diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,7100 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 14.059452701764133,
+  "global_step": 50600,
+  "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": 7.817,
+      "step": 100
+    },
+    {
+      "epoch": 0.03,
+      "eval_loss": 3.791971206665039,
+      "eval_runtime": 17.0609,
+      "eval_samples_per_second": 5.803,
+      "eval_wer": 1.0,
+      "step": 100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.999444259197512e-05,
+      "loss": 3.0102,
+      "step": 200
+    },
+    {
+      "epoch": 0.06,
+      "eval_loss": 3.3802108764648438,
+      "eval_runtime": 17.1385,
+      "eval_samples_per_second": 5.776,
+      "eval_wer": 1.0,
+      "step": 200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.99888851839502e-05,
+      "loss": 2.929,
+      "step": 300
+    },
+    {
+      "epoch": 0.08,
+      "eval_loss": 3.4951775074005127,
+      "eval_runtime": 17.7008,
+      "eval_samples_per_second": 5.593,
+      "eval_wer": 1.0,
+      "step": 300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.998332777592532e-05,
+      "loss": 2.918,
+      "step": 400
+    },
+    {
+      "epoch": 0.11,
+      "eval_loss": 3.2080225944519043,
+      "eval_runtime": 17.3326,
+      "eval_samples_per_second": 5.712,
+      "eval_wer": 1.0,
+      "step": 400
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.997777036790042e-05,
+      "loss": 2.8855,
+      "step": 500
+    },
+    {
+      "epoch": 0.14,
+      "eval_loss": 3.1997480392456055,
+      "eval_runtime": 17.2803,
+      "eval_samples_per_second": 5.729,
+      "eval_wer": 1.0,
+      "step": 500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.997221295987552e-05,
+      "loss": 2.8791,
+      "step": 600
+    },
+    {
+      "epoch": 0.17,
+      "eval_loss": 3.0948541164398193,
+      "eval_runtime": 17.1794,
+      "eval_samples_per_second": 5.763,
+      "eval_wer": 1.0,
+      "step": 600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.996665555185062e-05,
+      "loss": 2.8751,
+      "step": 700
+    },
+    {
+      "epoch": 0.19,
+      "eval_loss": 3.1061549186706543,
+      "eval_runtime": 17.4783,
+      "eval_samples_per_second": 5.664,
+      "eval_wer": 1.0,
+      "step": 700
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.996109814382572e-05,
+      "loss": 2.8643,
+      "step": 800
+    },
+    {
+      "epoch": 0.22,
+      "eval_loss": 3.0063648223876953,
+      "eval_runtime": 17.252,
+      "eval_samples_per_second": 5.738,
+      "eval_wer": 1.0,
+      "step": 800
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.995554073580082e-05,
+      "loss": 2.8587,
+      "step": 900
+    },
+    {
+      "epoch": 0.25,
+      "eval_loss": 2.9906227588653564,
+      "eval_runtime": 17.2628,
+      "eval_samples_per_second": 5.735,
+      "eval_wer": 1.0,
+      "step": 900
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.994998332777594e-05,
+      "loss": 2.7566,
+      "step": 1000
+    },
+    {
+      "epoch": 0.28,
+      "eval_loss": 2.6209235191345215,
+      "eval_runtime": 17.3238,
+      "eval_samples_per_second": 5.715,
+      "eval_wer": 1.0,
+      "step": 1000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.994442591975102e-05,
+      "loss": 2.1468,
+      "step": 1100
+    },
+    {
+      "epoch": 0.31,
+      "eval_loss": 1.8421308994293213,
+      "eval_runtime": 17.1871,
+      "eval_samples_per_second": 5.76,
+      "eval_wer": 0.9896907216494846,
+      "step": 1100
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.993886851172614e-05,
+      "loss": 1.4695,
+      "step": 1200
+    },
+    {
+      "epoch": 0.33,
+      "eval_loss": 1.4311132431030273,
+      "eval_runtime": 17.4523,
+      "eval_samples_per_second": 5.673,
+      "eval_wer": 0.875,
+      "step": 1200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.993331110370124e-05,
+      "loss": 1.1334,
+      "step": 1300
+    },
+    {
+      "epoch": 0.36,
+      "eval_loss": 1.2780916690826416,
+      "eval_runtime": 17.1785,
+      "eval_samples_per_second": 5.763,
+      "eval_wer": 0.854381443298969,
+      "step": 1300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.992775369567634e-05,
+      "loss": 0.9951,
+      "step": 1400
+    },
+    {
+      "epoch": 0.39,
+      "eval_loss": 1.0466135740280151,
+      "eval_runtime": 16.9445,
+      "eval_samples_per_second": 5.843,
+      "eval_wer": 0.7306701030927835,
+      "step": 1400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.992219628765145e-05,
+      "loss": 0.9161,
+      "step": 1500
+    },
+    {
+      "epoch": 0.42,
+      "eval_loss": 1.0910567045211792,
+      "eval_runtime": 17.1529,
+      "eval_samples_per_second": 5.772,
+      "eval_wer": 0.7345360824742269,
+      "step": 1500
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.991663887962654e-05,
+      "loss": 0.8162,
+      "step": 1600
+    },
+    {
+      "epoch": 0.44,
+      "eval_loss": 0.9578825831413269,
+      "eval_runtime": 17.0661,
+      "eval_samples_per_second": 5.801,
+      "eval_wer": 0.6842783505154639,
+      "step": 1600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.991108147160165e-05,
+      "loss": 0.7552,
+      "step": 1700
+    },
+    {
+      "epoch": 0.47,
+      "eval_loss": 0.9620718955993652,
+      "eval_runtime": 17.1705,
+      "eval_samples_per_second": 5.766,
+      "eval_wer": 0.6288659793814433,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.990552406357675e-05,
+      "loss": 0.752,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5,
+      "eval_loss": 0.9395142197608948,
+      "eval_runtime": 17.1773,
+      "eval_samples_per_second": 5.763,
+      "eval_wer": 0.6172680412371134,
+      "step": 1800
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.989996665555186e-05,
+      "loss": 0.6846,
+      "step": 1900
+    },
+    {
+      "epoch": 0.53,
+      "eval_loss": 1.0008608102798462,
+      "eval_runtime": 17.3332,
+      "eval_samples_per_second": 5.712,
+      "eval_wer": 0.5786082474226805,
+      "step": 1900
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.989440924752696e-05,
+      "loss": 0.6402,
+      "step": 2000
+    },
+    {
+      "epoch": 0.56,
+      "eval_loss": 0.8784067034721375,
+      "eval_runtime": 17.1604,
+      "eval_samples_per_second": 5.769,
+      "eval_wer": 0.5489690721649485,
+      "step": 2000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.988885183950206e-05,
+      "loss": 0.6582,
+      "step": 2100
+    },
+    {
+      "epoch": 0.58,
+      "eval_loss": 0.9032992720603943,
+      "eval_runtime": 17.1447,
+      "eval_samples_per_second": 5.774,
+      "eval_wer": 0.5631443298969072,
+      "step": 2100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.988329443147716e-05,
+      "loss": 0.5984,
+      "step": 2200
+    },
+    {
+      "epoch": 0.61,
+      "eval_loss": 0.8451138138771057,
+      "eval_runtime": 17.25,
+      "eval_samples_per_second": 5.739,
+      "eval_wer": 0.5283505154639175,
+      "step": 2200
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.987773702345227e-05,
+      "loss": 0.6265,
+      "step": 2300
+    },
+    {
+      "epoch": 0.64,
+      "eval_loss": 0.7656041383743286,
+      "eval_runtime": 17.2036,
+      "eval_samples_per_second": 5.755,
+      "eval_wer": 0.5373711340206185,
+      "step": 2300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.987217961542737e-05,
+      "loss": 0.5903,
+      "step": 2400
+    },
+    {
+      "epoch": 0.67,
+      "eval_loss": 0.8390687108039856,
+      "eval_runtime": 17.1101,
+      "eval_samples_per_second": 5.786,
+      "eval_wer": 0.5167525773195877,
+      "step": 2400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.986662220740247e-05,
+      "loss": 0.5788,
+      "step": 2500
+    },
+    {
+      "epoch": 0.69,
+      "eval_loss": 0.8256884217262268,
+      "eval_runtime": 17.1636,
+      "eval_samples_per_second": 5.768,
+      "eval_wer": 0.5064432989690721,
+      "step": 2500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.986106479937759e-05,
+      "loss": 0.583,
+      "step": 2600
+    },
+    {
+      "epoch": 0.72,
+      "eval_loss": 0.7550874352455139,
+      "eval_runtime": 17.1812,
+      "eval_samples_per_second": 5.762,
+      "eval_wer": 0.5335051546391752,
+      "step": 2600
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.985550739135267e-05,
+      "loss": 0.5805,
+      "step": 2700
+    },
+    {
+      "epoch": 0.75,
+      "eval_loss": 0.8099715709686279,
+      "eval_runtime": 17.1988,
+      "eval_samples_per_second": 5.756,
+      "eval_wer": 0.5193298969072165,
+      "step": 2700
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.984994998332779e-05,
+      "loss": 0.5465,
+      "step": 2800
+    },
+    {
+      "epoch": 0.78,
+      "eval_loss": 0.7341137528419495,
+      "eval_runtime": 17.1767,
+      "eval_samples_per_second": 5.764,
+      "eval_wer": 0.4884020618556701,
+      "step": 2800
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.984439257530287e-05,
+      "loss": 0.556,
+      "step": 2900
+    },
+    {
+      "epoch": 0.81,
+      "eval_loss": 0.845912516117096,
+      "eval_runtime": 17.2763,
+      "eval_samples_per_second": 5.73,
+      "eval_wer": 0.5025773195876289,
+      "step": 2900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.983883516727799e-05,
+      "loss": 0.5141,
+      "step": 3000
+    },
+    {
+      "epoch": 0.83,
+      "eval_loss": 0.822607696056366,
+      "eval_runtime": 17.5294,
+      "eval_samples_per_second": 5.648,
+      "eval_wer": 0.5038659793814433,
+      "step": 3000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.983327775925309e-05,
+      "loss": 0.5257,
+      "step": 3100
+    },
+    {
+      "epoch": 0.86,
+      "eval_loss": 0.8175485134124756,
+      "eval_runtime": 17.1702,
+      "eval_samples_per_second": 5.766,
+      "eval_wer": 0.4948453608247423,
+      "step": 3100
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.982772035122819e-05,
+      "loss": 0.501,
+      "step": 3200
+    },
+    {
+      "epoch": 0.89,
+      "eval_loss": 0.7639488577842712,
+      "eval_runtime": 17.3806,
+      "eval_samples_per_second": 5.696,
+      "eval_wer": 0.4806701030927835,
+      "step": 3200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.982216294320329e-05,
+      "loss": 0.5081,
+      "step": 3300
+    },
+    {
+      "epoch": 0.92,
+      "eval_loss": 0.7297512888908386,
+      "eval_runtime": 17.1321,
+      "eval_samples_per_second": 5.779,
+      "eval_wer": 0.48711340206185566,
+      "step": 3300
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.98166055351784e-05,
+      "loss": 0.4741,
+      "step": 3400
+    },
+    {
+      "epoch": 0.94,
+      "eval_loss": 0.9208475947380066,
+      "eval_runtime": 17.2201,
+      "eval_samples_per_second": 5.749,
+      "eval_wer": 0.49355670103092786,
+      "step": 3400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.98110481271535e-05,
+      "loss": 0.4848,
+      "step": 3500
+    },
+    {
+      "epoch": 0.97,
+      "eval_loss": 0.8245986104011536,
+      "eval_runtime": 17.1126,
+      "eval_samples_per_second": 5.785,
+      "eval_wer": 0.48582474226804123,
+      "step": 3500
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 9.98054907191286e-05,
+      "loss": 0.4794,
+      "step": 3600
+    },
+    {
+      "epoch": 1.0,
+      "eval_loss": 0.7197224497795105,
+      "eval_runtime": 17.3562,
+      "eval_samples_per_second": 5.704,
+      "eval_wer": 0.4677835051546392,
+      "step": 3600
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.97999333111037e-05,
+      "loss": 0.4189,
+      "step": 3700
+    },
+    {
+      "epoch": 1.03,
+      "eval_loss": 0.7290377616882324,
+      "eval_runtime": 17.1757,
+      "eval_samples_per_second": 5.764,
+      "eval_wer": 0.48195876288659795,
+      "step": 3700
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.97943759030788e-05,
+      "loss": 0.4133,
+      "step": 3800
+    },
+    {
+      "epoch": 1.06,
+      "eval_loss": 0.8328794240951538,
+      "eval_runtime": 17.375,
+      "eval_samples_per_second": 5.698,
+      "eval_wer": 0.47164948453608246,
+      "step": 3800
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.978881849505392e-05,
+      "loss": 0.43,
+      "step": 3900
+    },
+    {
+      "epoch": 1.08,
+      "eval_loss": 0.7619199156761169,
+      "eval_runtime": 17.4527,
+      "eval_samples_per_second": 5.672,
+      "eval_wer": 0.46649484536082475,
+      "step": 3900
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.978326108702901e-05,
+      "loss": 0.4274,
+      "step": 4000
+    },
+    {
+      "epoch": 1.11,
+      "eval_loss": 0.8476623892784119,
+      "eval_runtime": 17.2396,
+      "eval_samples_per_second": 5.743,
+      "eval_wer": 0.5103092783505154,
+      "step": 4000
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 9.977770367900412e-05,
+      "loss": 0.4586,
+      "step": 4100
+    },
+    {
+      "epoch": 1.14,
+      "eval_loss": 0.8450866937637329,
+      "eval_runtime": 17.5121,
+      "eval_samples_per_second": 5.653,
+      "eval_wer": 0.48711340206185566,
+      "step": 4100
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 9.977214627097922e-05,
+      "loss": 0.3902,
+      "step": 4200
+    },
+    {
+      "epoch": 1.17,
+      "eval_loss": 1.0022594928741455,
+      "eval_runtime": 17.3099,
+      "eval_samples_per_second": 5.719,
+      "eval_wer": 0.520618556701031,
+      "step": 4200
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 9.976658886295432e-05,
+      "loss": 0.3995,
+      "step": 4300
+    },
+    {
+      "epoch": 1.19,
+      "eval_loss": 0.825476884841919,
+      "eval_runtime": 17.3054,
+      "eval_samples_per_second": 5.721,
+      "eval_wer": 0.5038659793814433,
+      "step": 4300
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 9.976103145492942e-05,
+      "loss": 0.3952,
+      "step": 4400
+    },
+    {
+      "epoch": 1.22,
+      "eval_loss": 0.7960893511772156,
+      "eval_runtime": 17.2244,
+      "eval_samples_per_second": 5.748,
+      "eval_wer": 0.48711340206185566,
+      "step": 4400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 9.975547404690452e-05,
+      "loss": 0.4177,
+      "step": 4500
+    },
+    {
+      "epoch": 1.25,
+      "eval_loss": 0.7501189112663269,
+      "eval_runtime": 17.0905,
+      "eval_samples_per_second": 5.793,
+      "eval_wer": 0.47164948453608246,
+      "step": 4500
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 9.974991663887962e-05,
+      "loss": 0.4054,
+      "step": 4600
+    },
+    {
+      "epoch": 1.28,
+      "eval_loss": 0.7024160027503967,
+      "eval_runtime": 17.2738,
+      "eval_samples_per_second": 5.731,
+      "eval_wer": 0.4884020618556701,
+      "step": 4600
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 9.974435923085474e-05,
+      "loss": 0.4029,
+      "step": 4700
+    },
+    {
+      "epoch": 1.31,
+      "eval_loss": 0.7287462949752808,
+      "eval_runtime": 17.6417,
+      "eval_samples_per_second": 5.612,
+      "eval_wer": 0.45103092783505155,
+      "step": 4700
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 9.973880182282984e-05,
+      "loss": 0.3967,
+      "step": 4800
+    },
+    {
+      "epoch": 1.33,
+      "eval_loss": 0.7488074898719788,
+      "eval_runtime": 17.3755,
+      "eval_samples_per_second": 5.698,
+      "eval_wer": 0.49097938144329895,
+      "step": 4800
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 9.973324441480494e-05,
+      "loss": 0.4053,
+      "step": 4900
+    },
+    {
+      "epoch": 1.36,
+      "eval_loss": 0.7065747976303101,
+      "eval_runtime": 17.3999,
+      "eval_samples_per_second": 5.69,
+      "eval_wer": 0.4652061855670103,
+      "step": 4900
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 9.972768700678004e-05,
+      "loss": 0.4165,
+      "step": 5000
+    },
+    {
+      "epoch": 1.39,
+      "eval_loss": 0.7919918894767761,
+      "eval_runtime": 17.0704,
+      "eval_samples_per_second": 5.799,
+      "eval_wer": 0.46649484536082475,
+      "step": 5000
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 9.972212959875514e-05,
+      "loss": 0.4001,
+      "step": 5100
+    },
+    {
+      "epoch": 1.42,
+      "eval_loss": 0.758572518825531,
+      "eval_runtime": 17.1616,
+      "eval_samples_per_second": 5.769,
+      "eval_wer": 0.47036082474226804,
+      "step": 5100
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 9.971657219073025e-05,
+      "loss": 0.3736,
+      "step": 5200
+    },
+    {
+      "epoch": 1.44,
+      "eval_loss": 0.8151732683181763,
+      "eval_runtime": 17.3692,
+      "eval_samples_per_second": 5.7,
+      "eval_wer": 0.5025773195876289,
+      "step": 5200
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 9.971101478270534e-05,
+      "loss": 0.3694,
+      "step": 5300
+    },
+    {
+      "epoch": 1.47,
+      "eval_loss": 0.823724627494812,
+      "eval_runtime": 17.36,
+      "eval_samples_per_second": 5.703,
+      "eval_wer": 0.5025773195876289,
+      "step": 5300
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.970545737468046e-05,
+      "loss": 0.3748,
+      "step": 5400
+    },
+    {
+      "epoch": 1.5,
+      "eval_loss": 0.7445100545883179,
+      "eval_runtime": 17.4369,
+      "eval_samples_per_second": 5.678,
+      "eval_wer": 0.44458762886597936,
+      "step": 5400
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.969989996665556e-05,
+      "loss": 0.3645,
+      "step": 5500
+    },
+    {
+      "epoch": 1.53,
+      "eval_loss": 0.7235777378082275,
+      "eval_runtime": 17.1581,
+      "eval_samples_per_second": 5.77,
+      "eval_wer": 0.47680412371134023,
+      "step": 5500
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.969434255863066e-05,
+      "loss": 0.3905,
+      "step": 5600
+    },
+    {
+      "epoch": 1.56,
+      "eval_loss": 0.739282488822937,
+      "eval_runtime": 17.2125,
+      "eval_samples_per_second": 5.752,
+      "eval_wer": 0.4793814432989691,
+      "step": 5600
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.968878515060576e-05,
+      "loss": 0.3723,
+      "step": 5700
+    },
+    {
+      "epoch": 1.58,
+      "eval_loss": 0.762248158454895,
+      "eval_runtime": 17.4572,
+      "eval_samples_per_second": 5.671,
+      "eval_wer": 0.4652061855670103,
+      "step": 5700
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.968322774258087e-05,
+      "loss": 0.3611,
+      "step": 5800
+    },
+    {
+      "epoch": 1.61,
+      "eval_loss": 0.7599887251853943,
+      "eval_runtime": 17.487,
+      "eval_samples_per_second": 5.661,
+      "eval_wer": 0.46649484536082475,
+      "step": 5800
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.967767033455597e-05,
+      "loss": 0.3747,
+      "step": 5900
+    },
+    {
+      "epoch": 1.64,
+      "eval_loss": 0.7197273373603821,
+      "eval_runtime": 17.4419,
+      "eval_samples_per_second": 5.676,
+      "eval_wer": 0.47680412371134023,
+      "step": 5900
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 9.967211292653107e-05,
+      "loss": 0.3736,
+      "step": 6000
+    },
+    {
+      "epoch": 1.67,
+      "eval_loss": 0.7508798241615295,
+      "eval_runtime": 17.1575,
+      "eval_samples_per_second": 5.77,
+      "eval_wer": 0.47036082474226804,
+      "step": 6000
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 9.966655551850617e-05,
+      "loss": 0.3828,
+      "step": 6100
+    },
+    {
+      "epoch": 1.69,
+      "eval_loss": 0.7594204545021057,
+      "eval_runtime": 17.2159,
+      "eval_samples_per_second": 5.75,
+      "eval_wer": 0.4652061855670103,
+      "step": 6100
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 9.966099811048127e-05,
+      "loss": 0.3717,
+      "step": 6200
+    },
+    {
+      "epoch": 1.72,
+      "eval_loss": 0.6831815838813782,
+      "eval_runtime": 17.2288,
+      "eval_samples_per_second": 5.746,
+      "eval_wer": 0.4639175257731959,
+      "step": 6200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 9.965544070245639e-05,
+      "loss": 0.3605,
+      "step": 6300
+    },
+    {
+      "epoch": 1.75,
+      "eval_loss": 0.8021068572998047,
+      "eval_runtime": 17.2637,
+      "eval_samples_per_second": 5.735,
+      "eval_wer": 0.46262886597938147,
+      "step": 6300
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 9.964988329443147e-05,
+      "loss": 0.3868,
+      "step": 6400
+    },
+    {
+      "epoch": 1.78,
+      "eval_loss": 0.7559285759925842,
+      "eval_runtime": 17.2295,
+      "eval_samples_per_second": 5.746,
+      "eval_wer": 0.46262886597938147,
+      "step": 6400
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 9.964432588640659e-05,
+      "loss": 0.3762,
+      "step": 6500
+    },
+    {
+      "epoch": 1.81,
+      "eval_loss": 0.6938344240188599,
+      "eval_runtime": 17.2273,
+      "eval_samples_per_second": 5.747,
+      "eval_wer": 0.44587628865979384,
