{
  "best_global_step": 24000,
  "best_metric": 85.56038719629122,
  "best_model_checkpoint": "checkpoints_7B_lora_translated/en-chv-final/checkpoint-24000",
  "epoch": 0.38206826286296486,
  "eval_steps": 1000,
  "global_step": 33000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.00011577826147362571,
      "grad_norm": 2.2461981773376465,
      "learning_rate": 1.389317690645261e-07,
      "loss": 3.2022,
      "step": 10
    },
    {
      "epoch": 0.00023155652294725142,
      "grad_norm": 1.8676271438598633,
      "learning_rate": 2.9330040135844395e-07,
      "loss": 3.0958,
      "step": 20
    },
    {
      "epoch": 0.00034733478442087714,
      "grad_norm": 2.413166046142578,
      "learning_rate": 4.476690336523618e-07,
      "loss": 3.2048,
      "step": 30
    },
    {
      "epoch": 0.00046311304589450285,
      "grad_norm": 2.3978662490844727,
      "learning_rate": 6.020376659462798e-07,
      "loss": 3.1917,
      "step": 40
    },
    {
      "epoch": 0.0005788913073681286,
      "grad_norm": 2.6323177814483643,
      "learning_rate": 7.564062982401976e-07,
      "loss": 3.1719,
      "step": 50
    },
    {
      "epoch": 0.0006946695688417543,
      "grad_norm": 1.8137117624282837,
      "learning_rate": 9.107749305341156e-07,
      "loss": 3.2232,
      "step": 60
    },
    {
      "epoch": 0.00081044783031538,
      "grad_norm": 2.2883598804473877,
      "learning_rate": 1.0651435628280334e-06,
      "loss": 3.1677,
      "step": 70
    },
    {
      "epoch": 0.0009262260917890057,
      "grad_norm": 2.310643434524536,
      "learning_rate": 1.2195121951219514e-06,
      "loss": 3.1682,
      "step": 80
    },
    {
      "epoch": 0.0010420043532626314,
      "grad_norm": 2.4111878871917725,
      "learning_rate": 1.3738808274158691e-06,
      "loss": 3.1628,
      "step": 90
    },
    {
      "epoch": 0.0011577826147362571,
      "grad_norm": 2.883052349090576,
      "learning_rate": 1.5282494597097871e-06,
      "loss": 3.1689,
      "step": 100
    },
    {
      "epoch": 0.0012735608762098828,
      "grad_norm": 2.283472776412964,
      "learning_rate": 1.6826180920037051e-06,
      "loss": 3.1567,
      "step": 110
    },
    {
      "epoch": 0.0013893391376835085,
      "grad_norm": 2.956798791885376,
      "learning_rate": 1.8369867242976229e-06,
      "loss": 3.0266,
      "step": 120
    },
    {
      "epoch": 0.0015051173991571343,
      "grad_norm": 2.6263251304626465,
      "learning_rate": 1.9913553565915407e-06,
      "loss": 2.913,
      "step": 130
    },
    {
      "epoch": 0.00162089566063076,
      "grad_norm": 2.811680316925049,
      "learning_rate": 2.1457239888854584e-06,
      "loss": 2.859,
      "step": 140
    },
    {
      "epoch": 0.0017366739221043857,
      "grad_norm": 2.3774526119232178,
      "learning_rate": 2.3000926211793766e-06,
      "loss": 2.8647,
      "step": 150
    },
    {
      "epoch": 0.0018524521835780114,
      "grad_norm": 2.3617496490478516,
      "learning_rate": 2.4544612534732944e-06,
      "loss": 2.7865,
      "step": 160
    },
    {
      "epoch": 0.001968230445051637,
      "grad_norm": 2.1448798179626465,
      "learning_rate": 2.608829885767212e-06,
      "loss": 2.6098,
      "step": 170
    },
    {
      "epoch": 0.002084008706525263,
      "grad_norm": 1.6232613325119019,
      "learning_rate": 2.7631985180611304e-06,
      "loss": 2.5649,
      "step": 180
    },
    {
      "epoch": 0.0021997869679988885,
      "grad_norm": 1.9341398477554321,
      "learning_rate": 2.917567150355048e-06,
      "loss": 2.4696,
      "step": 190
    },
    {
      "epoch": 0.0023155652294725142,
      "grad_norm": 1.791678547859192,
      "learning_rate": 3.071935782648966e-06,
      "loss": 2.4328,
      "step": 200
    },
    {
      "epoch": 0.00243134349094614,
      "grad_norm": 1.7509756088256836,
      "learning_rate": 3.2263044149428837e-06,
      "loss": 2.2921,
      "step": 210
    },
    {
      "epoch": 0.0025471217524197657,
      "grad_norm": 1.1987025737762451,
      "learning_rate": 3.3806730472368014e-06,
      "loss": 2.2021,
      "step": 220
    },
    {
      "epoch": 0.0026629000138933914,
      "grad_norm": 1.2671940326690674,
      "learning_rate": 3.5350416795307192e-06,
      "loss": 2.0922,
      "step": 230
    },
    {
      "epoch": 0.002778678275367017,
      "grad_norm": 1.0479227304458618,
      "learning_rate": 3.689410311824638e-06,
      "loss": 2.1015,
      "step": 240
    },
    {
      "epoch": 0.002894456536840643,
      "grad_norm": 0.9470033645629883,
      "learning_rate": 3.843778944118555e-06,
      "loss": 2.0249,
      "step": 250
    },
    {
      "epoch": 0.0030102347983142685,
      "grad_norm": 0.8377466797828674,
      "learning_rate": 3.998147576412473e-06,
      "loss": 1.9202,
      "step": 260
    },
    {
      "epoch": 0.0031260130597878942,
      "grad_norm": 0.7731795310974121,
      "learning_rate": 4.152516208706391e-06,
      "loss": 1.8559,
      "step": 270
    },
    {
      "epoch": 0.00324179132126152,
      "grad_norm": 0.776683509349823,
      "learning_rate": 4.306884841000309e-06,
      "loss": 1.8088,
      "step": 280
    },
    {
      "epoch": 0.0033575695827351456,
      "grad_norm": 0.6955391764640808,
      "learning_rate": 4.461253473294226e-06,
      "loss": 1.7343,
      "step": 290
    },
    {
      "epoch": 0.0034733478442087714,
      "grad_norm": 0.6504161953926086,
      "learning_rate": 4.6156221055881445e-06,
      "loss": 1.6248,
      "step": 300
    },
    {
      "epoch": 0.003589126105682397,
      "grad_norm": 0.7929391860961914,
      "learning_rate": 4.769990737882063e-06,
      "loss": 1.711,
      "step": 310
    },
    {
      "epoch": 0.0037049043671560228,
      "grad_norm": 0.8294330835342407,
      "learning_rate": 4.924359370175981e-06,
      "loss": 1.6191,
      "step": 320
    },
    {
      "epoch": 0.0038206826286296485,
      "grad_norm": 0.6876463890075684,
      "learning_rate": 5.078728002469898e-06,
      "loss": 1.6912,
      "step": 330
    },
    {
      "epoch": 0.003936460890103274,
      "grad_norm": 0.6395148038864136,
      "learning_rate": 5.233096634763816e-06,
      "loss": 1.5623,
      "step": 340
    },
    {
      "epoch": 0.0040522391515769,
      "grad_norm": 0.6318981647491455,
      "learning_rate": 5.387465267057734e-06,
      "loss": 1.5632,
      "step": 350
    },
    {
      "epoch": 0.004168017413050526,
      "grad_norm": 0.659462571144104,
      "learning_rate": 5.541833899351652e-06,
      "loss": 1.6424,
      "step": 360
    },
    {
      "epoch": 0.004283795674524152,
      "grad_norm": 0.6242923736572266,
      "learning_rate": 5.69620253164557e-06,
      "loss": 1.5602,
      "step": 370
    },
    {
      "epoch": 0.004399573935997777,
      "grad_norm": 0.7650619745254517,
      "learning_rate": 5.8505711639394875e-06,
      "loss": 1.612,
      "step": 380
    },
    {
      "epoch": 0.004515352197471403,
      "grad_norm": 0.634069561958313,
      "learning_rate": 6.004939796233406e-06,
      "loss": 1.4928,
      "step": 390
    },
    {
      "epoch": 0.0046311304589450285,
      "grad_norm": 0.7291055917739868,
      "learning_rate": 6.159308428527324e-06,
      "loss": 1.5118,
      "step": 400
    },
    {
      "epoch": 0.004746908720418655,
      "grad_norm": 0.7340362071990967,
      "learning_rate": 6.313677060821241e-06,
      "loss": 1.5039,
      "step": 410
    },
    {
      "epoch": 0.00486268698189228,
      "grad_norm": 0.7519592642784119,
      "learning_rate": 6.4680456931151586e-06,
      "loss": 1.5023,
      "step": 420
    },
    {
      "epoch": 0.004978465243365906,
      "grad_norm": 0.875359296798706,
      "learning_rate": 6.622414325409078e-06,
      "loss": 1.5312,
      "step": 430
    },
    {
      "epoch": 0.005094243504839531,
      "grad_norm": 0.7627996206283569,
      "learning_rate": 6.776782957702996e-06,
      "loss": 1.4769,
      "step": 440
    },
    {
      "epoch": 0.0052100217663131575,
      "grad_norm": 0.9831496477127075,
      "learning_rate": 6.931151589996913e-06,
      "loss": 1.5039,
      "step": 450
    },
    {
      "epoch": 0.005325800027786783,
      "grad_norm": 0.6336537003517151,
      "learning_rate": 7.085520222290831e-06,
      "loss": 1.4107,
      "step": 460
    },
    {
      "epoch": 0.005441578289260409,
      "grad_norm": 0.9347900152206421,
      "learning_rate": 7.239888854584749e-06,
      "loss": 1.4793,
      "step": 470
    },
    {
      "epoch": 0.005557356550734034,
      "grad_norm": 0.9351417422294617,
      "learning_rate": 7.394257486878667e-06,
      "loss": 1.4556,
      "step": 480
    },
    {
      "epoch": 0.00567313481220766,
      "grad_norm": 1.0249700546264648,
      "learning_rate": 7.548626119172584e-06,
      "loss": 1.5199,
      "step": 490
    },
    {
      "epoch": 0.005788913073681286,
      "grad_norm": 1.002763032913208,
      "learning_rate": 7.702994751466502e-06,
      "loss": 1.4544,
      "step": 500
    },
    {
      "epoch": 0.005904691335154912,
      "grad_norm": 0.8069408535957336,
      "learning_rate": 7.85736338376042e-06,
      "loss": 1.4169,
      "step": 510
    },
    {
      "epoch": 0.006020469596628537,
      "grad_norm": 0.9878456592559814,
      "learning_rate": 8.011732016054338e-06,
      "loss": 1.4122,
      "step": 520
    },
    {
      "epoch": 0.006136247858102163,
      "grad_norm": 1.0431718826293945,
      "learning_rate": 8.166100648348255e-06,
      "loss": 1.3979,
      "step": 530
    },
    {
      "epoch": 0.0062520261195757884,
      "grad_norm": 0.853024423122406,
      "learning_rate": 8.320469280642174e-06,
      "loss": 1.429,
      "step": 540
    },
    {
      "epoch": 0.006367804381049415,
      "grad_norm": 1.0999716520309448,
      "learning_rate": 8.474837912936092e-06,
      "loss": 1.3591,
      "step": 550
    },
    {
      "epoch": 0.00648358264252304,
      "grad_norm": 1.1147035360336304,
      "learning_rate": 8.629206545230009e-06,
      "loss": 1.3853,
      "step": 560
    },
    {
      "epoch": 0.006599360903996666,
      "grad_norm": 1.2002438306808472,
      "learning_rate": 8.783575177523926e-06,
      "loss": 1.3646,
      "step": 570
    },
    {
      "epoch": 0.006715139165470291,
      "grad_norm": 0.8041489720344543,
      "learning_rate": 8.937943809817845e-06,
      "loss": 1.3767,
      "step": 580
    },
    {
      "epoch": 0.0068309174269439174,
      "grad_norm": 1.0359758138656616,
      "learning_rate": 9.092312442111765e-06,
      "loss": 1.3744,
      "step": 590
    },
    {
      "epoch": 0.006946695688417543,
      "grad_norm": 1.066063642501831,
      "learning_rate": 9.246681074405682e-06,
      "loss": 1.3475,
      "step": 600
    },
    {
      "epoch": 0.007062473949891169,
      "grad_norm": 1.219103217124939,
      "learning_rate": 9.4010497066996e-06,
      "loss": 1.4439,
      "step": 610
    },
    {
      "epoch": 0.007178252211364794,
      "grad_norm": 1.5211087465286255,
      "learning_rate": 9.555418338993517e-06,
      "loss": 1.3859,
      "step": 620
    },
    {
      "epoch": 0.00729403047283842,
      "grad_norm": 1.2016996145248413,
      "learning_rate": 9.709786971287436e-06,
      "loss": 1.349,
      "step": 630
    },
    {
      "epoch": 0.0074098087343120456,
      "grad_norm": 1.355387806892395,
      "learning_rate": 9.864155603581353e-06,
      "loss": 1.3666,
      "step": 640
    },
    {
      "epoch": 0.007525586995785672,
      "grad_norm": 1.2157585620880127,
      "learning_rate": 1.001852423587527e-05,
      "loss": 1.3589,
      "step": 650
    },
    {
      "epoch": 0.007641365257259297,
      "grad_norm": 1.3706347942352295,
      "learning_rate": 1.0172892868169188e-05,
      "loss": 1.3604,
      "step": 660
    },
    {
      "epoch": 0.007757143518732923,
      "grad_norm": 0.9463343024253845,
      "learning_rate": 1.0327261500463107e-05,
      "loss": 1.32,
      "step": 670
    },
    {
      "epoch": 0.007872921780206548,
      "grad_norm": 1.4948500394821167,
      "learning_rate": 1.0481630132757024e-05,
      "loss": 1.307,
      "step": 680
    },
    {
      "epoch": 0.007988700041680175,
      "grad_norm": 1.2556153535842896,
      "learning_rate": 1.0635998765050941e-05,
      "loss": 1.3867,
      "step": 690
    },
    {
      "epoch": 0.0081044783031538,
      "grad_norm": 1.2945194244384766,
      "learning_rate": 1.079036739734486e-05,
      "loss": 1.311,
      "step": 700
    },
    {
      "epoch": 0.008220256564627425,
      "grad_norm": 1.147867202758789,
      "learning_rate": 1.0944736029638778e-05,
      "loss": 1.2464,
      "step": 710
    },
    {
      "epoch": 0.008336034826101051,
      "grad_norm": 1.147763967514038,
      "learning_rate": 1.1099104661932695e-05,
      "loss": 1.2917,
      "step": 720
    },
    {
      "epoch": 0.008451813087574677,
      "grad_norm": 1.3554691076278687,
      "learning_rate": 1.1253473294226612e-05,
      "loss": 1.333,
      "step": 730
    },
    {
      "epoch": 0.008567591349048304,
      "grad_norm": 1.6088576316833496,
      "learning_rate": 1.1407841926520532e-05,
      "loss": 1.3111,
      "step": 740
    },
    {
      "epoch": 0.008683369610521928,
      "grad_norm": 1.7919034957885742,
      "learning_rate": 1.156221055881445e-05,
      "loss": 1.2456,
      "step": 750
    },
    {
      "epoch": 0.008799147871995554,
      "grad_norm": 1.6351501941680908,
      "learning_rate": 1.1716579191108368e-05,
      "loss": 1.2741,
      "step": 760
    },
    {
      "epoch": 0.00891492613346918,
      "grad_norm": 1.5512449741363525,
      "learning_rate": 1.1870947823402285e-05,
      "loss": 1.2488,
      "step": 770
    },
    {
      "epoch": 0.009030704394942806,
      "grad_norm": 1.9897394180297852,
      "learning_rate": 1.2025316455696203e-05,
      "loss": 1.2266,
      "step": 780
    },
    {
      "epoch": 0.00914648265641643,
      "grad_norm": 1.7279316186904907,
      "learning_rate": 1.2179685087990122e-05,
      "loss": 1.2552,
      "step": 790
    },
    {
      "epoch": 0.009262260917890057,
      "grad_norm": 1.7041466236114502,
      "learning_rate": 1.2334053720284039e-05,
      "loss": 1.288,
      "step": 800
    },
    {
      "epoch": 0.009378039179363683,
      "grad_norm": 1.6459987163543701,
      "learning_rate": 1.2488422352577956e-05,
      "loss": 1.3163,
      "step": 810
    },
    {
      "epoch": 0.00949381744083731,
      "grad_norm": 1.9403908252716064,
      "learning_rate": 1.2642790984871874e-05,
      "loss": 1.2484,
      "step": 820
    },
    {
      "epoch": 0.009609595702310934,
      "grad_norm": 2.0014758110046387,
      "learning_rate": 1.2797159617165794e-05,
      "loss": 1.2432,
      "step": 830
    },
    {
      "epoch": 0.00972537396378456,
      "grad_norm": 1.9093619585037231,
      "learning_rate": 1.295152824945971e-05,
      "loss": 1.256,
      "step": 840
    },
    {
      "epoch": 0.009841152225258186,
      "grad_norm": 1.548734188079834,
      "learning_rate": 1.3105896881753629e-05,
      "loss": 1.2634,
      "step": 850
    },
    {
      "epoch": 0.009956930486731812,
      "grad_norm": 2.2801010608673096,
      "learning_rate": 1.3260265514047546e-05,
      "loss": 1.2874,
      "step": 860
    },
    {
      "epoch": 0.010072708748205437,
      "grad_norm": 2.1263022422790527,
      "learning_rate": 1.3414634146341466e-05,
      "loss": 1.2556,
      "step": 870
    },
    {
      "epoch": 0.010188487009679063,
      "grad_norm": 1.7874159812927246,
      "learning_rate": 1.3569002778635381e-05,
      "loss": 1.2273,
      "step": 880
    },
    {
      "epoch": 0.010304265271152689,
      "grad_norm": 1.8987889289855957,
      "learning_rate": 1.37233714109293e-05,
      "loss": 1.241,
      "step": 890
    },
    {
      "epoch": 0.010420043532626315,
      "grad_norm": 1.6198410987854004,
      "learning_rate": 1.3877740043223218e-05,
      "loss": 1.2792,
      "step": 900
    },
    {
      "epoch": 0.01053582179409994,
      "grad_norm": 1.868829369544983,
      "learning_rate": 1.4032108675517137e-05,
      "loss": 1.258,
      "step": 910
    },
    {
      "epoch": 0.010651600055573565,
      "grad_norm": 1.9847484827041626,
      "learning_rate": 1.4186477307811052e-05,
      "loss": 1.2298,
      "step": 920
    },
    {
      "epoch": 0.010767378317047192,
      "grad_norm": 2.0772342681884766,
      "learning_rate": 1.4340845940104971e-05,
      "loss": 1.215,
      "step": 930
    },
    {
      "epoch": 0.010883156578520818,
      "grad_norm": 2.3518121242523193,
      "learning_rate": 1.4495214572398889e-05,
      "loss": 1.217,
      "step": 940
    },
    {
      "epoch": 0.010998934839994442,
      "grad_norm": 2.6899712085723877,
      "learning_rate": 1.4649583204692808e-05,
      "loss": 1.2741,
      "step": 950
    },
    {
      "epoch": 0.011114713101468068,
      "grad_norm": 2.2902116775512695,
      "learning_rate": 1.4803951836986727e-05,
      "loss": 1.273,
      "step": 960
    },
    {
      "epoch": 0.011230491362941694,
      "grad_norm": 1.927801489830017,
      "learning_rate": 1.4958320469280642e-05,
      "loss": 1.2444,
      "step": 970
    },
    {
      "epoch": 0.01134626962441532,
      "grad_norm": 2.3730885982513428,
      "learning_rate": 1.5112689101574561e-05,
      "loss": 1.1811,
      "step": 980
    },
    {
      "epoch": 0.011462047885888945,
      "grad_norm": 2.474961280822754,
      "learning_rate": 1.526705773386848e-05,
      "loss": 1.1951,
      "step": 990
    },
    {
      "epoch": 0.011577826147362571,
      "grad_norm": 2.8506340980529785,
      "learning_rate": 1.5421426366162398e-05,
      "loss": 1.1713,
      "step": 1000
    },
    {
      "epoch": 0.011577826147362571,
      "eval_chrf": 70.673277017744,
      "eval_loss": 1.8353379964828491,
      "eval_runtime": 373.8216,
      "eval_samples_per_second": 0.268,
      "eval_steps_per_second": 0.011,
      "step": 1000
    },
    {
      "epoch": 0.011693604408836197,
      "grad_norm": 2.152484893798828,
      "learning_rate": 1.5575794998456313e-05,
      "loss": 1.1834,
      "step": 1010
    },
    {
      "epoch": 0.011809382670309823,
      "grad_norm": 1.8658829927444458,
      "learning_rate": 1.5730163630750233e-05,
      "loss": 1.1938,
      "step": 1020
    },
    {
      "epoch": 0.011925160931783448,
      "grad_norm": 2.368344783782959,
      "learning_rate": 1.5884532263044148e-05,
      "loss": 1.1408,
      "step": 1030
    },
    {
      "epoch": 0.012040939193257074,
      "grad_norm": 2.0136966705322266,
      "learning_rate": 1.6038900895338067e-05,
      "loss": 1.187,
      "step": 1040
    },
    {
      "epoch": 0.0121567174547307,
      "grad_norm": 2.2445461750030518,
      "learning_rate": 1.6193269527631986e-05,
      "loss": 1.2375,
      "step": 1050
    },
    {
      "epoch": 0.012272495716204326,
      "grad_norm": 2.9442145824432373,
      "learning_rate": 1.6347638159925905e-05,
      "loss": 1.207,
      "step": 1060
    },
    {
      "epoch": 0.01238827397767795,
      "grad_norm": 2.0814785957336426,
      "learning_rate": 1.650200679221982e-05,
      "loss": 1.1357,
      "step": 1070
    },
    {
      "epoch": 0.012504052239151577,
      "grad_norm": 1.9440767765045166,
      "learning_rate": 1.665637542451374e-05,
      "loss": 1.2145,
      "step": 1080
    },
    {
      "epoch": 0.012619830500625203,
      "grad_norm": 2.155322551727295,
      "learning_rate": 1.6810744056807656e-05,
      "loss": 1.234,
      "step": 1090
    },
    {
      "epoch": 0.01273560876209883,
      "grad_norm": 2.5245254039764404,
      "learning_rate": 1.6965112689101575e-05,
      "loss": 1.1983,
      "step": 1100
    },
    {
      "epoch": 0.012851387023572454,
      "grad_norm": 2.82124662399292,
      "learning_rate": 1.711948132139549e-05,
      "loss": 1.1659,
      "step": 1110
    },
    {
      "epoch": 0.01296716528504608,
      "grad_norm": 2.253861665725708,
      "learning_rate": 1.727384995368941e-05,
      "loss": 1.1145,
      "step": 1120
    },
    {
      "epoch": 0.013082943546519706,
      "grad_norm": 2.3893697261810303,
      "learning_rate": 1.7428218585983332e-05,
      "loss": 1.2086,
      "step": 1130
    },
    {
      "epoch": 0.013198721807993332,
      "grad_norm": 2.7133824825286865,
      "learning_rate": 1.7582587218277247e-05,
      "loss": 1.1192,
      "step": 1140
    },
    {
      "epoch": 0.013314500069466956,
      "grad_norm": 2.9232800006866455,
      "learning_rate": 1.7736955850571167e-05,
      "loss": 1.1828,
      "step": 1150
    },
    {
      "epoch": 0.013430278330940583,
      "grad_norm": 2.678086280822754,
      "learning_rate": 1.7891324482865082e-05,
      "loss": 1.1349,
      "step": 1160
    },
    {
      "epoch": 0.013546056592414209,
      "grad_norm": 2.729440689086914,
      "learning_rate": 1.8045693115159e-05,
      "loss": 1.1445,
      "step": 1170
    },
    {
      "epoch": 0.013661834853887835,
      "grad_norm": 2.464496374130249,
      "learning_rate": 1.8200061747452917e-05,
      "loss": 1.1899,
      "step": 1180
    },
    {
      "epoch": 0.01377761311536146,
      "grad_norm": 2.538877487182617,
      "learning_rate": 1.8354430379746836e-05,
      "loss": 1.1528,
      "step": 1190
    },
    {
      "epoch": 0.013893391376835085,
      "grad_norm": 1.8316330909729004,
      "learning_rate": 1.8508799012040755e-05,
      "loss": 1.1342,
      "step": 1200
    },
    {
      "epoch": 0.014009169638308712,
      "grad_norm": 2.8184266090393066,
      "learning_rate": 1.8663167644334674e-05,
      "loss": 1.1521,
      "step": 1210
    },
    {
      "epoch": 0.014124947899782338,
      "grad_norm": 2.5736935138702393,
      "learning_rate": 1.881753627662859e-05,
      "loss": 1.0831,
      "step": 1220
    },
    {
      "epoch": 0.014240726161255962,
      "grad_norm": 2.4134764671325684,
      "learning_rate": 1.897190490892251e-05,
      "loss": 1.1569,
      "step": 1230
    },
    {
      "epoch": 0.014356504422729588,
      "grad_norm": 2.6627583503723145,
      "learning_rate": 1.9126273541216424e-05,
      "loss": 1.168,
      "step": 1240
    },
    {
      "epoch": 0.014472282684203214,
      "grad_norm": 2.3337690830230713,
      "learning_rate": 1.9280642173510343e-05,
      "loss": 1.0915,
      "step": 1250
    },
    {
      "epoch": 0.01458806094567684,
      "grad_norm": 2.935228109359741,
      "learning_rate": 1.943501080580426e-05,
      "loss": 1.154,
      "step": 1260
    },
    {
      "epoch": 0.014703839207150465,
      "grad_norm": 2.7111053466796875,
      "learning_rate": 1.9589379438098178e-05,
      "loss": 1.0839,
      "step": 1270
    },
    {
      "epoch": 0.014819617468624091,
      "grad_norm": 1.9878060817718506,
      "learning_rate": 1.9743748070392097e-05,
      "loss": 1.1184,
      "step": 1280
    },
    {
      "epoch": 0.014935395730097717,
      "grad_norm": 2.0981931686401367,
      "learning_rate": 1.9898116702686016e-05,
      "loss": 1.1044,
      "step": 1290
    },
    {
      "epoch": 0.015051173991571343,
      "grad_norm": 2.7092490196228027,
      "learning_rate": 2.0052485334979935e-05,
      "loss": 1.0753,
      "step": 1300
    },
    {
      "epoch": 0.015166952253044968,
      "grad_norm": 2.7931082248687744,
      "learning_rate": 2.020685396727385e-05,
      "loss": 1.1314,
      "step": 1310
    },
    {
      "epoch": 0.015282730514518594,
      "grad_norm": 2.5221099853515625,
      "learning_rate": 2.036122259956777e-05,
      "loss": 1.0728,
      "step": 1320
    },
    {
      "epoch": 0.01539850877599222,
      "grad_norm": 2.549616813659668,
      "learning_rate": 2.0515591231861686e-05,
      "loss": 1.1195,
      "step": 1330
    },
    {
      "epoch": 0.015514287037465846,
      "grad_norm": 3.0201480388641357,
      "learning_rate": 2.0669959864155605e-05,
      "loss": 1.0502,
      "step": 1340
    },
    {
      "epoch": 0.015630065298939472,
      "grad_norm": 3.0371510982513428,
      "learning_rate": 2.082432849644952e-05,
      "loss": 1.0779,
      "step": 1350
    },
    {
      "epoch": 0.015745843560413097,
      "grad_norm": 3.3051536083221436,
      "learning_rate": 2.097869712874344e-05,
      "loss": 1.0712,
      "step": 1360
    },
    {
      "epoch": 0.01586162182188672,
      "grad_norm": 3.4807722568511963,
      "learning_rate": 2.113306576103736e-05,
      "loss": 1.169,
      "step": 1370
    },
    {
      "epoch": 0.01597740008336035,
      "grad_norm": 2.8707168102264404,
      "learning_rate": 2.1287434393331277e-05,
      "loss": 1.01,
      "step": 1380
    },
    {
      "epoch": 0.016093178344833974,
      "grad_norm": 3.265197515487671,
      "learning_rate": 2.1441803025625193e-05,
      "loss": 1.1368,
      "step": 1390
    },
    {
      "epoch": 0.0162089566063076,
      "grad_norm": 2.2836358547210693,
      "learning_rate": 2.1596171657919112e-05,
      "loss": 1.0972,
      "step": 1400
    },
    {
      "epoch": 0.016324734867781226,
      "grad_norm": 3.0558969974517822,
      "learning_rate": 2.1750540290213028e-05,
      "loss": 1.0455,
      "step": 1410
    },
    {
      "epoch": 0.01644051312925485,
      "grad_norm": 2.7946484088897705,
      "learning_rate": 2.1904908922506947e-05,
      "loss": 1.1051,
      "step": 1420
    },
    {
      "epoch": 0.016556291390728478,
      "grad_norm": 2.8424155712127686,
      "learning_rate": 2.2059277554800866e-05,
      "loss": 1.0662,
      "step": 1430
    },
    {
      "epoch": 0.016672069652202103,
      "grad_norm": 2.8506076335906982,
      "learning_rate": 2.2213646187094785e-05,
      "loss": 1.0649,
      "step": 1440
    },
    {
      "epoch": 0.016787847913675727,
      "grad_norm": 3.030529260635376,
      "learning_rate": 2.2368014819388704e-05,
      "loss": 1.0523,
      "step": 1450
    },
    {
      "epoch": 0.016903626175149355,
      "grad_norm": 2.6668131351470947,
      "learning_rate": 2.252238345168262e-05,
      "loss": 1.0682,
      "step": 1460
    },
    {
      "epoch": 0.01701940443662298,
      "grad_norm": 2.3751957416534424,
      "learning_rate": 2.267675208397654e-05,
      "loss": 1.0783,
      "step": 1470
    },
    {
      "epoch": 0.017135182698096607,
      "grad_norm": 2.4701077938079834,
      "learning_rate": 2.2831120716270454e-05,
      "loss": 1.1051,
      "step": 1480
    },
    {
      "epoch": 0.01725096095957023,
      "grad_norm": 3.0305330753326416,
      "learning_rate": 2.2985489348564373e-05,
      "loss": 1.1174,
      "step": 1490
    },
    {
      "epoch": 0.017366739221043856,
      "grad_norm": 2.4100780487060547,
      "learning_rate": 2.313985798085829e-05,
      "loss": 1.0384,
      "step": 1500
    },
    {
      "epoch": 0.017482517482517484,
      "grad_norm": 3.0715460777282715,
      "learning_rate": 2.3294226613152208e-05,
      "loss": 1.0816,
      "step": 1510
    },
    {
      "epoch": 0.017598295743991108,
      "grad_norm": 3.546391248703003,
      "learning_rate": 2.3448595245446127e-05,
      "loss": 1.0105,
      "step": 1520
    },
    {
      "epoch": 0.017714074005464733,
      "grad_norm": 3.3128504753112793,
      "learning_rate": 2.3602963877740046e-05,
      "loss": 1.0383,
      "step": 1530
    },
    {
      "epoch": 0.01782985226693836,
      "grad_norm": 2.636296510696411,
      "learning_rate": 2.3757332510033962e-05,
      "loss": 1.0495,
      "step": 1540
    },
    {
      "epoch": 0.017945630528411985,
      "grad_norm": 3.293440580368042,
      "learning_rate": 2.391170114232788e-05,
      "loss": 1.0215,
      "step": 1550
    },
    {
      "epoch": 0.018061408789885613,
      "grad_norm": 3.3361904621124268,
      "learning_rate": 2.4066069774621796e-05,
      "loss": 1.0356,
      "step": 1560
    },
    {
      "epoch": 0.018177187051359237,
      "grad_norm": 3.039552688598633,
      "learning_rate": 2.4220438406915715e-05,
      "loss": 1.0869,
      "step": 1570
    },
    {
      "epoch": 0.01829296531283286,
      "grad_norm": 2.745058298110962,
      "learning_rate": 2.4374807039209634e-05,
      "loss": 1.0952,
      "step": 1580
    },
    {
      "epoch": 0.01840874357430649,
      "grad_norm": 2.920832395553589,
      "learning_rate": 2.452917567150355e-05,
      "loss": 1.0461,
      "step": 1590
    },
    {
      "epoch": 0.018524521835780114,
      "grad_norm": 3.1426374912261963,
      "learning_rate": 2.468354430379747e-05,
      "loss": 1.0186,
      "step": 1600
    },
    {
      "epoch": 0.01864030009725374,
      "grad_norm": 3.40806245803833,
      "learning_rate": 2.4837912936091388e-05,
      "loss": 1.0087,
      "step": 1610
    },
    {
      "epoch": 0.018756078358727366,
      "grad_norm": 3.004892349243164,
      "learning_rate": 2.4992281568385307e-05,
      "loss": 1.0264,
      "step": 1620
    },
    {
      "epoch": 0.01887185662020099,
      "grad_norm": 3.021807909011841,
      "learning_rate": 2.5146650200679223e-05,
      "loss": 1.0895,
      "step": 1630
    },
    {
      "epoch": 0.01898763488167462,
      "grad_norm": 3.2554380893707275,
      "learning_rate": 2.530101883297314e-05,
      "loss": 1.0906,
      "step": 1640
    },
    {
      "epoch": 0.019103413143148243,
      "grad_norm": 3.3182787895202637,
      "learning_rate": 2.545538746526706e-05,
      "loss": 1.0478,
      "step": 1650
    },
    {
      "epoch": 0.019219191404621867,
      "grad_norm": 3.3247146606445312,
      "learning_rate": 2.5609756097560977e-05,
      "loss": 1.0827,
      "step": 1660
    },
    {
      "epoch": 0.019334969666095495,
      "grad_norm": 2.731297731399536,
      "learning_rate": 2.5764124729854892e-05,
      "loss": 1.0529,
      "step": 1670
    },
    {
      "epoch": 0.01945074792756912,
      "grad_norm": 3.013205051422119,
      "learning_rate": 2.591849336214881e-05,
      "loss": 1.0359,
      "step": 1680
    },
    {
      "epoch": 0.019566526189042744,
      "grad_norm": 3.552138328552246,
      "learning_rate": 2.6072861994442734e-05,
      "loss": 1.0454,
      "step": 1690
    },
    {
      "epoch": 0.019682304450516372,
      "grad_norm": 3.9224207401275635,
      "learning_rate": 2.622723062673665e-05,
      "loss": 1.0597,
      "step": 1700
    },
    {
      "epoch": 0.019798082711989996,
      "grad_norm": 3.3173727989196777,
      "learning_rate": 2.6381599259030565e-05,
      "loss": 1.049,
      "step": 1710
    },
    {
      "epoch": 0.019913860973463624,
      "grad_norm": 3.405595541000366,
      "learning_rate": 2.6535967891324488e-05,
      "loss": 1.0833,
      "step": 1720
    },
    {
      "epoch": 0.02002963923493725,
      "grad_norm": 2.7375831604003906,
      "learning_rate": 2.6690336523618403e-05,
      "loss": 1.003,
      "step": 1730
    },
    {
      "epoch": 0.020145417496410873,
      "grad_norm": 3.2522525787353516,
      "learning_rate": 2.684470515591232e-05,
      "loss": 1.0083,
      "step": 1740
    },
    {
      "epoch": 0.0202611957578845,
      "grad_norm": 4.265419960021973,
      "learning_rate": 2.6999073788206234e-05,
      "loss": 1.0428,
      "step": 1750
    },
    {
      "epoch": 0.020376974019358125,
      "grad_norm": 3.507509708404541,
      "learning_rate": 2.7153442420500157e-05,
      "loss": 1.0111,
      "step": 1760
    },
    {
      "epoch": 0.02049275228083175,
      "grad_norm": 3.173712968826294,
      "learning_rate": 2.7307811052794076e-05,
      "loss": 1.0135,
      "step": 1770
    },
    {
      "epoch": 0.020608530542305378,
      "grad_norm": 2.615588903427124,
      "learning_rate": 2.746217968508799e-05,
      "loss": 1.0133,
      "step": 1780
    },
    {
      "epoch": 0.020724308803779002,
      "grad_norm": 3.384016990661621,
      "learning_rate": 2.7616548317381907e-05,
      "loss": 0.9682,
      "step": 1790
    },
    {
      "epoch": 0.02084008706525263,
      "grad_norm": 3.2637808322906494,
      "learning_rate": 2.777091694967583e-05,
      "loss": 1.0344,
      "step": 1800
    },
    {
      "epoch": 0.020955865326726254,
      "grad_norm": 3.193427562713623,
      "learning_rate": 2.7925285581969745e-05,
      "loss": 0.9591,
      "step": 1810
    },
    {
      "epoch": 0.02107164358819988,
      "grad_norm": 3.3400566577911377,
      "learning_rate": 2.807965421426366e-05,
      "loss": 0.988,
      "step": 1820
    },
    {
      "epoch": 0.021187421849673507,
      "grad_norm": 4.1463189125061035,
      "learning_rate": 2.823402284655758e-05,
      "loss": 1.0367,
      "step": 1830
    },
    {
      "epoch": 0.02130320011114713,
      "grad_norm": 3.716966152191162,
      "learning_rate": 2.83883914788515e-05,
      "loss": 1.0261,
      "step": 1840
    },
    {
      "epoch": 0.021418978372620755,
      "grad_norm": 3.9861702919006348,
      "learning_rate": 2.8542760111145418e-05,
      "loss": 0.9644,
      "step": 1850
    },
    {
      "epoch": 0.021534756634094383,
      "grad_norm": 3.443084716796875,
      "learning_rate": 2.8697128743439334e-05,
      "loss": 0.9853,
      "step": 1860
    },
    {
      "epoch": 0.021650534895568008,
      "grad_norm": 3.5585291385650635,
      "learning_rate": 2.885149737573325e-05,
      "loss": 1.003,
      "step": 1870
    },
    {
      "epoch": 0.021766313157041636,
      "grad_norm": 3.945314645767212,
      "learning_rate": 2.9005866008027172e-05,
      "loss": 0.9488,
      "step": 1880
    },
    {
      "epoch": 0.02188209141851526,
      "grad_norm": 2.5524964332580566,
      "learning_rate": 2.9160234640321088e-05,
      "loss": 0.9994,
      "step": 1890
    },
    {
      "epoch": 0.021997869679988884,
      "grad_norm": 3.669727325439453,
      "learning_rate": 2.9314603272615003e-05,
      "loss": 1.0113,
      "step": 1900
    },
    {
      "epoch": 0.022113647941462512,
      "grad_norm": 3.417371988296509,
      "learning_rate": 2.9468971904908926e-05,
      "loss": 1.0186,
      "step": 1910
    },
    {
      "epoch": 0.022229426202936137,
      "grad_norm": 3.546546459197998,
      "learning_rate": 2.962334053720284e-05,
      "loss": 1.0105,
      "step": 1920
    },
    {
      "epoch": 0.02234520446440976,
      "grad_norm": 3.285242795944214,
      "learning_rate": 2.977770916949676e-05,
      "loss": 0.9804,
      "step": 1930
    },
    {
      "epoch": 0.02246098272588339,
      "grad_norm": 3.5312914848327637,
      "learning_rate": 2.9932077801790676e-05,
      "loss": 1.0216,
      "step": 1940
    },
    {
      "epoch": 0.022576760987357013,
      "grad_norm": 4.406423568725586,
      "learning_rate": 3.00864464340846e-05,
      "loss": 1.0774,
      "step": 1950
    },
    {
      "epoch": 0.02269253924883064,
      "grad_norm": 3.41166353225708,
      "learning_rate": 3.0240815066378514e-05,
      "loss": 1.0042,
      "step": 1960
    },
    {
      "epoch": 0.022808317510304266,
      "grad_norm": 3.5296311378479004,
      "learning_rate": 3.039518369867243e-05,
      "loss": 0.9685,
      "step": 1970
    },
    {
      "epoch": 0.02292409577177789,
      "grad_norm": 3.5020294189453125,
      "learning_rate": 3.0549552330966345e-05,
      "loss": 1.0113,
      "step": 1980
    },
    {
      "epoch": 0.023039874033251518,
      "grad_norm": 3.420305013656616,
      "learning_rate": 3.070392096326027e-05,
      "loss": 1.0196,
      "step": 1990
    },
    {
      "epoch": 0.023155652294725142,
      "grad_norm": 3.3157858848571777,
      "learning_rate": 3.0858289595554183e-05,
      "loss": 0.9866,
      "step": 2000
    },
    {
      "epoch": 0.023155652294725142,
      "eval_chrf": 66.63596109779625,
      "eval_loss": 1.5073097944259644,
      "eval_runtime": 360.1486,
      "eval_samples_per_second": 0.278,
      "eval_steps_per_second": 0.011,
      "step": 2000
    },
    {
      "epoch": 0.023271430556198767,
      "grad_norm": 3.2328410148620605,
      "learning_rate": 3.10126582278481e-05,
      "loss": 1.0005,
      "step": 2010
    },
    {
      "epoch": 0.023387208817672395,
      "grad_norm": 2.708080768585205,
      "learning_rate": 3.1167026860142015e-05,
      "loss": 1.019,
      "step": 2020
    },
    {
      "epoch": 0.02350298707914602,
      "grad_norm": 3.409646511077881,
      "learning_rate": 3.132139549243594e-05,
      "loss": 1.0132,
      "step": 2030
    },
    {
      "epoch": 0.023618765340619647,
      "grad_norm": 3.2843925952911377,
      "learning_rate": 3.147576412472986e-05,
      "loss": 0.9537,
      "step": 2040
    },
    {
      "epoch": 0.02373454360209327,
      "grad_norm": 3.7965049743652344,
      "learning_rate": 3.163013275702377e-05,
      "loss": 0.9789,
      "step": 2050
    },
    {
      "epoch": 0.023850321863566896,
      "grad_norm": 3.3306093215942383,
      "learning_rate": 3.17845013893177e-05,
      "loss": 1.0042,
      "step": 2060
    },
    {
      "epoch": 0.023966100125040524,
      "grad_norm": 2.9587419033050537,
      "learning_rate": 3.193887002161161e-05,
      "loss": 0.9922,
      "step": 2070
    },
    {
      "epoch": 0.024081878386514148,
      "grad_norm": 3.425163984298706,
      "learning_rate": 3.209323865390553e-05,
      "loss": 0.9196,
      "step": 2080
    },
    {
      "epoch": 0.024197656647987772,
      "grad_norm": 2.6284561157226562,
      "learning_rate": 3.224760728619944e-05,
      "loss": 0.9698,
      "step": 2090
    },
    {
      "epoch": 0.0243134349094614,
      "grad_norm": 3.568027973175049,
      "learning_rate": 3.240197591849337e-05,
      "loss": 0.9104,
      "step": 2100
    },
    {
      "epoch": 0.024429213170935025,
      "grad_norm": 3.238100528717041,
      "learning_rate": 3.255634455078728e-05,
      "loss": 1.0187,
      "step": 2110
    },
    {
      "epoch": 0.024544991432408653,
      "grad_norm": 3.090458631515503,
      "learning_rate": 3.27107131830812e-05,
      "loss": 0.961,
      "step": 2120
    },
    {
      "epoch": 0.024660769693882277,
      "grad_norm": 3.2551889419555664,
      "learning_rate": 3.286508181537512e-05,
      "loss": 0.9695,
      "step": 2130
    },
    {
      "epoch": 0.0247765479553559,
      "grad_norm": 3.4830145835876465,
      "learning_rate": 3.3019450447669036e-05,
      "loss": 0.9376,
      "step": 2140
    },
    {
      "epoch": 0.02489232621682953,
      "grad_norm": 3.2776224613189697,
      "learning_rate": 3.3173819079962956e-05,
      "loss": 0.9539,
      "step": 2150
    },
    {
      "epoch": 0.025008104478303154,
      "grad_norm": 3.374253273010254,
      "learning_rate": 3.332818771225687e-05,
      "loss": 0.9837,
      "step": 2160
    },
    {
      "epoch": 0.025123882739776778,
      "grad_norm": 3.049576997756958,
      "learning_rate": 3.348255634455079e-05,
      "loss": 0.8857,
      "step": 2170
    },
    {
      "epoch": 0.025239661001250406,
      "grad_norm": 3.490880250930786,
      "learning_rate": 3.3636924976844706e-05,
      "loss": 1.0045,
      "step": 2180
    },
    {
      "epoch": 0.02535543926272403,
      "grad_norm": 3.2462692260742188,
      "learning_rate": 3.3791293609138625e-05,
      "loss": 0.9645,
      "step": 2190
    },
    {
      "epoch": 0.02547121752419766,
      "grad_norm": 3.759615659713745,
      "learning_rate": 3.3945662241432544e-05,
      "loss": 0.9464,
      "step": 2200
    },
    {
      "epoch": 0.025586995785671283,
      "grad_norm": 3.589405059814453,
      "learning_rate": 3.410003087372646e-05,
      "loss": 0.9054,
      "step": 2210
    },
    {
      "epoch": 0.025702774047144907,
      "grad_norm": 4.128746032714844,
      "learning_rate": 3.425439950602038e-05,
      "loss": 0.9756,
      "step": 2220
    },
    {
      "epoch": 0.025818552308618535,
      "grad_norm": 3.282271385192871,
      "learning_rate": 3.4408768138314294e-05,
      "loss": 1.0265,
      "step": 2230
    },
    {
      "epoch": 0.02593433057009216,
      "grad_norm": 3.7959372997283936,
      "learning_rate": 3.456313677060821e-05,
      "loss": 0.9413,
      "step": 2240
    },
    {
      "epoch": 0.026050108831565784,
      "grad_norm": 4.119252681732178,
      "learning_rate": 3.471750540290213e-05,
      "loss": 1.0083,
      "step": 2250
    },
    {
      "epoch": 0.026165887093039412,
      "grad_norm": 3.5423946380615234,
      "learning_rate": 3.487187403519605e-05,
      "loss": 0.9939,
      "step": 2260
    },
    {
      "epoch": 0.026281665354513036,
      "grad_norm": 4.1071882247924805,
      "learning_rate": 3.5026242667489964e-05,
      "loss": 0.9486,
      "step": 2270
    },
    {
      "epoch": 0.026397443615986664,
      "grad_norm": 2.39046573638916,
      "learning_rate": 3.518061129978388e-05,
      "loss": 0.9348,
      "step": 2280
    },
    {
      "epoch": 0.02651322187746029,
      "grad_norm": 3.1564602851867676,
      "learning_rate": 3.533497993207781e-05,
      "loss": 0.9627,
      "step": 2290
    },
    {
      "epoch": 0.026629000138933913,
      "grad_norm": 3.6290247440338135,
      "learning_rate": 3.548934856437172e-05,
      "loss": 0.9517,
      "step": 2300
    },
    {
      "epoch": 0.02674477840040754,
      "grad_norm": 3.5867044925689697,
      "learning_rate": 3.564371719666564e-05,
      "loss": 0.9746,
      "step": 2310
    },
    {
      "epoch": 0.026860556661881165,
      "grad_norm": 4.0372114181518555,
      "learning_rate": 3.579808582895955e-05,
      "loss": 0.9921,
      "step": 2320
    },
    {
      "epoch": 0.02697633492335479,
      "grad_norm": 3.113901138305664,
      "learning_rate": 3.595245446125348e-05,
      "loss": 0.9084,
      "step": 2330
    },
    {
      "epoch": 0.027092113184828417,
      "grad_norm": 3.802921772003174,
      "learning_rate": 3.610682309354739e-05,
      "loss": 0.9022,
      "step": 2340
    },
    {
      "epoch": 0.027207891446302042,
      "grad_norm": 3.4507830142974854,
      "learning_rate": 3.626119172584131e-05,
      "loss": 0.9464,
      "step": 2350
    },
    {
      "epoch": 0.02732366970777567,
      "grad_norm": 4.381093502044678,
      "learning_rate": 3.641556035813523e-05,
      "loss": 0.9977,
      "step": 2360
    },
    {
      "epoch": 0.027439447969249294,
      "grad_norm": 2.51751446723938,
      "learning_rate": 3.656992899042915e-05,
      "loss": 0.9143,
      "step": 2370
    },
    {
      "epoch": 0.02755522623072292,
      "grad_norm": 3.534101724624634,
      "learning_rate": 3.6724297622723066e-05,
      "loss": 0.9669,
      "step": 2380
    },
    {
      "epoch": 0.027671004492196546,
      "grad_norm": 4.157338619232178,
      "learning_rate": 3.687866625501698e-05,
      "loss": 0.9401,
      "step": 2390
    },
    {
      "epoch": 0.02778678275367017,
      "grad_norm": 2.6612493991851807,
      "learning_rate": 3.7033034887310904e-05,
      "loss": 0.8944,
      "step": 2400
    },
    {
      "epoch": 0.027902561015143795,
      "grad_norm": 3.8302664756774902,
      "learning_rate": 3.718740351960482e-05,
      "loss": 0.9386,
      "step": 2410
    },
    {
      "epoch": 0.028018339276617423,
      "grad_norm": 3.27974009513855,
      "learning_rate": 3.7341772151898736e-05,
      "loss": 0.9775,
      "step": 2420
    },
    {
      "epoch": 0.028134117538091048,
      "grad_norm": 3.8576836585998535,
      "learning_rate": 3.7496140784192655e-05,
      "loss": 0.931,
      "step": 2430
    },
    {
      "epoch": 0.028249895799564675,
      "grad_norm": 3.122047185897827,
      "learning_rate": 3.7650509416486574e-05,
      "loss": 0.9224,
      "step": 2440
    },
    {
      "epoch": 0.0283656740610383,
      "grad_norm": 3.2555081844329834,
      "learning_rate": 3.780487804878049e-05,
      "loss": 0.9484,
      "step": 2450
    },
    {
      "epoch": 0.028481452322511924,
      "grad_norm": 4.440473556518555,
      "learning_rate": 3.7959246681074405e-05,
      "loss": 0.9022,
      "step": 2460
    },
    {
      "epoch": 0.028597230583985552,
      "grad_norm": 4.415891647338867,
      "learning_rate": 3.8113615313368324e-05,
      "loss": 0.9589,
      "step": 2470
    },
    {
      "epoch": 0.028713008845459177,
      "grad_norm": 3.2047066688537598,
      "learning_rate": 3.826798394566224e-05,
      "loss": 0.9532,
      "step": 2480
    },
    {
      "epoch": 0.0288287871069328,
      "grad_norm": 3.310581684112549,
      "learning_rate": 3.842235257795616e-05,
      "loss": 0.9224,
      "step": 2490
    },
    {
      "epoch": 0.02894456536840643,
      "grad_norm": 3.6949031352996826,
      "learning_rate": 3.8576721210250075e-05,
      "loss": 0.9472,
      "step": 2500
    },
    {
      "epoch": 0.029060343629880053,
      "grad_norm": 3.2445971965789795,
      "learning_rate": 3.8731089842544e-05,
      "loss": 0.9207,
      "step": 2510
    },
    {
      "epoch": 0.02917612189135368,
      "grad_norm": 3.180643320083618,
      "learning_rate": 3.888545847483791e-05,
      "loss": 0.9137,
      "step": 2520
    },
    {
      "epoch": 0.029291900152827306,
      "grad_norm": 3.478769063949585,
      "learning_rate": 3.903982710713183e-05,
      "loss": 0.9437,
      "step": 2530
    },
    {
      "epoch": 0.02940767841430093,
      "grad_norm": 3.4851291179656982,
      "learning_rate": 3.919419573942575e-05,
      "loss": 0.9451,
      "step": 2540
    },
    {
      "epoch": 0.029523456675774558,
      "grad_norm": 3.5303795337677,
      "learning_rate": 3.934856437171967e-05,
      "loss": 0.9534,
      "step": 2550
    },
    {
      "epoch": 0.029639234937248182,
      "grad_norm": 3.5847115516662598,
      "learning_rate": 3.950293300401359e-05,
      "loss": 0.8699,
      "step": 2560
    },
    {
      "epoch": 0.029755013198721807,
      "grad_norm": 4.173857688903809,
      "learning_rate": 3.96573016363075e-05,
      "loss": 0.9464,
      "step": 2570
    },
    {
      "epoch": 0.029870791460195435,
      "grad_norm": 3.6040806770324707,
      "learning_rate": 3.981167026860142e-05,
      "loss": 0.9125,
      "step": 2580
    },
    {
      "epoch": 0.02998656972166906,
      "grad_norm": 3.881631374359131,
      "learning_rate": 3.996603890089534e-05,
      "loss": 0.9155,
      "step": 2590
    },
    {
      "epoch": 0.030102347983142687,
      "grad_norm": 3.304704189300537,
      "learning_rate": 4.012040753318926e-05,
      "loss": 0.9379,
      "step": 2600
    },
    {
      "epoch": 0.03021812624461631,
      "grad_norm": 3.695401430130005,
      "learning_rate": 4.027477616548318e-05,
      "loss": 0.906,
      "step": 2610
    },
    {
      "epoch": 0.030333904506089936,
      "grad_norm": 2.877660036087036,
      "learning_rate": 4.042914479777709e-05,
      "loss": 0.8771,
      "step": 2620
    },
    {
      "epoch": 0.030449682767563564,
      "grad_norm": 3.3087306022644043,
      "learning_rate": 4.0583513430071015e-05,
      "loss": 0.9575,
      "step": 2630
    },
    {
      "epoch": 0.030565461029037188,
      "grad_norm": 2.551759719848633,
      "learning_rate": 4.073788206236493e-05,
      "loss": 0.8743,
      "step": 2640
    },
    {
      "epoch": 0.030681239290510812,
      "grad_norm": 3.864423990249634,
      "learning_rate": 4.089225069465885e-05,
      "loss": 0.908,
      "step": 2650
    },
    {
      "epoch": 0.03079701755198444,
      "grad_norm": 3.080941915512085,
      "learning_rate": 4.1046619326952766e-05,
      "loss": 0.8992,
      "step": 2660
    },
    {
      "epoch": 0.030912795813458065,
      "grad_norm": 3.5077943801879883,
      "learning_rate": 4.1200987959246685e-05,
      "loss": 0.9176,
      "step": 2670
    },
    {
      "epoch": 0.031028574074931693,
      "grad_norm": 3.443204641342163,
      "learning_rate": 4.1355356591540604e-05,
      "loss": 0.8974,
      "step": 2680
    },
    {
      "epoch": 0.031144352336405317,
      "grad_norm": 3.79953932762146,
      "learning_rate": 4.1509725223834516e-05,
      "loss": 0.8789,
      "step": 2690
    },
    {
      "epoch": 0.031260130597878945,
      "grad_norm": 2.7392163276672363,
      "learning_rate": 4.166409385612844e-05,
      "loss": 0.9331,
      "step": 2700
    },
    {
      "epoch": 0.03137590885935257,
      "grad_norm": 4.023155212402344,
      "learning_rate": 4.1818462488422354e-05,
      "loss": 0.9074,
      "step": 2710
    },
    {
      "epoch": 0.031491687120826194,
      "grad_norm": 3.5471322536468506,
      "learning_rate": 4.197283112071627e-05,
      "loss": 0.92,
      "step": 2720
    },
    {
      "epoch": 0.03160746538229982,
      "grad_norm": 3.721235513687134,
      "learning_rate": 4.2127199753010185e-05,
      "loss": 0.9536,
      "step": 2730
    },
    {
      "epoch": 0.03172324364377344,
      "grad_norm": 3.322194814682007,
      "learning_rate": 4.228156838530411e-05,
      "loss": 0.8973,
      "step": 2740
    },
    {
      "epoch": 0.031839021905247074,
      "grad_norm": 3.457873821258545,
      "learning_rate": 4.2435937017598024e-05,
      "loss": 0.9056,
      "step": 2750
    },
    {
      "epoch": 0.0319548001667207,
      "grad_norm": 3.555284023284912,
      "learning_rate": 4.259030564989194e-05,
      "loss": 0.8985,
      "step": 2760
    },
    {
      "epoch": 0.03207057842819432,
      "grad_norm": 3.6158485412597656,
      "learning_rate": 4.274467428218586e-05,
      "loss": 0.8745,
      "step": 2770
    },
    {
      "epoch": 0.03218635668966795,
      "grad_norm": 3.0284039974212646,
      "learning_rate": 4.289904291447978e-05,
      "loss": 0.9364,
      "step": 2780
    },
    {
      "epoch": 0.03230213495114157,
      "grad_norm": 3.9668619632720947,
      "learning_rate": 4.30534115467737e-05,
      "loss": 0.9246,
      "step": 2790
    },
    {
      "epoch": 0.0324179132126152,
      "grad_norm": 3.699307680130005,
      "learning_rate": 4.320778017906761e-05,
      "loss": 0.9134,
      "step": 2800
    },
    {
      "epoch": 0.03253369147408883,
      "grad_norm": 2.898625612258911,
      "learning_rate": 4.336214881136154e-05,
      "loss": 0.9559,
      "step": 2810
    },
    {
      "epoch": 0.03264946973556245,
      "grad_norm": 3.613478660583496,
      "learning_rate": 4.351651744365545e-05,
      "loss": 0.929,
      "step": 2820
    },
    {
      "epoch": 0.032765247997036076,
      "grad_norm": 3.6123273372650146,
      "learning_rate": 4.367088607594937e-05,
      "loss": 0.8947,
      "step": 2830
    },
    {
      "epoch": 0.0328810262585097,
      "grad_norm": 3.690481662750244,
      "learning_rate": 4.382525470824329e-05,
      "loss": 0.9236,
      "step": 2840
    },
    {
      "epoch": 0.032996804519983325,
      "grad_norm": 3.019315481185913,
      "learning_rate": 4.397962334053721e-05,
      "loss": 0.928,
      "step": 2850
    },
    {
      "epoch": 0.033112582781456956,
      "grad_norm": 3.4528145790100098,
      "learning_rate": 4.4133991972831126e-05,
      "loss": 0.8853,
      "step": 2860
    },
    {
      "epoch": 0.03322836104293058,
      "grad_norm": 3.455610752105713,
      "learning_rate": 4.428836060512504e-05,
      "loss": 0.8617,
      "step": 2870
    },
    {
      "epoch": 0.033344139304404205,
      "grad_norm": 3.1545498371124268,
      "learning_rate": 4.444272923741896e-05,
      "loss": 0.8815,
      "step": 2880
    },
    {
      "epoch": 0.03345991756587783,
      "grad_norm": 3.3913047313690186,
      "learning_rate": 4.4597097869712877e-05,
      "loss": 0.9454,
      "step": 2890
    },
    {
      "epoch": 0.033575695827351454,
      "grad_norm": 3.386596441268921,
      "learning_rate": 4.4751466502006796e-05,
      "loss": 0.9323,
      "step": 2900
    },
    {
      "epoch": 0.033691474088825085,
      "grad_norm": 3.378140687942505,
      "learning_rate": 4.490583513430071e-05,
      "loss": 0.9317,
      "step": 2910
    },
    {
      "epoch": 0.03380725235029871,
      "grad_norm": 3.180781841278076,
      "learning_rate": 4.506020376659463e-05,
      "loss": 0.8937,
      "step": 2920
    },
    {
      "epoch": 0.033923030611772334,
      "grad_norm": 3.7732934951782227,
      "learning_rate": 4.521457239888855e-05,
      "loss": 0.9222,
      "step": 2930
    },
    {
      "epoch": 0.03403880887324596,
      "grad_norm": 3.6625473499298096,
      "learning_rate": 4.5368941031182465e-05,
      "loss": 0.9137,
      "step": 2940
    },
    {
      "epoch": 0.03415458713471958,
      "grad_norm": 3.2060325145721436,
      "learning_rate": 4.5523309663476384e-05,
      "loss": 0.9044,
      "step": 2950
    },
    {
      "epoch": 0.034270365396193214,
      "grad_norm": 3.0926353931427,
      "learning_rate": 4.56776782957703e-05,
      "loss": 0.9117,
      "step": 2960
    },
    {
      "epoch": 0.03438614365766684,
      "grad_norm": 3.5341603755950928,
      "learning_rate": 4.583204692806422e-05,
      "loss": 0.8675,
      "step": 2970
    },
    {
      "epoch": 0.03450192191914046,
      "grad_norm": 2.6664083003997803,
      "learning_rate": 4.5986415560358134e-05,
      "loss": 0.8186,
      "step": 2980
    },
    {
      "epoch": 0.03461770018061409,
      "grad_norm": 3.1512250900268555,
      "learning_rate": 4.6140784192652053e-05,
      "loss": 0.8495,
      "step": 2990
    },
    {
      "epoch": 0.03473347844208771,
      "grad_norm": 3.44207763671875,
      "learning_rate": 4.629515282494597e-05,
      "loss": 0.8995,
      "step": 3000
    },
    {
      "epoch": 0.03473347844208771,
      "eval_chrf": 70.13015374581238,
      "eval_loss": 1.3304837942123413,
      "eval_runtime": 345.8057,
      "eval_samples_per_second": 0.289,
      "eval_steps_per_second": 0.012,
      "step": 3000
    },
    {
      "epoch": 0.034849256703561336,
      "grad_norm": 3.2957146167755127,
      "learning_rate": 4.644952145723989e-05,
      "loss": 0.9227,
      "step": 3010
    },
    {
      "epoch": 0.03496503496503497,
      "grad_norm": 3.4464991092681885,
      "learning_rate": 4.660389008953381e-05,
      "loss": 0.8515,
      "step": 3020
    },
    {
      "epoch": 0.03508081322650859,
      "grad_norm": 3.763947010040283,
      "learning_rate": 4.675825872182772e-05,
      "loss": 0.8865,
      "step": 3030
    },
    {
      "epoch": 0.035196591487982216,
      "grad_norm": 2.4498541355133057,
      "learning_rate": 4.691262735412165e-05,
      "loss": 0.8898,
      "step": 3040
    },
    {
      "epoch": 0.03531236974945584,
      "grad_norm": 3.7224974632263184,
      "learning_rate": 4.706699598641556e-05,
      "loss": 0.922,
      "step": 3050
    },
    {
      "epoch": 0.035428148010929465,
      "grad_norm": 2.902953863143921,
      "learning_rate": 4.722136461870948e-05,
      "loss": 0.8597,
      "step": 3060
    },
    {
      "epoch": 0.0355439262724031,
      "grad_norm": 2.5387275218963623,
      "learning_rate": 4.73757332510034e-05,
      "loss": 0.8588,
      "step": 3070
    },
    {
      "epoch": 0.03565970453387672,
      "grad_norm": 2.914139986038208,
      "learning_rate": 4.753010188329732e-05,
      "loss": 0.8211,
      "step": 3080
    },
    {
      "epoch": 0.035775482795350345,
      "grad_norm": 3.456786632537842,
      "learning_rate": 4.768447051559124e-05,
      "loss": 0.8712,
      "step": 3090
    },
    {
      "epoch": 0.03589126105682397,
      "grad_norm": 3.0146403312683105,
      "learning_rate": 4.783883914788515e-05,
      "loss": 0.8565,
      "step": 3100
    },
    {
      "epoch": 0.036007039318297594,
      "grad_norm": 2.8797733783721924,
      "learning_rate": 4.7993207780179075e-05,
      "loss": 0.8537,
      "step": 3110
    },
    {
      "epoch": 0.036122817579771226,
      "grad_norm": 3.228199005126953,
      "learning_rate": 4.814757641247299e-05,
      "loss": 0.8665,
      "step": 3120
    },
    {
      "epoch": 0.03623859584124485,
      "grad_norm": 3.2365682125091553,
      "learning_rate": 4.8301945044766906e-05,
      "loss": 0.8557,
      "step": 3130
    },
    {
      "epoch": 0.036354374102718474,
      "grad_norm": 2.7745420932769775,
      "learning_rate": 4.845631367706082e-05,
      "loss": 0.8259,
      "step": 3140
    },
    {
      "epoch": 0.0364701523641921,
      "grad_norm": 2.7285120487213135,
      "learning_rate": 4.8610682309354745e-05,
      "loss": 0.8716,
      "step": 3150
    },
    {
      "epoch": 0.03658593062566572,
      "grad_norm": 3.227248430252075,
      "learning_rate": 4.876505094164866e-05,
      "loss": 0.8937,
      "step": 3160
    },
    {
      "epoch": 0.03670170888713935,
      "grad_norm": 3.636167287826538,
      "learning_rate": 4.8919419573942576e-05,
      "loss": 0.8709,
      "step": 3170
    },
    {
      "epoch": 0.03681748714861298,
      "grad_norm": 2.8551406860351562,
      "learning_rate": 4.9073788206236495e-05,
      "loss": 0.9412,
      "step": 3180
    },
    {
      "epoch": 0.0369332654100866,
      "grad_norm": 3.8716530799865723,
      "learning_rate": 4.9228156838530414e-05,
      "loss": 0.8917,
      "step": 3190
    },
    {
      "epoch": 0.03704904367156023,
      "grad_norm": 2.37703275680542,
      "learning_rate": 4.938252547082433e-05,
      "loss": 0.8364,
      "step": 3200
    },
    {
      "epoch": 0.03716482193303385,
      "grad_norm": 2.83221435546875,
      "learning_rate": 4.9536894103118245e-05,
      "loss": 0.8734,
      "step": 3210
    },
    {
      "epoch": 0.03728060019450748,
      "grad_norm": 2.612856388092041,
      "learning_rate": 4.9691262735412164e-05,
      "loss": 0.7954,
      "step": 3220
    },
    {
      "epoch": 0.03739637845598111,
      "grad_norm": 2.5937700271606445,
      "learning_rate": 4.984563136770608e-05,
      "loss": 0.802,
      "step": 3230
    },
    {
      "epoch": 0.03751215671745473,
      "grad_norm": 3.6801981925964355,
      "learning_rate": 5e-05,
      "loss": 0.8902,
      "step": 3240
    },
    {
      "epoch": 0.03762793497892836,
      "grad_norm": 2.671705484390259,
      "learning_rate": 5.015436863229392e-05,
      "loss": 0.894,
      "step": 3250
    },
    {
      "epoch": 0.03774371324040198,
      "grad_norm": 3.6469271183013916,
      "learning_rate": 5.0308737264587834e-05,
      "loss": 0.9043,
      "step": 3260
    },
    {
      "epoch": 0.037859491501875606,
      "grad_norm": 2.4723293781280518,
      "learning_rate": 5.046310589688176e-05,
      "loss": 0.8541,
      "step": 3270
    },
    {
      "epoch": 0.03797526976334924,
      "grad_norm": 3.370495080947876,
      "learning_rate": 5.061747452917568e-05,
      "loss": 0.88,
      "step": 3280
    },
    {
      "epoch": 0.03809104802482286,
      "grad_norm": 3.356306314468384,
      "learning_rate": 5.077184316146959e-05,
      "loss": 0.8632,
      "step": 3290
    },
    {
      "epoch": 0.038206826286296486,
      "grad_norm": 3.166931390762329,
      "learning_rate": 5.092621179376351e-05,
      "loss": 0.873,
      "step": 3300
    },
    {
      "epoch": 0.03832260454777011,
      "grad_norm": 3.311189651489258,
      "learning_rate": 5.108058042605742e-05,
      "loss": 0.8115,
      "step": 3310
    },
    {
      "epoch": 0.038438382809243735,
      "grad_norm": 3.2933318614959717,
      "learning_rate": 5.123494905835135e-05,
      "loss": 0.8813,
      "step": 3320
    },
    {
      "epoch": 0.03855416107071736,
      "grad_norm": 3.240525245666504,
      "learning_rate": 5.138931769064527e-05,
      "loss": 0.8657,
      "step": 3330
    },
    {
      "epoch": 0.03866993933219099,
      "grad_norm": 3.965378999710083,
      "learning_rate": 5.154368632293918e-05,
      "loss": 0.8354,
      "step": 3340
    },
    {
      "epoch": 0.038785717593664615,
      "grad_norm": 3.094787359237671,
      "learning_rate": 5.16980549552331e-05,
      "loss": 0.8763,
      "step": 3350
    },
    {
      "epoch": 0.03890149585513824,
      "grad_norm": 2.820939779281616,
      "learning_rate": 5.1852423587527024e-05,
      "loss": 0.8765,
      "step": 3360
    },
    {
      "epoch": 0.039017274116611864,
      "grad_norm": 3.478208303451538,
      "learning_rate": 5.200679221982093e-05,
      "loss": 0.866,
      "step": 3370
    },
    {
      "epoch": 0.03913305237808549,
      "grad_norm": 2.8255369663238525,
      "learning_rate": 5.2161160852114855e-05,
      "loss": 0.8165,
      "step": 3380
    },
    {
      "epoch": 0.03924883063955912,
      "grad_norm": 3.1648597717285156,
      "learning_rate": 5.231552948440877e-05,
      "loss": 0.8875,
      "step": 3390
    },
    {
      "epoch": 0.039364608901032744,
      "grad_norm": 3.260051727294922,
      "learning_rate": 5.246989811670269e-05,
      "loss": 0.8923,
      "step": 3400
    },
    {
      "epoch": 0.03948038716250637,
      "grad_norm": 3.078834295272827,
      "learning_rate": 5.2624266748996606e-05,
      "loss": 0.8659,
      "step": 3410
    },
    {
      "epoch": 0.03959616542397999,
      "grad_norm": 3.1070268154144287,
      "learning_rate": 5.277863538129052e-05,
      "loss": 0.8446,
      "step": 3420
    },
    {
      "epoch": 0.03971194368545362,
      "grad_norm": 3.4536848068237305,
      "learning_rate": 5.2933004013584444e-05,
      "loss": 0.8689,
      "step": 3430
    },
    {
      "epoch": 0.03982772194692725,
      "grad_norm": 2.8955140113830566,
      "learning_rate": 5.308737264587836e-05,
      "loss": 0.8374,
      "step": 3440
    },
    {
      "epoch": 0.03994350020840087,
      "grad_norm": 4.241433143615723,
      "learning_rate": 5.3241741278172275e-05,
      "loss": 0.8498,
      "step": 3450
    },
    {
      "epoch": 0.0400592784698745,
      "grad_norm": 3.111825942993164,
      "learning_rate": 5.3396109910466194e-05,
      "loss": 0.8377,
      "step": 3460
    },
    {
      "epoch": 0.04017505673134812,
      "grad_norm": 3.1120669841766357,
      "learning_rate": 5.355047854276012e-05,
      "loss": 0.8353,
      "step": 3470
    },
    {
      "epoch": 0.040290834992821746,
      "grad_norm": 3.282771110534668,
      "learning_rate": 5.370484717505403e-05,
      "loss": 0.8499,
      "step": 3480
    },
    {
      "epoch": 0.04040661325429537,
      "grad_norm": 3.2864034175872803,
      "learning_rate": 5.385921580734795e-05,
      "loss": 0.8721,
      "step": 3490
    },
    {
      "epoch": 0.040522391515769,
      "grad_norm": 3.350191831588745,
      "learning_rate": 5.4013584439641864e-05,
      "loss": 0.8412,
      "step": 3500
    },
    {
      "epoch": 0.040638169777242626,
      "grad_norm": 2.9095778465270996,
      "learning_rate": 5.416795307193578e-05,
      "loss": 0.8362,
      "step": 3510
    },
    {
      "epoch": 0.04075394803871625,
      "grad_norm": 2.626868963241577,
      "learning_rate": 5.432232170422971e-05,
      "loss": 0.8676,
      "step": 3520
    },
    {
      "epoch": 0.040869726300189875,
      "grad_norm": 2.8825254440307617,
      "learning_rate": 5.4476690336523614e-05,
      "loss": 0.8388,
      "step": 3530
    },
    {
      "epoch": 0.0409855045616635,
      "grad_norm": 2.8298516273498535,
      "learning_rate": 5.463105896881754e-05,
      "loss": 0.7971,
      "step": 3540
    },
    {
      "epoch": 0.04110128282313713,
      "grad_norm": 3.4504618644714355,
      "learning_rate": 5.478542760111146e-05,
      "loss": 0.8803,
      "step": 3550
    },
    {
      "epoch": 0.041217061084610755,
      "grad_norm": 2.8215651512145996,
      "learning_rate": 5.493979623340537e-05,
      "loss": 0.8311,
      "step": 3560
    },
    {
      "epoch": 0.04133283934608438,
      "grad_norm": 2.7110390663146973,
      "learning_rate": 5.50941648656993e-05,
      "loss": 0.7798,
      "step": 3570
    },
    {
      "epoch": 0.041448617607558004,
      "grad_norm": 3.101635694503784,
      "learning_rate": 5.5248533497993216e-05,
      "loss": 0.8916,
      "step": 3580
    },
    {
      "epoch": 0.04156439586903163,
      "grad_norm": 2.4397974014282227,
      "learning_rate": 5.540290213028713e-05,
      "loss": 0.82,
      "step": 3590
    },
    {
      "epoch": 0.04168017413050526,
      "grad_norm": 3.3154051303863525,
      "learning_rate": 5.555727076258105e-05,
      "loss": 0.8032,
      "step": 3600
    },
    {
      "epoch": 0.041795952391978884,
      "grad_norm": 3.1992838382720947,
      "learning_rate": 5.571163939487496e-05,
      "loss": 0.8279,
      "step": 3610
    },
    {
      "epoch": 0.04191173065345251,
      "grad_norm": 3.188650369644165,
      "learning_rate": 5.586600802716888e-05,
      "loss": 0.8039,
      "step": 3620
    },
    {
      "epoch": 0.04202750891492613,
      "grad_norm": 2.711498737335205,
      "learning_rate": 5.6020376659462804e-05,
      "loss": 0.8561,
      "step": 3630
    },
    {
      "epoch": 0.04214328717639976,
      "grad_norm": 2.9499616622924805,
      "learning_rate": 5.617474529175672e-05,
      "loss": 0.8305,
      "step": 3640
    },
    {
      "epoch": 0.04225906543787338,
      "grad_norm": 3.1685194969177246,
      "learning_rate": 5.6329113924050636e-05,
      "loss": 0.8178,
      "step": 3650
    },
    {
      "epoch": 0.04237484369934701,
      "grad_norm": 2.8285470008850098,
      "learning_rate": 5.6483482556344555e-05,
      "loss": 0.825,
      "step": 3660
    },
    {
      "epoch": 0.04249062196082064,
      "grad_norm": 3.0122334957122803,
      "learning_rate": 5.663785118863847e-05,
      "loss": 0.8565,
      "step": 3670
    },
    {
      "epoch": 0.04260640022229426,
      "grad_norm": 2.9502828121185303,
      "learning_rate": 5.679221982093239e-05,
      "loss": 0.8358,
      "step": 3680
    },
    {
      "epoch": 0.042722178483767886,
      "grad_norm": 3.1885080337524414,
      "learning_rate": 5.69465884532263e-05,
      "loss": 0.8375,
      "step": 3690
    },
    {
      "epoch": 0.04283795674524151,
      "grad_norm": 2.9346580505371094,
      "learning_rate": 5.7100957085520224e-05,
      "loss": 0.8415,
      "step": 3700
    },
    {
      "epoch": 0.04295373500671514,
      "grad_norm": 2.2674832344055176,
      "learning_rate": 5.725532571781414e-05,
      "loss": 0.8082,
      "step": 3710
    },
    {
      "epoch": 0.04306951326818877,
      "grad_norm": 2.7285845279693604,
      "learning_rate": 5.7409694350108055e-05,
      "loss": 0.852,
      "step": 3720
    },
    {
      "epoch": 0.04318529152966239,
      "grad_norm": 3.1136832237243652,
      "learning_rate": 5.756406298240198e-05,
      "loss": 0.8124,
      "step": 3730
    },
    {
      "epoch": 0.043301069791136015,
      "grad_norm": 3.061062812805176,
      "learning_rate": 5.77184316146959e-05,
      "loss": 0.7996,
      "step": 3740
    },
    {
      "epoch": 0.04341684805260964,
      "grad_norm": 3.20890212059021,
      "learning_rate": 5.787280024698981e-05,
      "loss": 0.8482,
      "step": 3750
    },
    {
      "epoch": 0.04353262631408327,
      "grad_norm": 3.1474831104278564,
      "learning_rate": 5.802716887928373e-05,
      "loss": 0.7762,
      "step": 3760
    },
    {
      "epoch": 0.043648404575556896,
      "grad_norm": 2.5395827293395996,
      "learning_rate": 5.818153751157766e-05,
      "loss": 0.8322,
      "step": 3770
    },
    {
      "epoch": 0.04376418283703052,
      "grad_norm": 3.292640447616577,
      "learning_rate": 5.833590614387156e-05,
      "loss": 0.8601,
      "step": 3780
    },
    {
      "epoch": 0.043879961098504144,
      "grad_norm": 3.3988218307495117,
      "learning_rate": 5.849027477616549e-05,
      "loss": 0.7854,
      "step": 3790
    },
    {
      "epoch": 0.04399573935997777,
      "grad_norm": 2.664951801300049,
      "learning_rate": 5.86446434084594e-05,
      "loss": 0.8195,
      "step": 3800
    },
    {
      "epoch": 0.04411151762145139,
      "grad_norm": 3.0127804279327393,
      "learning_rate": 5.879901204075332e-05,
      "loss": 0.8289,
      "step": 3810
    },
    {
      "epoch": 0.044227295882925025,
      "grad_norm": 3.0910518169403076,
      "learning_rate": 5.8953380673047246e-05,
      "loss": 0.8144,
      "step": 3820
    },
    {
      "epoch": 0.04434307414439865,
      "grad_norm": 2.8214893341064453,
      "learning_rate": 5.910774930534115e-05,
      "loss": 0.8251,
      "step": 3830
    },
    {
      "epoch": 0.04445885240587227,
      "grad_norm": 2.9957244396209717,
      "learning_rate": 5.926211793763508e-05,
      "loss": 0.7836,
      "step": 3840
    },
    {
      "epoch": 0.0445746306673459,
      "grad_norm": 2.6521999835968018,
      "learning_rate": 5.9416486569928996e-05,
      "loss": 0.8301,
      "step": 3850
    },
    {
      "epoch": 0.04469040892881952,
      "grad_norm": 2.4596571922302246,
      "learning_rate": 5.957085520222291e-05,
      "loss": 0.7771,
      "step": 3860
    },
    {
      "epoch": 0.044806187190293154,
      "grad_norm": 2.8867480754852295,
      "learning_rate": 5.972522383451683e-05,
      "loss": 0.8202,
      "step": 3870
    },
    {
      "epoch": 0.04492196545176678,
      "grad_norm": 2.1872568130493164,
      "learning_rate": 5.987959246681075e-05,
      "loss": 0.8217,
      "step": 3880
    },
    {
      "epoch": 0.0450377437132404,
      "grad_norm": 3.0329911708831787,
      "learning_rate": 6.0033961099104666e-05,
      "loss": 0.7875,
      "step": 3890
    },
    {
      "epoch": 0.04515352197471403,
      "grad_norm": 2.1675074100494385,
      "learning_rate": 6.0188329731398585e-05,
      "loss": 0.8276,
      "step": 3900
    },
    {
      "epoch": 0.04526930023618765,
      "grad_norm": 3.205519437789917,
      "learning_rate": 6.03426983636925e-05,
      "loss": 0.7714,
      "step": 3910
    },
    {
      "epoch": 0.04538507849766128,
      "grad_norm": 3.124452829360962,
      "learning_rate": 6.0497066995986416e-05,
      "loss": 0.8003,
      "step": 3920
    },
    {
      "epoch": 0.04550085675913491,
      "grad_norm": 3.3183441162109375,
      "learning_rate": 6.065143562828034e-05,
      "loss": 0.8199,
      "step": 3930
    },
    {
      "epoch": 0.04561663502060853,
      "grad_norm": 2.4837539196014404,
      "learning_rate": 6.080580426057425e-05,
      "loss": 0.8187,
      "step": 3940
    },
    {
      "epoch": 0.045732413282082156,
      "grad_norm": 2.7787535190582275,
      "learning_rate": 6.096017289286817e-05,
      "loss": 0.823,
      "step": 3950
    },
    {
      "epoch": 0.04584819154355578,
      "grad_norm": 2.201709508895874,
      "learning_rate": 6.11145415251621e-05,
      "loss": 0.8299,
      "step": 3960
    },
    {
      "epoch": 0.045963969805029405,
      "grad_norm": 2.976959228515625,
      "learning_rate": 6.1268910157456e-05,
      "loss": 0.7301,
      "step": 3970
    },
    {
      "epoch": 0.046079748066503036,
      "grad_norm": 2.4852588176727295,
      "learning_rate": 6.142327878974992e-05,
      "loss": 0.7821,
      "step": 3980
    },
    {
      "epoch": 0.04619552632797666,
      "grad_norm": 2.3276824951171875,
      "learning_rate": 6.157764742204384e-05,
      "loss": 0.8311,
      "step": 3990
    },
    {
      "epoch": 0.046311304589450285,
      "grad_norm": 2.7353389263153076,
      "learning_rate": 6.173201605433776e-05,
      "loss": 0.7818,
      "step": 4000
    },
    {
      "epoch": 0.046311304589450285,
      "eval_chrf": 78.10267281278205,
      "eval_loss": 1.2255442142486572,
      "eval_runtime": 335.4954,
      "eval_samples_per_second": 0.298,
      "eval_steps_per_second": 0.012,
      "step": 4000
    },
    {
      "epoch": 0.04642708285092391,
      "grad_norm": 2.5188751220703125,
      "learning_rate": 6.188638468663168e-05,
      "loss": 0.8627,
      "step": 4010
    },
    {
      "epoch": 0.046542861112397534,
      "grad_norm": 2.4813973903656006,
      "learning_rate": 6.20407533189256e-05,
      "loss": 0.8123,
      "step": 4020
    },
    {
      "epoch": 0.046658639373871165,
      "grad_norm": 2.739382028579712,
      "learning_rate": 6.219512195121952e-05,
      "loss": 0.8058,
      "step": 4030
    },
    {
      "epoch": 0.04677441763534479,
      "grad_norm": 2.2167294025421143,
      "learning_rate": 6.234949058351344e-05,
      "loss": 0.7937,
      "step": 4040
    },
    {
      "epoch": 0.046890195896818414,
      "grad_norm": 2.842698097229004,
      "learning_rate": 6.250385921580734e-05,
      "loss": 0.7537,
      "step": 4050
    },
    {
      "epoch": 0.04700597415829204,
      "grad_norm": 1.9638904333114624,
      "learning_rate": 6.265822784810128e-05,
      "loss": 0.7799,
      "step": 4060
    },
    {
      "epoch": 0.04712175241976566,
      "grad_norm": 2.0738861560821533,
      "learning_rate": 6.28125964803952e-05,
      "loss": 0.8036,
      "step": 4070
    },
    {
      "epoch": 0.047237530681239294,
      "grad_norm": 2.02272367477417,
      "learning_rate": 6.29669651126891e-05,
      "loss": 0.816,
      "step": 4080
    },
    {
      "epoch": 0.04735330894271292,
      "grad_norm": 3.2900118827819824,
      "learning_rate": 6.312133374498302e-05,
      "loss": 0.8152,
      "step": 4090
    },
    {
      "epoch": 0.04746908720418654,
      "grad_norm": 2.37040114402771,
      "learning_rate": 6.327570237727694e-05,
      "loss": 0.7987,
      "step": 4100
    },
    {
      "epoch": 0.04758486546566017,
      "grad_norm": 2.4891092777252197,
      "learning_rate": 6.343007100957086e-05,
      "loss": 0.8258,
      "step": 4110
    },
    {
      "epoch": 0.04770064372713379,
      "grad_norm": 2.619110345840454,
      "learning_rate": 6.358443964186478e-05,
      "loss": 0.8003,
      "step": 4120
    },
    {
      "epoch": 0.047816421988607416,
      "grad_norm": 2.712148666381836,
      "learning_rate": 6.37388082741587e-05,
      "loss": 0.7675,
