{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 15.028415070511471,
  "eval_steps": 100,
  "global_step": 35700,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.04,
      "grad_norm": 2.6430506706237793,
      "learning_rate": 9.5e-06,
      "loss": 5.6141,
      "step": 100
    },
    {
      "epoch": 0.04,
      "eval_cer": 0.9770320871299519,
      "eval_loss": 2.355087995529175,
      "eval_runtime": 329.8418,
      "eval_samples_per_second": 28.735,
      "eval_steps_per_second": 3.593,
      "step": 100
    },
    {
      "epoch": 0.08,
      "grad_norm": 2.838773012161255,
      "learning_rate": 1.9500000000000003e-05,
      "loss": 2.1498,
      "step": 200
    },
    {
      "epoch": 0.08,
      "eval_cer": 0.6385162879824802,
      "eval_loss": 1.961960792541504,
      "eval_runtime": 334.0577,
      "eval_samples_per_second": 28.372,
      "eval_steps_per_second": 3.547,
      "step": 200
    },
    {
      "epoch": 0.13,
      "grad_norm": 3.848980665206909,
      "learning_rate": 2.95e-05,
      "loss": 2.6461,
      "step": 300
    },
    {
      "epoch": 0.13,
      "eval_cer": 0.5852479371162645,
      "eval_loss": 2.6564223766326904,
      "eval_runtime": 350.7705,
      "eval_samples_per_second": 27.021,
      "eval_steps_per_second": 3.378,
      "step": 300
    },
    {
      "epoch": 0.17,
      "grad_norm": 3.9299519062042236,
      "learning_rate": 3.9500000000000005e-05,
      "loss": 2.359,
      "step": 400
    },
    {
      "epoch": 0.17,
      "eval_cer": 0.5907913261116108,
      "eval_loss": 2.6195201873779297,
      "eval_runtime": 341.3159,
      "eval_samples_per_second": 27.769,
      "eval_steps_per_second": 3.472,
      "step": 400
    },
    {
      "epoch": 0.21,
      "grad_norm": 4.236645698547363,
      "learning_rate": 4.9500000000000004e-05,
      "loss": 4.2363,
      "step": 500
    },
    {
      "epoch": 0.21,
      "eval_cer": 0.624139650385202,
      "eval_loss": 2.457566261291504,
      "eval_runtime": 356.3471,
      "eval_samples_per_second": 26.598,
      "eval_steps_per_second": 3.325,
      "step": 500
    },
    {
      "epoch": 0.25,
      "grad_norm": 13.769584655761719,
      "learning_rate": 4.9932862190812725e-05,
      "loss": 2.6933,
      "step": 600
    },
    {
      "epoch": 0.25,
      "eval_cer": 0.6483369833013961,
      "eval_loss": 3.4720070362091064,
      "eval_runtime": 345.5879,
      "eval_samples_per_second": 27.426,
      "eval_steps_per_second": 3.429,
      "step": 600
    },
    {
      "epoch": 0.29,
      "grad_norm": 4.292990207672119,
      "learning_rate": 4.986289752650177e-05,
      "loss": 4.1075,
      "step": 700
    },
    {
      "epoch": 0.29,
      "eval_cer": 0.6181098705564898,
      "eval_loss": 3.3849761486053467,
      "eval_runtime": 353.3198,
      "eval_samples_per_second": 26.826,
      "eval_steps_per_second": 3.354,
      "step": 700
    },
    {
      "epoch": 0.34,
      "grad_norm": 4.32611083984375,
      "learning_rate": 4.97922261484099e-05,
      "loss": 2.6907,
      "step": 800
    },
    {
      "epoch": 0.34,
      "eval_cer": 0.6191510891243988,
      "eval_loss": 2.5412750244140625,
      "eval_runtime": 343.4967,
      "eval_samples_per_second": 27.593,
      "eval_steps_per_second": 3.45,
      "step": 800
    },
    {
      "epoch": 0.38,
      "grad_norm": 3.173403263092041,
      "learning_rate": 4.972155477031802e-05,
      "loss": 4.2294,
      "step": 900
    },
    {
      "epoch": 0.38,
      "eval_cer": 0.5840918423213797,
      "eval_loss": 2.6817102432250977,
      "eval_runtime": 354.5341,
      "eval_samples_per_second": 26.734,
      "eval_steps_per_second": 3.342,
      "step": 900
    },
    {
      "epoch": 0.42,
      "grad_norm": 4.8158440589904785,
      "learning_rate": 4.965088339222615e-05,
      "loss": 2.2949,
      "step": 1000
    },
    {
      "epoch": 0.42,
      "eval_cer": 0.5736649915920379,
      "eval_loss": 1.9513429403305054,
      "eval_runtime": 345.1224,
      "eval_samples_per_second": 27.463,
      "eval_steps_per_second": 3.434,
      "step": 1000
    },
    {
      "epoch": 0.46,
      "grad_norm": 3.0650887489318848,
      "learning_rate": 4.958021201413428e-05,
      "loss": 2.6115,
      "step": 1100
    },
    {
      "epoch": 0.46,
      "eval_cer": 0.5645799929607759,
      "eval_loss": 2.8409903049468994,
      "eval_runtime": 352.0824,
      "eval_samples_per_second": 26.92,
      "eval_steps_per_second": 3.366,
      "step": 1100
    },
    {
      "epoch": 0.51,
      "grad_norm": 3.0540664196014404,
      "learning_rate": 4.950954063604241e-05,
      "loss": 2.3453,
      "step": 1200
    },
    {
      "epoch": 0.51,
      "eval_cer": 0.598292499315631,
      "eval_loss": 1.8713030815124512,
      "eval_runtime": 344.1368,
      "eval_samples_per_second": 27.541,
      "eval_steps_per_second": 3.443,
      "step": 1200
    },
    {
      "epoch": 0.55,
      "grad_norm": 4.366485118865967,
      "learning_rate": 4.9438869257950535e-05,
      "loss": 3.2578,
      "step": 1300
    },
    {
      "epoch": 0.55,
      "eval_cer": 0.5734816784638849,
      "eval_loss": 2.000699758529663,
      "eval_runtime": 350.6902,
      "eval_samples_per_second": 27.027,
      "eval_steps_per_second": 3.379,
      "step": 1300
    },
    {
      "epoch": 0.59,
      "grad_norm": 81.67564392089844,
      "learning_rate": 4.936819787985866e-05,
      "loss": 1.9657,
      "step": 1400
    },
    {
      "epoch": 0.59,
      "eval_cer": 0.6325427241797349,
      "eval_loss": 3.403754472732544,
      "eval_runtime": 340.0489,
      "eval_samples_per_second": 27.872,
      "eval_steps_per_second": 3.485,
      "step": 1400
    },
    {
      "epoch": 0.63,
      "grad_norm": 3.837653875350952,
      "learning_rate": 4.929752650176679e-05,
      "loss": 2.6865,
      "step": 1500
    },
    {
      "epoch": 0.63,
      "eval_cer": 0.5574918853388604,
      "eval_loss": 3.5201330184936523,
      "eval_runtime": 351.432,
      "eval_samples_per_second": 26.97,
      "eval_steps_per_second": 3.372,
      "step": 1500
    },
    {
      "epoch": 0.67,
      "grad_norm": 2.9047458171844482,
      "learning_rate": 4.922685512367491e-05,
      "loss": 2.4105,
      "step": 1600
    },
    {
      "epoch": 0.67,
      "eval_cer": 0.5886624496499941,
      "eval_loss": 3.5269789695739746,
      "eval_runtime": 347.023,
      "eval_samples_per_second": 27.312,
      "eval_steps_per_second": 3.415,
      "step": 1600
    },
    {
      "epoch": 0.72,
      "grad_norm": 2.9270007610321045,
      "learning_rate": 4.915618374558304e-05,
      "loss": 2.418,
      "step": 1700
    },
    {
      "epoch": 0.72,
      "eval_cer": 0.624877791247898,
      "eval_loss": 2.498126983642578,
      "eval_runtime": 355.9419,
      "eval_samples_per_second": 26.628,
      "eval_steps_per_second": 3.329,
      "step": 1700
    },
    {
      "epoch": 0.76,
      "grad_norm": 3.458078622817993,
      "learning_rate": 4.908551236749117e-05,
      "loss": 2.6118,
      "step": 1800
    },
    {
      "epoch": 0.76,
      "eval_cer": 0.5718074185600877,
      "eval_loss": 3.407224655151367,
      "eval_runtime": 343.8365,
      "eval_samples_per_second": 27.565,
      "eval_steps_per_second": 3.446,
      "step": 1800
    },
    {
      "epoch": 0.8,
      "grad_norm": 1.9933278560638428,
      "learning_rate": 4.90148409893993e-05,
      "loss": 3.2046,
      "step": 1900
    },
    {
      "epoch": 0.8,
      "eval_cer": 0.5539844941535333,
      "eval_loss": 1.928993821144104,
      "eval_runtime": 357.2509,
      "eval_samples_per_second": 26.53,
      "eval_steps_per_second": 3.317,
      "step": 1900
    },
    {
      "epoch": 0.84,
      "grad_norm": 4.917328357696533,
      "learning_rate": 4.8944169611307425e-05,
      "loss": 3.1904,
      "step": 2000
    },
    {
      "epoch": 0.84,
      "eval_cer": 0.5568001838019632,
      "eval_loss": 2.729193925857544,
      "eval_runtime": 341.3463,
      "eval_samples_per_second": 27.767,
      "eval_steps_per_second": 3.472,
      "step": 2000
    },
    {
      "epoch": 0.88,
      "grad_norm": 2.842353343963623,
      "learning_rate": 4.8873498233215547e-05,
      "loss": 2.1649,
      "step": 2100
    },
    {
      "epoch": 0.88,
      "eval_cer": 0.544757733369833,
      "eval_loss": 2.421653985977173,
      "eval_runtime": 358.7261,
      "eval_samples_per_second": 26.421,
      "eval_steps_per_second": 3.303,
      "step": 2100
    },
    {
      "epoch": 0.93,
      "grad_norm": 4.818225383758545,
      "learning_rate": 4.880282685512368e-05,
      "loss": 2.2775,
      "step": 2200
    },
    {
      "epoch": 0.93,
      "eval_cer": 0.5533832270931915,
      "eval_loss": 2.188101291656494,
      "eval_runtime": 354.4215,
      "eval_samples_per_second": 26.742,
      "eval_steps_per_second": 3.343,
      "step": 2200
    },
    {
      "epoch": 0.97,
      "grad_norm": 2.9436004161834717,
      "learning_rate": 4.87321554770318e-05,
      "loss": 1.6308,
      "step": 2300
    },
    {
      "epoch": 0.97,
      "eval_cer": 0.5162586523796489,
      "eval_loss": 3.3528623580932617,
      "eval_runtime": 360.9925,
      "eval_samples_per_second": 26.255,
      "eval_steps_per_second": 3.283,
      "step": 2300
    },
    {
      "epoch": 1.01,
      "grad_norm": 7.935622692108154,
      "learning_rate": 4.866148409893993e-05,
      "loss": 2.4462,
      "step": 2400
    },
    {
      "epoch": 1.01,
      "eval_cer": 0.575390579171718,
      "eval_loss": 2.0690765380859375,
      "eval_runtime": 343.7001,
      "eval_samples_per_second": 27.576,
      "eval_steps_per_second": 3.448,
      "step": 2400
    },
    {
      "epoch": 1.05,
      "grad_norm": 2.40964937210083,
      "learning_rate": 4.859081272084806e-05,
      "loss": 2.3368,
      "step": 2500
    },
    {
      "epoch": 1.05,
      "eval_cer": 0.5851746118650033,
      "eval_loss": 1.9030027389526367,
      "eval_runtime": 352.7743,
      "eval_samples_per_second": 26.867,
      "eval_steps_per_second": 3.359,
      "step": 2500
    },
    {
      "epoch": 1.09,
      "grad_norm": 3.7618064880371094,
      "learning_rate": 4.852084805653711e-05,
      "loss": 2.4157,
      "step": 2600
    },
    {
      "epoch": 1.09,
      "eval_cer": 0.5566339798991045,
      "eval_loss": 3.213930368423462,
      "eval_runtime": 341.3273,
      "eval_samples_per_second": 27.768,
      "eval_steps_per_second": 3.472,
      "step": 2600
    },
    {
      "epoch": 1.14,
      "grad_norm": 9.61824893951416,
      "learning_rate": 4.845017667844523e-05,
      "loss": 2.7299,
      "step": 2700
    },
    {
      "epoch": 1.14,
      "eval_cer": 0.5819702983848891,
      "eval_loss": 1.7893775701522827,
      "eval_runtime": 353.5497,
      "eval_samples_per_second": 26.808,
      "eval_steps_per_second": 3.352,
      "step": 2700
    },
    {
      "epoch": 1.18,
      "grad_norm": 2.6824772357940674,
      "learning_rate": 4.8379505300353364e-05,
      "loss": 2.934,
      "step": 2800
    },
    {
      "epoch": 1.18,
      "eval_cer": 0.5119813460560791,
      "eval_loss": 2.804206609725952,
      "eval_runtime": 340.4666,
      "eval_samples_per_second": 27.838,
      "eval_steps_per_second": 3.481,
      "step": 2800
    },
    {
      "epoch": 1.22,
      "grad_norm": 4.691165924072266,
      "learning_rate": 4.8308833922261485e-05,
      "loss": 2.0012,
      "step": 2900
    },
    {
      "epoch": 1.22,
      "eval_cer": 0.5138047006374409,
      "eval_loss": 3.412604331970215,
      "eval_runtime": 349.7896,
      "eval_samples_per_second": 27.096,
      "eval_steps_per_second": 3.388,
      "step": 2900
    },
    {
      "epoch": 1.26,
      "grad_norm": 2.8883514404296875,
      "learning_rate": 4.823816254416961e-05,
      "loss": 3.5496,
      "step": 3000
    },
    {
      "epoch": 1.26,
      "eval_cer": 0.5417074029173673,
      "eval_loss": 1.7423540353775024,
      "eval_runtime": 339.3199,
      "eval_samples_per_second": 27.932,
      "eval_steps_per_second": 3.492,
      "step": 3000
    },
    {
      "epoch": 1.3,
      "grad_norm": 3.9499669075012207,
      "learning_rate": 4.816749116607774e-05,
      "loss": 2.4353,
      "step": 3100
    },
    {
      "epoch": 1.3,
      "eval_cer": 0.5479229400492746,
      "eval_loss": 3.4527170658111572,
      "eval_runtime": 354.1425,
      "eval_samples_per_second": 26.763,
      "eval_steps_per_second": 3.346,
      "step": 3100
    },
    {
      "epoch": 1.35,
      "grad_norm": 2.9536569118499756,
      "learning_rate": 4.809681978798587e-05,
      "loss": 2.9787,
      "step": 3200
    },
    {
      "epoch": 1.35,
      "eval_cer": 0.5677500879903015,
      "eval_loss": 2.6674227714538574,
      "eval_runtime": 344.0522,
      "eval_samples_per_second": 27.548,
      "eval_steps_per_second": 3.444,
      "step": 3200
    },
    {
      "epoch": 1.39,
      "grad_norm": 3.094930648803711,
      "learning_rate": 4.8026148409894e-05,
      "loss": 2.2166,
      "step": 3300
    },
    {
      "epoch": 1.39,
      "eval_cer": 0.5857514371749247,
      "eval_loss": 3.2282423973083496,
      "eval_runtime": 358.5567,
      "eval_samples_per_second": 26.434,
      "eval_steps_per_second": 3.305,
      "step": 3300
    },
    {
      "epoch": 1.43,
      "grad_norm": 5.626856803894043,
      "learning_rate": 4.795547703180212e-05,
      "loss": 2.6222,
      "step": 3400
    },
    {
      "epoch": 1.43,
      "eval_cer": 0.5495043213014743,
      "eval_loss": 1.7686785459518433,
      "eval_runtime": 342.362,
      "eval_samples_per_second": 27.684,
      "eval_steps_per_second": 3.461,
      "step": 3400
    },
    {
      "epoch": 1.47,
      "grad_norm": 5.24758768081665,
      "learning_rate": 4.7884805653710253e-05,
      "loss": 2.078,
      "step": 3500
    },
    {
      "epoch": 1.47,
      "eval_cer": 0.5179402448085723,
      "eval_loss": 2.244272470474243,
      "eval_runtime": 354.1685,
      "eval_samples_per_second": 26.761,
      "eval_steps_per_second": 3.346,
      "step": 3500
    },
    {
      "epoch": 1.52,
      "grad_norm": 6.822329044342041,
      "learning_rate": 4.7814134275618375e-05,
      "loss": 2.7694,
      "step": 3600
    },
    {
      "epoch": 1.52,
      "eval_cer": 0.5457598451370693,
      "eval_loss": 2.137202501296997,
      "eval_runtime": 345.4818,
      "eval_samples_per_second": 27.434,
      "eval_steps_per_second": 3.43,
      "step": 3600
    },
    {
      "epoch": 1.56,
      "grad_norm": 2.648332357406616,
      "learning_rate": 4.77434628975265e-05,
      "loss": 1.74,
      "step": 3700
    },
    {
      "epoch": 1.56,
      "eval_cer": 0.5568441789527199,
      "eval_loss": 3.1145806312561035,
      "eval_runtime": 362.0538,
      "eval_samples_per_second": 26.178,
      "eval_steps_per_second": 3.273,
      "step": 3700
    },
    {
      "epoch": 1.6,
      "grad_norm": 17.834945678710938,
      "learning_rate": 4.767279151943463e-05,
      "loss": 1.7586,
      "step": 3800
    },
    {
      "epoch": 1.6,
      "eval_cer": 0.5344653122678034,
      "eval_loss": 1.8897360563278198,
      "eval_runtime": 345.2891,
      "eval_samples_per_second": 27.449,
      "eval_steps_per_second": 3.432,
      "step": 3800
    },
    {
      "epoch": 1.64,
      "grad_norm": 7.307125568389893,
      "learning_rate": 4.760212014134276e-05,
      "loss": 1.6227,
      "step": 3900
    },
    {
      "epoch": 1.64,
      "eval_cer": 0.551083258378632,
      "eval_loss": 1.9913257360458374,
      "eval_runtime": 354.7374,
      "eval_samples_per_second": 26.718,
      "eval_steps_per_second": 3.34,
      "step": 3900
    },
    {
      "epoch": 1.68,
      "grad_norm": 4.344304084777832,
      "learning_rate": 4.753144876325089e-05,
      "loss": 3.1757,
      "step": 4000
    },
    {
      "epoch": 1.68,
      "eval_cer": 0.5003299636306754,
      "eval_loss": 1.8998807668685913,
      "eval_runtime": 348.4123,
      "eval_samples_per_second": 27.203,
      "eval_steps_per_second": 3.401,
      "step": 4000
    },
    {
      "epoch": 1.73,
      "grad_norm": 2.8654837608337402,
      "learning_rate": 4.746077738515901e-05,
      "loss": 2.226,
      "step": 4100
    },
    {
      "epoch": 1.73,
      "eval_cer": 0.5324904188338352,
      "eval_loss": 3.042088031768799,
      "eval_runtime": 354.1261,
      "eval_samples_per_second": 26.764,
      "eval_steps_per_second": 3.346,
      "step": 4100
    },
    {
      "epoch": 1.77,
      "grad_norm": 7.595367908477783,
      "learning_rate": 4.7390106007067143e-05,
      "loss": 2.7925,
      "step": 4200
    },
    {
      "epoch": 1.77,
      "eval_cer": 0.5470479253842243,
      "eval_loss": 3.0831844806671143,
      "eval_runtime": 344.2619,
      "eval_samples_per_second": 27.531,
      "eval_steps_per_second": 3.442,
      "step": 4200
    },
    {
      "epoch": 1.81,
      "grad_norm": 7.425869941711426,
      "learning_rate": 4.7319434628975265e-05,
      "loss": 3.2591,
      "step": 4300
    },
    {
      "epoch": 1.81,
      "eval_cer": 0.5522955691994837,
      "eval_loss": 1.7015308141708374,
      "eval_runtime": 353.5748,
      "eval_samples_per_second": 26.806,
      "eval_steps_per_second": 3.351,
      "step": 4300
    },
    {
      "epoch": 1.85,
      "grad_norm": 2.405174493789673,
      "learning_rate": 4.724876325088339e-05,
      "loss": 2.373,
      "step": 4400
    },
    {
      "epoch": 1.85,
      "eval_cer": 0.5220977865550819,
      "eval_loss": 2.095752477645874,
      "eval_runtime": 344.3236,
      "eval_samples_per_second": 27.526,
      "eval_steps_per_second": 3.442,
      "step": 4400
    },
    {
      "epoch": 1.89,
      "grad_norm": 3.823211908340454,
      "learning_rate": 4.717809187279153e-05,
      "loss": 2.3919,
      "step": 4500
    },
    {
      "epoch": 1.89,
      "eval_cer": 0.5331259043447656,
      "eval_loss": 2.41994571685791,
      "eval_runtime": 353.9416,