+      "step": 6500
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 9.963876847838169e-05,
+      "loss": 0.3335,
+      "step": 6600
+    },
+    {
+      "epoch": 1.83,
+      "eval_loss": 0.7515353560447693,
+      "eval_runtime": 17.2748,
+      "eval_samples_per_second": 5.731,
+      "eval_wer": 0.4652061855670103,
+      "step": 6600
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 9.963321107035679e-05,
+      "loss": 0.3761,
+      "step": 6700
+    },
+    {
+      "epoch": 1.86,
+      "eval_loss": 0.6331177353858948,
+      "eval_runtime": 17.2836,
+      "eval_samples_per_second": 5.728,
+      "eval_wer": 0.43556701030927836,
+      "step": 6700
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 9.962765366233189e-05,
+      "loss": 0.3597,
+      "step": 6800
+    },
+    {
+      "epoch": 1.89,
+      "eval_loss": 0.7062612771987915,
+      "eval_runtime": 17.0606,
+      "eval_samples_per_second": 5.803,
+      "eval_wer": 0.46262886597938147,
+      "step": 6800
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 9.962209625430699e-05,
+      "loss": 0.3489,
+      "step": 6900
+    },
+    {
+      "epoch": 1.92,
+      "eval_loss": 0.7171001434326172,
+      "eval_runtime": 17.3915,
+      "eval_samples_per_second": 5.692,
+      "eval_wer": 0.4587628865979381,
+      "step": 6900
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 9.96165388462821e-05,
+      "loss": 0.3492,
+      "step": 7000
+    },
+    {
+      "epoch": 1.94,
+      "eval_loss": 0.8369248509407043,
+      "eval_runtime": 17.2963,
+      "eval_samples_per_second": 5.724,
+      "eval_wer": 0.4845360824742268,
+      "step": 7000
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 9.96109814382572e-05,
+      "loss": 0.3581,
+      "step": 7100
+    },
+    {
+      "epoch": 1.97,
+      "eval_loss": 0.7614104747772217,
+      "eval_runtime": 17.1514,
+      "eval_samples_per_second": 5.772,
+      "eval_wer": 0.4484536082474227,
+      "step": 7100
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 9.96054240302323e-05,
+      "loss": 0.3617,
+      "step": 7200
+    },
+    {
+      "epoch": 2.0,
+      "eval_loss": 0.7410733699798584,
+      "eval_runtime": 16.8873,
+      "eval_samples_per_second": 5.862,
+      "eval_wer": 0.45103092783505155,
+      "step": 7200
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 9.95998666222074e-05,
+      "loss": 0.3044,
+      "step": 7300
+    },
+    {
+      "epoch": 2.03,
+      "eval_loss": 0.8537248373031616,
+      "eval_runtime": 16.8753,
+      "eval_samples_per_second": 5.867,
+      "eval_wer": 0.4793814432989691,
+      "step": 7300
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 9.959430921418251e-05,
+      "loss": 0.2982,
+      "step": 7400
+    },
+    {
+      "epoch": 2.06,
+      "eval_loss": 0.7687365412712097,
+      "eval_runtime": 17.2864,
+      "eval_samples_per_second": 5.727,
+      "eval_wer": 0.48582474226804123,
+      "step": 7400
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 9.958875180615761e-05,
+      "loss": 0.3035,
+      "step": 7500
+    },
+    {
+      "epoch": 2.08,
+      "eval_loss": 0.8554800152778625,
+      "eval_runtime": 17.1923,
+      "eval_samples_per_second": 5.758,
+      "eval_wer": 0.4536082474226804,
+      "step": 7500
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 9.958319439813272e-05,
+      "loss": 0.3032,
+      "step": 7600
+    },
+    {
+      "epoch": 2.11,
+      "eval_loss": 0.8073357343673706,
+      "eval_runtime": 17.2402,
+      "eval_samples_per_second": 5.742,
+      "eval_wer": 0.47551546391752575,
+      "step": 7600
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 9.957763699010781e-05,
+      "loss": 0.3336,
+      "step": 7700
+    },
+    {
+      "epoch": 2.14,
+      "eval_loss": 0.8232630491256714,
+      "eval_runtime": 17.1702,
+      "eval_samples_per_second": 5.766,
+      "eval_wer": 0.4574742268041237,
+      "step": 7700
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 9.957207958208292e-05,
+      "loss": 0.3062,
+      "step": 7800
+    },
+    {
+      "epoch": 2.17,
+      "eval_loss": 0.73056960105896,
+      "eval_runtime": 17.1533,
+      "eval_samples_per_second": 5.771,
+      "eval_wer": 0.45489690721649484,
+      "step": 7800
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 9.956652217405802e-05,
+      "loss": 0.3278,
+      "step": 7900
+    },
+    {
+      "epoch": 2.2,
+      "eval_loss": 0.7208071351051331,
+      "eval_runtime": 17.2805,
+      "eval_samples_per_second": 5.729,
+      "eval_wer": 0.4742268041237113,
+      "step": 7900
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 9.956096476603312e-05,
+      "loss": 0.3116,
+      "step": 8000
+    },
+    {
+      "epoch": 2.22,
+      "eval_loss": 0.7529891133308411,
+      "eval_runtime": 17.1791,
+      "eval_samples_per_second": 5.763,
+      "eval_wer": 0.4574742268041237,
+      "step": 8000
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 9.955540735800824e-05,
+      "loss": 0.2909,
+      "step": 8100
+    },
+    {
+      "epoch": 2.25,
+      "eval_loss": 0.7208893299102783,
+      "eval_runtime": 17.1849,
+      "eval_samples_per_second": 5.761,
+      "eval_wer": 0.4536082474226804,
+      "step": 8100
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 9.954984994998332e-05,
+      "loss": 0.2998,
+      "step": 8200
+    },
+    {
+      "epoch": 2.28,
+      "eval_loss": 0.7474488615989685,
+      "eval_runtime": 17.1962,
+      "eval_samples_per_second": 5.757,
+      "eval_wer": 0.44072164948453607,
+      "step": 8200
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 9.954429254195844e-05,
+      "loss": 0.3007,
+      "step": 8300
+    },
+    {
+      "epoch": 2.31,
+      "eval_loss": 0.8237383365631104,
+      "eval_runtime": 17.1897,
+      "eval_samples_per_second": 5.759,
+      "eval_wer": 0.47164948453608246,
+      "step": 8300
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 9.953873513393354e-05,
+      "loss": 0.3047,
+      "step": 8400
+    },
+    {
+      "epoch": 2.33,
+      "eval_loss": 0.7258074879646301,
+      "eval_runtime": 17.2814,
+      "eval_samples_per_second": 5.729,
+      "eval_wer": 0.46005154639175255,
+      "step": 8400
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 9.953317772590864e-05,
+      "loss": 0.3224,
+      "step": 8500
+    },
+    {
+      "epoch": 2.36,
+      "eval_loss": 0.9189648032188416,
+      "eval_runtime": 17.2837,
+      "eval_samples_per_second": 5.728,
+      "eval_wer": 0.4652061855670103,
+      "step": 8500
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 9.952762031788374e-05,
+      "loss": 0.3126,
+      "step": 8600
+    },
+    {
+      "epoch": 2.39,
+      "eval_loss": 0.7851585745811462,
+      "eval_runtime": 17.2252,
+      "eval_samples_per_second": 5.747,
+      "eval_wer": 0.4574742268041237,
+      "step": 8600
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 9.952206290985885e-05,
+      "loss": 0.2806,
+      "step": 8700
+    },
+    {
+      "epoch": 2.42,
+      "eval_loss": 0.9620941877365112,
+      "eval_runtime": 17.2564,
+      "eval_samples_per_second": 5.737,
+      "eval_wer": 0.46649484536082475,
+      "step": 8700
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 9.951650550183394e-05,
+      "loss": 0.3057,
+      "step": 8800
+    },
+    {
+      "epoch": 2.45,
+      "eval_loss": 0.6678251624107361,
+      "eval_runtime": 17.2795,
+      "eval_samples_per_second": 5.729,
+      "eval_wer": 0.422680412371134,
+      "step": 8800
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 9.951094809380906e-05,
+      "loss": 0.3066,
+      "step": 8900
+    },
+    {
+      "epoch": 2.47,
+      "eval_loss": 0.7364962100982666,
+      "eval_runtime": 17.3954,
+      "eval_samples_per_second": 5.691,
+      "eval_wer": 0.43427835051546393,
+      "step": 8900
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 9.950539068578416e-05,
+      "loss": 0.2891,
+      "step": 9000
+    },
+    {
+      "epoch": 2.5,
+      "eval_loss": 0.7602680921554565,
+      "eval_runtime": 17.2544,
+      "eval_samples_per_second": 5.738,
+      "eval_wer": 0.4574742268041237,
+      "step": 9000
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 9.949983327775926e-05,
+      "loss": 0.2857,
+      "step": 9100
+    },
+    {
+      "epoch": 2.53,
+      "eval_loss": 0.7446552515029907,
+      "eval_runtime": 17.5308,
+      "eval_samples_per_second": 5.647,
+      "eval_wer": 0.4497422680412371,
+      "step": 9100
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 9.949427586973437e-05,
+      "loss": 0.333,
+      "step": 9200
+    },
+    {
+      "epoch": 2.56,
+      "eval_loss": 0.7369379997253418,
+      "eval_runtime": 17.5237,
+      "eval_samples_per_second": 5.649,
+      "eval_wer": 0.4484536082474227,
+      "step": 9200
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 9.948871846170946e-05,
+      "loss": 0.3125,
+      "step": 9300
+    },
+    {
+      "epoch": 2.58,
+      "eval_loss": 0.7238438725471497,
+      "eval_runtime": 17.2568,
+      "eval_samples_per_second": 5.737,
+      "eval_wer": 0.452319587628866,
+      "step": 9300
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 9.948316105368457e-05,
+      "loss": 0.3182,
+      "step": 9400
+    },
+    {
+      "epoch": 2.61,
+      "eval_loss": 0.7532787919044495,
+      "eval_runtime": 17.267,
+      "eval_samples_per_second": 5.733,
+      "eval_wer": 0.4536082474226804,
+      "step": 9400
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 9.947760364565967e-05,
+      "loss": 0.2925,
+      "step": 9500
+    },
+    {
+      "epoch": 2.64,
+      "eval_loss": 0.7420103549957275,
+      "eval_runtime": 17.2284,
+      "eval_samples_per_second": 5.746,
+      "eval_wer": 0.45489690721649484,
+      "step": 9500
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 9.947204623763477e-05,
+      "loss": 0.3078,
+      "step": 9600
+    },
+    {
+      "epoch": 2.67,
+      "eval_loss": 0.7169169187545776,
+      "eval_runtime": 17.501,
+      "eval_samples_per_second": 5.657,
+      "eval_wer": 0.4536082474226804,
+      "step": 9600
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 9.946648882960987e-05,
+      "loss": 0.2922,
+      "step": 9700
+    },
+    {
+      "epoch": 2.7,
+      "eval_loss": 0.8229072093963623,
+      "eval_runtime": 17.3292,
+      "eval_samples_per_second": 5.713,
+      "eval_wer": 0.4832474226804124,
+      "step": 9700
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 9.946093142158497e-05,
+      "loss": 0.2934,
+      "step": 9800
+    },
+    {
+      "epoch": 2.72,
+      "eval_loss": 0.6904311776161194,
+      "eval_runtime": 17.3862,
+      "eval_samples_per_second": 5.694,
+      "eval_wer": 0.4536082474226804,
+      "step": 9800
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 9.945537401356007e-05,
+      "loss": 0.2655,
+      "step": 9900
+    },
+    {
+      "epoch": 2.75,
+      "eval_loss": 0.7936742901802063,
+      "eval_runtime": 17.2939,
+      "eval_samples_per_second": 5.725,
+      "eval_wer": 0.46134020618556704,
+      "step": 9900
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 9.944981660553519e-05,
+      "loss": 0.2764,
+      "step": 10000
+    },
+    {
+      "epoch": 2.78,
+      "eval_loss": 0.7943949103355408,
+      "eval_runtime": 17.0681,
+      "eval_samples_per_second": 5.8,
+      "eval_wer": 0.45103092783505155,
+      "step": 10000
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 9.944425919751028e-05,
+      "loss": 0.2991,
+      "step": 10100
+    },
+    {
+      "epoch": 2.81,
+      "eval_loss": 0.7811460494995117,
+      "eval_runtime": 17.292,
+      "eval_samples_per_second": 5.725,
+      "eval_wer": 0.46005154639175255,
+      "step": 10100
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 9.943870178948539e-05,
+      "loss": 0.2866,
+      "step": 10200
+    },
+    {
+      "epoch": 2.83,
+      "eval_loss": 0.7307493686676025,
+      "eval_runtime": 17.5153,
+      "eval_samples_per_second": 5.652,
+      "eval_wer": 0.47164948453608246,
+      "step": 10200
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 9.943314438146049e-05,
+      "loss": 0.2973,
+      "step": 10300
+    },
+    {
+      "epoch": 2.86,
+      "eval_loss": 0.7479050755500793,
+      "eval_runtime": 17.4352,
+      "eval_samples_per_second": 5.678,
+      "eval_wer": 0.4420103092783505,
+      "step": 10300
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 9.942758697343559e-05,
+      "loss": 0.3063,
+      "step": 10400
+    },
+    {
+      "epoch": 2.89,
+      "eval_loss": 0.6997365951538086,
+      "eval_runtime": 17.2367,
+      "eval_samples_per_second": 5.744,
+      "eval_wer": 0.44072164948453607,
+      "step": 10400
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 9.94220295654107e-05,
+      "loss": 0.2738,
+      "step": 10500
+    },
+    {
+      "epoch": 2.92,
+      "eval_loss": 0.6006211042404175,
+      "eval_runtime": 17.0511,
+      "eval_samples_per_second": 5.806,
+      "eval_wer": 0.43170103092783507,
+      "step": 10500
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 9.941647215738579e-05,
+      "loss": 0.282,
+      "step": 10600
+    },
+    {
+      "epoch": 2.95,
+      "eval_loss": 0.7482351660728455,
+      "eval_runtime": 17.5673,
+      "eval_samples_per_second": 5.635,
+      "eval_wer": 0.47164948453608246,
+      "step": 10600
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 9.94109147493609e-05,
+      "loss": 0.2666,
+      "step": 10700
+    },
+    {
+      "epoch": 2.97,
+      "eval_loss": 0.7779076099395752,
+      "eval_runtime": 17.5922,
+      "eval_samples_per_second": 5.628,
+      "eval_wer": 0.45489690721649484,
+      "step": 10700
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 9.9405357341336e-05,
+      "loss": 0.28,
+      "step": 10800
+    },
+    {
+      "epoch": 3.0,
+      "eval_loss": 0.7526705265045166,
+      "eval_runtime": 17.6203,
+      "eval_samples_per_second": 5.619,
+      "eval_wer": 0.44587628865979384,
+      "step": 10800
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 9.939979993331111e-05,
+      "loss": 0.2516,
+      "step": 10900
+    },
+    {
+      "epoch": 3.03,
+      "eval_loss": 0.7915493249893188,
+      "eval_runtime": 17.3797,
+      "eval_samples_per_second": 5.696,
+      "eval_wer": 0.45489690721649484,
+      "step": 10900
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 9.939424252528621e-05,
+      "loss": 0.2777,
+      "step": 11000
+    },
+    {
+      "epoch": 3.06,
+      "eval_loss": 0.7125186920166016,
+      "eval_runtime": 17.4703,
+      "eval_samples_per_second": 5.667,
+      "eval_wer": 0.44458762886597936,
+      "step": 11000
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 9.938868511726132e-05,
+      "loss": 0.2502,
+      "step": 11100
+    },
+    {
+      "epoch": 3.08,
+      "eval_loss": 0.7668015360832214,
+      "eval_runtime": 17.4017,
+      "eval_samples_per_second": 5.689,
+      "eval_wer": 0.4265463917525773,
+      "step": 11100
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 9.938312770923641e-05,
+      "loss": 0.2432,
+      "step": 11200
+    },
+    {
+      "epoch": 3.11,
+      "eval_loss": 0.8193698525428772,
+      "eval_runtime": 17.3292,
+      "eval_samples_per_second": 5.713,
+      "eval_wer": 0.47164948453608246,
+      "step": 11200
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 9.937757030121152e-05,
+      "loss": 0.2593,
+      "step": 11300
+    },
+    {
+      "epoch": 3.14,
+      "eval_loss": 0.831807017326355,
+      "eval_runtime": 17.591,
+      "eval_samples_per_second": 5.628,
+      "eval_wer": 0.44329896907216493,
+      "step": 11300
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 9.937201289318662e-05,
+      "loss": 0.2706,
+      "step": 11400
+    },
+    {
+      "epoch": 3.17,
+      "eval_loss": 0.7298266291618347,
+      "eval_runtime": 17.4027,
+      "eval_samples_per_second": 5.689,
+      "eval_wer": 0.4420103092783505,
+      "step": 11400
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 9.936645548516172e-05,
+      "loss": 0.2561,
+      "step": 11500
+    },
+    {
+      "epoch": 3.2,
+      "eval_loss": 0.695881724357605,
+      "eval_runtime": 17.3517,
+      "eval_samples_per_second": 5.705,
+      "eval_wer": 0.45618556701030927,
+      "step": 11500
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 9.936089807713684e-05,
+      "loss": 0.2648,
+      "step": 11600
+    },
+    {
+      "epoch": 3.22,
+      "eval_loss": 0.7764458656311035,
+      "eval_runtime": 17.4577,
+      "eval_samples_per_second": 5.671,
+      "eval_wer": 0.46134020618556704,
+      "step": 11600
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 9.935534066911193e-05,
+      "loss": 0.2547,
+      "step": 11700
+    },
+    {
+      "epoch": 3.25,
+      "eval_loss": 0.6394737958908081,
+      "eval_runtime": 17.2603,
+      "eval_samples_per_second": 5.736,
+      "eval_wer": 0.45489690721649484,
+      "step": 11700
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 9.934978326108704e-05,
+      "loss": 0.2746,
+      "step": 11800
+    },
+    {
+      "epoch": 3.28,
+      "eval_loss": 0.7559604048728943,
+      "eval_runtime": 17.4736,
+      "eval_samples_per_second": 5.666,
+      "eval_wer": 0.4484536082474227,
+      "step": 11800
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 9.934422585306214e-05,
+      "loss": 0.2406,
+      "step": 11900
+    },
+    {
+      "epoch": 3.31,
+      "eval_loss": 0.7239082455635071,
+      "eval_runtime": 17.7291,
+      "eval_samples_per_second": 5.584,
+      "eval_wer": 0.43170103092783507,
+      "step": 11900
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 9.933866844503724e-05,
+      "loss": 0.2394,
+      "step": 12000
+    },
+    {
+      "epoch": 3.33,
+      "eval_loss": 0.7092143297195435,
+      "eval_runtime": 17.5081,
+      "eval_samples_per_second": 5.655,
+      "eval_wer": 0.4420103092783505,
+      "step": 12000
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 9.933311103701234e-05,
+      "loss": 0.2696,
+      "step": 12100
+    },
+    {
+      "epoch": 3.36,
+      "eval_loss": 0.7482930421829224,
+      "eval_runtime": 17.1766,
+      "eval_samples_per_second": 5.764,
+      "eval_wer": 0.4381443298969072,
+      "step": 12100
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 9.932755362898744e-05,
+      "loss": 0.2737,
+      "step": 12200
+    },
+    {
+      "epoch": 3.39,
+      "eval_loss": 0.7534049153327942,
+      "eval_runtime": 17.5047,
+      "eval_samples_per_second": 5.656,
+      "eval_wer": 0.45103092783505155,
+      "step": 12200
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 9.932199622096254e-05,
+      "loss": 0.2796,
+      "step": 12300
+    },
+    {
+      "epoch": 3.42,
+      "eval_loss": 0.7405951619148254,
+      "eval_runtime": 17.3851,
+      "eval_samples_per_second": 5.695,
+      "eval_wer": 0.4536082474226804,
+      "step": 12300
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 9.931643881293766e-05,
+      "loss": 0.2442,
+      "step": 12400
+    },
+    {
+      "epoch": 3.45,
+      "eval_loss": 0.7737710475921631,
+      "eval_runtime": 17.4053,
+      "eval_samples_per_second": 5.688,
+      "eval_wer": 0.45618556701030927,
+      "step": 12400
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 9.931088140491276e-05,
+      "loss": 0.2286,
+      "step": 12500
+    },
+    {
+      "epoch": 3.47,
+      "eval_loss": 0.7984783053398132,
+      "eval_runtime": 17.2754,
+      "eval_samples_per_second": 5.731,
+      "eval_wer": 0.46005154639175255,
+      "step": 12500
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 9.930532399688786e-05,
+      "loss": 0.2778,
+      "step": 12600
+    },
+    {
+      "epoch": 3.5,
+      "eval_loss": 0.73711758852005,
+      "eval_runtime": 17.2963,
+      "eval_samples_per_second": 5.724,
+      "eval_wer": 0.44072164948453607,
+      "step": 12600
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 9.929976658886296e-05,
+      "loss": 0.2456,
+      "step": 12700
+    },
+    {
+      "epoch": 3.53,
+      "eval_loss": 0.7433577179908752,
+      "eval_runtime": 17.1925,
+      "eval_samples_per_second": 5.758,
+      "eval_wer": 0.4368556701030928,
+      "step": 12700
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 9.929420918083806e-05,
+      "loss": 0.252,
+      "step": 12800
+    },
+    {
+      "epoch": 3.56,
+      "eval_loss": 0.6638050079345703,
+      "eval_runtime": 17.2159,
+      "eval_samples_per_second": 5.75,
+      "eval_wer": 0.4329896907216495,
+      "step": 12800
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 9.928865177281317e-05,
+      "loss": 0.2733,
+      "step": 12900
+    },
+    {
+      "epoch": 3.58,
+      "eval_loss": 0.6307669878005981,
+      "eval_runtime": 17.3572,
+      "eval_samples_per_second": 5.704,
+      "eval_wer": 0.44587628865979384,
+      "step": 12900