      "step": 4130
    },
    {
      "epoch": 0.04793220025008105,
      "grad_norm": 2.545774221420288,
      "learning_rate": 6.389317690645261e-05,
      "loss": 0.8105,
      "step": 4140
    },
    {
      "epoch": 0.04804797851155467,
      "grad_norm": 2.7815654277801514,
      "learning_rate": 6.404754553874653e-05,
      "loss": 0.7548,
      "step": 4150
    },
    {
      "epoch": 0.048163756773028296,
      "grad_norm": 2.2365033626556396,
      "learning_rate": 6.420191417104044e-05,
      "loss": 0.817,
      "step": 4160
    },
    {
      "epoch": 0.04827953503450192,
      "grad_norm": 2.854419469833374,
      "learning_rate": 6.435628280333437e-05,
      "loss": 0.7837,
      "step": 4170
    },
    {
      "epoch": 0.048395313295975545,
      "grad_norm": 2.573758125305176,
      "learning_rate": 6.451065143562829e-05,
      "loss": 0.7692,
      "step": 4180
    },
    {
      "epoch": 0.048511091557449176,
      "grad_norm": 1.9099613428115845,
      "learning_rate": 6.46650200679222e-05,
      "loss": 0.7568,
      "step": 4190
    },
    {
      "epoch": 0.0486268698189228,
      "grad_norm": 2.696302890777588,
      "learning_rate": 6.481938870021612e-05,
      "loss": 0.7838,
      "step": 4200
    },
    {
      "epoch": 0.048742648080396425,
      "grad_norm": 2.6622250080108643,
      "learning_rate": 6.497375733251003e-05,
      "loss": 0.806,
      "step": 4210
    },
    {
      "epoch": 0.04885842634187005,
      "grad_norm": 2.6504921913146973,
      "learning_rate": 6.512812596480395e-05,
      "loss": 0.6977,
      "step": 4220
    },
    {
      "epoch": 0.048974204603343674,
      "grad_norm": 2.4777019023895264,
      "learning_rate": 6.528249459709787e-05,
      "loss": 0.7442,
      "step": 4230
    },
    {
      "epoch": 0.049089982864817305,
      "grad_norm": 1.877619743347168,
      "learning_rate": 6.543686322939179e-05,
      "loss": 0.7939,
      "step": 4240
    },
    {
      "epoch": 0.04920576112629093,
      "grad_norm": 2.833174467086792,
      "learning_rate": 6.559123186168571e-05,
      "loss": 0.7944,
      "step": 4250
    },
    {
      "epoch": 0.049321539387764554,
      "grad_norm": 2.5579330921173096,
      "learning_rate": 6.574560049397963e-05,
      "loss": 0.7618,
      "step": 4260
    },
    {
      "epoch": 0.04943731764923818,
      "grad_norm": 2.9838063716888428,
      "learning_rate": 6.589996912627355e-05,
      "loss": 0.8335,
      "step": 4270
    },
    {
      "epoch": 0.0495530959107118,
      "grad_norm": 2.114177942276001,
      "learning_rate": 6.605433775856747e-05,
      "loss": 0.708,
      "step": 4280
    },
    {
      "epoch": 0.04966887417218543,
      "grad_norm": 2.3989551067352295,
      "learning_rate": 6.620870639086137e-05,
      "loss": 0.7868,
      "step": 4290
    },
    {
      "epoch": 0.04978465243365906,
      "grad_norm": 2.2982163429260254,
      "learning_rate": 6.636307502315529e-05,
      "loss": 0.8518,
      "step": 4300
    },
    {
      "epoch": 0.04990043069513268,
      "grad_norm": 2.146652936935425,
      "learning_rate": 6.651744365544922e-05,
      "loss": 0.7471,
      "step": 4310
    },
    {
      "epoch": 0.05001620895660631,
      "grad_norm": 2.5075058937072754,
      "learning_rate": 6.667181228774313e-05,
      "loss": 0.7963,
      "step": 4320
    },
    {
      "epoch": 0.05013198721807993,
      "grad_norm": 2.463543176651001,
      "learning_rate": 6.682618092003705e-05,
      "loss": 0.768,
      "step": 4330
    },
    {
      "epoch": 0.050247765479553556,
      "grad_norm": 2.5465927124023438,
      "learning_rate": 6.698054955233097e-05,
      "loss": 0.7699,
      "step": 4340
    },
    {
      "epoch": 0.05036354374102719,
      "grad_norm": 2.5057928562164307,
      "learning_rate": 6.713491818462489e-05,
      "loss": 0.7773,
      "step": 4350
    },
    {
      "epoch": 0.05047932200250081,
      "grad_norm": 2.135197639465332,
      "learning_rate": 6.72892868169188e-05,
      "loss": 0.7582,
      "step": 4360
    },
    {
      "epoch": 0.05059510026397444,
      "grad_norm": 2.8874289989471436,
      "learning_rate": 6.744365544921273e-05,
      "loss": 0.7849,
      "step": 4370
    },
    {
      "epoch": 0.05071087852544806,
      "grad_norm": 2.699106216430664,
      "learning_rate": 6.759802408150664e-05,
      "loss": 0.7414,
      "step": 4380
    },
    {
      "epoch": 0.050826656786921685,
      "grad_norm": 2.337489366531372,
      "learning_rate": 6.775239271380056e-05,
      "loss": 0.8291,
      "step": 4390
    },
    {
      "epoch": 0.05094243504839532,
      "grad_norm": 2.667902946472168,
      "learning_rate": 6.790676134609447e-05,
      "loss": 0.797,
      "step": 4400
    },
    {
      "epoch": 0.05105821330986894,
      "grad_norm": 2.971599578857422,
      "learning_rate": 6.806112997838839e-05,
      "loss": 0.7692,
      "step": 4410
    },
    {
      "epoch": 0.051173991571342566,
      "grad_norm": 2.433220624923706,
      "learning_rate": 6.821549861068232e-05,
      "loss": 0.7582,
      "step": 4420
    },
    {
      "epoch": 0.05128976983281619,
      "grad_norm": 2.4022865295410156,
      "learning_rate": 6.836986724297623e-05,
      "loss": 0.7812,
      "step": 4430
    },
    {
      "epoch": 0.051405548094289814,
      "grad_norm": 2.868147850036621,
      "learning_rate": 6.852423587527015e-05,
      "loss": 0.7301,
      "step": 4440
    },
    {
      "epoch": 0.05152132635576344,
      "grad_norm": 2.374920129776001,
      "learning_rate": 6.867860450756406e-05,
      "loss": 0.7682,
      "step": 4450
    },
    {
      "epoch": 0.05163710461723707,
      "grad_norm": 1.8938639163970947,
      "learning_rate": 6.883297313985798e-05,
      "loss": 0.7755,
      "step": 4460
    },
    {
      "epoch": 0.051752882878710695,
      "grad_norm": 2.5213005542755127,
      "learning_rate": 6.89873417721519e-05,
      "loss": 0.7409,
      "step": 4470
    },
    {
      "epoch": 0.05186866114018432,
      "grad_norm": 2.381471633911133,
      "learning_rate": 6.914171040444582e-05,
      "loss": 0.7829,
      "step": 4480
    },
    {
      "epoch": 0.05198443940165794,
      "grad_norm": 2.480947256088257,
      "learning_rate": 6.929607903673974e-05,
      "loss": 0.7664,
      "step": 4490
    },
    {
      "epoch": 0.05210021766313157,
      "grad_norm": 2.5081028938293457,
      "learning_rate": 6.945044766903366e-05,
      "loss": 0.7392,
      "step": 4500
    },
    {
      "epoch": 0.0522159959246052,
      "grad_norm": 2.8040149211883545,
      "learning_rate": 6.960481630132756e-05,
      "loss": 0.7464,
      "step": 4510
    },
    {
      "epoch": 0.052331774186078824,
      "grad_norm": 2.6462340354919434,
      "learning_rate": 6.97591849336215e-05,
      "loss": 0.7852,
      "step": 4520
    },
    {
      "epoch": 0.05244755244755245,
      "grad_norm": 1.8709461688995361,
      "learning_rate": 6.991355356591542e-05,
      "loss": 0.7589,
      "step": 4530
    },
    {
      "epoch": 0.05256333070902607,
      "grad_norm": 2.481294631958008,
      "learning_rate": 7.006792219820932e-05,
      "loss": 0.7656,
      "step": 4540
    },
    {
      "epoch": 0.0526791089704997,
      "grad_norm": 2.5509347915649414,
      "learning_rate": 7.022229083050324e-05,
      "loss": 0.7926,
      "step": 4550
    },
    {
      "epoch": 0.05279488723197333,
      "grad_norm": 2.2504377365112305,
      "learning_rate": 7.037665946279717e-05,
      "loss": 0.7344,
      "step": 4560
    },
    {
      "epoch": 0.05291066549344695,
      "grad_norm": 2.534723997116089,
      "learning_rate": 7.053102809509108e-05,
      "loss": 0.7686,
      "step": 4570
    },
    {
      "epoch": 0.05302644375492058,
      "grad_norm": 2.498837947845459,
      "learning_rate": 7.0685396727385e-05,
      "loss": 0.7859,
      "step": 4580
    },
    {
      "epoch": 0.0531422220163942,
      "grad_norm": 2.3672940731048584,
      "learning_rate": 7.083976535967892e-05,
      "loss": 0.7904,
      "step": 4590
    },
    {
      "epoch": 0.053258000277867826,
      "grad_norm": 2.424630641937256,
      "learning_rate": 7.099413399197284e-05,
      "loss": 0.7482,
      "step": 4600
    },
    {
      "epoch": 0.05337377853934145,
      "grad_norm": 2.677380084991455,
      "learning_rate": 7.114850262426676e-05,
      "loss": 0.7088,
      "step": 4610
    },
    {
      "epoch": 0.05348955680081508,
      "grad_norm": 2.422776699066162,
      "learning_rate": 7.130287125656066e-05,
      "loss": 0.7646,
      "step": 4620
    },
    {
      "epoch": 0.053605335062288706,
      "grad_norm": 2.2435402870178223,
      "learning_rate": 7.14572398888546e-05,
      "loss": 0.7661,
      "step": 4630
    },
    {
      "epoch": 0.05372111332376233,
      "grad_norm": 2.1754071712493896,
      "learning_rate": 7.161160852114851e-05,
      "loss": 0.7185,
      "step": 4640
    },
    {
      "epoch": 0.053836891585235955,
      "grad_norm": 2.1318485736846924,
      "learning_rate": 7.176597715344242e-05,
      "loss": 0.7419,
      "step": 4650
    },
    {
      "epoch": 0.05395266984670958,
      "grad_norm": 2.4326610565185547,
      "learning_rate": 7.192034578573634e-05,
      "loss": 0.7811,
      "step": 4660
    },
    {
      "epoch": 0.05406844810818321,
      "grad_norm": 2.451073408126831,
      "learning_rate": 7.207471441803027e-05,
      "loss": 0.7287,
      "step": 4670
    },
    {
      "epoch": 0.054184226369656835,
      "grad_norm": 2.6873838901519775,
      "learning_rate": 7.222908305032418e-05,
      "loss": 0.7497,
      "step": 4680
    },
    {
      "epoch": 0.05430000463113046,
      "grad_norm": 2.3071513175964355,
      "learning_rate": 7.23834516826181e-05,
      "loss": 0.7473,
      "step": 4690
    },
    {
      "epoch": 0.054415782892604084,
      "grad_norm": 2.7855887413024902,
      "learning_rate": 7.253782031491201e-05,
      "loss": 0.7484,
      "step": 4700
    },
    {
      "epoch": 0.05453156115407771,
      "grad_norm": 2.7396068572998047,
      "learning_rate": 7.269218894720593e-05,
      "loss": 0.7428,
      "step": 4710
    },
    {
      "epoch": 0.05464733941555134,
      "grad_norm": 2.5614960193634033,
      "learning_rate": 7.284655757949985e-05,
      "loss": 0.7378,
      "step": 4720
    },
    {
      "epoch": 0.054763117677024964,
      "grad_norm": 2.317500352859497,
      "learning_rate": 7.300092621179376e-05,
      "loss": 0.7336,
      "step": 4730
    },
    {
      "epoch": 0.05487889593849859,
      "grad_norm": 2.811732769012451,
      "learning_rate": 7.315529484408769e-05,
      "loss": 0.7815,
      "step": 4740
    },
    {
      "epoch": 0.05499467419997221,
      "grad_norm": 2.2732126712799072,
      "learning_rate": 7.330966347638161e-05,
      "loss": 0.7828,
      "step": 4750
    },
    {
      "epoch": 0.05511045246144584,
      "grad_norm": 2.6949050426483154,
      "learning_rate": 7.346403210867551e-05,
      "loss": 0.7451,
      "step": 4760
    },
    {
      "epoch": 0.05522623072291946,
      "grad_norm": 2.434868335723877,
      "learning_rate": 7.361840074096945e-05,
      "loss": 0.7165,
      "step": 4770
    },
    {
      "epoch": 0.05534200898439309,
      "grad_norm": 2.574965715408325,
      "learning_rate": 7.377276937326337e-05,
      "loss": 0.7833,
      "step": 4780
    },
    {
      "epoch": 0.05545778724586672,
      "grad_norm": 2.523674726486206,
      "learning_rate": 7.392713800555727e-05,
      "loss": 0.7812,
      "step": 4790
    },
    {
      "epoch": 0.05557356550734034,
      "grad_norm": 2.5483083724975586,
      "learning_rate": 7.408150663785119e-05,
      "loss": 0.6918,
      "step": 4800
    },
    {
      "epoch": 0.055689343768813966,
      "grad_norm": 1.6692194938659668,
      "learning_rate": 7.423587527014511e-05,
      "loss": 0.681,
      "step": 4810
    },
    {
      "epoch": 0.05580512203028759,
      "grad_norm": 2.2478420734405518,
      "learning_rate": 7.439024390243903e-05,
      "loss": 0.7449,
      "step": 4820
    },
    {
      "epoch": 0.05592090029176122,
      "grad_norm": 2.313654899597168,
      "learning_rate": 7.454461253473295e-05,
      "loss": 0.7744,
      "step": 4830
    },
    {
      "epoch": 0.056036678553234846,
      "grad_norm": 2.223144769668579,
      "learning_rate": 7.469898116702687e-05,
      "loss": 0.7917,
      "step": 4840
    },
    {
      "epoch": 0.05615245681470847,
      "grad_norm": 2.536388874053955,
      "learning_rate": 7.485334979932079e-05,
      "loss": 0.7377,
      "step": 4850
    },
    {
      "epoch": 0.056268235076182095,
      "grad_norm": 2.5985777378082275,
      "learning_rate": 7.50077184316147e-05,
      "loss": 0.741,
      "step": 4860
    },
    {
      "epoch": 0.05638401333765572,
      "grad_norm": 2.3443901538848877,
      "learning_rate": 7.516208706390861e-05,
      "loss": 0.7134,
      "step": 4870
    },
    {
      "epoch": 0.05649979159912935,
      "grad_norm": 1.9422653913497925,
      "learning_rate": 7.531645569620254e-05,
      "loss": 0.719,
      "step": 4880
    },
    {
      "epoch": 0.056615569860602975,
      "grad_norm": 2.352421998977661,
      "learning_rate": 7.547082432849645e-05,
      "loss": 0.8023,
      "step": 4890
    },
    {
      "epoch": 0.0567313481220766,
      "grad_norm": 2.353652000427246,
      "learning_rate": 7.562519296079037e-05,
      "loss": 0.7451,
      "step": 4900
    },
    {
      "epoch": 0.056847126383550224,
      "grad_norm": 2.4540939331054688,
      "learning_rate": 7.577956159308429e-05,
      "loss": 0.7751,
      "step": 4910
    },
    {
      "epoch": 0.05696290464502385,
      "grad_norm": 2.339121103286743,
      "learning_rate": 7.59339302253782e-05,
      "loss": 0.6886,
      "step": 4920
    },
    {
      "epoch": 0.05707868290649747,
      "grad_norm": 2.1383137702941895,
      "learning_rate": 7.608829885767212e-05,
      "loss": 0.7626,
      "step": 4930
    },
    {
      "epoch": 0.057194461167971104,
      "grad_norm": 2.392554521560669,
      "learning_rate": 7.624266748996604e-05,
      "loss": 0.768,
      "step": 4940
    },
    {
      "epoch": 0.05731023942944473,
      "grad_norm": 1.97488534450531,
      "learning_rate": 7.639703612225996e-05,
      "loss": 0.7584,
      "step": 4950
    },
    {
      "epoch": 0.05742601769091835,
      "grad_norm": 2.549555540084839,
      "learning_rate": 7.655140475455388e-05,
      "loss": 0.7074,
      "step": 4960
    },
    {
      "epoch": 0.05754179595239198,
      "grad_norm": 2.3444621562957764,
      "learning_rate": 7.67057733868478e-05,
      "loss": 0.7439,
      "step": 4970
    },
    {
      "epoch": 0.0576575742138656,
      "grad_norm": 2.538506269454956,
      "learning_rate": 7.68601420191417e-05,
      "loss": 0.7035,
      "step": 4980
    },
    {
      "epoch": 0.05777335247533923,
      "grad_norm": 2.229445219039917,
      "learning_rate": 7.701451065143564e-05,
      "loss": 0.7294,
      "step": 4990
    },
    {
      "epoch": 0.05788913073681286,
      "grad_norm": 2.288196086883545,
      "learning_rate": 7.716887928372954e-05,
      "loss": 0.7059,
      "step": 5000
    },
    {
      "epoch": 0.05788913073681286,
      "eval_chrf": 68.1041571970253,
      "eval_loss": 1.119584083557129,
      "eval_runtime": 372.9758,
      "eval_samples_per_second": 0.268,
      "eval_steps_per_second": 0.011,
      "step": 5000
    },
    {
      "epoch": 0.05800490899828648,
      "grad_norm": 2.2075209617614746,
      "learning_rate": 7.732324791602346e-05,
      "loss": 0.6979,
      "step": 5010
    },
    {
      "epoch": 0.058120687259760107,
      "grad_norm": 2.3697617053985596,
      "learning_rate": 7.74776165483174e-05,
      "loss": 0.6848,
      "step": 5020
    },
    {
      "epoch": 0.05823646552123373,
      "grad_norm": 2.216676950454712,
      "learning_rate": 7.76319851806113e-05,
      "loss": 0.7588,
      "step": 5030
    },
    {
      "epoch": 0.05835224378270736,
      "grad_norm": 2.1751492023468018,
      "learning_rate": 7.778635381290522e-05,
      "loss": 0.7202,
      "step": 5040
    },
    {
      "epoch": 0.05846802204418099,
      "grad_norm": 2.577639102935791,
      "learning_rate": 7.794072244519914e-05,
      "loss": 0.7168,
      "step": 5050
    },
    {
      "epoch": 0.05858380030565461,
      "grad_norm": 2.2894887924194336,
      "learning_rate": 7.809509107749306e-05,
      "loss": 0.7501,
      "step": 5060
    },
    {
      "epoch": 0.058699578567128236,
      "grad_norm": 2.4248814582824707,
      "learning_rate": 7.824945970978698e-05,
      "loss": 0.7171,
      "step": 5070
    },
    {
      "epoch": 0.05881535682860186,
      "grad_norm": 2.2205724716186523,
      "learning_rate": 7.84038283420809e-05,
      "loss": 0.7034,
      "step": 5080
    },
    {
      "epoch": 0.058931135090075484,
      "grad_norm": 2.135868549346924,
      "learning_rate": 7.855819697437482e-05,
      "loss": 0.7344,
      "step": 5090
    },
    {
      "epoch": 0.059046913351549116,
      "grad_norm": 2.3834145069122314,
      "learning_rate": 7.871256560666873e-05,
      "loss": 0.7241,
      "step": 5100
    },
    {
      "epoch": 0.05916269161302274,
      "grad_norm": 2.800309896469116,
      "learning_rate": 7.886693423896264e-05,
      "loss": 0.6932,
      "step": 5110
    },
    {
      "epoch": 0.059278469874496365,
      "grad_norm": 2.100050926208496,
      "learning_rate": 7.902130287125656e-05,
      "loss": 0.7334,
      "step": 5120
    },
    {
      "epoch": 0.05939424813596999,
      "grad_norm": 2.2075440883636475,
      "learning_rate": 7.917567150355049e-05,
      "loss": 0.718,
      "step": 5130
    },
    {
      "epoch": 0.05951002639744361,
      "grad_norm": 2.4797189235687256,
      "learning_rate": 7.93300401358444e-05,
      "loss": 0.7518,
      "step": 5140
    },
    {
      "epoch": 0.059625804658917245,
      "grad_norm": 2.4825644493103027,
      "learning_rate": 7.948440876813832e-05,
      "loss": 0.7088,
      "step": 5150
    },
    {
      "epoch": 0.05974158292039087,
      "grad_norm": 1.8724825382232666,
      "learning_rate": 7.963877740043223e-05,
      "loss": 0.7398,
      "step": 5160
    },
    {
      "epoch": 0.059857361181864494,
      "grad_norm": 2.2066917419433594,
      "learning_rate": 7.979314603272615e-05,
      "loss": 0.7018,
      "step": 5170
    },
    {
      "epoch": 0.05997313944333812,
      "grad_norm": 1.7651664018630981,
      "learning_rate": 7.994751466502007e-05,
      "loss": 0.7379,
      "step": 5180
    },
    {
      "epoch": 0.06008891770481174,
      "grad_norm": 2.3731231689453125,
      "learning_rate": 8.010188329731398e-05,
      "loss": 0.7332,
      "step": 5190
    },
    {
      "epoch": 0.060204695966285374,
      "grad_norm": 2.377887725830078,
      "learning_rate": 8.025625192960791e-05,
      "loss": 0.7211,
      "step": 5200
    },
    {
      "epoch": 0.060320474227759,
      "grad_norm": 2.151663303375244,
      "learning_rate": 8.041062056190183e-05,
      "loss": 0.6708,
      "step": 5210
    },
    {
      "epoch": 0.06043625248923262,
      "grad_norm": 2.24853253364563,
      "learning_rate": 8.056498919419574e-05,
      "loss": 0.6876,
      "step": 5220
    },
    {
      "epoch": 0.06055203075070625,
      "grad_norm": 2.00726580619812,
      "learning_rate": 8.071935782648965e-05,
      "loss": 0.6922,
      "step": 5230
    },
    {
      "epoch": 0.06066780901217987,
      "grad_norm": 2.350165367126465,
      "learning_rate": 8.087372645878359e-05,
      "loss": 0.7169,
      "step": 5240
    },
    {
      "epoch": 0.060783587273653496,
      "grad_norm": 1.947046160697937,
      "learning_rate": 8.102809509107749e-05,
      "loss": 0.7085,
      "step": 5250
    },
    {
      "epoch": 0.06089936553512713,
      "grad_norm": 2.224485397338867,
      "learning_rate": 8.118246372337141e-05,
      "loss": 0.6767,
      "step": 5260
    },
    {
      "epoch": 0.06101514379660075,
      "grad_norm": 2.2891488075256348,
      "learning_rate": 8.133683235566534e-05,
      "loss": 0.7551,
      "step": 5270
    },
    {
      "epoch": 0.061130922058074376,
      "grad_norm": 2.366861343383789,
      "learning_rate": 8.149120098795925e-05,
      "loss": 0.7257,
      "step": 5280
    },
    {
      "epoch": 0.061246700319548,
      "grad_norm": 2.1227612495422363,
      "learning_rate": 8.164556962025317e-05,
      "loss": 0.715,
      "step": 5290
    },
    {
      "epoch": 0.061362478581021625,
      "grad_norm": 2.105232000350952,
      "learning_rate": 8.179993825254709e-05,
      "loss": 0.7493,
      "step": 5300
    },
    {
      "epoch": 0.061478256842495256,
      "grad_norm": 1.9392329454421997,
      "learning_rate": 8.195430688484101e-05,
      "loss": 0.6857,
      "step": 5310
    },
    {
      "epoch": 0.06159403510396888,
      "grad_norm": 2.0439629554748535,
      "learning_rate": 8.210867551713493e-05,
      "loss": 0.7476,
      "step": 5320
    },
    {
      "epoch": 0.061709813365442505,
      "grad_norm": 2.4530372619628906,
      "learning_rate": 8.226304414942883e-05,
      "loss": 0.7237,
      "step": 5330
    },
    {
      "epoch": 0.06182559162691613,
      "grad_norm": 2.3262078762054443,
      "learning_rate": 8.241741278172276e-05,
      "loss": 0.6685,
      "step": 5340
    },
    {
      "epoch": 0.061941369888389754,
      "grad_norm": 2.394943952560425,
      "learning_rate": 8.257178141401668e-05,
      "loss": 0.693,
      "step": 5350
    },
    {
      "epoch": 0.062057148149863385,
      "grad_norm": 2.129511594772339,
      "learning_rate": 8.272615004631059e-05,
      "loss": 0.7286,
      "step": 5360
    },
    {
      "epoch": 0.06217292641133701,
      "grad_norm": 2.0786781311035156,
      "learning_rate": 8.288051867860451e-05,
      "loss": 0.7012,
      "step": 5370
    },
    {
      "epoch": 0.062288704672810634,
      "grad_norm": 2.4915850162506104,
      "learning_rate": 8.303488731089844e-05,
      "loss": 0.6995,
      "step": 5380
    },
    {
      "epoch": 0.06240448293428426,
      "grad_norm": 1.8084322214126587,
      "learning_rate": 8.318925594319235e-05,
      "loss": 0.7064,
      "step": 5390
    },
    {
      "epoch": 0.06252026119575789,
      "grad_norm": 2.1116020679473877,
      "learning_rate": 8.334362457548626e-05,
      "loss": 0.6557,
      "step": 5400
    },
    {
      "epoch": 0.06263603945723151,
      "grad_norm": 2.2924602031707764,
      "learning_rate": 8.349799320778018e-05,
      "loss": 0.751,
      "step": 5410
    },
    {
      "epoch": 0.06275181771870514,
      "grad_norm": 2.040170907974243,
      "learning_rate": 8.36523618400741e-05,
      "loss": 0.7363,
      "step": 5420
    },
    {
      "epoch": 0.06286759598017876,
      "grad_norm": 2.3817520141601562,
      "learning_rate": 8.380673047236802e-05,
      "loss": 0.7296,
      "step": 5430
    },
    {
      "epoch": 0.06298337424165239,
      "grad_norm": 2.015960931777954,
      "learning_rate": 8.396109910466193e-05,
      "loss": 0.6141,
      "step": 5440
    },
    {
      "epoch": 0.06309915250312602,
      "grad_norm": 1.9143738746643066,
      "learning_rate": 8.411546773695586e-05,
      "loss": 0.7281,
      "step": 5450
    },
    {
      "epoch": 0.06321493076459964,
      "grad_norm": 2.176298141479492,
      "learning_rate": 8.426983636924978e-05,
      "loss": 0.7207,
      "step": 5460
    },
    {
      "epoch": 0.06333070902607327,
      "grad_norm": 1.8675295114517212,
      "learning_rate": 8.442420500154368e-05,
      "loss": 0.6465,
      "step": 5470
    },
    {
      "epoch": 0.06344648728754688,
      "grad_norm": 1.937580943107605,
      "learning_rate": 8.45785736338376e-05,
      "loss": 0.6902,
      "step": 5480
    },
    {
      "epoch": 0.06356226554902052,
      "grad_norm": 2.13000750541687,
      "learning_rate": 8.473294226613152e-05,
      "loss": 0.7225,
      "step": 5490
    },
    {
      "epoch": 0.06367804381049415,
      "grad_norm": 2.099123477935791,
      "learning_rate": 8.488731089842544e-05,
      "loss": 0.6857,
      "step": 5500
    },
    {
      "epoch": 0.06379382207196777,
      "grad_norm": 1.4179965257644653,
      "learning_rate": 8.504167953071936e-05,
      "loss": 0.6895,
      "step": 5510
    },
    {
      "epoch": 0.0639096003334414,
      "grad_norm": 1.969341516494751,
      "learning_rate": 8.519604816301328e-05,
      "loss": 0.7304,
      "step": 5520
    },
    {
      "epoch": 0.06402537859491501,
      "grad_norm": 1.8461629152297974,
      "learning_rate": 8.53504167953072e-05,
      "loss": 0.7309,
      "step": 5530
    },
    {
      "epoch": 0.06414115685638865,
      "grad_norm": 1.9074101448059082,
      "learning_rate": 8.550478542760112e-05,
      "loss": 0.6792,
      "step": 5540
    },
    {
      "epoch": 0.06425693511786228,
      "grad_norm": 1.8870418071746826,
      "learning_rate": 8.565915405989504e-05,
      "loss": 0.7239,
      "step": 5550
    },
    {
      "epoch": 0.0643727133793359,
      "grad_norm": 2.2057502269744873,
      "learning_rate": 8.581352269218896e-05,
      "loss": 0.6806,
      "step": 5560
    },
    {
      "epoch": 0.06448849164080953,
      "grad_norm": 1.8167786598205566,
      "learning_rate": 8.596789132448287e-05,
      "loss": 0.7036,
      "step": 5570
    },
    {
      "epoch": 0.06460426990228314,
      "grad_norm": 2.0966055393218994,
      "learning_rate": 8.612225995677678e-05,
      "loss": 0.7306,
      "step": 5580
    },
    {
      "epoch": 0.06472004816375677,
      "grad_norm": 2.016758680343628,
      "learning_rate": 8.627662858907071e-05,
      "loss": 0.7188,
      "step": 5590
    },
    {
      "epoch": 0.0648358264252304,
      "grad_norm": 1.7833069562911987,
      "learning_rate": 8.643099722136462e-05,
      "loss": 0.6754,
      "step": 5600
    },
    {
      "epoch": 0.06495160468670402,
      "grad_norm": 1.8291857242584229,
      "learning_rate": 8.658536585365854e-05,
      "loss": 0.7223,
      "step": 5610
    },
    {
      "epoch": 0.06506738294817765,
      "grad_norm": 2.064657688140869,
      "learning_rate": 8.673973448595246e-05,
      "loss": 0.6898,
      "step": 5620
    },
    {
      "epoch": 0.06518316120965127,
      "grad_norm": 2.343154191970825,
      "learning_rate": 8.689410311824638e-05,
      "loss": 0.7121,
      "step": 5630
    },
    {
      "epoch": 0.0652989394711249,
      "grad_norm": 2.1844403743743896,
      "learning_rate": 8.70484717505403e-05,
      "loss": 0.7176,
      "step": 5640
    },
    {
      "epoch": 0.06541471773259852,
      "grad_norm": 2.175757646560669,
      "learning_rate": 8.720284038283421e-05,
      "loss": 0.6757,
      "step": 5650
    },
    {
      "epoch": 0.06553049599407215,
      "grad_norm": 2.3068065643310547,
      "learning_rate": 8.735720901512813e-05,
      "loss": 0.6628,
      "step": 5660
    },
    {
      "epoch": 0.06564627425554578,
      "grad_norm": 1.8900717496871948,
      "learning_rate": 8.751157764742205e-05,
      "loss": 0.7111,
      "step": 5670
    },
    {
      "epoch": 0.0657620525170194,
      "grad_norm": 1.975999355316162,
      "learning_rate": 8.766594627971597e-05,
      "loss": 0.6405,
      "step": 5680
    },
    {
      "epoch": 0.06587783077849303,
      "grad_norm": 2.3730521202087402,
      "learning_rate": 8.782031491200988e-05,
      "loss": 0.6701,
      "step": 5690
    },
    {
      "epoch": 0.06599360903996665,
      "grad_norm": 1.9095561504364014,
      "learning_rate": 8.797468354430381e-05,
      "loss": 0.6625,
      "step": 5700
    },
    {
      "epoch": 0.06610938730144028,
      "grad_norm": 1.8720043897628784,
      "learning_rate": 8.812905217659771e-05,
      "loss": 0.681,
      "step": 5710
    },
    {
      "epoch": 0.06622516556291391,
      "grad_norm": 2.0537919998168945,
      "learning_rate": 8.828342080889163e-05,
      "loss": 0.6937,
      "step": 5720
    },
    {
      "epoch": 0.06634094382438753,
      "grad_norm": 1.5734232664108276,
      "learning_rate": 8.843778944118555e-05,
      "loss": 0.7001,
      "step": 5730
    },
    {
      "epoch": 0.06645672208586116,
      "grad_norm": 2.167491912841797,
      "learning_rate": 8.859215807347947e-05,
      "loss": 0.6348,
      "step": 5740
    },
    {
      "epoch": 0.06657250034733478,
      "grad_norm": 1.6173847913742065,
      "learning_rate": 8.874652670577339e-05,
      "loss": 0.7171,
      "step": 5750
    },
    {
      "epoch": 0.06668827860880841,
      "grad_norm": 2.2063612937927246,
      "learning_rate": 8.890089533806731e-05,
      "loss": 0.6922,
      "step": 5760
    },
    {
      "epoch": 0.06680405687028204,
      "grad_norm": 2.069704294204712,
      "learning_rate": 8.905526397036123e-05,
      "loss": 0.6679,
      "step": 5770
    },
    {
      "epoch": 0.06691983513175566,
      "grad_norm": 1.8291423320770264,
      "learning_rate": 8.920963260265515e-05,
      "loss": 0.6771,
      "step": 5780
    },
    {
      "epoch": 0.06703561339322929,
      "grad_norm": 1.96626877784729,
      "learning_rate": 8.936400123494905e-05,
      "loss": 0.6323,
      "step": 5790
    },
    {
      "epoch": 0.06715139165470291,
      "grad_norm": 1.854010820388794,
      "learning_rate": 8.951836986724299e-05,
      "loss": 0.6628,
      "step": 5800
    },
    {
      "epoch": 0.06726716991617654,
      "grad_norm": 2.073439121246338,
      "learning_rate": 8.96727384995369e-05,
      "loss": 0.7042,
      "step": 5810
    },
    {
      "epoch": 0.06738294817765017,
      "grad_norm": 1.8596131801605225,
      "learning_rate": 8.982710713183081e-05,
      "loss": 0.656,
      "step": 5820
    },
    {
      "epoch": 0.06749872643912379,
      "grad_norm": 1.978877067565918,
      "learning_rate": 8.998147576412473e-05,
      "loss": 0.6626,
      "step": 5830
    },
    {
      "epoch": 0.06761450470059742,
      "grad_norm": 1.5365252494812012,
      "learning_rate": 9.013584439641866e-05,
      "loss": 0.7086,
      "step": 5840
    },
    {
      "epoch": 0.06773028296207104,
      "grad_norm": 1.7259769439697266,
      "learning_rate": 9.029021302871257e-05,
      "loss": 0.6991,
      "step": 5850
    },
    {
      "epoch": 0.06784606122354467,
      "grad_norm": 1.5910617113113403,
      "learning_rate": 9.044458166100649e-05,
      "loss": 0.6643,
      "step": 5860
    },
    {
      "epoch": 0.0679618394850183,
      "grad_norm": 2.133162498474121,
      "learning_rate": 9.05989502933004e-05,
      "loss": 0.6835,
      "step": 5870
    },
    {
      "epoch": 0.06807761774649192,
      "grad_norm": 1.6416977643966675,
      "learning_rate": 9.075331892559432e-05,
      "loss": 0.6663,
      "step": 5880
    },
    {
      "epoch": 0.06819339600796555,
      "grad_norm": 1.7024892568588257,
      "learning_rate": 9.090768755788824e-05,
      "loss": 0.7164,
      "step": 5890
    },
    {
      "epoch": 0.06830917426943917,
      "grad_norm": 2.0171029567718506,
      "learning_rate": 9.106205619018215e-05,
      "loss": 0.683,
      "step": 5900
    },
    {
      "epoch": 0.0684249525309128,
      "grad_norm": 1.9070229530334473,
      "learning_rate": 9.121642482247608e-05,
      "loss": 0.6946,
      "step": 5910
    },
    {
      "epoch": 0.06854073079238643,
      "grad_norm": 1.8838920593261719,
      "learning_rate": 9.137079345477e-05,
      "loss": 0.6907,
      "step": 5920
    },
    {
      "epoch": 0.06865650905386005,
      "grad_norm": 1.930190086364746,
      "learning_rate": 9.15251620870639e-05,
      "loss": 0.6846,
      "step": 5930
    },
    {
      "epoch": 0.06877228731533368,
      "grad_norm": 2.060683012008667,
      "learning_rate": 9.167953071935783e-05,
      "loss": 0.6526,
      "step": 5940
    },
    {
      "epoch": 0.0688880655768073,
      "grad_norm": 2.043300151824951,
      "learning_rate": 9.183389935165176e-05,
      "loss": 0.6677,
      "step": 5950
    },
    {
      "epoch": 0.06900384383828093,
      "grad_norm": 1.9208158254623413,
      "learning_rate": 9.198826798394566e-05,
      "loss": 0.6815,
      "step": 5960
    },
    {
      "epoch": 0.06911962209975454,
      "grad_norm": 1.7970118522644043,
      "learning_rate": 9.214263661623958e-05,
      "loss": 0.665,
      "step": 5970
    },
    {
      "epoch": 0.06923540036122817,
      "grad_norm": 1.8869653940200806,
      "learning_rate": 9.22970052485335e-05,
      "loss": 0.6529,
      "step": 5980
    },
    {
      "epoch": 0.0693511786227018,
      "grad_norm": 1.8884750604629517,
      "learning_rate": 9.245137388082742e-05,
      "loss": 0.6602,
      "step": 5990
    },
    {
      "epoch": 0.06946695688417542,
      "grad_norm": 1.7585941553115845,
      "learning_rate": 9.260574251312134e-05,
      "loss": 0.7157,
      "step": 6000
    },
    {
      "epoch": 0.06946695688417542,
      "eval_chrf": 77.35801631389266,
      "eval_loss": 1.0322370529174805,
      "eval_runtime": 346.6588,
      "eval_samples_per_second": 0.288,
      "eval_steps_per_second": 0.012,
      "step": 6000
    },
    {
      "epoch": 0.06958273514564906,
      "grad_norm": 1.3285300731658936,
      "learning_rate": 9.276011114541525e-05,
      "loss": 0.6524,
      "step": 6010
    },
    {
      "epoch": 0.06969851340712267,
      "grad_norm": 1.652048945426941,
      "learning_rate": 9.291447977770918e-05,
      "loss": 0.6673,
      "step": 6020
    },
    {
      "epoch": 0.0698142916685963,
      "grad_norm": 2.03122615814209,
      "learning_rate": 9.30688484100031e-05,
      "loss": 0.7098,
      "step": 6030
    },
    {
      "epoch": 0.06993006993006994,
      "grad_norm": 2.149397373199463,
      "learning_rate": 9.3223217042297e-05,
      "loss": 0.6784,
      "step": 6040
    },
    {
      "epoch": 0.07004584819154355,
      "grad_norm": 1.895490050315857,
      "learning_rate": 9.337758567459093e-05,
      "loss": 0.6849,
      "step": 6050
    },
    {
      "epoch": 0.07016162645301718,
      "grad_norm": 1.8744556903839111,
      "learning_rate": 9.353195430688485e-05,
      "loss": 0.6621,
      "step": 6060
    },
    {
      "epoch": 0.0702774047144908,
      "grad_norm": 1.6119526624679565,
      "learning_rate": 9.368632293917876e-05,
      "loss": 0.6741,
      "step": 6070
    },
    {
      "epoch": 0.07039318297596443,
      "grad_norm": 1.7818323373794556,
      "learning_rate": 9.384069157147268e-05,
      "loss": 0.676,
      "step": 6080
    },
    {
      "epoch": 0.07050896123743806,
      "grad_norm": 2.0417943000793457,
      "learning_rate": 9.39950602037666e-05,
      "loss": 0.6644,
      "step": 6090
    },
    {
      "epoch": 0.07062473949891168,
      "grad_norm": 1.7990567684173584,
      "learning_rate": 9.414942883606052e-05,
      "loss": 0.6457,
      "step": 6100
    },
    {
      "epoch": 0.07074051776038531,
      "grad_norm": 2.0464746952056885,
      "learning_rate": 9.430379746835444e-05,
      "loss": 0.7082,
      "step": 6110
    },
    {
      "epoch": 0.07085629602185893,
      "grad_norm": 1.8701130151748657,
      "learning_rate": 9.445816610064835e-05,
      "loss": 0.6289,
      "step": 6120
    },
    {
      "epoch": 0.07097207428333256,
      "grad_norm": 1.685515284538269,
      "learning_rate": 9.461253473294227e-05,
      "loss": 0.658,
      "step": 6130
    },
    {
      "epoch": 0.0710878525448062,
      "grad_norm": 1.3846675157546997,
      "learning_rate": 9.476690336523619e-05,
      "loss": 0.6543,
      "step": 6140
    },
    {
      "epoch": 0.07120363080627981,
      "grad_norm": 1.9423093795776367,
      "learning_rate": 9.49212719975301e-05,
      "loss": 0.6854,
      "step": 6150
    },
    {
      "epoch": 0.07131940906775344,
      "grad_norm": 2.0019657611846924,
      "learning_rate": 9.507564062982403e-05,
      "loss": 0.633,
      "step": 6160
    },
    {
      "epoch": 0.07143518732922706,
      "grad_norm": 1.8133070468902588,
      "learning_rate": 9.523000926211795e-05,
      "loss": 0.6293,
      "step": 6170
    },
    {
      "epoch": 0.07155096559070069,
      "grad_norm": 1.6339232921600342,
      "learning_rate": 9.538437789441186e-05,
      "loss": 0.6923,
      "step": 6180
    },
    {
      "epoch": 0.07166674385217432,
      "grad_norm": 1.6258397102355957,
      "learning_rate": 9.553874652670577e-05,
      "loss": 0.6674,
      "step": 6190
    },
    {
      "epoch": 0.07178252211364794,
      "grad_norm": 1.7933048009872437,
      "learning_rate": 9.56931151589997e-05,
      "loss": 0.676,
      "step": 6200
    },
    {
      "epoch": 0.07189830037512157,
      "grad_norm": 1.7312748432159424,
      "learning_rate": 9.584748379129361e-05,
      "loss": 0.6616,
      "step": 6210
    },
    {
      "epoch": 0.07201407863659519,
      "grad_norm": 1.8328841924667358,
      "learning_rate": 9.600185242358753e-05,
      "loss": 0.6387,
      "step": 6220
    },
    {
      "epoch": 0.07212985689806882,
      "grad_norm": 1.982265591621399,
      "learning_rate": 9.615622105588145e-05,
      "loss": 0.6548,
      "step": 6230
    },
    {
      "epoch": 0.07224563515954245,
      "grad_norm": 1.8598612546920776,
      "learning_rate": 9.631058968817537e-05,
      "loss": 0.6652,
      "step": 6240
    },
    {
      "epoch": 0.07236141342101607,
      "grad_norm": 1.7716618776321411,
      "learning_rate": 9.646495832046929e-05,
      "loss": 0.6585,
      "step": 6250
    },
    {
      "epoch": 0.0724771916824897,
      "grad_norm": 1.708778738975525,
      "learning_rate": 9.66193269527632e-05,
      "loss": 0.6569,
      "step": 6260
    },
    {
      "epoch": 0.07259296994396332,
      "grad_norm": 1.8234045505523682,
      "learning_rate": 9.677369558505713e-05,
      "loss": 0.6378,
      "step": 6270
    },
    {
      "epoch": 0.07270874820543695,
      "grad_norm": 1.9949054718017578,
      "learning_rate": 9.692806421735103e-05,
      "loss": 0.6505,
      "step": 6280
    },
    {
      "epoch": 0.07282452646691057,
      "grad_norm": 2.0804336071014404,
      "learning_rate": 9.708243284964495e-05,
      "loss": 0.6193,
      "step": 6290
    },
    {
      "epoch": 0.0729403047283842,
      "grad_norm": 1.501078486442566,
      "learning_rate": 9.723680148193888e-05,
      "loss": 0.669,
      "step": 6300
    },
    {
      "epoch": 0.07305608298985783,
      "grad_norm": 1.410736083984375,
      "learning_rate": 9.739117011423279e-05,
      "loss": 0.7089,
      "step": 6310
    },
    {
      "epoch": 0.07317186125133145,
      "grad_norm": 1.65840744972229,
      "learning_rate": 9.754553874652671e-05,
      "loss": 0.686,
      "step": 6320
    },
    {
      "epoch": 0.07328763951280508,
      "grad_norm": 1.8569484949111938,
      "learning_rate": 9.769990737882063e-05,
      "loss": 0.6426,
      "step": 6330
    },
    {
      "epoch": 0.0734034177742787,
      "grad_norm": 1.770147442817688,
      "learning_rate": 9.785427601111455e-05,
      "loss": 0.6409,
      "step": 6340
    },
    {
      "epoch": 0.07351919603575233,
      "grad_norm": 1.8238513469696045,
      "learning_rate": 9.800864464340847e-05,
      "loss": 0.6453,
      "step": 6350
    },
    {
      "epoch": 0.07363497429722596,
      "grad_norm": 1.641202688217163,
      "learning_rate": 9.816301327570238e-05,
      "loss": 0.6341,
      "step": 6360
    },
    {
      "epoch": 0.07375075255869958,
      "grad_norm": 1.791733741760254,
      "learning_rate": 9.83173819079963e-05,
      "loss": 0.6288,
      "step": 6370
    },
    {
      "epoch": 0.0738665308201732,
      "grad_norm": 1.7087746858596802,
      "learning_rate": 9.847175054029022e-05,
      "loss": 0.6619,
      "step": 6380
    },
    {
      "epoch": 0.07398230908164682,
      "grad_norm": 1.6210119724273682,
      "learning_rate": 9.862611917258413e-05,
      "loss": 0.6911,
      "step": 6390
    },
    {
      "epoch": 0.07409808734312046,
      "grad_norm": 1.6889116764068604,
      "learning_rate": 9.878048780487805e-05,
      "loss": 0.6798,
      "step": 6400
    },
    {
      "epoch": 0.07421386560459409,
      "grad_norm": 1.9681158065795898,
      "learning_rate": 9.893485643717198e-05,
      "loss": 0.6133,
      "step": 6410
    },
    {
      "epoch": 0.0743296438660677,
      "grad_norm": 1.7416598796844482,
      "learning_rate": 9.908922506946589e-05,
      "loss": 0.6307,
      "step": 6420
    },
    {
      "epoch": 0.07444542212754134,
      "grad_norm": 1.6089060306549072,
      "learning_rate": 9.92435937017598e-05,
      "loss": 0.6464,
      "step": 6430
    },
    {
      "epoch": 0.07456120038901495,
      "grad_norm": 1.7866533994674683,
      "learning_rate": 9.939796233405372e-05,
      "loss": 0.6392,
      "step": 6440
    },
    {
      "epoch": 0.07467697865048858,
      "grad_norm": 1.6842327117919922,
      "learning_rate": 9.955233096634764e-05,
      "loss": 0.6453,
      "step": 6450
    },
    {
      "epoch": 0.07479275691196222,
      "grad_norm": 1.7159595489501953,
      "learning_rate": 9.970669959864156e-05,
      "loss": 0.6356,
      "step": 6460
    },
    {
      "epoch": 0.07490853517343583,
      "grad_norm": 1.9100960493087769,
      "learning_rate": 9.986106823093548e-05,
      "loss": 0.6543,
      "step": 6470
    },
    {
      "epoch": 0.07502431343490946,
      "grad_norm": 1.9316266775131226,
      "learning_rate": 0.0001000154368632294,
      "loss": 0.6438,
      "step": 6480
    },
    {
      "epoch": 0.07514009169638308,
      "grad_norm": 1.6655845642089844,
      "learning_rate": 0.0001001698054955233,
      "loss": 0.6288,
      "step": 6490
    },
    {
      "epoch": 0.07525586995785671,
      "grad_norm": 1.2088539600372314,
      "learning_rate": 0.00010032417412781722,
      "loss": 0.6163,
      "step": 6500
    },
    {
      "epoch": 0.07537164821933034,
      "grad_norm": 1.6582393646240234,
      "learning_rate": 0.00010047854276011114,
      "loss": 0.6158,
      "step": 6510
    },
    {
      "epoch": 0.07548742648080396,
      "grad_norm": 1.978683352470398,
      "learning_rate": 0.00010063291139240508,
      "loss": 0.6355,
      "step": 6520
    },
    {
      "epoch": 0.0756032047422776,
      "grad_norm": 1.8670979738235474,
      "learning_rate": 0.000100787280024699,
      "loss": 0.6742,
      "step": 6530
    },
    {
      "epoch": 0.07571898300375121,
      "grad_norm": 1.7858290672302246,
      "learning_rate": 0.00010094164865699289,
      "loss": 0.6186,
      "step": 6540
    },
    {
      "epoch": 0.07583476126522484,
      "grad_norm": 1.8994193077087402,
      "learning_rate": 0.00010109601728928682,
      "loss": 0.6253,
      "step": 6550
    },
    {
      "epoch": 0.07595053952669847,
      "grad_norm": 1.5134297609329224,
      "learning_rate": 0.00010125038592158074,
      "loss": 0.6484,
      "step": 6560
    },
    {
      "epoch": 0.07606631778817209,
      "grad_norm": 1.7415775060653687,
      "learning_rate": 0.00010140475455387466,
      "loss": 0.6594,
      "step": 6570
    },
    {
      "epoch": 0.07618209604964572,
      "grad_norm": 1.5696642398834229,
      "learning_rate": 0.00010155912318616858,
      "loss": 0.671,
      "step": 6580
    },
    {
      "epoch": 0.07629787431111934,
      "grad_norm": 1.5940064191818237,
      "learning_rate": 0.00010171349181846251,
      "loss": 0.6236,
      "step": 6590
    },
    {
      "epoch": 0.07641365257259297,
      "grad_norm": 1.937171220779419,
      "learning_rate": 0.0001018678604507564,
      "loss": 0.6295,
      "step": 6600
    },
    {
      "epoch": 0.07652943083406659,
      "grad_norm": 1.5327917337417603,
      "learning_rate": 0.00010202222908305032,
      "loss": 0.6579,
      "step": 6610
    },
    {
      "epoch": 0.07664520909554022,
      "grad_norm": 1.7471036911010742,
      "learning_rate": 0.00010217659771534425,
      "loss": 0.6712,
      "step": 6620
    },
    {
      "epoch": 0.07676098735701385,
      "grad_norm": 1.5663530826568604,
      "learning_rate": 0.00010233096634763817,
      "loss": 0.6341,
      "step": 6630
    },
    {
      "epoch": 0.07687676561848747,
      "grad_norm": 1.6636271476745605,
      "learning_rate": 0.00010248533497993209,
      "loss": 0.6562,
      "step": 6640
    },
    {
      "epoch": 0.0769925438799611,
      "grad_norm": 1.8766775131225586,
      "learning_rate": 0.000102639703612226,
      "loss": 0.6055,
      "step": 6650
    },
    {
      "epoch": 0.07710832214143472,
      "grad_norm": 1.5774224996566772,
      "learning_rate": 0.00010279407224451992,
      "loss": 0.6323,
      "step": 6660
    },
    {
      "epoch": 0.07722410040290835,
      "grad_norm": 1.676321268081665,
      "learning_rate": 0.00010294844087681383,
      "loss": 0.6477,
      "step": 6670
    },
    {
      "epoch": 0.07733987866438198,
      "grad_norm": 1.6336504220962524,
      "learning_rate": 0.00010310280950910775,
      "loss": 0.636,
      "step": 6680
    },
    {
      "epoch": 0.0774556569258556,
      "grad_norm": 1.5671923160552979,
      "learning_rate": 0.00010325717814140167,
      "loss": 0.6283,
      "step": 6690
    },
    {
      "epoch": 0.07757143518732923,
      "grad_norm": 2.129922866821289,
      "learning_rate": 0.0001034115467736956,
      "loss": 0.6285,
      "step": 6700
    },
    {
      "epoch": 0.07768721344880285,
      "grad_norm": 1.6730690002441406,
      "learning_rate": 0.0001035659154059895,
      "loss": 0.6697,
      "step": 6710
    },
    {
      "epoch": 0.07780299171027648,
      "grad_norm": 1.8635300397872925,
      "learning_rate": 0.00010372028403828342,
      "loss": 0.6522,
      "step": 6720
    },
    {
      "epoch": 0.07791876997175011,
      "grad_norm": 1.6877497434616089,
      "learning_rate": 0.00010387465267057735,
      "loss": 0.6962,
      "step": 6730
    },
    {
      "epoch": 0.07803454823322373,
      "grad_norm": 1.301317572593689,
      "learning_rate": 0.00010402902130287127,
      "loss": 0.6564,
      "step": 6740
    },
    {
      "epoch": 0.07815032649469736,
      "grad_norm": 1.2891700267791748,
      "learning_rate": 0.00010418338993516519,
      "loss": 0.6255,
      "step": 6750
    },
    {
      "epoch": 0.07826610475617098,
      "grad_norm": 1.6739258766174316,
      "learning_rate": 0.00010433775856745909,
      "loss": 0.6477,
      "step": 6760
    },
    {
      "epoch": 0.07838188301764461,
      "grad_norm": 1.4691791534423828,
      "learning_rate": 0.00010449212719975301,
      "loss": 0.626,
      "step": 6770
    },
    {
      "epoch": 0.07849766127911824,
      "grad_norm": 1.7983348369598389,
      "learning_rate": 0.00010464649583204693,
      "loss": 0.6411,
      "step": 6780
    },
    {
      "epoch": 0.07861343954059186,
      "grad_norm": 1.7828774452209473,
      "learning_rate": 0.00010480086446434085,
      "loss": 0.6596,
      "step": 6790
    },
    {
      "epoch": 0.07872921780206549,
      "grad_norm": 1.6177442073822021,
      "learning_rate": 0.00010495523309663478,
      "loss": 0.6568,
      "step": 6800
    },
    {
      "epoch": 0.0788449960635391,
      "grad_norm": 1.5620723962783813,
      "learning_rate": 0.0001051096017289287,
      "loss": 0.6078,
      "step": 6810
    },
    {
      "epoch": 0.07896077432501274,
      "grad_norm": 1.6690117120742798,
      "learning_rate": 0.00010526397036122259,
      "loss": 0.6515,
      "step": 6820
    },
    {
      "epoch": 0.07907655258648637,
      "grad_norm": 1.6047637462615967,
      "learning_rate": 0.00010541833899351653,
      "loss": 0.624,
      "step": 6830
    },
    {
      "epoch": 0.07919233084795999,
      "grad_norm": 1.6738725900650024,
      "learning_rate": 0.00010557270762581044,
      "loss": 0.6863,
      "step": 6840
    },
    {
      "epoch": 0.07930810910943362,
      "grad_norm": 1.6993305683135986,
      "learning_rate": 0.00010572707625810436,
      "loss": 0.6439,
      "step": 6850
    },
    {
      "epoch": 0.07942388737090723,
      "grad_norm": 1.7397960424423218,
      "learning_rate": 0.00010588144489039828,
      "loss": 0.6463,
      "step": 6860
    },
    {
      "epoch": 0.07953966563238087,
      "grad_norm": 1.7887377738952637,
      "learning_rate": 0.00010603581352269219,
      "loss": 0.634,
      "step": 6870
    },
    {
      "epoch": 0.0796554438938545,
      "grad_norm": 1.6251155138015747,
      "learning_rate": 0.00010619018215498611,
      "loss": 0.646,
      "step": 6880
    },
    {
      "epoch": 0.07977122215532811,
      "grad_norm": 1.9751814603805542,
      "learning_rate": 0.00010634455078728003,
      "loss": 0.592,
      "step": 6890
    },
    {
      "epoch": 0.07988700041680175,
      "grad_norm": 1.8172136545181274,
      "learning_rate": 0.00010649891941957394,
      "loss": 0.6708,
      "step": 6900
    },
    {
      "epoch": 0.08000277867827536,
      "grad_norm": 1.7657650709152222,
      "learning_rate": 0.00010665328805186788,
      "loss": 0.6044,
      "step": 6910
    },
    {
      "epoch": 0.080118556939749,
      "grad_norm": 1.883344292640686,
      "learning_rate": 0.0001068076566841618,
      "loss": 0.6487,
      "step": 6920
    },
    {
      "epoch": 0.08023433520122261,
      "grad_norm": 1.8151952028274536,
      "learning_rate": 0.00010696202531645569,
      "loss": 0.5774,
      "step": 6930
    },
    {
      "epoch": 0.08035011346269624,
      "grad_norm": 1.6558256149291992,
      "learning_rate": 0.00010711639394874962,
      "loss": 0.605,
      "step": 6940
    },
    {
      "epoch": 0.08046589172416987,
      "grad_norm": 1.4244712591171265,
      "learning_rate": 0.00010727076258104354,
      "loss": 0.6425,
      "step": 6950
    },
    {
      "epoch": 0.08058166998564349,
      "grad_norm": 1.6205344200134277,
      "learning_rate": 0.00010742513121333746,
      "loss": 0.6271,
      "step": 6960
    },
    {
      "epoch": 0.08069744824711712,
      "grad_norm": 1.7787622213363647,
      "learning_rate": 0.00010757949984563138,
      "loss": 0.6597,
      "step": 6970
    },
    {
      "epoch": 0.08081322650859074,
      "grad_norm": 1.4365336894989014,
      "learning_rate": 0.00010773386847792528,
      "loss": 0.6342,
      "step": 6980
    },
    {
      "epoch": 0.08092900477006437,
      "grad_norm": 1.8610031604766846,
      "learning_rate": 0.0001078882371102192,
      "loss": 0.6075,
      "step": 6990
    },
    {
      "epoch": 0.081044783031538,
      "grad_norm": 1.6070164442062378,
      "learning_rate": 0.00010804260574251312,
      "loss": 0.6092,
      "step": 7000
    },
    {
      "epoch": 0.081044783031538,
      "eval_chrf": 77.5078644779532,
      "eval_loss": 0.9781879186630249,
      "eval_runtime": 288.9861,
      "eval_samples_per_second": 0.346,
      "eval_steps_per_second": 0.014,
      "step": 7000
    },
    {
      "epoch": 0.08116056129301162,
      "grad_norm": 1.9760247468948364,
      "learning_rate": 0.00010819697437480704,
      "loss": 0.6607,
      "step": 7010
    },
    {
      "epoch": 0.08127633955448525,
      "grad_norm": 1.7072561979293823,
      "learning_rate": 0.00010835134300710097,
      "loss": 0.6457,
      "step": 7020
    },
    {
      "epoch": 0.08139211781595887,
      "grad_norm": 1.1423382759094238,
      "learning_rate": 0.00010850571163939487,
      "loss": 0.6077,
      "step": 7030
    },
    {
      "epoch": 0.0815078960774325,
      "grad_norm": 1.8201282024383545,
      "learning_rate": 0.00010866008027168878,
      "loss": 0.6134,
      "step": 7040
    },
    {
      "epoch": 0.08162367433890613,
      "grad_norm": 1.670220136642456,
      "learning_rate": 0.00010881444890398272,
      "loss": 0.618,
      "step": 7050
    },
    {
      "epoch": 0.08173945260037975,
      "grad_norm": 1.679893970489502,
      "learning_rate": 0.00010896881753627664,
      "loss": 0.6427,
      "step": 7060
    },
    {
      "epoch": 0.08185523086185338,
      "grad_norm": 1.770871639251709,
      "learning_rate": 0.00010912318616857056,
      "loss": 0.6009,
      "step": 7070
    },
    {
      "epoch": 0.081971009123327,
      "grad_norm": 1.300878643989563,
      "learning_rate": 0.00010927755480086447,
      "loss": 0.6365,
      "step": 7080
    },
    {
      "epoch": 0.08208678738480063,
      "grad_norm": 1.7360892295837402,
      "learning_rate": 0.00010943192343315838,
      "loss": 0.5799,
      "step": 7090
    },
    {
      "epoch": 0.08220256564627426,
      "grad_norm": 1.4603334665298462,
      "learning_rate": 0.0001095862920654523,
      "loss": 0.6345,
      "step": 7100
    },
    {
      "epoch": 0.08231834390774788,
      "grad_norm": 1.6318318843841553,
      "learning_rate": 0.00010974066069774622,
      "loss": 0.637,
      "step": 7110
    },
    {
      "epoch": 0.08243412216922151,
      "grad_norm": 1.1125714778900146,
      "learning_rate": 0.00010989502933004015,
      "loss": 0.5513,
      "step": 7120
    },
    {
      "epoch": 0.08254990043069513,
      "grad_norm": 1.4028571844100952,
      "learning_rate": 0.00011004939796233407,
      "loss": 0.6456,
      "step": 7130
    },
    {
      "epoch": 0.08266567869216876,
      "grad_norm": 1.685958981513977,
      "learning_rate": 0.00011020376659462796,
      "loss": 0.6268,
      "step": 7140
    },
    {
      "epoch": 0.08278145695364239,
      "grad_norm": 1.6421723365783691,
      "learning_rate": 0.0001103581352269219,
      "loss": 0.6307,
      "step": 7150
    },
    {
      "epoch": 0.08289723521511601,
      "grad_norm": 1.5831868648529053,
      "learning_rate": 0.00011051250385921581,
      "loss": 0.6581,
      "step": 7160
    },
    {
      "epoch": 0.08301301347658964,
      "grad_norm": 1.552307367324829,
      "learning_rate": 0.00011066687249150973,
      "loss": 0.6187,
      "step": 7170
    },
    {
      "epoch": 0.08312879173806326,
      "grad_norm": 1.8470345735549927,
      "learning_rate": 0.00011082124112380365,
      "loss": 0.6344,
      "step": 7180
    },
    {
      "epoch": 0.08324456999953689,
      "grad_norm": 1.4212722778320312,
      "learning_rate": 0.00011097560975609757,
      "loss": 0.6249,
      "step": 7190
    },
    {
      "epoch": 0.08336034826101052,
      "grad_norm": 1.5102179050445557,
      "learning_rate": 0.00011112997838839148,
      "loss": 0.6786,
      "step": 7200
    },
    {
      "epoch": 0.08347612652248414,
      "grad_norm": 1.4406299591064453,
      "learning_rate": 0.0001112843470206854,
      "loss": 0.634,
      "step": 7210
    },
    {
      "epoch": 0.08359190478395777,
      "grad_norm": 1.5112054347991943,
      "learning_rate": 0.00011143871565297931,
      "loss": 0.6184,
      "step": 7220
    },
    {
      "epoch": 0.08370768304543139,
      "grad_norm": 1.8691558837890625,
      "learning_rate": 0.00011159308428527325,
      "loss": 0.609,
      "step": 7230
    },
    {
      "epoch": 0.08382346130690502,
      "grad_norm": 1.6059640645980835,
      "learning_rate": 0.00011174745291756717,
      "loss": 0.6323,
      "step": 7240
    },
    {
      "epoch": 0.08393923956837863,
      "grad_norm": 1.6076180934906006,
      "learning_rate": 0.00011190182154986106,
      "loss": 0.6072,
      "step": 7250
    },
    {
      "epoch": 0.08405501782985227,
      "grad_norm": 1.495972752571106,
      "learning_rate": 0.00011205619018215499,
      "loss": 0.5986,
      "step": 7260
    },
    {
      "epoch": 0.0841707960913259,
      "grad_norm": 1.6854097843170166,
      "learning_rate": 0.00011221055881444891,
      "loss": 0.5849,
      "step": 7270
    },
    {
      "epoch": 0.08428657435279951,
      "grad_norm": 1.2864974737167358,
      "learning_rate": 0.00011236492744674283,
      "loss": 0.6189,
      "step": 7280
    },
    {
      "epoch": 0.08440235261427315,
      "grad_norm": 1.587349772453308,
      "learning_rate": 0.00011251929607903675,
      "loss": 0.6366,
      "step": 7290
    },
    {
      "epoch": 0.08451813087574676,
      "grad_norm": 1.5801544189453125,
      "learning_rate": 0.00011267366471133068,
      "loss": 0.5971,
      "step": 7300
    },
    {
      "epoch": 0.0846339091372204,
      "grad_norm": 1.655573844909668,
      "learning_rate": 0.00011282803334362457,
      "loss": 0.6089,
      "step": 7310
    },
    {
      "epoch": 0.08474968739869403,
      "grad_norm": 1.6956554651260376,
      "learning_rate": 0.00011298240197591849,
      "loss": 0.6097,
      "step": 7320
    },
    {
      "epoch": 0.08486546566016764,
      "grad_norm": 1.423643946647644,
      "learning_rate": 0.00011313677060821242,
      "loss": 0.65,
      "step": 7330
    },
    {
      "epoch": 0.08498124392164128,
      "grad_norm": 1.6977276802062988,
      "learning_rate": 0.00011329113924050634,
      "loss": 0.6087,
      "step": 7340
    },
    {
      "epoch": 0.08509702218311489,
      "grad_norm": 1.547904372215271,
      "learning_rate": 0.00011344550787280026,
      "loss": 0.6229,
      "step": 7350
    },
    {
      "epoch": 0.08521280044458852,
      "grad_norm": 1.526370644569397,
      "learning_rate": 0.00011359987650509417,
      "loss": 0.5783,
      "step": 7360
    },
    {
      "epoch": 0.08532857870606216,
      "grad_norm": 1.516898512840271,
      "learning_rate": 0.00011375424513738809,
      "loss": 0.6316,
      "step": 7370
    },
    {
      "epoch": 0.08544435696753577,
      "grad_norm": 1.36735999584198,
      "learning_rate": 0.000113908613769682,
      "loss": 0.5945,
      "step": 7380
    },
    {
      "epoch": 0.0855601352290094,
      "grad_norm": 1.5668267011642456,
      "learning_rate": 0.00011406298240197592,
      "loss": 0.6167,
      "step": 7390
    },
    {
      "epoch": 0.08567591349048302,
      "grad_norm": 1.4169343709945679,
      "learning_rate": 0.00011421735103426984,
      "loss": 0.65,
      "step": 7400
    },
    {
      "epoch": 0.08579169175195665,
      "grad_norm": 1.229204773902893,
      "learning_rate": 0.00011437171966656378,
      "loss": 0.6223,
      "step": 7410
    },
    {
      "epoch": 0.08590747001343028,
      "grad_norm": 1.7227059602737427,
      "learning_rate": 0.00011452608829885767,
      "loss": 0.6522,
      "step": 7420
    },
    {
      "epoch": 0.0860232482749039,
      "grad_norm": 1.2250372171401978,
      "learning_rate": 0.00011468045693115159,
      "loss": 0.5795,
      "step": 7430
    },
    {
      "epoch": 0.08613902653637753,
      "grad_norm": 1.1256134510040283,
      "learning_rate": 0.00011483482556344552,
      "loss": 0.577,
      "step": 7440
    },
    {
      "epoch": 0.08625480479785115,
      "grad_norm": 1.6174768209457397,
      "learning_rate": 0.00011498919419573944,
      "loss": 0.6206,
      "step": 7450
    },
    {
      "epoch": 0.08637058305932478,
      "grad_norm": 1.5744134187698364,
      "learning_rate": 0.00011514356282803336,
      "loss": 0.6299,
      "step": 7460
    },
    {
      "epoch": 0.08648636132079841,
      "grad_norm": 1.511730670928955,
      "learning_rate": 0.00011529793146032726,
      "loss": 0.6113,
      "step": 7470
    },
    {
      "epoch": 0.08660213958227203,
      "grad_norm": 1.0650959014892578,
      "learning_rate": 0.00011545230009262118,
      "loss": 0.6214,
      "step": 7480
    },
    {
      "epoch": 0.08671791784374566,
      "grad_norm": 1.4732906818389893,
      "learning_rate": 0.0001156066687249151,
      "loss": 0.6,
      "step": 7490
    },
    {
      "epoch": 0.08683369610521928,
      "grad_norm": 1.5763410329818726,
      "learning_rate": 0.00011576103735720902,
      "loss": 0.6316,
      "step": 7500
    },
    {
      "epoch": 0.08694947436669291,
      "grad_norm": 1.3677054643630981,
      "learning_rate": 0.00011591540598950294,
      "loss": 0.6364,
      "step": 7510
    },
    {
      "epoch": 0.08706525262816654,
      "grad_norm": 1.6371885538101196,
      "learning_rate": 0.00011606977462179687,
      "loss": 0.6186,
      "step": 7520
    },
    {
      "epoch": 0.08718103088964016,
      "grad_norm": 1.5173264741897583,
      "learning_rate": 0.00011622414325409076,
      "loss": 0.5895,
      "step": 7530
    },
    {
      "epoch": 0.08729680915111379,
      "grad_norm": 1.356893539428711,
      "learning_rate": 0.00011637851188638468,
      "loss": 0.5909,
      "step": 7540
    },
    {
      "epoch": 0.08741258741258741,
      "grad_norm": 1.4249447584152222,
      "learning_rate": 0.00011653288051867861,
      "loss": 0.6154,
      "step": 7550
    },
    {
      "epoch": 0.08752836567406104,
      "grad_norm": 1.709651231765747,
      "learning_rate": 0.00011668724915097253,
      "loss": 0.645,
      "step": 7560
    },
    {
      "epoch": 0.08764414393553466,
      "grad_norm": 1.1198627948760986,
      "learning_rate": 0.00011684161778326645,
      "loss": 0.6013,
      "step": 7570
    },
    {
      "epoch": 0.08775992219700829,
      "grad_norm": 1.6907554864883423,
      "learning_rate": 0.00011699598641556036,
      "loss": 0.5891,
      "step": 7580
    },
    {
      "epoch": 0.08787570045848192,
      "grad_norm": 1.4322824478149414,
      "learning_rate": 0.00011715035504785428,
      "loss": 0.6245,
      "step": 7590
    },
    {
      "epoch": 0.08799147871995554,
      "grad_norm": 1.5642427206039429,
      "learning_rate": 0.0001173047236801482,
      "loss": 0.5773,
      "step": 7600
    },
    {
      "epoch": 0.08810725698142917,
      "grad_norm": 1.4554600715637207,
      "learning_rate": 0.00011745909231244212,
      "loss": 0.5773,
      "step": 7610
    },
    {
      "epoch": 0.08822303524290279,
      "grad_norm": 1.5032012462615967,
      "learning_rate": 0.00011761346094473605,
      "loss": 0.6052,
      "step": 7620
    },
    {
      "epoch": 0.08833881350437642,
      "grad_norm": 1.4786330461502075,
      "learning_rate": 0.00011776782957702994,
      "loss": 0.6391,
      "step": 7630
    },
    {
      "epoch": 0.08845459176585005,
      "grad_norm": 1.6468702554702759,
      "learning_rate": 0.00011792219820932386,
      "loss": 0.5931,
      "step": 7640
    },
    {
      "epoch": 0.08857037002732367,
      "grad_norm": 1.1737236976623535,
      "learning_rate": 0.00011807656684161779,
      "loss": 0.5508,
      "step": 7650
    },
    {
      "epoch": 0.0886861482887973,
      "grad_norm": 1.3755773305892944,
      "learning_rate": 0.00011823093547391171,
      "loss": 0.617,
      "step": 7660
    },
    {
      "epoch": 0.08880192655027092,
      "grad_norm": 1.4953662157058716,
      "learning_rate": 0.00011838530410620563,
      "loss": 0.5913,
      "step": 7670
    },
    {
      "epoch": 0.08891770481174455,
      "grad_norm": 1.3691350221633911,
      "learning_rate": 0.00011853967273849955,
      "loss": 0.6014,
      "step": 7680
    },
    {
      "epoch": 0.08903348307321818,
      "grad_norm": 1.5081861019134521,
      "learning_rate": 0.00011869404137079345,
      "loss": 0.6098,
      "step": 7690
    },
    {
      "epoch": 0.0891492613346918,
      "grad_norm": 1.0612900257110596,
      "learning_rate": 0.00011884841000308737,
      "loss": 0.5549,
      "step": 7700
    },
    {
      "epoch": 0.08926503959616543,
      "grad_norm": 1.838680624961853,
      "learning_rate": 0.00011900277863538129,
      "loss": 0.5754,
      "step": 7710
    },
    {
      "epoch": 0.08938081785763904,
      "grad_norm": 1.3452513217926025,
      "learning_rate": 0.00011915714726767521,
      "loss": 0.5955,
      "step": 7720
    },
    {
      "epoch": 0.08949659611911268,
      "grad_norm": 1.6873469352722168,
      "learning_rate": 0.00011931151589996914,
      "loss": 0.643,
      "step": 7730
    },
    {
      "epoch": 0.08961237438058631,
      "grad_norm": 1.3446232080459595,
      "learning_rate": 0.00011946588453226304,
      "loss": 0.6297,
      "step": 7740
    },
    {
      "epoch": 0.08972815264205992,
      "grad_norm": 1.4106768369674683,
      "learning_rate": 0.00011962025316455696,
      "loss": 0.6031,
      "step": 7750
    },
    {
      "epoch": 0.08984393090353356,
      "grad_norm": 1.3622088432312012,
      "learning_rate": 0.00011977462179685089,
      "loss": 0.5842,
      "step": 7760
    },
    {
      "epoch": 0.08995970916500717,
      "grad_norm": 1.4048413038253784,
      "learning_rate": 0.00011992899042914481,
      "loss": 0.627,
      "step": 7770
    },
    {
      "epoch": 0.0900754874264808,
      "grad_norm": 1.5896952152252197,
      "learning_rate": 0.00012008335906143873,
      "loss": 0.5832,
      "step": 7780
    },
    {
      "epoch": 0.09019126568795444,
      "grad_norm": 1.556296706199646,
      "learning_rate": 0.00012023772769373264,
      "loss": 0.6067,
      "step": 7790
    },
    {
      "epoch": 0.09030704394942805,
      "grad_norm": 1.18021821975708,
      "learning_rate": 0.00012039209632602655,
      "loss": 0.5565,
      "step": 7800
    },
    {
      "epoch": 0.09042282221090168,
      "grad_norm": 1.2262533903121948,
      "learning_rate": 0.00012054646495832047,
      "loss": 0.5884,
      "step": 7810
    },
    {
      "epoch": 0.0905386004723753,
      "grad_norm": 1.611538052558899,
      "learning_rate": 0.00012070083359061439,
      "loss": 0.6222,
      "step": 7820
    },
    {
      "epoch": 0.09065437873384893,
      "grad_norm": 1.4492065906524658,
      "learning_rate": 0.00012085520222290832,
      "loss": 0.6462,
      "step": 7830
    },
    {
      "epoch": 0.09077015699532257,
      "grad_norm": 1.315789818763733,
      "learning_rate": 0.00012100957085520224,
      "loss": 0.6197,
      "step": 7840
    },
    {
      "epoch": 0.09088593525679618,
      "grad_norm": 1.4071003198623657,
      "learning_rate": 0.00012116393948749613,
      "loss": 0.6291,
      "step": 7850
    },
    {
      "epoch": 0.09100171351826981,
      "grad_norm": 1.6243146657943726,
      "learning_rate": 0.00012131830811979006,
      "loss": 0.6315,
      "step": 7860
    },
    {
      "epoch": 0.09111749177974343,
      "grad_norm": 1.271146535873413,
      "learning_rate": 0.00012147267675208398,
      "loss": 0.5774,
      "step": 7870
    },
    {
      "epoch": 0.09123327004121706,
      "grad_norm": 1.40945565700531,
      "learning_rate": 0.0001216270453843779,
      "loss": 0.5967,
      "step": 7880
    },
    {
      "epoch": 0.09134904830269068,
      "grad_norm": 1.4639146327972412,
      "learning_rate": 0.00012178141401667182,
      "loss": 0.6151,
      "step": 7890
    },
    {
      "epoch": 0.09146482656416431,
      "grad_norm": 1.2797602415084839,
      "learning_rate": 0.00012193578264896574,
      "loss": 0.6145,
      "step": 7900
    },
    {
      "epoch": 0.09158060482563794,
      "grad_norm": 1.627112865447998,
      "learning_rate": 0.00012209015128125965,
      "loss": 0.577,
      "step": 7910
    },
    {
      "epoch": 0.09169638308711156,
      "grad_norm": 1.6429121494293213,
      "learning_rate": 0.00012224451991355357,
      "loss": 0.6166,
      "step": 7920
    },
    {
      "epoch": 0.09181216134858519,
      "grad_norm": 1.3286457061767578,
      "learning_rate": 0.00012239888854584748,
      "loss": 0.6161,
      "step": 7930
    },
    {
      "epoch": 0.09192793961005881,
      "grad_norm": 1.244727373123169,
      "learning_rate": 0.0001225532571781414,
      "loss": 0.6118,
      "step": 7940
    },
    {
      "epoch": 0.09204371787153244,
      "grad_norm": 1.2672816514968872,
      "learning_rate": 0.00012270762581043532,
      "loss": 0.6022,
      "step": 7950
    },
    {
      "epoch": 0.09215949613300607,
      "grad_norm": 1.8414747714996338,
      "learning_rate": 0.00012286199444272924,
      "loss": 0.5904,
      "step": 7960
    },
    {
      "epoch": 0.09227527439447969,
      "grad_norm": 1.3384901285171509,
      "learning_rate": 0.00012301636307502316,
      "loss": 0.6023,
      "step": 7970
    },
    {
      "epoch": 0.09239105265595332,
      "grad_norm": 1.671437382698059,
      "learning_rate": 0.00012317073170731708,
      "loss": 0.6095,
      "step": 7980
    },
    {
      "epoch": 0.09250683091742694,
      "grad_norm": 1.3710498809814453,
      "learning_rate": 0.000123325100339611,
      "loss": 0.5638,
      "step": 7990
    },
    {
      "epoch": 0.09262260917890057,
      "grad_norm": 1.2797064781188965,
      "learning_rate": 0.00012347946897190492,
      "loss": 0.6445,
      "step": 8000
    },
    {
      "epoch": 0.09262260917890057,
      "eval_chrf": 81.30991015560114,
      "eval_loss": 0.9110472202301025,
      "eval_runtime": 196.7086,
      "eval_samples_per_second": 0.508,
      "eval_steps_per_second": 0.02,
      "step": 8000
    },
    {
      "epoch": 0.0927383874403742,
      "grad_norm": 1.5262354612350464,
      "learning_rate": 0.00012363383760419884,
      "loss": 0.5924,
      "step": 8010
    },
    {
      "epoch": 0.09285416570184782,
      "grad_norm": 1.577624797821045,
      "learning_rate": 0.00012378820623649276,
      "loss": 0.5941,
      "step": 8020
    },
    {
      "epoch": 0.09296994396332145,
      "grad_norm": 1.403977632522583,
      "learning_rate": 0.00012394257486878667,
      "loss": 0.5668,
      "step": 8030
    },
    {
      "epoch": 0.09308572222479507,
      "grad_norm": 1.437752366065979,
      "learning_rate": 0.0001240969435010806,
      "loss": 0.5817,
      "step": 8040
    },
    {
      "epoch": 0.0932015004862687,
      "grad_norm": 1.4250694513320923,
      "learning_rate": 0.0001242513121333745,
      "loss": 0.5788,
      "step": 8050
    },
    {
      "epoch": 0.09331727874774233,
      "grad_norm": 1.5299127101898193,
      "learning_rate": 0.00012440568076566843,
      "loss": 0.5923,
      "step": 8060
    },
    {
      "epoch": 0.09343305700921595,
      "grad_norm": 1.3042469024658203,
      "learning_rate": 0.00012456004939796232,
      "loss": 0.588,
      "step": 8070
    },
    {
      "epoch": 0.09354883527068958,
      "grad_norm": 1.6172361373901367,
      "learning_rate": 0.00012471441803025624,
      "loss": 0.6053,
      "step": 8080
    },
    {
      "epoch": 0.0936646135321632,
      "grad_norm": 1.2567100524902344,
      "learning_rate": 0.00012486878666255016,
      "loss": 0.606,
      "step": 8090
    },
    {
      "epoch": 0.09378039179363683,
      "grad_norm": 1.529083251953125,
      "learning_rate": 0.0001250231552948441,
      "loss": 0.5402,
      "step": 8100
    },
    {
      "epoch": 0.09389617005511046,
      "grad_norm": 1.569347858428955,
      "learning_rate": 0.00012517752392713803,
      "loss": 0.6213,
      "step": 8110
    },
    {
      "epoch": 0.09401194831658408,
      "grad_norm": 1.103323221206665,
      "learning_rate": 0.00012533189255943195,
      "loss": 0.5959,
      "step": 8120
    },
    {
      "epoch": 0.09412772657805771,
      "grad_norm": 1.4739731550216675,
      "learning_rate": 0.00012548626119172584,
      "loss": 0.6149,
      "step": 8130
    },
    {
      "epoch": 0.09424350483953133,
      "grad_norm": 1.3428549766540527,
      "learning_rate": 0.00012564062982401976,
      "loss": 0.5954,
      "step": 8140
    },
    {
      "epoch": 0.09435928310100496,
      "grad_norm": 1.268978238105774,
      "learning_rate": 0.00012579499845631368,
      "loss": 0.6771,
      "step": 8150
    },
    {
      "epoch": 0.09447506136247859,
      "grad_norm": 1.729616641998291,
      "learning_rate": 0.0001259493670886076,
      "loss": 0.6056,
      "step": 8160
    },
    {
      "epoch": 0.0945908396239522,
      "grad_norm": 1.1826121807098389,
      "learning_rate": 0.00012610373572090154,
      "loss": 0.5792,
      "step": 8170
    },
    {
      "epoch": 0.09470661788542584,
      "grad_norm": 1.5031476020812988,
      "learning_rate": 0.00012625810435319543,
      "loss": 0.6175,
      "step": 8180
    },
    {
      "epoch": 0.09482239614689945,
      "grad_norm": 1.440767526626587,
      "learning_rate": 0.00012641247298548935,
      "loss": 0.5999,
      "step": 8190
    },
    {
      "epoch": 0.09493817440837309,
      "grad_norm": 1.4809963703155518,
      "learning_rate": 0.00012656684161778327,
      "loss": 0.6238,
      "step": 8200
    },
    {
      "epoch": 0.0950539526698467,
      "grad_norm": 1.3183481693267822,
      "learning_rate": 0.0001267212102500772,
      "loss": 0.6007,
      "step": 8210
    },
    {
      "epoch": 0.09516973093132033,
      "grad_norm": 1.6722873449325562,
      "learning_rate": 0.0001268755788823711,
      "loss": 0.6092,
      "step": 8220
    },
    {
      "epoch": 0.09528550919279397,
      "grad_norm": 1.3660516738891602,
      "learning_rate": 0.00012702994751466503,
      "loss": 0.6316,
      "step": 8230
    },
    {
      "epoch": 0.09540128745426758,
      "grad_norm": 1.3543256521224976,
      "learning_rate": 0.00012718431614695895,
      "loss": 0.6008,
      "step": 8240
    },
    {
      "epoch": 0.09551706571574121,
      "grad_norm": 1.416183352470398,
      "learning_rate": 0.00012733868477925287,
      "loss": 0.5911,
      "step": 8250
    },
    {
      "epoch": 0.09563284397721483,
      "grad_norm": 1.6643449068069458,
      "learning_rate": 0.00012749305341154679,
      "loss": 0.5987,
      "step": 8260
    },
    {
      "epoch": 0.09574862223868846,
      "grad_norm": 1.0764458179473877,
      "learning_rate": 0.0001276474220438407,
      "loss": 0.5964,
      "step": 8270
    },
    {
      "epoch": 0.0958644005001621,
      "grad_norm": 0.9228554964065552,
      "learning_rate": 0.00012780179067613462,
      "loss": 0.5634,
      "step": 8280
    },
    {
