{
  "best_global_step": 9500,
  "best_metric": 16.040464106107944,
  "best_model_checkpoint": "/content/drive/MyDrive/models/whisper_small_ru_model_trainer_3ep/checkpoint-9500",
  "epoch": 3.0,
  "eval_steps": 500,
  "global_step": 9894,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.007580351728320194,
      "grad_norm": 10.993572235107422,
      "learning_rate": 9.975742874469376e-06,
      "loss": 0.4792,
      "step": 25
    },
    {
      "epoch": 0.015160703456640388,
      "grad_norm": 7.92294979095459,
      "learning_rate": 9.950475035374976e-06,
      "loss": 0.27,
      "step": 50
    },
    {
      "epoch": 0.02274105518496058,
      "grad_norm": 11.214887619018555,
      "learning_rate": 9.925207196280575e-06,
      "loss": 0.3111,
      "step": 75
    },
    {
      "epoch": 0.030321406913280776,
      "grad_norm": 6.483564376831055,
      "learning_rate": 9.899939357186174e-06,
      "loss": 0.2659,
      "step": 100
    },
    {
      "epoch": 0.03790175864160097,
      "grad_norm": 13.091401100158691,
      "learning_rate": 9.874671518091774e-06,
      "loss": 0.2524,
      "step": 125
    },
    {
      "epoch": 0.04548211036992116,
      "grad_norm": 9.941374778747559,
      "learning_rate": 9.849403678997373e-06,
      "loss": 0.2601,
      "step": 150
    },
    {
      "epoch": 0.05306246209824136,
      "grad_norm": 10.436762809753418,
      "learning_rate": 9.824135839902972e-06,
      "loss": 0.2364,
      "step": 175
    },
    {
      "epoch": 0.06064281382656155,
      "grad_norm": 6.529853343963623,
      "learning_rate": 9.798868000808572e-06,
      "loss": 0.2406,
      "step": 200
    },
    {
      "epoch": 0.06822316555488174,
      "grad_norm": 6.7477641105651855,
      "learning_rate": 9.77360016171417e-06,
      "loss": 0.2628,
      "step": 225
    },
    {
      "epoch": 0.07580351728320193,
      "grad_norm": 8.057855606079102,
      "learning_rate": 9.74833232261977e-06,
      "loss": 0.2343,
      "step": 250
    },
    {
      "epoch": 0.08338386901152213,
      "grad_norm": 8.777417182922363,
      "learning_rate": 9.72306448352537e-06,
      "loss": 0.2544,
      "step": 275
    },
    {
      "epoch": 0.09096422073984232,
      "grad_norm": 9.590123176574707,
      "learning_rate": 9.697796644430968e-06,
      "loss": 0.2533,
      "step": 300
    },
    {
      "epoch": 0.09854457246816252,
      "grad_norm": 12.065073013305664,
      "learning_rate": 9.672528805336569e-06,
      "loss": 0.248,
      "step": 325
    },
    {
      "epoch": 0.10612492419648271,
      "grad_norm": 6.656917095184326,
      "learning_rate": 9.647260966242168e-06,
      "loss": 0.245,
      "step": 350
    },
    {
      "epoch": 0.11370527592480291,
      "grad_norm": 13.028727531433105,
      "learning_rate": 9.621993127147768e-06,
      "loss": 0.2245,
      "step": 375
    },
    {
      "epoch": 0.1212856276531231,
      "grad_norm": 9.074626922607422,
      "learning_rate": 9.596725288053367e-06,
      "loss": 0.2464,
      "step": 400
    },
    {
      "epoch": 0.12886597938144329,
      "grad_norm": 7.7365217208862305,
      "learning_rate": 9.571457448958967e-06,
      "loss": 0.2233,
      "step": 425
    },
    {
      "epoch": 0.13644633110976348,
      "grad_norm": 7.931459426879883,
      "learning_rate": 9.546189609864566e-06,
      "loss": 0.2522,
      "step": 450
    },
    {
      "epoch": 0.14402668283808367,
      "grad_norm": 10.054927825927734,
      "learning_rate": 9.520921770770164e-06,
      "loss": 0.2288,
      "step": 475
    },
    {
      "epoch": 0.15160703456640387,
      "grad_norm": 11.330881118774414,
      "learning_rate": 9.495653931675763e-06,
      "loss": 0.2206,
      "step": 500
    },
    {
      "epoch": 0.15160703456640387,
      "eval_cer": 5.496337965229563,
      "eval_loss": 0.26026931405067444,
      "eval_runtime": 4183.6066,
      "eval_samples_per_second": 2.439,
      "eval_ser": 69.43055963932177,
      "eval_steps_per_second": 0.61,
      "eval_wer": 21.266884240601154,
      "step": 500
    },
    {
      "epoch": 0.15918738629472406,
      "grad_norm": 22.062847137451172,
      "learning_rate": 9.470386092581364e-06,
      "loss": 0.2404,
      "step": 525
    },
    {
      "epoch": 0.16676773802304426,
      "grad_norm": 6.779280662536621,
      "learning_rate": 9.445118253486962e-06,
      "loss": 0.2299,
      "step": 550
    },
    {
      "epoch": 0.17434808975136445,
      "grad_norm": 10.592273712158203,
      "learning_rate": 9.419850414392561e-06,
      "loss": 0.2064,
      "step": 575
    },
    {
      "epoch": 0.18192844147968465,
      "grad_norm": 5.647468566894531,
      "learning_rate": 9.394582575298161e-06,
      "loss": 0.1878,
      "step": 600
    },
    {
      "epoch": 0.18950879320800484,
      "grad_norm": 12.663495063781738,
      "learning_rate": 9.36931473620376e-06,
      "loss": 0.2302,
      "step": 625
    },
    {
      "epoch": 0.19708914493632504,
      "grad_norm": 11.353273391723633,
      "learning_rate": 9.34404689710936e-06,
      "loss": 0.263,
      "step": 650
    },
    {
      "epoch": 0.20466949666464523,
      "grad_norm": 10.60062026977539,
      "learning_rate": 9.318779058014959e-06,
      "loss": 0.2276,
      "step": 675
    },
    {
      "epoch": 0.21224984839296543,
      "grad_norm": 6.700550556182861,
      "learning_rate": 9.29351121892056e-06,
      "loss": 0.2314,
      "step": 700
    },
    {
      "epoch": 0.21983020012128562,
      "grad_norm": 11.847448348999023,
      "learning_rate": 9.268243379826158e-06,
      "loss": 0.2282,
      "step": 725
    },
    {
      "epoch": 0.22741055184960582,
      "grad_norm": 10.603355407714844,
      "learning_rate": 9.242975540731759e-06,
      "loss": 0.1968,
      "step": 750
    },
    {
      "epoch": 0.234990903577926,
      "grad_norm": 10.278948783874512,
      "learning_rate": 9.217707701637356e-06,
      "loss": 0.2209,
      "step": 775
    },
    {
      "epoch": 0.2425712553062462,
      "grad_norm": 7.834460735321045,
      "learning_rate": 9.192439862542956e-06,
      "loss": 0.189,
      "step": 800
    },
    {
      "epoch": 0.2501516070345664,
      "grad_norm": 7.9238810539245605,
      "learning_rate": 9.167172023448555e-06,
      "loss": 0.1758,
      "step": 825
    },
    {
      "epoch": 0.25773195876288657,
      "grad_norm": 16.181570053100586,
      "learning_rate": 9.141904184354155e-06,
      "loss": 0.1999,
      "step": 850
    },
    {
      "epoch": 0.2653123104912068,
      "grad_norm": 5.385849475860596,
      "learning_rate": 9.116636345259754e-06,
      "loss": 0.2353,
      "step": 875
    },
    {
      "epoch": 0.27289266221952696,
      "grad_norm": 14.86568832397461,
      "learning_rate": 9.091368506165354e-06,
      "loss": 0.2039,
      "step": 900
    },
    {
      "epoch": 0.2804730139478472,
      "grad_norm": 8.38940715789795,
      "learning_rate": 9.066100667070953e-06,
      "loss": 0.2048,
      "step": 925
    },
    {
      "epoch": 0.28805336567616735,
      "grad_norm": 6.71006965637207,
      "learning_rate": 9.040832827976552e-06,
      "loss": 0.2016,
      "step": 950
    },
    {
      "epoch": 0.2956337174044876,
      "grad_norm": 9.024951934814453,
      "learning_rate": 9.015564988882152e-06,
      "loss": 0.2272,
      "step": 975
    },
    {
      "epoch": 0.30321406913280774,
      "grad_norm": 6.729151725769043,
      "learning_rate": 8.99029714978775e-06,
      "loss": 0.22,
      "step": 1000
    },
    {
      "epoch": 0.30321406913280774,
      "eval_cer": 5.382320373005102,
      "eval_loss": 0.24674250185489655,
      "eval_runtime": 4121.9591,
      "eval_samples_per_second": 2.475,
      "eval_ser": 67.35273939037538,
      "eval_steps_per_second": 0.619,
      "eval_wer": 20.297092948978527,
      "step": 1000
    },
    {
      "epoch": 0.31079442086112796,
      "grad_norm": 9.8275785446167,
      "learning_rate": 8.965029310693351e-06,
      "loss": 0.2513,
      "step": 1025
    },
    {
      "epoch": 0.31837477258944813,
      "grad_norm": 8.723328590393066,
      "learning_rate": 8.93976147159895e-06,
      "loss": 0.2095,
      "step": 1050
    },
    {
      "epoch": 0.32595512431776835,
      "grad_norm": 10.17667007446289,
      "learning_rate": 8.91449363250455e-06,
      "loss": 0.1945,
      "step": 1075
    },
    {
      "epoch": 0.3335354760460885,
      "grad_norm": 9.78806209564209,
      "learning_rate": 8.889225793410147e-06,
      "loss": 0.2357,
      "step": 1100
    },
    {
      "epoch": 0.34111582777440874,
      "grad_norm": 7.773096084594727,
      "learning_rate": 8.863957954315748e-06,
      "loss": 0.1901,
      "step": 1125
    },
    {
      "epoch": 0.3486961795027289,
      "grad_norm": 5.588106155395508,
      "learning_rate": 8.838690115221346e-06,
      "loss": 0.2156,
      "step": 1150
    },
    {
      "epoch": 0.35627653123104913,
      "grad_norm": 9.3190336227417,
      "learning_rate": 8.813422276126947e-06,
      "loss": 0.2014,
      "step": 1175
    },
    {
      "epoch": 0.3638568829593693,
      "grad_norm": 7.115675926208496,
      "learning_rate": 8.788154437032545e-06,
      "loss": 0.2073,
      "step": 1200
    },
    {
      "epoch": 0.3714372346876895,
      "grad_norm": 8.060104370117188,
      "learning_rate": 8.762886597938146e-06,
      "loss": 0.1789,
      "step": 1225