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 9.928309436478826e-05,
+      "loss": 0.255,
+      "step": 13000
+    },
+    {
+      "epoch": 3.61,
+      "eval_loss": 0.6689628958702087,
+      "eval_runtime": 17.5354,
+      "eval_samples_per_second": 5.646,
+      "eval_wer": 0.4420103092783505,
+      "step": 13000
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 9.927753695676337e-05,
+      "loss": 0.2663,
+      "step": 13100
+    },
+    {
+      "epoch": 3.64,
+      "eval_loss": 0.6840199828147888,
+      "eval_runtime": 17.5476,
+      "eval_samples_per_second": 5.642,
+      "eval_wer": 0.4536082474226804,
+      "step": 13100
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 9.927197954873847e-05,
+      "loss": 0.2666,
+      "step": 13200
+    },
+    {
+      "epoch": 3.67,
+      "eval_loss": 0.6798214912414551,
+      "eval_runtime": 17.5852,
+      "eval_samples_per_second": 5.63,
+      "eval_wer": 0.4368556701030928,
+      "step": 13200
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 9.926642214071357e-05,
+      "loss": 0.2515,
+      "step": 13300
+    },
+    {
+      "epoch": 3.7,
+      "eval_loss": 0.7858713269233704,
+      "eval_runtime": 17.2073,
+      "eval_samples_per_second": 5.753,
+      "eval_wer": 0.4536082474226804,
+      "step": 13300
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 9.926086473268867e-05,
+      "loss": 0.2474,
+      "step": 13400
+    },
+    {
+      "epoch": 3.72,
+      "eval_loss": 0.6592583060264587,
+      "eval_runtime": 17.613,
+      "eval_samples_per_second": 5.621,
+      "eval_wer": 0.45489690721649484,
+      "step": 13400
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 9.925530732466378e-05,
+      "loss": 0.2619,
+      "step": 13500
+    },
+    {
+      "epoch": 3.75,
+      "eval_loss": 0.7771138548851013,
+      "eval_runtime": 17.4838,
+      "eval_samples_per_second": 5.662,
+      "eval_wer": 0.4484536082474227,
+      "step": 13500
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 9.924974991663888e-05,
+      "loss": 0.2541,
+      "step": 13600
+    },
+    {
+      "epoch": 3.78,
+      "eval_loss": 0.6582715511322021,
+      "eval_runtime": 17.4038,
+      "eval_samples_per_second": 5.688,
+      "eval_wer": 0.42783505154639173,
+      "step": 13600
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 9.924419250861399e-05,
+      "loss": 0.2335,
+      "step": 13700
+    },
+    {
+      "epoch": 3.81,
+      "eval_loss": 0.7495954036712646,
+      "eval_runtime": 17.2322,
+      "eval_samples_per_second": 5.745,
+      "eval_wer": 0.4381443298969072,
+      "step": 13700
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 9.923863510058909e-05,
+      "loss": 0.2546,
+      "step": 13800
+    },
+    {
+      "epoch": 3.83,
+      "eval_loss": 0.7367487549781799,
+      "eval_runtime": 17.4399,
+      "eval_samples_per_second": 5.677,
+      "eval_wer": 0.45103092783505155,
+      "step": 13800
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 9.923307769256419e-05,
+      "loss": 0.2369,
+      "step": 13900
+    },
+    {
+      "epoch": 3.86,
+      "eval_loss": 0.7236208915710449,
+      "eval_runtime": 17.2973,
+      "eval_samples_per_second": 5.723,
+      "eval_wer": 0.43943298969072164,
+      "step": 13900
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 9.92275202845393e-05,
+      "loss": 0.2518,
+      "step": 14000
+    },
+    {
+      "epoch": 3.89,
+      "eval_loss": 0.6668509840965271,
+      "eval_runtime": 17.3127,
+      "eval_samples_per_second": 5.718,
+      "eval_wer": 0.44072164948453607,
+      "step": 14000
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 9.922196287651439e-05,
+      "loss": 0.2336,
+      "step": 14100
+    },
+    {
+      "epoch": 3.92,
+      "eval_loss": 0.6360946297645569,
+      "eval_runtime": 17.457,
+      "eval_samples_per_second": 5.671,
+      "eval_wer": 0.43170103092783507,
+      "step": 14100
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 9.92164054684895e-05,
+      "loss": 0.2519,
+      "step": 14200
+    },
+    {
+      "epoch": 3.95,
+      "eval_loss": 0.7355924844741821,
+      "eval_runtime": 17.2736,
+      "eval_samples_per_second": 5.731,
+      "eval_wer": 0.4381443298969072,
+      "step": 14200
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 9.921084806046461e-05,
+      "loss": 0.2505,
+      "step": 14300
+    },
+    {
+      "epoch": 3.97,
+      "eval_loss": 0.6415805816650391,
+      "eval_runtime": 17.406,
+      "eval_samples_per_second": 5.688,
+      "eval_wer": 0.44458762886597936,
+      "step": 14300
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 9.920529065243971e-05,
+      "loss": 0.2671,
+      "step": 14400
+    },
+    {
+      "epoch": 4.0,
+      "eval_loss": 0.764227032661438,
+      "eval_runtime": 17.2386,
+      "eval_samples_per_second": 5.743,
+      "eval_wer": 0.43556701030927836,
+      "step": 14400
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 9.919973324441481e-05,
+      "loss": 0.202,
+      "step": 14500
+    },
+    {
+      "epoch": 4.03,
+      "eval_loss": 0.7409430742263794,
+      "eval_runtime": 17.3261,
+      "eval_samples_per_second": 5.714,
+      "eval_wer": 0.4484536082474227,
+      "step": 14500
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 9.919417583638991e-05,
+      "loss": 0.2304,
+      "step": 14600
+    },
+    {
+      "epoch": 4.06,
+      "eval_loss": 0.7179251909255981,
+      "eval_runtime": 17.3235,
+      "eval_samples_per_second": 5.715,
+      "eval_wer": 0.43943298969072164,
+      "step": 14600
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 9.918861842836501e-05,
+      "loss": 0.2391,
+      "step": 14700
+    },
+    {
+      "epoch": 4.08,
+      "eval_loss": 0.6860632300376892,
+      "eval_runtime": 17.2579,
+      "eval_samples_per_second": 5.737,
+      "eval_wer": 0.43556701030927836,
+      "step": 14700
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 9.918306102034012e-05,
+      "loss": 0.2302,
+      "step": 14800
+    },
+    {
+      "epoch": 4.11,
+      "eval_loss": 0.7088943719863892,
+      "eval_runtime": 17.4142,
+      "eval_samples_per_second": 5.685,
+      "eval_wer": 0.44072164948453607,
+      "step": 14800
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 9.917750361231522e-05,
+      "loss": 0.2395,
+      "step": 14900
+    },
+    {
+      "epoch": 4.14,
+      "eval_loss": 0.8547905683517456,
+      "eval_runtime": 17.2308,
+      "eval_samples_per_second": 5.746,
+      "eval_wer": 0.44587628865979384,
+      "step": 14900
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 9.917194620429032e-05,
+      "loss": 0.2108,
+      "step": 15000
+    },
+    {
+      "epoch": 4.17,
+      "eval_loss": 0.7243941426277161,
+      "eval_runtime": 17.2836,
+      "eval_samples_per_second": 5.728,
+      "eval_wer": 0.42783505154639173,
+      "step": 15000
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 9.916638879626542e-05,
+      "loss": 0.2479,
+      "step": 15100
+    },
+    {
+      "epoch": 4.2,
+      "eval_loss": 0.721064567565918,
+      "eval_runtime": 17.3846,
+      "eval_samples_per_second": 5.695,
+      "eval_wer": 0.43170103092783507,
+      "step": 15100
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 9.916083138824053e-05,
+      "loss": 0.228,
+      "step": 15200
+    },
+    {
+      "epoch": 4.22,
+      "eval_loss": 0.6933774352073669,
+      "eval_runtime": 17.4588,
+      "eval_samples_per_second": 5.671,
+      "eval_wer": 0.44587628865979384,
+      "step": 15200
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 9.915527398021564e-05,
+      "loss": 0.2089,
+      "step": 15300
+    },
+    {
+      "epoch": 4.25,
+      "eval_loss": 0.6892577409744263,
+      "eval_runtime": 17.334,
+      "eval_samples_per_second": 5.711,
+      "eval_wer": 0.43427835051546393,
+      "step": 15300
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 9.914971657219073e-05,
+      "loss": 0.2647,
+      "step": 15400
+    },
+    {
+      "epoch": 4.28,
+      "eval_loss": 0.7205445170402527,
+      "eval_runtime": 17.3769,
+      "eval_samples_per_second": 5.697,
+      "eval_wer": 0.43943298969072164,
+      "step": 15400
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 9.914415916416584e-05,
+      "loss": 0.2111,
+      "step": 15500
+    },
+    {
+      "epoch": 4.31,
+      "eval_loss": 0.7887137532234192,
+      "eval_runtime": 17.5118,
+      "eval_samples_per_second": 5.653,
+      "eval_wer": 0.4484536082474227,
+      "step": 15500
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 9.913860175614094e-05,
+      "loss": 0.2395,
+      "step": 15600
+    },
+    {
+      "epoch": 4.33,
+      "eval_loss": 0.7880820631980896,
+      "eval_runtime": 17.4336,
+      "eval_samples_per_second": 5.679,
+      "eval_wer": 0.4574742268041237,
+      "step": 15600
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 9.913304434811604e-05,
+      "loss": 0.2283,
+      "step": 15700
+    },
+    {
+      "epoch": 4.36,
+      "eval_loss": 0.7957388162612915,
+      "eval_runtime": 17.4422,
+      "eval_samples_per_second": 5.676,
+      "eval_wer": 0.45618556701030927,
+      "step": 15700
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 9.912748694009114e-05,
+      "loss": 0.2315,
+      "step": 15800
+    },
+    {
+      "epoch": 4.39,
+      "eval_loss": 0.7855367064476013,
+      "eval_runtime": 17.8336,
+      "eval_samples_per_second": 5.551,
+      "eval_wer": 0.44716494845360827,
+      "step": 15800
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 9.912192953206624e-05,
+      "loss": 0.2161,
+      "step": 15900
+    },
+    {
+      "epoch": 4.42,
+      "eval_loss": 0.8106959462165833,
+      "eval_runtime": 17.4952,
+      "eval_samples_per_second": 5.659,
+      "eval_wer": 0.4420103092783505,
+      "step": 15900
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 9.911637212404136e-05,
+      "loss": 0.21,
+      "step": 16000
+    },
+    {
+      "epoch": 4.45,
+      "eval_loss": 0.7481738924980164,
+      "eval_runtime": 17.3874,
+      "eval_samples_per_second": 5.694,
+      "eval_wer": 0.42396907216494845,
+      "step": 16000
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 9.911081471601646e-05,
+      "loss": 0.2208,
+      "step": 16100
+    },
+    {
+      "epoch": 4.47,
+      "eval_loss": 0.8416795134544373,
+      "eval_runtime": 17.3637,
+      "eval_samples_per_second": 5.702,
+      "eval_wer": 0.44458762886597936,
+      "step": 16100
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 9.910525730799156e-05,
+      "loss": 0.2339,
+      "step": 16200
+    },
+    {
+      "epoch": 4.5,
+      "eval_loss": 0.7456889152526855,
+      "eval_runtime": 17.6094,
+      "eval_samples_per_second": 5.622,
+      "eval_wer": 0.4574742268041237,
+      "step": 16200
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 9.909969989996666e-05,
+      "loss": 0.2184,
+      "step": 16300
+    },
+    {
+      "epoch": 4.53,
+      "eval_loss": 0.7151892185211182,
+      "eval_runtime": 17.6465,
+      "eval_samples_per_second": 5.61,
+      "eval_wer": 0.42783505154639173,
+      "step": 16300
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 9.909414249194177e-05,
+      "loss": 0.2538,
+      "step": 16400
+    },
+    {
+      "epoch": 4.56,
+      "eval_loss": 0.6038363575935364,
+      "eval_runtime": 17.6191,
+      "eval_samples_per_second": 5.619,
+      "eval_wer": 0.43556701030927836,
+      "step": 16400
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 9.908858508391686e-05,
+      "loss": 0.2241,
+      "step": 16500
+    },
+    {
+      "epoch": 4.58,
+      "eval_loss": 0.7377514243125916,
+      "eval_runtime": 17.3266,
+      "eval_samples_per_second": 5.714,
+      "eval_wer": 0.43943298969072164,
+      "step": 16500
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 9.908302767589197e-05,
+      "loss": 0.2301,
+      "step": 16600
+    },
+    {
+      "epoch": 4.61,
+      "eval_loss": 0.72115558385849,
+      "eval_runtime": 17.2905,
+      "eval_samples_per_second": 5.726,
+      "eval_wer": 0.44587628865979384,
+      "step": 16600
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 9.907747026786707e-05,
+      "loss": 0.2154,
+      "step": 16700
+    },
+    {
+      "epoch": 4.64,
+      "eval_loss": 0.6517682075500488,
+      "eval_runtime": 17.2953,
+      "eval_samples_per_second": 5.724,
+      "eval_wer": 0.44716494845360827,
+      "step": 16700
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 9.907191285984217e-05,
+      "loss": 0.2293,
+      "step": 16800
+    },
+    {
+      "epoch": 4.67,
+      "eval_loss": 0.7426177263259888,
+      "eval_runtime": 17.4692,
+      "eval_samples_per_second": 5.667,
+      "eval_wer": 0.44458762886597936,
+      "step": 16800
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 9.906635545181728e-05,
+      "loss": 0.2189,
+      "step": 16900
+    },
+    {
+      "epoch": 4.7,
+      "eval_loss": 0.7251791954040527,
+      "eval_runtime": 17.5699,
+      "eval_samples_per_second": 5.635,
+      "eval_wer": 0.44072164948453607,
+      "step": 16900
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 9.906079804379238e-05,
+      "loss": 0.2137,
+      "step": 17000
+    },
+    {
+      "epoch": 4.72,
+      "eval_loss": 0.810121476650238,
+      "eval_runtime": 17.5304,
+      "eval_samples_per_second": 5.647,
+      "eval_wer": 0.45103092783505155,
+      "step": 17000
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 9.905524063576749e-05,
+      "loss": 0.2294,
+      "step": 17100
+    },
+    {
+      "epoch": 4.75,
+      "eval_loss": 0.7991353273391724,
+      "eval_runtime": 17.5411,
+      "eval_samples_per_second": 5.644,
+      "eval_wer": 0.422680412371134,
+      "step": 17100
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 9.904968322774259e-05,
+      "loss": 0.1986,
+      "step": 17200
+    },
+    {
+      "epoch": 4.78,
+      "eval_loss": 0.7718966603279114,
+      "eval_runtime": 17.3818,
+      "eval_samples_per_second": 5.696,
+      "eval_wer": 0.4381443298969072,
+      "step": 17200
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 9.904412581971769e-05,
+      "loss": 0.2058,
+      "step": 17300
+    },
+    {
+      "epoch": 4.81,
+      "eval_loss": 0.7458451986312866,
+      "eval_runtime": 17.525,
+      "eval_samples_per_second": 5.649,
+      "eval_wer": 0.43427835051546393,
+      "step": 17300
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 9.903856841169279e-05,
+      "loss": 0.2611,
+      "step": 17400
+    },
+    {
+      "epoch": 4.83,
+      "eval_loss": 0.694915235042572,
+      "eval_runtime": 17.7415,
+      "eval_samples_per_second": 5.58,
+      "eval_wer": 0.422680412371134,
+      "step": 17400
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 9.903301100366789e-05,
+      "loss": 0.2072,
+      "step": 17500
+    },
+    {
+      "epoch": 4.86,
+      "eval_loss": 0.7110035419464111,
+      "eval_runtime": 17.4436,
+      "eval_samples_per_second": 5.675,
+      "eval_wer": 0.43170103092783507,
+      "step": 17500
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 9.902745359564299e-05,
+      "loss": 0.2245,
+      "step": 17600
+    },
+    {
+      "epoch": 4.89,
+      "eval_loss": 0.7854979038238525,
+      "eval_runtime": 17.6469,
+      "eval_samples_per_second": 5.61,
+      "eval_wer": 0.4536082474226804,
+      "step": 17600
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 9.90218961876181e-05,
+      "loss": 0.2456,
+      "step": 17700
+    },
+    {
+      "epoch": 4.92,
+      "eval_loss": 0.860173761844635,
+      "eval_runtime": 17.567,
+      "eval_samples_per_second": 5.636,
+      "eval_wer": 0.4652061855670103,
+      "step": 17700
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 9.90163387795932e-05,
+      "loss": 0.2196,
+      "step": 17800
+    },
+    {
+      "epoch": 4.95,
+      "eval_loss": 0.7404292225837708,
+      "eval_runtime": 17.3988,
+      "eval_samples_per_second": 5.69,
+      "eval_wer": 0.44329896907216493,
+      "step": 17800
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 9.901078137156831e-05,
+      "loss": 0.2153,
+      "step": 17900
+    },
+    {
+      "epoch": 4.97,
+      "eval_loss": 0.8624671101570129,
+      "eval_runtime": 17.5005,
+      "eval_samples_per_second": 5.657,
+      "eval_wer": 0.4652061855670103,
+      "step": 17900
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 9.900522396354341e-05,
+      "loss": 0.2307,
+      "step": 18000
+    },
+    {
+      "epoch": 5.0,
+      "eval_loss": 0.702769935131073,
+      "eval_runtime": 17.8612,
+      "eval_samples_per_second": 5.543,
+      "eval_wer": 0.44329896907216493,
+      "step": 18000
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 9.899966655551851e-05,
+      "loss": 0.2018,
+      "step": 18100
+    },
+    {
+      "epoch": 5.03,
+      "eval_loss": 0.7851367592811584,
+      "eval_runtime": 17.3766,
+      "eval_samples_per_second": 5.697,
+      "eval_wer": 0.44458762886597936,
+      "step": 18100
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 9.899410914749362e-05,
+      "loss": 0.2057,
+      "step": 18200
+    },
+    {
+      "epoch": 5.06,
+      "eval_loss": 0.7941185832023621,
+      "eval_runtime": 17.5072,
+      "eval_samples_per_second": 5.655,
+      "eval_wer": 0.4536082474226804,
+      "step": 18200
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 9.898855173946871e-05,
+      "loss": 0.2075,
+      "step": 18300
+    },
+    {
+      "epoch": 5.08,
+      "eval_loss": 0.6440731287002563,
+      "eval_runtime": 17.3683,
+      "eval_samples_per_second": 5.7,
+      "eval_wer": 0.4368556701030928,
+      "step": 18300
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 9.898299433144382e-05,
+      "loss": 0.2042,
+      "step": 18400
+    },
+    {
+      "epoch": 5.11,
+      "eval_loss": 0.7736424803733826,
+      "eval_runtime": 17.3679,
+      "eval_samples_per_second": 5.7,
+      "eval_wer": 0.4329896907216495,
+      "step": 18400
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 9.897743692341892e-05,
+      "loss": 0.208,
+      "step": 18500
+    },
+    {
+      "epoch": 5.14,
+      "eval_loss": 0.7068067789077759,
+      "eval_runtime": 17.8158,
+      "eval_samples_per_second": 5.557,
+      "eval_wer": 0.43943298969072164,
+      "step": 18500
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 9.897187951539403e-05,
+      "loss": 0.2056,
+      "step": 18600
+    },
+    {
+      "epoch": 5.17,
+      "eval_loss": 0.8301738500595093,
+      "eval_runtime": 17.4918,
+      "eval_samples_per_second": 5.66,
+      "eval_wer": 0.45103092783505155,
+      "step": 18600
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 9.896632210736913e-05,
+      "loss": 0.2037,
+      "step": 18700
+    },
+    {
+      "epoch": 5.2,
+      "eval_loss": 0.7583587169647217,
+      "eval_runtime": 17.3452,
+      "eval_samples_per_second": 5.708,
+      "eval_wer": 0.4381443298969072,
+      "step": 18700
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 9.896076469934423e-05,
+      "loss": 0.204,
+      "step": 18800
+    },
+    {
+      "epoch": 5.22,
+      "eval_loss": 0.7550833821296692,
+      "eval_runtime": 17.4638,
+      "eval_samples_per_second": 5.669,
+      "eval_wer": 0.4742268041237113,
+      "step": 18800
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 9.895520729131933e-05,
+      "loss": 0.2118,
+      "step": 18900
+    },
+    {
+      "epoch": 5.25,
+      "eval_loss": 0.7639862895011902,
+      "eval_runtime": 17.3009,
+      "eval_samples_per_second": 5.722,
+      "eval_wer": 0.4420103092783505,
+      "step": 18900
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 9.894964988329444e-05,
+      "loss": 0.2005,
+      "step": 19000
+    },
+    {
+      "epoch": 5.28,
+      "eval_loss": 0.7661750912666321,
+      "eval_runtime": 17.8661,
+      "eval_samples_per_second": 5.541,
+      "eval_wer": 0.45103092783505155,
+      "step": 19000
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 9.894409247526953e-05,
+      "loss": 0.2234,
+      "step": 19100
+    },
+    {
+      "epoch": 5.31,
+      "eval_loss": 0.7485219836235046,
+      "eval_runtime": 17.886,
+      "eval_samples_per_second": 5.535,
+      "eval_wer": 0.44716494845360827,
+      "step": 19100
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 9.893853506724464e-05,
+      "loss": 0.2204,
+      "step": 19200
+    },
+    {
+      "epoch": 5.33,
+      "eval_loss": 0.827965497970581,
+      "eval_runtime": 17.4582,
+      "eval_samples_per_second": 5.671,
+      "eval_wer": 0.4420103092783505,
+      "step": 19200
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 9.893297765921976e-05,
+      "loss": 0.233,
+      "step": 19300
+    },
+    {
+      "epoch": 5.36,
+      "eval_loss": 0.750889778137207,
+      "eval_runtime": 17.5168,
+      "eval_samples_per_second": 5.652,
+      "eval_wer": 0.452319587628866,
+      "step": 19300