      "eval_samples_per_second": 26.778,
      "eval_steps_per_second": 3.348,
      "step": 4500
    },
    {
      "epoch": 1.94,
      "grad_norm": 2.0702412128448486,
      "learning_rate": 4.710742049469965e-05,
      "loss": 2.4842,
      "step": 4600
    },
    {
      "epoch": 1.94,
      "eval_cer": 0.5329523679167807,
      "eval_loss": 3.652872085571289,
      "eval_runtime": 339.8921,
      "eval_samples_per_second": 27.885,
      "eval_steps_per_second": 3.486,
      "step": 4600
    },
    {
      "epoch": 1.98,
      "grad_norm": 3.5580544471740723,
      "learning_rate": 4.703745583038869e-05,
      "loss": 3.577,
      "step": 4700
    },
    {
      "epoch": 1.98,
      "eval_cer": 0.5378284971256502,
      "eval_loss": 2.777487277984619,
      "eval_runtime": 356.1415,
      "eval_samples_per_second": 26.613,
      "eval_steps_per_second": 3.327,
      "step": 4700
    },
    {
      "epoch": 2.02,
      "grad_norm": 2.3930351734161377,
      "learning_rate": 4.6966784452296826e-05,
      "loss": 2.733,
      "step": 4800
    },
    {
      "epoch": 2.02,
      "eval_cer": 0.5305424113253294,
      "eval_loss": 3.2367630004882812,
      "eval_runtime": 346.0446,
      "eval_samples_per_second": 27.39,
      "eval_steps_per_second": 3.424,
      "step": 4800
    },
    {
      "epoch": 2.06,
      "grad_norm": 4.198298931121826,
      "learning_rate": 4.6896113074204954e-05,
      "loss": 1.616,
      "step": 4900
    },
    {
      "epoch": 2.06,
      "eval_cer": 0.523026573071057,
      "eval_loss": 2.150926351547241,
      "eval_runtime": 371.9418,
      "eval_samples_per_second": 25.482,
      "eval_steps_per_second": 3.186,
      "step": 4900
    },
    {
      "epoch": 2.1,
      "grad_norm": 3.6696269512176514,
      "learning_rate": 4.6825441696113075e-05,
      "loss": 1.8252,
      "step": 5000
    },
    {
      "epoch": 2.1,
      "eval_cer": 0.5257933792186461,
      "eval_loss": 3.1695752143859863,
      "eval_runtime": 351.2533,
      "eval_samples_per_second": 26.983,
      "eval_steps_per_second": 3.374,
      "step": 5000
    },
    {
      "epoch": 2.15,
      "grad_norm": 5.002595901489258,
      "learning_rate": 4.67547703180212e-05,
      "loss": 2.0594,
      "step": 5100
    },
    {
      "epoch": 2.15,
      "eval_cer": 0.5161682179030933,
      "eval_loss": 2.4667067527770996,
      "eval_runtime": 377.3711,
      "eval_samples_per_second": 25.116,
      "eval_steps_per_second": 3.14,
      "step": 5100
    },
    {
      "epoch": 2.19,
      "grad_norm": 3.2300162315368652,
      "learning_rate": 4.668409893992933e-05,
      "loss": 1.8512,
      "step": 5200
    },
    {
      "epoch": 2.19,
      "eval_cer": 0.5493185639982793,
      "eval_loss": 1.743632435798645,
      "eval_runtime": 361.1881,
      "eval_samples_per_second": 26.241,
      "eval_steps_per_second": 3.281,
      "step": 5200
    },
    {
      "epoch": 2.23,
      "grad_norm": 2.109961986541748,
      "learning_rate": 4.661342756183746e-05,
      "loss": 2.4585,
      "step": 5300
    },
    {
      "epoch": 2.23,
      "eval_cer": 0.5222273278323101,
      "eval_loss": 1.6703613996505737,
      "eval_runtime": 374.7053,
      "eval_samples_per_second": 25.295,
      "eval_steps_per_second": 3.162,
      "step": 5300
    },
    {
      "epoch": 2.27,
      "grad_norm": 3.7831201553344727,
      "learning_rate": 4.654275618374558e-05,
      "loss": 1.6107,
      "step": 5400
    },
    {
      "epoch": 2.27,
      "eval_cer": 0.5470601462594346,
      "eval_loss": 2.0870959758758545,
      "eval_runtime": 367.5169,
      "eval_samples_per_second": 25.789,
      "eval_steps_per_second": 3.224,
      "step": 5400
    },
    {
      "epoch": 2.32,
      "grad_norm": 12.008744239807129,
      "learning_rate": 4.6472084805653715e-05,
      "loss": 2.3883,
      "step": 5500
    },
    {
      "epoch": 2.32,
      "eval_cer": 0.5516771929138478,
      "eval_loss": 3.582747459411621,
      "eval_runtime": 376.938,
      "eval_samples_per_second": 25.145,
      "eval_steps_per_second": 3.144,
      "step": 5500
    },
    {
      "epoch": 2.36,
      "grad_norm": 2.4807026386260986,
      "learning_rate": 4.6401413427561844e-05,
      "loss": 2.1991,
      "step": 5600
    },
    {
      "epoch": 2.36,
      "eval_cer": 0.5373152203668218,
      "eval_loss": 3.5865066051483154,
      "eval_runtime": 360.1614,
      "eval_samples_per_second": 26.316,
      "eval_steps_per_second": 3.29,
      "step": 5600
    },
    {
      "epoch": 2.4,
      "grad_norm": 6.418155670166016,
      "learning_rate": 4.6330742049469965e-05,
      "loss": 2.3905,
      "step": 5700
    },
    {
      "epoch": 2.4,
      "eval_cer": 0.527748719252278,
      "eval_loss": 2.6406588554382324,
      "eval_runtime": 379.816,
      "eval_samples_per_second": 24.954,
      "eval_steps_per_second": 3.12,
      "step": 5700
    },
    {
      "epoch": 2.44,
      "grad_norm": 2.4806301593780518,
      "learning_rate": 4.626007067137809e-05,
      "loss": 3.3218,
      "step": 5800
    },
    {
      "epoch": 2.44,
      "eval_cer": 0.5217727112744907,
      "eval_loss": 1.8109639883041382,
      "eval_runtime": 351.8213,
      "eval_samples_per_second": 26.94,
      "eval_steps_per_second": 3.368,
      "step": 5800
    },
    {
      "epoch": 2.48,
      "grad_norm": 2.4776511192321777,
      "learning_rate": 4.618939929328622e-05,
      "loss": 2.3876,
      "step": 5900
    },
    {
      "epoch": 2.48,
      "eval_cer": 0.5089310156036134,
      "eval_loss": 3.134875774383545,
      "eval_runtime": 372.2436,
      "eval_samples_per_second": 25.462,
      "eval_steps_per_second": 3.183,
      "step": 5900
    },
    {
      "epoch": 2.53,
      "grad_norm": 9.97956657409668,
      "learning_rate": 4.611872791519435e-05,
      "loss": 4.1527,
      "step": 6000
    },
    {
      "epoch": 2.53,
      "eval_cer": 0.4981815337687224,
      "eval_loss": 1.7425427436828613,
      "eval_runtime": 363.658,
      "eval_samples_per_second": 26.063,
      "eval_steps_per_second": 3.259,
      "step": 6000
    },
    {
      "epoch": 2.57,
      "grad_norm": 3.353976249694824,
      "learning_rate": 4.604805653710248e-05,
      "loss": 1.7398,
      "step": 6100
    },
    {
      "epoch": 2.57,
      "eval_cer": 0.5255465175394001,
      "eval_loss": 2.6437551975250244,
      "eval_runtime": 380.4559,
      "eval_samples_per_second": 24.912,
      "eval_steps_per_second": 3.115,
      "step": 6100
    },
    {
      "epoch": 2.61,
      "grad_norm": 2.671917676925659,
      "learning_rate": 4.5978091872791526e-05,
      "loss": 2.2359,
      "step": 6200
    },
    {
      "epoch": 2.61,
      "eval_cer": 0.537452094169176,
      "eval_loss": 3.4509007930755615,
      "eval_runtime": 364.3511,
      "eval_samples_per_second": 26.013,
      "eval_steps_per_second": 3.252,
      "step": 6200
    },
    {
      "epoch": 2.65,
      "grad_norm": 10.607972145080566,
      "learning_rate": 4.590742049469965e-05,
      "loss": 3.6419,
      "step": 6300
    },
    {
      "epoch": 2.65,
      "eval_cer": 0.5374105431934614,
      "eval_loss": 2.2622523307800293,
      "eval_runtime": 370.8325,
      "eval_samples_per_second": 25.559,
      "eval_steps_per_second": 3.196,
      "step": 6300
    },
    {
      "epoch": 2.69,
      "grad_norm": 4.678300857543945,
      "learning_rate": 4.5836749116607775e-05,
      "loss": 1.7309,
      "step": 6400
    },
    {
      "epoch": 2.69,
      "eval_cer": 0.5418393883696375,
      "eval_loss": 1.8007577657699585,
      "eval_runtime": 357.8741,
      "eval_samples_per_second": 26.484,
      "eval_steps_per_second": 3.311,
      "step": 6400
    },
    {
      "epoch": 2.74,
      "grad_norm": 3.6259689331054688,
      "learning_rate": 4.57660777385159e-05,
      "loss": 2.784,
      "step": 6500
    },
    {
      "epoch": 2.74,
      "eval_cer": 0.5228799225685347,
      "eval_loss": 2.9312586784362793,
      "eval_runtime": 371.6433,
      "eval_samples_per_second": 25.503,
      "eval_steps_per_second": 3.189,
      "step": 6500
    },
    {
      "epoch": 2.78,
      "grad_norm": 1.8296568393707275,
      "learning_rate": 4.569540636042403e-05,
      "loss": 2.4269,
      "step": 6600
    },
    {
      "epoch": 2.78,
      "eval_cer": 0.5211445582886864,
      "eval_loss": 3.522822618484497,
      "eval_runtime": 359.0335,
      "eval_samples_per_second": 26.399,
      "eval_steps_per_second": 3.301,
      "step": 6600
    },
    {
      "epoch": 2.82,
      "grad_norm": 23.777395248413086,
      "learning_rate": 4.562473498233216e-05,
      "loss": 1.5294,
      "step": 6700
    },
    {
      "epoch": 2.82,
      "eval_cer": 0.5097864768683275,
      "eval_loss": 2.758403778076172,
      "eval_runtime": 375.8984,
      "eval_samples_per_second": 25.214,
      "eval_steps_per_second": 3.152,
      "step": 6700
    },
    {
      "epoch": 2.86,
      "grad_norm": 5.520616054534912,
      "learning_rate": 4.555406360424028e-05,
      "loss": 1.6177,
      "step": 6800
    },
    {
      "epoch": 2.86,
      "eval_cer": 0.5417905048687967,
      "eval_loss": 2.8678946495056152,
      "eval_runtime": 360.269,
      "eval_samples_per_second": 26.308,
      "eval_steps_per_second": 3.289,
      "step": 6800
    },
    {
      "epoch": 2.9,
      "grad_norm": 12.704540252685547,
      "learning_rate": 4.5483392226148416e-05,
      "loss": 2.6892,
      "step": 6900
    },
    {
      "epoch": 2.9,
      "eval_cer": 0.5306426225020531,
      "eval_loss": 2.7854502201080322,
      "eval_runtime": 372.5558,
      "eval_samples_per_second": 25.44,
      "eval_steps_per_second": 3.181,
      "step": 6900
    },
    {
      "epoch": 2.95,
      "grad_norm": 3.1302738189697266,
      "learning_rate": 4.541272084805654e-05,
      "loss": 3.1467,
      "step": 7000
    },
    {
      "epoch": 2.95,
      "eval_cer": 0.5148948027061906,
      "eval_loss": 2.9890220165252686,
      "eval_runtime": 359.2287,
      "eval_samples_per_second": 26.384,
      "eval_steps_per_second": 3.299,
      "step": 7000
    },
    {
      "epoch": 2.99,
      "grad_norm": 1.750848412513733,
      "learning_rate": 4.5342049469964665e-05,
      "loss": 2.1825,
      "step": 7100
    },
    {
      "epoch": 2.99,
      "eval_cer": 0.5565606546478432,
      "eval_loss": 3.195390224456787,
      "eval_runtime": 366.6828,
      "eval_samples_per_second": 25.848,
      "eval_steps_per_second": 3.232,
      "step": 7100
    },
    {
      "epoch": 3.03,
      "grad_norm": 237.01815795898438,
      "learning_rate": 4.527137809187279e-05,
      "loss": 3.0545,
      "step": 7200
    },
    {
      "epoch": 3.03,
      "eval_cer": 0.5361493488717688,
      "eval_loss": 2.276334524154663,
      "eval_runtime": 373.4974,
      "eval_samples_per_second": 25.376,
      "eval_steps_per_second": 3.173,
      "step": 7200
    },
    {
      "epoch": 3.07,
      "grad_norm": 2.9628891944885254,
      "learning_rate": 4.520070671378092e-05,
      "loss": 1.9898,
      "step": 7300
    },
    {
      "epoch": 3.07,
      "eval_cer": 0.5218631457510461,
      "eval_loss": 1.94766104221344,
      "eval_runtime": 364.5953,
      "eval_samples_per_second": 25.996,
      "eval_steps_per_second": 3.25,
      "step": 7300
    },
    {
      "epoch": 3.12,
      "grad_norm": 19.71111488342285,
      "learning_rate": 4.513003533568905e-05,
      "loss": 1.7744,
      "step": 7400
    },
    {
      "epoch": 3.12,
      "eval_cer": 0.5048052481326503,
      "eval_loss": 2.699909210205078,
      "eval_runtime": 377.4252,
      "eval_samples_per_second": 25.112,
      "eval_steps_per_second": 3.14,
      "step": 7400
    },
    {
      "epoch": 3.16,
      "grad_norm": 3.2357943058013916,
      "learning_rate": 4.505936395759717e-05,
      "loss": 1.8872,
      "step": 7500
    },
    {
      "epoch": 3.16,
      "eval_cer": 0.5123968558132259,
      "eval_loss": 2.0029726028442383,
      "eval_runtime": 372.4224,
      "eval_samples_per_second": 25.45,
      "eval_steps_per_second": 3.182,
      "step": 7500
    },
    {
      "epoch": 3.2,
      "grad_norm": 2.642812967300415,
      "learning_rate": 4.4988692579505305e-05,
      "loss": 3.1774,
      "step": 7600
    },
    {
      "epoch": 3.2,
      "eval_cer": 0.5154300770403973,
      "eval_loss": 2.36077618598938,
      "eval_runtime": 375.4381,
      "eval_samples_per_second": 25.245,
      "eval_steps_per_second": 3.156,
      "step": 7600
    },
    {
      "epoch": 3.24,
      "grad_norm": 4.245635509490967,
      "learning_rate": 4.491802120141343e-05,
      "loss": 1.7844,
      "step": 7700
    },
    {
      "epoch": 3.24,
      "eval_cer": 0.5169381330413358,
      "eval_loss": 2.6845099925994873,
      "eval_runtime": 355.3776,
      "eval_samples_per_second": 26.67,
      "eval_steps_per_second": 3.334,
      "step": 7700
    },
    {
      "epoch": 3.28,
      "grad_norm": 2.799055576324463,
      "learning_rate": 4.4847349823321555e-05,
      "loss": 3.4812,
      "step": 7800
    },
    {
      "epoch": 3.28,
      "eval_cer": 0.49670036369324627,
      "eval_loss": 3.4256324768066406,
      "eval_runtime": 374.5672,
      "eval_samples_per_second": 25.304,
      "eval_steps_per_second": 3.164,
      "step": 7800
    },
    {
      "epoch": 3.33,
      "grad_norm": 2.1845884323120117,
      "learning_rate": 4.477667844522968e-05,
      "loss": 3.0965,
      "step": 7900
    },
    {
      "epoch": 3.33,
      "eval_cer": 0.5227894880919792,
      "eval_loss": 1.759318232536316,
      "eval_runtime": 366.7135,
      "eval_samples_per_second": 25.846,
      "eval_steps_per_second": 3.231,
      "step": 7900
    },
    {
      "epoch": 3.37,
      "grad_norm": 4.01229190826416,
      "learning_rate": 4.470600706713781e-05,
      "loss": 2.008,
      "step": 8000
    },
    {
      "epoch": 3.37,
      "eval_cer": 0.5001319854522701,
      "eval_loss": 1.97179114818573,
      "eval_runtime": 377.7336,
      "eval_samples_per_second": 25.092,
      "eval_steps_per_second": 3.137,
      "step": 8000
    },
    {
      "epoch": 3.41,
      "grad_norm": 1.3265995979309082,
      "learning_rate": 4.463533568904594e-05,
      "loss": 1.7854,
      "step": 8100
    },
    {
      "epoch": 3.41,
      "eval_cer": 0.48717785772945915,
      "eval_loss": 1.6064122915267944,
      "eval_runtime": 358.3216,
      "eval_samples_per_second": 26.451,
      "eval_steps_per_second": 3.307,
      "step": 8100
    },
    {
      "epoch": 3.45,
      "grad_norm": 3.6079511642456055,
      "learning_rate": 4.456466431095407e-05,
      "loss": 1.7713,
      "step": 8200
    },
    {
      "epoch": 3.45,
      "eval_cer": 0.495703140276094,
      "eval_loss": 2.7158584594726562,
      "eval_runtime": 373.6599,
      "eval_samples_per_second": 25.365,
      "eval_steps_per_second": 3.171,
      "step": 8200
    },
    {
      "epoch": 3.49,
      "grad_norm": 3.969160318374634,
      "learning_rate": 4.4493992932862195e-05,
      "loss": 3.3072,
      "step": 8300
    },
    {
      "epoch": 3.49,
      "eval_cer": 0.49535606742012434,
      "eval_loss": 1.9461767673492432,
      "eval_runtime": 357.0354,
      "eval_samples_per_second": 26.546,
      "eval_steps_per_second": 3.319,
      "step": 8300
    },
    {
      "epoch": 3.54,
      "grad_norm": 1.3938627243041992,
      "learning_rate": 4.442332155477032e-05,
      "loss": 2.1757,
      "step": 8400
    },
    {
      "epoch": 3.54,
      "eval_cer": 0.5164395213327597,
      "eval_loss": 1.6015843152999878,
      "eval_runtime": 375.0822,
      "eval_samples_per_second": 25.269,
      "eval_steps_per_second": 3.159,
      "step": 8400
    },
    {
      "epoch": 3.58,
      "grad_norm": 7.331507682800293,
      "learning_rate": 4.4352650176678445e-05,
      "loss": 2.3463,
      "step": 8500
    },
    {
      "epoch": 3.58,
      "eval_cer": 0.4924132806695084,
      "eval_loss": 3.185905933380127,
      "eval_runtime": 357.5938,
      "eval_samples_per_second": 26.505,
      "eval_steps_per_second": 3.314,
      "step": 8500
    },
    {
      "epoch": 3.62,
      "grad_norm": 2.3522162437438965,
      "learning_rate": 4.428197879858658e-05,
      "loss": 1.9984,
      "step": 8600
    },
    {
      "epoch": 3.62,
      "eval_cer": 0.5185977278948809,
      "eval_loss": 2.115999221801758,
      "eval_runtime": 377.0919,
      "eval_samples_per_second": 25.134,
      "eval_steps_per_second": 3.142,
      "step": 8600
    },
    {
      "epoch": 3.66,
      "grad_norm": 5.214267730712891,
      "learning_rate": 4.42113074204947e-05,
      "loss": 1.5957,
      "step": 8700
    },
    {
      "epoch": 3.66,
      "eval_cer": 0.5082393140667162,
      "eval_loss": 1.6540412902832031,
      "eval_runtime": 357.3421,
      "eval_samples_per_second": 26.524,
      "eval_steps_per_second": 3.316,
      "step": 8700
    },
    {
      "epoch": 3.7,
      "grad_norm": 14.154474258422852,
      "learning_rate": 4.414063604240283e-05,
      "loss": 1.6646,
      "step": 8800
    },
    {
      "epoch": 3.7,
      "eval_cer": 0.49412175902389427,
      "eval_loss": 2.7898120880126953,
      "eval_runtime": 372.8281,
      "eval_samples_per_second": 25.422,
      "eval_steps_per_second": 3.178,
      "step": 8800
    },
    {
      "epoch": 3.75,
      "grad_norm": 6.580379486083984,
      "learning_rate": 4.406996466431096e-05,
      "loss": 1.9858,
      "step": 8900
    },
    {
      "epoch": 3.75,
      "eval_cer": 0.5368801572093387,
      "eval_loss": 3.0827128887176514,
      "eval_runtime": 359.3692,
      "eval_samples_per_second": 26.374,
      "eval_steps_per_second": 3.297,
      "step": 8900
    },
    {
      "epoch": 3.79,
      "grad_norm": 4.16035270690918,
      "learning_rate": 4.3999293286219085e-05,
      "loss": 2.312,
      "step": 9000
    },
    {
      "epoch": 3.79,
      "eval_cer": 0.503619823237261,
      "eval_loss": 1.6672168970108032,
      "eval_runtime": 375.3507,