    },
    {
      "epoch": 0.3790175864160097,
      "grad_norm": 8.937747955322266,
      "learning_rate": 8.737618758843744e-06,
      "loss": 0.2115,
      "step": 1250
    },
    {
      "epoch": 0.3865979381443299,
      "grad_norm": 10.151251792907715,
      "learning_rate": 8.712350919749343e-06,
      "loss": 0.216,
      "step": 1275
    },
    {
      "epoch": 0.3941782898726501,
      "grad_norm": 6.023889064788818,
      "learning_rate": 8.687083080654943e-06,
      "loss": 0.1836,
      "step": 1300
    },
    {
      "epoch": 0.4017586416009703,
      "grad_norm": 9.286721229553223,
      "learning_rate": 8.661815241560542e-06,
      "loss": 0.1799,
      "step": 1325
    },
    {
      "epoch": 0.40933899332929047,
      "grad_norm": 10.163074493408203,
      "learning_rate": 8.636547402466143e-06,
      "loss": 0.183,
      "step": 1350
    },
    {
      "epoch": 0.4169193450576107,
      "grad_norm": 11.878681182861328,
      "learning_rate": 8.611279563371741e-06,
      "loss": 0.2121,
      "step": 1375
    },
    {
      "epoch": 0.42449969678593086,
      "grad_norm": 10.150252342224121,
      "learning_rate": 8.58601172427734e-06,
      "loss": 0.1837,
      "step": 1400
    },
    {
      "epoch": 0.4320800485142511,
      "grad_norm": 6.533999443054199,
      "learning_rate": 8.560743885182939e-06,
      "loss": 0.219,
      "step": 1425
    },
    {
      "epoch": 0.43966040024257125,
      "grad_norm": 10.426810264587402,
      "learning_rate": 8.535476046088539e-06,
      "loss": 0.2285,
      "step": 1450
    },
    {
      "epoch": 0.44724075197089147,
      "grad_norm": 10.712567329406738,
      "learning_rate": 8.510208206994138e-06,
      "loss": 0.1837,
      "step": 1475
    },
    {
      "epoch": 0.45482110369921164,
      "grad_norm": 8.469382286071777,
      "learning_rate": 8.484940367899738e-06,
      "loss": 0.1901,
      "step": 1500
    },
    {
      "epoch": 0.45482110369921164,
      "eval_cer": 5.116005507016847,
      "eval_loss": 0.23773027956485748,
      "eval_runtime": 4147.5807,
      "eval_samples_per_second": 2.46,
      "eval_ser": 66.17661472116045,
      "eval_steps_per_second": 0.615,
      "eval_wer": 19.564232703452923,
      "step": 1500
    },
    {
      "epoch": 0.46240145542753186,
      "grad_norm": 11.634538650512695,
      "learning_rate": 8.459672528805337e-06,
      "loss": 0.1774,
      "step": 1525
    },
    {
      "epoch": 0.469981807155852,
      "grad_norm": 9.295145034790039,
      "learning_rate": 8.434404689710937e-06,
      "loss": 0.2293,
      "step": 1550
    },
    {
      "epoch": 0.47756215888417225,
      "grad_norm": 9.195737838745117,
      "learning_rate": 8.409136850616536e-06,
      "loss": 0.195,
      "step": 1575
    },
    {
      "epoch": 0.4851425106124924,
      "grad_norm": 7.1239776611328125,
      "learning_rate": 8.383869011522136e-06,
      "loss": 0.1902,
      "step": 1600
    },
    {
      "epoch": 0.49272286234081264,
      "grad_norm": 8.784886360168457,
      "learning_rate": 8.358601172427735e-06,
      "loss": 0.1708,
      "step": 1625
    },
    {
      "epoch": 0.5003032140691328,
      "grad_norm": 9.43509578704834,
      "learning_rate": 8.333333333333334e-06,
      "loss": 0.1978,
      "step": 1650
    },
    {
      "epoch": 0.507883565797453,
      "grad_norm": 9.499982833862305,
      "learning_rate": 8.308065494238934e-06,
      "loss": 0.2036,
      "step": 1675
    },
    {
      "epoch": 0.5154639175257731,
      "grad_norm": 7.286120414733887,
      "learning_rate": 8.282797655144533e-06,
      "loss": 0.2003,
      "step": 1700
    },
    {
      "epoch": 0.5230442692540934,
      "grad_norm": 6.725291728973389,
      "learning_rate": 8.257529816050132e-06,
      "loss": 0.2139,
      "step": 1725
    },
    {
      "epoch": 0.5306246209824136,
      "grad_norm": 5.5903401374816895,
      "learning_rate": 8.23226197695573e-06,
      "loss": 0.1869,
      "step": 1750
    },
    {
      "epoch": 0.5382049727107338,
      "grad_norm": 7.579101085662842,
      "learning_rate": 8.20699413786133e-06,
      "loss": 0.2097,
      "step": 1775
    },
    {
      "epoch": 0.5457853244390539,
      "grad_norm": 4.244465351104736,
      "learning_rate": 8.18172629876693e-06,
      "loss": 0.1906,
      "step": 1800
    },
    {
      "epoch": 0.5533656761673742,
      "grad_norm": 5.5115580558776855,
      "learning_rate": 8.15645845967253e-06,
      "loss": 0.197,
      "step": 1825
    },
    {
      "epoch": 0.5609460278956944,
      "grad_norm": 6.544187068939209,
      "learning_rate": 8.131190620578128e-06,
      "loss": 0.2061,
      "step": 1850
    },
    {
      "epoch": 0.5685263796240145,
      "grad_norm": 6.87299108505249,
      "learning_rate": 8.105922781483729e-06,
      "loss": 0.2018,
      "step": 1875
    },
    {
      "epoch": 0.5761067313523347,
      "grad_norm": 11.84753704071045,
      "learning_rate": 8.080654942389327e-06,
      "loss": 0.2018,
      "step": 1900
    },
    {
      "epoch": 0.583687083080655,
      "grad_norm": 7.050841331481934,
      "learning_rate": 8.055387103294928e-06,
      "loss": 0.1903,
      "step": 1925
    },
    {
      "epoch": 0.5912674348089751,
      "grad_norm": 7.1467461585998535,
      "learning_rate": 8.030119264200527e-06,
      "loss": 0.2079,
      "step": 1950
    },
    {
      "epoch": 0.5988477865372953,
      "grad_norm": 7.622501373291016,
      "learning_rate": 8.004851425106125e-06,
      "loss": 0.2047,
      "step": 1975
    },
    {
      "epoch": 0.6064281382656155,
      "grad_norm": 10.488551139831543,
      "learning_rate": 7.979583586011726e-06,
      "loss": 0.1969,
      "step": 2000
    },
    {
      "epoch": 0.6064281382656155,
      "eval_cer": 5.075425758775576,
      "eval_loss": 0.22730673849582672,
      "eval_runtime": 4147.4744,
      "eval_samples_per_second": 2.46,
      "eval_ser": 64.32421836714691,
      "eval_steps_per_second": 0.615,
      "eval_wer": 19.050882103576033,
      "step": 2000
    },
    {
      "epoch": 0.6140084899939358,
      "grad_norm": 6.823671817779541,
      "learning_rate": 7.954315746917324e-06,
      "loss": 0.2002,
      "step": 2025
    },
    {
      "epoch": 0.6215888417222559,
      "grad_norm": 7.486941814422607,
      "learning_rate": 7.929047907822923e-06,
      "loss": 0.2046,
      "step": 2050
    },
    {
      "epoch": 0.6291691934505761,
      "grad_norm": 6.544958114624023,
      "learning_rate": 7.903780068728523e-06,
      "loss": 0.1736,
      "step": 2075
    },
    {
      "epoch": 0.6367495451788963,
      "grad_norm": 7.139219760894775,
      "learning_rate": 7.878512229634122e-06,
      "loss": 0.1844,
      "step": 2100
    },
    {
      "epoch": 0.6443298969072165,
      "grad_norm": 10.455595970153809,
      "learning_rate": 7.85324439053972e-06,
      "loss": 0.1916,
      "step": 2125
    },
    {
      "epoch": 0.6519102486355367,
      "grad_norm": 6.0816216468811035,
      "learning_rate": 7.827976551445321e-06,
      "loss": 0.2102,
      "step": 2150
    },
    {
      "epoch": 0.6594906003638569,
      "grad_norm": 4.258129119873047,
      "learning_rate": 7.80270871235092e-06,
      "loss": 0.1785,
      "step": 2175
    },
    {
      "epoch": 0.667070952092177,
      "grad_norm": 8.886266708374023,
      "learning_rate": 7.77744087325652e-06,
      "loss": 0.1916,
      "step": 2200
    },
    {
      "epoch": 0.6746513038204973,
      "grad_norm": 8.997633934020996,
      "learning_rate": 7.752173034162119e-06,
      "loss": 0.2088,
      "step": 2225
    },
    {
      "epoch": 0.6822316555488175,
      "grad_norm": 9.341837882995605,
      "learning_rate": 7.72690519506772e-06,
      "loss": 0.1996,
      "step": 2250
    },
    {
      "epoch": 0.6898120072771377,
      "grad_norm": 11.538143157958984,
      "learning_rate": 7.701637355973318e-06,
      "loss": 0.1766,
      "step": 2275
    },
    {
      "epoch": 0.6973923590054578,
      "grad_norm": 6.78935432434082,
      "learning_rate": 7.676369516878917e-06,
      "loss": 0.1748,
      "step": 2300
    },
    {
      "epoch": 0.7049727107337781,
      "grad_norm": 9.051713943481445,
      "learning_rate": 7.651101677784516e-06,
      "loss": 0.173,
      "step": 2325
    },
    {
      "epoch": 0.7125530624620983,
      "grad_norm": 8.62745189666748,
      "learning_rate": 7.625833838690115e-06,
      "loss": 0.1936,
      "step": 2350
    },
    {
      "epoch": 0.7201334141904184,
      "grad_norm": 9.529070854187012,
      "learning_rate": 7.600565999595715e-06,
      "loss": 0.2004,
      "step": 2375
    },
    {
      "epoch": 0.7277137659187386,
      "grad_norm": 7.745057106018066,
      "learning_rate": 7.575298160501314e-06,
      "loss": 0.1832,
      "step": 2400
    },
    {
      "epoch": 0.7352941176470589,
      "grad_norm": 8.844827651977539,
      "learning_rate": 7.550030321406914e-06,
      "loss": 0.2056,
      "step": 2425
    },
    {
      "epoch": 0.742874469375379,
      "grad_norm": 13.322829246520996,
      "learning_rate": 7.524762482312513e-06,
      "loss": 0.1825,
      "step": 2450
    },
    {
      "epoch": 0.7504548211036992,
      "grad_norm": 10.450922012329102,
      "learning_rate": 7.499494643218113e-06,
      "loss": 0.1798,
      "step": 2475
    },
    {
      "epoch": 0.7580351728320194,