      "epoch": 0.09598017876163571,
      "grad_norm": 1.8675537109375,
      "learning_rate": 0.00012795615930842852,
      "loss": 0.5647,
      "step": 8290
    },
    {
      "epoch": 0.09609595702310934,
      "grad_norm": 1.2835443019866943,
      "learning_rate": 0.00012811052794072243,
      "loss": 0.6156,
      "step": 8300
    },
    {
      "epoch": 0.09621173528458296,
      "grad_norm": 1.5996533632278442,
      "learning_rate": 0.00012826489657301638,
      "loss": 0.5868,
      "step": 8310
    },
    {
      "epoch": 0.09632751354605659,
      "grad_norm": 1.4104992151260376,
      "learning_rate": 0.0001284192652053103,
      "loss": 0.5801,
      "step": 8320
    },
    {
      "epoch": 0.09644329180753022,
      "grad_norm": 1.3446601629257202,
      "learning_rate": 0.00012857363383760422,
      "loss": 0.5937,
      "step": 8330
    },
    {
      "epoch": 0.09655907006900384,
      "grad_norm": 1.3941295146942139,
      "learning_rate": 0.0001287280024698981,
      "loss": 0.63,
      "step": 8340
    },
    {
      "epoch": 0.09667484833047747,
      "grad_norm": 1.4139586687088013,
      "learning_rate": 0.00012888237110219203,
      "loss": 0.5843,
      "step": 8350
    },
    {
      "epoch": 0.09679062659195109,
      "grad_norm": 1.4007705450057983,
      "learning_rate": 0.00012903673973448595,
      "loss": 0.6374,
      "step": 8360
    },
    {
      "epoch": 0.09690640485342472,
      "grad_norm": 1.3027223348617554,
      "learning_rate": 0.00012919110836677987,
      "loss": 0.5998,
      "step": 8370
    },
    {
      "epoch": 0.09702218311489835,
      "grad_norm": 1.2569539546966553,
      "learning_rate": 0.0001293454769990738,
      "loss": 0.5713,
      "step": 8380
    },
    {
      "epoch": 0.09713796137637197,
      "grad_norm": 1.8045423030853271,
      "learning_rate": 0.00012949984563136773,
      "loss": 0.5923,
      "step": 8390
    },
    {
      "epoch": 0.0972537396378456,
      "grad_norm": 1.31775963306427,
      "learning_rate": 0.00012965421426366163,
      "loss": 0.5698,
      "step": 8400
    },
    {
      "epoch": 0.09736951789931922,
      "grad_norm": 1.3272954225540161,
      "learning_rate": 0.00012980858289595554,
      "loss": 0.6066,
      "step": 8410
    },
    {
      "epoch": 0.09748529616079285,
      "grad_norm": 1.4235128164291382,
      "learning_rate": 0.00012996295152824946,
      "loss": 0.6103,
      "step": 8420
    },
    {
      "epoch": 0.09760107442226648,
      "grad_norm": 1.5209492444992065,
      "learning_rate": 0.00013011732016054338,
      "loss": 0.5898,
      "step": 8430
    },
    {
      "epoch": 0.0977168526837401,
      "grad_norm": 1.377868890762329,
      "learning_rate": 0.0001302716887928373,
      "loss": 0.5395,
      "step": 8440
    },
    {
      "epoch": 0.09783263094521373,
      "grad_norm": 1.3593305349349976,
      "learning_rate": 0.00013042605742513122,
      "loss": 0.5797,
      "step": 8450
    },
    {
      "epoch": 0.09794840920668735,
      "grad_norm": 1.4212123155593872,
      "learning_rate": 0.00013058042605742514,
      "loss": 0.5823,
      "step": 8460
    },
    {
      "epoch": 0.09806418746816098,
      "grad_norm": 1.3460135459899902,
      "learning_rate": 0.00013073479468971906,
      "loss": 0.5999,
      "step": 8470
    },
    {
      "epoch": 0.09817996572963461,
      "grad_norm": 1.0044142007827759,
      "learning_rate": 0.00013088916332201298,
      "loss": 0.5579,
      "step": 8480
    },
    {
      "epoch": 0.09829574399110823,
      "grad_norm": 1.573012113571167,
      "learning_rate": 0.0001310435319543069,
      "loss": 0.5778,
      "step": 8490
    },
    {
      "epoch": 0.09841152225258186,
      "grad_norm": 1.374923586845398,
      "learning_rate": 0.00013119790058660082,
      "loss": 0.5413,
      "step": 8500
    },
    {
      "epoch": 0.09852730051405548,
      "grad_norm": 1.5594185590744019,
      "learning_rate": 0.0001313522692188947,
      "loss": 0.602,
      "step": 8510
    },
    {
      "epoch": 0.09864307877552911,
      "grad_norm": 1.262014627456665,
      "learning_rate": 0.00013150663785118865,
      "loss": 0.5776,
      "step": 8520
    },
    {
      "epoch": 0.09875885703700273,
      "grad_norm": 1.3214367628097534,
      "learning_rate": 0.00013166100648348257,
      "loss": 0.5749,
      "step": 8530
    },
    {
      "epoch": 0.09887463529847636,
      "grad_norm": 1.3003894090652466,
      "learning_rate": 0.0001318153751157765,
      "loss": 0.5972,
      "step": 8540
    },
    {
      "epoch": 0.09899041355994999,
      "grad_norm": 1.3032628297805786,
      "learning_rate": 0.0001319697437480704,
      "loss": 0.6025,
      "step": 8550
    },
    {
      "epoch": 0.0991061918214236,
      "grad_norm": 1.1279217004776,
      "learning_rate": 0.0001321241123803643,
      "loss": 0.6171,
      "step": 8560
    },
    {
      "epoch": 0.09922197008289724,
      "grad_norm": 1.5024949312210083,
      "learning_rate": 0.00013227848101265822,
      "loss": 0.5989,
      "step": 8570
    },
    {
      "epoch": 0.09933774834437085,
      "grad_norm": 1.2569438219070435,
      "learning_rate": 0.00013243284964495214,
      "loss": 0.5829,
      "step": 8580
    },
    {
      "epoch": 0.09945352660584449,
      "grad_norm": 2.529432535171509,
      "learning_rate": 0.00013258721827724606,
      "loss": 0.5594,
      "step": 8590
    },
    {
      "epoch": 0.09956930486731812,
      "grad_norm": 1.4286943674087524,
      "learning_rate": 0.00013274158690954,
      "loss": 0.5716,
      "step": 8600
    },
    {
      "epoch": 0.09968508312879173,
      "grad_norm": 0.9087528586387634,
      "learning_rate": 0.00013289595554183393,
      "loss": 0.5778,
      "step": 8610
    },
    {
      "epoch": 0.09980086139026537,
      "grad_norm": 1.2014377117156982,
      "learning_rate": 0.00013305032417412782,
      "loss": 0.5037,
      "step": 8620
    },
    {
      "epoch": 0.09991663965173898,
      "grad_norm": 0.9955915808677673,
      "learning_rate": 0.00013320469280642174,
      "loss": 0.5882,
      "step": 8630
    },
    {
      "epoch": 0.10003241791321262,
      "grad_norm": 1.3380610942840576,
      "learning_rate": 0.00013335906143871566,
      "loss": 0.5788,
      "step": 8640
    },
    {
      "epoch": 0.10014819617468625,
      "grad_norm": 1.39886474609375,
      "learning_rate": 0.00013351343007100957,
      "loss": 0.5603,
      "step": 8650
    },
    {
      "epoch": 0.10026397443615986,
      "grad_norm": 1.0779165029525757,
      "learning_rate": 0.0001336677987033035,
      "loss": 0.5589,
      "step": 8660
    },
    {
      "epoch": 0.1003797526976335,
      "grad_norm": 1.258846640586853,
      "learning_rate": 0.0001338221673355974,
      "loss": 0.5421,
      "step": 8670
    },
    {
      "epoch": 0.10049553095910711,
      "grad_norm": 1.3092535734176636,
      "learning_rate": 0.00013397653596789133,
      "loss": 0.6237,
      "step": 8680
    },
    {
      "epoch": 0.10061130922058074,
      "grad_norm": 1.2816115617752075,
      "learning_rate": 0.00013413090460018525,
      "loss": 0.6091,
      "step": 8690
    },
    {
      "epoch": 0.10072708748205438,
      "grad_norm": 1.2709742784500122,
      "learning_rate": 0.00013428527323247917,
      "loss": 0.5725,
      "step": 8700
    },
    {
      "epoch": 0.10084286574352799,
      "grad_norm": 0.8437705636024475,
      "learning_rate": 0.0001344396418647731,
      "loss": 0.5561,
      "step": 8710
    },
    {
      "epoch": 0.10095864400500162,
      "grad_norm": 1.5167697668075562,
      "learning_rate": 0.000134594010497067,
      "loss": 0.5879,
      "step": 8720
    },
    {
      "epoch": 0.10107442226647524,
      "grad_norm": 1.2363792657852173,
      "learning_rate": 0.00013474837912936093,
      "loss": 0.5552,
      "step": 8730
    },
    {
      "epoch": 0.10119020052794887,
      "grad_norm": 1.4121090173721313,
      "learning_rate": 0.00013490274776165485,
      "loss": 0.5799,
      "step": 8740
    },
    {
      "epoch": 0.1013059787894225,
      "grad_norm": 1.3154006004333496,
      "learning_rate": 0.00013505711639394876,
      "loss": 0.5816,
      "step": 8750
    },
    {
      "epoch": 0.10142175705089612,
      "grad_norm": 1.5440751314163208,
      "learning_rate": 0.00013521148502624268,
      "loss": 0.5653,
      "step": 8760
    },
    {
      "epoch": 0.10153753531236975,
      "grad_norm": 1.2971959114074707,
      "learning_rate": 0.0001353658536585366,
      "loss": 0.5651,
      "step": 8770
    },
    {
      "epoch": 0.10165331357384337,
      "grad_norm": 1.615846037864685,
      "learning_rate": 0.0001355202222908305,
      "loss": 0.6039,
      "step": 8780
    },
    {
      "epoch": 0.101769091835317,
      "grad_norm": 1.3674498796463013,
      "learning_rate": 0.00013567459092312441,
      "loss": 0.629,
      "step": 8790
    },
    {
      "epoch": 0.10188487009679063,
      "grad_norm": 1.279589056968689,
      "learning_rate": 0.00013582895955541833,
      "loss": 0.5647,
      "step": 8800
    },
    {
      "epoch": 0.10200064835826425,
      "grad_norm": 1.3363662958145142,
      "learning_rate": 0.00013598332818771228,
      "loss": 0.5757,
      "step": 8810
    },
    {
      "epoch": 0.10211642661973788,
      "grad_norm": 1.4299280643463135,
      "learning_rate": 0.0001361376968200062,
      "loss": 0.5597,
      "step": 8820
    },
    {
      "epoch": 0.1022322048812115,
      "grad_norm": 1.4442036151885986,
      "learning_rate": 0.0001362920654523001,
      "loss": 0.587,
      "step": 8830
    },
    {
      "epoch": 0.10234798314268513,
      "grad_norm": 1.5106290578842163,
      "learning_rate": 0.000136446434084594,
      "loss": 0.6177,
      "step": 8840
    },
    {
      "epoch": 0.10246376140415875,
      "grad_norm": 1.4563581943511963,
      "learning_rate": 0.00013660080271688793,
      "loss": 0.5966,
      "step": 8850
    },
    {
      "epoch": 0.10257953966563238,
      "grad_norm": 1.373263955116272,
      "learning_rate": 0.00013675517134918185,
      "loss": 0.594,
      "step": 8860
    },
    {
      "epoch": 0.10269531792710601,
      "grad_norm": 1.3267521858215332,
      "learning_rate": 0.00013690953998147577,
      "loss": 0.5384,
      "step": 8870
    },
    {
      "epoch": 0.10281109618857963,
      "grad_norm": 1.08661949634552,
      "learning_rate": 0.00013706390861376968,
      "loss": 0.59,
      "step": 8880
    },
    {
      "epoch": 0.10292687445005326,
      "grad_norm": 1.2311760187149048,
      "learning_rate": 0.0001372182772460636,
      "loss": 0.5724,
      "step": 8890
    },
    {
      "epoch": 0.10304265271152688,
      "grad_norm": 1.1044796705245972,
      "learning_rate": 0.00013737264587835752,
      "loss": 0.5803,
      "step": 8900
    },
    {
      "epoch": 0.10315843097300051,
      "grad_norm": 1.4755815267562866,
      "learning_rate": 0.00013752701451065144,
      "loss": 0.5938,
      "step": 8910
    },
    {
      "epoch": 0.10327420923447414,
      "grad_norm": 1.1818537712097168,
      "learning_rate": 0.00013768138314294536,
      "loss": 0.6255,
      "step": 8920
    },
    {
      "epoch": 0.10338998749594776,
      "grad_norm": 1.1751924753189087,
      "learning_rate": 0.00013783575177523928,
      "loss": 0.5437,
      "step": 8930
    },
    {
      "epoch": 0.10350576575742139,
      "grad_norm": 1.4293769598007202,
      "learning_rate": 0.00013799012040753317,
      "loss": 0.5531,
      "step": 8940
    },
    {
      "epoch": 0.103621544018895,
      "grad_norm": 1.4072513580322266,
      "learning_rate": 0.00013814448903982712,
      "loss": 0.5822,
      "step": 8950
    },
    {
      "epoch": 0.10373732228036864,
      "grad_norm": 1.3803318738937378,
      "learning_rate": 0.00013829885767212104,
      "loss": 0.5506,
      "step": 8960
    },
    {
      "epoch": 0.10385310054184227,
      "grad_norm": 1.2164582014083862,
      "learning_rate": 0.00013845322630441496,
      "loss": 0.6014,
      "step": 8970
    },
    {
      "epoch": 0.10396887880331589,
      "grad_norm": 1.150188684463501,
      "learning_rate": 0.00013860759493670888,
      "loss": 0.5452,
      "step": 8980
    },
    {
      "epoch": 0.10408465706478952,
      "grad_norm": 1.3069425821304321,
      "learning_rate": 0.0001387619635690028,
      "loss": 0.5472,
      "step": 8990
    },
    {
      "epoch": 0.10420043532626314,
      "grad_norm": 0.8681260347366333,
      "learning_rate": 0.00013891633220129669,
      "loss": 0.514,
      "step": 9000
    },
    {
      "epoch": 0.10420043532626314,
      "eval_chrf": 80.49578165416115,
      "eval_loss": 0.8683168888092041,
      "eval_runtime": 349.3705,
      "eval_samples_per_second": 0.286,
      "eval_steps_per_second": 0.011,
      "step": 9000
    },
    {
      "epoch": 0.10431621358773677,
      "grad_norm": 1.2877256870269775,
      "learning_rate": 0.0001390707008335906,
      "loss": 0.6079,
      "step": 9010
    },
    {
      "epoch": 0.1044319918492104,
      "grad_norm": 1.2810430526733398,
      "learning_rate": 0.00013922506946588455,
      "loss": 0.5683,
      "step": 9020
    },
    {
      "epoch": 0.10454777011068402,
      "grad_norm": 1.2778046131134033,
      "learning_rate": 0.00013937943809817847,
      "loss": 0.5905,
      "step": 9030
    },
    {
      "epoch": 0.10466354837215765,
      "grad_norm": 1.2640587091445923,
      "learning_rate": 0.0001395338067304724,
      "loss": 0.5385,
      "step": 9040
    },
    {
      "epoch": 0.10477932663363126,
      "grad_norm": 1.288314938545227,
      "learning_rate": 0.00013968817536276628,
      "loss": 0.5956,
      "step": 9050
    },
    {
      "epoch": 0.1048951048951049,
      "grad_norm": 1.3123701810836792,
      "learning_rate": 0.0001398425439950602,
      "loss": 0.5521,
      "step": 9060
    },
    {
      "epoch": 0.10501088315657853,
      "grad_norm": 1.1706702709197998,
      "learning_rate": 0.00013999691262735412,
      "loss": 0.5635,
      "step": 9070
    },
    {
      "epoch": 0.10512666141805214,
      "grad_norm": 1.2759759426116943,
      "learning_rate": 0.00014015128125964804,
      "loss": 0.5647,
      "step": 9080
    },
    {
      "epoch": 0.10524243967952578,
      "grad_norm": 1.2104543447494507,
      "learning_rate": 0.00014030564989194196,
      "loss": 0.5641,
      "step": 9090
    },
    {
      "epoch": 0.1053582179409994,
      "grad_norm": 1.4197536706924438,
      "learning_rate": 0.0001404600185242359,
      "loss": 0.608,
      "step": 9100
    },
    {
      "epoch": 0.10547399620247302,
      "grad_norm": 1.3618944883346558,
      "learning_rate": 0.0001406143871565298,
      "loss": 0.5961,
      "step": 9110
    },
    {
      "epoch": 0.10558977446394666,
      "grad_norm": 1.3979140520095825,
      "learning_rate": 0.00014076875578882371,
      "loss": 0.5534,
      "step": 9120
    },
    {
      "epoch": 0.10570555272542027,
      "grad_norm": 1.2438030242919922,
      "learning_rate": 0.00014092312442111763,
      "loss": 0.5645,
      "step": 9130
    },
    {
      "epoch": 0.1058213309868939,
      "grad_norm": 1.3422703742980957,
      "learning_rate": 0.00014107749305341155,
      "loss": 0.6022,
      "step": 9140
    },
    {
      "epoch": 0.10593710924836752,
      "grad_norm": 1.2845385074615479,
      "learning_rate": 0.00014123186168570547,
      "loss": 0.5796,
      "step": 9150
    },
    {
      "epoch": 0.10605288750984115,
      "grad_norm": 1.369936466217041,
      "learning_rate": 0.0001413862303179994,
      "loss": 0.6256,
      "step": 9160
    },
    {
      "epoch": 0.10616866577131477,
      "grad_norm": 1.415257215499878,
      "learning_rate": 0.0001415405989502933,
      "loss": 0.5833,
      "step": 9170
    },
    {
      "epoch": 0.1062844440327884,
      "grad_norm": 1.2637728452682495,
      "learning_rate": 0.00014169496758258723,
      "loss": 0.5591,
      "step": 9180
    },
    {
      "epoch": 0.10640022229426203,
      "grad_norm": 1.3961775302886963,
      "learning_rate": 0.00014184933621488115,
      "loss": 0.5871,
      "step": 9190
    },
    {
      "epoch": 0.10651600055573565,
      "grad_norm": 1.3229669332504272,
      "learning_rate": 0.00014200370484717507,
      "loss": 0.5524,
      "step": 9200
    },
    {
      "epoch": 0.10663177881720928,
      "grad_norm": 1.2066811323165894,
      "learning_rate": 0.00014215807347946899,
      "loss": 0.5421,
      "step": 9210
    },
    {
      "epoch": 0.1067475570786829,
      "grad_norm": 1.528618335723877,
      "learning_rate": 0.00014231244211176288,
      "loss": 0.5878,
      "step": 9220
    },
    {
      "epoch": 0.10686333534015653,
      "grad_norm": 1.3731118440628052,
      "learning_rate": 0.00014246681074405682,
      "loss": 0.5883,
      "step": 9230
    },
    {
      "epoch": 0.10697911360163016,
      "grad_norm": 1.430810570716858,
      "learning_rate": 0.00014262117937635074,
      "loss": 0.571,
      "step": 9240
    },
    {
      "epoch": 0.10709489186310378,
      "grad_norm": 1.2471226453781128,
      "learning_rate": 0.00014277554800864466,
      "loss": 0.5479,
      "step": 9250
    },
    {
      "epoch": 0.10721067012457741,
      "grad_norm": 1.3020845651626587,
      "learning_rate": 0.00014292991664093858,
      "loss": 0.5917,
      "step": 9260
    },
    {
      "epoch": 0.10732644838605103,
      "grad_norm": 1.3037395477294922,
      "learning_rate": 0.00014308428527323247,
      "loss": 0.5787,
      "step": 9270
    },
    {
      "epoch": 0.10744222664752466,
      "grad_norm": 1.2151628732681274,
      "learning_rate": 0.0001432386539055264,
      "loss": 0.5855,
      "step": 9280
    },
    {
      "epoch": 0.10755800490899829,
      "grad_norm": 1.2218369245529175,
      "learning_rate": 0.0001433930225378203,
      "loss": 0.5505,
      "step": 9290
    },
    {
      "epoch": 0.10767378317047191,
      "grad_norm": 1.1271525621414185,
      "learning_rate": 0.00014354739117011423,
      "loss": 0.5586,
      "step": 9300
    },
    {
      "epoch": 0.10778956143194554,
      "grad_norm": 1.2802969217300415,
      "learning_rate": 0.00014370175980240818,
      "loss": 0.5864,
      "step": 9310
    },
    {
      "epoch": 0.10790533969341916,
      "grad_norm": 1.2270424365997314,
      "learning_rate": 0.00014385612843470207,
      "loss": 0.5869,
      "step": 9320
    },
    {
      "epoch": 0.10802111795489279,
      "grad_norm": 1.2666281461715698,
      "learning_rate": 0.000144010497066996,
      "loss": 0.5723,
      "step": 9330
    },
    {
      "epoch": 0.10813689621636642,
      "grad_norm": 1.2750580310821533,
      "learning_rate": 0.0001441648656992899,
      "loss": 0.5624,
      "step": 9340
    },
    {
      "epoch": 0.10825267447784004,
      "grad_norm": 1.2398009300231934,
      "learning_rate": 0.00014431923433158383,
      "loss": 0.5702,
      "step": 9350
    },
    {
      "epoch": 0.10836845273931367,
      "grad_norm": 1.1327173709869385,
      "learning_rate": 0.00014447360296387774,
      "loss": 0.5369,
      "step": 9360
    },
    {
      "epoch": 0.10848423100078729,
      "grad_norm": 1.3858134746551514,
      "learning_rate": 0.00014462797159617166,
      "loss": 0.6502,
      "step": 9370
    },
    {
      "epoch": 0.10860000926226092,
      "grad_norm": 1.3318617343902588,
      "learning_rate": 0.00014478234022846558,
      "loss": 0.5877,
      "step": 9380
    },
    {
      "epoch": 0.10871578752373455,
      "grad_norm": 1.3926358222961426,
      "learning_rate": 0.0001449367088607595,
      "loss": 0.5974,
      "step": 9390
    },
    {
      "epoch": 0.10883156578520817,
      "grad_norm": 1.5860700607299805,
      "learning_rate": 0.00014509107749305342,
      "loss": 0.6003,
      "step": 9400
    },
    {
      "epoch": 0.1089473440466818,
      "grad_norm": 1.2389644384384155,
      "learning_rate": 0.00014524544612534734,
      "loss": 0.602,
      "step": 9410
    },
    {
      "epoch": 0.10906312230815542,
      "grad_norm": 1.2141090631484985,
      "learning_rate": 0.00014539981475764126,
      "loss": 0.537,
      "step": 9420
    },
    {
      "epoch": 0.10917890056962905,
      "grad_norm": 1.2273863554000854,
      "learning_rate": 0.00014555418338993515,
      "loss": 0.5828,
      "step": 9430
    },
    {
      "epoch": 0.10929467883110268,
      "grad_norm": 1.179559588432312,
      "learning_rate": 0.00014570855202222907,
      "loss": 0.6051,
      "step": 9440
    },
    {
      "epoch": 0.1094104570925763,
      "grad_norm": 1.3141292333602905,
      "learning_rate": 0.00014586292065452302,
      "loss": 0.5817,
      "step": 9450
    },
    {
      "epoch": 0.10952623535404993,
      "grad_norm": 1.328954815864563,
      "learning_rate": 0.00014601728928681694,
      "loss": 0.5679,
      "step": 9460
    },
    {
      "epoch": 0.10964201361552355,
      "grad_norm": 1.5461632013320923,
      "learning_rate": 0.00014617165791911085,
      "loss": 0.5483,
      "step": 9470
    },
    {
      "epoch": 0.10975779187699718,
      "grad_norm": 1.4829483032226562,
      "learning_rate": 0.00014632602655140477,
      "loss": 0.5539,
      "step": 9480
    },
    {
      "epoch": 0.1098735701384708,
      "grad_norm": 1.3567897081375122,
      "learning_rate": 0.00014648039518369867,
      "loss": 0.5556,
      "step": 9490
    },
    {
      "epoch": 0.10998934839994443,
      "grad_norm": 1.104142427444458,
      "learning_rate": 0.00014663476381599258,
      "loss": 0.6215,
      "step": 9500
    },
    {
      "epoch": 0.11010512666141806,
      "grad_norm": 1.3122401237487793,
      "learning_rate": 0.0001467891324482865,
      "loss": 0.5685,
      "step": 9510
    },
    {
      "epoch": 0.11022090492289167,
      "grad_norm": 1.4448779821395874,
      "learning_rate": 0.00014694350108058045,
      "loss": 0.5635,
      "step": 9520
    },
    {
      "epoch": 0.1103366831843653,
      "grad_norm": 1.151503324508667,
      "learning_rate": 0.00014709786971287437,
      "loss": 0.5701,
      "step": 9530
    },
    {
      "epoch": 0.11045246144583892,
      "grad_norm": 1.3240145444869995,
      "learning_rate": 0.00014725223834516826,
      "loss": 0.5519,
      "step": 9540
    },
    {
      "epoch": 0.11056823970731255,
      "grad_norm": 1.1416387557983398,
      "learning_rate": 0.00014740660697746218,
      "loss": 0.5548,
      "step": 9550
    },
    {
      "epoch": 0.11068401796878619,
      "grad_norm": 1.630345106124878,
      "learning_rate": 0.0001475609756097561,
      "loss": 0.589,
      "step": 9560
    },
    {
      "epoch": 0.1107997962302598,
      "grad_norm": 1.276701807975769,
      "learning_rate": 0.00014771534424205002,
      "loss": 0.5667,
      "step": 9570
    },
    {
      "epoch": 0.11091557449173343,
      "grad_norm": 0.7283602952957153,
      "learning_rate": 0.00014786971287434394,
      "loss": 0.5594,
      "step": 9580
    },
    {
      "epoch": 0.11103135275320705,
      "grad_norm": 1.2396055459976196,
      "learning_rate": 0.00014802408150663786,
      "loss": 0.5745,
      "step": 9590
    },
    {
      "epoch": 0.11114713101468068,
      "grad_norm": 1.2529735565185547,
      "learning_rate": 0.00014817845013893177,
      "loss": 0.5734,
      "step": 9600
    },
    {
      "epoch": 0.11126290927615431,
      "grad_norm": 1.3989439010620117,
      "learning_rate": 0.0001483328187712257,
      "loss": 0.6088,
      "step": 9610
    },
    {
      "epoch": 0.11137868753762793,
      "grad_norm": 1.3028852939605713,
      "learning_rate": 0.0001484871874035196,
      "loss": 0.5693,
      "step": 9620
    },
    {
      "epoch": 0.11149446579910156,
      "grad_norm": 1.315389633178711,
      "learning_rate": 0.00014864155603581353,
      "loss": 0.5542,
      "step": 9630
    },
    {
      "epoch": 0.11161024406057518,
      "grad_norm": 1.3869110345840454,
      "learning_rate": 0.00014879592466810745,
      "loss": 0.5712,
      "step": 9640
    },
    {
      "epoch": 0.11172602232204881,
      "grad_norm": 1.3050357103347778,
      "learning_rate": 0.00014895029330040134,
      "loss": 0.5527,
      "step": 9650
    },
    {
      "epoch": 0.11184180058352244,
      "grad_norm": 1.2478455305099487,
      "learning_rate": 0.0001491046619326953,
      "loss": 0.5343,
      "step": 9660
    },
    {
      "epoch": 0.11195757884499606,
      "grad_norm": 1.173696756362915,
      "learning_rate": 0.0001492590305649892,
      "loss": 0.5837,
      "step": 9670
    },
    {
      "epoch": 0.11207335710646969,
      "grad_norm": 1.0571203231811523,
      "learning_rate": 0.00014941339919728313,
      "loss": 0.5847,
      "step": 9680
    },
    {
      "epoch": 0.11218913536794331,
      "grad_norm": 1.2881155014038086,
      "learning_rate": 0.00014956776782957705,
      "loss": 0.5816,
      "step": 9690
    },
    {
      "epoch": 0.11230491362941694,
      "grad_norm": 1.19301438331604,
      "learning_rate": 0.00014972213646187097,
      "loss": 0.5027,
      "step": 9700
    },
    {
      "epoch": 0.11242069189089057,
      "grad_norm": 0.8694561123847961,
      "learning_rate": 0.00014987650509416486,
      "loss": 0.5323,
      "step": 9710
    },
    {
      "epoch": 0.11253647015236419,
      "grad_norm": 1.2321244478225708,
      "learning_rate": 0.00015003087372645878,
      "loss": 0.5573,
      "step": 9720
    },
    {
      "epoch": 0.11265224841383782,
      "grad_norm": 1.078788161277771,
      "learning_rate": 0.00015018524235875272,
      "loss": 0.5648,
      "step": 9730
    },
    {
      "epoch": 0.11276802667531144,
      "grad_norm": 0.8772777915000916,
      "learning_rate": 0.00015033961099104664,
      "loss": 0.4924,
      "step": 9740
    },
    {
      "epoch": 0.11288380493678507,
      "grad_norm": 1.4024667739868164,
      "learning_rate": 0.00015049397962334056,
      "loss": 0.5817,
      "step": 9750
    },
    {
      "epoch": 0.1129995831982587,
      "grad_norm": 1.1767064332962036,
      "learning_rate": 0.00015064834825563445,
      "loss": 0.5338,
      "step": 9760
    },
    {
      "epoch": 0.11311536145973232,
      "grad_norm": 1.348203420639038,
      "learning_rate": 0.00015080271688792837,
      "loss": 0.5618,
      "step": 9770
    },
    {
      "epoch": 0.11323113972120595,
      "grad_norm": 1.195645809173584,
      "learning_rate": 0.0001509570855202223,
      "loss": 0.5786,
      "step": 9780
    },
    {
      "epoch": 0.11334691798267957,
      "grad_norm": 1.1947705745697021,
      "learning_rate": 0.0001511114541525162,
      "loss": 0.5669,
      "step": 9790
    },
    {
      "epoch": 0.1134626962441532,
      "grad_norm": 1.0327008962631226,
      "learning_rate": 0.00015126582278481013,
      "loss": 0.5822,
      "step": 9800
    },
    {
      "epoch": 0.11357847450562682,
      "grad_norm": 1.0953099727630615,
      "learning_rate": 0.00015142019141710407,
      "loss": 0.5642,
      "step": 9810
    },
    {
      "epoch": 0.11369425276710045,
      "grad_norm": 1.1357226371765137,
      "learning_rate": 0.00015157456004939797,
      "loss": 0.5204,
      "step": 9820
    },
    {
      "epoch": 0.11381003102857408,
      "grad_norm": 1.1709736585617065,
      "learning_rate": 0.00015172892868169189,
      "loss": 0.5532,
      "step": 9830
    },
    {
      "epoch": 0.1139258092900477,
      "grad_norm": 1.457765817642212,
      "learning_rate": 0.0001518832973139858,
      "loss": 0.5645,
      "step": 9840
    },
    {
      "epoch": 0.11404158755152133,
      "grad_norm": 1.2417305707931519,
      "learning_rate": 0.00015203766594627972,
      "loss": 0.5925,
      "step": 9850
    },
    {
      "epoch": 0.11415736581299495,
      "grad_norm": 1.1331074237823486,
      "learning_rate": 0.00015219203457857364,
      "loss": 0.5902,
      "step": 9860
    },
    {
      "epoch": 0.11427314407446858,
      "grad_norm": 1.1287226676940918,
      "learning_rate": 0.00015234640321086756,
      "loss": 0.5959,
      "step": 9870
    },
    {
      "epoch": 0.11438892233594221,
      "grad_norm": 1.2500649690628052,
      "learning_rate": 0.00015250077184316148,
      "loss": 0.6099,
      "step": 9880
    },
    {
      "epoch": 0.11450470059741583,
      "grad_norm": 1.2514734268188477,
      "learning_rate": 0.0001526551404754554,
      "loss": 0.541,
      "step": 9890
    },
    {
      "epoch": 0.11462047885888946,
      "grad_norm": 1.2987282276153564,
      "learning_rate": 0.00015280950910774932,
      "loss": 0.5874,
      "step": 9900
    },
    {
      "epoch": 0.11473625712036307,
      "grad_norm": 1.2376899719238281,
      "learning_rate": 0.00015296387774004324,
      "loss": 0.5443,
      "step": 9910
    },
    {
      "epoch": 0.1148520353818367,
      "grad_norm": 1.0586546659469604,
      "learning_rate": 0.00015311824637233713,
      "loss": 0.5576,
      "step": 9920
    },
    {
      "epoch": 0.11496781364331034,
      "grad_norm": 1.2568001747131348,
      "learning_rate": 0.00015327261500463105,
      "loss": 0.56,
      "step": 9930
    },
    {
      "epoch": 0.11508359190478396,
      "grad_norm": 1.0285547971725464,
      "learning_rate": 0.00015342698363692497,
      "loss": 0.563,
      "step": 9940
    },
    {
      "epoch": 0.11519937016625759,
      "grad_norm": 1.4542441368103027,
      "learning_rate": 0.00015358135226921891,
      "loss": 0.5849,
      "step": 9950
    },
    {
      "epoch": 0.1153151484277312,
      "grad_norm": 1.1740046739578247,
      "learning_rate": 0.00015373572090151283,
      "loss": 0.5787,
      "step": 9960
    },
    {
      "epoch": 0.11543092668920484,
      "grad_norm": 1.2496739625930786,
      "learning_rate": 0.00015389008953380675,
      "loss": 0.5311,
      "step": 9970
    },
    {
      "epoch": 0.11554670495067847,
      "grad_norm": 1.2663365602493286,
      "learning_rate": 0.00015404445816610064,
      "loss": 0.5589,
      "step": 9980
    },
    {
      "epoch": 0.11566248321215208,
      "grad_norm": 1.1600028276443481,
      "learning_rate": 0.00015419882679839456,
      "loss": 0.5149,
      "step": 9990
    },
    {
      "epoch": 0.11577826147362572,
      "grad_norm": 1.1629509925842285,
      "learning_rate": 0.00015435319543068848,
      "loss": 0.5537,
      "step": 10000
    },
    {
      "epoch": 0.11577826147362572,
      "eval_chrf": 84.638197697528,
      "eval_loss": 0.8278533816337585,
      "eval_runtime": 130.0577,
      "eval_samples_per_second": 0.769,
      "eval_steps_per_second": 0.031,
      "step": 10000
    },
    {
      "epoch": 0.11589403973509933,
      "grad_norm": 1.1861213445663452,
      "learning_rate": 0.0001545075640629824,
      "loss": 0.5576,
      "step": 10010
    },
    {
      "epoch": 0.11600981799657296,
      "grad_norm": 1.050795316696167,
      "learning_rate": 0.00015466193269527635,
      "loss": 0.5668,
      "step": 10020
    },
    {
      "epoch": 0.1161255962580466,
      "grad_norm": 1.1185582876205444,
      "learning_rate": 0.00015481630132757024,
      "loss": 0.5434,
      "step": 10030
    },
    {
      "epoch": 0.11624137451952021,
      "grad_norm": 1.116737961769104,
      "learning_rate": 0.00015497066995986416,
      "loss": 0.5336,
      "step": 10040
    },
    {
      "epoch": 0.11635715278099384,
      "grad_norm": 0.8800601959228516,
      "learning_rate": 0.00015512503859215808,
      "loss": 0.565,
      "step": 10050
    },
    {
      "epoch": 0.11647293104246746,
      "grad_norm": 1.2967404127120972,
      "learning_rate": 0.000155279407224452,
      "loss": 0.5465,
      "step": 10060
    },
    {
      "epoch": 0.1165887093039411,
      "grad_norm": 1.1198583841323853,
      "learning_rate": 0.00015543377585674592,
      "loss": 0.5601,
      "step": 10070
    },
    {
      "epoch": 0.11670448756541472,
      "grad_norm": 1.3123164176940918,
      "learning_rate": 0.00015558814448903983,
      "loss": 0.5623,
      "step": 10080
    },
    {
      "epoch": 0.11682026582688834,
      "grad_norm": 1.3212881088256836,
      "learning_rate": 0.00015574251312133375,
      "loss": 0.5587,
      "step": 10090
    },
    {
      "epoch": 0.11693604408836197,
      "grad_norm": 1.1772260665893555,
      "learning_rate": 0.00015589688175362767,
      "loss": 0.5482,
      "step": 10100
    },
    {
      "epoch": 0.11705182234983559,
      "grad_norm": 1.3611674308776855,
      "learning_rate": 0.0001560512503859216,
      "loss": 0.5797,
      "step": 10110
    },
    {
      "epoch": 0.11716760061130922,
      "grad_norm": 1.2402210235595703,
      "learning_rate": 0.0001562056190182155,
      "loss": 0.5826,
      "step": 10120
    },
    {
      "epoch": 0.11728337887278284,
      "grad_norm": 1.1671757698059082,
      "learning_rate": 0.00015635998765050943,
      "loss": 0.5527,
      "step": 10130
    },
    {
      "epoch": 0.11739915713425647,
      "grad_norm": 1.2186027765274048,
      "learning_rate": 0.00015651435628280332,
      "loss": 0.5107,
      "step": 10140
    },
    {
      "epoch": 0.1175149353957301,
      "grad_norm": 0.8878567814826965,
      "learning_rate": 0.00015666872491509724,
      "loss": 0.5451,
      "step": 10150
    },
    {
      "epoch": 0.11763071365720372,
      "grad_norm": 1.2072186470031738,
      "learning_rate": 0.0001568230935473912,
      "loss": 0.575,
      "step": 10160
    },
    {
      "epoch": 0.11774649191867735,
      "grad_norm": 1.0611921548843384,
      "learning_rate": 0.0001569774621796851,
      "loss": 0.5637,
      "step": 10170
    },
    {
      "epoch": 0.11786227018015097,
      "grad_norm": 1.129486083984375,
      "learning_rate": 0.00015713183081197902,
      "loss": 0.56,
      "step": 10180
    },
    {
      "epoch": 0.1179780484416246,
      "grad_norm": 1.0836197137832642,
      "learning_rate": 0.00015728619944427294,
      "loss": 0.5332,
      "step": 10190
    },
    {
      "epoch": 0.11809382670309823,
      "grad_norm": 1.149590015411377,
      "learning_rate": 0.00015744056807656684,
      "loss": 0.5665,
      "step": 10200
    },
    {
      "epoch": 0.11820960496457185,
      "grad_norm": 1.039841651916504,
      "learning_rate": 0.00015759493670886075,
      "loss": 0.5666,
      "step": 10210
    },
    {
      "epoch": 0.11832538322604548,
      "grad_norm": 1.2359282970428467,
      "learning_rate": 0.00015774930534115467,
      "loss": 0.5774,
      "step": 10220
    },
    {
      "epoch": 0.1184411614875191,
      "grad_norm": 1.1911756992340088,
      "learning_rate": 0.00015790367397344862,
      "loss": 0.6242,
      "step": 10230
    },
    {
      "epoch": 0.11855693974899273,
      "grad_norm": 1.0651416778564453,
      "learning_rate": 0.00015805804260574254,
      "loss": 0.5214,
      "step": 10240
    },
    {
      "epoch": 0.11867271801046636,
      "grad_norm": 1.1892896890640259,
      "learning_rate": 0.00015821241123803643,
      "loss": 0.5544,
      "step": 10250
    },
    {
      "epoch": 0.11878849627193998,
      "grad_norm": 1.1992381811141968,
      "learning_rate": 0.00015836677987033035,
      "loss": 0.5596,
      "step": 10260
    },
    {
      "epoch": 0.11890427453341361,
      "grad_norm": 1.320581316947937,
      "learning_rate": 0.00015852114850262427,
      "loss": 0.5642,
      "step": 10270
    },
    {
      "epoch": 0.11902005279488723,
      "grad_norm": 1.2349812984466553,
      "learning_rate": 0.0001586755171349182,
      "loss": 0.5277,
      "step": 10280
    },
    {
      "epoch": 0.11913583105636086,
      "grad_norm": 1.3728326559066772,
      "learning_rate": 0.0001588298857672121,
      "loss": 0.5688,
      "step": 10290
    },
    {
      "epoch": 0.11925160931783449,
      "grad_norm": 1.085245966911316,
      "learning_rate": 0.00015898425439950603,
      "loss": 0.5629,
      "step": 10300
    },
    {
      "epoch": 0.1193673875793081,
      "grad_norm": 1.0424472093582153,
      "learning_rate": 0.00015913862303179995,
      "loss": 0.5813,
      "step": 10310
    },
    {
      "epoch": 0.11948316584078174,
      "grad_norm": 1.1915884017944336,
      "learning_rate": 0.00015929299166409386,
      "loss": 0.5753,
      "step": 10320
    },
    {
      "epoch": 0.11959894410225536,
      "grad_norm": 1.059067964553833,
      "learning_rate": 0.00015944736029638778,
      "loss": 0.5296,
      "step": 10330
    },
    {
      "epoch": 0.11971472236372899,
      "grad_norm": 0.9762371778488159,
      "learning_rate": 0.0001596017289286817,
      "loss": 0.5748,
      "step": 10340
    },
    {
      "epoch": 0.11983050062520262,
      "grad_norm": 1.168055772781372,
      "learning_rate": 0.00015975609756097562,
      "loss": 0.5447,
      "step": 10350
    },
    {
      "epoch": 0.11994627888667624,
      "grad_norm": 1.1109033823013306,
      "learning_rate": 0.0001599104661932695,
      "loss": 0.5771,
      "step": 10360
    },
    {
      "epoch": 0.12006205714814987,
      "grad_norm": 1.4880404472351074,
      "learning_rate": 0.00016006483482556346,
      "loss": 0.5418,
      "step": 10370
    },
    {
      "epoch": 0.12017783540962348,
      "grad_norm": 1.2616000175476074,
      "learning_rate": 0.00016021920345785738,
      "loss": 0.5306,
      "step": 10380
    },
    {
      "epoch": 0.12029361367109712,
      "grad_norm": 1.141715407371521,
      "learning_rate": 0.0001603735720901513,
      "loss": 0.5536,
      "step": 10390
    },
    {
      "epoch": 0.12040939193257075,
      "grad_norm": 1.0811337232589722,
      "learning_rate": 0.00016052794072244522,
      "loss": 0.5698,
      "step": 10400
    },
    {
      "epoch": 0.12052517019404436,
      "grad_norm": 1.1806341409683228,
      "learning_rate": 0.00016068230935473914,
      "loss": 0.5538,
      "step": 10410
    },
    {
      "epoch": 0.120640948455518,
      "grad_norm": 1.2402247190475464,
      "learning_rate": 0.00016083667798703303,
      "loss": 0.5628,
      "step": 10420
    },
    {
      "epoch": 0.12075672671699161,
      "grad_norm": 1.0643714666366577,
      "learning_rate": 0.00016099104661932695,
      "loss": 0.5335,
      "step": 10430
    },
    {
      "epoch": 0.12087250497846524,
      "grad_norm": 1.2642136812210083,
      "learning_rate": 0.00016114541525162087,
      "loss": 0.5176,
      "step": 10440
    },
    {
      "epoch": 0.12098828323993886,
      "grad_norm": 1.1404954195022583,
      "learning_rate": 0.0001612997838839148,
      "loss": 0.5596,
      "step": 10450
    },
    {
      "epoch": 0.1211040615014125,
      "grad_norm": 1.1933493614196777,
      "learning_rate": 0.00016145415251620873,
      "loss": 0.5763,
      "step": 10460
    },
    {
      "epoch": 0.12121983976288613,
      "grad_norm": 1.1639156341552734,
      "learning_rate": 0.00016160852114850262,
      "loss": 0.5465,
      "step": 10470
    },
    {
      "epoch": 0.12133561802435974,
      "grad_norm": 1.1753767728805542,
      "learning_rate": 0.00016176288978079654,
      "loss": 0.5268,
      "step": 10480
    },
    {
      "epoch": 0.12145139628583337,
      "grad_norm": 1.1199126243591309,
      "learning_rate": 0.00016191725841309046,
      "loss": 0.5464,
      "step": 10490
    },
    {
      "epoch": 0.12156717454730699,
      "grad_norm": 1.1204979419708252,
      "learning_rate": 0.00016207162704538438,
      "loss": 0.5632,
      "step": 10500
    },
    {
      "epoch": 0.12168295280878062,
      "grad_norm": 1.3289235830307007,
      "learning_rate": 0.0001622259956776783,
      "loss": 0.5738,
      "step": 10510
    },
    {
      "epoch": 0.12179873107025425,
      "grad_norm": 1.3842880725860596,
      "learning_rate": 0.00016238036430997222,
      "loss": 0.5596,
      "step": 10520
    },
    {
      "epoch": 0.12191450933172787,
      "grad_norm": 1.0078904628753662,
      "learning_rate": 0.00016253473294226614,
      "loss": 0.5113,
      "step": 10530
    },
    {
      "epoch": 0.1220302875932015,
      "grad_norm": 1.1998320817947388,
      "learning_rate": 0.00016268910157456006,
      "loss": 0.524,
      "step": 10540
    },
    {
      "epoch": 0.12214606585467512,
      "grad_norm": 1.0711064338684082,
      "learning_rate": 0.00016284347020685398,
      "loss": 0.5834,
      "step": 10550
    },
    {
      "epoch": 0.12226184411614875,
      "grad_norm": 1.2888844013214111,
      "learning_rate": 0.0001629978388391479,
      "loss": 0.5644,
      "step": 10560
    },
    {
      "epoch": 0.12237762237762238,
      "grad_norm": 1.1922492980957031,
      "learning_rate": 0.0001631522074714418,
      "loss": 0.5798,
      "step": 10570
    },
    {
      "epoch": 0.122493400639096,
      "grad_norm": 1.1588951349258423,
      "learning_rate": 0.00016330657610373573,
      "loss": 0.5432,
      "step": 10580
    },
    {
      "epoch": 0.12260917890056963,
      "grad_norm": 1.0158624649047852,
      "learning_rate": 0.00016346094473602965,
      "loss": 0.5374,
      "step": 10590
    },
    {
      "epoch": 0.12272495716204325,
      "grad_norm": 1.221661925315857,
      "learning_rate": 0.00016361531336832357,
      "loss": 0.6158,
      "step": 10600
    },
    {
      "epoch": 0.12284073542351688,
      "grad_norm": 1.276640772819519,
      "learning_rate": 0.0001637696820006175,
      "loss": 0.5408,
      "step": 10610
    },
    {
      "epoch": 0.12295651368499051,
      "grad_norm": 0.9623101949691772,
      "learning_rate": 0.0001639240506329114,
      "loss": 0.562,
      "step": 10620
    },
    {
      "epoch": 0.12307229194646413,
      "grad_norm": 1.1899003982543945,
      "learning_rate": 0.0001640784192652053,
      "loss": 0.577,
      "step": 10630
    },
    {
      "epoch": 0.12318807020793776,
      "grad_norm": 1.1083171367645264,
      "learning_rate": 0.00016423278789749922,
      "loss": 0.5248,
      "step": 10640
    },
    {
      "epoch": 0.12330384846941138,
      "grad_norm": 0.945651650428772,
      "learning_rate": 0.00016438715652979314,
      "loss": 0.5576,
      "step": 10650
    },
    {
      "epoch": 0.12341962673088501,
      "grad_norm": 1.1229653358459473,
      "learning_rate": 0.00016454152516208708,
      "loss": 0.5506,
      "step": 10660
    },
    {
      "epoch": 0.12353540499235864,
      "grad_norm": 1.126285433769226,
      "learning_rate": 0.000164695893794381,
      "loss": 0.5586,
      "step": 10670
    },
    {
      "epoch": 0.12365118325383226,
      "grad_norm": 1.1712100505828857,
      "learning_rate": 0.00016485026242667492,
      "loss": 0.5512,
      "step": 10680
    },
    {
      "epoch": 0.12376696151530589,
      "grad_norm": 1.2397513389587402,
      "learning_rate": 0.00016500463105896881,
      "loss": 0.5823,
      "step": 10690
    },
    {
      "epoch": 0.12388273977677951,
      "grad_norm": 1.1539273262023926,
      "learning_rate": 0.00016515899969126273,
      "loss": 0.5426,
      "step": 10700
    },
    {
      "epoch": 0.12399851803825314,
      "grad_norm": 1.3479893207550049,
      "learning_rate": 0.00016531336832355665,
      "loss": 0.5407,
      "step": 10710
    },
    {
      "epoch": 0.12411429629972677,
      "grad_norm": 1.0497949123382568,
      "learning_rate": 0.00016546773695585057,
      "loss": 0.5763,
      "step": 10720
    },
    {
      "epoch": 0.12423007456120039,
      "grad_norm": 1.5127085447311401,
      "learning_rate": 0.00016562210558814452,
      "loss": 0.5126,
      "step": 10730
    },
    {
      "epoch": 0.12434585282267402,
      "grad_norm": 1.3668838739395142,
      "learning_rate": 0.0001657764742204384,
      "loss": 0.5827,
      "step": 10740
    },
    {
      "epoch": 0.12446163108414764,
      "grad_norm": 1.3154889345169067,
      "learning_rate": 0.00016593084285273233,
      "loss": 0.5621,
      "step": 10750
    },
    {
      "epoch": 0.12457740934562127,
      "grad_norm": 1.0343090295791626,
      "learning_rate": 0.00016608521148502625,
      "loss": 0.5287,
      "step": 10760
    },
    {
      "epoch": 0.12469318760709489,
      "grad_norm": 1.1013846397399902,
      "learning_rate": 0.00016623958011732017,
      "loss": 0.5613,
      "step": 10770
    },
    {
      "epoch": 0.12480896586856852,
      "grad_norm": 1.0979907512664795,
      "learning_rate": 0.00016639394874961409,
      "loss": 0.569,
      "step": 10780
    },
    {
      "epoch": 0.12492474413004215,
      "grad_norm": 1.4586269855499268,
      "learning_rate": 0.000166548317381908,
      "loss": 0.5544,
      "step": 10790
    },
    {
      "epoch": 0.12504052239151578,
      "grad_norm": 0.8446677327156067,
      "learning_rate": 0.00016670268601420192,
      "loss": 0.5613,
      "step": 10800
    },
    {
      "epoch": 0.12515630065298938,
      "grad_norm": 1.0182301998138428,
      "learning_rate": 0.00016685705464649584,
      "loss": 0.597,
      "step": 10810
    },
    {
      "epoch": 0.12527207891446301,
      "grad_norm": 1.5233309268951416,
      "learning_rate": 0.00016701142327878976,
      "loss": 0.597,
      "step": 10820
    },
    {
      "epoch": 0.12538785717593665,
      "grad_norm": 1.3310496807098389,
      "learning_rate": 0.00016716579191108368,
      "loss": 0.5362,
      "step": 10830
    },
    {
      "epoch": 0.12550363543741028,
      "grad_norm": 1.1259944438934326,
      "learning_rate": 0.0001673201605433776,
      "loss": 0.5224,
      "step": 10840
    },
    {
      "epoch": 0.1256194136988839,
      "grad_norm": 0.9475477337837219,
      "learning_rate": 0.0001674745291756715,
      "loss": 0.5637,
      "step": 10850
    },
    {
      "epoch": 0.1257351919603575,
      "grad_norm": 0.7506548762321472,
      "learning_rate": 0.0001676288978079654,
      "loss": 0.5621,
      "step": 10860
    },
    {
      "epoch": 0.12585097022183114,
      "grad_norm": 1.293733835220337,
      "learning_rate": 0.00016778326644025936,
      "loss": 0.6133,
      "step": 10870
    },
    {
      "epoch": 0.12596674848330477,
      "grad_norm": 1.132787823677063,
      "learning_rate": 0.00016793763507255328,
      "loss": 0.5535,
      "step": 10880
    },
    {
      "epoch": 0.1260825267447784,
      "grad_norm": 0.9605393409729004,
      "learning_rate": 0.0001680920037048472,
      "loss": 0.5293,
      "step": 10890
    },
    {
      "epoch": 0.12619830500625204,
      "grad_norm": 1.193603754043579,
      "learning_rate": 0.00016824637233714111,
      "loss": 0.5348,
      "step": 10900
    },
    {
      "epoch": 0.12631408326772564,
      "grad_norm": 0.9524046778678894,
      "learning_rate": 0.000168400740969435,
      "loss": 0.5466,
      "step": 10910
    },
    {
      "epoch": 0.12642986152919927,
      "grad_norm": 0.9687754511833191,
      "learning_rate": 0.00016855510960172893,
      "loss": 0.5615,
      "step": 10920
    },
    {
      "epoch": 0.1265456397906729,
      "grad_norm": 1.1673355102539062,
      "learning_rate": 0.00016870947823402284,
      "loss": 0.5191,
      "step": 10930
    },
    {
      "epoch": 0.12666141805214653,
      "grad_norm": 1.1563589572906494,
      "learning_rate": 0.00016886384686631676,
      "loss": 0.5525,
      "step": 10940
    },
    {
      "epoch": 0.12677719631362017,
      "grad_norm": 1.19508695602417,
      "learning_rate": 0.0001690182154986107,
      "loss": 0.5689,
      "step": 10950
    },
    {
      "epoch": 0.12689297457509377,
      "grad_norm": 1.2190064191818237,
      "learning_rate": 0.0001691725841309046,
      "loss": 0.5316,
      "step": 10960
    },
    {
      "epoch": 0.1270087528365674,
      "grad_norm": 1.0793712139129639,
      "learning_rate": 0.00016932695276319852,
      "loss": 0.5668,
      "step": 10970
    },
    {
      "epoch": 0.12712453109804103,
      "grad_norm": 1.0726120471954346,
      "learning_rate": 0.00016948132139549244,
      "loss": 0.5692,
      "step": 10980
    },
    {
      "epoch": 0.12724030935951466,
      "grad_norm": 1.1050915718078613,
      "learning_rate": 0.00016963569002778636,
      "loss": 0.542,
      "step": 10990
    },
    {
      "epoch": 0.1273560876209883,
      "grad_norm": 1.142502784729004,
      "learning_rate": 0.00016979005866008028,
      "loss": 0.5347,
      "step": 11000
    },
    {
      "epoch": 0.1273560876209883,
      "eval_chrf": 83.23927244399603,
      "eval_loss": 0.8144679665565491,
      "eval_runtime": 266.851,
      "eval_samples_per_second": 0.375,
      "eval_steps_per_second": 0.015,
      "step": 11000
    },
    {
      "epoch": 0.1274718658824619,
      "grad_norm": 1.0333619117736816,
      "learning_rate": 0.0001699444272923742,
      "loss": 0.5571,
      "step": 11010
    },
    {
      "epoch": 0.12758764414393553,
      "grad_norm": 1.0597230195999146,
      "learning_rate": 0.00017009879592466812,
      "loss": 0.508,
      "step": 11020
    },
    {
      "epoch": 0.12770342240540916,
      "grad_norm": 1.0315881967544556,
      "learning_rate": 0.00017025316455696204,
      "loss": 0.5524,
      "step": 11030
    },
    {
      "epoch": 0.1278192006668828,
      "grad_norm": 1.2254493236541748,
      "learning_rate": 0.00017040753318925595,
      "loss": 0.5297,
      "step": 11040
    },
    {
      "epoch": 0.12793497892835642,
      "grad_norm": 0.9213074445724487,
      "learning_rate": 0.00017056190182154987,
      "loss": 0.5336,
      "step": 11050
    },
    {
      "epoch": 0.12805075718983003,
      "grad_norm": 1.309411644935608,
      "learning_rate": 0.0001707162704538438,
      "loss": 0.5068,
      "step": 11060
    },
    {
      "epoch": 0.12816653545130366,
      "grad_norm": 1.152342677116394,
      "learning_rate": 0.00017087063908613768,
      "loss": 0.5822,
      "step": 11070
    },
    {
      "epoch": 0.1282823137127773,
      "grad_norm": 1.1705478429794312,
      "learning_rate": 0.00017102500771843163,
      "loss": 0.532,
      "step": 11080
    },
    {
      "epoch": 0.12839809197425092,
      "grad_norm": 1.048901915550232,
      "learning_rate": 0.00017117937635072555,
      "loss": 0.4915,
      "step": 11090
    },
    {
      "epoch": 0.12851387023572455,
      "grad_norm": 1.1591132879257202,
      "learning_rate": 0.00017133374498301947,
      "loss": 0.577,
      "step": 11100
    },
    {
      "epoch": 0.12862964849719816,
      "grad_norm": 0.8415403962135315,
      "learning_rate": 0.0001714881136153134,
      "loss": 0.5413,
      "step": 11110
    },
    {
      "epoch": 0.1287454267586718,
      "grad_norm": 1.2463173866271973,
      "learning_rate": 0.00017164248224760728,
      "loss": 0.5185,
      "step": 11120
    },
    {
      "epoch": 0.12886120502014542,
      "grad_norm": 1.1211379766464233,
      "learning_rate": 0.0001717968508799012,
      "loss": 0.5642,
      "step": 11130
    },
    {
      "epoch": 0.12897698328161905,
      "grad_norm": 0.8294934034347534,
      "learning_rate": 0.00017195121951219512,
      "loss": 0.5478,
      "step": 11140
    },
    {
      "epoch": 0.12909276154309268,
      "grad_norm": 0.9989081025123596,
      "learning_rate": 0.00017210558814448904,
      "loss": 0.5407,
      "step": 11150
    },
    {
      "epoch": 0.12920853980456629,
      "grad_norm": 1.027762532234192,
      "learning_rate": 0.00017225995677678298,
      "loss": 0.5974,
      "step": 11160
    },
    {
      "epoch": 0.12932431806603992,
      "grad_norm": 1.0650486946105957,
      "learning_rate": 0.0001724143254090769,
      "loss": 0.5158,
      "step": 11170
    },
    {
      "epoch": 0.12944009632751355,
      "grad_norm": 1.0328350067138672,
      "learning_rate": 0.0001725686940413708,
      "loss": 0.5319,
      "step": 11180
    },
    {
      "epoch": 0.12955587458898718,
      "grad_norm": 1.0063869953155518,
      "learning_rate": 0.0001727230626736647,
      "loss": 0.5116,
      "step": 11190
    },
    {
      "epoch": 0.1296716528504608,
      "grad_norm": 1.168185830116272,
      "learning_rate": 0.00017287743130595863,
      "loss": 0.5596,
      "step": 11200
    },
    {
      "epoch": 0.12978743111193441,
      "grad_norm": 1.115496277809143,
      "learning_rate": 0.00017303179993825255,
      "loss": 0.5578,
      "step": 11210
    },
    {
      "epoch": 0.12990320937340805,
      "grad_norm": 1.2830357551574707,
      "learning_rate": 0.00017318616857054647,
      "loss": 0.5645,
      "step": 11220
    },
    {
      "epoch": 0.13001898763488168,
      "grad_norm": 1.221888780593872,
      "learning_rate": 0.0001733405372028404,
      "loss": 0.5344,
      "step": 11230
    },
    {
      "epoch": 0.1301347658963553,
      "grad_norm": 1.1060632467269897,
      "learning_rate": 0.0001734949058351343,
      "loss": 0.5318,
      "step": 11240
    },
    {
      "epoch": 0.1302505441578289,
      "grad_norm": 0.777014970779419,
      "learning_rate": 0.00017364927446742823,
      "loss": 0.5489,
      "step": 11250
    },
    {
      "epoch": 0.13036632241930254,
      "grad_norm": 1.0885053873062134,
      "learning_rate": 0.00017380364309972215,
      "loss": 0.5687,
      "step": 11260
    },
    {
      "epoch": 0.13048210068077618,
      "grad_norm": 1.1854456663131714,
      "learning_rate": 0.00017395801173201607,
      "loss": 0.5516,
      "step": 11270
    },
    {
      "epoch": 0.1305978789422498,
      "grad_norm": 1.1444556713104248,
      "learning_rate": 0.00017411238036430998,
      "loss": 0.5766,
      "step": 11280
    },
    {
      "epoch": 0.13071365720372344,
      "grad_norm": 1.273538589477539,
      "learning_rate": 0.0001742667489966039,
      "loss": 0.5706,
      "step": 11290
    },
    {
      "epoch": 0.13082943546519704,
      "grad_norm": 0.9996747970581055,
      "learning_rate": 0.00017442111762889782,
      "loss": 0.525,
      "step": 11300
    },
    {
      "epoch": 0.13094521372667067,
      "grad_norm": 1.1024609804153442,
      "learning_rate": 0.00017457548626119174,
      "loss": 0.5166,
      "step": 11310
    },
    {
      "epoch": 0.1310609919881443,
      "grad_norm": 1.111045002937317,
      "learning_rate": 0.00017472985489348566,
      "loss": 0.5581,
      "step": 11320
    },
    {
      "epoch": 0.13117677024961794,
      "grad_norm": 1.2413759231567383,
      "learning_rate": 0.00017488422352577958,
      "loss": 0.5832,
      "step": 11330
    },
    {
      "epoch": 0.13129254851109157,
      "grad_norm": 1.2619898319244385,
      "learning_rate": 0.00017503859215807347,
      "loss": 0.5329,
      "step": 11340
    },
    {
      "epoch": 0.13140832677256517,
      "grad_norm": 1.0059072971343994,
      "learning_rate": 0.0001751929607903674,
      "loss": 0.5782,
      "step": 11350
    },
    {
      "epoch": 0.1315241050340388,
      "grad_norm": 1.025973916053772,
      "learning_rate": 0.0001753473294226613,
      "loss": 0.5213,
      "step": 11360
    },
    {
      "epoch": 0.13163988329551243,
      "grad_norm": 1.1603498458862305,
      "learning_rate": 0.00017550169805495526,
      "loss": 0.5222,
      "step": 11370
    },
    {
      "epoch": 0.13175566155698606,
      "grad_norm": 1.2103770971298218,
      "learning_rate": 0.00017565606668724917,
      "loss": 0.5256,
      "step": 11380
    },
    {
      "epoch": 0.1318714398184597,
      "grad_norm": 1.1276682615280151,
      "learning_rate": 0.0001758104353195431,
      "loss": 0.5853,
      "step": 11390
    },
    {
      "epoch": 0.1319872180799333,
      "grad_norm": 1.136440634727478,
      "learning_rate": 0.00017596480395183699,
      "loss": 0.5244,
      "step": 11400
    },
    {
      "epoch": 0.13210299634140693,
      "grad_norm": 1.136427402496338,
      "learning_rate": 0.0001761191725841309,
      "loss": 0.5478,
      "step": 11410
    },
    {
      "epoch": 0.13221877460288056,
      "grad_norm": 1.0788636207580566,
      "learning_rate": 0.00017627354121642482,
      "loss": 0.5525,
      "step": 11420
    },
    {
      "epoch": 0.1323345528643542,
      "grad_norm": 1.2557777166366577,
      "learning_rate": 0.00017642790984871874,
      "loss": 0.5372,
      "step": 11430
    },
    {
      "epoch": 0.13245033112582782,
      "grad_norm": 1.040418267250061,
      "learning_rate": 0.00017658227848101266,
      "loss": 0.5575,
      "step": 11440
    },
    {
      "epoch": 0.13256610938730143,
      "grad_norm": 1.2515392303466797,
      "learning_rate": 0.00017673664711330658,
      "loss": 0.5553,
      "step": 11450
    },
    {
      "epoch": 0.13268188764877506,
      "grad_norm": 0.915041983127594,
      "learning_rate": 0.0001768910157456005,
      "loss": 0.5841,
      "step": 11460
    },
    {
      "epoch": 0.1327976659102487,
      "grad_norm": 1.2221473455429077,
      "learning_rate": 0.00017704538437789442,
      "loss": 0.5139,
      "step": 11470
    },
    {
      "epoch": 0.13291344417172232,
      "grad_norm": 0.9782427549362183,
      "learning_rate": 0.00017719975301018834,
      "loss": 0.504,
      "step": 11480
    },
    {
      "epoch": 0.13302922243319595,
      "grad_norm": 1.1323063373565674,
      "learning_rate": 0.00017735412164248226,
      "loss": 0.5084,
      "step": 11490
    },
    {
      "epoch": 0.13314500069466956,
      "grad_norm": 1.3384109735488892,
      "learning_rate": 0.00017750849027477618,
      "loss": 0.53,
      "step": 11500
    },
    {
      "epoch": 0.1332607789561432,
      "grad_norm": 0.8835853338241577,
      "learning_rate": 0.0001776628589070701,
      "loss": 0.5243,
      "step": 11510
    },
    {
      "epoch": 0.13337655721761682,
      "grad_norm": 1.2791578769683838,
      "learning_rate": 0.00017781722753936401,
      "loss": 0.5791,
      "step": 11520
    },
    {
      "epoch": 0.13349233547909045,
      "grad_norm": 0.806826651096344,
      "learning_rate": 0.00017797159617165793,
      "loss": 0.5318,
      "step": 11530
    },
    {
      "epoch": 0.13360811374056408,
      "grad_norm": 1.084456443786621,
      "learning_rate": 0.00017812596480395185,
      "loss": 0.5219,
      "step": 11540
    },
    {
      "epoch": 0.1337238920020377,
      "grad_norm": 1.0938810110092163,
      "learning_rate": 0.00017828033343624577,
      "loss": 0.528,
      "step": 11550
    },
    {
      "epoch": 0.13383967026351132,
      "grad_norm": 0.9293761849403381,
      "learning_rate": 0.00017843470206853966,
      "loss": 0.4871,
      "step": 11560
    },
    {
      "epoch": 0.13395544852498495,
      "grad_norm": 1.3259512186050415,
      "learning_rate": 0.00017858907070083358,
      "loss": 0.5218,
      "step": 11570
    },
    {
      "epoch": 0.13407122678645858,
      "grad_norm": 1.0393692255020142,
      "learning_rate": 0.00017874343933312753,
      "loss": 0.5824,
      "step": 11580
    },
    {
      "epoch": 0.1341870050479322,
      "grad_norm": 1.091562271118164,
      "learning_rate": 0.00017889780796542145,
      "loss": 0.5312,
      "step": 11590
    },
    {
      "epoch": 0.13430278330940582,
      "grad_norm": 0.9571833610534668,
      "learning_rate": 0.00017905217659771537,
      "loss": 0.4809,
      "step": 11600
    },
    {
      "epoch": 0.13441856157087945,
      "grad_norm": 0.9122828841209412,
      "learning_rate": 0.00017920654523000929,
      "loss": 0.5268,
      "step": 11610
    },
    {
      "epoch": 0.13453433983235308,
      "grad_norm": 1.106863021850586,
      "learning_rate": 0.00017936091386230318,
      "loss": 0.4979,
      "step": 11620
    },
    {
      "epoch": 0.1346501180938267,
      "grad_norm": 0.9412614107131958,
      "learning_rate": 0.0001795152824945971,
      "loss": 0.5449,
      "step": 11630
    },
    {
      "epoch": 0.13476589635530034,
      "grad_norm": 1.068676233291626,
      "learning_rate": 0.00017966965112689102,
      "loss": 0.5392,
      "step": 11640
    },
    {
      "epoch": 0.13488167461677394,
      "grad_norm": 0.9199798703193665,
      "learning_rate": 0.00017982401975918493,
      "loss": 0.553,
      "step": 11650
    },
    {
      "epoch": 0.13499745287824758,
      "grad_norm": 1.0406941175460815,
      "learning_rate": 0.00017997838839147888,
      "loss": 0.5119,
      "step": 11660
    },
    {
      "epoch": 0.1351132311397212,
      "grad_norm": 1.0165780782699585,
      "learning_rate": 0.00018013275702377277,
      "loss": 0.5964,
      "step": 11670
    },
    {
      "epoch": 0.13522900940119484,
      "grad_norm": 1.1605157852172852,
      "learning_rate": 0.0001802871256560667,
      "loss": 0.5357,
      "step": 11680
    },
    {
      "epoch": 0.13534478766266847,
      "grad_norm": 1.2324901819229126,
      "learning_rate": 0.0001804414942883606,
      "loss": 0.5654,
      "step": 11690
    },
    {
      "epoch": 0.13546056592414207,
      "grad_norm": 1.203242301940918,
      "learning_rate": 0.00018059586292065453,
      "loss": 0.546,
      "step": 11700
    },
    {
      "epoch": 0.1355763441856157,
      "grad_norm": 0.955553412437439,
      "learning_rate": 0.00018075023155294845,
      "loss": 0.5792,
      "step": 11710
    },
    {
      "epoch": 0.13569212244708934,
      "grad_norm": 1.0696978569030762,
      "learning_rate": 0.00018090460018524237,
      "loss": 0.497,
      "step": 11720
    },
    {
      "epoch": 0.13580790070856297,
      "grad_norm": 0.9858132004737854,
      "learning_rate": 0.0001810589688175363,
      "loss": 0.5201,
      "step": 11730
    },
    {
      "epoch": 0.1359236789700366,
      "grad_norm": 1.1289206743240356,
      "learning_rate": 0.0001812133374498302,
      "loss": 0.5504,
      "step": 11740
    },
    {
      "epoch": 0.1360394572315102,
      "grad_norm": 0.9448925852775574,
      "learning_rate": 0.00018136770608212412,
      "loss": 0.5174,
      "step": 11750
    },
    {
      "epoch": 0.13615523549298383,
      "grad_norm": 0.9734740853309631,
      "learning_rate": 0.00018152207471441804,
      "loss": 0.4766,
      "step": 11760
    },
    {
      "epoch": 0.13627101375445747,
      "grad_norm": 1.3037446737289429,
      "learning_rate": 0.00018167644334671196,
      "loss": 0.5263,
      "step": 11770
    },
    {
      "epoch": 0.1363867920159311,
      "grad_norm": 1.0228137969970703,
      "learning_rate": 0.00018183081197900585,
      "loss": 0.5461,
      "step": 11780
    },
    {
      "epoch": 0.13650257027740473,
      "grad_norm": 1.0178076028823853,
      "learning_rate": 0.0001819851806112998,
      "loss": 0.5633,
      "step": 11790
    },
    {
      "epoch": 0.13661834853887833,
      "grad_norm": 1.0332564115524292,
      "learning_rate": 0.00018213954924359372,
      "loss": 0.5081,
      "step": 11800
    },
    {
      "epoch": 0.13673412680035196,
      "grad_norm": 1.2683144807815552,
      "learning_rate": 0.00018229391787588764,
      "loss": 0.5323,
      "step": 11810
    },
    {
      "epoch": 0.1368499050618256,
      "grad_norm": 1.0793180465698242,
      "learning_rate": 0.00018244828650818156,
      "loss": 0.5571,
      "step": 11820
    },
    {
      "epoch": 0.13696568332329923,
      "grad_norm": 1.2837165594100952,
      "learning_rate": 0.00018260265514047545,
      "loss": 0.5191,
      "step": 11830
    },
    {
      "epoch": 0.13708146158477286,
      "grad_norm": 1.0644558668136597,
      "learning_rate": 0.00018275702377276937,
      "loss": 0.5665,
      "step": 11840
    },
    {
      "epoch": 0.13719723984624646,
      "grad_norm": 0.9577881097793579,
      "learning_rate": 0.0001829113924050633,
      "loss": 0.5169,
      "step": 11850
    },
    {
      "epoch": 0.1373130181077201,
      "grad_norm": 1.0442769527435303,
      "learning_rate": 0.0001830657610373572,
      "loss": 0.5538,
      "step": 11860
    },
    {
      "epoch": 0.13742879636919372,
      "grad_norm": 1.006024718284607,
      "learning_rate": 0.00018322012966965115,
      "loss": 0.5206,
      "step": 11870
    },
    {
      "epoch": 0.13754457463066735,
      "grad_norm": 1.124611735343933,
      "learning_rate": 0.00018337449830194507,
      "loss": 0.5528,
      "step": 11880
    },
    {
      "epoch": 0.13766035289214096,
      "grad_norm": 1.016611933708191,
      "learning_rate": 0.00018352886693423896,
      "loss": 0.514,
      "step": 11890
    },
    {
      "epoch": 0.1377761311536146,
      "grad_norm": 1.1806875467300415,
      "learning_rate": 0.00018368323556653288,
      "loss": 0.5298,
      "step": 11900
    },
    {
      "epoch": 0.13789190941508822,
      "grad_norm": 1.2107545137405396,
      "learning_rate": 0.0001838376041988268,
      "loss": 0.5286,
      "step": 11910
    },
    {
      "epoch": 0.13800768767656185,
      "grad_norm": 1.073745608329773,
      "learning_rate": 0.00018399197283112072,
      "loss": 0.5396,
      "step": 11920
    },
    {
      "epoch": 0.13812346593803548,
      "grad_norm": 0.9077839851379395,
      "learning_rate": 0.00018414634146341464,
      "loss": 0.5423,
      "step": 11930
    },
    {
      "epoch": 0.1382392441995091,
      "grad_norm": 0.799449622631073,
      "learning_rate": 0.00018430071009570856,
      "loss": 0.5247,
      "step": 11940
    },
    {
      "epoch": 0.13835502246098272,
      "grad_norm": 0.9549907445907593,
      "learning_rate": 0.00018445507872800248,
      "loss": 0.521,
      "step": 11950
    },
    {
      "epoch": 0.13847080072245635,
      "grad_norm": 1.1224545240402222,
      "learning_rate": 0.0001846094473602964,
      "loss": 0.5819,
      "step": 11960
    },
    {
      "epoch": 0.13858657898392998,
      "grad_norm": 1.1244319677352905,
      "learning_rate": 0.00018476381599259032,
      "loss": 0.5349,
      "step": 11970
    },
    {
      "epoch": 0.1387023572454036,
      "grad_norm": 0.9261943101882935,
      "learning_rate": 0.00018491818462488424,
      "loss": 0.5422,
      "step": 11980
    },
    {
      "epoch": 0.13881813550687722,
      "grad_norm": 0.9949973225593567,
      "learning_rate": 0.00018507255325717815,
      "loss": 0.5169,
      "step": 11990
    },
    {
      "epoch": 0.13893391376835085,
      "grad_norm": 1.7659807205200195,
      "learning_rate": 0.00018522692188947207,
      "loss": 0.536,
      "step": 12000
    },
    {
      "epoch": 0.13893391376835085,
      "eval_chrf": 82.54549731700665,
      "eval_loss": 0.8094607591629028,
      "eval_runtime": 202.044,
      "eval_samples_per_second": 0.495,
      "eval_steps_per_second": 0.02,
      "step": 12000
    },
    {
      "epoch": 0.13904969202982448,
      "grad_norm": 1.013721227645874,
      "learning_rate": 0.000185381290521766,
      "loss": 0.534,
      "step": 12010
    },
    {
      "epoch": 0.1391654702912981,
      "grad_norm": 0.9654140472412109,
      "learning_rate": 0.0001855356591540599,
      "loss": 0.5169,
      "step": 12020
    },
    {
      "epoch": 0.13928124855277174,
      "grad_norm": 1.178539514541626,
      "learning_rate": 0.00018569002778635383,
      "loss": 0.5652,
      "step": 12030
    },
    {
      "epoch": 0.13939702681424535,
      "grad_norm": 1.261973261833191,
      "learning_rate": 0.00018584439641864775,
      "loss": 0.4845,
      "step": 12040
    },
    {
      "epoch": 0.13951280507571898,
      "grad_norm": 0.8742697834968567,
      "learning_rate": 0.00018599876505094164,
      "loss": 0.527,
      "step": 12050
    },
    {
      "epoch": 0.1396285833371926,
      "grad_norm": 1.3083447217941284,
      "learning_rate": 0.00018615313368323556,
      "loss": 0.5299,
      "step": 12060
    },
    {
      "epoch": 0.13974436159866624,
      "grad_norm": 1.1984888315200806,
      "learning_rate": 0.00018630750231552948,
      "loss": 0.5648,
      "step": 12070
    },
    {
      "epoch": 0.13986013986013987,
      "grad_norm": 0.8685519099235535,
      "learning_rate": 0.00018646187094782343,
      "loss": 0.539,
      "step": 12080
    },
    {
      "epoch": 0.13997591812161347,
      "grad_norm": 0.8454098701477051,
      "learning_rate": 0.00018661623958011735,
      "loss": 0.5362,
      "step": 12090
    },
    {
      "epoch": 0.1400916963830871,
      "grad_norm": 1.0055502653121948,
      "learning_rate": 0.00018677060821241126,
      "loss": 0.5422,
      "step": 12100
    },
    {
      "epoch": 0.14020747464456074,
      "grad_norm": 1.48152494430542,
      "learning_rate": 0.00018692497684470516,
      "loss": 0.5189,
      "step": 12110
    },
    {
      "epoch": 0.14032325290603437,
      "grad_norm": 0.9812230467796326,
      "learning_rate": 0.00018707934547699908,
      "loss": 0.5244,
      "step": 12120
    },
    {
      "epoch": 0.140439031167508,
      "grad_norm": 1.140608310699463,
      "learning_rate": 0.000187233714109293,
      "loss": 0.5373,
      "step": 12130
    },
    {
      "epoch": 0.1405548094289816,
      "grad_norm": 0.9877954721450806,
      "learning_rate": 0.0001873880827415869,
      "loss": 0.5195,
      "step": 12140
    },
    {
      "epoch": 0.14067058769045523,
      "grad_norm": 1.1157594919204712,
      "learning_rate": 0.00018754245137388083,
      "loss": 0.5606,
      "step": 12150
    },
    {
      "epoch": 0.14078636595192887,
      "grad_norm": 0.8598747849464417,
      "learning_rate": 0.00018769682000617475,
      "loss": 0.5366,
      "step": 12160
    },
    {
      "epoch": 0.1409021442134025,
      "grad_norm": 0.6583921909332275,
      "learning_rate": 0.00018785118863846867,
      "loss": 0.5138,
      "step": 12170
    },
    {
      "epoch": 0.14101792247487613,
      "grad_norm": 1.0767642259597778,
      "learning_rate": 0.0001880055572707626,
      "loss": 0.538,
      "step": 12180
    },
    {
      "epoch": 0.14113370073634973,
      "grad_norm": 1.048838496208191,
      "learning_rate": 0.0001881599259030565,
      "loss": 0.5153,
      "step": 12190
    },
    {
      "epoch": 0.14124947899782336,
      "grad_norm": 1.035570502281189,
      "learning_rate": 0.00018831429453535043,
      "loss": 0.5404,
      "step": 12200
    },
    {
      "epoch": 0.141365257259297,
      "grad_norm": 1.0647659301757812,
      "learning_rate": 0.00018846866316764432,
      "loss": 0.5562,
      "step": 12210
    },
    {
      "epoch": 0.14148103552077063,
      "grad_norm": 0.991861879825592,
      "learning_rate": 0.00018862303179993827,
      "loss": 0.5005,
      "step": 12220
    },
    {
      "epoch": 0.14159681378224426,
      "grad_norm": 0.977424681186676,
      "learning_rate": 0.00018877740043223218,
      "loss": 0.5177,
      "step": 12230
    },
    {
      "epoch": 0.14171259204371786,
      "grad_norm": 1.1738947629928589,
      "learning_rate": 0.0001889317690645261,
      "loss": 0.5283,
      "step": 12240
    },
    {
      "epoch": 0.1418283703051915,
      "grad_norm": 1.055560827255249,
      "learning_rate": 0.00018908613769682002,
      "loss": 0.5506,
      "step": 12250
    },
    {
      "epoch": 0.14194414856666512,
      "grad_norm": 1.1133482456207275,
      "learning_rate": 0.00018924050632911394,
      "loss": 0.5264,
      "step": 12260
    },
    {
      "epoch": 0.14205992682813876,
      "grad_norm": 1.02242112159729,
      "learning_rate": 0.00018939487496140783,
      "loss": 0.4913,
      "step": 12270
    },
    {
      "epoch": 0.1421757050896124,
      "grad_norm": 0.967154860496521,
      "learning_rate": 0.00018954924359370175,
      "loss": 0.5375,
      "step": 12280
    },
    {
      "epoch": 0.142291483351086,
      "grad_norm": 1.0422741174697876,
      "learning_rate": 0.0001897036122259957,
      "loss": 0.5827,
      "step": 12290
    },
    {
      "epoch": 0.14240726161255962,
      "grad_norm": 1.1207256317138672,
      "learning_rate": 0.00018985798085828962,
      "loss": 0.5528,
      "step": 12300
    },
    {
      "epoch": 0.14252303987403325,
      "grad_norm": 1.1637494564056396,
      "learning_rate": 0.00019001234949058354,
      "loss": 0.5411,
      "step": 12310
    },
    {
      "epoch": 0.14263881813550688,
      "grad_norm": 0.9992698431015015,
      "learning_rate": 0.00019016671812287743,
      "loss": 0.575,
      "step": 12320
    },
    {
      "epoch": 0.14275459639698052,
      "grad_norm": 1.29230535030365,
      "learning_rate": 0.00019032108675517135,
      "loss": 0.5392,
      "step": 12330
    },
    {
      "epoch": 0.14287037465845412,
      "grad_norm": 0.9479355812072754,
      "learning_rate": 0.00019047545538746527,
      "loss": 0.5542,
      "step": 12340
    },
    {
      "epoch": 0.14298615291992775,
      "grad_norm": 1.1067039966583252,
      "learning_rate": 0.00019062982401975919,
      "loss": 0.531,
      "step": 12350
    },
    {
      "epoch": 0.14310193118140138,
      "grad_norm": 1.1938244104385376,
      "learning_rate": 0.0001907841926520531,
      "loss": 0.5431,
      "step": 12360
    },
    {
      "epoch": 0.143217709442875,
      "grad_norm": 1.029172420501709,
      "learning_rate": 0.00019093856128434705,
      "loss": 0.52,
      "step": 12370
    },
    {
      "epoch": 0.14333348770434864,
      "grad_norm": 1.0918617248535156,
      "learning_rate": 0.00019109292991664094,
      "loss": 0.5157,
      "step": 12380
    },
    {
      "epoch": 0.14344926596582225,
      "grad_norm": 1.1285583972930908,
      "learning_rate": 0.00019124729854893486,
      "loss": 0.5182,
      "step": 12390
    },
    {
      "epoch": 0.14356504422729588,
      "grad_norm": 0.9844357371330261,
      "learning_rate": 0.00019140166718122878,
      "loss": 0.5348,
      "step": 12400
    },
    {
      "epoch": 0.1436808224887695,
      "grad_norm": 0.8017510175704956,
      "learning_rate": 0.0001915560358135227,
      "loss": 0.535,
      "step": 12410
    },
    {
