{
  "best_global_step": 39000,
  "best_metric": 84.03459253857896,
  "best_model_checkpoint": "checkpoints_7B_lora_translated/en-chv-final/checkpoint-39000",
  "epoch": 0.7027900766041183,
  "eval_steps": 1000,
  "global_step": 50000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.00014055801532082368,
      "grad_norm": 2.6348493099212646,
      "learning_rate": 5.05902192242833e-07,
      "loss": 2.5918,
      "step": 10
    },
    {
      "epoch": 0.00028111603064164735,
      "grad_norm": 2.4136617183685303,
      "learning_rate": 1.0680157391793143e-06,
      "loss": 2.4399,
      "step": 20
    },
    {
      "epoch": 0.000421674045962471,
      "grad_norm": 2.4269416332244873,
      "learning_rate": 1.6301292861157953e-06,
      "loss": 2.5371,
      "step": 30
    },
    {
      "epoch": 0.0005622320612832947,
      "grad_norm": 2.088914394378662,
      "learning_rate": 2.1922428330522765e-06,
      "loss": 2.5674,
      "step": 40
    },
    {
      "epoch": 0.0007027900766041184,
      "grad_norm": 2.4543466567993164,
      "learning_rate": 2.7543563799887578e-06,
      "loss": 2.4914,
      "step": 50
    },
    {
      "epoch": 0.000843348091924942,
      "grad_norm": 2.4548532962799072,
      "learning_rate": 3.316469926925239e-06,
      "loss": 2.4309,
      "step": 60
    },
    {
      "epoch": 0.0009839061072457658,
      "grad_norm": 2.680846691131592,
      "learning_rate": 3.878583473861721e-06,
      "loss": 2.3076,
      "step": 70
    },
    {
      "epoch": 0.0011244641225665894,
      "grad_norm": 2.1085638999938965,
      "learning_rate": 4.440697020798201e-06,
      "loss": 2.1619,
      "step": 80
    },
    {
      "epoch": 0.001265022137887413,
      "grad_norm": 2.349139928817749,
      "learning_rate": 5.002810567734682e-06,
      "loss": 2.1182,
      "step": 90
    },
    {
      "epoch": 0.0014055801532082367,
      "grad_norm": 2.2252542972564697,
      "learning_rate": 5.564924114671164e-06,
      "loss": 1.9373,
      "step": 100
    },
    {
      "epoch": 0.0015461381685290604,
      "grad_norm": 1.816213846206665,
      "learning_rate": 6.1270376616076454e-06,
      "loss": 1.7312,
      "step": 110
    },
    {
      "epoch": 0.001686696183849884,
      "grad_norm": 1.467105507850647,
      "learning_rate": 6.689151208544127e-06,
      "loss": 1.6066,
      "step": 120
    },
    {
      "epoch": 0.0018272541991707076,
      "grad_norm": 1.057456612586975,
      "learning_rate": 7.251264755480608e-06,
      "loss": 1.399,
      "step": 130
    },
    {
      "epoch": 0.0019678122144915315,
      "grad_norm": 0.7919467091560364,
      "learning_rate": 7.813378302417088e-06,
      "loss": 1.2557,
      "step": 140
    },
    {
      "epoch": 0.002108370229812355,
      "grad_norm": 0.7170568108558655,
      "learning_rate": 8.37549184935357e-06,
      "loss": 1.1852,
      "step": 150
    },
    {
      "epoch": 0.002248928245133179,
      "grad_norm": 0.6167587637901306,
      "learning_rate": 8.93760539629005e-06,
      "loss": 1.0936,
      "step": 160
    },
    {
      "epoch": 0.0023894862604540022,
      "grad_norm": 0.5959345698356628,
      "learning_rate": 9.499718943226533e-06,
      "loss": 1.04,
      "step": 170
    },
    {
      "epoch": 0.002530044275774826,
      "grad_norm": 0.6112813949584961,
      "learning_rate": 1.0061832490163013e-05,
      "loss": 1.0254,
      "step": 180
    },
    {
      "epoch": 0.0026706022910956495,
      "grad_norm": 0.45967984199523926,
      "learning_rate": 1.0623946037099493e-05,
      "loss": 0.9459,
      "step": 190
    },
    {
      "epoch": 0.0028111603064164734,
      "grad_norm": 0.4986984431743622,
      "learning_rate": 1.1186059584035975e-05,
      "loss": 0.8789,
      "step": 200
    },
    {
      "epoch": 0.0029517183217372973,
      "grad_norm": 0.5307064056396484,
      "learning_rate": 1.1748173130972456e-05,
      "loss": 0.9222,
      "step": 210
    },
    {
      "epoch": 0.0030922763370581207,
      "grad_norm": 0.5603976249694824,
      "learning_rate": 1.231028667790894e-05,
      "loss": 0.8335,
      "step": 220
    },
    {
      "epoch": 0.0032328343523789446,
      "grad_norm": 0.5682457089424133,
      "learning_rate": 1.2872400224845418e-05,
      "loss": 0.8723,
      "step": 230
    },
    {
      "epoch": 0.003373392367699768,
      "grad_norm": 0.5278539657592773,
      "learning_rate": 1.34345137717819e-05,
      "loss": 0.8296,
      "step": 240
    },
    {
      "epoch": 0.003513950383020592,
      "grad_norm": 0.5293506383895874,
      "learning_rate": 1.399662731871838e-05,
      "loss": 0.8298,
      "step": 250
    },
    {
      "epoch": 0.0036545083983414153,
      "grad_norm": 0.5066598653793335,
      "learning_rate": 1.4558740865654863e-05,
      "loss": 0.7801,
      "step": 260
    },
    {
      "epoch": 0.003795066413662239,
      "grad_norm": 0.617091178894043,
      "learning_rate": 1.5120854412591345e-05,
      "loss": 0.7985,
      "step": 270
    },
    {
      "epoch": 0.003935624428983063,
      "grad_norm": 0.5896926522254944,
      "learning_rate": 1.5682967959527825e-05,
      "loss": 0.8191,
      "step": 280
    },
    {
      "epoch": 0.0040761824443038865,
      "grad_norm": 0.5436782836914062,
      "learning_rate": 1.624508150646431e-05,
      "loss": 0.7488,
      "step": 290
    },
    {
      "epoch": 0.00421674045962471,
      "grad_norm": 0.6122232675552368,
      "learning_rate": 1.6807195053400786e-05,
      "loss": 0.7994,
      "step": 300
    },
    {
      "epoch": 0.004357298474945534,
      "grad_norm": 0.6472449898719788,
      "learning_rate": 1.736930860033727e-05,
      "loss": 0.7276,
      "step": 310
    },
    {
      "epoch": 0.004497856490266358,
      "grad_norm": 0.6526391506195068,
      "learning_rate": 1.793142214727375e-05,
      "loss": 0.7184,
      "step": 320
    },
    {
      "epoch": 0.004638414505587181,
      "grad_norm": 1.2312449216842651,
      "learning_rate": 1.849353569421023e-05,
      "loss": 0.7544,
      "step": 330
    },
    {
      "epoch": 0.0047789725209080045,
      "grad_norm": 0.6525812745094299,
      "learning_rate": 1.905564924114671e-05,
      "loss": 0.694,
      "step": 340
    },
    {
      "epoch": 0.004919530536228829,
      "grad_norm": 0.7462595701217651,
      "learning_rate": 1.9617762788083194e-05,
      "loss": 0.7105,
      "step": 350
    },
    {
      "epoch": 0.005060088551549652,
      "grad_norm": 0.7727270126342773,
      "learning_rate": 2.0179876335019675e-05,
      "loss": 0.6715,
      "step": 360
    },
    {
      "epoch": 0.005200646566870476,
      "grad_norm": 0.715003252029419,
      "learning_rate": 2.0741989881956155e-05,
      "loss": 0.6575,
      "step": 370
    },
    {
      "epoch": 0.005341204582191299,
      "grad_norm": 0.7247048020362854,
      "learning_rate": 2.130410342889264e-05,
      "loss": 0.6635,
      "step": 380
    },
    {
      "epoch": 0.005481762597512123,
      "grad_norm": 0.8433679938316345,
      "learning_rate": 2.186621697582912e-05,
      "loss": 0.6494,
      "step": 390
    },
    {
      "epoch": 0.005622320612832947,
      "grad_norm": 0.9507123827934265,
      "learning_rate": 2.24283305227656e-05,
      "loss": 0.6629,
      "step": 400
    },
    {
      "epoch": 0.00576287862815377,
      "grad_norm": 0.9074307084083557,
      "learning_rate": 2.299044406970208e-05,
      "loss": 0.7001,
      "step": 410
    },
    {
      "epoch": 0.0059034366434745945,
      "grad_norm": 0.9715143442153931,
      "learning_rate": 2.3552557616638564e-05,
      "loss": 0.6768,
      "step": 420
    },
    {
      "epoch": 0.006043994658795418,
      "grad_norm": 0.9060778617858887,
      "learning_rate": 2.411467116357504e-05,
      "loss": 0.6512,
      "step": 430
    },
    {
      "epoch": 0.006184552674116241,
      "grad_norm": 0.9930410385131836,
      "learning_rate": 2.4676784710511524e-05,
      "loss": 0.6667,
      "step": 440
    },
    {
      "epoch": 0.006325110689437065,
      "grad_norm": 1.1297030448913574,
      "learning_rate": 2.5238898257448008e-05,
      "loss": 0.631,
      "step": 450
    },
    {
      "epoch": 0.006465668704757889,
      "grad_norm": 1.0275930166244507,
      "learning_rate": 2.580101180438449e-05,
      "loss": 0.5607,
      "step": 460
    },
    {
      "epoch": 0.006606226720078713,
      "grad_norm": 1.1458061933517456,
      "learning_rate": 2.6363125351320966e-05,
      "loss": 0.6091,
      "step": 470
    },
    {
      "epoch": 0.006746784735399536,
      "grad_norm": 1.185503363609314,
      "learning_rate": 2.692523889825745e-05,
      "loss": 0.6215,
      "step": 480
    },
    {
      "epoch": 0.006887342750720359,
      "grad_norm": 1.04978346824646,
      "learning_rate": 2.748735244519393e-05,
      "loss": 0.6037,
      "step": 490
    },
    {
      "epoch": 0.007027900766041184,
      "grad_norm": 1.1438580751419067,
      "learning_rate": 2.8049465992130413e-05,
      "loss": 0.5887,
      "step": 500
    },
    {
      "epoch": 0.007168458781362007,
      "grad_norm": 0.9910487532615662,
      "learning_rate": 2.861157953906689e-05,
      "loss": 0.5972,
      "step": 510
    },
    {
      "epoch": 0.007309016796682831,
      "grad_norm": 1.381588339805603,
      "learning_rate": 2.9173693086003374e-05,
      "loss": 0.5942,
      "step": 520
    },
    {
      "epoch": 0.007449574812003655,
      "grad_norm": 0.9777755737304688,
      "learning_rate": 2.9735806632939855e-05,
      "loss": 0.5574,
      "step": 530
    },
    {
      "epoch": 0.007590132827324478,
      "grad_norm": 1.1398427486419678,
      "learning_rate": 3.0297920179876338e-05,
      "loss": 0.566,
      "step": 540
    },
    {
      "epoch": 0.007730690842645302,
      "grad_norm": 1.6577353477478027,
      "learning_rate": 3.086003372681282e-05,
      "loss": 0.5462,
      "step": 550
    },
    {
      "epoch": 0.007871248857966126,
      "grad_norm": 1.4741021394729614,
      "learning_rate": 3.14221472737493e-05,
      "loss": 0.573,
      "step": 560
    },
    {
      "epoch": 0.00801180687328695,
      "grad_norm": 1.2243646383285522,
      "learning_rate": 3.198426082068578e-05,
      "loss": 0.5566,
      "step": 570
    },
    {
      "epoch": 0.008152364888607773,
      "grad_norm": 1.2751414775848389,
      "learning_rate": 3.254637436762226e-05,
      "loss": 0.5994,
      "step": 580
    },
    {
      "epoch": 0.008292922903928596,
      "grad_norm": 1.2607536315917969,
      "learning_rate": 3.310848791455875e-05,
      "loss": 0.5064,
      "step": 590
    },
    {
      "epoch": 0.00843348091924942,
      "grad_norm": 1.5125900506973267,
      "learning_rate": 3.367060146149522e-05,
      "loss": 0.5533,
      "step": 600
    },
    {
      "epoch": 0.008574038934570243,
      "grad_norm": 1.3755593299865723,
      "learning_rate": 3.42327150084317e-05,
      "loss": 0.5528,
      "step": 610
    },
    {
      "epoch": 0.008714596949891068,
      "grad_norm": 1.279312252998352,
      "learning_rate": 3.479482855536819e-05,
      "loss": 0.5626,
      "step": 620
    },
    {
      "epoch": 0.008855154965211892,
      "grad_norm": 1.4999586343765259,
      "learning_rate": 3.535694210230467e-05,
      "loss": 0.5571,
      "step": 630
    },
    {
      "epoch": 0.008995712980532715,
      "grad_norm": 1.3935269117355347,
      "learning_rate": 3.591905564924115e-05,
      "loss": 0.5743,
      "step": 640
    },
    {
      "epoch": 0.009136270995853539,
      "grad_norm": 1.4465786218643188,
      "learning_rate": 3.648116919617763e-05,
      "loss": 0.5431,
      "step": 650
    },
    {
      "epoch": 0.009276829011174362,
      "grad_norm": 1.6643967628479004,
      "learning_rate": 3.704328274311411e-05,
      "loss": 0.546,
      "step": 660
    },
    {
      "epoch": 0.009417387026495186,
      "grad_norm": 1.3728482723236084,
      "learning_rate": 3.760539629005059e-05,
      "loss": 0.546,
      "step": 670
    },
    {
      "epoch": 0.009557945041816009,
      "grad_norm": 1.3901290893554688,
      "learning_rate": 3.816750983698708e-05,
      "loss": 0.5309,
      "step": 680
    },
    {
      "epoch": 0.009698503057136832,
      "grad_norm": 1.4969929456710815,
      "learning_rate": 3.872962338392355e-05,
      "loss": 0.5615,
      "step": 690
    },
    {
      "epoch": 0.009839061072457658,
      "grad_norm": 1.3987517356872559,
      "learning_rate": 3.929173693086004e-05,
      "loss": 0.504,
      "step": 700
    },
    {
      "epoch": 0.009979619087778481,
      "grad_norm": 1.2550172805786133,
      "learning_rate": 3.985385047779652e-05,
      "loss": 0.5662,
      "step": 710
    },
    {
      "epoch": 0.010120177103099304,
      "grad_norm": 1.4774284362792969,
      "learning_rate": 4.0415964024733e-05,
      "loss": 0.4979,
      "step": 720
    },
    {
      "epoch": 0.010260735118420128,
      "grad_norm": 1.6103997230529785,
      "learning_rate": 4.097807757166948e-05,
      "loss": 0.5223,
      "step": 730
    },
    {
      "epoch": 0.010401293133740951,
      "grad_norm": 1.2553802728652954,
      "learning_rate": 4.154019111860596e-05,
      "loss": 0.5115,
      "step": 740
    },
    {
      "epoch": 0.010541851149061775,
      "grad_norm": 1.5846428871154785,
      "learning_rate": 4.210230466554244e-05,
      "loss": 0.52,
      "step": 750
    },
    {
      "epoch": 0.010682409164382598,
      "grad_norm": 1.5420085191726685,
      "learning_rate": 4.2664418212478927e-05,
      "loss": 0.5173,
      "step": 760
    },
    {
      "epoch": 0.010822967179703423,
      "grad_norm": 1.3390594720840454,
      "learning_rate": 4.322653175941541e-05,
      "loss": 0.4842,
      "step": 770
    },
    {
      "epoch": 0.010963525195024247,
      "grad_norm": 1.249771237373352,
      "learning_rate": 4.378864530635188e-05,
      "loss": 0.5459,
      "step": 780
    },
    {
      "epoch": 0.01110408321034507,
      "grad_norm": 1.3949081897735596,
      "learning_rate": 4.435075885328837e-05,
      "loss": 0.5317,
      "step": 790
    },
    {
      "epoch": 0.011244641225665894,
      "grad_norm": 1.27957022190094,
      "learning_rate": 4.491287240022485e-05,
      "loss": 0.5237,
      "step": 800
    },
    {
      "epoch": 0.011385199240986717,
      "grad_norm": 1.7414036989212036,
      "learning_rate": 4.547498594716133e-05,
      "loss": 0.4906,
      "step": 810
    },
    {
      "epoch": 0.01152575725630754,
      "grad_norm": 1.4559071063995361,
      "learning_rate": 4.603709949409781e-05,
      "loss": 0.5103,
      "step": 820
    },
    {
      "epoch": 0.011666315271628364,
      "grad_norm": 1.3347688913345337,
      "learning_rate": 4.659921304103429e-05,
      "loss": 0.5145,
      "step": 830
    },
    {
      "epoch": 0.011806873286949189,
      "grad_norm": 1.288408875465393,
      "learning_rate": 4.716132658797077e-05,
      "loss": 0.5062,
      "step": 840
    },
    {
      "epoch": 0.011947431302270013,
      "grad_norm": 1.7022039890289307,
      "learning_rate": 4.7723440134907257e-05,
      "loss": 0.5169,
      "step": 850
    },
    {
      "epoch": 0.012087989317590836,
      "grad_norm": 1.4527498483657837,
      "learning_rate": 4.828555368184374e-05,
      "loss": 0.4784,
      "step": 860
    },
    {
      "epoch": 0.01222854733291166,
      "grad_norm": 1.1472194194793701,
      "learning_rate": 4.884766722878022e-05,
      "loss": 0.4804,
      "step": 870
    },
    {
      "epoch": 0.012369105348232483,
      "grad_norm": 1.1323292255401611,
      "learning_rate": 4.94097807757167e-05,
      "loss": 0.4463,
      "step": 880
    },
    {
      "epoch": 0.012509663363553306,
      "grad_norm": 1.5541162490844727,
      "learning_rate": 4.997189432265318e-05,
      "loss": 0.5003,
      "step": 890
    },
    {
      "epoch": 0.01265022137887413,
      "grad_norm": 1.435592532157898,
      "learning_rate": 5.053400786958966e-05,
      "loss": 0.5199,
      "step": 900
    },
    {
      "epoch": 0.012790779394194953,
      "grad_norm": 1.5248143672943115,
      "learning_rate": 5.1096121416526146e-05,
      "loss": 0.4724,
      "step": 910
    },
    {
      "epoch": 0.012931337409515778,
      "grad_norm": 1.5094494819641113,
      "learning_rate": 5.1658234963462626e-05,
      "loss": 0.4463,
      "step": 920
    },
    {
      "epoch": 0.013071895424836602,
      "grad_norm": 1.7850382328033447,
      "learning_rate": 5.2220348510399106e-05,
      "loss": 0.5005,
      "step": 930
    },
    {
      "epoch": 0.013212453440157425,
      "grad_norm": 1.5300614833831787,
      "learning_rate": 5.278246205733558e-05,
      "loss": 0.4515,
      "step": 940
    },
    {
      "epoch": 0.013353011455478249,
      "grad_norm": 1.3888568878173828,
      "learning_rate": 5.334457560427206e-05,
      "loss": 0.4704,
      "step": 950
    },
    {
      "epoch": 0.013493569470799072,
      "grad_norm": 1.5508018732070923,
      "learning_rate": 5.390668915120855e-05,
      "loss": 0.5048,
      "step": 960
    },
    {
      "epoch": 0.013634127486119895,
      "grad_norm": 1.7522460222244263,
      "learning_rate": 5.446880269814503e-05,
      "loss": 0.4869,
      "step": 970
    },
    {
      "epoch": 0.013774685501440719,
      "grad_norm": 1.1714389324188232,
      "learning_rate": 5.503091624508151e-05,
      "loss": 0.4906,
      "step": 980
    },
    {
      "epoch": 0.013915243516761544,
      "grad_norm": 1.3763507604599,
      "learning_rate": 5.5593029792017995e-05,
      "loss": 0.4758,
      "step": 990
    },
    {
      "epoch": 0.014055801532082367,
      "grad_norm": 1.4176082611083984,
      "learning_rate": 5.6155143338954476e-05,
      "loss": 0.5115,
      "step": 1000
    },
    {
      "epoch": 0.014055801532082367,
      "eval_chrf": 79.5330761969255,
      "eval_loss": 0.8439362049102783,
      "eval_runtime": 327.6679,
      "eval_samples_per_second": 0.305,
      "eval_steps_per_second": 0.012,
      "step": 1000
    },
    {
      "epoch": 0.014196359547403191,
      "grad_norm": 1.3907123804092407,
      "learning_rate": 5.6717256885890956e-05,
      "loss": 0.5051,
      "step": 1010
    },
    {
      "epoch": 0.014336917562724014,
      "grad_norm": 1.5536024570465088,
      "learning_rate": 5.7279370432827436e-05,
      "loss": 0.4984,
      "step": 1020
    },
    {
      "epoch": 0.014477475578044838,
      "grad_norm": 1.141889214515686,
      "learning_rate": 5.784148397976391e-05,
      "loss": 0.4467,
      "step": 1030
    },
    {
      "epoch": 0.014618033593365661,
      "grad_norm": 1.4408601522445679,
      "learning_rate": 5.840359752670039e-05,
      "loss": 0.4604,
      "step": 1040
    },
    {
      "epoch": 0.014758591608686485,
      "grad_norm": 1.8335459232330322,
      "learning_rate": 5.896571107363688e-05,
      "loss": 0.4617,
      "step": 1050
    },
    {
      "epoch": 0.01489914962400731,
      "grad_norm": 1.133646845817566,
      "learning_rate": 5.952782462057336e-05,
      "loss": 0.4644,
      "step": 1060
    },
    {
      "epoch": 0.015039707639328133,
      "grad_norm": 1.1528613567352295,
      "learning_rate": 6.008993816750984e-05,
      "loss": 0.4685,
      "step": 1070
    },
    {
      "epoch": 0.015180265654648957,
      "grad_norm": 1.8868284225463867,
      "learning_rate": 6.0652051714446325e-05,
      "loss": 0.4715,
      "step": 1080
    },
    {
      "epoch": 0.01532082366996978,
      "grad_norm": 1.3340322971343994,
      "learning_rate": 6.12141652613828e-05,
      "loss": 0.4428,
      "step": 1090
    },
    {
      "epoch": 0.015461381685290604,
      "grad_norm": 1.3002533912658691,
      "learning_rate": 6.177627880831929e-05,
      "loss": 0.4071,
      "step": 1100
    },
    {
      "epoch": 0.015601939700611427,
      "grad_norm": 1.4698681831359863,
      "learning_rate": 6.233839235525576e-05,
      "loss": 0.4395,
      "step": 1110
    },
    {
      "epoch": 0.015742497715932252,
      "grad_norm": 1.5502619743347168,
      "learning_rate": 6.290050590219225e-05,
      "loss": 0.4712,
      "step": 1120
    },
    {
      "epoch": 0.015883055731253076,
      "grad_norm": 1.1336027383804321,
      "learning_rate": 6.346261944912872e-05,
      "loss": 0.4367,
      "step": 1130
    },
    {
      "epoch": 0.0160236137465739,
      "grad_norm": 1.5595983266830444,
      "learning_rate": 6.402473299606521e-05,
      "loss": 0.4312,
      "step": 1140
    },
    {
      "epoch": 0.016164171761894722,
      "grad_norm": 1.5467026233673096,
      "learning_rate": 6.45868465430017e-05,
      "loss": 0.439,
      "step": 1150
    },
    {
      "epoch": 0.016304729777215546,
      "grad_norm": 1.471920371055603,
      "learning_rate": 6.514896008993817e-05,
      "loss": 0.4728,
      "step": 1160
    },
    {
      "epoch": 0.01644528779253637,
      "grad_norm": 1.8791810274124146,
      "learning_rate": 6.571107363687466e-05,
      "loss": 0.4584,
      "step": 1170
    },
    {
      "epoch": 0.016585845807857193,
      "grad_norm": 1.5976896286010742,
      "learning_rate": 6.627318718381114e-05,
      "loss": 0.463,
      "step": 1180
    },
    {
      "epoch": 0.016726403823178016,
      "grad_norm": 1.1342140436172485,
      "learning_rate": 6.683530073074762e-05,
      "loss": 0.4611,
      "step": 1190
    },
    {
      "epoch": 0.01686696183849884,
      "grad_norm": 1.3907729387283325,
      "learning_rate": 6.739741427768409e-05,
      "loss": 0.4471,
      "step": 1200
    },
    {
      "epoch": 0.017007519853819663,
      "grad_norm": 1.6492199897766113,
      "learning_rate": 6.795952782462058e-05,
      "loss": 0.4637,
      "step": 1210
    },
    {
      "epoch": 0.017148077869140486,
      "grad_norm": 1.5096964836120605,
      "learning_rate": 6.852164137155705e-05,
      "loss": 0.4498,
      "step": 1220
    },
    {
      "epoch": 0.01728863588446131,
      "grad_norm": 1.5577164888381958,
      "learning_rate": 6.908375491849354e-05,
      "loss": 0.4747,
      "step": 1230
    },
    {
      "epoch": 0.017429193899782137,
      "grad_norm": 1.2539633512496948,
      "learning_rate": 6.964586846543002e-05,
      "loss": 0.4391,
      "step": 1240
    },
    {
      "epoch": 0.01756975191510296,
      "grad_norm": 1.2233649492263794,
      "learning_rate": 7.02079820123665e-05,
      "loss": 0.4138,
      "step": 1250
    },
    {
      "epoch": 0.017710309930423784,
      "grad_norm": 1.5597176551818848,
      "learning_rate": 7.077009555930299e-05,
      "loss": 0.4144,
      "step": 1260
    },
    {
      "epoch": 0.017850867945744607,
      "grad_norm": 1.2913601398468018,
      "learning_rate": 7.133220910623947e-05,
      "loss": 0.4579,
      "step": 1270
    },
    {
      "epoch": 0.01799142596106543,
      "grad_norm": 1.2796281576156616,
      "learning_rate": 7.189432265317595e-05,
      "loss": 0.441,
      "step": 1280
    },
    {
      "epoch": 0.018131983976386254,
      "grad_norm": 1.7188844680786133,
      "learning_rate": 7.245643620011242e-05,
      "loss": 0.4527,
      "step": 1290
    },
    {
      "epoch": 0.018272541991707077,
      "grad_norm": 1.4409807920455933,
      "learning_rate": 7.30185497470489e-05,
      "loss": 0.4398,
      "step": 1300
    },
    {
      "epoch": 0.0184131000070279,
      "grad_norm": 1.4035305976867676,
      "learning_rate": 7.358066329398538e-05,
      "loss": 0.4337,
      "step": 1310
    },
    {
      "epoch": 0.018553658022348724,
      "grad_norm": 1.3769419193267822,
      "learning_rate": 7.414277684092187e-05,
      "loss": 0.4226,
      "step": 1320
    },
    {
      "epoch": 0.018694216037669548,
      "grad_norm": 1.3842592239379883,
      "learning_rate": 7.470489038785835e-05,
      "loss": 0.4376,
      "step": 1330
    },
    {
      "epoch": 0.01883477405299037,
      "grad_norm": 1.2802128791809082,
      "learning_rate": 7.526700393479483e-05,
      "loss": 0.4434,
      "step": 1340
    },
    {
      "epoch": 0.018975332068311195,
      "grad_norm": 1.3373446464538574,
      "learning_rate": 7.582911748173132e-05,
      "loss": 0.4339,
      "step": 1350
    },
    {
      "epoch": 0.019115890083632018,
      "grad_norm": 1.5376042127609253,
      "learning_rate": 7.63912310286678e-05,
      "loss": 0.4371,
      "step": 1360
    },
    {
      "epoch": 0.01925644809895284,
      "grad_norm": 1.5754722356796265,
      "learning_rate": 7.695334457560428e-05,
      "loss": 0.4052,
      "step": 1370
    },
    {
      "epoch": 0.019397006114273665,
      "grad_norm": 1.1984570026397705,
      "learning_rate": 7.751545812254075e-05,
      "loss": 0.433,
      "step": 1380
    },
    {
      "epoch": 0.01953756412959449,
      "grad_norm": 1.137647271156311,
      "learning_rate": 7.807757166947724e-05,
      "loss": 0.4159,
      "step": 1390
    },
    {
      "epoch": 0.019678122144915315,
      "grad_norm": 1.257773995399475,
      "learning_rate": 7.863968521641371e-05,
      "loss": 0.4287,
      "step": 1400
    },
    {
      "epoch": 0.01981868016023614,
      "grad_norm": 1.1570932865142822,
      "learning_rate": 7.92017987633502e-05,
      "loss": 0.4532,
      "step": 1410
    },
    {
      "epoch": 0.019959238175556962,
      "grad_norm": 1.426266074180603,
      "learning_rate": 7.976391231028668e-05,
      "loss": 0.4345,
      "step": 1420
    },
    {
      "epoch": 0.020099796190877785,
      "grad_norm": 1.1181676387786865,
      "learning_rate": 8.032602585722316e-05,
      "loss": 0.4368,
      "step": 1430
    },
    {
      "epoch": 0.02024035420619861,
      "grad_norm": 1.487839937210083,
      "learning_rate": 8.088813940415965e-05,
      "loss": 0.4388,
      "step": 1440
    },
    {
      "epoch": 0.020380912221519432,
      "grad_norm": 1.7598848342895508,
      "learning_rate": 8.145025295109613e-05,
      "loss": 0.437,
      "step": 1450
    },
    {
      "epoch": 0.020521470236840256,
      "grad_norm": 2.225933313369751,
      "learning_rate": 8.20123664980326e-05,
      "loss": 0.4562,
      "step": 1460
    },
    {
      "epoch": 0.02066202825216108,
      "grad_norm": 1.6490896940231323,
      "learning_rate": 8.257448004496908e-05,
      "loss": 0.4389,
      "step": 1470
    },
    {
      "epoch": 0.020802586267481903,
      "grad_norm": 1.0282951593399048,
      "learning_rate": 8.313659359190557e-05,
      "loss": 0.4185,
      "step": 1480
    },
    {
      "epoch": 0.020943144282802726,
      "grad_norm": 1.0933427810668945,
      "learning_rate": 8.369870713884205e-05,
      "loss": 0.3744,
      "step": 1490
    },
    {
      "epoch": 0.02108370229812355,
      "grad_norm": 1.3399862051010132,
      "learning_rate": 8.426082068577853e-05,
      "loss": 0.4078,
      "step": 1500
    },
    {
      "epoch": 0.021224260313444373,
      "grad_norm": 0.9295111894607544,
      "learning_rate": 8.482293423271501e-05,
      "loss": 0.3846,
      "step": 1510
    },
    {
      "epoch": 0.021364818328765196,
      "grad_norm": 1.304292917251587,
      "learning_rate": 8.538504777965149e-05,
      "loss": 0.3944,
      "step": 1520
    },
    {
      "epoch": 0.021505376344086023,
      "grad_norm": 1.3154966831207275,
      "learning_rate": 8.594716132658798e-05,
      "loss": 0.4197,
      "step": 1530
    },
    {
      "epoch": 0.021645934359406847,
      "grad_norm": 1.1910078525543213,
      "learning_rate": 8.650927487352446e-05,
      "loss": 0.4154,
      "step": 1540
    },
    {
      "epoch": 0.02178649237472767,
      "grad_norm": 1.4511432647705078,
      "learning_rate": 8.707138842046094e-05,
      "loss": 0.3892,
      "step": 1550
    },
    {
      "epoch": 0.021927050390048494,
      "grad_norm": 1.166874647140503,
      "learning_rate": 8.763350196739741e-05,
      "loss": 0.4098,
      "step": 1560
    },
    {
      "epoch": 0.022067608405369317,
      "grad_norm": 1.02645742893219,
      "learning_rate": 8.81956155143339e-05,
      "loss": 0.4274,
      "step": 1570
    },
    {
      "epoch": 0.02220816642069014,
      "grad_norm": 1.072724461555481,
      "learning_rate": 8.875772906127038e-05,
      "loss": 0.3831,
      "step": 1580
    },
    {
      "epoch": 0.022348724436010964,
      "grad_norm": 1.0810933113098145,
      "learning_rate": 8.931984260820686e-05,
      "loss": 0.4282,
      "step": 1590
    },
    {
      "epoch": 0.022489282451331787,
      "grad_norm": 1.1835025548934937,
      "learning_rate": 8.988195615514334e-05,
      "loss": 0.3466,
      "step": 1600
    },
    {
      "epoch": 0.02262984046665261,
      "grad_norm": 1.2185114622116089,