      "grad_norm": 9.188721656799316,
      "learning_rate": 7.474226804123712e-06,
      "loss": 0.1743,
      "step": 2500
    },
    {
      "epoch": 0.7580351728320194,
      "eval_cer": 4.852319288687943,
      "eval_loss": 0.2188422828912735,
      "eval_runtime": 4181.1454,
      "eval_samples_per_second": 2.44,
      "eval_ser": 63.148093697931984,
      "eval_steps_per_second": 0.61,
      "eval_wer": 18.228592002415766,
      "step": 2500
    },
    {
      "epoch": 0.7656155245603395,
      "grad_norm": 5.665472030639648,
      "learning_rate": 7.448958965029312e-06,
      "loss": 0.1979,
      "step": 2525
    },
    {
      "epoch": 0.7731958762886598,
      "grad_norm": 4.752591133117676,
      "learning_rate": 7.4236911259349106e-06,
      "loss": 0.191,
      "step": 2550
    },
    {
      "epoch": 0.78077622801698,
      "grad_norm": 9.350257873535156,
      "learning_rate": 7.39842328684051e-06,
      "loss": 0.1735,
      "step": 2575
    },
    {
      "epoch": 0.7883565797453002,
      "grad_norm": 7.437028408050537,
      "learning_rate": 7.37315544774611e-06,
      "loss": 0.2365,
      "step": 2600
    },
    {
      "epoch": 0.7959369314736203,
      "grad_norm": 9.3013916015625,
      "learning_rate": 7.347887608651709e-06,
      "loss": 0.1888,
      "step": 2625
    },
    {
      "epoch": 0.8035172832019406,
      "grad_norm": 7.452167987823486,
      "learning_rate": 7.322619769557308e-06,
      "loss": 0.1752,
      "step": 2650
    },
    {
      "epoch": 0.8110976349302608,
      "grad_norm": 6.42844820022583,
      "learning_rate": 7.2973519304629074e-06,
      "loss": 0.1991,
      "step": 2675
    },
    {
      "epoch": 0.8186779866585809,
      "grad_norm": 6.838171482086182,
      "learning_rate": 7.272084091368506e-06,
      "loss": 0.1748,
      "step": 2700
    },
    {
      "epoch": 0.8262583383869011,
      "grad_norm": 7.954586505889893,
      "learning_rate": 7.246816252274106e-06,
      "loss": 0.1905,
      "step": 2725
    },
    {
      "epoch": 0.8338386901152214,
      "grad_norm": 6.925525188446045,
      "learning_rate": 7.221548413179705e-06,
      "loss": 0.1784,
      "step": 2750
    },
    {
      "epoch": 0.8414190418435415,
      "grad_norm": 7.18998384475708,
      "learning_rate": 7.196280574085305e-06,
      "loss": 0.1693,
      "step": 2775
    },
    {
      "epoch": 0.8489993935718617,
      "grad_norm": 8.666403770446777,
      "learning_rate": 7.171012734990904e-06,
      "loss": 0.1684,
      "step": 2800
    },
    {
      "epoch": 0.8565797453001819,
      "grad_norm": 7.043132781982422,
      "learning_rate": 7.145744895896504e-06,
      "loss": 0.1781,
      "step": 2825
    },
    {
      "epoch": 0.8641600970285022,
      "grad_norm": 5.09641695022583,
      "learning_rate": 7.120477056802103e-06,
      "loss": 0.1778,
      "step": 2850
    },
    {
      "epoch": 0.8717404487568223,
      "grad_norm": 6.9336748123168945,
      "learning_rate": 7.095209217707703e-06,
      "loss": 0.1607,
      "step": 2875
    },
    {
      "epoch": 0.8793208004851425,
      "grad_norm": 5.951363563537598,
      "learning_rate": 7.069941378613302e-06,
      "loss": 0.1793,
      "step": 2900
    },
    {
      "epoch": 0.8869011522134627,
      "grad_norm": 5.363964557647705,
      "learning_rate": 7.044673539518901e-06,
      "loss": 0.1742,
      "step": 2925
    },
    {
      "epoch": 0.8944815039417829,
      "grad_norm": 8.336238861083984,
      "learning_rate": 7.0194057004245e-06,
      "loss": 0.2014,
      "step": 2950
    },
    {
      "epoch": 0.9020618556701031,
      "grad_norm": 6.5310587882995605,
      "learning_rate": 6.9941378613300994e-06,
      "loss": 0.2074,
      "step": 2975
    },
    {
      "epoch": 0.9096422073984233,
      "grad_norm": 9.294402122497559,
      "learning_rate": 6.968870022235699e-06,
      "loss": 0.1747,
      "step": 3000
    },
    {
      "epoch": 0.9096422073984233,
      "eval_cer": 4.88665599873825,
      "eval_loss": 0.21674878895282745,
      "eval_runtime": 4171.9101,
      "eval_samples_per_second": 2.446,
      "eval_ser": 62.40321474076252,
      "eval_steps_per_second": 0.611,
      "eval_wer": 18.098512212401715,
      "step": 3000
    },
    {
      "epoch": 0.9172225591267434,
      "grad_norm": 8.7633056640625,
      "learning_rate": 6.943602183141298e-06,
      "loss": 0.1575,
      "step": 3025
    },
    {
      "epoch": 0.9248029108550637,
      "grad_norm": 13.642629623413086,
      "learning_rate": 6.918334344046897e-06,
      "loss": 0.1817,
      "step": 3050
    },
    {
      "epoch": 0.9323832625833839,
      "grad_norm": 6.795938968658447,
      "learning_rate": 6.893066504952497e-06,
      "loss": 0.1887,
      "step": 3075
    },
    {
      "epoch": 0.939963614311704,
      "grad_norm": 7.764118671417236,
      "learning_rate": 6.867798665858096e-06,
      "loss": 0.1611,
      "step": 3100
    },
    {
      "epoch": 0.9475439660400242,
      "grad_norm": 7.532794952392578,
      "learning_rate": 6.842530826763696e-06,
      "loss": 0.1869,
      "step": 3125
    },
    {
      "epoch": 0.9551243177683445,
      "grad_norm": 6.741429328918457,
      "learning_rate": 6.817262987669295e-06,
      "loss": 0.1846,
      "step": 3150
    },
    {
      "epoch": 0.9627046694966647,
      "grad_norm": 9.013187408447266,
      "learning_rate": 6.791995148574895e-06,
      "loss": 0.1649,
      "step": 3175
    },
    {
      "epoch": 0.9702850212249848,
      "grad_norm": 11.595109939575195,
      "learning_rate": 6.7667273094804945e-06,
      "loss": 0.1818,
      "step": 3200
    },
    {
      "epoch": 0.977865372953305,
      "grad_norm": 12.244895935058594,
      "learning_rate": 6.741459470386094e-06,
      "loss": 0.1906,
      "step": 3225
    },
    {
      "epoch": 0.9854457246816253,
      "grad_norm": 6.988622188568115,
      "learning_rate": 6.716191631291692e-06,
      "loss": 0.173,
      "step": 3250
    },
    {
      "epoch": 0.9930260764099454,
      "grad_norm": 4.436202526092529,
      "learning_rate": 6.690923792197291e-06,
      "loss": 0.1838,
      "step": 3275
    },
    {
      "epoch": 1.0006064281382656,
      "grad_norm": 6.243381977081299,
      "learning_rate": 6.665655953102891e-06,
      "loss": 0.1718,
      "step": 3300
    },
    {
      "epoch": 1.0081867798665858,
      "grad_norm": 5.074841022491455,
      "learning_rate": 6.6403881140084905e-06,
      "loss": 0.076,
      "step": 3325
    },
    {
      "epoch": 1.015767131594906,
      "grad_norm": 4.297674655914307,
      "learning_rate": 6.61512027491409e-06,
      "loss": 0.078,
      "step": 3350
    },
    {
      "epoch": 1.0233474833232261,
      "grad_norm": 3.33673095703125,
      "learning_rate": 6.589852435819689e-06,
      "loss": 0.0774,
      "step": 3375
    },
    {
      "epoch": 1.0309278350515463,
      "grad_norm": 7.982945442199707,
      "learning_rate": 6.564584596725288e-06,
      "loss": 0.0856,
      "step": 3400
    },
    {
      "epoch": 1.0385081867798667,
      "grad_norm": 5.4341888427734375,
      "learning_rate": 6.539316757630888e-06,
      "loss": 0.0752,
      "step": 3425
    },
    {
      "epoch": 1.0460885385081868,
      "grad_norm": 3.646911144256592,
      "learning_rate": 6.514048918536487e-06,
      "loss": 0.0657,
      "step": 3450
    },
    {
      "epoch": 1.053668890236507,
      "grad_norm": 2.50636625289917,
      "learning_rate": 6.488781079442087e-06,
      "loss": 0.0678,
      "step": 3475
    },
    {
      "epoch": 1.0612492419648272,
      "grad_norm": 2.662381172180176,
      "learning_rate": 6.4635132403476865e-06,
      "loss": 0.077,
      "step": 3500
    },
    {
      "epoch": 1.0612492419648272,
      "eval_cer": 4.527188431321651,
      "eval_loss": 0.2141607105731964,
      "eval_runtime": 4130.8959,
      "eval_samples_per_second": 2.47,
      "eval_ser": 60.59982358129962,
      "eval_steps_per_second": 0.618,
      "eval_wer": 17.200729375965437,
      "step": 3500
    },
    {
      "epoch": 1.0688295936931473,
      "grad_norm": 2.4962751865386963,
      "learning_rate": 6.438245401253286e-06,
      "loss": 0.0885,
      "step": 3525
    },
    {
      "epoch": 1.0764099454214675,
      "grad_norm": 5.805286884307861,
      "learning_rate": 6.4129775621588856e-06,
      "loss": 0.0644,
      "step": 3550
    },
    {
      "epoch": 1.0839902971497877,
      "grad_norm": 3.763657331466675,
      "learning_rate": 6.387709723064483e-06,
      "loss": 0.0647,
      "step": 3575
    },
    {
      "epoch": 1.0915706488781078,
      "grad_norm": 3.8962442874908447,
      "learning_rate": 6.362441883970083e-06,
      "loss": 0.0759,
      "step": 3600
    },
    {
      "epoch": 1.0991510006064282,
      "grad_norm": 6.139822483062744,
      "learning_rate": 6.3371740448756825e-06,
      "loss": 0.0707,
      "step": 3625
    },
    {
      "epoch": 1.1067313523347484,
      "grad_norm": 4.11650276184082,
      "learning_rate": 6.311906205781282e-06,
      "loss": 0.0699,
      "step": 3650
    },
    {
      "epoch": 1.1143117040630686,
      "grad_norm": 6.533496379852295,
      "learning_rate": 6.286638366686882e-06,
      "loss": 0.069,
      "step": 3675
    },
    {
      "epoch": 1.1218920557913887,
      "grad_norm": 3.1129257678985596,
      "learning_rate": 6.261370527592481e-06,
      "loss": 0.0707,
      "step": 3700
