634,
+      "eval_samples_per_second": 4.794,
+      "eval_wer": 0.37991858887381275,
+      "step": 84100
+    },
+    {
+      "epoch": 22.45,
+      "learning_rate": 9.252377989154591e-05,
+      "loss": 0.085,
+      "step": 84200
+    },
+    {
+      "epoch": 22.45,
+      "eval_loss": 0.7314173579216003,
+      "eval_runtime": 18.7643,
+      "eval_samples_per_second": 4.743,
+      "eval_wer": 0.3853459972862958,
+      "step": 84200
+    },
+    {
+      "epoch": 22.48,
+      "learning_rate": 9.251489021246333e-05,
+      "loss": 0.0951,
+      "step": 84300
+    },
+    {
+      "epoch": 22.48,
+      "eval_loss": 0.6925246119499207,
+      "eval_runtime": 18.6839,
+      "eval_samples_per_second": 4.763,
+      "eval_wer": 0.3704206241519674,
+      "step": 84300
+    },
+    {
+      "epoch": 22.51,
+      "learning_rate": 9.250600053338075e-05,
+      "loss": 0.0823,
+      "step": 84400
+    },
+    {
+      "epoch": 22.51,
+      "eval_loss": 0.7289780378341675,
+      "eval_runtime": 18.8295,
+      "eval_samples_per_second": 4.727,
+      "eval_wer": 0.3894165535956581,
+      "step": 84400
+    },
+    {
+      "epoch": 22.53,
+      "learning_rate": 9.249711085429816e-05,
+      "loss": 0.0821,
+      "step": 84500
+    },
+    {
+      "epoch": 22.53,
+      "eval_loss": 0.7559340596199036,
+      "eval_runtime": 18.7208,
+      "eval_samples_per_second": 4.754,
+      "eval_wer": 0.3704206241519674,
+      "step": 84500
+    },
+    {
+      "epoch": 22.56,
+      "learning_rate": 9.248822117521558e-05,
+      "loss": 0.0855,
+      "step": 84600
+    },
+    {
+      "epoch": 22.56,
+      "eval_loss": 0.7435261607170105,
+      "eval_runtime": 18.8028,
+      "eval_samples_per_second": 4.733,
+      "eval_wer": 0.3894165535956581,
+      "step": 84600
+    },
+    {
+      "epoch": 22.59,
+      "learning_rate": 9.247933149613299e-05,
+      "loss": 0.0842,
+      "step": 84700
+    },
+    {
+      "epoch": 22.59,
+      "eval_loss": 0.7084380388259888,
+      "eval_runtime": 18.7519,
+      "eval_samples_per_second": 4.746,
+      "eval_wer": 0.3853459972862958,
+      "step": 84700
+    },
+    {
+      "epoch": 22.61,
+      "learning_rate": 9.247044181705041e-05,
+      "loss": 0.0777,
+      "step": 84800
+    },
+    {
+      "epoch": 22.61,
+      "eval_loss": 0.7924440503120422,
+      "eval_runtime": 18.707,
+      "eval_samples_per_second": 4.758,
+      "eval_wer": 0.3853459972862958,
+      "step": 84800
+    },
+    {
+      "epoch": 22.64,
+      "learning_rate": 9.246155213796783e-05,
+      "loss": 0.0822,
+      "step": 84900
+    },
+    {
+      "epoch": 22.64,
+      "eval_loss": 0.6965318322181702,
+      "eval_runtime": 18.6269,
+      "eval_samples_per_second": 4.778,
+      "eval_wer": 0.3704206241519674,
+      "step": 84900
+    },
+    {
+      "epoch": 22.67,
+      "learning_rate": 9.245266245888525e-05,
+      "loss": 0.0892,
+      "step": 85000
+    },
+    {
+      "epoch": 22.67,
+      "eval_loss": 0.7296701669692993,
+      "eval_runtime": 18.7881,
+      "eval_samples_per_second": 4.737,
+      "eval_wer": 0.383989145183175,
+      "step": 85000
+    },
+    {
+      "epoch": 22.69,
+      "learning_rate": 9.244377277980266e-05,
+      "loss": 0.0761,
+      "step": 85100
+    },
+    {
+      "epoch": 22.69,
+      "eval_loss": 0.7442314028739929,
+      "eval_runtime": 18.7063,
+      "eval_samples_per_second": 4.758,
+      "eval_wer": 0.37991858887381275,
+      "step": 85100
+    },
+    {
+      "epoch": 22.72,
+      "learning_rate": 9.243488310072008e-05,
+      "loss": 0.0817,
+      "step": 85200
+    },
+    {
+      "epoch": 22.72,
+      "eval_loss": 0.7786024212837219,