      "learning_rate": 9.044406970207983e-05,
      "loss": 0.3761,
      "step": 1610
    },
    {
      "epoch": 0.022770398481973434,
      "grad_norm": 1.1819682121276855,
      "learning_rate": 9.10061832490163e-05,
      "loss": 0.3896,
      "step": 1620
    },
    {
      "epoch": 0.022910956497294258,
      "grad_norm": 1.5933457612991333,
      "learning_rate": 9.156829679595279e-05,
      "loss": 0.416,
      "step": 1630
    },
    {
      "epoch": 0.02305151451261508,
      "grad_norm": 1.5445760488510132,
      "learning_rate": 9.213041034288927e-05,
      "loss": 0.4051,
      "step": 1640
    },
    {
      "epoch": 0.023192072527935904,
      "grad_norm": 1.101395606994629,
      "learning_rate": 9.269252388982574e-05,
      "loss": 0.4149,
      "step": 1650
    },
    {
      "epoch": 0.023332630543256728,
      "grad_norm": 1.1006306409835815,
      "learning_rate": 9.325463743676223e-05,
      "loss": 0.4041,
      "step": 1660
    },
    {
      "epoch": 0.02347318855857755,
      "grad_norm": 1.545573115348816,
      "learning_rate": 9.381675098369871e-05,
      "loss": 0.3858,
      "step": 1670
    },
    {
      "epoch": 0.023613746573898378,
      "grad_norm": 1.2185055017471313,
      "learning_rate": 9.437886453063519e-05,
      "loss": 0.4642,
      "step": 1680
    },
    {
      "epoch": 0.0237543045892192,
      "grad_norm": 1.1222918033599854,
      "learning_rate": 9.494097807757167e-05,
      "loss": 0.3786,
      "step": 1690
    },
    {
      "epoch": 0.023894862604540025,
      "grad_norm": 1.0804522037506104,
      "learning_rate": 9.550309162450816e-05,
      "loss": 0.4327,
      "step": 1700
    },
    {
      "epoch": 0.02403542061986085,
      "grad_norm": 1.225407361984253,
      "learning_rate": 9.606520517144464e-05,
      "loss": 0.3724,
      "step": 1710
    },
    {
      "epoch": 0.024175978635181672,
      "grad_norm": 1.0895999670028687,
      "learning_rate": 9.662731871838112e-05,
      "loss": 0.3786,
      "step": 1720
    },
    {
      "epoch": 0.024316536650502495,
      "grad_norm": 1.0843249559402466,
      "learning_rate": 9.718943226531761e-05,
      "loss": 0.3818,
      "step": 1730
    },
    {
      "epoch": 0.02445709466582332,
      "grad_norm": 1.1851115226745605,
      "learning_rate": 9.775154581225407e-05,
      "loss": 0.4002,
      "step": 1740
    },
    {
      "epoch": 0.024597652681144142,
      "grad_norm": 1.2212036848068237,
      "learning_rate": 9.831365935919056e-05,
      "loss": 0.4062,
      "step": 1750
    },
    {
      "epoch": 0.024738210696464966,
      "grad_norm": 1.0576879978179932,
      "learning_rate": 9.887577290612704e-05,
      "loss": 0.3912,
      "step": 1760
    },
    {
      "epoch": 0.02487876871178579,
      "grad_norm": 0.8810908198356628,
      "learning_rate": 9.943788645306352e-05,
      "loss": 0.384,
      "step": 1770
    },
    {
      "epoch": 0.025019326727106612,
      "grad_norm": 1.2164630889892578,
      "learning_rate": 0.0001,
      "loss": 0.3911,
      "step": 1780
    },
    {
      "epoch": 0.025159884742427436,
      "grad_norm": 1.2411495447158813,
      "learning_rate": 0.00010056211354693648,
      "loss": 0.3891,
      "step": 1790
    },
    {
      "epoch": 0.02530044275774826,
      "grad_norm": 1.175451397895813,
      "learning_rate": 0.00010112422709387297,
      "loss": 0.3795,
      "step": 1800
    },
    {
      "epoch": 0.025441000773069083,
      "grad_norm": 0.9957692623138428,
      "learning_rate": 0.00010168634064080944,
      "loss": 0.3749,
      "step": 1810
    },
    {
      "epoch": 0.025581558788389906,
      "grad_norm": 0.9644479751586914,
      "learning_rate": 0.00010224845418774594,
      "loss": 0.3787,
      "step": 1820
    },
    {
      "epoch": 0.025722116803710733,
      "grad_norm": 0.9513217210769653,
      "learning_rate": 0.0001028105677346824,
      "loss": 0.4283,
      "step": 1830
    },
    {
      "epoch": 0.025862674819031557,
      "grad_norm": 0.9092191457748413,
      "learning_rate": 0.0001033726812816189,
      "loss": 0.3784,
      "step": 1840
    },
    {
      "epoch": 0.02600323283435238,
      "grad_norm": 1.4808021783828735,
      "learning_rate": 0.00010393479482855537,
      "loss": 0.4201,
      "step": 1850
    },
    {
      "epoch": 0.026143790849673203,
      "grad_norm": 1.2115821838378906,
      "learning_rate": 0.00010449690837549186,
      "loss": 0.4046,
      "step": 1860
    },
    {
      "epoch": 0.026284348864994027,
      "grad_norm": 1.122286081314087,
      "learning_rate": 0.00010505902192242833,
      "loss": 0.384,
      "step": 1870
    },
    {
      "epoch": 0.02642490688031485,
      "grad_norm": 1.0566567182540894,
      "learning_rate": 0.00010562113546936481,
      "loss": 0.3915,
      "step": 1880
    },
    {
      "epoch": 0.026565464895635674,
      "grad_norm": 0.9356314539909363,
      "learning_rate": 0.0001061832490163013,
      "loss": 0.3931,
      "step": 1890
    },
    {
      "epoch": 0.026706022910956497,
      "grad_norm": 0.906581461429596,
      "learning_rate": 0.00010674536256323777,
      "loss": 0.3777,
      "step": 1900
    },
    {
      "epoch": 0.02684658092627732,
      "grad_norm": 1.248866319656372,
      "learning_rate": 0.00010730747611017427,
      "loss": 0.3682,
      "step": 1910
    },
    {
      "epoch": 0.026987138941598144,
      "grad_norm": 1.0143498182296753,
      "learning_rate": 0.00010786958965711074,
      "loss": 0.3706,
      "step": 1920
    },
    {
      "epoch": 0.027127696956918967,
      "grad_norm": 1.083511471748352,
      "learning_rate": 0.00010843170320404723,
      "loss": 0.3921,
      "step": 1930
    },
    {
      "epoch": 0.02726825497223979,
      "grad_norm": 1.1626296043395996,
      "learning_rate": 0.0001089938167509837,
      "loss": 0.4035,
      "step": 1940
    },
    {
      "epoch": 0.027408812987560614,
      "grad_norm": 1.2879900932312012,
      "learning_rate": 0.00010955593029792019,
      "loss": 0.4166,
      "step": 1950
    },
    {
      "epoch": 0.027549371002881438,
      "grad_norm": 1.0682278871536255,
      "learning_rate": 0.00011011804384485666,
      "loss": 0.3695,
      "step": 1960
    },
    {
      "epoch": 0.027689929018202265,
      "grad_norm": 0.9039905667304993,
      "learning_rate": 0.00011068015739179314,
      "loss": 0.3632,
      "step": 1970
    },
    {
      "epoch": 0.027830487033523088,
      "grad_norm": 1.5052381753921509,
      "learning_rate": 0.00011124227093872963,
      "loss": 0.3791,
      "step": 1980
    },
    {
      "epoch": 0.02797104504884391,
      "grad_norm": 0.8607678413391113,
      "learning_rate": 0.0001118043844856661,
      "loss": 0.363,
      "step": 1990
    },
    {
      "epoch": 0.028111603064164735,
      "grad_norm": 1.0287585258483887,
      "learning_rate": 0.0001123664980326026,
      "loss": 0.3258,
      "step": 2000
    },
    {
      "epoch": 0.028111603064164735,
      "eval_chrf": 79.95813802049287,
      "eval_loss": 0.6827706098556519,
      "eval_runtime": 328.1457,
      "eval_samples_per_second": 0.305,
      "eval_steps_per_second": 0.012,
      "step": 2000
    },
    {
      "epoch": 0.02825216107948556,
      "grad_norm": 0.9364410638809204,
      "learning_rate": 0.00011292861157953907,
      "loss": 0.3226,
      "step": 2010
    },
    {
      "epoch": 0.028392719094806382,
      "grad_norm": 0.9163166284561157,
      "learning_rate": 0.00011349072512647556,
      "loss": 0.3789,
      "step": 2020
    },
    {
      "epoch": 0.028533277110127205,
      "grad_norm": 0.9576098322868347,
      "learning_rate": 0.00011405283867341203,
      "loss": 0.385,
      "step": 2030
    },
    {
      "epoch": 0.02867383512544803,
      "grad_norm": 1.0045305490493774,
      "learning_rate": 0.00011461495222034852,
      "loss": 0.3638,
      "step": 2040
    },
    {
      "epoch": 0.028814393140768852,
      "grad_norm": 0.879092276096344,
      "learning_rate": 0.000115177065767285,
      "loss": 0.3359,
      "step": 2050
    },
    {
      "epoch": 0.028954951156089676,
      "grad_norm": 1.1309394836425781,
      "learning_rate": 0.00011573917931422147,
      "loss": 0.3656,
      "step": 2060
    },
    {
      "epoch": 0.0290955091714105,
      "grad_norm": 0.8896138668060303,
      "learning_rate": 0.00011630129286115796,
      "loss": 0.3606,
      "step": 2070
    },
    {
      "epoch": 0.029236067186731322,
      "grad_norm": 0.968147873878479,
      "learning_rate": 0.00011686340640809443,
      "loss": 0.3619,
      "step": 2080
    },
    {
      "epoch": 0.029376625202052146,
      "grad_norm": 0.9790638089179993,
      "learning_rate": 0.00011742551995503093,
      "loss": 0.3823,
      "step": 2090
    },
    {
      "epoch": 0.02951718321737297,
      "grad_norm": 0.9956974983215332,
      "learning_rate": 0.0001179876335019674,
      "loss": 0.35,
      "step": 2100
    },
    {
      "epoch": 0.029657741232693793,
      "grad_norm": 0.8596989512443542,
      "learning_rate": 0.00011854974704890389,
      "loss": 0.3352,
      "step": 2110
    },
    {
      "epoch": 0.02979829924801462,
      "grad_norm": 0.8734012842178345,
      "learning_rate": 0.00011911186059584036,
      "loss": 0.3636,
      "step": 2120
    },
    {
      "epoch": 0.029938857263335443,
      "grad_norm": 0.9950189590454102,
      "learning_rate": 0.00011967397414277685,
      "loss": 0.3519,
      "step": 2130
    },
    {
      "epoch": 0.030079415278656266,
      "grad_norm": 0.8868091106414795,
      "learning_rate": 0.00012023608768971332,
      "loss": 0.35,
      "step": 2140
    },
    {
      "epoch": 0.03021997329397709,
      "grad_norm": 1.1114511489868164,
      "learning_rate": 0.0001207982012366498,
      "loss": 0.3787,
      "step": 2150
    },
    {
      "epoch": 0.030360531309297913,
      "grad_norm": 0.9478281736373901,
      "learning_rate": 0.0001213603147835863,
      "loss": 0.3804,
      "step": 2160
    },
    {
      "epoch": 0.030501089324618737,
      "grad_norm": 1.0067881345748901,
      "learning_rate": 0.00012192242833052276,
      "loss": 0.3483,
      "step": 2170
    },
    {
      "epoch": 0.03064164733993956,
      "grad_norm": 0.8068206906318665,
      "learning_rate": 0.00012248454187745926,
      "loss": 0.346,
      "step": 2180
    },
    {
      "epoch": 0.030782205355260384,
      "grad_norm": 1.019374966621399,
      "learning_rate": 0.00012304665542439572,
      "loss": 0.3719,
      "step": 2190
    },
    {
      "epoch": 0.030922763370581207,
      "grad_norm": 0.9281585812568665,
      "learning_rate": 0.00012360876897133223,
      "loss": 0.3848,
      "step": 2200
    },
    {
      "epoch": 0.03106332138590203,
      "grad_norm": 1.1539422273635864,
      "learning_rate": 0.0001241708825182687,
      "loss": 0.3375,
      "step": 2210
    },
    {
      "epoch": 0.031203879401222854,
      "grad_norm": 0.738135814666748,
      "learning_rate": 0.00012473299606520518,
      "loss": 0.3362,
      "step": 2220
    },
    {
      "epoch": 0.03134443741654368,
      "grad_norm": 0.9841747879981995,
      "learning_rate": 0.00012529510961214167,
      "loss": 0.345,
      "step": 2230
    },
    {
      "epoch": 0.031484995431864504,
      "grad_norm": 1.066952109336853,
      "learning_rate": 0.00012585722315907813,
      "loss": 0.3728,
      "step": 2240
    },
    {
      "epoch": 0.031625553447185324,
      "grad_norm": 0.8559726476669312,
      "learning_rate": 0.00012641933670601462,
      "loss": 0.3485,
      "step": 2250
    },
    {
      "epoch": 0.03176611146250615,
      "grad_norm": 0.8930100798606873,
      "learning_rate": 0.0001269814502529511,
      "loss": 0.3533,
      "step": 2260
    },
    {
      "epoch": 0.03190666947782697,
      "grad_norm": 0.862795889377594,
      "learning_rate": 0.0001275435637998876,
      "loss": 0.3372,
      "step": 2270
    },
    {
      "epoch": 0.0320472274931478,
      "grad_norm": 1.2475395202636719,
      "learning_rate": 0.00012810567734682405,
      "loss": 0.334,
      "step": 2280
    },
    {
      "epoch": 0.03218778550846862,
      "grad_norm": 1.1247025728225708,
      "learning_rate": 0.00012866779089376056,
      "loss": 0.3321,
      "step": 2290
    },
    {
      "epoch": 0.032328343523789445,
      "grad_norm": 0.8312082290649414,
      "learning_rate": 0.00012922990444069702,
      "loss": 0.3547,
      "step": 2300
    },
    {
      "epoch": 0.032468901539110265,
      "grad_norm": 1.132459044456482,
      "learning_rate": 0.0001297920179876335,
      "loss": 0.346,
      "step": 2310
    },
    {
      "epoch": 0.03260945955443109,
      "grad_norm": 1.0509437322616577,
      "learning_rate": 0.00013035413153457,
      "loss": 0.353,
      "step": 2320
    },
    {
      "epoch": 0.03275001756975191,
      "grad_norm": 0.6921985745429993,
      "learning_rate": 0.00013091624508150646,
      "loss": 0.3266,
      "step": 2330
    },
    {
      "epoch": 0.03289057558507274,
      "grad_norm": 1.0222564935684204,
      "learning_rate": 0.00013147835862844295,
      "loss": 0.3283,
      "step": 2340
    },
    {
      "epoch": 0.033031133600393565,
      "grad_norm": 0.9538044929504395,
      "learning_rate": 0.00013204047217537943,
      "loss": 0.3507,
      "step": 2350
    },
    {
      "epoch": 0.033171691615714385,
      "grad_norm": 0.9025123715400696,
      "learning_rate": 0.00013260258572231592,
      "loss": 0.3235,
      "step": 2360
    },
    {
      "epoch": 0.03331224963103521,
      "grad_norm": 0.8488820195198059,
      "learning_rate": 0.00013316469926925238,
      "loss": 0.3101,
      "step": 2370
    },
    {
      "epoch": 0.03345280764635603,
      "grad_norm": 0.8986126780509949,
      "learning_rate": 0.0001337268128161889,
      "loss": 0.3591,
      "step": 2380
    },
    {
      "epoch": 0.03359336566167686,
      "grad_norm": 0.9442368149757385,
      "learning_rate": 0.00013428892636312535,
      "loss": 0.3341,
      "step": 2390
    },
    {
      "epoch": 0.03373392367699768,
      "grad_norm": 0.8823990225791931,
      "learning_rate": 0.00013485103991006184,
      "loss": 0.3431,
      "step": 2400
    },
    {
      "epoch": 0.033874481692318506,
      "grad_norm": 0.9334458112716675,
      "learning_rate": 0.00013541315345699833,
      "loss": 0.3524,
      "step": 2410
    },
    {
      "epoch": 0.034015039707639326,
      "grad_norm": 1.0057344436645508,
      "learning_rate": 0.0001359752670039348,
      "loss": 0.4021,
      "step": 2420
    },
    {
      "epoch": 0.03415559772296015,
      "grad_norm": 0.6858983635902405,
      "learning_rate": 0.00013653738055087128,
      "loss": 0.3678,
      "step": 2430
    },
    {
      "epoch": 0.03429615573828097,
      "grad_norm": 1.089292287826538,
      "learning_rate": 0.00013709949409780776,
      "loss": 0.3315,
      "step": 2440
    },
    {
      "epoch": 0.0344367137536018,
      "grad_norm": 0.9135834574699402,
      "learning_rate": 0.00013766160764474425,
      "loss": 0.3675,
      "step": 2450
    },
    {
      "epoch": 0.03457727176892262,
      "grad_norm": 0.9899437427520752,
      "learning_rate": 0.0001382237211916807,
      "loss": 0.3682,
      "step": 2460
    },
    {
      "epoch": 0.03471782978424345,
      "grad_norm": 1.0549930334091187,
      "learning_rate": 0.00013878583473861722,
      "loss": 0.3739,
      "step": 2470
    },
    {
      "epoch": 0.034858387799564274,
      "grad_norm": 0.9098493456840515,
      "learning_rate": 0.00013934794828555368,
      "loss": 0.3201,
      "step": 2480
    },
    {
      "epoch": 0.034998945814885093,
      "grad_norm": 0.8151450157165527,
      "learning_rate": 0.00013991006183249017,
      "loss": 0.3572,
      "step": 2490
    },
    {
      "epoch": 0.03513950383020592,
      "grad_norm": 0.8182278871536255,
      "learning_rate": 0.00014047217537942666,
      "loss": 0.347,
      "step": 2500
    },
    {
      "epoch": 0.03528006184552674,
      "grad_norm": 0.8542600870132446,
      "learning_rate": 0.00014103428892636312,
      "loss": 0.3541,
      "step": 2510
    },
    {
      "epoch": 0.03542061986084757,
      "grad_norm": 0.7973415851593018,
      "learning_rate": 0.0001415964024732996,
      "loss": 0.3006,
      "step": 2520
    },
    {
      "epoch": 0.03556117787616839,
      "grad_norm": 0.8098814487457275,
      "learning_rate": 0.0001421585160202361,
      "loss": 0.3527,
      "step": 2530
    },
    {
      "epoch": 0.035701735891489214,
      "grad_norm": 1.0114139318466187,
      "learning_rate": 0.00014272062956717258,
      "loss": 0.3239,
      "step": 2540
    },
    {
      "epoch": 0.035842293906810034,
      "grad_norm": 0.9623722434043884,
      "learning_rate": 0.00014328274311410904,
      "loss": 0.3452,
      "step": 2550
    },
    {
      "epoch": 0.03598285192213086,
      "grad_norm": 0.7429597973823547,
      "learning_rate": 0.00014384485666104555,
      "loss": 0.3391,
      "step": 2560
    },
    {
      "epoch": 0.03612340993745168,
      "grad_norm": 0.9132286906242371,
      "learning_rate": 0.00014440697020798201,
      "loss": 0.3569,
      "step": 2570
    },
    {
      "epoch": 0.03626396795277251,
      "grad_norm": 0.8367570638656616,
      "learning_rate": 0.0001449690837549185,
      "loss": 0.3389,
      "step": 2580
    },
    {
      "epoch": 0.03640452596809333,
      "grad_norm": 0.6398904919624329,
      "learning_rate": 0.000145531197301855,
      "loss": 0.3537,
      "step": 2590
    },
    {
      "epoch": 0.036545083983414155,
      "grad_norm": 0.720858097076416,
      "learning_rate": 0.00014609331084879145,
      "loss": 0.3445,
      "step": 2600
    },
    {
      "epoch": 0.036685641998734975,
      "grad_norm": 0.7617928385734558,
      "learning_rate": 0.00014665542439572794,
      "loss": 0.3368,
      "step": 2610
    },
    {
      "epoch": 0.0368262000140558,
      "grad_norm": 0.7096990942955017,
      "learning_rate": 0.00014721753794266442,
      "loss": 0.3279,
      "step": 2620
    },
    {
      "epoch": 0.03696675802937663,
      "grad_norm": 0.8983321189880371,
      "learning_rate": 0.0001477796514896009,
      "loss": 0.3285,
      "step": 2630
    },
    {
      "epoch": 0.03710731604469745,
      "grad_norm": 0.7671912908554077,
      "learning_rate": 0.00014834176503653737,
      "loss": 0.3237,
      "step": 2640
    },
    {
      "epoch": 0.037247874060018275,
      "grad_norm": 0.9126039147377014,
      "learning_rate": 0.00014890387858347388,
      "loss": 0.3819,
      "step": 2650
    },
    {
      "epoch": 0.037388432075339095,
      "grad_norm": 0.8447277545928955,
      "learning_rate": 0.00014946599213041034,
      "loss": 0.3585,
      "step": 2660
    },
    {
      "epoch": 0.03752899009065992,
      "grad_norm": 0.8905502557754517,
      "learning_rate": 0.00015002810567734683,
      "loss": 0.3018,
      "step": 2670
    },
    {
      "epoch": 0.03766954810598074,
      "grad_norm": 0.8705244064331055,
      "learning_rate": 0.00015059021922428332,
      "loss": 0.3162,
      "step": 2680
    },
    {
      "epoch": 0.03781010612130157,
      "grad_norm": 0.8502803444862366,
      "learning_rate": 0.00015115233277121978,
      "loss": 0.3344,
      "step": 2690
    },
    {
      "epoch": 0.03795066413662239,
      "grad_norm": 0.7822072505950928,
      "learning_rate": 0.00015171444631815627,
      "loss": 0.3248,
      "step": 2700
    },
    {
      "epoch": 0.038091222151943216,
      "grad_norm": 0.7527584433555603,
      "learning_rate": 0.00015227655986509275,
      "loss": 0.3294,
      "step": 2710
    },
    {
      "epoch": 0.038231780167264036,
      "grad_norm": 1.0729758739471436,
      "learning_rate": 0.00015283867341202924,
      "loss": 0.3372,
      "step": 2720
    },
    {
      "epoch": 0.03837233818258486,
      "grad_norm": 0.8660837411880493,
      "learning_rate": 0.00015340078695896573,
      "loss": 0.3804,
      "step": 2730
    },
    {
      "epoch": 0.03851289619790568,
      "grad_norm": 1.1277107000350952,
      "learning_rate": 0.00015396290050590221,
      "loss": 0.3386,
      "step": 2740
    },
    {
      "epoch": 0.03865345421322651,
      "grad_norm": 0.769478976726532,
      "learning_rate": 0.00015452501405283867,
      "loss": 0.303,
      "step": 2750
    },
    {
      "epoch": 0.03879401222854733,
      "grad_norm": 0.8349766135215759,
      "learning_rate": 0.00015508712759977516,
      "loss": 0.3209,
      "step": 2760
    },
    {
      "epoch": 0.038934570243868157,
      "grad_norm": 1.119290828704834,
      "learning_rate": 0.00015564924114671165,
      "loss": 0.3496,
      "step": 2770
    },
    {
      "epoch": 0.03907512825918898,
      "grad_norm": 0.9543491005897522,
      "learning_rate": 0.0001562113546936481,
      "loss": 0.3111,
      "step": 2780
    },
    {
      "epoch": 0.0392156862745098,
      "grad_norm": 0.7911608219146729,
      "learning_rate": 0.0001567734682405846,
      "loss": 0.3173,
      "step": 2790
    },
    {
      "epoch": 0.03935624428983063,
      "grad_norm": 0.8939515352249146,
      "learning_rate": 0.00015733558178752108,
      "loss": 0.3213,
      "step": 2800
    },
    {
      "epoch": 0.03949680230515145,
      "grad_norm": 0.6804023385047913,
      "learning_rate": 0.00015789769533445757,
      "loss": 0.2831,
      "step": 2810
    },
    {
      "epoch": 0.03963736032047228,
      "grad_norm": 0.894201934337616,
      "learning_rate": 0.00015845980888139406,
      "loss": 0.3261,
      "step": 2820
    },
    {
      "epoch": 0.0397779183357931,
      "grad_norm": 0.6332712173461914,
      "learning_rate": 0.00015902192242833054,
      "loss": 0.3138,
      "step": 2830
    },
    {
      "epoch": 0.039918476351113924,
      "grad_norm": 0.8150140643119812,
      "learning_rate": 0.000159584035975267,
      "loss": 0.3085,
      "step": 2840
    },
    {
      "epoch": 0.040059034366434744,
      "grad_norm": 1.022434115409851,
      "learning_rate": 0.0001601461495222035,
      "loss": 0.3481,
      "step": 2850
    },
    {
      "epoch": 0.04019959238175557,
      "grad_norm": 0.8622433543205261,
      "learning_rate": 0.00016070826306913998,
      "loss": 0.2984,
      "step": 2860
    },
    {
      "epoch": 0.04034015039707639,
      "grad_norm": 0.7442618608474731,
      "learning_rate": 0.00016127037661607644,
      "loss": 0.3327,
      "step": 2870
    },
    {
      "epoch": 0.04048070841239722,
      "grad_norm": 0.7326914668083191,
      "learning_rate": 0.00016183249016301295,
      "loss": 0.3162,
      "step": 2880
    },
    {
      "epoch": 0.04062126642771804,
      "grad_norm": 1.0088212490081787,
      "learning_rate": 0.0001623946037099494,
      "loss": 0.3155,
      "step": 2890
    },
    {
      "epoch": 0.040761824443038865,
      "grad_norm": 0.6482020616531372,
      "learning_rate": 0.0001629567172568859,
      "loss": 0.3216,
      "step": 2900
    },
    {
      "epoch": 0.040902382458359685,
      "grad_norm": 0.5968291759490967,
      "learning_rate": 0.0001635188308038224,
      "loss": 0.2942,
      "step": 2910
    },
    {
      "epoch": 0.04104294047368051,
      "grad_norm": 0.715417206287384,
      "learning_rate": 0.00016408094435075887,
      "loss": 0.3336,
      "step": 2920
    },
    {
      "epoch": 0.04118349848900134,
      "grad_norm": 0.7715390920639038,
      "learning_rate": 0.00016464305789769533,
      "loss": 0.3328,
      "step": 2930
    },
    {
      "epoch": 0.04132405650432216,
      "grad_norm": 0.8046168684959412,
      "learning_rate": 0.00016520517144463182,
      "loss": 0.3092,
      "step": 2940
    },
    {
      "epoch": 0.041464614519642985,
      "grad_norm": 0.8180148005485535,
      "learning_rate": 0.0001657672849915683,
      "loss": 0.3111,
      "step": 2950
    },
    {
      "epoch": 0.041605172534963805,
      "grad_norm": 0.664538562297821,
      "learning_rate": 0.00016632939853850477,
      "loss": 0.3258,
      "step": 2960
    },
    {
      "epoch": 0.04174573055028463,
      "grad_norm": 0.8169111013412476,
      "learning_rate": 0.00016689151208544128,
      "loss": 0.3313,
      "step": 2970
    },
    {
      "epoch": 0.04188628856560545,
      "grad_norm": 0.7573257684707642,
      "learning_rate": 0.00016745362563237774,
      "loss": 0.34,
      "step": 2980
    },
    {
      "epoch": 0.04202684658092628,
      "grad_norm": 0.9846084117889404,
      "learning_rate": 0.00016801573917931423,
      "loss": 0.3267,
      "step": 2990
    },
    {
      "epoch": 0.0421674045962471,
      "grad_norm": 0.5704976916313171,
      "learning_rate": 0.00016857785272625072,
      "loss": 0.3344,
      "step": 3000
    },
    {
      "epoch": 0.0421674045962471,
      "eval_chrf": 72.83505853609785,
      "eval_loss": 0.5984305739402771,
      "eval_runtime": 340.9131,
      "eval_samples_per_second": 0.293,
      "eval_steps_per_second": 0.012,
      "step": 3000
    },
    {
      "epoch": 0.042307962611567926,
      "grad_norm": 0.7193232774734497,
      "learning_rate": 0.0001691399662731872,
      "loss": 0.3326,
      "step": 3010
    },
    {
      "epoch": 0.042448520626888746,
      "grad_norm": 0.6804803013801575,
      "learning_rate": 0.00016970207982012366,
      "loss": 0.2982,
      "step": 3020
    },
    {
      "epoch": 0.04258907864220957,
      "grad_norm": 0.8106788396835327,
      "learning_rate": 0.00017026419336706015,
      "loss": 0.3429,
      "step": 3030
    },
    {
      "epoch": 0.04272963665753039,
      "grad_norm": 0.6692339777946472,
      "learning_rate": 0.00017082630691399664,
      "loss": 0.289,
      "step": 3040
    },
    {
      "epoch": 0.04287019467285122,
      "grad_norm": 0.61186683177948,
      "learning_rate": 0.0001713884204609331,
      "loss": 0.3138,
      "step": 3050
    },
    {
      "epoch": 0.043010752688172046,
      "grad_norm": 0.8928340077400208,
      "learning_rate": 0.0001719505340078696,
      "loss": 0.2933,
      "step": 3060
    },
    {
      "epoch": 0.043151310703492866,
      "grad_norm": 0.6500282287597656,
      "learning_rate": 0.00017251264755480607,
      "loss": 0.3044,
      "step": 3070
    },
    {
      "epoch": 0.04329186871881369,
      "grad_norm": 0.6658849716186523,
      "learning_rate": 0.00017307476110174256,
      "loss": 0.2989,
      "step": 3080
    },
    {
      "epoch": 0.04343242673413451,
      "grad_norm": 0.5688410997390747,
      "learning_rate": 0.00017363687464867905,
      "loss": 0.3307,
      "step": 3090
    },
    {
      "epoch": 0.04357298474945534,
      "grad_norm": 0.7900064587593079,
      "learning_rate": 0.00017419898819561553,
      "loss": 0.3258,
      "step": 3100
    },
    {
      "epoch": 0.04371354276477616,
      "grad_norm": 0.5709797739982605,
      "learning_rate": 0.000174761101742552,
      "loss": 0.3107,
      "step": 3110
    },
    {
      "epoch": 0.04385410078009699,
      "grad_norm": 0.7320407629013062,
      "learning_rate": 0.0001753232152894885,
      "loss": 0.2967,
      "step": 3120
    },
    {
      "epoch": 0.04399465879541781,
      "grad_norm": 0.6004393696784973,
      "learning_rate": 0.00017588532883642497,
      "loss": 0.2797,
      "step": 3130
    },
    {
      "epoch": 0.044135216810738634,
      "grad_norm": 0.6750929951667786,
      "learning_rate": 0.00017644744238336143,
      "loss": 0.2963,
      "step": 3140
    },
    {
      "epoch": 0.044275774826059454,
      "grad_norm": 0.8092765212059021,
      "learning_rate": 0.00017700955593029794,
      "loss": 0.3199,
      "step": 3150
    },
    {
      "epoch": 0.04441633284138028,
      "grad_norm": 0.712920069694519,
      "learning_rate": 0.0001775716694772344,
      "loss": 0.3357,
      "step": 3160
    },
    {
      "epoch": 0.0445568908567011,
      "grad_norm": 0.7346781492233276,
      "learning_rate": 0.0001781337830241709,
      "loss": 0.2914,
      "step": 3170
    },
    {
      "epoch": 0.04469744887202193,
      "grad_norm": 0.8555288314819336,
      "learning_rate": 0.00017869589657110738,
      "loss": 0.3262,
      "step": 3180
    },
    {
      "epoch": 0.04483800688734275,
      "grad_norm": 0.6750013828277588,
      "learning_rate": 0.00017925801011804386,
      "loss": 0.276,
      "step": 3190
    },
    {
      "epoch": 0.044978564902663574,
      "grad_norm": 0.7254294157028198,
      "learning_rate": 0.00017982012366498032,
      "loss": 0.3026,
      "step": 3200
    },
    {
      "epoch": 0.0451191229179844,
      "grad_norm": 0.7228041887283325,
      "learning_rate": 0.00018038223721191684,