      "epoch": 0.14379660075024314,
      "grad_norm": 1.1872317790985107,
      "learning_rate": 0.00019171040444581662,
      "loss": 0.5435,
      "step": 12420
    },
    {
      "epoch": 0.14391237901171677,
      "grad_norm": 1.1137135028839111,
      "learning_rate": 0.00019186477307811054,
      "loss": 0.5706,
      "step": 12430
    },
    {
      "epoch": 0.14402815727319038,
      "grad_norm": 0.9512019157409668,
      "learning_rate": 0.00019201914171040446,
      "loss": 0.5228,
      "step": 12440
    },
    {
      "epoch": 0.144143935534664,
      "grad_norm": 1.0688989162445068,
      "learning_rate": 0.00019217351034269838,
      "loss": 0.5394,
      "step": 12450
    },
    {
      "epoch": 0.14425971379613764,
      "grad_norm": 1.0268332958221436,
      "learning_rate": 0.0001923278789749923,
      "loss": 0.5156,
      "step": 12460
    },
    {
      "epoch": 0.14437549205761127,
      "grad_norm": 1.5388939380645752,
      "learning_rate": 0.00019248224760728621,
      "loss": 0.5755,
      "step": 12470
    },
    {
      "epoch": 0.1444912703190849,
      "grad_norm": 0.9914690852165222,
      "learning_rate": 0.00019263661623958013,
      "loss": 0.548,
      "step": 12480
    },
    {
      "epoch": 0.1446070485805585,
      "grad_norm": 0.959027886390686,
      "learning_rate": 0.00019279098487187403,
      "loss": 0.5403,
      "step": 12490
    },
    {
      "epoch": 0.14472282684203214,
      "grad_norm": 0.9708199501037598,
      "learning_rate": 0.00019294535350416797,
      "loss": 0.5218,
      "step": 12500
    },
    {
      "epoch": 0.14483860510350577,
      "grad_norm": 1.0253245830535889,
      "learning_rate": 0.0001930997221364619,
      "loss": 0.549,
      "step": 12510
    },
    {
      "epoch": 0.1449543833649794,
      "grad_norm": 1.1119471788406372,
      "learning_rate": 0.0001932540907687558,
      "loss": 0.5643,
      "step": 12520
    },
    {
      "epoch": 0.145070161626453,
      "grad_norm": 0.9286425709724426,
      "learning_rate": 0.00019340845940104973,
      "loss": 0.515,
      "step": 12530
    },
    {
      "epoch": 0.14518593988792663,
      "grad_norm": 0.9852153062820435,
      "learning_rate": 0.00019356282803334362,
      "loss": 0.5124,
      "step": 12540
    },
    {
      "epoch": 0.14530171814940027,
      "grad_norm": 1.0258785486221313,
      "learning_rate": 0.00019371719666563754,
      "loss": 0.498,
      "step": 12550
    },
    {
      "epoch": 0.1454174964108739,
      "grad_norm": 1.0606341361999512,
      "learning_rate": 0.00019387156529793146,
      "loss": 0.5348,
      "step": 12560
    },
    {
      "epoch": 0.14553327467234753,
      "grad_norm": 0.9994125366210938,
      "learning_rate": 0.00019402593393022538,
      "loss": 0.5179,
      "step": 12570
    },
    {
      "epoch": 0.14564905293382113,
      "grad_norm": 0.9378136396408081,
      "learning_rate": 0.00019418030256251932,
      "loss": 0.5007,
      "step": 12580
    },
    {
      "epoch": 0.14576483119529476,
      "grad_norm": 0.981959342956543,
      "learning_rate": 0.00019433467119481324,
      "loss": 0.541,
      "step": 12590
    },
    {
      "epoch": 0.1458806094567684,
      "grad_norm": 1.0490880012512207,
      "learning_rate": 0.00019448903982710714,
      "loss": 0.5489,
      "step": 12600
    },
    {
      "epoch": 0.14599638771824203,
      "grad_norm": 1.0558048486709595,
      "learning_rate": 0.00019464340845940105,
      "loss": 0.5072,
      "step": 12610
    },
    {
      "epoch": 0.14611216597971566,
      "grad_norm": 1.2084742784500122,
      "learning_rate": 0.00019479777709169497,
      "loss": 0.5364,
      "step": 12620
    },
    {
      "epoch": 0.14622794424118926,
      "grad_norm": 1.0054224729537964,
      "learning_rate": 0.0001949521457239889,
      "loss": 0.5353,
      "step": 12630
    },
    {
      "epoch": 0.1463437225026629,
      "grad_norm": 1.053541898727417,
      "learning_rate": 0.0001951065143562828,
      "loss": 0.5115,
      "step": 12640
    },
    {
      "epoch": 0.14645950076413652,
      "grad_norm": 1.075671911239624,
      "learning_rate": 0.00019526088298857673,
      "loss": 0.548,
      "step": 12650
    },
    {
      "epoch": 0.14657527902561016,
      "grad_norm": 1.0095752477645874,
      "learning_rate": 0.00019541525162087065,
      "loss": 0.5233,
      "step": 12660
    },
    {
      "epoch": 0.1466910572870838,
      "grad_norm": 1.2185146808624268,
      "learning_rate": 0.00019556962025316457,
      "loss": 0.5293,
      "step": 12670
    },
    {
      "epoch": 0.1468068355485574,
      "grad_norm": 0.9857822060585022,
      "learning_rate": 0.0001957239888854585,
      "loss": 0.5244,
      "step": 12680
    },
    {
      "epoch": 0.14692261381003102,
      "grad_norm": 1.0182740688323975,
      "learning_rate": 0.0001958783575177524,
      "loss": 0.5435,
      "step": 12690
    },
    {
      "epoch": 0.14703839207150465,
      "grad_norm": 0.9528821706771851,
      "learning_rate": 0.00019603272615004633,
      "loss": 0.5242,
      "step": 12700
    },
    {
      "epoch": 0.14715417033297828,
      "grad_norm": 0.8967875838279724,
      "learning_rate": 0.00019618709478234022,
      "loss": 0.5303,
      "step": 12710
    },
    {
      "epoch": 0.14726994859445192,
      "grad_norm": 0.9793334603309631,
      "learning_rate": 0.00019634146341463416,
      "loss": 0.529,
      "step": 12720
    },
    {
      "epoch": 0.14738572685592552,
      "grad_norm": 1.0513063669204712,
      "learning_rate": 0.00019649583204692808,
      "loss": 0.5534,
      "step": 12730
    },
    {
      "epoch": 0.14750150511739915,
      "grad_norm": 1.0110771656036377,
      "learning_rate": 0.000196650200679222,
      "loss": 0.5449,
      "step": 12740
    },
    {
      "epoch": 0.14761728337887278,
      "grad_norm": 1.05746328830719,
      "learning_rate": 0.00019680456931151592,
      "loss": 0.5395,
      "step": 12750
    },
    {
      "epoch": 0.1477330616403464,
      "grad_norm": 0.9189900755882263,
      "learning_rate": 0.0001969589379438098,
      "loss": 0.5224,
      "step": 12760
    },
    {
      "epoch": 0.14784883990182004,
      "grad_norm": 1.0486233234405518,
      "learning_rate": 0.00019711330657610373,
      "loss": 0.5275,
      "step": 12770
    },
    {
      "epoch": 0.14796461816329365,
      "grad_norm": 0.9569796323776245,
      "learning_rate": 0.00019726767520839765,
      "loss": 0.4871,
      "step": 12780
    },
    {
      "epoch": 0.14808039642476728,
      "grad_norm": 0.894376277923584,
      "learning_rate": 0.0001974220438406916,
      "loss": 0.5224,
      "step": 12790
    },
    {
      "epoch": 0.1481961746862409,
      "grad_norm": 0.9812209010124207,
      "learning_rate": 0.00019757641247298552,
      "loss": 0.5305,
      "step": 12800
    },
    {
      "epoch": 0.14831195294771454,
      "grad_norm": 1.0676387548446655,
      "learning_rate": 0.0001977307811052794,
      "loss": 0.5152,
      "step": 12810
    },
    {
      "epoch": 0.14842773120918817,
      "grad_norm": 0.9315291047096252,
      "learning_rate": 0.00019788514973757333,
      "loss": 0.5546,
      "step": 12820
    },
    {
      "epoch": 0.14854350947066178,
      "grad_norm": 0.7768070697784424,
      "learning_rate": 0.00019803951836986725,
      "loss": 0.5038,
      "step": 12830
    },
    {
      "epoch": 0.1486592877321354,
      "grad_norm": 0.975273609161377,
      "learning_rate": 0.00019819388700216116,
      "loss": 0.5401,
      "step": 12840
    },
    {
      "epoch": 0.14877506599360904,
      "grad_norm": 1.1329939365386963,
      "learning_rate": 0.00019834825563445508,
      "loss": 0.5497,
      "step": 12850
    },
    {
      "epoch": 0.14889084425508267,
      "grad_norm": 1.0359381437301636,
      "learning_rate": 0.000198502624266749,
      "loss": 0.4876,
      "step": 12860
    },
    {
      "epoch": 0.1490066225165563,
      "grad_norm": 0.9096277952194214,
      "learning_rate": 0.00019865699289904292,
      "loss": 0.4975,
      "step": 12870
    },
    {
      "epoch": 0.1491224007780299,
      "grad_norm": 0.8534004092216492,
      "learning_rate": 0.00019881136153133684,
      "loss": 0.5196,
      "step": 12880
    },
    {
      "epoch": 0.14923817903950354,
      "grad_norm": 1.071838140487671,
      "learning_rate": 0.00019896573016363076,
      "loss": 0.5364,
      "step": 12890
    },
    {
      "epoch": 0.14935395730097717,
      "grad_norm": 0.9981087446212769,
      "learning_rate": 0.00019912009879592468,
      "loss": 0.5281,
      "step": 12900
    },
    {
      "epoch": 0.1494697355624508,
      "grad_norm": 0.8559280633926392,
      "learning_rate": 0.0001992744674282186,
      "loss": 0.4899,
      "step": 12910
    },
    {
      "epoch": 0.14958551382392443,
      "grad_norm": 1.156166434288025,
      "learning_rate": 0.0001994288360605125,
      "loss": 0.5433,
      "step": 12920
    },
    {
      "epoch": 0.14970129208539804,
      "grad_norm": 0.9986305832862854,
      "learning_rate": 0.00019958320469280644,
      "loss": 0.5355,
      "step": 12930
    },
    {
      "epoch": 0.14981707034687167,
      "grad_norm": 1.0870987176895142,
      "learning_rate": 0.00019973757332510036,
      "loss": 0.5211,
      "step": 12940
    },
    {
      "epoch": 0.1499328486083453,
      "grad_norm": 0.9895318746566772,
      "learning_rate": 0.00019989194195739427,
      "loss": 0.5423,
      "step": 12950
    },
    {
      "epoch": 0.15004862686981893,
      "grad_norm": 1.019115924835205,
      "learning_rate": 0.0001999999999267045,
      "loss": 0.4864,
      "step": 12960
    },
    {
      "epoch": 0.15016440513129256,
      "grad_norm": 0.9224790930747986,
      "learning_rate": 0.00019999999862367344,
      "loss": 0.5464,
      "step": 12970
    },
    {
      "epoch": 0.15028018339276616,
      "grad_norm": 1.016212821006775,
      "learning_rate": 0.0001999999956918536,
      "loss": 0.5633,
      "step": 12980
    },
    {
      "epoch": 0.1503959616542398,
      "grad_norm": 0.9151487946510315,
      "learning_rate": 0.00019999999113124502,
      "loss": 0.53,
      "step": 12990
    },
    {
      "epoch": 0.15051173991571343,
      "grad_norm": 1.353114128112793,
      "learning_rate": 0.00019999998494184777,
      "loss": 0.5318,
      "step": 13000
    },
    {
      "epoch": 0.15051173991571343,
      "eval_chrf": 83.9289537905008,
      "eval_loss": 0.7918225526809692,
      "eval_runtime": 211.6443,
      "eval_samples_per_second": 0.472,
      "eval_steps_per_second": 0.019,
      "step": 13000
    },
    {
      "epoch": 0.15062751817718706,
      "grad_norm": 1.2324007749557495,
      "learning_rate": 0.00019999997712366195,
      "loss": 0.5601,
      "step": 13010
    },
    {
      "epoch": 0.1507432964386607,
      "grad_norm": 1.0476680994033813,
      "learning_rate": 0.00019999996767668766,
      "loss": 0.5061,
      "step": 13020
    },
    {
      "epoch": 0.1508590747001343,
      "grad_norm": 0.9604538083076477,
      "learning_rate": 0.00019999995660092512,
      "loss": 0.4906,
      "step": 13030
    },
    {
      "epoch": 0.15097485296160792,
      "grad_norm": 0.952919602394104,
      "learning_rate": 0.00019999994389637444,
      "loss": 0.522,
      "step": 13040
    },
    {
      "epoch": 0.15109063122308156,
      "grad_norm": 0.9822264909744263,
      "learning_rate": 0.0001999999295630359,
      "loss": 0.5136,
      "step": 13050
    },
    {
      "epoch": 0.1512064094845552,
      "grad_norm": 1.010438323020935,
      "learning_rate": 0.00019999991360090963,
      "loss": 0.5169,
      "step": 13060
    },
    {
      "epoch": 0.15132218774602882,
      "grad_norm": 1.0127623081207275,
      "learning_rate": 0.000199999896009996,
      "loss": 0.5399,
      "step": 13070
    },
    {
      "epoch": 0.15143796600750242,
      "grad_norm": 0.9422155022621155,
      "learning_rate": 0.00019999987679029523,
      "loss": 0.5439,
      "step": 13080
    },
    {
      "epoch": 0.15155374426897605,
      "grad_norm": 0.8040795922279358,
      "learning_rate": 0.00019999985594180766,
      "loss": 0.5234,
      "step": 13090
    },
    {
      "epoch": 0.15166952253044969,
      "grad_norm": 1.2307194471359253,
      "learning_rate": 0.0001999998334645336,
      "loss": 0.4947,
      "step": 13100
    },
    {
      "epoch": 0.15178530079192332,
      "grad_norm": 0.9566722512245178,
      "learning_rate": 0.0001999998093584735,
      "loss": 0.5329,
      "step": 13110
    },
    {
      "epoch": 0.15190107905339695,
      "grad_norm": 0.9066735506057739,
      "learning_rate": 0.00019999978362362762,
      "loss": 0.5071,
      "step": 13120
    },
    {
      "epoch": 0.15201685731487055,
      "grad_norm": 1.259695291519165,
      "learning_rate": 0.0001999997562599965,
      "loss": 0.547,
      "step": 13130
    },
    {
      "epoch": 0.15213263557634418,
      "grad_norm": 0.8511571884155273,
      "learning_rate": 0.0001999997272675805,
      "loss": 0.4821,
      "step": 13140
    },
    {
      "epoch": 0.15224841383781781,
      "grad_norm": 0.9403756260871887,
      "learning_rate": 0.00019999969664638014,
      "loss": 0.5393,
      "step": 13150
    },
    {
      "epoch": 0.15236419209929145,
      "grad_norm": 1.1131285429000854,
      "learning_rate": 0.0001999996643963959,
      "loss": 0.5155,
      "step": 13160
    },
    {
      "epoch": 0.15247997036076505,
      "grad_norm": 0.9601148962974548,
      "learning_rate": 0.0001999996305176283,
      "loss": 0.5146,
      "step": 13170
    },
    {
      "epoch": 0.15259574862223868,
      "grad_norm": 0.798585832118988,
      "learning_rate": 0.00019999959501007793,
      "loss": 0.5366,
      "step": 13180
    },
    {
      "epoch": 0.1527115268837123,
      "grad_norm": 1.0919116735458374,
      "learning_rate": 0.00019999955787374535,
      "loss": 0.5241,
      "step": 13190
    },
    {
      "epoch": 0.15282730514518594,
      "grad_norm": 0.9964574575424194,
      "learning_rate": 0.0001999995191086311,
      "loss": 0.5363,
      "step": 13200
    },
    {
      "epoch": 0.15294308340665957,
      "grad_norm": 1.0141183137893677,
      "learning_rate": 0.00019999947871473592,
      "loss": 0.5157,
      "step": 13210
    },
    {
      "epoch": 0.15305886166813318,
      "grad_norm": 1.152384877204895,
      "learning_rate": 0.0001999994366920604,
      "loss": 0.5328,
      "step": 13220
    },
    {
      "epoch": 0.1531746399296068,
      "grad_norm": 1.1605803966522217,
      "learning_rate": 0.00019999939304060524,
      "loss": 0.5297,
      "step": 13230
    },
    {
      "epoch": 0.15329041819108044,
      "grad_norm": 0.8750791549682617,
      "learning_rate": 0.00019999934776037117,
      "loss": 0.5387,
      "step": 13240
    },
    {
      "epoch": 0.15340619645255407,
      "grad_norm": 0.986000120639801,
      "learning_rate": 0.00019999930085135887,
      "loss": 0.5055,
      "step": 13250
    },
    {
      "epoch": 0.1535219747140277,
      "grad_norm": 0.9871428608894348,
      "learning_rate": 0.00019999925231356917,
      "loss": 0.525,
      "step": 13260
    },
    {
      "epoch": 0.1536377529755013,
      "grad_norm": 1.1668473482131958,
      "learning_rate": 0.00019999920214700283,
      "loss": 0.5097,
      "step": 13270
    },
    {
      "epoch": 0.15375353123697494,
      "grad_norm": 0.807448148727417,
      "learning_rate": 0.00019999915035166067,
      "loss": 0.5416,
      "step": 13280
    },
    {
      "epoch": 0.15386930949844857,
      "grad_norm": 0.9240652918815613,
      "learning_rate": 0.00019999909692754354,
      "loss": 0.5622,
      "step": 13290
    },
    {
      "epoch": 0.1539850877599222,
      "grad_norm": 0.8307350277900696,
      "learning_rate": 0.00019999904187465229,
      "loss": 0.4935,
      "step": 13300
    },
    {
      "epoch": 0.15410086602139583,
      "grad_norm": 0.808761477470398,
      "learning_rate": 0.00019999898519298782,
      "loss": 0.5065,
      "step": 13310
    },
    {
      "epoch": 0.15421664428286944,
      "grad_norm": 0.878026008605957,
      "learning_rate": 0.00019999892688255108,
      "loss": 0.4912,
      "step": 13320
    },
    {
      "epoch": 0.15433242254434307,
      "grad_norm": 0.9887682199478149,
      "learning_rate": 0.00019999886694334302,
      "loss": 0.5182,
      "step": 13330
    },
    {
      "epoch": 0.1544482008058167,
      "grad_norm": 1.0253299474716187,
      "learning_rate": 0.00019999880537536458,
      "loss": 0.528,
      "step": 13340
    },
    {
      "epoch": 0.15456397906729033,
      "grad_norm": 0.8339182734489441,
      "learning_rate": 0.00019999874217861677,
      "loss": 0.4959,
      "step": 13350
    },
    {
      "epoch": 0.15467975732876396,
      "grad_norm": 1.1303633451461792,
      "learning_rate": 0.00019999867735310067,
      "loss": 0.5416,
      "step": 13360
    },
    {
      "epoch": 0.15479553559023757,
      "grad_norm": 1.045961856842041,
      "learning_rate": 0.00019999861089881727,
      "loss": 0.5228,
      "step": 13370
    },
    {
      "epoch": 0.1549113138517112,
      "grad_norm": 1.3213156461715698,
      "learning_rate": 0.00019999854281576768,
      "loss": 0.5223,
      "step": 13380
    },
    {
      "epoch": 0.15502709211318483,
      "grad_norm": 0.9789355993270874,
      "learning_rate": 0.00019999847310395305,
      "loss": 0.5075,
      "step": 13390
    },
    {
      "epoch": 0.15514287037465846,
      "grad_norm": 1.0346044301986694,
      "learning_rate": 0.00019999840176337442,
      "loss": 0.5329,
      "step": 13400
    },
    {
      "epoch": 0.1552586486361321,
      "grad_norm": 1.0019253492355347,
      "learning_rate": 0.00019999832879403302,
      "loss": 0.5126,
      "step": 13410
    },
    {
      "epoch": 0.1553744268976057,
      "grad_norm": 1.1664997339248657,
      "learning_rate": 0.00019999825419593004,
      "loss": 0.5561,
      "step": 13420
    },
    {
      "epoch": 0.15549020515907933,
      "grad_norm": 1.0309034585952759,
      "learning_rate": 0.00019999817796906668,
      "loss": 0.495,
      "step": 13430
    },
    {
      "epoch": 0.15560598342055296,
      "grad_norm": 1.0485267639160156,
      "learning_rate": 0.00019999810011344418,
      "loss": 0.5099,
      "step": 13440
    },
    {
      "epoch": 0.1557217616820266,
      "grad_norm": 0.6741553544998169,
      "learning_rate": 0.0001999980206290638,
      "loss": 0.5028,
      "step": 13450
    },
    {
      "epoch": 0.15583753994350022,
      "grad_norm": 0.9793708324432373,
      "learning_rate": 0.00019999793951592688,
      "loss": 0.5219,
      "step": 13460
    },
    {
      "epoch": 0.15595331820497382,
      "grad_norm": 0.8301864266395569,
      "learning_rate": 0.00019999785677403463,
      "loss": 0.572,
      "step": 13470
    },
    {
      "epoch": 0.15606909646644745,
      "grad_norm": 0.9063313603401184,
      "learning_rate": 0.00019999777240338855,
      "loss": 0.5587,
      "step": 13480
    },
    {
      "epoch": 0.15618487472792109,
      "grad_norm": 0.9104887843132019,
      "learning_rate": 0.0001999976864039899,
      "loss": 0.5409,
      "step": 13490
    },
    {
      "epoch": 0.15630065298939472,
      "grad_norm": 0.9351223707199097,
      "learning_rate": 0.00019999759877584015,
      "loss": 0.4699,
      "step": 13500
    },
    {
      "epoch": 0.15641643125086835,
      "grad_norm": 1.0543067455291748,
      "learning_rate": 0.00019999750951894067,
      "loss": 0.5453,
      "step": 13510
    },
    {
      "epoch": 0.15653220951234195,
      "grad_norm": 0.950364351272583,
      "learning_rate": 0.00019999741863329295,
      "loss": 0.5182,
      "step": 13520
    },
    {
      "epoch": 0.15664798777381558,
      "grad_norm": 0.9859875440597534,
      "learning_rate": 0.00019999732611889845,
      "loss": 0.522,
      "step": 13530
    },
    {
      "epoch": 0.15676376603528921,
      "grad_norm": 1.1516504287719727,
      "learning_rate": 0.0001999972319757587,
      "loss": 0.5281,
      "step": 13540
    },
    {
      "epoch": 0.15687954429676285,
      "grad_norm": 0.9546409249305725,
      "learning_rate": 0.00019999713620387518,
      "loss": 0.542,
      "step": 13550
    },
    {
      "epoch": 0.15699532255823648,
      "grad_norm": 0.8956830501556396,
      "learning_rate": 0.00019999703880324953,
      "loss": 0.5171,
      "step": 13560
    },
    {
      "epoch": 0.15711110081971008,
      "grad_norm": 0.9716196060180664,
      "learning_rate": 0.0001999969397738833,
      "loss": 0.5131,
      "step": 13570
    },
    {
      "epoch": 0.1572268790811837,
      "grad_norm": 1.1810029745101929,
      "learning_rate": 0.00019999683911577808,
      "loss": 0.5289,
      "step": 13580
    },
    {
      "epoch": 0.15734265734265734,
      "grad_norm": 1.1359663009643555,
      "learning_rate": 0.00019999673682893554,
      "loss": 0.4823,
      "step": 13590
    },
    {
      "epoch": 0.15745843560413098,
      "grad_norm": 1.0935301780700684,
      "learning_rate": 0.00019999663291335736,
      "loss": 0.4927,
      "step": 13600
    },
    {
      "epoch": 0.1575742138656046,
      "grad_norm": 0.9837374091148376,
      "learning_rate": 0.00019999652736904516,
      "loss": 0.4688,
      "step": 13610
    },
    {
      "epoch": 0.1576899921270782,
      "grad_norm": 1.077224612236023,
      "learning_rate": 0.00019999642019600074,
      "loss": 0.5423,
      "step": 13620
    },
    {
      "epoch": 0.15780577038855184,
      "grad_norm": 0.9799237251281738,
      "learning_rate": 0.00019999631139422578,
      "loss": 0.5028,
      "step": 13630
    },
    {
      "epoch": 0.15792154865002547,
      "grad_norm": 1.0222835540771484,
      "learning_rate": 0.00019999620096372212,
      "loss": 0.5321,
      "step": 13640
    },
    {
      "epoch": 0.1580373269114991,
      "grad_norm": 0.7173193097114563,
      "learning_rate": 0.00019999608890449152,
      "loss": 0.5225,
      "step": 13650
    },
    {
      "epoch": 0.15815310517297274,
      "grad_norm": 0.8962286710739136,
      "learning_rate": 0.00019999597521653577,
      "loss": 0.4806,
      "step": 13660
    },
    {
      "epoch": 0.15826888343444634,
      "grad_norm": 1.0847452878952026,
      "learning_rate": 0.00019999585989985682,
      "loss": 0.5528,
      "step": 13670
    },
    {
      "epoch": 0.15838466169591997,
      "grad_norm": 0.8121541142463684,
      "learning_rate": 0.00019999574295445643,
      "loss": 0.552,
      "step": 13680
    },
    {
      "epoch": 0.1585004399573936,
      "grad_norm": 0.8168137073516846,
      "learning_rate": 0.0001999956243803366,
      "loss": 0.5343,
      "step": 13690
    },
    {
      "epoch": 0.15861621821886723,
      "grad_norm": 0.984264612197876,
      "learning_rate": 0.0001999955041774992,
      "loss": 0.5343,
      "step": 13700
    },
    {
      "epoch": 0.15873199648034086,
      "grad_norm": 0.9627836346626282,
      "learning_rate": 0.00019999538234594623,
      "loss": 0.499,
      "step": 13710
    },
    {
      "epoch": 0.15884777474181447,
      "grad_norm": 1.0787767171859741,
      "learning_rate": 0.00019999525888567967,
      "loss": 0.5169,
      "step": 13720
    },
    {
      "epoch": 0.1589635530032881,
      "grad_norm": 1.0938972234725952,
      "learning_rate": 0.00019999513379670146,
      "loss": 0.5098,
      "step": 13730
    },
    {
      "epoch": 0.15907933126476173,
      "grad_norm": 0.7843903303146362,
      "learning_rate": 0.00019999500707901377,
      "loss": 0.5162,
      "step": 13740
    },
    {
      "epoch": 0.15919510952623536,
      "grad_norm": 0.99320387840271,
      "learning_rate": 0.00019999487873261854,
      "loss": 0.5313,
      "step": 13750
    },
    {
      "epoch": 0.159310887787709,
      "grad_norm": 1.0005385875701904,
      "learning_rate": 0.00019999474875751793,
      "loss": 0.5345,
      "step": 13760
    },
    {
      "epoch": 0.1594266660491826,
      "grad_norm": 0.8248289227485657,
      "learning_rate": 0.00019999461715371403,
      "loss": 0.5342,
      "step": 13770
    },
    {
      "epoch": 0.15954244431065623,
      "grad_norm": 1.2158162593841553,
      "learning_rate": 0.00019999448392120898,
      "loss": 0.4959,
      "step": 13780
    },
    {
      "epoch": 0.15965822257212986,
      "grad_norm": 0.9897249937057495,
      "learning_rate": 0.000199994349060005,
      "loss": 0.5572,
      "step": 13790
    },
    {
      "epoch": 0.1597740008336035,
      "grad_norm": 0.9399228692054749,
      "learning_rate": 0.0001999942125701042,
      "loss": 0.5656,
      "step": 13800
    },
    {
      "epoch": 0.1598897790950771,
      "grad_norm": 1.1157760620117188,
      "learning_rate": 0.0001999940744515089,
      "loss": 0.5138,
      "step": 13810
    },
    {
      "epoch": 0.16000555735655073,
      "grad_norm": 1.0835034847259521,
      "learning_rate": 0.00019999393470422126,
      "loss": 0.484,
      "step": 13820
    },
    {
      "epoch": 0.16012133561802436,
      "grad_norm": 0.9482656717300415,
      "learning_rate": 0.0001999937933282436,
      "loss": 0.5382,
      "step": 13830
    },
    {
      "epoch": 0.160237113879498,
      "grad_norm": 0.8925505876541138,
      "learning_rate": 0.00019999365032357823,
      "loss": 0.5222,
      "step": 13840
    },
    {
      "epoch": 0.16035289214097162,
      "grad_norm": 0.9050923585891724,
      "learning_rate": 0.00019999350569022747,
      "loss": 0.5213,
      "step": 13850
    },
    {
      "epoch": 0.16046867040244522,
      "grad_norm": 1.175371766090393,
      "learning_rate": 0.00019999335942819365,
      "loss": 0.5126,
      "step": 13860
    },
    {
      "epoch": 0.16058444866391886,
      "grad_norm": 0.9089219570159912,
      "learning_rate": 0.0001999932115374792,
      "loss": 0.5079,
      "step": 13870
    },
    {
      "epoch": 0.1607002269253925,
      "grad_norm": 0.9556267261505127,
      "learning_rate": 0.0001999930620180865,
      "loss": 0.5313,
      "step": 13880
    },
    {
      "epoch": 0.16081600518686612,
      "grad_norm": 1.0205243825912476,
      "learning_rate": 0.00019999291087001797,
      "loss": 0.4936,
      "step": 13890
    },
    {
      "epoch": 0.16093178344833975,
      "grad_norm": 1.4007447957992554,
      "learning_rate": 0.0001999927580932761,
      "loss": 0.5548,
      "step": 13900
    },
    {
      "epoch": 0.16104756170981335,
      "grad_norm": 1.0816422700881958,
      "learning_rate": 0.0001999926036878634,
      "loss": 0.4974,
      "step": 13910
    },
    {
      "epoch": 0.16116333997128698,
      "grad_norm": 1.1491740942001343,
      "learning_rate": 0.0001999924476537823,
      "loss": 0.5298,
      "step": 13920
    },
    {
      "epoch": 0.16127911823276062,
      "grad_norm": 0.7216561436653137,
      "learning_rate": 0.00019999228999103546,
      "loss": 0.4947,
      "step": 13930
    },
    {
      "epoch": 0.16139489649423425,
      "grad_norm": 0.9310114979743958,
      "learning_rate": 0.00019999213069962535,
      "loss": 0.5117,
      "step": 13940
    },
    {
      "epoch": 0.16151067475570788,
      "grad_norm": 0.8562695980072021,
      "learning_rate": 0.00019999196977955462,
      "loss": 0.5051,
      "step": 13950
    },
    {
      "epoch": 0.16162645301718148,
      "grad_norm": 1.0261586904525757,
      "learning_rate": 0.00019999180723082583,
      "loss": 0.506,
      "step": 13960
    },
    {
      "epoch": 0.1617422312786551,
      "grad_norm": 0.9514285326004028,
      "learning_rate": 0.0001999916430534417,
      "loss": 0.5328,
      "step": 13970
    },
    {
      "epoch": 0.16185800954012874,
      "grad_norm": 0.9260467886924744,
      "learning_rate": 0.00019999147724740487,
      "loss": 0.5432,
      "step": 13980
    },
    {
      "epoch": 0.16197378780160238,
      "grad_norm": 0.9063887000083923,
      "learning_rate": 0.00019999130981271805,
      "loss": 0.491,
      "step": 13990
    },
    {
      "epoch": 0.162089566063076,
      "grad_norm": 1.0284674167633057,
      "learning_rate": 0.0001999911407493839,
      "loss": 0.5596,
      "step": 14000
    },
    {
      "epoch": 0.162089566063076,
      "eval_chrf": 81.43673640677459,
      "eval_loss": 0.7931199073791504,
      "eval_runtime": 349.9972,
      "eval_samples_per_second": 0.286,
      "eval_steps_per_second": 0.011,
      "step": 14000
    },
    {
      "epoch": 0.1622053443245496,
      "grad_norm": 0.8409836888313293,
      "learning_rate": 0.00019999097005740532,
      "loss": 0.5129,
      "step": 14010
    },
    {
      "epoch": 0.16232112258602324,
      "grad_norm": 0.8699890971183777,
      "learning_rate": 0.00019999079773678498,
      "loss": 0.4965,
      "step": 14020
    },
    {
      "epoch": 0.16243690084749687,
      "grad_norm": 0.8601651787757874,
      "learning_rate": 0.00019999062378752566,
      "loss": 0.5138,
      "step": 14030
    },
    {
      "epoch": 0.1625526791089705,
      "grad_norm": 0.9185842275619507,
      "learning_rate": 0.0001999904482096303,
      "loss": 0.4865,
      "step": 14040
    },
    {
      "epoch": 0.16266845737044414,
      "grad_norm": 0.781516432762146,
      "learning_rate": 0.00019999027100310167,
      "loss": 0.5206,
      "step": 14050
    },
    {
      "epoch": 0.16278423563191774,
      "grad_norm": 0.9632774591445923,
      "learning_rate": 0.0001999900921679427,
      "loss": 0.5376,
      "step": 14060
    },
    {
      "epoch": 0.16290001389339137,
      "grad_norm": 0.9048920273780823,
      "learning_rate": 0.00019998991170415632,
      "loss": 0.507,
      "step": 14070
    },
    {
      "epoch": 0.163015792154865,
      "grad_norm": 1.1058522462844849,
      "learning_rate": 0.0001999897296117454,
      "loss": 0.4937,
      "step": 14080
    },
    {
      "epoch": 0.16313157041633863,
      "grad_norm": 0.8302661180496216,
      "learning_rate": 0.00019998954589071298,
      "loss": 0.5234,
      "step": 14090
    },
    {
      "epoch": 0.16324734867781227,
      "grad_norm": 0.8306782245635986,
      "learning_rate": 0.000199989360541062,
      "loss": 0.5215,
      "step": 14100
    },
    {
      "epoch": 0.16336312693928587,
      "grad_norm": 1.0566506385803223,
      "learning_rate": 0.00019998917356279548,
      "loss": 0.5306,
      "step": 14110
    },
    {
      "epoch": 0.1634789052007595,
      "grad_norm": 1.0093319416046143,
      "learning_rate": 0.00019998898495591653,
      "loss": 0.5165,
      "step": 14120
    },
    {
      "epoch": 0.16359468346223313,
      "grad_norm": 0.8950533866882324,
      "learning_rate": 0.00019998879472042815,
      "loss": 0.5321,
      "step": 14130
    },
    {
      "epoch": 0.16371046172370676,
      "grad_norm": 0.926193118095398,
      "learning_rate": 0.0001999886028563335,
      "loss": 0.4834,
      "step": 14140
    },
    {
      "epoch": 0.1638262399851804,
      "grad_norm": 0.7696795463562012,
      "learning_rate": 0.00019998840936363564,
      "loss": 0.5467,
      "step": 14150
    },
    {
      "epoch": 0.163942018246654,
      "grad_norm": 0.9309307932853699,
      "learning_rate": 0.00019998821424233775,
      "loss": 0.5062,
      "step": 14160
    },
    {
      "epoch": 0.16405779650812763,
      "grad_norm": 0.8849621415138245,
      "learning_rate": 0.00019998801749244302,
      "loss": 0.5309,
      "step": 14170
    },
    {
      "epoch": 0.16417357476960126,
      "grad_norm": 1.0118907690048218,
      "learning_rate": 0.00019998781911395464,
      "loss": 0.5307,
      "step": 14180
    },
    {
      "epoch": 0.1642893530310749,
      "grad_norm": 0.9496971368789673,
      "learning_rate": 0.00019998761910687584,
      "loss": 0.5341,
      "step": 14190
    },
    {
      "epoch": 0.16440513129254852,
      "grad_norm": 0.8178584575653076,
      "learning_rate": 0.00019998741747120989,
      "loss": 0.4948,
      "step": 14200
    },
    {
      "epoch": 0.16452090955402213,
      "grad_norm": 0.8048959374427795,
      "learning_rate": 0.00019998721420696008,
      "loss": 0.5327,
      "step": 14210
    },
    {
      "epoch": 0.16463668781549576,
      "grad_norm": 0.9143779277801514,
      "learning_rate": 0.0001999870093141297,
      "loss": 0.5188,
      "step": 14220
    },
    {
      "epoch": 0.1647524660769694,
      "grad_norm": 0.8038369417190552,
      "learning_rate": 0.0001999868027927221,
      "loss": 0.5485,
      "step": 14230
    },
    {
      "epoch": 0.16486824433844302,
      "grad_norm": 0.859785795211792,
      "learning_rate": 0.0001999865946427406,
      "loss": 0.4995,
      "step": 14240
    },
    {
      "epoch": 0.16498402259991665,
      "grad_norm": 0.8410471081733704,
      "learning_rate": 0.00019998638486418867,
      "loss": 0.5308,
      "step": 14250
    },
    {
      "epoch": 0.16509980086139026,
      "grad_norm": 1.0899213552474976,
      "learning_rate": 0.0001999861734570697,
      "loss": 0.5215,
      "step": 14260
    },
    {
      "epoch": 0.1652155791228639,
      "grad_norm": 0.8579217791557312,
      "learning_rate": 0.00019998596042138708,
      "loss": 0.5162,
      "step": 14270
    },
    {
      "epoch": 0.16533135738433752,
      "grad_norm": 0.8680914044380188,
      "learning_rate": 0.00019998574575714434,
      "loss": 0.4844,
      "step": 14280
    },
    {
      "epoch": 0.16544713564581115,
      "grad_norm": 0.9365717172622681,
      "learning_rate": 0.00019998552946434494,
      "loss": 0.5281,
      "step": 14290
    },
    {
      "epoch": 0.16556291390728478,
      "grad_norm": 0.9904703497886658,
      "learning_rate": 0.00019998531154299246,
      "loss": 0.4714,
      "step": 14300
    },
    {
      "epoch": 0.16567869216875838,
      "grad_norm": 0.9480510950088501,
      "learning_rate": 0.00019998509199309038,
      "loss": 0.5137,
      "step": 14310
    },
    {
      "epoch": 0.16579447043023202,
      "grad_norm": 0.9532126188278198,
      "learning_rate": 0.0001999848708146423,
      "loss": 0.5092,
      "step": 14320
    },
    {
      "epoch": 0.16591024869170565,
      "grad_norm": 1.0289478302001953,
      "learning_rate": 0.00019998464800765185,
      "loss": 0.4897,
      "step": 14330
    },
    {
      "epoch": 0.16602602695317928,
      "grad_norm": 0.9527698159217834,
      "learning_rate": 0.0001999844235721226,
      "loss": 0.5203,
      "step": 14340
    },
    {
      "epoch": 0.1661418052146529,
      "grad_norm": 0.8923049569129944,
      "learning_rate": 0.0001999841975080583,
      "loss": 0.4317,
      "step": 14350
    },
    {
      "epoch": 0.1662575834761265,
      "grad_norm": 0.9820677042007446,
      "learning_rate": 0.00019998396981546251,
      "loss": 0.514,
      "step": 14360
    },
    {
      "epoch": 0.16637336173760015,
      "grad_norm": 0.8312664031982422,
      "learning_rate": 0.00019998374049433906,
      "loss": 0.5393,
      "step": 14370
    },
    {
      "epoch": 0.16648913999907378,
      "grad_norm": 0.8548251390457153,
      "learning_rate": 0.00019998350954469158,
      "loss": 0.507,
      "step": 14380
    },
    {
      "epoch": 0.1666049182605474,
      "grad_norm": 0.835356593132019,
      "learning_rate": 0.00019998327696652392,
      "loss": 0.5075,
      "step": 14390
    },
    {
      "epoch": 0.16672069652202104,
      "grad_norm": 0.7591782212257385,
      "learning_rate": 0.0001999830427598398,
      "loss": 0.4992,
      "step": 14400
    },
    {
      "epoch": 0.16683647478349464,
      "grad_norm": 0.845977783203125,
      "learning_rate": 0.0001999828069246431,
      "loss": 0.5341,
      "step": 14410
    },
    {
      "epoch": 0.16695225304496827,
      "grad_norm": 0.9690206050872803,
      "learning_rate": 0.00019998256946093758,
      "loss": 0.5106,
      "step": 14420
    },
    {
      "epoch": 0.1670680313064419,
      "grad_norm": 1.0200231075286865,
      "learning_rate": 0.00019998233036872716,
      "loss": 0.5564,
      "step": 14430
    },
    {
      "epoch": 0.16718380956791554,
      "grad_norm": 1.0304076671600342,
      "learning_rate": 0.00019998208964801573,
      "loss": 0.5094,
      "step": 14440
    },
    {
      "epoch": 0.16729958782938914,
      "grad_norm": 0.9124101996421814,
      "learning_rate": 0.00019998184729880722,
      "loss": 0.4673,
      "step": 14450
    },
    {
      "epoch": 0.16741536609086277,
      "grad_norm": 1.0167433023452759,
      "learning_rate": 0.00019998160332110554,
      "loss": 0.5141,
      "step": 14460
    },
    {
      "epoch": 0.1675311443523364,
      "grad_norm": 1.0613348484039307,
      "learning_rate": 0.00019998135771491472,
      "loss": 0.4761,
      "step": 14470
    },
    {
      "epoch": 0.16764692261381003,
      "grad_norm": 1.0821583271026611,
      "learning_rate": 0.0001999811104802387,
      "loss": 0.5348,
      "step": 14480
    },
    {
      "epoch": 0.16776270087528367,
      "grad_norm": 0.9306504726409912,
      "learning_rate": 0.00019998086161708155,
      "loss": 0.5123,
      "step": 14490
    },
    {
      "epoch": 0.16787847913675727,
      "grad_norm": 0.7311174273490906,
      "learning_rate": 0.0001999806111254473,
      "loss": 0.4772,
      "step": 14500
    },
    {
      "epoch": 0.1679942573982309,
      "grad_norm": 1.2023673057556152,
      "learning_rate": 0.00019998035900534004,
      "loss": 0.4854,
      "step": 14510
    },
    {
      "epoch": 0.16811003565970453,
      "grad_norm": 1.234139323234558,
      "learning_rate": 0.00019998010525676388,
      "loss": 0.4833,
      "step": 14520
    },
    {
      "epoch": 0.16822581392117816,
      "grad_norm": 0.8681570291519165,
      "learning_rate": 0.00019997984987972297,
      "loss": 0.5141,
      "step": 14530
    },
    {
      "epoch": 0.1683415921826518,
      "grad_norm": 0.805265486240387,
      "learning_rate": 0.0001999795928742214,
      "loss": 0.5431,
      "step": 14540
    },
    {
      "epoch": 0.1684573704441254,
      "grad_norm": 0.8005607724189758,
      "learning_rate": 0.00019997933424026345,
      "loss": 0.4831,
      "step": 14550
    },
    {
      "epoch": 0.16857314870559903,
      "grad_norm": 1.0039647817611694,
      "learning_rate": 0.00019997907397785328,
      "loss": 0.5067,
      "step": 14560
    },
    {
      "epoch": 0.16868892696707266,
      "grad_norm": 0.8453294038772583,
      "learning_rate": 0.0001999788120869951,
      "loss": 0.4764,
      "step": 14570
    },
    {
      "epoch": 0.1688047052285463,
      "grad_norm": 0.9476231336593628,
      "learning_rate": 0.0001999785485676932,
      "loss": 0.5278,
      "step": 14580
    },
    {
      "epoch": 0.16892048349001992,
      "grad_norm": 1.1090285778045654,
      "learning_rate": 0.00019997828341995194,
      "loss": 0.4858,
      "step": 14590
    },
    {
      "epoch": 0.16903626175149353,
      "grad_norm": 0.9511401057243347,
      "learning_rate": 0.00019997801664377557,
      "loss": 0.5136,
      "step": 14600
    },
    {
      "epoch": 0.16915204001296716,
      "grad_norm": 0.8264870643615723,
      "learning_rate": 0.00019997774823916843,
      "loss": 0.5244,
      "step": 14610
    },
    {
      "epoch": 0.1692678182744408,
      "grad_norm": 0.8192489743232727,
      "learning_rate": 0.00019997747820613496,
      "loss": 0.4999,
      "step": 14620
    },
    {
      "epoch": 0.16938359653591442,
      "grad_norm": 0.8035051822662354,
      "learning_rate": 0.00019997720654467944,
      "loss": 0.5281,
      "step": 14630
    },
    {
      "epoch": 0.16949937479738805,
      "grad_norm": 0.9535481333732605,
      "learning_rate": 0.0001999769332548064,
      "loss": 0.5127,
      "step": 14640
    },
    {
      "epoch": 0.16961515305886166,
      "grad_norm": 0.9488846659660339,
      "learning_rate": 0.0001999766583365202,
      "loss": 0.5075,
      "step": 14650
    },
    {
      "epoch": 0.1697309313203353,
      "grad_norm": 0.7904040217399597,
      "learning_rate": 0.00019997638178982543,
      "loss": 0.5244,
      "step": 14660
    },
    {
      "epoch": 0.16984670958180892,
      "grad_norm": 0.9358911514282227,
      "learning_rate": 0.0001999761036147265,
      "loss": 0.5164,
      "step": 14670
    },
    {
      "epoch": 0.16996248784328255,
      "grad_norm": 0.9743123054504395,
      "learning_rate": 0.00019997582381122802,
      "loss": 0.5334,
      "step": 14680
    },
    {
      "epoch": 0.17007826610475618,
      "grad_norm": 1.0591320991516113,
      "learning_rate": 0.0001999755423793345,
      "loss": 0.5063,
      "step": 14690
    },
    {
      "epoch": 0.17019404436622979,
      "grad_norm": 0.8685837984085083,
      "learning_rate": 0.00019997525931905047,
      "loss": 0.4992,
      "step": 14700
    },
    {
      "epoch": 0.17030982262770342,
      "grad_norm": 0.9020100831985474,
      "learning_rate": 0.00019997497463038064,
      "loss": 0.4914,
      "step": 14710
    },
    {
      "epoch": 0.17042560088917705,
      "grad_norm": 1.0350594520568848,
      "learning_rate": 0.0001999746883133296,
      "loss": 0.4918,
      "step": 14720
    },
    {
      "epoch": 0.17054137915065068,
      "grad_norm": 0.8129405379295349,
      "learning_rate": 0.00019997440036790203,
      "loss": 0.5009,
      "step": 14730
    },
    {
      "epoch": 0.1706571574121243,
      "grad_norm": 0.8105813264846802,
      "learning_rate": 0.0001999741107941026,
      "loss": 0.5411,
      "step": 14740
    },
    {
      "epoch": 0.17077293567359791,
      "grad_norm": 0.9189984202384949,
      "learning_rate": 0.00019997381959193603,
      "loss": 0.5139,
      "step": 14750
    },
    {
      "epoch": 0.17088871393507155,
      "grad_norm": 0.8906582593917847,
      "learning_rate": 0.00019997352676140708,
      "loss": 0.4861,
      "step": 14760
    },
    {
      "epoch": 0.17100449219654518,
      "grad_norm": 0.7896034121513367,
      "learning_rate": 0.0001999732323025205,
      "loss": 0.5163,
      "step": 14770
    },
    {
      "epoch": 0.1711202704580188,
      "grad_norm": 0.9161441326141357,
      "learning_rate": 0.00019997293621528108,
      "loss": 0.4858,
      "step": 14780
    },
    {
      "epoch": 0.17123604871949244,
      "grad_norm": 0.8519560694694519,
      "learning_rate": 0.0001999726384996937,
      "loss": 0.5316,
      "step": 14790
    },
    {
      "epoch": 0.17135182698096604,
      "grad_norm": 0.759761393070221,
      "learning_rate": 0.00019997233915576313,
      "loss": 0.5132,
      "step": 14800
    },
    {
      "epoch": 0.17146760524243967,
      "grad_norm": 0.6669837236404419,
      "learning_rate": 0.0001999720381834943,
      "loss": 0.4624,
      "step": 14810
    },
    {
      "epoch": 0.1715833835039133,
      "grad_norm": 0.8967482447624207,
      "learning_rate": 0.00019997173558289206,
      "loss": 0.4829,
      "step": 14820
    },
    {
      "epoch": 0.17169916176538694,
      "grad_norm": 0.7932374477386475,
      "learning_rate": 0.00019997143135396138,
      "loss": 0.4793,
      "step": 14830
    },
    {
      "epoch": 0.17181494002686057,
      "grad_norm": 0.9862871766090393,
      "learning_rate": 0.00019997112549670725,
      "loss": 0.5125,
      "step": 14840
    },
    {
      "epoch": 0.17193071828833417,
      "grad_norm": 0.8630059957504272,
      "learning_rate": 0.0001999708180111346,
      "loss": 0.5138,
      "step": 14850
    },
    {
      "epoch": 0.1720464965498078,
      "grad_norm": 1.0792642831802368,
      "learning_rate": 0.00019997050889724842,
      "loss": 0.5013,
      "step": 14860
    },
    {
      "epoch": 0.17216227481128143,
      "grad_norm": 0.7956786155700684,
      "learning_rate": 0.00019997019815505378,
      "loss": 0.494,
      "step": 14870
    },
    {
      "epoch": 0.17227805307275507,
      "grad_norm": 1.120779037475586,
      "learning_rate": 0.00019996988578455574,
      "loss": 0.5057,
      "step": 14880
    },
    {
      "epoch": 0.1723938313342287,
      "grad_norm": 1.0774561166763306,
      "learning_rate": 0.0001999695717857594,
      "loss": 0.5308,
      "step": 14890
    },
    {
      "epoch": 0.1725096095957023,
      "grad_norm": 1.391444444656372,
      "learning_rate": 0.0001999692561586698,
      "loss": 0.4611,
      "step": 14900
    },
    {
      "epoch": 0.17262538785717593,
      "grad_norm": 1.1209179162979126,
      "learning_rate": 0.0001999689389032922,
      "loss": 0.5135,
      "step": 14910
    },
    {
      "epoch": 0.17274116611864956,
      "grad_norm": 0.8599587678909302,
      "learning_rate": 0.00019996862001963169,
      "loss": 0.5263,
      "step": 14920
    },
    {
      "epoch": 0.1728569443801232,
      "grad_norm": 1.065420389175415,
      "learning_rate": 0.00019996829950769346,
      "loss": 0.5095,
      "step": 14930
    },
    {
      "epoch": 0.17297272264159683,
      "grad_norm": 1.0574932098388672,
      "learning_rate": 0.00019996797736748277,
      "loss": 0.5101,
      "step": 14940
    },
    {
      "epoch": 0.17308850090307043,
      "grad_norm": 1.0462032556533813,
      "learning_rate": 0.0001999676535990048,
      "loss": 0.5052,
      "step": 14950
    },
    {
      "epoch": 0.17320427916454406,
      "grad_norm": 0.8948199152946472,
      "learning_rate": 0.00019996732820226493,
      "loss": 0.5418,
      "step": 14960
    },
    {
      "epoch": 0.1733200574260177,
      "grad_norm": 0.8862332701683044,
      "learning_rate": 0.00019996700117726836,
      "loss": 0.5027,
      "step": 14970
    },
    {
      "epoch": 0.17343583568749132,
      "grad_norm": 0.9259552359580994,
      "learning_rate": 0.00019996667252402048,
      "loss": 0.4958,
      "step": 14980
    },
    {
      "epoch": 0.17355161394896496,
      "grad_norm": 0.913835883140564,
      "learning_rate": 0.0001999663422425266,
      "loss": 0.5157,
      "step": 14990
    },
    {
      "epoch": 0.17366739221043856,
      "grad_norm": 0.9743481278419495,
      "learning_rate": 0.00019996601033279215,
      "loss": 0.509,
      "step": 15000
    },
    {
      "epoch": 0.17366739221043856,
      "eval_chrf": 85.47568328388861,
      "eval_loss": 0.7757729291915894,
      "eval_runtime": 105.9694,
      "eval_samples_per_second": 0.944,
      "eval_steps_per_second": 0.038,
      "step": 15000
    },
    {
      "epoch": 0.1737831704719122,
      "grad_norm": 0.8948907852172852,
      "learning_rate": 0.00019996567679482244,
      "loss": 0.5093,
      "step": 15010
    },
    {
      "epoch": 0.17389894873338582,
      "grad_norm": 0.8802430033683777,
      "learning_rate": 0.00019996534162862304,
      "loss": 0.5024,
      "step": 15020
    },
    {
      "epoch": 0.17401472699485945,
      "grad_norm": 0.9146379828453064,
      "learning_rate": 0.0001999650048341993,
      "loss": 0.5022,
      "step": 15030
    },
    {
      "epoch": 0.17413050525633308,
      "grad_norm": 0.8232753276824951,
      "learning_rate": 0.00019996466641155675,
      "loss": 0.4949,
      "step": 15040
    },
    {
      "epoch": 0.1742462835178067,
      "grad_norm": 0.8490627408027649,
      "learning_rate": 0.0001999643263607009,
      "loss": 0.5332,
      "step": 15050
    },
    {
      "epoch": 0.17436206177928032,
      "grad_norm": 1.0217550992965698,
      "learning_rate": 0.00019996398468163728,
      "loss": 0.5035,
      "step": 15060
    },
    {
      "epoch": 0.17447784004075395,
      "grad_norm": 0.8498371243476868,
      "learning_rate": 0.00019996364137437147,
      "loss": 0.4705,
      "step": 15070
    },
    {
      "epoch": 0.17459361830222758,
      "grad_norm": 1.047986388206482,
      "learning_rate": 0.000199963296438909,
      "loss": 0.5127,
      "step": 15080
    },
    {
      "epoch": 0.17470939656370119,
      "grad_norm": 1.009361982345581,
      "learning_rate": 0.00019996294987525558,
      "loss": 0.4933,
      "step": 15090
    },
    {
      "epoch": 0.17482517482517482,
      "grad_norm": 0.49553871154785156,
      "learning_rate": 0.00019996260168341683,
      "loss": 0.4879,
      "step": 15100
    },
    {
      "epoch": 0.17494095308664845,
      "grad_norm": 1.1071234941482544,
      "learning_rate": 0.0001999622518633984,
      "loss": 0.5134,
      "step": 15110
    },
    {
      "epoch": 0.17505673134812208,
      "grad_norm": 0.8918356895446777,
      "learning_rate": 0.00019996190041520596,
      "loss": 0.4804,
      "step": 15120
    },
    {
      "epoch": 0.1751725096095957,
      "grad_norm": 0.9107331037521362,
      "learning_rate": 0.0001999615473388453,
      "loss": 0.4847,
      "step": 15130
    },
    {
      "epoch": 0.17528828787106931,
      "grad_norm": 0.9363231062889099,
      "learning_rate": 0.0001999611926343221,
      "loss": 0.4694,
      "step": 15140
    },
    {
      "epoch": 0.17540406613254295,
      "grad_norm": 1.1326897144317627,
      "learning_rate": 0.00019996083630164217,
      "loss": 0.5201,
      "step": 15150
    },
    {
      "epoch": 0.17551984439401658,
      "grad_norm": 0.9087540507316589,
      "learning_rate": 0.00019996047834081138,
      "loss": 0.5203,
      "step": 15160
    },
    {
      "epoch": 0.1756356226554902,
      "grad_norm": 0.9782746434211731,
      "learning_rate": 0.00019996011875183544,
      "loss": 0.5282,
      "step": 15170
    },
    {
      "epoch": 0.17575140091696384,
      "grad_norm": 0.9152000546455383,
      "learning_rate": 0.0001999597575347203,
      "loss": 0.4961,
      "step": 15180
    },
    {
      "epoch": 0.17586717917843744,
      "grad_norm": 0.9302141070365906,
      "learning_rate": 0.0001999593946894718,
      "loss": 0.5099,
      "step": 15190
    },
    {
      "epoch": 0.17598295743991108,
      "grad_norm": 0.860459566116333,
      "learning_rate": 0.00019995903021609586,
      "loss": 0.5068,
      "step": 15200
    },
    {
      "epoch": 0.1760987357013847,
      "grad_norm": 0.8768830299377441,
      "learning_rate": 0.00019995866411459844,
      "loss": 0.5152,
      "step": 15210
    },
    {
      "epoch": 0.17621451396285834,
      "grad_norm": 0.6508819460868835,
      "learning_rate": 0.00019995829638498545,
      "loss": 0.4922,
      "step": 15220
    },
    {
      "epoch": 0.17633029222433197,
      "grad_norm": 0.9003118872642517,
      "learning_rate": 0.0001999579270272629,
      "loss": 0.4852,
      "step": 15230
    },
    {
      "epoch": 0.17644607048580557,
      "grad_norm": 0.8923856019973755,
      "learning_rate": 0.0001999575560414368,
      "loss": 0.4897,
      "step": 15240
    },
    {
      "epoch": 0.1765618487472792,
      "grad_norm": 1.007230520248413,
      "learning_rate": 0.00019995718342751326,
      "loss": 0.51,
      "step": 15250
    },
    {
      "epoch": 0.17667762700875284,
      "grad_norm": 1.056197166442871,
      "learning_rate": 0.00019995680918549828,
      "loss": 0.5056,
      "step": 15260
    },
    {
      "epoch": 0.17679340527022647,
      "grad_norm": 0.9003465175628662,
      "learning_rate": 0.00019995643331539796,
      "loss": 0.5005,
      "step": 15270
    },
    {
      "epoch": 0.1769091835317001,
      "grad_norm": 0.7962557673454285,
      "learning_rate": 0.00019995605581721844,
      "loss": 0.4852,
      "step": 15280
    },
    {
      "epoch": 0.1770249617931737,
      "grad_norm": 1.1030645370483398,
      "learning_rate": 0.00019995567669096585,
      "loss": 0.4941,
      "step": 15290
    },
    {
      "epoch": 0.17714074005464733,
      "grad_norm": 0.9637556672096252,
      "learning_rate": 0.00019995529593664642,
      "loss": 0.4518,
      "step": 15300
    },
    {
      "epoch": 0.17725651831612096,
      "grad_norm": 1.0759456157684326,
      "learning_rate": 0.00019995491355426624,
      "loss": 0.4993,
      "step": 15310
    },
    {
      "epoch": 0.1773722965775946,
      "grad_norm": 1.3181650638580322,
      "learning_rate": 0.00019995452954383168,
      "loss": 0.5236,
      "step": 15320
    },
    {
      "epoch": 0.17748807483906823,
      "grad_norm": 0.907558798789978,
      "learning_rate": 0.0001999541439053489,
      "loss": 0.4732,
      "step": 15330
    },
    {
      "epoch": 0.17760385310054183,
      "grad_norm": 0.8819958567619324,
      "learning_rate": 0.0001999537566388242,
      "loss": 0.481,
      "step": 15340
    },
    {
      "epoch": 0.17771963136201546,
      "grad_norm": 0.9222123622894287,
      "learning_rate": 0.00019995336774426386,
      "loss": 0.5429,
      "step": 15350
    },
    {
      "epoch": 0.1778354096234891,
      "grad_norm": 0.916992723941803,
      "learning_rate": 0.00019995297722167428,
      "loss": 0.5154,
      "step": 15360
    },
    {
      "epoch": 0.17795118788496272,
      "grad_norm": 0.9565227627754211,
      "learning_rate": 0.0001999525850710618,
      "loss": 0.5035,
      "step": 15370
    },
    {
      "epoch": 0.17806696614643636,
      "grad_norm": 1.0885106325149536,
      "learning_rate": 0.00019995219129243275,
      "loss": 0.5016,
      "step": 15380
    },
    {
      "epoch": 0.17818274440790996,
      "grad_norm": 0.9579272270202637,
      "learning_rate": 0.00019995179588579363,
      "loss": 0.5178,
      "step": 15390
    },
    {
      "epoch": 0.1782985226693836,
      "grad_norm": 0.8931837677955627,
      "learning_rate": 0.0001999513988511508,
      "loss": 0.4913,
      "step": 15400
    },
    {
      "epoch": 0.17841430093085722,
      "grad_norm": 0.842904269695282,
      "learning_rate": 0.0001999510001885108,
      "loss": 0.524,
      "step": 15410
    },
    {
      "epoch": 0.17853007919233085,
      "grad_norm": 1.0296918153762817,
      "learning_rate": 0.00019995059989788008,
      "loss": 0.5002,
      "step": 15420
    },
    {
      "epoch": 0.17864585745380449,
      "grad_norm": 0.8142427206039429,
      "learning_rate": 0.00019995019797926515,
      "loss": 0.489,
      "step": 15430
    },
    {
      "epoch": 0.1787616357152781,
      "grad_norm": 0.955742359161377,
      "learning_rate": 0.00019994979443267258,
      "loss": 0.5261,
      "step": 15440
    },
    {
      "epoch": 0.17887741397675172,
      "grad_norm": 0.9862329959869385,
      "learning_rate": 0.00019994938925810895,
      "loss": 0.4873,
      "step": 15450
    },
    {
      "epoch": 0.17899319223822535,
      "grad_norm": 1.1219494342803955,
      "learning_rate": 0.00019994898245558084,
      "loss": 0.5111,
      "step": 15460
    },
    {
      "epoch": 0.17910897049969898,
      "grad_norm": 0.8652596473693848,
      "learning_rate": 0.0001999485740250949,
      "loss": 0.5194,
      "step": 15470
    },
    {
      "epoch": 0.17922474876117261,
      "grad_norm": 0.9113269448280334,
      "learning_rate": 0.00019994816396665775,
      "loss": 0.5139,
      "step": 15480
    },
    {
      "epoch": 0.17934052702264622,
      "grad_norm": 1.0372772216796875,
      "learning_rate": 0.0001999477522802761,
      "loss": 0.4604,
      "step": 15490
    },
    {
      "epoch": 0.17945630528411985,
      "grad_norm": 0.8772414922714233,
      "learning_rate": 0.0001999473389659566,
      "loss": 0.4644,
      "step": 15500
    },
    {
      "epoch": 0.17957208354559348,
      "grad_norm": 0.9783409833908081,
      "learning_rate": 0.00019994692402370605,
      "loss": 0.521,
      "step": 15510
    },
    {
      "epoch": 0.1796878618070671,
      "grad_norm": 1.0312986373901367,
      "learning_rate": 0.00019994650745353115,
      "loss": 0.5288,
      "step": 15520
    },
    {
      "epoch": 0.17980364006854074,
      "grad_norm": 0.951167106628418,
      "learning_rate": 0.0001999460892554388,
      "loss": 0.5269,
      "step": 15530
    },
    {
      "epoch": 0.17991941833001435,
      "grad_norm": 0.9194570779800415,
      "learning_rate": 0.00019994566942943564,
      "loss": 0.5372,
      "step": 15540
    },
    {
      "epoch": 0.18003519659148798,
      "grad_norm": 0.8009816408157349,
      "learning_rate": 0.00019994524797552863,
      "loss": 0.5026,
      "step": 15550
    },
    {
      "epoch": 0.1801509748529616,
      "grad_norm": 0.7390750646591187,
      "learning_rate": 0.0001999448248937246,
      "loss": 0.493,
      "step": 15560
    },
    {
      "epoch": 0.18026675311443524,
      "grad_norm": 0.908737063407898,
      "learning_rate": 0.00019994440018403044,
      "loss": 0.4882,
      "step": 15570
    },
    {
      "epoch": 0.18038253137590887,
      "grad_norm": 0.838344931602478,
      "learning_rate": 0.00019994397384645306,
      "loss": 0.4914,
      "step": 15580
    },
    {
      "epoch": 0.18049830963738248,
      "grad_norm": 0.8958513736724854,
      "learning_rate": 0.00019994354588099943,
      "loss": 0.499,
      "step": 15590
    },
    {
      "epoch": 0.1806140878988561,
      "grad_norm": 0.8887691497802734,
      "learning_rate": 0.0001999431162876765,
      "loss": 0.5049,
      "step": 15600
    },
    {
      "epoch": 0.18072986616032974,
      "grad_norm": 0.8820658326148987,
      "learning_rate": 0.00019994268506649126,
      "loss": 0.4739,
      "step": 15610
    },
    {
      "epoch": 0.18084564442180337,
      "grad_norm": 0.8708551526069641,
      "learning_rate": 0.00019994225221745072,
      "loss": 0.498,
      "step": 15620
    },
    {
      "epoch": 0.180961422683277,
      "grad_norm": 0.5771028995513916,
      "learning_rate": 0.000199941817740562,
      "loss": 0.4589,
      "step": 15630
    },
    {
      "epoch": 0.1810772009447506,
      "grad_norm": 0.9449729919433594,
      "learning_rate": 0.00019994138163583212,
      "loss": 0.4865,
      "step": 15640
    },
    {
      "epoch": 0.18119297920622424,
      "grad_norm": 0.8463087677955627,
      "learning_rate": 0.0001999409439032682,
      "loss": 0.4967,
      "step": 15650
    },
    {
      "epoch": 0.18130875746769787,
      "grad_norm": 0.8847309350967407,
      "learning_rate": 0.00019994050454287733,
      "loss": 0.4768,
      "step": 15660
    },
    {
      "epoch": 0.1814245357291715,
      "grad_norm": 0.8302356600761414,
      "learning_rate": 0.00019994006355466672,
      "loss": 0.4865,
      "step": 15670
    },
    {
      "epoch": 0.18154031399064513,
      "grad_norm": 0.6166642904281616,
      "learning_rate": 0.00019993962093864355,
      "loss": 0.4882,
      "step": 15680
    },
    {
      "epoch": 0.18165609225211873,
      "grad_norm": 0.9548183083534241,
      "learning_rate": 0.000199939176694815,
      "loss": 0.513,
      "step": 15690
    },
    {
      "epoch": 0.18177187051359237,
      "grad_norm": 0.6811615228652954,
      "learning_rate": 0.0001999387308231883,
      "loss": 0.488,
      "step": 15700
    },
    {
      "epoch": 0.181887648775066,
      "grad_norm": 0.829233705997467,
      "learning_rate": 0.00019993828332377076,
      "loss": 0.4918,
      "step": 15710
    },
    {
      "epoch": 0.18200342703653963,
      "grad_norm": 0.8032342791557312,
      "learning_rate": 0.00019993783419656962,
      "loss": 0.4877,
      "step": 15720
    },
    {
      "epoch": 0.18211920529801323,
      "grad_norm": 1.0349808931350708,
      "learning_rate": 0.00019993738344159222,
      "loss": 0.5181,
      "step": 15730
    },
    {
      "epoch": 0.18223498355948686,
      "grad_norm": 0.9067845940589905,
      "learning_rate": 0.00019993693105884588,
      "loss": 0.5159,
      "step": 15740
    },
    {
      "epoch": 0.1823507618209605,
      "grad_norm": 0.8226310014724731,
      "learning_rate": 0.000199936477048338,
      "loss": 0.4973,
      "step": 15750
    },
    {
      "epoch": 0.18246654008243413,
      "grad_norm": 0.9729318618774414,
      "learning_rate": 0.00019993602141007597,
      "loss": 0.4934,
      "step": 15760
    },
    {
      "epoch": 0.18258231834390776,
      "grad_norm": 0.6730903387069702,
      "learning_rate": 0.0001999355641440672,
      "loss": 0.4999,
      "step": 15770
    },
    {
      "epoch": 0.18269809660538136,
      "grad_norm": 0.8104208707809448,
      "learning_rate": 0.00019993510525031914,
      "loss": 0.5241,
      "step": 15780
    },
    {
      "epoch": 0.182813874866855,
      "grad_norm": 0.7883507609367371,
      "learning_rate": 0.00019993464472883925,
      "loss": 0.4929,
      "step": 15790
    },
    {
      "epoch": 0.18292965312832862,
      "grad_norm": 0.7942865490913391,
      "learning_rate": 0.00019993418257963506,
      "loss": 0.4741,
      "step": 15800
    },
    {
      "epoch": 0.18304543138980225,
      "grad_norm": 0.7424351572990417,
      "learning_rate": 0.00019993371880271409,
      "loss": 0.4863,
      "step": 15810
    },
    {
      "epoch": 0.18316120965127589,
      "grad_norm": 0.9395301938056946,
      "learning_rate": 0.00019993325339808383,
      "loss": 0.5142,
      "step": 15820
    },
    {
      "epoch": 0.1832769879127495,
      "grad_norm": 1.0837334394454956,
      "learning_rate": 0.00019993278636575197,
      "loss": 0.5143,
      "step": 15830
    },
    {
      "epoch": 0.18339276617422312,
      "grad_norm": 0.8544259071350098,
      "learning_rate": 0.00019993231770572604,
      "loss": 0.4793,
      "step": 15840
    },
    {
      "epoch": 0.18350854443569675,
      "grad_norm": 0.9475684762001038,
      "learning_rate": 0.00019993184741801372,
      "loss": 0.4949,
      "step": 15850
    },
    {
      "epoch": 0.18362432269717038,
      "grad_norm": 1.132094144821167,
      "learning_rate": 0.00019993137550262263,
      "loss": 0.4985,
      "step": 15860
    },
    {
      "epoch": 0.18374010095864401,
      "grad_norm": 0.8986677527427673,
      "learning_rate": 0.00019993090195956044,
      "loss": 0.4791,
      "step": 15870
    },
    {
      "epoch": 0.18385587922011762,
      "grad_norm": 0.8654846549034119,
      "learning_rate": 0.00019993042678883492,
      "loss": 0.5135,
      "step": 15880
    },
    {
      "epoch": 0.18397165748159125,
      "grad_norm": 0.8354236483573914,
      "learning_rate": 0.00019992994999045382,
      "loss": 0.4945,
      "step": 15890
    },
    {
      "epoch": 0.18408743574306488,
      "grad_norm": 0.9234731197357178,
      "learning_rate": 0.00019992947156442485,
      "loss": 0.5155,
      "step": 15900
    },
    {
      "epoch": 0.1842032140045385,
      "grad_norm": 0.7724660038948059,
      "learning_rate": 0.00019992899151075584,
      "loss": 0.4996,
      "step": 15910
    },
    {
      "epoch": 0.18431899226601214,
      "grad_norm": 0.8467785716056824,
      "learning_rate": 0.00019992850982945455,
      "loss": 0.5351,
      "step": 15920
    },
    {
      "epoch": 0.18443477052748575,
      "grad_norm": 0.7764392495155334,
      "learning_rate": 0.0001999280265205289,
      "loss": 0.4691,
      "step": 15930
    },
    {
      "epoch": 0.18455054878895938,
      "grad_norm": 1.137548804283142,
      "learning_rate": 0.00019992754158398675,
      "loss": 0.5135,
      "step": 15940
    },
    {
      "epoch": 0.184666327050433,
      "grad_norm": 0.918748676776886,
      "learning_rate": 0.00019992705501983596,
      "loss": 0.505,
      "step": 15950
    },
    {
      "epoch": 0.18478210531190664,
      "grad_norm": 0.6120473742485046,
      "learning_rate": 0.00019992656682808443,
      "loss": 0.4766,
      "step": 15960
    },
    {
      "epoch": 0.18489788357338027,
      "grad_norm": 1.143615484237671,
      "learning_rate": 0.0001999260770087402,
      "loss": 0.4559,
      "step": 15970
    },
    {
      "epoch": 0.18501366183485388,
      "grad_norm": 0.8118082284927368,
      "learning_rate": 0.00019992558556181122,
      "loss": 0.4886,
      "step": 15980
    },
    {
      "epoch": 0.1851294400963275,
      "grad_norm": 0.7622039914131165,
      "learning_rate": 0.00019992509248730546,
      "loss": 0.4969,
      "step": 15990
    },
    {
      "epoch": 0.18524521835780114,
      "grad_norm": 1.1264930963516235,
      "learning_rate": 0.00019992459778523096,
      "loss": 0.5206,
      "step": 16000
    },
    {
      "epoch": 0.18524521835780114,
      "eval_chrf": 77.97940200596634,
      "eval_loss": 0.7463719248771667,
      "eval_runtime": 246.7953,
      "eval_samples_per_second": 0.405,
      "eval_steps_per_second": 0.016,
      "step": 16000
    },
    {
      "epoch": 0.18536099661927477,
      "grad_norm": 0.8578152060508728,
      "learning_rate": 0.0001999241014555958,
      "loss": 0.5409,
      "step": 16010
    },
    {
      "epoch": 0.1854767748807484,
      "grad_norm": 0.9739165902137756,
      "learning_rate": 0.00019992360349840803,
      "loss": 0.462,
      "step": 16020
    },
    {
      "epoch": 0.185592553142222,
      "grad_norm": 0.7403093576431274,
      "learning_rate": 0.00019992310391367583,
      "loss": 0.4987,
      "step": 16030
    },
    {
      "epoch": 0.18570833140369564,
      "grad_norm": 0.797931969165802,
      "learning_rate": 0.00019992260270140726,
      "loss": 0.5003,
      "step": 16040
    },
    {
      "epoch": 0.18582410966516927,
      "grad_norm": 0.9602776169776917,
      "learning_rate": 0.00019992209986161052,
      "loss": 0.4905,
      "step": 16050
    },
    {
      "epoch": 0.1859398879266429,
      "grad_norm": 0.8908615708351135,
      "learning_rate": 0.00019992159539429378,
      "loss": 0.4865,
      "step": 16060
    },
    {
      "epoch": 0.18605566618811653,
      "grad_norm": 0.9004862904548645,
      "learning_rate": 0.00019992108929946527,
      "loss": 0.5411,
      "step": 16070
    },
    {
      "epoch": 0.18617144444959013,
      "grad_norm": 0.8991667628288269,
      "learning_rate": 0.00019992058157713326,
      "loss": 0.4813,
      "step": 16080
    },
    {
      "epoch": 0.18628722271106377,
      "grad_norm": 0.7413689494132996,
      "learning_rate": 0.00019992007222730597,
      "loss": 0.5018,
      "step": 16090
    },
    {
      "epoch": 0.1864030009725374,
      "grad_norm": 0.931563138961792,
      "learning_rate": 0.00019991956124999175,
      "loss": 0.4659,
      "step": 16100
    },
    {
      "epoch": 0.18651877923401103,
      "grad_norm": 0.7988595366477966,
      "learning_rate": 0.00019991904864519885,
      "loss": 0.4937,
      "step": 16110
    },
    {
      "epoch": 0.18663455749548466,
      "grad_norm": 0.9304957389831543,
      "learning_rate": 0.0001999185344129357,
      "loss": 0.5235,
      "step": 16120
    },
    {
      "epoch": 0.18675033575695826,
      "grad_norm": 0.8325493931770325,
      "learning_rate": 0.0001999180185532106,
      "loss": 0.5058,
      "step": 16130
    },
    {
      "epoch": 0.1868661140184319,
      "grad_norm": 0.9377426505088806,
      "learning_rate": 0.000199917501066032,
      "loss": 0.5201,
      "step": 16140
    },
    {
      "epoch": 0.18698189227990553,
      "grad_norm": 0.876797616481781,
      "learning_rate": 0.00019991698195140836,
      "loss": 0.4893,
      "step": 16150
    },
    {
      "epoch": 0.18709767054137916,
      "grad_norm": 0.922610878944397,
      "learning_rate": 0.00019991646120934804,
      "loss": 0.5103,
      "step": 16160
    },
    {
      "epoch": 0.1872134488028528,
      "grad_norm": 0.8169971108436584,
      "learning_rate": 0.0001999159388398596,
      "loss": 0.486,
      "step": 16170
    },
    {
      "epoch": 0.1873292270643264,
      "grad_norm": 0.7925812602043152,
      "learning_rate": 0.00019991541484295152,
      "loss": 0.4962,
      "step": 16180
    },
    {
      "epoch": 0.18744500532580002,
      "grad_norm": 0.8858729004859924,
      "learning_rate": 0.00019991488921863236,
      "loss": 0.4933,
      "step": 16190
    },
    {
      "epoch": 0.18756078358727366,
      "grad_norm": 0.7224738001823425,
      "learning_rate": 0.0001999143619669106,
      "loss": 0.5051,
      "step": 16200
    },
    {
      "epoch": 0.1876765618487473,
      "grad_norm": 0.7886340618133545,
      "learning_rate": 0.00019991383308779495,
      "loss": 0.5091,
      "step": 16210
    },
    {
      "epoch": 0.18779234011022092,
      "grad_norm": 0.8229026794433594,
      "learning_rate": 0.00019991330258129395,
      "loss": 0.5004,
      "step": 16220
    },
    {
      "epoch": 0.18790811837169452,
      "grad_norm": 0.7977997660636902,
      "learning_rate": 0.00019991277044741624,
      "loss": 0.5029,
      "step": 16230
    },
    {
      "epoch": 0.18802389663316815,
      "grad_norm": 0.9360793828964233,
      "learning_rate": 0.0001999122366861705,
      "loss": 0.4849,
      "step": 16240
    },
    {
      "epoch": 0.18813967489464178,
      "grad_norm": 1.0242421627044678,
      "learning_rate": 0.0001999117012975654,
      "loss": 0.4977,
      "step": 16250
    },
    {
      "epoch": 0.18825545315611542,
      "grad_norm": 0.6850507259368896,
      "learning_rate": 0.00019991116428160974,
      "loss": 0.4827,
      "step": 16260
    },
    {
      "epoch": 0.18837123141758905,
      "grad_norm": 1.1516903638839722,
      "learning_rate": 0.00019991062563831218,
      "loss": 0.5193,
      "step": 16270
    },
    {
      "epoch": 0.18848700967906265,
      "grad_norm": 1.1835291385650635,
      "learning_rate": 0.00019991008536768149,
      "loss": 0.4797,
      "step": 16280
    },
    {
      "epoch": 0.18860278794053628,
      "grad_norm": 0.9349057078361511,
      "learning_rate": 0.00019990954346972656,
      "loss": 0.522,
      "step": 16290
    },
    {
      "epoch": 0.1887185662020099,
      "grad_norm": 0.9771071076393127,
      "learning_rate": 0.00019990899994445615,
      "loss": 0.5127,
      "step": 16300
    },
    {
      "epoch": 0.18883434446348354,
      "grad_norm": 0.801821231842041,
      "learning_rate": 0.00019990845479187912,
      "loss": 0.5034,
      "step": 16310
    },
    {
      "epoch": 0.18895012272495718,
      "grad_norm": 0.8890246748924255,
      "learning_rate": 0.00019990790801200435,
      "loss": 0.5167,
      "step": 16320
    },
    {
      "epoch": 0.18906590098643078,
      "grad_norm": 0.9110584855079651,
      "learning_rate": 0.00019990735960484073,
      "loss": 0.497,
      "step": 16330
    },
    {
      "epoch": 0.1891816792479044,
      "grad_norm": 0.8443339467048645,
      "learning_rate": 0.00019990680957039726,
      "loss": 0.4911,
      "step": 16340
    },
    {
      "epoch": 0.18929745750937804,
      "grad_norm": 0.9822980761528015,
      "learning_rate": 0.0001999062579086828,
      "loss": 0.4723,
      "step": 16350
    },
    {
      "epoch": 0.18941323577085167,
      "grad_norm": 1.1315723657608032,
      "learning_rate": 0.00019990570461970642,
      "loss": 0.4852,
      "step": 16360
    },
    {
      "epoch": 0.18952901403232528,
      "grad_norm": 0.9471389055252075,
      "learning_rate": 0.00019990514970347708,
      "loss": 0.4802,
      "step": 16370
    },
    {
      "epoch": 0.1896447922937989,
      "grad_norm": 0.9146233201026917,
      "learning_rate": 0.00019990459316000382,
      "loss": 0.5348,
      "step": 16380
    },
    {
      "epoch": 0.18976057055527254,
      "grad_norm": 0.8330798745155334,
      "learning_rate": 0.00019990403498929576,
      "loss": 0.5,
      "step": 16390
    },
    {
      "epoch": 0.18987634881674617,
      "grad_norm": 0.9550514221191406,
      "learning_rate": 0.00019990347519136193,
      "loss": 0.509,
      "step": 16400
    },
    {
      "epoch": 0.1899921270782198,
      "grad_norm": 1.097962498664856,
      "learning_rate": 0.00019990291376621147,
      "loss": 0.5237,
      "step": 16410
    },
    {
      "epoch": 0.1901079053396934,
      "grad_norm": 0.9184927940368652,
      "learning_rate": 0.00019990235071385353,
      "loss": 0.5036,
      "step": 16420
    },
    {
      "epoch": 0.19022368360116704,
      "grad_norm": 0.7849079966545105,
      "learning_rate": 0.00019990178603429727,
      "loss": 0.5163,
      "step": 16430
    },
    {
      "epoch": 0.19033946186264067,
      "grad_norm": 0.8647149205207825,
      "learning_rate": 0.0001999012197275519,
      "loss": 0.509,
      "step": 16440
    },
    {
      "epoch": 0.1904552401241143,
      "grad_norm": 0.9983821511268616,
      "learning_rate": 0.00019990065179362662,
      "loss": 0.4997,
      "step": 16450
    },
    {
      "epoch": 0.19057101838558793,
      "grad_norm": 0.7585316300392151,
      "learning_rate": 0.00019990008223253071,
      "loss": 0.5101,
      "step": 16460
    },
    {
      "epoch": 0.19068679664706153,
      "grad_norm": 1.0215591192245483,
      "learning_rate": 0.00019989951104427346,
      "loss": 0.4907,
      "step": 16470
    },
    {
      "epoch": 0.19080257490853517,
      "grad_norm": 0.8907660245895386,
      "learning_rate": 0.00019989893822886412,
      "loss": 0.5184,
      "step": 16480
    },
    {
      "epoch": 0.1909183531700088,
      "grad_norm": 0.8353692293167114,
      "learning_rate": 0.00019989836378631206,
      "loss": 0.478,
      "step": 16490
    },
    {
      "epoch": 0.19103413143148243,
      "grad_norm": 0.7806422114372253,
      "learning_rate": 0.0001998977877166266,
      "loss": 0.4648,
      "step": 16500
    },
    {
      "epoch": 0.19114990969295606,
      "grad_norm": 0.8553751111030579,
      "learning_rate": 0.00019989721001981717,
      "loss": 0.5173,
      "step": 16510
    },
    {
      "epoch": 0.19126568795442966,
      "grad_norm": 0.960426390171051,
      "learning_rate": 0.00019989663069589316,
      "loss": 0.4601,
      "step": 16520
    },
    {
      "epoch": 0.1913814662159033,
      "grad_norm": 0.8828798532485962,
      "learning_rate": 0.000199896049744864,
      "loss": 0.4783,
      "step": 16530
    },
    {
