{
  "best_global_step": 5000,
  "best_metric": 20.124409102568798,
  "best_model_checkpoint": "./checkpoint-5000",
  "epoch": 140.18691588785046,
  "eval_steps": 1000,
  "global_step": 15000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.2336448598130841,
      "grad_norm": 18.17540740966797,
      "learning_rate": 1.575e-06,
      "loss": 2.461,
      "step": 25
    },
    {
      "epoch": 0.4672897196261682,
      "grad_norm": 6.570761203765869,
      "learning_rate": 3.4499999999999996e-06,
      "loss": 1.8802,
      "step": 50
    },
    {
      "epoch": 0.7009345794392523,
      "grad_norm": 4.048244953155518,
      "learning_rate": 5.324999999999999e-06,
      "loss": 1.3297,
      "step": 75
    },
    {
      "epoch": 0.9345794392523364,
      "grad_norm": 3.7204740047454834,
      "learning_rate": 7.2e-06,
      "loss": 1.0145,
      "step": 100
    },
    {
      "epoch": 1.1682242990654206,
      "grad_norm": 3.382382869720459,
      "learning_rate": 9.074999999999999e-06,
      "loss": 0.8354,
      "step": 125
    },
    {
      "epoch": 1.4018691588785046,
      "grad_norm": 3.214405059814453,
      "learning_rate": 1.0949999999999998e-05,
      "loss": 0.7314,
      "step": 150
    },
    {
      "epoch": 1.6355140186915889,
      "grad_norm": 3.0894336700439453,
      "learning_rate": 1.2825e-05,
      "loss": 0.662,
      "step": 175
    },
    {
      "epoch": 1.8691588785046729,
      "grad_norm": 2.8020288944244385,
      "learning_rate": 1.47e-05,
      "loss": 0.5881,
      "step": 200
    },
    {
      "epoch": 2.102803738317757,
      "grad_norm": 2.8243134021759033,
      "learning_rate": 1.6575e-05,
      "loss": 0.532,
      "step": 225
    },
    {
      "epoch": 2.336448598130841,
      "grad_norm": 3.0090668201446533,
      "learning_rate": 1.8449999999999998e-05,
      "loss": 0.4756,
      "step": 250
    },
    {
      "epoch": 2.5700934579439254,
      "grad_norm": 3.1206634044647217,
      "learning_rate": 2.0325e-05,
      "loss": 0.4452,
      "step": 275
    },
    {
      "epoch": 2.803738317757009,
      "grad_norm": 2.780308246612549,
      "learning_rate": 2.2199999999999998e-05,
      "loss": 0.423,
      "step": 300
    },
    {
      "epoch": 3.0373831775700935,
      "grad_norm": 2.3730380535125732,
      "learning_rate": 2.4075e-05,
      "loss": 0.3939,
      "step": 325
    },
    {
      "epoch": 3.2710280373831777,
      "grad_norm": 2.537344455718994,
      "learning_rate": 2.5949999999999997e-05,
      "loss": 0.3378,
      "step": 350
    },
    {
      "epoch": 3.5046728971962615,
      "grad_norm": 2.265047073364258,
      "learning_rate": 2.7825e-05,
      "loss": 0.3194,
      "step": 375
    },
    {
      "epoch": 3.7383177570093458,
      "grad_norm": 2.3696253299713135,
      "learning_rate": 2.97e-05,
      "loss": 0.3164,
      "step": 400
    },
    {
      "epoch": 3.97196261682243,
      "grad_norm": 2.213193893432617,
      "learning_rate": 3.1574999999999995e-05,
      "loss": 0.3078,
      "step": 425
    },
    {
      "epoch": 4.205607476635514,
      "grad_norm": 2.1573235988616943,
      "learning_rate": 3.345e-05,
      "loss": 0.2461,
      "step": 450
    },
    {
      "epoch": 4.4392523364485985,
      "grad_norm": 2.1853525638580322,
      "learning_rate": 3.5325e-05,
      "loss": 0.2355,
      "step": 475
    },
    {
      "epoch": 4.672897196261682,
      "grad_norm": 2.2035555839538574,
      "learning_rate": 3.7199999999999996e-05,
      "loss": 0.2318,
      "step": 500
    },
    {
      "epoch": 4.906542056074766,
      "grad_norm": 2.218616008758545,
      "learning_rate": 3.7480063291139237e-05,
      "loss": 0.2327,
      "step": 525
    },
    {
      "epoch": 5.140186915887851,
      "grad_norm": 1.9302279949188232,
      "learning_rate": 3.745632911392405e-05,
      "loss": 0.1966,
      "step": 550
    },
    {
      "epoch": 5.373831775700935,
      "grad_norm": 1.9767264127731323,
      "learning_rate": 3.7432594936708856e-05,
      "loss": 0.1645,
      "step": 575
    },
    {
      "epoch": 5.607476635514018,
      "grad_norm": 1.9298858642578125,
      "learning_rate": 3.740886075949367e-05,
      "loss": 0.1696,
      "step": 600
    },
    {
      "epoch": 5.841121495327103,
      "grad_norm": 2.433237075805664,
      "learning_rate": 3.7385126582278475e-05,
      "loss": 0.1683,
      "step": 625
    },
    {
      "epoch": 6.074766355140187,
      "grad_norm": 1.4901180267333984,
      "learning_rate": 3.736139240506329e-05,
      "loss": 0.1492,
      "step": 650
    },
    {
      "epoch": 6.308411214953271,
      "grad_norm": 1.7219280004501343,
      "learning_rate": 3.7337658227848094e-05,
      "loss": 0.1141,
      "step": 675
    },
    {
      "epoch": 6.542056074766355,
      "grad_norm": 1.9466065168380737,
      "learning_rate": 3.7313924050632907e-05,
      "loss": 0.118,
      "step": 700
    },
    {
      "epoch": 6.775700934579439,
      "grad_norm": 2.024010419845581,
      "learning_rate": 3.729018987341772e-05,
      "loss": 0.1218,
      "step": 725
    },
    {
      "epoch": 7.009345794392523,
      "grad_norm": 1.4249122142791748,
      "learning_rate": 3.7266455696202526e-05,
      "loss": 0.1196,
      "step": 750
    },
    {
      "epoch": 7.242990654205608,
      "grad_norm": 1.6397173404693604,
      "learning_rate": 3.724272151898734e-05,
      "loss": 0.0778,
      "step": 775
    },
    {
      "epoch": 7.4766355140186915,
      "grad_norm": 1.3958916664123535,
      "learning_rate": 3.7218987341772145e-05,
      "loss": 0.0792,
      "step": 800
    },
    {
      "epoch": 7.710280373831775,
      "grad_norm": 1.7240949869155884,
      "learning_rate": 3.719525316455696e-05,
      "loss": 0.083,
      "step": 825
    },
    {
      "epoch": 7.94392523364486,
      "grad_norm": 1.5714045763015747,
      "learning_rate": 3.717151898734177e-05,
      "loss": 0.0844,
      "step": 850
    },
    {
      "epoch": 8.177570093457945,
      "grad_norm": 1.1597058773040771,
      "learning_rate": 3.714778481012658e-05,
      "loss": 0.0594,
      "step": 875
    },
    {
      "epoch": 8.411214953271028,
      "grad_norm": 1.7305407524108887,
      "learning_rate": 3.712405063291139e-05,
      "loss": 0.0541,
      "step": 900
    },
    {
      "epoch": 8.644859813084112,
      "grad_norm": 1.1935184001922607,
      "learning_rate": 3.71003164556962e-05,
      "loss": 0.054,
      "step": 925
    },
    {
      "epoch": 8.878504672897197,
      "grad_norm": 1.3133002519607544,
      "learning_rate": 3.707658227848101e-05,
      "loss": 0.0562,
      "step": 950
    },
    {
      "epoch": 9.11214953271028,
      "grad_norm": 0.8144969344139099,
      "learning_rate": 3.705284810126582e-05,
      "loss": 0.0465,
      "step": 975
    },
    {
      "epoch": 9.345794392523365,
      "grad_norm": 1.2241060733795166,
      "learning_rate": 3.7029113924050634e-05,
      "loss": 0.0348,
      "step": 1000
    },
    {
      "epoch": 9.345794392523365,
      "eval_loss": 0.3381936550140381,
      "eval_runtime": 64.1556,
      "eval_samples_per_second": 212.452,
      "eval_steps_per_second": 1.668,
      "eval_wer": 22.71519659936238,
      "step": 1000
    },
    {
      "epoch": 9.57943925233645,
      "grad_norm": 1.095574140548706,
      "learning_rate": 3.700537974683544e-05,
      "loss": 0.0362,
      "step": 1025
    },
    {
      "epoch": 9.813084112149532,
      "grad_norm": 2.174593687057495,
      "learning_rate": 3.698164556962025e-05,
      "loss": 0.0374,
      "step": 1050
    },
    {
      "epoch": 10.046728971962617,
      "grad_norm": 0.7143180966377258,
      "learning_rate": 3.695791139240506e-05,
      "loss": 0.0348,
      "step": 1075
    },
    {
      "epoch": 10.280373831775702,
      "grad_norm": 0.8688119649887085,
      "learning_rate": 3.693417721518987e-05,
      "loss": 0.0226,
      "step": 1100
    },
    {
      "epoch": 10.514018691588785,
      "grad_norm": 1.339827299118042,
      "learning_rate": 3.691044303797468e-05,
      "loss": 0.023,
      "step": 1125
    },
    {
      "epoch": 10.74766355140187,
      "grad_norm": 0.998110830783844,
      "learning_rate": 3.688670886075949e-05,
      "loss": 0.0238,
      "step": 1150
    },
    {
      "epoch": 10.981308411214954,
      "grad_norm": 1.0500718355178833,
      "learning_rate": 3.6862974683544304e-05,
      "loss": 0.0258,
      "step": 1175
    },
    {
      "epoch": 11.214953271028037,
      "grad_norm": 1.6924108266830444,
      "learning_rate": 3.683924050632911e-05,
      "loss": 0.0169,
      "step": 1200
    },
    {
      "epoch": 11.448598130841122,
      "grad_norm": 0.8866608142852783,
      "learning_rate": 3.681550632911392e-05,
      "loss": 0.0163,
      "step": 1225
    },
    {
      "epoch": 11.682242990654206,
      "grad_norm": 0.9268547296524048,
      "learning_rate": 3.679177215189873e-05,
      "loss": 0.0166,
      "step": 1250
    },
    {
      "epoch": 11.91588785046729,
      "grad_norm": 1.0789356231689453,
      "learning_rate": 3.676803797468354e-05,
      "loss": 0.0179,
      "step": 1275
    },
    {
      "epoch": 12.149532710280374,
      "grad_norm": 0.5105511546134949,
      "learning_rate": 3.674430379746835e-05,
      "loss": 0.0139,
      "step": 1300
    },
    {
      "epoch": 12.383177570093459,
      "grad_norm": 0.9963575005531311,
      "learning_rate": 3.672056962025316e-05,
      "loss": 0.0112,
      "step": 1325
    },
    {
      "epoch": 12.616822429906541,
      "grad_norm": 0.46727660298347473,
      "learning_rate": 3.6696835443037974e-05,
      "loss": 0.0111,
      "step": 1350
    },
    {
      "epoch": 12.850467289719626,
      "grad_norm": 2.102184534072876,
      "learning_rate": 3.667310126582278e-05,
      "loss": 0.0115,
      "step": 1375
    },
    {
      "epoch": 13.08411214953271,
      "grad_norm": 0.44597211480140686,
      "learning_rate": 3.664936708860759e-05,
      "loss": 0.0102,
      "step": 1400
    },
    {
      "epoch": 13.317757009345794,
      "grad_norm": 0.2395753413438797,
      "learning_rate": 3.66256329113924e-05,
      "loss": 0.0073,
      "step": 1425
    },
    {
      "epoch": 13.551401869158878,
      "grad_norm": 0.2823430895805359,
      "learning_rate": 3.660189873417721e-05,
      "loss": 0.007,
      "step": 1450
    },
    {
      "epoch": 13.785046728971963,
      "grad_norm": 0.22637470066547394,
      "learning_rate": 3.657816455696202e-05,
      "loss": 0.0069,
      "step": 1475
    },
    {
      "epoch": 14.018691588785046,
      "grad_norm": 0.15601249039173126,
      "learning_rate": 3.655443037974683e-05,
      "loss": 0.0068,
      "step": 1500
    },
    {
      "epoch": 14.25233644859813,
      "grad_norm": 0.14189423620700836,
      "learning_rate": 3.6530696202531644e-05,
      "loss": 0.005,
      "step": 1525
    },
    {
      "epoch": 14.485981308411215,
      "grad_norm": 0.19872644543647766,
      "learning_rate": 3.650696202531645e-05,
      "loss": 0.0047,
      "step": 1550
    },
    {
      "epoch": 14.719626168224298,
      "grad_norm": 0.11419831961393356,
      "learning_rate": 3.648322784810126e-05,
      "loss": 0.0046,
      "step": 1575
    },
    {
      "epoch": 14.953271028037383,
      "grad_norm": 0.13338784873485565,
      "learning_rate": 3.645949367088607e-05,
      "loss": 0.0047,
      "step": 1600
    },
    {
      "epoch": 15.186915887850468,
      "grad_norm": 0.08034642785787582,
      "learning_rate": 3.643575949367088e-05,
      "loss": 0.0037,
      "step": 1625
    },
    {
      "epoch": 15.42056074766355,
      "grad_norm": 0.07917828857898712,
      "learning_rate": 3.641202531645569e-05,
      "loss": 0.0036,
      "step": 1650
    },
    {
      "epoch": 15.654205607476635,
      "grad_norm": 0.09836310893297195,
      "learning_rate": 3.638829113924051e-05,
      "loss": 0.0036,
      "step": 1675
    },
    {
      "epoch": 15.88785046728972,
      "grad_norm": 0.08892955631017685,
      "learning_rate": 3.6364556962025314e-05,
      "loss": 0.0036,
      "step": 1700
    },
    {
      "epoch": 16.121495327102803,
      "grad_norm": 0.07386955618858337,
      "learning_rate": 3.634082278481013e-05,
      "loss": 0.0032,
      "step": 1725
    },
    {
      "epoch": 16.35514018691589,
      "grad_norm": 0.06931279599666595,
      "learning_rate": 3.631708860759493e-05,
      "loss": 0.0031,
      "step": 1750
    },
    {
      "epoch": 16.588785046728972,
      "grad_norm": 0.07556220144033432,
      "learning_rate": 3.6293354430379746e-05,
      "loss": 0.0029,
      "step": 1775
    },
    {
      "epoch": 16.822429906542055,
      "grad_norm": 0.07463835179805756,
      "learning_rate": 3.626962025316455e-05,
      "loss": 0.0029,
      "step": 1800
    },
    {
      "epoch": 17.05607476635514,
      "grad_norm": 0.05796229466795921,
      "learning_rate": 3.6245886075949365e-05,
      "loss": 0.0028,
      "step": 1825
    },
    {
      "epoch": 17.289719626168225,
      "grad_norm": 0.055481746792793274,
      "learning_rate": 3.622215189873418e-05,
      "loss": 0.0025,
      "step": 1850
    },
    {
      "epoch": 17.523364485981308,
      "grad_norm": 0.06257276237010956,
      "learning_rate": 3.6198417721518984e-05,
      "loss": 0.0025,
      "step": 1875
    },
    {
      "epoch": 17.757009345794394,
      "grad_norm": 0.06188027188181877,
      "learning_rate": 3.61746835443038e-05,
      "loss": 0.0025,
      "step": 1900
    },
    {
      "epoch": 17.990654205607477,
      "grad_norm": 0.06342104822397232,
      "learning_rate": 3.61509493670886e-05,
      "loss": 0.0025,
      "step": 1925
    },
    {