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 9.892742025119484e-05,
+      "loss": 0.217,
+      "step": 19400
+    },
+    {
+      "epoch": 5.39,
+      "eval_loss": 0.8290805220603943,
+      "eval_runtime": 17.3789,
+      "eval_samples_per_second": 5.697,
+      "eval_wer": 0.452319587628866,
+      "step": 19400
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 9.892186284316996e-05,
+      "loss": 0.2061,
+      "step": 19500
+    },
+    {
+      "epoch": 5.42,
+      "eval_loss": 0.7057229280471802,
+      "eval_runtime": 17.3775,
+      "eval_samples_per_second": 5.697,
+      "eval_wer": 0.45103092783505155,
+      "step": 19500
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 9.891630543514506e-05,
+      "loss": 0.2044,
+      "step": 19600
+    },
+    {
+      "epoch": 5.45,
+      "eval_loss": 0.7502852082252502,
+      "eval_runtime": 17.4524,
+      "eval_samples_per_second": 5.673,
+      "eval_wer": 0.4484536082474227,
+      "step": 19600
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 9.891074802712016e-05,
+      "loss": 0.2063,
+      "step": 19700
+    },
+    {
+      "epoch": 5.47,
+      "eval_loss": 0.7316065430641174,
+      "eval_runtime": 17.2186,
+      "eval_samples_per_second": 5.75,
+      "eval_wer": 0.4368556701030928,
+      "step": 19700
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 9.890519061909526e-05,
+      "loss": 0.1778,
+      "step": 19800
+    },
+    {
+      "epoch": 5.5,
+      "eval_loss": 0.7007728219032288,
+      "eval_runtime": 17.1646,
+      "eval_samples_per_second": 5.768,
+      "eval_wer": 0.44587628865979384,
+      "step": 19800
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 9.889963321107036e-05,
+      "loss": 0.2153,
+      "step": 19900
+    },
+    {
+      "epoch": 5.53,
+      "eval_loss": 0.6902391314506531,
+      "eval_runtime": 17.7569,
+      "eval_samples_per_second": 5.575,
+      "eval_wer": 0.43170103092783507,
+      "step": 19900
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 9.889407580304546e-05,
+      "loss": 0.1939,
+      "step": 20000
+    },
+    {
+      "epoch": 5.56,
+      "eval_loss": 0.7636317014694214,
+      "eval_runtime": 17.4399,
+      "eval_samples_per_second": 5.677,
+      "eval_wer": 0.4484536082474227,
+      "step": 20000
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 9.888851839502057e-05,
+      "loss": 0.204,
+      "step": 20100
+    },
+    {
+      "epoch": 5.58,
+      "eval_loss": 0.7957937121391296,
+      "eval_runtime": 17.6575,
+      "eval_samples_per_second": 5.607,
+      "eval_wer": 0.44072164948453607,
+      "step": 20100
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 9.888296098699566e-05,
+      "loss": 0.1844,
+      "step": 20200
+    },
+    {
+      "epoch": 5.61,
+      "eval_loss": 0.7200835347175598,
+      "eval_runtime": 17.5413,
+      "eval_samples_per_second": 5.644,
+      "eval_wer": 0.4420103092783505,
+      "step": 20200
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 9.887740357897077e-05,
+      "loss": 0.2165,
+      "step": 20300
+    },
+    {
+      "epoch": 5.64,
+      "eval_loss": 0.8324032425880432,
+      "eval_runtime": 17.4764,
+      "eval_samples_per_second": 5.665,
+      "eval_wer": 0.45489690721649484,
+      "step": 20300
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 9.887184617094588e-05,
+      "loss": 0.2087,
+      "step": 20400
+    },
+    {
+      "epoch": 5.67,
+      "eval_loss": 0.7996511459350586,
+      "eval_runtime": 17.4391,
+      "eval_samples_per_second": 5.677,
+      "eval_wer": 0.45618556701030927,
+      "step": 20400
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 9.886628876292098e-05,
+      "loss": 0.2297,
+      "step": 20500
+    },
+    {
+      "epoch": 5.7,
+      "eval_loss": 0.8123847842216492,
+      "eval_runtime": 17.4148,
+      "eval_samples_per_second": 5.685,
+      "eval_wer": 0.4574742268041237,
+      "step": 20500
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 9.886073135489609e-05,
+      "loss": 0.2202,
+      "step": 20600
+    },
+    {
+      "epoch": 5.72,
+      "eval_loss": 0.8624526262283325,
+      "eval_runtime": 17.5176,
+      "eval_samples_per_second": 5.651,
+      "eval_wer": 0.452319587628866,
+      "step": 20600
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 9.885517394687118e-05,
+      "loss": 0.2093,
+      "step": 20700
+    },
+    {
+      "epoch": 5.75,
+      "eval_loss": 0.8562197685241699,
+      "eval_runtime": 17.5151,
+      "eval_samples_per_second": 5.652,
+      "eval_wer": 0.46262886597938147,
+      "step": 20700
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 9.884961653884629e-05,
+      "loss": 0.2093,
+      "step": 20800
+    },
+    {
+      "epoch": 5.78,
+      "eval_loss": 0.8221404552459717,
+      "eval_runtime": 17.6305,
+      "eval_samples_per_second": 5.615,
+      "eval_wer": 0.47164948453608246,
+      "step": 20800
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 9.884405913082139e-05,
+      "loss": 0.2303,
+      "step": 20900
+    },
+    {
+      "epoch": 5.81,
+      "eval_loss": 0.7461365461349487,
+      "eval_runtime": 17.3023,
+      "eval_samples_per_second": 5.722,
+      "eval_wer": 0.43943298969072164,
+      "step": 20900
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 9.883850172279649e-05,
+      "loss": 0.2137,
+      "step": 21000
+    },
+    {
+      "epoch": 5.83,
+      "eval_loss": 0.8054025769233704,
+      "eval_runtime": 17.4479,
+      "eval_samples_per_second": 5.674,
+      "eval_wer": 0.4536082474226804,
+      "step": 21000
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 9.883294431477159e-05,
+      "loss": 0.2042,
+      "step": 21100
+    },
+    {
+      "epoch": 5.86,
+      "eval_loss": 0.7558605074882507,
+      "eval_runtime": 17.4077,
+      "eval_samples_per_second": 5.687,
+      "eval_wer": 0.44716494845360827,
+      "step": 21100
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 9.88273869067467e-05,
+      "loss": 0.1955,
+      "step": 21200
+    },
+    {
+      "epoch": 5.89,
+      "eval_loss": 0.8814973831176758,
+      "eval_runtime": 17.5114,
+      "eval_samples_per_second": 5.653,
+      "eval_wer": 0.44329896907216493,
+      "step": 21200
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 9.88218294987218e-05,
+      "loss": 0.2004,
+      "step": 21300
+    },
+    {
+      "epoch": 5.92,
+      "eval_loss": 0.8037664294242859,
+      "eval_runtime": 17.5704,
+      "eval_samples_per_second": 5.634,
+      "eval_wer": 0.46649484536082475,
+      "step": 21300
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 9.881627209069691e-05,
+      "loss": 0.2085,
+      "step": 21400
+    },
+    {
+      "epoch": 5.95,
+      "eval_loss": 0.7956470251083374,
+      "eval_runtime": 17.3299,
+      "eval_samples_per_second": 5.713,
+      "eval_wer": 0.452319587628866,
+      "step": 21400
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 9.881071468267201e-05,
+      "loss": 0.2049,
+      "step": 21500
+    },
+    {
+      "epoch": 5.97,
+      "eval_loss": 0.7397593259811401,
+      "eval_runtime": 17.4776,
+      "eval_samples_per_second": 5.664,
+      "eval_wer": 0.4368556701030928,
+      "step": 21500
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 9.880515727464711e-05,
+      "loss": 0.1847,
+      "step": 21600
+    },
+    {
+      "epoch": 6.0,
+      "eval_loss": 0.7041661143302917,
+      "eval_runtime": 17.5194,
+      "eval_samples_per_second": 5.651,
+      "eval_wer": 0.42396907216494845,
+      "step": 21600
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 9.879959986662222e-05,
+      "loss": 0.1834,
+      "step": 21700
+    },
+    {
+      "epoch": 6.03,
+      "eval_loss": 0.7885836362838745,
+      "eval_runtime": 17.3117,
+      "eval_samples_per_second": 5.719,
+      "eval_wer": 0.4729381443298969,
+      "step": 21700
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 9.879404245859731e-05,
+      "loss": 0.1682,
+      "step": 21800
+    },
+    {
+      "epoch": 6.06,
+      "eval_loss": 0.7541030049324036,
+      "eval_runtime": 17.3759,
+      "eval_samples_per_second": 5.698,
+      "eval_wer": 0.43041237113402064,
+      "step": 21800
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 9.878848505057242e-05,
+      "loss": 0.1658,
+      "step": 21900
+    },
+    {
+      "epoch": 6.09,
+      "eval_loss": 0.7061160206794739,
+      "eval_runtime": 17.5618,
+      "eval_samples_per_second": 5.637,
+      "eval_wer": 0.44458762886597936,
+      "step": 21900
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 9.878292764254752e-05,
+      "loss": 0.1986,
+      "step": 22000
+    },
+    {
+      "epoch": 6.11,
+      "eval_loss": 0.7886027097702026,
+      "eval_runtime": 17.6462,
+      "eval_samples_per_second": 5.61,
+      "eval_wer": 0.4329896907216495,
+      "step": 22000
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 9.877737023452263e-05,
+      "loss": 0.1833,
+      "step": 22100
+    },
+    {
+      "epoch": 6.14,
+      "eval_loss": 0.6864758729934692,
+      "eval_runtime": 17.5319,
+      "eval_samples_per_second": 5.647,
+      "eval_wer": 0.43041237113402064,
+      "step": 22100
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 9.877181282649773e-05,
+      "loss": 0.2008,
+      "step": 22200
+    },
+    {
+      "epoch": 6.17,
+      "eval_loss": 0.7436667680740356,
+      "eval_runtime": 17.6401,
+      "eval_samples_per_second": 5.612,
+      "eval_wer": 0.42010309278350516,
+      "step": 22200
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 9.876625541847283e-05,
+      "loss": 0.2114,
+      "step": 22300
+    },
+    {
+      "epoch": 6.2,
+      "eval_loss": 0.7265353798866272,
+      "eval_runtime": 17.6855,
+      "eval_samples_per_second": 5.598,
+      "eval_wer": 0.44329896907216493,
+      "step": 22300
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 9.876069801044793e-05,
+      "loss": 0.1848,
+      "step": 22400
+    },
+    {
+      "epoch": 6.22,
+      "eval_loss": 0.663320779800415,
+      "eval_runtime": 17.7987,
+      "eval_samples_per_second": 5.562,
+      "eval_wer": 0.43170103092783507,
+      "step": 22400
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 9.875514060242304e-05,
+      "loss": 0.1877,
+      "step": 22500
+    },
+    {
+      "epoch": 6.25,
+      "eval_loss": 0.6840032935142517,
+      "eval_runtime": 18.0706,
+      "eval_samples_per_second": 5.479,
+      "eval_wer": 0.45489690721649484,
+      "step": 22500
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 9.874958319439813e-05,
+      "loss": 0.1667,
+      "step": 22600
+    },
+    {
+      "epoch": 6.28,
+      "eval_loss": 0.6797974109649658,
+      "eval_runtime": 17.8615,
+      "eval_samples_per_second": 5.543,
+      "eval_wer": 0.44329896907216493,
+      "step": 22600
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 9.874402578637324e-05,
+      "loss": 0.1786,
+      "step": 22700
+    },
+    {
+      "epoch": 6.31,
+      "eval_loss": 0.7095053195953369,
+      "eval_runtime": 18.0172,
+      "eval_samples_per_second": 5.495,
+      "eval_wer": 0.4536082474226804,
+      "step": 22700
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 9.873846837834834e-05,
+      "loss": 0.1952,
+      "step": 22800
+    },
+    {
+      "epoch": 6.34,
+      "eval_loss": 0.6215536594390869,
+      "eval_runtime": 17.9111,
+      "eval_samples_per_second": 5.527,
+      "eval_wer": 0.43170103092783507,
+      "step": 22800
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 9.873291097032344e-05,
+      "loss": 0.2022,
+      "step": 22900
+    },
+    {
+      "epoch": 6.36,
+      "eval_loss": 0.6748619675636292,
+      "eval_runtime": 17.9802,
+      "eval_samples_per_second": 5.506,
+      "eval_wer": 0.4587628865979381,
+      "step": 22900
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 9.872735356229856e-05,
+      "loss": 0.1906,
+      "step": 23000
+    },
+    {
+      "epoch": 6.39,
+      "eval_loss": 0.6207524538040161,
+      "eval_runtime": 17.8119,
+      "eval_samples_per_second": 5.558,
+      "eval_wer": 0.44072164948453607,
+      "step": 23000
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 9.872179615427364e-05,
+      "loss": 0.1926,
+      "step": 23100
+    },
+    {
+      "epoch": 6.42,
+      "eval_loss": 0.7496910691261292,
+      "eval_runtime": 18.1906,
+      "eval_samples_per_second": 5.442,
+      "eval_wer": 0.4574742268041237,
+      "step": 23100
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 9.871623874624876e-05,
+      "loss": 0.1854,
+      "step": 23200
+    },
+    {
+      "epoch": 6.45,
+      "eval_loss": 0.685153603553772,
+      "eval_runtime": 17.9086,
+      "eval_samples_per_second": 5.528,
+      "eval_wer": 0.46005154639175255,
+      "step": 23200
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 9.871068133822386e-05,
+      "loss": 0.1576,
+      "step": 23300
+    },
+    {
+      "epoch": 6.47,
+      "eval_loss": 0.649712085723877,
+      "eval_runtime": 17.9237,
+      "eval_samples_per_second": 5.523,
+      "eval_wer": 0.4497422680412371,
+      "step": 23300
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 9.870512393019896e-05,
+      "loss": 0.1823,
+      "step": 23400
+    },
+    {
+      "epoch": 6.5,
+      "eval_loss": 0.7309712767601013,
+      "eval_runtime": 18.1848,
+      "eval_samples_per_second": 5.444,
+      "eval_wer": 0.44329896907216493,
+      "step": 23400
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 9.869956652217406e-05,
+      "loss": 0.1693,
+      "step": 23500
+    },
+    {
+      "epoch": 6.53,
+      "eval_loss": 0.7329779863357544,
+      "eval_runtime": 17.7531,
+      "eval_samples_per_second": 5.576,
+      "eval_wer": 0.4652061855670103,
+      "step": 23500
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 9.869400911414916e-05,
+      "loss": 0.1861,
+      "step": 23600
+    },
+    {
+      "epoch": 6.56,
+      "eval_loss": 0.8161032795906067,
+      "eval_runtime": 17.6576,
+      "eval_samples_per_second": 5.607,
+      "eval_wer": 0.46649484536082475,
+      "step": 23600
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 9.868845170612426e-05,
+      "loss": 0.1955,
+      "step": 23700
+    },
+    {
+      "epoch": 6.59,
+      "eval_loss": 0.7811844944953918,
+      "eval_runtime": 18.1639,
+      "eval_samples_per_second": 5.45,
+      "eval_wer": 0.4652061855670103,
+      "step": 23700
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 9.868289429809938e-05,
+      "loss": 0.1801,
+      "step": 23800
+    },
+    {
+      "epoch": 6.61,
+      "eval_loss": 0.7747323513031006,
+      "eval_runtime": 17.4915,
+      "eval_samples_per_second": 5.66,
+      "eval_wer": 0.4652061855670103,
+      "step": 23800
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 9.867733689007448e-05,
+      "loss": 0.1954,
+      "step": 23900
+    },
+    {
+      "epoch": 6.64,
+      "eval_loss": 0.8298807144165039,
+      "eval_runtime": 17.8544,
+      "eval_samples_per_second": 5.545,
+      "eval_wer": 0.4690721649484536,
+      "step": 23900
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 9.867177948204958e-05,
+      "loss": 0.1976,
+      "step": 24000
+    },
+    {
+      "epoch": 6.67,
+      "eval_loss": 0.8623689413070679,
+      "eval_runtime": 17.5465,
+      "eval_samples_per_second": 5.642,
+      "eval_wer": 0.452319587628866,
+      "step": 24000
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 9.866622207402468e-05,
+      "loss": 0.1839,
+      "step": 24100
+    },
+    {
+      "epoch": 6.7,
+      "eval_loss": 0.8633375763893127,
+      "eval_runtime": 17.7487,
+      "eval_samples_per_second": 5.578,
+      "eval_wer": 0.45103092783505155,
+      "step": 24100
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 9.866066466599978e-05,
+      "loss": 0.1858,
+      "step": 24200
+    },
+    {
+      "epoch": 6.72,
+      "eval_loss": 1.0021711587905884,
+      "eval_runtime": 17.7581,
+      "eval_samples_per_second": 5.575,
+      "eval_wer": 0.4845360824742268,
+      "step": 24200
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 9.865510725797489e-05,
+      "loss": 0.1753,
+      "step": 24300
+    },
+    {
+      "epoch": 6.75,
+      "eval_loss": 0.904293417930603,
+      "eval_runtime": 18.4007,
+      "eval_samples_per_second": 5.38,
+      "eval_wer": 0.46649484536082475,
+      "step": 24300
+    },
+    {
+      "epoch": 6.78,
+      "learning_rate": 9.864954984994998e-05,
+      "loss": 0.1829,
+      "step": 24400
+    },
+    {
+      "epoch": 6.78,
+      "eval_loss": 0.7328855991363525,
+      "eval_runtime": 17.5994,
+      "eval_samples_per_second": 5.625,
+      "eval_wer": 0.44458762886597936,
+      "step": 24400
+    },
+    {
+      "epoch": 6.81,
+      "learning_rate": 9.864399244192509e-05,
+      "loss": 0.1984,
+      "step": 24500
+    },
+    {
+      "epoch": 6.81,
+      "eval_loss": 0.7215154767036438,
+      "eval_runtime": 17.6361,
+      "eval_samples_per_second": 5.613,
+      "eval_wer": 0.43556701030927836,
+      "step": 24500
+    },
+    {
+      "epoch": 6.84,
+      "learning_rate": 9.863843503390019e-05,
+      "loss": 0.2195,
+      "step": 24600
+    },
+    {
+      "epoch": 6.84,
+      "eval_loss": 0.7022905945777893,
+      "eval_runtime": 17.8301,
+      "eval_samples_per_second": 5.552,
+      "eval_wer": 0.46005154639175255,
+      "step": 24600
+    },
+    {
+      "epoch": 6.86,
+      "learning_rate": 9.86328776258753e-05,
+      "loss": 0.2079,
+      "step": 24700
+    },
+    {
+      "epoch": 6.86,
+      "eval_loss": 0.7629209756851196,
+      "eval_runtime": 18.0665,
+      "eval_samples_per_second": 5.48,
+      "eval_wer": 0.4639175257731959,
+      "step": 24700
+    },
+    {
+      "epoch": 6.89,
+      "learning_rate": 9.86273202178504e-05,
+      "loss": 0.2118,
+      "step": 24800
+    },
+    {
+      "epoch": 6.89,
+      "eval_loss": 0.6994808316230774,
+      "eval_runtime": 17.5504,
+      "eval_samples_per_second": 5.641,
+      "eval_wer": 0.43556701030927836,
+      "step": 24800
+    },
+    {
+      "epoch": 6.92,
+      "learning_rate": 9.862176280982551e-05,
+      "loss": 0.2154,
+      "step": 24900
+    },
+    {
+      "epoch": 6.92,
+      "eval_loss": 0.6871984601020813,
+      "eval_runtime": 18.0511,
+      "eval_samples_per_second": 5.484,
+      "eval_wer": 0.4484536082474227,
+      "step": 24900
+    },
+    {
+      "epoch": 6.95,
+      "learning_rate": 9.861620540180061e-05,
+      "loss": 0.1823,
+      "step": 25000
+    },
+    {
+      "epoch": 6.95,
+      "eval_loss": 0.5884273052215576,
+      "eval_runtime": 17.6454,
+      "eval_samples_per_second": 5.611,
+      "eval_wer": 0.4420103092783505,
+      "step": 25000
+    },
+    {
+      "epoch": 6.97,
+      "learning_rate": 9.861064799377571e-05,
+      "loss": 0.1786,
+      "step": 25100
+    },
+    {
+      "epoch": 6.97,
+      "eval_loss": 0.7519210577011108,
+      "eval_runtime": 17.8606,
+      "eval_samples_per_second": 5.543,
+      "eval_wer": 0.46262886597938147,
+      "step": 25100
+    },
+    {
+      "epoch": 7.0,
+      "learning_rate": 9.860509058575081e-05,
+      "loss": 0.1841,
+      "step": 25200
+    },
+    {
+      "epoch": 7.0,
+      "eval_loss": 0.7630220651626587,
+      "eval_runtime": 17.5638,
+      "eval_samples_per_second": 5.637,
+      "eval_wer": 0.4420103092783505,
+      "step": 25200
+    },
+    {
+      "epoch": 7.03,
+      "learning_rate": 9.859953317772591e-05,
+      "loss": 0.1866,
+      "step": 25300
+    },
+    {
+      "epoch": 7.03,
+      "eval_loss": 0.7063058614730835,
+      "eval_runtime": 17.2492,
+      "eval_samples_per_second": 5.739,
+      "eval_wer": 0.45618556701030927,
+      "step": 25300
+    },
+    {
+      "epoch": 7.06,
+      "learning_rate": 9.859397576970102e-05,
+      "loss": 0.1756,
+      "step": 25400
+    },
+    {
+      "epoch": 7.06,
+      "eval_loss": 0.6924075484275818,
+      "eval_runtime": 18.1923,
+      "eval_samples_per_second": 5.442,
+      "eval_wer": 0.45618556701030927,
+      "step": 25400
+    },
+    {
+      "epoch": 7.09,
+      "learning_rate": 9.858841836167611e-05,
+      "loss": 0.1834,
+      "step": 25500
+    },
+    {
+      "epoch": 7.09,
+      "eval_loss": 0.7205408215522766,
+      "eval_runtime": 17.5655,
+      "eval_samples_per_second": 5.636,
+      "eval_wer": 0.4420103092783505,
+      "step": 25500
+    },
+    {
+      "epoch": 7.11,
+      "learning_rate": 9.858286095365123e-05,
+      "loss": 0.1556,
+      "step": 25600
+    },
+    {
+      "epoch": 7.11,
+      "eval_loss": 0.9185993671417236,
+      "eval_runtime": 17.8154,
+      "eval_samples_per_second": 5.557,
+      "eval_wer": 0.44458762886597936,
+      "step": 25600
+    },
+    {
+      "epoch": 7.14,
+      "learning_rate": 9.857730354562633e-05,
+      "loss": 0.1736,
+      "step": 25700
+    },
+    {
+      "epoch": 7.14,
+      "eval_loss": 0.7671197652816772,
+      "eval_runtime": 17.8195,