      "epoch": 0.19149724447737693,
      "grad_norm": 0.9241445660591125,
      "learning_rate": 0.00019989546716673911,
      "loss": 0.4973,
      "step": 16540
    },
    {
      "epoch": 0.19161302273885056,
      "grad_norm": 0.9370666742324829,
      "learning_rate": 0.00019989488296152808,
      "loss": 0.5126,
      "step": 16550
    },
    {
      "epoch": 0.1917288010003242,
      "grad_norm": 0.7989693880081177,
      "learning_rate": 0.00019989429712924034,
      "loss": 0.4761,
      "step": 16560
    },
    {
      "epoch": 0.1918445792617978,
      "grad_norm": 0.8280283212661743,
      "learning_rate": 0.0001998937096698855,
      "loss": 0.5093,
      "step": 16570
    },
    {
      "epoch": 0.19196035752327142,
      "grad_norm": 0.8971747159957886,
      "learning_rate": 0.00019989312058347304,
      "loss": 0.4789,
      "step": 16580
    },
    {
      "epoch": 0.19207613578474506,
      "grad_norm": 0.7944716215133667,
      "learning_rate": 0.00019989252987001263,
      "loss": 0.4637,
      "step": 16590
    },
    {
      "epoch": 0.1921919140462187,
      "grad_norm": 0.7768067121505737,
      "learning_rate": 0.00019989193752951388,
      "loss": 0.5141,
      "step": 16600
    },
    {
      "epoch": 0.19230769230769232,
      "grad_norm": 0.8733528256416321,
      "learning_rate": 0.0001998913435619864,
      "loss": 0.5265,
      "step": 16610
    },
    {
      "epoch": 0.19242347056916592,
      "grad_norm": 0.8538109064102173,
      "learning_rate": 0.0001998907479674399,
      "loss": 0.5208,
      "step": 16620
    },
    {
      "epoch": 0.19253924883063955,
      "grad_norm": 0.7162742614746094,
      "learning_rate": 0.00019989015074588408,
      "loss": 0.4527,
      "step": 16630
    },
    {
      "epoch": 0.19265502709211318,
      "grad_norm": 0.8274531364440918,
      "learning_rate": 0.00019988955189732864,
      "loss": 0.4602,
      "step": 16640
    },
    {
      "epoch": 0.19277080535358682,
      "grad_norm": 0.8357710242271423,
      "learning_rate": 0.00019988895142178335,
      "loss": 0.5099,
      "step": 16650
    },
    {
      "epoch": 0.19288658361506045,
      "grad_norm": 0.905566394329071,
      "learning_rate": 0.000199888349319258,
      "loss": 0.4664,
      "step": 16660
    },
    {
      "epoch": 0.19300236187653405,
      "grad_norm": 0.7365356683731079,
      "learning_rate": 0.0001998877455897624,
      "loss": 0.4684,
      "step": 16670
    },
    {
      "epoch": 0.19311814013800768,
      "grad_norm": 0.9022588729858398,
      "learning_rate": 0.00019988714023330638,
      "loss": 0.5147,
      "step": 16680
    },
    {
      "epoch": 0.1932339183994813,
      "grad_norm": 0.9667847156524658,
      "learning_rate": 0.00019988653324989975,
      "loss": 0.4792,
      "step": 16690
    },
    {
      "epoch": 0.19334969666095495,
      "grad_norm": 0.8369743227958679,
      "learning_rate": 0.00019988592463955244,
      "loss": 0.5217,
      "step": 16700
    },
    {
      "epoch": 0.19346547492242858,
      "grad_norm": 0.9116846323013306,
      "learning_rate": 0.0001998853144022744,
      "loss": 0.4942,
      "step": 16710
    },
    {
      "epoch": 0.19358125318390218,
      "grad_norm": 0.877573549747467,
      "learning_rate": 0.0001998847025380755,
      "loss": 0.5424,
      "step": 16720
    },
    {
      "epoch": 0.1936970314453758,
      "grad_norm": 0.9384279847145081,
      "learning_rate": 0.00019988408904696576,
      "loss": 0.4611,
      "step": 16730
    },
    {
      "epoch": 0.19381280970684944,
      "grad_norm": 0.8224205374717712,
      "learning_rate": 0.00019988347392895515,
      "loss": 0.4718,
      "step": 16740
    },
    {
      "epoch": 0.19392858796832307,
      "grad_norm": 0.7887263298034668,
      "learning_rate": 0.00019988285718405369,
      "loss": 0.4862,
      "step": 16750
    },
    {
      "epoch": 0.1940443662297967,
      "grad_norm": 0.882061243057251,
      "learning_rate": 0.0001998822388122714,
      "loss": 0.4502,
      "step": 16760
    },
    {
      "epoch": 0.1941601444912703,
      "grad_norm": 0.8657705783843994,
      "learning_rate": 0.0001998816188136184,
      "loss": 0.4894,
      "step": 16770
    },
    {
      "epoch": 0.19427592275274394,
      "grad_norm": 0.7500507235527039,
      "learning_rate": 0.00019988099718810476,
      "loss": 0.4899,
      "step": 16780
    },
    {
      "epoch": 0.19439170101421757,
      "grad_norm": 1.0131944417953491,
      "learning_rate": 0.0001998803739357406,
      "loss": 0.4917,
      "step": 16790
    },
    {
      "epoch": 0.1945074792756912,
      "grad_norm": 0.9293574690818787,
      "learning_rate": 0.0001998797490565361,
      "loss": 0.5272,
      "step": 16800
    },
    {
      "epoch": 0.19462325753716483,
      "grad_norm": 0.8435137867927551,
      "learning_rate": 0.00019987912255050138,
      "loss": 0.4913,
      "step": 16810
    },
    {
      "epoch": 0.19473903579863844,
      "grad_norm": 0.86867356300354,
      "learning_rate": 0.0001998784944176467,
      "loss": 0.5008,
      "step": 16820
    },
    {
      "epoch": 0.19485481406011207,
      "grad_norm": 0.7702768445014954,
      "learning_rate": 0.00019987786465798228,
      "loss": 0.4924,
      "step": 16830
    },
    {
      "epoch": 0.1949705923215857,
      "grad_norm": 0.930141806602478,
      "learning_rate": 0.0001998772332715184,
      "loss": 0.5379,
      "step": 16840
    },
    {
      "epoch": 0.19508637058305933,
      "grad_norm": 0.9449251294136047,
      "learning_rate": 0.0001998766002582653,
      "loss": 0.479,
      "step": 16850
    },
    {
      "epoch": 0.19520214884453296,
      "grad_norm": 0.8602539896965027,
      "learning_rate": 0.00019987596561823326,
      "loss": 0.5001,
      "step": 16860
    },
    {
      "epoch": 0.19531792710600657,
      "grad_norm": 0.8756052851676941,
      "learning_rate": 0.00019987532935143273,
      "loss": 0.5113,
      "step": 16870
    },
    {
      "epoch": 0.1954337053674802,
      "grad_norm": 1.0247206687927246,
      "learning_rate": 0.00019987469145787395,
      "loss": 0.5097,
      "step": 16880
    },
    {
      "epoch": 0.19554948362895383,
      "grad_norm": 0.7333729863166809,
      "learning_rate": 0.0001998740519375674,
      "loss": 0.4928,
      "step": 16890
    },
    {
      "epoch": 0.19566526189042746,
      "grad_norm": 0.8338857293128967,
      "learning_rate": 0.00019987341079052347,
      "loss": 0.5179,
      "step": 16900
    },
    {
      "epoch": 0.1957810401519011,
      "grad_norm": 0.9072408676147461,
      "learning_rate": 0.00019987276801675257,
      "loss": 0.5086,
      "step": 16910
    },
    {
      "epoch": 0.1958968184133747,
      "grad_norm": 0.7794625163078308,
      "learning_rate": 0.00019987212361626523,
      "loss": 0.466,
      "step": 16920
    },
    {
      "epoch": 0.19601259667484833,
      "grad_norm": 0.8542028069496155,
      "learning_rate": 0.00019987147758907187,
      "loss": 0.5003,
      "step": 16930
    },
    {
      "epoch": 0.19612837493632196,
      "grad_norm": 0.8598357439041138,
      "learning_rate": 0.00019987082993518307,
      "loss": 0.4752,
      "step": 16940
    },
    {
      "epoch": 0.1962441531977956,
      "grad_norm": 1.001768946647644,
      "learning_rate": 0.00019987018065460935,
      "loss": 0.4686,
      "step": 16950
    },
    {
      "epoch": 0.19635993145926922,
      "grad_norm": 0.9171622395515442,
      "learning_rate": 0.00019986952974736135,
      "loss": 0.51,
      "step": 16960
    },
    {
      "epoch": 0.19647570972074282,
      "grad_norm": 0.8419103622436523,
      "learning_rate": 0.00019986887721344955,
      "loss": 0.4809,
      "step": 16970
    },
    {
      "epoch": 0.19659148798221646,
      "grad_norm": 0.7191686034202576,
      "learning_rate": 0.0001998682230528847,
      "loss": 0.4762,
      "step": 16980
    },
    {
      "epoch": 0.1967072662436901,
      "grad_norm": 0.8218206167221069,
      "learning_rate": 0.00019986756726567739,
      "loss": 0.479,
      "step": 16990
    },
    {
      "epoch": 0.19682304450516372,
      "grad_norm": 0.9415873885154724,
      "learning_rate": 0.00019986690985183828,
      "loss": 0.4945,
      "step": 17000
    },
    {
      "epoch": 0.19682304450516372,
      "eval_chrf": 79.72724006061951,
      "eval_loss": 0.7424191832542419,
      "eval_runtime": 217.3798,
      "eval_samples_per_second": 0.46,
      "eval_steps_per_second": 0.018,
      "step": 17000
    },
    {
      "epoch": 0.19693882276663732,
      "grad_norm": 0.9736429452896118,
      "learning_rate": 0.00019986625081137816,
      "loss": 0.4739,
      "step": 17010
    },
    {
      "epoch": 0.19705460102811095,
      "grad_norm": 0.825629711151123,
      "learning_rate": 0.00019986559014430771,
      "loss": 0.4804,
      "step": 17020
    },
    {
      "epoch": 0.19717037928958459,
      "grad_norm": 0.8559385538101196,
      "learning_rate": 0.0001998649278506377,
      "loss": 0.4479,
      "step": 17030
    },
    {
      "epoch": 0.19728615755105822,
      "grad_norm": 0.812768816947937,
      "learning_rate": 0.0001998642639303789,
      "loss": 0.5129,
      "step": 17040
    },
    {
      "epoch": 0.19740193581253185,
      "grad_norm": 0.920117974281311,
      "learning_rate": 0.00019986359838354215,
      "loss": 0.5177,
      "step": 17050
    },
    {
      "epoch": 0.19751771407400545,
      "grad_norm": 0.7651894092559814,
      "learning_rate": 0.00019986293121013825,
      "loss": 0.4596,
      "step": 17060
    },
    {
      "epoch": 0.19763349233547908,
      "grad_norm": 0.9783793091773987,
      "learning_rate": 0.00019986226241017812,
      "loss": 0.5417,
      "step": 17070
    },
    {
      "epoch": 0.19774927059695271,
      "grad_norm": 0.88678377866745,
      "learning_rate": 0.0001998615919836726,
      "loss": 0.4947,
      "step": 17080
    },
    {
      "epoch": 0.19786504885842635,
      "grad_norm": 0.8654029965400696,
      "learning_rate": 0.00019986091993063267,
      "loss": 0.4614,
      "step": 17090
    },
    {
      "epoch": 0.19798082711989998,
      "grad_norm": 0.9105015993118286,
      "learning_rate": 0.00019986024625106926,
      "loss": 0.4645,
      "step": 17100
    },
    {
      "epoch": 0.19809660538137358,
      "grad_norm": 1.0340826511383057,
      "learning_rate": 0.0001998595709449933,
      "loss": 0.4776,
      "step": 17110
    },
    {
      "epoch": 0.1982123836428472,
      "grad_norm": 0.7128646969795227,
      "learning_rate": 0.00019985889401241585,
      "loss": 0.4806,
      "step": 17120
    },
    {
      "epoch": 0.19832816190432084,
      "grad_norm": 0.893755316734314,
      "learning_rate": 0.00019985821545334787,
      "loss": 0.4744,
      "step": 17130
    },
    {
      "epoch": 0.19844394016579447,
      "grad_norm": 0.7779598236083984,
      "learning_rate": 0.00019985753526780044,
      "loss": 0.475,
      "step": 17140
    },
    {
      "epoch": 0.1985597184272681,
      "grad_norm": 1.0338503122329712,
      "learning_rate": 0.00019985685345578466,
      "loss": 0.4962,
      "step": 17150
    },
    {
      "epoch": 0.1986754966887417,
      "grad_norm": 0.6588304042816162,
      "learning_rate": 0.00019985617001731164,
      "loss": 0.4717,
      "step": 17160
    },
    {
      "epoch": 0.19879127495021534,
      "grad_norm": 0.7454726696014404,
      "learning_rate": 0.00019985548495239246,
      "loss": 0.4616,
      "step": 17170
    },
    {
      "epoch": 0.19890705321168897,
      "grad_norm": 0.8275521397590637,
      "learning_rate": 0.0001998547982610383,
      "loss": 0.4672,
      "step": 17180
    },
    {
      "epoch": 0.1990228314731626,
      "grad_norm": 0.9495667815208435,
      "learning_rate": 0.0001998541099432604,
      "loss": 0.5069,
      "step": 17190
    },
    {
      "epoch": 0.19913860973463623,
      "grad_norm": 1.0234605073928833,
      "learning_rate": 0.0001998534199990699,
      "loss": 0.4999,
      "step": 17200
    },
    {
      "epoch": 0.19925438799610984,
      "grad_norm": 0.9439220428466797,
      "learning_rate": 0.00019985272842847808,
      "loss": 0.4952,
      "step": 17210
    },
    {
      "epoch": 0.19937016625758347,
      "grad_norm": 0.7583577036857605,
      "learning_rate": 0.00019985203523149617,
      "loss": 0.4943,
      "step": 17220
    },
    {
      "epoch": 0.1994859445190571,
      "grad_norm": 0.9253888726234436,
      "learning_rate": 0.00019985134040813546,
      "loss": 0.4831,
      "step": 17230
    },
    {
      "epoch": 0.19960172278053073,
      "grad_norm": 0.8285670876502991,
      "learning_rate": 0.00019985064395840733,
      "loss": 0.4872,
      "step": 17240
    },
    {
      "epoch": 0.19971750104200436,
      "grad_norm": 0.8756377100944519,
      "learning_rate": 0.00019984994588232306,
      "loss": 0.5228,
      "step": 17250
    },
    {
      "epoch": 0.19983327930347797,
      "grad_norm": 0.9798482060432434,
      "learning_rate": 0.00019984924617989403,
      "loss": 0.5004,
      "step": 17260
    },
    {
      "epoch": 0.1999490575649516,
      "grad_norm": 0.7798986434936523,
      "learning_rate": 0.00019984854485113166,
      "loss": 0.4693,
      "step": 17270
    },
    {
      "epoch": 0.20006483582642523,
      "grad_norm": 0.7864567041397095,
      "learning_rate": 0.00019984784189604736,
      "loss": 0.5029,
      "step": 17280
    },
    {
      "epoch": 0.20018061408789886,
      "grad_norm": 0.8364198207855225,
      "learning_rate": 0.00019984713731465255,
      "loss": 0.4862,
      "step": 17290
    },
    {
      "epoch": 0.2002963923493725,
      "grad_norm": 0.8812498450279236,
      "learning_rate": 0.00019984643110695878,
      "loss": 0.4693,
      "step": 17300
    },
    {
      "epoch": 0.2004121706108461,
      "grad_norm": 0.7478137016296387,
      "learning_rate": 0.0001998457232729775,
      "loss": 0.5226,
      "step": 17310
    },
    {
      "epoch": 0.20052794887231973,
      "grad_norm": 0.8203095197677612,
      "learning_rate": 0.00019984501381272024,
      "loss": 0.5393,
      "step": 17320
    },
    {
      "epoch": 0.20064372713379336,
      "grad_norm": 0.760674238204956,
      "learning_rate": 0.00019984430272619857,
      "loss": 0.4953,
      "step": 17330
    },
    {
      "epoch": 0.200759505395267,
      "grad_norm": 1.0259523391723633,
      "learning_rate": 0.00019984359001342403,
      "loss": 0.4814,
      "step": 17340
    },
    {
      "epoch": 0.20087528365674062,
      "grad_norm": 0.8328045010566711,
      "learning_rate": 0.0001998428756744083,
      "loss": 0.4711,
      "step": 17350
    },
    {
      "epoch": 0.20099106191821423,
      "grad_norm": 0.8433694839477539,
      "learning_rate": 0.00019984215970916296,
      "loss": 0.4699,
      "step": 17360
    },
    {
      "epoch": 0.20110684017968786,
      "grad_norm": 0.8189342021942139,
      "learning_rate": 0.00019984144211769968,
      "loss": 0.4815,
      "step": 17370
    },
    {
      "epoch": 0.2012226184411615,
      "grad_norm": 0.7954742908477783,
      "learning_rate": 0.00019984072290003019,
      "loss": 0.5222,
      "step": 17380
    },
    {
      "epoch": 0.20133839670263512,
      "grad_norm": 0.7419716715812683,
      "learning_rate": 0.00019984000205616615,
      "loss": 0.5511,
      "step": 17390
    },
    {
      "epoch": 0.20145417496410875,
      "grad_norm": 0.771791934967041,
      "learning_rate": 0.00019983927958611933,
      "loss": 0.4767,
      "step": 17400
    },
    {
      "epoch": 0.20156995322558235,
      "grad_norm": 0.7349338531494141,
      "learning_rate": 0.00019983855548990148,
      "loss": 0.4666,
      "step": 17410
    },
    {
      "epoch": 0.20168573148705599,
      "grad_norm": 0.6626814007759094,
      "learning_rate": 0.00019983782976752443,
      "loss": 0.4636,
      "step": 17420
    },
    {
      "epoch": 0.20180150974852962,
      "grad_norm": 0.7729905247688293,
      "learning_rate": 0.00019983710241899995,
      "loss": 0.471,
      "step": 17430
    },
    {
      "epoch": 0.20191728801000325,
      "grad_norm": 0.9113923907279968,
      "learning_rate": 0.00019983637344433993,
      "loss": 0.4673,
      "step": 17440
    },
    {
      "epoch": 0.20203306627147688,
      "grad_norm": 0.867956280708313,
      "learning_rate": 0.0001998356428435562,
      "loss": 0.4894,
      "step": 17450
    },
    {
      "epoch": 0.20214884453295048,
      "grad_norm": 0.9342520833015442,
      "learning_rate": 0.00019983491061666074,
      "loss": 0.4456,
      "step": 17460
    },
    {
      "epoch": 0.20226462279442411,
      "grad_norm": 0.7759660482406616,
      "learning_rate": 0.0001998341767636654,
      "loss": 0.494,
      "step": 17470
    },
    {
      "epoch": 0.20238040105589775,
      "grad_norm": 0.639121413230896,
      "learning_rate": 0.0001998334412845821,
      "loss": 0.5031,
      "step": 17480
    },
    {
      "epoch": 0.20249617931737138,
      "grad_norm": 0.8816165924072266,
      "learning_rate": 0.00019983270417942294,
      "loss": 0.4924,
      "step": 17490
    },
    {
      "epoch": 0.202611957578845,
      "grad_norm": 0.9371243715286255,
      "learning_rate": 0.00019983196544819985,
      "loss": 0.5222,
      "step": 17500
    },
    {
      "epoch": 0.2027277358403186,
      "grad_norm": 0.8375434875488281,
      "learning_rate": 0.00019983122509092485,
      "loss": 0.5014,
      "step": 17510
    },
    {
      "epoch": 0.20284351410179224,
      "grad_norm": 0.7366251349449158,
      "learning_rate": 0.00019983048310761003,
      "loss": 0.4943,
      "step": 17520
    },
    {
      "epoch": 0.20295929236326588,
      "grad_norm": 0.9275548458099365,
      "learning_rate": 0.00019982973949826747,
      "loss": 0.4966,
      "step": 17530
    },
    {
      "epoch": 0.2030750706247395,
      "grad_norm": 0.9041328430175781,
      "learning_rate": 0.0001998289942629093,
      "loss": 0.461,
      "step": 17540
    },
    {
      "epoch": 0.20319084888621314,
      "grad_norm": 0.8144553303718567,
      "learning_rate": 0.00019982824740154758,
      "loss": 0.5008,
      "step": 17550
    },
    {
      "epoch": 0.20330662714768674,
      "grad_norm": 0.8701579570770264,
      "learning_rate": 0.0001998274989141946,
      "loss": 0.5044,
      "step": 17560
    },
    {
      "epoch": 0.20342240540916037,
      "grad_norm": 0.8747161626815796,
      "learning_rate": 0.00019982674880086244,
      "loss": 0.5066,
      "step": 17570
    },
    {
      "epoch": 0.203538183670634,
      "grad_norm": 0.9712308645248413,
      "learning_rate": 0.00019982599706156338,
      "loss": 0.4881,
      "step": 17580
    },
    {
      "epoch": 0.20365396193210764,
      "grad_norm": 0.8672597408294678,
      "learning_rate": 0.00019982524369630965,
      "loss": 0.4826,
      "step": 17590
    },
    {
      "epoch": 0.20376974019358127,
      "grad_norm": 1.026075839996338,
      "learning_rate": 0.0001998244887051135,
      "loss": 0.5139,
      "step": 17600
    },
    {
      "epoch": 0.20388551845505487,
      "grad_norm": 0.7845612168312073,
      "learning_rate": 0.00019982373208798725,
      "loss": 0.4905,
      "step": 17610
    },
    {
      "epoch": 0.2040012967165285,
      "grad_norm": 0.8791155219078064,
      "learning_rate": 0.00019982297384494322,
      "loss": 0.4786,
      "step": 17620
    },
    {
      "epoch": 0.20411707497800213,
      "grad_norm": 0.8501670956611633,
      "learning_rate": 0.00019982221397599374,
      "loss": 0.5248,
      "step": 17630
    },
    {
      "epoch": 0.20423285323947576,
      "grad_norm": 0.7558290362358093,
      "learning_rate": 0.00019982145248115122,
      "loss": 0.4341,
      "step": 17640
    },
    {
      "epoch": 0.20434863150094937,
      "grad_norm": 0.9865317940711975,
      "learning_rate": 0.00019982068936042806,
      "loss": 0.4704,
      "step": 17650
    },
    {
      "epoch": 0.204464409762423,
      "grad_norm": 0.837709367275238,
      "learning_rate": 0.00019981992461383665,
      "loss": 0.4333,
      "step": 17660
    },
    {
      "epoch": 0.20458018802389663,
      "grad_norm": 0.9652982950210571,
      "learning_rate": 0.00019981915824138948,
      "loss": 0.4914,
      "step": 17670
    },
    {
      "epoch": 0.20469596628537026,
      "grad_norm": 0.9558224678039551,
      "learning_rate": 0.00019981839024309906,
      "loss": 0.4998,
      "step": 17680
    },
    {
      "epoch": 0.2048117445468439,
      "grad_norm": 0.8586015105247498,
      "learning_rate": 0.00019981762061897781,
      "loss": 0.532,
      "step": 17690
    },
    {
      "epoch": 0.2049275228083175,
      "grad_norm": 0.7897942662239075,
      "learning_rate": 0.00019981684936903836,
      "loss": 0.4904,
      "step": 17700
    },
    {
      "epoch": 0.20504330106979113,
      "grad_norm": 0.8042513728141785,
      "learning_rate": 0.0001998160764932932,
      "loss": 0.4754,
      "step": 17710
    },
    {
      "epoch": 0.20515907933126476,
      "grad_norm": 0.8123723268508911,
      "learning_rate": 0.00019981530199175498,
      "loss": 0.5266,
      "step": 17720
    },
    {
      "epoch": 0.2052748575927384,
      "grad_norm": 0.7880979180335999,
      "learning_rate": 0.00019981452586443626,
      "loss": 0.5238,
      "step": 17730
    },
    {
      "epoch": 0.20539063585421202,
      "grad_norm": 0.7563484907150269,
      "learning_rate": 0.00019981374811134974,
      "loss": 0.4434,
      "step": 17740
    },
    {
      "epoch": 0.20550641411568563,
      "grad_norm": 0.8079185485839844,
      "learning_rate": 0.000199812968732508,
      "loss": 0.469,
      "step": 17750
    },
    {
      "epoch": 0.20562219237715926,
      "grad_norm": 0.8503212928771973,
      "learning_rate": 0.00019981218772792384,
      "loss": 0.46,
      "step": 17760
    },
    {
      "epoch": 0.2057379706386329,
      "grad_norm": 0.7510192394256592,
      "learning_rate": 0.0001998114050976099,
      "loss": 0.4752,
      "step": 17770
    },
    {
      "epoch": 0.20585374890010652,
      "grad_norm": 0.7266948819160461,
      "learning_rate": 0.000199810620841579,
      "loss": 0.5079,
      "step": 17780
    },
    {
      "epoch": 0.20596952716158015,
      "grad_norm": 0.8466536402702332,
      "learning_rate": 0.00019980983495984384,
      "loss": 0.5255,
      "step": 17790
    },
    {
      "epoch": 0.20608530542305376,
      "grad_norm": 0.940623939037323,
      "learning_rate": 0.00019980904745241725,
      "loss": 0.5063,
      "step": 17800
    },
    {
      "epoch": 0.2062010836845274,
      "grad_norm": 0.5855724811553955,
      "learning_rate": 0.00019980825831931206,
      "loss": 0.4807,
      "step": 17810
    },
    {
      "epoch": 0.20631686194600102,
      "grad_norm": 0.8663539886474609,
      "learning_rate": 0.00019980746756054113,
      "loss": 0.483,
      "step": 17820
    },
    {
      "epoch": 0.20643264020747465,
      "grad_norm": 0.8698544502258301,
      "learning_rate": 0.00019980667517611732,
      "loss": 0.505,
      "step": 17830
    },
    {
      "epoch": 0.20654841846894828,
      "grad_norm": 0.8699235916137695,
      "learning_rate": 0.00019980588116605356,
      "loss": 0.4785,
      "step": 17840
    },
    {
      "epoch": 0.20666419673042188,
      "grad_norm": 0.8365659117698669,
      "learning_rate": 0.00019980508553036277,
      "loss": 0.5078,
      "step": 17850
    },
    {
      "epoch": 0.20677997499189552,
      "grad_norm": 0.905035674571991,
      "learning_rate": 0.00019980428826905792,
      "loss": 0.5099,
      "step": 17860
    },
    {
      "epoch": 0.20689575325336915,
      "grad_norm": 0.8892579078674316,
      "learning_rate": 0.00019980348938215196,
      "loss": 0.5173,
      "step": 17870
    },
    {
      "epoch": 0.20701153151484278,
      "grad_norm": 0.8087046146392822,
      "learning_rate": 0.00019980268886965794,
      "loss": 0.4651,
      "step": 17880
    },
    {
      "epoch": 0.2071273097763164,
      "grad_norm": 0.8616148233413696,
      "learning_rate": 0.0001998018867315889,
      "loss": 0.4701,
      "step": 17890
    },
    {
      "epoch": 0.20724308803779,
      "grad_norm": 0.8882416486740112,
      "learning_rate": 0.00019980108296795788,
      "loss": 0.4596,
      "step": 17900
    },
    {
      "epoch": 0.20735886629926364,
      "grad_norm": 0.9214577078819275,
      "learning_rate": 0.000199800277578778,
      "loss": 0.4697,
      "step": 17910
    },
    {
      "epoch": 0.20747464456073728,
      "grad_norm": 0.976055920124054,
      "learning_rate": 0.00019979947056406231,
      "loss": 0.509,
      "step": 17920
    },
    {
      "epoch": 0.2075904228222109,
      "grad_norm": 0.8006547689437866,
      "learning_rate": 0.00019979866192382408,
      "loss": 0.4616,
      "step": 17930
    },
    {
      "epoch": 0.20770620108368454,
      "grad_norm": 0.8892197608947754,
      "learning_rate": 0.00019979785165807635,
      "loss": 0.4711,
      "step": 17940
    },
    {
      "epoch": 0.20782197934515814,
      "grad_norm": 1.1781281232833862,
      "learning_rate": 0.0001997970397668324,
      "loss": 0.5026,
      "step": 17950
    },
    {
      "epoch": 0.20793775760663177,
      "grad_norm": 0.8484392762184143,
      "learning_rate": 0.00019979622625010543,
      "loss": 0.5454,
      "step": 17960
    },
    {
      "epoch": 0.2080535358681054,
      "grad_norm": 0.953849196434021,
      "learning_rate": 0.00019979541110790868,
      "loss": 0.5073,
      "step": 17970
    },
    {
      "epoch": 0.20816931412957904,
      "grad_norm": 0.8298689126968384,
      "learning_rate": 0.00019979459434025542,
      "loss": 0.5069,
      "step": 17980
    },
    {
      "epoch": 0.20828509239105267,
      "grad_norm": 0.8336108326911926,
      "learning_rate": 0.00019979377594715901,
      "loss": 0.4409,
      "step": 17990
    },
    {
      "epoch": 0.20840087065252627,
      "grad_norm": 0.8721938729286194,
      "learning_rate": 0.0001997929559286327,
      "loss": 0.4647,
      "step": 18000
    },
    {
      "epoch": 0.20840087065252627,
      "eval_chrf": 79.00803519103351,
      "eval_loss": 0.7419794201850891,
      "eval_runtime": 300.3688,
      "eval_samples_per_second": 0.333,
      "eval_steps_per_second": 0.013,
      "step": 18000
    },
    {
      "epoch": 0.2085166489139999,
      "grad_norm": 0.8734286427497864,
      "learning_rate": 0.0001997921342846899,
      "loss": 0.4872,
      "step": 18010
    },
    {
      "epoch": 0.20863242717547353,
      "grad_norm": 0.7984320521354675,
      "learning_rate": 0.00019979131101534398,
      "loss": 0.5083,
      "step": 18020
    },
    {
      "epoch": 0.20874820543694717,
      "grad_norm": 0.7219537496566772,
      "learning_rate": 0.00019979048612060835,
      "loss": 0.5315,
      "step": 18030
    },
    {
      "epoch": 0.2088639836984208,
      "grad_norm": 0.909748911857605,
      "learning_rate": 0.00019978965960049645,
      "loss": 0.4868,
      "step": 18040
    },
    {
      "epoch": 0.2089797619598944,
      "grad_norm": 0.8623420596122742,
      "learning_rate": 0.00019978883145502173,
      "loss": 0.4723,
      "step": 18050
    },
    {
      "epoch": 0.20909554022136803,
      "grad_norm": 0.7724198698997498,
      "learning_rate": 0.00019978800168419765,
      "loss": 0.4802,
      "step": 18060
    },
    {
      "epoch": 0.20921131848284166,
      "grad_norm": 0.806288480758667,
      "learning_rate": 0.0001997871702880378,
      "loss": 0.519,
      "step": 18070
    },
    {
      "epoch": 0.2093270967443153,
      "grad_norm": 0.8182432055473328,
      "learning_rate": 0.0001997863372665557,
      "loss": 0.4862,
      "step": 18080
    },
    {
      "epoch": 0.20944287500578893,
      "grad_norm": 1.0057041645050049,
      "learning_rate": 0.00019978550261976486,
      "loss": 0.481,
      "step": 18090
    },
    {
      "epoch": 0.20955865326726253,
      "grad_norm": 0.8643611669540405,
      "learning_rate": 0.00019978466634767894,
      "loss": 0.4607,
      "step": 18100
    },
    {
      "epoch": 0.20967443152873616,
      "grad_norm": 0.7916561961174011,
      "learning_rate": 0.0001997838284503115,
      "loss": 0.4953,
      "step": 18110
    },
    {
      "epoch": 0.2097902097902098,
      "grad_norm": 0.8533985018730164,
      "learning_rate": 0.00019978298892767626,
      "loss": 0.4907,
      "step": 18120
    },
    {
      "epoch": 0.20990598805168342,
      "grad_norm": 0.7255854606628418,
      "learning_rate": 0.00019978214777978685,
      "loss": 0.5264,
      "step": 18130
    },
    {
      "epoch": 0.21002176631315705,
      "grad_norm": 0.6964989304542542,
      "learning_rate": 0.00019978130500665698,
      "loss": 0.484,
      "step": 18140
    },
    {
      "epoch": 0.21013754457463066,
      "grad_norm": 0.8206535577774048,
      "learning_rate": 0.00019978046060830035,
      "loss": 0.4833,
      "step": 18150
    },
    {
      "epoch": 0.2102533228361043,
      "grad_norm": 0.9355806112289429,
      "learning_rate": 0.0001997796145847308,
      "loss": 0.4838,
      "step": 18160
    },
    {
      "epoch": 0.21036910109757792,
      "grad_norm": 0.9480417370796204,
      "learning_rate": 0.00019977876693596196,
      "loss": 0.4952,
      "step": 18170
    },
    {
      "epoch": 0.21048487935905155,
      "grad_norm": 0.7078197002410889,
      "learning_rate": 0.00019977791766200776,
      "loss": 0.4418,
      "step": 18180
    },
    {
      "epoch": 0.21060065762052518,
      "grad_norm": 0.9125195741653442,
      "learning_rate": 0.000199777066762882,
      "loss": 0.5197,
      "step": 18190
    },
    {
      "epoch": 0.2107164358819988,
      "grad_norm": 0.959878146648407,
      "learning_rate": 0.00019977621423859855,
      "loss": 0.4901,
      "step": 18200
    },
    {
      "epoch": 0.21083221414347242,
      "grad_norm": 0.9553471803665161,
      "learning_rate": 0.00019977536008917128,
      "loss": 0.4638,
      "step": 18210
    },
    {
      "epoch": 0.21094799240494605,
      "grad_norm": 0.744727611541748,
      "learning_rate": 0.0001997745043146141,
      "loss": 0.4888,
      "step": 18220
    },
    {
      "epoch": 0.21106377066641968,
      "grad_norm": 0.9068858623504639,
      "learning_rate": 0.00019977364691494095,
      "loss": 0.5268,
      "step": 18230
    },
    {
      "epoch": 0.2111795489278933,
      "grad_norm": 0.9972188472747803,
      "learning_rate": 0.0001997727878901658,
      "loss": 0.4882,
      "step": 18240
    },
    {
      "epoch": 0.21129532718936692,
      "grad_norm": 0.8237280249595642,
      "learning_rate": 0.00019977192724030264,
      "loss": 0.4895,
      "step": 18250
    },
    {
      "epoch": 0.21141110545084055,
      "grad_norm": 0.900851309299469,
      "learning_rate": 0.0001997710649653655,
      "loss": 0.4715,
      "step": 18260
    },
    {
      "epoch": 0.21152688371231418,
      "grad_norm": 0.9079288840293884,
      "learning_rate": 0.00019977020106536842,
      "loss": 0.4624,
      "step": 18270
    },
    {
      "epoch": 0.2116426619737878,
      "grad_norm": 0.8853783011436462,
      "learning_rate": 0.00019976933554032546,
      "loss": 0.486,
      "step": 18280
    },
    {
      "epoch": 0.2117584402352614,
      "grad_norm": 0.7894467115402222,
      "learning_rate": 0.00019976846839025067,
      "loss": 0.5234,
      "step": 18290
    },
    {
      "epoch": 0.21187421849673505,
      "grad_norm": 0.9138008952140808,
      "learning_rate": 0.00019976759961515828,
      "loss": 0.5016,
      "step": 18300
    },
    {
      "epoch": 0.21198999675820868,
      "grad_norm": 1.0169129371643066,
      "learning_rate": 0.00019976672921506237,
      "loss": 0.4866,
      "step": 18310
    },
    {
      "epoch": 0.2121057750196823,
      "grad_norm": 0.8770890831947327,
      "learning_rate": 0.00019976585718997714,
      "loss": 0.4903,
      "step": 18320
    },
    {
      "epoch": 0.21222155328115594,
      "grad_norm": 0.8056181073188782,
      "learning_rate": 0.00019976498353991678,
      "loss": 0.5208,
      "step": 18330
    },
    {
      "epoch": 0.21233733154262954,
      "grad_norm": 0.8492462635040283,
      "learning_rate": 0.0001997641082648955,
      "loss": 0.4765,
      "step": 18340
    },
    {
      "epoch": 0.21245310980410317,
      "grad_norm": 0.8984133005142212,
      "learning_rate": 0.00019976323136492761,
      "loss": 0.4735,
      "step": 18350
    },
    {
      "epoch": 0.2125688880655768,
      "grad_norm": 0.8451970815658569,
      "learning_rate": 0.0001997623528400274,
      "loss": 0.4831,
      "step": 18360
    },
    {
      "epoch": 0.21268466632705044,
      "grad_norm": 0.8115382194519043,
      "learning_rate": 0.00019976147269020912,
      "loss": 0.5128,
      "step": 18370
    },
    {
      "epoch": 0.21280044458852407,
      "grad_norm": 0.8258618712425232,
      "learning_rate": 0.0001997605909154871,
      "loss": 0.4956,
      "step": 18380
    },
    {
      "epoch": 0.21291622284999767,
      "grad_norm": 0.8482604622840881,
      "learning_rate": 0.00019975970751587578,
      "loss": 0.4774,
      "step": 18390
    },
    {
      "epoch": 0.2130320011114713,
      "grad_norm": 0.767049252986908,
      "learning_rate": 0.00019975882249138947,
      "loss": 0.5041,
      "step": 18400
    },
    {
      "epoch": 0.21314777937294493,
      "grad_norm": 0.8439600467681885,
      "learning_rate": 0.00019975793584204262,
      "loss": 0.478,
      "step": 18410
    },
    {
      "epoch": 0.21326355763441857,
      "grad_norm": 0.9389786720275879,
      "learning_rate": 0.0001997570475678497,
      "loss": 0.4666,
      "step": 18420
    },
    {
      "epoch": 0.2133793358958922,
      "grad_norm": 0.843723475933075,
      "learning_rate": 0.00019975615766882512,
      "loss": 0.4556,
      "step": 18430
    },
    {
      "epoch": 0.2134951141573658,
      "grad_norm": 0.931433379650116,
      "learning_rate": 0.0001997552661449834,
      "loss": 0.5002,
      "step": 18440
    },
    {
      "epoch": 0.21361089241883943,
      "grad_norm": 0.8732482194900513,
      "learning_rate": 0.00019975437299633907,
      "loss": 0.4442,
      "step": 18450
    },
    {
      "epoch": 0.21372667068031306,
      "grad_norm": 0.8885776400566101,
      "learning_rate": 0.0001997534782229067,
      "loss": 0.4902,
      "step": 18460
    },
    {
      "epoch": 0.2138424489417867,
      "grad_norm": 0.8338961601257324,
      "learning_rate": 0.00019975258182470079,
      "loss": 0.4869,
      "step": 18470
    },
    {
      "epoch": 0.21395822720326033,
      "grad_norm": 0.8058544993400574,
      "learning_rate": 0.00019975168380173598,
      "loss": 0.495,
      "step": 18480
    },
    {
      "epoch": 0.21407400546473393,
      "grad_norm": 0.8691881895065308,
      "learning_rate": 0.00019975078415402693,
      "loss": 0.4589,
      "step": 18490
    },
    {
      "epoch": 0.21418978372620756,
      "grad_norm": 0.8180102705955505,
      "learning_rate": 0.00019974988288158825,
      "loss": 0.5018,
      "step": 18500
    },
    {
      "epoch": 0.2143055619876812,
      "grad_norm": 0.7971519827842712,
      "learning_rate": 0.00019974897998443464,
      "loss": 0.4406,
      "step": 18510
    },
    {
      "epoch": 0.21442134024915482,
      "grad_norm": 0.9783874154090881,
      "learning_rate": 0.00019974807546258082,
      "loss": 0.485,
      "step": 18520
    },
    {
      "epoch": 0.21453711851062846,
      "grad_norm": 0.9014028310775757,
      "learning_rate": 0.00019974716931604148,
      "loss": 0.5166,
      "step": 18530
    },
    {
      "epoch": 0.21465289677210206,
      "grad_norm": 0.8176358342170715,
      "learning_rate": 0.0001997462615448314,
      "loss": 0.4718,
      "step": 18540
    },
    {
      "epoch": 0.2147686750335757,
      "grad_norm": 0.8621872663497925,
      "learning_rate": 0.00019974535214896538,
      "loss": 0.4862,
      "step": 18550
    },
    {
      "epoch": 0.21488445329504932,
      "grad_norm": 0.9799246191978455,
      "learning_rate": 0.00019974444112845821,
      "loss": 0.4704,
      "step": 18560
    },
    {
      "epoch": 0.21500023155652295,
      "grad_norm": 0.8795552253723145,
      "learning_rate": 0.00019974352848332476,
      "loss": 0.5188,
      "step": 18570
    },
    {
      "epoch": 0.21511600981799658,
      "grad_norm": 0.8586767315864563,
      "learning_rate": 0.00019974261421357986,
      "loss": 0.4803,
      "step": 18580
    },
    {
      "epoch": 0.2152317880794702,
      "grad_norm": 0.9071685075759888,
      "learning_rate": 0.0001997416983192384,
      "loss": 0.4896,
      "step": 18590
    },
    {
      "epoch": 0.21534756634094382,
      "grad_norm": 0.7288635969161987,
      "learning_rate": 0.00019974078080031534,
      "loss": 0.5119,
      "step": 18600
    },
    {
      "epoch": 0.21546334460241745,
      "grad_norm": 0.7109447717666626,
      "learning_rate": 0.00019973986165682555,
      "loss": 0.4497,
      "step": 18610
    },
    {
      "epoch": 0.21557912286389108,
      "grad_norm": 0.8410625457763672,
      "learning_rate": 0.0001997389408887841,
      "loss": 0.5218,
      "step": 18620
    },
    {
      "epoch": 0.2156949011253647,
      "grad_norm": 0.9533514380455017,
      "learning_rate": 0.00019973801849620595,
      "loss": 0.4755,
      "step": 18630
    },
    {
      "epoch": 0.21581067938683832,
      "grad_norm": 0.7994857430458069,
      "learning_rate": 0.00019973709447910606,
      "loss": 0.4778,
      "step": 18640
    },
    {
      "epoch": 0.21592645764831195,
      "grad_norm": 0.7716197371482849,
      "learning_rate": 0.00019973616883749956,
      "loss": 0.4969,
      "step": 18650
    },
    {
      "epoch": 0.21604223590978558,
      "grad_norm": 0.7360429167747498,
      "learning_rate": 0.0001997352415714015,
      "loss": 0.4673,
      "step": 18660
    },
    {
      "epoch": 0.2161580141712592,
      "grad_norm": 0.9170987606048584,
      "learning_rate": 0.00019973431268082697,
      "loss": 0.4777,
      "step": 18670
    },
    {
      "epoch": 0.21627379243273284,
      "grad_norm": 0.9852982759475708,
      "learning_rate": 0.00019973338216579113,
      "loss": 0.481,
      "step": 18680
    },
    {
      "epoch": 0.21638957069420645,
      "grad_norm": 0.9320043921470642,
      "learning_rate": 0.0001997324500263091,
      "loss": 0.4411,
      "step": 18690
    },
    {
      "epoch": 0.21650534895568008,
      "grad_norm": 0.836481511592865,
      "learning_rate": 0.0001997315162623961,
      "loss": 0.4544,
      "step": 18700
    },
    {
      "epoch": 0.2166211272171537,
      "grad_norm": 0.848663866519928,
      "learning_rate": 0.00019973058087406726,
      "loss": 0.4626,
      "step": 18710
    },
    {
      "epoch": 0.21673690547862734,
      "grad_norm": 0.8453987240791321,
      "learning_rate": 0.00019972964386133794,
      "loss": 0.4705,
      "step": 18720
    },
    {
      "epoch": 0.21685268374010097,
      "grad_norm": 0.7466351389884949,
      "learning_rate": 0.0001997287052242233,
      "loss": 0.4813,
      "step": 18730
    },
    {
      "epoch": 0.21696846200157457,
      "grad_norm": 0.8212053775787354,
      "learning_rate": 0.00019972776496273865,
      "loss": 0.4623,
      "step": 18740
    },
    {
      "epoch": 0.2170842402630482,
      "grad_norm": 0.8152648210525513,
      "learning_rate": 0.00019972682307689937,
      "loss": 0.4841,
      "step": 18750
    },
    {
      "epoch": 0.21720001852452184,
      "grad_norm": 0.6814244985580444,
      "learning_rate": 0.00019972587956672072,
      "loss": 0.4945,
      "step": 18760
    },
    {
      "epoch": 0.21731579678599547,
      "grad_norm": 0.8202094435691833,
      "learning_rate": 0.0001997249344322181,
      "loss": 0.4876,
      "step": 18770
    },
    {
      "epoch": 0.2174315750474691,
      "grad_norm": 1.152093529701233,
      "learning_rate": 0.00019972398767340689,
      "loss": 0.476,
      "step": 18780
    },
    {
      "epoch": 0.2175473533089427,
      "grad_norm": 0.9209386706352234,
      "learning_rate": 0.00019972303929030255,
      "loss": 0.516,
      "step": 18790
    },
    {
      "epoch": 0.21766313157041633,
      "grad_norm": 0.9491552710533142,
      "learning_rate": 0.00019972208928292046,
      "loss": 0.4581,
      "step": 18800
    },
    {
      "epoch": 0.21777890983188997,
      "grad_norm": 0.8183296322822571,
      "learning_rate": 0.00019972113765127615,
      "loss": 0.4965,
      "step": 18810
    },
    {
      "epoch": 0.2178946880933636,
      "grad_norm": 0.8461251854896545,
      "learning_rate": 0.00019972018439538513,
      "loss": 0.4803,
      "step": 18820
    },
    {
      "epoch": 0.21801046635483723,
      "grad_norm": 0.7568778991699219,
      "learning_rate": 0.00019971922951526286,
      "loss": 0.4944,
      "step": 18830
    },
    {
      "epoch": 0.21812624461631083,
      "grad_norm": 1.012679934501648,
      "learning_rate": 0.00019971827301092497,
      "loss": 0.4773,
      "step": 18840
    },
    {
      "epoch": 0.21824202287778446,
      "grad_norm": 0.7288202047348022,
      "learning_rate": 0.00019971731488238698,
      "loss": 0.4933,
      "step": 18850
    },
    {
      "epoch": 0.2183578011392581,
      "grad_norm": 0.799356997013092,
      "learning_rate": 0.00019971635512966454,
      "loss": 0.5043,
      "step": 18860
    },
    {
      "epoch": 0.21847357940073173,
      "grad_norm": 0.796561598777771,
      "learning_rate": 0.00019971539375277325,
      "loss": 0.5033,
      "step": 18870
    },
    {
      "epoch": 0.21858935766220536,
      "grad_norm": 0.9243367314338684,
      "learning_rate": 0.00019971443075172878,
      "loss": 0.4881,
      "step": 18880
    },
    {
      "epoch": 0.21870513592367896,
      "grad_norm": 0.8448047637939453,
      "learning_rate": 0.00019971346612654683,
      "loss": 0.4917,
      "step": 18890
    },
    {
      "epoch": 0.2188209141851526,
      "grad_norm": 0.8143089413642883,
      "learning_rate": 0.00019971249987724306,
      "loss": 0.4709,
      "step": 18900
    },
    {
      "epoch": 0.21893669244662622,
      "grad_norm": 0.7472609281539917,
      "learning_rate": 0.00019971153200383328,
      "loss": 0.462,
      "step": 18910
    },
    {
      "epoch": 0.21905247070809986,
      "grad_norm": 0.8620847463607788,
      "learning_rate": 0.00019971056250633318,
      "loss": 0.4915,
      "step": 18920
    },
    {
      "epoch": 0.21916824896957346,
      "grad_norm": 0.8203686475753784,
      "learning_rate": 0.0001997095913847586,
      "loss": 0.4868,
      "step": 18930
    },
    {
      "epoch": 0.2192840272310471,
      "grad_norm": 0.8799416422843933,
      "learning_rate": 0.0001997086186391254,
      "loss": 0.4491,
      "step": 18940
    },
    {
      "epoch": 0.21939980549252072,
      "grad_norm": 0.7626237869262695,
      "learning_rate": 0.00019970764426944932,
      "loss": 0.4582,
      "step": 18950
    },
    {
      "epoch": 0.21951558375399435,
      "grad_norm": 0.9030816555023193,
      "learning_rate": 0.00019970666827574628,
      "loss": 0.4721,
      "step": 18960
    },
    {
      "epoch": 0.21963136201546798,
      "grad_norm": 1.0223565101623535,
      "learning_rate": 0.00019970569065803223,
      "loss": 0.4669,
      "step": 18970
    },
    {
      "epoch": 0.2197471402769416,
      "grad_norm": 0.7570050358772278,
      "learning_rate": 0.000199704711416323,
      "loss": 0.5351,
      "step": 18980
    },
    {
      "epoch": 0.21986291853841522,
      "grad_norm": 0.8550995588302612,
      "learning_rate": 0.00019970373055063458,
      "loss": 0.4914,
      "step": 18990
    },
    {
      "epoch": 0.21997869679988885,
      "grad_norm": 0.7267685532569885,
      "learning_rate": 0.00019970274806098296,
      "loss": 0.4644,
      "step": 19000
    },
    {
      "epoch": 0.21997869679988885,
      "eval_chrf": 84.59165115715571,
      "eval_loss": 0.7306492924690247,
      "eval_runtime": 241.9136,
      "eval_samples_per_second": 0.413,
      "eval_steps_per_second": 0.017,
      "step": 19000
    },
    {
      "epoch": 0.22009447506136248,
      "grad_norm": 0.9555398225784302,
      "learning_rate": 0.00019970176394738413,
      "loss": 0.5118,
      "step": 19010
    },
    {
      "epoch": 0.2202102533228361,
      "grad_norm": 0.9046825766563416,
      "learning_rate": 0.00019970077820985414,
      "loss": 0.5087,
      "step": 19020
    },
    {
      "epoch": 0.22032603158430972,
      "grad_norm": 0.9978909492492676,
      "learning_rate": 0.000199699790848409,
      "loss": 0.4749,
      "step": 19030
    },
    {
      "epoch": 0.22044180984578335,
      "grad_norm": 0.753707230091095,
      "learning_rate": 0.00019969880186306482,
      "loss": 0.4486,
      "step": 19040
    },
    {
      "epoch": 0.22055758810725698,
      "grad_norm": 0.7626103758811951,
      "learning_rate": 0.00019969781125383772,
      "loss": 0.4875,
      "step": 19050
    },
    {
      "epoch": 0.2206733663687306,
      "grad_norm": 0.8806884288787842,
      "learning_rate": 0.00019969681902074382,
      "loss": 0.4667,
      "step": 19060
    },
    {
      "epoch": 0.22078914463020424,
      "grad_norm": 0.8446436524391174,
      "learning_rate": 0.00019969582516379927,
      "loss": 0.4508,
      "step": 19070
    },
    {
      "epoch": 0.22090492289167785,
      "grad_norm": 0.7920851111412048,
      "learning_rate": 0.0001996948296830203,
      "loss": 0.4924,
      "step": 19080
    },
    {
      "epoch": 0.22102070115315148,
      "grad_norm": 0.8616964221000671,
      "learning_rate": 0.00019969383257842303,
      "loss": 0.4902,
      "step": 19090
    },
    {
      "epoch": 0.2211364794146251,
      "grad_norm": 0.9056552052497864,
      "learning_rate": 0.00019969283385002384,
      "loss": 0.4603,
      "step": 19100
    },
    {
      "epoch": 0.22125225767609874,
      "grad_norm": 0.8576582670211792,
      "learning_rate": 0.00019969183349783889,
      "loss": 0.4778,
      "step": 19110
    },
    {
      "epoch": 0.22136803593757237,
      "grad_norm": 0.9580732583999634,
      "learning_rate": 0.00019969083152188453,
      "loss": 0.5163,
      "step": 19120
    },
    {
      "epoch": 0.22148381419904598,
      "grad_norm": 0.8541563153266907,
      "learning_rate": 0.00019968982792217702,
      "loss": 0.4718,
      "step": 19130
    },
    {
      "epoch": 0.2215995924605196,
      "grad_norm": 0.7276837825775146,
      "learning_rate": 0.00019968882269873274,
      "loss": 0.4637,
      "step": 19140
    },
    {
      "epoch": 0.22171537072199324,
      "grad_norm": 0.9170542359352112,
      "learning_rate": 0.0001996878158515681,
      "loss": 0.4708,
      "step": 19150
    },
    {
      "epoch": 0.22183114898346687,
      "grad_norm": 0.843813419342041,
      "learning_rate": 0.00019968680738069946,
      "loss": 0.4738,
      "step": 19160
    },
    {
      "epoch": 0.2219469272449405,
      "grad_norm": 0.8809876441955566,
      "learning_rate": 0.00019968579728614323,
      "loss": 0.4588,
      "step": 19170
    },
    {
      "epoch": 0.2220627055064141,
      "grad_norm": 0.9676681160926819,
      "learning_rate": 0.0001996847855679159,
      "loss": 0.4981,
      "step": 19180
    },
    {
      "epoch": 0.22217848376788774,
      "grad_norm": 0.9200544357299805,
      "learning_rate": 0.00019968377222603394,
      "loss": 0.4495,
      "step": 19190
    },
    {
      "epoch": 0.22229426202936137,
      "grad_norm": 0.6957087516784668,
      "learning_rate": 0.0001996827572605138,
      "loss": 0.4642,
      "step": 19200
    },
    {
      "epoch": 0.222410040290835,
      "grad_norm": 0.7724372744560242,
      "learning_rate": 0.0001996817406713721,
      "loss": 0.481,
      "step": 19210
    },
    {
      "epoch": 0.22252581855230863,
      "grad_norm": 0.7524381875991821,
      "learning_rate": 0.00019968072245862538,
      "loss": 0.4778,
      "step": 19220
    },
    {
      "epoch": 0.22264159681378223,
      "grad_norm": 0.8573737740516663,
      "learning_rate": 0.00019967970262229015,
      "loss": 0.4721,
      "step": 19230
    },
    {
      "epoch": 0.22275737507525586,
      "grad_norm": 0.7550068497657776,
      "learning_rate": 0.0001996786811623831,
      "loss": 0.4876,
      "step": 19240
    },
    {
      "epoch": 0.2228731533367295,
      "grad_norm": 0.7729037404060364,
      "learning_rate": 0.00019967765807892086,
      "loss": 0.4875,
      "step": 19250
    },
    {
      "epoch": 0.22298893159820313,
      "grad_norm": 0.7761761546134949,
      "learning_rate": 0.00019967663337192005,
      "loss": 0.4485,
      "step": 19260
    },
    {
      "epoch": 0.22310470985967676,
      "grad_norm": 0.9230116009712219,
      "learning_rate": 0.0001996756070413974,
      "loss": 0.4884,
      "step": 19270
    },
    {
      "epoch": 0.22322048812115036,
      "grad_norm": 0.928087592124939,
      "learning_rate": 0.00019967457908736958,
      "loss": 0.5024,
      "step": 19280
    },
    {
      "epoch": 0.223336266382624,
      "grad_norm": 0.7160900235176086,
      "learning_rate": 0.00019967354950985338,
      "loss": 0.4713,
      "step": 19290
    },
    {
      "epoch": 0.22345204464409762,
      "grad_norm": 0.8185821175575256,
      "learning_rate": 0.00019967251830886555,
      "loss": 0.4789,
      "step": 19300
    },
    {
      "epoch": 0.22356782290557126,
      "grad_norm": 0.8640700578689575,
      "learning_rate": 0.0001996714854844229,
      "loss": 0.4669,
      "step": 19310
    },
    {
      "epoch": 0.2236836011670449,
      "grad_norm": 0.9034169316291809,
      "learning_rate": 0.00019967045103654223,
      "loss": 0.5005,
      "step": 19320
    },
    {
      "epoch": 0.2237993794285185,
      "grad_norm": 0.6736032962799072,
      "learning_rate": 0.00019966941496524042,
      "loss": 0.496,
      "step": 19330
    },
    {
      "epoch": 0.22391515768999212,
      "grad_norm": 0.8957875370979309,
      "learning_rate": 0.00019966837727053433,
      "loss": 0.542,
      "step": 19340
    },
    {
      "epoch": 0.22403093595146575,
      "grad_norm": 0.8206246495246887,
      "learning_rate": 0.00019966733795244082,
      "loss": 0.4916,
      "step": 19350
    },
    {
      "epoch": 0.22414671421293939,
      "grad_norm": 0.694081723690033,
      "learning_rate": 0.0001996662970109769,
      "loss": 0.4897,
      "step": 19360
    },
    {
      "epoch": 0.22426249247441302,
      "grad_norm": 0.8788129091262817,
      "learning_rate": 0.00019966525444615948,
      "loss": 0.4645,
      "step": 19370
    },
    {
      "epoch": 0.22437827073588662,
      "grad_norm": 0.9289214015007019,
      "learning_rate": 0.00019966421025800554,
      "loss": 0.4688,
      "step": 19380
    },
    {
      "epoch": 0.22449404899736025,
      "grad_norm": 0.8340096473693848,
      "learning_rate": 0.00019966316444653206,
      "loss": 0.4795,
      "step": 19390
    },
    {
      "epoch": 0.22460982725883388,
      "grad_norm": 0.8196728229522705,
      "learning_rate": 0.00019966211701175613,
      "loss": 0.5069,
      "step": 19400
    },
    {
      "epoch": 0.22472560552030751,
      "grad_norm": 0.7735697031021118,
      "learning_rate": 0.00019966106795369477,
      "loss": 0.5033,
      "step": 19410
    },
    {
      "epoch": 0.22484138378178115,
      "grad_norm": 0.7703429460525513,
      "learning_rate": 0.0001996600172723651,
      "loss": 0.501,
      "step": 19420
    },
    {
      "epoch": 0.22495716204325475,
      "grad_norm": 0.9691808223724365,
      "learning_rate": 0.0001996589649677842,
      "loss": 0.4858,
      "step": 19430
    },
    {
      "epoch": 0.22507294030472838,
      "grad_norm": 0.9339166879653931,
      "learning_rate": 0.00019965791103996924,
      "loss": 0.4672,
      "step": 19440
    },
    {
      "epoch": 0.225188718566202,
      "grad_norm": 0.9605696797370911,
      "learning_rate": 0.00019965685548893738,
      "loss": 0.504,
      "step": 19450
    },
    {
      "epoch": 0.22530449682767564,
      "grad_norm": 0.7763170599937439,
      "learning_rate": 0.0001996557983147058,
      "loss": 0.4923,
      "step": 19460
    },
    {
      "epoch": 0.22542027508914927,
      "grad_norm": 0.7404714226722717,
      "learning_rate": 0.0001996547395172917,
      "loss": 0.4811,
      "step": 19470
    },
    {
      "epoch": 0.22553605335062288,
      "grad_norm": 0.9595430493354797,
      "learning_rate": 0.00019965367909671237,
      "loss": 0.4914,
      "step": 19480
    },
    {
      "epoch": 0.2256518316120965,
      "grad_norm": 1.0273886919021606,
      "learning_rate": 0.00019965261705298505,
      "loss": 0.4628,
      "step": 19490
    },
    {
      "epoch": 0.22576760987357014,
      "grad_norm": 0.7963073253631592,
      "learning_rate": 0.00019965155338612703,
      "loss": 0.4713,
      "step": 19500
    },
    {
      "epoch": 0.22588338813504377,
      "grad_norm": 0.7545601725578308,
      "learning_rate": 0.0001996504880961557,
      "loss": 0.4757,
      "step": 19510
    },
    {
      "epoch": 0.2259991663965174,
      "grad_norm": 0.9121971726417542,
      "learning_rate": 0.00019964942118308834,
      "loss": 0.4904,
      "step": 19520
    },
    {
      "epoch": 0.226114944657991,
      "grad_norm": 0.822879433631897,
      "learning_rate": 0.00019964835264694238,
      "loss": 0.4899,
      "step": 19530
    },
    {
      "epoch": 0.22623072291946464,
      "grad_norm": 1.001206636428833,
      "learning_rate": 0.00019964728248773517,
      "loss": 0.475,
      "step": 19540
    },
    {
      "epoch": 0.22634650118093827,
      "grad_norm": 0.7777689695358276,
      "learning_rate": 0.00019964621070548417,
      "loss": 0.493,
      "step": 19550
    },
    {
      "epoch": 0.2264622794424119,
      "grad_norm": 0.7859448194503784,
      "learning_rate": 0.00019964513730020685,
      "loss": 0.4725,
      "step": 19560
    },
    {
      "epoch": 0.2265780577038855,
      "grad_norm": 0.7579594850540161,
      "learning_rate": 0.00019964406227192068,
      "loss": 0.4634,
      "step": 19570
    },
    {
      "epoch": 0.22669383596535914,
      "grad_norm": 0.775324285030365,
      "learning_rate": 0.0001996429856206431,
      "loss": 0.5071,
      "step": 19580
    },
    {
      "epoch": 0.22680961422683277,
      "grad_norm": 1.0083378553390503,
      "learning_rate": 0.0001996419073463918,
      "loss": 0.4539,
      "step": 19590
    },
    {
      "epoch": 0.2269253924883064,
      "grad_norm": 0.823955237865448,
      "learning_rate": 0.00019964082744918424,
      "loss": 0.4959,
      "step": 19600
    },
    {
      "epoch": 0.22704117074978003,
      "grad_norm": 0.8164582252502441,
      "learning_rate": 0.00019963974592903804,
      "loss": 0.4872,
      "step": 19610
    },
    {
      "epoch": 0.22715694901125363,
      "grad_norm": 1.0181759595870972,
      "learning_rate": 0.00019963866278597078,
      "loss": 0.5058,
      "step": 19620
    },
    {
      "epoch": 0.22727272727272727,
      "grad_norm": 0.8415988683700562,
      "learning_rate": 0.00019963757802000012,
      "loss": 0.451,
      "step": 19630
    },
    {
      "epoch": 0.2273885055342009,
      "grad_norm": 0.8228923678398132,
      "learning_rate": 0.00019963649163114372,
      "loss": 0.4475,
      "step": 19640
    },
    {
      "epoch": 0.22750428379567453,
      "grad_norm": 0.7857764363288879,
      "learning_rate": 0.00019963540361941934,
      "loss": 0.4866,
      "step": 19650
    },
    {
      "epoch": 0.22762006205714816,
      "grad_norm": 0.7542510628700256,
      "learning_rate": 0.0001996343139848446,
      "loss": 0.4396,
      "step": 19660
    },
    {
      "epoch": 0.22773584031862176,
      "grad_norm": 0.811672031879425,
      "learning_rate": 0.00019963322272743733,
      "loss": 0.4919,
      "step": 19670
    },
    {
      "epoch": 0.2278516185800954,
      "grad_norm": 1.0724990367889404,
      "learning_rate": 0.00019963212984721524,
      "loss": 0.5029,
      "step": 19680
    },
    {
      "epoch": 0.22796739684156903,
      "grad_norm": 0.7121882438659668,
      "learning_rate": 0.00019963103534419618,
      "loss": 0.482,
      "step": 19690
    },
    {
      "epoch": 0.22808317510304266,
      "grad_norm": 0.8476919531822205,
      "learning_rate": 0.00019962993921839796,
      "loss": 0.4674,
      "step": 19700
    },
    {
      "epoch": 0.2281989533645163,
      "grad_norm": 0.7922258973121643,
      "learning_rate": 0.00019962884146983845,
      "loss": 0.4814,
      "step": 19710
    },
    {
      "epoch": 0.2283147316259899,
      "grad_norm": 0.884506344795227,
      "learning_rate": 0.00019962774209853553,
      "loss": 0.4585,
      "step": 19720
    },
    {
      "epoch": 0.22843050988746352,
      "grad_norm": 0.8416809439659119,
      "learning_rate": 0.00019962664110450705,
      "loss": 0.5042,
      "step": 19730
    },
    {
      "epoch": 0.22854628814893715,
      "grad_norm": 0.8013078570365906,
      "learning_rate": 0.000199625538487771,
      "loss": 0.498,
      "step": 19740
    },
    {
      "epoch": 0.22866206641041079,
      "grad_norm": 0.9373008608818054,
      "learning_rate": 0.0001996244342483453,
      "loss": 0.5011,
      "step": 19750
    },
    {
      "epoch": 0.22877784467188442,
      "grad_norm": 0.9532410502433777,
      "learning_rate": 0.00019962332838624797,
      "loss": 0.4613,
      "step": 19760
    },
    {
      "epoch": 0.22889362293335802,
      "grad_norm": 0.8349032402038574,
      "learning_rate": 0.000199622220901497,
      "loss": 0.4719,
      "step": 19770
    },
    {
      "epoch": 0.22900940119483165,
      "grad_norm": 0.9330806732177734,
      "learning_rate": 0.00019962111179411046,
      "loss": 0.4935,
      "step": 19780
    },
    {
      "epoch": 0.22912517945630528,
      "grad_norm": 0.9019778370857239,
      "learning_rate": 0.00019962000106410642,
      "loss": 0.4989,
      "step": 19790
    },
    {
      "epoch": 0.22924095771777891,
      "grad_norm": 0.7046825289726257,
      "learning_rate": 0.0001996188887115029,
      "loss": 0.4836,
      "step": 19800
    },
    {
      "epoch": 0.22935673597925255,
      "grad_norm": 0.8934338092803955,
      "learning_rate": 0.00019961777473631806,
      "loss": 0.468,
      "step": 19810
    },
    {
      "epoch": 0.22947251424072615,
      "grad_norm": 0.828784704208374,
      "learning_rate": 0.00019961665913857007,
      "loss": 0.4678,
      "step": 19820
    },
    {
      "epoch": 0.22958829250219978,
      "grad_norm": 0.718712329864502,
      "learning_rate": 0.00019961554191827706,
      "loss": 0.465,
      "step": 19830
    },
    {
      "epoch": 0.2297040707636734,
      "grad_norm": 0.8049870729446411,
      "learning_rate": 0.00019961442307545727,
      "loss": 0.4828,
      "step": 19840
    },
    {
      "epoch": 0.22981984902514704,
      "grad_norm": 0.7962097525596619,
      "learning_rate": 0.00019961330261012886,
      "loss": 0.5186,
      "step": 19850
    },
    {
      "epoch": 0.22993562728662068,
      "grad_norm": 0.697943925857544,
      "learning_rate": 0.00019961218052231015,
      "loss": 0.485,
      "step": 19860
    },
    {
      "epoch": 0.23005140554809428,
      "grad_norm": 0.7481710314750671,
      "learning_rate": 0.0001996110568120194,
      "loss": 0.504,
      "step": 19870
    },
    {
      "epoch": 0.2301671838095679,
      "grad_norm": 0.8054175972938538,
      "learning_rate": 0.00019960993147927488,
      "loss": 0.4685,
      "step": 19880
    },
    {
      "epoch": 0.23028296207104154,
      "grad_norm": 0.8510473966598511,
      "learning_rate": 0.0001996088045240949,
      "loss": 0.4976,
      "step": 19890
    },
    {
      "epoch": 0.23039874033251517,
      "grad_norm": 0.8469729423522949,
      "learning_rate": 0.0001996076759464979,
      "loss": 0.4591,
      "step": 19900
    },
    {
      "epoch": 0.2305145185939888,
      "grad_norm": 0.8469066619873047,
      "learning_rate": 0.00019960654574650223,
      "loss": 0.4941,
      "step": 19910
    },
    {
      "epoch": 0.2306302968554624,
      "grad_norm": 0.8097986578941345,
      "learning_rate": 0.00019960541392412625,
      "loss": 0.487,
      "step": 19920
    },
    {
      "epoch": 0.23074607511693604,
      "grad_norm": 0.8033235669136047,
      "learning_rate": 0.00019960428047938845,
      "loss": 0.4709,
      "step": 19930
    },
    {
      "epoch": 0.23086185337840967,
      "grad_norm": 0.9559374451637268,
      "learning_rate": 0.00019960314541230724,
      "loss": 0.4784,
      "step": 19940
    },
    {
      "epoch": 0.2309776316398833,
      "grad_norm": 0.7702792882919312,
      "learning_rate": 0.00019960200872290115,
      "loss": 0.4649,
      "step": 19950
    },
    {
      "epoch": 0.23109340990135693,
      "grad_norm": 0.8488996028900146,
      "learning_rate": 0.00019960087041118868,
      "loss": 0.4914,
      "step": 19960
    },
    {
      "epoch": 0.23120918816283054,
      "grad_norm": 0.8142027258872986,
      "learning_rate": 0.00019959973047718838,
      "loss": 0.507,
      "step": 19970
    },
    {
      "epoch": 0.23132496642430417,
      "grad_norm": 0.8428452014923096,
      "learning_rate": 0.00019959858892091884,
      "loss": 0.4734,
      "step": 19980
    },
    {
      "epoch": 0.2314407446857778,
      "grad_norm": 0.9110616445541382,
      "learning_rate": 0.0001995974457423986,
      "loss": 0.5022,
      "step": 19990
    },
    {
      "epoch": 0.23155652294725143,
      "grad_norm": 0.8447798490524292,
      "learning_rate": 0.00019959630094164627,
      "loss": 0.4502,
      "step": 20000
    },
    {
      "epoch": 0.23155652294725143,
      "eval_chrf": 80.23654691744446,
      "eval_loss": 0.7331246137619019,
      "eval_runtime": 286.9776,
      "eval_samples_per_second": 0.348,
      "eval_steps_per_second": 0.014,
      "step": 20000
    },
    {
      "epoch": 0.23167230120872506,
      "grad_norm": 0.8785196542739868,
      "learning_rate": 0.0001995951545186806,
      "loss": 0.5204,
      "step": 20010
    },
    {
      "epoch": 0.23178807947019867,
      "grad_norm": 0.8217412233352661,
      "learning_rate": 0.00019959400647352014,
      "loss": 0.4694,
      "step": 20020
    },
    {
      "epoch": 0.2319038577316723,
      "grad_norm": 0.8321861624717712,
      "learning_rate": 0.00019959285680618366,
      "loss": 0.4965,
      "step": 20030
    },
    {
      "epoch": 0.23201963599314593,
      "grad_norm": 0.9536188840866089,
      "learning_rate": 0.00019959170551668982,
      "loss": 0.4945,
      "step": 20040
    },
    {
      "epoch": 0.23213541425461956,
      "grad_norm": 0.8933611512184143,
      "learning_rate": 0.00019959055260505748,
      "loss": 0.4873,
      "step": 20050
    },
    {
      "epoch": 0.2322511925160932,
      "grad_norm": 0.768867552280426,
      "learning_rate": 0.00019958939807130532,
      "loss": 0.4659,
      "step": 20060
    },
    {
      "epoch": 0.2323669707775668,
      "grad_norm": 0.8929364085197449,
      "learning_rate": 0.00019958824191545222,
      "loss": 0.4981,
      "step": 20070
    },
    {
      "epoch": 0.23248274903904043,
      "grad_norm": 0.7275567054748535,
      "learning_rate": 0.00019958708413751698,
      "loss": 0.4691,
      "step": 20080
    },
    {
      "epoch": 0.23259852730051406,
      "grad_norm": 0.6579349637031555,
      "learning_rate": 0.00019958592473751842,
      "loss": 0.4903,
      "step": 20090
    },
    {
      "epoch": 0.2327143055619877,
      "grad_norm": 1.21674382686615,
      "learning_rate": 0.00019958476371547548,
      "loss": 0.4657,
      "step": 20100
    },
    {
      "epoch": 0.23283008382346132,
      "grad_norm": 0.8369050025939941,
      "learning_rate": 0.000199583601071407,
      "loss": 0.4864,
      "step": 20110
    },
    {
      "epoch": 0.23294586208493492,
      "grad_norm": 0.7792148590087891,
      "learning_rate": 0.00019958243680533202,
      "loss": 0.4734,
      "step": 20120
    },
    {
      "epoch": 0.23306164034640856,
      "grad_norm": 0.839080274105072,
      "learning_rate": 0.0001995812709172694,
      "loss": 0.4442,
      "step": 20130
    },
    {
      "epoch": 0.2331774186078822,
      "grad_norm": 0.8466188907623291,
      "learning_rate": 0.00019958010340723824,
      "loss": 0.4684,
      "step": 20140
    },
    {
      "epoch": 0.23329319686935582,
      "grad_norm": 0.8918906450271606,
      "learning_rate": 0.00019957893427525745,
      "loss": 0.4473,
      "step": 20150
    },
    {
      "epoch": 0.23340897513082945,
      "grad_norm": 0.9140417575836182,
      "learning_rate": 0.00019957776352134613,
      "loss": 0.4882,
      "step": 20160
    },
    {
      "epoch": 0.23352475339230305,
      "grad_norm": 0.762043833732605,
      "learning_rate": 0.00019957659114552332,
      "loss": 0.4965,
      "step": 20170
    },
    {
      "epoch": 0.23364053165377668,
      "grad_norm": 0.7834113240242004,
      "learning_rate": 0.00019957541714780814,
      "loss": 0.466,
      "step": 20180
    },
    {
      "epoch": 0.23375630991525032,
      "grad_norm": 0.813154935836792,
      "learning_rate": 0.0001995742415282197,
      "loss": 0.493,
      "step": 20190
    },
    {
      "epoch": 0.23387208817672395,
      "grad_norm": 0.618257462978363,
      "learning_rate": 0.00019957306428677716,
      "loss": 0.4677,
      "step": 20200
    },
    {
      "epoch": 0.23398786643819755,
      "grad_norm": 0.7799302935600281,
      "learning_rate": 0.00019957188542349966,
      "loss": 0.4666,
      "step": 20210
    },
    {
      "epoch": 0.23410364469967118,
      "grad_norm": 0.9057375192642212,
      "learning_rate": 0.0001995707049384065,
      "loss": 0.4742,
      "step": 20220
    },
    {
      "epoch": 0.2342194229611448,
      "grad_norm": 0.7182900309562683,
      "learning_rate": 0.00019956952283151674,
      "loss": 0.486,
      "step": 20230
    },
    {
      "epoch": 0.23433520122261844,
      "grad_norm": 0.7508489489555359,
      "learning_rate": 0.0001995683391028498,
      "loss": 0.478,
      "step": 20240
    },
    {
      "epoch": 0.23445097948409208,
      "grad_norm": 0.7437032461166382,
      "learning_rate": 0.00019956715375242485,
      "loss": 0.4787,
      "step": 20250
    },
    {
      "epoch": 0.23456675774556568,
      "grad_norm": 0.8082559108734131,
      "learning_rate": 0.00019956596678026125,
      "loss": 0.5177,
      "step": 20260
    },
    {
      "epoch": 0.2346825360070393,
      "grad_norm": 0.8698830008506775,
      "learning_rate": 0.00019956477818637832,
      "loss": 0.4927,
      "step": 20270
    },
    {
      "epoch": 0.23479831426851294,
      "grad_norm": 0.7740945219993591,
      "learning_rate": 0.0001995635879707954,
      "loss": 0.4821,
      "step": 20280
    },
    {
      "epoch": 0.23491409252998657,
      "grad_norm": 0.8962322473526001,
      "learning_rate": 0.00019956239613353193,
      "loss": 0.4709,
      "step": 20290
    },
    {
      "epoch": 0.2350298707914602,
      "grad_norm": 0.8895878195762634,
      "learning_rate": 0.00019956120267460728,
      "loss": 0.4994,
      "step": 20300
    },
    {
      "epoch": 0.2351456490529338,
      "grad_norm": 0.8879282474517822,
      "learning_rate": 0.00019956000759404087,
      "loss": 0.5123,
      "step": 20310
    },
    {
      "epoch": 0.23526142731440744,
      "grad_norm": 0.8099833130836487,
      "learning_rate": 0.00019955881089185222,
      "loss": 0.4671,
      "step": 20320
    },
    {
      "epoch": 0.23537720557588107,
      "grad_norm": 0.7878211140632629,
      "learning_rate": 0.00019955761256806078,
      "loss": 0.4717,
      "step": 20330
    },
    {
      "epoch": 0.2354929838373547,
      "grad_norm": 0.8423677086830139,
      "learning_rate": 0.00019955641262268609,
      "loss": 0.465,
      "step": 20340
    },
    {
      "epoch": 0.23560876209882833,
      "grad_norm": 0.9814093112945557,
      "learning_rate": 0.0001995552110557477,
      "loss": 0.4418,
      "step": 20350
    },
    {
      "epoch": 0.23572454036030194,
      "grad_norm": 0.8047503232955933,
      "learning_rate": 0.00019955400786726512,
      "loss": 0.4855,
      "step": 20360
    },
    {
      "epoch": 0.23584031862177557,
      "grad_norm": 0.6284241080284119,
      "learning_rate": 0.00019955280305725804,
      "loss": 0.4741,
      "step": 20370
    },
    {
      "epoch": 0.2359560968832492,
      "grad_norm": 0.6875554919242859,
      "learning_rate": 0.00019955159662574603,
      "loss": 0.4552,
      "step": 20380
    },
    {
      "epoch": 0.23607187514472283,
      "grad_norm": 0.732741117477417,
      "learning_rate": 0.00019955038857274874,
      "loss": 0.5087,
      "step": 20390
    },
    {
      "epoch": 0.23618765340619646,
      "grad_norm": 0.8414047956466675,
      "learning_rate": 0.00019954917889828587,
      "loss": 0.4742,
      "step": 20400
    },
    {
      "epoch": 0.23630343166767007,
      "grad_norm": 0.8171016573905945,
      "learning_rate": 0.00019954796760237708,
      "loss": 0.4601,
      "step": 20410
    },
    {
      "epoch": 0.2364192099291437,
      "grad_norm": 0.7645574808120728,
      "learning_rate": 0.00019954675468504215,
      "loss": 0.4893,
      "step": 20420
    },
    {
      "epoch": 0.23653498819061733,
      "grad_norm": 1.0034972429275513,
      "learning_rate": 0.0001995455401463008,
      "loss": 0.5233,
      "step": 20430
    },
    {
      "epoch": 0.23665076645209096,
      "grad_norm": 0.9084933996200562,
      "learning_rate": 0.00019954432398617284,
      "loss": 0.5087,
      "step": 20440
    },
    {
      "epoch": 0.2367665447135646,
      "grad_norm": 0.8002224564552307,
      "learning_rate": 0.00019954310620467808,
      "loss": 0.4835,
      "step": 20450
    },
    {
      "epoch": 0.2368823229750382,
      "grad_norm": 0.9463514089584351,
      "learning_rate": 0.0001995418868018363,
      "loss": 0.5134,
      "step": 20460
    },
    {
      "epoch": 0.23699810123651183,
      "grad_norm": 0.7522697448730469,
      "learning_rate": 0.00019954066577766743,
      "loss": 0.5045,
      "step": 20470
    },
    {
      "epoch": 0.23711387949798546,
      "grad_norm": 0.9325494766235352,
      "learning_rate": 0.00019953944313219133,
      "loss": 0.5108,
      "step": 20480
    },
    {
      "epoch": 0.2372296577594591,
      "grad_norm": 0.8868902325630188,
      "learning_rate": 0.00019953821886542792,
      "loss": 0.4505,
      "step": 20490
    },
    {
      "epoch": 0.23734543602093272,
      "grad_norm": 0.7130038142204285,
      "learning_rate": 0.0001995369929773971,
      "loss": 0.4787,
      "step": 20500
    },
    {
      "epoch": 0.23746121428240632,
      "grad_norm": 0.7780782580375671,
      "learning_rate": 0.00019953576546811888,
      "loss": 0.4649,
      "step": 20510
    },
    {
      "epoch": 0.23757699254387996,
      "grad_norm": 0.8343157768249512,
      "learning_rate": 0.00019953453633761328,
      "loss": 0.4744,
      "step": 20520
    },
    {
      "epoch": 0.2376927708053536,
      "grad_norm": 0.8360598683357239,
      "learning_rate": 0.00019953330558590025,
      "loss": 0.4615,
      "step": 20530
    },
    {
      "epoch": 0.23780854906682722,
      "grad_norm": 0.7804827690124512,
      "learning_rate": 0.00019953207321299988,
      "loss": 0.4845,
      "step": 20540
    },
    {
      "epoch": 0.23792432732830085,
      "grad_norm": 0.6212272047996521,
      "learning_rate": 0.00019953083921893226,
      "loss": 0.4591,
      "step": 20550
    },
    {
      "epoch": 0.23804010558977445,
      "grad_norm": 0.7534270286560059,
      "learning_rate": 0.00019952960360371743,
      "loss": 0.4961,
      "step": 20560
    },
    {
      "epoch": 0.23815588385124808,
      "grad_norm": 0.7674233317375183,
      "learning_rate": 0.00019952836636737557,
      "loss": 0.4365,
      "step": 20570
    },
    {
      "epoch": 0.23827166211272172,
      "grad_norm": 0.801550567150116,
      "learning_rate": 0.00019952712750992681,
      "loss": 0.4585,
      "step": 20580
    },
    {
      "epoch": 0.23838744037419535,
      "grad_norm": 0.7644691467285156,
      "learning_rate": 0.00019952588703139133,
      "loss": 0.4733,
      "step": 20590
    },
    {
      "epoch": 0.23850321863566898,
      "grad_norm": 0.8335167169570923,
      "learning_rate": 0.0001995246449317893,
      "loss": 0.45,
      "step": 20600
    },
    {
      "epoch": 0.23861899689714258,
      "grad_norm": 0.9844253659248352,
      "learning_rate": 0.000199523401211141,
      "loss": 0.4561,
      "step": 20610
    },
    {
      "epoch": 0.2387347751586162,
      "grad_norm": 0.7373968362808228,
      "learning_rate": 0.00019952215586946668,
      "loss": 0.4824,
      "step": 20620
    },
    {
      "epoch": 0.23885055342008985,
      "grad_norm": 0.9232361912727356,
      "learning_rate": 0.00019952090890678665,
      "loss": 0.4896,
      "step": 20630
    },
    {
      "epoch": 0.23896633168156348,
      "grad_norm": 0.7922347187995911,
      "learning_rate": 0.00019951966032312116,
      "loss": 0.4742,
      "step": 20640
    },
    {
      "epoch": 0.2390821099430371,
      "grad_norm": 0.9157789945602417,
      "learning_rate": 0.00019951841011849056,
      "loss": 0.4673,
      "step": 20650