      "epoch": 18.22429906542056,
      "grad_norm": 0.051547735929489136,
      "learning_rate": 3.6127215189873416e-05,
      "loss": 0.0022,
      "step": 1950
    },
    {
      "epoch": 18.457943925233646,
      "grad_norm": 0.05366269499063492,
      "learning_rate": 3.610348101265823e-05,
      "loss": 0.0021,
      "step": 1975
    },
    {
      "epoch": 18.69158878504673,
      "grad_norm": 0.05343176797032356,
      "learning_rate": 3.6079746835443035e-05,
      "loss": 0.0021,
      "step": 2000
    },
    {
      "epoch": 18.69158878504673,
      "eval_loss": 0.40916550159454346,
      "eval_runtime": 62.9859,
      "eval_samples_per_second": 216.398,
      "eval_steps_per_second": 1.699,
      "eval_wer": 21.784418630217303,
      "step": 2000
    },
    {
      "epoch": 18.925233644859812,
      "grad_norm": 0.05404841899871826,
      "learning_rate": 3.605601265822785e-05,
      "loss": 0.0022,
      "step": 2025
    },
    {
      "epoch": 19.1588785046729,
      "grad_norm": 0.044902652502059937,
      "learning_rate": 3.6032278481012654e-05,
      "loss": 0.002,
      "step": 2050
    },
    {
      "epoch": 19.39252336448598,
      "grad_norm": 0.04406387358903885,
      "learning_rate": 3.600854430379747e-05,
      "loss": 0.0019,
      "step": 2075
    },
    {
      "epoch": 19.626168224299064,
      "grad_norm": 0.04272112250328064,
      "learning_rate": 3.598481012658227e-05,
      "loss": 0.0019,
      "step": 2100
    },
    {
      "epoch": 19.85981308411215,
      "grad_norm": 0.044560402631759644,
      "learning_rate": 3.5961075949367086e-05,
      "loss": 0.0019,
      "step": 2125
    },
    {
      "epoch": 20.093457943925234,
      "grad_norm": 0.03814489766955376,
      "learning_rate": 3.59373417721519e-05,
      "loss": 0.0018,
      "step": 2150
    },
    {
      "epoch": 20.327102803738317,
      "grad_norm": 0.0408877357840538,
      "learning_rate": 3.5913607594936705e-05,
      "loss": 0.0017,
      "step": 2175
    },
    {
      "epoch": 20.560747663551403,
      "grad_norm": 0.03910227492451668,
      "learning_rate": 3.588987341772152e-05,
      "loss": 0.0017,
      "step": 2200
    },
    {
      "epoch": 20.794392523364486,
      "grad_norm": 0.04326726868748665,
      "learning_rate": 3.5866139240506324e-05,
      "loss": 0.0017,
      "step": 2225
    },
    {
      "epoch": 21.02803738317757,
      "grad_norm": 0.03501986712217331,
      "learning_rate": 3.5842405063291137e-05,
      "loss": 0.0016,
      "step": 2250
    },
    {
      "epoch": 21.261682242990656,
      "grad_norm": 0.03702597692608833,
      "learning_rate": 3.581867088607594e-05,
      "loss": 0.0015,
      "step": 2275
    },
    {
      "epoch": 21.49532710280374,
      "grad_norm": 0.035930950194597244,
      "learning_rate": 3.5794936708860756e-05,
      "loss": 0.0014,
      "step": 2300
    },
    {
      "epoch": 21.72897196261682,
      "grad_norm": 0.0382557213306427,
      "learning_rate": 3.577120253164557e-05,
      "loss": 0.0015,
      "step": 2325
    },
    {
      "epoch": 21.962616822429908,
      "grad_norm": 0.03772348538041115,
      "learning_rate": 3.5747468354430375e-05,
      "loss": 0.0015,
      "step": 2350
    },
    {
      "epoch": 22.19626168224299,
      "grad_norm": 0.03188503533601761,
      "learning_rate": 3.572373417721519e-05,
      "loss": 0.0014,
      "step": 2375
    },
    {
      "epoch": 22.429906542056074,
      "grad_norm": 0.03148446977138519,
      "learning_rate": 3.5699999999999994e-05,
      "loss": 0.0013,
      "step": 2400
    },
    {
      "epoch": 22.66355140186916,
      "grad_norm": 0.034659937024116516,
      "learning_rate": 3.5676265822784807e-05,
      "loss": 0.0013,
      "step": 2425
    },
    {
      "epoch": 22.897196261682243,
      "grad_norm": 0.033603236079216,
      "learning_rate": 3.565253164556961e-05,
      "loss": 0.0013,
      "step": 2450
    },
    {
      "epoch": 23.130841121495326,
      "grad_norm": 0.03048083372414112,
      "learning_rate": 3.5628797468354426e-05,
      "loss": 0.0013,
      "step": 2475
    },
    {
      "epoch": 23.364485981308412,
      "grad_norm": 0.03026960976421833,
      "learning_rate": 3.560506329113924e-05,
      "loss": 0.0012,
      "step": 2500
    },
    {
      "epoch": 23.598130841121495,
      "grad_norm": 0.028185931965708733,
      "learning_rate": 3.558132911392405e-05,
      "loss": 0.0012,
      "step": 2525
    },
    {
      "epoch": 23.83177570093458,
      "grad_norm": 0.030694087967276573,
      "learning_rate": 3.555759493670886e-05,
      "loss": 0.0012,
      "step": 2550
    },
    {
      "epoch": 24.065420560747665,
      "grad_norm": 0.024795854464173317,
      "learning_rate": 3.553386075949367e-05,
      "loss": 0.0011,
      "step": 2575
    },
    {
      "epoch": 24.299065420560748,
      "grad_norm": 0.026742959395051003,
      "learning_rate": 3.5510126582278476e-05,
      "loss": 0.0011,
      "step": 2600
    },
    {
      "epoch": 24.53271028037383,
      "grad_norm": 0.025302302092313766,
      "learning_rate": 3.548639240506329e-05,
      "loss": 0.0011,
      "step": 2625
    },
    {
      "epoch": 24.766355140186917,
      "grad_norm": 0.027101745828986168,
      "learning_rate": 3.54626582278481e-05,
      "loss": 0.0012,
      "step": 2650
    },
    {
      "epoch": 25.0,
      "grad_norm": 0.038450077176094055,
      "learning_rate": 3.543892405063291e-05,
      "loss": 0.0011,
      "step": 2675
    },
    {
      "epoch": 25.233644859813083,
      "grad_norm": 0.02212081104516983,
      "learning_rate": 3.541518987341772e-05,
      "loss": 0.001,
      "step": 2700
    },
    {
      "epoch": 25.46728971962617,
      "grad_norm": 0.02410231903195381,
      "learning_rate": 3.539145569620253e-05,
      "loss": 0.001,
      "step": 2725
    },
    {
      "epoch": 25.700934579439252,
      "grad_norm": 0.02458999678492546,
      "learning_rate": 3.536772151898734e-05,
      "loss": 0.001,
      "step": 2750
    },
    {
      "epoch": 25.934579439252335,
      "grad_norm": 0.024960873648524284,
      "learning_rate": 3.5343987341772146e-05,
      "loss": 0.001,
      "step": 2775
    },
    {
      "epoch": 26.16822429906542,
      "grad_norm": 0.02074788324534893,
      "learning_rate": 3.532025316455696e-05,
      "loss": 0.0009,
      "step": 2800
    },
    {
      "epoch": 26.401869158878505,
      "grad_norm": 0.022805292159318924,
      "learning_rate": 3.529651898734177e-05,
      "loss": 0.0009,
      "step": 2825
    },
    {
      "epoch": 26.635514018691588,
      "grad_norm": 0.022323718294501305,
      "learning_rate": 3.527278481012658e-05,
      "loss": 0.0009,
      "step": 2850
    },
    {
      "epoch": 26.869158878504674,
      "grad_norm": 0.02316974848508835,
      "learning_rate": 3.524905063291139e-05,
      "loss": 0.0009,
      "step": 2875
    },
    {
      "epoch": 27.102803738317757,
      "grad_norm": 0.019991261884570122,
      "learning_rate": 3.52253164556962e-05,
      "loss": 0.0009,
      "step": 2900
    },
    {
      "epoch": 27.33644859813084,
      "grad_norm": 0.020080773159861565,
      "learning_rate": 3.520158227848101e-05,
      "loss": 0.0008,
      "step": 2925
    },
    {
      "epoch": 27.570093457943926,
      "grad_norm": 0.020259954035282135,
      "learning_rate": 3.5177848101265816e-05,
      "loss": 0.0008,
      "step": 2950
    },
    {
      "epoch": 27.80373831775701,
      "grad_norm": 0.019856547936797142,
      "learning_rate": 3.515411392405063e-05,
      "loss": 0.0008,
      "step": 2975
    },
    {
      "epoch": 28.037383177570092,
      "grad_norm": 0.018801957368850708,
      "learning_rate": 3.513037974683544e-05,
      "loss": 0.0009,
      "step": 3000
    },
    {
      "epoch": 28.037383177570092,
      "eval_loss": 0.4509190320968628,
      "eval_runtime": 62.6433,
      "eval_samples_per_second": 217.581,
      "eval_steps_per_second": 1.708,
      "eval_wer": 21.90259811645718,
      "step": 3000
    },
    {
      "epoch": 28.27102803738318,
      "grad_norm": 0.017624234780669212,
      "learning_rate": 3.510664556962025e-05,
      "loss": 0.0008,
      "step": 3025
    },
    {
      "epoch": 28.50467289719626,
      "grad_norm": 0.01893887110054493,
      "learning_rate": 3.508291139240506e-05,
      "loss": 0.0008,
      "step": 3050
    },
    {
      "epoch": 28.738317757009344,
      "grad_norm": 0.01899472065269947,
      "learning_rate": 3.505917721518987e-05,
      "loss": 0.0008,
      "step": 3075
    },
    {
      "epoch": 28.97196261682243,
      "grad_norm": 0.018534839153289795,
      "learning_rate": 3.503544303797468e-05,
      "loss": 0.0007,
      "step": 3100
    },
    {
      "epoch": 29.205607476635514,
      "grad_norm": 0.016035275533795357,
      "learning_rate": 3.5011708860759486e-05,
      "loss": 0.0007,
      "step": 3125
    },
    {
      "epoch": 29.439252336448597,
      "grad_norm": 0.016668906435370445,
      "learning_rate": 3.49879746835443e-05,
      "loss": 0.0007,
      "step": 3150
    },
    {
      "epoch": 29.672897196261683,
      "grad_norm": 0.01565367542207241,
      "learning_rate": 3.496424050632911e-05,
      "loss": 0.0007,
      "step": 3175
    },
    {
      "epoch": 29.906542056074766,
      "grad_norm": 0.6403020024299622,
      "learning_rate": 3.494050632911392e-05,
      "loss": 0.001,
      "step": 3200
    },
    {
      "epoch": 30.14018691588785,
      "grad_norm": 1.84394109249115,
      "learning_rate": 3.491677215189873e-05,
      "loss": 0.0129,
      "step": 3225
    },
    {
      "epoch": 30.373831775700936,
      "grad_norm": 1.9401473999023438,
      "learning_rate": 3.489303797468354e-05,
      "loss": 0.0291,
      "step": 3250
    },
    {
      "epoch": 30.60747663551402,
      "grad_norm": 2.3162951469421387,
      "learning_rate": 3.486930379746835e-05,
      "loss": 0.0319,
      "step": 3275
    },
    {
      "epoch": 30.8411214953271,
      "grad_norm": 1.940976858139038,
      "learning_rate": 3.4845569620253156e-05,
      "loss": 0.0318,
      "step": 3300
    },
    {
      "epoch": 31.074766355140188,
      "grad_norm": 1.7011630535125732,
      "learning_rate": 3.4821835443037976e-05,
      "loss": 0.0283,
      "step": 3325
    },
    {
      "epoch": 31.30841121495327,
      "grad_norm": 2.005868434906006,
      "learning_rate": 3.479810126582278e-05,
      "loss": 0.0224,
      "step": 3350
    },
    {
      "epoch": 31.542056074766354,
      "grad_norm": 1.7186815738677979,
      "learning_rate": 3.4774367088607595e-05,
      "loss": 0.0233,
      "step": 3375
    },
    {
      "epoch": 31.77570093457944,
      "grad_norm": 1.352404236793518,
      "learning_rate": 3.47506329113924e-05,
      "loss": 0.024,
      "step": 3400
    },
    {
      "epoch": 32.00934579439252,
      "grad_norm": 1.0012911558151245,
      "learning_rate": 3.4726898734177214e-05,
      "loss": 0.022,
      "step": 3425
    },
    {
      "epoch": 32.242990654205606,
      "grad_norm": 1.6940593719482422,
      "learning_rate": 3.470316455696203e-05,
      "loss": 0.0148,
      "step": 3450
    },
    {
      "epoch": 32.47663551401869,
      "grad_norm": 1.2554163932800293,
      "learning_rate": 3.467943037974683e-05,
      "loss": 0.0153,
      "step": 3475
    },
    {
      "epoch": 32.71028037383178,
      "grad_norm": 1.4891505241394043,
      "learning_rate": 3.4655696202531646e-05,
      "loss": 0.0162,
      "step": 3500
    },
    {
      "epoch": 32.94392523364486,
      "grad_norm": 1.08815598487854,
      "learning_rate": 3.463196202531645e-05,
      "loss": 0.0156,
      "step": 3525
    },
    {
      "epoch": 33.177570093457945,
      "grad_norm": 0.8786664605140686,
      "learning_rate": 3.4608227848101265e-05,
      "loss": 0.0115,
      "step": 3550
    },
    {
      "epoch": 33.41121495327103,
      "grad_norm": 0.8262961506843567,
      "learning_rate": 3.458449367088607e-05,
      "loss": 0.0099,
      "step": 3575
    },
    {
      "epoch": 33.64485981308411,
      "grad_norm": 1.0154938697814941,
      "learning_rate": 3.4560759493670884e-05,
      "loss": 0.0099,
      "step": 3600
    },
    {
      "epoch": 33.87850467289719,
      "grad_norm": 1.3082599639892578,
      "learning_rate": 3.45370253164557e-05,
      "loss": 0.0117,
      "step": 3625
    },
    {
      "epoch": 34.11214953271028,
      "grad_norm": 0.8540720343589783,
      "learning_rate": 3.45132911392405e-05,
      "loss": 0.0098,
      "step": 3650
    },
    {
      "epoch": 34.345794392523366,
      "grad_norm": 0.6072850823402405,
      "learning_rate": 3.4489556962025316e-05,
      "loss": 0.0071,
      "step": 3675
    },
    {
      "epoch": 34.57943925233645,
      "grad_norm": 0.8056187033653259,
      "learning_rate": 3.446582278481012e-05,
      "loss": 0.006,
      "step": 3700
    },
    {
      "epoch": 34.81308411214953,
      "grad_norm": 0.7188331484794617,
      "learning_rate": 3.4442088607594935e-05,
      "loss": 0.0072,
      "step": 3725
    },
    {
      "epoch": 35.046728971962615,
      "grad_norm": 1.001788854598999,
      "learning_rate": 3.441835443037974e-05,
      "loss": 0.0061,
      "step": 3750
    },
    {
      "epoch": 35.2803738317757,
      "grad_norm": 0.34956827759742737,
      "learning_rate": 3.4394620253164554e-05,
      "loss": 0.0045,
      "step": 3775
    },
    {
      "epoch": 35.51401869158879,
      "grad_norm": 0.5170698761940002,
      "learning_rate": 3.437088607594937e-05,
      "loss": 0.0044,
      "step": 3800
    },
    {
      "epoch": 35.74766355140187,
      "grad_norm": 0.6102795600891113,
      "learning_rate": 3.434715189873417e-05,