      "eval_samples_per_second": 25.251,
      "eval_steps_per_second": 3.157,
      "step": 9000
    },
    {
      "epoch": 3.83,
      "grad_norm": 1.9299957752227783,
      "learning_rate": 4.392862190812721e-05,
      "loss": 2.0043,
      "step": 9100
    },
    {
      "epoch": 3.83,
      "eval_cer": 0.4974531696061945,
      "eval_loss": 3.044623613357544,
      "eval_runtime": 365.1986,
      "eval_samples_per_second": 25.953,
      "eval_steps_per_second": 3.245,
      "step": 9100
    },
    {
      "epoch": 3.87,
      "grad_norm": 1.68573796749115,
      "learning_rate": 4.3857950530035335e-05,
      "loss": 2.5474,
      "step": 9200
    },
    {
      "epoch": 3.87,
      "eval_cer": 0.5048394665832389,
      "eval_loss": 2.7017831802368164,
      "eval_runtime": 374.6095,
      "eval_samples_per_second": 25.301,
      "eval_steps_per_second": 3.163,
      "step": 9200
    },
    {
      "epoch": 3.91,
      "grad_norm": 4.48902702331543,
      "learning_rate": 4.378727915194347e-05,
      "loss": 2.2683,
      "step": 9300
    },
    {
      "epoch": 3.91,
      "eval_cer": 0.49138917132689375,
      "eval_loss": 1.5836848020553589,
      "eval_runtime": 360.7206,
      "eval_samples_per_second": 26.275,
      "eval_steps_per_second": 3.285,
      "step": 9300
    },
    {
      "epoch": 3.96,
      "grad_norm": 2.788163661956787,
      "learning_rate": 4.371731448763251e-05,
      "loss": 2.4278,
      "step": 9400
    },
    {
      "epoch": 3.96,
      "eval_cer": 0.5205579562785968,
      "eval_loss": 2.4102938175201416,
      "eval_runtime": 376.1786,
      "eval_samples_per_second": 25.195,
      "eval_steps_per_second": 3.15,
      "step": 9400
    },
    {
      "epoch": 4.0,
      "grad_norm": 2.6063647270202637,
      "learning_rate": 4.364664310954063e-05,
      "loss": 1.5116,
      "step": 9500
    },
    {
      "epoch": 4.0,
      "eval_cer": 0.4881310859958547,
      "eval_loss": 1.728480339050293,
      "eval_runtime": 352.8942,
      "eval_samples_per_second": 26.858,
      "eval_steps_per_second": 3.358,
      "step": 9500
    },
    {
      "epoch": 4.04,
      "grad_norm": 5.240590572357178,
      "learning_rate": 4.357597173144877e-05,
      "loss": 2.2426,
      "step": 9600
    },
    {
      "epoch": 4.04,
      "eval_cer": 0.49515320089163506,
      "eval_loss": 2.496321201324463,
      "eval_runtime": 373.8495,
      "eval_samples_per_second": 25.352,
      "eval_steps_per_second": 3.17,
      "step": 9600
    },
    {
      "epoch": 4.08,
      "grad_norm": 2.557533025741577,
      "learning_rate": 4.3505300353356896e-05,
      "loss": 1.8649,
      "step": 9700
    },
    {
      "epoch": 4.08,
      "eval_cer": 0.5054896171444214,
      "eval_loss": 3.00071120262146,
      "eval_runtime": 362.1744,
      "eval_samples_per_second": 26.17,
      "eval_steps_per_second": 3.272,
      "step": 9700
    },
    {
      "epoch": 4.13,
      "grad_norm": 27.368144989013672,
      "learning_rate": 4.343462897526502e-05,
      "loss": 1.6976,
      "step": 9800
    },
    {
      "epoch": 4.13,
      "eval_cer": 0.4800359782566188,
      "eval_loss": 2.5932295322418213,
      "eval_runtime": 371.4743,
      "eval_samples_per_second": 25.515,
      "eval_steps_per_second": 3.19,
      "step": 9800
    },
    {
      "epoch": 4.17,
      "grad_norm": 4.671775817871094,
      "learning_rate": 4.336395759717315e-05,
      "loss": 1.7712,
      "step": 9900
    },
    {
      "epoch": 4.17,
      "eval_cer": 0.4871509718039967,
      "eval_loss": 2.37919282913208,
      "eval_runtime": 357.2849,
      "eval_samples_per_second": 26.528,
      "eval_steps_per_second": 3.317,
      "step": 9900
    },
    {
      "epoch": 4.21,
      "grad_norm": 2.77481746673584,
      "learning_rate": 4.329328621908127e-05,
      "loss": 1.45,
      "step": 10000
    },
    {
      "epoch": 4.21,
      "eval_cer": 0.49421952602557584,
      "eval_loss": 1.6775341033935547,
      "eval_runtime": 376.5454,
      "eval_samples_per_second": 25.171,
      "eval_steps_per_second": 3.147,
      "step": 10000
    },
    {
      "epoch": 4.25,
      "grad_norm": 7.3151140213012695,
      "learning_rate": 4.32226148409894e-05,
      "loss": 1.5163,
      "step": 10100
    },
    {
      "epoch": 4.25,
      "eval_cer": 0.49458370810683977,
      "eval_loss": 2.463066577911377,
      "eval_runtime": 349.9183,
      "eval_samples_per_second": 27.086,
      "eval_steps_per_second": 3.387,
      "step": 10100
    },
    {
      "epoch": 4.29,
      "grad_norm": 3.947378635406494,
      "learning_rate": 4.315194346289753e-05,
      "loss": 1.4527,
      "step": 10200
    },
    {
      "epoch": 4.29,
      "eval_cer": 0.49508965234054203,
      "eval_loss": 3.2406046390533447,
      "eval_runtime": 377.1597,
      "eval_samples_per_second": 25.13,
      "eval_steps_per_second": 3.142,
      "step": 10200
    },
    {
      "epoch": 4.34,
      "grad_norm": 7.427024841308594,
      "learning_rate": 4.308127208480566e-05,
      "loss": 1.4446,
      "step": 10300
    },
    {
      "epoch": 4.34,
      "eval_cer": 0.48428639865472606,
      "eval_loss": 2.7247695922851562,
      "eval_runtime": 356.4628,
      "eval_samples_per_second": 26.589,
      "eval_steps_per_second": 3.324,
      "step": 10300
    },
    {
      "epoch": 4.38,
      "grad_norm": 3.1798312664031982,
      "learning_rate": 4.3010600706713785e-05,
      "loss": 2.4877,
      "step": 10400
    },
    {
      "epoch": 4.38,
      "eval_cer": 0.48304475773336986,
      "eval_loss": 2.2158772945404053,
      "eval_runtime": 376.4543,
      "eval_samples_per_second": 25.177,
      "eval_steps_per_second": 3.148,
      "step": 10400
    },
    {
      "epoch": 4.42,
      "grad_norm": 2.0509414672851562,
      "learning_rate": 4.293992932862191e-05,
      "loss": 1.9214,
      "step": 10500
    },
    {
      "epoch": 4.42,
      "eval_cer": 0.4835262602166517,
      "eval_loss": 1.906830906867981,
      "eval_runtime": 357.384,
      "eval_samples_per_second": 26.52,
      "eval_steps_per_second": 3.316,
      "step": 10500
    },
    {
      "epoch": 4.46,
      "grad_norm": 3.71893048286438,
      "learning_rate": 4.286925795053004e-05,
      "loss": 1.4426,
      "step": 10600
    },
    {
      "epoch": 4.46,
      "eval_cer": 0.48283211450471236,
      "eval_loss": 2.5203747749328613,
      "eval_runtime": 378.5225,
      "eval_samples_per_second": 25.039,
      "eval_steps_per_second": 3.131,
      "step": 10600
    },
    {
      "epoch": 4.5,
      "grad_norm": 3.5400538444519043,
      "learning_rate": 4.279858657243816e-05,
      "loss": 2.5204,
      "step": 10700
    },
    {
      "epoch": 4.5,
      "eval_cer": 0.496866567596105,
      "eval_loss": 3.1510612964630127,
      "eval_runtime": 360.9029,
      "eval_samples_per_second": 26.262,
      "eval_steps_per_second": 3.283,
      "step": 10700
    },
    {
      "epoch": 4.55,
      "grad_norm": 3.434525966644287,
      "learning_rate": 4.272791519434629e-05,
      "loss": 1.605,
      "step": 10800
    },
    {
      "epoch": 4.55,
      "eval_cer": 0.4771103007312972,
      "eval_loss": 3.0746238231658936,
      "eval_runtime": 370.845,
      "eval_samples_per_second": 25.558,
      "eval_steps_per_second": 3.195,
      "step": 10800
    },
    {
      "epoch": 4.59,
      "grad_norm": 6.7493815422058105,
      "learning_rate": 4.265724381625442e-05,
      "loss": 2.153,
      "step": 10900
    },
    {
      "epoch": 4.59,
      "eval_cer": 0.4859728794337335,
      "eval_loss": 2.0545401573181152,
      "eval_runtime": 361.9049,
      "eval_samples_per_second": 26.189,
      "eval_steps_per_second": 3.274,
      "step": 10900
    },
    {
      "epoch": 4.63,
      "grad_norm": 1.7070709466934204,
      "learning_rate": 4.258657243816255e-05,
      "loss": 1.8732,
      "step": 11000
    },
    {
      "epoch": 4.63,
      "eval_cer": 0.4698926518321536,
      "eval_loss": 2.4386019706726074,
      "eval_runtime": 373.6335,
      "eval_samples_per_second": 25.367,
      "eval_steps_per_second": 3.172,
      "step": 11000
    },
    {
      "epoch": 4.67,
      "grad_norm": 11.575754165649414,
      "learning_rate": 4.2515901060070675e-05,
      "loss": 1.9211,
      "step": 11100
    },
    {
      "epoch": 4.67,
      "eval_cer": 0.5030527746275077,
      "eval_loss": 2.1613056659698486,
      "eval_runtime": 353.1639,
      "eval_samples_per_second": 26.837,
      "eval_steps_per_second": 3.355,
      "step": 11100
    },
    {
      "epoch": 4.71,
      "grad_norm": 5.623499393463135,
      "learning_rate": 4.24452296819788e-05,
      "loss": 2.6473,
      "step": 11200
    },
    {
      "epoch": 4.71,
      "eval_cer": 0.5010998787689179,
      "eval_loss": 2.048527240753174,
      "eval_runtime": 372.3781,
      "eval_samples_per_second": 25.453,
      "eval_steps_per_second": 3.182,
      "step": 11200
    },
    {
      "epoch": 4.76,
      "grad_norm": 4.054663181304932,
      "learning_rate": 4.237455830388693e-05,
      "loss": 1.4888,
      "step": 11300
    },
    {
      "epoch": 4.76,
      "eval_cer": 0.48732939658206564,
      "eval_loss": 2.4912757873535156,
      "eval_runtime": 353.9893,
      "eval_samples_per_second": 26.775,
      "eval_steps_per_second": 3.348,
      "step": 11300
    },
    {
      "epoch": 4.8,
      "grad_norm": 1.9226583242416382,
      "learning_rate": 4.230388692579505e-05,
      "loss": 1.4556,
      "step": 11400
    },
    {
      "epoch": 4.8,
      "eval_cer": 0.4807105705682218,
      "eval_loss": 1.4841840267181396,
      "eval_runtime": 376.2491,
      "eval_samples_per_second": 25.191,
      "eval_steps_per_second": 3.15,
      "step": 11400
    },
    {
      "epoch": 4.84,
      "grad_norm": 6.85697603225708,
      "learning_rate": 4.223321554770318e-05,
      "loss": 1.4391,
      "step": 11500
    },
    {
      "epoch": 4.84,
      "eval_cer": 0.48008730593250165,
      "eval_loss": 3.111499547958374,
      "eval_runtime": 358.9952,
      "eval_samples_per_second": 26.401,
      "eval_steps_per_second": 3.301,
      "step": 11500
    },
    {
      "epoch": 4.88,
      "grad_norm": 12.522397994995117,
      "learning_rate": 4.216254416961131e-05,
      "loss": 1.4244,
      "step": 11600
    },
    {
      "epoch": 4.88,
      "eval_cer": 0.48068612881780143,
      "eval_loss": 2.598745822906494,
      "eval_runtime": 385.9814,
      "eval_samples_per_second": 24.556,
      "eval_steps_per_second": 3.07,
      "step": 11600
    },
    {
      "epoch": 4.93,
      "grad_norm": 3.1015026569366455,
      "learning_rate": 4.209187279151944e-05,
      "loss": 3.7378,
      "step": 11700
    },
    {
      "epoch": 4.93,
      "eval_cer": 0.47458791208791207,
      "eval_loss": 2.3908824920654297,
      "eval_runtime": 373.2148,
      "eval_samples_per_second": 25.396,
      "eval_steps_per_second": 3.175,
      "step": 11700
    },
    {
      "epoch": 4.97,
      "grad_norm": 86.87032318115234,
      "learning_rate": 4.2021201413427565e-05,
      "loss": 2.8329,
      "step": 11800
    },
    {
      "epoch": 4.97,
      "eval_cer": 0.4754898126784248,
      "eval_loss": 2.441450357437134,
      "eval_runtime": 446.0173,
      "eval_samples_per_second": 21.25,
      "eval_steps_per_second": 2.657,
      "step": 11800
    },
    {
      "epoch": 5.01,
      "grad_norm": 2.7503468990325928,
      "learning_rate": 4.195053003533569e-05,
      "loss": 2.4912,
      "step": 11900
    },
    {
      "epoch": 5.01,
      "eval_cer": 0.488764127331743,
      "eval_loss": 1.6247801780700684,
      "eval_runtime": 361.3079,
      "eval_samples_per_second": 26.232,
      "eval_steps_per_second": 3.28,
      "step": 11900
    },
    {
      "epoch": 5.05,
      "grad_norm": 2.511701822280884,
      "learning_rate": 4.187985865724382e-05,
      "loss": 2.009,
      "step": 12000
    },
    {
      "epoch": 5.05,
      "eval_cer": 0.46521938915177347,
      "eval_loss": 1.8090691566467285,
      "eval_runtime": 401.8599,
      "eval_samples_per_second": 23.585,
      "eval_steps_per_second": 2.949,
      "step": 12000
    },
    {
      "epoch": 5.09,
      "grad_norm": 4.231322765350342,
      "learning_rate": 4.180918727915194e-05,
      "loss": 1.6484,
      "step": 12100
    },
    {
      "epoch": 5.09,
      "eval_cer": 0.483240291736733,
      "eval_loss": 1.89494788646698,
      "eval_runtime": 367.7673,
      "eval_samples_per_second": 25.772,
      "eval_steps_per_second": 3.222,
      "step": 12100
    },
    {
      "epoch": 5.14,
      "grad_norm": 3.436452865600586,
      "learning_rate": 4.173851590106007e-05,
      "loss": 1.4205,
      "step": 12200
    },
    {
      "epoch": 5.14,
      "eval_cer": 0.48146582065621213,
      "eval_loss": 2.3205745220184326,
      "eval_runtime": 384.4394,
      "eval_samples_per_second": 24.654,
      "eval_steps_per_second": 3.082,
      "step": 12200
    },
    {
      "epoch": 5.18,
      "grad_norm": 1.96918523311615,
      "learning_rate": 4.16678445229682e-05,
      "loss": 1.3964,
      "step": 12300
    },
    {
      "epoch": 5.18,
      "eval_cer": 0.4692033944702984,
      "eval_loss": 2.9126245975494385,
      "eval_runtime": 358.8304,
      "eval_samples_per_second": 26.414,
      "eval_steps_per_second": 3.302,
      "step": 12300
    },
    {
      "epoch": 5.22,
      "grad_norm": 6.869575500488281,
      "learning_rate": 4.159717314487633e-05,
      "loss": 2.0721,
      "step": 12400
    },
    {
      "epoch": 5.22,
      "eval_cer": 0.4858066755308748,
      "eval_loss": 3.2426090240478516,
      "eval_runtime": 377.7703,
      "eval_samples_per_second": 25.089,
      "eval_steps_per_second": 3.137,
      "step": 12400
    },
    {
      "epoch": 5.26,
      "grad_norm": 2.511401414871216,
      "learning_rate": 4.1526501766784455e-05,
      "loss": 1.747,
      "step": 12500
    },
    {
      "epoch": 5.26,
      "eval_cer": 0.5091607680575652,
      "eval_loss": 3.057870864868164,
      "eval_runtime": 381.9144,
      "eval_samples_per_second": 24.817,
      "eval_steps_per_second": 3.103,
      "step": 12500
    },
    {
      "epoch": 5.3,
      "grad_norm": 9.076600074768066,
      "learning_rate": 4.1455830388692577e-05,
      "loss": 2.0271,
      "step": 12600
    },
    {
      "epoch": 5.3,
      "eval_cer": 0.48266591060185365,
      "eval_loss": 2.858898639678955,
      "eval_runtime": 405.4609,
      "eval_samples_per_second": 23.376,
      "eval_steps_per_second": 2.923,
      "step": 12600
    },
    {
      "epoch": 5.35,
      "grad_norm": 1.6710706949234009,
      "learning_rate": 4.138515901060071e-05,
      "loss": 1.7331,
      "step": 12700
    },
    {
      "epoch": 5.35,
      "eval_cer": 0.501048551093035,
      "eval_loss": 2.9208521842956543,
      "eval_runtime": 361.6906,
      "eval_samples_per_second": 26.205,
      "eval_steps_per_second": 3.276,
      "step": 12700
    },
    {
      "epoch": 5.39,
      "grad_norm": 1.902030348777771,
      "learning_rate": 4.131448763250883e-05,
      "loss": 2.1405,
      "step": 12800
    },
    {
      "epoch": 5.39,
      "eval_cer": 0.4769636502287748,
      "eval_loss": 2.750805139541626,
      "eval_runtime": 383.778,
      "eval_samples_per_second": 24.697,
      "eval_steps_per_second": 3.088,
      "step": 12800
    },
    {
      "epoch": 5.43,
      "grad_norm": 5.2548041343688965,
      "learning_rate": 4.124381625441696e-05,
      "loss": 1.967,
      "step": 12900
    },
    {
      "epoch": 5.43,
      "eval_cer": 0.4813216143287318,
      "eval_loss": 2.39349627494812,
      "eval_runtime": 366.8876,
      "eval_samples_per_second": 25.834,
      "eval_steps_per_second": 3.23,
      "step": 12900
    },
    {
      "epoch": 5.47,
      "grad_norm": 3.4912497997283936,
      "learning_rate": 4.1173144876325096e-05,
      "loss": 2.5075,
      "step": 13000
    },
    {
      "epoch": 5.47,
      "eval_cer": 0.48205731101638577,
      "eval_loss": 2.9853949546813965,
      "eval_runtime": 382.1207,
      "eval_samples_per_second": 24.804,
      "eval_steps_per_second": 3.101,
      "step": 13000
    },
    {
      "epoch": 5.51,
      "grad_norm": 3.671482801437378,
      "learning_rate": 4.110247349823322e-05,
      "loss": 2.3712,
      "step": 13100
    },
    {
      "epoch": 5.51,
      "eval_cer": 0.47388154550076256,
      "eval_loss": 2.6945626735687256,
      "eval_runtime": 374.115,
      "eval_samples_per_second": 25.334,
      "eval_steps_per_second": 3.167,
      "step": 13100
    },
    {
      "epoch": 5.56,
      "grad_norm": 3.400202512741089,
      "learning_rate": 4.1031802120141345e-05,
      "loss": 2.035,
      "step": 13200
    },
    {
      "epoch": 5.56,
      "eval_cer": 0.4721559579210825,
      "eval_loss": 2.723193645477295,
      "eval_runtime": 442.6463,
      "eval_samples_per_second": 21.412,
      "eval_steps_per_second": 2.677,
      "step": 13200
    },
    {
      "epoch": 5.6,
      "grad_norm": 28.22166633605957,
      "learning_rate": 4.0961130742049467e-05,
      "loss": 2.583,
      "step": 13300
    },
    {
      "epoch": 5.6,
      "eval_cer": 0.4679446443236479,
      "eval_loss": 2.848484516143799,
      "eval_runtime": 385.7533,
      "eval_samples_per_second": 24.57,
      "eval_steps_per_second": 3.072,
      "step": 13300
    },
    {
      "epoch": 5.64,
      "grad_norm": 2.6984710693359375,
      "learning_rate": 4.08904593639576e-05,
      "loss": 1.4041,
      "step": 13400
    },
    {
      "epoch": 5.64,
      "eval_cer": 0.46995620038324665,
      "eval_loss": 2.8926751613616943,
      "eval_runtime": 379.5665,
      "eval_samples_per_second": 24.971,
      "eval_steps_per_second": 3.122,
      "step": 13400
    },
    {
      "epoch": 5.68,
      "grad_norm": 28.124893188476562,
      "learning_rate": 4.081978798586573e-05,
      "loss": 1.3813,
      "step": 13500
    },
    {
      "epoch": 5.68,
      "eval_cer": 0.453462907199562,