    },
    {
      "epoch": 0.2391978882045107,
      "grad_norm": 0.7545085549354553,
      "learning_rate": 0.00019951715829291523,
      "loss": 0.4469,
      "step": 20660
    },
    {
      "epoch": 0.23931366646598434,
      "grad_norm": 0.8135688900947571,
      "learning_rate": 0.00019951590484641558,
      "loss": 0.461,
      "step": 20670
    },
    {
      "epoch": 0.23942944472745797,
      "grad_norm": 0.8085801601409912,
      "learning_rate": 0.000199514649779012,
      "loss": 0.4579,
      "step": 20680
    },
    {
      "epoch": 0.2395452229889316,
      "grad_norm": 0.8696886301040649,
      "learning_rate": 0.0001995133930907249,
      "loss": 0.4788,
      "step": 20690
    },
    {
      "epoch": 0.23966100125040524,
      "grad_norm": 0.9587036371231079,
      "learning_rate": 0.0001995121347815748,
      "loss": 0.4839,
      "step": 20700
    },
    {
      "epoch": 0.23977677951187884,
      "grad_norm": 0.9053393006324768,
      "learning_rate": 0.0001995108748515822,
      "loss": 0.4807,
      "step": 20710
    },
    {
      "epoch": 0.23989255777335247,
      "grad_norm": 0.8441702127456665,
      "learning_rate": 0.0001995096133007676,
      "loss": 0.4695,
      "step": 20720
    },
    {
      "epoch": 0.2400083360348261,
      "grad_norm": 0.8357120752334595,
      "learning_rate": 0.00019950835012915152,
      "loss": 0.4609,
      "step": 20730
    },
    {
      "epoch": 0.24012411429629973,
      "grad_norm": 0.7920693755149841,
      "learning_rate": 0.00019950708533675456,
      "loss": 0.4981,
      "step": 20740
    },
    {
      "epoch": 0.24023989255777337,
      "grad_norm": 0.835800290107727,
      "learning_rate": 0.00019950581892359736,
      "loss": 0.4719,
      "step": 20750
    },
    {
      "epoch": 0.24035567081924697,
      "grad_norm": 0.7899350523948669,
      "learning_rate": 0.0001995045508897005,
      "loss": 0.475,
      "step": 20760
    },
    {
      "epoch": 0.2404714490807206,
      "grad_norm": 0.8127520084381104,
      "learning_rate": 0.00019950328123508464,
      "loss": 0.4711,
      "step": 20770
    },
    {
      "epoch": 0.24058722734219423,
      "grad_norm": 0.895070493221283,
      "learning_rate": 0.00019950200995977045,
      "loss": 0.5064,
      "step": 20780
    },
    {
      "epoch": 0.24070300560366786,
      "grad_norm": 0.7311599850654602,
      "learning_rate": 0.0001995007370637787,
      "loss": 0.4539,
      "step": 20790
    },
    {
      "epoch": 0.2408187838651415,
      "grad_norm": 0.7276439070701599,
      "learning_rate": 0.00019949946254713003,
      "loss": 0.4945,
      "step": 20800
    },
    {
      "epoch": 0.2409345621266151,
      "grad_norm": 0.9493795037269592,
      "learning_rate": 0.00019949818640984526,
      "loss": 0.4795,
      "step": 20810
    },
    {
      "epoch": 0.24105034038808873,
      "grad_norm": 0.8072208762168884,
      "learning_rate": 0.00019949690865194515,
      "loss": 0.5143,
      "step": 20820
    },
    {
      "epoch": 0.24116611864956236,
      "grad_norm": 0.8493167757987976,
      "learning_rate": 0.0001994956292734505,
      "loss": 0.4468,
      "step": 20830
    },
    {
      "epoch": 0.241281896911036,
      "grad_norm": 0.7841184735298157,
      "learning_rate": 0.0001994943482743822,
      "loss": 0.4572,
      "step": 20840
    },
    {
      "epoch": 0.2413976751725096,
      "grad_norm": 0.8121345043182373,
      "learning_rate": 0.00019949306565476106,
      "loss": 0.4669,
      "step": 20850
    },
    {
      "epoch": 0.24151345343398323,
      "grad_norm": 0.8304721117019653,
      "learning_rate": 0.00019949178141460803,
      "loss": 0.4345,
      "step": 20860
    },
    {
      "epoch": 0.24162923169545686,
      "grad_norm": 0.7565070986747742,
      "learning_rate": 0.00019949049555394397,
      "loss": 0.4826,
      "step": 20870
    },
    {
      "epoch": 0.2417450099569305,
      "grad_norm": 0.8977930545806885,
      "learning_rate": 0.00019948920807278983,
      "loss": 0.5107,
      "step": 20880
    },
    {
      "epoch": 0.24186078821840412,
      "grad_norm": 0.8773626089096069,
      "learning_rate": 0.00019948791897116664,
      "loss": 0.4874,
      "step": 20890
    },
    {
      "epoch": 0.24197656647987772,
      "grad_norm": 0.7891585230827332,
      "learning_rate": 0.00019948662824909532,
      "loss": 0.5013,
      "step": 20900
    },
    {
      "epoch": 0.24209234474135136,
      "grad_norm": 1.0026624202728271,
      "learning_rate": 0.00019948533590659695,
      "loss": 0.4816,
      "step": 20910
    },
    {
      "epoch": 0.242208123002825,
      "grad_norm": 0.8887225389480591,
      "learning_rate": 0.00019948404194369251,
      "loss": 0.497,
      "step": 20920
    },
    {
      "epoch": 0.24232390126429862,
      "grad_norm": 0.7298649549484253,
      "learning_rate": 0.00019948274636040315,
      "loss": 0.491,
      "step": 20930
    },
    {
      "epoch": 0.24243967952577225,
      "grad_norm": 0.8127931356430054,
      "learning_rate": 0.00019948144915674996,
      "loss": 0.4472,
      "step": 20940
    },
    {
      "epoch": 0.24255545778724585,
      "grad_norm": 0.766441285610199,
      "learning_rate": 0.00019948015033275404,
      "loss": 0.4745,
      "step": 20950
    },
    {
      "epoch": 0.24267123604871949,
      "grad_norm": 0.8924943804740906,
      "learning_rate": 0.00019947884988843657,
      "loss": 0.4689,
      "step": 20960
    },
    {
      "epoch": 0.24278701431019312,
      "grad_norm": 0.9387590885162354,
      "learning_rate": 0.0001994775478238187,
      "loss": 0.4496,
      "step": 20970
    },
    {
      "epoch": 0.24290279257166675,
      "grad_norm": 0.7949775457382202,
      "learning_rate": 0.00019947624413892167,
      "loss": 0.475,
      "step": 20980
    },
    {
      "epoch": 0.24301857083314038,
      "grad_norm": 1.0228512287139893,
      "learning_rate": 0.00019947493883376672,
      "loss": 0.4646,
      "step": 20990
    },
    {
      "epoch": 0.24313434909461398,
      "grad_norm": 0.8532053232192993,
      "learning_rate": 0.00019947363190837508,
      "loss": 0.4906,
      "step": 21000
    },
    {
      "epoch": 0.24313434909461398,
      "eval_chrf": 72.27352234458223,
      "eval_loss": 0.7107462882995605,
      "eval_runtime": 349.0144,
      "eval_samples_per_second": 0.287,
      "eval_steps_per_second": 0.011,
      "step": 21000
    },
    {
      "epoch": 0.24325012735608761,
      "grad_norm": 0.9279664158821106,
      "learning_rate": 0.00019947232336276803,
      "loss": 0.4713,
      "step": 21010
    },
    {
      "epoch": 0.24336590561756125,
      "grad_norm": 0.8047536015510559,
      "learning_rate": 0.0001994710131969669,
      "loss": 0.4396,
      "step": 21020
    },
    {
      "epoch": 0.24348168387903488,
      "grad_norm": 0.7375866770744324,
      "learning_rate": 0.00019946970141099306,
      "loss": 0.4583,
      "step": 21030
    },
    {
      "epoch": 0.2435974621405085,
      "grad_norm": 0.9027757048606873,
      "learning_rate": 0.00019946838800486783,
      "loss": 0.4886,
      "step": 21040
    },
    {
      "epoch": 0.2437132404019821,
      "grad_norm": 0.6699516773223877,
      "learning_rate": 0.00019946707297861265,
      "loss": 0.4528,
      "step": 21050
    },
    {
      "epoch": 0.24382901866345574,
      "grad_norm": 0.7650065422058105,
      "learning_rate": 0.00019946575633224888,
      "loss": 0.502,
      "step": 21060
    },
    {
      "epoch": 0.24394479692492937,
      "grad_norm": 0.8151065111160278,
      "learning_rate": 0.000199464438065798,
      "loss": 0.5019,
      "step": 21070
    },
    {
      "epoch": 0.244060575186403,
      "grad_norm": 0.8434141874313354,
      "learning_rate": 0.00019946311817928148,
      "loss": 0.4799,
      "step": 21080
    },
    {
      "epoch": 0.24417635344787664,
      "grad_norm": 0.80101078748703,
      "learning_rate": 0.00019946179667272083,
      "loss": 0.4216,
      "step": 21090
    },
    {
      "epoch": 0.24429213170935024,
      "grad_norm": 0.5379853844642639,
      "learning_rate": 0.00019946047354613752,
      "loss": 0.4629,
      "step": 21100
    },
    {
      "epoch": 0.24440790997082387,
      "grad_norm": 0.8522878289222717,
      "learning_rate": 0.0001994591487995532,
      "loss": 0.4587,
      "step": 21110
    },
    {
      "epoch": 0.2445236882322975,
      "grad_norm": 1.004294991493225,
      "learning_rate": 0.00019945782243298936,
      "loss": 0.4456,
      "step": 21120
    },
    {
      "epoch": 0.24463946649377113,
      "grad_norm": 0.8788337111473083,
      "learning_rate": 0.00019945649444646764,
      "loss": 0.4678,
      "step": 21130
    },
    {
      "epoch": 0.24475524475524477,
      "grad_norm": 0.8699989914894104,
      "learning_rate": 0.00019945516484000964,
      "loss": 0.4558,
      "step": 21140
    },
    {
      "epoch": 0.24487102301671837,
      "grad_norm": 0.783763587474823,
      "learning_rate": 0.00019945383361363707,
      "loss": 0.4571,
      "step": 21150
    },
    {
      "epoch": 0.244986801278192,
      "grad_norm": 0.8254796266555786,
      "learning_rate": 0.00019945250076737155,
      "loss": 0.4845,
      "step": 21160
    },
    {
      "epoch": 0.24510257953966563,
      "grad_norm": 0.8270418643951416,
      "learning_rate": 0.00019945116630123484,
      "loss": 0.4934,
      "step": 21170
    },
    {
      "epoch": 0.24521835780113926,
      "grad_norm": 0.8202553987503052,
      "learning_rate": 0.0001994498302152487,
      "loss": 0.4605,
      "step": 21180
    },
    {
      "epoch": 0.2453341360626129,
      "grad_norm": 0.9639976024627686,
      "learning_rate": 0.00019944849250943478,
      "loss": 0.4349,
      "step": 21190
    },
    {
      "epoch": 0.2454499143240865,
      "grad_norm": 0.7821773290634155,
      "learning_rate": 0.00019944715318381495,
      "loss": 0.4975,
      "step": 21200
    },
    {
      "epoch": 0.24556569258556013,
      "grad_norm": 0.8731173872947693,
      "learning_rate": 0.00019944581223841102,
      "loss": 0.4909,
      "step": 21210
    },
    {
      "epoch": 0.24568147084703376,
      "grad_norm": 0.6022710204124451,
      "learning_rate": 0.00019944446967324483,
      "loss": 0.4654,
      "step": 21220
    },
    {
      "epoch": 0.2457972491085074,
      "grad_norm": 0.8561477661132812,
      "learning_rate": 0.00019944312548833825,
      "loss": 0.4781,
      "step": 21230
    },
    {
      "epoch": 0.24591302736998102,
      "grad_norm": 0.8511117696762085,
      "learning_rate": 0.00019944177968371314,
      "loss": 0.4597,
      "step": 21240
    },
    {
      "epoch": 0.24602880563145463,
      "grad_norm": 0.9651407599449158,
      "learning_rate": 0.00019944043225939148,
      "loss": 0.4877,
      "step": 21250
    },
    {
      "epoch": 0.24614458389292826,
      "grad_norm": 0.6892158389091492,
      "learning_rate": 0.00019943908321539515,
      "loss": 0.4843,
      "step": 21260
    },
    {
      "epoch": 0.2462603621544019,
      "grad_norm": 0.8957681655883789,
      "learning_rate": 0.00019943773255174616,
      "loss": 0.4981,
      "step": 21270
    },
    {
      "epoch": 0.24637614041587552,
      "grad_norm": 0.8759444355964661,
      "learning_rate": 0.00019943638026846653,
      "loss": 0.4702,
      "step": 21280
    },
    {
      "epoch": 0.24649191867734915,
      "grad_norm": 0.7360313534736633,
      "learning_rate": 0.00019943502636557824,
      "loss": 0.4502,
      "step": 21290
    },
    {
      "epoch": 0.24660769693882276,
      "grad_norm": 1.097824215888977,
      "learning_rate": 0.00019943367084310337,
      "loss": 0.4718,
      "step": 21300
    },
    {
      "epoch": 0.2467234752002964,
      "grad_norm": 0.8211607336997986,
      "learning_rate": 0.00019943231370106397,
      "loss": 0.4684,
      "step": 21310
    },
    {
      "epoch": 0.24683925346177002,
      "grad_norm": 0.9314284324645996,
      "learning_rate": 0.00019943095493948215,
      "loss": 0.4657,
      "step": 21320
    },
    {
      "epoch": 0.24695503172324365,
      "grad_norm": 0.925950825214386,
      "learning_rate": 0.0001994295945583801,
      "loss": 0.4793,
      "step": 21330
    },
    {
      "epoch": 0.24707080998471728,
      "grad_norm": 0.9872474670410156,
      "learning_rate": 0.00019942823255777992,
      "loss": 0.4526,
      "step": 21340
    },
    {
      "epoch": 0.24718658824619089,
      "grad_norm": 0.8325170278549194,
      "learning_rate": 0.00019942686893770383,
      "loss": 0.4783,
      "step": 21350
    },
    {
      "epoch": 0.24730236650766452,
      "grad_norm": 0.77191561460495,
      "learning_rate": 0.000199425503698174,
      "loss": 0.4882,
      "step": 21360
    },
    {
      "epoch": 0.24741814476913815,
      "grad_norm": 0.7750169634819031,
      "learning_rate": 0.00019942413683921267,
      "loss": 0.4581,
      "step": 21370
    },
    {
      "epoch": 0.24753392303061178,
      "grad_norm": 0.9309362769126892,
      "learning_rate": 0.00019942276836084214,
      "loss": 0.4524,
      "step": 21380
    },
    {
      "epoch": 0.2476497012920854,
      "grad_norm": 0.821103036403656,
      "learning_rate": 0.00019942139826308468,
      "loss": 0.4876,
      "step": 21390
    },
    {
      "epoch": 0.24776547955355901,
      "grad_norm": 0.8426516652107239,
      "learning_rate": 0.00019942002654596262,
      "loss": 0.4518,
      "step": 21400
    },
    {
      "epoch": 0.24788125781503265,
      "grad_norm": 0.9636382460594177,
      "learning_rate": 0.00019941865320949827,
      "loss": 0.4669,
      "step": 21410
    },
    {
      "epoch": 0.24799703607650628,
      "grad_norm": 0.7734730243682861,
      "learning_rate": 0.00019941727825371398,
      "loss": 0.4619,
      "step": 21420
    },
    {
      "epoch": 0.2481128143379799,
      "grad_norm": 0.9146222472190857,
      "learning_rate": 0.00019941590167863225,
      "loss": 0.489,
      "step": 21430
    },
    {
      "epoch": 0.24822859259945354,
      "grad_norm": 0.9050183296203613,
      "learning_rate": 0.00019941452348427538,
      "loss": 0.5036,
      "step": 21440
    },
    {
      "epoch": 0.24834437086092714,
      "grad_norm": 0.5671589970588684,
      "learning_rate": 0.0001994131436706659,
      "loss": 0.4298,
      "step": 21450
    },
    {
      "epoch": 0.24846014912240078,
      "grad_norm": 0.7186804413795471,
      "learning_rate": 0.00019941176223782625,
      "loss": 0.4757,
      "step": 21460
    },
    {
      "epoch": 0.2485759273838744,
      "grad_norm": 0.7497647404670715,
      "learning_rate": 0.00019941037918577893,
      "loss": 0.4913,
      "step": 21470
    },
    {
      "epoch": 0.24869170564534804,
      "grad_norm": 0.840472400188446,
      "learning_rate": 0.0001994089945145465,
      "loss": 0.4651,
      "step": 21480
    },
    {
      "epoch": 0.24880748390682164,
      "grad_norm": 0.764240026473999,
      "learning_rate": 0.00019940760822415146,
      "loss": 0.498,
      "step": 21490
    },
    {
      "epoch": 0.24892326216829527,
      "grad_norm": 0.8429006934165955,
      "learning_rate": 0.0001994062203146164,
      "loss": 0.4926,
      "step": 21500
    },
    {
      "epoch": 0.2490390404297689,
      "grad_norm": 0.702929675579071,
      "learning_rate": 0.00019940483078596398,
      "loss": 0.4713,
      "step": 21510
    },
    {
      "epoch": 0.24915481869124254,
      "grad_norm": 0.6671960949897766,
      "learning_rate": 0.00019940343963821677,
      "loss": 0.4451,
      "step": 21520
    },
    {
      "epoch": 0.24927059695271617,
      "grad_norm": 0.8115770816802979,
      "learning_rate": 0.00019940204687139744,
      "loss": 0.4823,
      "step": 21530
    },
    {
      "epoch": 0.24938637521418977,
      "grad_norm": 0.9203357696533203,
      "learning_rate": 0.00019940065248552874,
      "loss": 0.4935,
      "step": 21540
    },
    {
      "epoch": 0.2495021534756634,
      "grad_norm": 0.8061996698379517,
      "learning_rate": 0.0001993992564806333,
      "loss": 0.4614,
      "step": 21550
    },
    {
      "epoch": 0.24961793173713703,
      "grad_norm": 0.6012446284294128,
      "learning_rate": 0.0001993978588567339,
      "loss": 0.4713,
      "step": 21560
    },
    {
      "epoch": 0.24973370999861066,
      "grad_norm": 0.8243346214294434,
      "learning_rate": 0.00019939645961385326,
      "loss": 0.4648,
      "step": 21570
    },
    {
      "epoch": 0.2498494882600843,
      "grad_norm": 0.8228342533111572,
      "learning_rate": 0.0001993950587520142,
      "loss": 0.481,
      "step": 21580
    },
    {
      "epoch": 0.2499652665215579,
      "grad_norm": 0.8094422221183777,
      "learning_rate": 0.00019939365627123957,
      "loss": 0.4977,
      "step": 21590
    },
    {
      "epoch": 0.25008104478303156,
      "grad_norm": 0.7727957963943481,
      "learning_rate": 0.0001993922521715522,
      "loss": 0.4329,
      "step": 21600
    },
    {
      "epoch": 0.2501968230445052,
      "grad_norm": 0.8428878784179688,
      "learning_rate": 0.0001993908464529749,
      "loss": 0.4573,
      "step": 21610
    },
    {
      "epoch": 0.25031260130597877,
      "grad_norm": 0.8136334419250488,
      "learning_rate": 0.00019938943911553065,
      "loss": 0.4714,
      "step": 21620
    },
    {
      "epoch": 0.2504283795674524,
      "grad_norm": 0.6701997518539429,
      "learning_rate": 0.00019938803015924231,
      "loss": 0.5137,
      "step": 21630
    },
    {
      "epoch": 0.25054415782892603,
      "grad_norm": 0.7936267256736755,
      "learning_rate": 0.00019938661958413282,
      "loss": 0.4945,
      "step": 21640
    },
    {
      "epoch": 0.25065993609039966,
      "grad_norm": 0.8736158609390259,
      "learning_rate": 0.00019938520739022525,
      "loss": 0.497,
      "step": 21650
    },
    {
      "epoch": 0.2507757143518733,
      "grad_norm": 0.6858282089233398,
      "learning_rate": 0.00019938379357754252,
      "loss": 0.4433,
      "step": 21660
    },
    {
      "epoch": 0.2508914926133469,
      "grad_norm": 0.6970891356468201,
      "learning_rate": 0.00019938237814610764,
      "loss": 0.4865,
      "step": 21670
    },
    {
      "epoch": 0.25100727087482055,
      "grad_norm": 0.9730721712112427,
      "learning_rate": 0.00019938096109594375,
      "loss": 0.4848,
      "step": 21680
    },
    {
      "epoch": 0.2511230491362942,
      "grad_norm": 0.7481698393821716,
      "learning_rate": 0.00019937954242707387,
      "loss": 0.4829,
      "step": 21690
    },
    {
      "epoch": 0.2512388273977678,
      "grad_norm": 0.9864442348480225,
      "learning_rate": 0.0001993781221395211,
      "loss": 0.4484,
      "step": 21700
    },
    {
      "epoch": 0.25135460565924145,
      "grad_norm": 0.7727804183959961,
      "learning_rate": 0.0001993767002333086,
      "loss": 0.4431,
      "step": 21710
    },
    {
      "epoch": 0.251470383920715,
      "grad_norm": 0.8138746023178101,
      "learning_rate": 0.0001993752767084595,
      "loss": 0.4761,
      "step": 21720
    },
    {
      "epoch": 0.25158616218218866,
      "grad_norm": 0.7588403224945068,
      "learning_rate": 0.00019937385156499704,
      "loss": 0.4732,
      "step": 21730
    },
    {
      "epoch": 0.2517019404436623,
      "grad_norm": 0.8095316290855408,
      "learning_rate": 0.00019937242480294438,
      "loss": 0.509,
      "step": 21740
    },
    {
      "epoch": 0.2518177187051359,
      "grad_norm": 0.6787450313568115,
      "learning_rate": 0.0001993709964223248,
      "loss": 0.4875,
      "step": 21750
    },
    {
      "epoch": 0.25193349696660955,
      "grad_norm": 0.7589474320411682,
      "learning_rate": 0.00019936956642316154,
      "loss": 0.4774,
      "step": 21760
    },
    {
      "epoch": 0.2520492752280832,
      "grad_norm": 0.7602103352546692,
      "learning_rate": 0.0001993681348054779,
      "loss": 0.499,
      "step": 21770
    },
    {
      "epoch": 0.2521650534895568,
      "grad_norm": 0.8325792551040649,
      "learning_rate": 0.00019936670156929718,
      "loss": 0.4796,
      "step": 21780
    },
    {
      "epoch": 0.25228083175103044,
      "grad_norm": 0.7022053599357605,
      "learning_rate": 0.00019936526671464273,
      "loss": 0.4733,
      "step": 21790
    },
    {
      "epoch": 0.2523966100125041,
      "grad_norm": 0.8736443519592285,
      "learning_rate": 0.00019936383024153796,
      "loss": 0.487,
      "step": 21800
    },
    {
      "epoch": 0.25251238827397765,
      "grad_norm": 0.6411970853805542,
      "learning_rate": 0.0001993623921500062,
      "loss": 0.4415,
      "step": 21810
    },
    {
      "epoch": 0.2526281665354513,
      "grad_norm": 0.9100762605667114,
      "learning_rate": 0.00019936095244007095,
      "loss": 0.5196,
      "step": 21820
    },
    {
      "epoch": 0.2527439447969249,
      "grad_norm": 0.8505499362945557,
      "learning_rate": 0.0001993595111117556,
      "loss": 0.4791,
      "step": 21830
    },
    {
      "epoch": 0.25285972305839854,
      "grad_norm": 0.7599483132362366,
      "learning_rate": 0.00019935806816508364,
      "loss": 0.4553,
      "step": 21840
    },
    {
      "epoch": 0.2529755013198722,
      "grad_norm": 0.8440554141998291,
      "learning_rate": 0.00019935662360007862,
      "loss": 0.5143,
      "step": 21850
    },
    {
      "epoch": 0.2530912795813458,
      "grad_norm": 0.5842945575714111,
      "learning_rate": 0.00019935517741676397,
      "loss": 0.4722,
      "step": 21860
    },
    {
      "epoch": 0.25320705784281944,
      "grad_norm": 0.734700083732605,
      "learning_rate": 0.00019935372961516333,
      "loss": 0.4861,
      "step": 21870
    },
    {
      "epoch": 0.25332283610429307,
      "grad_norm": 0.8261541128158569,
      "learning_rate": 0.00019935228019530023,
      "loss": 0.4453,
      "step": 21880
    },
    {
      "epoch": 0.2534386143657667,
      "grad_norm": 0.8472097516059875,
      "learning_rate": 0.00019935082915719833,
      "loss": 0.4683,
      "step": 21890
    },
    {
      "epoch": 0.25355439262724033,
      "grad_norm": 0.8597475290298462,
      "learning_rate": 0.00019934937650088123,
      "loss": 0.4675,
      "step": 21900
    },
    {
      "epoch": 0.2536701708887139,
      "grad_norm": 0.7931632995605469,
      "learning_rate": 0.0001993479222263726,
      "loss": 0.4528,
      "step": 21910
    },
    {
      "epoch": 0.25378594915018754,
      "grad_norm": 0.7407145500183105,
      "learning_rate": 0.00019934646633369614,
      "loss": 0.4589,
      "step": 21920
    },
    {
      "epoch": 0.25390172741166117,
      "grad_norm": 0.7493576407432556,
      "learning_rate": 0.00019934500882287553,
      "loss": 0.4665,
      "step": 21930
    },
    {
      "epoch": 0.2540175056731348,
      "grad_norm": 0.7910557985305786,
      "learning_rate": 0.00019934354969393457,
      "loss": 0.4462,
      "step": 21940
    },
    {
      "epoch": 0.25413328393460843,
      "grad_norm": 0.617962658405304,
      "learning_rate": 0.00019934208894689692,
      "loss": 0.4332,
      "step": 21950
    },
    {
      "epoch": 0.25424906219608207,
      "grad_norm": 0.7557736039161682,
      "learning_rate": 0.00019934062658178648,
      "loss": 0.4714,
      "step": 21960
    },
    {
      "epoch": 0.2543648404575557,
      "grad_norm": 0.9161111116409302,
      "learning_rate": 0.00019933916259862698,
      "loss": 0.4622,
      "step": 21970
    },
    {
      "epoch": 0.25448061871902933,
      "grad_norm": 0.8381891846656799,
      "learning_rate": 0.00019933769699744235,
      "loss": 0.4812,
      "step": 21980
    },
    {
      "epoch": 0.25459639698050296,
      "grad_norm": 0.7691875696182251,
      "learning_rate": 0.00019933622977825644,
      "loss": 0.4555,
      "step": 21990
    },
    {
      "epoch": 0.2547121752419766,
      "grad_norm": 0.6883540749549866,
      "learning_rate": 0.00019933476094109308,
      "loss": 0.4753,
      "step": 22000
    },
    {
      "epoch": 0.2547121752419766,
      "eval_chrf": 75.9492548954396,
      "eval_loss": 0.7260739803314209,
      "eval_runtime": 345.7023,
      "eval_samples_per_second": 0.289,
      "eval_steps_per_second": 0.012,
      "step": 22000
    },
    {
      "epoch": 0.25482795350345017,
      "grad_norm": 0.7204933762550354,
      "learning_rate": 0.00019933329048597625,
      "loss": 0.4661,
      "step": 22010
    },
    {
      "epoch": 0.2549437317649238,
      "grad_norm": 0.7734494209289551,
      "learning_rate": 0.00019933181841292995,
      "loss": 0.4531,
      "step": 22020
    },
    {
      "epoch": 0.25505951002639743,
      "grad_norm": 0.8410596251487732,
      "learning_rate": 0.00019933034472197804,
      "loss": 0.4724,
      "step": 22030
    },
    {
      "epoch": 0.25517528828787106,
      "grad_norm": 0.8805966377258301,
      "learning_rate": 0.0001993288694131446,
      "loss": 0.505,
      "step": 22040
    },
    {
      "epoch": 0.2552910665493447,
      "grad_norm": 0.8467724323272705,
      "learning_rate": 0.00019932739248645367,
      "loss": 0.4869,
      "step": 22050
    },
    {
      "epoch": 0.2554068448108183,
      "grad_norm": 0.74871826171875,
      "learning_rate": 0.00019932591394192928,
      "loss": 0.4997,
      "step": 22060
    },
    {
      "epoch": 0.25552262307229195,
      "grad_norm": 0.6654328107833862,
      "learning_rate": 0.00019932443377959549,
      "loss": 0.4745,
      "step": 22070
    },
    {
      "epoch": 0.2556384013337656,
      "grad_norm": 0.7821851968765259,
      "learning_rate": 0.00019932295199947644,
      "loss": 0.4505,
      "step": 22080
    },
    {
      "epoch": 0.2557541795952392,
      "grad_norm": 0.8141744136810303,
      "learning_rate": 0.00019932146860159624,
      "loss": 0.4815,
      "step": 22090
    },
    {
      "epoch": 0.25586995785671285,
      "grad_norm": 0.9703858494758606,
      "learning_rate": 0.0001993199835859791,
      "loss": 0.4784,
      "step": 22100
    },
    {
      "epoch": 0.2559857361181864,
      "grad_norm": 0.6984914541244507,
      "learning_rate": 0.00019931849695264918,
      "loss": 0.4871,
      "step": 22110
    },
    {
      "epoch": 0.25610151437966006,
      "grad_norm": 0.7303805947303772,
      "learning_rate": 0.00019931700870163066,
      "loss": 0.4643,
      "step": 22120
    },
    {
      "epoch": 0.2562172926411337,
      "grad_norm": 1.0366698503494263,
      "learning_rate": 0.0001993155188329478,
      "loss": 0.4792,
      "step": 22130
    },
    {
      "epoch": 0.2563330709026073,
      "grad_norm": 0.7457212209701538,
      "learning_rate": 0.00019931402734662493,
      "loss": 0.4921,
      "step": 22140
    },
    {
      "epoch": 0.25644884916408095,
      "grad_norm": 0.7296402454376221,
      "learning_rate": 0.00019931253424268627,
      "loss": 0.4553,
      "step": 22150
    },
    {
      "epoch": 0.2565646274255546,
      "grad_norm": 0.9333901405334473,
      "learning_rate": 0.00019931103952115615,
      "loss": 0.4818,
      "step": 22160
    },
    {
      "epoch": 0.2566804056870282,
      "grad_norm": 0.7972484827041626,
      "learning_rate": 0.00019930954318205893,
      "loss": 0.4667,
      "step": 22170
    },
    {
      "epoch": 0.25679618394850184,
      "grad_norm": 0.7563261389732361,
      "learning_rate": 0.00019930804522541896,
      "loss": 0.5111,
      "step": 22180
    },
    {
      "epoch": 0.2569119622099755,
      "grad_norm": 0.7163794040679932,
      "learning_rate": 0.0001993065456512607,
      "loss": 0.4539,
      "step": 22190
    },
    {
      "epoch": 0.2570277404714491,
      "grad_norm": 0.7803443670272827,
      "learning_rate": 0.00019930504445960848,
      "loss": 0.4456,
      "step": 22200
    },
    {
      "epoch": 0.2571435187329227,
      "grad_norm": 0.8861815929412842,
      "learning_rate": 0.00019930354165048685,
      "loss": 0.4574,
      "step": 22210
    },
    {
      "epoch": 0.2572592969943963,
      "grad_norm": 0.84950190782547,
      "learning_rate": 0.00019930203722392023,
      "loss": 0.4711,
      "step": 22220
    },
    {
      "epoch": 0.25737507525586995,
      "grad_norm": 0.6551049947738647,
      "learning_rate": 0.0001993005311799331,
      "loss": 0.4649,
      "step": 22230
    },
    {
      "epoch": 0.2574908535173436,
      "grad_norm": 0.8828011155128479,
      "learning_rate": 0.00019929902351855007,
      "loss": 0.4679,
      "step": 22240
    },
    {
      "epoch": 0.2576066317788172,
      "grad_norm": 0.5688215494155884,
      "learning_rate": 0.0001992975142397956,
      "loss": 0.4583,
      "step": 22250
    },
    {
      "epoch": 0.25772241004029084,
      "grad_norm": 0.7960038781166077,
      "learning_rate": 0.0001992960033436944,
      "loss": 0.4614,
      "step": 22260
    },
    {
      "epoch": 0.25783818830176447,
      "grad_norm": 0.6251015663146973,
      "learning_rate": 0.00019929449083027094,
      "loss": 0.4495,
      "step": 22270
    },
    {
      "epoch": 0.2579539665632381,
      "grad_norm": 0.6125766634941101,
      "learning_rate": 0.00019929297669954993,
      "loss": 0.4775,
      "step": 22280
    },
    {
      "epoch": 0.25806974482471173,
      "grad_norm": 0.6822196245193481,
      "learning_rate": 0.00019929146095155607,
      "loss": 0.4689,
      "step": 22290
    },
    {
      "epoch": 0.25818552308618536,
      "grad_norm": 0.6716240048408508,
      "learning_rate": 0.00019928994358631394,
      "loss": 0.4899,
      "step": 22300
    },
    {
      "epoch": 0.25830130134765894,
      "grad_norm": 0.9194414019584656,
      "learning_rate": 0.00019928842460384833,
      "loss": 0.4918,
      "step": 22310
    },
    {
      "epoch": 0.25841707960913257,
      "grad_norm": 0.6665068864822388,
      "learning_rate": 0.000199286904004184,
      "loss": 0.4822,
      "step": 22320
    },
    {
      "epoch": 0.2585328578706062,
      "grad_norm": 0.77483731508255,
      "learning_rate": 0.00019928538178734563,
      "loss": 0.457,
      "step": 22330
    },
    {
      "epoch": 0.25864863613207983,
      "grad_norm": 0.7766614556312561,
      "learning_rate": 0.00019928385795335812,
      "loss": 0.4912,
      "step": 22340
    },
    {
      "epoch": 0.25876441439355347,
      "grad_norm": 0.7784126400947571,
      "learning_rate": 0.0001992823325022462,
      "loss": 0.4855,
      "step": 22350
    },
    {
      "epoch": 0.2588801926550271,
      "grad_norm": 0.6947225332260132,
      "learning_rate": 0.00019928080543403476,
      "loss": 0.496,
      "step": 22360
    },
    {
      "epoch": 0.25899597091650073,
      "grad_norm": 0.675093412399292,
      "learning_rate": 0.00019927927674874866,
      "loss": 0.4467,
      "step": 22370
    },
    {
      "epoch": 0.25911174917797436,
      "grad_norm": 0.6956536769866943,
      "learning_rate": 0.0001992777464464128,
      "loss": 0.4612,
      "step": 22380
    },
    {
      "epoch": 0.259227527439448,
      "grad_norm": 0.7089182734489441,
      "learning_rate": 0.00019927621452705215,
      "loss": 0.4629,
      "step": 22390
    },
    {
      "epoch": 0.2593433057009216,
      "grad_norm": 0.7213327884674072,
      "learning_rate": 0.0001992746809906916,
      "loss": 0.456,
      "step": 22400
    },
    {
      "epoch": 0.2594590839623952,
      "grad_norm": 0.847394585609436,
      "learning_rate": 0.0001992731458373561,
      "loss": 0.4935,
      "step": 22410
    },
    {
      "epoch": 0.25957486222386883,
      "grad_norm": 0.6759060025215149,
      "learning_rate": 0.00019927160906707075,
      "loss": 0.4527,
      "step": 22420
    },
    {
      "epoch": 0.25969064048534246,
      "grad_norm": 1.4248216152191162,
      "learning_rate": 0.00019927007067986055,
      "loss": 0.4705,
      "step": 22430
    },
    {
      "epoch": 0.2598064187468161,
      "grad_norm": 0.8298565745353699,
      "learning_rate": 0.0001992685306757505,
      "loss": 0.4621,
      "step": 22440
    },
    {
      "epoch": 0.2599221970082897,
      "grad_norm": 0.735045850276947,
      "learning_rate": 0.00019926698905476577,
      "loss": 0.4994,
      "step": 22450
    },
    {
      "epoch": 0.26003797526976336,
      "grad_norm": 0.7810381650924683,
      "learning_rate": 0.0001992654458169314,
      "loss": 0.4553,
      "step": 22460
    },
    {
      "epoch": 0.260153753531237,
      "grad_norm": 0.9417708516120911,
      "learning_rate": 0.00019926390096227257,
      "loss": 0.4583,
      "step": 22470
    },
    {
      "epoch": 0.2602695317927106,
      "grad_norm": 0.8525703549385071,
      "learning_rate": 0.0001992623544908144,
      "loss": 0.4621,
      "step": 22480
    },
    {
      "epoch": 0.26038531005418425,
      "grad_norm": 0.8209181427955627,
      "learning_rate": 0.00019926080640258215,
      "loss": 0.4768,
      "step": 22490
    },
    {
      "epoch": 0.2605010883156578,
      "grad_norm": 0.7920557260513306,
      "learning_rate": 0.00019925925669760092,
      "loss": 0.4422,
      "step": 22500
    },
    {
      "epoch": 0.26061686657713146,
      "grad_norm": 0.7790732383728027,
      "learning_rate": 0.00019925770537589606,
      "loss": 0.46,
      "step": 22510
    },
    {
      "epoch": 0.2607326448386051,
      "grad_norm": 0.7818410992622375,
      "learning_rate": 0.00019925615243749278,
      "loss": 0.4721,
      "step": 22520
    },
    {
      "epoch": 0.2608484231000787,
      "grad_norm": 0.9033719897270203,
      "learning_rate": 0.0001992545978824164,
      "loss": 0.4763,
      "step": 22530
    },
    {
      "epoch": 0.26096420136155235,
      "grad_norm": 0.8526149988174438,
      "learning_rate": 0.00019925304171069223,
      "loss": 0.4666,
      "step": 22540
    },
    {
      "epoch": 0.261079979623026,
      "grad_norm": 0.7692511677742004,
      "learning_rate": 0.00019925148392234562,
      "loss": 0.4616,
      "step": 22550
    },
    {
      "epoch": 0.2611957578844996,
      "grad_norm": 0.926516592502594,
      "learning_rate": 0.00019924992451740195,
      "loss": 0.4872,
      "step": 22560
    },
    {
      "epoch": 0.26131153614597324,
      "grad_norm": 0.9964279532432556,
      "learning_rate": 0.00019924836349588662,
      "loss": 0.4507,
      "step": 22570
    },
    {
      "epoch": 0.2614273144074469,
      "grad_norm": 0.7452640533447266,
      "learning_rate": 0.000199246800857825,
      "loss": 0.4398,
      "step": 22580
    },
    {
      "epoch": 0.2615430926689205,
      "grad_norm": 0.7037336826324463,
      "learning_rate": 0.0001992452366032426,
      "loss": 0.4625,
      "step": 22590
    },
    {
      "epoch": 0.2616588709303941,
      "grad_norm": 0.763759970664978,
      "learning_rate": 0.0001992436707321649,
      "loss": 0.4477,
      "step": 22600
    },
    {
      "epoch": 0.2617746491918677,
      "grad_norm": 0.7811423540115356,
      "learning_rate": 0.00019924210324461742,
      "loss": 0.4678,
      "step": 22610
    },
    {
      "epoch": 0.26189042745334135,
      "grad_norm": 0.8694429993629456,
      "learning_rate": 0.00019924053414062563,
      "loss": 0.4475,
      "step": 22620
    },
    {
      "epoch": 0.262006205714815,
      "grad_norm": 0.7192367911338806,
      "learning_rate": 0.0001992389634202151,
      "loss": 0.4471,
      "step": 22630
    },
    {
      "epoch": 0.2621219839762886,
      "grad_norm": 0.7533650994300842,
      "learning_rate": 0.00019923739108341148,
      "loss": 0.4707,
      "step": 22640
    },
    {
      "epoch": 0.26223776223776224,
      "grad_norm": 0.8486232757568359,
      "learning_rate": 0.00019923581713024032,
      "loss": 0.4706,
      "step": 22650
    },
    {
      "epoch": 0.26235354049923587,
      "grad_norm": 0.6653669476509094,
      "learning_rate": 0.00019923424156072724,
      "loss": 0.4486,
      "step": 22660
    },
    {
      "epoch": 0.2624693187607095,
      "grad_norm": 0.7561994194984436,
      "learning_rate": 0.00019923266437489798,
      "loss": 0.4873,
      "step": 22670
    },
    {
      "epoch": 0.26258509702218313,
      "grad_norm": 0.7854382395744324,
      "learning_rate": 0.00019923108557277813,
      "loss": 0.4639,
      "step": 22680
    },
    {
      "epoch": 0.26270087528365677,
      "grad_norm": 0.7397041320800781,
      "learning_rate": 0.00019922950515439348,
      "loss": 0.4456,
      "step": 22690
    },
    {
      "epoch": 0.26281665354513034,
      "grad_norm": 0.8341320753097534,
      "learning_rate": 0.0001992279231197698,
      "loss": 0.4848,
      "step": 22700
    },
    {
      "epoch": 0.262932431806604,
      "grad_norm": 0.8480237722396851,
      "learning_rate": 0.00019922633946893275,
      "loss": 0.4301,
      "step": 22710
    },
    {
      "epoch": 0.2630482100680776,
      "grad_norm": 1.00328528881073,
      "learning_rate": 0.0001992247542019082,
      "loss": 0.4899,
      "step": 22720
    },
    {
      "epoch": 0.26316398832955123,
      "grad_norm": 0.7801504731178284,
      "learning_rate": 0.00019922316731872192,
      "loss": 0.4937,
      "step": 22730
    },
    {
      "epoch": 0.26327976659102487,
      "grad_norm": 0.7921590805053711,
      "learning_rate": 0.00019922157881939982,
      "loss": 0.4949,
      "step": 22740
    },
    {
      "epoch": 0.2633955448524985,
      "grad_norm": 0.5391880869865417,
      "learning_rate": 0.00019921998870396775,
      "loss": 0.4723,
      "step": 22750
    },
    {
      "epoch": 0.26351132311397213,
      "grad_norm": 0.8103504180908203,
      "learning_rate": 0.00019921839697245158,
      "loss": 0.5043,
      "step": 22760
    },
    {
      "epoch": 0.26362710137544576,
      "grad_norm": 0.9150616526603699,
      "learning_rate": 0.00019921680362487728,
      "loss": 0.5382,
      "step": 22770
    },
    {
      "epoch": 0.2637428796369194,
      "grad_norm": 0.9048618078231812,
      "learning_rate": 0.00019921520866127078,
      "loss": 0.5033,
      "step": 22780
    },
    {
      "epoch": 0.263858657898393,
      "grad_norm": 0.8628024458885193,
      "learning_rate": 0.00019921361208165804,
      "loss": 0.4746,
      "step": 22790
    },
    {
      "epoch": 0.2639744361598666,
      "grad_norm": 0.7364067435264587,
      "learning_rate": 0.00019921201388606506,
      "loss": 0.4604,
      "step": 22800
    },
    {
      "epoch": 0.26409021442134023,
      "grad_norm": 0.8640881180763245,
      "learning_rate": 0.00019921041407451793,
      "loss": 0.4433,
      "step": 22810
    },
    {
      "epoch": 0.26420599268281386,
      "grad_norm": 0.8773605227470398,
      "learning_rate": 0.00019920881264704266,
      "loss": 0.4701,
      "step": 22820
    },
    {
      "epoch": 0.2643217709442875,
      "grad_norm": 0.8174883127212524,
      "learning_rate": 0.00019920720960366536,
      "loss": 0.4318,
      "step": 22830
    },
    {
      "epoch": 0.2644375492057611,
      "grad_norm": 0.7369967103004456,
      "learning_rate": 0.00019920560494441212,
      "loss": 0.4481,
      "step": 22840
    },
    {
      "epoch": 0.26455332746723476,
      "grad_norm": 0.8240845799446106,
      "learning_rate": 0.00019920399866930904,
      "loss": 0.4604,
      "step": 22850
    },
    {
      "epoch": 0.2646691057287084,
      "grad_norm": 0.5480002164840698,
      "learning_rate": 0.00019920239077838239,
      "loss": 0.4587,
      "step": 22860
    },
    {
      "epoch": 0.264784883990182,
      "grad_norm": 0.8237443566322327,
      "learning_rate": 0.00019920078127165825,
      "loss": 0.4721,
      "step": 22870
    },
    {
      "epoch": 0.26490066225165565,
      "grad_norm": 0.7899153232574463,
      "learning_rate": 0.00019919917014916294,
      "loss": 0.5098,
      "step": 22880
    },
    {
      "epoch": 0.2650164405131293,
      "grad_norm": 0.843370258808136,
      "learning_rate": 0.00019919755741092262,
      "loss": 0.4825,
      "step": 22890
    },
    {
      "epoch": 0.26513221877460286,
      "grad_norm": 0.8073774576187134,
      "learning_rate": 0.00019919594305696354,
      "loss": 0.417,
      "step": 22900
    },
    {
      "epoch": 0.2652479970360765,
      "grad_norm": 0.8770199418067932,
      "learning_rate": 0.00019919432708731205,
      "loss": 0.4449,
      "step": 22910
    },
    {
      "epoch": 0.2653637752975501,
      "grad_norm": 0.6097248196601868,
      "learning_rate": 0.00019919270950199445,
      "loss": 0.4898,
      "step": 22920
    },
    {
      "epoch": 0.26547955355902375,
      "grad_norm": 0.8626429438591003,
      "learning_rate": 0.00019919109030103714,
      "loss": 0.451,
      "step": 22930
    },
    {
      "epoch": 0.2655953318204974,
      "grad_norm": 0.9127733111381531,
      "learning_rate": 0.0001991894694844664,
      "loss": 0.4507,
      "step": 22940
    },
    {
      "epoch": 0.265711110081971,
      "grad_norm": 0.853125810623169,
      "learning_rate": 0.0001991878470523087,
      "loss": 0.4632,
      "step": 22950
    },
    {
      "epoch": 0.26582688834344465,
      "grad_norm": 0.8941962122917175,
      "learning_rate": 0.0001991862230045904,
      "loss": 0.4596,
      "step": 22960
    },
    {
      "epoch": 0.2659426666049183,
      "grad_norm": 0.7871774435043335,
      "learning_rate": 0.00019918459734133802,
      "loss": 0.4729,
      "step": 22970
    },
    {
      "epoch": 0.2660584448663919,
      "grad_norm": 0.806376576423645,
      "learning_rate": 0.000199182970062578,
      "loss": 0.4821,
      "step": 22980
    },
    {
      "epoch": 0.26617422312786554,
      "grad_norm": 0.8112010359764099,
      "learning_rate": 0.0001991813411683369,
      "loss": 0.458,
      "step": 22990
    },
    {
      "epoch": 0.2662900013893391,
      "grad_norm": 0.8459729552268982,
      "learning_rate": 0.00019917971065864119,
      "loss": 0.4379,
      "step": 23000
    },
    {
      "epoch": 0.2662900013893391,
      "eval_chrf": 83.42543709513681,
      "eval_loss": 0.7086561322212219,
      "eval_runtime": 195.9424,
      "eval_samples_per_second": 0.51,
      "eval_steps_per_second": 0.02,
      "step": 23000
    },
    {
      "epoch": 0.26640577965081275,
      "grad_norm": 0.8854068517684937,
      "learning_rate": 0.0001991780785335174,
      "loss": 0.4698,
      "step": 23010
    },
    {
      "epoch": 0.2665215579122864,
      "grad_norm": 0.7783320546150208,
      "learning_rate": 0.00019917644479299222,
      "loss": 0.4933,
      "step": 23020
    },
    {
      "epoch": 0.26663733617376,
      "grad_norm": 0.811390221118927,
      "learning_rate": 0.00019917480943709216,
      "loss": 0.4801,
      "step": 23030
    },
    {
      "epoch": 0.26675311443523364,
      "grad_norm": 0.8113588094711304,
      "learning_rate": 0.00019917317246584393,
      "loss": 0.4427,
      "step": 23040
    },
    {
      "epoch": 0.26686889269670727,
      "grad_norm": 0.8738912343978882,
      "learning_rate": 0.00019917153387927417,
      "loss": 0.438,
      "step": 23050
    },
    {
      "epoch": 0.2669846709581809,
      "grad_norm": 0.7134190201759338,
      "learning_rate": 0.00019916989367740952,
      "loss": 0.456,
      "step": 23060
    },
    {
      "epoch": 0.26710044921965453,
      "grad_norm": 0.7650332450866699,
      "learning_rate": 0.00019916825186027675,
      "loss": 0.4806,
      "step": 23070
    },
    {
      "epoch": 0.26721622748112817,
      "grad_norm": 0.8252532482147217,
      "learning_rate": 0.00019916660842790257,
      "loss": 0.4686,
      "step": 23080
    },
    {
      "epoch": 0.26733200574260174,
      "grad_norm": 0.8009850978851318,
      "learning_rate": 0.00019916496338031378,
      "loss": 0.4497,
      "step": 23090
    },
    {
      "epoch": 0.2674477840040754,
      "grad_norm": 0.7717449069023132,
      "learning_rate": 0.00019916331671753713,
      "loss": 0.4125,
      "step": 23100
    },
    {
      "epoch": 0.267563562265549,
      "grad_norm": 0.8198238015174866,
      "learning_rate": 0.00019916166843959952,
      "loss": 0.4651,
      "step": 23110
    },
    {
      "epoch": 0.26767934052702264,
      "grad_norm": 0.8147943019866943,
      "learning_rate": 0.00019916001854652771,
      "loss": 0.4538,
      "step": 23120
    },
    {
      "epoch": 0.26779511878849627,
      "grad_norm": 0.6511247158050537,
      "learning_rate": 0.00019915836703834861,
      "loss": 0.499,
      "step": 23130
    },
    {
      "epoch": 0.2679108970499699,
      "grad_norm": 0.7690098881721497,
      "learning_rate": 0.00019915671391508913,
      "loss": 0.4322,
      "step": 23140
    },
    {
      "epoch": 0.26802667531144353,
      "grad_norm": 0.8259420394897461,
      "learning_rate": 0.00019915505917677618,
      "loss": 0.4421,
      "step": 23150
    },
    {
      "epoch": 0.26814245357291716,
      "grad_norm": 0.7501933574676514,
      "learning_rate": 0.0001991534028234367,
      "loss": 0.4475,
      "step": 23160
    },
    {
      "epoch": 0.2682582318343908,
      "grad_norm": 0.6512454748153687,
      "learning_rate": 0.00019915174485509772,
      "loss": 0.4072,
      "step": 23170
    },
    {
      "epoch": 0.2683740100958644,
      "grad_norm": 0.806197464466095,
      "learning_rate": 0.0001991500852717862,
      "loss": 0.4716,
      "step": 23180
    },
    {
      "epoch": 0.268489788357338,
      "grad_norm": 0.7797796726226807,
      "learning_rate": 0.00019914842407352915,
      "loss": 0.4957,
      "step": 23190
    },
    {
      "epoch": 0.26860556661881163,
      "grad_norm": 0.7226470708847046,
      "learning_rate": 0.0001991467612603537,
      "loss": 0.4495,
      "step": 23200
    },
    {
      "epoch": 0.26872134488028526,
      "grad_norm": 0.8793154358863831,
      "learning_rate": 0.00019914509683228686,
      "loss": 0.4713,
      "step": 23210
    },
    {
      "epoch": 0.2688371231417589,
      "grad_norm": 0.6427780985832214,
      "learning_rate": 0.00019914343078935578,
      "loss": 0.4439,
      "step": 23220
    },
    {
      "epoch": 0.2689529014032325,
      "grad_norm": 0.7348723411560059,
      "learning_rate": 0.00019914176313158758,
      "loss": 0.4319,
      "step": 23230
    },
    {
      "epoch": 0.26906867966470616,
      "grad_norm": 0.7228298783302307,
      "learning_rate": 0.00019914009385900948,
      "loss": 0.4662,
      "step": 23240
    },
    {
      "epoch": 0.2691844579261798,
      "grad_norm": 0.8180272579193115,
      "learning_rate": 0.00019913842297164858,
      "loss": 0.4853,
      "step": 23250
    },
    {
      "epoch": 0.2693002361876534,
      "grad_norm": 0.5619539022445679,
      "learning_rate": 0.00019913675046953215,
      "loss": 0.4434,
      "step": 23260
    },
    {
      "epoch": 0.26941601444912705,
      "grad_norm": 0.7949888110160828,
      "learning_rate": 0.0001991350763526874,
      "loss": 0.4535,
      "step": 23270
    },
    {
      "epoch": 0.2695317927106007,
      "grad_norm": 0.804740309715271,
      "learning_rate": 0.0001991334006211416,
      "loss": 0.4743,
      "step": 23280
    },
    {
      "epoch": 0.26964757097207426,
      "grad_norm": 0.7593053579330444,
      "learning_rate": 0.0001991317232749221,
      "loss": 0.45,
      "step": 23290
    },
    {
      "epoch": 0.2697633492335479,
      "grad_norm": 0.7307462692260742,
      "learning_rate": 0.00019913004431405614,
      "loss": 0.4754,
      "step": 23300
    },
    {
      "epoch": 0.2698791274950215,
      "grad_norm": 0.8215537667274475,
      "learning_rate": 0.00019912836373857116,
      "loss": 0.4781,
      "step": 23310
    },
    {
      "epoch": 0.26999490575649515,
      "grad_norm": 0.8160101175308228,
      "learning_rate": 0.00019912668154849444,
      "loss": 0.4654,
      "step": 23320
    },
    {
      "epoch": 0.2701106840179688,
      "grad_norm": 0.8154664635658264,
      "learning_rate": 0.0001991249977438534,
      "loss": 0.4698,
      "step": 23330
    },
    {
      "epoch": 0.2702264622794424,
      "grad_norm": 0.7719538807868958,
      "learning_rate": 0.00019912331232467553,
      "loss": 0.457,
      "step": 23340
    },
    {
      "epoch": 0.27034224054091605,
      "grad_norm": 0.6792475581169128,
      "learning_rate": 0.00019912162529098822,
      "loss": 0.4638,
      "step": 23350
    },
    {
      "epoch": 0.2704580188023897,
      "grad_norm": 0.700428307056427,
      "learning_rate": 0.00019911993664281896,
      "loss": 0.4896,
      "step": 23360
    },
    {
      "epoch": 0.2705737970638633,
      "grad_norm": 0.6948319673538208,
      "learning_rate": 0.00019911824638019522,
      "loss": 0.4809,
      "step": 23370
    },
    {
      "epoch": 0.27068957532533694,
      "grad_norm": 0.8017446994781494,
      "learning_rate": 0.00019911655450314465,
      "loss": 0.4544,
      "step": 23380
    },
    {
      "epoch": 0.2708053535868105,
      "grad_norm": 0.8272501230239868,
      "learning_rate": 0.00019911486101169466,
      "loss": 0.4468,
      "step": 23390
    },
    {
      "epoch": 0.27092113184828415,
      "grad_norm": 0.7762539386749268,
      "learning_rate": 0.00019911316590587292,
      "loss": 0.4735,
      "step": 23400
    },
    {
      "epoch": 0.2710369101097578,
      "grad_norm": 0.7475146651268005,
      "learning_rate": 0.00019911146918570702,
      "loss": 0.4528,
      "step": 23410
    },
    {
      "epoch": 0.2711526883712314,
      "grad_norm": 0.6763512492179871,
      "learning_rate": 0.0001991097708512246,
      "loss": 0.4755,
      "step": 23420
    },
    {
      "epoch": 0.27126846663270504,
      "grad_norm": 0.933892011642456,
      "learning_rate": 0.00019910807090245332,
      "loss": 0.4846,
      "step": 23430
    },
    {
      "epoch": 0.2713842448941787,
      "grad_norm": 0.800424337387085,
      "learning_rate": 0.00019910636933942086,
      "loss": 0.4333,
      "step": 23440
    },
    {
      "epoch": 0.2715000231556523,
      "grad_norm": 0.9707953929901123,
      "learning_rate": 0.00019910466616215495,
      "loss": 0.4411,
      "step": 23450
    },
    {
      "epoch": 0.27161580141712593,
      "grad_norm": 0.8666991591453552,
      "learning_rate": 0.0001991029613706833,
      "loss": 0.481,
      "step": 23460
    },
    {
      "epoch": 0.27173157967859957,
      "grad_norm": 0.7029927968978882,
      "learning_rate": 0.00019910125496503374,
      "loss": 0.4613,
      "step": 23470
    },
    {
      "epoch": 0.2718473579400732,
      "grad_norm": 0.8219718337059021,
      "learning_rate": 0.00019909954694523398,
      "loss": 0.4427,
      "step": 23480
    },
    {
      "epoch": 0.2719631362015468,
      "grad_norm": 0.7498128414154053,
      "learning_rate": 0.00019909783731131191,
      "loss": 0.5028,
      "step": 23490
    },
    {
      "epoch": 0.2720789144630204,
      "grad_norm": 0.8823832869529724,
      "learning_rate": 0.00019909612606329536,
      "loss": 0.4675,
      "step": 23500
    },
    {
      "epoch": 0.27219469272449404,
      "grad_norm": 0.8480955958366394,
      "learning_rate": 0.00019909441320121215,
      "loss": 0.4718,
      "step": 23510
    },
    {
      "epoch": 0.27231047098596767,
      "grad_norm": 0.8292425870895386,
      "learning_rate": 0.00019909269872509024,
      "loss": 0.4632,
      "step": 23520
    },
    {
      "epoch": 0.2724262492474413,
      "grad_norm": 0.7262698411941528,
      "learning_rate": 0.00019909098263495755,
      "loss": 0.4736,
      "step": 23530
    },
    {
      "epoch": 0.27254202750891493,
      "grad_norm": 0.910237729549408,
      "learning_rate": 0.00019908926493084203,
      "loss": 0.4416,
      "step": 23540
    },
    {
      "epoch": 0.27265780577038856,
      "grad_norm": 0.9715549945831299,
      "learning_rate": 0.0001990875456127716,
      "loss": 0.4589,
      "step": 23550
    },
    {
      "epoch": 0.2727735840318622,
      "grad_norm": 0.767346203327179,
      "learning_rate": 0.00019908582468077435,
      "loss": 0.4341,
      "step": 23560
    },
    {
      "epoch": 0.2728893622933358,
      "grad_norm": 0.8054583668708801,
      "learning_rate": 0.00019908410213487825,
      "loss": 0.4563,
      "step": 23570
    },
    {
      "epoch": 0.27300514055480946,
      "grad_norm": 0.8324166536331177,
      "learning_rate": 0.0001990823779751114,
      "loss": 0.5082,
      "step": 23580
    },
    {
      "epoch": 0.27312091881628303,
      "grad_norm": 1.0959804058074951,
      "learning_rate": 0.00019908065220150184,
      "loss": 0.4788,
      "step": 23590
    },
    {
      "epoch": 0.27323669707775666,
      "grad_norm": 0.800231397151947,
      "learning_rate": 0.00019907892481407772,
      "loss": 0.4492,
      "step": 23600
    },
    {
      "epoch": 0.2733524753392303,
      "grad_norm": 0.789693295955658,
      "learning_rate": 0.00019907719581286712,
      "loss": 0.505,
      "step": 23610
    },
    {
      "epoch": 0.2734682536007039,
      "grad_norm": 0.719916582107544,
      "learning_rate": 0.00019907546519789827,
      "loss": 0.4735,
      "step": 23620
    },
    {
      "epoch": 0.27358403186217756,
      "grad_norm": 0.8244103789329529,
      "learning_rate": 0.0001990737329691993,
      "loss": 0.4967,
      "step": 23630
    },
    {
      "epoch": 0.2736998101236512,
      "grad_norm": 0.818156898021698,
      "learning_rate": 0.00019907199912679846,
      "loss": 0.4655,
      "step": 23640
    },
    {
      "epoch": 0.2738155883851248,
      "grad_norm": 0.7187866568565369,
      "learning_rate": 0.00019907026367072396,
      "loss": 0.4873,
      "step": 23650
    },
    {
      "epoch": 0.27393136664659845,
      "grad_norm": 0.8921357989311218,
      "learning_rate": 0.00019906852660100412,
      "loss": 0.445,
      "step": 23660
    },
    {
      "epoch": 0.2740471449080721,
      "grad_norm": 0.898240864276886,
      "learning_rate": 0.00019906678791766717,
      "loss": 0.4881,
      "step": 23670
    },
    {
      "epoch": 0.2741629231695457,
      "grad_norm": 0.8911371827125549,
      "learning_rate": 0.0001990650476207415,
      "loss": 0.4803,
      "step": 23680
    },
    {
      "epoch": 0.2742787014310193,
      "grad_norm": 0.9182499647140503,
      "learning_rate": 0.00019906330571025537,
      "loss": 0.4325,
      "step": 23690
    },
    {
      "epoch": 0.2743944796924929,
      "grad_norm": 0.7490956783294678,
      "learning_rate": 0.00019906156218623723,
      "loss": 0.4376,
      "step": 23700
    },
    {
      "epoch": 0.27451025795396655,
      "grad_norm": 0.739075243473053,
      "learning_rate": 0.0001990598170487154,
      "loss": 0.4593,
      "step": 23710
    },
    {
      "epoch": 0.2746260362154402,
      "grad_norm": 0.7499712705612183,
      "learning_rate": 0.00019905807029771842,
      "loss": 0.4695,
      "step": 23720
    },
    {
      "epoch": 0.2747418144769138,
      "grad_norm": 0.8054463863372803,
      "learning_rate": 0.0001990563219332746,
      "loss": 0.4476,
      "step": 23730
    },
    {
      "epoch": 0.27485759273838745,
      "grad_norm": 0.7969599366188049,
      "learning_rate": 0.00019905457195541252,
      "loss": 0.4675,
      "step": 23740
    },
    {
      "epoch": 0.2749733709998611,
      "grad_norm": 0.7260996699333191,
      "learning_rate": 0.00019905282036416065,
      "loss": 0.4654,
      "step": 23750
    },
    {
      "epoch": 0.2750891492613347,
      "grad_norm": 1.2067216634750366,
      "learning_rate": 0.00019905106715954754,
      "loss": 0.4651,
      "step": 23760
    },
    {
      "epoch": 0.27520492752280834,
      "grad_norm": 0.8120461702346802,
      "learning_rate": 0.00019904931234160172,
      "loss": 0.4659,
      "step": 23770
    },
    {
      "epoch": 0.2753207057842819,
      "grad_norm": 0.8080825805664062,
      "learning_rate": 0.00019904755591035176,
      "loss": 0.4726,
      "step": 23780
    },
    {
      "epoch": 0.27543648404575555,
      "grad_norm": 0.8611451983451843,
      "learning_rate": 0.0001990457978658263,
      "loss": 0.479,
      "step": 23790
    },
    {
      "epoch": 0.2755522623072292,
      "grad_norm": 0.7268285155296326,
      "learning_rate": 0.00019904403820805396,
      "loss": 0.4904,
      "step": 23800
    },
    {
      "epoch": 0.2756680405687028,
      "grad_norm": 0.5887938141822815,
      "learning_rate": 0.00019904227693706342,
      "loss": 0.4441,
      "step": 23810
    },
    {
      "epoch": 0.27578381883017644,
      "grad_norm": 0.75795978307724,
      "learning_rate": 0.00019904051405288331,
      "loss": 0.4746,
      "step": 23820
    },
    {
      "epoch": 0.2758995970916501,
      "grad_norm": 0.7585770487785339,
      "learning_rate": 0.0001990387495555424,
      "loss": 0.4754,
      "step": 23830
    },
    {
      "epoch": 0.2760153753531237,
      "grad_norm": 0.8079079985618591,
      "learning_rate": 0.00019903698344506942,
      "loss": 0.46,
      "step": 23840
    },
    {
      "epoch": 0.27613115361459734,
      "grad_norm": 0.7400149703025818,
      "learning_rate": 0.00019903521572149315,
      "loss": 0.4719,
      "step": 23850
    },
    {
      "epoch": 0.27624693187607097,
      "grad_norm": 0.6678773760795593,
      "learning_rate": 0.00019903344638484233,
      "loss": 0.4365,
      "step": 23860
    },
    {
      "epoch": 0.2763627101375446,
      "grad_norm": 0.828233540058136,
      "learning_rate": 0.00019903167543514581,
      "loss": 0.467,
      "step": 23870
    },
    {
      "epoch": 0.2764784883990182,
      "grad_norm": 0.7450108528137207,
      "learning_rate": 0.00019902990287243248,
      "loss": 0.4694,
      "step": 23880
    },
    {
      "epoch": 0.2765942666604918,
      "grad_norm": 0.8591787219047546,
      "learning_rate": 0.00019902812869673116,
      "loss": 0.4487,
      "step": 23890
    },
    {
      "epoch": 0.27671004492196544,
      "grad_norm": 0.7450703382492065,
      "learning_rate": 0.00019902635290807076,
      "loss": 0.4408,
      "step": 23900
    },
    {
      "epoch": 0.27682582318343907,
      "grad_norm": 0.850604236125946,
      "learning_rate": 0.00019902457550648018,
      "loss": 0.4954,
      "step": 23910
    },
    {
      "epoch": 0.2769416014449127,
      "grad_norm": 0.9734000563621521,
      "learning_rate": 0.00019902279649198836,
      "loss": 0.4977,
      "step": 23920
    },
    {
      "epoch": 0.27705737970638633,
      "grad_norm": 0.6951797604560852,
      "learning_rate": 0.00019902101586462436,
      "loss": 0.4522,
      "step": 23930
    },
    {
      "epoch": 0.27717315796785996,
      "grad_norm": 0.7991729378700256,
      "learning_rate": 0.00019901923362441708,
      "loss": 0.4731,
      "step": 23940
    },
    {
      "epoch": 0.2772889362293336,
      "grad_norm": 0.6922335028648376,
      "learning_rate": 0.00019901744977139565,
      "loss": 0.4691,
      "step": 23950
    },
    {
      "epoch": 0.2774047144908072,
      "grad_norm": 0.6763920187950134,
      "learning_rate": 0.00019901566430558903,
      "loss": 0.466,
      "step": 23960
    },
    {
      "epoch": 0.27752049275228086,
      "grad_norm": 0.7108004689216614,
      "learning_rate": 0.00019901387722702637,
      "loss": 0.4467,
      "step": 23970
    },
    {
      "epoch": 0.27763627101375443,
      "grad_norm": 0.7842297554016113,
      "learning_rate": 0.00019901208853573675,
      "loss": 0.445,
      "step": 23980
    },
    {
      "epoch": 0.27775204927522806,
      "grad_norm": 0.854042112827301,
      "learning_rate": 0.0001990102982317493,
      "loss": 0.4606,
      "step": 23990
    },
    {
      "epoch": 0.2778678275367017,
      "grad_norm": 0.7814891934394836,
      "learning_rate": 0.00019900850631509318,
      "loss": 0.4947,
      "step": 24000
    },
    {
      "epoch": 0.2778678275367017,
      "eval_chrf": 85.56038719629122,
      "eval_loss": 0.7117055654525757,
      "eval_runtime": 196.3131,
      "eval_samples_per_second": 0.509,
      "eval_steps_per_second": 0.02,
      "step": 24000
    },
    {
      "epoch": 0.2779836057981753,
      "grad_norm": 0.7856383919715881,
      "learning_rate": 0.00019900671278579761,
      "loss": 0.423,
      "step": 24010
    },
    {
      "epoch": 0.27809938405964896,
      "grad_norm": 0.7621785402297974,
      "learning_rate": 0.00019900491764389175,
      "loss": 0.4836,
      "step": 24020
    },
    {
      "epoch": 0.2782151623211226,
      "grad_norm": 0.789074718952179,
      "learning_rate": 0.00019900312088940488,
      "loss": 0.4815,
      "step": 24030
    },
    {
      "epoch": 0.2783309405825962,
      "grad_norm": 0.728230357170105,
      "learning_rate": 0.00019900132252236624,
      "loss": 0.4615,
      "step": 24040
    },
    {
      "epoch": 0.27844671884406985,
      "grad_norm": 0.8728474378585815,
      "learning_rate": 0.00019899952254280514,
      "loss": 0.4514,
      "step": 24050
    },
    {
      "epoch": 0.2785624971055435,
      "grad_norm": 0.8630943894386292,
      "learning_rate": 0.0001989977209507509,
      "loss": 0.4681,
      "step": 24060
    },
    {
      "epoch": 0.2786782753670171,
      "grad_norm": 0.931373655796051,
      "learning_rate": 0.00019899591774623284,
      "loss": 0.4931,
      "step": 24070
    },
    {
      "epoch": 0.2787940536284907,
      "grad_norm": 0.5404430031776428,
      "learning_rate": 0.00019899411292928032,
      "loss": 0.4285,
      "step": 24080
    },
    {
      "epoch": 0.2789098318899643,
      "grad_norm": 0.7951238751411438,
      "learning_rate": 0.0001989923064999228,
      "loss": 0.4866,
      "step": 24090
    },
    {
      "epoch": 0.27902561015143795,
      "grad_norm": 0.8958339691162109,
      "learning_rate": 0.00019899049845818964,
      "loss": 0.4791,
      "step": 24100
    },
    {
      "epoch": 0.2791413884129116,
      "grad_norm": 0.688094973564148,
      "learning_rate": 0.00019898868880411033,
      "loss": 0.4269,
      "step": 24110
    },
    {
      "epoch": 0.2792571666743852,
      "grad_norm": 0.7957499027252197,
      "learning_rate": 0.00019898687753771432,
      "loss": 0.4259,
      "step": 24120
    },
    {
      "epoch": 0.27937294493585885,
      "grad_norm": 0.6640834808349609,
      "learning_rate": 0.00019898506465903113,
      "loss": 0.4807,
      "step": 24130
    },
    {
      "epoch": 0.2794887231973325,
      "grad_norm": 0.6267457008361816,
      "learning_rate": 0.00019898325016809028,
      "loss": 0.4854,
      "step": 24140
    },
    {
      "epoch": 0.2796045014588061,
      "grad_norm": 0.7446527481079102,
      "learning_rate": 0.00019898143406492128,
      "loss": 0.4627,
      "step": 24150
    },
    {
      "epoch": 0.27972027972027974,
      "grad_norm": 0.787577211856842,
      "learning_rate": 0.0001989796163495538,
      "loss": 0.4501,
      "step": 24160
    },
    {
      "epoch": 0.27983605798175337,
      "grad_norm": 0.7797445058822632,
      "learning_rate": 0.0001989777970220174,
      "loss": 0.4859,
      "step": 24170
    },
    {
      "epoch": 0.27995183624322695,
      "grad_norm": 0.7492689490318298,
      "learning_rate": 0.00019897597608234171,
      "loss": 0.4472,
      "step": 24180
    },
    {
      "epoch": 0.2800676145047006,
      "grad_norm": 0.7863853573799133,
      "learning_rate": 0.00019897415353055636,
      "loss": 0.4835,
      "step": 24190
    },
    {
      "epoch": 0.2801833927661742,
      "grad_norm": 0.7232769727706909,
      "learning_rate": 0.0001989723293666911,
      "loss": 0.4603,
      "step": 24200
    },
    {
      "epoch": 0.28029917102764784,
      "grad_norm": 0.70110684633255,
      "learning_rate": 0.00019897050359077558,
      "loss": 0.4713,
      "step": 24210
    },
    {
      "epoch": 0.2804149492891215,
      "grad_norm": 0.8409924507141113,
      "learning_rate": 0.0001989686762028396,
      "loss": 0.4133,
      "step": 24220
    },
    {
      "epoch": 0.2805307275505951,
      "grad_norm": 0.8462908864021301,
      "learning_rate": 0.00019896684720291288,
      "loss": 0.4721,
      "step": 24230
    },
    {
      "epoch": 0.28064650581206874,
      "grad_norm": 0.8973506093025208,
      "learning_rate": 0.00019896501659102522,
      "loss": 0.4737,
      "step": 24240
    },
    {
      "epoch": 0.28076228407354237,
      "grad_norm": 0.9143046736717224,
      "learning_rate": 0.00019896318436720645,
      "loss": 0.4649,
      "step": 24250
    },
    {
      "epoch": 0.280878062335016,
      "grad_norm": 0.7367831468582153,
      "learning_rate": 0.00019896135053148637,
      "loss": 0.4433,
      "step": 24260
    },
    {
      "epoch": 0.28099384059648963,
      "grad_norm": 0.8570603132247925,
      "learning_rate": 0.00019895951508389493,
      "loss": 0.4962,
      "step": 24270
    },
    {
      "epoch": 0.2811096188579632,
      "grad_norm": 0.8422549962997437,
      "learning_rate": 0.00019895767802446192,
      "loss": 0.4869,
      "step": 24280
    },
    {
      "epoch": 0.28122539711943684,
      "grad_norm": 0.6377737522125244,
      "learning_rate": 0.00019895583935321736,
      "loss": 0.4469,
      "step": 24290
    },
    {
      "epoch": 0.28134117538091047,
      "grad_norm": 0.7327654361724854,
      "learning_rate": 0.00019895399907019114,
      "loss": 0.5111,
      "step": 24300
    },
    {
      "epoch": 0.2814569536423841,
      "grad_norm": 0.7901021242141724,
      "learning_rate": 0.00019895215717541327,
      "loss": 0.4488,
      "step": 24310
    },
    {
      "epoch": 0.28157273190385773,
      "grad_norm": 0.7474518418312073,
      "learning_rate": 0.0001989503136689137,
      "loss": 0.4546,
      "step": 24320
    },
    {
      "epoch": 0.28168851016533136,
      "grad_norm": 0.7942332625389099,
      "learning_rate": 0.00019894846855072253,
      "loss": 0.5048,
      "step": 24330
    },
    {
      "epoch": 0.281804288426805,
      "grad_norm": 0.7080331444740295,
      "learning_rate": 0.00019894662182086973,
      "loss": 0.4663,
      "step": 24340
    },
    {
      "epoch": 0.2819200666882786,
      "grad_norm": 0.7460561394691467,
      "learning_rate": 0.00019894477347938544,
      "loss": 0.4765,
      "step": 24350
    },
    {
      "epoch": 0.28203584494975226,
      "grad_norm": 0.8166491389274597,
      "learning_rate": 0.00019894292352629975,
      "loss": 0.4769,
      "step": 24360
    },
    {
      "epoch": 0.28215162321122583,
      "grad_norm": 0.7675840258598328,
      "learning_rate": 0.00019894107196164279,
      "loss": 0.473,
      "step": 24370
    },
    {
      "epoch": 0.28226740147269946,
      "grad_norm": 0.8223817348480225,
      "learning_rate": 0.00019893921878544473,
      "loss": 0.4489,
      "step": 24380
    },
    {
      "epoch": 0.2823831797341731,
      "grad_norm": 0.8057273626327515,
      "learning_rate": 0.0001989373639977357,
      "loss": 0.4679,
      "step": 24390
    },
    {
      "epoch": 0.2824989579956467,
      "grad_norm": 0.9901914000511169,
      "learning_rate": 0.00019893550759854598,
      "loss": 0.4766,
      "step": 24400
    },
    {
      "epoch": 0.28261473625712036,
      "grad_norm": 0.589401125907898,
      "learning_rate": 0.00019893364958790578,
      "loss": 0.4873,
      "step": 24410
    },
    {
      "epoch": 0.282730514518594,
      "grad_norm": 0.9426624774932861,
      "learning_rate": 0.00019893178996584533,
      "loss": 0.4433,
      "step": 24420
    },
    {
      "epoch": 0.2828462927800676,
      "grad_norm": 0.7681499123573303,
      "learning_rate": 0.000198929928732395,
      "loss": 0.4601,
      "step": 24430
    },
    {
      "epoch": 0.28296207104154125,
      "grad_norm": 0.6982399225234985,
      "learning_rate": 0.00019892806588758504,
      "loss": 0.4303,
      "step": 24440
    },
    {
      "epoch": 0.2830778493030149,
      "grad_norm": 0.8492203950881958,
      "learning_rate": 0.00019892620143144578,
      "loss": 0.494,
      "step": 24450
    },
    {
      "epoch": 0.2831936275644885,
      "grad_norm": 0.6173506379127502,
      "learning_rate": 0.00019892433536400762,
      "loss": 0.4477,
      "step": 24460
    },
    {
      "epoch": 0.2833094058259621,
      "grad_norm": 0.7925242185592651,
      "learning_rate": 0.00019892246768530097,
      "loss": 0.4468,
      "step": 24470
    },
    {
      "epoch": 0.2834251840874357,
      "grad_norm": 0.6825738549232483,
      "learning_rate": 0.00019892059839535624,
      "loss": 0.4374,
      "step": 24480
    },
    {
      "epoch": 0.28354096234890935,
      "grad_norm": 0.8898781538009644,
      "learning_rate": 0.00019891872749420384,
      "loss": 0.4671,
      "step": 24490
    },
    {
      "epoch": 0.283656740610383,
      "grad_norm": 0.8128108978271484,
      "learning_rate": 0.0001989168549818743,
      "loss": 0.429,
      "step": 24500
    },
    {
      "epoch": 0.2837725188718566,
      "grad_norm": 0.6619224548339844,
      "learning_rate": 0.00019891498085839807,
      "loss": 0.4636,
      "step": 24510
    },
    {
      "epoch": 0.28388829713333025,
      "grad_norm": 0.9081688523292542,
      "learning_rate": 0.0001989131051238057,
      "loss": 0.4666,
      "step": 24520
    },
    {
      "epoch": 0.2840040753948039,
      "grad_norm": 0.7981681227684021,
      "learning_rate": 0.0001989112277781277,
      "loss": 0.4642,
      "step": 24530
    },
    {
      "epoch": 0.2841198536562775,
      "grad_norm": 0.6607206463813782,
      "learning_rate": 0.00019890934882139477,
      "loss": 0.4868,
      "step": 24540
    },
    {
      "epoch": 0.28423563191775114,
      "grad_norm": 0.9970320463180542,
      "learning_rate": 0.00019890746825363737,
      "loss": 0.4686,
      "step": 24550
    },
    {
      "epoch": 0.2843514101792248,
      "grad_norm": 0.7873477935791016,
      "learning_rate": 0.00019890558607488622,
      "loss": 0.4781,
      "step": 24560
    },
    {
      "epoch": 0.28446718844069835,
      "grad_norm": 0.7801017165184021,
      "learning_rate": 0.0001989037022851719,
      "loss": 0.493,
      "step": 24570
    },
    {
      "epoch": 0.284582966702172,
      "grad_norm": 0.7829385995864868,
      "learning_rate": 0.0001989018168845252,
      "loss": 0.4723,
      "step": 24580
    },
    {
      "epoch": 0.2846987449636456,
      "grad_norm": 0.7387408018112183,
      "learning_rate": 0.0001988999298729767,
      "loss": 0.4633,
      "step": 24590
    },
    {
      "epoch": 0.28481452322511924,
      "grad_norm": 0.7763007283210754,
      "learning_rate": 0.00019889804125055726,
      "loss": 0.4604,
      "step": 24600
    },
    {
      "epoch": 0.2849303014865929,
      "grad_norm": 0.8328849673271179,
      "learning_rate": 0.00019889615101729755,
      "loss": 0.458,
      "step": 24610
    },
    {
      "epoch": 0.2850460797480665,
      "grad_norm": 0.7388263940811157,
      "learning_rate": 0.00019889425917322842,
      "loss": 0.4495,
      "step": 24620
    },
    {
      "epoch": 0.28516185800954014,
      "grad_norm": 0.658170223236084,
      "learning_rate": 0.00019889236571838065,
      "loss": 0.4284,
      "step": 24630
    },
    {
      "epoch": 0.28527763627101377,
      "grad_norm": 0.7450670599937439,
      "learning_rate": 0.00019889047065278507,
      "loss": 0.4401,
      "step": 24640
    },
    {
      "epoch": 0.2853934145324874,
      "grad_norm": 0.9235295653343201,
      "learning_rate": 0.0001988885739764726,
      "loss": 0.4443,
      "step": 24650
    },
    {
      "epoch": 0.28550919279396103,
      "grad_norm": 0.8104484677314758,
      "learning_rate": 0.00019888667568947405,
      "loss": 0.475,
      "step": 24660
    },
    {
      "epoch": 0.2856249710554346,
      "grad_norm": 0.7793791890144348,
      "learning_rate": 0.00019888477579182043,
      "loss": 0.4617,
      "step": 24670
    },
    {
      "epoch": 0.28574074931690824,
      "grad_norm": 0.8632597327232361,
      "learning_rate": 0.00019888287428354264,
      "loss": 0.4786,
      "step": 24680
    },
    {
      "epoch": 0.28585652757838187,
      "grad_norm": 0.9871675968170166,
      "learning_rate": 0.00019888097116467162,
      "loss": 0.4408,
      "step": 24690
    },
    {
      "epoch": 0.2859723058398555,
      "grad_norm": 0.7056323289871216,
      "learning_rate": 0.00019887906643523845,
      "loss": 0.4515,
      "step": 24700
    },
    {
      "epoch": 0.28608808410132913,
      "grad_norm": 0.7838559746742249,
      "learning_rate": 0.0001988771600952741,
      "loss": 0.4804,
      "step": 24710
    },
    {
      "epoch": 0.28620386236280276,
      "grad_norm": 1.0138752460479736,
      "learning_rate": 0.00019887525214480958,
      "loss": 0.4448,
      "step": 24720
    },
    {
      "epoch": 0.2863196406242764,
      "grad_norm": 0.740874171257019,
      "learning_rate": 0.00019887334258387603,
      "loss": 0.4401,
      "step": 24730
    },
    {
      "epoch": 0.28643541888575,
      "grad_norm": 0.8224324584007263,
      "learning_rate": 0.00019887143141250456,
      "loss": 0.4729,
      "step": 24740
    },
    {
      "epoch": 0.28655119714722366,
      "grad_norm": 0.6226980090141296,
      "learning_rate": 0.00019886951863072624,
      "loss": 0.4261,
      "step": 24750
    },
    {
      "epoch": 0.2866669754086973,
      "grad_norm": 0.6898742914199829,
      "learning_rate": 0.0001988676042385723,
      "loss": 0.4599,
      "step": 24760
    },
    {
      "epoch": 0.28678275367017086,
      "grad_norm": 0.792374312877655,
      "learning_rate": 0.00019886568823607385,
      "loss": 0.4299,
      "step": 24770
    },
    {
      "epoch": 0.2868985319316445,
      "grad_norm": 0.8704492449760437,