    },
    {
      "epoch": 1.129472407519709,
      "grad_norm": 3.159837245941162,
      "learning_rate": 6.23610268849808e-06,
      "loss": 0.0766,
      "step": 3725
    },
    {
      "epoch": 1.137052759248029,
      "grad_norm": 6.998505592346191,
      "learning_rate": 6.210834849403679e-06,
      "loss": 0.0799,
      "step": 3750
    },
    {
      "epoch": 1.1446331109763492,
      "grad_norm": 5.051515102386475,
      "learning_rate": 6.185567010309279e-06,
      "loss": 0.08,
      "step": 3775
    },
    {
      "epoch": 1.1522134627046694,
      "grad_norm": 4.849501132965088,
      "learning_rate": 6.1602991712148785e-06,
      "loss": 0.0694,
      "step": 3800
    },
    {
      "epoch": 1.1597938144329896,
      "grad_norm": 4.301568984985352,
      "learning_rate": 6.135031332120478e-06,
      "loss": 0.0801,
      "step": 3825
    },
    {
      "epoch": 1.16737416616131,
      "grad_norm": 7.137290954589844,
      "learning_rate": 6.1097634930260776e-06,
      "loss": 0.0707,
      "step": 3850
    },
    {
      "epoch": 1.1749545178896301,
      "grad_norm": 1.7661166191101074,
      "learning_rate": 6.084495653931675e-06,
      "loss": 0.0671,
      "step": 3875
    },
    {
      "epoch": 1.1825348696179503,
      "grad_norm": 3.448512077331543,
      "learning_rate": 6.059227814837275e-06,
      "loss": 0.0689,
      "step": 3900
    },
    {
      "epoch": 1.1901152213462705,
      "grad_norm": 3.7001595497131348,
      "learning_rate": 6.0339599757428745e-06,
      "loss": 0.0764,
      "step": 3925
    },
    {
      "epoch": 1.1976955730745906,
      "grad_norm": 6.902029037475586,
      "learning_rate": 6.008692136648474e-06,
      "loss": 0.0661,
      "step": 3950
    },
    {
      "epoch": 1.2052759248029108,
      "grad_norm": 4.167758464813232,
      "learning_rate": 5.983424297554074e-06,
      "loss": 0.071,
      "step": 3975
    },
    {
      "epoch": 1.212856276531231,
      "grad_norm": 3.972062110900879,
      "learning_rate": 5.958156458459673e-06,
      "loss": 0.0839,
      "step": 4000
    },
    {
      "epoch": 1.212856276531231,
      "eval_cer": 4.462786563667489,
      "eval_loss": 0.21264785528182983,
      "eval_runtime": 4149.8292,
      "eval_samples_per_second": 2.459,
      "eval_ser": 60.8742526707831,
      "eval_steps_per_second": 0.615,
      "eval_wer": 17.160079441586042,
      "step": 4000
    },
    {
      "epoch": 1.2204366282595513,
      "grad_norm": 4.492404937744141,
      "learning_rate": 5.932888619365273e-06,
      "loss": 0.0894,
      "step": 4025
    },
    {
      "epoch": 1.2280169799878715,
      "grad_norm": 3.026043176651001,
      "learning_rate": 5.907620780270872e-06,
      "loss": 0.072,
      "step": 4050
    },
    {
      "epoch": 1.2355973317161917,
      "grad_norm": 2.404448986053467,
      "learning_rate": 5.882352941176471e-06,
      "loss": 0.0769,
      "step": 4075
    },
    {
      "epoch": 1.2431776834445118,
      "grad_norm": 5.609960556030273,
      "learning_rate": 5.8570851020820705e-06,
      "loss": 0.0617,
      "step": 4100
    },
    {
      "epoch": 1.250758035172832,
      "grad_norm": 4.348845481872559,
      "learning_rate": 5.83181726298767e-06,
      "loss": 0.0761,
      "step": 4125
    },
    {
      "epoch": 1.2583383869011522,
      "grad_norm": 6.77551794052124,
      "learning_rate": 5.8065494238932696e-06,
      "loss": 0.0776,
      "step": 4150
    },
    {
      "epoch": 1.2659187386294724,
      "grad_norm": 5.368580341339111,
      "learning_rate": 5.781281584798869e-06,
      "loss": 0.0677,
      "step": 4175
    },
    {
      "epoch": 1.2734990903577925,
      "grad_norm": 6.325013637542725,
      "learning_rate": 5.756013745704467e-06,
      "loss": 0.0858,
      "step": 4200
    },
    {
      "epoch": 1.2810794420861127,
      "grad_norm": 7.039133548736572,
      "learning_rate": 5.7307459066100665e-06,
      "loss": 0.0917,
      "step": 4225
    },
    {
      "epoch": 1.2886597938144329,
      "grad_norm": 6.305510520935059,
      "learning_rate": 5.705478067515666e-06,
      "loss": 0.0815,
      "step": 4250
    },
    {
      "epoch": 1.2962401455427532,
      "grad_norm": 4.5553789138793945,
      "learning_rate": 5.680210228421266e-06,
      "loss": 0.0746,
      "step": 4275
    },
    {
      "epoch": 1.3038204972710734,
      "grad_norm": 14.23348331451416,
      "learning_rate": 5.654942389326865e-06,
      "loss": 0.0635,
      "step": 4300
    },
    {
      "epoch": 1.3114008489993936,
      "grad_norm": 4.656774997711182,
      "learning_rate": 5.629674550232465e-06,
      "loss": 0.0905,
      "step": 4325
    },
    {
      "epoch": 1.3189812007277137,
      "grad_norm": 7.67761754989624,
      "learning_rate": 5.604406711138064e-06,
      "loss": 0.0699,
      "step": 4350
    },
    {
      "epoch": 1.326561552456034,
      "grad_norm": 3.044973850250244,
      "learning_rate": 5.579138872043664e-06,
      "loss": 0.0684,
      "step": 4375
    },
    {
      "epoch": 1.334141904184354,
      "grad_norm": 12.772285461425781,
      "learning_rate": 5.553871032949263e-06,
      "loss": 0.0832,
      "step": 4400
    },
    {
      "epoch": 1.3417222559126745,
      "grad_norm": 4.697045803070068,
      "learning_rate": 5.528603193854862e-06,
      "loss": 0.0841,
      "step": 4425
    },
    {
      "epoch": 1.3493026076409946,
      "grad_norm": 3.967209815979004,
      "learning_rate": 5.5033353547604615e-06,
      "loss": 0.0653,
      "step": 4450
    },
    {
      "epoch": 1.3568829593693148,
      "grad_norm": 2.5490405559539795,
      "learning_rate": 5.478067515666061e-06,
      "loss": 0.0805,
      "step": 4475
    },
    {
      "epoch": 1.364463311097635,
      "grad_norm": 7.244118690490723,
      "learning_rate": 5.45279967657166e-06,
      "loss": 0.0888,
      "step": 4500
    },
    {
      "epoch": 1.364463311097635,
      "eval_cer": 4.486444392601671,
      "eval_loss": 0.2091592252254486,
      "eval_runtime": 4167.1809,
      "eval_samples_per_second": 2.448,
      "eval_ser": 60.394001764187,
      "eval_steps_per_second": 0.612,
      "eval_wer": 17.352876273214015,
      "step": 4500
    },
    {
      "epoch": 1.3720436628259551,
      "grad_norm": 3.822779655456543,
      "learning_rate": 5.427531837477259e-06,
      "loss": 0.0703,
      "step": 4525
    },
    {
      "epoch": 1.3796240145542753,
      "grad_norm": 4.564328193664551,
      "learning_rate": 5.402263998382858e-06,
      "loss": 0.0608,
      "step": 4550
    },
    {
      "epoch": 1.3872043662825955,
      "grad_norm": 1.9041502475738525,
      "learning_rate": 5.3769961592884576e-06,
      "loss": 0.0633,
      "step": 4575
    },
    {
      "epoch": 1.3947847180109156,
      "grad_norm": 3.6892504692077637,
      "learning_rate": 5.351728320194057e-06,
      "loss": 0.0703,
      "step": 4600
    },
    {
      "epoch": 1.4023650697392358,
      "grad_norm": 1.5316110849380493,
      "learning_rate": 5.326460481099657e-06,
      "loss": 0.0699,
      "step": 4625
    },
    {
      "epoch": 1.409945421467556,
      "grad_norm": 4.096865653991699,
      "learning_rate": 5.301192642005256e-06,
      "loss": 0.0713,
      "step": 4650
    },
    {
      "epoch": 1.4175257731958764,
      "grad_norm": 4.801632404327393,
      "learning_rate": 5.275924802910856e-06,
      "loss": 0.0678,
      "step": 4675
    },
    {
      "epoch": 1.4251061249241965,
      "grad_norm": 7.240773677825928,
      "learning_rate": 5.250656963816455e-06,
      "loss": 0.0683,
      "step": 4700
    },
    {
      "epoch": 1.4326864766525167,
      "grad_norm": 5.203670978546143,
      "learning_rate": 5.225389124722055e-06,
      "loss": 0.087,
      "step": 4725
    },
    {
      "epoch": 1.4402668283808369,
      "grad_norm": 5.950301170349121,
      "learning_rate": 5.2001212856276535e-06,
      "loss": 0.0824,
      "step": 4750
    },
    {
      "epoch": 1.447847180109157,
      "grad_norm": 8.965677261352539,
      "learning_rate": 5.174853446533253e-06,
      "loss": 0.0705,
      "step": 4775
    },
    {
      "epoch": 1.4554275318374772,
      "grad_norm": 4.188201427459717,
      "learning_rate": 5.149585607438852e-06,
      "loss": 0.0605,
      "step": 4800
    },
    {
      "epoch": 1.4630078835657976,
      "grad_norm": 5.050042152404785,
      "learning_rate": 5.124317768344451e-06,
      "loss": 0.0743,
      "step": 4825
    },
    {
      "epoch": 1.4705882352941178,
      "grad_norm": 8.579322814941406,
      "learning_rate": 5.099049929250051e-06,
      "loss": 0.0725,
      "step": 4850
    },
    {
      "epoch": 1.478168587022438,
      "grad_norm": 3.81740140914917,
      "learning_rate": 5.07378209015565e-06,
      "loss": 0.0585,
      "step": 4875
    },
    {
      "epoch": 1.485748938750758,
      "grad_norm": 2.082289218902588,
      "learning_rate": 5.048514251061249e-06,
      "loss": 0.0766,
      "step": 4900
    },
    {
      "epoch": 1.4933292904790783,
      "grad_norm": 6.351588249206543,
      "learning_rate": 5.023246411966849e-06,
      "loss": 0.0746,
      "step": 4925
    },
    {
      "epoch": 1.5009096422073984,
      "grad_norm": 3.2933731079101562,
      "learning_rate": 4.997978572872448e-06,
      "loss": 0.0762,
      "step": 4950
    },
    {
      "epoch": 1.5084899939357186,
      "grad_norm": 2.749415874481201,