      "eval_loss": 3.064070701599121,
      "eval_runtime": 364.1898,
      "eval_samples_per_second": 26.025,
      "eval_steps_per_second": 3.254,
      "step": 13500
    },
    {
      "epoch": 5.73,
      "grad_norm": 4.778401851654053,
      "learning_rate": 4.074911660777385e-05,
      "loss": 1.9508,
      "step": 13600
    },
    {
      "epoch": 5.73,
      "eval_cer": 0.4695724649016464,
      "eval_loss": 1.914655327796936,
      "eval_runtime": 378.9026,
      "eval_samples_per_second": 25.014,
      "eval_steps_per_second": 3.127,
      "step": 13600
    },
    {
      "epoch": 5.77,
      "grad_norm": 4.1898345947265625,
      "learning_rate": 4.0678445229681986e-05,
      "loss": 2.1823,
      "step": 13700
    },
    {
      "epoch": 5.77,
      "eval_cer": 0.47729850220953424,
      "eval_loss": 1.71451997756958,
      "eval_runtime": 365.8718,
      "eval_samples_per_second": 25.905,
      "eval_steps_per_second": 3.239,
      "step": 13700
    },
    {
      "epoch": 5.81,
      "grad_norm": 1.8314547538757324,
      "learning_rate": 4.060777385159011e-05,
      "loss": 2.3976,
      "step": 13800
    },
    {
      "epoch": 5.81,
      "eval_cer": 0.4735149192444566,
      "eval_loss": 2.7786667346954346,
      "eval_runtime": 377.8869,
      "eval_samples_per_second": 25.082,
      "eval_steps_per_second": 3.136,
      "step": 13800
    },
    {
      "epoch": 5.85,
      "grad_norm": 3.4417295455932617,
      "learning_rate": 4.0537102473498235e-05,
      "loss": 2.0057,
      "step": 13900
    },
    {
      "epoch": 5.85,
      "eval_cer": 0.47049147471745334,
      "eval_loss": 3.252101182937622,
      "eval_runtime": 365.7598,
      "eval_samples_per_second": 25.913,
      "eval_steps_per_second": 3.24,
      "step": 13900
    },
    {
      "epoch": 5.89,
      "grad_norm": 4.99106502532959,
      "learning_rate": 4.0466431095406356e-05,
      "loss": 1.7663,
      "step": 14000
    },
    {
      "epoch": 5.89,
      "eval_cer": 0.4712345039302335,
      "eval_loss": 2.6297221183776855,
      "eval_runtime": 401.5004,
      "eval_samples_per_second": 23.606,
      "eval_steps_per_second": 2.951,
      "step": 14000
    },
    {
      "epoch": 5.94,
      "grad_norm": 4.448285102844238,
      "learning_rate": 4.039575971731449e-05,
      "loss": 1.842,
      "step": 14100
    },
    {
      "epoch": 5.94,
      "eval_cer": 0.46505562942395684,
      "eval_loss": 2.5416791439056396,
      "eval_runtime": 383.7378,
      "eval_samples_per_second": 24.699,
      "eval_steps_per_second": 3.088,
      "step": 14100
    },
    {
      "epoch": 5.98,
      "grad_norm": 2.196765184402466,
      "learning_rate": 4.032579505300353e-05,
      "loss": 2.0249,
      "step": 14200
    },
    {
      "epoch": 5.98,
      "eval_cer": 0.4723050525986469,
      "eval_loss": 1.9896740913391113,
      "eval_runtime": 382.0154,
      "eval_samples_per_second": 24.811,
      "eval_steps_per_second": 3.102,
      "step": 14200
    },
    {
      "epoch": 6.02,
      "grad_norm": 2.4070873260498047,
      "learning_rate": 4.025512367491166e-05,
      "loss": 2.6354,
      "step": 14300
    },
    {
      "epoch": 6.02,
      "eval_cer": 0.46424660748504165,
      "eval_loss": 1.524404525756836,
      "eval_runtime": 361.2522,
      "eval_samples_per_second": 26.237,
      "eval_steps_per_second": 3.28,
      "step": 14300
    },
    {
      "epoch": 6.06,
      "grad_norm": 2.6403534412384033,
      "learning_rate": 4.018445229681979e-05,
      "loss": 1.4007,
      "step": 14400
    },
    {
      "epoch": 6.06,
      "eval_cer": 0.4687561104376051,
      "eval_loss": 2.6799304485321045,
      "eval_runtime": 377.8259,
      "eval_samples_per_second": 25.086,
      "eval_steps_per_second": 3.136,
      "step": 14400
    },
    {
      "epoch": 6.1,
      "grad_norm": 16.431684494018555,
      "learning_rate": 4.011378091872792e-05,
      "loss": 1.3673,
      "step": 14500
    },
    {
      "epoch": 6.1,
      "eval_cer": 0.4627141097336827,
      "eval_loss": 2.739257335662842,
      "eval_runtime": 365.917,
      "eval_samples_per_second": 25.902,
      "eval_steps_per_second": 3.238,
      "step": 14500
    },
    {
      "epoch": 6.15,
      "grad_norm": 1.2545260190963745,
      "learning_rate": 4.0043109540636045e-05,
      "loss": 1.398,
      "step": 14600
    },
    {
      "epoch": 6.15,
      "eval_cer": 0.46276054905948144,
      "eval_loss": 2.623035430908203,
      "eval_runtime": 380.5058,
      "eval_samples_per_second": 24.909,
      "eval_steps_per_second": 3.114,
      "step": 14600
    },
    {
      "epoch": 6.19,
      "grad_norm": 4.757970333099365,
      "learning_rate": 3.9972438162544173e-05,
      "loss": 1.7177,
      "step": 14700
    },
    {
      "epoch": 6.19,
      "eval_cer": 0.4671600641351531,
      "eval_loss": 3.1201841831207275,
      "eval_runtime": 364.6269,
      "eval_samples_per_second": 25.994,
      "eval_steps_per_second": 3.25,
      "step": 14700
    },
    {
      "epoch": 6.23,
      "grad_norm": 2.604834794998169,
      "learning_rate": 3.99017667844523e-05,
      "loss": 1.4781,
      "step": 14800
    },
    {
      "epoch": 6.23,
      "eval_cer": 0.4741137421297564,
      "eval_loss": 2.4959936141967773,
      "eval_runtime": 380.3259,
      "eval_samples_per_second": 24.921,
      "eval_steps_per_second": 3.116,
      "step": 14800
    },
    {
      "epoch": 6.27,
      "grad_norm": 5.067722797393799,
      "learning_rate": 3.983109540636042e-05,
      "loss": 1.6941,
      "step": 14900
    },
    {
      "epoch": 6.27,
      "eval_cer": 0.4551322787532752,
      "eval_loss": 2.24949049949646,
      "eval_runtime": 359.186,
      "eval_samples_per_second": 26.387,
      "eval_steps_per_second": 3.299,
      "step": 14900
    },
    {
      "epoch": 6.31,
      "grad_norm": 4.323084354400635,
      "learning_rate": 3.976042402826856e-05,
      "loss": 1.4552,
      "step": 15000
    },
    {
      "epoch": 6.31,
      "eval_cer": 0.459111395721716,
      "eval_loss": 2.0859220027923584,
      "eval_runtime": 380.4099,
      "eval_samples_per_second": 24.915,
      "eval_steps_per_second": 3.115,
      "step": 15000
    },
    {
      "epoch": 6.36,
      "grad_norm": 3.9104325771331787,
      "learning_rate": 3.968975265017668e-05,
      "loss": 2.836,
      "step": 15100
    },
    {
      "epoch": 6.36,
      "eval_cer": 0.47354180516991906,
      "eval_loss": 1.5567090511322021,
      "eval_runtime": 368.3958,
      "eval_samples_per_second": 25.728,
      "eval_steps_per_second": 3.217,
      "step": 15100
    },
    {
      "epoch": 6.4,
      "grad_norm": 22.448034286499023,
      "learning_rate": 3.961908127208481e-05,
      "loss": 1.3858,
      "step": 15200
    },
    {
      "epoch": 6.4,
      "eval_cer": 0.45766688827187046,
      "eval_loss": 1.999880313873291,
      "eval_runtime": 380.7492,
      "eval_samples_per_second": 24.893,
      "eval_steps_per_second": 3.112,
      "step": 15200
    },
    {
      "epoch": 6.44,
      "grad_norm": 10.558205604553223,
      "learning_rate": 3.9548409893992935e-05,
      "loss": 1.3921,
      "step": 15300
    },
    {
      "epoch": 6.44,
      "eval_cer": 0.47037904266551955,
      "eval_loss": 2.6411163806915283,
      "eval_runtime": 363.6122,
      "eval_samples_per_second": 26.066,
      "eval_steps_per_second": 3.259,
      "step": 15300
    },
    {
      "epoch": 6.48,
      "grad_norm": 32.08698654174805,
      "learning_rate": 3.947773851590106e-05,
      "loss": 2.0638,
      "step": 15400
    },
    {
      "epoch": 6.48,
      "eval_cer": 0.4723588244495718,
      "eval_loss": 2.9272561073303223,
      "eval_runtime": 386.8412,
      "eval_samples_per_second": 24.501,
      "eval_steps_per_second": 3.063,
      "step": 15400
    },
    {
      "epoch": 6.52,
      "grad_norm": 1.8787578344345093,
      "learning_rate": 3.940706713780919e-05,
      "loss": 1.3608,
      "step": 15500
    },
    {
      "epoch": 6.52,
      "eval_cer": 0.4626187869070431,
      "eval_loss": 3.0016987323760986,
      "eval_runtime": 353.2265,
      "eval_samples_per_second": 26.833,
      "eval_steps_per_second": 3.355,
      "step": 15500
    },
    {
      "epoch": 6.57,
      "grad_norm": 5.287090301513672,
      "learning_rate": 3.933639575971731e-05,
      "loss": 1.6601,
      "step": 15600
    },
    {
      "epoch": 6.57,
      "eval_cer": 0.4719262054671307,
      "eval_loss": 2.0440595149993896,
      "eval_runtime": 377.8824,
      "eval_samples_per_second": 25.082,
      "eval_steps_per_second": 3.136,
      "step": 15600
    },
    {
      "epoch": 6.61,
      "grad_norm": 2.0426106452941895,
      "learning_rate": 3.926572438162545e-05,
      "loss": 1.9773,
      "step": 15700
    },
    {
      "epoch": 6.61,
      "eval_cer": 0.4639239763794924,
      "eval_loss": 2.3498713970184326,
      "eval_runtime": 365.2548,
      "eval_samples_per_second": 25.949,
      "eval_steps_per_second": 3.244,
      "step": 15700
    },
    {
      "epoch": 6.65,
      "grad_norm": 2.1095080375671387,
      "learning_rate": 3.919505300353357e-05,
      "loss": 2.2505,
      "step": 15800
    },
    {
      "epoch": 6.65,
      "eval_cer": 0.4647232216182394,
      "eval_loss": 1.8471035957336426,
      "eval_runtime": 388.913,
      "eval_samples_per_second": 24.37,
      "eval_steps_per_second": 3.047,
      "step": 15800
    },
    {
      "epoch": 6.69,
      "grad_norm": 3.6614439487457275,
      "learning_rate": 3.91243816254417e-05,
      "loss": 1.9341,
      "step": 15900
    },
    {
      "epoch": 6.69,
      "eval_cer": 0.4613698134605608,
      "eval_loss": 2.8519198894500732,
      "eval_runtime": 363.022,
      "eval_samples_per_second": 26.109,
      "eval_steps_per_second": 3.264,
      "step": 15900
    },
    {
      "epoch": 6.74,
      "grad_norm": 8.302891731262207,
      "learning_rate": 3.9053710247349825e-05,
      "loss": 1.4603,
      "step": 16000
    },
    {
      "epoch": 6.74,
      "eval_cer": 0.47225128074772205,
      "eval_loss": 1.5901261568069458,
      "eval_runtime": 376.5314,
      "eval_samples_per_second": 25.172,
      "eval_steps_per_second": 3.147,
      "step": 16000
    },
    {
      "epoch": 6.78,
      "grad_norm": 3.2967019081115723,
      "learning_rate": 3.898303886925795e-05,
      "loss": 1.5493,
      "step": 16100
    },
    {
      "epoch": 6.78,
      "eval_cer": 0.47416751398068124,
      "eval_loss": 1.6550699472427368,
      "eval_runtime": 363.1067,
      "eval_samples_per_second": 26.103,
      "eval_steps_per_second": 3.264,
      "step": 16100
    },
    {
      "epoch": 6.82,
      "grad_norm": 2.848189115524292,
      "learning_rate": 3.891236749116608e-05,
      "loss": 1.5642,
      "step": 16200
    },
    {
      "epoch": 6.82,
      "eval_cer": 0.4807154589183059,
      "eval_loss": 2.1456046104431152,
      "eval_runtime": 376.8904,
      "eval_samples_per_second": 25.148,
      "eval_steps_per_second": 3.144,
      "step": 16200
    },
    {
      "epoch": 6.86,
      "grad_norm": 4.148609638214111,
      "learning_rate": 3.88416961130742e-05,
      "loss": 1.379,
      "step": 16300
    },
    {
      "epoch": 6.86,
      "eval_cer": 0.4555746744358844,
      "eval_loss": 1.7261931896209717,
      "eval_runtime": 360.5421,
      "eval_samples_per_second": 26.288,
      "eval_steps_per_second": 3.287,
      "step": 16300
    },
    {
      "epoch": 6.9,
      "grad_norm": 8.180316925048828,
      "learning_rate": 3.877102473498234e-05,
      "loss": 1.6474,
      "step": 16400
    },
    {
      "epoch": 6.9,
      "eval_cer": 0.4800946384576278,
      "eval_loss": 2.309068202972412,
      "eval_runtime": 377.3872,
      "eval_samples_per_second": 25.115,
      "eval_steps_per_second": 3.14,
      "step": 16400
    },
    {
      "epoch": 6.95,
      "grad_norm": 1.7500585317611694,
      "learning_rate": 3.870035335689046e-05,
      "loss": 1.3755,
      "step": 16500
    },
    {
      "epoch": 6.95,
      "eval_cer": 0.4641219545578976,
      "eval_loss": 1.5506832599639893,
      "eval_runtime": 359.5594,
      "eval_samples_per_second": 26.36,
      "eval_steps_per_second": 3.296,
      "step": 16500
    },
    {
      "epoch": 6.99,
      "grad_norm": 1.8204373121261597,
      "learning_rate": 3.862968197879859e-05,
      "loss": 1.3562,
      "step": 16600
    },
    {
      "epoch": 6.99,
      "eval_cer": 0.46081498572601776,
      "eval_loss": 2.0778937339782715,
      "eval_runtime": 381.9051,
      "eval_samples_per_second": 24.818,
      "eval_steps_per_second": 3.103,
      "step": 16600
    },
    {
      "epoch": 7.03,
      "grad_norm": 2.9875051975250244,
      "learning_rate": 3.8559010600706715e-05,
      "loss": 1.5916,
      "step": 16700
    },
    {
      "epoch": 7.03,
      "eval_cer": 0.4623352626021665,
      "eval_loss": 1.5354750156402588,
      "eval_runtime": 364.3123,
      "eval_samples_per_second": 26.016,
      "eval_steps_per_second": 3.253,
      "step": 16700
    },
    {
      "epoch": 7.07,
      "grad_norm": 4.643332004547119,
      "learning_rate": 3.848833922261484e-05,
      "loss": 1.3365,
      "step": 16800
    },
    {
      "epoch": 7.07,
      "eval_cer": 0.46146024793711626,
      "eval_loss": 1.8595181703567505,
      "eval_runtime": 374.6805,
      "eval_samples_per_second": 25.296,
      "eval_steps_per_second": 3.163,
      "step": 16800
    },
    {
      "epoch": 7.11,
      "grad_norm": 2.433307409286499,
      "learning_rate": 3.841766784452297e-05,
      "loss": 1.6175,
      "step": 16900
    },
    {
      "epoch": 7.11,
      "eval_cer": 0.468870986664581,
      "eval_loss": 1.7079046964645386,
      "eval_runtime": 364.7345,
      "eval_samples_per_second": 25.986,
      "eval_steps_per_second": 3.249,
      "step": 16900
    },
    {
      "epoch": 7.16,
      "grad_norm": 4.391873359680176,
      "learning_rate": 3.834699646643109e-05,
      "loss": 1.3334,
      "step": 17000
    },
    {
      "epoch": 7.16,
      "eval_cer": 0.4652682726526143,
      "eval_loss": 2.7629072666168213,
      "eval_runtime": 380.8752,
      "eval_samples_per_second": 24.885,
      "eval_steps_per_second": 3.111,
      "step": 17000
    },
    {
      "epoch": 7.2,
      "grad_norm": 7.738865375518799,
      "learning_rate": 3.827632508833923e-05,
      "loss": 1.4858,
      "step": 17100
    },
    {
      "epoch": 7.2,
      "eval_cer": 0.4701737319619882,
      "eval_loss": 2.189877986907959,
      "eval_runtime": 363.6078,
      "eval_samples_per_second": 26.067,
      "eval_steps_per_second": 3.259,
      "step": 17100
    },
    {
      "epoch": 7.24,
      "grad_norm": Infinity,
      "learning_rate": 3.820636042402827e-05,
      "loss": 1.5317,
      "step": 17200
    },
    {
      "epoch": 7.24,
      "eval_cer": 0.47163290446208594,
      "eval_loss": 1.9506334066390991,
      "eval_runtime": 380.9688,
      "eval_samples_per_second": 24.879,
      "eval_steps_per_second": 3.11,
      "step": 17200
    },
    {
      "epoch": 7.28,
      "grad_norm": 19.556819915771484,
      "learning_rate": 3.81356890459364e-05,
      "loss": 1.3515,
      "step": 17300
    },
    {
      "epoch": 7.28,
      "eval_cer": 0.45843191506002895,
      "eval_loss": 2.2836623191833496,
      "eval_runtime": 361.9084,
      "eval_samples_per_second": 26.189,
      "eval_steps_per_second": 3.274,
      "step": 17300
    },
    {
      "epoch": 7.32,
      "grad_norm": 3.9697859287261963,
      "learning_rate": 3.8065017667844525e-05,
      "loss": 1.3348,
      "step": 17400
    },
    {
      "epoch": 7.32,
      "eval_cer": 0.460401920143913,
      "eval_loss": 2.626167058944702,
      "eval_runtime": 377.1422,
      "eval_samples_per_second": 25.131,
      "eval_steps_per_second": 3.142,
      "step": 17400
    },
    {
      "epoch": 7.37,
      "grad_norm": 2.342355251312256,
      "learning_rate": 3.7994346289752653e-05,
      "loss": 2.0519,
      "step": 17500
    },
    {
      "epoch": 7.37,
      "eval_cer": 0.4643712604121857,
      "eval_loss": 1.923600196838379,
      "eval_runtime": 360.7805,
      "eval_samples_per_second": 26.271,
      "eval_steps_per_second": 3.285,
      "step": 17500
    },
    {
      "epoch": 7.41,
      "grad_norm": 1.692319393157959,
      "learning_rate": 3.7923674911660775e-05,
      "loss": 1.4731,
      "step": 17600
    },
    {
      "epoch": 7.41,
      "eval_cer": 0.4658597630127879,
      "eval_loss": 1.808614730834961,
      "eval_runtime": 377.9004,
      "eval_samples_per_second": 25.081,
      "eval_steps_per_second": 3.136,
      "step": 17600
    },
    {
      "epoch": 7.45,
      "grad_norm": 2.243326187133789,
      "learning_rate": 3.78530035335689e-05,
      "loss": 1.3005,
      "step": 17700
    },
    {
      "epoch": 7.45,
      "eval_cer": 0.45969066520667945,
      "eval_loss": 1.8014905452728271,
      "eval_runtime": 369.7047,
      "eval_samples_per_second": 25.637,
      "eval_steps_per_second": 3.205,
      "step": 17700
    },
    {
      "epoch": 7.49,
      "grad_norm": 3.0026488304138184,
      "learning_rate": 3.778233215547704e-05,
      "loss": 1.3286,
      "step": 17800
    },
    {
      "epoch": 7.49,
      "eval_cer": 0.44898517852254505,
      "eval_loss": 1.3857167959213257,
      "eval_runtime": 387.1419,
      "eval_samples_per_second": 24.482,
      "eval_steps_per_second": 3.061,
      "step": 17800
    },
    {
      "epoch": 7.54,
      "grad_norm": 4.501519203186035,
      "learning_rate": 3.771166077738516e-05,
      "loss": 1.3409,
      "step": 17900
    },
    {
      "epoch": 7.54,
      "eval_cer": 0.4575813421453991,
      "eval_loss": 2.416992664337158,
      "eval_runtime": 359.7708,
      "eval_samples_per_second": 26.345,
      "eval_steps_per_second": 3.294,
      "step": 17900
    },
    {
      "epoch": 7.58,
      "grad_norm": 10.005617141723633,
      "learning_rate": 3.764098939929329e-05,
      "loss": 2.1197,