      "learning_rate": 0.0001988637706232621,
      "loss": 0.4329,
      "step": 24780
    },
    {
      "epoch": 0.2870143101931181,
      "grad_norm": 0.8298460245132446,
      "learning_rate": 0.00019886185140016838,
      "loss": 0.4673,
      "step": 24790
    },
    {
      "epoch": 0.28713008845459176,
      "grad_norm": 0.8129158020019531,
      "learning_rate": 0.00019885993056682384,
      "loss": 0.4697,
      "step": 24800
    },
    {
      "epoch": 0.2872458667160654,
      "grad_norm": 0.8497953414916992,
      "learning_rate": 0.00019885800812325982,
      "loss": 0.4758,
      "step": 24810
    },
    {
      "epoch": 0.287361644977539,
      "grad_norm": 0.7291585206985474,
      "learning_rate": 0.0001988560840695076,
      "loss": 0.4651,
      "step": 24820
    },
    {
      "epoch": 0.28747742323901265,
      "grad_norm": 0.7276846170425415,
      "learning_rate": 0.00019885415840559855,
      "loss": 0.4444,
      "step": 24830
    },
    {
      "epoch": 0.2875932015004863,
      "grad_norm": 0.8416005969047546,
      "learning_rate": 0.00019885223113156407,
      "loss": 0.4598,
      "step": 24840
    },
    {
      "epoch": 0.2877089797619599,
      "grad_norm": 0.7282233834266663,
      "learning_rate": 0.00019885030224743545,
      "loss": 0.455,
      "step": 24850
    },
    {
      "epoch": 0.28782475802343355,
      "grad_norm": 0.8418949842453003,
      "learning_rate": 0.00019884837175324421,
      "loss": 0.4815,
      "step": 24860
    },
    {
      "epoch": 0.2879405362849071,
      "grad_norm": 0.9172649383544922,
      "learning_rate": 0.0001988464396490217,
      "loss": 0.4585,
      "step": 24870
    },
    {
      "epoch": 0.28805631454638075,
      "grad_norm": 0.8532072305679321,
      "learning_rate": 0.00019884450593479948,
      "loss": 0.465,
      "step": 24880
    },
    {
      "epoch": 0.2881720928078544,
      "grad_norm": 0.6876764297485352,
      "learning_rate": 0.00019884257061060897,
      "loss": 0.4853,
      "step": 24890
    },
    {
      "epoch": 0.288287871069328,
      "grad_norm": 0.8229912519454956,
      "learning_rate": 0.00019884063367648172,
      "loss": 0.4282,
      "step": 24900
    },
    {
      "epoch": 0.28840364933080165,
      "grad_norm": 1.0079822540283203,
      "learning_rate": 0.00019883869513244928,
      "loss": 0.4752,
      "step": 24910
    },
    {
      "epoch": 0.2885194275922753,
      "grad_norm": 0.7636349201202393,
      "learning_rate": 0.00019883675497854326,
      "loss": 0.4775,
      "step": 24920
    },
    {
      "epoch": 0.2886352058537489,
      "grad_norm": 0.663273811340332,
      "learning_rate": 0.00019883481321479523,
      "loss": 0.4346,
      "step": 24930
    },
    {
      "epoch": 0.28875098411522254,
      "grad_norm": 0.8784067034721375,
      "learning_rate": 0.00019883286984123682,
      "loss": 0.4824,
      "step": 24940
    },
    {
      "epoch": 0.2888667623766962,
      "grad_norm": 0.7610532641410828,
      "learning_rate": 0.00019883092485789965,
      "loss": 0.437,
      "step": 24950
    },
    {
      "epoch": 0.2889825406381698,
      "grad_norm": 0.6961755752563477,
      "learning_rate": 0.00019882897826481544,
      "loss": 0.4494,
      "step": 24960
    },
    {
      "epoch": 0.2890983188996434,
      "grad_norm": 0.8274893760681152,
      "learning_rate": 0.00019882703006201592,
      "loss": 0.4395,
      "step": 24970
    },
    {
      "epoch": 0.289214097161117,
      "grad_norm": 0.7429479360580444,
      "learning_rate": 0.00019882508024953274,
      "loss": 0.492,
      "step": 24980
    },
    {
      "epoch": 0.28932987542259064,
      "grad_norm": 0.8372470140457153,
      "learning_rate": 0.00019882312882739771,
      "loss": 0.4698,
      "step": 24990
    },
    {
      "epoch": 0.2894456536840643,
      "grad_norm": 0.932794988155365,
      "learning_rate": 0.00019882117579564262,
      "loss": 0.4482,
      "step": 25000
    },
    {
      "epoch": 0.2894456536840643,
      "eval_chrf": 73.6551260485451,
      "eval_loss": 0.6995149254798889,
      "eval_runtime": 348.6979,
      "eval_samples_per_second": 0.287,
      "eval_steps_per_second": 0.011,
      "step": 25000
    },
    {
      "epoch": 0.2895614319455379,
      "grad_norm": 0.7529760599136353,
      "learning_rate": 0.00019881922115429928,
      "loss": 0.4535,
      "step": 25010
    },
    {
      "epoch": 0.28967721020701154,
      "grad_norm": 0.7579430341720581,
      "learning_rate": 0.00019881726490339954,
      "loss": 0.4411,
      "step": 25020
    },
    {
      "epoch": 0.28979298846848517,
      "grad_norm": 0.8080278635025024,
      "learning_rate": 0.00019881530704297522,
      "loss": 0.4636,
      "step": 25030
    },
    {
      "epoch": 0.2899087667299588,
      "grad_norm": 0.7168752551078796,
      "learning_rate": 0.00019881334757305822,
      "loss": 0.4464,
      "step": 25040
    },
    {
      "epoch": 0.29002454499143243,
      "grad_norm": 0.7604632377624512,
      "learning_rate": 0.00019881138649368048,
      "loss": 0.4738,
      "step": 25050
    },
    {
      "epoch": 0.290140323252906,
      "grad_norm": 0.7843213677406311,
      "learning_rate": 0.00019880942380487394,
      "loss": 0.4833,
      "step": 25060
    },
    {
      "epoch": 0.29025610151437964,
      "grad_norm": 0.9687686562538147,
      "learning_rate": 0.00019880745950667054,
      "loss": 0.4678,
      "step": 25070
    },
    {
      "epoch": 0.29037187977585327,
      "grad_norm": 0.8397155404090881,
      "learning_rate": 0.0001988054935991023,
      "loss": 0.4622,
      "step": 25080
    },
    {
      "epoch": 0.2904876580373269,
      "grad_norm": 0.9570215940475464,
      "learning_rate": 0.00019880352608220124,
      "loss": 0.4591,
      "step": 25090
    },
    {
      "epoch": 0.29060343629880053,
      "grad_norm": 0.85113126039505,
      "learning_rate": 0.00019880155695599938,
      "loss": 0.4422,
      "step": 25100
    },
    {
      "epoch": 0.29071921456027416,
      "grad_norm": 0.5425119996070862,
      "learning_rate": 0.00019879958622052882,
      "loss": 0.4483,
      "step": 25110
    },
    {
      "epoch": 0.2908349928217478,
      "grad_norm": 0.8262937664985657,
      "learning_rate": 0.00019879761387582164,
      "loss": 0.4712,
      "step": 25120
    },
    {
      "epoch": 0.2909507710832214,
      "grad_norm": 0.7128187417984009,
      "learning_rate": 0.00019879563992190996,
      "loss": 0.4689,
      "step": 25130
    },
    {
      "epoch": 0.29106654934469506,
      "grad_norm": 0.7753236293792725,
      "learning_rate": 0.00019879366435882601,
      "loss": 0.4646,
      "step": 25140
    },
    {
      "epoch": 0.2911823276061687,
      "grad_norm": 0.8457900285720825,
      "learning_rate": 0.00019879168718660184,
      "loss": 0.4736,
      "step": 25150
    },
    {
      "epoch": 0.29129810586764227,
      "grad_norm": 0.8182496428489685,
      "learning_rate": 0.00019878970840526977,
      "loss": 0.4717,
      "step": 25160
    },
    {
      "epoch": 0.2914138841291159,
      "grad_norm": 0.709197998046875,
      "learning_rate": 0.00019878772801486193,
      "loss": 0.4635,
      "step": 25170
    },
    {
      "epoch": 0.29152966239058953,
      "grad_norm": 0.7016478776931763,
      "learning_rate": 0.00019878574601541065,
      "loss": 0.4442,
      "step": 25180
    },
    {
      "epoch": 0.29164544065206316,
      "grad_norm": 0.7485529780387878,
      "learning_rate": 0.0001987837624069482,
      "loss": 0.4934,
      "step": 25190
    },
    {
      "epoch": 0.2917612189135368,
      "grad_norm": 1.0375511646270752,
      "learning_rate": 0.00019878177718950685,
      "loss": 0.4424,
      "step": 25200
    },
    {
      "epoch": 0.2918769971750104,
      "grad_norm": 0.7467118501663208,
      "learning_rate": 0.000198779790363119,
      "loss": 0.4706,
      "step": 25210
    },
    {
      "epoch": 0.29199277543648405,
      "grad_norm": 0.712745189666748,
      "learning_rate": 0.00019877780192781696,
      "loss": 0.457,
      "step": 25220
    },
    {
      "epoch": 0.2921085536979577,
      "grad_norm": 0.7153512835502625,
      "learning_rate": 0.0001987758118836331,
      "loss": 0.4567,
      "step": 25230
    },
    {
      "epoch": 0.2922243319594313,
      "grad_norm": 0.8477054834365845,
      "learning_rate": 0.0001987738202305999,
      "loss": 0.433,
      "step": 25240
    },
    {
      "epoch": 0.29234011022090495,
      "grad_norm": 1.225630521774292,
      "learning_rate": 0.00019877182696874973,
      "loss": 0.4437,
      "step": 25250
    },
    {
      "epoch": 0.2924558884823785,
      "grad_norm": 0.6832100749015808,
      "learning_rate": 0.0001987698320981151,
      "loss": 0.4771,
      "step": 25260
    },
    {
      "epoch": 0.29257166674385215,
      "grad_norm": 0.7465460896492004,
      "learning_rate": 0.0001987678356187285,
      "loss": 0.4749,
      "step": 25270
    },
    {
      "epoch": 0.2926874450053258,
      "grad_norm": 0.9117196798324585,
      "learning_rate": 0.00019876583753062246,
      "loss": 0.445,
      "step": 25280
    },
    {
      "epoch": 0.2928032232667994,
      "grad_norm": 0.7943907380104065,
      "learning_rate": 0.0001987638378338295,
      "loss": 0.4913,
      "step": 25290
    },
    {
      "epoch": 0.29291900152827305,
      "grad_norm": 0.8349709510803223,
      "learning_rate": 0.00019876183652838217,
      "loss": 0.4657,
      "step": 25300
    },
    {
      "epoch": 0.2930347797897467,
      "grad_norm": 0.7472047209739685,
      "learning_rate": 0.0001987598336143131,
      "loss": 0.4477,
      "step": 25310
    },
    {
      "epoch": 0.2931505580512203,
      "grad_norm": 0.7242029905319214,
      "learning_rate": 0.00019875782909165493,
      "loss": 0.4557,
      "step": 25320
    },
    {
      "epoch": 0.29326633631269394,
      "grad_norm": 0.5495631098747253,
      "learning_rate": 0.00019875582296044028,
      "loss": 0.4794,
      "step": 25330
    },
    {
      "epoch": 0.2933821145741676,
      "grad_norm": 0.7208079695701599,
      "learning_rate": 0.0001987538152207018,
      "loss": 0.4479,
      "step": 25340
    },
    {
      "epoch": 0.2934978928356412,
      "grad_norm": 0.7286500930786133,
      "learning_rate": 0.00019875180587247224,
      "loss": 0.4412,
      "step": 25350
    },
    {
      "epoch": 0.2936136710971148,
      "grad_norm": 0.7493752241134644,
      "learning_rate": 0.0001987497949157843,
      "loss": 0.4625,
      "step": 25360
    },
    {
      "epoch": 0.2937294493585884,
      "grad_norm": 0.7854233980178833,
      "learning_rate": 0.0001987477823506708,
      "loss": 0.4334,
      "step": 25370
    },
    {
      "epoch": 0.29384522762006204,
      "grad_norm": 0.6529062986373901,
      "learning_rate": 0.0001987457681771644,
      "loss": 0.4467,
      "step": 25380
    },
    {
      "epoch": 0.2939610058815357,
      "grad_norm": 0.7695010304450989,
      "learning_rate": 0.000198743752395298,
      "loss": 0.464,
      "step": 25390
    },
    {
      "epoch": 0.2940767841430093,
      "grad_norm": 0.70816969871521,
      "learning_rate": 0.00019874173500510441,
      "loss": 0.4354,
      "step": 25400
    },
    {
      "epoch": 0.29419256240448294,
      "grad_norm": 0.7394952178001404,
      "learning_rate": 0.0001987397160066165,
      "loss": 0.4595,
      "step": 25410
    },
    {
      "epoch": 0.29430834066595657,
      "grad_norm": 0.7608230113983154,
      "learning_rate": 0.0001987376953998671,
      "loss": 0.4819,
      "step": 25420
    },
    {
      "epoch": 0.2944241189274302,
      "grad_norm": 0.8459185361862183,
      "learning_rate": 0.0001987356731848892,
      "loss": 0.4563,
      "step": 25430
    },
    {
      "epoch": 0.29453989718890383,
      "grad_norm": 0.8780574202537537,
      "learning_rate": 0.00019873364936171567,
      "loss": 0.4759,
      "step": 25440
    },
    {
      "epoch": 0.29465567545037746,
      "grad_norm": 0.7876760363578796,
      "learning_rate": 0.0001987316239303795,
      "loss": 0.453,
      "step": 25450
    },
    {
      "epoch": 0.29477145371185104,
      "grad_norm": 0.7067634463310242,
      "learning_rate": 0.0001987295968909137,
      "loss": 0.4401,
      "step": 25460
    },
    {
      "epoch": 0.29488723197332467,
      "grad_norm": 0.9772910475730896,
      "learning_rate": 0.00019872756824335127,
      "loss": 0.4196,
      "step": 25470
    },
    {
      "epoch": 0.2950030102347983,
      "grad_norm": 0.7131596803665161,
      "learning_rate": 0.00019872553798772524,
      "loss": 0.4456,
      "step": 25480
    },
    {
      "epoch": 0.29511878849627193,
      "grad_norm": 0.6839360594749451,
      "learning_rate": 0.0001987235061240687,
      "loss": 0.4659,
      "step": 25490
    },
    {
      "epoch": 0.29523456675774556,
      "grad_norm": 0.7787697911262512,
      "learning_rate": 0.00019872147265241472,
      "loss": 0.4278,
      "step": 25500
    },
    {
      "epoch": 0.2953503450192192,
      "grad_norm": 0.9587351679801941,
      "learning_rate": 0.00019871943757279644,
      "loss": 0.4712,
      "step": 25510
    },
    {
      "epoch": 0.2954661232806928,
      "grad_norm": 0.8071374893188477,
      "learning_rate": 0.000198717400885247,
      "loss": 0.4035,
      "step": 25520
    },
    {
      "epoch": 0.29558190154216646,
      "grad_norm": 0.9236906170845032,
      "learning_rate": 0.0001987153625897996,
      "loss": 0.4902,
      "step": 25530
    },
    {
      "epoch": 0.2956976798036401,
      "grad_norm": 0.7164406776428223,
      "learning_rate": 0.00019871332268648741,
      "loss": 0.4584,
      "step": 25540
    },
    {
      "epoch": 0.2958134580651137,
      "grad_norm": 0.6880311965942383,
      "learning_rate": 0.00019871128117534363,
      "loss": 0.4598,
      "step": 25550
    },
    {
      "epoch": 0.2959292363265873,
      "grad_norm": 0.9487251043319702,
      "learning_rate": 0.00019870923805640156,
      "loss": 0.4638,
      "step": 25560
    },
    {
      "epoch": 0.29604501458806093,
      "grad_norm": 0.7381792068481445,
      "learning_rate": 0.00019870719332969446,
      "loss": 0.4835,
      "step": 25570
    },
    {
      "epoch": 0.29616079284953456,
      "grad_norm": 0.8834322094917297,
      "learning_rate": 0.00019870514699525565,
      "loss": 0.4818,
      "step": 25580
    },
    {
      "epoch": 0.2962765711110082,
      "grad_norm": 1.2315737009048462,
      "learning_rate": 0.00019870309905311846,
      "loss": 0.461,
      "step": 25590
    },
    {
      "epoch": 0.2963923493724818,
      "grad_norm": 0.8836092948913574,
      "learning_rate": 0.00019870104950331618,
      "loss": 0.461,
      "step": 25600
    },
    {
      "epoch": 0.29650812763395545,
      "grad_norm": 0.5545638203620911,
      "learning_rate": 0.00019869899834588228,
      "loss": 0.4154,
      "step": 25610
    },
    {
      "epoch": 0.2966239058954291,
      "grad_norm": 0.8589716553688049,
      "learning_rate": 0.00019869694558085013,
      "loss": 0.4476,
      "step": 25620
    },
    {
      "epoch": 0.2967396841569027,
      "grad_norm": 0.8191109895706177,
      "learning_rate": 0.00019869489120825318,
      "loss": 0.4439,
      "step": 25630
    },
    {
      "epoch": 0.29685546241837635,
      "grad_norm": 0.6284889578819275,
      "learning_rate": 0.00019869283522812485,
      "loss": 0.48,
      "step": 25640
    },
    {
      "epoch": 0.2969712406798499,
      "grad_norm": 0.7687880396842957,
      "learning_rate": 0.00019869077764049872,
      "loss": 0.4544,
      "step": 25650
    },
    {
      "epoch": 0.29708701894132356,
      "grad_norm": 0.6644411087036133,
      "learning_rate": 0.00019868871844540818,
      "loss": 0.4136,
      "step": 25660
    },
    {
      "epoch": 0.2972027972027972,
      "grad_norm": 0.5750458836555481,
      "learning_rate": 0.00019868665764288682,
      "loss": 0.4139,
      "step": 25670
    },
    {
      "epoch": 0.2973185754642708,
      "grad_norm": 0.7164966464042664,
      "learning_rate": 0.00019868459523296827,
      "loss": 0.507,
      "step": 25680
    },
    {
      "epoch": 0.29743435372574445,
      "grad_norm": 0.6462358236312866,
      "learning_rate": 0.00019868253121568604,
      "loss": 0.4783,
      "step": 25690
    },
    {
      "epoch": 0.2975501319872181,
      "grad_norm": 0.7665681838989258,
      "learning_rate": 0.0001986804655910738,
      "loss": 0.4546,
      "step": 25700
    },
    {
      "epoch": 0.2976659102486917,
      "grad_norm": 0.6653458476066589,
      "learning_rate": 0.00019867839835916516,
      "loss": 0.4491,
      "step": 25710
    },
    {
      "epoch": 0.29778168851016534,
      "grad_norm": 0.9308329224586487,
      "learning_rate": 0.00019867632951999382,
      "loss": 0.4757,
      "step": 25720
    },
    {
      "epoch": 0.297897466771639,
      "grad_norm": 0.7271718382835388,
      "learning_rate": 0.00019867425907359345,
      "loss": 0.4553,
      "step": 25730
    },
    {
      "epoch": 0.2980132450331126,
      "grad_norm": 0.8564931154251099,
      "learning_rate": 0.0001986721870199978,
      "loss": 0.4943,
      "step": 25740
    },
    {
      "epoch": 0.2981290232945862,
      "grad_norm": 0.782024621963501,
      "learning_rate": 0.00019867011335924058,
      "loss": 0.4281,
      "step": 25750
    },
    {
      "epoch": 0.2982448015560598,
      "grad_norm": 0.7796012163162231,
      "learning_rate": 0.0001986680380913556,
      "loss": 0.4844,
      "step": 25760
    },
    {
      "epoch": 0.29836057981753344,
      "grad_norm": 0.8314896821975708,
      "learning_rate": 0.00019866596121637667,
      "loss": 0.4414,
      "step": 25770
    },
    {
      "epoch": 0.2984763580790071,
      "grad_norm": 0.8576761484146118,
      "learning_rate": 0.00019866388273433758,
      "loss": 0.4927,
      "step": 25780
    },
    {
      "epoch": 0.2985921363404807,
      "grad_norm": 0.7437127232551575,
      "learning_rate": 0.0001986618026452722,
      "loss": 0.4408,
      "step": 25790
    },
    {
      "epoch": 0.29870791460195434,
      "grad_norm": 0.7944348454475403,
      "learning_rate": 0.00019865972094921443,
      "loss": 0.4267,
      "step": 25800
    },
    {
      "epoch": 0.29882369286342797,
      "grad_norm": 0.6385840177536011,
      "learning_rate": 0.00019865763764619814,
      "loss": 0.4587,
      "step": 25810
    },
    {
      "epoch": 0.2989394711249016,
      "grad_norm": 0.5998877882957458,
      "learning_rate": 0.0001986555527362573,
      "loss": 0.4418,
      "step": 25820
    },
    {
      "epoch": 0.29905524938637523,
      "grad_norm": 0.7375982999801636,
      "learning_rate": 0.00019865346621942586,
      "loss": 0.4457,
      "step": 25830
    },
    {
      "epoch": 0.29917102764784886,
      "grad_norm": 0.7713582515716553,
      "learning_rate": 0.00019865137809573776,
      "loss": 0.478,
      "step": 25840
    },
    {
      "epoch": 0.29928680590932244,
      "grad_norm": 0.7261648178100586,
      "learning_rate": 0.00019864928836522708,
      "loss": 0.4456,
      "step": 25850
    },
    {
      "epoch": 0.29940258417079607,
      "grad_norm": 0.6686517596244812,
      "learning_rate": 0.0001986471970279278,
      "loss": 0.4914,
      "step": 25860
    },
    {
      "epoch": 0.2995183624322697,
      "grad_norm": 0.7416560053825378,
      "learning_rate": 0.000198645104083874,
      "loss": 0.4685,
      "step": 25870
    },
    {
      "epoch": 0.29963414069374333,
      "grad_norm": 0.6594707369804382,
      "learning_rate": 0.00019864300953309982,
      "loss": 0.4418,
      "step": 25880
    },
    {
      "epoch": 0.29974991895521697,
      "grad_norm": 0.7810648083686829,
      "learning_rate": 0.00019864091337563932,
      "loss": 0.4894,
      "step": 25890
    },
    {
      "epoch": 0.2998656972166906,
      "grad_norm": 0.7893812656402588,
      "learning_rate": 0.00019863881561152664,
      "loss": 0.4651,
      "step": 25900
    },
    {
      "epoch": 0.29998147547816423,
      "grad_norm": 0.8130543231964111,
      "learning_rate": 0.000198636716240796,
      "loss": 0.4852,
      "step": 25910
    },
    {
      "epoch": 0.30009725373963786,
      "grad_norm": 0.7054315209388733,
      "learning_rate": 0.00019863461526348152,
      "loss": 0.4326,
      "step": 25920
    },
    {
      "epoch": 0.3002130320011115,
      "grad_norm": 0.9094119071960449,
      "learning_rate": 0.0001986325126796175,
      "loss": 0.4634,
      "step": 25930
    },
    {
      "epoch": 0.3003288102625851,
      "grad_norm": 0.8726437091827393,
      "learning_rate": 0.00019863040848923808,
      "loss": 0.4657,
      "step": 25940
    },
    {
      "epoch": 0.3004445885240587,
      "grad_norm": 0.7491232752799988,
      "learning_rate": 0.00019862830269237766,
      "loss": 0.4307,
      "step": 25950
    },
    {
      "epoch": 0.30056036678553233,
      "grad_norm": 0.6829655766487122,
      "learning_rate": 0.00019862619528907046,
      "loss": 0.4518,
      "step": 25960
    },
    {
      "epoch": 0.30067614504700596,
      "grad_norm": 0.4723912179470062,
      "learning_rate": 0.0001986240862793508,
      "loss": 0.4225,
      "step": 25970
    },
    {
      "epoch": 0.3007919233084796,
      "grad_norm": 0.7675802111625671,
      "learning_rate": 0.00019862197566325308,
      "loss": 0.4748,
      "step": 25980
    },
    {
      "epoch": 0.3009077015699532,
      "grad_norm": 0.8357563018798828,
      "learning_rate": 0.00019861986344081164,
      "loss": 0.4614,
      "step": 25990
    },
    {
      "epoch": 0.30102347983142685,
      "grad_norm": 0.7614332437515259,
      "learning_rate": 0.00019861774961206087,
      "loss": 0.4507,
      "step": 26000
    },
    {
      "epoch": 0.30102347983142685,
      "eval_chrf": 76.55202599732712,
      "eval_loss": 0.6856644153594971,
      "eval_runtime": 349.7237,
      "eval_samples_per_second": 0.286,
      "eval_steps_per_second": 0.011,
      "step": 26000
    },
    {
      "epoch": 0.3011392580929005,
      "grad_norm": 0.9072996377944946,
      "learning_rate": 0.0001986156341770353,
      "loss": 0.4475,
      "step": 26010
    },
    {
      "epoch": 0.3012550363543741,
      "grad_norm": 0.7568226456642151,
      "learning_rate": 0.00019861351713576923,
      "loss": 0.4469,
      "step": 26020
    },
    {
      "epoch": 0.30137081461584775,
      "grad_norm": 0.6582204103469849,
      "learning_rate": 0.00019861139848829725,
      "loss": 0.4668,
      "step": 26030
    },
    {
      "epoch": 0.3014865928773214,
      "grad_norm": 0.8352224826812744,
      "learning_rate": 0.00019860927823465386,
      "loss": 0.4859,
      "step": 26040
    },
    {
      "epoch": 0.30160237113879496,
      "grad_norm": 0.7534937262535095,
      "learning_rate": 0.00019860715637487356,
      "loss": 0.4927,
      "step": 26050
    },
    {
      "epoch": 0.3017181494002686,
      "grad_norm": 0.7749175429344177,
      "learning_rate": 0.00019860503290899093,
      "loss": 0.4535,
      "step": 26060
    },
    {
      "epoch": 0.3018339276617422,
      "grad_norm": 0.8303593993186951,
      "learning_rate": 0.00019860290783704058,
      "loss": 0.486,
      "step": 26070
    },
    {
      "epoch": 0.30194970592321585,
      "grad_norm": 0.7758985757827759,
      "learning_rate": 0.00019860078115905704,
      "loss": 0.4451,
      "step": 26080
    },
    {
      "epoch": 0.3020654841846895,
      "grad_norm": 0.8329030275344849,
      "learning_rate": 0.00019859865287507505,
      "loss": 0.4631,
      "step": 26090
    },
    {
      "epoch": 0.3021812624461631,
      "grad_norm": 0.541795551776886,
      "learning_rate": 0.0001985965229851292,
      "loss": 0.4667,
      "step": 26100
    },
    {
      "epoch": 0.30229704070763674,
      "grad_norm": 0.6967641115188599,
      "learning_rate": 0.0001985943914892542,
      "loss": 0.4491,
      "step": 26110
    },
    {
      "epoch": 0.3024128189691104,
      "grad_norm": 0.813157856464386,
      "learning_rate": 0.00019859225838748481,
      "loss": 0.5039,
      "step": 26120
    },
    {
      "epoch": 0.302528597230584,
      "grad_norm": 0.7288057804107666,
      "learning_rate": 0.00019859012367985575,
      "loss": 0.5061,
      "step": 26130
    },
    {
      "epoch": 0.30264437549205764,
      "grad_norm": 0.8384978175163269,
      "learning_rate": 0.0001985879873664018,
      "loss": 0.4781,
      "step": 26140
    },
    {
      "epoch": 0.3027601537535312,
      "grad_norm": 0.7077749371528625,
      "learning_rate": 0.00019858584944715767,
      "loss": 0.4413,
      "step": 26150
    },
    {
      "epoch": 0.30287593201500485,
      "grad_norm": 0.8025346398353577,
      "learning_rate": 0.0001985837099221583,
      "loss": 0.4518,
      "step": 26160
    },
    {
      "epoch": 0.3029917102764785,
      "grad_norm": 0.8154165148735046,
      "learning_rate": 0.00019858156879143848,
      "loss": 0.4544,
      "step": 26170
    },
    {
      "epoch": 0.3031074885379521,
      "grad_norm": 0.7080431580543518,
      "learning_rate": 0.0001985794260550331,
      "loss": 0.4692,
      "step": 26180
    },
    {
      "epoch": 0.30322326679942574,
      "grad_norm": 0.7371339201927185,
      "learning_rate": 0.00019857728171297705,
      "loss": 0.4654,
      "step": 26190
    },
    {
      "epoch": 0.30333904506089937,
      "grad_norm": 0.7605065703392029,
      "learning_rate": 0.00019857513576530524,
      "loss": 0.4424,
      "step": 26200
    },
    {
      "epoch": 0.303454823322373,
      "grad_norm": 0.8257567286491394,
      "learning_rate": 0.00019857298821205266,
      "loss": 0.4726,
      "step": 26210
    },
    {
      "epoch": 0.30357060158384663,
      "grad_norm": 0.827387809753418,
      "learning_rate": 0.00019857083905325428,
      "loss": 0.4483,
      "step": 26220
    },
    {
      "epoch": 0.30368637984532026,
      "grad_norm": 0.6968565583229065,
      "learning_rate": 0.0001985686882889451,
      "loss": 0.4263,
      "step": 26230
    },
    {
      "epoch": 0.3038021581067939,
      "grad_norm": 0.7551820874214172,
      "learning_rate": 0.00019856653591916013,
      "loss": 0.4492,
      "step": 26240
    },
    {
      "epoch": 0.30391793636826747,
      "grad_norm": 0.7242693305015564,
      "learning_rate": 0.0001985643819439345,
      "loss": 0.4898,
      "step": 26250
    },
    {
      "epoch": 0.3040337146297411,
      "grad_norm": 0.6734589338302612,
      "learning_rate": 0.00019856222636330318,
      "loss": 0.4778,
      "step": 26260
    },
    {
      "epoch": 0.30414949289121473,
      "grad_norm": 0.9003152847290039,
      "learning_rate": 0.00019856006917730136,
      "loss": 0.4381,
      "step": 26270
    },
    {
      "epoch": 0.30426527115268837,
      "grad_norm": 0.7741272449493408,
      "learning_rate": 0.00019855791038596418,
      "loss": 0.4827,
      "step": 26280
    },
    {
      "epoch": 0.304381049414162,
      "grad_norm": 0.5133253931999207,
      "learning_rate": 0.00019855574998932675,
      "loss": 0.416,
      "step": 26290
    },
    {
      "epoch": 0.30449682767563563,
      "grad_norm": 0.8481763005256653,
      "learning_rate": 0.0001985535879874243,
      "loss": 0.4661,
      "step": 26300
    },
    {
      "epoch": 0.30461260593710926,
      "grad_norm": 0.671328604221344,
      "learning_rate": 0.00019855142438029205,
      "loss": 0.477,
      "step": 26310
    },
    {
      "epoch": 0.3047283841985829,
      "grad_norm": 0.6453095078468323,
      "learning_rate": 0.00019854925916796518,
      "loss": 0.4579,
      "step": 26320
    },
    {
      "epoch": 0.3048441624600565,
      "grad_norm": 1.0940552949905396,
      "learning_rate": 0.00019854709235047902,
      "loss": 0.4694,
      "step": 26330
    },
    {
      "epoch": 0.3049599407215301,
      "grad_norm": 0.7535539269447327,
      "learning_rate": 0.00019854492392786885,
      "loss": 0.4686,
      "step": 26340
    },
    {
      "epoch": 0.30507571898300373,
      "grad_norm": 0.8653777837753296,
      "learning_rate": 0.00019854275390016998,
      "loss": 0.4336,
      "step": 26350
    },
    {
      "epoch": 0.30519149724447736,
      "grad_norm": 0.6675240397453308,
      "learning_rate": 0.00019854058226741777,
      "loss": 0.4464,
      "step": 26360
    },
    {
      "epoch": 0.305307275505951,
      "grad_norm": 0.7508164048194885,
      "learning_rate": 0.00019853840902964756,
      "loss": 0.4171,
      "step": 26370
    },
    {
      "epoch": 0.3054230537674246,
      "grad_norm": 0.7988144755363464,
      "learning_rate": 0.00019853623418689478,
      "loss": 0.4583,
      "step": 26380
    },
    {
      "epoch": 0.30553883202889826,
      "grad_norm": 0.8029592633247375,
      "learning_rate": 0.0001985340577391948,
      "loss": 0.4374,
      "step": 26390
    },
    {
      "epoch": 0.3056546102903719,
      "grad_norm": 0.8196548819541931,
      "learning_rate": 0.00019853187968658315,
      "loss": 0.4535,
      "step": 26400
    },
    {
      "epoch": 0.3057703885518455,
      "grad_norm": 0.708761990070343,
      "learning_rate": 0.00019852970002909526,
      "loss": 0.448,
      "step": 26410
    },
    {
      "epoch": 0.30588616681331915,
      "grad_norm": 0.6414585709571838,
      "learning_rate": 0.0001985275187667666,
      "loss": 0.4568,
      "step": 26420
    },
    {
      "epoch": 0.3060019450747928,
      "grad_norm": 0.7461490035057068,
      "learning_rate": 0.00019852533589963278,
      "loss": 0.4502,
      "step": 26430
    },
    {
      "epoch": 0.30611772333626636,
      "grad_norm": 0.7462594509124756,
      "learning_rate": 0.0001985231514277293,
      "loss": 0.46,
      "step": 26440
    },
    {
      "epoch": 0.30623350159774,
      "grad_norm": 0.7091456055641174,
      "learning_rate": 0.00019852096535109172,
      "loss": 0.4569,
      "step": 26450
    },
    {
      "epoch": 0.3063492798592136,
      "grad_norm": 0.7239569425582886,
      "learning_rate": 0.0001985187776697557,
      "loss": 0.4467,
      "step": 26460
    },
    {
      "epoch": 0.30646505812068725,
      "grad_norm": 0.7347016930580139,
      "learning_rate": 0.00019851658838375685,
      "loss": 0.4489,
      "step": 26470
    },
    {
      "epoch": 0.3065808363821609,
      "grad_norm": 0.8039568662643433,
      "learning_rate": 0.0001985143974931308,
      "loss": 0.4118,
      "step": 26480
    },
    {
      "epoch": 0.3066966146436345,
      "grad_norm": 0.7843155264854431,
      "learning_rate": 0.0001985122049979133,
      "loss": 0.4426,
      "step": 26490
    },
    {
      "epoch": 0.30681239290510814,
      "grad_norm": 0.5724373459815979,
      "learning_rate": 0.00019851001089814,
      "loss": 0.4835,
      "step": 26500
    },
    {
      "epoch": 0.3069281711665818,
      "grad_norm": 0.8228092193603516,
      "learning_rate": 0.0001985078151938467,
      "loss": 0.4574,
      "step": 26510
    },
    {
      "epoch": 0.3070439494280554,
      "grad_norm": 0.6921157240867615,
      "learning_rate": 0.0001985056178850691,
      "loss": 0.4658,
      "step": 26520
    },
    {
      "epoch": 0.30715972768952904,
      "grad_norm": 0.8331666588783264,
      "learning_rate": 0.000198503418971843,
      "loss": 0.467,
      "step": 26530
    },
    {
      "epoch": 0.3072755059510026,
      "grad_norm": 0.8443690538406372,
      "learning_rate": 0.00019850121845420423,
      "loss": 0.4681,
      "step": 26540
    },
    {
      "epoch": 0.30739128421247625,
      "grad_norm": 0.7651891112327576,
      "learning_rate": 0.00019849901633218862,
      "loss": 0.4576,
      "step": 26550
    },
    {
      "epoch": 0.3075070624739499,
      "grad_norm": 0.7527976632118225,
      "learning_rate": 0.0001984968126058321,
      "loss": 0.4611,
      "step": 26560
    },
    {
      "epoch": 0.3076228407354235,
      "grad_norm": 0.7612469792366028,
      "learning_rate": 0.00019849460727517046,
      "loss": 0.4512,
      "step": 26570
    },
    {
      "epoch": 0.30773861899689714,
      "grad_norm": 0.8089647889137268,
      "learning_rate": 0.0001984924003402397,
      "loss": 0.4182,
      "step": 26580
    },
    {
      "epoch": 0.30785439725837077,
      "grad_norm": 0.7524958848953247,
      "learning_rate": 0.00019849019180107573,
      "loss": 0.4364,
      "step": 26590
    },
    {
      "epoch": 0.3079701755198444,
      "grad_norm": 0.7265125513076782,
      "learning_rate": 0.00019848798165771455,
      "loss": 0.4515,
      "step": 26600
    },
    {
      "epoch": 0.30808595378131803,
      "grad_norm": 0.8231269121170044,
      "learning_rate": 0.0001984857699101921,
      "loss": 0.4867,
      "step": 26610
    },
    {
      "epoch": 0.30820173204279167,
      "grad_norm": 0.829472005367279,
      "learning_rate": 0.00019848355655854448,
      "loss": 0.4419,
      "step": 26620
    },
    {
      "epoch": 0.3083175103042653,
      "grad_norm": 0.7649688124656677,
      "learning_rate": 0.0001984813416028077,
      "loss": 0.4686,
      "step": 26630
    },
    {
      "epoch": 0.3084332885657389,
      "grad_norm": 0.5475059151649475,
      "learning_rate": 0.00019847912504301785,
      "loss": 0.4663,
      "step": 26640
    },
    {
      "epoch": 0.3085490668272125,
      "grad_norm": 0.7118447422981262,
      "learning_rate": 0.00019847690687921102,
      "loss": 0.4534,
      "step": 26650
    },
    {
      "epoch": 0.30866484508868614,
      "grad_norm": 0.7882592678070068,
      "learning_rate": 0.0001984746871114233,
      "loss": 0.441,
      "step": 26660
    },
    {
      "epoch": 0.30878062335015977,
      "grad_norm": 0.6085837483406067,
      "learning_rate": 0.00019847246573969095,
      "loss": 0.4329,
      "step": 26670
    },
    {
      "epoch": 0.3088964016116334,
      "grad_norm": 0.9138346910476685,
      "learning_rate": 0.0001984702427640501,
      "loss": 0.4463,
      "step": 26680
    },
    {
      "epoch": 0.30901217987310703,
      "grad_norm": 0.8461740612983704,
      "learning_rate": 0.0001984680181845369,
      "loss": 0.4341,
      "step": 26690
    },
    {
      "epoch": 0.30912795813458066,
      "grad_norm": 0.7021976113319397,
      "learning_rate": 0.0001984657920011877,
      "loss": 0.4718,
      "step": 26700
    },
    {
      "epoch": 0.3092437363960543,
      "grad_norm": 0.7682430148124695,
      "learning_rate": 0.00019846356421403862,
      "loss": 0.4582,
      "step": 26710
    },
    {
      "epoch": 0.3093595146575279,
      "grad_norm": 0.7685060501098633,
      "learning_rate": 0.00019846133482312605,
      "loss": 0.4395,
      "step": 26720
    },
    {
      "epoch": 0.30947529291900155,
      "grad_norm": 0.7221096158027649,
      "learning_rate": 0.0001984591038284863,
      "loss": 0.464,
      "step": 26730
    },
    {
      "epoch": 0.30959107118047513,
      "grad_norm": 0.67400062084198,
      "learning_rate": 0.00019845687123015563,
      "loss": 0.4644,
      "step": 26740
    },
    {
      "epoch": 0.30970684944194876,
      "grad_norm": 0.596010148525238,
      "learning_rate": 0.00019845463702817048,
      "loss": 0.4319,
      "step": 26750
    },
    {
      "epoch": 0.3098226277034224,
      "grad_norm": 0.8571289777755737,
      "learning_rate": 0.00019845240122256722,
      "loss": 0.4568,
      "step": 26760
    },
    {
      "epoch": 0.309938405964896,
      "grad_norm": 0.7959387302398682,
      "learning_rate": 0.00019845016381338224,
      "loss": 0.4596,
      "step": 26770
    },
    {
      "epoch": 0.31005418422636966,
      "grad_norm": 0.7275651693344116,
      "learning_rate": 0.00019844792480065203,
      "loss": 0.4402,
      "step": 26780
    },
    {
      "epoch": 0.3101699624878433,
      "grad_norm": 0.7132201194763184,
      "learning_rate": 0.000198445684184413,
      "loss": 0.4566,
      "step": 26790
    },
    {
      "epoch": 0.3102857407493169,
      "grad_norm": 0.6844136714935303,
      "learning_rate": 0.0001984434419647017,
      "loss": 0.4523,
      "step": 26800
    },
    {
      "epoch": 0.31040151901079055,
      "grad_norm": 0.9616363048553467,
      "learning_rate": 0.00019844119814155464,
      "loss": 0.4472,
      "step": 26810
    },
    {
      "epoch": 0.3105172972722642,
      "grad_norm": 0.6819164752960205,
      "learning_rate": 0.0001984389527150083,
      "loss": 0.46,
      "step": 26820
    },
    {
      "epoch": 0.3106330755337378,
      "grad_norm": 0.7731338739395142,
      "learning_rate": 0.00019843670568509935,
      "loss": 0.4441,
      "step": 26830
    },
    {
      "epoch": 0.3107488537952114,
      "grad_norm": 0.6755056381225586,
      "learning_rate": 0.00019843445705186436,
      "loss": 0.4542,
      "step": 26840
    },
    {
      "epoch": 0.310864632056685,
      "grad_norm": 0.8113489151000977,
      "learning_rate": 0.00019843220681533993,
      "loss": 0.48,
      "step": 26850
    },
    {
      "epoch": 0.31098041031815865,
      "grad_norm": 0.8629299998283386,
      "learning_rate": 0.00019842995497556273,
      "loss": 0.4458,
      "step": 26860
    },
    {
      "epoch": 0.3110961885796323,
      "grad_norm": 0.7523194551467896,
      "learning_rate": 0.0001984277015325694,
      "loss": 0.4346,
      "step": 26870
    },
    {
      "epoch": 0.3112119668411059,
      "grad_norm": 0.647796630859375,
      "learning_rate": 0.00019842544648639675,
      "loss": 0.4773,
      "step": 26880
    },
    {
      "epoch": 0.31132774510257955,
      "grad_norm": 0.8183492422103882,
      "learning_rate": 0.00019842318983708136,
      "loss": 0.457,
      "step": 26890
    },
    {
      "epoch": 0.3114435233640532,
      "grad_norm": 0.7981014847755432,
      "learning_rate": 0.0001984209315846601,
      "loss": 0.4778,
      "step": 26900
    },
    {
      "epoch": 0.3115593016255268,
      "grad_norm": 0.5752872824668884,
      "learning_rate": 0.0001984186717291697,
      "loss": 0.4416,
      "step": 26910
    },
    {
      "epoch": 0.31167507988700044,
      "grad_norm": 0.8017776012420654,
      "learning_rate": 0.000198416410270647,
      "loss": 0.4837,
      "step": 26920
    },
    {
      "epoch": 0.311790858148474,
      "grad_norm": 0.7454376816749573,
      "learning_rate": 0.0001984141472091288,
      "loss": 0.443,
      "step": 26930
    },
    {
      "epoch": 0.31190663640994765,
      "grad_norm": 0.7297793626785278,
      "learning_rate": 0.000198411882544652,
      "loss": 0.4415,
      "step": 26940
    },
    {
      "epoch": 0.3120224146714213,
      "grad_norm": 0.7995197772979736,
      "learning_rate": 0.00019840961627725344,
      "loss": 0.4308,
      "step": 26950
    },
    {
      "epoch": 0.3121381929328949,
      "grad_norm": 0.698663055896759,
      "learning_rate": 0.00019840734840697008,
      "loss": 0.4123,
      "step": 26960
    },
    {
      "epoch": 0.31225397119436854,
      "grad_norm": 0.7404296398162842,
      "learning_rate": 0.00019840507893383882,
      "loss": 0.4399,
      "step": 26970
    },
    {
      "epoch": 0.31236974945584217,
      "grad_norm": 0.8033159375190735,
      "learning_rate": 0.00019840280785789666,
      "loss": 0.4575,
      "step": 26980
    },
    {
      "epoch": 0.3124855277173158,
      "grad_norm": 0.7322561740875244,
      "learning_rate": 0.00019840053517918054,
      "loss": 0.4495,
      "step": 26990
    },
    {
      "epoch": 0.31260130597878943,
      "grad_norm": 0.7145832180976868,
      "learning_rate": 0.00019839826089772751,
      "loss": 0.4385,
      "step": 27000
    },
    {
      "epoch": 0.31260130597878943,
      "eval_chrf": 81.7699740110989,
      "eval_loss": 0.693875253200531,
      "eval_runtime": 218.5509,
      "eval_samples_per_second": 0.458,
      "eval_steps_per_second": 0.018,
      "step": 27000
    },
    {
      "epoch": 0.31271708424026307,
      "grad_norm": 0.5868543982505798,
      "learning_rate": 0.00019839598501357464,
      "loss": 0.4697,
      "step": 27010
    },
    {
      "epoch": 0.3128328625017367,
      "grad_norm": 1.0674078464508057,
      "learning_rate": 0.00019839370752675896,
      "loss": 0.4471,
      "step": 27020
    },
    {
      "epoch": 0.3129486407632103,
      "grad_norm": 0.832271158695221,
      "learning_rate": 0.00019839142843731757,
      "loss": 0.4725,
      "step": 27030
    },
    {
      "epoch": 0.3130644190246839,
      "grad_norm": 0.8049610257148743,
      "learning_rate": 0.00019838914774528762,
      "loss": 0.4375,
      "step": 27040
    },
    {
      "epoch": 0.31318019728615754,
      "grad_norm": 0.720533013343811,
      "learning_rate": 0.0001983868654507062,
      "loss": 0.4732,
      "step": 27050
    },
    {
      "epoch": 0.31329597554763117,
      "grad_norm": 0.7142242789268494,
      "learning_rate": 0.00019838458155361054,
      "loss": 0.4555,
      "step": 27060
    },
    {
      "epoch": 0.3134117538091048,
      "grad_norm": 0.7077202796936035,
      "learning_rate": 0.00019838229605403785,
      "loss": 0.482,
      "step": 27070
    },
    {
      "epoch": 0.31352753207057843,
      "grad_norm": 0.7336899638175964,
      "learning_rate": 0.0001983800089520253,
      "loss": 0.4545,
      "step": 27080
    },
    {
      "epoch": 0.31364331033205206,
      "grad_norm": 0.8226677775382996,
      "learning_rate": 0.00019837772024761016,
      "loss": 0.4723,
      "step": 27090
    },
    {
      "epoch": 0.3137590885935257,
      "grad_norm": 0.6755543947219849,
      "learning_rate": 0.00019837542994082972,
      "loss": 0.4627,
      "step": 27100
    },
    {
      "epoch": 0.3138748668549993,
      "grad_norm": 0.6180155873298645,
      "learning_rate": 0.00019837313803172129,
      "loss": 0.4595,
      "step": 27110
    },
    {
      "epoch": 0.31399064511647296,
      "grad_norm": 0.8046608567237854,
      "learning_rate": 0.00019837084452032218,
      "loss": 0.4628,
      "step": 27120
    },
    {
      "epoch": 0.31410642337794653,
      "grad_norm": 0.8131647109985352,
      "learning_rate": 0.00019836854940666974,
      "loss": 0.4545,
      "step": 27130
    },
    {
      "epoch": 0.31422220163942016,
      "grad_norm": 0.7022969722747803,
      "learning_rate": 0.0001983662526908014,
      "loss": 0.5048,
      "step": 27140
    },
    {
      "epoch": 0.3143379799008938,
      "grad_norm": 0.779664158821106,
      "learning_rate": 0.00019836395437275456,
      "loss": 0.4742,
      "step": 27150
    },
    {
      "epoch": 0.3144537581623674,
      "grad_norm": 0.9005945324897766,
      "learning_rate": 0.00019836165445256658,
      "loss": 0.4638,
      "step": 27160
    },
    {
      "epoch": 0.31456953642384106,
      "grad_norm": 0.6903424859046936,
      "learning_rate": 0.00019835935293027501,
      "loss": 0.4421,
      "step": 27170
    },
    {
      "epoch": 0.3146853146853147,
      "grad_norm": 0.703633189201355,
      "learning_rate": 0.00019835704980591728,
      "loss": 0.432,
      "step": 27180
    },
    {
      "epoch": 0.3148010929467883,
      "grad_norm": 0.819511353969574,
      "learning_rate": 0.00019835474507953093,
      "loss": 0.5039,
      "step": 27190
    },
    {
      "epoch": 0.31491687120826195,
      "grad_norm": 0.8665363788604736,
      "learning_rate": 0.0001983524387511535,
      "loss": 0.474,
      "step": 27200
    },
    {
      "epoch": 0.3150326494697356,
      "grad_norm": 0.7022074460983276,
      "learning_rate": 0.00019835013082082251,
      "loss": 0.4564,
      "step": 27210
    },
    {
      "epoch": 0.3151484277312092,
      "grad_norm": 0.6639718413352966,
      "learning_rate": 0.00019834782128857564,
      "loss": 0.4455,
      "step": 27220
    },
    {
      "epoch": 0.3152642059926828,
      "grad_norm": 0.7446359395980835,
      "learning_rate": 0.00019834551015445048,
      "loss": 0.4513,
      "step": 27230
    },
    {
      "epoch": 0.3153799842541564,
      "grad_norm": 0.7960302233695984,
      "learning_rate": 0.00019834319741848458,
      "loss": 0.4572,
      "step": 27240
    },
    {
      "epoch": 0.31549576251563005,
      "grad_norm": 0.7684571743011475,
      "learning_rate": 0.00019834088308071573,
      "loss": 0.4554,
      "step": 27250
    },
    {
      "epoch": 0.3156115407771037,
      "grad_norm": 0.7665683627128601,
      "learning_rate": 0.00019833856714118154,
      "loss": 0.4246,
      "step": 27260
    },
    {
      "epoch": 0.3157273190385773,
      "grad_norm": 0.7351424694061279,
      "learning_rate": 0.0001983362495999198,
      "loss": 0.4202,
      "step": 27270
    },
    {
      "epoch": 0.31584309730005095,
      "grad_norm": 0.8117152452468872,
      "learning_rate": 0.0001983339304569682,
      "loss": 0.4568,
      "step": 27280
    },
    {
      "epoch": 0.3159588755615246,
      "grad_norm": 0.709430456161499,
      "learning_rate": 0.00019833160971236455,
      "loss": 0.4897,
      "step": 27290
    },
    {
      "epoch": 0.3160746538229982,
      "grad_norm": 0.7107734680175781,
      "learning_rate": 0.00019832928736614663,
      "loss": 0.468,
      "step": 27300
    },
    {
      "epoch": 0.31619043208447184,
      "grad_norm": 0.6033035516738892,
      "learning_rate": 0.0001983269634183523,
      "loss": 0.4446,
      "step": 27310
    },
    {
      "epoch": 0.31630621034594547,
      "grad_norm": 0.7143062949180603,
      "learning_rate": 0.00019832463786901937,
      "loss": 0.443,
      "step": 27320
    },
    {
      "epoch": 0.31642198860741905,
      "grad_norm": 0.8292445540428162,
      "learning_rate": 0.00019832231071818574,
      "loss": 0.4701,
      "step": 27330
    },
    {
      "epoch": 0.3165377668688927,
      "grad_norm": 0.6561749577522278,
      "learning_rate": 0.0001983199819658893,
      "loss": 0.4481,
      "step": 27340
    },
    {
      "epoch": 0.3166535451303663,
      "grad_norm": 0.8589171171188354,
      "learning_rate": 0.000198317651612168,
      "loss": 0.4562,
      "step": 27350
    },
    {
      "epoch": 0.31676932339183994,
      "grad_norm": 0.5710141658782959,
      "learning_rate": 0.0001983153196570598,
      "loss": 0.4779,
      "step": 27360
    },
    {
      "epoch": 0.3168851016533136,
      "grad_norm": 0.8408932685852051,
      "learning_rate": 0.00019831298610060264,
      "loss": 0.4754,
      "step": 27370
    },
    {
      "epoch": 0.3170008799147872,
      "grad_norm": 0.8772223591804504,
      "learning_rate": 0.0001983106509428346,
      "loss": 0.4604,
      "step": 27380
    },
    {
      "epoch": 0.31711665817626083,
      "grad_norm": 0.7065361738204956,
      "learning_rate": 0.00019830831418379363,
      "loss": 0.4164,
      "step": 27390
    },
    {
      "epoch": 0.31723243643773447,
      "grad_norm": 0.7624000310897827,
      "learning_rate": 0.00019830597582351788,
      "loss": 0.4288,
      "step": 27400
    },
    {
      "epoch": 0.3173482146992081,
      "grad_norm": 0.7695420384407043,
      "learning_rate": 0.00019830363586204535,
      "loss": 0.4787,
      "step": 27410
    },
    {
      "epoch": 0.31746399296068173,
      "grad_norm": 0.7112919092178345,
      "learning_rate": 0.0001983012942994142,
      "loss": 0.439,
      "step": 27420
    },
    {
      "epoch": 0.3175797712221553,
      "grad_norm": 0.9614130854606628,
      "learning_rate": 0.0001982989511356626,
      "loss": 0.4717,
      "step": 27430
    },
    {
      "epoch": 0.31769554948362894,
      "grad_norm": 0.8063520193099976,
      "learning_rate": 0.00019829660637082864,
      "loss": 0.4551,
      "step": 27440
    },
    {
      "epoch": 0.31781132774510257,
      "grad_norm": 0.7403252720832825,
      "learning_rate": 0.00019829426000495055,
      "loss": 0.4603,
      "step": 27450
    },
    {
      "epoch": 0.3179271060065762,
      "grad_norm": 0.5429824590682983,
      "learning_rate": 0.00019829191203806656,
      "loss": 0.4419,
      "step": 27460
    },
    {
      "epoch": 0.31804288426804983,
      "grad_norm": 0.654305100440979,
      "learning_rate": 0.00019828956247021492,
      "loss": 0.4816,
      "step": 27470
    },
    {
      "epoch": 0.31815866252952346,
      "grad_norm": 0.5370567440986633,
      "learning_rate": 0.00019828721130143385,
      "loss": 0.4399,
      "step": 27480
    },
    {
      "epoch": 0.3182744407909971,
      "grad_norm": 0.8804116249084473,
      "learning_rate": 0.00019828485853176167,
      "loss": 0.4632,
      "step": 27490
    },
    {
      "epoch": 0.3183902190524707,
      "grad_norm": 0.8157013654708862,
      "learning_rate": 0.00019828250416123675,
      "loss": 0.4136,
      "step": 27500
    },
    {
      "epoch": 0.31850599731394436,
      "grad_norm": 0.7248451709747314,
      "learning_rate": 0.00019828014818989734,
      "loss": 0.449,
      "step": 27510
    },
    {
      "epoch": 0.318621775575418,
      "grad_norm": 0.7211676239967346,
      "learning_rate": 0.00019827779061778192,
      "loss": 0.4805,
      "step": 27520
    },
    {
      "epoch": 0.31873755383689156,
      "grad_norm": 0.7188708782196045,
      "learning_rate": 0.00019827543144492878,
      "loss": 0.4383,
      "step": 27530
    },
    {
      "epoch": 0.3188533320983652,
      "grad_norm": 0.8172465562820435,
      "learning_rate": 0.00019827307067137647,
      "loss": 0.4597,
      "step": 27540
    },
    {
      "epoch": 0.3189691103598388,
      "grad_norm": 0.8475064039230347,
      "learning_rate": 0.00019827070829716336,
      "loss": 0.456,
      "step": 27550
    },
    {
      "epoch": 0.31908488862131246,
      "grad_norm": 0.7453060150146484,
      "learning_rate": 0.00019826834432232793,
      "loss": 0.442,
      "step": 27560
    },
    {
      "epoch": 0.3192006668827861,
      "grad_norm": 0.8786172270774841,
      "learning_rate": 0.00019826597874690867,
      "loss": 0.4642,
      "step": 27570
    },
    {
      "epoch": 0.3193164451442597,
      "grad_norm": 0.739477813243866,
      "learning_rate": 0.0001982636115709442,
      "loss": 0.4445,
      "step": 27580
    },
    {
      "epoch": 0.31943222340573335,
      "grad_norm": 0.5354799032211304,
      "learning_rate": 0.000198261242794473,
      "loss": 0.4527,
      "step": 27590
    },
    {
      "epoch": 0.319548001667207,
      "grad_norm": 0.7867555618286133,
      "learning_rate": 0.00019825887241753365,
      "loss": 0.4462,
      "step": 27600
    },
    {
      "epoch": 0.3196637799286806,
      "grad_norm": 0.90069979429245,
      "learning_rate": 0.00019825650044016477,
      "loss": 0.4519,
      "step": 27610
    },
    {
      "epoch": 0.3197795581901542,
      "grad_norm": 0.8946885466575623,
      "learning_rate": 0.000198254126862405,
      "loss": 0.4782,
      "step": 27620
    },
    {
      "epoch": 0.3198953364516278,
      "grad_norm": 0.8960312604904175,
      "learning_rate": 0.00019825175168429305,
      "loss": 0.4645,
      "step": 27630
    },
    {
      "epoch": 0.32001111471310145,
      "grad_norm": 0.7747102379798889,
      "learning_rate": 0.0001982493749058675,
      "loss": 0.411,
      "step": 27640
    },
    {
      "epoch": 0.3201268929745751,
      "grad_norm": 0.7728378772735596,
      "learning_rate": 0.00019824699652716714,
      "loss": 0.44,
      "step": 27650
    },
    {
      "epoch": 0.3202426712360487,
      "grad_norm": 0.7578223347663879,
      "learning_rate": 0.0001982446165482307,
      "loss": 0.4576,
      "step": 27660
    },
    {
      "epoch": 0.32035844949752235,
      "grad_norm": 0.71628338098526,
      "learning_rate": 0.00019824223496909692,
      "loss": 0.4697,
      "step": 27670
    },
    {
      "epoch": 0.320474227758996,
      "grad_norm": 0.7491822838783264,
      "learning_rate": 0.0001982398517898046,
      "loss": 0.4641,
      "step": 27680
    },
    {
      "epoch": 0.3205900060204696,
      "grad_norm": 0.8317740559577942,
      "learning_rate": 0.00019823746701039258,
      "loss": 0.4566,
      "step": 27690
    },
    {
      "epoch": 0.32070578428194324,
      "grad_norm": 0.8668158650398254,
      "learning_rate": 0.0001982350806308997,
      "loss": 0.4647,
      "step": 27700
    },
    {
      "epoch": 0.32082156254341687,
      "grad_norm": 0.8157652616500854,
      "learning_rate": 0.00019823269265136477,
      "loss": 0.4393,
      "step": 27710
    },
    {
      "epoch": 0.32093734080489045,
      "grad_norm": 0.7637067437171936,
      "learning_rate": 0.00019823030307182678,
      "loss": 0.4794,
      "step": 27720
    },
    {
      "epoch": 0.3210531190663641,
      "grad_norm": 0.6508481502532959,
      "learning_rate": 0.00019822791189232457,
      "loss": 0.434,
      "step": 27730
    },
    {
      "epoch": 0.3211688973278377,
      "grad_norm": 0.7629069089889526,
      "learning_rate": 0.00019822551911289713,
      "loss": 0.4675,
      "step": 27740
    },
    {
      "epoch": 0.32128467558931134,
      "grad_norm": 0.7749221324920654,
      "learning_rate": 0.0001982231247335834,
      "loss": 0.462,
      "step": 27750
    },
    {
      "epoch": 0.321400453850785,
      "grad_norm": 0.8904512524604797,
      "learning_rate": 0.0001982207287544224,
      "loss": 0.4898,
      "step": 27760
    },
    {
      "epoch": 0.3215162321122586,
      "grad_norm": 0.7557669281959534,
      "learning_rate": 0.0001982183311754532,
      "loss": 0.4494,
      "step": 27770
    },
    {
      "epoch": 0.32163201037373224,
      "grad_norm": 0.8687517642974854,
      "learning_rate": 0.0001982159319967148,
      "loss": 0.442,
      "step": 27780
    },
    {
      "epoch": 0.32174778863520587,
      "grad_norm": 0.6912224888801575,
      "learning_rate": 0.0001982135312182462,
      "loss": 0.4746,
      "step": 27790
    },
    {
      "epoch": 0.3218635668966795,
      "grad_norm": 0.7170769572257996,
      "learning_rate": 0.0001982111288400867,
      "loss": 0.4567,
      "step": 27800
    },
    {
      "epoch": 0.32197934515815313,
      "grad_norm": 0.7778599262237549,
      "learning_rate": 0.00019820872486227525,
      "loss": 0.4493,
      "step": 27810
    },
    {
      "epoch": 0.3220951234196267,
      "grad_norm": 0.8060257434844971,
      "learning_rate": 0.0001982063192848511,
      "loss": 0.439,
      "step": 27820
    },
    {
      "epoch": 0.32221090168110034,
      "grad_norm": 0.8419181704521179,
      "learning_rate": 0.0001982039121078534,
      "loss": 0.438,
      "step": 27830
    },
    {
      "epoch": 0.32232667994257397,
      "grad_norm": 1.0731698274612427,
      "learning_rate": 0.00019820150333132137,
      "loss": 0.4705,
      "step": 27840
    },
    {
      "epoch": 0.3224424582040476,
      "grad_norm": 0.7142056822776794,
      "learning_rate": 0.0001981990929552942,
      "loss": 0.4596,
      "step": 27850
    },
    {
      "epoch": 0.32255823646552123,
      "grad_norm": 0.690332293510437,
      "learning_rate": 0.00019819668097981123,
      "loss": 0.461,
      "step": 27860
    },
    {
      "epoch": 0.32267401472699486,
      "grad_norm": 0.7570613622665405,
      "learning_rate": 0.00019819426740491172,
      "loss": 0.4323,
      "step": 27870
    },
    {
      "epoch": 0.3227897929884685,
      "grad_norm": 0.7912685871124268,
      "learning_rate": 0.00019819185223063493,
      "loss": 0.4661,
      "step": 27880
    },
    {
      "epoch": 0.3229055712499421,
      "grad_norm": 0.6318713426589966,
      "learning_rate": 0.00019818943545702022,
      "loss": 0.4537,
      "step": 27890
    },
    {
      "epoch": 0.32302134951141576,
      "grad_norm": 0.7996380925178528,
      "learning_rate": 0.000198187017084107,
      "loss": 0.4708,
      "step": 27900
    },
    {
      "epoch": 0.3231371277728894,
      "grad_norm": 0.7950218319892883,
      "learning_rate": 0.00019818459711193462,
      "loss": 0.4341,
      "step": 27910
    },
    {
      "epoch": 0.32325290603436296,
      "grad_norm": 0.8847008347511292,
      "learning_rate": 0.0001981821755405425,
      "loss": 0.4937,
      "step": 27920
    },
    {
      "epoch": 0.3233686842958366,
      "grad_norm": 0.5279824733734131,
      "learning_rate": 0.00019817975236997008,
      "loss": 0.4492,
      "step": 27930
    },
    {
      "epoch": 0.3234844625573102,
      "grad_norm": 0.7658901810646057,
      "learning_rate": 0.00019817732760025686,
      "loss": 0.4335,
      "step": 27940
    },
    {
      "epoch": 0.32360024081878386,
      "grad_norm": 0.9087826013565063,
      "learning_rate": 0.0001981749012314423,
      "loss": 0.4623,
      "step": 27950
    },
    {
      "epoch": 0.3237160190802575,
      "grad_norm": 0.7327570915222168,
      "learning_rate": 0.0001981724732635659,
      "loss": 0.4456,
      "step": 27960
    },
    {
      "epoch": 0.3238317973417311,
      "grad_norm": 0.8276339173316956,
      "learning_rate": 0.00019817004369666728,
      "loss": 0.4617,
      "step": 27970
    },
    {
      "epoch": 0.32394757560320475,
      "grad_norm": 0.6314492225646973,
      "learning_rate": 0.00019816761253078596,
      "loss": 0.4099,
      "step": 27980
    },
    {
      "epoch": 0.3240633538646784,
      "grad_norm": 0.9477747678756714,
      "learning_rate": 0.0001981651797659615,
      "loss": 0.4657,
      "step": 27990
    },
    {
      "epoch": 0.324179132126152,
      "grad_norm": 0.7269591689109802,
      "learning_rate": 0.00019816274540223363,
      "loss": 0.4677,
      "step": 28000
    },
    {
      "epoch": 0.324179132126152,
      "eval_chrf": 72.20928585918801,
      "eval_loss": 0.6838599443435669,
      "eval_runtime": 279.063,
      "eval_samples_per_second": 0.358,
      "eval_steps_per_second": 0.014,
      "step": 28000
    },
    {
      "epoch": 0.32429491038762565,
      "grad_norm": 0.7083125710487366,
      "learning_rate": 0.0001981603094396419,
      "loss": 0.4615,
      "step": 28010
    },
    {
      "epoch": 0.3244106886490992,
      "grad_norm": 0.7080761790275574,
      "learning_rate": 0.00019815787187822605,
      "loss": 0.4653,
      "step": 28020
    },
    {
      "epoch": 0.32452646691057285,
      "grad_norm": 0.711959719657898,
      "learning_rate": 0.00019815543271802576,
      "loss": 0.419,
      "step": 28030
    },
    {
      "epoch": 0.3246422451720465,
      "grad_norm": 0.8714075088500977,
      "learning_rate": 0.00019815299195908076,
      "loss": 0.4829,
      "step": 28040
    },
    {
      "epoch": 0.3247580234335201,
      "grad_norm": 0.6999655365943909,
      "learning_rate": 0.00019815054960143082,
      "loss": 0.4281,
      "step": 28050
    },
    {
      "epoch": 0.32487380169499375,
      "grad_norm": 0.8160327076911926,
      "learning_rate": 0.0001981481056451157,
      "loss": 0.4191,
      "step": 28060
    },
    {
      "epoch": 0.3249895799564674,
      "grad_norm": 0.7244841456413269,
      "learning_rate": 0.0001981456600901752,
      "loss": 0.4435,
      "step": 28070
    },
    {
      "epoch": 0.325105358217941,
      "grad_norm": 0.7441781163215637,
      "learning_rate": 0.00019814321293664918,
      "loss": 0.4232,
      "step": 28080
    },
    {
      "epoch": 0.32522113647941464,
      "grad_norm": 0.8062184453010559,
      "learning_rate": 0.00019814076418457746,
      "loss": 0.4353,
      "step": 28090
    },
    {
      "epoch": 0.32533691474088827,
      "grad_norm": 0.8151717782020569,
      "learning_rate": 0.00019813831383399996,
      "loss": 0.4546,
      "step": 28100
    },
    {
      "epoch": 0.3254526930023619,
      "grad_norm": 0.6115370988845825,
      "learning_rate": 0.00019813586188495658,
      "loss": 0.4454,
      "step": 28110
    },
    {
      "epoch": 0.3255684712638355,
      "grad_norm": 0.5664581060409546,
      "learning_rate": 0.00019813340833748727,
      "loss": 0.463,
      "step": 28120
    },
    {
      "epoch": 0.3256842495253091,
      "grad_norm": 0.8573354482650757,
      "learning_rate": 0.000198130953191632,
      "loss": 0.4688,
      "step": 28130
    },
    {
      "epoch": 0.32580002778678274,
      "grad_norm": 0.5428782105445862,
      "learning_rate": 0.0001981284964474307,
      "loss": 0.453,
      "step": 28140
    },
    {
      "epoch": 0.3259158060482564,
      "grad_norm": 0.7835739850997925,
      "learning_rate": 0.00019812603810492343,
      "loss": 0.4764,
      "step": 28150
    },
    {
      "epoch": 0.32603158430973,
      "grad_norm": 0.6520804762840271,
      "learning_rate": 0.00019812357816415026,
      "loss": 0.4029,
      "step": 28160
    },
    {
      "epoch": 0.32614736257120364,
      "grad_norm": 0.7318693995475769,
      "learning_rate": 0.0001981211166251512,
      "loss": 0.4532,
      "step": 28170
    },
    {
      "epoch": 0.32626314083267727,
      "grad_norm": 0.8157362937927246,
      "learning_rate": 0.00019811865348796635,
      "loss": 0.4572,
      "step": 28180
    },
    {
      "epoch": 0.3263789190941509,
      "grad_norm": 0.8469082117080688,
      "learning_rate": 0.0001981161887526359,
      "loss": 0.4679,
      "step": 28190
    },
    {
      "epoch": 0.32649469735562453,
      "grad_norm": 0.7483899593353271,
      "learning_rate": 0.0001981137224191999,
      "loss": 0.4418,
      "step": 28200
    },
    {
      "epoch": 0.3266104756170981,
      "grad_norm": 0.7446368932723999,
      "learning_rate": 0.00019811125448769857,
      "loss": 0.4408,
      "step": 28210
    },
    {
      "epoch": 0.32672625387857174,
      "grad_norm": 0.8507893681526184,
      "learning_rate": 0.00019810878495817212,
      "loss": 0.4732,
      "step": 28220
    },
    {
      "epoch": 0.32684203214004537,
      "grad_norm": 0.8535634875297546,
      "learning_rate": 0.00019810631383066074,
      "loss": 0.4414,
      "step": 28230
    },
    {
      "epoch": 0.326957810401519,
      "grad_norm": 0.8075177073478699,
      "learning_rate": 0.00019810384110520467,
      "loss": 0.494,
      "step": 28240
    },
    {
      "epoch": 0.32707358866299263,
      "grad_norm": 0.7646968364715576,
      "learning_rate": 0.00019810136678184424,
      "loss": 0.426,
      "step": 28250
    },
    {
      "epoch": 0.32718936692446626,
      "grad_norm": 0.6302253007888794,
      "learning_rate": 0.0001980988908606197,
      "loss": 0.4815,
      "step": 28260
    },
    {
      "epoch": 0.3273051451859399,
      "grad_norm": 0.8549550771713257,
      "learning_rate": 0.00019809641334157143,
      "loss": 0.4603,
      "step": 28270
    },
    {
      "epoch": 0.3274209234474135,
      "grad_norm": 0.7948998808860779,
      "learning_rate": 0.0001980939342247397,
      "loss": 0.4556,
      "step": 28280
    },
    {
      "epoch": 0.32753670170888716,
      "grad_norm": 0.8027181029319763,
      "learning_rate": 0.000198091453510165,
      "loss": 0.4684,
      "step": 28290
    },
    {
      "epoch": 0.3276524799703608,
      "grad_norm": 0.7599784135818481,
      "learning_rate": 0.00019808897119788762,
      "loss": 0.4443,
      "step": 28300
    },
    {
      "epoch": 0.32776825823183436,
      "grad_norm": 0.8203538656234741,
      "learning_rate": 0.0001980864872879481,
      "loss": 0.4707,
      "step": 28310
    },
    {
      "epoch": 0.327884036493308,
      "grad_norm": 0.8016511797904968,
      "learning_rate": 0.00019808400178038683,
      "loss": 0.4688,
      "step": 28320
    },
    {
      "epoch": 0.3279998147547816,
      "grad_norm": 0.6890262961387634,
      "learning_rate": 0.0001980815146752443,
      "loss": 0.4154,
      "step": 28330
    },
    {
      "epoch": 0.32811559301625526,
      "grad_norm": 0.8005924820899963,
      "learning_rate": 0.00019807902597256103,
      "loss": 0.4742,
      "step": 28340
    },
    {
      "epoch": 0.3282313712777289,
      "grad_norm": 0.7315450310707092,
      "learning_rate": 0.00019807653567237757,
      "loss": 0.4368,
      "step": 28350
    },
    {
      "epoch": 0.3283471495392025,
      "grad_norm": 0.7996712327003479,
      "learning_rate": 0.00019807404377473445,
      "loss": 0.4893,
      "step": 28360
    },
    {
      "epoch": 0.32846292780067615,
      "grad_norm": 0.794567883014679,
      "learning_rate": 0.00019807155027967225,
      "loss": 0.4438,
      "step": 28370
    },
    {
      "epoch": 0.3285787060621498,
      "grad_norm": 0.9396027326583862,
      "learning_rate": 0.00019806905518723167,
      "loss": 0.4588,
      "step": 28380
    },
    {
      "epoch": 0.3286944843236234,
      "grad_norm": 0.7226946949958801,
      "learning_rate": 0.00019806655849745325,
      "loss": 0.4672,
      "step": 28390
    },
    {
      "epoch": 0.32881026258509705,
      "grad_norm": 0.7486585974693298,
      "learning_rate": 0.0001980640602103777,
      "loss": 0.4401,
      "step": 28400
    },
    {
      "epoch": 0.3289260408465706,
      "grad_norm": 0.7915642857551575,
      "learning_rate": 0.0001980615603260457,
      "loss": 0.451,
      "step": 28410
    },
    {
      "epoch": 0.32904181910804425,
      "grad_norm": 0.7100210785865784,
      "learning_rate": 0.00019805905884449798,
      "loss": 0.4529,
      "step": 28420
    },
    {
      "epoch": 0.3291575973695179,
      "grad_norm": 0.7301798462867737,
      "learning_rate": 0.00019805655576577528,
      "loss": 0.4579,
      "step": 28430
    },
    {
      "epoch": 0.3292733756309915,
      "grad_norm": 0.8199750185012817,
      "learning_rate": 0.00019805405108991837,
      "loss": 0.4611,
      "step": 28440
    },
    {
      "epoch": 0.32938915389246515,
      "grad_norm": 0.7680374383926392,
      "learning_rate": 0.00019805154481696805,
      "loss": 0.4536,
      "step": 28450
    },
    {
      "epoch": 0.3295049321539388,
      "grad_norm": 0.7141996622085571,
      "learning_rate": 0.0001980490369469651,
      "loss": 0.4659,
      "step": 28460
    },
    {
      "epoch": 0.3296207104154124,
      "grad_norm": 0.8633837103843689,
      "learning_rate": 0.00019804652747995043,
      "loss": 0.458,
      "step": 28470
    },
    {
      "epoch": 0.32973648867688604,
      "grad_norm": 0.8147624731063843,
      "learning_rate": 0.0001980440164159649,
      "loss": 0.491,
      "step": 28480
    },
    {
      "epoch": 0.3298522669383597,
      "grad_norm": 0.741216778755188,
      "learning_rate": 0.00019804150375504938,
      "loss": 0.4706,
      "step": 28490
    },
    {
      "epoch": 0.3299680451998333,
      "grad_norm": 0.7062682509422302,
      "learning_rate": 0.0001980389894972448,
      "loss": 0.4395,
      "step": 28500
    },
    {
      "epoch": 0.3300838234613069,
      "grad_norm": 0.6140871047973633,
      "learning_rate": 0.00019803647364259215,
      "loss": 0.4383,
      "step": 28510
    },
    {
      "epoch": 0.3301996017227805,
      "grad_norm": 0.8302459716796875,
      "learning_rate": 0.00019803395619113237,
      "loss": 0.449,
      "step": 28520
    },
    {
      "epoch": 0.33031537998425414,
      "grad_norm": 0.7300112247467041,
      "learning_rate": 0.00019803143714290647,
      "loss": 0.4883,
      "step": 28530
    },
    {
      "epoch": 0.3304311582457278,
      "grad_norm": 0.7336609959602356,
      "learning_rate": 0.00019802891649795549,
      "loss": 0.4883,
      "step": 28540
    },
    {
      "epoch": 0.3305469365072014,
      "grad_norm": 0.5155927538871765,
      "learning_rate": 0.00019802639425632048,
      "loss": 0.4536,
      "step": 28550
    },
    {
      "epoch": 0.33066271476867504,
      "grad_norm": 0.7415698766708374,
      "learning_rate": 0.00019802387041804254,
      "loss": 0.4322,
      "step": 28560
    },
    {
      "epoch": 0.33077849303014867,
      "grad_norm": 0.7414401769638062,
      "learning_rate": 0.00019802134498316277,
      "loss": 0.4622,
      "step": 28570
    },
    {
      "epoch": 0.3308942712916223,
      "grad_norm": 0.8409310579299927,
      "learning_rate": 0.00019801881795172225,
      "loss": 0.4454,
      "step": 28580
    },
    {
      "epoch": 0.33101004955309593,
      "grad_norm": 0.6393105983734131,
      "learning_rate": 0.0001980162893237622,
      "loss": 0.4252,
      "step": 28590
    },
    {
      "epoch": 0.33112582781456956,
      "grad_norm": 0.7128394842147827,
      "learning_rate": 0.00019801375909932385,
      "loss": 0.4631,
      "step": 28600
    },
    {
      "epoch": 0.33124160607604314,
      "grad_norm": 0.7944321036338806,
      "learning_rate": 0.0001980112272784483,
      "loss": 0.4658,
      "step": 28610
    },
    {
      "epoch": 0.33135738433751677,
      "grad_norm": 0.6957277655601501,
      "learning_rate": 0.00019800869386117685,
      "loss": 0.4486,
      "step": 28620
    },
    {
      "epoch": 0.3314731625989904,
      "grad_norm": 1.0048245191574097,
      "learning_rate": 0.00019800615884755076,
      "loss": 0.4235,
      "step": 28630
    },
    {
      "epoch": 0.33158894086046403,
      "grad_norm": 0.7389762997627258,
      "learning_rate": 0.00019800362223761133,
      "loss": 0.426,
      "step": 28640
    },
    {
      "epoch": 0.33170471912193766,
      "grad_norm": 0.8358013033866882,
      "learning_rate": 0.00019800108403139984,
      "loss": 0.4481,
      "step": 28650
    },
    {
      "epoch": 0.3318204973834113,
      "grad_norm": 0.7752150297164917,
      "learning_rate": 0.00019799854422895767,
      "loss": 0.4602,
      "step": 28660
    },
    {
      "epoch": 0.3319362756448849,
      "grad_norm": 0.7261896133422852,
      "learning_rate": 0.00019799600283032618,
      "loss": 0.4502,
      "step": 28670
    },
    {
      "epoch": 0.33205205390635856,
      "grad_norm": 0.8711327314376831,
      "learning_rate": 0.00019799345983554677,
      "loss": 0.4487,
      "step": 28680
    },
    {
      "epoch": 0.3321678321678322,
      "grad_norm": 0.6738938689231873,
      "learning_rate": 0.00019799091524466081,
      "loss": 0.4594,
      "step": 28690
    },
    {
      "epoch": 0.3322836104293058,
      "grad_norm": 0.7297458648681641,
      "learning_rate": 0.00019798836905770978,
      "loss": 0.4177,
      "step": 28700
    },
    {
      "epoch": 0.3323993886907794,
      "grad_norm": 0.831659734249115,
      "learning_rate": 0.0001979858212747352,
      "loss": 0.4731,
      "step": 28710
    },
    {
      "epoch": 0.332515166952253,
      "grad_norm": 0.7975872159004211,
      "learning_rate": 0.0001979832718957785,
      "loss": 0.4703,
      "step": 28720
    },
    {
      "epoch": 0.33263094521372666,
      "grad_norm": 0.843238115310669,
      "learning_rate": 0.00019798072092088123,
      "loss": 0.4705,
      "step": 28730
    },
    {
      "epoch": 0.3327467234752003,
      "grad_norm": 0.8644689917564392,
      "learning_rate": 0.00019797816835008492,
      "loss": 0.4634,
      "step": 28740
    },
    {
      "epoch": 0.3328625017366739,
      "grad_norm": 0.724216639995575,
      "learning_rate": 0.00019797561418343117,
      "loss": 0.4474,
      "step": 28750
    },
    {
      "epoch": 0.33297827999814755,
      "grad_norm": 0.8634474873542786,
      "learning_rate": 0.00019797305842096156,
      "loss": 0.4737,
      "step": 28760
    },
    {
      "epoch": 0.3330940582596212,
      "grad_norm": 0.7459732890129089,
      "learning_rate": 0.00019797050106271774,
      "loss": 0.4482,
      "step": 28770
    },
    {
      "epoch": 0.3332098365210948,
      "grad_norm": 0.8599674105644226,
      "learning_rate": 0.0001979679421087414,
      "loss": 0.4792,
      "step": 28780
    },
    {
      "epoch": 0.33332561478256845,
      "grad_norm": 0.7562803030014038,
      "learning_rate": 0.0001979653815590741,
      "loss": 0.4429,
      "step": 28790
    },
    {
      "epoch": 0.3334413930440421,
      "grad_norm": 0.5722392797470093,
      "learning_rate": 0.0001979628194137577,
      "loss": 0.4477,
      "step": 28800
    },
    {
      "epoch": 0.33355717130551565,
      "grad_norm": 0.9015145301818848,
      "learning_rate": 0.0001979602556728338,
      "loss": 0.4436,
      "step": 28810
    },
    {
      "epoch": 0.3336729495669893,
      "grad_norm": 0.708286702632904,
      "learning_rate": 0.00019795769033634426,
      "loss": 0.4375,
      "step": 28820
    },
    {
      "epoch": 0.3337887278284629,
      "grad_norm": 0.7045858502388,
      "learning_rate": 0.00019795512340433077,
      "loss": 0.4309,
      "step": 28830
    },
    {
      "epoch": 0.33390450608993655,
      "grad_norm": 0.5952958464622498,
      "learning_rate": 0.0001979525548768352,
      "loss": 0.4651,
      "step": 28840
    },
    {
      "epoch": 0.3340202843514102,
      "grad_norm": 0.8038861751556396,
      "learning_rate": 0.00019794998475389938,
      "loss": 0.4359,
      "step": 28850
    },
    {
      "epoch": 0.3341360626128838,
      "grad_norm": 0.750450849533081,
      "learning_rate": 0.00019794741303556516,
      "loss": 0.4545,
      "step": 28860
    },
    {
      "epoch": 0.33425184087435744,
      "grad_norm": 0.7617290019989014,
      "learning_rate": 0.00019794483972187442,
      "loss": 0.4659,
      "step": 28870
    },
    {
      "epoch": 0.3343676191358311,
      "grad_norm": 0.7738581299781799,
      "learning_rate": 0.00019794226481286914,
      "loss": 0.4279,
      "step": 28880
    },
    {
      "epoch": 0.3344833973973047,
      "grad_norm": 0.6191737651824951,
      "learning_rate": 0.00019793968830859114,
      "loss": 0.4197,
      "step": 28890
    },
    {
      "epoch": 0.3345991756587783,
      "grad_norm": 0.7525450587272644,
      "learning_rate": 0.00019793711020908247,
      "loss": 0.4557,
      "step": 28900
    },
    {
      "epoch": 0.3347149539202519,