      "learning_rate": 4.972710733778048e-06,
      "loss": 0.0872,
      "step": 4975
    },
    {
      "epoch": 1.5160703456640388,
      "grad_norm": 3.414822578430176,
      "learning_rate": 4.947442894683647e-06,
      "loss": 0.069,
      "step": 5000
    },
    {
      "epoch": 1.5160703456640388,
      "eval_cer": 4.46672953515652,
      "eval_loss": 0.2117619663476944,
      "eval_runtime": 4150.2171,
      "eval_samples_per_second": 2.458,
      "eval_ser": 60.15877683034402,
      "eval_steps_per_second": 0.615,
      "eval_wer": 17.157756588192935,
      "step": 5000
    },
    {
      "epoch": 1.523650697392359,
      "grad_norm": 3.9905097484588623,
      "learning_rate": 4.922175055589247e-06,
      "loss": 0.0647,
      "step": 5025
    },
    {
      "epoch": 1.531231049120679,
      "grad_norm": 6.00367546081543,
      "learning_rate": 4.8969072164948455e-06,
      "loss": 0.0609,
      "step": 5050
    },
    {
      "epoch": 1.5388114008489993,
      "grad_norm": 3.5311403274536133,
      "learning_rate": 4.871639377400445e-06,
      "loss": 0.0682,
      "step": 5075
    },
    {
      "epoch": 1.5463917525773194,
      "grad_norm": 4.727973461151123,
      "learning_rate": 4.846371538306045e-06,
      "loss": 0.0649,
      "step": 5100
    },
    {
      "epoch": 1.5539721043056398,
      "grad_norm": 5.63254976272583,
      "learning_rate": 4.821103699211644e-06,
      "loss": 0.0774,
      "step": 5125
    },
    {
      "epoch": 1.56155245603396,
      "grad_norm": 2.976087808609009,
      "learning_rate": 4.795835860117244e-06,
      "loss": 0.0711,
      "step": 5150
    },
    {
      "epoch": 1.5691328077622801,
      "grad_norm": 4.30213737487793,
      "learning_rate": 4.770568021022842e-06,
      "loss": 0.0726,
      "step": 5175
    },
    {
      "epoch": 1.5767131594906003,
      "grad_norm": 4.121530055999756,
      "learning_rate": 4.745300181928442e-06,
      "loss": 0.0676,
      "step": 5200
    },
    {
      "epoch": 1.5842935112189207,
      "grad_norm": 3.224336624145508,
      "learning_rate": 4.7200323428340415e-06,
      "loss": 0.0708,
      "step": 5225
    },
    {
      "epoch": 1.5918738629472409,
      "grad_norm": 4.8691935539245605,
      "learning_rate": 4.69476450373964e-06,
      "loss": 0.0829,
      "step": 5250
    },
    {
      "epoch": 1.599454214675561,
      "grad_norm": 5.627283573150635,
      "learning_rate": 4.66949666464524e-06,
      "loss": 0.0676,
      "step": 5275
    },
    {
      "epoch": 1.6070345664038812,
      "grad_norm": 3.1544229984283447,
      "learning_rate": 4.644228825550839e-06,
      "loss": 0.0699,
      "step": 5300
    },
    {
      "epoch": 1.6146149181322014,
      "grad_norm": 3.1078245639801025,
      "learning_rate": 4.618960986456439e-06,
      "loss": 0.0791,
      "step": 5325
    },
    {
      "epoch": 1.6221952698605215,
      "grad_norm": 3.3688204288482666,
      "learning_rate": 4.5936931473620375e-06,
      "loss": 0.0688,
      "step": 5350
    },
    {
      "epoch": 1.6297756215888417,
      "grad_norm": 4.0804524421691895,
      "learning_rate": 4.568425308267637e-06,
      "loss": 0.0778,
      "step": 5375
    },
    {
      "epoch": 1.6373559733171619,
      "grad_norm": 3.7207510471343994,
      "learning_rate": 4.543157469173237e-06,
      "loss": 0.0642,
      "step": 5400
    },
    {
      "epoch": 1.644936325045482,
      "grad_norm": 4.2266316413879395,
      "learning_rate": 4.517889630078836e-06,
      "loss": 0.0736,
      "step": 5425
    },
    {
      "epoch": 1.6525166767738022,
      "grad_norm": 6.174389839172363,
      "learning_rate": 4.492621790984436e-06,
      "loss": 0.0825,
      "step": 5450
    },
    {
      "epoch": 1.6600970285021224,
      "grad_norm": 5.419666290283203,
      "learning_rate": 4.467353951890035e-06,
      "loss": 0.0708,
      "step": 5475
    },
    {
      "epoch": 1.6676773802304425,
      "grad_norm": 3.7497785091400146,
      "learning_rate": 4.442086112795634e-06,
      "loss": 0.0609,
      "step": 5500
    },
    {
      "epoch": 1.6676773802304425,
      "eval_cer": 4.4297641774468595,
      "eval_loss": 0.20770902931690216,
      "eval_runtime": 4161.696,
      "eval_samples_per_second": 2.452,
      "eval_ser": 59.33548956189356,
      "eval_steps_per_second": 0.613,
      "eval_wer": 16.85462422039233,
      "step": 5500
    },
    {
      "epoch": 1.675257731958763,
      "grad_norm": 2.9338574409484863,
      "learning_rate": 4.4168182737012335e-06,
      "loss": 0.0667,
      "step": 5525
    },
    {
      "epoch": 1.682838083687083,
      "grad_norm": 2.5198843479156494,
      "learning_rate": 4.391550434606833e-06,
      "loss": 0.0676,
      "step": 5550
    },
    {
      "epoch": 1.6904184354154033,
      "grad_norm": 5.370730876922607,
      "learning_rate": 4.366282595512433e-06,
      "loss": 0.0786,
      "step": 5575
    },
    {
      "epoch": 1.6979987871437234,
      "grad_norm": 3.7155959606170654,
      "learning_rate": 4.341014756418031e-06,
      "loss": 0.0713,
      "step": 5600
    },
    {
      "epoch": 1.7055791388720438,
      "grad_norm": 5.575277805328369,
      "learning_rate": 4.315746917323631e-06,
      "loss": 0.0708,
      "step": 5625
    },
    {
      "epoch": 1.713159490600364,
      "grad_norm": 2.6846461296081543,
      "learning_rate": 4.2904790782292295e-06,
      "loss": 0.0681,
      "step": 5650
    },
    {
      "epoch": 1.7207398423286842,
      "grad_norm": 6.605522632598877,
      "learning_rate": 4.265211239134829e-06,
      "loss": 0.0604,
      "step": 5675
    },
    {
      "epoch": 1.7283201940570043,
      "grad_norm": 9.453361511230469,
      "learning_rate": 4.239943400040429e-06,
      "loss": 0.0795,
      "step": 5700
    },
    {
      "epoch": 1.7359005457853245,
      "grad_norm": 3.2720046043395996,
      "learning_rate": 4.214675560946028e-06,
      "loss": 0.0717,
      "step": 5725
    },
    {
      "epoch": 1.7434808975136447,
      "grad_norm": 4.286426544189453,
      "learning_rate": 4.189407721851628e-06,
      "loss": 0.0721,
      "step": 5750
    },
    {
      "epoch": 1.7510612492419648,
      "grad_norm": 4.679434776306152,
      "learning_rate": 4.164139882757227e-06,
      "loss": 0.0718,
      "step": 5775
    },
    {
      "epoch": 1.758641600970285,
      "grad_norm": 4.515007972717285,
      "learning_rate": 4.138872043662826e-06,
      "loss": 0.0665,
      "step": 5800
    },
    {
      "epoch": 1.7662219526986052,
      "grad_norm": 4.324512958526611,
      "learning_rate": 4.1136042045684255e-06,
      "loss": 0.0692,
      "step": 5825
    },
    {
      "epoch": 1.7738023044269253,
      "grad_norm": 4.820830821990967,
      "learning_rate": 4.088336365474025e-06,
      "loss": 0.058,
      "step": 5850
    },
    {
      "epoch": 1.7813826561552455,
      "grad_norm": 4.152068614959717,
      "learning_rate": 4.0630685263796246e-06,
      "loss": 0.0596,
      "step": 5875
    },
    {
      "epoch": 1.7889630078835657,
      "grad_norm": 7.0391082763671875,
      "learning_rate": 4.037800687285224e-06,
      "loss": 0.0638,
      "step": 5900
    },
    {
      "epoch": 1.7965433596118858,
      "grad_norm": 4.026524066925049,
      "learning_rate": 4.012532848190823e-06,
      "loss": 0.0704,
      "step": 5925
    },
    {
      "epoch": 1.8041237113402062,
      "grad_norm": 5.237268447875977,
      "learning_rate": 3.987265009096422e-06,
      "loss": 0.0721,
      "step": 5950
    },
    {
      "epoch": 1.8117040630685264,
      "grad_norm": 3.4499592781066895,
      "learning_rate": 3.961997170002022e-06,
      "loss": 0.0705,
      "step": 5975
    },
    {
      "epoch": 1.8192844147968465,
      "grad_norm": 5.843986511230469,
      "learning_rate": 3.936729330907621e-06,
      "loss": 0.0721,
      "step": 6000
    },
    {
      "epoch": 1.8192844147968465,
      "eval_cer": 4.3441688380391605,
      "eval_loss": 0.20601269602775574,
      "eval_runtime": 4156.9236,
      "eval_samples_per_second": 2.454,
      "eval_ser": 58.659217877094974,
      "eval_steps_per_second": 0.614,
      "eval_wer": 16.552653279288275,
      "step": 6000
    },
    {
      "epoch": 1.8268647665251667,
      "grad_norm": 4.0732831954956055,
      "learning_rate": 3.91146149181322e-06,
      "loss": 0.0694,
      "step": 6025
    },
    {
      "epoch": 1.834445118253487,
      "grad_norm": 3.2202863693237305,
      "learning_rate": 3.88619365271882e-06,
      "loss": 0.0724,
      "step": 6050
    },
    {
      "epoch": 1.8420254699818073,
      "grad_norm": 5.942609786987305,
      "learning_rate": 3.860925813624419e-06,
      "loss": 0.0717,
      "step": 6075
    },
    {
      "epoch": 1.8496058217101274,
      "grad_norm": 4.279210090637207,
      "learning_rate": 3.835657974530019e-06,
      "loss": 0.0715,
      "step": 6100
    },
    {
      "epoch": 1.8571861734384476,
      "grad_norm": 3.8644626140594482,
      "learning_rate": 3.8103901354356175e-06,
      "loss": 0.0739,
      "step": 6125
    },
    {
      "epoch": 1.8647665251667678,
      "grad_norm": 7.767407417297363,
      "learning_rate": 3.785122296341217e-06,
      "loss": 0.0689,
      "step": 6150
    },
    {
      "epoch": 1.872346876895088,
      "grad_norm": 6.898746967315674,
      "learning_rate": 3.7598544572468166e-06,
      "loss": 0.0834,
      "step": 6175