      "loss": 0.004,
      "step": 3825
    },
    {
      "epoch": 35.981308411214954,
      "grad_norm": 0.4813729226589203,
      "learning_rate": 3.4323417721518986e-05,
      "loss": 0.0039,
      "step": 3850
    },
    {
      "epoch": 36.21495327102804,
      "grad_norm": 0.37823376059532166,
      "learning_rate": 3.429968354430379e-05,
      "loss": 0.0034,
      "step": 3875
    },
    {
      "epoch": 36.44859813084112,
      "grad_norm": 0.3653680384159088,
      "learning_rate": 3.4275949367088605e-05,
      "loss": 0.0036,
      "step": 3900
    },
    {
      "epoch": 36.6822429906542,
      "grad_norm": 0.24608762562274933,
      "learning_rate": 3.425221518987341e-05,
      "loss": 0.0027,
      "step": 3925
    },
    {
      "epoch": 36.91588785046729,
      "grad_norm": 0.3653218150138855,
      "learning_rate": 3.4228481012658224e-05,
      "loss": 0.0028,
      "step": 3950
    },
    {
      "epoch": 37.149532710280376,
      "grad_norm": 0.1712789684534073,
      "learning_rate": 3.4204746835443037e-05,
      "loss": 0.0025,
      "step": 3975
    },
    {
      "epoch": 37.38317757009346,
      "grad_norm": 0.24373282492160797,
      "learning_rate": 3.418101265822784e-05,
      "loss": 0.0023,
      "step": 4000
    },
    {
      "epoch": 37.38317757009346,
      "eval_loss": 0.40623340010643005,
      "eval_runtime": 62.4579,
      "eval_samples_per_second": 218.227,
      "eval_steps_per_second": 1.713,
      "eval_wer": 20.718054893913298,
      "step": 4000
    },
    {
      "epoch": 37.61682242990654,
      "grad_norm": 0.21459850668907166,
      "learning_rate": 3.4157278481012656e-05,
      "loss": 0.0022,
      "step": 4025
    },
    {
      "epoch": 37.850467289719624,
      "grad_norm": 0.19664266705513,
      "learning_rate": 3.413354430379746e-05,
      "loss": 0.0018,
      "step": 4050
    },
    {
      "epoch": 38.08411214953271,
      "grad_norm": 0.19081653654575348,
      "learning_rate": 3.4109810126582275e-05,
      "loss": 0.0018,
      "step": 4075
    },
    {
      "epoch": 38.3177570093458,
      "grad_norm": 0.0588967390358448,
      "learning_rate": 3.408607594936708e-05,
      "loss": 0.0015,
      "step": 4100
    },
    {
      "epoch": 38.55140186915888,
      "grad_norm": 0.059834085404872894,
      "learning_rate": 3.4062341772151894e-05,
      "loss": 0.0016,
      "step": 4125
    },
    {
      "epoch": 38.78504672897196,
      "grad_norm": 0.3412402272224426,
      "learning_rate": 3.4038607594936707e-05,
      "loss": 0.0014,
      "step": 4150
    },
    {
      "epoch": 39.018691588785046,
      "grad_norm": 0.08089874684810638,
      "learning_rate": 3.401487341772152e-05,
      "loss": 0.0012,
      "step": 4175
    },
    {
      "epoch": 39.25233644859813,
      "grad_norm": 0.23941798508167267,
      "learning_rate": 3.3991139240506326e-05,
      "loss": 0.001,
      "step": 4200
    },
    {
      "epoch": 39.48598130841121,
      "grad_norm": 0.03656350076198578,
      "learning_rate": 3.396740506329114e-05,
      "loss": 0.001,
      "step": 4225
    },
    {
      "epoch": 39.7196261682243,
      "grad_norm": 0.030744420364499092,
      "learning_rate": 3.3943670886075945e-05,
      "loss": 0.001,
      "step": 4250
    },
    {
      "epoch": 39.953271028037385,
      "grad_norm": 0.03299107402563095,
      "learning_rate": 3.391993670886076e-05,
      "loss": 0.0008,
      "step": 4275
    },
    {
      "epoch": 40.18691588785047,
      "grad_norm": 0.029756462201476097,
      "learning_rate": 3.389620253164557e-05,
      "loss": 0.0009,
      "step": 4300
    },
    {
      "epoch": 40.42056074766355,
      "grad_norm": 0.05355725809931755,
      "learning_rate": 3.3872468354430376e-05,
      "loss": 0.0007,
      "step": 4325
    },
    {
      "epoch": 40.654205607476634,
      "grad_norm": 0.04675650596618652,
      "learning_rate": 3.384873417721519e-05,
      "loss": 0.0008,
      "step": 4350
    },
    {
      "epoch": 40.88785046728972,
      "grad_norm": 0.035253919661045074,
      "learning_rate": 3.3824999999999995e-05,
      "loss": 0.0007,
      "step": 4375
    },
    {
      "epoch": 41.12149532710281,
      "grad_norm": 0.028731150552630424,
      "learning_rate": 3.380126582278481e-05,
      "loss": 0.0009,
      "step": 4400
    },
    {
      "epoch": 41.35514018691589,
      "grad_norm": 0.19011081755161285,
      "learning_rate": 3.377753164556962e-05,
      "loss": 0.0007,
      "step": 4425
    },
    {
      "epoch": 41.58878504672897,
      "grad_norm": 0.049792371690273285,
      "learning_rate": 3.375379746835443e-05,
      "loss": 0.0009,
      "step": 4450
    },
    {
      "epoch": 41.822429906542055,
      "grad_norm": 0.04106704890727997,
      "learning_rate": 3.373006329113924e-05,
      "loss": 0.0007,
      "step": 4475
    },
    {
      "epoch": 42.05607476635514,
      "grad_norm": 0.020676936954259872,
      "learning_rate": 3.3706329113924046e-05,
      "loss": 0.0007,
      "step": 4500
    },
    {
      "epoch": 42.28971962616822,
      "grad_norm": 0.020809998735785484,
      "learning_rate": 3.368259493670886e-05,
      "loss": 0.0006,
      "step": 4525
    },
    {
      "epoch": 42.52336448598131,
      "grad_norm": 0.019781475886702538,
      "learning_rate": 3.3658860759493665e-05,
      "loss": 0.0006,
      "step": 4550
    },
    {
      "epoch": 42.757009345794394,
      "grad_norm": 0.01729559525847435,
      "learning_rate": 3.363512658227848e-05,
      "loss": 0.0006,
      "step": 4575
    },
    {
      "epoch": 42.99065420560748,
      "grad_norm": 0.025380814447999,
      "learning_rate": 3.361139240506329e-05,
      "loss": 0.0006,
      "step": 4600
    },
    {
      "epoch": 43.22429906542056,
      "grad_norm": 0.019424615427851677,
      "learning_rate": 3.35876582278481e-05,
      "loss": 0.0007,
      "step": 4625
    },
    {
      "epoch": 43.45794392523364,
      "grad_norm": 0.04872697964310646,
      "learning_rate": 3.356392405063291e-05,
      "loss": 0.0005,
      "step": 4650
    },
    {
      "epoch": 43.691588785046726,
      "grad_norm": 0.01503314170986414,
      "learning_rate": 3.3540189873417716e-05,
      "loss": 0.0005,
      "step": 4675
    },
    {
      "epoch": 43.925233644859816,
      "grad_norm": 0.013821057043969631,
      "learning_rate": 3.351645569620253e-05,
      "loss": 0.0006,
      "step": 4700
    },
    {
      "epoch": 44.1588785046729,
      "grad_norm": 0.015061864629387856,
      "learning_rate": 3.3492721518987335e-05,
      "loss": 0.0004,
      "step": 4725
    },
    {
      "epoch": 44.39252336448598,
      "grad_norm": 0.011650556698441505,
      "learning_rate": 3.346898734177215e-05,
      "loss": 0.0004,
      "step": 4750
    },
    {
      "epoch": 44.626168224299064,
      "grad_norm": 0.012532194145023823,
      "learning_rate": 3.344525316455696e-05,
      "loss": 0.0004,
      "step": 4775
    },
    {
      "epoch": 44.85981308411215,
      "grad_norm": 0.010689922608435154,
      "learning_rate": 3.342151898734177e-05,
      "loss": 0.0004,
      "step": 4800
    },
    {
      "epoch": 45.09345794392523,
      "grad_norm": 0.010240688920021057,
      "learning_rate": 3.339778481012658e-05,
      "loss": 0.0004,
      "step": 4825
    },
    {
      "epoch": 45.32710280373832,
      "grad_norm": 0.010845264419913292,
      "learning_rate": 3.3374050632911386e-05,
      "loss": 0.0004,
      "step": 4850
    },
    {
      "epoch": 45.5607476635514,
      "grad_norm": 0.010732454247772694,
      "learning_rate": 3.33503164556962e-05,
      "loss": 0.0004,
      "step": 4875
    },
    {
      "epoch": 45.794392523364486,
      "grad_norm": 0.010572795756161213,
      "learning_rate": 3.3326582278481005e-05,
      "loss": 0.0004,
      "step": 4900
    },
    {
      "epoch": 46.02803738317757,
      "grad_norm": 0.009393970482051373,
      "learning_rate": 3.330284810126582e-05,
      "loss": 0.0004,
      "step": 4925
    },
    {
      "epoch": 46.26168224299065,
      "grad_norm": 0.009326569736003876,
      "learning_rate": 3.327911392405063e-05,
      "loss": 0.0003,
      "step": 4950
    },
    {
      "epoch": 46.495327102803735,
      "grad_norm": 0.009698985144495964,
      "learning_rate": 3.3255379746835444e-05,
      "loss": 0.0004,
      "step": 4975
    },
    {
      "epoch": 46.728971962616825,
      "grad_norm": 0.009771204553544521,
      "learning_rate": 3.323164556962025e-05,
      "loss": 0.0003,
      "step": 5000
    },
    {
      "epoch": 46.728971962616825,
      "eval_loss": 0.4349908232688904,
      "eval_runtime": 63.0951,
      "eval_samples_per_second": 216.023,
      "eval_steps_per_second": 1.696,
      "eval_wer": 20.124409102568798,
      "step": 5000
    },
    {
      "epoch": 46.96261682242991,
      "grad_norm": 0.009263860061764717,
      "learning_rate": 3.320791139240506e-05,
      "loss": 0.0004,
      "step": 5025
    },
    {
      "epoch": 47.19626168224299,
      "grad_norm": 0.009048182517290115,
      "learning_rate": 3.318417721518987e-05,
      "loss": 0.0003,
      "step": 5050
    },
    {
      "epoch": 47.429906542056074,
      "grad_norm": 0.008781012147665024,
      "learning_rate": 3.316044303797468e-05,
      "loss": 0.0003,
      "step": 5075
    },
    {
      "epoch": 47.66355140186916,
      "grad_norm": 0.008727292530238628,
      "learning_rate": 3.3136708860759495e-05,
      "loss": 0.0003,
      "step": 5100
    },
    {
      "epoch": 47.89719626168224,
      "grad_norm": 0.00861198827624321,
      "learning_rate": 3.31129746835443e-05,
      "loss": 0.0003,
      "step": 5125
    },
    {
      "epoch": 48.13084112149533,
      "grad_norm": 0.008097299374639988,
      "learning_rate": 3.3089240506329114e-05,
      "loss": 0.0003,
      "step": 5150
    },
    {
      "epoch": 48.36448598130841,
      "grad_norm": 0.007882571779191494,
      "learning_rate": 3.306550632911392e-05,
      "loss": 0.0003,
      "step": 5175
    },
    {
      "epoch": 48.598130841121495,
      "grad_norm": 0.007794591132551432,
      "learning_rate": 3.304177215189873e-05,
      "loss": 0.0003,
      "step": 5200
    },
    {
      "epoch": 48.83177570093458,
      "grad_norm": 0.008559081703424454,
      "learning_rate": 3.301803797468354e-05,
      "loss": 0.0003,
      "step": 5225
    },
    {
      "epoch": 49.06542056074766,
      "grad_norm": 0.007447989657521248,
      "learning_rate": 3.299430379746835e-05,
      "loss": 0.0003,
      "step": 5250
    },
    {
      "epoch": 49.299065420560744,
      "grad_norm": 0.007480722386389971,
      "learning_rate": 3.2970569620253165e-05,
      "loss": 0.0003,
      "step": 5275
    },
    {
      "epoch": 49.532710280373834,
      "grad_norm": 0.007868384011089802,
      "learning_rate": 3.294683544303797e-05,
      "loss": 0.0003,
      "step": 5300
    },
    {
      "epoch": 49.76635514018692,
      "grad_norm": 0.007319408468902111,
      "learning_rate": 3.2923101265822784e-05,
      "loss": 0.0003,
      "step": 5325
    },
    {
      "epoch": 50.0,
      "grad_norm": 0.010627989657223225,
      "learning_rate": 3.289936708860759e-05,
      "loss": 0.0003,
      "step": 5350
    },
    {
      "epoch": 50.23364485981308,
      "grad_norm": 0.006707332096993923,
      "learning_rate": 3.28756329113924e-05,
      "loss": 0.0003,
      "step": 5375
    },
    {
      "epoch": 50.467289719626166,
      "grad_norm": 0.007135647349059582,
      "learning_rate": 3.285189873417721e-05,
      "loss": 0.0003,
      "step": 5400
    },
    {
      "epoch": 50.70093457943925,
      "grad_norm": 0.007132141385227442,
      "learning_rate": 3.282816455696202e-05,
      "loss": 0.0003,
      "step": 5425
    },
    {
      "epoch": 50.93457943925234,
      "grad_norm": 0.007261497899889946,
      "learning_rate": 3.2804430379746835e-05,
      "loss": 0.0003,
      "step": 5450
    },
    {
      "epoch": 51.16822429906542,
      "grad_norm": 0.006365330424159765,
      "learning_rate": 3.278069620253164e-05,
      "loss": 0.0003,
      "step": 5475
    },
    {
      "epoch": 51.401869158878505,
      "grad_norm": 0.0064588082022964954,
      "learning_rate": 3.2756962025316454e-05,
      "loss": 0.0002,
      "step": 5500
    },
    {
      "epoch": 51.63551401869159,
      "grad_norm": 0.0068720560520887375,
      "learning_rate": 3.273322784810126e-05,
      "loss": 0.0003,
      "step": 5525
    },
    {
      "epoch": 51.86915887850467,
      "grad_norm": 0.00682293064892292,
      "learning_rate": 3.270949367088607e-05,
      "loss": 0.0002,
      "step": 5550
    },
    {
      "epoch": 52.10280373831776,
      "grad_norm": 0.006176824681460857,
      "learning_rate": 3.268575949367088e-05,
      "loss": 0.0002,
      "step": 5575
    },
    {
      "epoch": 52.33644859813084,
      "grad_norm": 0.006278651766479015,
      "learning_rate": 3.266202531645569e-05,
      "loss": 0.0002,
      "step": 5600
    },
    {
      "epoch": 52.570093457943926,
      "grad_norm": 0.006093575153499842,
      "learning_rate": 3.2638291139240505e-05,
      "loss": 0.0002,
      "step": 5625
    },
    {
      "epoch": 52.80373831775701,
      "grad_norm": 0.006240413058549166,
      "learning_rate": 3.261455696202531e-05,
      "loss": 0.0003,
      "step": 5650
    },
    {
      "epoch": 53.03738317757009,
      "grad_norm": 0.005718667525798082,
      "learning_rate": 3.2590822784810124e-05,
      "loss": 0.0002,
      "step": 5675
    },
    {
      "epoch": 53.271028037383175,
      "grad_norm": 0.006080987397581339,
      "learning_rate": 3.256708860759493e-05,
      "loss": 0.0002,
      "step": 5700