      "grad_norm": 0.8681761622428894,
      "learning_rate": 0.00019793453051438514,
      "loss": 0.4452,
      "step": 28910
    },
    {
      "epoch": 0.33483073218172554,
      "grad_norm": 0.8026580810546875,
      "learning_rate": 0.0001979319492245411,
      "loss": 0.4687,
      "step": 28920
    },
    {
      "epoch": 0.3349465104431992,
      "grad_norm": 0.7069211602210999,
      "learning_rate": 0.00019792936633959243,
      "loss": 0.4483,
      "step": 28930
    },
    {
      "epoch": 0.3350622887046728,
      "grad_norm": 0.7879505157470703,
      "learning_rate": 0.0001979267818595812,
      "loss": 0.439,
      "step": 28940
    },
    {
      "epoch": 0.33517806696614644,
      "grad_norm": 0.8035557866096497,
      "learning_rate": 0.00019792419578454947,
      "loss": 0.4152,
      "step": 28950
    },
    {
      "epoch": 0.33529384522762007,
      "grad_norm": 0.8120391964912415,
      "learning_rate": 0.0001979216081145394,
      "loss": 0.4531,
      "step": 28960
    },
    {
      "epoch": 0.3354096234890937,
      "grad_norm": 0.8428208231925964,
      "learning_rate": 0.00019791901884959315,
      "loss": 0.4238,
      "step": 28970
    },
    {
      "epoch": 0.33552540175056733,
      "grad_norm": 0.7718029022216797,
      "learning_rate": 0.00019791642798975287,
      "loss": 0.4463,
      "step": 28980
    },
    {
      "epoch": 0.33564118001204096,
      "grad_norm": 0.7140524983406067,
      "learning_rate": 0.00019791383553506076,
      "loss": 0.4389,
      "step": 28990
    },
    {
      "epoch": 0.33575695827351454,
      "grad_norm": 0.7626215219497681,
      "learning_rate": 0.00019791124148555903,
      "loss": 0.4059,
      "step": 29000
    },
    {
      "epoch": 0.33575695827351454,
      "eval_chrf": 79.30603599493149,
      "eval_loss": 0.7022404670715332,
      "eval_runtime": 348.5992,
      "eval_samples_per_second": 0.287,
      "eval_steps_per_second": 0.011,
      "step": 29000
    },
    {
      "epoch": 0.33587273653498817,
      "grad_norm": 0.7247495055198669,
      "learning_rate": 0.00019790864584128998,
      "loss": 0.3932,
      "step": 29010
    },
    {
      "epoch": 0.3359885147964618,
      "grad_norm": 0.8484256863594055,
      "learning_rate": 0.00019790604860229585,
      "loss": 0.4566,
      "step": 29020
    },
    {
      "epoch": 0.33610429305793543,
      "grad_norm": 0.7210016250610352,
      "learning_rate": 0.00019790344976861895,
      "loss": 0.4445,
      "step": 29030
    },
    {
      "epoch": 0.33622007131940906,
      "grad_norm": 0.6728957891464233,
      "learning_rate": 0.0001979008493403016,
      "loss": 0.4357,
      "step": 29040
    },
    {
      "epoch": 0.3363358495808827,
      "grad_norm": 0.7190279364585876,
      "learning_rate": 0.00019789824731738617,
      "loss": 0.4679,
      "step": 29050
    },
    {
      "epoch": 0.3364516278423563,
      "grad_norm": 0.7486171722412109,
      "learning_rate": 0.00019789564369991505,
      "loss": 0.479,
      "step": 29060
    },
    {
      "epoch": 0.33656740610382996,
      "grad_norm": 0.7148741483688354,
      "learning_rate": 0.00019789303848793063,
      "loss": 0.4361,
      "step": 29070
    },
    {
      "epoch": 0.3366831843653036,
      "grad_norm": 0.6511372327804565,
      "learning_rate": 0.0001978904316814753,
      "loss": 0.4669,
      "step": 29080
    },
    {
      "epoch": 0.3367989626267772,
      "grad_norm": 0.9360610246658325,
      "learning_rate": 0.00019788782328059165,
      "loss": 0.4719,
      "step": 29090
    },
    {
      "epoch": 0.3369147408882508,
      "grad_norm": 0.900389552116394,
      "learning_rate": 0.00019788521328532204,
      "loss": 0.4254,
      "step": 29100
    },
    {
      "epoch": 0.33703051914972443,
      "grad_norm": 0.7916105389595032,
      "learning_rate": 0.00019788260169570902,
      "loss": 0.4634,
      "step": 29110
    },
    {
      "epoch": 0.33714629741119806,
      "grad_norm": 0.6842415928840637,
      "learning_rate": 0.00019787998851179513,
      "loss": 0.4727,
      "step": 29120
    },
    {
      "epoch": 0.3372620756726717,
      "grad_norm": 1.0879751443862915,
      "learning_rate": 0.00019787737373362296,
      "loss": 0.4772,
      "step": 29130
    },
    {
      "epoch": 0.3373778539341453,
      "grad_norm": 0.742527186870575,
      "learning_rate": 0.00019787475736123507,
      "loss": 0.4523,
      "step": 29140
    },
    {
      "epoch": 0.33749363219561895,
      "grad_norm": 0.9419659376144409,
      "learning_rate": 0.00019787213939467406,
      "loss": 0.4492,
      "step": 29150
    },
    {
      "epoch": 0.3376094104570926,
      "grad_norm": 0.6919956803321838,
      "learning_rate": 0.00019786951983398262,
      "loss": 0.4202,
      "step": 29160
    },
    {
      "epoch": 0.3377251887185662,
      "grad_norm": 0.6615548133850098,
      "learning_rate": 0.00019786689867920334,
      "loss": 0.4369,
      "step": 29170
    },
    {
      "epoch": 0.33784096698003985,
      "grad_norm": 0.8853069543838501,
      "learning_rate": 0.00019786427593037898,
      "loss": 0.4187,
      "step": 29180
    },
    {
      "epoch": 0.3379567452415135,
      "grad_norm": 0.699215292930603,
      "learning_rate": 0.00019786165158755227,
      "loss": 0.4412,
      "step": 29190
    },
    {
      "epoch": 0.33807252350298705,
      "grad_norm": 0.7400662899017334,
      "learning_rate": 0.0001978590256507659,
      "loss": 0.4448,
      "step": 29200
    },
    {
      "epoch": 0.3381883017644607,
      "grad_norm": 0.4990818500518799,
      "learning_rate": 0.00019785639812006268,
      "loss": 0.451,
      "step": 29210
    },
    {
      "epoch": 0.3383040800259343,
      "grad_norm": 0.8322715163230896,
      "learning_rate": 0.00019785376899548538,
      "loss": 0.4405,
      "step": 29220
    },
    {
      "epoch": 0.33841985828740795,
      "grad_norm": 0.854572057723999,
      "learning_rate": 0.00019785113827707685,
      "loss": 0.4601,
      "step": 29230
    },
    {
      "epoch": 0.3385356365488816,
      "grad_norm": 0.7197252511978149,
      "learning_rate": 0.0001978485059648799,
      "loss": 0.4834,
      "step": 29240
    },
    {
      "epoch": 0.3386514148103552,
      "grad_norm": 0.8339878916740417,
      "learning_rate": 0.00019784587205893742,
      "loss": 0.4524,
      "step": 29250
    },
    {
      "epoch": 0.33876719307182884,
      "grad_norm": 0.728815495967865,
      "learning_rate": 0.00019784323655929237,
      "loss": 0.4471,
      "step": 29260
    },
    {
      "epoch": 0.3388829713333025,
      "grad_norm": 0.845120370388031,
      "learning_rate": 0.00019784059946598758,
      "loss": 0.468,
      "step": 29270
    },
    {
      "epoch": 0.3389987495947761,
      "grad_norm": 0.8616262674331665,
      "learning_rate": 0.00019783796077906608,
      "loss": 0.4688,
      "step": 29280
    },
    {
      "epoch": 0.33911452785624974,
      "grad_norm": 0.7504338622093201,
      "learning_rate": 0.00019783532049857082,
      "loss": 0.4663,
      "step": 29290
    },
    {
      "epoch": 0.3392303061177233,
      "grad_norm": 0.7446357607841492,
      "learning_rate": 0.00019783267862454477,
      "loss": 0.4626,
      "step": 29300
    },
    {
      "epoch": 0.33934608437919694,
      "grad_norm": 0.7037805318832397,
      "learning_rate": 0.00019783003515703104,
      "loss": 0.4518,
      "step": 29310
    },
    {
      "epoch": 0.3394618626406706,
      "grad_norm": 0.7374095916748047,
      "learning_rate": 0.00019782739009607258,
      "loss": 0.4527,
      "step": 29320
    },
    {
      "epoch": 0.3395776409021442,
      "grad_norm": 0.6088375449180603,
      "learning_rate": 0.00019782474344171258,
      "loss": 0.4402,
      "step": 29330
    },
    {
      "epoch": 0.33969341916361784,
      "grad_norm": 0.8041601777076721,
      "learning_rate": 0.00019782209519399408,
      "loss": 0.4567,
      "step": 29340
    },
    {
      "epoch": 0.33980919742509147,
      "grad_norm": 0.7952395677566528,
      "learning_rate": 0.00019781944535296025,
      "loss": 0.4533,
      "step": 29350
    },
    {
      "epoch": 0.3399249756865651,
      "grad_norm": 0.7202016711235046,
      "learning_rate": 0.00019781679391865423,
      "loss": 0.4331,
      "step": 29360
    },
    {
      "epoch": 0.34004075394803873,
      "grad_norm": 0.8369538187980652,
      "learning_rate": 0.0001978141408911192,
      "loss": 0.4491,
      "step": 29370
    },
    {
      "epoch": 0.34015653220951236,
      "grad_norm": 0.6768138408660889,
      "learning_rate": 0.00019781148627039837,
      "loss": 0.4605,
      "step": 29380
    },
    {
      "epoch": 0.340272310470986,
      "grad_norm": 0.7428684234619141,
      "learning_rate": 0.000197808830056535,
      "loss": 0.4917,
      "step": 29390
    },
    {
      "epoch": 0.34038808873245957,
      "grad_norm": 0.973679780960083,
      "learning_rate": 0.00019780617224957238,
      "loss": 0.457,
      "step": 29400
    },
    {
      "epoch": 0.3405038669939332,
      "grad_norm": 0.785365879535675,
      "learning_rate": 0.00019780351284955376,
      "loss": 0.4575,
      "step": 29410
    },
    {
      "epoch": 0.34061964525540683,
      "grad_norm": 0.6418831944465637,
      "learning_rate": 0.00019780085185652245,
      "loss": 0.4742,
      "step": 29420
    },
    {
      "epoch": 0.34073542351688046,
      "grad_norm": 0.7536094188690186,
      "learning_rate": 0.0001977981892705218,
      "loss": 0.4387,
      "step": 29430
    },
    {
      "epoch": 0.3408512017783541,
      "grad_norm": 0.7309443950653076,
      "learning_rate": 0.0001977955250915952,
      "loss": 0.4493,
      "step": 29440
    },
    {
      "epoch": 0.3409669800398277,
      "grad_norm": 0.7243247628211975,
      "learning_rate": 0.00019779285931978598,
      "loss": 0.4198,
      "step": 29450
    },
    {
      "epoch": 0.34108275830130136,
      "grad_norm": 0.7006673216819763,
      "learning_rate": 0.00019779019195513762,
      "loss": 0.4267,
      "step": 29460
    },
    {
      "epoch": 0.341198536562775,
      "grad_norm": 0.6837512254714966,
      "learning_rate": 0.0001977875229976936,
      "loss": 0.4587,
      "step": 29470
    },
    {
      "epoch": 0.3413143148242486,
      "grad_norm": 0.8737258315086365,
      "learning_rate": 0.00019778485244749728,
      "loss": 0.4792,
      "step": 29480
    },
    {
      "epoch": 0.3414300930857222,
      "grad_norm": 0.7804811000823975,
      "learning_rate": 0.00019778218030459225,
      "loss": 0.4481,
      "step": 29490
    },
    {
      "epoch": 0.34154587134719583,
      "grad_norm": 0.774608314037323,
      "learning_rate": 0.000197779506569022,
      "loss": 0.4453,
      "step": 29500
    },
    {
      "epoch": 0.34166164960866946,
      "grad_norm": 0.5803654789924622,
      "learning_rate": 0.00019777683124083009,
      "loss": 0.4517,
      "step": 29510
    },
    {
      "epoch": 0.3417774278701431,
      "grad_norm": 0.7635154724121094,
      "learning_rate": 0.0001977741543200601,
      "loss": 0.4569,
      "step": 29520
    },
    {
      "epoch": 0.3418932061316167,
      "grad_norm": 0.7757710814476013,
      "learning_rate": 0.00019777147580675558,
      "loss": 0.432,
      "step": 29530
    },
    {
      "epoch": 0.34200898439309035,
      "grad_norm": 0.7158822417259216,
      "learning_rate": 0.00019776879570096022,
      "loss": 0.4493,
      "step": 29540
    },
    {
      "epoch": 0.342124762654564,
      "grad_norm": 0.7578673362731934,
      "learning_rate": 0.00019776611400271768,
      "loss": 0.452,
      "step": 29550
    },
    {
      "epoch": 0.3422405409160376,
      "grad_norm": 0.9474319219589233,
      "learning_rate": 0.00019776343071207158,
      "loss": 0.4704,
      "step": 29560
    },
    {
      "epoch": 0.34235631917751125,
      "grad_norm": 0.7483941316604614,
      "learning_rate": 0.00019776074582906564,
      "loss": 0.4652,
      "step": 29570
    },
    {
      "epoch": 0.3424720974389849,
      "grad_norm": 0.6780919432640076,
      "learning_rate": 0.00019775805935374365,
      "loss": 0.4699,
      "step": 29580
    },
    {
      "epoch": 0.34258787570045846,
      "grad_norm": 0.7262126803398132,
      "learning_rate": 0.0001977553712861493,
      "loss": 0.4865,
      "step": 29590
    },
    {
      "epoch": 0.3427036539619321,
      "grad_norm": 0.7847318649291992,
      "learning_rate": 0.00019775268162632639,
      "loss": 0.4735,
      "step": 29600
    },
    {
      "epoch": 0.3428194322234057,
      "grad_norm": 0.7372470498085022,
      "learning_rate": 0.00019774999037431873,
      "loss": 0.4448,
      "step": 29610
    },
    {
      "epoch": 0.34293521048487935,
      "grad_norm": 0.7142220139503479,
      "learning_rate": 0.0001977472975301702,
      "loss": 0.4192,
      "step": 29620
    },
    {
      "epoch": 0.343050988746353,
      "grad_norm": 0.633639395236969,
      "learning_rate": 0.0001977446030939246,
      "loss": 0.4212,
      "step": 29630
    },
    {
      "epoch": 0.3431667670078266,
      "grad_norm": 0.8264889717102051,
      "learning_rate": 0.00019774190706562582,
      "loss": 0.4497,
      "step": 29640
    },
    {
      "epoch": 0.34328254526930024,
      "grad_norm": 0.6413941979408264,
      "learning_rate": 0.00019773920944531778,
      "loss": 0.4604,
      "step": 29650
    },
    {
      "epoch": 0.3433983235307739,
      "grad_norm": 0.9008981585502625,
      "learning_rate": 0.00019773651023304446,
      "loss": 0.4372,
      "step": 29660
    },
    {
      "epoch": 0.3435141017922475,
      "grad_norm": 0.7054181098937988,
      "learning_rate": 0.0001977338094288498,
      "loss": 0.4449,
      "step": 29670
    },
    {
      "epoch": 0.34362988005372114,
      "grad_norm": 0.6859874129295349,
      "learning_rate": 0.00019773110703277778,
      "loss": 0.4227,
      "step": 29680
    },
    {
      "epoch": 0.3437456583151947,
      "grad_norm": 0.8675273656845093,
      "learning_rate": 0.0001977284030448724,
      "loss": 0.4179,
      "step": 29690
    },
    {
      "epoch": 0.34386143657666834,
      "grad_norm": 0.9087483882904053,
      "learning_rate": 0.00019772569746517773,
      "loss": 0.4253,
      "step": 29700
    },
    {
      "epoch": 0.343977214838142,
      "grad_norm": 0.6598452925682068,
      "learning_rate": 0.00019772299029373783,
      "loss": 0.4568,
      "step": 29710
    },
    {
      "epoch": 0.3440929930996156,
      "grad_norm": 0.7858065962791443,
      "learning_rate": 0.0001977202815305968,
      "loss": 0.4544,
      "step": 29720
    },
    {
      "epoch": 0.34420877136108924,
      "grad_norm": 0.8278815150260925,
      "learning_rate": 0.00019771757117579872,
      "loss": 0.4619,
      "step": 29730
    },
    {
      "epoch": 0.34432454962256287,
      "grad_norm": 0.7692482471466064,
      "learning_rate": 0.00019771485922938782,
      "loss": 0.4659,
      "step": 29740
    },
    {
      "epoch": 0.3444403278840365,
      "grad_norm": 0.7624439001083374,
      "learning_rate": 0.00019771214569140822,
      "loss": 0.445,
      "step": 29750
    },
    {
      "epoch": 0.34455610614551013,
      "grad_norm": 0.739372193813324,
      "learning_rate": 0.0001977094305619041,
      "loss": 0.4529,
      "step": 29760
    },
    {
      "epoch": 0.34467188440698376,
      "grad_norm": 0.6535555124282837,
      "learning_rate": 0.0001977067138409197,
      "loss": 0.503,
      "step": 29770
    },
    {
      "epoch": 0.3447876626684574,
      "grad_norm": 0.8062235116958618,
      "learning_rate": 0.00019770399552849926,
      "loss": 0.4537,
      "step": 29780
    },
    {
      "epoch": 0.34490344092993097,
      "grad_norm": 0.7946727275848389,
      "learning_rate": 0.0001977012756246871,
      "loss": 0.4443,
      "step": 29790
    },
    {
      "epoch": 0.3450192191914046,
      "grad_norm": 0.7691386938095093,
      "learning_rate": 0.00019769855412952748,
      "loss": 0.4302,
      "step": 29800
    },
    {
      "epoch": 0.34513499745287823,
      "grad_norm": 0.7168875336647034,
      "learning_rate": 0.00019769583104306472,
      "loss": 0.4599,
      "step": 29810
    },
    {
      "epoch": 0.34525077571435187,
      "grad_norm": 0.7042225003242493,
      "learning_rate": 0.00019769310636534322,
      "loss": 0.46,
      "step": 29820
    },
    {
      "epoch": 0.3453665539758255,
      "grad_norm": 0.7086291909217834,
      "learning_rate": 0.0001976903800964073,
      "loss": 0.4163,
      "step": 29830
    },
    {
      "epoch": 0.34548233223729913,
      "grad_norm": 0.8071205615997314,
      "learning_rate": 0.00019768765223630145,
      "loss": 0.4245,
      "step": 29840
    },
    {
      "epoch": 0.34559811049877276,
      "grad_norm": 0.9719737768173218,
      "learning_rate": 0.00019768492278506998,
      "loss": 0.4434,
      "step": 29850
    },
    {
      "epoch": 0.3457138887602464,
      "grad_norm": 0.7370936274528503,
      "learning_rate": 0.00019768219174275746,
      "loss": 0.4269,
      "step": 29860
    },
    {
      "epoch": 0.34582966702172,
      "grad_norm": 0.6700521111488342,
      "learning_rate": 0.00019767945910940832,
      "loss": 0.3961,
      "step": 29870
    },
    {
      "epoch": 0.34594544528319365,
      "grad_norm": 0.6146963238716125,
      "learning_rate": 0.00019767672488506705,
      "loss": 0.4928,
      "step": 29880
    },
    {
      "epoch": 0.34606122354466723,
      "grad_norm": 0.6936514377593994,
      "learning_rate": 0.00019767398906977824,
      "loss": 0.4466,
      "step": 29890
    },
    {
      "epoch": 0.34617700180614086,
      "grad_norm": 0.9286683797836304,
      "learning_rate": 0.00019767125166358639,
      "loss": 0.4621,
      "step": 29900
    },
    {
      "epoch": 0.3462927800676145,
      "grad_norm": 0.8040321469306946,
      "learning_rate": 0.00019766851266653617,
      "loss": 0.4254,
      "step": 29910
    },
    {
      "epoch": 0.3464085583290881,
      "grad_norm": 0.6007877588272095,
      "learning_rate": 0.0001976657720786721,
      "loss": 0.4359,
      "step": 29920
    },
    {
      "epoch": 0.34652433659056175,
      "grad_norm": 0.691734790802002,
      "learning_rate": 0.00019766302990003887,
      "loss": 0.4522,
      "step": 29930
    },
    {
      "epoch": 0.3466401148520354,
      "grad_norm": 0.778103768825531,
      "learning_rate": 0.0001976602861306811,
      "loss": 0.421,
      "step": 29940
    },
    {
      "epoch": 0.346755893113509,
      "grad_norm": 0.8104856014251709,
      "learning_rate": 0.00019765754077064355,
      "loss": 0.4594,
      "step": 29950
    },
    {
      "epoch": 0.34687167137498265,
      "grad_norm": 0.7215676307678223,
      "learning_rate": 0.00019765479381997087,
      "loss": 0.4426,
      "step": 29960
    },
    {
      "epoch": 0.3469874496364563,
      "grad_norm": 0.7769370675086975,
      "learning_rate": 0.00019765204527870788,
      "loss": 0.4394,
      "step": 29970
    },
    {
      "epoch": 0.3471032278979299,
      "grad_norm": 0.6785605549812317,
      "learning_rate": 0.00019764929514689926,
      "loss": 0.4404,
      "step": 29980
    },
    {
      "epoch": 0.3472190061594035,
      "grad_norm": 0.6804041862487793,
      "learning_rate": 0.00019764654342458986,
      "loss": 0.4735,
      "step": 29990
    },
    {
      "epoch": 0.3473347844208771,
      "grad_norm": 0.6676384210586548,
      "learning_rate": 0.00019764379011182447,
      "loss": 0.4283,
      "step": 30000
    },
    {
      "epoch": 0.3473347844208771,
      "eval_chrf": 80.61622692012071,
      "eval_loss": 0.6926522254943848,
      "eval_runtime": 197.0152,
      "eval_samples_per_second": 0.508,
      "eval_steps_per_second": 0.02,
      "step": 30000
    },
    {
      "epoch": 0.34745056268235075,
      "grad_norm": 0.9586856961250305,
      "learning_rate": 0.00019764103520864796,
      "loss": 0.4792,
      "step": 30010
    },
    {
      "epoch": 0.3475663409438244,
      "grad_norm": 0.5639489889144897,
      "learning_rate": 0.00019763827871510517,
      "loss": 0.4224,
      "step": 30020
    },
    {
      "epoch": 0.347682119205298,
      "grad_norm": 0.9713011980056763,
      "learning_rate": 0.00019763552063124101,
      "loss": 0.4641,
      "step": 30030
    },
    {
      "epoch": 0.34779789746677164,
      "grad_norm": 0.5006837248802185,
      "learning_rate": 0.00019763276095710044,
      "loss": 0.4235,
      "step": 30040
    },
    {
      "epoch": 0.3479136757282453,
      "grad_norm": 0.8714874982833862,
      "learning_rate": 0.0001976299996927284,
      "loss": 0.4555,
      "step": 30050
    },
    {
      "epoch": 0.3480294539897189,
      "grad_norm": 0.8021594882011414,
      "learning_rate": 0.0001976272368381698,
      "loss": 0.4542,
      "step": 30060
    },
    {
      "epoch": 0.34814523225119254,
      "grad_norm": 0.7853084802627563,
      "learning_rate": 0.0001976244723934697,
      "loss": 0.4639,
      "step": 30070
    },
    {
      "epoch": 0.34826101051266617,
      "grad_norm": 0.8078621625900269,
      "learning_rate": 0.00019762170635867312,
      "loss": 0.4639,
      "step": 30080
    },
    {
      "epoch": 0.34837678877413975,
      "grad_norm": 0.6852286458015442,
      "learning_rate": 0.0001976189387338251,
      "loss": 0.4596,
      "step": 30090
    },
    {
      "epoch": 0.3484925670356134,
      "grad_norm": 0.8198529481887817,
      "learning_rate": 0.0001976161695189707,
      "loss": 0.4224,
      "step": 30100
    },
    {
      "epoch": 0.348608345297087,
      "grad_norm": 0.8080971837043762,
      "learning_rate": 0.00019761339871415508,
      "loss": 0.4252,
      "step": 30110
    },
    {
      "epoch": 0.34872412355856064,
      "grad_norm": 0.8420395851135254,
      "learning_rate": 0.00019761062631942333,
      "loss": 0.4187,
      "step": 30120
    },
    {
      "epoch": 0.34883990182003427,
      "grad_norm": 0.6990167498588562,
      "learning_rate": 0.00019760785233482065,
      "loss": 0.4797,
      "step": 30130
    },
    {
      "epoch": 0.3489556800815079,
      "grad_norm": 0.6790205240249634,
      "learning_rate": 0.00019760507676039213,
      "loss": 0.4332,
      "step": 30140
    },
    {
      "epoch": 0.34907145834298153,
      "grad_norm": 0.6785823106765747,
      "learning_rate": 0.00019760229959618308,
      "loss": 0.4605,
      "step": 30150
    },
    {
      "epoch": 0.34918723660445516,
      "grad_norm": 0.8260369300842285,
      "learning_rate": 0.00019759952084223869,
      "loss": 0.4402,
      "step": 30160
    },
    {
      "epoch": 0.3493030148659288,
      "grad_norm": 0.7020462155342102,
      "learning_rate": 0.0001975967404986042,
      "loss": 0.4645,
      "step": 30170
    },
    {
      "epoch": 0.34941879312740237,
      "grad_norm": 0.8621097803115845,
      "learning_rate": 0.0001975939585653249,
      "loss": 0.4818,
      "step": 30180
    },
    {
      "epoch": 0.349534571388876,
      "grad_norm": 0.9369802474975586,
      "learning_rate": 0.00019759117504244617,
      "loss": 0.4194,
      "step": 30190
    },
    {
      "epoch": 0.34965034965034963,
      "grad_norm": 0.7883701920509338,
      "learning_rate": 0.00019758838993001328,
      "loss": 0.487,
      "step": 30200
    },
    {
      "epoch": 0.34976612791182327,
      "grad_norm": 0.7517625689506531,
      "learning_rate": 0.0001975856032280716,
      "loss": 0.4145,
      "step": 30210
    },
    {
      "epoch": 0.3498819061732969,
      "grad_norm": 0.7183372974395752,
      "learning_rate": 0.00019758281493666654,
      "loss": 0.4513,
      "step": 30220
    },
    {
      "epoch": 0.34999768443477053,
      "grad_norm": 1.0402839183807373,
      "learning_rate": 0.00019758002505584346,
      "loss": 0.3996,
      "step": 30230
    },
    {
      "epoch": 0.35011346269624416,
      "grad_norm": 0.7967755794525146,
      "learning_rate": 0.0001975772335856479,
      "loss": 0.3919,
      "step": 30240
    },
    {
      "epoch": 0.3502292409577178,
      "grad_norm": 0.7831071615219116,
      "learning_rate": 0.00019757444052612524,
      "loss": 0.4312,
      "step": 30250
    },
    {
      "epoch": 0.3503450192191914,
      "grad_norm": 0.7764838933944702,
      "learning_rate": 0.000197571645877321,
      "loss": 0.4604,
      "step": 30260
    },
    {
      "epoch": 0.35046079748066505,
      "grad_norm": 0.8191924691200256,
      "learning_rate": 0.00019756884963928074,
      "loss": 0.4338,
      "step": 30270
    },
    {
      "epoch": 0.35057657574213863,
      "grad_norm": 0.8969576954841614,
      "learning_rate": 0.00019756605181204997,
      "loss": 0.4791,
      "step": 30280
    },
    {
      "epoch": 0.35069235400361226,
      "grad_norm": 0.5914546847343445,
      "learning_rate": 0.00019756325239567426,
      "loss": 0.432,
      "step": 30290
    },
    {
      "epoch": 0.3508081322650859,
      "grad_norm": 0.7140315175056458,
      "learning_rate": 0.0001975604513901992,
      "loss": 0.4279,
      "step": 30300
    },
    {
      "epoch": 0.3509239105265595,
      "grad_norm": 0.7911560535430908,
      "learning_rate": 0.00019755764879567038,
      "loss": 0.4474,
      "step": 30310
    },
    {
      "epoch": 0.35103968878803316,
      "grad_norm": 0.7750484943389893,
      "learning_rate": 0.00019755484461213353,
      "loss": 0.4625,
      "step": 30320
    },
    {
      "epoch": 0.3511554670495068,
      "grad_norm": 0.8546374440193176,
      "learning_rate": 0.0001975520388396343,
      "loss": 0.505,
      "step": 30330
    },
    {
      "epoch": 0.3512712453109804,
      "grad_norm": 1.0121471881866455,
      "learning_rate": 0.00019754923147821834,
      "loss": 0.4737,
      "step": 30340
    },
    {
      "epoch": 0.35138702357245405,
      "grad_norm": 0.7756264805793762,
      "learning_rate": 0.0001975464225279314,
      "loss": 0.4451,
      "step": 30350
    },
    {
      "epoch": 0.3515028018339277,
      "grad_norm": 0.7790165543556213,
      "learning_rate": 0.00019754361198881924,
      "loss": 0.4125,
      "step": 30360
    },
    {
      "epoch": 0.3516185800954013,
      "grad_norm": 0.9156092405319214,
      "learning_rate": 0.0001975407998609277,
      "loss": 0.4547,
      "step": 30370
    },
    {
      "epoch": 0.3517343583568749,
      "grad_norm": 0.7473875284194946,
      "learning_rate": 0.0001975379861443025,
      "loss": 0.421,
      "step": 30380
    },
    {
      "epoch": 0.3518501366183485,
      "grad_norm": 0.6192530393600464,
      "learning_rate": 0.00019753517083898942,
      "loss": 0.4271,
      "step": 30390
    },
    {
      "epoch": 0.35196591487982215,
      "grad_norm": 0.749930202960968,
      "learning_rate": 0.00019753235394503443,
      "loss": 0.4719,
      "step": 30400
    },
    {
      "epoch": 0.3520816931412958,
      "grad_norm": 0.8089101314544678,
      "learning_rate": 0.0001975295354624834,
      "loss": 0.4234,
      "step": 30410
    },
    {
      "epoch": 0.3521974714027694,
      "grad_norm": 0.7470836043357849,
      "learning_rate": 0.00019752671539138218,
      "loss": 0.4627,
      "step": 30420
    },
    {
      "epoch": 0.35231324966424304,
      "grad_norm": 0.7350556254386902,
      "learning_rate": 0.00019752389373177672,
      "loss": 0.4667,
      "step": 30430
    },
    {
      "epoch": 0.3524290279257167,
      "grad_norm": 0.8652623891830444,
      "learning_rate": 0.00019752107048371298,
      "loss": 0.4819,
      "step": 30440
    },
    {
      "epoch": 0.3525448061871903,
      "grad_norm": 0.6465603113174438,
      "learning_rate": 0.00019751824564723697,
      "loss": 0.4403,
      "step": 30450
    },
    {
      "epoch": 0.35266058444866394,
      "grad_norm": 0.7292660474777222,
      "learning_rate": 0.0001975154192223947,
      "loss": 0.4178,
      "step": 30460
    },
    {
      "epoch": 0.35277636271013757,
      "grad_norm": 0.9069141149520874,
      "learning_rate": 0.00019751259120923215,
      "loss": 0.4381,
      "step": 30470
    },
    {
      "epoch": 0.35289214097161115,
      "grad_norm": 0.8609533905982971,
      "learning_rate": 0.00019750976160779545,
      "loss": 0.423,
      "step": 30480
    },
    {
      "epoch": 0.3530079192330848,
      "grad_norm": 0.7562017440795898,
      "learning_rate": 0.00019750693041813063,
      "loss": 0.4364,
      "step": 30490
    },
    {
      "epoch": 0.3531236974945584,
      "grad_norm": 0.9907569885253906,
      "learning_rate": 0.00019750409764028386,
      "loss": 0.4564,
      "step": 30500
    },
    {
      "epoch": 0.35323947575603204,
      "grad_norm": 0.7624794840812683,
      "learning_rate": 0.00019750126327430125,
      "loss": 0.4497,
      "step": 30510
    },
    {
      "epoch": 0.35335525401750567,
      "grad_norm": 0.6962045431137085,
      "learning_rate": 0.00019749842732022895,
      "loss": 0.4709,
      "step": 30520
    },
    {
      "epoch": 0.3534710322789793,
      "grad_norm": 0.6380375027656555,
      "learning_rate": 0.0001974955897781132,
      "loss": 0.4577,
      "step": 30530
    },
    {
      "epoch": 0.35358681054045293,
      "grad_norm": 0.8284725546836853,
      "learning_rate": 0.00019749275064800018,
      "loss": 0.4691,
      "step": 30540
    },
    {
      "epoch": 0.35370258880192657,
      "grad_norm": 0.7660990953445435,
      "learning_rate": 0.00019748990992993613,
      "loss": 0.436,
      "step": 30550
    },
    {
      "epoch": 0.3538183670634002,
      "grad_norm": 0.8985053896903992,
      "learning_rate": 0.00019748706762396734,
      "loss": 0.4626,
      "step": 30560
    },
    {
      "epoch": 0.35393414532487383,
      "grad_norm": 0.7087849974632263,
      "learning_rate": 0.0001974842237301401,
      "loss": 0.418,
      "step": 30570
    },
    {
      "epoch": 0.3540499235863474,
      "grad_norm": 0.7461625933647156,
      "learning_rate": 0.00019748137824850072,
      "loss": 0.4462,
      "step": 30580
    },
    {
      "epoch": 0.35416570184782104,
      "grad_norm": 0.9012613892555237,
      "learning_rate": 0.00019747853117909556,
      "loss": 0.4615,
      "step": 30590
    },
    {
      "epoch": 0.35428148010929467,
      "grad_norm": 0.9633703827857971,
      "learning_rate": 0.00019747568252197098,
      "loss": 0.4091,
      "step": 30600
    },
    {
      "epoch": 0.3543972583707683,
      "grad_norm": 0.7843459844589233,
      "learning_rate": 0.00019747283227717336,
      "loss": 0.4461,
      "step": 30610
    },
    {
      "epoch": 0.35451303663224193,
      "grad_norm": 0.7528501152992249,
      "learning_rate": 0.0001974699804447492,
      "loss": 0.5251,
      "step": 30620
    },
    {
      "epoch": 0.35462881489371556,
      "grad_norm": 0.726162850856781,
      "learning_rate": 0.0001974671270247449,
      "loss": 0.4266,
      "step": 30630
    },
    {
      "epoch": 0.3547445931551892,
      "grad_norm": 0.729716956615448,
      "learning_rate": 0.0001974642720172069,
      "loss": 0.4808,
      "step": 30640
    },
    {
      "epoch": 0.3548603714166628,
      "grad_norm": 0.7938385605812073,
      "learning_rate": 0.00019746141542218175,
      "loss": 0.4584,
      "step": 30650
    },
    {
      "epoch": 0.35497614967813645,
      "grad_norm": 0.7267888188362122,
      "learning_rate": 0.00019745855723971598,
      "loss": 0.4256,
      "step": 30660
    },
    {
      "epoch": 0.3550919279396101,
      "grad_norm": 0.7421509027481079,
      "learning_rate": 0.00019745569746985612,
      "loss": 0.4851,
      "step": 30670
    },
    {
      "epoch": 0.35520770620108366,
      "grad_norm": 0.7564364671707153,
      "learning_rate": 0.00019745283611264877,
      "loss": 0.4532,
      "step": 30680
    },
    {
      "epoch": 0.3553234844625573,
      "grad_norm": 0.5512920618057251,
      "learning_rate": 0.00019744997316814051,
      "loss": 0.4214,
      "step": 30690
    },
    {
      "epoch": 0.3554392627240309,
      "grad_norm": 0.7705976963043213,
      "learning_rate": 0.000197447108636378,
      "loss": 0.4598,
      "step": 30700
    },
    {
      "epoch": 0.35555504098550456,
      "grad_norm": 1.0558487176895142,
      "learning_rate": 0.00019744424251740786,
      "loss": 0.4019,
      "step": 30710
    },
    {
      "epoch": 0.3556708192469782,
      "grad_norm": 0.7541359066963196,
      "learning_rate": 0.00019744137481127685,
      "loss": 0.4967,
      "step": 30720
    },
    {
      "epoch": 0.3557865975084518,
      "grad_norm": 0.6676609516143799,
      "learning_rate": 0.00019743850551803157,
      "loss": 0.4654,
      "step": 30730
    },
    {
      "epoch": 0.35590237576992545,
      "grad_norm": 0.792260468006134,
      "learning_rate": 0.00019743563463771882,
      "loss": 0.4302,
      "step": 30740
    },
    {
      "epoch": 0.3560181540313991,
      "grad_norm": 0.7691541910171509,
      "learning_rate": 0.0001974327621703854,
      "loss": 0.4515,
      "step": 30750
    },
    {
      "epoch": 0.3561339322928727,
      "grad_norm": 0.7016454339027405,
      "learning_rate": 0.00019742988811607802,
      "loss": 0.4616,
      "step": 30760
    },
    {
      "epoch": 0.3562497105543463,
      "grad_norm": 0.7144368886947632,
      "learning_rate": 0.0001974270124748435,
      "loss": 0.4486,
      "step": 30770
    },
    {
      "epoch": 0.3563654888158199,
      "grad_norm": 0.7607747912406921,
      "learning_rate": 0.00019742413524672872,
      "loss": 0.4339,
      "step": 30780
    },
    {
      "epoch": 0.35648126707729355,
      "grad_norm": 0.7062933444976807,
      "learning_rate": 0.0001974212564317805,
      "loss": 0.4376,
      "step": 30790
    },
    {
      "epoch": 0.3565970453387672,
      "grad_norm": 0.6070752739906311,
      "learning_rate": 0.00019741837603004576,
      "loss": 0.4301,
      "step": 30800
    },
    {
      "epoch": 0.3567128236002408,
      "grad_norm": 0.7491345405578613,
      "learning_rate": 0.0001974154940415714,
      "loss": 0.4838,
      "step": 30810
    },
    {
      "epoch": 0.35682860186171445,
      "grad_norm": 0.9208494424819946,
      "learning_rate": 0.00019741261046640437,
      "loss": 0.4285,
      "step": 30820
    },
    {
      "epoch": 0.3569443801231881,
      "grad_norm": 0.7147670984268188,
      "learning_rate": 0.00019740972530459168,
      "loss": 0.4393,
      "step": 30830
    },
    {
      "epoch": 0.3570601583846617,
      "grad_norm": 0.7651496529579163,
      "learning_rate": 0.00019740683855618024,
      "loss": 0.4606,
      "step": 30840
    },
    {
      "epoch": 0.35717593664613534,
      "grad_norm": 0.7650686502456665,
      "learning_rate": 0.00019740395022121708,
      "loss": 0.4356,
      "step": 30850
    },
    {
      "epoch": 0.35729171490760897,
      "grad_norm": 0.7028475403785706,
      "learning_rate": 0.00019740106029974934,
      "loss": 0.488,
      "step": 30860
    },
    {
      "epoch": 0.35740749316908255,
      "grad_norm": 0.7773790955543518,
      "learning_rate": 0.000197398168791824,
      "loss": 0.433,
      "step": 30870
    },
    {
      "epoch": 0.3575232714305562,
      "grad_norm": 0.8732051849365234,
      "learning_rate": 0.00019739527569748817,
      "loss": 0.4675,
      "step": 30880
    },
    {
      "epoch": 0.3576390496920298,
      "grad_norm": 0.7385003566741943,
      "learning_rate": 0.00019739238101678898,
      "loss": 0.4893,
      "step": 30890
    },
    {
      "epoch": 0.35775482795350344,
      "grad_norm": 0.6992689967155457,
      "learning_rate": 0.0001973894847497736,
      "loss": 0.4346,
      "step": 30900
    },
    {
      "epoch": 0.35787060621497707,
      "grad_norm": 0.6133136749267578,
      "learning_rate": 0.00019738658689648917,
      "loss": 0.482,
      "step": 30910
    },
    {
      "epoch": 0.3579863844764507,
      "grad_norm": 0.7630688548088074,
      "learning_rate": 0.0001973836874569829,
      "loss": 0.4392,
      "step": 30920
    },
    {
      "epoch": 0.35810216273792433,
      "grad_norm": 0.8155734539031982,
      "learning_rate": 0.00019738078643130204,
      "loss": 0.4649,
      "step": 30930
    },
    {
      "epoch": 0.35821794099939797,
      "grad_norm": 0.9518517255783081,
      "learning_rate": 0.0001973778838194938,
      "loss": 0.4562,
      "step": 30940
    },
    {
      "epoch": 0.3583337192608716,
      "grad_norm": 0.7084271311759949,
      "learning_rate": 0.00019737497962160549,
      "loss": 0.4329,
      "step": 30950
    },
    {
      "epoch": 0.35844949752234523,
      "grad_norm": 0.5387693047523499,
      "learning_rate": 0.00019737207383768441,
      "loss": 0.4397,
      "step": 30960
    },
    {
      "epoch": 0.3585652757838188,
      "grad_norm": 0.6200643181800842,
      "learning_rate": 0.00019736916646777788,
      "loss": 0.4187,
      "step": 30970
    },
    {
      "epoch": 0.35868105404529244,
      "grad_norm": 0.5123502612113953,
      "learning_rate": 0.00019736625751193328,
      "loss": 0.4131,
      "step": 30980
    },
    {
      "epoch": 0.35879683230676607,
      "grad_norm": 0.5148366093635559,
      "learning_rate": 0.00019736334697019792,
      "loss": 0.4398,
      "step": 30990
    },
    {
      "epoch": 0.3589126105682397,
      "grad_norm": 0.6812059879302979,
      "learning_rate": 0.0001973604348426193,
      "loss": 0.4133,
      "step": 31000
    },
    {
      "epoch": 0.3589126105682397,
      "eval_chrf": 83.24967828273468,
      "eval_loss": 0.6759328842163086,
      "eval_runtime": 216.1219,
      "eval_samples_per_second": 0.463,
      "eval_steps_per_second": 0.019,
      "step": 31000
    },
    {
      "epoch": 0.35902838882971333,
      "grad_norm": 0.7028791904449463,
      "learning_rate": 0.00019735752112924477,
      "loss": 0.4171,
      "step": 31010
    },
    {
      "epoch": 0.35914416709118696,
      "grad_norm": 0.6313734650611877,
      "learning_rate": 0.00019735460583012184,
      "loss": 0.477,
      "step": 31020
    },
    {
      "epoch": 0.3592599453526606,
      "grad_norm": 0.8126435875892639,
      "learning_rate": 0.000197351688945298,
      "loss": 0.4533,
      "step": 31030
    },
    {
      "epoch": 0.3593757236141342,
      "grad_norm": 0.8385278582572937,
      "learning_rate": 0.0001973487704748207,
      "loss": 0.4409,
      "step": 31040
    },
    {
      "epoch": 0.35949150187560786,
      "grad_norm": 0.6164735555648804,
      "learning_rate": 0.00019734585041873755,
      "loss": 0.4543,
      "step": 31050
    },
    {
      "epoch": 0.3596072801370815,
      "grad_norm": 0.8127566576004028,
      "learning_rate": 0.00019734292877709604,
      "loss": 0.4384,
      "step": 31060
    },
    {
      "epoch": 0.35972305839855506,
      "grad_norm": 0.7247339487075806,
      "learning_rate": 0.00019734000554994384,
      "loss": 0.4416,
      "step": 31070
    },
    {
      "epoch": 0.3598388366600287,
      "grad_norm": 0.7585310339927673,
      "learning_rate": 0.00019733708073732848,
      "loss": 0.442,
      "step": 31080
    },
    {
      "epoch": 0.3599546149215023,
      "grad_norm": 0.7440115213394165,
      "learning_rate": 0.00019733415433929765,
      "loss": 0.4422,
      "step": 31090
    },
    {
      "epoch": 0.36007039318297596,
      "grad_norm": 0.59879070520401,
      "learning_rate": 0.000197331226355899,
      "loss": 0.4436,
      "step": 31100
    },
    {
      "epoch": 0.3601861714444496,
      "grad_norm": 0.6699204444885254,
      "learning_rate": 0.0001973282967871802,
      "loss": 0.4383,
      "step": 31110
    },
    {
      "epoch": 0.3603019497059232,
      "grad_norm": 0.8452684879302979,
      "learning_rate": 0.00019732536563318905,
      "loss": 0.447,
      "step": 31120
    },
    {
      "epoch": 0.36041772796739685,
      "grad_norm": 0.6436376571655273,
      "learning_rate": 0.00019732243289397316,
      "loss": 0.4109,
      "step": 31130
    },
    {
      "epoch": 0.3605335062288705,
      "grad_norm": 0.6838762760162354,
      "learning_rate": 0.0001973194985695804,
      "loss": 0.4367,
      "step": 31140
    },
    {
      "epoch": 0.3606492844903441,
      "grad_norm": 0.5743063688278198,
      "learning_rate": 0.00019731656266005855,
      "loss": 0.4134,
      "step": 31150
    },
    {
      "epoch": 0.36076506275181774,
      "grad_norm": 0.8025876879692078,
      "learning_rate": 0.0001973136251654554,
      "loss": 0.4349,
      "step": 31160
    },
    {
      "epoch": 0.3608808410132913,
      "grad_norm": 0.8272978067398071,
      "learning_rate": 0.0001973106860858188,
      "loss": 0.4532,
      "step": 31170
    },
    {
      "epoch": 0.36099661927476495,
      "grad_norm": 0.7785543203353882,
      "learning_rate": 0.00019730774542119664,
      "loss": 0.4558,
      "step": 31180
    },
    {
      "epoch": 0.3611123975362386,
      "grad_norm": 0.6963177919387817,
      "learning_rate": 0.00019730480317163676,
      "loss": 0.4561,
      "step": 31190
    },
    {
      "epoch": 0.3612281757977122,
      "grad_norm": 0.7514274716377258,
      "learning_rate": 0.0001973018593371872,
      "loss": 0.4569,
      "step": 31200
    },
    {
      "epoch": 0.36134395405918585,
      "grad_norm": 0.8091863393783569,
      "learning_rate": 0.00019729891391789578,
      "loss": 0.4404,
      "step": 31210
    },
    {
      "epoch": 0.3614597323206595,
      "grad_norm": 0.7302409410476685,
      "learning_rate": 0.00019729596691381058,
      "loss": 0.4278,
      "step": 31220
    },
    {
      "epoch": 0.3615755105821331,
      "grad_norm": 0.8030725121498108,
      "learning_rate": 0.00019729301832497952,
      "loss": 0.4661,
      "step": 31230
    },
    {
      "epoch": 0.36169128884360674,
      "grad_norm": 0.6427509188652039,
      "learning_rate": 0.00019729006815145069,
      "loss": 0.424,
      "step": 31240
    },
    {
      "epoch": 0.36180706710508037,
      "grad_norm": 0.6757525205612183,
      "learning_rate": 0.0001972871163932721,
      "loss": 0.4417,
      "step": 31250
    },
    {
      "epoch": 0.361922845366554,
      "grad_norm": 0.7824217677116394,
      "learning_rate": 0.00019728416305049183,
      "loss": 0.4366,
      "step": 31260
    },
    {
      "epoch": 0.3620386236280276,
      "grad_norm": 0.7894025444984436,
      "learning_rate": 0.00019728120812315797,
      "loss": 0.473,
      "step": 31270
    },
    {
      "epoch": 0.3621544018895012,
      "grad_norm": 0.7695927023887634,
      "learning_rate": 0.0001972782516113187,
      "loss": 0.4689,
      "step": 31280
    },
    {
      "epoch": 0.36227018015097484,
      "grad_norm": 0.6761317849159241,
      "learning_rate": 0.00019727529351502217,
      "loss": 0.4713,
      "step": 31290
    },
    {
      "epoch": 0.3623859584124485,
      "grad_norm": 0.7752683162689209,
      "learning_rate": 0.00019727233383431652,
      "loss": 0.459,
      "step": 31300
    },
    {
      "epoch": 0.3625017366739221,
      "grad_norm": 0.6869176626205444,
      "learning_rate": 0.00019726937256924994,
      "loss": 0.4621,
      "step": 31310
    },
    {
      "epoch": 0.36261751493539573,
      "grad_norm": 0.7642596960067749,
      "learning_rate": 0.00019726640971987075,
      "loss": 0.4613,
      "step": 31320
    },
    {
      "epoch": 0.36273329319686937,
      "grad_norm": 0.9158223271369934,
      "learning_rate": 0.00019726344528622715,
      "loss": 0.4443,
      "step": 31330
    },
    {
      "epoch": 0.362849071458343,
      "grad_norm": 0.8089000582695007,
      "learning_rate": 0.00019726047926836744,
      "loss": 0.4498,
      "step": 31340
    },
    {
      "epoch": 0.36296484971981663,
      "grad_norm": 0.8005766868591309,
      "learning_rate": 0.0001972575116663399,
      "loss": 0.4333,
      "step": 31350
    },
    {
      "epoch": 0.36308062798129026,
      "grad_norm": 0.7373596429824829,
      "learning_rate": 0.0001972545424801929,
      "loss": 0.4364,
      "step": 31360
    },
    {
      "epoch": 0.36319640624276384,
      "grad_norm": 0.5547783970832825,
      "learning_rate": 0.00019725157170997476,
      "loss": 0.4477,
      "step": 31370
    },
    {
      "epoch": 0.36331218450423747,
      "grad_norm": 0.8195983171463013,
      "learning_rate": 0.00019724859935573395,
      "loss": 0.4236,
      "step": 31380
    },
    {
      "epoch": 0.3634279627657111,
      "grad_norm": 0.741077184677124,
      "learning_rate": 0.0001972456254175188,
      "loss": 0.4064,
      "step": 31390
    },
    {
      "epoch": 0.36354374102718473,
      "grad_norm": 0.69661545753479,
      "learning_rate": 0.00019724264989537778,
      "loss": 0.4881,
      "step": 31400
    },
    {
      "epoch": 0.36365951928865836,
      "grad_norm": 0.7796928882598877,
      "learning_rate": 0.00019723967278935937,
      "loss": 0.4407,
      "step": 31410
    },
    {
      "epoch": 0.363775297550132,
      "grad_norm": 0.6568746566772461,
      "learning_rate": 0.00019723669409951202,
      "loss": 0.449,
      "step": 31420
    },
    {
      "epoch": 0.3638910758116056,
      "grad_norm": 0.6616988182067871,
      "learning_rate": 0.00019723371382588428,
      "loss": 0.4122,
      "step": 31430
    },
    {
      "epoch": 0.36400685407307926,
      "grad_norm": 0.7161182761192322,
      "learning_rate": 0.0001972307319685247,
      "loss": 0.4274,
      "step": 31440
    },
    {
      "epoch": 0.3641226323345529,
      "grad_norm": 0.7637435793876648,
      "learning_rate": 0.00019722774852748182,
      "loss": 0.4568,
      "step": 31450
    },
    {
      "epoch": 0.36423841059602646,
      "grad_norm": 0.6790422201156616,
      "learning_rate": 0.00019722476350280427,
      "loss": 0.4373,
      "step": 31460
    },
    {
      "epoch": 0.3643541888575001,
      "grad_norm": 0.8845119476318359,
      "learning_rate": 0.00019722177689454062,
      "loss": 0.4239,
      "step": 31470
    },
    {
      "epoch": 0.3644699671189737,
      "grad_norm": 0.7735661268234253,
      "learning_rate": 0.00019721878870273955,
      "loss": 0.4393,
      "step": 31480
    },
    {
      "epoch": 0.36458574538044736,
      "grad_norm": 0.6819901466369629,
      "learning_rate": 0.0001972157989274497,
      "loss": 0.4684,
      "step": 31490
    },
    {
      "epoch": 0.364701523641921,
      "grad_norm": 0.5607456564903259,
      "learning_rate": 0.00019721280756871984,
      "loss": 0.4635,
      "step": 31500
    },
    {
      "epoch": 0.3648173019033946,
      "grad_norm": 0.8919089436531067,
      "learning_rate": 0.0001972098146265986,
      "loss": 0.443,
      "step": 31510
    },
    {
      "epoch": 0.36493308016486825,
      "grad_norm": 0.905838668346405,
      "learning_rate": 0.0001972068201011348,
      "loss": 0.4554,
      "step": 31520
    },
    {
      "epoch": 0.3650488584263419,
      "grad_norm": 0.6962574124336243,
      "learning_rate": 0.00019720382399237717,
      "loss": 0.492,
      "step": 31530
    },
    {
      "epoch": 0.3651646366878155,
      "grad_norm": 0.8887657523155212,
      "learning_rate": 0.00019720082630037455,
      "loss": 0.4637,
      "step": 31540
    },
    {
      "epoch": 0.36528041494928915,
      "grad_norm": 0.8585859537124634,
      "learning_rate": 0.00019719782702517572,
      "loss": 0.4687,
      "step": 31550
    },
    {
      "epoch": 0.3653961932107627,
      "grad_norm": 0.644352912902832,
      "learning_rate": 0.0001971948261668296,
      "loss": 0.4414,
      "step": 31560
    },
    {
      "epoch": 0.36551197147223635,
      "grad_norm": 0.7631421089172363,
      "learning_rate": 0.00019719182372538496,
      "loss": 0.4304,
      "step": 31570
    },
    {
      "epoch": 0.36562774973371,
      "grad_norm": 0.8129732012748718,
      "learning_rate": 0.0001971888197008908,
      "loss": 0.4148,
      "step": 31580
    },
    {
      "epoch": 0.3657435279951836,
      "grad_norm": 0.7444393038749695,
      "learning_rate": 0.00019718581409339597,
      "loss": 0.4036,
      "step": 31590
    },
    {
      "epoch": 0.36585930625665725,
      "grad_norm": 0.7733768224716187,
      "learning_rate": 0.00019718280690294955,
      "loss": 0.4546,
      "step": 31600
    },
    {
      "epoch": 0.3659750845181309,
      "grad_norm": 0.571118950843811,
      "learning_rate": 0.00019717979812960038,
      "loss": 0.4128,
      "step": 31610
    },
    {
      "epoch": 0.3660908627796045,
      "grad_norm": 0.8078710436820984,
      "learning_rate": 0.00019717678777339756,
      "loss": 0.4316,
      "step": 31620
    },
    {
      "epoch": 0.36620664104107814,
      "grad_norm": 0.7164070010185242,
      "learning_rate": 0.00019717377583439006,
      "loss": 0.5278,
      "step": 31630
    },
    {
      "epoch": 0.36632241930255177,
      "grad_norm": 0.7216572761535645,
      "learning_rate": 0.00019717076231262702,
      "loss": 0.458,
      "step": 31640
    },
    {
      "epoch": 0.3664381975640254,
      "grad_norm": 0.5927894115447998,
      "learning_rate": 0.0001971677472081574,
      "loss": 0.4314,
      "step": 31650
    },
    {
      "epoch": 0.366553975825499,
      "grad_norm": 0.7894261479377747,
      "learning_rate": 0.0001971647305210304,
      "loss": 0.4654,
      "step": 31660
    },
    {
      "epoch": 0.3666697540869726,
      "grad_norm": 0.8569123148918152,
      "learning_rate": 0.0001971617122512952,
      "loss": 0.4324,
      "step": 31670
    },
    {
      "epoch": 0.36678553234844624,
      "grad_norm": 0.8199725151062012,
      "learning_rate": 0.00019715869239900083,
      "loss": 0.4436,
      "step": 31680
    },
    {
      "epoch": 0.3669013106099199,
      "grad_norm": 0.8042467832565308,
      "learning_rate": 0.00019715567096419658,
      "loss": 0.4123,
      "step": 31690
    },
    {
      "epoch": 0.3670170888713935,
      "grad_norm": 0.6773293614387512,
      "learning_rate": 0.00019715264794693162,
      "loss": 0.4533,
      "step": 31700
    },
    {
      "epoch": 0.36713286713286714,
      "grad_norm": 0.7476542592048645,
      "learning_rate": 0.0001971496233472552,
      "loss": 0.4428,
      "step": 31710
    },
    {
      "epoch": 0.36724864539434077,
      "grad_norm": 0.7265835404396057,
      "learning_rate": 0.00019714659716521657,
      "loss": 0.45,
      "step": 31720
    },
    {
      "epoch": 0.3673644236558144,
      "grad_norm": 0.6283068656921387,
      "learning_rate": 0.00019714356940086502,
      "loss": 0.4523,
      "step": 31730
    },
    {
      "epoch": 0.36748020191728803,
      "grad_norm": 0.6916100382804871,
      "learning_rate": 0.0001971405400542499,
      "loss": 0.4246,
      "step": 31740
    },
    {
      "epoch": 0.36759598017876166,
      "grad_norm": 0.8223037719726562,
      "learning_rate": 0.00019713750912542054,
      "loss": 0.445,
      "step": 31750
    },
    {
      "epoch": 0.36771175844023524,
      "grad_norm": 0.7205837965011597,
      "learning_rate": 0.00019713447661442627,
      "loss": 0.4268,
      "step": 31760
    },
    {
      "epoch": 0.36782753670170887,
      "grad_norm": 0.8070530891418457,
      "learning_rate": 0.0001971314425213165,
      "loss": 0.4504,
      "step": 31770
    },
    {
      "epoch": 0.3679433149631825,
      "grad_norm": 0.6983947157859802,
      "learning_rate": 0.0001971284068461407,
      "loss": 0.4372,
      "step": 31780
    },
    {
      "epoch": 0.36805909322465613,
      "grad_norm": 0.8791850209236145,
      "learning_rate": 0.00019712536958894824,
      "loss": 0.4422,
      "step": 31790
    },
    {
      "epoch": 0.36817487148612976,
      "grad_norm": 0.7085838913917542,
      "learning_rate": 0.00019712233074978866,
      "loss": 0.4371,
      "step": 31800
    },
    {
      "epoch": 0.3682906497476034,
      "grad_norm": 0.8415221571922302,
      "learning_rate": 0.0001971192903287114,
      "loss": 0.4451,
      "step": 31810
    },
    {
      "epoch": 0.368406428009077,
      "grad_norm": 0.838901698589325,
      "learning_rate": 0.00019711624832576598,
      "loss": 0.4456,
      "step": 31820
    },
    {
      "epoch": 0.36852220627055066,
      "grad_norm": 0.6288304924964905,
      "learning_rate": 0.000197113204741002,
      "loss": 0.4579,
      "step": 31830
    },
    {
      "epoch": 0.3686379845320243,
      "grad_norm": 0.7653651833534241,
      "learning_rate": 0.00019711015957446897,
      "loss": 0.4392,
      "step": 31840
    },
    {
      "epoch": 0.3687537627934979,
      "grad_norm": 0.8340009450912476,
      "learning_rate": 0.00019710711282621657,
      "loss": 0.4904,
      "step": 31850
    },
    {
      "epoch": 0.3688695410549715,
      "grad_norm": 0.6676978468894958,
      "learning_rate": 0.00019710406449629435,
      "loss": 0.4324,
      "step": 31860
    },
    {
      "epoch": 0.3689853193164451,
      "grad_norm": 0.7971967458724976,
      "learning_rate": 0.00019710101458475195,
      "loss": 0.449,
      "step": 31870
    },
    {
      "epoch": 0.36910109757791876,
      "grad_norm": 0.6887031197547913,
      "learning_rate": 0.00019709796309163915,
      "loss": 0.4467,
      "step": 31880
    },
    {
      "epoch": 0.3692168758393924,
      "grad_norm": 0.818535566329956,
      "learning_rate": 0.00019709491001700555,
      "loss": 0.4103,
      "step": 31890
    },
    {
      "epoch": 0.369332654100866,
      "grad_norm": 0.6199958920478821,
      "learning_rate": 0.00019709185536090094,
      "loss": 0.4392,
      "step": 31900
    },
    {
      "epoch": 0.36944843236233965,
      "grad_norm": 0.8440913558006287,
      "learning_rate": 0.00019708879912337502,
      "loss": 0.4703,
      "step": 31910
    },
    {
      "epoch": 0.3695642106238133,
      "grad_norm": 0.64154052734375,
      "learning_rate": 0.00019708574130447763,
      "loss": 0.4427,
      "step": 31920
    },
    {
      "epoch": 0.3696799888852869,
      "grad_norm": 0.7256616353988647,
      "learning_rate": 0.00019708268190425852,
      "loss": 0.4704,
      "step": 31930
    },
    {
      "epoch": 0.36979576714676055,
      "grad_norm": 0.862727701663971,
      "learning_rate": 0.00019707962092276756,
      "loss": 0.4578,
      "step": 31940
    },
    {
      "epoch": 0.3699115454082342,
      "grad_norm": 0.7893557548522949,
      "learning_rate": 0.00019707655836005458,
      "loss": 0.4331,
      "step": 31950
    },
    {
      "epoch": 0.37002732366970775,
      "grad_norm": 0.7825723886489868,
      "learning_rate": 0.0001970734942161695,
      "loss": 0.4292,
      "step": 31960
    },
    {
      "epoch": 0.3701431019311814,
      "grad_norm": 0.7858340740203857,
      "learning_rate": 0.00019707042849116218,
      "loss": 0.4705,
      "step": 31970
    },
    {
      "epoch": 0.370258880192655,
      "grad_norm": 0.8253327012062073,
      "learning_rate": 0.0001970673611850826,
      "loss": 0.4501,
      "step": 31980
    },
    {
      "epoch": 0.37037465845412865,
      "grad_norm": 0.8884454369544983,
      "learning_rate": 0.00019706429229798068,
      "loss": 0.4702,
      "step": 31990
    },
    {
      "epoch": 0.3704904367156023,
      "grad_norm": 0.6838307976722717,
      "learning_rate": 0.00019706122182990642,
      "loss": 0.4772,
      "step": 32000
    },
    {
      "epoch": 0.3704904367156023,
      "eval_chrf": 75.47916311219912,
      "eval_loss": 0.6640332341194153,
      "eval_runtime": 279.4112,
      "eval_samples_per_second": 0.358,
      "eval_steps_per_second": 0.014,
      "step": 32000
    },
    {
      "epoch": 0.3706062149770759,
      "grad_norm": 0.747809112071991,
      "learning_rate": 0.00019705814978090985,
      "loss": 0.4442,
      "step": 32010
    },
    {
      "epoch": 0.37072199323854954,
      "grad_norm": 0.7655125856399536,
      "learning_rate": 0.000197055076151041,
      "loss": 0.4248,
      "step": 32020
    },
    {
      "epoch": 0.37083777150002317,
      "grad_norm": 0.6934143304824829,
      "learning_rate": 0.0001970520009403499,
      "loss": 0.4384,
      "step": 32030
    },
    {
      "epoch": 0.3709535497614968,
      "grad_norm": 0.8173362016677856,
      "learning_rate": 0.0001970489241488867,
      "loss": 0.4171,
      "step": 32040
    },
    {
      "epoch": 0.3710693280229704,
      "grad_norm": 0.49783217906951904,
      "learning_rate": 0.00019704584577670144,
      "loss": 0.4432,
      "step": 32050
    },
    {
      "epoch": 0.371185106284444,
      "grad_norm": 0.7017600536346436,
      "learning_rate": 0.0001970427658238443,
      "loss": 0.4431,
      "step": 32060
    },
    {
      "epoch": 0.37130088454591764,
      "grad_norm": 0.8327655792236328,
      "learning_rate": 0.00019703968429036548,
      "loss": 0.4391,
      "step": 32070
    },
    {
      "epoch": 0.3714166628073913,
      "grad_norm": 0.7152509689331055,
      "learning_rate": 0.0001970366011763151,
      "loss": 0.479,
      "step": 32080
    },
    {
      "epoch": 0.3715324410688649,
      "grad_norm": 0.7732742428779602,
      "learning_rate": 0.00019703351648174344,
      "loss": 0.4484,
      "step": 32090
    },
    {
      "epoch": 0.37164821933033854,
      "grad_norm": 0.8258805871009827,
      "learning_rate": 0.00019703043020670072,
      "loss": 0.4442,
      "step": 32100
    },
    {
      "epoch": 0.37176399759181217,
      "grad_norm": 0.8022509813308716,
      "learning_rate": 0.00019702734235123716,
      "loss": 0.4484,
      "step": 32110
    },
    {
      "epoch": 0.3718797758532858,
      "grad_norm": 0.7884577512741089,
      "learning_rate": 0.00019702425291540313,
      "loss": 0.4445,
      "step": 32120
    },
    {
      "epoch": 0.37199555411475943,
      "grad_norm": 0.8316375017166138,
      "learning_rate": 0.00019702116189924894,
      "loss": 0.4605,
      "step": 32130
    },
    {
      "epoch": 0.37211133237623306,
      "grad_norm": 0.8084789514541626,
      "learning_rate": 0.0001970180693028249,
      "loss": 0.4421,
      "step": 32140
    },
    {
      "epoch": 0.37222711063770664,
      "grad_norm": 0.7886109352111816,
      "learning_rate": 0.0001970149751261814,
      "loss": 0.4409,
      "step": 32150
    },
    {
      "epoch": 0.37234288889918027,
      "grad_norm": 0.798734188079834,
      "learning_rate": 0.00019701187936936883,
      "loss": 0.4362,
      "step": 32160
    },
    {
      "epoch": 0.3724586671606539,
      "grad_norm": 0.6926020383834839,
      "learning_rate": 0.0001970087820324376,
      "loss": 0.4391,
      "step": 32170
    },
    {
      "epoch": 0.37257444542212753,
      "grad_norm": 0.698354959487915,
      "learning_rate": 0.00019700568311543823,
      "loss": 0.4347,
      "step": 32180
    },
    {
      "epoch": 0.37269022368360116,
      "grad_norm": 0.7995192408561707,
      "learning_rate": 0.00019700258261842111,
      "loss": 0.432,
      "step": 32190
    },
    {
      "epoch": 0.3728060019450748,
      "grad_norm": 0.813062310218811,
      "learning_rate": 0.00019699948054143678,
      "loss": 0.4462,
      "step": 32200
    },
    {
      "epoch": 0.3729217802065484,
      "grad_norm": 0.8295624852180481,
      "learning_rate": 0.00019699637688453577,
      "loss": 0.4467,
      "step": 32210
    },
    {
      "epoch": 0.37303755846802206,
      "grad_norm": 0.7042912244796753,
      "learning_rate": 0.00019699327164776859,
      "loss": 0.4246,
      "step": 32220
    },
    {
      "epoch": 0.3731533367294957,
      "grad_norm": 0.8360041975975037,
      "learning_rate": 0.00019699016483118586,
      "loss": 0.4786,
      "step": 32230
    },
    {
      "epoch": 0.3732691149909693,
      "grad_norm": 0.6229578852653503,
      "learning_rate": 0.00019698705643483822,
      "loss": 0.4356,
      "step": 32240
    },
    {
      "epoch": 0.3733848932524429,
      "grad_norm": 0.7104181051254272,
      "learning_rate": 0.0001969839464587762,
      "loss": 0.4642,
      "step": 32250
    },
    {
      "epoch": 0.3735006715139165,
      "grad_norm": 0.9683583974838257,
      "learning_rate": 0.00019698083490305053,
      "loss": 0.5241,
      "step": 32260
    },
    {
      "epoch": 0.37361644977539016,
      "grad_norm": 0.8482490181922913,
      "learning_rate": 0.00019697772176771188,
      "loss": 0.436,
      "step": 32270
    },
    {
      "epoch": 0.3737322280368638,
      "grad_norm": 0.7092196345329285,
      "learning_rate": 0.0001969746070528109,
      "loss": 0.4535,
      "step": 32280
    },
    {
      "epoch": 0.3738480062983374,
      "grad_norm": 0.6221027374267578,
      "learning_rate": 0.00019697149075839843,
      "loss": 0.4378,
      "step": 32290
    },
    {
      "epoch": 0.37396378455981105,
      "grad_norm": 0.8272190093994141,
      "learning_rate": 0.00019696837288452513,
      "loss": 0.4173,
      "step": 32300
    },
    {
      "epoch": 0.3740795628212847,
      "grad_norm": 0.761194109916687,
      "learning_rate": 0.0001969652534312418,
      "loss": 0.4563,
      "step": 32310
    },
    {
      "epoch": 0.3741953410827583,
      "grad_norm": 0.7332513332366943,
      "learning_rate": 0.0001969621323985993,
      "loss": 0.4401,
      "step": 32320
    },
    {
      "epoch": 0.37431111934423195,
      "grad_norm": 0.9964307546615601,
      "learning_rate": 0.0001969590097866484,
      "loss": 0.4308,
      "step": 32330
    },
    {
      "epoch": 0.3744268976057056,
      "grad_norm": 0.7437024116516113,
      "learning_rate": 0.00019695588559544,
      "loss": 0.4622,
      "step": 32340
    },
    {
      "epoch": 0.37454267586717915,
      "grad_norm": 0.7568386793136597,
      "learning_rate": 0.00019695275982502502,
      "loss": 0.459,
      "step": 32350
    },
    {
      "epoch": 0.3746584541286528,
      "grad_norm": 0.851673424243927,
      "learning_rate": 0.00019694963247545433,
      "loss": 0.4281,
      "step": 32360
    },
    {
      "epoch": 0.3747742323901264,
      "grad_norm": 0.6231158375740051,
      "learning_rate": 0.00019694650354677883,
      "loss": 0.4399,
      "step": 32370
    },
    {
      "epoch": 0.37489001065160005,
      "grad_norm": 0.7695604562759399,
      "learning_rate": 0.00019694337303904954,
      "loss": 0.4552,
      "step": 32380
    },
    {
      "epoch": 0.3750057889130737,
      "grad_norm": 0.8278711438179016,
      "learning_rate": 0.00019694024095231742,
      "loss": 0.4657,
      "step": 32390
    },
    {
      "epoch": 0.3751215671745473,
      "grad_norm": 0.6331350207328796,
      "learning_rate": 0.00019693710728663352,
      "loss": 0.4239,
      "step": 32400
    },
    {
      "epoch": 0.37523734543602094,
      "grad_norm": 0.6943710446357727,
      "learning_rate": 0.00019693397204204884,
      "loss": 0.4451,
      "step": 32410
    },
    {
      "epoch": 0.3753531236974946,
      "grad_norm": 0.8185776472091675,
      "learning_rate": 0.0001969308352186145,
      "loss": 0.4592,
      "step": 32420
    },
    {
      "epoch": 0.3754689019589682,
      "grad_norm": 0.9405238628387451,
      "learning_rate": 0.0001969276968163815,
      "loss": 0.4507,
      "step": 32430
    },
    {
      "epoch": 0.37558468022044184,
      "grad_norm": 0.8298096656799316,
      "learning_rate": 0.00019692455683540107,
      "loss": 0.4665,
      "step": 32440
    },
    {
      "epoch": 0.3757004584819154,
      "grad_norm": 0.7867631316184998,
      "learning_rate": 0.00019692141527572428,
      "loss": 0.4354,
      "step": 32450
    },
    {
      "epoch": 0.37581623674338904,
      "grad_norm": 0.9160256385803223,
      "learning_rate": 0.00019691827213740232,
      "loss": 0.4528,
      "step": 32460
    },
    {
      "epoch": 0.3759320150048627,
      "grad_norm": 0.7393675446510315,
      "learning_rate": 0.0001969151274204864,
      "loss": 0.4249,
      "step": 32470
    },
    {
      "epoch": 0.3760477932663363,
      "grad_norm": 0.7604638338088989,
      "learning_rate": 0.00019691198112502768,
      "loss": 0.4408,
      "step": 32480
    },
    {
      "epoch": 0.37616357152780994,
      "grad_norm": 0.8976063132286072,
      "learning_rate": 0.00019690883325107746,
      "loss": 0.466,
      "step": 32490
    },
    {
      "epoch": 0.37627934978928357,
      "grad_norm": 0.7089973092079163,
      "learning_rate": 0.00019690568379868704,
      "loss": 0.4348,
      "step": 32500
    },
    {
      "epoch": 0.3763951280507572,
      "grad_norm": 0.7004563808441162,
      "learning_rate": 0.00019690253276790766,
      "loss": 0.4449,
      "step": 32510
    },
    {
      "epoch": 0.37651090631223083,
      "grad_norm": 0.7332916259765625,
      "learning_rate": 0.00019689938015879068,
      "loss": 0.4376,
      "step": 32520
    },
    {
      "epoch": 0.37662668457370446,
      "grad_norm": 0.9055464267730713,
      "learning_rate": 0.00019689622597138741,
      "loss": 0.4795,
      "step": 32530
    },
    {
      "epoch": 0.3767424628351781,
      "grad_norm": 0.9704527854919434,
      "learning_rate": 0.00019689307020574924,
      "loss": 0.4415,
      "step": 32540
    },
    {
      "epoch": 0.37685824109665167,
      "grad_norm": 0.6158437728881836,
      "learning_rate": 0.00019688991286192763,
      "loss": 0.4087,
      "step": 32550
    },
    {
      "epoch": 0.3769740193581253,
      "grad_norm": 0.7678256630897522,
      "learning_rate": 0.0001968867539399739,
      "loss": 0.4239,
      "step": 32560
    },
    {
      "epoch": 0.37708979761959893,
      "grad_norm": 0.8120213150978088,
      "learning_rate": 0.0001968835934399396,
      "loss": 0.4626,
      "step": 32570
    },
    {
      "epoch": 0.37720557588107256,
      "grad_norm": 0.6679076552391052,
      "learning_rate": 0.00019688043136187618,
      "loss": 0.446,
      "step": 32580
    },
    {
      "epoch": 0.3773213541425462,
      "grad_norm": 0.8104445934295654,
      "learning_rate": 0.00019687726770583507,
      "loss": 0.4415,
      "step": 32590
    },
    {
      "epoch": 0.3774371324040198,
      "grad_norm": 0.5499981045722961,
      "learning_rate": 0.00019687410247186792,
      "loss": 0.4334,
      "step": 32600
    },
    {
      "epoch": 0.37755291066549346,
      "grad_norm": 0.6923993229866028,
      "learning_rate": 0.0001968709356600262,
      "loss": 0.4479,
      "step": 32610
    },
    {
      "epoch": 0.3776686889269671,
      "grad_norm": 0.6233457326889038,
      "learning_rate": 0.0001968677672703615,
      "loss": 0.4332,
      "step": 32620
    },
    {
      "epoch": 0.3777844671884407,
      "grad_norm": 0.8203330636024475,
      "learning_rate": 0.00019686459730292545,
      "loss": 0.4345,
      "step": 32630
    },
    {
      "epoch": 0.37790024544991435,
      "grad_norm": 0.7847434878349304,
      "learning_rate": 0.00019686142575776967,
      "loss": 0.4357,
      "step": 32640
    },
    {
      "epoch": 0.3780160237113879,
      "grad_norm": 0.7158651351928711,
      "learning_rate": 0.00019685825263494585,
      "loss": 0.4883,
      "step": 32650
    },
    {
      "epoch": 0.37813180197286156,
      "grad_norm": 0.8530229926109314,
      "learning_rate": 0.00019685507793450562,
      "loss": 0.4687,
      "step": 32660
    },
    {
      "epoch": 0.3782475802343352,
      "grad_norm": 0.7206749320030212,
      "learning_rate": 0.00019685190165650069,
      "loss": 0.4869,
      "step": 32670
    },
    {
      "epoch": 0.3783633584958088,
      "grad_norm": 0.581028401851654,
      "learning_rate": 0.00019684872380098284,
      "loss": 0.4264,
      "step": 32680
    },
    {
      "epoch": 0.37847913675728245,
      "grad_norm": 0.7559664845466614,
      "learning_rate": 0.00019684554436800382,
      "loss": 0.4444,
      "step": 32690
    },
    {
      "epoch": 0.3785949150187561,
      "grad_norm": 0.7686482071876526,
      "learning_rate": 0.00019684236335761542,
      "loss": 0.4581,
      "step": 32700
    },
    {
      "epoch": 0.3787106932802297,
      "grad_norm": 0.7029942870140076,
      "learning_rate": 0.0001968391807698694,
      "loss": 0.4493,
      "step": 32710
    },
    {
      "epoch": 0.37882647154170335,
      "grad_norm": 0.7855258584022522,
      "learning_rate": 0.00019683599660481765,
      "loss": 0.4119,
      "step": 32720
    },
    {
      "epoch": 0.378942249803177,
      "grad_norm": 1.0246684551239014,
      "learning_rate": 0.000196832810862512,
      "loss": 0.4379,
      "step": 32730
    },
    {
      "epoch": 0.37905802806465055,
      "grad_norm": 0.8838835954666138,
      "learning_rate": 0.0001968296235430044,
      "loss": 0.4478,
      "step": 32740
    },
    {
      "epoch": 0.3791738063261242,
      "grad_norm": 0.7728408575057983,
      "learning_rate": 0.0001968264346463467,
      "loss": 0.4323,
      "step": 32750
    },
    {
      "epoch": 0.3792895845875978,
      "grad_norm": 0.8997697830200195,
      "learning_rate": 0.0001968232441725909,
      "loss": 0.432,
      "step": 32760
    },
    {
      "epoch": 0.37940536284907145,
      "grad_norm": 0.750637412071228,
      "learning_rate": 0.00019682005212178888,
      "loss": 0.3989,
      "step": 32770
    },
    {
      "epoch": 0.3795211411105451,
      "grad_norm": 0.742658257484436,
      "learning_rate": 0.00019681685849399269,
      "loss": 0.4544,
      "step": 32780
    },
    {
      "epoch": 0.3796369193720187,
      "grad_norm": 0.7581350803375244,
      "learning_rate": 0.00019681366328925433,
      "loss": 0.4632,
      "step": 32790
    },
    {
      "epoch": 0.37975269763349234,
      "grad_norm": 0.7807260155677795,
      "learning_rate": 0.00019681046650762586,
      "loss": 0.4561,
      "step": 32800
    },
    {
      "epoch": 0.379868475894966,
      "grad_norm": 0.7902418375015259,
      "learning_rate": 0.00019680726814915933,
      "loss": 0.4227,
      "step": 32810
    },
    {
      "epoch": 0.3799842541564396,
      "grad_norm": 0.7732905745506287,
      "learning_rate": 0.00019680406821390688,
      "loss": 0.4632,
      "step": 32820
    },
    {
      "epoch": 0.38010003241791324,
      "grad_norm": 0.7653182744979858,
      "learning_rate": 0.00019680086670192056,
      "loss": 0.4516,
      "step": 32830
    },
    {
      "epoch": 0.3802158106793868,
      "grad_norm": 0.8937899470329285,
      "learning_rate": 0.00019679766361325256,
      "loss": 0.4608,
      "step": 32840
    },
    {
      "epoch": 0.38033158894086044,
      "grad_norm": 0.6731113791465759,
      "learning_rate": 0.00019679445894795503,
      "loss": 0.4805,
      "step": 32850
    },
    {
      "epoch": 0.3804473672023341,
      "grad_norm": 0.6992696523666382,
      "learning_rate": 0.0001967912527060802,
      "loss": 0.4506,
      "step": 32860
    },
    {
      "epoch": 0.3805631454638077,
      "grad_norm": 0.6685723662376404,
      "learning_rate": 0.00019678804488768026,
      "loss": 0.4709,
      "step": 32870
    },
    {
      "epoch": 0.38067892372528134,
      "grad_norm": 0.6553595066070557,
      "learning_rate": 0.00019678483549280746,
      "loss": 0.4332,
      "step": 32880
    },
    {
      "epoch": 0.38079470198675497,
      "grad_norm": 0.8781308531761169,
      "learning_rate": 0.0001967816245215141,
      "loss": 0.4297,
      "step": 32890
    },
    {
      "epoch": 0.3809104802482286,
      "grad_norm": 0.8023228645324707,
      "learning_rate": 0.00019677841197385249,
      "loss": 0.4486,
      "step": 32900
    },
    {
      "epoch": 0.38102625850970223,
      "grad_norm": 0.8838029503822327,
      "learning_rate": 0.0001967751978498749,
      "loss": 0.4223,
      "step": 32910
    },
    {
      "epoch": 0.38114203677117586,
      "grad_norm": 1.0198270082473755,
      "learning_rate": 0.0001967719821496337,
      "loss": 0.4528,
      "step": 32920
    },
    {
      "epoch": 0.3812578150326495,
      "grad_norm": 0.7790926098823547,
      "learning_rate": 0.0001967687648731813,
      "loss": 0.4405,
      "step": 32930
    },
    {
      "epoch": 0.38137359329412307,
      "grad_norm": 0.7534083724021912,
      "learning_rate": 0.0001967655460205701,
      "loss": 0.4656,
      "step": 32940
    },
    {
      "epoch": 0.3814893715555967,
      "grad_norm": 0.7717673778533936,
      "learning_rate": 0.00019676232559185245,
      "loss": 0.4749,
      "step": 32950
    },
    {
      "epoch": 0.38160514981707033,
      "grad_norm": 0.8359370827674866,
      "learning_rate": 0.0001967591035870809,
      "loss": 0.466,
      "step": 32960
    },
    {
      "epoch": 0.38172092807854396,
      "grad_norm": 0.653666615486145,
      "learning_rate": 0.00019675588000630788,
      "loss": 0.4603,
      "step": 32970
    },
    {
      "epoch": 0.3818367063400176,
      "grad_norm": 0.7439927458763123,
      "learning_rate": 0.00019675265484958595,
      "loss": 0.4681,
      "step": 32980
    },
    {
      "epoch": 0.3819524846014912,
      "grad_norm": 0.5780789852142334,
      "learning_rate": 0.00019674942811696757,
      "loss": 0.3726,
      "step": 32990
    },
    {
      "epoch": 0.38206826286296486,
      "grad_norm": 0.8285802602767944,
      "learning_rate": 0.00019674619980850533,
      "loss": 0.4609,
      "step": 33000
    },
    {
      "epoch": 0.38206826286296486,
      "eval_chrf": 70.65395075493964,
      "eval_loss": 0.6838914752006531,
      "eval_runtime": 348.4953,
      "eval_samples_per_second": 0.287,
      "eval_steps_per_second": 0.011,
      "step": 33000
    }
  ],
  "logging_steps": 10,
  "max_steps": 259116,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 1000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 5.593740795633992e+18,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}