      "loss": 0.3006,
      "step": 3210
    },
    {
      "epoch": 0.04525968093330522,
      "grad_norm": 0.7393637299537659,
      "learning_rate": 0.0001809443507588533,
      "loss": 0.2931,
      "step": 3220
    },
    {
      "epoch": 0.04540023894862605,
      "grad_norm": 0.6972795724868774,
      "learning_rate": 0.00018150646430578976,
      "loss": 0.3016,
      "step": 3230
    },
    {
      "epoch": 0.04554079696394687,
      "grad_norm": 0.6396249532699585,
      "learning_rate": 0.00018206857785272627,
      "loss": 0.3473,
      "step": 3240
    },
    {
      "epoch": 0.045681354979267695,
      "grad_norm": 0.6952264308929443,
      "learning_rate": 0.00018263069139966273,
      "loss": 0.2949,
      "step": 3250
    },
    {
      "epoch": 0.045821912994588515,
      "grad_norm": 0.701374888420105,
      "learning_rate": 0.00018319280494659922,
      "loss": 0.2838,
      "step": 3260
    },
    {
      "epoch": 0.04596247100990934,
      "grad_norm": 0.7950251698493958,
      "learning_rate": 0.0001837549184935357,
      "loss": 0.3195,
      "step": 3270
    },
    {
      "epoch": 0.04610302902523016,
      "grad_norm": 0.6929548978805542,
      "learning_rate": 0.0001843170320404722,
      "loss": 0.317,
      "step": 3280
    },
    {
      "epoch": 0.04624358704055099,
      "grad_norm": 0.8195067644119263,
      "learning_rate": 0.00018487914558740865,
      "loss": 0.2885,
      "step": 3290
    },
    {
      "epoch": 0.04638414505587181,
      "grad_norm": 0.7615912556648254,
      "learning_rate": 0.00018544125913434514,
      "loss": 0.3118,
      "step": 3300
    },
    {
      "epoch": 0.046524703071192636,
      "grad_norm": 0.7470051050186157,
      "learning_rate": 0.00018600337268128163,
      "loss": 0.3003,
      "step": 3310
    },
    {
      "epoch": 0.046665261086513456,
      "grad_norm": 0.6888974905014038,
      "learning_rate": 0.0001865654862282181,
      "loss": 0.3051,
      "step": 3320
    },
    {
      "epoch": 0.04680581910183428,
      "grad_norm": 0.7712236046791077,
      "learning_rate": 0.0001871275997751546,
      "loss": 0.3053,
      "step": 3330
    },
    {
      "epoch": 0.0469463771171551,
      "grad_norm": 0.7824040651321411,
      "learning_rate": 0.00018768971332209106,
      "loss": 0.2838,
      "step": 3340
    },
    {
      "epoch": 0.04708693513247593,
      "grad_norm": 0.7765060663223267,
      "learning_rate": 0.00018825182686902755,
      "loss": 0.2985,
      "step": 3350
    },
    {
      "epoch": 0.047227493147796756,
      "grad_norm": 0.6892460584640503,
      "learning_rate": 0.00018881394041596404,
      "loss": 0.3083,
      "step": 3360
    },
    {
      "epoch": 0.047368051163117576,
      "grad_norm": 0.8150136470794678,
      "learning_rate": 0.00018937605396290052,
      "loss": 0.2859,
      "step": 3370
    },
    {
      "epoch": 0.0475086091784384,
      "grad_norm": 0.743232250213623,
      "learning_rate": 0.00018993816750983698,
      "loss": 0.2857,
      "step": 3380
    },
    {
      "epoch": 0.04764916719375922,
      "grad_norm": 0.7354878783226013,
      "learning_rate": 0.00019050028105677347,
      "loss": 0.3026,
      "step": 3390
    },
    {
      "epoch": 0.04778972520908005,
      "grad_norm": 0.5742655396461487,
      "learning_rate": 0.00019106239460370996,
      "loss": 0.3129,
      "step": 3400
    },
    {
      "epoch": 0.04793028322440087,
      "grad_norm": 0.6180288791656494,
      "learning_rate": 0.00019162450815064642,
      "loss": 0.2948,
      "step": 3410
    },
    {
      "epoch": 0.0480708412397217,
      "grad_norm": 0.6718013882637024,
      "learning_rate": 0.00019218662169758293,
      "loss": 0.2985,
      "step": 3420
    },
    {
      "epoch": 0.04821139925504252,
      "grad_norm": 0.5934156179428101,
      "learning_rate": 0.0001927487352445194,
      "loss": 0.2858,
      "step": 3430
    },
    {
      "epoch": 0.048351957270363344,
      "grad_norm": 0.5521450638771057,
      "learning_rate": 0.00019331084879145588,
      "loss": 0.3007,
      "step": 3440
    },
    {
      "epoch": 0.048492515285684164,
      "grad_norm": 0.678540825843811,
      "learning_rate": 0.00019387296233839237,
      "loss": 0.2795,
      "step": 3450
    },
    {
      "epoch": 0.04863307330100499,
      "grad_norm": 0.6061440706253052,
      "learning_rate": 0.00019443507588532885,
      "loss": 0.2996,
      "step": 3460
    },
    {
      "epoch": 0.04877363131632581,
      "grad_norm": 0.6506759524345398,
      "learning_rate": 0.00019499718943226531,
      "loss": 0.3141,
      "step": 3470
    },
    {
      "epoch": 0.04891418933164664,
      "grad_norm": 0.5880570411682129,
      "learning_rate": 0.0001955593029792018,
      "loss": 0.3123,
      "step": 3480
    },
    {
      "epoch": 0.04905474734696746,
      "grad_norm": 0.8206682205200195,
      "learning_rate": 0.0001961214165261383,
      "loss": 0.3009,
      "step": 3490
    },
    {
      "epoch": 0.049195305362288284,
      "grad_norm": 0.7400031089782715,
      "learning_rate": 0.00019668353007307475,
      "loss": 0.2944,
      "step": 3500
    },
    {
      "epoch": 0.04933586337760911,
      "grad_norm": 0.6021711826324463,
      "learning_rate": 0.00019724564362001126,
      "loss": 0.3242,
      "step": 3510
    },
    {
      "epoch": 0.04947642139292993,
      "grad_norm": 0.6773003339767456,
      "learning_rate": 0.00019780775716694772,
      "loss": 0.3115,
      "step": 3520
    },
    {
      "epoch": 0.04961697940825076,
      "grad_norm": 0.600722074508667,
      "learning_rate": 0.0001983698707138842,
      "loss": 0.2828,
      "step": 3530
    },
    {
      "epoch": 0.04975753742357158,
      "grad_norm": 0.6942675113677979,
      "learning_rate": 0.0001989319842608207,
      "loss": 0.2918,
      "step": 3540
    },
    {
      "epoch": 0.049898095438892405,
      "grad_norm": 0.7262158393859863,
      "learning_rate": 0.00019949409780775718,
      "loss": 0.2892,
      "step": 3550
    },
    {
      "epoch": 0.050038653454213225,
      "grad_norm": 0.7155596613883972,
      "learning_rate": 0.00019999999989197024,
      "loss": 0.2958,
      "step": 3560
    },
    {
      "epoch": 0.05017921146953405,
      "grad_norm": 0.6224256157875061,
      "learning_rate": 0.00019999998692839935,
      "loss": 0.2975,
      "step": 3570
    },
    {
      "epoch": 0.05031976948485487,
      "grad_norm": 0.6881932020187378,
      "learning_rate": 0.00019999995235887974,
      "loss": 0.3063,
      "step": 3580
    },
    {
      "epoch": 0.0504603275001757,
      "grad_norm": 0.595230221748352,
      "learning_rate": 0.00019999989618341886,
      "loss": 0.2761,
      "step": 3590
    },
    {
      "epoch": 0.05060088551549652,
      "grad_norm": 0.7748909592628479,
      "learning_rate": 0.00019999981840202884,
      "loss": 0.2684,
      "step": 3600
    },
    {
      "epoch": 0.050741443530817346,
      "grad_norm": 0.6825932264328003,
      "learning_rate": 0.0001999997190147265,
      "loss": 0.3061,
      "step": 3610
    },
    {
      "epoch": 0.050882001546138166,
      "grad_norm": 0.7129770517349243,
      "learning_rate": 0.0001999995980215333,
      "loss": 0.3035,
      "step": 3620
    },
    {
      "epoch": 0.05102255956145899,
      "grad_norm": 0.7043442130088806,
      "learning_rate": 0.00019999945542247541,
      "loss": 0.2797,
      "step": 3630
    },
    {
      "epoch": 0.05116311757677981,
      "grad_norm": 0.5520175695419312,
      "learning_rate": 0.0001999992912175836,
      "loss": 0.2845,
      "step": 3640
    },
    {
      "epoch": 0.05130367559210064,
      "grad_norm": 0.5965823531150818,
      "learning_rate": 0.00019999910540689339,
      "loss": 0.3074,
      "step": 3650
    },
    {
      "epoch": 0.051444233607421466,
      "grad_norm": 0.578132152557373,
      "learning_rate": 0.0001999988979904449,
      "loss": 0.285,
      "step": 3660
    },
    {
      "epoch": 0.051584791622742286,
      "grad_norm": 0.6404640674591064,
      "learning_rate": 0.0001999986689682829,
      "loss": 0.3044,
      "step": 3670
    },
    {
      "epoch": 0.05172534963806311,
      "grad_norm": 0.6365982294082642,
      "learning_rate": 0.00019999841834045698,
      "loss": 0.2775,
      "step": 3680
    },
    {
      "epoch": 0.05186590765338393,
      "grad_norm": 0.7316439747810364,
      "learning_rate": 0.0001999981461070212,
      "loss": 0.3128,
      "step": 3690
    },
    {
      "epoch": 0.05200646566870476,
      "grad_norm": 0.5798088312149048,
      "learning_rate": 0.00019999785226803446,
      "loss": 0.2856,
      "step": 3700
    },
    {
      "epoch": 0.05214702368402558,
      "grad_norm": 0.5912203192710876,
      "learning_rate": 0.00019999753682356012,
      "loss": 0.3062,
      "step": 3710
    },
    {
      "epoch": 0.05228758169934641,
      "grad_norm": 0.8300290107727051,
      "learning_rate": 0.00019999719977366648,
      "loss": 0.293,
      "step": 3720
    },
    {
      "epoch": 0.05242813971466723,
      "grad_norm": 0.6406461000442505,
      "learning_rate": 0.00019999684111842627,
      "loss": 0.2573,
      "step": 3730
    },
    {
      "epoch": 0.052568697729988054,
      "grad_norm": 0.6481155157089233,
      "learning_rate": 0.00019999646085791701,
      "loss": 0.2763,
      "step": 3740
    },
    {
      "epoch": 0.052709255745308874,
      "grad_norm": 0.6483648419380188,
      "learning_rate": 0.00019999605899222083,
      "loss": 0.2845,
      "step": 3750
    },
    {
      "epoch": 0.0528498137606297,
      "grad_norm": 0.5151050686836243,
      "learning_rate": 0.00019999563552142461,
      "loss": 0.2493,
      "step": 3760
    },
    {
      "epoch": 0.05299037177595052,
      "grad_norm": 0.6392170190811157,
      "learning_rate": 0.00019999519044561982,
      "loss": 0.2946,
      "step": 3770
    },
    {
      "epoch": 0.05313092979127135,
      "grad_norm": 0.573092520236969,
      "learning_rate": 0.0001999947237649026,
      "loss": 0.2933,
      "step": 3780
    },
    {
      "epoch": 0.053271487806592174,
      "grad_norm": 0.6438156366348267,
      "learning_rate": 0.00019999423547937382,
      "loss": 0.2462,
      "step": 3790
    },
    {
      "epoch": 0.053412045821912994,
      "grad_norm": 0.6364404559135437,
      "learning_rate": 0.00019999372558913894,
      "loss": 0.3092,
      "step": 3800
    },
    {
      "epoch": 0.05355260383723382,
      "grad_norm": 0.5950109958648682,
      "learning_rate": 0.00019999319409430815,
      "loss": 0.2821,
      "step": 3810
    },
    {
      "epoch": 0.05369316185255464,
      "grad_norm": 0.5111678242683411,
      "learning_rate": 0.00019999264099499628,
      "loss": 0.3083,
      "step": 3820
    },
    {
      "epoch": 0.05383371986787547,
      "grad_norm": 0.5594717264175415,
      "learning_rate": 0.00019999206629132282,
      "loss": 0.3032,
      "step": 3830
    },
    {
      "epoch": 0.05397427788319629,
      "grad_norm": 0.7607744932174683,
      "learning_rate": 0.000199991469983412,
      "loss": 0.3019,
      "step": 3840
    },
    {
      "epoch": 0.054114835898517115,
      "grad_norm": 1.16470468044281,
      "learning_rate": 0.0001999908520713926,
      "loss": 0.287,
      "step": 3850
    },
    {
      "epoch": 0.054255393913837935,
      "grad_norm": 0.6292256712913513,
      "learning_rate": 0.0001999902125553981,
      "loss": 0.2816,
      "step": 3860
    },
    {
      "epoch": 0.05439595192915876,
      "grad_norm": 0.49813827872276306,
      "learning_rate": 0.0001999895514355667,
      "loss": 0.2729,
      "step": 3870
    },
    {
      "epoch": 0.05453650994447958,
      "grad_norm": 0.8331692814826965,
      "learning_rate": 0.0001999888687120413,
      "loss": 0.3057,
      "step": 3880
    },
    {
      "epoch": 0.05467706795980041,
      "grad_norm": 0.5014657974243164,
      "learning_rate": 0.00019998816438496935,
      "loss": 0.2808,
      "step": 3890
    },
    {
      "epoch": 0.05481762597512123,
      "grad_norm": 0.7092079520225525,
      "learning_rate": 0.000199987438454503,
      "loss": 0.2895,
      "step": 3900
    },
    {
      "epoch": 0.054958183990442055,
      "grad_norm": 0.7176849246025085,
      "learning_rate": 0.00019998669092079915,
      "loss": 0.2883,
      "step": 3910
    },
    {
      "epoch": 0.055098742005762875,
      "grad_norm": 0.7874364852905273,
      "learning_rate": 0.0001999859217840193,
      "loss": 0.2901,
      "step": 3920
    },
    {
      "epoch": 0.0552393000210837,
      "grad_norm": 0.6541767120361328,
      "learning_rate": 0.00019998513104432962,
      "loss": 0.3062,
      "step": 3930
    },
    {
      "epoch": 0.05537985803640453,
      "grad_norm": 0.6360107064247131,
      "learning_rate": 0.00019998431870190097,
      "loss": 0.2651,
      "step": 3940
    },
    {
      "epoch": 0.05552041605172535,
      "grad_norm": 0.6698468327522278,
      "learning_rate": 0.00019998348475690883,
      "loss": 0.2763,
      "step": 3950
    },
    {
      "epoch": 0.055660974067046176,
      "grad_norm": 0.6481212973594666,
      "learning_rate": 0.00019998262920953343,
      "loss": 0.2844,
      "step": 3960
    },
    {
      "epoch": 0.055801532082366996,
      "grad_norm": 0.6569493412971497,
      "learning_rate": 0.0001999817520599596,
      "loss": 0.2694,
      "step": 3970
    },
    {
      "epoch": 0.05594209009768782,
      "grad_norm": 0.6541939377784729,
      "learning_rate": 0.0001999808533083768,
      "loss": 0.3063,
      "step": 3980
    },
    {
      "epoch": 0.05608264811300864,
      "grad_norm": 0.6222134232521057,
      "learning_rate": 0.00019997993295497933,
      "loss": 0.2811,
      "step": 3990
    },
    {
      "epoch": 0.05622320612832947,
      "grad_norm": 0.537887454032898,
      "learning_rate": 0.00019997899099996594,
      "loss": 0.2882,
      "step": 4000
    },
    {
      "epoch": 0.05622320612832947,
      "eval_chrf": 80.12473807899819,
      "eval_loss": 0.5646688342094421,
      "eval_runtime": 269.6909,
      "eval_samples_per_second": 0.371,
      "eval_steps_per_second": 0.015,
      "step": 4000
    },
    {
      "epoch": 0.05636376414365029,
      "grad_norm": 0.7329326868057251,
      "learning_rate": 0.00019997802744354019,
      "loss": 0.2727,
      "step": 4010
    },
    {
      "epoch": 0.05650432215897112,
      "grad_norm": 0.5730016827583313,
      "learning_rate": 0.00019997704228591023,
      "loss": 0.3122,
      "step": 4020
    },
    {
      "epoch": 0.05664488017429194,
      "grad_norm": 0.6560089588165283,
      "learning_rate": 0.000199976035527289,
      "loss": 0.2894,
      "step": 4030
    },
    {
      "epoch": 0.056785438189612764,
      "grad_norm": 0.6476408243179321,
      "learning_rate": 0.0001999750071678939,
      "loss": 0.2553,
      "step": 4040
    },
    {
      "epoch": 0.056925996204933584,
      "grad_norm": 0.4794682562351227,
      "learning_rate": 0.00019997395720794723,
      "loss": 0.2737,
      "step": 4050
    },
    {
      "epoch": 0.05706655422025441,
      "grad_norm": 0.5600661635398865,
      "learning_rate": 0.00019997288564767578,
      "loss": 0.2772,
      "step": 4060
    },
    {
      "epoch": 0.05720711223557523,
      "grad_norm": 0.7425886392593384,
      "learning_rate": 0.00019997179248731106,
      "loss": 0.3064,
      "step": 4070
    },
    {
      "epoch": 0.05734767025089606,
      "grad_norm": 0.7069060206413269,
      "learning_rate": 0.0001999706777270893,
      "loss": 0.3061,
      "step": 4080
    },
    {
      "epoch": 0.057488228266216884,
      "grad_norm": 0.6282569766044617,
      "learning_rate": 0.00019996954136725133,
      "loss": 0.2908,
      "step": 4090
    },
    {
      "epoch": 0.057628786281537704,
      "grad_norm": 0.7161028385162354,
      "learning_rate": 0.00019996838340804267,
      "loss": 0.2728,
      "step": 4100
    },
    {
      "epoch": 0.05776934429685853,
      "grad_norm": 0.7045167684555054,
      "learning_rate": 0.00019996720384971354,
      "loss": 0.2568,
      "step": 4110
    },
    {
      "epoch": 0.05790990231217935,
      "grad_norm": 0.569190502166748,
      "learning_rate": 0.00019996600269251875,
      "loss": 0.2459,
      "step": 4120
    },
    {
      "epoch": 0.05805046032750018,
      "grad_norm": 0.6955873370170593,
      "learning_rate": 0.00019996477993671781,
      "loss": 0.2681,
      "step": 4130
    },
    {
      "epoch": 0.058191018342821,
      "grad_norm": 0.6311067938804626,
      "learning_rate": 0.000199963535582575,
      "loss": 0.2687,
      "step": 4140
    },
    {
      "epoch": 0.058331576358141825,
      "grad_norm": 0.5590523481369019,
      "learning_rate": 0.0001999622696303591,
      "loss": 0.313,
      "step": 4150
    },
    {
      "epoch": 0.058472134373462645,
      "grad_norm": 0.6355190277099609,
      "learning_rate": 0.00019996098208034365,
      "loss": 0.2935,
      "step": 4160
    },
    {
      "epoch": 0.05861269238878347,
      "grad_norm": 0.7000370025634766,
      "learning_rate": 0.00019995967293280677,
      "loss": 0.2969,
      "step": 4170
    },
    {
      "epoch": 0.05875325040410429,
      "grad_norm": 0.7441065311431885,
      "learning_rate": 0.00019995834218803144,
      "loss": 0.2705,
      "step": 4180
    },
    {
      "epoch": 0.05889380841942512,
      "grad_norm": 0.6410252451896667,
      "learning_rate": 0.0001999569898463051,
      "loss": 0.2813,
      "step": 4190
    },
    {
      "epoch": 0.05903436643474594,
      "grad_norm": 0.5887627005577087,
      "learning_rate": 0.00019995561590791994,
      "loss": 0.3098,
      "step": 4200
    },
    {
      "epoch": 0.059174924450066765,
      "grad_norm": 0.6581453680992126,
      "learning_rate": 0.00019995422037317285,
      "loss": 0.2945,
      "step": 4210
    },
    {
      "epoch": 0.059315482465387585,
      "grad_norm": 0.8213226199150085,
      "learning_rate": 0.0001999528032423653,
      "loss": 0.3026,
      "step": 4220
    },
    {
      "epoch": 0.05945604048070841,
      "grad_norm": 0.6150140762329102,
      "learning_rate": 0.0001999513645158035,
      "loss": 0.2856,
      "step": 4230
    },
    {
      "epoch": 0.05959659849602924,
      "grad_norm": 0.7967432737350464,
      "learning_rate": 0.0001999499041937983,
      "loss": 0.3084,
      "step": 4240
    },
    {
      "epoch": 0.05973715651135006,
      "grad_norm": 0.7099254131317139,
      "learning_rate": 0.00019994842227666524,
      "loss": 0.323,
      "step": 4250
    },
    {
      "epoch": 0.059877714526670886,
      "grad_norm": 0.6161561608314514,
      "learning_rate": 0.00019994691876472446,
      "loss": 0.3072,
      "step": 4260
    },
    {
      "epoch": 0.060018272541991706,
      "grad_norm": 0.5760515332221985,
      "learning_rate": 0.00019994539365830082,
      "loss": 0.2331,
      "step": 4270
    },
    {
      "epoch": 0.06015883055731253,
      "grad_norm": 0.619286060333252,
      "learning_rate": 0.00019994384695772385,
      "loss": 0.275,
      "step": 4280
    },
    {
      "epoch": 0.06029938857263335,
      "grad_norm": 0.5424856543540955,
      "learning_rate": 0.0001999422786633277,
      "loss": 0.2968,
      "step": 4290
    },
    {
      "epoch": 0.06043994658795418,
      "grad_norm": 0.5366678237915039,
      "learning_rate": 0.00019994068877545124,
      "loss": 0.2816,
      "step": 4300
    },
    {
      "epoch": 0.060580504603275,
      "grad_norm": 0.6595004796981812,
      "learning_rate": 0.000199939077294438,
      "loss": 0.2827,
      "step": 4310
    },
    {
      "epoch": 0.06072106261859583,
      "grad_norm": 0.6365996599197388,
      "learning_rate": 0.0001999374442206361,
      "loss": 0.2903,
      "step": 4320
    },
    {
      "epoch": 0.06086162063391665,
      "grad_norm": 0.6086582541465759,
      "learning_rate": 0.0001999357895543984,
      "loss": 0.2812,
      "step": 4330
    },
    {
      "epoch": 0.06100217864923747,
      "grad_norm": 0.623997151851654,
      "learning_rate": 0.00019993411329608245,
      "loss": 0.2985,
      "step": 4340
    },
    {
      "epoch": 0.06114273666455829,
      "grad_norm": 0.5988753437995911,
      "learning_rate": 0.00019993241544605036,
      "loss": 0.2632,
      "step": 4350
    },
    {
      "epoch": 0.06128329467987912,
      "grad_norm": 0.5856814384460449,
      "learning_rate": 0.00019993069600466904,
      "loss": 0.2703,
      "step": 4360
    },
    {
      "epoch": 0.06142385269519995,
      "grad_norm": 0.6997985243797302,
      "learning_rate": 0.00019992895497230993,
      "loss": 0.2943,
      "step": 4370
    },
    {
      "epoch": 0.06156441071052077,
      "grad_norm": 0.6265032887458801,
      "learning_rate": 0.00019992719234934917,
      "loss": 0.278,
      "step": 4380
    },
    {
      "epoch": 0.061704968725841594,
      "grad_norm": 0.5652262568473816,
      "learning_rate": 0.00019992540813616768,
      "loss": 0.2548,
      "step": 4390
    },
    {
      "epoch": 0.061845526741162414,
      "grad_norm": 0.5852500796318054,
      "learning_rate": 0.0001999236023331509,
      "loss": 0.3365,
      "step": 4400
    },
    {
      "epoch": 0.06198608475648324,
      "grad_norm": 0.7313827276229858,
      "learning_rate": 0.000199921774940689,
      "loss": 0.287,
      "step": 4410
    },
    {
      "epoch": 0.06212664277180406,
      "grad_norm": 0.6800966858863831,
      "learning_rate": 0.0001999199259591768,
      "loss": 0.2714,
      "step": 4420
    },
    {
      "epoch": 0.06226720078712489,
      "grad_norm": 0.563739538192749,
      "learning_rate": 0.00019991805538901382,
      "loss": 0.2744,
      "step": 4430
    },
    {
      "epoch": 0.06240775880244571,
      "grad_norm": 0.5896499752998352,
      "learning_rate": 0.00019991616323060418,
      "loss": 0.2732,
      "step": 4440
    },
    {
      "epoch": 0.06254831681776653,
      "grad_norm": 0.5427751541137695,
      "learning_rate": 0.00019991424948435671,
      "loss": 0.2753,
      "step": 4450
    },
    {
      "epoch": 0.06268887483308735,
      "grad_norm": 0.6296198964118958,
      "learning_rate": 0.0001999123141506849,
      "loss": 0.3029,
      "step": 4460
    },
    {
      "epoch": 0.06282943284840818,
      "grad_norm": 0.6063293218612671,
      "learning_rate": 0.0001999103572300069,
      "loss": 0.2979,
      "step": 4470
    },
    {
      "epoch": 0.06296999086372901,
      "grad_norm": 0.5386064052581787,
      "learning_rate": 0.00019990837872274552,
      "loss": 0.2657,
      "step": 4480
    },
    {
      "epoch": 0.06311054887904982,
      "grad_norm": 0.6669312715530396,
      "learning_rate": 0.00019990637862932816,
      "loss": 0.2939,
      "step": 4490
    },
    {
      "epoch": 0.06325110689437065,
      "grad_norm": 0.5777736306190491,
      "learning_rate": 0.0001999043569501871,
      "loss": 0.2665,
      "step": 4500
    },
    {
      "epoch": 0.06339166490969148,
      "grad_norm": 0.6691046357154846,
      "learning_rate": 0.00019990231368575905,
      "loss": 0.2958,
      "step": 4510
    },
    {
      "epoch": 0.0635322229250123,
      "grad_norm": 0.7404462695121765,
      "learning_rate": 0.0001999002488364855,
      "loss": 0.3013,
      "step": 4520
    },
    {
      "epoch": 0.06367278094033313,
      "grad_norm": 0.7259801626205444,
      "learning_rate": 0.00019989816240281255,
      "loss": 0.327,
      "step": 4530
    },
    {
      "epoch": 0.06381333895565394,
      "grad_norm": 0.5758121013641357,
      "learning_rate": 0.00019989605438519106,
      "loss": 0.2552,
      "step": 4540
    },
    {
      "epoch": 0.06395389697097477,
      "grad_norm": 0.6627341508865356,
      "learning_rate": 0.00019989392478407642,
      "loss": 0.2732,
      "step": 4550
    },
    {
      "epoch": 0.0640944549862956,
      "grad_norm": 0.6391485929489136,
      "learning_rate": 0.0001998917735999288,
      "loss": 0.2782,
      "step": 4560
    },
    {
      "epoch": 0.06423501300161642,
      "grad_norm": 0.6049222946166992,
      "learning_rate": 0.00019988960083321296,
      "loss": 0.2793,
      "step": 4570
    },
    {
      "epoch": 0.06437557101693724,
      "grad_norm": 0.5777444839477539,
      "learning_rate": 0.00019988740648439835,
      "loss": 0.2675,
      "step": 4580
    },
    {
      "epoch": 0.06451612903225806,
      "grad_norm": 0.43069472908973694,
      "learning_rate": 0.00019988519055395907,
      "loss": 0.2698,
      "step": 4590
    },
    {
      "epoch": 0.06465668704757889,
      "grad_norm": 0.5715444087982178,
      "learning_rate": 0.0001998829530423739,
      "loss": 0.2706,
      "step": 4600
    },
    {
      "epoch": 0.06479724506289972,
      "grad_norm": 0.7473022937774658,
      "learning_rate": 0.00019988069395012628,
      "loss": 0.2802,
      "step": 4610
    },
    {
      "epoch": 0.06493780307822053,
      "grad_norm": 0.5839947462081909,
      "learning_rate": 0.0001998784132777043,
      "loss": 0.2734,
      "step": 4620
    },
    {
      "epoch": 0.06507836109354136,
      "grad_norm": 0.7734289169311523,
      "learning_rate": 0.00019987611102560075,
      "loss": 0.2694,
      "step": 4630
    },
    {
      "epoch": 0.06521891910886218,
      "grad_norm": 0.5843231081962585,
      "learning_rate": 0.00019987378719431303,
      "loss": 0.2721,
      "step": 4640
    },
    {
      "epoch": 0.06535947712418301,
      "grad_norm": 0.7259225249290466,
      "learning_rate": 0.0001998714417843432,
      "loss": 0.2629,
      "step": 4650
    },
    {
      "epoch": 0.06550003513950382,
      "grad_norm": 0.8221696615219116,
      "learning_rate": 0.00019986907479619808,
      "loss": 0.2801,
      "step": 4660
    },
    {
      "epoch": 0.06564059315482465,
      "grad_norm": 0.4843771159648895,
      "learning_rate": 0.00019986668623038901,
      "loss": 0.2645,
      "step": 4670
    },
    {
      "epoch": 0.06578115117014548,
      "grad_norm": 0.3744613826274872,
      "learning_rate": 0.0001998642760874321,
      "loss": 0.2492,
      "step": 4680
    },
    {
      "epoch": 0.0659217091854663,
      "grad_norm": 0.5203977227210999,
      "learning_rate": 0.00019986184436784804,
      "loss": 0.2524,
      "step": 4690
    },
    {
      "epoch": 0.06606226720078713,
      "grad_norm": 0.5358824729919434,
      "learning_rate": 0.00019985939107216227,
      "loss": 0.2906,
      "step": 4700
    },
    {
      "epoch": 0.06620282521610794,
      "grad_norm": 0.6105577945709229,
      "learning_rate": 0.00019985691620090488,
      "loss": 0.2964,
      "step": 4710
    },
    {
      "epoch": 0.06634338323142877,
      "grad_norm": 0.6519573330879211,
      "learning_rate": 0.00019985441975461048,
      "loss": 0.2943,
      "step": 4720
    },
    {
      "epoch": 0.0664839412467496,
      "grad_norm": 0.6661750674247742,
      "learning_rate": 0.00019985190173381855,
      "loss": 0.2872,
      "step": 4730
    },
    {
      "epoch": 0.06662449926207042,
      "grad_norm": 0.5077973008155823,
      "learning_rate": 0.00019984936213907311,
      "loss": 0.2424,
      "step": 4740
    },
    {
      "epoch": 0.06676505727739124,
      "grad_norm": 0.5132244825363159,
      "learning_rate": 0.00019984680097092283,
      "loss": 0.2675,
      "step": 4750
    },
    {
      "epoch": 0.06690561529271206,
      "grad_norm": 0.6807255148887634,
      "learning_rate": 0.00019984421822992113,
      "loss": 0.2519,
      "step": 4760
    },
    {
      "epoch": 0.06704617330803289,
      "grad_norm": 0.6630058288574219,
      "learning_rate": 0.00019984161391662598,
      "loss": 0.2565,
      "step": 4770
    },
    {
      "epoch": 0.06718673132335372,
      "grad_norm": 0.5398477911949158,
      "learning_rate": 0.00019983898803160006,
      "loss": 0.2892,
      "step": 4780
    },
    {
      "epoch": 0.06732728933867453,
      "grad_norm": 0.4638122022151947,
      "learning_rate": 0.0001998363405754108,
      "loss": 0.3054,
      "step": 4790
    },
    {
      "epoch": 0.06746784735399536,
      "grad_norm": 0.6484492421150208,
      "learning_rate": 0.00019983367154863014,
      "loss": 0.287,
      "step": 4800
    },
    {
      "epoch": 0.06760840536931619,
      "grad_norm": 0.6471605896949768,
      "learning_rate": 0.00019983098095183475,
      "loss": 0.2659,
      "step": 4810
    },
    {
      "epoch": 0.06774896338463701,
      "grad_norm": 0.5113816857337952,
      "learning_rate": 0.00019982826878560598,
      "loss": 0.2951,
      "step": 4820
    },
    {
      "epoch": 0.06788952139995784,