      "step": 18000
    },
    {
      "epoch": 7.58,
      "eval_cer": 0.4567307692307692,
      "eval_loss": 1.900498628616333,
      "eval_runtime": 386.3412,
      "eval_samples_per_second": 24.533,
      "eval_steps_per_second": 3.067,
      "step": 18000
    },
    {
      "epoch": 7.62,
      "grad_norm": 3.4333086013793945,
      "learning_rate": 3.7570318021201415e-05,
      "loss": 1.3268,
      "step": 18100
    },
    {
      "epoch": 7.62,
      "eval_cer": 0.4478413046028704,
      "eval_loss": 1.8988970518112183,
      "eval_runtime": 368.4314,
      "eval_samples_per_second": 25.725,
      "eval_steps_per_second": 3.216,
      "step": 18100
    },
    {
      "epoch": 7.66,
      "grad_norm": 4.548975944519043,
      "learning_rate": 3.749964664310954e-05,
      "loss": 1.3146,
      "step": 18200
    },
    {
      "epoch": 7.66,
      "eval_cer": 0.4571047280122013,
      "eval_loss": 2.1630055904388428,
      "eval_runtime": 388.4273,
      "eval_samples_per_second": 24.401,
      "eval_steps_per_second": 3.051,
      "step": 18200
    },
    {
      "epoch": 7.7,
      "grad_norm": 1.7991045713424683,
      "learning_rate": 3.742897526501767e-05,
      "loss": 2.0035,
      "step": 18300
    },
    {
      "epoch": 7.7,
      "eval_cer": 0.45147579289038364,
      "eval_loss": 2.4393255710601807,
      "eval_runtime": 378.3611,
      "eval_samples_per_second": 25.05,
      "eval_steps_per_second": 3.132,
      "step": 18300
    },
    {
      "epoch": 7.75,
      "grad_norm": 2.6425940990448,
      "learning_rate": 3.73583038869258e-05,
      "loss": 1.3294,
      "step": 18400
    },
    {
      "epoch": 7.75,
      "eval_cer": 0.4522603730788784,
      "eval_loss": 1.833477258682251,
      "eval_runtime": 381.8605,
      "eval_samples_per_second": 24.821,
      "eval_steps_per_second": 3.103,
      "step": 18400
    },
    {
      "epoch": 7.79,
      "grad_norm": 4.755228042602539,
      "learning_rate": 3.728763250883393e-05,
      "loss": 1.417,
      "step": 18500
    },
    {
      "epoch": 7.79,
      "eval_cer": 0.47154491416057254,
      "eval_loss": 1.8104172945022583,
      "eval_runtime": 382.6075,
      "eval_samples_per_second": 24.772,
      "eval_steps_per_second": 3.097,
      "step": 18500
    },
    {
      "epoch": 7.83,
      "grad_norm": 2.874833822250366,
      "learning_rate": 3.721696113074205e-05,
      "loss": 1.3215,
      "step": 18600
    },
    {
      "epoch": 7.83,
      "eval_cer": 0.45071076610222516,
      "eval_loss": 3.382107973098755,
      "eval_runtime": 367.7162,
      "eval_samples_per_second": 25.775,
      "eval_steps_per_second": 3.223,
      "step": 18600
    },
    {
      "epoch": 7.87,
      "grad_norm": 2.091411590576172,
      "learning_rate": 3.714628975265018e-05,
      "loss": 1.7498,
      "step": 18700
    },
    {
      "epoch": 7.87,
      "eval_cer": 0.44587129951898635,
      "eval_loss": 2.4873616695404053,
      "eval_runtime": 381.5796,
      "eval_samples_per_second": 24.839,
      "eval_steps_per_second": 3.106,
      "step": 18700
    },
    {
      "epoch": 7.91,
      "grad_norm": 2.1210122108459473,
      "learning_rate": 3.7075618374558305e-05,
      "loss": 1.4463,
      "step": 18800
    },
    {
      "epoch": 7.91,
      "eval_cer": 0.4508696374799578,
      "eval_loss": 2.1447994709014893,
      "eval_runtime": 360.7307,
      "eval_samples_per_second": 26.274,
      "eval_steps_per_second": 3.285,
      "step": 18800
    },
    {
      "epoch": 7.96,
      "grad_norm": 9.81248950958252,
      "learning_rate": 3.700494699646643e-05,
      "loss": 1.327,
      "step": 18900
    },
    {
      "epoch": 7.96,
      "eval_cer": 0.45252678815846076,
      "eval_loss": 2.999907970428467,
      "eval_runtime": 382.5678,
      "eval_samples_per_second": 24.775,
      "eval_steps_per_second": 3.097,
      "step": 18900
    },
    {
      "epoch": 8.0,
      "grad_norm": 2.9528186321258545,
      "learning_rate": 3.693427561837456e-05,
      "loss": 1.3228,
      "step": 19000
    },
    {
      "epoch": 8.0,
      "eval_cer": 0.4471960423917719,
      "eval_loss": 1.6979167461395264,
      "eval_runtime": 363.6541,
      "eval_samples_per_second": 26.063,
      "eval_steps_per_second": 3.259,
      "step": 19000
    },
    {
      "epoch": 8.04,
      "grad_norm": 3.4124321937561035,
      "learning_rate": 3.686360424028269e-05,
      "loss": 1.2656,
      "step": 19100
    },
    {
      "epoch": 8.04,
      "eval_cer": 0.4476115521489187,
      "eval_loss": 2.0440304279327393,
      "eval_runtime": 381.9207,
      "eval_samples_per_second": 24.817,
      "eval_steps_per_second": 3.103,
      "step": 19100
    },
    {
      "epoch": 8.08,
      "grad_norm": 4.451693058013916,
      "learning_rate": 3.679293286219082e-05,
      "loss": 1.2961,
      "step": 19200
    },
    {
      "epoch": 8.08,
      "eval_cer": 0.44471520472410153,
      "eval_loss": 1.5803910493850708,
      "eval_runtime": 361.691,
      "eval_samples_per_second": 26.205,
      "eval_steps_per_second": 3.276,
      "step": 19200
    },
    {
      "epoch": 8.12,
      "grad_norm": 2.6323907375335693,
      "learning_rate": 3.672226148409894e-05,
      "loss": 1.2874,
      "step": 19300
    },
    {
      "epoch": 8.12,
      "eval_cer": 0.4615849008642603,
      "eval_loss": 1.754168152809143,
      "eval_runtime": 385.1298,
      "eval_samples_per_second": 24.61,
      "eval_steps_per_second": 3.077,
      "step": 19300
    },
    {
      "epoch": 8.17,
      "grad_norm": 2.5475683212280273,
      "learning_rate": 3.665159010600707e-05,
      "loss": 1.5227,
      "step": 19400
    },
    {
      "epoch": 8.17,
      "eval_cer": 0.45406906260998786,
      "eval_loss": 1.954167127609253,
      "eval_runtime": 364.6983,
      "eval_samples_per_second": 25.989,
      "eval_steps_per_second": 3.249,
      "step": 19400
    },
    {
      "epoch": 8.21,
      "grad_norm": 1.623938798904419,
      "learning_rate": 3.6580918727915195e-05,
      "loss": 1.2988,
      "step": 19500
    },
    {
      "epoch": 8.21,
      "eval_cer": 0.4476164404990028,
      "eval_loss": 2.2084550857543945,
      "eval_runtime": 384.5947,
      "eval_samples_per_second": 24.644,
      "eval_steps_per_second": 3.081,
      "step": 19500
    },
    {
      "epoch": 8.25,
      "grad_norm": 3.2966043949127197,
      "learning_rate": 3.6510954063604243e-05,
      "loss": 1.5739,
      "step": 19600
    },
    {
      "epoch": 8.25,
      "eval_cer": 0.451500234640804,
      "eval_loss": 1.4389057159423828,
      "eval_runtime": 366.8675,
      "eval_samples_per_second": 25.835,
      "eval_steps_per_second": 3.23,
      "step": 19600
    },
    {
      "epoch": 8.29,
      "grad_norm": 2.2112064361572266,
      "learning_rate": 3.6440282685512365e-05,
      "loss": 1.4271,
      "step": 19700
    },
    {
      "epoch": 8.29,
      "eval_cer": 0.4538075358804896,
      "eval_loss": 2.505011796951294,
      "eval_runtime": 384.1903,
      "eval_samples_per_second": 24.67,
      "eval_steps_per_second": 3.084,
      "step": 19700
    },
    {
      "epoch": 8.34,
      "grad_norm": 3.7650928497314453,
      "learning_rate": 3.63696113074205e-05,
      "loss": 1.665,
      "step": 19800
    },
    {
      "epoch": 8.34,
      "eval_cer": 0.4431655977474483,
      "eval_loss": 2.0151588916778564,
      "eval_runtime": 364.0023,
      "eval_samples_per_second": 26.038,
      "eval_steps_per_second": 3.255,
      "step": 19800
    },
    {
      "epoch": 8.38,
      "grad_norm": 1.848604679107666,
      "learning_rate": 3.629893992932862e-05,
      "loss": 2.2543,
      "step": 19900
    },
    {
      "epoch": 8.38,
      "eval_cer": 0.4581728325055727,
      "eval_loss": 1.8045251369476318,
      "eval_runtime": 381.1374,
      "eval_samples_per_second": 24.868,
      "eval_steps_per_second": 3.109,
      "step": 19900
    },
    {
      "epoch": 8.42,
      "grad_norm": 34.23902893066406,
      "learning_rate": 3.622826855123675e-05,
      "loss": 1.313,
      "step": 20000
    },
    {
      "epoch": 8.42,
      "eval_cer": 0.4448887411520864,
      "eval_loss": 2.140085458755493,
      "eval_runtime": 364.9023,
      "eval_samples_per_second": 25.974,
      "eval_steps_per_second": 3.247,
      "step": 20000
    },
    {
      "epoch": 8.46,
      "grad_norm": 1.8063780069351196,
      "learning_rate": 3.615759717314488e-05,
      "loss": 2.0001,
      "step": 20100
    },
    {
      "epoch": 8.46,
      "eval_cer": 0.44904628289859605,
      "eval_loss": 3.0254170894622803,
      "eval_runtime": 380.5352,
      "eval_samples_per_second": 24.907,
      "eval_steps_per_second": 3.114,
      "step": 20100
    },
    {
      "epoch": 8.5,
      "grad_norm": 4.191483974456787,
      "learning_rate": 3.6086925795053005e-05,
      "loss": 1.2751,
      "step": 20200
    },
    {
      "epoch": 8.5,
      "eval_cer": 0.4562517109225294,
      "eval_loss": 4.0329742431640625,
      "eval_runtime": 362.926,
      "eval_samples_per_second": 26.116,
      "eval_steps_per_second": 3.265,
      "step": 20200
    },
    {
      "epoch": 8.55,
      "grad_norm": 6.447911262512207,
      "learning_rate": 3.6016254416961133e-05,
      "loss": 2.11,
      "step": 20300
    },
    {
      "epoch": 8.55,
      "eval_cer": 0.43633657267998904,
      "eval_loss": 3.3446898460388184,
      "eval_runtime": 385.9035,
      "eval_samples_per_second": 24.561,
      "eval_steps_per_second": 3.071,
      "step": 20300
    },
    {
      "epoch": 8.59,
      "grad_norm": 2.172891139984131,
      "learning_rate": 3.5945583038869255e-05,
      "loss": 1.4999,
      "step": 20400
    },
    {
      "epoch": 8.59,
      "eval_cer": 0.4393355754565719,
      "eval_loss": 3.0930521488189697,
      "eval_runtime": 363.4753,
      "eval_samples_per_second": 26.076,
      "eval_steps_per_second": 3.26,
      "step": 20400
    },
    {
      "epoch": 8.63,
      "grad_norm": 3.379791498184204,
      "learning_rate": 3.587491166077739e-05,
      "loss": 1.4381,
      "step": 20500
    },
    {
      "epoch": 8.63,
      "eval_cer": 0.4450402800046928,
      "eval_loss": 2.9096951484680176,
      "eval_runtime": 384.973,
      "eval_samples_per_second": 24.62,
      "eval_steps_per_second": 3.078,
      "step": 20500
    },
    {
      "epoch": 8.67,
      "grad_norm": 2.5090067386627197,
      "learning_rate": 3.580424028268551e-05,
      "loss": 1.6658,
      "step": 20600
    },
    {
      "epoch": 8.67,
      "eval_cer": 0.4514855695905518,
      "eval_loss": 2.3211405277252197,
      "eval_runtime": 367.2057,
      "eval_samples_per_second": 25.811,
      "eval_steps_per_second": 3.227,
      "step": 20600
    },
    {
      "epoch": 8.71,
      "grad_norm": 2.2183985710144043,
      "learning_rate": 3.573356890459364e-05,
      "loss": 2.1266,
      "step": 20700
    },
    {
      "epoch": 8.71,
      "eval_cer": 0.45651323765202767,
      "eval_loss": 1.70187246799469,
      "eval_runtime": 382.2617,
      "eval_samples_per_second": 24.795,
      "eval_steps_per_second": 3.1,
      "step": 20700
    },
    {
      "epoch": 8.76,
      "grad_norm": 1.713640570640564,
      "learning_rate": 3.5662897526501774e-05,
      "loss": 1.3038,
      "step": 20800
    },
    {
      "epoch": 8.76,
      "eval_cer": 0.44218548355559034,
      "eval_loss": 1.8959708213806152,
      "eval_runtime": 362.6249,
      "eval_samples_per_second": 26.137,
      "eval_steps_per_second": 3.268,
      "step": 20800
    },
    {
      "epoch": 8.8,
      "grad_norm": 10.084565162658691,
      "learning_rate": 3.5592226148409895e-05,
      "loss": 1.3031,
      "step": 20900
    },
    {
      "epoch": 8.8,
      "eval_cer": 0.44013237652027687,
      "eval_loss": 1.9289778470993042,
      "eval_runtime": 383.5806,
      "eval_samples_per_second": 24.709,
      "eval_steps_per_second": 3.089,
      "step": 20900
    },
    {
      "epoch": 8.84,
      "grad_norm": 2.5064008235931396,
      "learning_rate": 3.552155477031802e-05,
      "loss": 2.0089,
      "step": 21000
    },
    {
      "epoch": 8.84,
      "eval_cer": 0.4491318290250675,
      "eval_loss": 2.749382734298706,
      "eval_runtime": 372.5311,
      "eval_samples_per_second": 25.442,
      "eval_steps_per_second": 3.181,
      "step": 21000
    },
    {
      "epoch": 8.88,
      "grad_norm": 1.8727614879608154,
      "learning_rate": 3.5450883392226145e-05,
      "loss": 2.5929,
      "step": 21100
    },
    {
      "epoch": 8.88,
      "eval_cer": 0.4467780884595831,
      "eval_loss": 2.1603004932403564,
      "eval_runtime": 382.1598,
      "eval_samples_per_second": 24.801,
      "eval_steps_per_second": 3.101,
      "step": 21100
    },
    {
      "epoch": 8.92,
      "grad_norm": 2.88053560256958,
      "learning_rate": 3.538021201413428e-05,
      "loss": 1.2652,
      "step": 21200
    },
    {
      "epoch": 8.92,
      "eval_cer": 0.44302872394509407,
      "eval_loss": 2.7272582054138184,
      "eval_runtime": 364.629,
      "eval_samples_per_second": 25.994,
      "eval_steps_per_second": 3.25,
      "step": 21200
    },
    {
      "epoch": 8.97,
      "grad_norm": 3.520254135131836,
      "learning_rate": 3.53095406360424e-05,
      "loss": 1.2819,
      "step": 21300
    },
    {
      "epoch": 8.97,
      "eval_cer": 0.44700784091353485,
      "eval_loss": 1.840844988822937,
      "eval_runtime": 393.1165,
      "eval_samples_per_second": 24.11,
      "eval_steps_per_second": 3.014,
      "step": 21300
    },
    {
      "epoch": 9.01,
      "grad_norm": 1.5368082523345947,
      "learning_rate": 3.523886925795053e-05,
      "loss": 1.2587,
      "step": 21400
    },
    {
      "epoch": 9.01,
      "eval_cer": 0.43732890774705724,
      "eval_loss": 2.2236878871917725,
      "eval_runtime": 363.5106,
      "eval_samples_per_second": 26.074,
      "eval_steps_per_second": 3.26,
      "step": 21400
    },
    {
      "epoch": 9.05,
      "grad_norm": 2.522552251815796,
      "learning_rate": 3.5168197879858664e-05,
      "loss": 1.2512,
      "step": 21500
    },
    {
      "epoch": 9.05,
      "eval_cer": 0.44793418325446793,
      "eval_loss": 2.5103232860565186,
      "eval_runtime": 384.7009,
      "eval_samples_per_second": 24.637,
      "eval_steps_per_second": 3.08,
      "step": 21500
    },
    {
      "epoch": 9.09,
      "grad_norm": 17.708566665649414,
      "learning_rate": 3.5097526501766785e-05,
      "loss": 1.2248,
      "step": 21600
    },
    {
      "epoch": 9.09,
      "eval_cer": 0.4432878064995503,
      "eval_loss": 2.5621256828308105,
      "eval_runtime": 370.664,
      "eval_samples_per_second": 25.57,
      "eval_steps_per_second": 3.197,
      "step": 21600
    },
    {
      "epoch": 9.13,
      "grad_norm": 2.740123987197876,
      "learning_rate": 3.502685512367491e-05,
      "loss": 1.3298,
      "step": 21700
    },
    {
      "epoch": 9.13,
      "eval_cer": 0.441982617027101,
      "eval_loss": 2.49664568901062,
      "eval_runtime": 372.6231,
      "eval_samples_per_second": 25.436,
      "eval_steps_per_second": 3.18,
      "step": 21700
    },
    {
      "epoch": 9.18,
      "grad_norm": 7.561341762542725,
      "learning_rate": 3.495689045936396e-05,
      "loss": 1.2998,
      "step": 21800
    },
    {
      "epoch": 9.18,
      "eval_cer": 0.43526113566149155,
      "eval_loss": 3.151437282562256,
      "eval_runtime": 381.1035,
      "eval_samples_per_second": 24.87,
      "eval_steps_per_second": 3.109,
      "step": 21800
    },
    {
      "epoch": 9.22,
      "grad_norm": 2.588192939758301,
      "learning_rate": 3.488621908127209e-05,
      "loss": 1.2493,
      "step": 21900
    },
    {
      "epoch": 9.22,
      "eval_cer": 0.46399241328066954,
      "eval_loss": 1.8955408334732056,
      "eval_runtime": 369.5234,
      "eval_samples_per_second": 25.649,
      "eval_steps_per_second": 3.207,
      "step": 21900
    },
    {
      "epoch": 9.26,
      "grad_norm": 2.3700144290924072,
      "learning_rate": 3.481554770318021e-05,
      "loss": 1.3379,
      "step": 22000
    },
    {
      "epoch": 9.26,
      "eval_cer": 0.4379032888819366,
      "eval_loss": 1.7394232749938965,
      "eval_runtime": 383.407,
      "eval_samples_per_second": 24.72,
      "eval_steps_per_second": 3.091,
      "step": 22000
    },
    {
      "epoch": 9.3,
      "grad_norm": 2.381225824356079,
      "learning_rate": 3.4744876325088346e-05,
      "loss": 2.2422,
      "step": 22100
    },
    {
      "epoch": 9.3,
      "eval_cer": 0.43050965937976615,
      "eval_loss": 1.691564917564392,
      "eval_runtime": 362.7906,
      "eval_samples_per_second": 26.125,
      "eval_steps_per_second": 3.266,
      "step": 22100
    },
    {
      "epoch": 9.35,
      "grad_norm": 4.20334005355835,
      "learning_rate": 3.467420494699647e-05,
      "loss": 1.239,
      "step": 22200
    },
    {
      "epoch": 9.35,
      "eval_cer": 0.44948623440616325,
      "eval_loss": 3.0320708751678467,
      "eval_runtime": 386.6159,
      "eval_samples_per_second": 24.515,
      "eval_steps_per_second": 3.065,
      "step": 22200
    },
    {
      "epoch": 9.39,
      "grad_norm": 12.748970985412598,
      "learning_rate": 3.4603533568904595e-05,
      "loss": 1.622,
      "step": 22300
    },
    {
      "epoch": 9.39,
      "eval_cer": 0.4492564819522115,
      "eval_loss": 2.2297956943511963,
      "eval_runtime": 368.8399,
      "eval_samples_per_second": 25.697,
      "eval_steps_per_second": 3.213,
      "step": 22300
    },
    {
      "epoch": 9.43,
      "grad_norm": 10.591804504394531,
      "learning_rate": 3.4532862190812723e-05,
      "loss": 1.4811,
      "step": 22400
    },
    {
      "epoch": 9.43,
      "eval_cer": 0.44453922412107466,
      "eval_loss": 2.3979556560516357,
      "eval_runtime": 383.7637,
      "eval_samples_per_second": 24.697,
      "eval_steps_per_second": 3.088,
      "step": 22400
    },
    {
      "epoch": 9.47,