+      "eval_samples_per_second": 5.556,
+      "eval_wer": 0.44329896907216493,
+      "step": 25700
+    },
+    {
+      "epoch": 7.17,
+      "learning_rate": 9.857174613760143e-05,
+      "loss": 0.1535,
+      "step": 25800
+    },
+    {
+      "epoch": 7.17,
+      "eval_loss": 0.813725471496582,
+      "eval_runtime": 17.458,
+      "eval_samples_per_second": 5.671,
+      "eval_wer": 0.44716494845360827,
+      "step": 25800
+    },
+    {
+      "epoch": 7.2,
+      "learning_rate": 9.856618872957653e-05,
+      "loss": 0.1725,
+      "step": 25900
+    },
+    {
+      "epoch": 7.2,
+      "eval_loss": 0.860133707523346,
+      "eval_runtime": 17.7915,
+      "eval_samples_per_second": 5.564,
+      "eval_wer": 0.45618556701030927,
+      "step": 25900
+    },
+    {
+      "epoch": 7.22,
+      "learning_rate": 9.856063132155163e-05,
+      "loss": 0.1883,
+      "step": 26000
+    },
+    {
+      "epoch": 7.22,
+      "eval_loss": 0.846496045589447,
+      "eval_runtime": 17.5964,
+      "eval_samples_per_second": 5.626,
+      "eval_wer": 0.4690721649484536,
+      "step": 26000
+    },
+    {
+      "epoch": 7.25,
+      "learning_rate": 9.855507391352674e-05,
+      "loss": 0.1626,
+      "step": 26100
+    },
+    {
+      "epoch": 7.25,
+      "eval_loss": 0.6927329301834106,
+      "eval_runtime": 17.582,
+      "eval_samples_per_second": 5.631,
+      "eval_wer": 0.42783505154639173,
+      "step": 26100
+    },
+    {
+      "epoch": 7.28,
+      "learning_rate": 9.854951650550184e-05,
+      "loss": 0.1758,
+      "step": 26200
+    },
+    {
+      "epoch": 7.28,
+      "eval_loss": 0.7043600082397461,
+      "eval_runtime": 18.8361,
+      "eval_samples_per_second": 5.256,
+      "eval_wer": 0.4252577319587629,
+      "step": 26200
+    },
+    {
+      "epoch": 7.31,
+      "learning_rate": 9.854395909747694e-05,
+      "loss": 0.1568,
+      "step": 26300
+    },
+    {
+      "epoch": 7.31,
+      "eval_loss": 0.7357842326164246,
+      "eval_runtime": 18.3695,
+      "eval_samples_per_second": 5.389,
+      "eval_wer": 0.44587628865979384,
+      "step": 26300
+    },
+    {
+      "epoch": 7.34,
+      "learning_rate": 9.853840168945204e-05,
+      "loss": 0.1583,
+      "step": 26400
+    },
+    {
+      "epoch": 7.34,
+      "eval_loss": 0.7848875522613525,
+      "eval_runtime": 17.9559,
+      "eval_samples_per_second": 5.513,
+      "eval_wer": 0.4420103092783505,
+      "step": 26400
+    },
+    {
+      "epoch": 7.36,
+      "learning_rate": 9.853284428142714e-05,
+      "loss": 0.167,
+      "step": 26500
+    },
+    {
+      "epoch": 7.36,
+      "eval_loss": 0.7771514654159546,
+      "eval_runtime": 17.8217,
+      "eval_samples_per_second": 5.555,
+      "eval_wer": 0.4420103092783505,
+      "step": 26500
+    },
+    {
+      "epoch": 7.39,
+      "learning_rate": 9.852728687340224e-05,
+      "loss": 0.1659,
+      "step": 26600
+    },
+    {
+      "epoch": 7.39,
+      "eval_loss": 0.8126515746116638,
+      "eval_runtime": 18.1033,
+      "eval_samples_per_second": 5.469,
+      "eval_wer": 0.4420103092783505,
+      "step": 26600
+    },
+    {
+      "epoch": 7.42,
+      "learning_rate": 9.852172946537736e-05,
+      "loss": 0.1909,
+      "step": 26700
+    },
+    {
+      "epoch": 7.42,
+      "eval_loss": 0.7457703948020935,
+      "eval_runtime": 18.2622,
+      "eval_samples_per_second": 5.421,
+      "eval_wer": 0.43427835051546393,
+      "step": 26700
+    },
+    {
+      "epoch": 7.45,
+      "learning_rate": 9.851617205735245e-05,
+      "loss": 0.1842,
+      "step": 26800
+    },
+    {
+      "epoch": 7.45,
+      "eval_loss": 0.7543534636497498,
+      "eval_runtime": 18.2488,
+      "eval_samples_per_second": 5.425,
+      "eval_wer": 0.45103092783505155,
+      "step": 26800
+    },
+    {
+      "epoch": 7.47,
+      "learning_rate": 9.851061464932756e-05,
+      "loss": 0.1839,
+      "step": 26900
+    },
+    {
+      "epoch": 7.47,
+      "eval_loss": 0.7522377371788025,
+      "eval_runtime": 18.0753,
+      "eval_samples_per_second": 5.477,
+      "eval_wer": 0.43170103092783507,
+      "step": 26900
+    },
+    {
+      "epoch": 7.5,
+      "learning_rate": 9.850505724130266e-05,
+      "loss": 0.158,
+      "step": 27000
+    },
+    {
+      "epoch": 7.5,
+      "eval_loss": 0.7622743844985962,
+      "eval_runtime": 17.6407,
+      "eval_samples_per_second": 5.612,
+      "eval_wer": 0.4265463917525773,
+      "step": 27000
+    },
+    {
+      "epoch": 7.53,
+      "learning_rate": 9.849949983327776e-05,
+      "loss": 0.1685,
+      "step": 27100
+    },
+    {
+      "epoch": 7.53,
+      "eval_loss": 0.7140101194381714,
+      "eval_runtime": 17.3617,
+      "eval_samples_per_second": 5.702,
+      "eval_wer": 0.4368556701030928,
+      "step": 27100
+    },
+    {
+      "epoch": 7.56,
+      "learning_rate": 9.849394242525287e-05,
+      "loss": 0.1785,
+      "step": 27200
+    },
+    {
+      "epoch": 7.56,
+      "eval_loss": 0.7777717709541321,
+      "eval_runtime": 17.2778,
+      "eval_samples_per_second": 5.73,
+      "eval_wer": 0.4097938144329897,
+      "step": 27200
+    },
+    {
+      "epoch": 7.59,
+      "learning_rate": 9.848838501722798e-05,
+      "loss": 0.1633,
+      "step": 27300
+    },
+    {
+      "epoch": 7.59,
+      "eval_loss": 0.7742789387702942,
+      "eval_runtime": 17.358,
+      "eval_samples_per_second": 5.703,
+      "eval_wer": 0.41881443298969073,
+      "step": 27300
+    },
+    {
+      "epoch": 7.61,
+      "learning_rate": 9.848282760920308e-05,
+      "loss": 0.1841,
+      "step": 27400
+    },
+    {
+      "epoch": 7.61,
+      "eval_loss": 0.794254720211029,
+      "eval_runtime": 17.5933,
+      "eval_samples_per_second": 5.627,
+      "eval_wer": 0.4368556701030928,
+      "step": 27400
+    },
+    {
+      "epoch": 7.64,
+      "learning_rate": 9.847727020117818e-05,
+      "loss": 0.176,
+      "step": 27500
+    },
+    {
+      "epoch": 7.64,
+      "eval_loss": 0.7638439536094666,
+      "eval_runtime": 17.3793,
+      "eval_samples_per_second": 5.696,
+      "eval_wer": 0.45489690721649484,
+      "step": 27500
+    },
+    {
+      "epoch": 7.67,
+      "learning_rate": 9.847171279315328e-05,
+      "loss": 0.158,
+      "step": 27600
+    },
+    {
+      "epoch": 7.67,
+      "eval_loss": 0.7681997418403625,
+      "eval_runtime": 17.46,
+      "eval_samples_per_second": 5.67,
+      "eval_wer": 0.4574742268041237,
+      "step": 27600
+    },
+    {
+      "epoch": 7.7,
+      "learning_rate": 9.846615538512838e-05,
+      "loss": 0.1915,
+      "step": 27700
+    },
+    {
+      "epoch": 7.7,
+      "eval_loss": 0.7488833069801331,
+      "eval_runtime": 17.731,
+      "eval_samples_per_second": 5.583,
+      "eval_wer": 0.44072164948453607,
+      "step": 27700
+    },
+    {
+      "epoch": 7.72,
+      "learning_rate": 9.846059797710349e-05,
+      "loss": 0.1634,
+      "step": 27800
+    },
+    {
+      "epoch": 7.72,
+      "eval_loss": 0.7865298390388489,
+      "eval_runtime": 17.4124,
+      "eval_samples_per_second": 5.686,
+      "eval_wer": 0.43556701030927836,
+      "step": 27800
+    },
+    {
+      "epoch": 7.75,
+      "learning_rate": 9.845504056907858e-05,
+      "loss": 0.2035,
+      "step": 27900
+    },
+    {
+      "epoch": 7.75,
+      "eval_loss": 0.7207059860229492,
+      "eval_runtime": 17.7628,
+      "eval_samples_per_second": 5.573,
+      "eval_wer": 0.4252577319587629,
+      "step": 27900
+    },
+    {
+      "epoch": 7.78,
+      "learning_rate": 9.844948316105369e-05,
+      "loss": 0.1679,
+      "step": 28000
+    },
+    {
+      "epoch": 7.78,
+      "eval_loss": 0.7208133339881897,
+      "eval_runtime": 17.4884,
+      "eval_samples_per_second": 5.661,
+      "eval_wer": 0.4484536082474227,
+      "step": 28000
+    },
+    {
+      "epoch": 7.81,
+      "learning_rate": 9.84439257530288e-05,
+      "loss": 0.1646,
+      "step": 28100
+    },
+    {
+      "epoch": 7.81,
+      "eval_loss": 0.7599995136260986,
+      "eval_runtime": 17.4136,
+      "eval_samples_per_second": 5.685,
+      "eval_wer": 0.45103092783505155,
+      "step": 28100
+    },
+    {
+      "epoch": 7.84,
+      "learning_rate": 9.84383683450039e-05,
+      "loss": 0.1781,
+      "step": 28200
+    },
+    {
+      "epoch": 7.84,
+      "eval_loss": 0.7319245934486389,
+      "eval_runtime": 17.5926,
+      "eval_samples_per_second": 5.627,
+      "eval_wer": 0.4265463917525773,
+      "step": 28200
+    },
+    {
+      "epoch": 7.86,
+      "learning_rate": 9.843281093697901e-05,
+      "loss": 0.19,
+      "step": 28300
+    },
+    {
+      "epoch": 7.86,
+      "eval_loss": 0.7298288345336914,
+      "eval_runtime": 17.6158,
+      "eval_samples_per_second": 5.62,
+      "eval_wer": 0.43170103092783507,
+      "step": 28300
+    },
+    {
+      "epoch": 7.89,
+      "learning_rate": 9.84272535289541e-05,
+      "loss": 0.1792,
+      "step": 28400
+    },
+    {
+      "epoch": 7.89,
+      "eval_loss": 0.7053973078727722,
+      "eval_runtime": 17.3878,
+      "eval_samples_per_second": 5.694,
+      "eval_wer": 0.4381443298969072,
+      "step": 28400
+    },
+    {
+      "epoch": 7.92,
+      "learning_rate": 9.842169612092921e-05,
+      "loss": 0.1709,
+      "step": 28500
+    },
+    {
+      "epoch": 7.92,
+      "eval_loss": 0.8577209115028381,
+      "eval_runtime": 17.4901,
+      "eval_samples_per_second": 5.66,
+      "eval_wer": 0.4265463917525773,
+      "step": 28500
+    },
+    {
+      "epoch": 7.95,
+      "learning_rate": 9.841613871290431e-05,
+      "loss": 0.2003,
+      "step": 28600
+    },
+    {
+      "epoch": 7.95,
+      "eval_loss": 0.7145041227340698,
+      "eval_runtime": 17.3386,
+      "eval_samples_per_second": 5.71,
+      "eval_wer": 0.41881443298969073,
+      "step": 28600
+    },
+    {
+      "epoch": 7.97,
+      "learning_rate": 9.841058130487941e-05,
+      "loss": 0.1722,
+      "step": 28700
+    },
+    {
+      "epoch": 7.97,
+      "eval_loss": 0.7263432741165161,
+      "eval_runtime": 17.3569,
+      "eval_samples_per_second": 5.704,
+      "eval_wer": 0.4175257731958763,
+      "step": 28700
+    },
+    {
+      "epoch": 8.0,
+      "learning_rate": 9.840502389685451e-05,
+      "loss": 0.157,
+      "step": 28800
+    },
+    {
+      "epoch": 8.0,
+      "eval_loss": 0.7807848453521729,
+      "eval_runtime": 18.0075,
+      "eval_samples_per_second": 5.498,
+      "eval_wer": 0.41881443298969073,
+      "step": 28800
+    },
+    {
+      "epoch": 8.03,
+      "learning_rate": 9.839946648882961e-05,
+      "loss": 0.1728,
+      "step": 28900
+    },
+    {
+      "epoch": 8.03,
+      "eval_loss": 0.7600880861282349,
+      "eval_runtime": 17.8814,
+      "eval_samples_per_second": 5.536,
+      "eval_wer": 0.4329896907216495,
+      "step": 28900
+    },
+    {
+      "epoch": 8.06,
+      "learning_rate": 9.839390908080471e-05,
+      "loss": 0.1859,
+      "step": 29000
+    },
+    {
+      "epoch": 8.06,
+      "eval_loss": 0.7904886603355408,
+      "eval_runtime": 17.4997,
+      "eval_samples_per_second": 5.657,
+      "eval_wer": 0.41365979381443296,
+      "step": 29000
+    },
+    {
+      "epoch": 8.09,
+      "learning_rate": 9.838835167277983e-05,
+      "loss": 0.1759,
+      "step": 29100
+    },
+    {
+      "epoch": 8.09,
+      "eval_loss": 0.8925608396530151,
+      "eval_runtime": 17.4355,
+      "eval_samples_per_second": 5.678,
+      "eval_wer": 0.43556701030927836,
+      "step": 29100
+    },
+    {
+      "epoch": 8.11,
+      "learning_rate": 9.838279426475491e-05,
+      "loss": 0.1689,
+      "step": 29200
+    },
+    {
+      "epoch": 8.11,
+      "eval_loss": 0.7167654037475586,
+      "eval_runtime": 17.7948,
+      "eval_samples_per_second": 5.563,
+      "eval_wer": 0.43427835051546393,
+      "step": 29200
+    },
+    {
+      "epoch": 8.14,
+      "learning_rate": 9.837723685673003e-05,
+      "loss": 0.1708,
+      "step": 29300
+    },
+    {
+      "epoch": 8.14,
+      "eval_loss": 0.7209995985031128,
+      "eval_runtime": 17.5008,
+      "eval_samples_per_second": 5.657,
+      "eval_wer": 0.42912371134020616,
+      "step": 29300
+    },
+    {
+      "epoch": 8.17,
+      "learning_rate": 9.837167944870513e-05,
+      "loss": 0.1549,
+      "step": 29400
+    },
+    {
+      "epoch": 8.17,
+      "eval_loss": 0.7742015719413757,
+      "eval_runtime": 17.5354,
+      "eval_samples_per_second": 5.646,
+      "eval_wer": 0.4420103092783505,
+      "step": 29400
+    },
+    {
+      "epoch": 8.2,
+      "learning_rate": 9.836612204068023e-05,
+      "loss": 0.1605,
+      "step": 29500
+    },
+    {
+      "epoch": 8.2,
+      "eval_loss": 0.7685571312904358,
+      "eval_runtime": 17.5794,
+      "eval_samples_per_second": 5.632,
+      "eval_wer": 0.42783505154639173,
+      "step": 29500
+    },
+    {
+      "epoch": 8.22,
+      "learning_rate": 9.836056463265534e-05,
+      "loss": 0.1443,
+      "step": 29600
+    },
+    {
+      "epoch": 8.22,
+      "eval_loss": 0.8935885429382324,
+      "eval_runtime": 17.5375,
+      "eval_samples_per_second": 5.645,
+      "eval_wer": 0.43556701030927836,
+      "step": 29600
+    },
+    {
+      "epoch": 8.25,
+      "learning_rate": 9.835500722463043e-05,
+      "loss": 0.1492,
+      "step": 29700
+    },
+    {
+      "epoch": 8.25,
+      "eval_loss": 0.7365788221359253,
+      "eval_runtime": 17.3626,
+      "eval_samples_per_second": 5.702,
+      "eval_wer": 0.42912371134020616,
+      "step": 29700
+    },
+    {
+      "epoch": 8.28,
+      "learning_rate": 9.834944981660554e-05,
+      "loss": 0.1505,
+      "step": 29800
+    },
+    {
+      "epoch": 8.28,
+      "eval_loss": 0.7134827375411987,
+      "eval_runtime": 17.6924,
+      "eval_samples_per_second": 5.596,
+      "eval_wer": 0.42010309278350516,
+      "step": 29800
+    },
+    {
+      "epoch": 8.31,
+      "learning_rate": 9.834389240858064e-05,
+      "loss": 0.1662,
+      "step": 29900
+    },
+    {
+      "epoch": 8.31,
+      "eval_loss": 0.7517086863517761,
+      "eval_runtime": 17.6753,
+      "eval_samples_per_second": 5.601,
+      "eval_wer": 0.422680412371134,
+      "step": 29900
+    },
+    {
+      "epoch": 8.34,
+      "learning_rate": 9.833833500055574e-05,
+      "loss": 0.1557,
+      "step": 30000
+    },
+    {
+      "epoch": 8.34,
+      "eval_loss": 0.8542296290397644,
+      "eval_runtime": 17.6307,
+      "eval_samples_per_second": 5.615,
+      "eval_wer": 0.4368556701030928,
+      "step": 30000
+    },
+    {
+      "epoch": 8.36,
+      "learning_rate": 9.833277759253084e-05,
+      "loss": 0.1532,
+      "step": 30100
+    },
+    {
+      "epoch": 8.36,
+      "eval_loss": 0.7641463875770569,
+      "eval_runtime": 17.5001,
+      "eval_samples_per_second": 5.657,
+      "eval_wer": 0.42396907216494845,
+      "step": 30100
+    },
+    {
+      "epoch": 8.39,
+      "learning_rate": 9.832722018450596e-05,
+      "loss": 0.1636,
+      "step": 30200
+    },
+    {
+      "epoch": 8.39,
+      "eval_loss": 0.8011400699615479,
+      "eval_runtime": 17.3521,
+      "eval_samples_per_second": 5.705,
+      "eval_wer": 0.44458762886597936,
+      "step": 30200
+    },
+    {
+      "epoch": 8.42,
+      "learning_rate": 9.832166277648105e-05,
+      "loss": 0.1503,
+      "step": 30300
+    },
+    {
+      "epoch": 8.42,
+      "eval_loss": 0.7441977858543396,
+      "eval_runtime": 17.6318,
+      "eval_samples_per_second": 5.615,
+      "eval_wer": 0.42912371134020616,
+      "step": 30300
+    },
+    {
+      "epoch": 8.45,
+      "learning_rate": 9.831610536845616e-05,
+      "loss": 0.157,
+      "step": 30400
+    },
+    {
+      "epoch": 8.45,
+      "eval_loss": 0.7959823608398438,
+      "eval_runtime": 17.5355,
+      "eval_samples_per_second": 5.646,
+      "eval_wer": 0.43170103092783507,
+      "step": 30400
+    },
+    {
+      "epoch": 8.47,
+      "learning_rate": 9.831054796043126e-05,
+      "loss": 0.1934,
+      "step": 30500
+    },
+    {
+      "epoch": 8.47,
+      "eval_loss": 0.7789689898490906,
+      "eval_runtime": 17.4458,
+      "eval_samples_per_second": 5.675,
+      "eval_wer": 0.4497422680412371,
+      "step": 30500
+    },
+    {
+      "epoch": 8.5,
+      "learning_rate": 9.830499055240636e-05,
+      "loss": 0.1822,
+      "step": 30600
+    },
+    {
+      "epoch": 8.5,
+      "eval_loss": 0.6615021228790283,
+      "eval_runtime": 17.5369,
+      "eval_samples_per_second": 5.645,
+      "eval_wer": 0.4252577319587629,
+      "step": 30600
+    },
+    {
+      "epoch": 8.53,
+      "learning_rate": 9.829943314438148e-05,
+      "loss": 0.155,
+      "step": 30700
+    },
+    {
+      "epoch": 8.53,
+      "eval_loss": 0.7741401791572571,
+      "eval_runtime": 17.7325,
+      "eval_samples_per_second": 5.583,
+      "eval_wer": 0.43041237113402064,
+      "step": 30700
+    },
+    {
+      "epoch": 8.56,
+      "learning_rate": 9.829387573635656e-05,
+      "loss": 0.1626,
+      "step": 30800
+    },
+    {
+      "epoch": 8.56,
+      "eval_loss": 0.7389179468154907,
+      "eval_runtime": 17.6894,
+      "eval_samples_per_second": 5.597,
+      "eval_wer": 0.4381443298969072,
+      "step": 30800
+    },
+    {
+      "epoch": 8.59,
+      "learning_rate": 9.828831832833168e-05,
+      "loss": 0.1872,
+      "step": 30900
+    },
+    {
+      "epoch": 8.59,
+      "eval_loss": 0.6793811917304993,
+      "eval_runtime": 17.4718,
+      "eval_samples_per_second": 5.666,
+      "eval_wer": 0.4381443298969072,
+      "step": 30900
+    },
+    {
+      "epoch": 8.61,
+      "learning_rate": 9.828276092030678e-05,
+      "loss": 0.1521,
+      "step": 31000
+    },
+    {
+      "epoch": 8.61,
+      "eval_loss": 0.720003068447113,
+      "eval_runtime": 17.8272,
+      "eval_samples_per_second": 5.553,
+      "eval_wer": 0.42396907216494845,
+      "step": 31000
+    },
+    {
+      "epoch": 8.64,
+      "learning_rate": 9.827720351228188e-05,
+      "loss": 0.1776,
+      "step": 31100
+    },
+    {
+      "epoch": 8.64,
+      "eval_loss": 0.7978392243385315,
+      "eval_runtime": 17.9635,
+      "eval_samples_per_second": 5.511,
+      "eval_wer": 0.43427835051546393,
+      "step": 31100
+    },
+    {
+      "epoch": 8.67,
+      "learning_rate": 9.827164610425698e-05,
+      "loss": 0.1639,
+      "step": 31200
+    },
+    {
+      "epoch": 8.67,
+      "eval_loss": 0.6895660161972046,
+      "eval_runtime": 17.433,
+      "eval_samples_per_second": 5.679,
+      "eval_wer": 0.43170103092783507,
+      "step": 31200
+    },
+    {
+      "epoch": 8.7,
+      "learning_rate": 9.826608869623208e-05,
+      "loss": 0.1811,
+      "step": 31300
+    },
+    {
+      "epoch": 8.7,
+      "eval_loss": 0.6539278030395508,
+      "eval_runtime": 17.6188,
+      "eval_samples_per_second": 5.619,
+      "eval_wer": 0.4497422680412371,
+      "step": 31300
+    },
+    {
+      "epoch": 8.72,
+      "learning_rate": 9.826053128820718e-05,
+      "loss": 0.1783,
+      "step": 31400
+    },
+    {
+      "epoch": 8.72,
+      "eval_loss": 0.8889223337173462,
+      "eval_runtime": 17.4796,
+      "eval_samples_per_second": 5.664,
+      "eval_wer": 0.44458762886597936,
+      "step": 31400
+    },
+    {
+      "epoch": 8.75,
+      "learning_rate": 9.825497388018229e-05,
+      "loss": 0.1706,
+      "step": 31500
+    },
+    {
+      "epoch": 8.75,
+      "eval_loss": 0.7976836562156677,
+      "eval_runtime": 17.3817,
+      "eval_samples_per_second": 5.696,
+      "eval_wer": 0.42396907216494845,
+      "step": 31500
+    },
+    {
+      "epoch": 8.78,
+      "learning_rate": 9.824941647215738e-05,
+      "loss": 0.1666,
+      "step": 31600
+    },
+    {
+      "epoch": 8.78,
+      "eval_loss": 0.8039941787719727,
+      "eval_runtime": 17.7238,
+      "eval_samples_per_second": 5.586,
+      "eval_wer": 0.43556701030927836,
+      "step": 31600
+    },
+    {
+      "epoch": 8.81,
+      "learning_rate": 9.82438590641325e-05,
+      "loss": 0.1493,
+      "step": 31700
+    },
+    {
+      "epoch": 8.81,
+      "eval_loss": 0.9158120155334473,
+      "eval_runtime": 17.3559,
+      "eval_samples_per_second": 5.704,
+      "eval_wer": 0.4368556701030928,
+      "step": 31700
+    },
+    {
+      "epoch": 8.84,
+      "learning_rate": 9.82383016561076e-05,
+      "loss": 0.1558,
+      "step": 31800
+    },
+    {
+      "epoch": 8.84,
+      "eval_loss": 0.7148767113685608,
+      "eval_runtime": 17.5657,
+      "eval_samples_per_second": 5.636,
+      "eval_wer": 0.4213917525773196,
+      "step": 31800
+    },
+    {
+      "epoch": 8.86,
+      "learning_rate": 9.82327442480827e-05,
+      "loss": 0.1689,
+      "step": 31900
+    },
+    {
+      "epoch": 8.86,
+      "eval_loss": 0.6876005530357361,
+      "eval_runtime": 17.4155,
+      "eval_samples_per_second": 5.685,
+      "eval_wer": 0.43170103092783507,
+      "step": 31900
+    },
+    {
+      "epoch": 8.89,
+      "learning_rate": 9.822718684005781e-05,
+      "loss": 0.1657,