      "grad_norm": 0.6426228880882263,
      "learning_rate": 0.0001998255350505298,
      "loss": 0.2928,
      "step": 4830
    },
    {
      "epoch": 0.06803007941527865,
      "grad_norm": 0.5491995811462402,
      "learning_rate": 0.0001998227797471969,
      "loss": 0.2524,
      "step": 4840
    },
    {
      "epoch": 0.06817063743059948,
      "grad_norm": 0.5573206543922424,
      "learning_rate": 0.00019982000287620255,
      "loss": 0.2792,
      "step": 4850
    },
    {
      "epoch": 0.0683111954459203,
      "grad_norm": 0.605418860912323,
      "learning_rate": 0.00019981720443814672,
      "loss": 0.2673,
      "step": 4860
    },
    {
      "epoch": 0.06845175346124113,
      "grad_norm": 0.4459189474582672,
      "learning_rate": 0.00019981438443363408,
      "loss": 0.2716,
      "step": 4870
    },
    {
      "epoch": 0.06859231147656195,
      "grad_norm": 0.5312105417251587,
      "learning_rate": 0.00019981154286327385,
      "loss": 0.2874,
      "step": 4880
    },
    {
      "epoch": 0.06873286949188277,
      "grad_norm": 0.5258371233940125,
      "learning_rate": 0.00019980867972768006,
      "loss": 0.2656,
      "step": 4890
    },
    {
      "epoch": 0.0688734275072036,
      "grad_norm": 0.6958247423171997,
      "learning_rate": 0.00019980579502747124,
      "loss": 0.2722,
      "step": 4900
    },
    {
      "epoch": 0.06901398552252443,
      "grad_norm": 0.6064774394035339,
      "learning_rate": 0.0001998028887632707,
      "loss": 0.2788,
      "step": 4910
    },
    {
      "epoch": 0.06915454353784524,
      "grad_norm": 0.5372047424316406,
      "learning_rate": 0.00019979996093570635,
      "loss": 0.2554,
      "step": 4920
    },
    {
      "epoch": 0.06929510155316607,
      "grad_norm": 0.42729809880256653,
      "learning_rate": 0.0001997970115454108,
      "loss": 0.2599,
      "step": 4930
    },
    {
      "epoch": 0.0694356595684869,
      "grad_norm": 0.394705206155777,
      "learning_rate": 0.00019979404059302126,
      "loss": 0.2753,
      "step": 4940
    },
    {
      "epoch": 0.06957621758380772,
      "grad_norm": 0.6101125478744507,
      "learning_rate": 0.00019979104807917966,
      "loss": 0.258,
      "step": 4950
    },
    {
      "epoch": 0.06971677559912855,
      "grad_norm": 0.5726629495620728,
      "learning_rate": 0.00019978803400453254,
      "loss": 0.2968,
      "step": 4960
    },
    {
      "epoch": 0.06985733361444936,
      "grad_norm": 0.5277143120765686,
      "learning_rate": 0.00019978499836973115,
      "loss": 0.2384,
      "step": 4970
    },
    {
      "epoch": 0.06999789162977019,
      "grad_norm": 0.501110851764679,
      "learning_rate": 0.00019978194117543132,
      "loss": 0.2515,
      "step": 4980
    },
    {
      "epoch": 0.07013844964509101,
      "grad_norm": 0.5731441378593445,
      "learning_rate": 0.00019977886242229364,
      "loss": 0.2585,
      "step": 4990
    },
    {
      "epoch": 0.07027900766041184,
      "grad_norm": 0.46598953008651733,
      "learning_rate": 0.00019977576211098325,
      "loss": 0.2707,
      "step": 5000
    },
    {
      "epoch": 0.07027900766041184,
      "eval_chrf": 75.03560060246959,
      "eval_loss": 0.5062581300735474,
      "eval_runtime": 277.3737,
      "eval_samples_per_second": 0.361,
      "eval_steps_per_second": 0.014,
      "step": 5000
    },
    {
      "epoch": 0.07041956567573265,
      "grad_norm": 0.6802473664283752,
      "learning_rate": 0.00019977264024217007,
      "loss": 0.2533,
      "step": 5010
    },
    {
      "epoch": 0.07056012369105348,
      "grad_norm": 0.5293552875518799,
      "learning_rate": 0.00019976949681652854,
      "loss": 0.2683,
      "step": 5020
    },
    {
      "epoch": 0.07070068170637431,
      "grad_norm": 0.48123079538345337,
      "learning_rate": 0.00019976633183473784,
      "loss": 0.258,
      "step": 5030
    },
    {
      "epoch": 0.07084123972169513,
      "grad_norm": 0.5725906491279602,
      "learning_rate": 0.00019976314529748183,
      "loss": 0.2627,
      "step": 5040
    },
    {
      "epoch": 0.07098179773701595,
      "grad_norm": 0.5668321847915649,
      "learning_rate": 0.000199759937205449,
      "loss": 0.2823,
      "step": 5050
    },
    {
      "epoch": 0.07112235575233677,
      "grad_norm": 0.934905469417572,
      "learning_rate": 0.00019975670755933243,
      "loss": 0.2561,
      "step": 5060
    },
    {
      "epoch": 0.0712629137676576,
      "grad_norm": 0.6161137819290161,
      "learning_rate": 0.00019975345635982994,
      "loss": 0.2741,
      "step": 5070
    },
    {
      "epoch": 0.07140347178297843,
      "grad_norm": 0.6338168978691101,
      "learning_rate": 0.00019975018360764398,
      "loss": 0.2402,
      "step": 5080
    },
    {
      "epoch": 0.07154402979829924,
      "grad_norm": 0.48079991340637207,
      "learning_rate": 0.0001997468893034817,
      "loss": 0.2948,
      "step": 5090
    },
    {
      "epoch": 0.07168458781362007,
      "grad_norm": 0.5044981241226196,
      "learning_rate": 0.00019974357344805483,
      "loss": 0.2754,
      "step": 5100
    },
    {
      "epoch": 0.0718251458289409,
      "grad_norm": 0.6133807897567749,
      "learning_rate": 0.0001997402360420798,
      "loss": 0.2572,
      "step": 5110
    },
    {
      "epoch": 0.07196570384426172,
      "grad_norm": 0.500830888748169,
      "learning_rate": 0.00019973687708627766,
      "loss": 0.2973,
      "step": 5120
    },
    {
      "epoch": 0.07210626185958255,
      "grad_norm": 0.5794868469238281,
      "learning_rate": 0.0001997334965813742,
      "loss": 0.2631,
      "step": 5130
    },
    {
      "epoch": 0.07224681987490336,
      "grad_norm": 0.562156081199646,
      "learning_rate": 0.00019973009452809974,
      "loss": 0.2863,
      "step": 5140
    },
    {
      "epoch": 0.07238737789022419,
      "grad_norm": 0.5636243224143982,
      "learning_rate": 0.0001997266709271894,
      "loss": 0.2801,
      "step": 5150
    },
    {
      "epoch": 0.07252793590554502,
      "grad_norm": 0.5850925445556641,
      "learning_rate": 0.0001997232257793828,
      "loss": 0.2621,
      "step": 5160
    },
    {
      "epoch": 0.07266849392086584,
      "grad_norm": 0.4883113503456116,
      "learning_rate": 0.00019971975908542438,
      "loss": 0.2549,
      "step": 5170
    },
    {
      "epoch": 0.07280905193618666,
      "grad_norm": 0.6881444454193115,
      "learning_rate": 0.0001997162708460631,
      "loss": 0.2476,
      "step": 5180
    },
    {
      "epoch": 0.07294960995150748,
      "grad_norm": 0.6542561054229736,
      "learning_rate": 0.00019971276106205268,
      "loss": 0.2735,
      "step": 5190
    },
    {
      "epoch": 0.07309016796682831,
      "grad_norm": 0.5345896482467651,
      "learning_rate": 0.00019970922973415136,
      "loss": 0.2419,
      "step": 5200
    },
    {
      "epoch": 0.07323072598214914,
      "grad_norm": 0.5475381016731262,
      "learning_rate": 0.0001997056768631222,
      "loss": 0.278,
      "step": 5210
    },
    {
      "epoch": 0.07337128399746995,
      "grad_norm": 0.5337420105934143,
      "learning_rate": 0.00019970210244973276,
      "loss": 0.2463,
      "step": 5220
    },
    {
      "epoch": 0.07351184201279078,
      "grad_norm": 0.8173430562019348,
      "learning_rate": 0.00019969850649475537,
      "loss": 0.293,
      "step": 5230
    },
    {
      "epoch": 0.0736524000281116,
      "grad_norm": 0.5612601637840271,
      "learning_rate": 0.000199694888998967,
      "loss": 0.2689,
      "step": 5240
    },
    {
      "epoch": 0.07379295804343243,
      "grad_norm": 0.47143304347991943,
      "learning_rate": 0.00019969124996314913,
      "loss": 0.2485,
      "step": 5250
    },
    {
      "epoch": 0.07393351605875326,
      "grad_norm": 0.6005493402481079,
      "learning_rate": 0.00019968758938808813,
      "loss": 0.2505,
      "step": 5260
    },
    {
      "epoch": 0.07407407407407407,
      "grad_norm": 0.49098819494247437,
      "learning_rate": 0.00019968390727457486,
      "loss": 0.2411,
      "step": 5270
    },
    {
      "epoch": 0.0742146320893949,
      "grad_norm": 0.5847885608673096,
      "learning_rate": 0.00019968020362340485,
      "loss": 0.2698,
      "step": 5280
    },
    {
      "epoch": 0.07435519010471572,
      "grad_norm": 0.5103782415390015,
      "learning_rate": 0.00019967647843537837,
      "loss": 0.2587,
      "step": 5290
    },
    {
      "epoch": 0.07449574812003655,
      "grad_norm": 0.6858464479446411,
      "learning_rate": 0.0001996727317113002,
      "loss": 0.262,
      "step": 5300
    },
    {
      "epoch": 0.07463630613535736,
      "grad_norm": 0.7934888005256653,
      "learning_rate": 0.00019966896345197989,
      "loss": 0.2364,
      "step": 5310
    },
    {
      "epoch": 0.07477686415067819,
      "grad_norm": 0.5066108107566833,
      "learning_rate": 0.00019966517365823164,
      "loss": 0.2535,
      "step": 5320
    },
    {
      "epoch": 0.07491742216599902,
      "grad_norm": 0.6084860563278198,
      "learning_rate": 0.00019966136233087425,
      "loss": 0.2713,
      "step": 5330
    },
    {
      "epoch": 0.07505798018131984,
      "grad_norm": 0.5973907709121704,
      "learning_rate": 0.00019965752947073115,
      "loss": 0.286,
      "step": 5340
    },
    {
      "epoch": 0.07519853819664066,
      "grad_norm": 0.5671558380126953,
      "learning_rate": 0.00019965367507863055,
      "loss": 0.2694,
      "step": 5350
    },
    {
      "epoch": 0.07533909621196148,
      "grad_norm": 0.6024190783500671,
      "learning_rate": 0.0001996497991554052,
      "loss": 0.2651,
      "step": 5360
    },
    {
      "epoch": 0.07547965422728231,
      "grad_norm": 0.5919899940490723,
      "learning_rate": 0.00019964590170189246,
      "loss": 0.2575,
      "step": 5370
    },
    {
      "epoch": 0.07562021224260314,
      "grad_norm": 0.6757079362869263,
      "learning_rate": 0.00019964198271893452,
      "loss": 0.24,
      "step": 5380
    },
    {
      "epoch": 0.07576077025792397,
      "grad_norm": 0.5828556418418884,
      "learning_rate": 0.000199638042207378,
      "loss": 0.2384,
      "step": 5390
    },
    {
      "epoch": 0.07590132827324478,
      "grad_norm": 0.6026638746261597,
      "learning_rate": 0.00019963408016807437,
      "loss": 0.2532,
      "step": 5400
    },
    {
      "epoch": 0.0760418862885656,
      "grad_norm": 0.5327057838439941,
      "learning_rate": 0.00019963009660187966,
      "loss": 0.2867,
      "step": 5410
    },
    {
      "epoch": 0.07618244430388643,
      "grad_norm": 0.6412426829338074,
      "learning_rate": 0.0001996260915096545,
      "loss": 0.2697,
      "step": 5420
    },
    {
      "epoch": 0.07632300231920726,
      "grad_norm": 0.5836541652679443,
      "learning_rate": 0.00019962206489226433,
      "loss": 0.2459,
      "step": 5430
    },
    {
      "epoch": 0.07646356033452807,
      "grad_norm": 0.7453212141990662,
      "learning_rate": 0.00019961801675057902,
      "loss": 0.2481,
      "step": 5440
    },
    {
      "epoch": 0.0766041183498489,
      "grad_norm": 0.5052883625030518,
      "learning_rate": 0.0001996139470854733,
      "loss": 0.278,
      "step": 5450
    },
    {
      "epoch": 0.07674467636516973,
      "grad_norm": 0.5511698722839355,
      "learning_rate": 0.00019960985589782641,
      "loss": 0.2621,
      "step": 5460
    },
    {
      "epoch": 0.07688523438049055,
      "grad_norm": 0.7070469856262207,
      "learning_rate": 0.0001996057431885223,
      "loss": 0.2965,
      "step": 5470
    },
    {
      "epoch": 0.07702579239581137,
      "grad_norm": 0.5122061371803284,
      "learning_rate": 0.00019960160895844957,
      "loss": 0.2566,
      "step": 5480
    },
    {
      "epoch": 0.07716635041113219,
      "grad_norm": 0.6440472602844238,
      "learning_rate": 0.00019959745320850144,
      "loss": 0.2799,
      "step": 5490
    },
    {
      "epoch": 0.07730690842645302,
      "grad_norm": 0.5466797351837158,
      "learning_rate": 0.00019959327593957585,
      "loss": 0.2383,
      "step": 5500
    },
    {
      "epoch": 0.07744746644177385,
      "grad_norm": 0.5561037659645081,
      "learning_rate": 0.00019958907715257528,
      "loss": 0.266,
      "step": 5510
    },
    {
      "epoch": 0.07758802445709466,
      "grad_norm": 0.6058019399642944,
      "learning_rate": 0.00019958485684840694,
      "loss": 0.2586,
      "step": 5520
    },
    {
      "epoch": 0.07772858247241549,
      "grad_norm": 0.48346906900405884,
      "learning_rate": 0.00019958061502798268,
      "loss": 0.2653,
      "step": 5530
    },
    {
      "epoch": 0.07786914048773631,
      "grad_norm": 0.5441072583198547,
      "learning_rate": 0.00019957635169221894,
      "loss": 0.2466,
      "step": 5540
    },
    {
      "epoch": 0.07800969850305714,
      "grad_norm": 0.6066705584526062,
      "learning_rate": 0.00019957206684203694,
      "loss": 0.2736,
      "step": 5550
    },
    {
      "epoch": 0.07815025651837797,
      "grad_norm": 0.5180489420890808,
      "learning_rate": 0.00019956776047836238,
      "loss": 0.235,
      "step": 5560
    },
    {
      "epoch": 0.07829081453369878,
      "grad_norm": 0.5487321019172668,
      "learning_rate": 0.0001995634326021257,
      "loss": 0.2201,
      "step": 5570
    },
    {
      "epoch": 0.0784313725490196,
      "grad_norm": 0.5634502172470093,
      "learning_rate": 0.000199559083214262,
      "loss": 0.2468,
      "step": 5580
    },
    {
      "epoch": 0.07857193056434043,
      "grad_norm": 0.5380911827087402,
      "learning_rate": 0.00019955471231571106,
      "loss": 0.2681,
      "step": 5590
    },
    {
      "epoch": 0.07871248857966126,
      "grad_norm": 0.5714244842529297,
      "learning_rate": 0.00019955031990741718,
      "loss": 0.2692,
      "step": 5600
    },
    {
      "epoch": 0.07885304659498207,
      "grad_norm": 0.5239103436470032,
      "learning_rate": 0.0001995459059903294,
      "loss": 0.2682,
      "step": 5610
    },
    {
      "epoch": 0.0789936046103029,
      "grad_norm": 0.6180918216705322,
      "learning_rate": 0.0001995414705654014,
      "loss": 0.2551,
      "step": 5620
    },
    {
      "epoch": 0.07913416262562373,
      "grad_norm": 0.5177292823791504,
      "learning_rate": 0.00019953701363359145,
      "loss": 0.2817,
      "step": 5630
    },
    {
      "epoch": 0.07927472064094455,
      "grad_norm": 0.48171669244766235,
      "learning_rate": 0.00019953253519586258,
      "loss": 0.2391,
      "step": 5640
    },
    {
      "epoch": 0.07941527865626537,
      "grad_norm": 0.6228054761886597,
      "learning_rate": 0.00019952803525318235,
      "loss": 0.2723,
      "step": 5650
    },
    {
      "epoch": 0.0795558366715862,
      "grad_norm": 0.41678664088249207,
      "learning_rate": 0.00019952351380652305,
      "loss": 0.2701,
      "step": 5660
    },
    {
      "epoch": 0.07969639468690702,
      "grad_norm": 0.44831570982933044,
      "learning_rate": 0.00019951897085686158,
      "loss": 0.2713,
      "step": 5670
    },
    {
      "epoch": 0.07983695270222785,
      "grad_norm": 0.5627865195274353,
      "learning_rate": 0.00019951440640517945,
      "loss": 0.2689,
      "step": 5680
    },
    {
      "epoch": 0.07997751071754867,
      "grad_norm": 0.5584320425987244,
      "learning_rate": 0.00019950982045246286,
      "loss": 0.2636,
      "step": 5690
    },
    {
      "epoch": 0.08011806873286949,
      "grad_norm": 0.4371662735939026,
      "learning_rate": 0.0001995052129997027,
      "loss": 0.2666,
      "step": 5700
    },
    {
      "epoch": 0.08025862674819031,
      "grad_norm": 0.6560251116752625,
      "learning_rate": 0.00019950058404789438,
      "loss": 0.2774,
      "step": 5710
    },
    {
      "epoch": 0.08039918476351114,
      "grad_norm": 0.6930556297302246,
      "learning_rate": 0.0001994959335980381,
      "loss": 0.2758,
      "step": 5720
    },
    {
      "epoch": 0.08053974277883197,
      "grad_norm": 0.5577618479728699,
      "learning_rate": 0.0001994912616511386,
      "loss": 0.258,
      "step": 5730
    },
    {
      "epoch": 0.08068030079415278,
      "grad_norm": 0.45983192324638367,
      "learning_rate": 0.0001994865682082053,
      "loss": 0.269,
      "step": 5740
    },
    {
      "epoch": 0.08082085880947361,
      "grad_norm": 0.5181092023849487,
      "learning_rate": 0.00019948185327025225,
      "loss": 0.2596,
      "step": 5750
    },
    {
      "epoch": 0.08096141682479444,
      "grad_norm": 0.5390303730964661,
      "learning_rate": 0.0001994771168382982,
      "loss": 0.2678,
      "step": 5760
    },
    {
      "epoch": 0.08110197484011526,
      "grad_norm": 0.48422420024871826,
      "learning_rate": 0.00019947235891336644,
      "loss": 0.253,
      "step": 5770
    },
    {
      "epoch": 0.08124253285543608,
      "grad_norm": 0.5115424990653992,
      "learning_rate": 0.000199467579496485,
      "loss": 0.2384,
      "step": 5780
    },
    {
      "epoch": 0.0813830908707569,
      "grad_norm": 0.507433295249939,
      "learning_rate": 0.00019946277858868655,
      "loss": 0.2822,
      "step": 5790
    },
    {
      "epoch": 0.08152364888607773,
      "grad_norm": 0.45908185839653015,
      "learning_rate": 0.0001994579561910083,
      "loss": 0.243,
      "step": 5800
    },
    {
      "epoch": 0.08166420690139856,
      "grad_norm": 0.6661983728408813,
      "learning_rate": 0.00019945311230449222,
      "loss": 0.2263,
      "step": 5810
    },
    {
      "epoch": 0.08180476491671937,
      "grad_norm": 0.6793397665023804,
      "learning_rate": 0.00019944824693018487,
      "loss": 0.2618,
      "step": 5820
    },
    {
      "epoch": 0.0819453229320402,
      "grad_norm": 0.5420805811882019,
      "learning_rate": 0.00019944336006913748,
      "loss": 0.2581,
      "step": 5830
    },
    {
      "epoch": 0.08208588094736102,
      "grad_norm": 0.542273759841919,
      "learning_rate": 0.00019943845172240584,
      "loss": 0.2634,
      "step": 5840
    },
    {
      "epoch": 0.08222643896268185,
      "grad_norm": 0.5444084405899048,
      "learning_rate": 0.00019943352189105053,
      "loss": 0.2698,
      "step": 5850
    },
    {
      "epoch": 0.08236699697800268,
      "grad_norm": 0.43644988536834717,
      "learning_rate": 0.00019942857057613663,
      "loss": 0.25,
      "step": 5860
    },
    {
      "epoch": 0.08250755499332349,
      "grad_norm": 0.39292803406715393,
      "learning_rate": 0.00019942359777873393,
      "loss": 0.2566,
      "step": 5870
    },
    {
      "epoch": 0.08264811300864432,
      "grad_norm": 0.736021101474762,
      "learning_rate": 0.00019941860349991688,
      "loss": 0.256,
      "step": 5880
    },
    {
      "epoch": 0.08278867102396514,
      "grad_norm": 0.5877601504325867,
      "learning_rate": 0.0001994135877407645,
      "loss": 0.274,
      "step": 5890
    },
    {
      "epoch": 0.08292922903928597,
      "grad_norm": 0.6260401010513306,
      "learning_rate": 0.00019940855050236048,
      "loss": 0.2566,
      "step": 5900
    },
    {
      "epoch": 0.08306978705460678,
      "grad_norm": 0.6187877655029297,
      "learning_rate": 0.00019940349178579326,
      "loss": 0.249,
      "step": 5910
    },
    {
      "epoch": 0.08321034506992761,
      "grad_norm": 0.5491911768913269,
      "learning_rate": 0.0001993984115921557,
      "loss": 0.2561,
      "step": 5920
    },
    {
      "epoch": 0.08335090308524844,
      "grad_norm": 0.6714176535606384,
      "learning_rate": 0.0001993933099225455,
      "loss": 0.259,
      "step": 5930
    },
    {
      "epoch": 0.08349146110056926,
      "grad_norm": 0.5543636083602905,
      "learning_rate": 0.00019938818677806488,
      "loss": 0.2774,
      "step": 5940
    },
    {
      "epoch": 0.08363201911589008,
      "grad_norm": 0.582870602607727,
      "learning_rate": 0.0001993830421598208,
      "loss": 0.2503,
      "step": 5950
    },
    {
      "epoch": 0.0837725771312109,
      "grad_norm": 0.5580389499664307,
      "learning_rate": 0.00019937787606892475,
      "loss": 0.2302,
      "step": 5960
    },
    {
      "epoch": 0.08391313514653173,
      "grad_norm": 0.5439727902412415,
      "learning_rate": 0.00019937268850649294,
      "loss": 0.2659,
      "step": 5970
    },
    {
      "epoch": 0.08405369316185256,
      "grad_norm": 0.4882200062274933,
      "learning_rate": 0.0001993674794736462,
      "loss": 0.2467,
      "step": 5980
    },
    {
      "epoch": 0.08419425117717338,
      "grad_norm": 0.5599921345710754,
      "learning_rate": 0.00019936224897150996,
      "loss": 0.2742,
      "step": 5990
    },
    {
      "epoch": 0.0843348091924942,
      "grad_norm": 0.728784441947937,
      "learning_rate": 0.00019935699700121431,
      "loss": 0.2446,
      "step": 6000
    },
    {
      "epoch": 0.0843348091924942,
      "eval_chrf": 77.78453910676699,
      "eval_loss": 0.502712607383728,
      "eval_runtime": 272.5964,
      "eval_samples_per_second": 0.367,
      "eval_steps_per_second": 0.015,
      "step": 6000
    },
    {
      "epoch": 0.08447536720781502,
      "grad_norm": 0.42536473274230957,
      "learning_rate": 0.00019935172356389406,
      "loss": 0.2493,
      "step": 6010
    },
    {
      "epoch": 0.08461592522313585,
      "grad_norm": 0.4811732769012451,
      "learning_rate": 0.00019934642866068853,
      "loss": 0.2551,
      "step": 6020
    },
    {
      "epoch": 0.08475648323845668,
      "grad_norm": 0.4817690849304199,
      "learning_rate": 0.00019934111229274174,
      "loss": 0.2505,
      "step": 6030
    },
    {
      "epoch": 0.08489704125377749,
      "grad_norm": 0.515292227268219,
      "learning_rate": 0.00019933577446120236,
      "loss": 0.2572,
      "step": 6040
    },
    {
      "epoch": 0.08503759926909832,
      "grad_norm": 0.5563884377479553,
      "learning_rate": 0.00019933041516722368,
      "loss": 0.2403,
      "step": 6050
    },
    {
      "epoch": 0.08517815728441915,
      "grad_norm": 0.47406917810440063,
      "learning_rate": 0.00019932503441196357,
      "loss": 0.2698,
      "step": 6060
    },
    {
      "epoch": 0.08531871529973997,
      "grad_norm": 0.6517256498336792,
      "learning_rate": 0.00019931963219658468,
      "loss": 0.2986,
      "step": 6070
    },
    {
      "epoch": 0.08545927331506079,
      "grad_norm": 0.5201681852340698,
      "learning_rate": 0.00019931420852225415,
      "loss": 0.2511,
      "step": 6080
    },
    {
      "epoch": 0.08559983133038161,
      "grad_norm": 0.42335227131843567,
      "learning_rate": 0.00019930876339014383,
      "loss": 0.2451,
      "step": 6090
    },
    {
      "epoch": 0.08574038934570244,
      "grad_norm": 0.5125184655189514,
      "learning_rate": 0.0001993032968014302,
      "loss": 0.2551,
      "step": 6100
    },
    {
      "epoch": 0.08588094736102327,
      "grad_norm": 0.6325479745864868,
      "learning_rate": 0.00019929780875729437,
      "loss": 0.262,
      "step": 6110
    },
    {
      "epoch": 0.08602150537634409,
      "grad_norm": 0.4180940091609955,
      "learning_rate": 0.0001992922992589221,
      "loss": 0.2548,
      "step": 6120
    },
    {
      "epoch": 0.0861620633916649,
      "grad_norm": 0.64287269115448,
      "learning_rate": 0.0001992867683075037,
      "loss": 0.2479,
      "step": 6130
    },
    {
      "epoch": 0.08630262140698573,
      "grad_norm": 0.523990273475647,
      "learning_rate": 0.0001992812159042343,
      "loss": 0.2832,
      "step": 6140
    },
    {
      "epoch": 0.08644317942230656,
      "grad_norm": 0.5508052706718445,
      "learning_rate": 0.00019927564205031342,
      "loss": 0.2494,
      "step": 6150
    },
    {
      "epoch": 0.08658373743762739,
      "grad_norm": 0.6613081097602844,
      "learning_rate": 0.00019927004674694543,
      "loss": 0.2536,
      "step": 6160
    },
    {
      "epoch": 0.0867242954529482,
      "grad_norm": 0.627884030342102,
      "learning_rate": 0.00019926442999533924,
      "loss": 0.272,
      "step": 6170
    },
    {
      "epoch": 0.08686485346826903,
      "grad_norm": 0.511834442615509,
      "learning_rate": 0.00019925879179670837,
      "loss": 0.2475,
      "step": 6180
    },
    {
      "epoch": 0.08700541148358985,
      "grad_norm": 0.7373877167701721,
      "learning_rate": 0.00019925313215227102,
      "loss": 0.2579,
      "step": 6190
    },
    {
      "epoch": 0.08714596949891068,
      "grad_norm": 0.5701001286506653,
      "learning_rate": 0.00019924745106325004,
      "loss": 0.2461,
      "step": 6200
    },
    {
      "epoch": 0.0872865275142315,
      "grad_norm": 0.41175466775894165,
      "learning_rate": 0.00019924174853087282,
      "loss": 0.2366,
      "step": 6210
    },
    {
      "epoch": 0.08742708552955232,
      "grad_norm": 0.676011323928833,
      "learning_rate": 0.00019923602455637152,
      "loss": 0.2642,
      "step": 6220
    },
    {
      "epoch": 0.08756764354487315,
      "grad_norm": 0.5405893325805664,
      "learning_rate": 0.0001992302791409828,
      "loss": 0.2344,
      "step": 6230
    },
    {
      "epoch": 0.08770820156019397,
      "grad_norm": 0.5449483394622803,
      "learning_rate": 0.00019922451228594805,
      "loss": 0.2557,
      "step": 6240
    },
    {
      "epoch": 0.08784875957551479,
      "grad_norm": 0.6101166605949402,
      "learning_rate": 0.00019921872399251321,
      "loss": 0.232,
      "step": 6250
    },
    {
      "epoch": 0.08798931759083561,
      "grad_norm": 0.5353050827980042,
      "learning_rate": 0.00019921291426192898,
      "loss": 0.2606,
      "step": 6260
    },
    {
      "epoch": 0.08812987560615644,
      "grad_norm": 0.4918290376663208,
      "learning_rate": 0.00019920708309545054,
      "loss": 0.2597,
      "step": 6270
    },
    {
      "epoch": 0.08827043362147727,
      "grad_norm": 0.5629746913909912,
      "learning_rate": 0.00019920123049433776,
      "loss": 0.2756,
      "step": 6280
    },
    {
      "epoch": 0.0884109916367981,
      "grad_norm": 0.5814275145530701,
      "learning_rate": 0.00019919535645985517,
      "loss": 0.2447,
      "step": 6290
    },
    {
      "epoch": 0.08855154965211891,
      "grad_norm": 0.5752664804458618,
      "learning_rate": 0.00019918946099327194,
      "loss": 0.254,
      "step": 6300
    },
    {
      "epoch": 0.08869210766743973,
      "grad_norm": 0.5221841335296631,
      "learning_rate": 0.00019918354409586178,
      "loss": 0.2445,
      "step": 6310
    },
    {
      "epoch": 0.08883266568276056,
      "grad_norm": 0.5126676559448242,
      "learning_rate": 0.00019917760576890315,
      "loss": 0.2354,
      "step": 6320
    },
    {
      "epoch": 0.08897322369808139,
      "grad_norm": 0.4682322144508362,
      "learning_rate": 0.00019917164601367904,
      "loss": 0.2286,
      "step": 6330
    },
    {
      "epoch": 0.0891137817134022,
      "grad_norm": 0.44143393635749817,
      "learning_rate": 0.00019916566483147715,
      "loss": 0.2326,
      "step": 6340
    },
    {
      "epoch": 0.08925433972872303,
      "grad_norm": 0.5489510893821716,
      "learning_rate": 0.00019915966222358976,
      "loss": 0.2772,
      "step": 6350
    },
    {
      "epoch": 0.08939489774404386,
      "grad_norm": 0.5929551124572754,
      "learning_rate": 0.00019915363819131374,
      "loss": 0.2331,
      "step": 6360
    },
    {
      "epoch": 0.08953545575936468,
      "grad_norm": 0.47340741753578186,
      "learning_rate": 0.0001991475927359507,
      "loss": 0.21,
      "step": 6370
    },
    {
      "epoch": 0.0896760137746855,
      "grad_norm": 0.6008301973342896,
      "learning_rate": 0.00019914152585880684,
      "loss": 0.2413,
      "step": 6380
    },
    {
      "epoch": 0.08981657179000632,
      "grad_norm": 0.5169734954833984,
      "learning_rate": 0.00019913543756119287,
      "loss": 0.239,
      "step": 6390
    },
    {
      "epoch": 0.08995712980532715,
      "grad_norm": 0.4037923216819763,
      "learning_rate": 0.0001991293278444243,
      "loss": 0.2459,
      "step": 6400
    },
    {
      "epoch": 0.09009768782064798,
      "grad_norm": 0.5685014724731445,
      "learning_rate": 0.00019912319670982117,
      "loss": 0.2245,
      "step": 6410
    },
    {
      "epoch": 0.0902382458359688,
      "grad_norm": 0.4846496880054474,
      "learning_rate": 0.0001991170441587082,
      "loss": 0.2427,
      "step": 6420
    },
    {
      "epoch": 0.09037880385128962,
      "grad_norm": 0.5893009305000305,