      "grad_norm": 7.108492374420166,
      "learning_rate": 3.446219081272085e-05,
      "loss": 1.5169,
      "step": 22500
    },
    {
      "epoch": 9.47,
      "eval_cer": 0.4425838840874428,
      "eval_loss": 1.763051152229309,
      "eval_runtime": 371.2126,
      "eval_samples_per_second": 25.533,
      "eval_steps_per_second": 3.192,
      "step": 22500
    },
    {
      "epoch": 9.51,
      "grad_norm": 2.171220302581787,
      "learning_rate": 3.439151943462898e-05,
      "loss": 1.2667,
      "step": 22600
    },
    {
      "epoch": 9.51,
      "eval_cer": 0.44492784795275897,
      "eval_loss": 1.6504524946212769,
      "eval_runtime": 384.6648,
      "eval_samples_per_second": 24.64,
      "eval_steps_per_second": 3.081,
      "step": 22600
    },
    {
      "epoch": 9.56,
      "grad_norm": 4.868325233459473,
      "learning_rate": 3.43208480565371e-05,
      "loss": 1.2425,
      "step": 22700
    },
    {
      "epoch": 9.56,
      "eval_cer": 0.4337261937350905,
      "eval_loss": 1.6726810932159424,
      "eval_runtime": 360.9955,
      "eval_samples_per_second": 26.255,
      "eval_steps_per_second": 3.283,
      "step": 22700
    },
    {
      "epoch": 9.6,
      "grad_norm": 4.784325122833252,
      "learning_rate": 3.4250176678445236e-05,
      "loss": 1.2519,
      "step": 22800
    },
    {
      "epoch": 9.6,
      "eval_cer": 0.4394284541081694,
      "eval_loss": 2.026627540588379,
      "eval_runtime": 383.883,
      "eval_samples_per_second": 24.69,
      "eval_steps_per_second": 3.087,
      "step": 22800
    },
    {
      "epoch": 9.64,
      "grad_norm": 34.55485534667969,
      "learning_rate": 3.417950530035336e-05,
      "loss": 1.2678,
      "step": 22900
    },
    {
      "epoch": 9.64,
      "eval_cer": 0.45279075906300104,
      "eval_loss": 1.6792824268341064,
      "eval_runtime": 362.9993,
      "eval_samples_per_second": 26.11,
      "eval_steps_per_second": 3.264,
      "step": 22900
    },
    {
      "epoch": 9.68,
      "grad_norm": 8.905746459960938,
      "learning_rate": 3.4108833922261485e-05,
      "loss": 1.2559,
      "step": 23000
    },
    {
      "epoch": 9.68,
      "eval_cer": 0.452297035704509,
      "eval_loss": 1.6737189292907715,
      "eval_runtime": 383.7761,
      "eval_samples_per_second": 24.697,
      "eval_steps_per_second": 3.088,
      "step": 23000
    },
    {
      "epoch": 9.72,
      "grad_norm": 3.8121345043182373,
      "learning_rate": 3.4038162544169613e-05,
      "loss": 1.2646,
      "step": 23100
    },
    {
      "epoch": 9.72,
      "eval_cer": 0.45129247976223064,
      "eval_loss": 1.4810179471969604,
      "eval_runtime": 361.7355,
      "eval_samples_per_second": 26.201,
      "eval_steps_per_second": 3.276,
      "step": 23100
    },
    {
      "epoch": 9.77,
      "grad_norm": 1.6551660299301147,
      "learning_rate": 3.396749116607774e-05,
      "loss": 1.2812,
      "step": 23200
    },
    {
      "epoch": 9.77,
      "eval_cer": 0.4401079347698565,
      "eval_loss": 1.4080007076263428,
      "eval_runtime": 387.5504,
      "eval_samples_per_second": 24.456,
      "eval_steps_per_second": 3.058,
      "step": 23200
    },
    {
      "epoch": 9.81,
      "grad_norm": 1.3945401906967163,
      "learning_rate": 3.389681978798587e-05,
      "loss": 1.2539,
      "step": 23300
    },
    {
      "epoch": 9.81,
      "eval_cer": 0.44230280395760824,
      "eval_loss": 1.8802071809768677,
      "eval_runtime": 376.3421,
      "eval_samples_per_second": 25.185,
      "eval_steps_per_second": 3.149,
      "step": 23300
    },
    {
      "epoch": 9.85,
      "grad_norm": 19.793508529663086,
      "learning_rate": 3.382614840989399e-05,
      "loss": 1.9455,
      "step": 23400
    },
    {
      "epoch": 9.85,
      "eval_cer": 0.45229947987955105,
      "eval_loss": 1.607908010482788,
      "eval_runtime": 388.4585,
      "eval_samples_per_second": 24.399,
      "eval_steps_per_second": 3.051,
      "step": 23400
    },
    {
      "epoch": 9.89,
      "grad_norm": 25.258953094482422,
      "learning_rate": 3.3755477031802126e-05,
      "loss": 1.2352,
      "step": 23500
    },
    {
      "epoch": 9.89,
      "eval_cer": 0.43993439834187165,
      "eval_loss": 1.3818862438201904,
      "eval_runtime": 363.0875,
      "eval_samples_per_second": 26.104,
      "eval_steps_per_second": 3.264,
      "step": 23500
    },
    {
      "epoch": 9.93,
      "grad_norm": 9.951347351074219,
      "learning_rate": 3.368480565371025e-05,
      "loss": 1.5735,
      "step": 23600
    },
    {
      "epoch": 9.93,
      "eval_cer": 0.4422074811309687,
      "eval_loss": 1.6464565992355347,
      "eval_runtime": 386.065,
      "eval_samples_per_second": 24.55,
      "eval_steps_per_second": 3.069,
      "step": 23600
    },
    {
      "epoch": 9.98,
      "grad_norm": 4.360719203948975,
      "learning_rate": 3.3614134275618375e-05,
      "loss": 1.6236,
      "step": 23700
    },
    {
      "epoch": 9.98,
      "eval_cer": 0.44366909780610847,
      "eval_loss": 1.770129919052124,
      "eval_runtime": 361.9008,
      "eval_samples_per_second": 26.189,
      "eval_steps_per_second": 3.274,
      "step": 23700
    },
    {
      "epoch": 10.02,
      "grad_norm": 12.59276294708252,
      "learning_rate": 3.35434628975265e-05,
      "loss": 1.2522,
      "step": 23800
    },
    {
      "epoch": 10.02,
      "eval_cer": 0.43885651714833207,
      "eval_loss": 2.233562707901001,
      "eval_runtime": 381.3188,
      "eval_samples_per_second": 24.856,
      "eval_steps_per_second": 3.108,
      "step": 23800
    },
    {
      "epoch": 10.06,
      "grad_norm": 6.893917560577393,
      "learning_rate": 3.347279151943463e-05,
      "loss": 1.2052,
      "step": 23900
    },
    {
      "epoch": 10.06,
      "eval_cer": 0.44252522388643384,
      "eval_loss": 2.4695940017700195,
      "eval_runtime": 366.61,
      "eval_samples_per_second": 25.853,
      "eval_steps_per_second": 3.232,
      "step": 23900
    },
    {
      "epoch": 10.1,
      "grad_norm": 2.430925130844116,
      "learning_rate": 3.340212014134276e-05,
      "loss": 1.185,
      "step": 24000
    },
    {
      "epoch": 10.1,
      "eval_cer": 0.4357133080442689,
      "eval_loss": 1.5434983968734741,
      "eval_runtime": 385.0716,
      "eval_samples_per_second": 24.614,
      "eval_steps_per_second": 3.077,
      "step": 24000
    },
    {
      "epoch": 10.15,
      "grad_norm": 2.367770195007324,
      "learning_rate": 3.333144876325088e-05,
      "loss": 1.2225,
      "step": 24100
    },
    {
      "epoch": 10.15,
      "eval_cer": 0.4268898361425052,
      "eval_loss": 1.602295160293579,
      "eval_runtime": 368.0631,
      "eval_samples_per_second": 25.751,
      "eval_steps_per_second": 3.22,
      "step": 24100
    },
    {
      "epoch": 10.19,
      "grad_norm": 2.6913535594940186,
      "learning_rate": 3.3260777385159016e-05,
      "loss": 1.2071,
      "step": 24200
    },
    {
      "epoch": 10.19,
      "eval_cer": 0.43098382933792184,
      "eval_loss": 1.9375296831130981,
      "eval_runtime": 388.5754,
      "eval_samples_per_second": 24.392,
      "eval_steps_per_second": 3.05,
      "step": 24200
    },
    {
      "epoch": 10.23,
      "grad_norm": 2.8861985206604004,
      "learning_rate": 3.319010600706714e-05,
      "loss": 1.2745,
      "step": 24300
    },
    {
      "epoch": 10.23,
      "eval_cer": 0.43681563098822884,
      "eval_loss": 1.8604010343551636,
      "eval_runtime": 365.4924,
      "eval_samples_per_second": 25.932,
      "eval_steps_per_second": 3.242,
      "step": 24300
    },
    {
      "epoch": 10.27,
      "grad_norm": 3.54598069190979,
      "learning_rate": 3.3119434628975265e-05,
      "loss": 1.3486,
      "step": 24400
    },
    {
      "epoch": 10.27,
      "eval_cer": 0.4357133080442689,
      "eval_loss": 1.3347864151000977,
      "eval_runtime": 380.5736,
      "eval_samples_per_second": 24.905,
      "eval_steps_per_second": 3.114,
      "step": 24400
    },
    {
      "epoch": 10.31,
      "grad_norm": 2.1198437213897705,
      "learning_rate": 3.304876325088339e-05,
      "loss": 1.1866,
      "step": 24500
    },
    {
      "epoch": 10.31,
      "eval_cer": 0.4299792733956435,
      "eval_loss": 1.308254361152649,
      "eval_runtime": 368.9525,
      "eval_samples_per_second": 25.689,
      "eval_steps_per_second": 3.212,
      "step": 24500
    },
    {
      "epoch": 10.36,
      "grad_norm": 2.029782295227051,
      "learning_rate": 3.297809187279152e-05,
      "loss": 1.1697,
      "step": 24600
    },
    {
      "epoch": 10.36,
      "eval_cer": 0.42697293809393455,
      "eval_loss": 1.4594690799713135,
      "eval_runtime": 389.4322,
      "eval_samples_per_second": 24.338,
      "eval_steps_per_second": 3.043,
      "step": 24600
    },
    {
      "epoch": 10.4,
      "grad_norm": 2.0395631790161133,
      "learning_rate": 3.290742049469965e-05,
      "loss": 1.1793,
      "step": 24700
    },
    {
      "epoch": 10.4,
      "eval_cer": 0.42983506706816316,
      "eval_loss": 1.475653886795044,
      "eval_runtime": 374.0567,
      "eval_samples_per_second": 25.338,
      "eval_steps_per_second": 3.168,
      "step": 24700
    },
    {
      "epoch": 10.44,
      "grad_norm": 6.735607624053955,
      "learning_rate": 3.283674911660777e-05,
      "loss": 1.4708,
      "step": 24800
    },
    {
      "epoch": 10.44,
      "eval_cer": 0.4388076336474913,
      "eval_loss": 1.9250705242156982,
      "eval_runtime": 386.263,
      "eval_samples_per_second": 24.538,
      "eval_steps_per_second": 3.068,
      "step": 24800
    },
    {
      "epoch": 10.48,
      "grad_norm": 1.8092114925384521,
      "learning_rate": 3.2766077738515906e-05,
      "loss": 1.1854,
      "step": 24900
    },
    {
      "epoch": 10.48,
      "eval_cer": 0.4322132493840679,
      "eval_loss": 1.8587048053741455,
      "eval_runtime": 371.4948,
      "eval_samples_per_second": 25.513,
      "eval_steps_per_second": 3.19,
      "step": 24900
    },
    {
      "epoch": 10.52,
      "grad_norm": 3.406538724899292,
      "learning_rate": 3.269540636042403e-05,
      "loss": 1.2144,
      "step": 25000
    },
    {
      "epoch": 10.52,
      "eval_cer": 0.4353711235383833,
      "eval_loss": 1.5310605764389038,
      "eval_runtime": 388.4081,
      "eval_samples_per_second": 24.402,
      "eval_steps_per_second": 3.051,
      "step": 25000
    },
    {
      "epoch": 10.57,
      "grad_norm": 21.588300704956055,
      "learning_rate": 3.2624734982332155e-05,
      "loss": 1.1869,
      "step": 25100
    },
    {
      "epoch": 10.57,
      "eval_cer": 0.41679294904383873,
      "eval_loss": 1.753686785697937,
      "eval_runtime": 363.8235,
      "eval_samples_per_second": 26.051,
      "eval_steps_per_second": 3.257,
      "step": 25100
    },
    {
      "epoch": 10.61,
      "grad_norm": 2.433943033218384,
      "learning_rate": 3.255406360424029e-05,
      "loss": 1.186,
      "step": 25200
    },
    {
      "epoch": 10.61,
      "eval_cer": 0.4293755621602597,
      "eval_loss": 2.9843032360076904,
      "eval_runtime": 383.8866,
      "eval_samples_per_second": 24.69,
      "eval_steps_per_second": 3.087,
      "step": 25200
    },
    {
      "epoch": 10.65,
      "grad_norm": 2.10840106010437,
      "learning_rate": 3.248339222614841e-05,
      "loss": 1.2008,
      "step": 25300
    },
    {
      "epoch": 10.65,
      "eval_cer": 0.42626657150678504,
      "eval_loss": 2.721179485321045,
      "eval_runtime": 367.0761,
      "eval_samples_per_second": 25.82,
      "eval_steps_per_second": 3.228,
      "step": 25300
    },
    {
      "epoch": 10.69,
      "grad_norm": 2.1997592449188232,
      "learning_rate": 3.241272084805654e-05,
      "loss": 1.1923,
      "step": 25400
    },
    {
      "epoch": 10.69,
      "eval_cer": 0.42187438895623947,
      "eval_loss": 2.4305357933044434,
      "eval_runtime": 385.3951,
      "eval_samples_per_second": 24.593,
      "eval_steps_per_second": 3.075,
      "step": 25400
    },
    {
      "epoch": 10.73,
      "grad_norm": 1.7977826595306396,
      "learning_rate": 3.234204946996466e-05,
      "loss": 1.7363,
      "step": 25500
    },
    {
      "epoch": 10.73,
      "eval_cer": 0.4337506354855109,
      "eval_loss": 2.236682176589966,
      "eval_runtime": 370.2296,
      "eval_samples_per_second": 25.6,
      "eval_steps_per_second": 3.201,
      "step": 25500
    },
    {
      "epoch": 10.78,
      "grad_norm": 3.47428035736084,
      "learning_rate": 3.2271378091872796e-05,
      "loss": 1.3785,
      "step": 25600
    },
    {
      "epoch": 10.78,
      "eval_cer": 0.4252571272144226,
      "eval_loss": 2.5334064960479736,
      "eval_runtime": 380.018,
      "eval_samples_per_second": 24.941,
      "eval_steps_per_second": 3.118,
      "step": 25600
    },
    {
      "epoch": 10.82,
      "grad_norm": 2.5327584743499756,
      "learning_rate": 3.2200706713780924e-05,
      "loss": 1.518,
      "step": 25700
    },
    {
      "epoch": 10.82,
      "eval_cer": 0.42553087481913104,
      "eval_loss": 2.6629116535186768,
      "eval_runtime": 359.4773,
      "eval_samples_per_second": 26.366,
      "eval_steps_per_second": 3.296,
      "step": 25700
    },
    {
      "epoch": 10.86,
      "grad_norm": 5.47965145111084,
      "learning_rate": 3.2130035335689045e-05,
      "loss": 1.1865,
      "step": 25800
    },
    {
      "epoch": 10.86,
      "eval_cer": 0.4287083023737828,
      "eval_loss": 1.6172243356704712,
      "eval_runtime": 386.7546,
      "eval_samples_per_second": 24.506,
      "eval_steps_per_second": 3.064,
      "step": 25800
    },
    {
      "epoch": 10.9,
      "grad_norm": 7.293168544769287,
      "learning_rate": 3.2060070671378093e-05,
      "loss": 1.3999,
      "step": 25900
    },
    {
      "epoch": 10.9,
      "eval_cer": 0.42187194478119744,
      "eval_loss": 1.8544808626174927,
      "eval_runtime": 365.482,
      "eval_samples_per_second": 25.933,
      "eval_steps_per_second": 3.242,
      "step": 25900
    },
    {
      "epoch": 10.95,
      "grad_norm": 3.8992362022399902,
      "learning_rate": 3.198939929328622e-05,
      "loss": 1.1808,
      "step": 26000
    },
    {
      "epoch": 10.95,
      "eval_cer": 0.419022036682179,
      "eval_loss": 1.6932332515716553,
      "eval_runtime": 384.5535,
      "eval_samples_per_second": 24.647,
      "eval_steps_per_second": 3.081,
      "step": 26000
    },
    {
      "epoch": 10.99,
      "grad_norm": 64.95602416992188,
      "learning_rate": 3.191872791519435e-05,
      "loss": 1.2023,
      "step": 26100
    },
    {
      "epoch": 10.99,
      "eval_cer": 0.4221408040358218,
      "eval_loss": 1.5864986181259155,
      "eval_runtime": 363.5244,
      "eval_samples_per_second": 26.073,
      "eval_steps_per_second": 3.26,
      "step": 26100
    },
    {
      "epoch": 11.03,
      "grad_norm": 1.6677500009536743,
      "learning_rate": 3.184805653710248e-05,
      "loss": 1.1585,
      "step": 26200
    },
    {
      "epoch": 11.03,
      "eval_cer": 0.4356106526925032,
      "eval_loss": 1.4082372188568115,
      "eval_runtime": 386.3668,
      "eval_samples_per_second": 24.531,
      "eval_steps_per_second": 3.067,
      "step": 26200
    },
    {
      "epoch": 11.07,
      "grad_norm": 1.4230600595474243,
      "learning_rate": 3.1777385159010606e-05,
      "loss": 1.2774,
      "step": 26300
    },
    {
      "epoch": 11.07,
      "eval_cer": 0.41568573774979467,
      "eval_loss": 1.6947673559188843,
      "eval_runtime": 370.2625,
      "eval_samples_per_second": 25.598,
      "eval_steps_per_second": 3.2,
      "step": 26300
    },
    {
      "epoch": 11.11,
      "grad_norm": 10.38554859161377,
      "learning_rate": 3.170671378091873e-05,
      "loss": 1.1303,
      "step": 26400
    },
    {
      "epoch": 11.11,
      "eval_cer": 0.42248543271674943,
      "eval_loss": 2.5170605182647705,
      "eval_runtime": 383.496,
      "eval_samples_per_second": 24.715,
      "eval_steps_per_second": 3.09,
      "step": 26400
    },
    {
      "epoch": 11.16,
      "grad_norm": 1.8860437870025635,
      "learning_rate": 3.1636042402826855e-05,
      "loss": 1.1186,
      "step": 26500
    },
    {
      "epoch": 11.16,
      "eval_cer": 0.4318832857533925,
      "eval_loss": 1.2920070886611938,
      "eval_runtime": 374.2965,
      "eval_samples_per_second": 25.322,
      "eval_steps_per_second": 3.166,
      "step": 26500
    },
    {
      "epoch": 11.2,
      "grad_norm": 2.1193289756774902,
      "learning_rate": 3.156537102473498e-05,
      "loss": 1.1587,
      "step": 26600
    },
    {
      "epoch": 11.2,
      "eval_cer": 0.4244921004262641,
      "eval_loss": 2.2303755283355713,
      "eval_runtime": 390.1983,
      "eval_samples_per_second": 24.29,
      "eval_steps_per_second": 3.037,
      "step": 26600
    },
    {
      "epoch": 11.24,
      "grad_norm": 13.682442665100098,
      "learning_rate": 3.149469964664311e-05,
      "loss": 1.1569,
      "step": 26700
    },
    {
      "epoch": 11.24,
      "eval_cer": 0.4308762856360721,
      "eval_loss": 1.715321660041809,
      "eval_runtime": 368.2313,
      "eval_samples_per_second": 25.739,
      "eval_steps_per_second": 3.218,
      "step": 26700
    },
    {
      "epoch": 11.28,
      "grad_norm": 1.7043545246124268,
      "learning_rate": 3.142402826855124e-05,
      "loss": 1.1348,
      "step": 26800
    },
    {
      "epoch": 11.28,
      "eval_cer": 0.41776573071057055,
      "eval_loss": 2.2900288105010986,
      "eval_runtime": 394.8533,
      "eval_samples_per_second": 24.004,
      "eval_steps_per_second": 3.001,
      "step": 26800
    },
    {
      "epoch": 11.32,
      "grad_norm": 1.8619517087936401,
      "learning_rate": 3.135335689045937e-05,
      "loss": 1.1366,
      "step": 26900
    },
    {
      "epoch": 11.32,
      "eval_cer": 0.42347532360877554,