    },
    {
      "epoch": 53.504672897196265,
      "grad_norm": 0.006001737434417009,
      "learning_rate": 3.254335443037974e-05,
      "loss": 0.0002,
      "step": 5725
    },
    {
      "epoch": 53.73831775700935,
      "grad_norm": 0.005847087129950523,
      "learning_rate": 3.251962025316455e-05,
      "loss": 0.0002,
      "step": 5750
    },
    {
      "epoch": 53.97196261682243,
      "grad_norm": 0.0060851857997477055,
      "learning_rate": 3.249588607594936e-05,
      "loss": 0.0002,
      "step": 5775
    },
    {
      "epoch": 54.205607476635514,
      "grad_norm": 0.005599698051810265,
      "learning_rate": 3.2472151898734175e-05,
      "loss": 0.0002,
      "step": 5800
    },
    {
      "epoch": 54.4392523364486,
      "grad_norm": 0.005331767722964287,
      "learning_rate": 3.244841772151899e-05,
      "loss": 0.0003,
      "step": 5825
    },
    {
      "epoch": 54.67289719626168,
      "grad_norm": 0.005450521130114794,
      "learning_rate": 3.2424683544303794e-05,
      "loss": 0.0002,
      "step": 5850
    },
    {
      "epoch": 54.90654205607477,
      "grad_norm": 0.005088282749056816,
      "learning_rate": 3.2400949367088607e-05,
      "loss": 0.0002,
      "step": 5875
    },
    {
      "epoch": 55.14018691588785,
      "grad_norm": 0.0053012073040008545,
      "learning_rate": 3.237721518987342e-05,
      "loss": 0.0002,
      "step": 5900
    },
    {
      "epoch": 55.373831775700936,
      "grad_norm": 0.004951627925038338,
      "learning_rate": 3.2353481012658226e-05,
      "loss": 0.0002,
      "step": 5925
    },
    {
      "epoch": 55.60747663551402,
      "grad_norm": 0.0051350980065763,
      "learning_rate": 3.232974683544304e-05,
      "loss": 0.0002,
      "step": 5950
    },
    {
      "epoch": 55.8411214953271,
      "grad_norm": 0.00548820523545146,
      "learning_rate": 3.2306012658227845e-05,
      "loss": 0.0002,
      "step": 5975
    },
    {
      "epoch": 56.074766355140184,
      "grad_norm": 0.004812245722860098,
      "learning_rate": 3.228227848101266e-05,
      "loss": 0.0002,
      "step": 6000
    },
    {
      "epoch": 56.074766355140184,
      "eval_loss": 0.45458748936653137,
      "eval_runtime": 63.1244,
      "eval_samples_per_second": 215.923,
      "eval_steps_per_second": 1.695,
      "eval_wer": 20.170215104987356,
      "step": 6000
    },
    {
      "epoch": 56.308411214953274,
      "grad_norm": 0.00498458556830883,
      "learning_rate": 3.2258544303797464e-05,
      "loss": 0.0002,
      "step": 6025
    },
    {
      "epoch": 56.54205607476636,
      "grad_norm": 0.0048084803856909275,
      "learning_rate": 3.2234810126582276e-05,
      "loss": 0.0002,
      "step": 6050
    },
    {
      "epoch": 56.77570093457944,
      "grad_norm": 0.004982407670468092,
      "learning_rate": 3.221107594936709e-05,
      "loss": 0.0002,
      "step": 6075
    },
    {
      "epoch": 57.00934579439252,
      "grad_norm": 0.004549420904368162,
      "learning_rate": 3.2187341772151895e-05,
      "loss": 0.0002,
      "step": 6100
    },
    {
      "epoch": 57.242990654205606,
      "grad_norm": 0.004289143718779087,
      "learning_rate": 3.216360759493671e-05,
      "loss": 0.0002,
      "step": 6125
    },
    {
      "epoch": 57.47663551401869,
      "grad_norm": 0.0047737290151417255,
      "learning_rate": 3.2139873417721515e-05,
      "loss": 0.0002,
      "step": 6150
    },
    {
      "epoch": 57.71028037383178,
      "grad_norm": 0.00470471428707242,
      "learning_rate": 3.211613924050633e-05,
      "loss": 0.0002,
      "step": 6175
    },
    {
      "epoch": 57.94392523364486,
      "grad_norm": 0.004886131267994642,
      "learning_rate": 3.2092405063291134e-05,
      "loss": 0.0002,
      "step": 6200
    },
    {
      "epoch": 58.177570093457945,
      "grad_norm": 0.004642590414732695,
      "learning_rate": 3.2068670886075946e-05,
      "loss": 0.0002,
      "step": 6225
    },
    {
      "epoch": 58.41121495327103,
      "grad_norm": 0.004558751359581947,
      "learning_rate": 3.204493670886076e-05,
      "loss": 0.0002,
      "step": 6250
    },
    {
      "epoch": 58.64485981308411,
      "grad_norm": 0.004408851731568575,
      "learning_rate": 3.2021202531645565e-05,
      "loss": 0.0002,
      "step": 6275
    },
    {
      "epoch": 58.87850467289719,
      "grad_norm": 0.0045586684718728065,
      "learning_rate": 3.199746835443038e-05,
      "loss": 0.0002,
      "step": 6300
    },
    {
      "epoch": 59.11214953271028,
      "grad_norm": 0.0039480444975197315,
      "learning_rate": 3.1973734177215184e-05,
      "loss": 0.0002,
      "step": 6325
    },
    {
      "epoch": 59.345794392523366,
      "grad_norm": 0.004225443117320538,
      "learning_rate": 3.195e-05,
      "loss": 0.0002,
      "step": 6350
    },
    {
      "epoch": 59.57943925233645,
      "grad_norm": 0.004148137290030718,
      "learning_rate": 3.1926265822784803e-05,
      "loss": 0.0002,
      "step": 6375
    },
    {
      "epoch": 59.81308411214953,
      "grad_norm": 0.0041252258233726025,
      "learning_rate": 3.1902531645569616e-05,
      "loss": 0.0002,
      "step": 6400
    },
    {
      "epoch": 60.046728971962615,
      "grad_norm": 0.004037890117615461,
      "learning_rate": 3.187879746835443e-05,
      "loss": 0.0002,
      "step": 6425
    },
    {
      "epoch": 60.2803738317757,
      "grad_norm": 0.0038345332723110914,
      "learning_rate": 3.1855063291139235e-05,
      "loss": 0.0001,
      "step": 6450
    },
    {
      "epoch": 60.51401869158879,
      "grad_norm": 0.00392970209941268,
      "learning_rate": 3.183132911392405e-05,
      "loss": 0.0001,
      "step": 6475
    },
    {
      "epoch": 60.74766355140187,
      "grad_norm": 0.003990889992564917,
      "learning_rate": 3.1807594936708854e-05,
      "loss": 0.0001,
      "step": 6500
    },
    {
      "epoch": 60.981308411214954,
      "grad_norm": 0.004070672672241926,
      "learning_rate": 3.178386075949367e-05,
      "loss": 0.0002,
      "step": 6525
    },
    {
      "epoch": 61.21495327102804,
      "grad_norm": 0.0036899985279887915,
      "learning_rate": 3.1760126582278473e-05,
      "loss": 0.0001,
      "step": 6550
    },
    {
      "epoch": 61.44859813084112,
      "grad_norm": 0.003702005837112665,
      "learning_rate": 3.1736392405063286e-05,
      "loss": 0.0001,
      "step": 6575
    },
    {
      "epoch": 61.6822429906542,
      "grad_norm": 0.003752394113689661,
      "learning_rate": 3.17126582278481e-05,
      "loss": 0.0001,
      "step": 6600
    },
    {
      "epoch": 61.91588785046729,
      "grad_norm": 0.0037900095339864492,
      "learning_rate": 3.168892405063291e-05,
      "loss": 0.0001,
      "step": 6625
    },
    {
      "epoch": 62.149532710280376,
      "grad_norm": 0.0035399282351136208,
      "learning_rate": 3.166518987341772e-05,
      "loss": 0.0001,
      "step": 6650
    },
    {
      "epoch": 62.38317757009346,
      "grad_norm": 0.003451088210567832,
      "learning_rate": 3.164145569620253e-05,
      "loss": 0.0001,
      "step": 6675
    },
    {
      "epoch": 62.61682242990654,
      "grad_norm": 0.0036648407112807035,
      "learning_rate": 3.1617721518987344e-05,
      "loss": 0.0001,
      "step": 6700
    },
    {
      "epoch": 62.850467289719624,
      "grad_norm": 0.003599660936743021,
      "learning_rate": 3.159398734177215e-05,
      "loss": 0.0001,
      "step": 6725
    },
    {
      "epoch": 63.08411214953271,
      "grad_norm": 0.0032180894631892443,
      "learning_rate": 3.157025316455696e-05,
      "loss": 0.0001,
      "step": 6750
    },
    {
      "epoch": 63.3177570093458,
      "grad_norm": 0.0031950839329510927,
      "learning_rate": 3.154651898734177e-05,
      "loss": 0.0001,
      "step": 6775
    },
    {
      "epoch": 63.55140186915888,
      "grad_norm": 0.0033521002624183893,
      "learning_rate": 3.152278481012658e-05,
      "loss": 0.0001,
      "step": 6800
    },
    {
      "epoch": 63.78504672897196,
      "grad_norm": 0.0032866327092051506,
      "learning_rate": 3.149905063291139e-05,
      "loss": 0.0001,
      "step": 6825
    },
    {
      "epoch": 64.01869158878505,
      "grad_norm": 0.003140072338283062,
      "learning_rate": 3.14753164556962e-05,
      "loss": 0.0001,
      "step": 6850
    },
    {
      "epoch": 64.25233644859813,
      "grad_norm": 0.003216780023649335,
      "learning_rate": 3.1451582278481014e-05,
      "loss": 0.0001,
      "step": 6875
    },
    {
      "epoch": 64.48598130841121,
      "grad_norm": 0.0032517279032617807,
      "learning_rate": 3.142784810126582e-05,
      "loss": 0.0001,
      "step": 6900
    },
    {
      "epoch": 64.7196261682243,
      "grad_norm": 0.0032107916194945574,
      "learning_rate": 3.140411392405063e-05,
      "loss": 0.0001,
      "step": 6925
    },
    {
      "epoch": 64.95327102803738,
      "grad_norm": 0.0031511501874774694,
      "learning_rate": 3.138037974683544e-05,
      "loss": 0.0001,
      "step": 6950
    },
    {
      "epoch": 65.18691588785046,
      "grad_norm": 0.002939599333330989,
      "learning_rate": 3.135664556962025e-05,
      "loss": 0.0001,
      "step": 6975
    },
    {
      "epoch": 65.42056074766356,
      "grad_norm": 0.0030510067008435726,
      "learning_rate": 3.133291139240506e-05,
      "loss": 0.0001,
      "step": 7000
    },
    {
      "epoch": 65.42056074766356,
      "eval_loss": 0.4744671881198883,
      "eval_runtime": 63.2415,
      "eval_samples_per_second": 215.523,
      "eval_steps_per_second": 1.692,
      "eval_wer": 20.217853347502658,
      "step": 7000
    },
    {
      "epoch": 65.65420560747664,
      "grad_norm": 0.0032058239448815584,
      "learning_rate": 3.130917721518987e-05,
      "loss": 0.0001,
      "step": 7025
    },
    {
      "epoch": 65.88785046728972,
      "grad_norm": 0.0030427316669374704,
      "learning_rate": 3.1285443037974684e-05,
      "loss": 0.0001,
      "step": 7050
    },
    {
      "epoch": 66.1214953271028,
      "grad_norm": 0.0028826987836509943,
      "learning_rate": 3.126170886075949e-05,
      "loss": 0.0001,
      "step": 7075
    },
    {
      "epoch": 66.35514018691589,
      "grad_norm": 0.0028522161301225424,
      "learning_rate": 3.12379746835443e-05,
      "loss": 0.0001,
      "step": 7100
    },
    {
      "epoch": 66.58878504672897,
      "grad_norm": 0.002932359464466572,
      "learning_rate": 3.121424050632911e-05,
      "loss": 0.0001,
      "step": 7125
    },
    {
      "epoch": 66.82242990654206,
      "grad_norm": 0.003014651359990239,
      "learning_rate": 3.119050632911392e-05,
      "loss": 0.0001,
      "step": 7150
    },
    {
      "epoch": 67.05607476635514,
      "grad_norm": 0.0027976776473224163,
      "learning_rate": 3.116677215189873e-05,
      "loss": 0.0001,
      "step": 7175
    },
    {
      "epoch": 67.28971962616822,
      "grad_norm": 0.0027002531569451094,
      "learning_rate": 3.114303797468354e-05,
      "loss": 0.0001,
      "step": 7200
    },
    {
      "epoch": 67.5233644859813,
      "grad_norm": 0.002804843010380864,
      "learning_rate": 3.1119303797468354e-05,
      "loss": 0.0001,
      "step": 7225
    },
    {
      "epoch": 67.75700934579439,
      "grad_norm": 0.0026762145571410656,
      "learning_rate": 3.109556962025316e-05,
      "loss": 0.0001,
      "step": 7250
    },
    {
      "epoch": 67.99065420560747,
      "grad_norm": 0.0027852137573063374,
      "learning_rate": 3.107183544303797e-05,
      "loss": 0.0001,
      "step": 7275
    },
    {
      "epoch": 68.22429906542057,
      "grad_norm": 0.002483020769432187,
      "learning_rate": 3.104810126582278e-05,
      "loss": 0.0001,
      "step": 7300
    },
    {
      "epoch": 68.45794392523365,
      "grad_norm": 0.0026285869535058737,
      "learning_rate": 3.102436708860759e-05,
      "loss": 0.0001,
      "step": 7325
    },
    {
      "epoch": 68.69158878504673,
      "grad_norm": 0.002632495714351535,
      "learning_rate": 3.10006329113924e-05,
      "loss": 0.0001,
      "step": 7350
    },
    {
      "epoch": 68.92523364485982,
      "grad_norm": 0.002577392850071192,
      "learning_rate": 3.097689873417721e-05,
      "loss": 0.0001,
      "step": 7375
    },
    {
      "epoch": 69.1588785046729,
      "grad_norm": 0.0023590901400893927,
      "learning_rate": 3.0953164556962024e-05,
      "loss": 0.0001,
      "step": 7400
    },
    {
      "epoch": 69.39252336448598,
      "grad_norm": 0.002418393502011895,
      "learning_rate": 3.092943037974683e-05,
      "loss": 0.0001,
      "step": 7425
    },
    {
      "epoch": 69.62616822429906,
      "grad_norm": 0.0023916559293866158,
      "learning_rate": 3.090569620253164e-05,
      "loss": 0.0001,
      "step": 7450
    },
    {
      "epoch": 69.85981308411215,
      "grad_norm": 0.002579368883743882,
      "learning_rate": 3.0881962025316456e-05,
      "loss": 0.0001,
      "step": 7475
    },
    {
      "epoch": 70.09345794392523,
      "grad_norm": 0.00250100321136415,
      "learning_rate": 3.085822784810126e-05,
      "loss": 0.0001,
      "step": 7500
    },
    {
      "epoch": 70.32710280373831,
      "grad_norm": 0.002333480166271329,
      "learning_rate": 3.0834493670886075e-05,
      "loss": 0.0001,
      "step": 7525
    },
    {
      "epoch": 70.5607476635514,
      "grad_norm": 0.0024163569323718548,
      "learning_rate": 3.081075949367089e-05,
      "loss": 0.0001,
      "step": 7550
    },
    {
      "epoch": 70.79439252336448,
      "grad_norm": 0.0024094872642308474,
      "learning_rate": 3.0787025316455694e-05,
      "loss": 0.0001,
      "step": 7575