    },
    {
      "epoch": 1.879927228623408,
      "grad_norm": 3.7618629932403564,
      "learning_rate": 3.734586618152416e-06,
      "loss": 0.067,
      "step": 6200
    },
    {
      "epoch": 1.8875075803517283,
      "grad_norm": 5.345553398132324,
      "learning_rate": 3.7093187790580152e-06,
      "loss": 0.0807,
      "step": 6225
    },
    {
      "epoch": 1.8950879320800484,
      "grad_norm": 1.9484589099884033,
      "learning_rate": 3.6840509399636148e-06,
      "loss": 0.0577,
      "step": 6250
    },
    {
      "epoch": 1.9026682838083686,
      "grad_norm": 8.23530101776123,
      "learning_rate": 3.658783100869214e-06,
      "loss": 0.0668,
      "step": 6275
    },
    {
      "epoch": 1.9102486355366888,
      "grad_norm": 3.302901029586792,
      "learning_rate": 3.633515261774813e-06,
      "loss": 0.0712,
      "step": 6300
    },
    {
      "epoch": 1.917828987265009,
      "grad_norm": 6.6083855628967285,
      "learning_rate": 3.6082474226804126e-06,
      "loss": 0.0784,
      "step": 6325
    },
    {
      "epoch": 1.9254093389933293,
      "grad_norm": 4.849329948425293,
      "learning_rate": 3.582979583586012e-06,
      "loss": 0.0578,
      "step": 6350
    },
    {
      "epoch": 1.9329896907216495,
      "grad_norm": 3.2557804584503174,
      "learning_rate": 3.5577117444916117e-06,
      "loss": 0.062,
      "step": 6375
    },
    {
      "epoch": 1.9405700424499697,
      "grad_norm": 4.66004753112793,
      "learning_rate": 3.5324439053972108e-06,
      "loss": 0.0652,
      "step": 6400
    },
    {
      "epoch": 1.9481503941782898,
      "grad_norm": 3.7304911613464355,
      "learning_rate": 3.50717606630281e-06,
      "loss": 0.0733,
      "step": 6425
    },
    {
      "epoch": 1.9557307459066102,
      "grad_norm": 6.284624099731445,
      "learning_rate": 3.4819082272084094e-06,
      "loss": 0.0749,
      "step": 6450
    },
    {
      "epoch": 1.9633110976349304,
      "grad_norm": 5.241745948791504,
      "learning_rate": 3.4566403881140086e-06,
      "loss": 0.072,
      "step": 6475
    },
    {
      "epoch": 1.9708914493632506,
      "grad_norm": 4.935563087463379,
      "learning_rate": 3.431372549019608e-06,
      "loss": 0.0681,
      "step": 6500
    },
    {
      "epoch": 1.9708914493632506,
      "eval_cer": 4.328396952083039,
      "eval_loss": 0.20379287004470825,
      "eval_runtime": 4180.665,
      "eval_samples_per_second": 2.441,
      "eval_ser": 58.16916593158875,
      "eval_steps_per_second": 0.61,
      "eval_wer": 16.3575335942672,
      "step": 6500
    },
    {
      "epoch": 1.9784718010915707,
      "grad_norm": 1.918434739112854,
      "learning_rate": 3.4061047099252077e-06,
      "loss": 0.0314,
      "step": 6525
    },
    {
      "epoch": 1.986052152819891,
      "grad_norm": 4.31190299987793,
      "learning_rate": 3.380836870830807e-06,
      "loss": 0.0266,
      "step": 6550
    },
    {
      "epoch": 1.993632504548211,
      "grad_norm": 3.206461191177368,
      "learning_rate": 3.355569031736406e-06,
      "loss": 0.0286,
      "step": 6575
    },
    {
      "epoch": 2.0012128562765312,
      "grad_norm": 1.4688981771469116,
      "learning_rate": 3.3303011926420054e-06,
      "loss": 0.0254,
      "step": 6600
    },
    {
      "epoch": 2.0087932080048514,
      "grad_norm": 2.9358630180358887,
      "learning_rate": 3.305033353547605e-06,
      "loss": 0.0298,
      "step": 6625
    },
    {
      "epoch": 2.0163735597331716,
      "grad_norm": 2.1530213356018066,
      "learning_rate": 3.279765514453204e-06,
      "loss": 0.0311,
      "step": 6650
    },
    {
      "epoch": 2.0239539114614917,
      "grad_norm": 18.083703994750977,
      "learning_rate": 3.2544976753588036e-06,
      "loss": 0.0328,
      "step": 6675
    },
    {
      "epoch": 2.031534263189812,
      "grad_norm": 3.114184856414795,
      "learning_rate": 3.229229836264403e-06,
      "loss": 0.0278,
      "step": 6700
    },
    {
      "epoch": 2.039114614918132,
      "grad_norm": 6.7866106033325195,
      "learning_rate": 3.2039619971700027e-06,
      "loss": 0.0259,
      "step": 6725
    },
    {
      "epoch": 2.0466949666464522,
      "grad_norm": 2.4553844928741455,
      "learning_rate": 3.1786941580756014e-06,
      "loss": 0.0273,
      "step": 6750
    },
    {
      "epoch": 2.0542753183747724,
      "grad_norm": 3.798393726348877,
      "learning_rate": 3.153426318981201e-06,
      "loss": 0.0318,
      "step": 6775
    },
    {
      "epoch": 2.0618556701030926,
      "grad_norm": 6.048929691314697,
      "learning_rate": 3.1281584798868005e-06,
      "loss": 0.034,
      "step": 6800
    },
    {
      "epoch": 2.069436021831413,
      "grad_norm": 0.9682230949401855,
      "learning_rate": 3.1028906407923996e-06,
      "loss": 0.0355,
      "step": 6825
    },
    {
      "epoch": 2.0770163735597333,
      "grad_norm": 0.7081900835037231,
      "learning_rate": 3.077622801697999e-06,
      "loss": 0.0257,
      "step": 6850
    },
    {
      "epoch": 2.0845967252880535,
      "grad_norm": 3.3265202045440674,
      "learning_rate": 3.0523549626035987e-06,
      "loss": 0.0257,
      "step": 6875
    },
    {
      "epoch": 2.0921770770163737,
      "grad_norm": 1.3119407892227173,
      "learning_rate": 3.0270871235091974e-06,
      "loss": 0.0263,
      "step": 6900
    },
    {
      "epoch": 2.099757428744694,
      "grad_norm": 3.3000783920288086,
      "learning_rate": 3.001819284414797e-06,
      "loss": 0.0273,
      "step": 6925
    },
    {
      "epoch": 2.107337780473014,
      "grad_norm": 1.735640048980713,
      "learning_rate": 2.9765514453203965e-06,
      "loss": 0.0314,
      "step": 6950
    },
    {
      "epoch": 2.114918132201334,
      "grad_norm": 4.68803596496582,
      "learning_rate": 2.951283606225996e-06,
      "loss": 0.0338,
      "step": 6975
    },
    {
      "epoch": 2.1224984839296543,
      "grad_norm": 3.5388286113739014,
      "learning_rate": 2.926015767131595e-06,
      "loss": 0.0322,
      "step": 7000
    },
    {
      "epoch": 2.1224984839296543,
      "eval_cer": 4.27089528453468,
      "eval_loss": 0.21301127970218658,
      "eval_runtime": 3826.4362,
      "eval_samples_per_second": 2.666,
      "eval_ser": 57.777124375183774,
      "eval_steps_per_second": 0.667,
      "eval_wer": 16.280879432294633,
      "step": 7000
    },
    {
      "epoch": 2.1300788356579745,
      "grad_norm": 5.576068878173828,
      "learning_rate": 2.9007479280371947e-06,
      "loss": 0.0284,
      "step": 7025
    },
    {
      "epoch": 2.1376591873862947,
      "grad_norm": 1.70852792263031,
      "learning_rate": 2.8754800889427934e-06,
      "loss": 0.0247,
      "step": 7050
    },
    {
      "epoch": 2.145239539114615,
      "grad_norm": 3.2939512729644775,
      "learning_rate": 2.850212249848393e-06,
      "loss": 0.0252,
      "step": 7075
    },
    {
      "epoch": 2.152819890842935,
      "grad_norm": 0.9231658577919006,
      "learning_rate": 2.8249444107539925e-06,
      "loss": 0.0288,
      "step": 7100
    },
    {
      "epoch": 2.160400242571255,
      "grad_norm": 5.767349720001221,
      "learning_rate": 2.799676571659592e-06,
      "loss": 0.0294,
      "step": 7125
    },
    {
      "epoch": 2.1679805942995753,
      "grad_norm": 3.638855457305908,
      "learning_rate": 2.774408732565191e-06,
      "loss": 0.0269,
      "step": 7150
    },
    {
      "epoch": 2.1755609460278955,
      "grad_norm": 4.53891658782959,
      "learning_rate": 2.7491408934707907e-06,
      "loss": 0.0289,
      "step": 7175
    },
    {
      "epoch": 2.1831412977562157,
      "grad_norm": 0.5863853693008423,
      "learning_rate": 2.72387305437639e-06,
      "loss": 0.0247,
      "step": 7200
    },
    {
      "epoch": 2.1907216494845363,
      "grad_norm": 3.777498483657837,
      "learning_rate": 2.698605215281989e-06,
      "loss": 0.0234,
      "step": 7225
    },
    {
      "epoch": 2.1983020012128565,
      "grad_norm": 2.23175048828125,
      "learning_rate": 2.6733373761875885e-06,
      "loss": 0.0314,
      "step": 7250
    },
    {
      "epoch": 2.2058823529411766,
      "grad_norm": 3.1541337966918945,
      "learning_rate": 2.648069537093188e-06,
      "loss": 0.0311,
      "step": 7275
    },
    {
      "epoch": 2.213462704669497,
      "grad_norm": 2.6496334075927734,
      "learning_rate": 2.6228016979987876e-06,
      "loss": 0.0351,
      "step": 7300
    },
    {
      "epoch": 2.221043056397817,
      "grad_norm": 1.7922499179840088,
      "learning_rate": 2.5975338589043867e-06,
      "loss": 0.0211,
      "step": 7325
    },
    {
      "epoch": 2.228623408126137,
      "grad_norm": 5.5097222328186035,
      "learning_rate": 2.572266019809986e-06,
      "loss": 0.0302,
      "step": 7350
    },
    {
      "epoch": 2.2362037598544573,
      "grad_norm": 2.997144937515259,
      "learning_rate": 2.5469981807155854e-06,
      "loss": 0.0296,
      "step": 7375
    },
    {
      "epoch": 2.2437841115827775,
      "grad_norm": 1.941422462463379,
      "learning_rate": 2.5217303416211845e-06,
      "loss": 0.0244,
      "step": 7400
    },
    {
      "epoch": 2.2513644633110976,
      "grad_norm": 1.389489769935608,
      "learning_rate": 2.496462502526784e-06,
      "loss": 0.0308,
      "step": 7425
    },
    {
      "epoch": 2.258944815039418,