      "eval_loss": 1.863010287284851,
      "eval_runtime": 370.2336,
      "eval_samples_per_second": 25.6,
      "eval_steps_per_second": 3.201,
      "step": 26900
    },
    {
      "epoch": 11.37,
      "grad_norm": 1.628806710243225,
      "learning_rate": 3.1282685512367496e-05,
      "loss": 1.1399,
      "step": 27000
    },
    {
      "epoch": 11.37,
      "eval_cer": 0.4168051699190489,
      "eval_loss": 1.782639741897583,
      "eval_runtime": 391.1594,
      "eval_samples_per_second": 24.231,
      "eval_steps_per_second": 3.029,
      "step": 27000
    },
    {
      "epoch": 11.41,
      "grad_norm": 3.158604621887207,
      "learning_rate": 3.121201413427562e-05,
      "loss": 1.1373,
      "step": 27100
    },
    {
      "epoch": 11.41,
      "eval_cer": 0.4201585780767275,
      "eval_loss": 1.5199880599975586,
      "eval_runtime": 368.0376,
      "eval_samples_per_second": 25.753,
      "eval_steps_per_second": 3.22,
      "step": 27100
    },
    {
      "epoch": 11.45,
      "grad_norm": 1.9599921703338623,
      "learning_rate": 3.1141342756183745e-05,
      "loss": 1.1358,
      "step": 27200
    },
    {
      "epoch": 11.45,
      "eval_cer": 0.4173404442532556,
      "eval_loss": 1.598440170288086,
      "eval_runtime": 388.8994,
      "eval_samples_per_second": 24.371,
      "eval_steps_per_second": 3.047,
      "step": 27200
    },
    {
      "epoch": 11.49,
      "grad_norm": 1.7773711681365967,
      "learning_rate": 3.107067137809187e-05,
      "loss": 1.1567,
      "step": 27300
    },
    {
      "epoch": 11.49,
      "eval_cer": 0.4288060693754644,
      "eval_loss": 3.0100185871124268,
      "eval_runtime": 392.5508,
      "eval_samples_per_second": 24.145,
      "eval_steps_per_second": 3.019,
      "step": 27300
    },
    {
      "epoch": 11.53,
      "grad_norm": 19.142953872680664,
      "learning_rate": 3.1e-05,
      "loss": 1.1284,
      "step": 27400
    },
    {
      "epoch": 11.53,
      "eval_cer": 0.43251877126432287,
      "eval_loss": 2.1405186653137207,
      "eval_runtime": 413.4099,
      "eval_samples_per_second": 22.926,
      "eval_steps_per_second": 2.866,
      "step": 27400
    },
    {
      "epoch": 11.58,
      "grad_norm": 9.353096961975098,
      "learning_rate": 3.092932862190813e-05,
      "loss": 1.3106,
      "step": 27500
    },
    {
      "epoch": 11.58,
      "eval_cer": 0.4250787024363537,
      "eval_loss": 2.7651588916778564,
      "eval_runtime": 372.5808,
      "eval_samples_per_second": 25.439,
      "eval_steps_per_second": 3.181,
      "step": 27500
    },
    {
      "epoch": 11.62,
      "grad_norm": 2.0867862701416016,
      "learning_rate": 3.085865724381626e-05,
      "loss": 1.1146,
      "step": 27600
    },
    {
      "epoch": 11.62,
      "eval_cer": 0.41358374721364044,
      "eval_loss": 2.7960143089294434,
      "eval_runtime": 392.1126,
      "eval_samples_per_second": 24.172,
      "eval_steps_per_second": 3.022,
      "step": 27600
    },
    {
      "epoch": 11.66,
      "grad_norm": 1.8952158689498901,
      "learning_rate": 3.0787985865724386e-05,
      "loss": 1.1186,
      "step": 27700
    },
    {
      "epoch": 11.66,
      "eval_cer": 0.41495004106214073,
      "eval_loss": 2.6641368865966797,
      "eval_runtime": 367.6488,
      "eval_samples_per_second": 25.78,
      "eval_steps_per_second": 3.223,
      "step": 27700
    },
    {
      "epoch": 11.7,
      "grad_norm": 6.817134380340576,
      "learning_rate": 3.071802120141343e-05,
      "loss": 1.581,
      "step": 27800
    },
    {
      "epoch": 11.7,
      "eval_cer": 0.4181641312424231,
      "eval_loss": 2.996872901916504,
      "eval_runtime": 382.8066,
      "eval_samples_per_second": 24.759,
      "eval_steps_per_second": 3.096,
      "step": 27800
    },
    {
      "epoch": 11.74,
      "grad_norm": 4.087586879730225,
      "learning_rate": 3.0647349823321555e-05,
      "loss": 1.1385,
      "step": 27900
    },
    {
      "epoch": 11.74,
      "eval_cer": 0.42436255914903603,
      "eval_loss": 2.688666343688965,
      "eval_runtime": 375.5428,
      "eval_samples_per_second": 25.238,
      "eval_steps_per_second": 3.155,
      "step": 27900
    },
    {
      "epoch": 11.79,
      "grad_norm": 2.216014862060547,
      "learning_rate": 3.0576678445229683e-05,
      "loss": 1.2095,
      "step": 28000
    },
    {
      "epoch": 11.79,
      "eval_cer": 0.42844188729420046,
      "eval_loss": 1.8659263849258423,
      "eval_runtime": 392.5017,
      "eval_samples_per_second": 24.148,
      "eval_steps_per_second": 3.019,
      "step": 28000
    },
    {
      "epoch": 11.83,
      "grad_norm": 6.676244258880615,
      "learning_rate": 3.0506007067137808e-05,
      "loss": 1.4826,
      "step": 28100
    },
    {
      "epoch": 11.83,
      "eval_cer": 0.4170275898478745,
      "eval_loss": 1.5196857452392578,
      "eval_runtime": 370.7857,
      "eval_samples_per_second": 25.562,
      "eval_steps_per_second": 3.196,
      "step": 28100
    },
    {
      "epoch": 11.87,
      "grad_norm": 5.378576755523682,
      "learning_rate": 3.0435335689045936e-05,
      "loss": 1.125,
      "step": 28200
    },
    {
      "epoch": 11.87,
      "eval_cer": 0.4203663329553009,
      "eval_loss": 1.5670437812805176,
      "eval_runtime": 388.8978,
      "eval_samples_per_second": 24.371,
      "eval_steps_per_second": 3.047,
      "step": 28200
    },
    {
      "epoch": 11.91,
      "grad_norm": 4.532820701599121,
      "learning_rate": 3.0364664310954068e-05,
      "loss": 1.1398,
      "step": 28300
    },
    {
      "epoch": 11.91,
      "eval_cer": 0.42046654413202456,
      "eval_loss": 1.8523951768875122,
      "eval_runtime": 367.1427,
      "eval_samples_per_second": 25.816,
      "eval_steps_per_second": 3.228,
      "step": 28300
    },
    {
      "epoch": 11.96,
      "grad_norm": 13.059289932250977,
      "learning_rate": 3.0293992932862192e-05,
      "loss": 1.1318,
      "step": 28400
    },
    {
      "epoch": 11.96,
      "eval_cer": 0.4224341050408666,
      "eval_loss": 2.1082444190979004,
      "eval_runtime": 393.3473,
      "eval_samples_per_second": 24.096,
      "eval_steps_per_second": 3.013,
      "step": 28400
    },
    {
      "epoch": 12.0,
      "grad_norm": 5.903038501739502,
      "learning_rate": 3.0223321554770317e-05,
      "loss": 1.1131,
      "step": 28500
    },
    {
      "epoch": 12.0,
      "eval_cer": 0.41807125259082556,
      "eval_loss": 1.7783021926879883,
      "eval_runtime": 365.8673,
      "eval_samples_per_second": 25.906,
      "eval_steps_per_second": 3.239,
      "step": 28500
    },
    {
      "epoch": 12.04,
      "grad_norm": 7.630153656005859,
      "learning_rate": 3.015265017667845e-05,
      "loss": 1.0524,
      "step": 28600
    },
    {
      "epoch": 12.04,
      "eval_cer": 0.4113008877243753,
      "eval_loss": 1.9460973739624023,
      "eval_runtime": 389.3684,
      "eval_samples_per_second": 24.342,
      "eval_steps_per_second": 3.043,
      "step": 28600
    },
    {
      "epoch": 12.08,
      "grad_norm": 16.8402099609375,
      "learning_rate": 3.0081978798586573e-05,
      "loss": 1.073,
      "step": 28700
    },
    {
      "epoch": 12.08,
      "eval_cer": 0.4151773493410504,
      "eval_loss": 2.0268595218658447,
      "eval_runtime": 367.5014,
      "eval_samples_per_second": 25.79,
      "eval_steps_per_second": 3.224,
      "step": 28700
    },
    {
      "epoch": 12.12,
      "grad_norm": 1.9278266429901123,
      "learning_rate": 3.0011307420494698e-05,
      "loss": 1.0857,
      "step": 28800
    },
    {
      "epoch": 12.12,
      "eval_cer": 0.4125180868953111,
      "eval_loss": 1.7026644945144653,
      "eval_runtime": 390.3027,
      "eval_samples_per_second": 24.284,
      "eval_steps_per_second": 3.036,
      "step": 28800
    },
    {
      "epoch": 12.17,
      "grad_norm": 1.4629472494125366,
      "learning_rate": 2.994063604240283e-05,
      "loss": 1.1243,
      "step": 28900
    },
    {
      "epoch": 12.17,
      "eval_cer": 0.411283778499081,
      "eval_loss": 2.061549663543701,
      "eval_runtime": 372.4148,
      "eval_samples_per_second": 25.45,
      "eval_steps_per_second": 3.182,
      "step": 28900
    },
    {
      "epoch": 12.21,
      "grad_norm": 15.632731437683105,
      "learning_rate": 2.9869964664310958e-05,
      "loss": 1.0708,
      "step": 29000
    },
    {
      "epoch": 12.21,
      "eval_cer": 0.41651920143913024,
      "eval_loss": 2.2883496284484863,
      "eval_runtime": 388.5509,
      "eval_samples_per_second": 24.393,
      "eval_steps_per_second": 3.05,
      "step": 29000
    },
    {
      "epoch": 12.25,
      "grad_norm": 6.1355180740356445,
      "learning_rate": 2.9799293286219082e-05,
      "loss": 1.0688,
      "step": 29100
    },
    {
      "epoch": 12.25,
      "eval_cer": 0.4098588244495718,
      "eval_loss": 2.406083822250366,
      "eval_runtime": 374.5529,
      "eval_samples_per_second": 25.305,
      "eval_steps_per_second": 3.164,
      "step": 29100
    },
    {
      "epoch": 12.29,
      "grad_norm": 26.448299407958984,
      "learning_rate": 2.9728621908127207e-05,
      "loss": 1.091,
      "step": 29200
    },
    {
      "epoch": 12.29,
      "eval_cer": 0.41021811818075166,
      "eval_loss": 1.8954789638519287,
      "eval_runtime": 388.9635,
      "eval_samples_per_second": 24.367,
      "eval_steps_per_second": 3.047,
      "step": 29200
    },
    {
      "epoch": 12.33,
      "grad_norm": 19.144874572753906,
      "learning_rate": 2.965795053003534e-05,
      "loss": 1.0624,
      "step": 29300
    },
    {
      "epoch": 12.33,
      "eval_cer": 0.41046497985999764,
      "eval_loss": 1.6875534057617188,
      "eval_runtime": 368.9225,
      "eval_samples_per_second": 25.691,
      "eval_steps_per_second": 3.212,
      "step": 29300
    },
    {
      "epoch": 12.38,
      "grad_norm": 2.1881866455078125,
      "learning_rate": 2.9587279151943463e-05,
      "loss": 1.0462,
      "step": 29400
    },
    {
      "epoch": 12.38,
      "eval_cer": 0.40699180712525906,
      "eval_loss": 1.977940320968628,
      "eval_runtime": 389.1348,
      "eval_samples_per_second": 24.357,
      "eval_steps_per_second": 3.045,
      "step": 29400
    },
    {
      "epoch": 12.42,
      "grad_norm": 2.38139009475708,
      "learning_rate": 2.951660777385159e-05,
      "loss": 1.0711,
      "step": 29500
    },
    {
      "epoch": 12.42,
      "eval_cer": 0.4067278362207188,
      "eval_loss": 1.981135606765747,
      "eval_runtime": 374.3024,
      "eval_samples_per_second": 25.322,
      "eval_steps_per_second": 3.166,
      "step": 29500
    },
    {
      "epoch": 12.46,
      "grad_norm": 13.329228401184082,
      "learning_rate": 2.944593639575972e-05,
      "loss": 1.041,
      "step": 29600
    },
    {
      "epoch": 12.46,
      "eval_cer": 0.41142309647647723,
      "eval_loss": 2.5235061645507812,
      "eval_runtime": 396.1248,
      "eval_samples_per_second": 23.927,
      "eval_steps_per_second": 2.991,
      "step": 29600
    },
    {
      "epoch": 12.5,
      "grad_norm": 3.899754524230957,
      "learning_rate": 2.9375265017667848e-05,
      "loss": 1.1686,
      "step": 29700
    },
    {
      "epoch": 12.5,
      "eval_cer": 0.4094359821672989,
      "eval_loss": 1.6854169368743896,
      "eval_runtime": 371.974,
      "eval_samples_per_second": 25.48,
      "eval_steps_per_second": 3.186,
      "step": 29700
    },
    {
      "epoch": 12.54,
      "grad_norm": 4.518542766571045,
      "learning_rate": 2.9304593639575972e-05,
      "loss": 1.0679,
      "step": 29800
    },
    {
      "epoch": 12.54,
      "eval_cer": 0.4047798287122131,
      "eval_loss": 2.345759868621826,
      "eval_runtime": 387.3324,
      "eval_samples_per_second": 24.47,
      "eval_steps_per_second": 3.059,
      "step": 29800
    },
    {
      "epoch": 12.59,
      "grad_norm": 3.159818410873413,
      "learning_rate": 2.9233922261484097e-05,
      "loss": 1.0724,
      "step": 29900
    },
    {
      "epoch": 12.59,
      "eval_cer": 0.4022965468695006,
      "eval_loss": 2.1892457008361816,
      "eval_runtime": 368.124,
      "eval_samples_per_second": 25.747,
      "eval_steps_per_second": 3.219,
      "step": 29900
    },
    {
      "epoch": 12.63,
      "grad_norm": 5.468438148498535,
      "learning_rate": 2.916395759717315e-05,
      "loss": 1.3109,
      "step": 30000
    },
    {
      "epoch": 12.63,
      "eval_cer": 0.4083043291228345,
      "eval_loss": 2.8631815910339355,
      "eval_runtime": 397.7294,
      "eval_samples_per_second": 23.83,
      "eval_steps_per_second": 2.979,
      "step": 30000
    },
    {
      "epoch": 12.67,
      "grad_norm": 3.4033985137939453,
      "learning_rate": 2.9093286219081274e-05,
      "loss": 1.0724,
      "step": 30100
    },
    {
      "epoch": 12.67,
      "eval_cer": 0.4132855578585116,
      "eval_loss": 2.082000732421875,
      "eval_runtime": 370.8747,
      "eval_samples_per_second": 25.556,
      "eval_steps_per_second": 3.195,
      "step": 30100
    },
    {
      "epoch": 12.71,
      "grad_norm": 8.163016319274902,
      "learning_rate": 2.9022614840989398e-05,
      "loss": 1.2078,
      "step": 30200
    },
    {
      "epoch": 12.71,
      "eval_cer": 0.41148908920261235,
      "eval_loss": 2.2178566455841064,
      "eval_runtime": 392.2715,
      "eval_samples_per_second": 24.162,
      "eval_steps_per_second": 3.021,
      "step": 30200
    },
    {
      "epoch": 12.76,
      "grad_norm": 7.318975448608398,
      "learning_rate": 2.895194346289753e-05,
      "loss": 1.0775,
      "step": 30300
    },
    {
      "epoch": 12.76,
      "eval_cer": 0.4067913847718118,
      "eval_loss": 2.1497671604156494,
      "eval_runtime": 374.4081,
      "eval_samples_per_second": 25.315,
      "eval_steps_per_second": 3.165,
      "step": 30300
    },
    {
      "epoch": 12.8,
      "grad_norm": 1.8343230485916138,
      "learning_rate": 2.8881272084805654e-05,
      "loss": 1.0572,
      "step": 30400
    },
    {
      "epoch": 12.8,
      "eval_cer": 0.40438142818036055,
      "eval_loss": 2.247307062149048,
      "eval_runtime": 396.7902,
      "eval_samples_per_second": 23.887,
      "eval_steps_per_second": 2.986,
      "step": 30400
    },
    {
      "epoch": 12.84,
      "grad_norm": 2.385056495666504,
      "learning_rate": 2.881060070671378e-05,
      "loss": 1.0668,
      "step": 30500
    },
    {
      "epoch": 12.84,
      "eval_cer": 0.40543242344843766,
      "eval_loss": 2.0315330028533936,
      "eval_runtime": 370.6236,
      "eval_samples_per_second": 25.573,
      "eval_steps_per_second": 3.197,
      "step": 30500
    },
    {
      "epoch": 12.88,
      "grad_norm": 3.699707508087158,
      "learning_rate": 2.873992932862191e-05,
      "loss": 1.2704,
      "step": 30600
    },
    {
      "epoch": 12.88,
      "eval_cer": 0.40483360056313794,
      "eval_loss": 3.221102476119995,
      "eval_runtime": 396.883,
      "eval_samples_per_second": 23.881,
      "eval_steps_per_second": 2.986,
      "step": 30600
    },
    {
      "epoch": 12.92,
      "grad_norm": 2.9071247577667236,
      "learning_rate": 2.866925795053004e-05,
      "loss": 2.4896,
      "step": 30700
    },
    {
      "epoch": 12.92,
      "eval_cer": 0.4067840522466857,
      "eval_loss": 2.3889622688293457,
      "eval_runtime": 374.4971,
      "eval_samples_per_second": 25.309,
      "eval_steps_per_second": 3.164,
      "step": 30700
    },
    {
      "epoch": 12.97,
      "grad_norm": 2.6603596210479736,
      "learning_rate": 2.8598586572438163e-05,
      "loss": 1.0564,
      "step": 30800
    },
    {
      "epoch": 12.97,
      "eval_cer": 0.40476760783700283,
      "eval_loss": 2.143598794937134,
      "eval_runtime": 399.5314,
      "eval_samples_per_second": 23.723,
      "eval_steps_per_second": 2.966,
      "step": 30800
    },
    {
      "epoch": 13.01,
      "grad_norm": 2.8563497066497803,
      "learning_rate": 2.8527915194346288e-05,
      "loss": 1.0411,
      "step": 30900
    },
    {
      "epoch": 13.01,
      "eval_cer": 0.40590170505650935,
      "eval_loss": 2.6135616302490234,
      "eval_runtime": 370.2003,
      "eval_samples_per_second": 25.602,
      "eval_steps_per_second": 3.201,
      "step": 30900
    },
    {
      "epoch": 13.05,
      "grad_norm": 3.738598585128784,
      "learning_rate": 2.845724381625442e-05,
      "loss": 0.9841,
      "step": 31000
    },
    {
      "epoch": 13.05,
      "eval_cer": 0.4061705643111337,
      "eval_loss": 2.7012014389038086,
      "eval_runtime": 394.4654,
      "eval_samples_per_second": 24.027,
      "eval_steps_per_second": 3.004,
      "step": 31000
    },
    {
      "epoch": 13.09,
      "grad_norm": 3.3957040309906006,
      "learning_rate": 2.8386572438162544e-05,
      "loss": 1.444,
      "step": 31100
    },
    {
      "epoch": 13.09,
      "eval_cer": 0.3994588596456924,
      "eval_loss": 2.732797861099243,
      "eval_runtime": 371.9235,
      "eval_samples_per_second": 25.484,
      "eval_steps_per_second": 3.186,
      "step": 31100
    },
    {
      "epoch": 13.13,
      "grad_norm": 2.130281448364258,
      "learning_rate": 2.8315901060070672e-05,
      "loss": 0.9865,
      "step": 31200
    },
    {
      "epoch": 13.13,
      "eval_cer": 0.4065005279418091,
      "eval_loss": 2.812870502471924,
      "eval_runtime": 395.591,
      "eval_samples_per_second": 23.959,
      "eval_steps_per_second": 2.996,
      "step": 31200
    },
    {
      "epoch": 13.18,
      "grad_norm": 3.105449676513672,
      "learning_rate": 2.82452296819788e-05,
      "loss": 0.9784,
      "step": 31300
    },
    {
      "epoch": 13.18,
      "eval_cer": 0.3964916311446561,
      "eval_loss": 2.1949923038482666,
      "eval_runtime": 377.7569,
      "eval_samples_per_second": 25.09,
      "eval_steps_per_second": 3.137,
      "step": 31300
    },
    {
      "epoch": 13.22,