    },
    {
      "epoch": 71.02803738317758,
      "grad_norm": 0.0022144243121147156,
      "learning_rate": 3.0763291139240507e-05,
      "loss": 0.0001,
      "step": 7600
    },
    {
      "epoch": 71.26168224299066,
      "grad_norm": 0.0022461097687482834,
      "learning_rate": 3.073955696202531e-05,
      "loss": 0.0001,
      "step": 7625
    },
    {
      "epoch": 71.49532710280374,
      "grad_norm": 0.002173445653170347,
      "learning_rate": 3.0715822784810126e-05,
      "loss": 0.0001,
      "step": 7650
    },
    {
      "epoch": 71.72897196261682,
      "grad_norm": 0.0022633324842900038,
      "learning_rate": 3.069208860759493e-05,
      "loss": 0.0001,
      "step": 7675
    },
    {
      "epoch": 71.96261682242991,
      "grad_norm": 0.002266242867335677,
      "learning_rate": 3.0668354430379745e-05,
      "loss": 0.0001,
      "step": 7700
    },
    {
      "epoch": 72.19626168224299,
      "grad_norm": 0.002105720341205597,
      "learning_rate": 3.064462025316456e-05,
      "loss": 0.0001,
      "step": 7725
    },
    {
      "epoch": 72.42990654205607,
      "grad_norm": 0.002207930898293853,
      "learning_rate": 3.0620886075949364e-05,
      "loss": 0.0001,
      "step": 7750
    },
    {
      "epoch": 72.66355140186916,
      "grad_norm": 0.0022758631967008114,
      "learning_rate": 3.0597151898734176e-05,
      "loss": 0.0001,
      "step": 7775
    },
    {
      "epoch": 72.89719626168224,
      "grad_norm": 0.002142318058758974,
      "learning_rate": 3.057341772151898e-05,
      "loss": 0.0001,
      "step": 7800
    },
    {
      "epoch": 73.13084112149532,
      "grad_norm": 0.001979030901566148,
      "learning_rate": 3.0549683544303796e-05,
      "loss": 0.0001,
      "step": 7825
    },
    {
      "epoch": 73.3644859813084,
      "grad_norm": 0.002008823910728097,
      "learning_rate": 3.05259493670886e-05,
      "loss": 0.0001,
      "step": 7850
    },
    {
      "epoch": 73.59813084112149,
      "grad_norm": 0.0020186167676001787,
      "learning_rate": 3.0502215189873415e-05,
      "loss": 0.0001,
      "step": 7875
    },
    {
      "epoch": 73.83177570093459,
      "grad_norm": 0.0020830687135457993,
      "learning_rate": 3.0478481012658224e-05,
      "loss": 0.0001,
      "step": 7900
    },
    {
      "epoch": 74.06542056074767,
      "grad_norm": 0.001894044573418796,
      "learning_rate": 3.0454746835443034e-05,
      "loss": 0.0001,
      "step": 7925
    },
    {
      "epoch": 74.29906542056075,
      "grad_norm": 0.0018573316046968102,
      "learning_rate": 3.0431012658227846e-05,
      "loss": 0.0001,
      "step": 7950
    },
    {
      "epoch": 74.53271028037383,
      "grad_norm": 0.0019013582495972514,
      "learning_rate": 3.0407278481012656e-05,
      "loss": 0.0001,
      "step": 7975
    },
    {
      "epoch": 74.76635514018692,
      "grad_norm": 0.0020685612689703703,
      "learning_rate": 3.0383544303797465e-05,
      "loss": 0.0001,
      "step": 8000
    },
    {
      "epoch": 74.76635514018692,
      "eval_loss": 0.494055837392807,
      "eval_runtime": 63.3615,
      "eval_samples_per_second": 215.115,
      "eval_steps_per_second": 1.689,
      "eval_wer": 20.199530946535234,
      "step": 8000
    },
    {
      "epoch": 75.0,
      "grad_norm": 0.0028726994059979916,
      "learning_rate": 3.0359810126582275e-05,
      "loss": 0.0001,
      "step": 8025
    },
    {
      "epoch": 75.23364485981308,
      "grad_norm": 0.0018449650378897786,
      "learning_rate": 3.0336075949367084e-05,
      "loss": 0.0001,
      "step": 8050
    },
    {
      "epoch": 75.46728971962617,
      "grad_norm": 0.001773087540641427,
      "learning_rate": 3.0312341772151894e-05,
      "loss": 0.0001,
      "step": 8075
    },
    {
      "epoch": 75.70093457943925,
      "grad_norm": 0.0019065611995756626,
      "learning_rate": 3.0288607594936703e-05,
      "loss": 0.0001,
      "step": 8100
    },
    {
      "epoch": 75.93457943925233,
      "grad_norm": 0.0018953917315229774,
      "learning_rate": 3.0264873417721516e-05,
      "loss": 0.0001,
      "step": 8125
    },
    {
      "epoch": 76.16822429906541,
      "grad_norm": 0.001699391519650817,
      "learning_rate": 3.0241139240506326e-05,
      "loss": 0.0001,
      "step": 8150
    },
    {
      "epoch": 76.40186915887851,
      "grad_norm": 0.0016678664833307266,
      "learning_rate": 3.0217405063291135e-05,
      "loss": 0.0001,
      "step": 8175
    },
    {
      "epoch": 76.6355140186916,
      "grad_norm": 0.0017894869670271873,
      "learning_rate": 3.0193670886075945e-05,
      "loss": 0.0001,
      "step": 8200
    },
    {
      "epoch": 76.86915887850468,
      "grad_norm": 0.001669207587838173,
      "learning_rate": 3.0169936708860754e-05,
      "loss": 0.0001,
      "step": 8225
    },
    {
      "epoch": 77.10280373831776,
      "grad_norm": 0.001707504503428936,
      "learning_rate": 3.0146202531645564e-05,
      "loss": 0.0001,
      "step": 8250
    },
    {
      "epoch": 77.33644859813084,
      "grad_norm": 0.0016919142799451947,
      "learning_rate": 3.012246835443038e-05,
      "loss": 0.0001,
      "step": 8275
    },
    {
      "epoch": 77.57009345794393,
      "grad_norm": 0.0016664247959852219,
      "learning_rate": 3.009873417721519e-05,
      "loss": 0.0001,
      "step": 8300
    },
    {
      "epoch": 77.80373831775701,
      "grad_norm": 0.0017846949631348252,
      "learning_rate": 3.0075e-05,
      "loss": 0.0001,
      "step": 8325
    },
    {
      "epoch": 78.03738317757009,
      "grad_norm": 0.0016335134860128164,
      "learning_rate": 3.005126582278481e-05,
      "loss": 0.0001,
      "step": 8350
    },
    {
      "epoch": 78.27102803738318,
      "grad_norm": 0.0015427300240844488,
      "learning_rate": 3.0027531645569618e-05,
      "loss": 0.0001,
      "step": 8375
    },
    {
      "epoch": 78.50467289719626,
      "grad_norm": 0.0016644004499539733,
      "learning_rate": 3.0003797468354428e-05,
      "loss": 0.0001,
      "step": 8400
    },
    {
      "epoch": 78.73831775700934,
      "grad_norm": 0.0015361284604296088,
      "learning_rate": 2.998006329113924e-05,
      "loss": 0.0001,
      "step": 8425
    },
    {
      "epoch": 78.97196261682242,
      "grad_norm": 0.0015355064533650875,
      "learning_rate": 2.995632911392405e-05,
      "loss": 0.0001,
      "step": 8450
    },
    {
      "epoch": 79.20560747663552,
      "grad_norm": 0.0014871564926579595,
      "learning_rate": 2.993259493670886e-05,
      "loss": 0.0001,
      "step": 8475
    },
    {
      "epoch": 79.4392523364486,
      "grad_norm": 0.0014554571826010942,
      "learning_rate": 2.990886075949367e-05,
      "loss": 0.0001,
      "step": 8500
    },
    {
      "epoch": 79.67289719626169,
      "grad_norm": 0.0014896997017785907,
      "learning_rate": 2.988512658227848e-05,
      "loss": 0.0001,
      "step": 8525
    },
    {
      "epoch": 79.90654205607477,
      "grad_norm": 0.0015526514034718275,
      "learning_rate": 2.9861392405063288e-05,
      "loss": 0.0001,
      "step": 8550
    },
    {
      "epoch": 80.14018691588785,
      "grad_norm": 0.0014903959818184376,
      "learning_rate": 2.9837658227848098e-05,
      "loss": 0.0001,
      "step": 8575
    },
    {
      "epoch": 80.37383177570094,
      "grad_norm": 0.001390935154631734,
      "learning_rate": 2.981392405063291e-05,
      "loss": 0.0001,
      "step": 8600
    },
    {
      "epoch": 80.60747663551402,
      "grad_norm": 0.0014142347499728203,
      "learning_rate": 2.979018987341772e-05,
      "loss": 0.0001,
      "step": 8625
    },
    {
      "epoch": 80.8411214953271,
      "grad_norm": 0.0014390508877113461,
      "learning_rate": 2.976645569620253e-05,
      "loss": 0.0001,
      "step": 8650
    },
    {
      "epoch": 81.07476635514018,
      "grad_norm": 0.0013639021199196577,
      "learning_rate": 2.974272151898734e-05,
      "loss": 0.0001,
      "step": 8675
    },
    {
      "epoch": 81.30841121495327,
      "grad_norm": 0.001334126922301948,
      "learning_rate": 2.971898734177215e-05,
      "loss": 0.0001,
      "step": 8700
    },
    {
      "epoch": 81.54205607476635,
      "grad_norm": 0.0014011391904205084,
      "learning_rate": 2.9695253164556958e-05,
      "loss": 0.0001,
      "step": 8725
    },
    {
      "epoch": 81.77570093457943,
      "grad_norm": 0.0013581993989646435,
      "learning_rate": 2.9671518987341768e-05,
      "loss": 0.0,
      "step": 8750
    },
    {
      "epoch": 82.00934579439253,
      "grad_norm": 0.0013259041588753462,
      "learning_rate": 2.964778481012658e-05,
      "loss": 0.0,
      "step": 8775
    },
    {
      "epoch": 82.24299065420561,
      "grad_norm": 0.0012890142388641834,
      "learning_rate": 2.962405063291139e-05,
      "loss": 0.0,
      "step": 8800
    },
    {
      "epoch": 82.4766355140187,
      "grad_norm": 0.0012812621425837278,
      "learning_rate": 2.96003164556962e-05,
      "loss": 0.0,
      "step": 8825
    },
    {
      "epoch": 82.71028037383178,
      "grad_norm": 0.0012719089863821864,
      "learning_rate": 2.957658227848101e-05,
      "loss": 0.0,
      "step": 8850
    },
    {
      "epoch": 82.94392523364486,
      "grad_norm": 0.0013342360034585,
      "learning_rate": 2.955284810126582e-05,
      "loss": 0.0,
      "step": 8875
    },
    {
      "epoch": 83.17757009345794,
      "grad_norm": 0.0011831733863800764,
      "learning_rate": 2.9529113924050628e-05,
      "loss": 0.0,
      "step": 8900
    },
    {
      "epoch": 83.41121495327103,
      "grad_norm": 0.0012514119734987617,
      "learning_rate": 2.9505379746835438e-05,
      "loss": 0.0,
      "step": 8925
    },
    {
      "epoch": 83.64485981308411,
      "grad_norm": 0.0013140386436134577,
      "learning_rate": 2.948164556962025e-05,
      "loss": 0.0,
      "step": 8950
    },
    {
      "epoch": 83.8785046728972,
      "grad_norm": 0.0012639579363167286,
      "learning_rate": 2.945791139240506e-05,
      "loss": 0.0,
      "step": 8975
    },
    {
      "epoch": 84.11214953271028,
      "grad_norm": 0.0011605726322159171,
      "learning_rate": 2.943417721518987e-05,
      "loss": 0.0,
      "step": 9000
    },
    {
      "epoch": 84.11214953271028,
      "eval_loss": 0.5142388939857483,
      "eval_runtime": 62.9672,
      "eval_samples_per_second": 216.462,
      "eval_steps_per_second": 1.699,
      "eval_wer": 20.33420059364579,
      "step": 9000
    },
    {
      "epoch": 84.34579439252336,
      "grad_norm": 0.0011589956702664495,
      "learning_rate": 2.941044303797468e-05,
      "loss": 0.0,
      "step": 9025
    },
    {
      "epoch": 84.57943925233644,
      "grad_norm": 0.001201428472995758,
      "learning_rate": 2.938670886075949e-05,
      "loss": 0.0,
      "step": 9050
    },
    {
      "epoch": 84.81308411214954,
      "grad_norm": 0.0012158616445958614,
      "learning_rate": 2.9362974683544298e-05,
      "loss": 0.0,
      "step": 9075
    },
    {
      "epoch": 85.04672897196262,
      "grad_norm": 0.0010845279321074486,
      "learning_rate": 2.9339240506329114e-05,
      "loss": 0.0,
      "step": 9100
    },
    {
      "epoch": 85.2803738317757,
      "grad_norm": 0.001130130491219461,
      "learning_rate": 2.9315506329113924e-05,
      "loss": 0.0,
      "step": 9125
    },
    {
      "epoch": 85.51401869158879,
      "grad_norm": 0.0011497740633785725,
      "learning_rate": 2.9291772151898733e-05,
      "loss": 0.0,
      "step": 9150
    },
    {
      "epoch": 85.74766355140187,
      "grad_norm": 0.001153731718659401,
      "learning_rate": 2.9268037974683543e-05,
      "loss": 0.0,
      "step": 9175
    },
    {
      "epoch": 85.98130841121495,
      "grad_norm": 0.0011366461403667927,
      "learning_rate": 2.9244303797468352e-05,
      "loss": 0.0,
      "step": 9200
    },
    {
      "epoch": 86.21495327102804,
      "grad_norm": 0.0010454368311911821,
      "learning_rate": 2.9220569620253162e-05,
      "loss": 0.0,
      "step": 9225
    },
    {
      "epoch": 86.44859813084112,
      "grad_norm": 0.0010803466429933906,
      "learning_rate": 2.9196835443037975e-05,
      "loss": 0.0,
      "step": 9250
    },
    {
      "epoch": 86.6822429906542,
      "grad_norm": 0.0010971734300255775,
      "learning_rate": 2.9173101265822784e-05,
      "loss": 0.0,
      "step": 9275
    },
    {
      "epoch": 86.91588785046729,
      "grad_norm": 0.0010489817941561341,
      "learning_rate": 2.9149367088607594e-05,
      "loss": 0.0,
      "step": 9300
    },
    {
      "epoch": 87.14953271028037,
      "grad_norm": 0.0010429786052554846,
      "learning_rate": 2.9125632911392403e-05,
      "loss": 0.0,
      "step": 9325
    },
    {
      "epoch": 87.38317757009345,
      "grad_norm": 0.0010035474551841617,
      "learning_rate": 2.9101898734177213e-05,
      "loss": 0.0,
      "step": 9350
    },
    {
      "epoch": 87.61682242990655,
      "grad_norm": 0.0010115285404026508,
      "learning_rate": 2.9078164556962022e-05,
      "loss": 0.0,
      "step": 9375
    },
    {
      "epoch": 87.85046728971963,
      "grad_norm": 0.0010733642848208547,
      "learning_rate": 2.9054430379746832e-05,
      "loss": 0.0,
      "step": 9400
    },
    {
      "epoch": 88.08411214953271,
      "grad_norm": 0.0009068447980098426,
      "learning_rate": 2.9030696202531645e-05,
      "loss": 0.0,
      "step": 9425
    },
    {
      "epoch": 88.3177570093458,
      "grad_norm": 0.001027858816087246,
      "learning_rate": 2.9006962025316454e-05,
      "loss": 0.0,
      "step": 9450
    },
    {
      "epoch": 88.55140186915888,