      "grad_norm": 1.393212914466858,
      "learning_rate": 2.4711946634323836e-06,
      "loss": 0.0262,
      "step": 7450
    },
    {
      "epoch": 2.266525166767738,
      "grad_norm": 2.6350386142730713,
      "learning_rate": 2.4459268243379827e-06,
      "loss": 0.0263,
      "step": 7475
    },
    {
      "epoch": 2.274105518496058,
      "grad_norm": 3.369213819503784,
      "learning_rate": 2.4206589852435823e-06,
      "loss": 0.0277,
      "step": 7500
    },
    {
      "epoch": 2.274105518496058,
      "eval_cer": 4.254301946185011,
      "eval_loss": 0.2150956392288208,
      "eval_runtime": 3817.2467,
      "eval_samples_per_second": 2.673,
      "eval_ser": 57.47329216896991,
      "eval_steps_per_second": 0.668,
      "eval_wer": 16.106665427811524,
      "step": 7500
    },
    {
      "epoch": 2.2816858702243783,
      "grad_norm": 2.3932759761810303,
      "learning_rate": 2.395391146149182e-06,
      "loss": 0.0321,
      "step": 7525
    },
    {
      "epoch": 2.2892662219526985,
      "grad_norm": 2.8787713050842285,
      "learning_rate": 2.370123307054781e-06,
      "loss": 0.0278,
      "step": 7550
    },
    {
      "epoch": 2.2968465736810186,
      "grad_norm": 2.570692539215088,
      "learning_rate": 2.34485546796038e-06,
      "loss": 0.0211,
      "step": 7575
    },
    {
      "epoch": 2.304426925409339,
      "grad_norm": 2.9834413528442383,
      "learning_rate": 2.3195876288659796e-06,
      "loss": 0.0308,
      "step": 7600
    },
    {
      "epoch": 2.312007277137659,
      "grad_norm": 3.4857709407806396,
      "learning_rate": 2.2943197897715787e-06,
      "loss": 0.0272,
      "step": 7625
    },
    {
      "epoch": 2.319587628865979,
      "grad_norm": 4.783783435821533,
      "learning_rate": 2.2690519506771783e-06,
      "loss": 0.0248,
      "step": 7650
    },
    {
      "epoch": 2.3271679805942997,
      "grad_norm": 3.023923873901367,
      "learning_rate": 2.243784111582778e-06,
      "loss": 0.0297,
      "step": 7675
    },
    {
      "epoch": 2.33474833232262,
      "grad_norm": 1.8888705968856812,
      "learning_rate": 2.218516272488377e-06,
      "loss": 0.0291,
      "step": 7700
    },
    {
      "epoch": 2.34232868405094,
      "grad_norm": 2.11702561378479,
      "learning_rate": 2.1932484333939765e-06,
      "loss": 0.0271,
      "step": 7725
    },
    {
      "epoch": 2.3499090357792602,
      "grad_norm": 2.8734652996063232,
      "learning_rate": 2.1679805942995756e-06,
      "loss": 0.0309,
      "step": 7750
    },
    {
      "epoch": 2.3574893875075804,
      "grad_norm": 1.2483388185501099,
      "learning_rate": 2.1427127552051747e-06,
      "loss": 0.0273,
      "step": 7775
    },
    {
      "epoch": 2.3650697392359006,
      "grad_norm": 1.2977802753448486,
      "learning_rate": 2.1174449161107743e-06,
      "loss": 0.0302,
      "step": 7800
    },
    {
      "epoch": 2.3726500909642207,
      "grad_norm": 3.370899200439453,
      "learning_rate": 2.092177077016374e-06,
      "loss": 0.0268,
      "step": 7825
    },
    {
      "epoch": 2.380230442692541,
      "grad_norm": 2.3007266521453857,
      "learning_rate": 2.066909237921973e-06,
      "loss": 0.0265,
      "step": 7850
    },
    {
      "epoch": 2.387810794420861,
      "grad_norm": 3.0940144062042236,
      "learning_rate": 2.0416413988275725e-06,
      "loss": 0.0221,
      "step": 7875
    },
    {
      "epoch": 2.3953911461491812,
      "grad_norm": 2.137110948562622,
      "learning_rate": 2.016373559733172e-06,
      "loss": 0.0273,
      "step": 7900
    },
    {
      "epoch": 2.4029714978775014,
      "grad_norm": 3.142803430557251,
      "learning_rate": 1.991105720638771e-06,
      "loss": 0.0309,
      "step": 7925
    },
    {
      "epoch": 2.4105518496058216,
      "grad_norm": 3.708592176437378,
      "learning_rate": 1.9658378815443703e-06,
      "loss": 0.0269,
      "step": 7950
    },
    {
      "epoch": 2.4181322013341418,
      "grad_norm": 2.705143451690674,
      "learning_rate": 1.94057004244997e-06,
      "loss": 0.0284,
      "step": 7975
    },
    {
      "epoch": 2.425712553062462,
      "grad_norm": 2.2267937660217285,
      "learning_rate": 1.9153022033555693e-06,
      "loss": 0.0249,
      "step": 8000
    },
    {
      "epoch": 2.425712553062462,
      "eval_cer": 4.251344717568238,
      "eval_loss": 0.21296031773090363,
      "eval_runtime": 3956.9891,
      "eval_samples_per_second": 2.578,
      "eval_ser": 57.463491130059786,
      "eval_steps_per_second": 0.645,
      "eval_wer": 16.07414548030801,
      "step": 8000
    },
    {
      "epoch": 2.4332929047907825,
      "grad_norm": 2.44812273979187,
      "learning_rate": 1.8900343642611685e-06,
      "loss": 0.0301,
      "step": 8025
    },
    {
      "epoch": 2.4408732565191027,
      "grad_norm": 3.240962028503418,
      "learning_rate": 1.864766525166768e-06,
      "loss": 0.0315,
      "step": 8050
    },
    {
      "epoch": 2.448453608247423,
      "grad_norm": 3.5091423988342285,
      "learning_rate": 1.8394986860723673e-06,
      "loss": 0.0218,
      "step": 8075
    },
    {
      "epoch": 2.456033959975743,
      "grad_norm": 2.2608911991119385,
      "learning_rate": 1.8142308469779665e-06,
      "loss": 0.026,
      "step": 8100
    },
    {
      "epoch": 2.463614311704063,
      "grad_norm": 2.7215895652770996,
      "learning_rate": 1.788963007883566e-06,
      "loss": 0.0323,
      "step": 8125
    },
    {
      "epoch": 2.4711946634323834,
      "grad_norm": 3.6915700435638428,
      "learning_rate": 1.7636951687891653e-06,
      "loss": 0.0397,
      "step": 8150
    },
    {
      "epoch": 2.4787750151607035,
      "grad_norm": 3.11942720413208,
      "learning_rate": 1.7384273296947647e-06,
      "loss": 0.032,
      "step": 8175
    },
    {
      "epoch": 2.4863553668890237,
      "grad_norm": 0.9445828795433044,
      "learning_rate": 1.713159490600364e-06,
      "loss": 0.0244,
      "step": 8200
    },
    {
      "epoch": 2.493935718617344,
      "grad_norm": 2.8565821647644043,
      "learning_rate": 1.6878916515059636e-06,
      "loss": 0.0283,
      "step": 8225
    },
    {
      "epoch": 2.501516070345664,
      "grad_norm": 0.4366835057735443,
      "learning_rate": 1.6626238124115627e-06,
      "loss": 0.0236,
      "step": 8250
    },
    {
      "epoch": 2.509096422073984,
      "grad_norm": 2.801251173019409,
      "learning_rate": 1.637355973317162e-06,
      "loss": 0.0278,
      "step": 8275
    },
    {
      "epoch": 2.5166767738023044,
      "grad_norm": 2.6153385639190674,
      "learning_rate": 1.6120881342227616e-06,
      "loss": 0.0274,
      "step": 8300
    },
    {
      "epoch": 2.5242571255306245,
      "grad_norm": 4.796694755554199,
      "learning_rate": 1.5868202951283607e-06,
      "loss": 0.023,
      "step": 8325
    },
    {
      "epoch": 2.5318374772589447,
      "grad_norm": 1.340336799621582,
      "learning_rate": 1.56155245603396e-06,
      "loss": 0.0266,
      "step": 8350
    },
    {
      "epoch": 2.539417828987265,
      "grad_norm": 2.282335042953491,
      "learning_rate": 1.5362846169395595e-06,
      "loss": 0.0221,
      "step": 8375
    },
    {
      "epoch": 2.546998180715585,
      "grad_norm": 0.9249318242073059,
      "learning_rate": 1.5110167778451587e-06,
      "loss": 0.0222,
      "step": 8400
    },
    {
      "epoch": 2.554578532443905,
      "grad_norm": 0.8499120473861694,
      "learning_rate": 1.4857489387507582e-06,
      "loss": 0.0219,
      "step": 8425
    },
    {
      "epoch": 2.5621588841722254,
      "grad_norm": 2.0028629302978516,
      "learning_rate": 1.4604810996563575e-06,
      "loss": 0.0352,
      "step": 8450
    },
    {
      "epoch": 2.5697392359005455,
      "grad_norm": 3.0861124992370605,
      "learning_rate": 1.4352132605619567e-06,
      "loss": 0.0255,
      "step": 8475
    },
    {
      "epoch": 2.5773195876288657,
      "grad_norm": 3.6241636276245117,
      "learning_rate": 1.4099454214675562e-06,
      "loss": 0.0234,
      "step": 8500
    },
    {
      "epoch": 2.5773195876288657,
      "eval_cer": 4.2832170704379,
      "eval_loss": 0.21502597630023956,
      "eval_runtime": 4007.4384,
      "eval_samples_per_second": 2.546,
      "eval_ser": 57.6693129471724,
      "eval_steps_per_second": 0.637,
      "eval_wer": 16.259973751756657,
      "step": 8500
    },
    {
      "epoch": 2.5848999393571863,
      "grad_norm": 3.110015869140625,
      "learning_rate": 1.3846775823731555e-06,
      "loss": 0.0286,
      "step": 8525
    },
    {
      "epoch": 2.5924802910855065,
      "grad_norm": 2.1836204528808594,
      "learning_rate": 1.3594097432787549e-06,
      "loss": 0.0283,
      "step": 8550
    },
    {
      "epoch": 2.6000606428138267,
      "grad_norm": 2.07279634475708,
      "learning_rate": 1.3341419041843542e-06,
      "loss": 0.0334,
      "step": 8575
    },
    {
      "epoch": 2.607640994542147,
      "grad_norm": 3.523423910140991,
      "learning_rate": 1.3088740650899538e-06,
      "loss": 0.0281,
      "step": 8600
    },
    {
      "epoch": 2.615221346270467,
      "grad_norm": 3.105787992477417,
      "learning_rate": 1.2836062259955529e-06,
      "loss": 0.0242,
      "step": 8625
    },
    {
      "epoch": 2.622801697998787,
      "grad_norm": 0.8383470177650452,
      "learning_rate": 1.2583383869011522e-06,