      "learning_rate": 0.00019911087019241466,
      "loss": 0.2562,
      "step": 6430
    },
    {
      "epoch": 0.09051936186661044,
      "grad_norm": 0.5451236963272095,
      "learning_rate": 0.00019910467481227452,
      "loss": 0.2269,
      "step": 6440
    },
    {
      "epoch": 0.09065991988193127,
      "grad_norm": 0.46445807814598083,
      "learning_rate": 0.00019909845801962636,
      "loss": 0.2583,
      "step": 6450
    },
    {
      "epoch": 0.0908004778972521,
      "grad_norm": 0.38349512219429016,
      "learning_rate": 0.00019909221981581334,
      "loss": 0.2493,
      "step": 6460
    },
    {
      "epoch": 0.09094103591257291,
      "grad_norm": 0.6188830137252808,
      "learning_rate": 0.00019908596020218333,
      "loss": 0.2664,
      "step": 6470
    },
    {
      "epoch": 0.09108159392789374,
      "grad_norm": 0.516822874546051,
      "learning_rate": 0.00019907967918008873,
      "loss": 0.2412,
      "step": 6480
    },
    {
      "epoch": 0.09122215194321456,
      "grad_norm": 0.42812368273735046,
      "learning_rate": 0.00019907337675088664,
      "loss": 0.2495,
      "step": 6490
    },
    {
      "epoch": 0.09136270995853539,
      "grad_norm": 0.5717242360115051,
      "learning_rate": 0.00019906705291593876,
      "loss": 0.249,
      "step": 6500
    },
    {
      "epoch": 0.0915032679738562,
      "grad_norm": 0.48675647377967834,
      "learning_rate": 0.00019906070767661144,
      "loss": 0.2476,
      "step": 6510
    },
    {
      "epoch": 0.09164382598917703,
      "grad_norm": 0.4916194975376129,
      "learning_rate": 0.0001990543410342756,
      "loss": 0.2774,
      "step": 6520
    },
    {
      "epoch": 0.09178438400449786,
      "grad_norm": 0.5784870982170105,
      "learning_rate": 0.0001990479529903068,
      "loss": 0.2368,
      "step": 6530
    },
    {
      "epoch": 0.09192494201981868,
      "grad_norm": 0.5581029653549194,
      "learning_rate": 0.00019904154354608526,
      "loss": 0.275,
      "step": 6540
    },
    {
      "epoch": 0.0920655000351395,
      "grad_norm": 0.4577856957912445,
      "learning_rate": 0.00019903511270299583,
      "loss": 0.2732,
      "step": 6550
    },
    {
      "epoch": 0.09220605805046032,
      "grad_norm": 0.6007733345031738,
      "learning_rate": 0.00019902866046242785,
      "loss": 0.2408,
      "step": 6560
    },
    {
      "epoch": 0.09234661606578115,
      "grad_norm": 0.4733324944972992,
      "learning_rate": 0.0001990221868257755,
      "loss": 0.2442,
      "step": 6570
    },
    {
      "epoch": 0.09248717408110198,
      "grad_norm": 0.5347083210945129,
      "learning_rate": 0.00019901569179443744,
      "loss": 0.2243,
      "step": 6580
    },
    {
      "epoch": 0.0926277320964228,
      "grad_norm": 0.5432078838348389,
      "learning_rate": 0.00019900917536981697,
      "loss": 0.2569,
      "step": 6590
    },
    {
      "epoch": 0.09276829011174362,
      "grad_norm": 0.47041982412338257,
      "learning_rate": 0.00019900263755332203,
      "loss": 0.2623,
      "step": 6600
    },
    {
      "epoch": 0.09290884812706444,
      "grad_norm": 0.709175705909729,
      "learning_rate": 0.00019899607834636518,
      "loss": 0.2454,
      "step": 6610
    },
    {
      "epoch": 0.09304940614238527,
      "grad_norm": 0.5433493256568909,
      "learning_rate": 0.00019898949775036355,
      "loss": 0.2594,
      "step": 6620
    },
    {
      "epoch": 0.0931899641577061,
      "grad_norm": 0.4944099485874176,
      "learning_rate": 0.00019898289576673905,
      "loss": 0.2749,
      "step": 6630
    },
    {
      "epoch": 0.09333052217302691,
      "grad_norm": 0.6038771867752075,
      "learning_rate": 0.000198976272396918,
      "loss": 0.2425,
      "step": 6640
    },
    {
      "epoch": 0.09347108018834774,
      "grad_norm": 0.6890851855278015,
      "learning_rate": 0.00019896962764233148,
      "loss": 0.2524,
      "step": 6650
    },
    {
      "epoch": 0.09361163820366857,
      "grad_norm": 0.6427812576293945,
      "learning_rate": 0.00019896296150441516,
      "loss": 0.2449,
      "step": 6660
    },
    {
      "epoch": 0.09375219621898939,
      "grad_norm": 0.5056873559951782,
      "learning_rate": 0.00019895627398460933,
      "loss": 0.2421,
      "step": 6670
    },
    {
      "epoch": 0.0938927542343102,
      "grad_norm": 0.7854136228561401,
      "learning_rate": 0.00019894956508435886,
      "loss": 0.2575,
      "step": 6680
    },
    {
      "epoch": 0.09403331224963103,
      "grad_norm": 0.5667088627815247,
      "learning_rate": 0.0001989428348051133,
      "loss": 0.2742,
      "step": 6690
    },
    {
      "epoch": 0.09417387026495186,
      "grad_norm": 0.6039605140686035,
      "learning_rate": 0.00019893608314832677,
      "loss": 0.2721,
      "step": 6700
    },
    {
      "epoch": 0.09431442828027269,
      "grad_norm": 0.49026021361351013,
      "learning_rate": 0.00019892931011545803,
      "loss": 0.2341,
      "step": 6710
    },
    {
      "epoch": 0.09445498629559351,
      "grad_norm": 0.4740043878555298,
      "learning_rate": 0.00019892251570797047,
      "loss": 0.2331,
      "step": 6720
    },
    {
      "epoch": 0.09459554431091433,
      "grad_norm": 0.4828124940395355,
      "learning_rate": 0.0001989156999273321,
      "loss": 0.2738,
      "step": 6730
    },
    {
      "epoch": 0.09473610232623515,
      "grad_norm": 0.5850508809089661,
      "learning_rate": 0.0001989088627750155,
      "loss": 0.2336,
      "step": 6740
    },
    {
      "epoch": 0.09487666034155598,
      "grad_norm": 0.48459509015083313,
      "learning_rate": 0.00019890200425249792,
      "loss": 0.252,
      "step": 6750
    },
    {
      "epoch": 0.0950172183568768,
      "grad_norm": 0.4098058342933655,
      "learning_rate": 0.0001988951243612612,
      "loss": 0.2502,
      "step": 6760
    },
    {
      "epoch": 0.09515777637219762,
      "grad_norm": 0.5001391172409058,
      "learning_rate": 0.00019888822310279184,
      "loss": 0.2418,
      "step": 6770
    },
    {
      "epoch": 0.09529833438751845,
      "grad_norm": 0.6043509244918823,
      "learning_rate": 0.0001988813004785809,
      "loss": 0.2552,
      "step": 6780
    },
    {
      "epoch": 0.09543889240283927,
      "grad_norm": 0.4252854287624359,
      "learning_rate": 0.00019887435649012408,
      "loss": 0.2252,
      "step": 6790
    },
    {
      "epoch": 0.0955794504181601,
      "grad_norm": 0.45443910360336304,
      "learning_rate": 0.00019886739113892166,
      "loss": 0.2399,
      "step": 6800
    },
    {
      "epoch": 0.09572000843348091,
      "grad_norm": 0.4590103328227997,
      "learning_rate": 0.0001988604044264786,
      "loss": 0.2294,
      "step": 6810
    },
    {
      "epoch": 0.09586056644880174,
      "grad_norm": 0.4579792320728302,
      "learning_rate": 0.0001988533963543045,
      "loss": 0.2328,
      "step": 6820
    },
    {
      "epoch": 0.09600112446412257,
      "grad_norm": 0.5196878910064697,
      "learning_rate": 0.00019884636692391343,
      "loss": 0.2263,
      "step": 6830
    },
    {
      "epoch": 0.0961416824794434,
      "grad_norm": 0.46159133315086365,
      "learning_rate": 0.0001988393161368242,
      "loss": 0.2534,
      "step": 6840
    },
    {
      "epoch": 0.09628224049476422,
      "grad_norm": 0.582403838634491,
      "learning_rate": 0.0001988322439945602,
      "loss": 0.2528,
      "step": 6850
    },
    {
      "epoch": 0.09642279851008503,
      "grad_norm": 0.5543406009674072,
      "learning_rate": 0.00019882515049864945,
      "loss": 0.2295,
      "step": 6860
    },
    {
      "epoch": 0.09656335652540586,
      "grad_norm": 0.42296671867370605,
      "learning_rate": 0.00019881803565062456,
      "loss": 0.2402,
      "step": 6870
    },
    {
      "epoch": 0.09670391454072669,
      "grad_norm": 0.5111995935440063,
      "learning_rate": 0.00019881089945202274,
      "loss": 0.2117,
      "step": 6880
    },
    {
      "epoch": 0.09684447255604751,
      "grad_norm": 0.40601876378059387,
      "learning_rate": 0.00019880374190438585,
      "loss": 0.2378,
      "step": 6890
    },
    {
      "epoch": 0.09698503057136833,
      "grad_norm": 0.5247124433517456,
      "learning_rate": 0.00019879656300926036,
      "loss": 0.2824,
      "step": 6900
    },
    {
      "epoch": 0.09712558858668915,
      "grad_norm": 0.49566319584846497,
      "learning_rate": 0.0001987893627681973,
      "loss": 0.2467,
      "step": 6910
    },
    {
      "epoch": 0.09726614660200998,
      "grad_norm": 0.582575798034668,
      "learning_rate": 0.0001987821411827524,
      "loss": 0.2451,
      "step": 6920
    },
    {
      "epoch": 0.09740670461733081,
      "grad_norm": 0.6084159016609192,
      "learning_rate": 0.00019877489825448588,
      "loss": 0.239,
      "step": 6930
    },
    {
      "epoch": 0.09754726263265162,
      "grad_norm": 0.5342608094215393,
      "learning_rate": 0.00019876763398496273,
      "loss": 0.2594,
      "step": 6940
    },
    {
      "epoch": 0.09768782064797245,
      "grad_norm": 0.5405324101448059,
      "learning_rate": 0.0001987603483757524,
      "loss": 0.2271,
      "step": 6950
    },
    {
      "epoch": 0.09782837866329328,
      "grad_norm": 0.505929708480835,
      "learning_rate": 0.00019875304142842905,
      "loss": 0.252,
      "step": 6960
    },
    {
      "epoch": 0.0979689366786141,
      "grad_norm": 0.5125372409820557,
      "learning_rate": 0.0001987457131445714,
      "loss": 0.2668,
      "step": 6970
    },
    {
      "epoch": 0.09810949469393491,
      "grad_norm": 0.5413892865180969,
      "learning_rate": 0.0001987383635257628,
      "loss": 0.2326,
      "step": 6980
    },
    {
      "epoch": 0.09825005270925574,
      "grad_norm": 0.6026713848114014,
      "learning_rate": 0.00019873099257359123,
      "loss": 0.2719,
      "step": 6990
    },
    {
      "epoch": 0.09839061072457657,
      "grad_norm": 0.5687562227249146,
      "learning_rate": 0.0001987236002896492,
      "loss": 0.2375,
      "step": 7000
    },
    {
      "epoch": 0.09839061072457657,
      "eval_chrf": 79.72353346065984,
      "eval_loss": 0.4819222390651703,
      "eval_runtime": 196.8304,
      "eval_samples_per_second": 0.508,
      "eval_steps_per_second": 0.02,
      "step": 7000
    },
    {
      "epoch": 0.0985311687398974,
      "grad_norm": 0.6272669434547424,
      "learning_rate": 0.0001987161866755339,
      "loss": 0.2573,
      "step": 7010
    },
    {
      "epoch": 0.09867172675521822,
      "grad_norm": 0.43905067443847656,
      "learning_rate": 0.00019870875173284716,
      "loss": 0.2667,
      "step": 7020
    },
    {
      "epoch": 0.09881228477053904,
      "grad_norm": 0.5508715510368347,
      "learning_rate": 0.00019870129546319533,
      "loss": 0.2484,
      "step": 7030
    },
    {
      "epoch": 0.09895284278585986,
      "grad_norm": 0.5342567563056946,
      "learning_rate": 0.00019869381786818938,
      "loss": 0.2521,
      "step": 7040
    },
    {
      "epoch": 0.09909340080118069,
      "grad_norm": 0.5154755711555481,
      "learning_rate": 0.00019868631894944498,
      "loss": 0.2399,
      "step": 7050
    },
    {
      "epoch": 0.09923395881650152,
      "grad_norm": 0.40197110176086426,
      "learning_rate": 0.0001986787987085823,
      "loss": 0.2216,
      "step": 7060
    },
    {
      "epoch": 0.09937451683182233,
      "grad_norm": 0.4556470811367035,
      "learning_rate": 0.00019867125714722618,
      "loss": 0.2395,
      "step": 7070
    },
    {
      "epoch": 0.09951507484714316,
      "grad_norm": 0.46725577116012573,
      "learning_rate": 0.00019866369426700602,
      "loss": 0.2501,
      "step": 7080
    },
    {
      "epoch": 0.09965563286246398,
      "grad_norm": 0.6671276688575745,
      "learning_rate": 0.0001986561100695559,
      "loss": 0.2496,
      "step": 7090
    },
    {
      "epoch": 0.09979619087778481,
      "grad_norm": 0.4392211437225342,
      "learning_rate": 0.00019864850455651438,
      "loss": 0.2292,
      "step": 7100
    },
    {
      "epoch": 0.09993674889310562,
      "grad_norm": 0.5096685886383057,
      "learning_rate": 0.00019864087772952479,
      "loss": 0.2348,
      "step": 7110
    },
    {
      "epoch": 0.10007730690842645,
      "grad_norm": 0.6072500944137573,
      "learning_rate": 0.00019863322959023493,
      "loss": 0.2564,
      "step": 7120
    },
    {
      "epoch": 0.10021786492374728,
      "grad_norm": 0.46237069368362427,
      "learning_rate": 0.00019862556014029725,
      "loss": 0.2228,
      "step": 7130
    },
    {
      "epoch": 0.1003584229390681,
      "grad_norm": 0.527052640914917,
      "learning_rate": 0.00019861786938136882,
      "loss": 0.2264,
      "step": 7140
    },
    {
      "epoch": 0.10049898095438893,
      "grad_norm": 0.5248671770095825,
      "learning_rate": 0.0001986101573151113,
      "loss": 0.2181,
      "step": 7150
    },
    {
      "epoch": 0.10063953896970974,
      "grad_norm": 0.5087975263595581,
      "learning_rate": 0.00019860242394319097,
      "loss": 0.2399,
      "step": 7160
    },
    {
      "epoch": 0.10078009698503057,
      "grad_norm": 0.6083720922470093,
      "learning_rate": 0.0001985946692672787,
      "loss": 0.25,
      "step": 7170
    },
    {
      "epoch": 0.1009206550003514,
      "grad_norm": 0.427260160446167,
      "learning_rate": 0.0001985868932890499,
      "loss": 0.2368,
      "step": 7180
    },
    {
      "epoch": 0.10106121301567222,
      "grad_norm": 0.5379019975662231,
      "learning_rate": 0.00019857909601018472,
      "loss": 0.2441,
      "step": 7190
    },
    {
      "epoch": 0.10120177103099304,
      "grad_norm": 0.4616182744503021,
      "learning_rate": 0.0001985712774323678,
      "loss": 0.2326,
      "step": 7200
    },
    {
      "epoch": 0.10134232904631386,
      "grad_norm": 0.509182870388031,
      "learning_rate": 0.00019856343755728846,
      "loss": 0.2416,
      "step": 7210
    },
    {
      "epoch": 0.10148288706163469,
      "grad_norm": 0.4154062569141388,
      "learning_rate": 0.0001985555763866405,
      "loss": 0.2091,
      "step": 7220
    },
    {
      "epoch": 0.10162344507695552,
      "grad_norm": 0.4568565785884857,
      "learning_rate": 0.00019854769392212246,
      "loss": 0.2101,
      "step": 7230
    },
    {
      "epoch": 0.10176400309227633,
      "grad_norm": 0.5902854800224304,
      "learning_rate": 0.0001985397901654374,
      "loss": 0.2583,
      "step": 7240
    },
    {
      "epoch": 0.10190456110759716,
      "grad_norm": 0.4994632303714752,
      "learning_rate": 0.00019853186511829303,
      "loss": 0.2377,
      "step": 7250
    },
    {
      "epoch": 0.10204511912291798,
      "grad_norm": 0.4629485011100769,
      "learning_rate": 0.0001985239187824016,
      "loss": 0.2637,
      "step": 7260
    },
    {
      "epoch": 0.10218567713823881,
      "grad_norm": 0.5010542273521423,
      "learning_rate": 0.00019851595115948,
      "loss": 0.2284,
      "step": 7270
    },
    {
      "epoch": 0.10232623515355962,
      "grad_norm": 0.5815369486808777,
      "learning_rate": 0.0001985079622512497,
      "loss": 0.2589,
      "step": 7280
    },
    {
      "epoch": 0.10246679316888045,
      "grad_norm": 0.5223342776298523,
      "learning_rate": 0.00019849995205943683,
      "loss": 0.1986,
      "step": 7290
    },
    {
      "epoch": 0.10260735118420128,
      "grad_norm": 0.47997236251831055,
      "learning_rate": 0.000198491920585772,
      "loss": 0.2295,
      "step": 7300
    },
    {
      "epoch": 0.1027479091995221,
      "grad_norm": 0.4088563621044159,
      "learning_rate": 0.00019848386783199055,
      "loss": 0.2579,
      "step": 7310
    },
    {
      "epoch": 0.10288846721484293,
      "grad_norm": 0.47599363327026367,
      "learning_rate": 0.00019847579379983228,
      "loss": 0.2112,
      "step": 7320
    },
    {
      "epoch": 0.10302902523016375,
      "grad_norm": 0.4933672249317169,
      "learning_rate": 0.00019846769849104172,
      "loss": 0.2191,
      "step": 7330
    },
    {
      "epoch": 0.10316958324548457,
      "grad_norm": 0.5240706205368042,
      "learning_rate": 0.00019845958190736792,
      "loss": 0.2438,
      "step": 7340
    },
    {
      "epoch": 0.1033101412608054,
      "grad_norm": 0.49246230721473694,
      "learning_rate": 0.00019845144405056456,
      "loss": 0.2165,
      "step": 7350
    },
    {
      "epoch": 0.10345069927612623,
      "grad_norm": 0.4593331813812256,
      "learning_rate": 0.00019844328492238985,
      "loss": 0.2268,
      "step": 7360
    },
    {
      "epoch": 0.10359125729144704,
      "grad_norm": 0.46987101435661316,
      "learning_rate": 0.00019843510452460672,
      "loss": 0.2433,
      "step": 7370
    },
    {
      "epoch": 0.10373181530676787,
      "grad_norm": 0.39112797379493713,
      "learning_rate": 0.00019842690285898256,
      "loss": 0.239,
      "step": 7380
    },
    {
      "epoch": 0.10387237332208869,
      "grad_norm": 0.43101245164871216,
      "learning_rate": 0.00019841867992728948,
      "loss": 0.2328,
      "step": 7390
    },
    {
      "epoch": 0.10401293133740952,
      "grad_norm": 0.4474431872367859,
      "learning_rate": 0.00019841043573130409,
      "loss": 0.2493,
      "step": 7400
    },
    {
      "epoch": 0.10415348935273033,
      "grad_norm": 0.6698712110519409,
      "learning_rate": 0.0001984021702728076,
      "loss": 0.2216,
      "step": 7410
    },
    {
      "epoch": 0.10429404736805116,
      "grad_norm": 0.6470811367034912,
      "learning_rate": 0.00019839388355358587,
      "loss": 0.2573,
      "step": 7420
    },
    {
      "epoch": 0.10443460538337199,
      "grad_norm": 0.5258649587631226,
      "learning_rate": 0.0001983855755754293,
      "loss": 0.2524,
      "step": 7430
    },
    {
      "epoch": 0.10457516339869281,
      "grad_norm": 0.4496793746948242,
      "learning_rate": 0.00019837724634013299,
      "loss": 0.2452,
      "step": 7440
    },
    {
      "epoch": 0.10471572141401364,
      "grad_norm": 0.4539211094379425,
      "learning_rate": 0.00019836889584949643,
      "loss": 0.2345,
      "step": 7450
    },
    {
      "epoch": 0.10485627942933445,
      "grad_norm": 0.44238775968551636,
      "learning_rate": 0.0001983605241053239,
      "loss": 0.2284,
      "step": 7460
    },
    {
      "epoch": 0.10499683744465528,
      "grad_norm": 0.7367810010910034,
      "learning_rate": 0.0001983521311094242,
      "loss": 0.2167,
      "step": 7470
    },
    {
      "epoch": 0.10513739545997611,
      "grad_norm": 0.6223915815353394,
      "learning_rate": 0.00019834371686361068,
      "loss": 0.2394,
      "step": 7480
    },
    {
      "epoch": 0.10527795347529693,
      "grad_norm": 0.5297920107841492,
      "learning_rate": 0.00019833528136970133,
      "loss": 0.2371,
      "step": 7490
    },
    {
      "epoch": 0.10541851149061775,
      "grad_norm": 0.6213352084159851,
      "learning_rate": 0.00019832682462951872,
      "loss": 0.2323,
      "step": 7500
    },
    {
      "epoch": 0.10555906950593857,
      "grad_norm": 0.6057639122009277,
      "learning_rate": 0.00019831834664489,
      "loss": 0.2644,
      "step": 7510
    },
    {
      "epoch": 0.1056996275212594,
      "grad_norm": 0.43835896253585815,
      "learning_rate": 0.00019830984741764693,
      "loss": 0.248,
      "step": 7520
    },
    {
      "epoch": 0.10584018553658023,
      "grad_norm": 0.40223389863967896,
      "learning_rate": 0.00019830132694962586,
      "loss": 0.2213,
      "step": 7530
    },
    {
      "epoch": 0.10598074355190104,
      "grad_norm": 0.545213520526886,
      "learning_rate": 0.0001982927852426677,
      "loss": 0.2464,
      "step": 7540
    },
    {
      "epoch": 0.10612130156722187,
      "grad_norm": 0.6962087750434875,
      "learning_rate": 0.00019828422229861798,
      "loss": 0.2557,
      "step": 7550
    },
    {
      "epoch": 0.1062618595825427,
      "grad_norm": 0.4490336775779724,
      "learning_rate": 0.00019827563811932678,
      "loss": 0.2205,
      "step": 7560
    },
    {
      "epoch": 0.10640241759786352,
      "grad_norm": 0.668609082698822,
      "learning_rate": 0.00019826703270664885,
      "loss": 0.2255,
      "step": 7570
    },
    {
      "epoch": 0.10654297561318435,
      "grad_norm": 0.577055811882019,
      "learning_rate": 0.00019825840606244342,
      "loss": 0.2412,
      "step": 7580
    },
    {
      "epoch": 0.10668353362850516,
      "grad_norm": 0.4610646665096283,
      "learning_rate": 0.00019824975818857435,
      "loss": 0.2206,
      "step": 7590
    },
    {
      "epoch": 0.10682409164382599,
      "grad_norm": 0.5945837497711182,
      "learning_rate": 0.0001982410890869101,
      "loss": 0.2382,
      "step": 7600
    },
    {
      "epoch": 0.10696464965914682,
      "grad_norm": 0.48687273263931274,
      "learning_rate": 0.00019823239875932377,
      "loss": 0.2252,
      "step": 7610
    },
    {
      "epoch": 0.10710520767446764,
      "grad_norm": 0.4183899462223053,
      "learning_rate": 0.00019822368720769293,
      "loss": 0.2288,
      "step": 7620
    },
    {
      "epoch": 0.10724576568978846,
      "grad_norm": 0.39779701828956604,
      "learning_rate": 0.0001982149544338998,
      "loss": 0.2239,
      "step": 7630
    },
    {
      "epoch": 0.10738632370510928,
      "grad_norm": 0.4888986349105835,
      "learning_rate": 0.00019820620043983124,
      "loss": 0.2267,
      "step": 7640
    },
    {
      "epoch": 0.10752688172043011,
      "grad_norm": 0.43206343054771423,
      "learning_rate": 0.00019819742522737854,
      "loss": 0.263,
      "step": 7650
    },
    {
      "epoch": 0.10766743973575094,
      "grad_norm": 0.49468815326690674,
      "learning_rate": 0.0001981886287984377,
      "loss": 0.2393,
      "step": 7660
    },
    {
      "epoch": 0.10780799775107175,
      "grad_norm": 0.4747503697872162,
      "learning_rate": 0.00019817981115490932,
      "loss": 0.2164,
      "step": 7670
    },
    {
      "epoch": 0.10794855576639258,
      "grad_norm": 0.5016836524009705,
      "learning_rate": 0.00019817097229869849,
      "loss": 0.2225,
      "step": 7680
    },
    {
      "epoch": 0.1080891137817134,
      "grad_norm": 0.47500646114349365,
      "learning_rate": 0.00019816211223171493,
      "loss": 0.2545,
      "step": 7690
    },
    {
      "epoch": 0.10822967179703423,
      "grad_norm": 0.4425160884857178,
      "learning_rate": 0.00019815323095587295,
      "loss": 0.2328,
      "step": 7700
    },
    {
      "epoch": 0.10837022981235504,
      "grad_norm": 0.4525761008262634,
      "learning_rate": 0.00019814432847309145,
      "loss": 0.215,
      "step": 7710
    },
    {
      "epoch": 0.10851078782767587,
      "grad_norm": 0.5745055079460144,
      "learning_rate": 0.00019813540478529385,
      "loss": 0.2292,
      "step": 7720
    },
    {
      "epoch": 0.1086513458429967,
      "grad_norm": 0.5081249475479126,
      "learning_rate": 0.00019812645989440827,
      "loss": 0.2418,
      "step": 7730
    },
    {
      "epoch": 0.10879190385831752,
      "grad_norm": 0.4926249384880066,
      "learning_rate": 0.00019811749380236728,
      "loss": 0.2463,
      "step": 7740
    },
    {
      "epoch": 0.10893246187363835,
      "grad_norm": 0.5166073441505432,
      "learning_rate": 0.00019810850651110807,
      "loss": 0.2275,
      "step": 7750
    },
    {
      "epoch": 0.10907301988895916,
      "grad_norm": 0.35917648673057556,
      "learning_rate": 0.0001980994980225725,
      "loss": 0.2274,
      "step": 7760
    },
    {
      "epoch": 0.10921357790427999,
      "grad_norm": 0.39934268593788147,
      "learning_rate": 0.00019809046833870692,
      "loss": 0.2674,
      "step": 7770
    },
    {
      "epoch": 0.10935413591960082,
      "grad_norm": 0.42677411437034607,
      "learning_rate": 0.00019808141746146225,
      "loss": 0.2292,
      "step": 7780
    },
    {
      "epoch": 0.10949469393492164,
      "grad_norm": 0.5277503132820129,
      "learning_rate": 0.00019807234539279405,
      "loss": 0.2441,
      "step": 7790
    },
    {
      "epoch": 0.10963525195024246,
      "grad_norm": 0.49637770652770996,
      "learning_rate": 0.00019806325213466238,
      "loss": 0.2547,
      "step": 7800
    },
    {
      "epoch": 0.10977580996556328,
      "grad_norm": 0.43198898434638977,
      "learning_rate": 0.00019805413768903197,
      "loss": 0.2427,
      "step": 7810
    },
    {
      "epoch": 0.10991636798088411,
      "grad_norm": 0.6287575364112854,
      "learning_rate": 0.00019804500205787207,
      "loss": 0.2561,
      "step": 7820
    },
    {
      "epoch": 0.11005692599620494,
      "grad_norm": 0.5387474894523621,
      "learning_rate": 0.0001980358452431565,
      "loss": 0.2219,
      "step": 7830
    },
    {
      "epoch": 0.11019748401152575,
      "grad_norm": 0.5207616686820984,
      "learning_rate": 0.0001980266672468637,
      "loss": 0.2252,
      "step": 7840
    },
    {
      "epoch": 0.11033804202684658,
      "grad_norm": 0.6341081261634827,
      "learning_rate": 0.00019801746807097668,
      "loss": 0.2415,
      "step": 7850
    },
    {
      "epoch": 0.1104786000421674,
      "grad_norm": 0.49170592427253723,
      "learning_rate": 0.00019800824771748298,
      "loss": 0.2483,
      "step": 7860
    },
    {
      "epoch": 0.11061915805748823,
      "grad_norm": 0.4697312116622925,
      "learning_rate": 0.00019799900618837474,
      "loss": 0.2623,
      "step": 7870
    },
    {
      "epoch": 0.11075971607280906,
      "grad_norm": 0.5614626407623291,
      "learning_rate": 0.0001979897434856487,
      "loss": 0.2268,
      "step": 7880
    },
    {
      "epoch": 0.11090027408812987,
      "grad_norm": 0.5126927495002747,
      "learning_rate": 0.00019798045961130615,
      "loss": 0.2269,
      "step": 7890
    },
    {
      "epoch": 0.1110408321034507,
      "grad_norm": 0.45797285437583923,
      "learning_rate": 0.00019797115456735297,
      "loss": 0.2573,
      "step": 7900
    },
    {
      "epoch": 0.11118139011877153,
      "grad_norm": 0.4749772250652313,
      "learning_rate": 0.00019796182835579956,
      "loss": 0.2135,
      "step": 7910
    },
    {
      "epoch": 0.11132194813409235,
      "grad_norm": 0.4598739743232727,
      "learning_rate": 0.000197952480978661,
      "loss": 0.2278,
      "step": 7920
    },
    {
      "epoch": 0.11146250614941317,
      "grad_norm": 0.40140441060066223,
      "learning_rate": 0.00019794311243795685,
      "loss": 0.232,
      "step": 7930
    },
    {
      "epoch": 0.11160306416473399,
      "grad_norm": 0.5466649532318115,
      "learning_rate": 0.00019793372273571125,
      "loss": 0.2229,
      "step": 7940
    },
    {
      "epoch": 0.11174362218005482,
      "grad_norm": 0.4565970301628113,
      "learning_rate": 0.00019792431187395294,
      "loss": 0.2664,
      "step": 7950
    },
    {
      "epoch": 0.11188418019537565,
      "grad_norm": 0.4595685303211212,
      "learning_rate": 0.00019791487985471525,
      "loss": 0.2316,
      "step": 7960
    },
    {
      "epoch": 0.11202473821069646,
      "grad_norm": 0.455619752407074,
      "learning_rate": 0.00019790542668003607,
      "loss": 0.2581,
      "step": 7970
    },
    {
      "epoch": 0.11216529622601729,
      "grad_norm": 0.4443734884262085,
      "learning_rate": 0.0001978959523519578,
      "loss": 0.2496,
      "step": 7980
    },
    {
      "epoch": 0.11230585424133811,
      "grad_norm": 0.45769697427749634,
      "learning_rate": 0.0001978864568725275,
      "loss": 0.2285,
      "step": 7990
    },
    {
      "epoch": 0.11244641225665894,
      "grad_norm": 0.4537741541862488,
      "learning_rate": 0.00019787694024379676,
      "loss": 0.2621,
      "step": 8000
    },
    {
      "epoch": 0.11244641225665894,
      "eval_chrf": 82.43581377835828,
      "eval_loss": 0.5067694783210754,
      "eval_runtime": 205.7966,
      "eval_samples_per_second": 0.486,
      "eval_steps_per_second": 0.019,
      "step": 8000
    },
    {
      "epoch": 0.11258697027197975,
      "grad_norm": 0.5132802128791809,
      "learning_rate": 0.0001978674024678217,
      "loss": 0.2356,
      "step": 8010
    },
    {
      "epoch": 0.11272752828730058,
      "grad_norm": 0.47775766253471375,
      "learning_rate": 0.00019785784354666306,
      "loss": 0.2392,
      "step": 8020
    },
    {
      "epoch": 0.1128680863026214,
      "grad_norm": 0.48678943514823914,
      "learning_rate": 0.00019784826348238617,