+      "step": 32000
+    },
+    {
+      "epoch": 8.89,
+      "eval_loss": 0.702477216720581,
+      "eval_runtime": 17.5273,
+      "eval_samples_per_second": 5.648,
+      "eval_wer": 0.44329896907216493,
+      "step": 32000
+    },
+    {
+      "epoch": 8.92,
+      "learning_rate": 9.82216294320329e-05,
+      "loss": 0.1887,
+      "step": 32100
+    },
+    {
+      "epoch": 8.92,
+      "eval_loss": 0.670263946056366,
+      "eval_runtime": 18.0974,
+      "eval_samples_per_second": 5.47,
+      "eval_wer": 0.4252577319587629,
+      "step": 32100
+    },
+    {
+      "epoch": 8.95,
+      "learning_rate": 9.821607202400801e-05,
+      "loss": 0.1743,
+      "step": 32200
+    },
+    {
+      "epoch": 8.95,
+      "eval_loss": 0.6662067174911499,
+      "eval_runtime": 17.5027,
+      "eval_samples_per_second": 5.656,
+      "eval_wer": 0.44072164948453607,
+      "step": 32200
+    },
+    {
+      "epoch": 8.97,
+      "learning_rate": 9.821051461598311e-05,
+      "loss": 0.1697,
+      "step": 32300
+    },
+    {
+      "epoch": 8.97,
+      "eval_loss": 0.6950443983078003,
+      "eval_runtime": 17.4463,
+      "eval_samples_per_second": 5.675,
+      "eval_wer": 0.42783505154639173,
+      "step": 32300
+    },
+    {
+      "epoch": 9.0,
+      "learning_rate": 9.820495720795821e-05,
+      "loss": 0.1565,
+      "step": 32400
+    },
+    {
+      "epoch": 9.0,
+      "eval_loss": 0.682771623134613,
+      "eval_runtime": 17.574,
+      "eval_samples_per_second": 5.633,
+      "eval_wer": 0.422680412371134,
+      "step": 32400
+    },
+    {
+      "epoch": 9.03,
+      "learning_rate": 9.819939979993331e-05,
+      "loss": 0.154,
+      "step": 32500
+    },
+    {
+      "epoch": 9.03,
+      "eval_loss": 0.7133552432060242,
+      "eval_runtime": 17.5006,
+      "eval_samples_per_second": 5.657,
+      "eval_wer": 0.4329896907216495,
+      "step": 32500
+    },
+    {
+      "epoch": 9.06,
+      "learning_rate": 9.819384239190843e-05,
+      "loss": 0.159,
+      "step": 32600
+    },
+    {
+      "epoch": 9.06,
+      "eval_loss": 0.7330553531646729,
+      "eval_runtime": 17.2801,
+      "eval_samples_per_second": 5.729,
+      "eval_wer": 0.4420103092783505,
+      "step": 32600
+    },
+    {
+      "epoch": 9.09,
+      "learning_rate": 9.818828498388351e-05,
+      "loss": 0.1585,
+      "step": 32700
+    },
+    {
+      "epoch": 9.09,
+      "eval_loss": 0.696050763130188,
+      "eval_runtime": 17.5298,
+      "eval_samples_per_second": 5.648,
+      "eval_wer": 0.4213917525773196,
+      "step": 32700
+    },
+    {
+      "epoch": 9.11,
+      "learning_rate": 9.818272757585863e-05,
+      "loss": 0.1595,
+      "step": 32800
+    },
+    {
+      "epoch": 9.11,
+      "eval_loss": 0.7466849088668823,
+      "eval_runtime": 17.3402,
+      "eval_samples_per_second": 5.709,
+      "eval_wer": 0.42783505154639173,
+      "step": 32800
+    },
+    {
+      "epoch": 9.14,
+      "learning_rate": 9.817717016783373e-05,
+      "loss": 0.1507,
+      "step": 32900
+    },
+    {
+      "epoch": 9.14,
+      "eval_loss": 0.7582687735557556,
+      "eval_runtime": 17.6824,
+      "eval_samples_per_second": 5.599,
+      "eval_wer": 0.43041237113402064,
+      "step": 32900
+    },
+    {
+      "epoch": 9.17,
+      "learning_rate": 9.817161275980883e-05,
+      "loss": 0.1586,
+      "step": 33000
+    },
+    {
+      "epoch": 9.17,
+      "eval_loss": 0.7383239269256592,
+      "eval_runtime": 17.5908,
+      "eval_samples_per_second": 5.628,
+      "eval_wer": 0.4265463917525773,
+      "step": 33000
+    },
+    {
+      "epoch": 9.2,
+      "learning_rate": 9.816605535178394e-05,
+      "loss": 0.1498,
+      "step": 33100
+    },
+    {
+      "epoch": 9.2,
+      "eval_loss": 0.9060437083244324,
+      "eval_runtime": 17.5173,
+      "eval_samples_per_second": 5.652,
+      "eval_wer": 0.43943298969072164,
+      "step": 33100
+    },
+    {
+      "epoch": 9.22,
+      "learning_rate": 9.816049794375903e-05,
+      "loss": 0.1415,
+      "step": 33200
+    },
+    {
+      "epoch": 9.22,
+      "eval_loss": 1.02446448802948,
+      "eval_runtime": 17.6833,
+      "eval_samples_per_second": 5.598,
+      "eval_wer": 0.46134020618556704,
+      "step": 33200
+    },
+    {
+      "epoch": 9.25,
+      "learning_rate": 9.815494053573414e-05,
+      "loss": 0.1465,
+      "step": 33300
+    },
+    {
+      "epoch": 9.25,
+      "eval_loss": 1.0142923593521118,
+      "eval_runtime": 17.6092,
+      "eval_samples_per_second": 5.622,
+      "eval_wer": 0.46005154639175255,
+      "step": 33300
+    },
+    {
+      "epoch": 9.28,
+      "learning_rate": 9.814938312770924e-05,
+      "loss": 0.155,
+      "step": 33400
+    },
+    {
+      "epoch": 9.28,
+      "eval_loss": 0.8718487620353699,
+      "eval_runtime": 17.6525,
+      "eval_samples_per_second": 5.608,
+      "eval_wer": 0.4536082474226804,
+      "step": 33400
+    },
+    {
+      "epoch": 9.31,
+      "learning_rate": 9.814382571968434e-05,
+      "loss": 0.1459,
+      "step": 33500
+    },
+    {
+      "epoch": 9.31,
+      "eval_loss": 0.6963269114494324,
+      "eval_runtime": 17.6886,
+      "eval_samples_per_second": 5.597,
+      "eval_wer": 0.42912371134020616,
+      "step": 33500
+    },
+    {
+      "epoch": 9.34,
+      "learning_rate": 9.813826831165944e-05,
+      "loss": 0.1598,
+      "step": 33600
+    },
+    {
+      "epoch": 9.34,
+      "eval_loss": 0.9281795024871826,
+      "eval_runtime": 17.3291,
+      "eval_samples_per_second": 5.713,
+      "eval_wer": 0.45103092783505155,
+      "step": 33600
+    },
+    {
+      "epoch": 9.36,
+      "learning_rate": 9.813271090363455e-05,
+      "loss": 0.1823,
+      "step": 33700
+    },
+    {
+      "epoch": 9.36,
+      "eval_loss": 0.9645785093307495,
+      "eval_runtime": 17.1301,
+      "eval_samples_per_second": 5.779,
+      "eval_wer": 0.4497422680412371,
+      "step": 33700
+    },
+    {
+      "epoch": 9.39,
+      "learning_rate": 9.812715349560965e-05,
+      "loss": 0.167,
+      "step": 33800
+    },
+    {
+      "epoch": 9.39,
+      "eval_loss": 0.9201110005378723,
+      "eval_runtime": 17.7043,
+      "eval_samples_per_second": 5.592,
+      "eval_wer": 0.4420103092783505,
+      "step": 33800
+    },
+    {
+      "epoch": 9.42,
+      "learning_rate": 9.812159608758476e-05,
+      "loss": 0.1381,
+      "step": 33900
+    },
+    {
+      "epoch": 9.42,
+      "eval_loss": 0.8799803853034973,
+      "eval_runtime": 17.5771,
+      "eval_samples_per_second": 5.632,
+      "eval_wer": 0.4420103092783505,
+      "step": 33900
+    },
+    {
+      "epoch": 9.45,
+      "learning_rate": 9.811603867955986e-05,
+      "loss": 0.1651,
+      "step": 34000
+    },
+    {
+      "epoch": 9.45,
+      "eval_loss": 0.8249523043632507,
+      "eval_runtime": 17.7685,
+      "eval_samples_per_second": 5.572,
+      "eval_wer": 0.44329896907216493,
+      "step": 34000
+    },
+    {
+      "epoch": 9.47,
+      "learning_rate": 9.811048127153496e-05,
+      "loss": 0.1583,
+      "step": 34100
+    },
+    {
+      "epoch": 9.47,
+      "eval_loss": 0.7931082248687744,
+      "eval_runtime": 17.569,
+      "eval_samples_per_second": 5.635,
+      "eval_wer": 0.42783505154639173,
+      "step": 34100
+    },
+    {
+      "epoch": 9.5,
+      "learning_rate": 9.810492386351006e-05,
+      "loss": 0.1443,
+      "step": 34200
+    },
+    {
+      "epoch": 9.5,
+      "eval_loss": 0.9122607111930847,
+      "eval_runtime": 17.381,
+      "eval_samples_per_second": 5.696,
+      "eval_wer": 0.44072164948453607,
+      "step": 34200
+    },
+    {
+      "epoch": 9.53,
+      "learning_rate": 9.809936645548516e-05,
+      "loss": 0.1597,
+      "step": 34300
+    },
+    {
+      "epoch": 9.53,
+      "eval_loss": 0.7589895725250244,
+      "eval_runtime": 17.8656,
+      "eval_samples_per_second": 5.541,
+      "eval_wer": 0.4162371134020619,
+      "step": 34300
+    },
+    {
+      "epoch": 9.56,
+      "learning_rate": 9.809380904746028e-05,
+      "loss": 0.1477,
+      "step": 34400
+    },
+    {
+      "epoch": 9.56,
+      "eval_loss": 0.7940624356269836,
+      "eval_runtime": 17.883,
+      "eval_samples_per_second": 5.536,
+      "eval_wer": 0.4420103092783505,
+      "step": 34400
+    },
+    {
+      "epoch": 9.59,
+      "learning_rate": 9.808825163943536e-05,
+      "loss": 0.1602,
+      "step": 34500
+    },
+    {
+      "epoch": 9.59,
+      "eval_loss": 0.8617969155311584,
+      "eval_runtime": 17.5334,
+      "eval_samples_per_second": 5.646,
+      "eval_wer": 0.4381443298969072,
+      "step": 34500
+    },
+    {
+      "epoch": 9.61,
+      "learning_rate": 9.808269423141048e-05,
+      "loss": 0.1788,
+      "step": 34600
+    },
+    {
+      "epoch": 9.61,
+      "eval_loss": 0.8317437767982483,
+      "eval_runtime": 17.4731,
+      "eval_samples_per_second": 5.666,
+      "eval_wer": 0.4175257731958763,
+      "step": 34600
+    },
+    {
+      "epoch": 9.64,
+      "learning_rate": 9.807713682338558e-05,
+      "loss": 0.1574,
+      "step": 34700
+    },
+    {
+      "epoch": 9.64,
+      "eval_loss": 0.8786133527755737,
+      "eval_runtime": 17.691,
+      "eval_samples_per_second": 5.596,
+      "eval_wer": 0.4420103092783505,
+      "step": 34700
+    },
+    {
+      "epoch": 9.67,
+      "learning_rate": 9.807157941536068e-05,
+      "loss": 0.1572,
+      "step": 34800
+    },
+    {
+      "epoch": 9.67,
+      "eval_loss": 0.9068703055381775,
+      "eval_runtime": 17.5185,
+      "eval_samples_per_second": 5.651,
+      "eval_wer": 0.46005154639175255,
+      "step": 34800
+    },
+    {
+      "epoch": 9.7,
+      "learning_rate": 9.806602200733578e-05,
+      "loss": 0.1644,
+      "step": 34900
+    },
+    {
+      "epoch": 9.7,
+      "eval_loss": 0.7928957939147949,
+      "eval_runtime": 17.6768,
+      "eval_samples_per_second": 5.601,
+      "eval_wer": 0.452319587628866,
+      "step": 34900
+    },
+    {
+      "epoch": 9.72,
+      "learning_rate": 9.806046459931088e-05,
+      "loss": 0.1592,
+      "step": 35000
+    },
+    {
+      "epoch": 9.72,
+      "eval_loss": 0.7851370573043823,
+      "eval_runtime": 17.4073,
+      "eval_samples_per_second": 5.687,
+      "eval_wer": 0.42783505154639173,
+      "step": 35000
+    },
+    {
+      "epoch": 9.75,
+      "learning_rate": 9.8054907191286e-05,
+      "loss": 0.1419,
+      "step": 35100
+    },
+    {
+      "epoch": 9.75,
+      "eval_loss": 0.7190160155296326,
+      "eval_runtime": 17.1768,
+      "eval_samples_per_second": 5.764,
+      "eval_wer": 0.42783505154639173,
+      "step": 35100
+    },
+    {
+      "epoch": 9.78,
+      "learning_rate": 9.80493497832611e-05,
+      "loss": 0.1658,
+      "step": 35200
+    },
+    {
+      "epoch": 9.78,
+      "eval_loss": 0.7918796539306641,
+      "eval_runtime": 17.3718,
+      "eval_samples_per_second": 5.699,
+      "eval_wer": 0.422680412371134,
+      "step": 35200
+    },
+    {
+      "epoch": 9.81,
+      "learning_rate": 9.80437923752362e-05,
+      "loss": 0.1711,
+      "step": 35300
+    },
+    {
+      "epoch": 9.81,
+      "eval_loss": 0.8889212012290955,
+      "eval_runtime": 17.1156,
+      "eval_samples_per_second": 5.784,
+      "eval_wer": 0.4652061855670103,
+      "step": 35300
+    },
+    {
+      "epoch": 9.84,
+      "learning_rate": 9.80382349672113e-05,
+      "loss": 0.162,
+      "step": 35400
+    },
+    {
+      "epoch": 9.84,
+      "eval_loss": 0.9033122062683105,
+      "eval_runtime": 17.4865,
+      "eval_samples_per_second": 5.662,
+      "eval_wer": 0.43943298969072164,
+      "step": 35400
+    },
+    {
+      "epoch": 9.86,
+      "learning_rate": 9.803267755918641e-05,
+      "loss": 0.1599,
+      "step": 35500
+    },
+    {
+      "epoch": 9.86,
+      "eval_loss": 0.7897907495498657,
+      "eval_runtime": 17.0639,
+      "eval_samples_per_second": 5.802,
+      "eval_wer": 0.45103092783505155,
+      "step": 35500
+    },
+    {
+      "epoch": 9.89,
+      "learning_rate": 9.80271201511615e-05,
+      "loss": 0.136,
+      "step": 35600
+    },
+    {
+      "epoch": 9.89,
+      "eval_loss": 0.8795129060745239,
+      "eval_runtime": 18.3203,
+      "eval_samples_per_second": 5.404,
+      "eval_wer": 0.43943298969072164,
+      "step": 35600
+    },
+    {
+      "epoch": 9.92,
+      "learning_rate": 9.802156274313661e-05,
+      "loss": 0.1543,
+      "step": 35700
+    },
+    {
+      "epoch": 9.92,
+      "eval_loss": 0.940313994884491,
+      "eval_runtime": 17.5084,
+      "eval_samples_per_second": 5.654,
+      "eval_wer": 0.4497422680412371,
+      "step": 35700
+    },
+    {
+      "epoch": 9.95,
+      "learning_rate": 9.801600533511171e-05,
+      "loss": 0.1531,
+      "step": 35800
+    },
+    {
+      "epoch": 9.95,
+      "eval_loss": 0.9860208034515381,
+      "eval_runtime": 17.6284,
+      "eval_samples_per_second": 5.616,
+      "eval_wer": 0.4484536082474227,
+      "step": 35800
+    },
+    {
+      "epoch": 9.97,
+      "learning_rate": 9.801044792708681e-05,
+      "loss": 0.1537,
+      "step": 35900
+    },
+    {
+      "epoch": 9.97,
+      "eval_loss": 0.910431444644928,
+      "eval_runtime": 17.5953,
+      "eval_samples_per_second": 5.626,
+      "eval_wer": 0.45489690721649484,
+      "step": 35900
+    },
+    {
+      "epoch": 10.0,
+      "learning_rate": 9.800489051906191e-05,
+      "loss": 0.1525,
+      "step": 36000
+    },
+    {
+      "epoch": 10.0,
+      "eval_loss": 0.9185177087783813,
+      "eval_runtime": 17.5837,
+      "eval_samples_per_second": 5.63,
+      "eval_wer": 0.4381443298969072,
+      "step": 36000
+    },
+    {
+      "epoch": 10.03,
+      "learning_rate": 9.799933311103701e-05,
+      "loss": 0.1323,
+      "step": 36100
+    },
+    {
+      "epoch": 10.03,
+      "eval_loss": 0.9509208798408508,
+      "eval_runtime": 17.3541,
+      "eval_samples_per_second": 5.705,
+      "eval_wer": 0.4265463917525773,
+      "step": 36100
+    },
+    {
+      "epoch": 10.06,
+      "learning_rate": 9.799377570301213e-05,
+      "loss": 0.1411,
+      "step": 36200
+    },
+    {
+      "epoch": 10.06,
+      "eval_loss": 0.9488706588745117,
+      "eval_runtime": 17.624,
+      "eval_samples_per_second": 5.617,
+      "eval_wer": 0.4420103092783505,
+      "step": 36200
+    },
+    {
+      "epoch": 10.09,
+      "learning_rate": 9.798821829498723e-05,
+      "loss": 0.1376,
+      "step": 36300
+    },
+    {
+      "epoch": 10.09,
+      "eval_loss": 0.9858886003494263,
+      "eval_runtime": 17.6205,
+      "eval_samples_per_second": 5.618,
+      "eval_wer": 0.47164948453608246,
+      "step": 36300
+    },
+    {
+      "epoch": 10.11,
+      "learning_rate": 9.798266088696233e-05,
+      "loss": 0.1711,
+      "step": 36400
+    },
+    {
+      "epoch": 10.11,
+      "eval_loss": 0.9317906498908997,
+      "eval_runtime": 17.6122,
+      "eval_samples_per_second": 5.621,
+      "eval_wer": 0.4587628865979381,
+      "step": 36400
+    },
+    {
+      "epoch": 10.14,
+      "learning_rate": 9.797710347893743e-05,
+      "loss": 0.138,
+      "step": 36500
+    },
+    {
+      "epoch": 10.14,
+      "eval_loss": 1.0478296279907227,
+      "eval_runtime": 17.614,
+      "eval_samples_per_second": 5.621,
+      "eval_wer": 0.45618556701030927,
+      "step": 36500
+    },
+    {
+      "epoch": 10.17,
+      "learning_rate": 9.797154607091253e-05,
+      "loss": 0.1456,
+      "step": 36600
+    },
+    {
+      "epoch": 10.17,
+      "eval_loss": 0.9326837062835693,
+      "eval_runtime": 17.7144,
+      "eval_samples_per_second": 5.589,
+      "eval_wer": 0.44587628865979384,
+      "step": 36600
+    },
+    {
+      "epoch": 10.2,
+      "learning_rate": 9.796598866288763e-05,
+      "loss": 0.1486,
+      "step": 36700
+    },
+    {
+      "epoch": 10.2,
+      "eval_loss": 0.9639229774475098,
+      "eval_runtime": 17.5424,
+      "eval_samples_per_second": 5.643,
+      "eval_wer": 0.44072164948453607,
+      "step": 36700
+    },
+    {
+      "epoch": 10.23,
+      "learning_rate": 9.796043125486274e-05,
+      "loss": 0.1385,
+      "step": 36800
+    },
+    {
+      "epoch": 10.23,
+      "eval_loss": 0.9051303863525391,
+      "eval_runtime": 17.5669,
+      "eval_samples_per_second": 5.636,
+      "eval_wer": 0.4329896907216495,
+      "step": 36800
+    },
+    {
+      "epoch": 10.25,
+      "learning_rate": 9.795487384683783e-05,
+      "loss": 0.1397,
+      "step": 36900
+    },
+    {
+      "epoch": 10.25,
+      "eval_loss": 0.9713443517684937,
+      "eval_runtime": 17.6453,
+      "eval_samples_per_second": 5.611,
+      "eval_wer": 0.4497422680412371,
+      "step": 36900
+    },
+    {
+      "epoch": 10.28,
+      "learning_rate": 9.794931643881294e-05,
+      "loss": 0.155,
+      "step": 37000
+    },
+    {
+      "epoch": 10.28,
+      "eval_loss": 0.9461785554885864,
+      "eval_runtime": 17.6547,
+      "eval_samples_per_second": 5.608,
+      "eval_wer": 0.44587628865979384,
+      "step": 37000
+    },
+    {
+      "epoch": 10.31,
+      "learning_rate": 9.794375903078805e-05,
+      "loss": 0.1484,
+      "step": 37100
+    },
+    {
+      "epoch": 10.31,
+      "eval_loss": 1.0771350860595703,
+      "eval_runtime": 17.4288,
+      "eval_samples_per_second": 5.68,
+      "eval_wer": 0.43943298969072164,
+      "step": 37100
+    },
+    {
+      "epoch": 10.34,
+      "learning_rate": 9.793820162276315e-05,
+      "loss": 0.134,
+      "step": 37200
+    },
+    {
+      "epoch": 10.34,
+      "eval_loss": 0.9590903520584106,
+      "eval_runtime": 17.6468,
+      "eval_samples_per_second": 5.61,
+      "eval_wer": 0.4265463917525773,
+      "step": 37200
+    },
+    {
+      "epoch": 10.36,
+      "learning_rate": 9.793264421473826e-05,
+      "loss": 0.1372,
+      "step": 37300
+    },
+    {
+      "epoch": 10.36,
+      "eval_loss": 0.9968072772026062,
+      "eval_runtime": 17.5496,
+      "eval_samples_per_second": 5.641,
+      "eval_wer": 0.43427835051546393,
+      "step": 37300
+    },
+    {
+      "epoch": 10.39,
+      "learning_rate": 9.792708680671335e-05,
+      "loss": 0.1412,
+      "step": 37400
+    },
+    {
+      "epoch": 10.39,
+      "eval_loss": 0.921562910079956,
+      "eval_runtime": 17.5184,
+      "eval_samples_per_second": 5.651,
+      "eval_wer": 0.4574742268041237,
+      "step": 37400
+    },
+    {
+      "epoch": 10.42,
+      "learning_rate": 9.792152939868846e-05,
+      "loss": 0.134,
+      "step": 37500
+    },
+    {
+      "epoch": 10.42,
+      "eval_loss": 0.9161884784698486,
+      "eval_runtime": 17.7692,
+      "eval_samples_per_second": 5.571,
+      "eval_wer": 0.46649484536082475,
+      "step": 37500
+    },
+    {
+      "epoch": 10.45,
+      "learning_rate": 9.791597199066356e-05,
+      "loss": 0.1467,
+      "step": 37600
+    },
+    {
+      "epoch": 10.45,
+      "eval_loss": 0.9519623517990112,
+      "eval_runtime": 17.501,
+      "eval_samples_per_second": 5.657,
+      "eval_wer": 0.44458762886597936,
+      "step": 37600
+    },
+    {
+      "epoch": 10.48,
+      "learning_rate": 9.791041458263866e-05,
+      "loss": 0.1297,
+      "step": 37700
+    },
+    {
+      "epoch": 10.48,
+      "eval_loss": 0.9233236312866211,
+      "eval_runtime": 17.6622,
+      "eval_samples_per_second": 5.605,
+      "eval_wer": 0.44458762886597936,
+      "step": 37700
+    },
+    {
+      "epoch": 10.5,
+      "learning_rate": 9.790485717461376e-05,
+      "loss": 0.1428,
+      "step": 37800
+    },
+    {
+      "epoch": 10.5,
+      "eval_loss": 0.9528768658638,
+      "eval_runtime": 17.5775,
+      "eval_samples_per_second": 5.632,
+      "eval_wer": 0.44329896907216493,
+      "step": 37800
+    },
+    {
+      "epoch": 10.53,
+      "learning_rate": 9.789929976658888e-05,
+      "loss": 0.1499,
+      "step": 37900
+    },
+    {
+      "epoch": 10.53,
+      "eval_loss": 0.8795115351676941,
+      "eval_runtime": 17.4947,
+      "eval_samples_per_second": 5.659,
+      "eval_wer": 0.45489690721649484,
+      "step": 37900
+    },
+    {
+      "epoch": 10.56,
+      "learning_rate": 9.789374235856396e-05,
+      "loss": 0.1492,
+      "step": 38000
+    },
+    {
+      "epoch": 10.56,
+      "eval_loss": 0.9267066121101379,
+      "eval_runtime": 17.6159,
+      "eval_samples_per_second": 5.62,
+      "eval_wer": 0.45103092783505155,
+      "step": 38000
+    },
+    {
+      "epoch": 10.59,
+      "learning_rate": 9.788818495053908e-05,
+      "loss": 0.1476,
+      "step": 38100
+    },
+    {
+      "epoch": 10.59,
+      "eval_loss": 0.9196488261222839,
+      "eval_runtime": 17.564,
+      "eval_samples_per_second": 5.637,
+      "eval_wer": 0.4690721649484536,
+      "step": 38100
+    },
+    {
+      "epoch": 10.61,
+      "learning_rate": 9.788262754251416e-05,
+      "loss": 0.1398,
+      "step": 38200
+    },
+    {
+      "epoch": 10.61,
+      "eval_loss": 0.928520143032074,
+      "eval_runtime": 17.5698,
+      "eval_samples_per_second": 5.635,