      "grad_norm": 0.0009572837734594941,
      "learning_rate": 2.8983227848101264e-05,
      "loss": 0.0,
      "step": 9475
    },
    {
      "epoch": 88.78504672897196,
      "grad_norm": 0.0010000813053920865,
      "learning_rate": 2.8959493670886073e-05,
      "loss": 0.0,
      "step": 9500
    },
    {
      "epoch": 89.01869158878505,
      "grad_norm": 0.0008946892339736223,
      "learning_rate": 2.8935759493670883e-05,
      "loss": 0.0,
      "step": 9525
    },
    {
      "epoch": 89.25233644859813,
      "grad_norm": 0.000919956190045923,
      "learning_rate": 2.8912025316455692e-05,
      "loss": 0.0,
      "step": 9550
    },
    {
      "epoch": 89.48598130841121,
      "grad_norm": 0.0008721091435290873,
      "learning_rate": 2.8888291139240505e-05,
      "loss": 0.0,
      "step": 9575
    },
    {
      "epoch": 89.7196261682243,
      "grad_norm": 0.0009453958482481539,
      "learning_rate": 2.8864556962025315e-05,
      "loss": 0.0,
      "step": 9600
    },
    {
      "epoch": 89.95327102803738,
      "grad_norm": 0.0009352854685857892,
      "learning_rate": 2.8840822784810124e-05,
      "loss": 0.0,
      "step": 9625
    },
    {
      "epoch": 90.18691588785046,
      "grad_norm": 0.0008482401608489454,
      "learning_rate": 2.8817088607594934e-05,
      "loss": 0.0,
      "step": 9650
    },
    {
      "epoch": 90.42056074766356,
      "grad_norm": 0.0008235819404944777,
      "learning_rate": 2.8793354430379743e-05,
      "loss": 0.0,
      "step": 9675
    },
    {
      "epoch": 90.65420560747664,
      "grad_norm": 0.0009272462921217084,
      "learning_rate": 2.8769620253164553e-05,
      "loss": 0.0,
      "step": 9700
    },
    {
      "epoch": 90.88785046728972,
      "grad_norm": 0.0009033215465024114,
      "learning_rate": 2.8745886075949362e-05,
      "loss": 0.0,
      "step": 9725
    },
    {
      "epoch": 91.1214953271028,
      "grad_norm": 0.0008244973723776639,
      "learning_rate": 2.8722151898734175e-05,
      "loss": 0.0,
      "step": 9750
    },
    {
      "epoch": 91.35514018691589,
      "grad_norm": 0.0008386572008021176,
      "learning_rate": 2.8698417721518984e-05,
      "loss": 0.0,
      "step": 9775
    },
    {
      "epoch": 91.58878504672897,
      "grad_norm": 0.000844955095089972,
      "learning_rate": 2.8674683544303794e-05,
      "loss": 0.0,
      "step": 9800
    },
    {
      "epoch": 91.82242990654206,
      "grad_norm": 0.0007930775755085051,
      "learning_rate": 2.8650949367088603e-05,
      "loss": 0.0,
      "step": 9825
    },
    {
      "epoch": 92.05607476635514,
      "grad_norm": 0.00079232023563236,
      "learning_rate": 2.8627215189873413e-05,
      "loss": 0.0,
      "step": 9850
    },
    {
      "epoch": 92.28971962616822,
      "grad_norm": 0.0007878366741351783,
      "learning_rate": 2.8603481012658222e-05,
      "loss": 0.0,
      "step": 9875
    },
    {
      "epoch": 92.5233644859813,
      "grad_norm": 0.0008155304822139442,
      "learning_rate": 2.8579746835443032e-05,
      "loss": 0.0,
      "step": 9900
    },
    {
      "epoch": 92.75700934579439,
      "grad_norm": 0.0007804965134710073,
      "learning_rate": 2.8556012658227848e-05,
      "loss": 0.0,
      "step": 9925
    },
    {
      "epoch": 92.99065420560747,
      "grad_norm": 0.000814582861494273,
      "learning_rate": 2.8532278481012658e-05,
      "loss": 0.0,
      "step": 9950
    },
    {
      "epoch": 93.22429906542057,
      "grad_norm": 0.000759601010940969,
      "learning_rate": 2.8508544303797467e-05,
      "loss": 0.0,
      "step": 9975
    },
    {
      "epoch": 93.45794392523365,
      "grad_norm": 0.0007659905240871012,
      "learning_rate": 2.8484810126582277e-05,
      "loss": 0.0,
      "step": 10000
    },
    {
      "epoch": 93.45794392523365,
      "eval_loss": 0.5353172421455383,
      "eval_runtime": 63.4927,
      "eval_samples_per_second": 214.671,
      "eval_steps_per_second": 1.685,
      "eval_wer": 20.438638279160102,
      "step": 10000
    },
    {
      "epoch": 93.69158878504673,
      "grad_norm": 0.0007560112280771136,
      "learning_rate": 2.8461075949367086e-05,
      "loss": 0.0,
      "step": 10025
    },
    {
      "epoch": 93.92523364485982,
      "grad_norm": 0.0007651768974028528,
      "learning_rate": 2.84373417721519e-05,
      "loss": 0.0,
      "step": 10050
    },
    {
      "epoch": 94.1588785046729,
      "grad_norm": 0.0007016289164312184,
      "learning_rate": 2.841360759493671e-05,
      "loss": 0.0,
      "step": 10075
    },
    {
      "epoch": 94.39252336448598,
      "grad_norm": 0.0007208699826151133,
      "learning_rate": 2.8389873417721518e-05,
      "loss": 0.0,
      "step": 10100
    },
    {
      "epoch": 94.62616822429906,
      "grad_norm": 0.0007434538565576077,
      "learning_rate": 2.8366139240506328e-05,
      "loss": 0.0,
      "step": 10125
    },
    {
      "epoch": 94.85981308411215,
      "grad_norm": 0.0007124239928089082,
      "learning_rate": 2.8342405063291137e-05,
      "loss": 0.0,
      "step": 10150
    },
    {
      "epoch": 95.09345794392523,
      "grad_norm": 0.0006733191548846662,
      "learning_rate": 2.8318670886075947e-05,
      "loss": 0.0,
      "step": 10175
    },
    {
      "epoch": 95.32710280373831,
      "grad_norm": 0.0006819823756814003,
      "learning_rate": 2.8294936708860756e-05,
      "loss": 0.0,
      "step": 10200
    },
    {
      "epoch": 95.5607476635514,
      "grad_norm": 0.0006861250731162727,
      "learning_rate": 2.827120253164557e-05,
      "loss": 0.0,
      "step": 10225
    },
    {
      "epoch": 95.79439252336448,
      "grad_norm": 0.0007079722126945853,
      "learning_rate": 2.824746835443038e-05,
      "loss": 0.0,
      "step": 10250
    },
    {
      "epoch": 96.02803738317758,
      "grad_norm": 0.0006552456179633737,
      "learning_rate": 2.8223734177215188e-05,
      "loss": 0.0,
      "step": 10275
    },
    {
      "epoch": 96.26168224299066,
      "grad_norm": 0.0006260311347432435,
      "learning_rate": 2.8199999999999998e-05,
      "loss": 0.0,
      "step": 10300
    },
    {
      "epoch": 96.49532710280374,
      "grad_norm": 0.0006413419614546001,
      "learning_rate": 2.8176265822784807e-05,
      "loss": 0.0,
      "step": 10325
    },
    {
      "epoch": 96.72897196261682,
      "grad_norm": 0.0006349146133288741,
      "learning_rate": 2.8152531645569617e-05,
      "loss": 0.0,
      "step": 10350
    },
    {
      "epoch": 96.96261682242991,
      "grad_norm": 0.0006550104590132833,
      "learning_rate": 2.8128797468354426e-05,
      "loss": 0.0,
      "step": 10375
    },
    {
      "epoch": 97.19626168224299,
      "grad_norm": 0.0006043660687282681,
      "learning_rate": 2.810506329113924e-05,
      "loss": 0.0,
      "step": 10400
    },
    {
      "epoch": 97.42990654205607,
      "grad_norm": 0.0005886057624593377,
      "learning_rate": 2.808132911392405e-05,
      "loss": 0.0,
      "step": 10425
    },
    {
      "epoch": 97.66355140186916,
      "grad_norm": 0.000634817173704505,
      "learning_rate": 2.8057594936708858e-05,
      "loss": 0.0,
      "step": 10450
    },
    {
      "epoch": 97.89719626168224,
      "grad_norm": 0.0006467931088991463,
      "learning_rate": 2.8033860759493668e-05,
      "loss": 0.0,
      "step": 10475
    },
    {
      "epoch": 98.13084112149532,
      "grad_norm": 0.00058325583813712,
      "learning_rate": 2.8010126582278477e-05,
      "loss": 0.0,
      "step": 10500
    },
    {
      "epoch": 98.3644859813084,
      "grad_norm": 0.0005941679701209068,
      "learning_rate": 2.7986392405063287e-05,
      "loss": 0.0,
      "step": 10525
    },
    {
      "epoch": 98.59813084112149,
      "grad_norm": 0.0005906676524318755,
      "learning_rate": 2.7962658227848096e-05,
      "loss": 0.0,
      "step": 10550
    },
    {
      "epoch": 98.83177570093459,
      "grad_norm": 0.0005812451709061861,
      "learning_rate": 2.793892405063291e-05,
      "loss": 0.0,
      "step": 10575
    },
    {
      "epoch": 99.06542056074767,
      "grad_norm": 0.0005404899129644036,
      "learning_rate": 2.791518987341772e-05,
      "loss": 0.0,
      "step": 10600
    },
    {
      "epoch": 99.29906542056075,
      "grad_norm": 0.0005504210712388158,
      "learning_rate": 2.7891455696202528e-05,
      "loss": 0.0,
      "step": 10625
    },
    {
      "epoch": 99.53271028037383,
      "grad_norm": 0.0005766512476839125,
      "learning_rate": 2.7867721518987338e-05,
      "loss": 0.0,
      "step": 10650
    },
    {
      "epoch": 99.76635514018692,
      "grad_norm": 0.0005945775774307549,
      "learning_rate": 2.7843987341772147e-05,
      "loss": 0.0,
      "step": 10675
    },
    {
      "epoch": 100.0,
      "grad_norm": 0.000835551240015775,
      "learning_rate": 2.7820253164556957e-05,
      "loss": 0.0,
      "step": 10700
    },
    {
      "epoch": 100.23364485981308,
      "grad_norm": 0.0005175265832804143,
      "learning_rate": 2.7796518987341773e-05,
      "loss": 0.0,
      "step": 10725
    },
    {
      "epoch": 100.46728971962617,
      "grad_norm": 0.0005436977371573448,
      "learning_rate": 2.7772784810126582e-05,
      "loss": 0.0,
      "step": 10750
    },
    {
      "epoch": 100.70093457943925,
      "grad_norm": 0.0005550860078074038,
      "learning_rate": 2.7749050632911392e-05,
      "loss": 0.0,
      "step": 10775
    },
    {
      "epoch": 100.93457943925233,
      "grad_norm": 0.0005275776493363082,
      "learning_rate": 2.77253164556962e-05,
      "loss": 0.0,
      "step": 10800
    },
    {
      "epoch": 101.16822429906541,
      "grad_norm": 0.0005124133895151317,
      "learning_rate": 2.770158227848101e-05,
      "loss": 0.0,
      "step": 10825
    },
    {
      "epoch": 101.40186915887851,
      "grad_norm": 0.0005023005069233477,
      "learning_rate": 2.767784810126582e-05,
      "loss": 0.0,
      "step": 10850
    },
    {
      "epoch": 101.6355140186916,
      "grad_norm": 0.0004804893978871405,
      "learning_rate": 2.7654113924050633e-05,
      "loss": 0.0,
      "step": 10875
    },
    {
      "epoch": 101.86915887850468,
      "grad_norm": 0.0005130966892465949,
      "learning_rate": 2.7630379746835443e-05,
      "loss": 0.0,
      "step": 10900
    },
    {
      "epoch": 102.10280373831776,
      "grad_norm": 0.0004586328286677599,
      "learning_rate": 2.7606645569620252e-05,
      "loss": 0.0,
      "step": 10925
    },
    {
      "epoch": 102.33644859813084,
      "grad_norm": 0.0004780337039846927,
      "learning_rate": 2.7582911392405062e-05,
      "loss": 0.0,
      "step": 10950
    },
    {
      "epoch": 102.57009345794393,
      "grad_norm": 0.0004985179984942079,
      "learning_rate": 2.755917721518987e-05,
      "loss": 0.0,
      "step": 10975
    },
    {
      "epoch": 102.80373831775701,
      "grad_norm": 0.0005429569282568991,
      "learning_rate": 2.753544303797468e-05,
      "loss": 0.0,
      "step": 11000
    },
    {
      "epoch": 102.80373831775701,
      "eval_loss": 0.5567328333854675,
      "eval_runtime": 63.6963,
      "eval_samples_per_second": 213.984,
      "eval_steps_per_second": 1.68,
      "eval_wer": 20.54948880501301,
      "step": 11000
    },
    {
      "epoch": 103.03738317757009,
      "grad_norm": 0.00046830964856781065,
      "learning_rate": 2.751170886075949e-05,
      "loss": 0.0,
      "step": 11025
    },
    {
      "epoch": 103.27102803738318,
      "grad_norm": 0.0004802467883564532,
      "learning_rate": 2.7487974683544303e-05,
      "loss": 0.0,
      "step": 11050
    },
    {
      "epoch": 103.50467289719626,
      "grad_norm": 0.0004503963573370129,
      "learning_rate": 2.7464240506329113e-05,
      "loss": 0.0,
      "step": 11075
    },
    {
      "epoch": 103.73831775700934,
      "grad_norm": 0.00045956505346111953,
      "learning_rate": 2.7440506329113922e-05,
      "loss": 0.0,
      "step": 11100
    },
    {
      "epoch": 103.97196261682242,
      "grad_norm": 0.0004916164325550199,
      "learning_rate": 2.7416772151898732e-05,
      "loss": 0.0,
      "step": 11125
    },
    {
      "epoch": 104.20560747663552,
      "grad_norm": 0.00044095932389609516,
      "learning_rate": 2.739303797468354e-05,
      "loss": 0.0,
      "step": 11150
    },
    {
      "epoch": 104.4392523364486,
      "grad_norm": 0.0004480903735384345,
      "learning_rate": 2.736930379746835e-05,
      "loss": 0.0,
      "step": 11175
    },
    {
      "epoch": 104.67289719626169,
      "grad_norm": 0.00044654516386799514,
      "learning_rate": 2.734556962025316e-05,
      "loss": 0.0,
      "step": 11200
    },
    {
      "epoch": 104.90654205607477,
      "grad_norm": 0.0004509761347435415,
      "learning_rate": 2.7321835443037973e-05,
      "loss": 0.0,
      "step": 11225
    },
    {
      "epoch": 105.14018691588785,
      "grad_norm": 0.0004024368245154619,
      "learning_rate": 2.7298101265822783e-05,
      "loss": 0.0,
      "step": 11250
    },
    {
      "epoch": 105.37383177570094,
      "grad_norm": 0.0004066548717673868,
      "learning_rate": 2.7274367088607592e-05,
      "loss": 0.0,
      "step": 11275
    },
    {
      "epoch": 105.60747663551402,
      "grad_norm": 0.00045356256305240095,
      "learning_rate": 2.72506329113924e-05,
      "loss": 0.0,
      "step": 11300
    },
    {
      "epoch": 105.8411214953271,
      "grad_norm": 0.000430036976467818,
      "learning_rate": 2.722689873417721e-05,
      "loss": 0.0,
      "step": 11325
    },
    {
      "epoch": 106.07476635514018,
      "grad_norm": 0.0003893323882948607,
      "learning_rate": 2.720316455696202e-05,
      "loss": 0.0,