      "loss": 0.2202,
      "step": 8030
    },
    {
      "epoch": 0.11300864431794223,
      "grad_norm": 0.5569357872009277,
      "learning_rate": 0.00019783866227706084,
      "loss": 0.2233,
      "step": 8040
    },
    {
      "epoch": 0.11314920233326306,
      "grad_norm": 0.46687600016593933,
      "learning_rate": 0.00019782903993276156,
      "loss": 0.2487,
      "step": 8050
    },
    {
      "epoch": 0.11328976034858387,
      "grad_norm": 0.4049544632434845,
      "learning_rate": 0.00019781939645156728,
      "loss": 0.2368,
      "step": 8060
    },
    {
      "epoch": 0.1134303183639047,
      "grad_norm": 0.5318085551261902,
      "learning_rate": 0.00019780973183556162,
      "loss": 0.2557,
      "step": 8070
    },
    {
      "epoch": 0.11357087637922553,
      "grad_norm": 0.46007734537124634,
      "learning_rate": 0.00019780004608683266,
      "loss": 0.2357,
      "step": 8080
    },
    {
      "epoch": 0.11371143439454635,
      "grad_norm": 0.5977789759635925,
      "learning_rate": 0.0001977903392074731,
      "loss": 0.2584,
      "step": 8090
    },
    {
      "epoch": 0.11385199240986717,
      "grad_norm": 0.5007039904594421,
      "learning_rate": 0.00019778061119958023,
      "loss": 0.2112,
      "step": 8100
    },
    {
      "epoch": 0.113992550425188,
      "grad_norm": 0.5760759711265564,
      "learning_rate": 0.00019777086206525585,
      "loss": 0.2459,
      "step": 8110
    },
    {
      "epoch": 0.11413310844050882,
      "grad_norm": 0.43324658274650574,
      "learning_rate": 0.0001977610918066064,
      "loss": 0.2449,
      "step": 8120
    },
    {
      "epoch": 0.11427366645582965,
      "grad_norm": 0.5646788477897644,
      "learning_rate": 0.0001977513004257428,
      "loss": 0.2107,
      "step": 8130
    },
    {
      "epoch": 0.11441422447115046,
      "grad_norm": 0.3953395187854767,
      "learning_rate": 0.00019774148792478058,
      "loss": 0.2319,
      "step": 8140
    },
    {
      "epoch": 0.11455478248647129,
      "grad_norm": 0.3455359637737274,
      "learning_rate": 0.00019773165430583982,
      "loss": 0.2154,
      "step": 8150
    },
    {
      "epoch": 0.11469534050179211,
      "grad_norm": 0.5823870897293091,
      "learning_rate": 0.00019772179957104518,
      "loss": 0.2302,
      "step": 8160
    },
    {
      "epoch": 0.11483589851711294,
      "grad_norm": 0.5114933252334595,
      "learning_rate": 0.00019771192372252584,
      "loss": 0.2567,
      "step": 8170
    },
    {
      "epoch": 0.11497645653243377,
      "grad_norm": 0.48762500286102295,
      "learning_rate": 0.00019770202676241562,
      "loss": 0.2375,
      "step": 8180
    },
    {
      "epoch": 0.11511701454775458,
      "grad_norm": 0.8657709360122681,
      "learning_rate": 0.0001976921086928528,
      "loss": 0.2359,
      "step": 8190
    },
    {
      "epoch": 0.11525757256307541,
      "grad_norm": 0.48863881826400757,
      "learning_rate": 0.00019768216951598032,
      "loss": 0.2253,
      "step": 8200
    },
    {
      "epoch": 0.11539813057839624,
      "grad_norm": 0.4692245125770569,
      "learning_rate": 0.0001976722092339456,
      "loss": 0.239,
      "step": 8210
    },
    {
      "epoch": 0.11553868859371706,
      "grad_norm": 0.5036208033561707,
      "learning_rate": 0.00019766222784890067,
      "loss": 0.2216,
      "step": 8220
    },
    {
      "epoch": 0.11567924660903788,
      "grad_norm": 0.5226303935050964,
      "learning_rate": 0.0001976522253630021,
      "loss": 0.2641,
      "step": 8230
    },
    {
      "epoch": 0.1158198046243587,
      "grad_norm": 0.5416714549064636,
      "learning_rate": 0.00019764220177841103,
      "loss": 0.255,
      "step": 8240
    },
    {
      "epoch": 0.11596036263967953,
      "grad_norm": 0.39930713176727295,
      "learning_rate": 0.0001976321570972931,
      "loss": 0.2399,
      "step": 8250
    },
    {
      "epoch": 0.11610092065500036,
      "grad_norm": 0.4201917350292206,
      "learning_rate": 0.00019762209132181862,
      "loss": 0.2565,
      "step": 8260
    },
    {
      "epoch": 0.11624147867032117,
      "grad_norm": 0.6495237946510315,
      "learning_rate": 0.00019761200445416239,
      "loss": 0.2286,
      "step": 8270
    },
    {
      "epoch": 0.116382036685642,
      "grad_norm": 0.4430442452430725,
      "learning_rate": 0.00019760189649650376,
      "loss": 0.2358,
      "step": 8280
    },
    {
      "epoch": 0.11652259470096282,
      "grad_norm": 0.498177707195282,
      "learning_rate": 0.00019759176745102664,
      "loss": 0.2272,
      "step": 8290
    },
    {
      "epoch": 0.11666315271628365,
      "grad_norm": 0.4552452862262726,
      "learning_rate": 0.0001975816173199195,
      "loss": 0.233,
      "step": 8300
    },
    {
      "epoch": 0.11680371073160448,
      "grad_norm": 0.5571867227554321,
      "learning_rate": 0.00019757144610537544,
      "loss": 0.217,
      "step": 8310
    },
    {
      "epoch": 0.11694426874692529,
      "grad_norm": 0.5300667881965637,
      "learning_rate": 0.00019756125380959193,
      "loss": 0.2378,
      "step": 8320
    },
    {
      "epoch": 0.11708482676224612,
      "grad_norm": 0.5119692087173462,
      "learning_rate": 0.00019755104043477125,
      "loss": 0.2262,
      "step": 8330
    },
    {
      "epoch": 0.11722538477756694,
      "grad_norm": 0.4793112874031067,
      "learning_rate": 0.00019754080598312,
      "loss": 0.2116,
      "step": 8340
    },
    {
      "epoch": 0.11736594279288777,
      "grad_norm": 0.5343843698501587,
      "learning_rate": 0.00019753055045684944,
      "loss": 0.2373,
      "step": 8350
    },
    {
      "epoch": 0.11750650080820858,
      "grad_norm": 0.4279910922050476,
      "learning_rate": 0.0001975202738581754,
      "loss": 0.2469,
      "step": 8360
    },
    {
      "epoch": 0.11764705882352941,
      "grad_norm": 0.49253949522972107,
      "learning_rate": 0.00019750997618931825,
      "loss": 0.245,
      "step": 8370
    },
    {
      "epoch": 0.11778761683885024,
      "grad_norm": 0.4585903584957123,
      "learning_rate": 0.00019749965745250285,
      "loss": 0.2319,
      "step": 8380
    },
    {
      "epoch": 0.11792817485417106,
      "grad_norm": 0.6014941930770874,
      "learning_rate": 0.00019748931764995868,
      "loss": 0.2493,
      "step": 8390
    },
    {
      "epoch": 0.11806873286949188,
      "grad_norm": 0.6054887175559998,
      "learning_rate": 0.0001974789567839198,
      "loss": 0.2592,
      "step": 8400
    },
    {
      "epoch": 0.1182092908848127,
      "grad_norm": 0.6954628229141235,
      "learning_rate": 0.00019746857485662475,
      "loss": 0.2638,
      "step": 8410
    },
    {
      "epoch": 0.11834984890013353,
      "grad_norm": 0.719398021697998,
      "learning_rate": 0.00019745817187031655,
      "loss": 0.2348,
      "step": 8420
    },
    {
      "epoch": 0.11849040691545436,
      "grad_norm": 0.38335222005844116,
      "learning_rate": 0.000197447747827243,
      "loss": 0.2278,
      "step": 8430
    },
    {
      "epoch": 0.11863096493077517,
      "grad_norm": 0.36306431889533997,
      "learning_rate": 0.00019743730272965627,
      "loss": 0.212,
      "step": 8440
    },
    {
      "epoch": 0.118771522946096,
      "grad_norm": 0.5828087329864502,
      "learning_rate": 0.00019742683657981308,
      "loss": 0.2816,
      "step": 8450
    },
    {
      "epoch": 0.11891208096141682,
      "grad_norm": 0.535893440246582,
      "learning_rate": 0.00019741634937997478,
      "loss": 0.236,
      "step": 8460
    },
    {
      "epoch": 0.11905263897673765,
      "grad_norm": 0.4915892779827118,
      "learning_rate": 0.0001974058411324072,
      "loss": 0.2408,
      "step": 8470
    },
    {
      "epoch": 0.11919319699205848,
      "grad_norm": 0.4109126925468445,
      "learning_rate": 0.0001973953118393808,
      "loss": 0.2082,
      "step": 8480
    },
    {
      "epoch": 0.11933375500737929,
      "grad_norm": 0.49847477674484253,
      "learning_rate": 0.0001973847615031705,
      "loss": 0.2236,
      "step": 8490
    },
    {
      "epoch": 0.11947431302270012,
      "grad_norm": 0.5509682297706604,
      "learning_rate": 0.0001973741901260558,
      "loss": 0.245,
      "step": 8500
    },
    {
      "epoch": 0.11961487103802095,
      "grad_norm": 0.5117948651313782,
      "learning_rate": 0.00019736359771032076,
      "loss": 0.2585,
      "step": 8510
    },
    {
      "epoch": 0.11975542905334177,
      "grad_norm": 0.4626653790473938,
      "learning_rate": 0.00019735298425825392,
      "loss": 0.2572,
      "step": 8520
    },
    {
      "epoch": 0.11989598706866259,
      "grad_norm": 0.7884001135826111,
      "learning_rate": 0.00019734234977214848,
      "loss": 0.2232,
      "step": 8530
    },
    {
      "epoch": 0.12003654508398341,
      "grad_norm": 0.43860548734664917,
      "learning_rate": 0.0001973316942543021,
      "loss": 0.1902,
      "step": 8540
    },
    {
      "epoch": 0.12017710309930424,
      "grad_norm": 0.6320831775665283,
      "learning_rate": 0.000197321017707017,
      "loss": 0.2194,
      "step": 8550
    },
    {
      "epoch": 0.12031766111462507,
      "grad_norm": 0.5021673440933228,
      "learning_rate": 0.0001973103201326,
      "loss": 0.2337,
      "step": 8560
    },
    {
      "epoch": 0.12045821912994588,
      "grad_norm": 0.5387911200523376,
      "learning_rate": 0.0001972996015333623,
      "loss": 0.2245,
      "step": 8570
    },
    {
      "epoch": 0.1205987771452667,
      "grad_norm": 0.5952897071838379,
      "learning_rate": 0.00019728886191161987,
      "loss": 0.2344,
      "step": 8580
    },
    {
      "epoch": 0.12073933516058753,
      "grad_norm": 0.3942081034183502,
      "learning_rate": 0.00019727810126969305,
      "loss": 0.2262,
      "step": 8590
    },
    {
      "epoch": 0.12087989317590836,
      "grad_norm": 0.5513126850128174,
      "learning_rate": 0.0001972673196099068,
      "loss": 0.2509,
      "step": 8600
    },
    {
      "epoch": 0.12102045119122919,
      "grad_norm": 0.5135902762413025,
      "learning_rate": 0.00019725651693459058,
      "loss": 0.2407,
      "step": 8610
    },
    {
      "epoch": 0.12116100920655,
      "grad_norm": 0.46190938353538513,
      "learning_rate": 0.00019724569324607843,
      "loss": 0.2418,
      "step": 8620
    },
    {
      "epoch": 0.12130156722187083,
      "grad_norm": 0.520503044128418,
      "learning_rate": 0.0001972348485467089,
      "loss": 0.2665,
      "step": 8630
    },
    {
      "epoch": 0.12144212523719165,
      "grad_norm": 0.5142205953598022,
      "learning_rate": 0.0001972239828388251,
      "loss": 0.2331,
      "step": 8640
    },
    {
      "epoch": 0.12158268325251248,
      "grad_norm": 0.496604859828949,
      "learning_rate": 0.00019721309612477463,
      "loss": 0.2185,
      "step": 8650
    },
    {
      "epoch": 0.1217232412678333,
      "grad_norm": 0.8093092441558838,
      "learning_rate": 0.00019720218840690973,
      "loss": 0.2721,
      "step": 8660
    },
    {
      "epoch": 0.12186379928315412,
      "grad_norm": 0.4930914044380188,
      "learning_rate": 0.00019719125968758707,
      "loss": 0.241,
      "step": 8670
    },
    {
      "epoch": 0.12200435729847495,
      "grad_norm": 0.44180360436439514,
      "learning_rate": 0.00019718030996916793,
      "loss": 0.2084,
      "step": 8680
    },
    {
      "epoch": 0.12214491531379577,
      "grad_norm": 0.47540080547332764,
      "learning_rate": 0.0001971693392540181,
      "loss": 0.2316,
      "step": 8690
    },
    {
      "epoch": 0.12228547332911659,
      "grad_norm": 0.49588194489479065,
      "learning_rate": 0.00019715834754450791,
      "loss": 0.2561,
      "step": 8700
    },
    {
      "epoch": 0.12242603134443741,
      "grad_norm": 0.33857402205467224,
      "learning_rate": 0.00019714733484301218,
      "loss": 0.2137,
      "step": 8710
    },
    {
      "epoch": 0.12256658935975824,
      "grad_norm": 0.5175547003746033,
      "learning_rate": 0.00019713630115191037,
      "loss": 0.2084,
      "step": 8720
    },
    {
      "epoch": 0.12270714737507907,
      "grad_norm": 0.5434607863426208,
      "learning_rate": 0.00019712524647358637,
      "loss": 0.2329,
      "step": 8730
    },
    {
      "epoch": 0.1228477053903999,
      "grad_norm": 0.5493789315223694,
      "learning_rate": 0.00019711417081042868,
      "loss": 0.2165,
      "step": 8740
    },
    {
      "epoch": 0.12298826340572071,
      "grad_norm": 0.48063433170318604,
      "learning_rate": 0.00019710307416483024,
      "loss": 0.2262,
      "step": 8750
    },
    {
      "epoch": 0.12312882142104153,
      "grad_norm": 0.5307974815368652,
      "learning_rate": 0.0001970919565391887,
      "loss": 0.2445,
      "step": 8760
    },
    {
      "epoch": 0.12326937943636236,
      "grad_norm": 0.8536155223846436,
      "learning_rate": 0.000197080817935906,
      "loss": 0.2391,
      "step": 8770
    },
    {
      "epoch": 0.12340993745168319,
      "grad_norm": 0.49438464641571045,
      "learning_rate": 0.0001970696583573888,
      "loss": 0.222,
      "step": 8780
    },
    {
      "epoch": 0.123550495467004,
      "grad_norm": 0.4848876893520355,
      "learning_rate": 0.0001970584778060483,
      "loss": 0.2287,
      "step": 8790
    },
    {
      "epoch": 0.12369105348232483,
      "grad_norm": 0.408035546541214,
      "learning_rate": 0.00019704727628430004,
      "loss": 0.2271,
      "step": 8800
    },
    {
      "epoch": 0.12383161149764566,
      "grad_norm": 0.5347570776939392,
      "learning_rate": 0.0001970360537945643,
      "loss": 0.2477,
      "step": 8810
    },
    {
      "epoch": 0.12397216951296648,
      "grad_norm": 0.5373921990394592,
      "learning_rate": 0.00019702481033926578,
      "loss": 0.2469,
      "step": 8820
    },
    {
      "epoch": 0.1241127275282873,
      "grad_norm": 0.5754639506340027,
      "learning_rate": 0.00019701354592083377,
      "loss": 0.2607,
      "step": 8830
    },
    {
      "epoch": 0.12425328554360812,
      "grad_norm": 0.5097301006317139,
      "learning_rate": 0.00019700226054170198,
      "loss": 0.2157,
      "step": 8840
    },
    {
      "epoch": 0.12439384355892895,
      "grad_norm": 0.42645829916000366,
      "learning_rate": 0.00019699095420430877,
      "loss": 0.2224,
      "step": 8850
    },
    {
      "epoch": 0.12453440157424978,
      "grad_norm": 0.5002944469451904,
      "learning_rate": 0.00019697962691109697,
      "loss": 0.2162,
      "step": 8860
    },
    {
      "epoch": 0.12467495958957059,
      "grad_norm": 0.42151981592178345,
      "learning_rate": 0.00019696827866451394,
      "loss": 0.2208,
      "step": 8870
    },
    {
      "epoch": 0.12481551760489142,
      "grad_norm": 0.5172638893127441,
      "learning_rate": 0.00019695690946701163,
      "loss": 0.2177,
      "step": 8880
    },
    {
      "epoch": 0.12495607562021224,
      "grad_norm": 0.48275619745254517,
      "learning_rate": 0.00019694551932104642,
      "loss": 0.2216,
      "step": 8890
    },
    {
      "epoch": 0.12509663363553306,
      "grad_norm": 0.4467068314552307,
      "learning_rate": 0.00019693410822907924,
      "loss": 0.2391,
      "step": 8900
    },
    {
      "epoch": 0.1252371916508539,
      "grad_norm": 0.4540744125843048,
      "learning_rate": 0.00019692267619357562,
      "loss": 0.2436,
      "step": 8910
    },
    {
      "epoch": 0.1253777496661747,
      "grad_norm": 0.4542349576950073,
      "learning_rate": 0.00019691122321700553,
      "loss": 0.2116,
      "step": 8920
    },
    {
      "epoch": 0.12551830768149555,
      "grad_norm": 0.48326826095581055,
      "learning_rate": 0.0001968997493018435,
      "loss": 0.2603,
      "step": 8930
    },
    {
      "epoch": 0.12565886569681636,
      "grad_norm": 0.45143458247184753,
      "learning_rate": 0.00019688825445056854,
      "loss": 0.223,
      "step": 8940
    },
    {
      "epoch": 0.12579942371213718,
      "grad_norm": 0.49860867857933044,
      "learning_rate": 0.00019687673866566427,
      "loss": 0.2305,
      "step": 8950
    },
    {
      "epoch": 0.12593998172745802,
      "grad_norm": 0.5772536396980286,
      "learning_rate": 0.0001968652019496188,
      "loss": 0.2452,
      "step": 8960
    },
    {
      "epoch": 0.12608053974277883,
      "grad_norm": 0.5652404427528381,
      "learning_rate": 0.0001968536443049247,
      "loss": 0.2551,
      "step": 8970
    },
    {
      "epoch": 0.12622109775809964,
      "grad_norm": 0.44158807396888733,
      "learning_rate": 0.0001968420657340791,
      "loss": 0.2139,
      "step": 8980
    },
    {
      "epoch": 0.12636165577342048,
      "grad_norm": 0.5304787755012512,
      "learning_rate": 0.0001968304662395837,
      "loss": 0.2375,
      "step": 8990
    },
    {
      "epoch": 0.1265022137887413,
      "grad_norm": 0.4007686376571655,
      "learning_rate": 0.00019681884582394468,
      "loss": 0.2681,
      "step": 9000
    },
    {
      "epoch": 0.1265022137887413,
      "eval_chrf": 83.5463488314016,
      "eval_loss": 0.4594023525714874,
      "eval_runtime": 205.5947,
      "eval_samples_per_second": 0.486,
      "eval_steps_per_second": 0.019,
      "step": 9000
    },
    {
      "epoch": 0.12664277180406214,
      "grad_norm": 0.5894201993942261,
      "learning_rate": 0.00019680720448967271,
      "loss": 0.2327,
      "step": 9010
    },
    {
      "epoch": 0.12678332981938295,
      "grad_norm": 0.5372081398963928,
      "learning_rate": 0.00019679554223928306,
      "loss": 0.2178,
      "step": 9020
    },
    {
      "epoch": 0.12692388783470376,
      "grad_norm": 0.4564507305622101,
      "learning_rate": 0.00019678385907529543,
      "loss": 0.2213,
      "step": 9030
    },
    {
      "epoch": 0.1270644458500246,
      "grad_norm": 0.3771321475505829,
      "learning_rate": 0.00019677215500023406,
      "loss": 0.2248,
      "step": 9040
    },
    {
      "epoch": 0.12720500386534542,
      "grad_norm": 0.5481129884719849,
      "learning_rate": 0.0001967604300166278,
      "loss": 0.2589,
      "step": 9050
    },
    {
      "epoch": 0.12734556188066626,
      "grad_norm": 0.5340000987052917,
      "learning_rate": 0.00019674868412700988,
      "loss": 0.2493,
      "step": 9060
    },
    {
      "epoch": 0.12748611989598707,
      "grad_norm": 0.6218507289886475,
      "learning_rate": 0.00019673691733391814,
      "loss": 0.2186,
      "step": 9070
    },
    {
      "epoch": 0.12762667791130788,
      "grad_norm": 0.4179977476596832,
      "learning_rate": 0.0001967251296398949,
      "loss": 0.2448,
      "step": 9080
    },
    {
      "epoch": 0.12776723592662872,
      "grad_norm": 0.626200795173645,
      "learning_rate": 0.00019671332104748701,
      "loss": 0.2592,
      "step": 9090
    },
    {
      "epoch": 0.12790779394194954,
      "grad_norm": 0.5540094375610352,
      "learning_rate": 0.0001967014915592458,
      "loss": 0.2281,
      "step": 9100
    },
    {
      "epoch": 0.12804835195727035,
      "grad_norm": 0.5310568809509277,
      "learning_rate": 0.00019668964117772718,
      "loss": 0.2474,
      "step": 9110
    },
    {
      "epoch": 0.1281889099725912,
      "grad_norm": 0.44051870703697205,
      "learning_rate": 0.00019667776990549155,
      "loss": 0.2523,
      "step": 9120
    },
    {
      "epoch": 0.128329467987912,
      "grad_norm": 0.39723333716392517,
      "learning_rate": 0.00019666587774510373,
      "loss": 0.2224,
      "step": 9130
    },
    {
      "epoch": 0.12847002600323285,
      "grad_norm": 0.40893715620040894,
      "learning_rate": 0.00019665396469913321,
      "loss": 0.2333,
      "step": 9140
    },
    {
      "epoch": 0.12861058401855366,
      "grad_norm": 0.4723608195781708,
      "learning_rate": 0.00019664203077015389,
      "loss": 0.2293,
      "step": 9150
    },
    {
      "epoch": 0.12875114203387447,
      "grad_norm": 0.4790593087673187,
      "learning_rate": 0.00019663007596074422,
      "loss": 0.2409,
      "step": 9160
    },
    {
      "epoch": 0.1288917000491953,
      "grad_norm": 0.5768985152244568,
      "learning_rate": 0.00019661810027348718,
      "loss": 0.2289,
      "step": 9170
    },
    {
      "epoch": 0.12903225806451613,
      "grad_norm": 0.45519647002220154,
      "learning_rate": 0.00019660610371097014,
      "loss": 0.2169,
      "step": 9180
    },
    {
      "epoch": 0.12917281607983697,
      "grad_norm": 0.5375517010688782,
      "learning_rate": 0.00019659408627578518,
      "loss": 0.2313,
      "step": 9190
    },
    {
      "epoch": 0.12931337409515778,
      "grad_norm": 0.5258593559265137,
      "learning_rate": 0.0001965820479705287,
      "loss": 0.2525,
      "step": 9200
    },
    {
      "epoch": 0.1294539321104786,
      "grad_norm": 0.4912596642971039,
      "learning_rate": 0.0001965699887978017,
      "loss": 0.2692,
      "step": 9210
    },
    {
      "epoch": 0.12959449012579943,
      "grad_norm": 0.5644695162773132,
      "learning_rate": 0.00019655790876020972,
      "loss": 0.2123,
      "step": 9220
    },
    {
      "epoch": 0.12973504814112025,
      "grad_norm": 0.5708533525466919,
      "learning_rate": 0.00019654580786036276,
      "loss": 0.2336,
      "step": 9230
    },
    {
      "epoch": 0.12987560615644106,
      "grad_norm": 0.4405006766319275,
      "learning_rate": 0.00019653368610087532,
      "loss": 0.2438,
      "step": 9240
    },
    {
      "epoch": 0.1300161641717619,
      "grad_norm": 0.6312296986579895,
      "learning_rate": 0.00019652154348436638,
      "loss": 0.2254,
      "step": 9250
    },
    {
      "epoch": 0.1301567221870827,
      "grad_norm": 0.5054198503494263,
      "learning_rate": 0.00019650938001345954,
      "loss": 0.238,
      "step": 9260
    },
    {
      "epoch": 0.13029728020240355,
      "grad_norm": 0.4843730330467224,
      "learning_rate": 0.0001964971956907828,
      "loss": 0.2237,
      "step": 9270
    },
    {
      "epoch": 0.13043783821772437,
      "grad_norm": 0.5578977465629578,
      "learning_rate": 0.00019648499051896868,
      "loss": 0.2551,
      "step": 9280
    },
    {
      "epoch": 0.13057839623304518,
      "grad_norm": 0.6799400448799133,
      "learning_rate": 0.0001964727645006543,
      "loss": 0.2286,
      "step": 9290
    },
    {
      "epoch": 0.13071895424836602,
      "grad_norm": 0.4192991554737091,
      "learning_rate": 0.0001964605176384811,
      "loss": 0.262,
      "step": 9300
    },
    {
      "epoch": 0.13085951226368683,
      "grad_norm": 0.4753544330596924,
      "learning_rate": 0.00019644824993509522,
      "loss": 0.2607,
      "step": 9310
    },
    {
      "epoch": 0.13100007027900765,
      "grad_norm": 0.4694324731826782,
      "learning_rate": 0.00019643596139314715,
      "loss": 0.2221,
      "step": 9320
    },
    {
      "epoch": 0.1311406282943285,
      "grad_norm": 0.43425655364990234,
      "learning_rate": 0.000196423652015292,
      "loss": 0.2448,
      "step": 9330
    },
    {
      "epoch": 0.1312811863096493,
      "grad_norm": 0.30343201756477356,
      "learning_rate": 0.0001964113218041893,
      "loss": 0.2271,
      "step": 9340
    },
    {
      "epoch": 0.13142174432497014,
      "grad_norm": 0.457977831363678,
      "learning_rate": 0.0001963989707625031,
      "loss": 0.2114,
      "step": 9350
    },
    {
      "epoch": 0.13156230234029095,
      "grad_norm": 0.45157763361930847,
      "learning_rate": 0.00019638659889290193,
      "loss": 0.2095,
      "step": 9360
    },
    {
      "epoch": 0.13170286035561177,
      "grad_norm": 0.44693228602409363,
      "learning_rate": 0.00019637420619805895,
      "loss": 0.2295,
      "step": 9370
    },
    {
      "epoch": 0.1318434183709326,
      "grad_norm": 0.5506470203399658,
      "learning_rate": 0.00019636179268065168,
      "loss": 0.2344,
      "step": 9380
    },
    {
      "epoch": 0.13198397638625342,
      "grad_norm": 0.3795108497142792,
      "learning_rate": 0.0001963493583433621,
      "loss": 0.2339,
      "step": 9390
    },
    {
      "epoch": 0.13212453440157426,
      "grad_norm": 0.42435768246650696,
      "learning_rate": 0.00019633690318887688,
      "loss": 0.2245,
      "step": 9400
    },
    {
      "epoch": 0.13226509241689507,
      "grad_norm": 0.48172616958618164,
      "learning_rate": 0.00019632442721988696,
      "loss": 0.2214,
      "step": 9410
    },
    {
      "epoch": 0.1324056504322159,
      "grad_norm": 0.4663597643375397,
      "learning_rate": 0.000196311930439088,
      "loss": 0.238,
      "step": 9420
    },
    {
      "epoch": 0.13254620844753673,
      "grad_norm": 0.43239930272102356,
      "learning_rate": 0.00019629941284918,
      "loss": 0.2342,
      "step": 9430
    },
    {
      "epoch": 0.13268676646285754,
      "grad_norm": 0.6404114961624146,
      "learning_rate": 0.0001962868744528675,
      "loss": 0.2242,
      "step": 9440
    },
    {
      "epoch": 0.13282732447817835,
      "grad_norm": 0.42753544449806213,
      "learning_rate": 0.00019627431525285954,
      "loss": 0.2237,
      "step": 9450
    },
    {
      "epoch": 0.1329678824934992,
      "grad_norm": 0.3860357999801636,
      "learning_rate": 0.00019626173525186968,
      "loss": 0.2327,
      "step": 9460
    },
    {
      "epoch": 0.13310844050882,
      "grad_norm": 0.5065526366233826,
      "learning_rate": 0.00019624913445261594,
      "loss": 0.213,
      "step": 9470
    },
    {
      "epoch": 0.13324899852414085,
      "grad_norm": 0.5617696642875671,
      "learning_rate": 0.0001962365128578208,
      "loss": 0.2382,
      "step": 9480
    },
    {
      "epoch": 0.13338955653946166,
      "grad_norm": 0.4992463290691376,
      "learning_rate": 0.0001962238704702113,
      "loss": 0.2364,
      "step": 9490
    },
    {
      "epoch": 0.13353011455478248,
      "grad_norm": 0.637082040309906,
      "learning_rate": 0.000196211207292519,
      "loss": 0.2254,
      "step": 9500
    },
    {
      "epoch": 0.13367067257010332,
      "grad_norm": 0.43641334772109985,
      "learning_rate": 0.00019619852332747984,
      "loss": 0.2171,
      "step": 9510
    },
    {
      "epoch": 0.13381123058542413,
      "grad_norm": 0.42612147331237793,
      "learning_rate": 0.00019618581857783432,
      "loss": 0.2484,
      "step": 9520
    },
    {
      "epoch": 0.13395178860074497,
      "grad_norm": 0.4482842981815338,
      "learning_rate": 0.00019617309304632743,
      "loss": 0.2294,
      "step": 9530
    },
    {
      "epoch": 0.13409234661606578,
      "grad_norm": 0.5693816542625427,
      "learning_rate": 0.00019616034673570863,
      "loss": 0.2484,
      "step": 9540
    },
    {
      "epoch": 0.1342329046313866,
      "grad_norm": 0.506377100944519,
      "learning_rate": 0.0001961475796487319,
      "loss": 0.1952,
      "step": 9550
    },
    {
      "epoch": 0.13437346264670744,
      "grad_norm": 0.5523819923400879,
      "learning_rate": 0.0001961347917881557,
      "loss": 0.2189,
      "step": 9560
    },
    {
      "epoch": 0.13451402066202825,
      "grad_norm": 0.4156286120414734,
      "learning_rate": 0.00019612198315674295,
      "loss": 0.2076,
      "step": 9570
    },
    {
      "epoch": 0.13465457867734906,
      "grad_norm": 0.38789063692092896,
      "learning_rate": 0.00019610915375726108,
      "loss": 0.2092,
      "step": 9580
    },
    {
      "epoch": 0.1347951366926699,
      "grad_norm": 0.5569458603858948,
      "learning_rate": 0.000196096303592482,
      "loss": 0.2292,
      "step": 9590
    },
    {
      "epoch": 0.13493569470799072,
      "grad_norm": 0.4448249638080597,
      "learning_rate": 0.0001960834326651821,
      "loss": 0.2044,
      "step": 9600
    },
    {
      "epoch": 0.13507625272331156,
      "grad_norm": 0.42959892749786377,
      "learning_rate": 0.0001960705409781423,
      "loss": 0.2375,
      "step": 9610
    },
    {
      "epoch": 0.13521681073863237,
      "grad_norm": 0.4909825325012207,
      "learning_rate": 0.00019605762853414796,
      "loss": 0.2007,
      "step": 9620
    },
    {
      "epoch": 0.13535736875395318,
      "grad_norm": 1.044018268585205,
      "learning_rate": 0.0001960446953359889,
      "loss": 0.2375,
      "step": 9630
    },
    {
      "epoch": 0.13549792676927402,
      "grad_norm": 0.43878230452537537,
      "learning_rate": 0.00019603174138645952,
      "loss": 0.2492,
      "step": 9640
    },
    {
      "epoch": 0.13563848478459484,
      "grad_norm": 0.4571862518787384,