+      "eval_wer": 0.46005154639175255,
+      "step": 38200
+    },
+    {
+      "epoch": 10.64,
+      "learning_rate": 9.787707013448928e-05,
+      "loss": 0.1512,
+      "step": 38300
+    },
+    {
+      "epoch": 10.64,
+      "eval_loss": 0.840891420841217,
+      "eval_runtime": 17.5719,
+      "eval_samples_per_second": 5.634,
+      "eval_wer": 0.43556701030927836,
+      "step": 38300
+    },
+    {
+      "epoch": 10.67,
+      "learning_rate": 9.787151272646439e-05,
+      "loss": 0.1627,
+      "step": 38400
+    },
+    {
+      "epoch": 10.67,
+      "eval_loss": 0.8385064601898193,
+      "eval_runtime": 17.6499,
+      "eval_samples_per_second": 5.609,
+      "eval_wer": 0.44716494845360827,
+      "step": 38400
+    },
+    {
+      "epoch": 10.7,
+      "learning_rate": 9.786595531843948e-05,
+      "loss": 0.1418,
+      "step": 38500
+    },
+    {
+      "epoch": 10.7,
+      "eval_loss": 0.8606293797492981,
+      "eval_runtime": 17.9059,
+      "eval_samples_per_second": 5.529,
+      "eval_wer": 0.44329896907216493,
+      "step": 38500
+    },
+    {
+      "epoch": 10.73,
+      "learning_rate": 9.78603979104146e-05,
+      "loss": 0.1587,
+      "step": 38600
+    },
+    {
+      "epoch": 10.73,
+      "eval_loss": 0.776897132396698,
+      "eval_runtime": 17.5532,
+      "eval_samples_per_second": 5.64,
+      "eval_wer": 0.45103092783505155,
+      "step": 38600
+    },
+    {
+      "epoch": 10.75,
+      "learning_rate": 9.78548405023897e-05,
+      "loss": 0.1416,
+      "step": 38700
+    },
+    {
+      "epoch": 10.75,
+      "eval_loss": 0.7967720627784729,
+      "eval_runtime": 17.409,
+      "eval_samples_per_second": 5.687,
+      "eval_wer": 0.4420103092783505,
+      "step": 38700
+    },
+    {
+      "epoch": 10.78,
+      "learning_rate": 9.78492830943648e-05,
+      "loss": 0.1651,
+      "step": 38800
+    },
+    {
+      "epoch": 10.78,
+      "eval_loss": 0.8126049637794495,
+      "eval_runtime": 17.6952,
+      "eval_samples_per_second": 5.595,
+      "eval_wer": 0.4368556701030928,
+      "step": 38800
+    },
+    {
+      "epoch": 10.81,
+      "learning_rate": 9.78437256863399e-05,
+      "loss": 0.1439,
+      "step": 38900
+    },
+    {
+      "epoch": 10.81,
+      "eval_loss": 0.8487725853919983,
+      "eval_runtime": 17.6223,
+      "eval_samples_per_second": 5.618,
+      "eval_wer": 0.4252577319587629,
+      "step": 38900
+    },
+    {
+      "epoch": 10.84,
+      "learning_rate": 9.7838168278315e-05,
+      "loss": 0.129,
+      "step": 39000
+    },
+    {
+      "epoch": 10.84,
+      "eval_loss": 0.9903653860092163,
+      "eval_runtime": 17.4638,
+      "eval_samples_per_second": 5.669,
+      "eval_wer": 0.43556701030927836,
+      "step": 39000
+    },
+    {
+      "epoch": 10.86,
+      "learning_rate": 9.78326108702901e-05,
+      "loss": 0.1299,
+      "step": 39100
+    },
+    {
+      "epoch": 10.86,
+      "eval_loss": 0.9441640973091125,
+      "eval_runtime": 17.6194,
+      "eval_samples_per_second": 5.619,
+      "eval_wer": 0.4381443298969072,
+      "step": 39100
+    },
+    {
+      "epoch": 10.89,
+      "learning_rate": 9.782705346226521e-05,
+      "loss": 0.1521,
+      "step": 39200
+    },
+    {
+      "epoch": 10.89,
+      "eval_loss": 0.9177586436271667,
+      "eval_runtime": 17.6014,
+      "eval_samples_per_second": 5.625,
+      "eval_wer": 0.452319587628866,
+      "step": 39200
+    },
+    {
+      "epoch": 10.92,
+      "learning_rate": 9.78214960542403e-05,
+      "loss": 0.1651,
+      "step": 39300
+    },
+    {
+      "epoch": 10.92,
+      "eval_loss": 0.7855709195137024,
+      "eval_runtime": 17.36,
+      "eval_samples_per_second": 5.703,
+      "eval_wer": 0.4484536082474227,
+      "step": 39300
+    },
+    {
+      "epoch": 10.95,
+      "learning_rate": 9.781593864621541e-05,
+      "loss": 0.16,
+      "step": 39400
+    },
+    {
+      "epoch": 10.95,
+      "eval_loss": 0.8589774370193481,
+      "eval_runtime": 17.4597,
+      "eval_samples_per_second": 5.67,
+      "eval_wer": 0.44587628865979384,
+      "step": 39400
+    },
+    {
+      "epoch": 10.98,
+      "learning_rate": 9.781038123819051e-05,
+      "loss": 0.1551,
+      "step": 39500
+    },
+    {
+      "epoch": 10.98,
+      "eval_loss": 0.8536927103996277,
+      "eval_runtime": 17.5722,
+      "eval_samples_per_second": 5.634,
+      "eval_wer": 0.4574742268041237,
+      "step": 39500
+    },
+    {
+      "epoch": 11.0,
+      "learning_rate": 9.780482383016561e-05,
+      "loss": 0.1656,
+      "step": 39600
+    },
+    {
+      "epoch": 11.0,
+      "eval_loss": 0.7709180116653442,
+      "eval_runtime": 17.5544,
+      "eval_samples_per_second": 5.64,
+      "eval_wer": 0.45103092783505155,
+      "step": 39600
+    },
+    {
+      "epoch": 11.03,
+      "learning_rate": 9.779926642214073e-05,
+      "loss": 0.1348,
+      "step": 39700
+    },
+    {
+      "epoch": 11.03,
+      "eval_loss": 0.8829686045646667,
+      "eval_runtime": 17.4954,
+      "eval_samples_per_second": 5.659,
+      "eval_wer": 0.43556701030927836,
+      "step": 39700
+    },
+    {
+      "epoch": 11.06,
+      "learning_rate": 9.779370901411581e-05,
+      "loss": 0.1344,
+      "step": 39800
+    },
+    {
+      "epoch": 11.06,
+      "eval_loss": 0.8350869417190552,
+      "eval_runtime": 17.6252,
+      "eval_samples_per_second": 5.617,
+      "eval_wer": 0.44716494845360827,
+      "step": 39800
+    },
+    {
+      "epoch": 11.09,
+      "learning_rate": 9.778815160609093e-05,
+      "loss": 0.1484,
+      "step": 39900
+    },
+    {
+      "epoch": 11.09,
+      "eval_loss": 0.8020666241645813,
+      "eval_runtime": 17.8124,
+      "eval_samples_per_second": 5.558,
+      "eval_wer": 0.4484536082474227,
+      "step": 39900
+    },
+    {
+      "epoch": 11.11,
+      "learning_rate": 9.778259419806603e-05,
+      "loss": 0.1294,
+      "step": 40000
+    },
+    {
+      "epoch": 11.11,
+      "eval_loss": 0.8609241247177124,
+      "eval_runtime": 17.6098,
+      "eval_samples_per_second": 5.622,
+      "eval_wer": 0.44329896907216493,
+      "step": 40000
+    },
+    {
+      "epoch": 11.14,
+      "learning_rate": 9.777703679004113e-05,
+      "loss": 0.1339,
+      "step": 40100
+    },
+    {
+      "epoch": 11.14,
+      "eval_loss": 0.9225992560386658,
+      "eval_runtime": 18.105,
+      "eval_samples_per_second": 5.468,
+      "eval_wer": 0.4652061855670103,
+      "step": 40100
+    },
+    {
+      "epoch": 11.17,
+      "learning_rate": 9.777147938201623e-05,
+      "loss": 0.1405,
+      "step": 40200
+    },
+    {
+      "epoch": 11.17,
+      "eval_loss": 0.8729308247566223,
+      "eval_runtime": 17.978,
+      "eval_samples_per_second": 5.507,
+      "eval_wer": 0.4484536082474227,
+      "step": 40200
+    },
+    {
+      "epoch": 11.2,
+      "learning_rate": 9.776592197399133e-05,
+      "loss": 0.1463,
+      "step": 40300
+    },
+    {
+      "epoch": 11.2,
+      "eval_loss": 0.8339123725891113,
+      "eval_runtime": 17.6199,
+      "eval_samples_per_second": 5.619,
+      "eval_wer": 0.45103092783505155,
+      "step": 40300
+    },
+    {
+      "epoch": 11.23,
+      "learning_rate": 9.776036456596643e-05,
+      "loss": 0.1476,
+      "step": 40400
+    },
+    {
+      "epoch": 11.23,
+      "eval_loss": 0.8812620639801025,
+      "eval_runtime": 17.7803,
+      "eval_samples_per_second": 5.568,
+      "eval_wer": 0.45618556701030927,
+      "step": 40400
+    },
+    {
+      "epoch": 11.25,
+      "learning_rate": 9.775480715794154e-05,
+      "loss": 0.1436,
+      "step": 40500
+    },
+    {
+      "epoch": 11.25,
+      "eval_loss": 0.8837606906890869,
+      "eval_runtime": 17.5897,
+      "eval_samples_per_second": 5.628,
+      "eval_wer": 0.45618556701030927,
+      "step": 40500
+    },
+    {
+      "epoch": 11.28,
+      "learning_rate": 9.774924974991663e-05,
+      "loss": 0.1536,
+      "step": 40600
+    },
+    {
+      "epoch": 11.28,
+      "eval_loss": 0.8110833168029785,
+      "eval_runtime": 17.6617,
+      "eval_samples_per_second": 5.605,
+      "eval_wer": 0.47680412371134023,
+      "step": 40600
+    },
+    {
+      "epoch": 11.31,
+      "learning_rate": 9.774369234189175e-05,
+      "loss": 0.1281,
+      "step": 40700
+    },
+    {
+      "epoch": 11.31,
+      "eval_loss": 0.9150132536888123,
+      "eval_runtime": 17.6707,
+      "eval_samples_per_second": 5.602,
+      "eval_wer": 0.4639175257731959,
+      "step": 40700
+    },
+    {
+      "epoch": 11.34,
+      "learning_rate": 9.773813493386686e-05,
+      "loss": 0.1399,
+      "step": 40800
+    },
+    {
+      "epoch": 11.34,
+      "eval_loss": 0.8512896299362183,
+      "eval_runtime": 17.4246,
+      "eval_samples_per_second": 5.682,
+      "eval_wer": 0.46005154639175255,
+      "step": 40800
+    },
+    {
+      "epoch": 11.36,
+      "learning_rate": 9.773257752584195e-05,
+      "loss": 0.133,
+      "step": 40900
+    },
+    {
+      "epoch": 11.36,
+      "eval_loss": 0.8384860754013062,
+      "eval_runtime": 17.5781,
+      "eval_samples_per_second": 5.632,
+      "eval_wer": 0.45489690721649484,
+      "step": 40900
+    },
+    {
+      "epoch": 11.39,
+      "learning_rate": 9.772702011781706e-05,
+      "loss": 0.1477,
+      "step": 41000
+    },
+    {
+      "epoch": 11.39,
+      "eval_loss": 0.8938087821006775,
+      "eval_runtime": 17.8894,
+      "eval_samples_per_second": 5.534,
+      "eval_wer": 0.4484536082474227,
+      "step": 41000
+    },
+    {
+      "epoch": 11.42,
+      "learning_rate": 9.772146270979216e-05,
+      "loss": 0.1478,
+      "step": 41100
+    },
+    {
+      "epoch": 11.42,
+      "eval_loss": 0.8160294890403748,
+      "eval_runtime": 17.6392,
+      "eval_samples_per_second": 5.613,
+      "eval_wer": 0.46005154639175255,
+      "step": 41100
+    },
+    {
+      "epoch": 11.45,
+      "learning_rate": 9.771590530176726e-05,
+      "loss": 0.1505,
+      "step": 41200
+    },
+    {
+      "epoch": 11.45,
+      "eval_loss": 0.7812536358833313,
+      "eval_runtime": 17.5733,
+      "eval_samples_per_second": 5.634,
+      "eval_wer": 0.46262886597938147,
+      "step": 41200
+    },
+    {
+      "epoch": 11.48,
+      "learning_rate": 9.771034789374236e-05,
+      "loss": 0.1545,
+      "step": 41300
+    },
+    {
+      "epoch": 11.48,
+      "eval_loss": 0.8860114216804504,
+      "eval_runtime": 17.7418,
+      "eval_samples_per_second": 5.58,
+      "eval_wer": 0.47036082474226804,
+      "step": 41300
+    },
+    {
+      "epoch": 11.5,
+      "learning_rate": 9.770479048571746e-05,
+      "loss": 0.1266,
+      "step": 41400
+    },
+    {
+      "epoch": 11.5,
+      "eval_loss": 0.8299418091773987,
+      "eval_runtime": 17.8271,
+      "eval_samples_per_second": 5.553,
+      "eval_wer": 0.4587628865979381,
+      "step": 41400
+    },
+    {
+      "epoch": 11.53,
+      "learning_rate": 9.769923307769256e-05,
+      "loss": 0.1405,
+      "step": 41500
+    },
+    {
+      "epoch": 11.53,
+      "eval_loss": 0.7701277136802673,
+      "eval_runtime": 17.5493,
+      "eval_samples_per_second": 5.641,
+      "eval_wer": 0.44329896907216493,
+      "step": 41500
+    },
+    {
+      "epoch": 11.56,
+      "learning_rate": 9.769367566966768e-05,
+      "loss": 0.1499,
+      "step": 41600
+    },
+    {
+      "epoch": 11.56,
+      "eval_loss": 0.7627600431442261,
+      "eval_runtime": 17.7427,
+      "eval_samples_per_second": 5.58,
+      "eval_wer": 0.45103092783505155,
+      "step": 41600
+    },
+    {
+      "epoch": 11.59,
+      "learning_rate": 9.768811826164276e-05,
+      "loss": 0.1362,
+      "step": 41700
+    },
+    {
+      "epoch": 11.59,
+      "eval_loss": 0.8455436825752258,
+      "eval_runtime": 17.8944,
+      "eval_samples_per_second": 5.532,
+      "eval_wer": 0.44458762886597936,
+      "step": 41700
+    },
+    {
+      "epoch": 11.61,
+      "learning_rate": 9.768256085361788e-05,
+      "loss": 0.1539,
+      "step": 41800
+    },
+    {
+      "epoch": 11.61,
+      "eval_loss": 0.7182540893554688,
+      "eval_runtime": 17.7016,
+      "eval_samples_per_second": 5.593,
+      "eval_wer": 0.44072164948453607,
+      "step": 41800
+    },
+    {
+      "epoch": 11.64,
+      "learning_rate": 9.767700344559298e-05,
+      "loss": 0.1471,
+      "step": 41900
+    },
+    {
+      "epoch": 11.64,
+      "eval_loss": 0.6807606816291809,
+      "eval_runtime": 17.7241,
+      "eval_samples_per_second": 5.586,
+      "eval_wer": 0.42783505154639173,
+      "step": 41900
+    },
+    {
+      "epoch": 11.67,
+      "learning_rate": 9.767144603756808e-05,
+      "loss": 0.135,
+      "step": 42000
+    },
+    {
+      "epoch": 11.67,
+      "eval_loss": 0.8377964496612549,
+      "eval_runtime": 17.7805,
+      "eval_samples_per_second": 5.568,
+      "eval_wer": 0.43556701030927836,
+      "step": 42000
+    },
+    {
+      "epoch": 11.7,
+      "learning_rate": 9.76658886295432e-05,
+      "loss": 0.1406,
+      "step": 42100
+    },
+    {
+      "epoch": 11.7,
+      "eval_loss": 0.8310317397117615,
+      "eval_runtime": 17.8675,
+      "eval_samples_per_second": 5.541,
+      "eval_wer": 0.4536082474226804,
+      "step": 42100
+    },
+    {
+      "epoch": 11.73,
+      "learning_rate": 9.766033122151828e-05,
+      "loss": 0.1544,
+      "step": 42200
+    },
+    {
+      "epoch": 11.73,
+      "eval_loss": 0.8470320105552673,
+      "eval_runtime": 17.6266,
+      "eval_samples_per_second": 5.617,
+      "eval_wer": 0.45618556701030927,
+      "step": 42200
+    },
+    {
+      "epoch": 11.75,
+      "learning_rate": 9.76547738134934e-05,
+      "loss": 0.1445,
+      "step": 42300
+    },
+    {
+      "epoch": 11.75,
+      "eval_loss": 0.9373657703399658,
+      "eval_runtime": 17.6007,
+      "eval_samples_per_second": 5.625,
+      "eval_wer": 0.44458762886597936,
+      "step": 42300
+    },
+    {
+      "epoch": 11.78,
+      "learning_rate": 9.76492164054685e-05,
+      "loss": 0.1409,
+      "step": 42400
+    },
+    {
+      "epoch": 11.78,
+      "eval_loss": 0.9654198884963989,
+      "eval_runtime": 17.6901,
+      "eval_samples_per_second": 5.596,
+      "eval_wer": 0.452319587628866,
+      "step": 42400
+    },
+    {
+      "epoch": 11.81,
+      "learning_rate": 9.76436589974436e-05,
+      "loss": 0.1332,
+      "step": 42500
+    },
+    {
+      "epoch": 11.81,
+      "eval_loss": 0.8946043252944946,
+      "eval_runtime": 17.6063,
+      "eval_samples_per_second": 5.623,
+      "eval_wer": 0.43427835051546393,
+      "step": 42500
+    },
+    {
+      "epoch": 11.84,
+      "learning_rate": 9.76381015894187e-05,
+      "loss": 0.1525,
+      "step": 42600
+    },
+    {
+      "epoch": 11.84,
+      "eval_loss": 1.005324363708496,
+      "eval_runtime": 17.5394,
+      "eval_samples_per_second": 5.644,
+      "eval_wer": 0.45103092783505155,
+      "step": 42600
+    },
+    {
+      "epoch": 11.86,
+      "learning_rate": 9.76325441813938e-05,
+      "loss": 0.1331,
+      "step": 42700
+    },
+    {
+      "epoch": 11.86,
+      "eval_loss": 0.7600377202033997,
+      "eval_runtime": 17.3487,
+      "eval_samples_per_second": 5.706,
+      "eval_wer": 0.42912371134020616,
+      "step": 42700
+    },
+    {
+      "epoch": 11.89,
+      "learning_rate": 9.76269867733689e-05,
+      "loss": 0.1375,
+      "step": 42800
+    },
+    {
+      "epoch": 11.89,
+      "eval_loss": 0.9018005132675171,
+      "eval_runtime": 17.6205,
+      "eval_samples_per_second": 5.618,
+      "eval_wer": 0.42783505154639173,
+      "step": 42800
+    },
+    {
+      "epoch": 11.92,
+      "learning_rate": 9.762142936534401e-05,
+      "loss": 0.1654,
+      "step": 42900
+    },
+    {
+      "epoch": 11.92,
+      "eval_loss": 0.8960022330284119,
+      "eval_runtime": 17.6311,
+      "eval_samples_per_second": 5.615,
+      "eval_wer": 0.4368556701030928,
+      "step": 42900
+    },
+    {
+      "epoch": 11.95,
+      "learning_rate": 9.761587195731911e-05,
+      "loss": 0.1514,
+      "step": 43000
+    },
+    {
+      "epoch": 11.95,
+      "eval_loss": 0.9692587852478027,
+      "eval_runtime": 17.8495,
+      "eval_samples_per_second": 5.546,
+      "eval_wer": 0.43041237113402064,
+      "step": 43000
+    },
+    {
+      "epoch": 11.98,
+      "learning_rate": 9.761031454929421e-05,
+      "loss": 0.1396,
+      "step": 43100
+    },
+    {
+      "epoch": 11.98,
+      "eval_loss": 1.0123811960220337,
+      "eval_runtime": 17.4191,
+      "eval_samples_per_second": 5.683,
+      "eval_wer": 0.44587628865979384,
+      "step": 43100
+    },
+    {
+      "epoch": 12.0,
+      "learning_rate": 9.760475714126933e-05,
+      "loss": 0.1542,
+      "step": 43200
+    },
+    {
+      "epoch": 12.0,
+      "eval_loss": 1.1317671537399292,
+      "eval_runtime": 17.8602,
+      "eval_samples_per_second": 5.543,
+      "eval_wer": 0.43556701030927836,
+      "step": 43200
+    },
+    {
+      "epoch": 12.03,
+      "learning_rate": 9.759919973324441e-05,
+      "loss": 0.1317,
+      "step": 43300
+    },
+    {
+      "epoch": 12.03,
+      "eval_loss": 0.9830509424209595,
+      "eval_runtime": 17.6547,
+      "eval_samples_per_second": 5.608,
+      "eval_wer": 0.4536082474226804,
+      "step": 43300
+    },
+    {
+      "epoch": 12.06,
+      "learning_rate": 9.759364232521953e-05,
+      "loss": 0.1245,
+      "step": 43400
+    },
+    {
+      "epoch": 12.06,
+      "eval_loss": 0.8731426000595093,
+      "eval_runtime": 17.5203,
+      "eval_samples_per_second": 5.651,
+      "eval_wer": 0.44329896907216493,
+      "step": 43400
+    },
+    {
+      "epoch": 12.09,
+      "learning_rate": 9.758808491719462e-05,
+      "loss": 0.1514,
+      "step": 43500
+    },
+    {
+      "epoch": 12.09,
+      "eval_loss": 0.9377870559692383,
+      "eval_runtime": 17.9338,
+      "eval_samples_per_second": 5.52,
+      "eval_wer": 0.452319587628866,
+      "step": 43500
+    },
+    {
+      "epoch": 12.11,
+      "learning_rate": 9.758252750916973e-05,
+      "loss": 0.1446,
+      "step": 43600
+    },
+    {
+      "epoch": 12.11,
+      "eval_loss": 0.7838338017463684,
+      "eval_runtime": 17.4422,
+      "eval_samples_per_second": 5.676,
+      "eval_wer": 0.42783505154639173,
+      "step": 43600
+    },
+    {
+      "epoch": 12.14,
+      "learning_rate": 9.757697010114483e-05,
+      "loss": 0.121,
+      "step": 43700
+    },
+    {
+      "epoch": 12.14,
+      "eval_loss": 0.9419594407081604,
+      "eval_runtime": 17.4214,
+      "eval_samples_per_second": 5.683,
+      "eval_wer": 0.43943298969072164,
+      "step": 43700
+    },
+    {
+      "epoch": 12.17,
+      "learning_rate": 9.757141269311993e-05,
+      "loss": 0.1423,
+      "step": 43800
+    },
+    {
+      "epoch": 12.17,
+      "eval_loss": 1.0655540227890015,
+      "eval_runtime": 17.57,
+      "eval_samples_per_second": 5.635,
+      "eval_wer": 0.4574742268041237,
+      "step": 43800
+    },
+    {
+      "epoch": 12.2,
+      "learning_rate": 9.756585528509503e-05,
+      "loss": 0.1337,
+      "step": 43900
+    },
+    {
+      "epoch": 12.2,
+      "eval_loss": 0.8492249250411987,
+      "eval_runtime": 17.7159,
+      "eval_samples_per_second": 5.588,
+      "eval_wer": 0.44716494845360827,
+      "step": 43900
+    },
+    {
+      "epoch": 12.23,
+      "learning_rate": 9.756029787707015e-05,
+      "loss": 0.1456,
+      "step": 44000
+    },
+    {
+      "epoch": 12.23,
+      "eval_loss": 0.9107276201248169,
+      "eval_runtime": 17.6083,
+      "eval_samples_per_second": 5.622,
+      "eval_wer": 0.47036082474226804,
+      "step": 44000
+    },
+    {
+      "epoch": 12.25,
+      "learning_rate": 9.755474046904525e-05,
+      "loss": 0.1489,
+      "step": 44100
+    },
+    {
+      "epoch": 12.25,
+      "eval_loss": 0.7715519070625305,
+      "eval_runtime": 17.578,
+      "eval_samples_per_second": 5.632,
+      "eval_wer": 0.44716494845360827,
+      "step": 44100
+    },
+    {
+      "epoch": 12.28,
+      "learning_rate": 9.754918306102035e-05,
+      "loss": 0.1377,
+      "step": 44200
+    },
+    {
+      "epoch": 12.28,
+      "eval_loss": 0.7675387263298035,
+      "eval_runtime": 17.6393,
+      "eval_samples_per_second": 5.612,
+      "eval_wer": 0.452319587628866,
+      "step": 44200
+    },
+    {
+      "epoch": 12.31,
+      "learning_rate": 9.754362565299545e-05,
+      "loss": 0.1372,
+      "step": 44300
+    },
+    {
+      "epoch": 12.31,
+      "eval_loss": 0.8274447321891785,
+      "eval_runtime": 18.0346,
+      "eval_samples_per_second": 5.489,
+      "eval_wer": 0.45103092783505155,
+      "step": 44300
+    },
+    {
+      "epoch": 12.34,
+      "learning_rate": 9.753806824497055e-05,
+      "loss": 0.1293,
+      "step": 44400
+    },
+    {
+      "epoch": 12.34,
+      "eval_loss": 0.7182275652885437,
+      "eval_runtime": 17.8809,
+      "eval_samples_per_second": 5.537,
+      "eval_wer": 0.44458762886597936,
+      "step": 44400
+    },
+    {
+      "epoch": 12.36,