+      "eval_runtime": 18.6943,
+      "eval_samples_per_second": 4.761,
+      "eval_wer": 0.39077340569877883,
+      "step": 85200
+    },
+    {
+      "epoch": 22.75,
+      "learning_rate": 9.242599342163748e-05,
+      "loss": 0.0905,
+      "step": 85300
+    },
+    {
+      "epoch": 22.75,
+      "eval_loss": 0.7861672043800354,
+      "eval_runtime": 18.7555,
+      "eval_samples_per_second": 4.745,
+      "eval_wer": 0.39348710990502034,
+      "step": 85300
+    },
+    {
+      "epoch": 22.77,
+      "learning_rate": 9.24171037425549e-05,
+      "loss": 0.0866,
+      "step": 85400
+    },
+    {
+      "epoch": 22.77,
+      "eval_loss": 0.6661286354064941,
+      "eval_runtime": 18.6383,
+      "eval_samples_per_second": 4.775,
+      "eval_wer": 0.38263229308005425,
+      "step": 85400
+    },
+    {
+      "epoch": 22.8,
+      "learning_rate": 9.240821406347231e-05,
+      "loss": 0.0895,
+      "step": 85500
+    },
+    {
+      "epoch": 22.8,
+      "eval_loss": 0.6603561639785767,
+      "eval_runtime": 18.7588,
+      "eval_samples_per_second": 4.744,
+      "eval_wer": 0.37720488466757124,
+      "step": 85500
+    },
+    {
+      "epoch": 22.83,
+      "learning_rate": 9.239932438438973e-05,
+      "loss": 0.0817,
+      "step": 85600
+    },
+    {
+      "epoch": 22.83,
+      "eval_loss": 0.6738704442977905,
+      "eval_runtime": 18.5978,
+      "eval_samples_per_second": 4.786,
+      "eval_wer": 0.3894165535956581,
+      "step": 85600
+    },
+    {
+      "epoch": 22.85,
+      "learning_rate": 9.239043470530714e-05,
+      "loss": 0.087,
+      "step": 85700
+    },
+    {
+      "epoch": 22.85,
+      "eval_loss": 0.6745472550392151,
+      "eval_runtime": 18.6286,
+      "eval_samples_per_second": 4.778,
+      "eval_wer": 0.38263229308005425,
+      "step": 85700
+    },
+    {
+      "epoch": 22.88,
+      "learning_rate": 9.238154502622456e-05,
+      "loss": 0.0837,
+      "step": 85800
+    },
+    {
+      "epoch": 22.88,
+      "eval_loss": 0.6245145797729492,
+      "eval_runtime": 18.6219,
+      "eval_samples_per_second": 4.779,
+      "eval_wer": 0.383989145183175,
+      "step": 85800
+    },
+    {
+      "epoch": 22.91,
+      "learning_rate": 9.237265534714197e-05,
+      "loss": 0.0873,
+      "step": 85900
+    },
+    {
+      "epoch": 22.91,
+      "eval_loss": 0.677052915096283,
+      "eval_runtime": 18.741,
+      "eval_samples_per_second": 4.749,
+      "eval_wer": 0.3704206241519674,
+      "step": 85900
+    },
+    {
+      "epoch": 22.93,
+      "learning_rate": 9.236376566805939e-05,
+      "loss": 0.0706,
+      "step": 86000
+    },
+    {
+      "epoch": 22.93,
+      "eval_loss": 0.6459066271781921,
+      "eval_runtime": 18.5637,
+      "eval_samples_per_second": 4.794,
+      "eval_wer": 0.3677069199457259,
+      "step": 86000
+    },
+    {
+      "epoch": 22.96,
+      "learning_rate": 9.235487598897681e-05,
+      "loss": 0.0867,
+      "step": 86100
+    },
+    {
+      "epoch": 22.96,
+      "eval_loss": 0.7289776802062988,
+      "eval_runtime": 18.569,
+      "eval_samples_per_second": 4.793,
+      "eval_wer": 0.38263229308005425,
+      "step": 86100
+    },
+    {
+      "epoch": 22.99,
+      "learning_rate": 9.234598630989422e-05,
+      "loss": 0.0849,
+      "step": 86200
+    },
+    {
+      "epoch": 22.99,
+      "eval_loss": 0.6735973954200745,
+      "eval_runtime": 18.6893,
+      "eval_samples_per_second": 4.762,
+      "eval_wer": 0.3921302578018996,
+      "step": 86200
+    },
+    {
+      "epoch": 23.01,
+      "learning_rate": 9.233709663081164e-05,
+      "loss": 0.0761,
+      "step": 86300
+    },
+    {
+      "epoch": 23.01,