      "grad_norm": 3.4888200759887695,
      "learning_rate": 2.817455830388693e-05,
      "loss": 0.9847,
      "step": 31400
    },
    {
      "epoch": 13.22,
      "eval_cer": 0.39935864846896874,
      "eval_loss": 2.095374584197998,
      "eval_runtime": 395.2097,
      "eval_samples_per_second": 23.982,
      "eval_steps_per_second": 2.998,
      "step": 31400
    },
    {
      "epoch": 13.26,
      "grad_norm": 1.688668966293335,
      "learning_rate": 2.8103886925795053e-05,
      "loss": 1.0167,
      "step": 31500
    },
    {
      "epoch": 13.26,
      "eval_cer": 0.40794747956669664,
      "eval_loss": 1.7657544612884521,
      "eval_runtime": 377.384,
      "eval_samples_per_second": 25.115,
      "eval_steps_per_second": 3.14,
      "step": 31500
    },
    {
      "epoch": 13.3,
      "grad_norm": 2.9201929569244385,
      "learning_rate": 2.8033215547703178e-05,
      "loss": 0.9921,
      "step": 31600
    },
    {
      "epoch": 13.3,
      "eval_cer": 0.40619256188651204,
      "eval_loss": 1.6753425598144531,
      "eval_runtime": 390.1339,
      "eval_samples_per_second": 24.294,
      "eval_steps_per_second": 3.037,
      "step": 31600
    },
    {
      "epoch": 13.34,
      "grad_norm": 3.985402822494507,
      "learning_rate": 2.796254416961131e-05,
      "loss": 0.9951,
      "step": 31700
    },
    {
      "epoch": 13.34,
      "eval_cer": 0.4047969379375073,
      "eval_loss": 1.8533233404159546,
      "eval_runtime": 373.1578,
      "eval_samples_per_second": 25.399,
      "eval_steps_per_second": 3.176,
      "step": 31700
    },
    {
      "epoch": 13.39,
      "grad_norm": 2.613786458969116,
      "learning_rate": 2.7891872791519434e-05,
      "loss": 1.002,
      "step": 31800
    },
    {
      "epoch": 13.39,
      "eval_cer": 0.39678737632474287,
      "eval_loss": 2.017918109893799,
      "eval_runtime": 398.8604,
      "eval_samples_per_second": 23.763,
      "eval_steps_per_second": 2.971,
      "step": 31800
    },
    {
      "epoch": 13.43,
      "grad_norm": 5.32037878036499,
      "learning_rate": 2.7821201413427562e-05,
      "loss": 1.0059,
      "step": 31900
    },
    {
      "epoch": 13.43,
      "eval_cer": 0.40224277501857575,
      "eval_loss": 2.3004064559936523,
      "eval_runtime": 375.8974,
      "eval_samples_per_second": 25.214,
      "eval_steps_per_second": 3.152,
      "step": 31900
    },
    {
      "epoch": 13.47,
      "grad_norm": 1.963639736175537,
      "learning_rate": 2.7750530035335694e-05,
      "loss": 0.9866,
      "step": 32000
    },
    {
      "epoch": 13.47,
      "eval_cer": 0.4031495639591725,
      "eval_loss": 2.7950069904327393,
      "eval_runtime": 392.4567,
      "eval_samples_per_second": 24.15,
      "eval_steps_per_second": 3.019,
      "step": 32000
    },
    {
      "epoch": 13.51,
      "grad_norm": 10.024706840515137,
      "learning_rate": 2.767985865724382e-05,
      "loss": 1.0195,
      "step": 32100
    },
    {
      "epoch": 13.51,
      "eval_cer": 0.409443314692425,
      "eval_loss": 2.4231760501861572,
      "eval_runtime": 370.8807,
      "eval_samples_per_second": 25.555,
      "eval_steps_per_second": 3.195,
      "step": 32100
    },
    {
      "epoch": 13.56,
      "grad_norm": 4.829713344573975,
      "learning_rate": 2.7609187279151943e-05,
      "loss": 1.0437,
      "step": 32200
    },
    {
      "epoch": 13.56,
      "eval_cer": 0.4043276563294357,
      "eval_loss": 1.8479336500167847,
      "eval_runtime": 398.7479,
      "eval_samples_per_second": 23.769,
      "eval_steps_per_second": 2.972,
      "step": 32200
    },
    {
      "epoch": 13.6,
      "grad_norm": 3.0680336952209473,
      "learning_rate": 2.7538515901060075e-05,
      "loss": 1.4222,
      "step": 32300
    },
    {
      "epoch": 13.6,
      "eval_cer": 0.4040881271753158,
      "eval_loss": 2.828200101852417,
      "eval_runtime": 367.3731,
      "eval_samples_per_second": 25.799,
      "eval_steps_per_second": 3.226,
      "step": 32300
    },
    {
      "epoch": 13.64,
      "grad_norm": 2.740971565246582,
      "learning_rate": 2.74678445229682e-05,
      "loss": 1.0443,
      "step": 32400
    },
    {
      "epoch": 13.64,
      "eval_cer": 0.422089476359939,
      "eval_loss": 2.9092776775360107,
      "eval_runtime": 389.1697,
      "eval_samples_per_second": 24.354,
      "eval_steps_per_second": 3.045,
      "step": 32400
    },
    {
      "epoch": 13.68,
      "grad_norm": 7.612214088439941,
      "learning_rate": 2.7397173144876324e-05,
      "loss": 1.0561,
      "step": 32500
    },
    {
      "epoch": 13.68,
      "eval_cer": 0.4040587970748113,
      "eval_loss": 1.9982529878616333,
      "eval_runtime": 371.4689,
      "eval_samples_per_second": 25.515,
      "eval_steps_per_second": 3.19,
      "step": 32500
    },
    {
      "epoch": 13.72,
      "grad_norm": 2.935277223587036,
      "learning_rate": 2.7326501766784452e-05,
      "loss": 1.0508,
      "step": 32600
    },
    {
      "epoch": 13.72,
      "eval_cer": 0.40732665910601856,
      "eval_loss": 2.4883534908294678,
      "eval_runtime": 395.2638,
      "eval_samples_per_second": 23.979,
      "eval_steps_per_second": 2.998,
      "step": 32600
    },
    {
      "epoch": 13.77,
      "grad_norm": 2.7527854442596436,
      "learning_rate": 2.7255830388692584e-05,
      "loss": 1.0314,
      "step": 32700
    },
    {
      "epoch": 13.77,
      "eval_cer": 0.40216944976731456,
      "eval_loss": 2.4499011039733887,
      "eval_runtime": 369.8082,
      "eval_samples_per_second": 25.63,
      "eval_steps_per_second": 3.204,
      "step": 32700
    },
    {
      "epoch": 13.81,
      "grad_norm": 1.3860234022140503,
      "learning_rate": 2.718515901060071e-05,
      "loss": 1.7756,
      "step": 32800
    },
    {
      "epoch": 13.81,
      "eval_cer": 0.41578839310156035,
      "eval_loss": 2.565376043319702,
      "eval_runtime": 390.8183,
      "eval_samples_per_second": 24.252,
      "eval_steps_per_second": 3.032,
      "step": 32800
    },
    {
      "epoch": 13.85,
      "grad_norm": 9.211983680725098,
      "learning_rate": 2.7114487632508833e-05,
      "loss": 1.0164,
      "step": 32900
    },
    {
      "epoch": 13.85,
      "eval_cer": 0.39727621133315083,
      "eval_loss": 2.448713779449463,
      "eval_runtime": 372.957,
      "eval_samples_per_second": 25.413,
      "eval_steps_per_second": 3.177,
      "step": 32900
    },
    {
      "epoch": 13.89,
      "grad_norm": 73.3850326538086,
      "learning_rate": 2.7043816254416965e-05,
      "loss": 1.0211,
      "step": 33000
    },
    {
      "epoch": 13.89,
      "eval_cer": 0.405021802041375,
      "eval_loss": 2.4291627407073975,
      "eval_runtime": 390.776,
      "eval_samples_per_second": 24.254,
      "eval_steps_per_second": 3.032,
      "step": 33000
    },
    {
      "epoch": 13.93,
      "grad_norm": 2.177152395248413,
      "learning_rate": 2.697314487632509e-05,
      "loss": 1.2605,
      "step": 33100
    },
    {
      "epoch": 13.93,
      "eval_cer": 0.40312512220875213,
      "eval_loss": 2.450680732727051,
      "eval_runtime": 369.1148,
      "eval_samples_per_second": 25.678,
      "eval_steps_per_second": 3.21,
      "step": 33100
    },
    {
      "epoch": 13.98,
      "grad_norm": 2.1010019779205322,
      "learning_rate": 2.6902473498233218e-05,
      "loss": 0.9974,
      "step": 33200
    },
    {
      "epoch": 13.98,
      "eval_cer": 0.4032106683352235,
      "eval_loss": 2.416074514389038,
      "eval_runtime": 395.8532,
      "eval_samples_per_second": 23.943,
      "eval_steps_per_second": 2.994,
      "step": 33200
    },
    {
      "epoch": 14.02,
      "grad_norm": 5.9587812423706055,
      "learning_rate": 2.6831802120141342e-05,
      "loss": 0.9724,
      "step": 33300
    },
    {
      "epoch": 14.02,
      "eval_cer": 0.3974839662117242,
      "eval_loss": 2.979396343231201,
      "eval_runtime": 370.8376,
      "eval_samples_per_second": 25.558,
      "eval_steps_per_second": 3.195,
      "step": 33300
    },
    {
      "epoch": 14.06,
      "grad_norm": 2.916482925415039,
      "learning_rate": 2.6761130742049474e-05,
      "loss": 0.9488,
      "step": 33400
    },
    {
      "epoch": 14.06,
      "eval_cer": 0.39579504125767473,
      "eval_loss": 2.687380075454712,
      "eval_runtime": 390.5164,
      "eval_samples_per_second": 24.27,
      "eval_steps_per_second": 3.034,
      "step": 33400
    },
    {
      "epoch": 14.1,
      "grad_norm": 5.404353618621826,
      "learning_rate": 2.66904593639576e-05,
      "loss": 0.9346,
      "step": 33500
    },
    {
      "epoch": 14.1,
      "eval_cer": 0.39129775918032145,
      "eval_loss": 2.3900837898254395,
      "eval_runtime": 372.0348,
      "eval_samples_per_second": 25.476,
      "eval_steps_per_second": 3.185,
      "step": 33500
    },
    {
      "epoch": 14.14,
      "grad_norm": 2.114933729171753,
      "learning_rate": 2.6619787985865723e-05,
      "loss": 0.9445,
      "step": 33600
    },
    {
      "epoch": 14.14,
      "eval_cer": 0.3915470650346095,
      "eval_loss": 2.192039728164673,
      "eval_runtime": 401.3991,
      "eval_samples_per_second": 23.612,
      "eval_steps_per_second": 2.952,
      "step": 33600
    },
    {
      "epoch": 14.19,
      "grad_norm": 22.767732620239258,
      "learning_rate": 2.6549116607773855e-05,
      "loss": 0.9435,
      "step": 33700
    },
    {
      "epoch": 14.19,
      "eval_cer": 0.39616655586406474,
      "eval_loss": 2.0086452960968018,
      "eval_runtime": 374.9441,
      "eval_samples_per_second": 25.278,
      "eval_steps_per_second": 3.16,
      "step": 33700
    },
    {
      "epoch": 14.23,
      "grad_norm": 2.4856455326080322,
      "learning_rate": 2.647844522968198e-05,
      "loss": 0.9387,
      "step": 33800
    },
    {
      "epoch": 14.23,
      "eval_cer": 0.3982514371749247,
      "eval_loss": 1.5301440954208374,
      "eval_runtime": 391.5929,
      "eval_samples_per_second": 24.204,
      "eval_steps_per_second": 3.026,
      "step": 33800
    },
    {
      "epoch": 14.27,
      "grad_norm": 21.76763916015625,
      "learning_rate": 2.6407773851590107e-05,
      "loss": 0.9334,
      "step": 33900
    },
    {
      "epoch": 14.27,
      "eval_cer": 0.3909311329240155,
      "eval_loss": 2.4011144638061523,
      "eval_runtime": 373.6109,
      "eval_samples_per_second": 25.369,
      "eval_steps_per_second": 3.172,
      "step": 33900
    },
    {
      "epoch": 14.31,
      "grad_norm": 6.195644855499268,
      "learning_rate": 2.6337102473498232e-05,
      "loss": 0.9263,
      "step": 34000
    },
    {
      "epoch": 14.31,
      "eval_cer": 0.39136375190645656,
      "eval_loss": 2.636327028274536,
      "eval_runtime": 397.3915,
      "eval_samples_per_second": 23.851,
      "eval_steps_per_second": 2.982,
      "step": 34000
    },
    {
      "epoch": 14.35,
      "grad_norm": 8.33462905883789,
      "learning_rate": 2.6266431095406364e-05,
      "loss": 1.2049,
      "step": 34100
    },
    {
      "epoch": 14.35,
      "eval_cer": 0.38994857455711546,
      "eval_loss": 1.6413642168045044,
      "eval_runtime": 373.2342,
      "eval_samples_per_second": 25.394,
      "eval_steps_per_second": 3.175,
      "step": 34100
    },
    {
      "epoch": 14.4,
      "grad_norm": 2.263068675994873,
      "learning_rate": 2.619575971731449e-05,
      "loss": 0.9314,
      "step": 34200
    },
    {
      "epoch": 14.4,
      "eval_cer": 0.3896846036525752,
      "eval_loss": 1.7732062339782715,
      "eval_runtime": 392.0357,
      "eval_samples_per_second": 24.176,
      "eval_steps_per_second": 3.023,
      "step": 34200
    },
    {
      "epoch": 14.44,
      "grad_norm": 3.9732091426849365,
      "learning_rate": 2.6125088339222613e-05,
      "loss": 0.9228,
      "step": 34300
    },
    {
      "epoch": 14.44,
      "eval_cer": 0.3950935630206093,
      "eval_loss": 1.7944254875183105,
      "eval_runtime": 369.9816,
      "eval_samples_per_second": 25.617,
      "eval_steps_per_second": 3.203,
      "step": 34300
    },
    {
      "epoch": 14.48,
      "grad_norm": 1.9228265285491943,
      "learning_rate": 2.6054416961130745e-05,
      "loss": 0.9234,
      "step": 34400
    },
    {
      "epoch": 14.48,
      "eval_cer": 0.39062316686871845,
      "eval_loss": 1.7525721788406372,
      "eval_runtime": 386.1261,
      "eval_samples_per_second": 24.546,
      "eval_steps_per_second": 3.069,
      "step": 34400
    },
    {
      "epoch": 14.52,
      "grad_norm": 10.801593780517578,
      "learning_rate": 2.598374558303887e-05,
      "loss": 0.9669,
      "step": 34500
    },
    {
      "epoch": 14.52,
      "eval_cer": 0.3913124242305737,
      "eval_loss": 1.5855803489685059,
      "eval_runtime": 376.3532,
      "eval_samples_per_second": 25.184,
      "eval_steps_per_second": 3.149,
      "step": 34500
    },
    {
      "epoch": 14.57,
      "grad_norm": 2.1704392433166504,
      "learning_rate": 2.5913074204946997e-05,
      "loss": 0.9261,
      "step": 34600
    },
    {
      "epoch": 14.57,
      "eval_cer": 0.38692268585507017,
      "eval_loss": 1.6806480884552002,
      "eval_runtime": 393.3479,
      "eval_samples_per_second": 24.096,
      "eval_steps_per_second": 3.013,
      "step": 34600
    },
    {
      "epoch": 14.61,
      "grad_norm": 4.049818515777588,
      "learning_rate": 2.5842402826855122e-05,
      "loss": 0.9017,
      "step": 34700
    },
    {
      "epoch": 14.61,
      "eval_cer": 0.3871744358844003,
      "eval_loss": 1.6744087934494019,
      "eval_runtime": 365.5955,
      "eval_samples_per_second": 25.925,
      "eval_steps_per_second": 3.241,
      "step": 34700
    },
    {
      "epoch": 14.65,
      "grad_norm": 3.3934333324432373,
      "learning_rate": 2.5771731448763254e-05,
      "loss": 0.9375,
      "step": 34800
    },
    {
      "epoch": 14.65,
      "eval_cer": 0.3903005357631692,
      "eval_loss": 1.944000005722046,
      "eval_runtime": 404.5333,
      "eval_samples_per_second": 23.429,
      "eval_steps_per_second": 2.929,
      "step": 34800
    },
    {
      "epoch": 14.69,
      "grad_norm": 2.0570108890533447,
      "learning_rate": 2.5701060070671378e-05,
      "loss": 0.9158,
      "step": 34900
    },
    {
      "epoch": 14.69,
      "eval_cer": 0.3864998435727973,
      "eval_loss": 1.8147820234298706,
      "eval_runtime": 366.6552,
      "eval_samples_per_second": 25.85,
      "eval_steps_per_second": 3.232,
      "step": 34900
    },
    {
      "epoch": 14.73,
      "grad_norm": 2.4123330116271973,
      "learning_rate": 2.5631095406360423e-05,
      "loss": 1.0998,
      "step": 35000
    },
    {
      "epoch": 14.73,
      "eval_cer": 0.3869789018810371,
      "eval_loss": 1.8494433164596558,
      "eval_runtime": 395.7827,
      "eval_samples_per_second": 23.947,
      "eval_steps_per_second": 2.994,
      "step": 35000
    },
    {
      "epoch": 14.78,
      "grad_norm": 2.7764666080474854,
      "learning_rate": 2.5560424028268555e-05,
      "loss": 0.9243,
      "step": 35100
    },
    {
      "epoch": 14.78,
      "eval_cer": 0.38634097219506475,
      "eval_loss": 1.8553264141082764,
      "eval_runtime": 368.5207,
      "eval_samples_per_second": 25.719,
      "eval_steps_per_second": 3.216,
      "step": 35100
    },
    {
      "epoch": 14.82,
      "grad_norm": 37.1422119140625,
      "learning_rate": 2.548975265017668e-05,
      "loss": 0.9169,
      "step": 35200
    },
    {
      "epoch": 14.82,
      "eval_cer": 0.3885431739079426,
      "eval_loss": 2.9854750633239746,
      "eval_runtime": 393.2259,
      "eval_samples_per_second": 24.103,
      "eval_steps_per_second": 3.014,
      "step": 35200
    },
    {
      "epoch": 14.86,
      "grad_norm": 2.0946056842803955,
      "learning_rate": 2.5419081272084804e-05,
      "loss": 0.9188,
      "step": 35300
    },
    {
      "epoch": 14.86,
      "eval_cer": 0.38827187047827616,
      "eval_loss": 2.4621119499206543,
      "eval_runtime": 369.124,
      "eval_samples_per_second": 25.677,
      "eval_steps_per_second": 3.21,
      "step": 35300
    },
    {
      "epoch": 14.9,
      "grad_norm": 2.0357742309570312,
      "learning_rate": 2.5348409893992936e-05,
      "loss": 0.9183,
      "step": 35400
    },
    {
      "epoch": 14.9,
      "eval_cer": 0.38374036995033434,
      "eval_loss": 1.656967043876648,
      "eval_runtime": 397.837,
      "eval_samples_per_second": 23.824,
      "eval_steps_per_second": 2.979,
      "step": 35400
    },
    {
      "epoch": 14.94,
      "grad_norm": 2.4356577396392822,
      "learning_rate": 2.527773851590106e-05,
      "loss": 0.9208,
      "step": 35500
    },
    {
      "epoch": 14.94,
      "eval_cer": 0.3846764889914356,
      "eval_loss": 1.6596330404281616,
      "eval_runtime": 368.5045,
      "eval_samples_per_second": 25.72,
      "eval_steps_per_second": 3.216,
      "step": 35500
    },
    {
      "epoch": 14.99,
      "grad_norm": 3.9734749794006348,
      "learning_rate": 2.520706713780919e-05,
      "loss": 0.9122,
      "step": 35600
    },
    {
      "epoch": 14.99,
      "eval_cer": 0.3868713581791874,
      "eval_loss": 2.068138360977173,
      "eval_runtime": 401.8562,
      "eval_samples_per_second": 23.586,
      "eval_steps_per_second": 2.949,
      "step": 35600
    },
    {
      "epoch": 15.03,
      "grad_norm": 1.6432957649230957,
      "learning_rate": 2.513639575971732e-05,
      "loss": 0.872,
      "step": 35700
    },
    {
      "epoch": 15.03,
      "eval_cer": 0.3834226271948692,
      "eval_loss": 2.0865533351898193,
      "eval_runtime": 372.0301,
      "eval_samples_per_second": 25.476,
      "eval_steps_per_second": 3.185,
      "step": 35700
    }
  ],
  "logging_steps": 100,
  "max_steps": 71250,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 30,
  "save_steps": 100,
  "total_flos": 3.9129036959681564e+20,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}