+      "learning_rate": 9.753251083694566e-05,
+      "loss": 0.1372,
+      "step": 44500
+    },
+    {
+      "epoch": 12.36,
+      "eval_loss": 0.7243654727935791,
+      "eval_runtime": 17.4045,
+      "eval_samples_per_second": 5.688,
+      "eval_wer": 0.4484536082474227,
+      "step": 44500
+    },
+    {
+      "epoch": 12.39,
+      "learning_rate": 9.752695342892075e-05,
+      "loss": 0.137,
+      "step": 44600
+    },
+    {
+      "epoch": 12.39,
+      "eval_loss": 0.9591490030288696,
+      "eval_runtime": 17.8398,
+      "eval_samples_per_second": 5.549,
+      "eval_wer": 0.44716494845360827,
+      "step": 44600
+    },
+    {
+      "epoch": 12.42,
+      "learning_rate": 9.752139602089586e-05,
+      "loss": 0.1402,
+      "step": 44700
+    },
+    {
+      "epoch": 12.42,
+      "eval_loss": 0.8182889223098755,
+      "eval_runtime": 17.7667,
+      "eval_samples_per_second": 5.572,
+      "eval_wer": 0.44458762886597936,
+      "step": 44700
+    },
+    {
+      "epoch": 12.45,
+      "learning_rate": 9.751583861287096e-05,
+      "loss": 0.1433,
+      "step": 44800
+    },
+    {
+      "epoch": 12.45,
+      "eval_loss": 0.8458079099655151,
+      "eval_runtime": 17.7774,
+      "eval_samples_per_second": 5.569,
+      "eval_wer": 0.4587628865979381,
+      "step": 44800
+    },
+    {
+      "epoch": 12.48,
+      "learning_rate": 9.751028120484606e-05,
+      "loss": 0.1507,
+      "step": 44900
+    },
+    {
+      "epoch": 12.48,
+      "eval_loss": 0.7557888627052307,
+      "eval_runtime": 17.5593,
+      "eval_samples_per_second": 5.638,
+      "eval_wer": 0.43943298969072164,
+      "step": 44900
+    },
+    {
+      "epoch": 12.5,
+      "learning_rate": 9.750472379682116e-05,
+      "loss": 0.1364,
+      "step": 45000
+    },
+    {
+      "epoch": 12.5,
+      "eval_loss": 0.7761731147766113,
+      "eval_runtime": 17.2076,
+      "eval_samples_per_second": 5.753,
+      "eval_wer": 0.43170103092783507,
+      "step": 45000
+    },
+    {
+      "epoch": 12.53,
+      "learning_rate": 9.749916638879626e-05,
+      "loss": 0.1352,
+      "step": 45100
+    },
+    {
+      "epoch": 12.53,
+      "eval_loss": 0.8945127725601196,
+      "eval_runtime": 17.1857,
+      "eval_samples_per_second": 5.761,
+      "eval_wer": 0.4497422680412371,
+      "step": 45100
+    },
+    {
+      "epoch": 12.56,
+      "learning_rate": 9.749360898077138e-05,
+      "loss": 0.1617,
+      "step": 45200
+    },
+    {
+      "epoch": 12.56,
+      "eval_loss": 0.9693708419799805,
+      "eval_runtime": 17.6296,
+      "eval_samples_per_second": 5.616,
+      "eval_wer": 0.4497422680412371,
+      "step": 45200
+    },
+    {
+      "epoch": 12.59,
+      "learning_rate": 9.748805157274648e-05,
+      "loss": 0.1391,
+      "step": 45300
+    },
+    {
+      "epoch": 12.59,
+      "eval_loss": 0.8889138102531433,
+      "eval_runtime": 17.7851,
+      "eval_samples_per_second": 5.566,
+      "eval_wer": 0.45618556701030927,
+      "step": 45300
+    },
+    {
+      "epoch": 12.61,
+      "learning_rate": 9.748249416472158e-05,
+      "loss": 0.1356,
+      "step": 45400
+    },
+    {
+      "epoch": 12.61,
+      "eval_loss": 0.8959171175956726,
+      "eval_runtime": 18.0488,
+      "eval_samples_per_second": 5.485,
+      "eval_wer": 0.4574742268041237,
+      "step": 45400
+    },
+    {
+      "epoch": 12.64,
+      "learning_rate": 9.747693675669668e-05,
+      "loss": 0.1316,
+      "step": 45500
+    },
+    {
+      "epoch": 12.64,
+      "eval_loss": 0.9146299362182617,
+      "eval_runtime": 17.9338,
+      "eval_samples_per_second": 5.52,
+      "eval_wer": 0.4497422680412371,
+      "step": 45500
+    },
+    {
+      "epoch": 12.67,
+      "learning_rate": 9.747137934867178e-05,
+      "loss": 0.1398,
+      "step": 45600
+    },
+    {
+      "epoch": 12.67,
+      "eval_loss": 0.9133784174919128,
+      "eval_runtime": 17.7605,
+      "eval_samples_per_second": 5.574,
+      "eval_wer": 0.4587628865979381,
+      "step": 45600
+    },
+    {
+      "epoch": 12.7,
+      "learning_rate": 9.746582194064688e-05,
+      "loss": 0.1508,
+      "step": 45700
+    },
+    {
+      "epoch": 12.7,
+      "eval_loss": 1.0588983297348022,
+      "eval_runtime": 17.9533,
+      "eval_samples_per_second": 5.514,
+      "eval_wer": 0.45618556701030927,
+      "step": 45700
+    },
+    {
+      "epoch": 12.73,
+      "learning_rate": 9.7460264532622e-05,
+      "loss": 0.1404,
+      "step": 45800
+    },
+    {
+      "epoch": 12.73,
+      "eval_loss": 0.902584433555603,
+      "eval_runtime": 17.8614,
+      "eval_samples_per_second": 5.543,
+      "eval_wer": 0.44072164948453607,
+      "step": 45800
+    },
+    {
+      "epoch": 12.75,
+      "learning_rate": 9.745470712459708e-05,
+      "loss": 0.1375,
+      "step": 45900
+    },
+    {
+      "epoch": 12.75,
+      "eval_loss": 0.8664600849151611,
+      "eval_runtime": 18.0529,
+      "eval_samples_per_second": 5.484,
+      "eval_wer": 0.44458762886597936,
+      "step": 45900
+    },
+    {
+      "epoch": 12.78,
+      "learning_rate": 9.74491497165722e-05,
+      "loss": 0.1189,
+      "step": 46000
+    },
+    {
+      "epoch": 12.78,
+      "eval_loss": 0.8664206266403198,
+      "eval_runtime": 17.4467,
+      "eval_samples_per_second": 5.674,
+      "eval_wer": 0.44329896907216493,
+      "step": 46000
+    },
+    {
+      "epoch": 12.81,
+      "learning_rate": 9.74435923085473e-05,
+      "loss": 0.1285,
+      "step": 46100
+    },
+    {
+      "epoch": 12.81,
+      "eval_loss": 0.7403653264045715,
+      "eval_runtime": 17.6209,
+      "eval_samples_per_second": 5.618,
+      "eval_wer": 0.4420103092783505,
+      "step": 46100
+    },
+    {
+      "epoch": 12.84,
+      "learning_rate": 9.74380349005224e-05,
+      "loss": 0.1276,
+      "step": 46200
+    },
+    {
+      "epoch": 12.84,
+      "eval_loss": 0.8062552213668823,
+      "eval_runtime": 18.0006,
+      "eval_samples_per_second": 5.5,
+      "eval_wer": 0.4639175257731959,
+      "step": 46200
+    },
+    {
+      "epoch": 12.86,
+      "learning_rate": 9.743247749249751e-05,
+      "loss": 0.1404,
+      "step": 46300
+    },
+    {
+      "epoch": 12.86,
+      "eval_loss": 0.8171700239181519,
+      "eval_runtime": 17.6983,
+      "eval_samples_per_second": 5.594,
+      "eval_wer": 0.4690721649484536,
+      "step": 46300
+    },
+    {
+      "epoch": 12.89,
+      "learning_rate": 9.742692008447261e-05,
+      "loss": 0.1425,
+      "step": 46400
+    },
+    {
+      "epoch": 12.89,
+      "eval_loss": 0.8227722644805908,
+      "eval_runtime": 17.5604,
+      "eval_samples_per_second": 5.638,
+      "eval_wer": 0.45103092783505155,
+      "step": 46400
+    },
+    {
+      "epoch": 12.92,
+      "learning_rate": 9.742136267644771e-05,
+      "loss": 0.1366,
+      "step": 46500
+    },
+    {
+      "epoch": 12.92,
+      "eval_loss": 0.904529869556427,
+      "eval_runtime": 18.125,
+      "eval_samples_per_second": 5.462,
+      "eval_wer": 0.44458762886597936,
+      "step": 46500
+    },
+    {
+      "epoch": 12.95,
+      "learning_rate": 9.741580526842281e-05,
+      "loss": 0.1278,
+      "step": 46600
+    },
+    {
+      "epoch": 12.95,
+      "eval_loss": 0.8617892265319824,
+      "eval_runtime": 17.7415,
+      "eval_samples_per_second": 5.58,
+      "eval_wer": 0.4574742268041237,
+      "step": 46600
+    },
+    {
+      "epoch": 12.98,
+      "learning_rate": 9.741024786039791e-05,
+      "loss": 0.1359,
+      "step": 46700
+    },
+    {
+      "epoch": 12.98,
+      "eval_loss": 0.9559828639030457,
+      "eval_runtime": 17.6272,
+      "eval_samples_per_second": 5.616,
+      "eval_wer": 0.4742268041237113,
+      "step": 46700
+    },
+    {
+      "epoch": 13.0,
+      "learning_rate": 9.740469045237301e-05,
+      "loss": 0.139,
+      "step": 46800
+    },
+    {
+      "epoch": 13.0,
+      "eval_loss": 0.8375121355056763,
+      "eval_runtime": 17.6913,
+      "eval_samples_per_second": 5.596,
+      "eval_wer": 0.4587628865979381,
+      "step": 46800
+    },
+    {
+      "epoch": 13.03,
+      "learning_rate": 9.739913304434813e-05,
+      "loss": 0.1354,
+      "step": 46900
+    },
+    {
+      "epoch": 13.03,
+      "eval_loss": 0.8707177639007568,
+      "eval_runtime": 17.7257,
+      "eval_samples_per_second": 5.585,
+      "eval_wer": 0.46262886597938147,
+      "step": 46900
+    },
+    {
+      "epoch": 13.06,
+      "learning_rate": 9.739357563632322e-05,
+      "loss": 0.1175,
+      "step": 47000
+    },
+    {
+      "epoch": 13.06,
+      "eval_loss": 0.8792176842689514,
+      "eval_runtime": 17.6797,
+      "eval_samples_per_second": 5.6,
+      "eval_wer": 0.45489690721649484,
+      "step": 47000
+    },
+    {
+      "epoch": 13.09,
+      "learning_rate": 9.738801822829833e-05,
+      "loss": 0.137,
+      "step": 47100
+    },
+    {
+      "epoch": 13.09,
+      "eval_loss": 0.9775089025497437,
+      "eval_runtime": 17.5385,
+      "eval_samples_per_second": 5.645,
+      "eval_wer": 0.4574742268041237,
+      "step": 47100
+    },
+    {
+      "epoch": 13.11,
+      "learning_rate": 9.738246082027343e-05,
+      "loss": 0.1223,
+      "step": 47200
+    },
+    {
+      "epoch": 13.11,
+      "eval_loss": 1.057368516921997,
+      "eval_runtime": 17.6222,
+      "eval_samples_per_second": 5.618,
+      "eval_wer": 0.46005154639175255,
+      "step": 47200
+    },
+    {
+      "epoch": 13.14,
+      "learning_rate": 9.737690341224853e-05,
+      "loss": 0.12,
+      "step": 47300
+    },
+    {
+      "epoch": 13.14,
+      "eval_loss": 0.9751449227333069,
+      "eval_runtime": 18.8995,
+      "eval_samples_per_second": 5.238,
+      "eval_wer": 0.4690721649484536,
+      "step": 47300
+    },
+    {
+      "epoch": 13.17,
+      "learning_rate": 9.737134600422364e-05,
+      "loss": 0.1428,
+      "step": 47400
+    },
+    {
+      "epoch": 13.17,
+      "eval_loss": 0.9404999017715454,
+      "eval_runtime": 18.2253,
+      "eval_samples_per_second": 5.432,
+      "eval_wer": 0.4484536082474227,
+      "step": 47400
+    },
+    {
+      "epoch": 13.2,
+      "learning_rate": 9.736578859619873e-05,
+      "loss": 0.1241,
+      "step": 47500
+    },
+    {
+      "epoch": 13.2,
+      "eval_loss": 0.849004328250885,
+      "eval_runtime": 18.7361,
+      "eval_samples_per_second": 5.284,
+      "eval_wer": 0.43943298969072164,
+      "step": 47500
+    },
+    {
+      "epoch": 13.23,
+      "learning_rate": 9.736023118817385e-05,
+      "loss": 0.1182,
+      "step": 47600
+    },
+    {
+      "epoch": 13.23,
+      "eval_loss": 0.9481346607208252,
+      "eval_runtime": 18.3385,
+      "eval_samples_per_second": 5.398,
+      "eval_wer": 0.4381443298969072,
+      "step": 47600
+    },
+    {
+      "epoch": 13.25,
+      "learning_rate": 9.735467378014895e-05,
+      "loss": 0.1358,
+      "step": 47700
+    },
+    {
+      "epoch": 13.25,
+      "eval_loss": 0.8224361538887024,
+      "eval_runtime": 18.1961,
+      "eval_samples_per_second": 5.441,
+      "eval_wer": 0.4329896907216495,
+      "step": 47700
+    },
+    {
+      "epoch": 13.28,
+      "learning_rate": 9.734911637212405e-05,
+      "loss": 0.1396,
+      "step": 47800
+    },
+    {
+      "epoch": 13.28,
+      "eval_loss": 0.8473471403121948,
+      "eval_runtime": 17.9527,
+      "eval_samples_per_second": 5.514,
+      "eval_wer": 0.46005154639175255,
+      "step": 47800
+    },
+    {
+      "epoch": 13.31,
+      "learning_rate": 9.734355896409915e-05,
+      "loss": 0.1267,
+      "step": 47900
+    },
+    {
+      "epoch": 13.31,
+      "eval_loss": 0.9468744993209839,
+      "eval_runtime": 17.4889,
+      "eval_samples_per_second": 5.661,
+      "eval_wer": 0.44072164948453607,
+      "step": 47900
+    },
+    {
+      "epoch": 13.34,
+      "learning_rate": 9.733800155607425e-05,
+      "loss": 0.1252,
+      "step": 48000
+    },
+    {
+      "epoch": 13.34,
+      "eval_loss": 0.8264307379722595,
+      "eval_runtime": 18.0995,
+      "eval_samples_per_second": 5.47,
+      "eval_wer": 0.45103092783505155,
+      "step": 48000
+    },
+    {
+      "epoch": 13.36,
+      "learning_rate": 9.733244414804935e-05,
+      "loss": 0.1226,
+      "step": 48100
+    },
+    {
+      "epoch": 13.36,
+      "eval_loss": 0.8667508363723755,
+      "eval_runtime": 18.589,
+      "eval_samples_per_second": 5.326,
+      "eval_wer": 0.44716494845360827,
+      "step": 48100
+    },
+    {
+      "epoch": 13.39,
+      "learning_rate": 9.732688674002446e-05,
+      "loss": 0.1328,
+      "step": 48200
+    },
+    {
+      "epoch": 13.39,
+      "eval_loss": 0.8937565088272095,
+      "eval_runtime": 18.255,
+      "eval_samples_per_second": 5.423,
+      "eval_wer": 0.4536082474226804,
+      "step": 48200
+    },
+    {
+      "epoch": 13.42,
+      "learning_rate": 9.732132933199955e-05,
+      "loss": 0.1425,
+      "step": 48300
+    },
+    {
+      "epoch": 13.42,
+      "eval_loss": 0.9718654155731201,
+      "eval_runtime": 18.0435,
+      "eval_samples_per_second": 5.487,
+      "eval_wer": 0.4639175257731959,
+      "step": 48300
+    },
+    {
+      "epoch": 13.45,
+      "learning_rate": 9.731577192397466e-05,
+      "loss": 0.114,
+      "step": 48400
+    },
+    {
+      "epoch": 13.45,
+      "eval_loss": 0.9460931420326233,
+      "eval_runtime": 18.0945,
+      "eval_samples_per_second": 5.471,
+      "eval_wer": 0.46262886597938147,
+      "step": 48400
+    },
+    {
+      "epoch": 13.48,
+      "learning_rate": 9.731021451594976e-05,
+      "loss": 0.1425,
+      "step": 48500
+    },
+    {
+      "epoch": 13.48,
+      "eval_loss": 0.7326005697250366,
+      "eval_runtime": 17.9853,
+      "eval_samples_per_second": 5.504,
+      "eval_wer": 0.44587628865979384,
+      "step": 48500
+    },
+    {
+      "epoch": 13.5,
+      "learning_rate": 9.730465710792486e-05,
+      "loss": 0.1164,
+      "step": 48600
+    },
+    {
+      "epoch": 13.5,
+      "eval_loss": 0.7906768321990967,
+      "eval_runtime": 18.2994,
+      "eval_samples_per_second": 5.41,
+      "eval_wer": 0.4536082474226804,
+      "step": 48600
+    },
+    {
+      "epoch": 13.53,
+      "learning_rate": 9.729909969989998e-05,
+      "loss": 0.1334,
+      "step": 48700
+    },
+    {
+      "epoch": 13.53,
+      "eval_loss": 0.8335945010185242,
+      "eval_runtime": 18.0063,
+      "eval_samples_per_second": 5.498,
+      "eval_wer": 0.4536082474226804,
+      "step": 48700
+    },
+    {
+      "epoch": 13.56,
+      "learning_rate": 9.729354229187507e-05,
+      "loss": 0.1213,
+      "step": 48800
+    },
+    {
+      "epoch": 13.56,
+      "eval_loss": 1.0961216688156128,
+      "eval_runtime": 18.2967,
+      "eval_samples_per_second": 5.411,
+      "eval_wer": 0.4793814432989691,
+      "step": 48800
+    },
+    {
+      "epoch": 13.59,
+      "learning_rate": 9.728798488385018e-05,
+      "loss": 0.1186,
+      "step": 48900
+    },
+    {
+      "epoch": 13.59,
+      "eval_loss": 1.0261573791503906,
+      "eval_runtime": 17.874,
+      "eval_samples_per_second": 5.539,
+      "eval_wer": 0.47551546391752575,
+      "step": 48900
+    },
+    {
+      "epoch": 13.61,
+      "learning_rate": 9.728242747582528e-05,
+      "loss": 0.1313,
+      "step": 49000
+    },
+    {
+      "epoch": 13.61,
+      "eval_loss": 0.9995627403259277,
+      "eval_runtime": 17.8346,
+      "eval_samples_per_second": 5.551,
+      "eval_wer": 0.44716494845360827,
+      "step": 49000
+    },
+    {
+      "epoch": 13.64,
+      "learning_rate": 9.727687006780038e-05,
+      "loss": 0.1372,
+      "step": 49100
+    },
+    {
+      "epoch": 13.64,
+      "eval_loss": 1.0337309837341309,
+      "eval_runtime": 17.768,
+      "eval_samples_per_second": 5.572,
+      "eval_wer": 0.46005154639175255,
+      "step": 49100
+    },
+    {
+      "epoch": 13.67,
+      "learning_rate": 9.727131265977548e-05,
+      "loss": 0.1183,
+      "step": 49200
+    },
+    {
+      "epoch": 13.67,
+      "eval_loss": 0.9927868247032166,
+      "eval_runtime": 18.1061,
+      "eval_samples_per_second": 5.468,
+      "eval_wer": 0.48195876288659795,
+      "step": 49200
+    },
+    {
+      "epoch": 13.7,
+      "learning_rate": 9.72657552517506e-05,
+      "loss": 0.1167,
+      "step": 49300
+    },
+    {
+      "epoch": 13.7,
+      "eval_loss": 0.9940055012702942,
+      "eval_runtime": 17.6771,
+      "eval_samples_per_second": 5.6,
+      "eval_wer": 0.49097938144329895,
+      "step": 49300
+    },
+    {
+      "epoch": 13.73,
+      "learning_rate": 9.726019784372568e-05,
+      "loss": 0.1324,
+      "step": 49400
+    },
+    {
+      "epoch": 13.73,
+      "eval_loss": 0.9435777068138123,
+      "eval_runtime": 17.8591,
+      "eval_samples_per_second": 5.543,
+      "eval_wer": 0.4729381443298969,
+      "step": 49400
+    },
+    {
+      "epoch": 13.75,
+      "learning_rate": 9.72546404357008e-05,
+      "loss": 0.1353,
+      "step": 49500
+    },
+    {
+      "epoch": 13.75,
+      "eval_loss": 1.0063272714614868,
+      "eval_runtime": 17.8941,
+      "eval_samples_per_second": 5.533,
+      "eval_wer": 0.4961340206185567,
+      "step": 49500
+    },
+    {
+      "epoch": 13.78,
+      "learning_rate": 9.72490830276759e-05,
+      "loss": 0.1193,
+      "step": 49600
+    },
+    {
+      "epoch": 13.78,
+      "eval_loss": 0.9747934341430664,
+      "eval_runtime": 17.8823,
+      "eval_samples_per_second": 5.536,
+      "eval_wer": 0.46134020618556704,
+      "step": 49600
+    },
+    {
+      "epoch": 13.81,
+      "learning_rate": 9.7243525619651e-05,
+      "loss": 0.132,
+      "step": 49700
+    },
+    {
+      "epoch": 13.81,
+      "eval_loss": 1.1134113073349,
+      "eval_runtime": 17.6443,
+      "eval_samples_per_second": 5.611,
+      "eval_wer": 0.4536082474226804,
+      "step": 49700
+    },
+    {
+      "epoch": 13.84,
+      "learning_rate": 9.723796821162611e-05,
+      "loss": 0.1244,
+      "step": 49800
+    },
+    {
+      "epoch": 13.84,
+      "eval_loss": 0.9971640706062317,
+      "eval_runtime": 17.5696,
+      "eval_samples_per_second": 5.635,
+      "eval_wer": 0.48195876288659795,
+      "step": 49800
+    },
+    {
+      "epoch": 13.86,
+      "learning_rate": 9.72324108036012e-05,
+      "loss": 0.1502,
+      "step": 49900
+    },
+    {
+      "epoch": 13.86,
+      "eval_loss": 1.0188100337982178,
+      "eval_runtime": 17.5476,
+      "eval_samples_per_second": 5.642,
+      "eval_wer": 0.47164948453608246,
+      "step": 49900
+    },
+    {
+      "epoch": 13.89,
+      "learning_rate": 9.722685339557631e-05,
+      "loss": 0.1361,
+      "step": 50000
+    },
+    {
+      "epoch": 13.89,
+      "eval_loss": 1.0230387449264526,
+      "eval_runtime": 17.5085,
+      "eval_samples_per_second": 5.654,
+      "eval_wer": 0.46649484536082475,
+      "step": 50000
+    },
+    {
+      "epoch": 13.92,
+      "learning_rate": 9.722129598755141e-05,
+      "loss": 0.1449,
+      "step": 50100
+    },
+    {
+      "epoch": 13.92,
+      "eval_loss": 0.9599841833114624,
+      "eval_runtime": 17.6089,
+      "eval_samples_per_second": 5.622,
+      "eval_wer": 0.46262886597938147,
+      "step": 50100
+    },
+    {
+      "epoch": 13.95,
+      "learning_rate": 9.721573857952651e-05,
+      "loss": 0.1295,
+      "step": 50200
+    },
+    {
+      "epoch": 13.95,
+      "eval_loss": 1.113866925239563,
+      "eval_runtime": 17.8779,
+      "eval_samples_per_second": 5.538,
+      "eval_wer": 0.45489690721649484,
+      "step": 50200
+    },
+    {
+      "epoch": 13.98,
+      "learning_rate": 9.721018117150161e-05,
+      "loss": 0.1299,
+      "step": 50300
+    },
+    {
+      "epoch": 13.98,
+      "eval_loss": 0.9001206159591675,
+      "eval_runtime": 17.7999,
+      "eval_samples_per_second": 5.562,
+      "eval_wer": 0.46134020618556704,
+      "step": 50300
+    },
+    {
+      "epoch": 14.0,
+      "learning_rate": 9.720462376347672e-05,
+      "loss": 0.1217,
+      "step": 50400
+    },
+    {
+      "epoch": 14.0,
+      "eval_loss": 1.0081254243850708,
+      "eval_runtime": 17.6755,
+      "eval_samples_per_second": 5.601,
+      "eval_wer": 0.4690721649484536,
+      "step": 50400
+    },
+    {
+      "epoch": 14.03,
+      "learning_rate": 9.719906635545182e-05,
+      "loss": 0.1383,
+      "step": 50500
+    },
+    {
+      "epoch": 14.03,
+      "eval_loss": 0.9827317595481873,
+      "eval_runtime": 17.7559,
+      "eval_samples_per_second": 5.576,
+      "eval_wer": 0.4536082474226804,
+      "step": 50500
+    },
+    {
+      "epoch": 14.06,
+      "learning_rate": 9.719350894742693e-05,
+      "loss": 0.1154,
+      "step": 50600
+    },
+    {
+      "epoch": 14.06,
+      "eval_loss": 1.0788767337799072,
+      "eval_runtime": 17.8636,
+      "eval_samples_per_second": 5.542,
+      "eval_wer": 0.47164948453608246,
+      "step": 50600
+    }
+  ],
+  "max_steps": 1799500,
+  "num_train_epochs": 500,
+  "total_flos": 2.6466141630383727e+19,
+  "trial_name": null,
+  "trial_params": null
+}