      "step": 11350
    },
    {
      "epoch": 106.30841121495327,
      "grad_norm": 0.0004118061042390764,
      "learning_rate": 2.717943037974683e-05,
      "loss": 0.0,
      "step": 11375
    },
    {
      "epoch": 106.54205607476635,
      "grad_norm": 0.0003958574088755995,
      "learning_rate": 2.7155696202531643e-05,
      "loss": 0.0,
      "step": 11400
    },
    {
      "epoch": 106.77570093457943,
      "grad_norm": 0.0003987887466792017,
      "learning_rate": 2.7131962025316453e-05,
      "loss": 0.0,
      "step": 11425
    },
    {
      "epoch": 107.00934579439253,
      "grad_norm": 0.0003438335261307657,
      "learning_rate": 2.7108227848101262e-05,
      "loss": 0.0,
      "step": 11450
    },
    {
      "epoch": 107.24299065420561,
      "grad_norm": 0.00037780802813358605,
      "learning_rate": 2.708449367088607e-05,
      "loss": 0.0,
      "step": 11475
    },
    {
      "epoch": 107.4766355140187,
      "grad_norm": 0.00037574151065200567,
      "learning_rate": 2.706075949367088e-05,
      "loss": 0.0,
      "step": 11500
    },
    {
      "epoch": 107.71028037383178,
      "grad_norm": 0.000367281143553555,
      "learning_rate": 2.703702531645569e-05,
      "loss": 0.0,
      "step": 11525
    },
    {
      "epoch": 107.94392523364486,
      "grad_norm": 0.00040768570033833385,
      "learning_rate": 2.7013291139240507e-05,
      "loss": 0.0,
      "step": 11550
    },
    {
      "epoch": 108.17757009345794,
      "grad_norm": 0.00037230472662486136,
      "learning_rate": 2.6989556962025316e-05,
      "loss": 0.0,
      "step": 11575
    },
    {
      "epoch": 108.41121495327103,
      "grad_norm": 0.00037245237035676837,
      "learning_rate": 2.6965822784810126e-05,
      "loss": 0.0,
      "step": 11600
    },
    {
      "epoch": 108.64485981308411,
      "grad_norm": 0.00035675018443726003,
      "learning_rate": 2.6942088607594935e-05,
      "loss": 0.0,
      "step": 11625
    },
    {
      "epoch": 108.8785046728972,
      "grad_norm": 0.0003777843958232552,
      "learning_rate": 2.6918354430379745e-05,
      "loss": 0.0,
      "step": 11650
    },
    {
      "epoch": 109.11214953271028,
      "grad_norm": 0.0003273714683018625,
      "learning_rate": 2.6894620253164554e-05,
      "loss": 0.0,
      "step": 11675
    },
    {
      "epoch": 109.34579439252336,
      "grad_norm": 0.0003377443936187774,
      "learning_rate": 2.6870886075949367e-05,
      "loss": 0.0,
      "step": 11700
    },
    {
      "epoch": 109.57943925233644,
      "grad_norm": 0.0003397573309484869,
      "learning_rate": 2.6847151898734177e-05,
      "loss": 0.0,
      "step": 11725
    },
    {
      "epoch": 109.81308411214954,
      "grad_norm": 0.000353008828824386,
      "learning_rate": 2.6823417721518986e-05,
      "loss": 0.0,
      "step": 11750
    },
    {
      "epoch": 110.04672897196262,
      "grad_norm": 0.0003144113579764962,
      "learning_rate": 2.6799683544303796e-05,
      "loss": 0.0,
      "step": 11775
    },
    {
      "epoch": 110.2803738317757,
      "grad_norm": 0.00032846504473127425,
      "learning_rate": 2.6775949367088605e-05,
      "loss": 0.0,
      "step": 11800
    },
    {
      "epoch": 110.51401869158879,
      "grad_norm": 0.00031359802233055234,
      "learning_rate": 2.6752215189873415e-05,
      "loss": 0.0,
      "step": 11825
    },
    {
      "epoch": 110.74766355140187,
      "grad_norm": 0.0003331272746436298,
      "learning_rate": 2.6728481012658228e-05,
      "loss": 0.0,
      "step": 11850
    },
    {
      "epoch": 110.98130841121495,
      "grad_norm": 0.0003427889314480126,
      "learning_rate": 2.6704746835443037e-05,
      "loss": 0.0,
      "step": 11875
    },
    {
      "epoch": 111.21495327102804,
      "grad_norm": 0.0003111446858383715,
      "learning_rate": 2.6681012658227847e-05,
      "loss": 0.0,
      "step": 11900
    },
    {
      "epoch": 111.44859813084112,
      "grad_norm": 0.00029258738504722714,
      "learning_rate": 2.6657278481012656e-05,
      "loss": 0.0,
      "step": 11925
    },
    {
      "epoch": 111.6822429906542,
      "grad_norm": 0.000318909646011889,
      "learning_rate": 2.6633544303797466e-05,
      "loss": 0.0,
      "step": 11950
    },
    {
      "epoch": 111.91588785046729,
      "grad_norm": 0.00031279714312404394,
      "learning_rate": 2.6609810126582275e-05,
      "loss": 0.0,
      "step": 11975
    },
    {
      "epoch": 112.14953271028037,
      "grad_norm": 0.00027763089747168124,
      "learning_rate": 2.6586075949367085e-05,
      "loss": 0.0,
      "step": 12000
    },
    {
      "epoch": 112.14953271028037,
      "eval_loss": 0.5788441896438599,
      "eval_runtime": 63.4461,
      "eval_samples_per_second": 214.828,
      "eval_steps_per_second": 1.686,
      "eval_wer": 20.609952728205506,
      "step": 12000
    },
    {
      "epoch": 112.38317757009345,
      "grad_norm": 0.0003049731603823602,
      "learning_rate": 2.6562341772151898e-05,
      "loss": 0.0,
      "step": 12025
    },
    {
      "epoch": 112.61682242990655,
      "grad_norm": 0.0002978947595693171,
      "learning_rate": 2.6538607594936707e-05,
      "loss": 0.0,
      "step": 12050
    },
    {
      "epoch": 112.85046728971963,
      "grad_norm": 0.0003005851758643985,
      "learning_rate": 2.6514873417721517e-05,
      "loss": 0.0,
      "step": 12075
    },
    {
      "epoch": 113.08411214953271,
      "grad_norm": 0.0002606875204946846,
      "learning_rate": 2.6491139240506326e-05,
      "loss": 0.0,
      "step": 12100
    },
    {
      "epoch": 113.3177570093458,
      "grad_norm": 0.00028541364008560777,
      "learning_rate": 2.6467405063291136e-05,
      "loss": 0.0,
      "step": 12125
    },
    {
      "epoch": 113.55140186915888,
      "grad_norm": 0.0002848714357241988,
      "learning_rate": 2.6443670886075945e-05,
      "loss": 0.0,
      "step": 12150
    },
    {
      "epoch": 113.78504672897196,
      "grad_norm": 0.00026482666726224124,
      "learning_rate": 2.6419936708860755e-05,
      "loss": 0.0,
      "step": 12175
    },
    {
      "epoch": 114.01869158878505,
      "grad_norm": 0.0002632560790516436,
      "learning_rate": 2.6396202531645568e-05,
      "loss": 0.0,
      "step": 12200
    },
    {
      "epoch": 114.25233644859813,
      "grad_norm": 0.0002834403421729803,
      "learning_rate": 2.6372468354430377e-05,
      "loss": 0.0,
      "step": 12225
    },
    {
      "epoch": 114.48598130841121,
      "grad_norm": 0.0002704873331822455,
      "learning_rate": 2.6348734177215187e-05,
      "loss": 0.0,
      "step": 12250
    },
    {
      "epoch": 114.7196261682243,
      "grad_norm": 0.00027725083054974675,
      "learning_rate": 2.6324999999999996e-05,
      "loss": 0.0,
      "step": 12275
    },
    {
      "epoch": 114.95327102803738,
      "grad_norm": 0.00026416871696710587,
      "learning_rate": 2.6301265822784806e-05,
      "loss": 0.0,
      "step": 12300
    },
    {
      "epoch": 115.18691588785046,
      "grad_norm": 0.0002661787730176002,
      "learning_rate": 2.6277531645569615e-05,
      "loss": 0.0,
      "step": 12325
    },
    {
      "epoch": 115.42056074766356,
      "grad_norm": 0.00026222606538794935,
      "learning_rate": 2.6253797468354425e-05,
      "loss": 0.0,
      "step": 12350
    },
    {
      "epoch": 115.65420560747664,
      "grad_norm": 0.00025618536164984107,
      "learning_rate": 2.623006329113924e-05,
      "loss": 0.0,
      "step": 12375
    },
    {
      "epoch": 115.88785046728972,
      "grad_norm": 0.000255120248766616,
      "learning_rate": 2.620632911392405e-05,
      "loss": 0.0,
      "step": 12400
    },
    {
      "epoch": 116.1214953271028,
      "grad_norm": 0.00024208318791352212,
      "learning_rate": 2.618259493670886e-05,
      "loss": 0.0,
      "step": 12425
    },
    {
      "epoch": 116.35514018691589,
      "grad_norm": 0.000250356417382136,
      "learning_rate": 2.615886075949367e-05,
      "loss": 0.0,
      "step": 12450
    },
    {
      "epoch": 116.58878504672897,
      "grad_norm": 0.00023948575835675,
      "learning_rate": 2.613512658227848e-05,
      "loss": 0.0,
      "step": 12475
    },
    {
      "epoch": 116.82242990654206,
      "grad_norm": 0.00023785764642525464,
      "learning_rate": 2.6111392405063292e-05,
      "loss": 0.0,
      "step": 12500
    },
    {
      "epoch": 117.05607476635514,
      "grad_norm": 0.00023564029834233224,
      "learning_rate": 2.60876582278481e-05,
      "loss": 0.0,
      "step": 12525
    },
    {
      "epoch": 117.28971962616822,
      "grad_norm": 0.00022972981969360262,
      "learning_rate": 2.606392405063291e-05,
      "loss": 0.0,
      "step": 12550
    },
    {
      "epoch": 117.5233644859813,
      "grad_norm": 0.0002341399813303724,
      "learning_rate": 2.604018987341772e-05,
      "loss": 0.0,
      "step": 12575
    },
    {
      "epoch": 117.75700934579439,
      "grad_norm": 0.00022405364143196493,
      "learning_rate": 2.601645569620253e-05,
      "loss": 0.0,
      "step": 12600
    },
    {
      "epoch": 117.99065420560747,
      "grad_norm": 0.0002332032599952072,
      "learning_rate": 2.599272151898734e-05,
      "loss": 0.0,
      "step": 12625
    },
    {
      "epoch": 118.22429906542057,
      "grad_norm": 0.00021942731109447777,
      "learning_rate": 2.596898734177215e-05,
      "loss": 0.0,
      "step": 12650
    },
    {
      "epoch": 118.45794392523365,
      "grad_norm": 0.00022152048768475652,
      "learning_rate": 2.5945253164556962e-05,
      "loss": 0.0,
      "step": 12675
    },
    {
      "epoch": 118.69158878504673,
      "grad_norm": 0.00022273638751357794,
      "learning_rate": 2.592151898734177e-05,
      "loss": 0.0,
      "step": 12700
    },
    {
      "epoch": 118.92523364485982,
      "grad_norm": 0.00021274616301525384,
      "learning_rate": 2.589778481012658e-05,
      "loss": 0.0,
      "step": 12725
    },
    {
      "epoch": 119.1588785046729,
      "grad_norm": 0.000209805992199108,
      "learning_rate": 2.587405063291139e-05,
      "loss": 0.0,
      "step": 12750
    },
    {
      "epoch": 119.39252336448598,
      "grad_norm": 0.00020782243518624455,
      "learning_rate": 2.58503164556962e-05,
      "loss": 0.0,
      "step": 12775
    },
    {
      "epoch": 119.62616822429906,
      "grad_norm": 0.00020385615061968565,
      "learning_rate": 2.582658227848101e-05,
      "loss": 0.0,
      "step": 12800
    },
    {
      "epoch": 119.85981308411215,
      "grad_norm": 0.000216295724385418,
      "learning_rate": 2.580284810126582e-05,
      "loss": 0.0,
      "step": 12825
    },
    {
      "epoch": 120.09345794392523,
      "grad_norm": 0.00018660743080545217,
      "learning_rate": 2.5779113924050632e-05,
      "loss": 0.0,
      "step": 12850
    },
    {
      "epoch": 120.32710280373831,
      "grad_norm": 0.00018773713964037597,
      "learning_rate": 2.575537974683544e-05,
      "loss": 0.0,
      "step": 12875
    },
    {
      "epoch": 120.5607476635514,
      "grad_norm": 0.00018672359874472022,
      "learning_rate": 2.573164556962025e-05,
      "loss": 0.0,
      "step": 12900
    },
    {
      "epoch": 120.79439252336448,
      "grad_norm": 0.00020279856107663363,
      "learning_rate": 2.570791139240506e-05,
      "loss": 0.0,
      "step": 12925
    },
    {
      "epoch": 121.02803738317758,
      "grad_norm": 0.00018718869250733405,
      "learning_rate": 2.568417721518987e-05,
      "loss": 0.0,
      "step": 12950
    },
    {
      "epoch": 121.26168224299066,
      "grad_norm": 0.00018918044224847108,
      "learning_rate": 2.566044303797468e-05,
      "loss": 0.0,
      "step": 12975
    },
    {
      "epoch": 121.49532710280374,
      "grad_norm": 0.00019031373085454106,
      "learning_rate": 2.563670886075949e-05,
      "loss": 0.0,
      "step": 13000
    },
    {
      "epoch": 121.49532710280374,
      "eval_loss": 0.6022807955741882,
      "eval_runtime": 60.8512,
      "eval_samples_per_second": 223.989,
      "eval_steps_per_second": 1.758,
      "eval_wer": 20.698816372897504,
      "step": 13000
    },
    {
      "epoch": 121.72897196261682,
      "grad_norm": 0.000184610893484205,
      "learning_rate": 2.56129746835443e-05,
      "loss": 0.0,
      "step": 13025
    },
    {
      "epoch": 121.96261682242991,
      "grad_norm": 0.00018427333270665258,
      "learning_rate": 2.558924050632911e-05,
      "loss": 0.0,
      "step": 13050
    },
    {
      "epoch": 122.19626168224299,
      "grad_norm": 0.00018340496171731502,
      "learning_rate": 2.556550632911392e-05,
      "loss": 0.0,
      "step": 13075
    },
    {
      "epoch": 122.42990654205607,
      "grad_norm": 0.00017801990907173604,
      "learning_rate": 2.554177215189873e-05,
      "loss": 0.0,
      "step": 13100
    },
    {
      "epoch": 122.66355140186916,
      "grad_norm": 0.0001757302088662982,
      "learning_rate": 2.551803797468354e-05,
      "loss": 0.0,
      "step": 13125
    },
    {
      "epoch": 122.89719626168224,
      "grad_norm": 0.0001856745220720768,
      "learning_rate": 2.549430379746835e-05,
      "loss": 0.0,
      "step": 13150
    },
    {
      "epoch": 123.13084112149532,
      "grad_norm": 0.00016652005433570594,
      "learning_rate": 2.547056962025316e-05,
      "loss": 0.0,
      "step": 13175
    },
    {
      "epoch": 123.3644859813084,
      "grad_norm": 0.0001729445648379624,
      "learning_rate": 2.5446835443037975e-05,
      "loss": 0.0,
      "step": 13200
    },
    {
      "epoch": 123.59813084112149,
      "grad_norm": 0.00017554008809383959,
      "learning_rate": 2.5423101265822784e-05,