      "loss": 0.0319,
      "step": 8650
    },
    {
      "epoch": 2.6303820497271073,
      "grad_norm": 0.714475691318512,
      "learning_rate": 1.2330705478067518e-06,
      "loss": 0.0265,
      "step": 8675
    },
    {
      "epoch": 2.6379624014554275,
      "grad_norm": 1.9016785621643066,
      "learning_rate": 1.207802708712351e-06,
      "loss": 0.0293,
      "step": 8700
    },
    {
      "epoch": 2.6455427531837477,
      "grad_norm": 4.238561153411865,
      "learning_rate": 1.1825348696179504e-06,
      "loss": 0.0342,
      "step": 8725
    },
    {
      "epoch": 2.653123104912068,
      "grad_norm": 2.495922803878784,
      "learning_rate": 1.1572670305235498e-06,
      "loss": 0.0223,
      "step": 8750
    },
    {
      "epoch": 2.660703456640388,
      "grad_norm": 2.157742738723755,
      "learning_rate": 1.131999191429149e-06,
      "loss": 0.0265,
      "step": 8775
    },
    {
      "epoch": 2.668283808368708,
      "grad_norm": 1.7393815517425537,
      "learning_rate": 1.1067313523347484e-06,
      "loss": 0.0238,
      "step": 8800
    },
    {
      "epoch": 2.6758641600970288,
      "grad_norm": 5.519290924072266,
      "learning_rate": 1.0814635132403478e-06,
      "loss": 0.0217,
      "step": 8825
    },
    {
      "epoch": 2.683444511825349,
      "grad_norm": 1.7298911809921265,
      "learning_rate": 1.056195674145947e-06,
      "loss": 0.0247,
      "step": 8850
    },
    {
      "epoch": 2.691024863553669,
      "grad_norm": 1.5079689025878906,
      "learning_rate": 1.0309278350515464e-06,
      "loss": 0.0258,
      "step": 8875
    },
    {
      "epoch": 2.6986052152819893,
      "grad_norm": 2.1406021118164062,
      "learning_rate": 1.005659995957146e-06,
      "loss": 0.0231,
      "step": 8900
    },
    {
      "epoch": 2.7061855670103094,
      "grad_norm": 2.1180737018585205,
      "learning_rate": 9.80392156862745e-07,
      "loss": 0.0245,
      "step": 8925
    },
    {
      "epoch": 2.7137659187386296,
      "grad_norm": 2.1627533435821533,
      "learning_rate": 9.551243177683444e-07,
      "loss": 0.0264,
      "step": 8950
    },
    {
      "epoch": 2.7213462704669498,
      "grad_norm": 0.6703803539276123,
      "learning_rate": 9.29856478673944e-07,
      "loss": 0.0206,
      "step": 8975
    },
    {
      "epoch": 2.72892662219527,
      "grad_norm": 3.99841046333313,
      "learning_rate": 9.045886395795432e-07,
      "loss": 0.0264,
      "step": 9000
    },
    {
      "epoch": 2.72892662219527,
      "eval_cer": 4.264487955865006,
      "eval_loss": 0.21449251472949982,
      "eval_runtime": 3978.4945,
      "eval_samples_per_second": 2.565,
      "eval_ser": 57.630108791531896,
      "eval_steps_per_second": 0.641,
      "eval_wer": 16.115956841383955,
      "step": 9000
    },
    {
      "epoch": 2.73650697392359,
      "grad_norm": 4.727873802185059,
      "learning_rate": 8.793208004851426e-07,
      "loss": 0.0215,
      "step": 9025
    },
    {
      "epoch": 2.7440873256519103,
      "grad_norm": 4.042278289794922,
      "learning_rate": 8.54052961390742e-07,
      "loss": 0.0167,
      "step": 9050
    },
    {
      "epoch": 2.7516676773802304,
      "grad_norm": 3.561555862426758,
      "learning_rate": 8.287851222963413e-07,
      "loss": 0.0318,
      "step": 9075
    },
    {
      "epoch": 2.7592480291085506,
      "grad_norm": 2.2994942665100098,
      "learning_rate": 8.035172832019407e-07,
      "loss": 0.0263,
      "step": 9100
    },
    {
      "epoch": 2.7668283808368708,
      "grad_norm": 1.5986050367355347,
      "learning_rate": 7.7824944410754e-07,
      "loss": 0.0232,
      "step": 9125
    },
    {
      "epoch": 2.774408732565191,
      "grad_norm": 1.9084292650222778,
      "learning_rate": 7.529816050131393e-07,
      "loss": 0.0315,
      "step": 9150
    },
    {
      "epoch": 2.781989084293511,
      "grad_norm": 2.2224199771881104,
      "learning_rate": 7.277137659187387e-07,
      "loss": 0.0223,
      "step": 9175
    },
    {
      "epoch": 2.7895694360218313,
      "grad_norm": 4.254256248474121,
      "learning_rate": 7.024459268243381e-07,
      "loss": 0.0214,
      "step": 9200
    },
    {
      "epoch": 2.7971497877501514,
      "grad_norm": 0.8206575512886047,
      "learning_rate": 6.771780877299374e-07,
      "loss": 0.023,
      "step": 9225
    },
    {
      "epoch": 2.8047301394784716,
      "grad_norm": 1.8297322988510132,
      "learning_rate": 6.519102486355368e-07,
      "loss": 0.0246,
      "step": 9250
    },
    {
      "epoch": 2.8123104912067918,
      "grad_norm": 2.382608413696289,
      "learning_rate": 6.266424095411361e-07,
      "loss": 0.0317,
      "step": 9275
    },
    {
      "epoch": 2.819890842935112,
      "grad_norm": 2.400407075881958,
      "learning_rate": 6.013745704467355e-07,
      "loss": 0.0233,
      "step": 9300
    },
    {
      "epoch": 2.827471194663432,
      "grad_norm": 1.5915170907974243,
      "learning_rate": 5.761067313523347e-07,
      "loss": 0.0185,
      "step": 9325
    },
    {
      "epoch": 2.8350515463917527,
      "grad_norm": 0.9242422580718994,
      "learning_rate": 5.508388922579342e-07,
      "loss": 0.0157,
      "step": 9350
    },
    {
      "epoch": 2.842631898120073,
      "grad_norm": 2.4773428440093994,
      "learning_rate": 5.255710531635335e-07,
      "loss": 0.022,
      "step": 9375
    },
    {
      "epoch": 2.850212249848393,
      "grad_norm": 1.8242741823196411,
      "learning_rate": 5.003032140691328e-07,
      "loss": 0.0228,
      "step": 9400
    },
    {
      "epoch": 2.857792601576713,
      "grad_norm": 1.73279869556427,
      "learning_rate": 4.7503537497473216e-07,
      "loss": 0.0261,
      "step": 9425
    },
    {
      "epoch": 2.8653729533050334,
      "grad_norm": 5.991119384765625,
      "learning_rate": 4.4976753588033155e-07,
      "loss": 0.0288,
      "step": 9450
    },
    {
      "epoch": 2.8729533050333536,
      "grad_norm": 6.558225154876709,
      "learning_rate": 4.2449969678593093e-07,
      "loss": 0.031,
      "step": 9475
    },
    {
      "epoch": 2.8805336567616737,
      "grad_norm": 1.9604460000991821,
      "learning_rate": 3.992318576915302e-07,
      "loss": 0.0268,
      "step": 9500
    },
    {
      "epoch": 2.8805336567616737,
      "eval_cer": 4.232122731559215,
      "eval_loss": 0.21253998577594757,
      "eval_runtime": 3873.6751,
      "eval_samples_per_second": 2.634,
      "eval_ser": 57.522297363520536,
      "eval_steps_per_second": 0.659,
      "eval_wer": 16.040464106107944,
      "step": 9500
    },
    {
      "epoch": 2.888114008489994,
      "grad_norm": 0.7039766907691956,
      "learning_rate": 3.739640185971296e-07,
      "loss": 0.0336,
      "step": 9525
    },
    {
      "epoch": 2.895694360218314,
      "grad_norm": 3.8160667419433594,
      "learning_rate": 3.48696179502729e-07,
      "loss": 0.0295,
      "step": 9550
    },
    {
      "epoch": 2.9032747119466342,
      "grad_norm": 2.7596523761749268,
      "learning_rate": 3.234283404083283e-07,
      "loss": 0.0239,
      "step": 9575
    },
    {
      "epoch": 2.9108550636749544,
      "grad_norm": 1.7568339109420776,
      "learning_rate": 2.9816050131392765e-07,
      "loss": 0.0344,
      "step": 9600
    },
    {
      "epoch": 2.9184354154032746,
      "grad_norm": 4.06329870223999,
      "learning_rate": 2.7289266221952704e-07,
      "loss": 0.0246,
      "step": 9625
    },
    {
      "epoch": 2.926015767131595,
      "grad_norm": 5.080215930938721,
      "learning_rate": 2.4762482312512637e-07,
      "loss": 0.0279,
      "step": 9650
    },
    {
      "epoch": 2.9335961188599153,
      "grad_norm": 0.7092571258544922,
      "learning_rate": 2.2235698403072573e-07,
      "loss": 0.0287,
      "step": 9675
    },
    {
      "epoch": 2.9411764705882355,
      "grad_norm": 4.8213911056518555,
      "learning_rate": 1.9708914493632506e-07,
      "loss": 0.0219,
      "step": 9700
    },
    {
      "epoch": 2.9487568223165557,
      "grad_norm": 1.9876303672790527,
      "learning_rate": 1.7182130584192442e-07,
      "loss": 0.0215,
      "step": 9725
    },
    {
      "epoch": 2.956337174044876,
      "grad_norm": 4.70910120010376,
      "learning_rate": 1.4655346674752375e-07,
      "loss": 0.0308,
      "step": 9750
    },
    {
      "epoch": 2.963917525773196,
      "grad_norm": 3.7995145320892334,
      "learning_rate": 1.212856276531231e-07,
      "loss": 0.0231,
      "step": 9775
    },
    {
      "epoch": 2.971497877501516,
      "grad_norm": 1.9700251817703247,
      "learning_rate": 9.601778855872247e-08,
      "loss": 0.0249,
      "step": 9800
    },
    {
      "epoch": 2.9790782292298363,
      "grad_norm": 2.985954761505127,
      "learning_rate": 7.074994946432182e-08,
      "loss": 0.0245,
      "step": 9825
    },
    {
      "epoch": 2.9866585809581565,
      "grad_norm": 5.655767917633057,
      "learning_rate": 4.548211036992117e-08,
      "loss": 0.0287,
      "step": 9850
    },
    {
      "epoch": 2.9942389326864767,
      "grad_norm": 2.333650588989258,
      "learning_rate": 2.021427127552052e-08,
      "loss": 0.0215,
      "step": 9875
    }
  ],
  "logging_steps": 25,
  "max_steps": 9894,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 2.283605131272192e+19,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}