      "learning_rate": 0.0001960187666883586,
      "loss": 0.2401,
      "step": 9650
    },
    {
      "epoch": 0.13577904279991568,
      "grad_norm": 0.510308027267456,
      "learning_rate": 0.00019600577124448946,
      "loss": 0.2349,
      "step": 9660
    },
    {
      "epoch": 0.1359196008152365,
      "grad_norm": 0.3984282910823822,
      "learning_rate": 0.0001959927550576599,
      "loss": 0.2202,
      "step": 9670
    },
    {
      "epoch": 0.1360601588305573,
      "grad_norm": 0.4630577862262726,
      "learning_rate": 0.00019597971813068216,
      "loss": 0.2337,
      "step": 9680
    },
    {
      "epoch": 0.13620071684587814,
      "grad_norm": 0.5899356007575989,
      "learning_rate": 0.00019596666046637303,
      "loss": 0.2292,
      "step": 9690
    },
    {
      "epoch": 0.13634127486119896,
      "grad_norm": 0.5899912714958191,
      "learning_rate": 0.00019595358206755373,
      "loss": 0.241,
      "step": 9700
    },
    {
      "epoch": 0.13648183287651977,
      "grad_norm": 0.5677922368049622,
      "learning_rate": 0.00019594048293704994,
      "loss": 0.2375,
      "step": 9710
    },
    {
      "epoch": 0.1366223908918406,
      "grad_norm": 0.4866439700126648,
      "learning_rate": 0.0001959273630776919,
      "loss": 0.2449,
      "step": 9720
    },
    {
      "epoch": 0.13676294890716142,
      "grad_norm": 0.4721696674823761,
      "learning_rate": 0.00019591422249231425,
      "loss": 0.2277,
      "step": 9730
    },
    {
      "epoch": 0.13690350692248227,
      "grad_norm": 0.3854721188545227,
      "learning_rate": 0.00019590106118375618,
      "loss": 0.2213,
      "step": 9740
    },
    {
      "epoch": 0.13704406493780308,
      "grad_norm": 0.4399317800998688,
      "learning_rate": 0.00019588787915486125,
      "loss": 0.2259,
      "step": 9750
    },
    {
      "epoch": 0.1371846229531239,
      "grad_norm": 0.49207353591918945,
      "learning_rate": 0.0001958746764084776,
      "loss": 0.2135,
      "step": 9760
    },
    {
      "epoch": 0.13732518096844473,
      "grad_norm": 0.47379225492477417,
      "learning_rate": 0.0001958614529474578,
      "loss": 0.2362,
      "step": 9770
    },
    {
      "epoch": 0.13746573898376555,
      "grad_norm": 0.5586134195327759,
      "learning_rate": 0.00019584820877465892,
      "loss": 0.2275,
      "step": 9780
    },
    {
      "epoch": 0.13760629699908639,
      "grad_norm": 0.49579349160194397,
      "learning_rate": 0.0001958349438929425,
      "loss": 0.2083,
      "step": 9790
    },
    {
      "epoch": 0.1377468550144072,
      "grad_norm": 0.38048985600471497,
      "learning_rate": 0.0001958216583051745,
      "loss": 0.2013,
      "step": 9800
    },
    {
      "epoch": 0.137887413029728,
      "grad_norm": 0.48474690318107605,
      "learning_rate": 0.00019580835201422545,
      "loss": 0.2571,
      "step": 9810
    },
    {
      "epoch": 0.13802797104504885,
      "grad_norm": 0.5224998593330383,
      "learning_rate": 0.00019579502502297026,
      "loss": 0.2331,
      "step": 9820
    },
    {
      "epoch": 0.13816852906036967,
      "grad_norm": 0.4451829493045807,
      "learning_rate": 0.00019578167733428834,
      "loss": 0.2139,
      "step": 9830
    },
    {
      "epoch": 0.13830908707569048,
      "grad_norm": 0.47491031885147095,
      "learning_rate": 0.00019576830895106365,
      "loss": 0.2188,
      "step": 9840
    },
    {
      "epoch": 0.13844964509101132,
      "grad_norm": 0.5613846778869629,
      "learning_rate": 0.0001957549198761845,
      "loss": 0.23,
      "step": 9850
    },
    {
      "epoch": 0.13859020310633213,
      "grad_norm": 0.5138101577758789,
      "learning_rate": 0.00019574151011254374,
      "loss": 0.2407,
      "step": 9860
    },
    {
      "epoch": 0.13873076112165297,
      "grad_norm": 0.6644909381866455,
      "learning_rate": 0.0001957280796630387,
      "loss": 0.2388,
      "step": 9870
    },
    {
      "epoch": 0.1388713191369738,
      "grad_norm": 0.4899885058403015,
      "learning_rate": 0.00019571462853057109,
      "loss": 0.2562,
      "step": 9880
    },
    {
      "epoch": 0.1390118771522946,
      "grad_norm": 0.5329784154891968,
      "learning_rate": 0.00019570115671804726,
      "loss": 0.2092,
      "step": 9890
    },
    {
      "epoch": 0.13915243516761544,
      "grad_norm": 0.4139322340488434,
      "learning_rate": 0.00019568766422837787,
      "loss": 0.23,
      "step": 9900
    },
    {
      "epoch": 0.13929299318293625,
      "grad_norm": 0.5224969983100891,
      "learning_rate": 0.00019567415106447805,
      "loss": 0.213,
      "step": 9910
    },
    {
      "epoch": 0.1394335511982571,
      "grad_norm": 0.387287974357605,
      "learning_rate": 0.00019566061722926751,
      "loss": 0.2257,
      "step": 9920
    },
    {
      "epoch": 0.1395741092135779,
      "grad_norm": 0.5436925292015076,
      "learning_rate": 0.00019564706272567035,
      "loss": 0.2511,
      "step": 9930
    },
    {
      "epoch": 0.13971466722889872,
      "grad_norm": 0.4864850342273712,
      "learning_rate": 0.00019563348755661516,
      "loss": 0.2112,
      "step": 9940
    },
    {
      "epoch": 0.13985522524421956,
      "grad_norm": 0.4781823456287384,
      "learning_rate": 0.00019561989172503497,
      "loss": 0.226,
      "step": 9950
    },
    {
      "epoch": 0.13999578325954037,
      "grad_norm": 0.5050092339515686,
      "learning_rate": 0.0001956062752338673,
      "loss": 0.2352,
      "step": 9960
    },
    {
      "epoch": 0.1401363412748612,
      "grad_norm": 0.4395652413368225,
      "learning_rate": 0.0001955926380860541,
      "loss": 0.2332,
      "step": 9970
    },
    {
      "epoch": 0.14027689929018203,
      "grad_norm": 0.5985183119773865,
      "learning_rate": 0.00019557898028454182,
      "loss": 0.239,
      "step": 9980
    },
    {
      "epoch": 0.14041745730550284,
      "grad_norm": 0.4194292724132538,
      "learning_rate": 0.00019556530183228137,
      "loss": 0.2056,
      "step": 9990
    },
    {
      "epoch": 0.14055801532082368,
      "grad_norm": 0.47930511832237244,
      "learning_rate": 0.00019555160273222809,
      "loss": 0.2028,
      "step": 10000
    },
    {
      "epoch": 0.14055801532082368,
      "eval_chrf": 83.31965993456161,
      "eval_loss": 0.46625107526779175,
      "eval_runtime": 204.6953,
      "eval_samples_per_second": 0.489,
      "eval_steps_per_second": 0.02,
      "step": 10000
    },
    {
      "epoch": 0.1406985733361445,
      "grad_norm": 0.7410067915916443,
      "learning_rate": 0.00019553788298734182,
      "loss": 0.2348,
      "step": 10010
    },
    {
      "epoch": 0.1408391313514653,
      "grad_norm": 0.5216307640075684,
      "learning_rate": 0.0001955241426005868,
      "loss": 0.2371,
      "step": 10020
    },
    {
      "epoch": 0.14097968936678615,
      "grad_norm": 0.5811477899551392,
      "learning_rate": 0.0001955103815749318,
      "loss": 0.2139,
      "step": 10030
    },
    {
      "epoch": 0.14112024738210696,
      "grad_norm": 0.49069467186927795,
      "learning_rate": 0.00019549659991335006,
      "loss": 0.2291,
      "step": 10040
    },
    {
      "epoch": 0.1412608053974278,
      "grad_norm": 0.46020469069480896,
      "learning_rate": 0.0001954827976188192,
      "loss": 0.1943,
      "step": 10050
    },
    {
      "epoch": 0.14140136341274862,
      "grad_norm": 0.48496192693710327,
      "learning_rate": 0.00019546897469432129,
      "loss": 0.2219,
      "step": 10060
    },
    {
      "epoch": 0.14154192142806943,
      "grad_norm": 0.5898727774620056,
      "learning_rate": 0.00019545513114284298,
      "loss": 0.2242,
      "step": 10070
    },
    {
      "epoch": 0.14168247944339027,
      "grad_norm": 0.4671427607536316,
      "learning_rate": 0.0001954412669673753,
      "loss": 0.2234,
      "step": 10080
    },
    {
      "epoch": 0.14182303745871108,
      "grad_norm": 0.6892075538635254,
      "learning_rate": 0.00019542738217091365,
      "loss": 0.2087,
      "step": 10090
    },
    {
      "epoch": 0.1419635954740319,
      "grad_norm": 0.5817259550094604,
      "learning_rate": 0.00019541347675645806,
      "loss": 0.2352,
      "step": 10100
    },
    {
      "epoch": 0.14210415348935274,
      "grad_norm": 0.4034591019153595,
      "learning_rate": 0.00019539955072701295,
      "loss": 0.2207,
      "step": 10110
    },
    {
      "epoch": 0.14224471150467355,
      "grad_norm": 0.47377240657806396,
      "learning_rate": 0.00019538560408558707,
      "loss": 0.2231,
      "step": 10120
    },
    {
      "epoch": 0.1423852695199944,
      "grad_norm": 0.4781285524368286,
      "learning_rate": 0.00019537163683519377,
      "loss": 0.22,
      "step": 10130
    },
    {
      "epoch": 0.1425258275353152,
      "grad_norm": 0.6363995671272278,
      "learning_rate": 0.00019535764897885084,
      "loss": 0.2312,
      "step": 10140
    },
    {
      "epoch": 0.14266638555063602,
      "grad_norm": 0.46255913376808167,
      "learning_rate": 0.00019534364051958044,
      "loss": 0.2199,
      "step": 10150
    },
    {
      "epoch": 0.14280694356595686,
      "grad_norm": 0.5095379948616028,
      "learning_rate": 0.00019532961146040927,
      "loss": 0.2301,
      "step": 10160
    },
    {
      "epoch": 0.14294750158127767,
      "grad_norm": 0.5038661360740662,
      "learning_rate": 0.0001953155618043684,
      "loss": 0.2323,
      "step": 10170
    },
    {
      "epoch": 0.14308805959659848,
      "grad_norm": 0.42738714814186096,
      "learning_rate": 0.00019530149155449344,
      "loss": 0.2105,
      "step": 10180
    },
    {
      "epoch": 0.14322861761191932,
      "grad_norm": 0.42202505469322205,
      "learning_rate": 0.0001952874007138244,
      "loss": 0.222,
      "step": 10190
    },
    {
      "epoch": 0.14336917562724014,
      "grad_norm": 0.562140941619873,
      "learning_rate": 0.00019527328928540567,
      "loss": 0.2117,
      "step": 10200
    },
    {
      "epoch": 0.14350973364256098,
      "grad_norm": 0.5138777494430542,
      "learning_rate": 0.00019525915727228624,
      "loss": 0.2098,
      "step": 10210
    },
    {
      "epoch": 0.1436502916578818,
      "grad_norm": 0.5792656540870667,
      "learning_rate": 0.00019524500467751942,
      "loss": 0.2492,
      "step": 10220
    },
    {
      "epoch": 0.1437908496732026,
      "grad_norm": 0.3801478147506714,
      "learning_rate": 0.000195230831504163,
      "loss": 0.2312,
      "step": 10230
    },
    {
      "epoch": 0.14393140768852344,
      "grad_norm": 0.4319283962249756,
      "learning_rate": 0.00019521663775527931,
      "loss": 0.2111,
      "step": 10240
    },
    {
      "epoch": 0.14407196570384426,
      "grad_norm": 0.594716489315033,
      "learning_rate": 0.00019520242343393495,
      "loss": 0.2148,
      "step": 10250
    },
    {
      "epoch": 0.1442125237191651,
      "grad_norm": 0.516743004322052,
      "learning_rate": 0.0001951881885432011,
      "loss": 0.2086,
      "step": 10260
    },
    {
      "epoch": 0.1443530817344859,
      "grad_norm": 0.5203841924667358,
      "learning_rate": 0.00019517393308615338,
      "loss": 0.2184,
      "step": 10270
    },
    {
      "epoch": 0.14449363974980672,
      "grad_norm": 0.5061612129211426,
      "learning_rate": 0.00019515965706587174,
      "loss": 0.2359,
      "step": 10280
    },
    {
      "epoch": 0.14463419776512756,
      "grad_norm": 0.49285292625427246,
      "learning_rate": 0.0001951453604854407,
      "loss": 0.2366,
      "step": 10290
    },
    {
      "epoch": 0.14477475578044838,
      "grad_norm": 0.4080356955528259,
      "learning_rate": 0.00019513104334794917,
      "loss": 0.2116,
      "step": 10300
    },
    {
      "epoch": 0.1449153137957692,
      "grad_norm": 0.460966020822525,
      "learning_rate": 0.00019511670565649047,
      "loss": 0.2254,
      "step": 10310
    },
    {
      "epoch": 0.14505587181109003,
      "grad_norm": 0.5037501454353333,
      "learning_rate": 0.00019510234741416244,
      "loss": 0.2269,
      "step": 10320
    },
    {
      "epoch": 0.14519642982641084,
      "grad_norm": 0.7021297812461853,
      "learning_rate": 0.00019508796862406728,
      "loss": 0.2282,
      "step": 10330
    },
    {
      "epoch": 0.14533698784173169,
      "grad_norm": 0.5365563035011292,
      "learning_rate": 0.00019507356928931172,
      "loss": 0.2094,
      "step": 10340
    },
    {
      "epoch": 0.1454775458570525,
      "grad_norm": 0.45180317759513855,
      "learning_rate": 0.0001950591494130068,
      "loss": 0.2235,
      "step": 10350
    },
    {
      "epoch": 0.1456181038723733,
      "grad_norm": 0.46922942996025085,
      "learning_rate": 0.0001950447089982681,
      "loss": 0.2483,
      "step": 10360
    },
    {
      "epoch": 0.14575866188769415,
      "grad_norm": 0.5505566000938416,
      "learning_rate": 0.0001950302480482156,
      "loss": 0.2225,
      "step": 10370
    },
    {
      "epoch": 0.14589921990301497,
      "grad_norm": 0.44335654377937317,
      "learning_rate": 0.00019501576656597377,
      "loss": 0.2624,
      "step": 10380
    },
    {
      "epoch": 0.1460397779183358,
      "grad_norm": 0.4967087507247925,
      "learning_rate": 0.00019500126455467145,
      "loss": 0.2072,
      "step": 10390
    },
    {
      "epoch": 0.14618033593365662,
      "grad_norm": 0.43531334400177,
      "learning_rate": 0.0001949867420174419,
      "loss": 0.215,
      "step": 10400
    },
    {
      "epoch": 0.14632089394897743,
      "grad_norm": 0.7007616758346558,
      "learning_rate": 0.00019497219895742287,
      "loss": 0.2255,
      "step": 10410
    },
    {
      "epoch": 0.14646145196429827,
      "grad_norm": 0.5119047164916992,
      "learning_rate": 0.00019495763537775656,
      "loss": 0.2443,
      "step": 10420
    },
    {
      "epoch": 0.14660200997961909,
      "grad_norm": 0.8673428297042847,
      "learning_rate": 0.00019494305128158956,
      "loss": 0.2403,
      "step": 10430
    },
    {
      "epoch": 0.1467425679949399,
      "grad_norm": 0.5624560117721558,
      "learning_rate": 0.0001949284466720729,
      "loss": 0.2311,
      "step": 10440
    },
    {
      "epoch": 0.14688312601026074,
      "grad_norm": 0.5158213973045349,
      "learning_rate": 0.00019491382155236198,
      "loss": 0.2141,
      "step": 10450
    },
    {
      "epoch": 0.14702368402558155,
      "grad_norm": 0.4443780183792114,
      "learning_rate": 0.00019489917592561677,
      "loss": 0.2156,
      "step": 10460
    },
    {
      "epoch": 0.1471642420409024,
      "grad_norm": 0.5963322520256042,
      "learning_rate": 0.0001948845097950016,
      "loss": 0.2358,
      "step": 10470
    },
    {
      "epoch": 0.1473048000562232,
      "grad_norm": 0.4588531255722046,
      "learning_rate": 0.00019486982316368519,
      "loss": 0.2029,
      "step": 10480
    },
    {
      "epoch": 0.14744535807154402,
      "grad_norm": 0.43157896399497986,
      "learning_rate": 0.00019485511603484074,
      "loss": 0.2093,
      "step": 10490
    },
    {
      "epoch": 0.14758591608686486,
      "grad_norm": 0.4570079445838928,
      "learning_rate": 0.00019484038841164586,
      "loss": 0.2317,
      "step": 10500
    },
    {
      "epoch": 0.14772647410218567,
      "grad_norm": 0.44846659898757935,
      "learning_rate": 0.00019482564029728262,
      "loss": 0.221,
      "step": 10510
    },
    {
      "epoch": 0.1478670321175065,
      "grad_norm": 0.48597201704978943,
      "learning_rate": 0.0001948108716949375,
      "loss": 0.2547,
      "step": 10520
    },
    {
      "epoch": 0.14800759013282733,
      "grad_norm": 0.5305898785591125,
      "learning_rate": 0.0001947960826078013,
      "loss": 0.2329,
      "step": 10530
    },
    {
      "epoch": 0.14814814814814814,
      "grad_norm": 0.4303765296936035,
      "learning_rate": 0.00019478127303906945,
      "loss": 0.2363,
      "step": 10540
    },
    {
      "epoch": 0.14828870616346898,
      "grad_norm": 0.6144363284111023,
      "learning_rate": 0.00019476644299194166,
      "loss": 0.2386,
      "step": 10550
    },
    {
      "epoch": 0.1484292641787898,
      "grad_norm": 0.5180664658546448,
      "learning_rate": 0.00019475159246962206,
      "loss": 0.2009,
      "step": 10560
    },
    {
      "epoch": 0.1485698221941106,
      "grad_norm": 0.8704946637153625,
      "learning_rate": 0.00019473672147531933,
      "loss": 0.2402,
      "step": 10570
    },
    {
      "epoch": 0.14871038020943145,
      "grad_norm": 0.5063996315002441,
      "learning_rate": 0.00019472183001224645,
      "loss": 0.2416,
      "step": 10580
    },
    {
      "epoch": 0.14885093822475226,
      "grad_norm": 0.4910792410373688,
      "learning_rate": 0.00019470691808362085,
      "loss": 0.2412,
      "step": 10590
    },
    {
      "epoch": 0.1489914962400731,
      "grad_norm": 0.43074488639831543,
      "learning_rate": 0.0001946919856926644,
      "loss": 0.2145,
      "step": 10600
    },
    {
      "epoch": 0.14913205425539391,
      "grad_norm": 0.5350972414016724,
      "learning_rate": 0.0001946770328426034,
      "loss": 0.2121,
      "step": 10610
    },
    {
      "epoch": 0.14927261227071473,
      "grad_norm": 0.5227901935577393,
      "learning_rate": 0.00019466205953666852,
      "loss": 0.2107,
      "step": 10620
    },
    {
      "epoch": 0.14941317028603557,
      "grad_norm": 0.4980103075504303,
      "learning_rate": 0.00019464706577809496,
      "loss": 0.2462,
      "step": 10630
    },
    {
      "epoch": 0.14955372830135638,
      "grad_norm": 0.5568163394927979,
      "learning_rate": 0.00019463205157012217,
      "loss": 0.2313,
      "step": 10640
    },
    {
      "epoch": 0.14969428631667722,
      "grad_norm": 0.4182775318622589,
      "learning_rate": 0.00019461701691599417,
      "loss": 0.2151,
      "step": 10650
    },
    {
      "epoch": 0.14983484433199803,
      "grad_norm": 0.5256010293960571,
      "learning_rate": 0.0001946019618189593,
      "loss": 0.2374,
      "step": 10660
    },
    {
      "epoch": 0.14997540234731885,
      "grad_norm": 0.4888829290866852,
      "learning_rate": 0.00019458688628227043,
      "loss": 0.2414,
      "step": 10670
    },
    {
      "epoch": 0.1501159603626397,
      "grad_norm": 0.35894840955734253,
      "learning_rate": 0.0001945717903091847,
      "loss": 0.2301,
      "step": 10680
    },
    {
      "epoch": 0.1502565183779605,
      "grad_norm": 0.6234340667724609,
      "learning_rate": 0.00019455667390296379,
      "loss": 0.2293,
      "step": 10690
    },
    {
      "epoch": 0.15039707639328131,
      "grad_norm": 0.5959022045135498,
      "learning_rate": 0.0001945415370668737,
      "loss": 0.2221,
      "step": 10700
    },
    {
      "epoch": 0.15053763440860216,
      "grad_norm": 0.5411731004714966,
      "learning_rate": 0.00019452637980418492,
      "loss": 0.2451,
      "step": 10710
    },
    {
      "epoch": 0.15067819242392297,
      "grad_norm": 0.41397252678871155,
      "learning_rate": 0.0001945112021181723,
      "loss": 0.226,
      "step": 10720
    },
    {
      "epoch": 0.1508187504392438,
      "grad_norm": 0.4831259846687317,
      "learning_rate": 0.0001944960040121151,
      "loss": 0.1949,
      "step": 10730
    },
    {
      "epoch": 0.15095930845456462,
      "grad_norm": 0.4825355112552643,
      "learning_rate": 0.00019448078548929708,
      "loss": 0.2511,
      "step": 10740
    },
    {
      "epoch": 0.15109986646988544,
      "grad_norm": 0.42742496728897095,
      "learning_rate": 0.0001944655465530063,
      "loss": 0.1975,
      "step": 10750
    },
    {
      "epoch": 0.15124042448520628,
      "grad_norm": 0.5690857768058777,
      "learning_rate": 0.00019445028720653534,
      "loss": 0.2182,
      "step": 10760
    },
    {
      "epoch": 0.1513809825005271,
      "grad_norm": 0.4305635392665863,
      "learning_rate": 0.00019443500745318102,
      "loss": 0.2287,
      "step": 10770
    },
    {
      "epoch": 0.15152154051584793,
      "grad_norm": 0.3888097107410431,
      "learning_rate": 0.00019441970729624475,
      "loss": 0.2393,
      "step": 10780
    },
    {
      "epoch": 0.15166209853116874,
      "grad_norm": 0.42094510793685913,
      "learning_rate": 0.00019440438673903226,
      "loss": 0.2327,
      "step": 10790
    },
    {
      "epoch": 0.15180265654648956,
      "grad_norm": 0.438627690076828,
      "learning_rate": 0.00019438904578485372,
      "loss": 0.2304,
      "step": 10800
    },
    {
      "epoch": 0.1519432145618104,
      "grad_norm": 0.507936954498291,
      "learning_rate": 0.00019437368443702362,
      "loss": 0.2064,
      "step": 10810
    },
    {
      "epoch": 0.1520837725771312,
      "grad_norm": 0.4277139902114868,
      "learning_rate": 0.00019435830269886098,
      "loss": 0.2124,
      "step": 10820
    },
    {
      "epoch": 0.15222433059245202,
      "grad_norm": 0.43021994829177856,
      "learning_rate": 0.0001943429005736892,
      "loss": 0.2263,
      "step": 10830
    },
    {
      "epoch": 0.15236488860777286,
      "grad_norm": 0.43604740500450134,
      "learning_rate": 0.00019432747806483603,
      "loss": 0.2275,
      "step": 10840
    },
    {
      "epoch": 0.15250544662309368,
      "grad_norm": 0.4549795985221863,
      "learning_rate": 0.0001943120351756336,
      "loss": 0.2049,
      "step": 10850
    },
    {
      "epoch": 0.15264600463841452,
      "grad_norm": 0.45931580662727356,
      "learning_rate": 0.00019429657190941856,
      "loss": 0.2235,
      "step": 10860
    },
    {
      "epoch": 0.15278656265373533,
      "grad_norm": 0.5587042570114136,
      "learning_rate": 0.00019428108826953188,
      "loss": 0.2682,
      "step": 10870
    },
    {
      "epoch": 0.15292712066905614,
      "grad_norm": 0.4498588442802429,
      "learning_rate": 0.00019426558425931893,
      "loss": 0.2162,
      "step": 10880
    },
    {
      "epoch": 0.15306767868437698,
      "grad_norm": 0.36715206503868103,
      "learning_rate": 0.00019425005988212948,
      "loss": 0.2055,
      "step": 10890
    },
    {
      "epoch": 0.1532082366996978,
      "grad_norm": 0.40799346566200256,
      "learning_rate": 0.00019423451514131776,
      "loss": 0.1994,
      "step": 10900
    },
    {
      "epoch": 0.1533487947150186,
      "grad_norm": 0.6820157170295715,
      "learning_rate": 0.00019421895004024235,
      "loss": 0.2721,
      "step": 10910
    },
    {
      "epoch": 0.15348935273033945,
      "grad_norm": 0.5497896075248718,
      "learning_rate": 0.00019420336458226626,
      "loss": 0.2473,
      "step": 10920
    },
    {
      "epoch": 0.15362991074566026,
      "grad_norm": 0.4838441014289856,
      "learning_rate": 0.0001941877587707568,
      "loss": 0.2316,
      "step": 10930
    },
    {
      "epoch": 0.1537704687609811,
      "grad_norm": 0.5930109024047852,
      "learning_rate": 0.00019417213260908585,
      "loss": 0.2177,
      "step": 10940
    },
    {
      "epoch": 0.15391102677630192,
      "grad_norm": 0.5457140207290649,
      "learning_rate": 0.00019415648610062952,
      "loss": 0.214,
      "step": 10950
    },
    {
      "epoch": 0.15405158479162273,
      "grad_norm": 0.49875709414482117,
      "learning_rate": 0.00019414081924876841,
      "loss": 0.2238,
      "step": 10960
    },
    {
      "epoch": 0.15419214280694357,
      "grad_norm": 0.5106085538864136,
      "learning_rate": 0.0001941251320568875,
      "loss": 0.2032,
      "step": 10970
    },
    {
      "epoch": 0.15433270082226438,
      "grad_norm": 0.5855301022529602,
      "learning_rate": 0.00019410942452837614,
      "loss": 0.2275,
      "step": 10980
    },
    {
      "epoch": 0.15447325883758523,
      "grad_norm": 0.45460623502731323,
      "learning_rate": 0.00019409369666662814,
      "loss": 0.1917,
      "step": 10990
    },
    {
      "epoch": 0.15461381685290604,
      "grad_norm": 0.46648094058036804,
      "learning_rate": 0.00019407794847504158,
      "loss": 0.2199,
      "step": 11000
    },
    {
      "epoch": 0.15461381685290604,
      "eval_chrf": 83.42666111421941,
      "eval_loss": 0.44805753231048584,
      "eval_runtime": 141.7912,
      "eval_samples_per_second": 0.705,
      "eval_steps_per_second": 0.028,
      "step": 11000
    },
    {
      "epoch": 0.15475437486822685,
      "grad_norm": 0.5129318237304688,
      "learning_rate": 0.00019406217995701904,
      "loss": 0.2431,
      "step": 11010
    },
    {
      "epoch": 0.1548949328835477,
      "grad_norm": 0.5577740669250488,
      "learning_rate": 0.00019404639111596749,
      "loss": 0.2244,
      "step": 11020
    },
    {
      "epoch": 0.1550354908988685,
      "grad_norm": 0.5699359774589539,
      "learning_rate": 0.00019403058195529823,
      "loss": 0.232,
      "step": 11030
    },
    {
      "epoch": 0.15517604891418932,
      "grad_norm": 0.37373116612434387,
      "learning_rate": 0.00019401475247842696,
      "loss": 0.229,
      "step": 11040
    },
    {
      "epoch": 0.15531660692951016,
      "grad_norm": 0.5445089340209961,
      "learning_rate": 0.00019399890268877382,
      "loss": 0.2208,
      "step": 11050
    },
    {
      "epoch": 0.15545716494483097,
      "grad_norm": 0.452178955078125,
      "learning_rate": 0.0001939830325897633,
      "loss": 0.2649,
      "step": 11060
    },
    {
      "epoch": 0.1555977229601518,
      "grad_norm": 0.5334861278533936,
      "learning_rate": 0.0001939671421848243,
      "loss": 0.2138,
      "step": 11070
    },
    {
      "epoch": 0.15573828097547263,
      "grad_norm": 0.45646482706069946,
      "learning_rate": 0.00019395123147739003,
      "loss": 0.2432,
      "step": 11080
    },
    {
      "epoch": 0.15587883899079344,
      "grad_norm": 0.541918158531189,
      "learning_rate": 0.00019393530047089822,
      "loss": 0.2224,
      "step": 11090
    },
    {
      "epoch": 0.15601939700611428,
      "grad_norm": 0.5018472671508789,
      "learning_rate": 0.0001939193491687909,
      "loss": 0.2117,
      "step": 11100
    },
    {
      "epoch": 0.1561599550214351,
      "grad_norm": 0.4296761453151703,
      "learning_rate": 0.00019390337757451448,
      "loss": 0.2125,
      "step": 11110
    },
    {
      "epoch": 0.15630051303675593,
      "grad_norm": 0.4516717791557312,
      "learning_rate": 0.0001938873856915198,
      "loss": 0.2294,
      "step": 11120
    },
    {
      "epoch": 0.15644107105207675,
      "grad_norm": 0.532499372959137,
      "learning_rate": 0.00019387137352326204,
      "loss": 0.2269,
      "step": 11130
    },
    {
      "epoch": 0.15658162906739756,
      "grad_norm": 0.5420687794685364,
      "learning_rate": 0.0001938553410732008,
      "loss": 0.2162,
      "step": 11140
    },
    {
      "epoch": 0.1567221870827184,
      "grad_norm": 0.5829330682754517,
      "learning_rate": 0.0001938392883448,
      "loss": 0.2294,
      "step": 11150
    },
    {
      "epoch": 0.1568627450980392,
      "grad_norm": 0.47940340638160706,
      "learning_rate": 0.00019382321534152803,
      "loss": 0.238,
      "step": 11160
    },
    {
      "epoch": 0.15700330311336003,
      "grad_norm": 0.4618227183818817,
      "learning_rate": 0.00019380712206685762,
      "loss": 0.2472,
      "step": 11170
    },
    {
      "epoch": 0.15714386112868087,
      "grad_norm": 0.451911985874176,
      "learning_rate": 0.00019379100852426584,
      "loss": 0.224,
      "step": 11180
    },
    {
      "epoch": 0.15728441914400168,
      "grad_norm": 0.5306965708732605,
      "learning_rate": 0.00019377487471723418,
      "loss": 0.2518,
      "step": 11190
    },
    {
      "epoch": 0.15742497715932252,
      "grad_norm": 0.6336318850517273,
      "learning_rate": 0.00019375872064924854,
      "loss": 0.2168,
      "step": 11200
    },
    {
      "epoch": 0.15756553517464333,
      "grad_norm": 0.6012146472930908,
      "learning_rate": 0.0001937425463237991,
      "loss": 0.239,
      "step": 11210
    },
    {
      "epoch": 0.15770609318996415,
      "grad_norm": 0.72569340467453,
      "learning_rate": 0.00019372635174438055,
      "loss": 0.219,
      "step": 11220
    },
    {
      "epoch": 0.157846651205285,
      "grad_norm": 0.41902977228164673,
      "learning_rate": 0.0001937101369144918,
      "loss": 0.2157,
      "step": 11230
    },
    {
      "epoch": 0.1579872092206058,
      "grad_norm": 0.6844558119773865,
      "learning_rate": 0.0001936939018376363,
      "loss": 0.2345,
      "step": 11240
    },