      "loss": 0.0,
      "step": 13225
    },
    {
      "epoch": 123.83177570093459,
      "grad_norm": 0.0001670542114879936,
      "learning_rate": 2.5399367088607594e-05,
      "loss": 0.0,
      "step": 13250
    },
    {
      "epoch": 124.06542056074767,
      "grad_norm": 0.00016005740326363593,
      "learning_rate": 2.5375632911392404e-05,
      "loss": 0.0,
      "step": 13275
    },
    {
      "epoch": 124.29906542056075,
      "grad_norm": 0.00016280432464554906,
      "learning_rate": 2.5351898734177213e-05,
      "loss": 0.0,
      "step": 13300
    },
    {
      "epoch": 124.53271028037383,
      "grad_norm": 0.00015820316912140697,
      "learning_rate": 2.5328164556962026e-05,
      "loss": 0.0,
      "step": 13325
    },
    {
      "epoch": 124.76635514018692,
      "grad_norm": 0.00015971559332683682,
      "learning_rate": 2.5304430379746835e-05,
      "loss": 0.0,
      "step": 13350
    },
    {
      "epoch": 125.0,
      "grad_norm": 0.00023178270203061402,
      "learning_rate": 2.5280696202531645e-05,
      "loss": 0.0,
      "step": 13375
    },
    {
      "epoch": 125.23364485981308,
      "grad_norm": 0.00015152622654568404,
      "learning_rate": 2.5256962025316454e-05,
      "loss": 0.0,
      "step": 13400
    },
    {
      "epoch": 125.46728971962617,
      "grad_norm": 0.00015582592459395528,
      "learning_rate": 2.5233227848101264e-05,
      "loss": 0.0,
      "step": 13425
    },
    {
      "epoch": 125.70093457943925,
      "grad_norm": 0.0001534547918708995,
      "learning_rate": 2.5209493670886073e-05,
      "loss": 0.0,
      "step": 13450
    },
    {
      "epoch": 125.93457943925233,
      "grad_norm": 0.0001554342161398381,
      "learning_rate": 2.5185759493670883e-05,
      "loss": 0.0,
      "step": 13475
    },
    {
      "epoch": 126.16822429906541,
      "grad_norm": 0.00013976512127555907,
      "learning_rate": 2.5162025316455696e-05,
      "loss": 0.0,
      "step": 13500
    },
    {
      "epoch": 126.40186915887851,
      "grad_norm": 0.00015099224401637912,
      "learning_rate": 2.5138291139240505e-05,
      "loss": 0.0,
      "step": 13525
    },
    {
      "epoch": 126.6355140186916,
      "grad_norm": 0.00014277642185334116,
      "learning_rate": 2.5114556962025315e-05,
      "loss": 0.0,
      "step": 13550
    },
    {
      "epoch": 126.86915887850468,
      "grad_norm": 0.00015035129035823047,
      "learning_rate": 2.5090822784810124e-05,
      "loss": 0.0,
      "step": 13575
    },
    {
      "epoch": 127.10280373831776,
      "grad_norm": 0.00014256956637836993,
      "learning_rate": 2.5067088607594934e-05,
      "loss": 0.0,
      "step": 13600
    },
    {
      "epoch": 127.33644859813084,
      "grad_norm": 0.00014243739133235067,
      "learning_rate": 2.5043354430379743e-05,
      "loss": 0.0,
      "step": 13625
    },
    {
      "epoch": 127.57009345794393,
      "grad_norm": 0.00013728807971347123,
      "learning_rate": 2.5019620253164553e-05,
      "loss": 0.0,
      "step": 13650
    },
    {
      "epoch": 127.80373831775701,
      "grad_norm": 0.00013955221220385283,
      "learning_rate": 2.4995886075949366e-05,
      "loss": 0.0,
      "step": 13675
    },
    {
      "epoch": 128.0373831775701,
      "grad_norm": 0.00012837977556046098,
      "learning_rate": 2.4972151898734175e-05,
      "loss": 0.0,
      "step": 13700
    },
    {
      "epoch": 128.2710280373832,
      "grad_norm": 0.00012790935579687357,
      "learning_rate": 2.4948417721518985e-05,
      "loss": 0.0,
      "step": 13725
    },
    {
      "epoch": 128.50467289719626,
      "grad_norm": 0.00013070220302324742,
      "learning_rate": 2.4924683544303794e-05,
      "loss": 0.0,
      "step": 13750
    },
    {
      "epoch": 128.73831775700936,
      "grad_norm": 0.00013144199328962713,
      "learning_rate": 2.4900949367088604e-05,
      "loss": 0.0,
      "step": 13775
    },
    {
      "epoch": 128.97196261682242,
      "grad_norm": 0.00013546706759370863,
      "learning_rate": 2.4877215189873413e-05,
      "loss": 0.0,
      "step": 13800
    },
    {
      "epoch": 129.20560747663552,
      "grad_norm": 0.00012009818601654842,
      "learning_rate": 2.4853481012658226e-05,
      "loss": 0.0,
      "step": 13825
    },
    {
      "epoch": 129.4392523364486,
      "grad_norm": 0.00012730721209663898,
      "learning_rate": 2.4829746835443036e-05,
      "loss": 0.0,
      "step": 13850
    },
    {
      "epoch": 129.6728971962617,
      "grad_norm": 0.00012798367242794484,
      "learning_rate": 2.4806012658227845e-05,
      "loss": 0.0,
      "step": 13875
    },
    {
      "epoch": 129.90654205607476,
      "grad_norm": 0.00012850937491748482,
      "learning_rate": 2.4782278481012655e-05,
      "loss": 0.0,
      "step": 13900
    },
    {
      "epoch": 130.14018691588785,
      "grad_norm": 0.0001231772912433371,
      "learning_rate": 2.4758544303797464e-05,
      "loss": 0.0,
      "step": 13925
    },
    {
      "epoch": 130.37383177570092,
      "grad_norm": 0.00011643731704680249,
      "learning_rate": 2.4734810126582274e-05,
      "loss": 0.0,
      "step": 13950
    },
    {
      "epoch": 130.60747663551402,
      "grad_norm": 0.00012311043974477798,
      "learning_rate": 2.4711075949367083e-05,
      "loss": 0.0,
      "step": 13975
    },
    {
      "epoch": 130.84112149532712,
      "grad_norm": 0.00012078093277523294,
      "learning_rate": 2.4687341772151896e-05,
      "loss": 0.0,
      "step": 14000
    },
    {
      "epoch": 130.84112149532712,
      "eval_loss": 0.625320553779602,
      "eval_runtime": 62.4102,
      "eval_samples_per_second": 218.394,
      "eval_steps_per_second": 1.714,
      "eval_wer": 20.776686577009052,
      "step": 14000
    },
    {
      "epoch": 131.07476635514018,
      "grad_norm": 0.00011285846994724125,
      "learning_rate": 2.466360759493671e-05,
      "loss": 0.0,
      "step": 14025
    },
    {
      "epoch": 131.30841121495328,
      "grad_norm": 0.00010817299335030839,
      "learning_rate": 2.463987341772152e-05,
      "loss": 0.0,
      "step": 14050
    },
    {
      "epoch": 131.54205607476635,
      "grad_norm": 0.00011426104174461216,
      "learning_rate": 2.4616139240506328e-05,
      "loss": 0.0,
      "step": 14075
    },
    {
      "epoch": 131.77570093457945,
      "grad_norm": 0.00011646000348264351,
      "learning_rate": 2.4592405063291138e-05,
      "loss": 0.0,
      "step": 14100
    },
    {
      "epoch": 132.00934579439252,
      "grad_norm": 0.00010999570076819509,
      "learning_rate": 2.4568670886075947e-05,
      "loss": 0.0,
      "step": 14125
    },
    {
      "epoch": 132.2429906542056,
      "grad_norm": 0.00010461529745953158,
      "learning_rate": 2.454493670886076e-05,
      "loss": 0.0,
      "step": 14150
    },
    {
      "epoch": 132.47663551401868,
      "grad_norm": 0.00010791856038849801,
      "learning_rate": 2.452120253164557e-05,
      "loss": 0.0,
      "step": 14175
    },
    {
      "epoch": 132.71028037383178,
      "grad_norm": 0.00010874889994738623,
      "learning_rate": 2.449746835443038e-05,
      "loss": 0.0,
      "step": 14200
    },
    {
      "epoch": 132.94392523364485,
      "grad_norm": 0.00011335639283061028,
      "learning_rate": 2.447373417721519e-05,
      "loss": 0.0,
      "step": 14225
    },
    {
      "epoch": 133.17757009345794,
      "grad_norm": 9.745185525389388e-05,
      "learning_rate": 2.4449999999999998e-05,
      "loss": 0.0,
      "step": 14250
    },
    {
      "epoch": 133.41121495327104,
      "grad_norm": 0.00010352137906011194,
      "learning_rate": 2.4426265822784808e-05,
      "loss": 0.0,
      "step": 14275
    },
    {
      "epoch": 133.6448598130841,
      "grad_norm": 0.00010501281940378249,
      "learning_rate": 2.440253164556962e-05,
      "loss": 0.0,
      "step": 14300
    },
    {
      "epoch": 133.8785046728972,
      "grad_norm": 0.00010406147339381278,
      "learning_rate": 2.437879746835443e-05,
      "loss": 0.0,
      "step": 14325
    },
    {
      "epoch": 134.11214953271028,
      "grad_norm": 9.530872921459377e-05,
      "learning_rate": 2.435506329113924e-05,
      "loss": 0.0,
      "step": 14350
    },
    {
      "epoch": 134.34579439252337,
      "grad_norm": 9.523833432467654e-05,
      "learning_rate": 2.433132911392405e-05,
      "loss": 0.0,
      "step": 14375
    },
    {
      "epoch": 134.57943925233644,
      "grad_norm": 9.822582069318742e-05,
      "learning_rate": 2.430759493670886e-05,
      "loss": 0.0,
      "step": 14400
    },
    {
      "epoch": 134.81308411214954,
      "grad_norm": 9.944786143023521e-05,
      "learning_rate": 2.4283860759493668e-05,
      "loss": 0.0,
      "step": 14425
    },
    {
      "epoch": 135.0467289719626,
      "grad_norm": 9.433593368157744e-05,
      "learning_rate": 2.4260126582278477e-05,
      "loss": 0.0,
      "step": 14450
    },
    {
      "epoch": 135.2803738317757,
      "grad_norm": 9.816836245590821e-05,
      "learning_rate": 2.423639240506329e-05,
      "loss": 0.0,
      "step": 14475
    },
    {
      "epoch": 135.51401869158877,
      "grad_norm": 9.458445128984749e-05,
      "learning_rate": 2.42126582278481e-05,
      "loss": 0.0,
      "step": 14500
    },
    {
      "epoch": 135.74766355140187,
      "grad_norm": 9.752513142302632e-05,
      "learning_rate": 2.418892405063291e-05,
      "loss": 0.0,
      "step": 14525
    },
    {
      "epoch": 135.98130841121494,
      "grad_norm": 9.620709897717461e-05,
      "learning_rate": 2.416518987341772e-05,
      "loss": 0.0,
      "step": 14550
    },
    {
      "epoch": 136.21495327102804,
      "grad_norm": 8.870051533449441e-05,
      "learning_rate": 2.414145569620253e-05,
      "loss": 0.0,
      "step": 14575
    },
    {
      "epoch": 136.44859813084113,
      "grad_norm": 8.476567745674402e-05,
      "learning_rate": 2.4117721518987338e-05,
      "loss": 0.0,
      "step": 14600
    },
    {
      "epoch": 136.6822429906542,
      "grad_norm": 9.104797209147364e-05,
      "learning_rate": 2.4093987341772147e-05,
      "loss": 0.0,
      "step": 14625
    },
    {
      "epoch": 136.9158878504673,
      "grad_norm": 8.767654799157754e-05,
      "learning_rate": 2.407025316455696e-05,
      "loss": 0.0,
      "step": 14650
    },
    {
      "epoch": 137.14953271028037,
      "grad_norm": 8.455555507680401e-05,
      "learning_rate": 2.404651898734177e-05,
      "loss": 0.0,
      "step": 14675
    },
    {
      "epoch": 137.38317757009347,
      "grad_norm": 8.793352026259527e-05,
      "learning_rate": 2.402278481012658e-05,
      "loss": 0.0,
      "step": 14700
    },
    {
      "epoch": 137.61682242990653,
      "grad_norm": 8.613619866082445e-05,
      "learning_rate": 2.399905063291139e-05,
      "loss": 0.0,
      "step": 14725
    },
    {
      "epoch": 137.85046728971963,
      "grad_norm": 8.538133261026815e-05,
      "learning_rate": 2.3975316455696198e-05,
      "loss": 0.0,
      "step": 14750
    },
    {
      "epoch": 138.0841121495327,
      "grad_norm": 8.429722947767004e-05,
      "learning_rate": 2.3951582278481008e-05,
      "loss": 0.0,
      "step": 14775
    },
    {
      "epoch": 138.3177570093458,
      "grad_norm": 7.928127160994336e-05,
      "learning_rate": 2.3927848101265817e-05,
      "loss": 0.0,
      "step": 14800
    },
    {
      "epoch": 138.55140186915887,
      "grad_norm": 8.374597382498905e-05,
      "learning_rate": 2.390411392405063e-05,
      "loss": 0.0,
      "step": 14825
    },
    {
      "epoch": 138.78504672897196,
      "grad_norm": 7.885757804615423e-05,
      "learning_rate": 2.3880379746835443e-05,
      "loss": 0.0,
      "step": 14850
    },
    {
      "epoch": 139.01869158878506,
      "grad_norm": 8.018634980544448e-05,
      "learning_rate": 2.3856645569620253e-05,
      "loss": 0.0,
      "step": 14875
    },
    {
      "epoch": 139.25233644859813,
      "grad_norm": 8.028265438042581e-05,
      "learning_rate": 2.3832911392405062e-05,
      "loss": 0.0,
      "step": 14900
    },
    {
      "epoch": 139.48598130841123,
      "grad_norm": 7.670955528737977e-05,
      "learning_rate": 2.380917721518987e-05,
      "loss": 0.0,
      "step": 14925
    },
    {
      "epoch": 139.7196261682243,
      "grad_norm": 8.080930274445564e-05,
      "learning_rate": 2.3785443037974685e-05,
      "loss": 0.0,
      "step": 14950
    },
    {
      "epoch": 139.9532710280374,
      "grad_norm": 7.850069960113615e-05,
      "learning_rate": 2.3761708860759494e-05,
      "loss": 0.0,
      "step": 14975
    },
    {
      "epoch": 140.18691588785046,
      "grad_norm": 7.2464783443138e-05,
      "learning_rate": 2.3737974683544304e-05,
      "loss": 0.0,
      "step": 15000
    },
    {
      "epoch": 140.18691588785046,
      "eval_loss": 0.6485691070556641,
      "eval_runtime": 62.8697,
      "eval_samples_per_second": 216.798,
      "eval_steps_per_second": 1.702,
      "eval_wer": 20.868298581846165,
      "step": 15000
    },
    {
      "epoch": 140.18691588785046,
      "step": 15000,
      "total_flos": 9.40815618637824e+19,
      "train_loss": 0.028148239994506972,
      "train_runtime": 11548.0276,
      "train_samples_per_second": 886.732,
      "train_steps_per_second": 3.464
    }
  ],
  "logging_steps": 25,
  "max_steps": 40000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 374,
  "save_steps": 1000,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 10,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 10
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 9.40815618637824e+19,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}