+      "eval_loss": 0.7038031220436096,
+      "eval_runtime": 18.6841,
+      "eval_samples_per_second": 4.763,
+      "eval_wer": 0.3677069199457259,
+      "step": 86300
+    },
+    {
+      "epoch": 23.04,
+      "learning_rate": 9.232820695172904e-05,
+      "loss": 0.0751,
+      "step": 86400
+    },
+    {
+      "epoch": 23.04,
+      "eval_loss": 0.7442944049835205,
+      "eval_runtime": 18.6111,
+      "eval_samples_per_second": 4.782,
+      "eval_wer": 0.3758480325644505,
+      "step": 86400
+    },
+    {
+      "epoch": 23.07,
+      "learning_rate": 9.231931727264646e-05,
+      "loss": 0.0841,
+      "step": 86500
+    },
+    {
+      "epoch": 23.07,
+      "eval_loss": 0.7802165746688843,
+      "eval_runtime": 18.7189,
+      "eval_samples_per_second": 4.755,
+      "eval_wer": 0.3758480325644505,
+      "step": 86500
+    },
+    {
+      "epoch": 23.09,
+      "learning_rate": 9.231042759356387e-05,
+      "loss": 0.074,
+      "step": 86600
+    },
+    {
+      "epoch": 23.09,
+      "eval_loss": 0.709088146686554,
+      "eval_runtime": 18.5288,
+      "eval_samples_per_second": 4.803,
+      "eval_wer": 0.37449118046132973,
+      "step": 86600
+    },
+    {
+      "epoch": 23.12,
+      "learning_rate": 9.230153791448129e-05,
+      "loss": 0.0818,
+      "step": 86700
+    },
+    {
+      "epoch": 23.12,
+      "eval_loss": 0.7752262949943542,
+      "eval_runtime": 18.5169,
+      "eval_samples_per_second": 4.806,
+      "eval_wer": 0.3758480325644505,
+      "step": 86700
+    },
+    {
+      "epoch": 23.15,
+      "learning_rate": 9.22926482353987e-05,
+      "loss": 0.0788,
+      "step": 86800
+    },
+    {
+      "epoch": 23.15,
+      "eval_loss": 0.7730712294578552,
+      "eval_runtime": 18.5811,
+      "eval_samples_per_second": 4.79,
+      "eval_wer": 0.37720488466757124,
+      "step": 86800
+    },
+    {
+      "epoch": 23.17,
+      "learning_rate": 9.228375855631612e-05,
+      "loss": 0.0746,
+      "step": 86900
+    },
+    {
+      "epoch": 23.17,
+      "eval_loss": 0.7588068246841431,
+      "eval_runtime": 18.7386,
+      "eval_samples_per_second": 4.75,
+      "eval_wer": 0.378561736770692,
+      "step": 86900
+    },
+    {
+      "epoch": 23.2,
+      "learning_rate": 9.227486887723354e-05,
+      "loss": 0.0809,
+      "step": 87000
+    },
+    {
+      "epoch": 23.2,
+      "eval_loss": 0.6910467147827148,
+      "eval_runtime": 18.6208,
+      "eval_samples_per_second": 4.78,
+      "eval_wer": 0.3812754409769335,
+      "step": 87000
+    },
+    {
+      "epoch": 23.23,
+      "learning_rate": 9.226597919815095e-05,
+      "loss": 0.0701,
+      "step": 87100
+    },
+    {
+      "epoch": 23.23,
+      "eval_loss": 0.7382462620735168,
+      "eval_runtime": 18.6734,
+      "eval_samples_per_second": 4.766,
+      "eval_wer": 0.38263229308005425,
+      "step": 87100
+    },
+    {
+      "epoch": 23.25,
+      "learning_rate": 9.225708951906837e-05,
+      "loss": 0.082,
+      "step": 87200
+    },
+    {
+      "epoch": 23.25,
+      "eval_loss": 0.6743248701095581,
+      "eval_runtime": 19.3858,
+      "eval_samples_per_second": 4.591,
+      "eval_wer": 0.37991858887381275,
+      "step": 87200
+    },
+    {
+      "epoch": 23.28,
+      "learning_rate": 9.224819983998579e-05,
+      "loss": 0.082,
+      "step": 87300
+    },
+    {
+      "epoch": 23.28,
+      "eval_loss": 0.712517499923706,
+      "eval_runtime": 19.2827,
+      "eval_samples_per_second": 4.616,
+      "eval_wer": 0.36363636363636365,
+      "step": 87300
+    }
+  ],
+  "max_steps": 1125000,
+  "num_train_epochs": 300,
+  "total_flos": 1.2027419823711332e+20,
+  "trial_name": null,
+  "trial_params": null
+}