    {
      "epoch": 0.15812776723592664,
      "grad_norm": 0.5705519318580627,
      "learning_rate": 0.00019367764651732173,
      "loss": 0.2267,
      "step": 11250
    },
    {
      "epoch": 0.15826832525124745,
      "grad_norm": 0.4265342950820923,
      "learning_rate": 0.00019366137095706024,
      "loss": 0.2178,
      "step": 11260
    },
    {
      "epoch": 0.15840888326656827,
      "grad_norm": 0.49786338210105896,
      "learning_rate": 0.0001936450751603683,
      "loss": 0.2194,
      "step": 11270
    },
    {
      "epoch": 0.1585494412818891,
      "grad_norm": 0.5172784328460693,
      "learning_rate": 0.0001936287591307668,
      "loss": 0.2397,
      "step": 11280
    },
    {
      "epoch": 0.15868999929720992,
      "grad_norm": 0.5547787547111511,
      "learning_rate": 0.00019361242287178093,
      "loss": 0.2122,
      "step": 11290
    },
    {
      "epoch": 0.15883055731253073,
      "grad_norm": 0.46685513854026794,
      "learning_rate": 0.00019359606638694033,
      "loss": 0.235,
      "step": 11300
    },
    {
      "epoch": 0.15897111532785158,
      "grad_norm": 0.5981021523475647,
      "learning_rate": 0.00019357968967977898,
      "loss": 0.243,
      "step": 11310
    },
    {
      "epoch": 0.1591116733431724,
      "grad_norm": 0.4840559661388397,
      "learning_rate": 0.0001935632927538352,
      "loss": 0.2093,
      "step": 11320
    },
    {
      "epoch": 0.15925223135849323,
      "grad_norm": 0.4729189872741699,
      "learning_rate": 0.00019354687561265167,
      "loss": 0.2144,
      "step": 11330
    },
    {
      "epoch": 0.15939278937381404,
      "grad_norm": 0.5277968645095825,
      "learning_rate": 0.00019353043825977558,
      "loss": 0.2071,
      "step": 11340
    },
    {
      "epoch": 0.15953334738913486,
      "grad_norm": 0.48004767298698425,
      "learning_rate": 0.00019351398069875826,
      "loss": 0.2289,
      "step": 11350
    },
    {
      "epoch": 0.1596739054044557,
      "grad_norm": 0.5471176505088806,
      "learning_rate": 0.00019349750293315557,
      "loss": 0.24,
      "step": 11360
    },
    {
      "epoch": 0.1598144634197765,
      "grad_norm": 0.49257561564445496,
      "learning_rate": 0.0001934810049665277,
      "loss": 0.2292,
      "step": 11370
    },
    {
      "epoch": 0.15995502143509735,
      "grad_norm": 0.394435316324234,
      "learning_rate": 0.00019346448680243917,
      "loss": 0.2066,
      "step": 11380
    },
    {
      "epoch": 0.16009557945041816,
      "grad_norm": 0.4983547031879425,
      "learning_rate": 0.0001934479484444589,
      "loss": 0.2114,
      "step": 11390
    },
    {
      "epoch": 0.16023613746573898,
      "grad_norm": 0.5153059363365173,
      "learning_rate": 0.00019343138989616013,
      "loss": 0.2369,
      "step": 11400
    },
    {
      "epoch": 0.16037669548105982,
      "grad_norm": 0.6187337636947632,
      "learning_rate": 0.00019341481116112052,
      "loss": 0.2314,
      "step": 11410
    },
    {
      "epoch": 0.16051725349638063,
      "grad_norm": 0.3402414619922638,
      "learning_rate": 0.00019339821224292208,
      "loss": 0.2134,
      "step": 11420
    },
    {
      "epoch": 0.16065781151170144,
      "grad_norm": 0.4472907483577728,
      "learning_rate": 0.00019338159314515117,
      "loss": 0.1956,
      "step": 11430
    },
    {
      "epoch": 0.16079836952702228,
      "grad_norm": 0.5496037006378174,
      "learning_rate": 0.00019336495387139844,
      "loss": 0.2259,
      "step": 11440
    },
    {
      "epoch": 0.1609389275423431,
      "grad_norm": 0.531448245048523,
      "learning_rate": 0.000193348294425259,
      "loss": 0.2356,
      "step": 11450
    },
    {
      "epoch": 0.16107948555766394,
      "grad_norm": 0.4533897936344147,
      "learning_rate": 0.00019333161481033232,
      "loss": 0.2438,
      "step": 11460
    },
    {
      "epoch": 0.16122004357298475,
      "grad_norm": 0.4858724772930145,
      "learning_rate": 0.00019331491503022212,
      "loss": 0.2291,
      "step": 11470
    },
    {
      "epoch": 0.16136060158830556,
      "grad_norm": 0.48073771595954895,
      "learning_rate": 0.00019329819508853658,
      "loss": 0.2346,
      "step": 11480
    },
    {
      "epoch": 0.1615011596036264,
      "grad_norm": 0.4850899875164032,
      "learning_rate": 0.0001932814549888882,
      "loss": 0.2124,
      "step": 11490
    },
    {
      "epoch": 0.16164171761894722,
      "grad_norm": 0.49765729904174805,
      "learning_rate": 0.00019326469473489387,
      "loss": 0.2031,
      "step": 11500
    },
    {
      "epoch": 0.16178227563426806,
      "grad_norm": 0.434810996055603,
      "learning_rate": 0.00019324791433017477,
      "loss": 0.2285,
      "step": 11510
    },
    {
      "epoch": 0.16192283364958887,
      "grad_norm": 0.4233197867870331,
      "learning_rate": 0.00019323111377835645,
      "loss": 0.1947,
      "step": 11520
    },
    {
      "epoch": 0.16206339166490968,
      "grad_norm": 0.4610973000526428,
      "learning_rate": 0.00019321429308306884,
      "loss": 0.2174,
      "step": 11530
    },
    {
      "epoch": 0.16220394968023052,
      "grad_norm": 0.4681873619556427,
      "learning_rate": 0.00019319745224794622,
      "loss": 0.2333,
      "step": 11540
    },
    {
      "epoch": 0.16234450769555134,
      "grad_norm": 0.3164941668510437,
      "learning_rate": 0.00019318059127662723,
      "loss": 0.2105,
      "step": 11550
    },
    {
      "epoch": 0.16248506571087215,
      "grad_norm": 0.4490436315536499,
      "learning_rate": 0.00019316371017275483,
      "loss": 0.2346,
      "step": 11560
    },
    {
      "epoch": 0.162625623726193,
      "grad_norm": 0.39336520433425903,
      "learning_rate": 0.0001931468089399763,
      "loss": 0.225,
      "step": 11570
    },
    {
      "epoch": 0.1627661817415138,
      "grad_norm": 0.48553505539894104,
      "learning_rate": 0.00019312988758194338,
      "loss": 0.2094,
      "step": 11580
    },
    {
      "epoch": 0.16290673975683465,
      "grad_norm": 0.5376838445663452,
      "learning_rate": 0.00019311294610231208,
      "loss": 0.2215,
      "step": 11590
    },
    {
      "epoch": 0.16304729777215546,
      "grad_norm": 0.3806610703468323,
      "learning_rate": 0.00019309598450474274,
      "loss": 0.24,
      "step": 11600
    },
    {
      "epoch": 0.16318785578747627,
      "grad_norm": 0.32200586795806885,
      "learning_rate": 0.00019307900279290005,
      "loss": 0.2218,
      "step": 11610
    },
    {
      "epoch": 0.1633284138027971,
      "grad_norm": 0.34597766399383545,
      "learning_rate": 0.00019306200097045313,
      "loss": 0.2129,
      "step": 11620
    },
    {
      "epoch": 0.16346897181811793,
      "grad_norm": 0.41980305314064026,
      "learning_rate": 0.00019304497904107533,
      "loss": 0.2381,
      "step": 11630
    },
    {
      "epoch": 0.16360952983343874,
      "grad_norm": 0.5225925445556641,
      "learning_rate": 0.00019302793700844446,
      "loss": 0.2235,
      "step": 11640
    },
    {
      "epoch": 0.16375008784875958,
      "grad_norm": 0.6067329049110413,
      "learning_rate": 0.00019301087487624257,
      "loss": 0.2363,
      "step": 11650
    },
    {
      "epoch": 0.1638906458640804,
      "grad_norm": 0.5046607255935669,
      "learning_rate": 0.00019299379264815612,
      "loss": 0.2086,
      "step": 11660
    },
    {
      "epoch": 0.16403120387940123,
      "grad_norm": 0.3685866892337799,
      "learning_rate": 0.00019297669032787587,
      "loss": 0.2232,
      "step": 11670
    },
    {
      "epoch": 0.16417176189472205,
      "grad_norm": 0.46026644110679626,
      "learning_rate": 0.00019295956791909693,
      "loss": 0.2348,
      "step": 11680
    },
    {
      "epoch": 0.16431231991004286,
      "grad_norm": 0.40158572793006897,
      "learning_rate": 0.0001929424254255188,
      "loss": 0.2264,
      "step": 11690
    },
    {
      "epoch": 0.1644528779253637,
      "grad_norm": 0.5387255549430847,
      "learning_rate": 0.00019292526285084524,
      "loss": 0.2055,
      "step": 11700
    },
    {
      "epoch": 0.1645934359406845,
      "grad_norm": 0.5109559893608093,
      "learning_rate": 0.0001929080801987844,
      "loss": 0.2088,
      "step": 11710
    },
    {
      "epoch": 0.16473399395600535,
      "grad_norm": 0.4983414113521576,
      "learning_rate": 0.00019289087747304873,
      "loss": 0.2203,
      "step": 11720
    },
    {
      "epoch": 0.16487455197132617,
      "grad_norm": 0.5244500041007996,
      "learning_rate": 0.00019287365467735511,
      "loss": 0.2393,
      "step": 11730
    },
    {
      "epoch": 0.16501510998664698,
      "grad_norm": 0.7381274104118347,
      "learning_rate": 0.00019285641181542465,
      "loss": 0.2293,
      "step": 11740
    },
    {
      "epoch": 0.16515566800196782,
      "grad_norm": 0.401611328125,
      "learning_rate": 0.0001928391488909828,
      "loss": 0.2418,
      "step": 11750
    },
    {
      "epoch": 0.16529622601728863,
      "grad_norm": 0.35754626989364624,
      "learning_rate": 0.00019282186590775942,
      "loss": 0.2519,
      "step": 11760
    },
    {
      "epoch": 0.16543678403260945,
      "grad_norm": 0.48142239451408386,
      "learning_rate": 0.00019280456286948868,
      "loss": 0.2001,
      "step": 11770
    },
    {
      "epoch": 0.1655773420479303,
      "grad_norm": 0.6075026392936707,
      "learning_rate": 0.00019278723977990902,
      "loss": 0.1858,
      "step": 11780
    },
    {
      "epoch": 0.1657179000632511,
      "grad_norm": 0.4034958779811859,
      "learning_rate": 0.0001927698966427633,
      "loss": 0.2119,
      "step": 11790
    },
    {
      "epoch": 0.16585845807857194,
      "grad_norm": 0.42134714126586914,
      "learning_rate": 0.0001927525334617986,
      "loss": 0.2428,
      "step": 11800
    },
    {
      "epoch": 0.16599901609389275,
      "grad_norm": 0.5965345501899719,
      "learning_rate": 0.0001927351502407665,
      "loss": 0.2184,
      "step": 11810
    },
    {
      "epoch": 0.16613957410921357,
      "grad_norm": 0.5037842392921448,
      "learning_rate": 0.00019271774698342273,
      "loss": 0.2136,
      "step": 11820
    },
    {
      "epoch": 0.1662801321245344,
      "grad_norm": 0.684178352355957,
      "learning_rate": 0.00019270032369352746,
      "loss": 0.2337,
      "step": 11830
    },
    {
      "epoch": 0.16642069013985522,
      "grad_norm": 0.45954614877700806,
      "learning_rate": 0.0001926828803748452,
      "loss": 0.2235,
      "step": 11840
    },
    {
      "epoch": 0.16656124815517606,
      "grad_norm": 0.4757162928581238,
      "learning_rate": 0.00019266541703114466,
      "loss": 0.214,
      "step": 11850
    },
    {
      "epoch": 0.16670180617049687,
      "grad_norm": 0.4088406264781952,
      "learning_rate": 0.000192647933666199,
      "loss": 0.2377,
      "step": 11860
    },
    {
      "epoch": 0.1668423641858177,
      "grad_norm": 0.5777791738510132,
      "learning_rate": 0.00019263043028378566,
      "loss": 0.2662,
      "step": 11870
    },
    {
      "epoch": 0.16698292220113853,
      "grad_norm": 0.4320667088031769,
      "learning_rate": 0.00019261290688768644,
      "loss": 0.2339,
      "step": 11880
    },
    {
      "epoch": 0.16712348021645934,
      "grad_norm": 0.4025791585445404,
      "learning_rate": 0.00019259536348168742,
      "loss": 0.2062,
      "step": 11890
    },
    {
      "epoch": 0.16726403823178015,
      "grad_norm": 0.5231763124465942,
      "learning_rate": 0.00019257780006957904,
      "loss": 0.2287,
      "step": 11900
    },
    {
      "epoch": 0.167404596247101,
      "grad_norm": 0.4425447881221771,
      "learning_rate": 0.00019256021665515601,
      "loss": 0.2174,
      "step": 11910
    },
    {
      "epoch": 0.1675451542624218,
      "grad_norm": 0.5530385971069336,
      "learning_rate": 0.0001925426132422174,
      "loss": 0.2384,
      "step": 11920
    },
    {
      "epoch": 0.16768571227774265,
      "grad_norm": 0.4237968921661377,
      "learning_rate": 0.0001925249898345666,
      "loss": 0.2378,
      "step": 11930
    },
    {
      "epoch": 0.16782627029306346,
      "grad_norm": 0.4966355264186859,
      "learning_rate": 0.0001925073464360113,
      "loss": 0.2284,
      "step": 11940
    },
    {
      "epoch": 0.16796682830838428,
      "grad_norm": 0.47906044125556946,
      "learning_rate": 0.00019248968305036354,
      "loss": 0.24,
      "step": 11950
    },
    {
      "epoch": 0.16810738632370512,
      "grad_norm": 0.3971858322620392,
      "learning_rate": 0.00019247199968143967,
      "loss": 0.2029,
      "step": 11960
    },
    {
      "epoch": 0.16824794433902593,
      "grad_norm": 0.40358278155326843,
      "learning_rate": 0.00019245429633306036,
      "loss": 0.2155,
      "step": 11970
    },
    {
      "epoch": 0.16838850235434677,
      "grad_norm": 0.4425409734249115,
      "learning_rate": 0.00019243657300905054,
      "loss": 0.2003,
      "step": 11980
    },
    {
      "epoch": 0.16852906036966758,
      "grad_norm": 0.5036493539810181,
      "learning_rate": 0.00019241882971323955,
      "loss": 0.2192,
      "step": 11990
    },
    {
      "epoch": 0.1686696183849884,
      "grad_norm": 0.5340859293937683,
      "learning_rate": 0.00019240106644946095,
      "loss": 0.2337,
      "step": 12000
    },
    {
      "epoch": 0.1686696183849884,
      "eval_chrf": 82.09448804661407,
      "eval_loss": 0.4382096529006958,
      "eval_runtime": 247.2295,
      "eval_samples_per_second": 0.404,
      "eval_steps_per_second": 0.016,
      "step": 12000
    },
    {
      "epoch": 0.16881017640030924,
      "grad_norm": 0.5135290026664734,
      "learning_rate": 0.00019238328322155273,
      "loss": 0.2238,
      "step": 12010
    },
    {
      "epoch": 0.16895073441563005,
      "grad_norm": 0.5166633725166321,
      "learning_rate": 0.0001923654800333571,
      "loss": 0.2511,
      "step": 12020
    },
    {
      "epoch": 0.16909129243095086,
      "grad_norm": 0.44228142499923706,
      "learning_rate": 0.00019234765688872055,
      "loss": 0.2041,
      "step": 12030
    },
    {
      "epoch": 0.1692318504462717,
      "grad_norm": 0.3702377378940582,
      "learning_rate": 0.000192329813791494,
      "loss": 0.2112,
      "step": 12040
    },
    {
      "epoch": 0.16937240846159252,
      "grad_norm": 0.4565224349498749,
      "learning_rate": 0.00019231195074553262,
      "loss": 0.2125,
      "step": 12050
    },
    {
      "epoch": 0.16951296647691336,
      "grad_norm": 0.4436790645122528,
      "learning_rate": 0.00019229406775469587,
      "loss": 0.1962,
      "step": 12060
    },
    {
      "epoch": 0.16965352449223417,
      "grad_norm": 0.5260986089706421,
      "learning_rate": 0.00019227616482284755,
      "loss": 0.2261,
      "step": 12070
    },
    {
      "epoch": 0.16979408250755498,
      "grad_norm": 0.3428475558757782,
      "learning_rate": 0.00019225824195385574,
      "loss": 0.231,
      "step": 12080
    },
    {
      "epoch": 0.16993464052287582,
      "grad_norm": 0.47546109557151794,
      "learning_rate": 0.00019224029915159286,
      "loss": 0.2129,
      "step": 12090
    },
    {
      "epoch": 0.17007519853819664,
      "grad_norm": 0.47432127594947815,
      "learning_rate": 0.00019222233641993567,
      "loss": 0.2029,
      "step": 12100
    },
    {
      "epoch": 0.17021575655351748,
      "grad_norm": 0.4969255030155182,
      "learning_rate": 0.00019220435376276509,
      "loss": 0.2369,
      "step": 12110
    },
    {
      "epoch": 0.1703563145688383,
      "grad_norm": 0.7607478499412537,
      "learning_rate": 0.00019218635118396652,
      "loss": 0.2149,
      "step": 12120
    },
    {
      "epoch": 0.1704968725841591,
      "grad_norm": 0.3294755220413208,
      "learning_rate": 0.0001921683286874296,
      "loss": 0.2078,
      "step": 12130
    },
    {
      "epoch": 0.17063743059947994,
      "grad_norm": 0.5928043723106384,
      "learning_rate": 0.0001921502862770482,
      "loss": 0.2304,
      "step": 12140
    },
    {
      "epoch": 0.17077798861480076,
      "grad_norm": 0.3646385371685028,
      "learning_rate": 0.00019213222395672057,
      "loss": 0.2069,
      "step": 12150
    },
    {
      "epoch": 0.17091854663012157,
      "grad_norm": 0.4349091053009033,
      "learning_rate": 0.00019211414173034928,
      "loss": 0.189,
      "step": 12160
    },
    {
      "epoch": 0.1710591046454424,
      "grad_norm": 0.41584905982017517,
      "learning_rate": 0.00019209603960184115,
      "loss": 0.2247,
      "step": 12170
    },
    {
      "epoch": 0.17119966266076322,
      "grad_norm": 0.42634835839271545,
      "learning_rate": 0.00019207791757510728,
      "loss": 0.2287,
      "step": 12180
    },
    {
      "epoch": 0.17134022067608407,
      "grad_norm": 0.6456392407417297,
      "learning_rate": 0.00019205977565406318,
      "loss": 0.2153,
      "step": 12190
    },
    {
      "epoch": 0.17148077869140488,
      "grad_norm": 0.44439736008644104,
      "learning_rate": 0.00019204161384262856,
      "loss": 0.203,
      "step": 12200
    },
    {
      "epoch": 0.1716213367067257,
      "grad_norm": 0.6120620965957642,
      "learning_rate": 0.0001920234321447274,
      "loss": 0.2369,
      "step": 12210
    },
    {
      "epoch": 0.17176189472204653,
      "grad_norm": 0.6957530379295349,
      "learning_rate": 0.0001920052305642881,
      "loss": 0.2361,
      "step": 12220
    },
    {
      "epoch": 0.17190245273736734,
      "grad_norm": 0.3580227494239807,
      "learning_rate": 0.00019198700910524321,
      "loss": 0.2263,
      "step": 12230
    },
    {
      "epoch": 0.17204301075268819,
      "grad_norm": 0.4982926547527313,
      "learning_rate": 0.0001919687677715297,
      "loss": 0.2188,
      "step": 12240
    },
    {
      "epoch": 0.172183568768009,
      "grad_norm": 0.4078795611858368,
      "learning_rate": 0.0001919505065670888,
      "loss": 0.2392,
      "step": 12250
    },
    {
      "epoch": 0.1723241267833298,
      "grad_norm": 0.4633350074291229,
      "learning_rate": 0.00019193222549586598,
      "loss": 0.221,
      "step": 12260
    },
    {
      "epoch": 0.17246468479865065,
      "grad_norm": 0.4645605683326721,
      "learning_rate": 0.00019191392456181105,
      "loss": 0.2296,
      "step": 12270
    },
    {
      "epoch": 0.17260524281397147,
      "grad_norm": 0.4471352994441986,
      "learning_rate": 0.0001918956037688781,
      "loss": 0.2,
      "step": 12280
    },
    {
      "epoch": 0.17274580082929228,
      "grad_norm": 0.4225039780139923,
      "learning_rate": 0.00019187726312102553,
      "loss": 0.1919,
      "step": 12290
    },
    {
      "epoch": 0.17288635884461312,
      "grad_norm": 0.44874119758605957,
      "learning_rate": 0.00019185890262221595,
      "loss": 0.2224,
      "step": 12300
    },
    {
      "epoch": 0.17302691685993393,
      "grad_norm": 0.4801863431930542,
      "learning_rate": 0.0001918405222764164,
      "loss": 0.2148,
      "step": 12310
    },
    {
      "epoch": 0.17316747487525477,
      "grad_norm": 0.43257445096969604,
      "learning_rate": 0.0001918221220875981,
      "loss": 0.2254,
      "step": 12320
    },
    {
      "epoch": 0.1733080328905756,
      "grad_norm": 0.4954662322998047,
      "learning_rate": 0.00019180370205973657,
      "loss": 0.2532,
      "step": 12330
    },
    {
      "epoch": 0.1734485909058964,
      "grad_norm": 0.47669610381126404,
      "learning_rate": 0.00019178526219681163,
      "loss": 0.1945,
      "step": 12340
    },
    {
      "epoch": 0.17358914892121724,
      "grad_norm": 0.3972938656806946,
      "learning_rate": 0.00019176680250280743,
      "loss": 0.2247,
      "step": 12350
    },
    {
      "epoch": 0.17372970693653805,
      "grad_norm": 0.5760671496391296,
      "learning_rate": 0.0001917483229817123,
      "loss": 0.2242,
      "step": 12360
    },
    {
      "epoch": 0.17387026495185887,
      "grad_norm": 0.4229896664619446,
      "learning_rate": 0.00019172982363751894,
      "loss": 0.2132,
      "step": 12370
    },
    {
      "epoch": 0.1740108229671797,
      "grad_norm": 0.5061476826667786,
      "learning_rate": 0.00019171130447422435,
      "loss": 0.2555,
      "step": 12380
    },
    {
      "epoch": 0.17415138098250052,
      "grad_norm": 0.5105674862861633,
      "learning_rate": 0.00019169276549582974,
      "loss": 0.2052,
      "step": 12390
    },
    {
      "epoch": 0.17429193899782136,
      "grad_norm": 0.4769287705421448,
      "learning_rate": 0.00019167420670634062,
      "loss": 0.1912,
      "step": 12400
    },
    {
      "epoch": 0.17443249701314217,
      "grad_norm": 0.430573433637619,
      "learning_rate": 0.00019165562810976683,
      "loss": 0.2602,
      "step": 12410
    },
    {
      "epoch": 0.174573055028463,
      "grad_norm": 0.4756665527820587,
      "learning_rate": 0.0001916370297101224,
      "loss": 0.2292,
      "step": 12420
    },
    {
      "epoch": 0.17471361304378383,
      "grad_norm": 0.4962463080883026,
      "learning_rate": 0.00019161841151142572,
      "loss": 0.1848,
      "step": 12430
    },
    {
      "epoch": 0.17485417105910464,
      "grad_norm": 0.4224216043949127,
      "learning_rate": 0.00019159977351769944,
      "loss": 0.2334,
      "step": 12440
    },
    {
      "epoch": 0.17499472907442548,
      "grad_norm": 0.45669710636138916,
      "learning_rate": 0.00019158111573297046,
      "loss": 0.2431,
      "step": 12450
    },
    {
      "epoch": 0.1751352870897463,
      "grad_norm": 0.4416327178478241,
      "learning_rate": 0.00019156243816126998,
      "loss": 0.2015,
      "step": 12460
    },
    {
      "epoch": 0.1752758451050671,
      "grad_norm": 0.558631956577301,
      "learning_rate": 0.00019154374080663345,
      "loss": 0.2086,
      "step": 12470
    },
    {
      "epoch": 0.17541640312038795,
      "grad_norm": 0.5783782601356506,
      "learning_rate": 0.00019152502367310062,
      "loss": 0.2295,
      "step": 12480
    },
    {
      "epoch": 0.17555696113570876,
      "grad_norm": 0.44455182552337646,
      "learning_rate": 0.00019150628676471555,
      "loss": 0.2517,
      "step": 12490
    },
    {
      "epoch": 0.17569751915102957,
      "grad_norm": 0.49356889724731445,
      "learning_rate": 0.00019148753008552645,
      "loss": 0.2256,
      "step": 12500
    },
    {
      "epoch": 0.17583807716635041,
      "grad_norm": 0.4006481468677521,
      "learning_rate": 0.0001914687536395859,
      "loss": 0.2079,
      "step": 12510
    },
    {
      "epoch": 0.17597863518167123,
      "grad_norm": 0.33360642194747925,
      "learning_rate": 0.00019144995743095077,
      "loss": 0.2366,
      "step": 12520
    },
    {
      "epoch": 0.17611919319699207,
      "grad_norm": 0.3714017868041992,
      "learning_rate": 0.00019143114146368215,
      "loss": 0.2267,
      "step": 12530
    },
    {
      "epoch": 0.17625975121231288,
      "grad_norm": 0.38289201259613037,
      "learning_rate": 0.00019141230574184538,
      "loss": 0.2202,
      "step": 12540
    },
    {
      "epoch": 0.1764003092276337,
      "grad_norm": 0.5035020112991333,
      "learning_rate": 0.00019139345026951012,
      "loss": 0.199,
      "step": 12550
    },
    {
      "epoch": 0.17654086724295454,
      "grad_norm": 0.38164594769477844,
      "learning_rate": 0.00019137457505075025,
      "loss": 0.2322,
      "step": 12560
    },
    {
      "epoch": 0.17668142525827535,
      "grad_norm": 0.5696097612380981,
      "learning_rate": 0.00019135568008964398,
      "loss": 0.1965,
      "step": 12570
    },
    {
      "epoch": 0.1768219832735962,
      "grad_norm": 0.5146777033805847,
      "learning_rate": 0.0001913367653902737,
      "loss": 0.221,
      "step": 12580
    },
    {
      "epoch": 0.176962541288917,
      "grad_norm": 0.4113754630088806,
      "learning_rate": 0.00019131783095672616,
      "loss": 0.2316,
      "step": 12590
    },
    {
      "epoch": 0.17710309930423782,
      "grad_norm": 0.38866838812828064,
      "learning_rate": 0.00019129887679309224,
      "loss": 0.2412,
      "step": 12600
    },
    {
      "epoch": 0.17724365731955866,
      "grad_norm": 0.5679992437362671,
      "learning_rate": 0.00019127990290346728,
      "loss": 0.2309,
      "step": 12610
    },
    {
      "epoch": 0.17738421533487947,
      "grad_norm": 0.3673727810382843,
      "learning_rate": 0.0001912609092919507,
      "loss": 0.2344,
      "step": 12620
    },
    {
      "epoch": 0.17752477335020028,
      "grad_norm": 0.48948225378990173,
      "learning_rate": 0.00019124189596264626,
      "loss": 0.2096,
      "step": 12630
    },
    {
      "epoch": 0.17766533136552112,
      "grad_norm": 0.4309072494506836,
      "learning_rate": 0.00019122286291966195,
      "loss": 0.2049,
      "step": 12640
    },
    {
      "epoch": 0.17780588938084194,
      "grad_norm": 0.4598619043827057,
      "learning_rate": 0.00019120381016711008,
      "loss": 0.1806,
      "step": 12650
    },
    {
      "epoch": 0.17794644739616278,
      "grad_norm": 0.405023455619812,
      "learning_rate": 0.0001911847377091072,
      "loss": 0.2084,
      "step": 12660
    },
    {
      "epoch": 0.1780870054114836,
      "grad_norm": 0.4332144260406494,
      "learning_rate": 0.00019116564554977404,
      "loss": 0.2318,
      "step": 12670
    },
    {
      "epoch": 0.1782275634268044,
      "grad_norm": 0.5438380837440491,
      "learning_rate": 0.00019114653369323566,
      "loss": 0.2146,
      "step": 12680
    },
    {
      "epoch": 0.17836812144212524,
      "grad_norm": 0.48414480686187744,
      "learning_rate": 0.00019112740214362136,
      "loss": 0.1936,
      "step": 12690
    },
    {
      "epoch": 0.17850867945744606,
      "grad_norm": 0.5146945118904114,
      "learning_rate": 0.00019110825090506468,
      "loss": 0.2198,
      "step": 12700
    },
    {
      "epoch": 0.1786492374727669,
      "grad_norm": 0.4110274314880371,
      "learning_rate": 0.00019108907998170344,
      "loss": 0.2242,
      "step": 12710
    },
    {
      "epoch": 0.1787897954880877,
      "grad_norm": 0.5148970484733582,
      "learning_rate": 0.00019106988937767972,
      "loss": 0.2179,
      "step": 12720
    },
    {
      "epoch": 0.17893035350340852,
      "grad_norm": 0.48713383078575134,
      "learning_rate": 0.0001910506790971398,
      "loss": 0.2264,
      "step": 12730
    },
    {
      "epoch": 0.17907091151872936,
      "grad_norm": 0.4454123377799988,
      "learning_rate": 0.0001910314491442343,
      "loss": 0.2162,
      "step": 12740
    },
    {
      "epoch": 0.17921146953405018,
      "grad_norm": 0.46519991755485535,
      "learning_rate": 0.00019101219952311795,
      "loss": 0.2241,
      "step": 12750
    },
    {
      "epoch": 0.179352027549371,
      "grad_norm": 0.48231005668640137,
      "learning_rate": 0.0001909929302379499,
      "loss": 0.2213,
      "step": 12760
    },
    {
      "epoch": 0.17949258556469183,
      "grad_norm": 0.49811065196990967,
      "learning_rate": 0.00019097364129289337,
      "loss": 0.2279,
      "step": 12770
    },
    {
      "epoch": 0.17963314358001264,
      "grad_norm": 0.3791274428367615,
      "learning_rate": 0.00019095433269211598,
      "loss": 0.2109,
      "step": 12780
    },
    {
      "epoch": 0.17977370159533348,
      "grad_norm": 0.4766414165496826,
      "learning_rate": 0.00019093500443978956,
      "loss": 0.2173,
      "step": 12790
    },
    {
      "epoch": 0.1799142596106543,
      "grad_norm": 0.49525031447410583,
      "learning_rate": 0.00019091565654009009,
      "loss": 0.2219,
      "step": 12800
    },
    {
      "epoch": 0.1800548176259751,
      "grad_norm": 0.529272735118866,
      "learning_rate": 0.00019089628899719791,
      "loss": 0.2211,
      "step": 12810
    },
    {
      "epoch": 0.18019537564129595,
      "grad_norm": 0.519967257976532,
      "learning_rate": 0.00019087690181529758,
      "loss": 0.2226,
      "step": 12820
    },
    {
      "epoch": 0.18033593365661676,
      "grad_norm": 0.4815217852592468,
      "learning_rate": 0.00019085749499857786,
      "loss": 0.2246,
      "step": 12830
    },
    {
      "epoch": 0.1804764916719376,
      "grad_norm": 0.4821476936340332,
      "learning_rate": 0.00019083806855123178,
      "loss": 0.2039,
      "step": 12840
    },
    {
      "epoch": 0.18061704968725842,
      "grad_norm": 0.6629970073699951,
      "learning_rate": 0.00019081862247745659,
      "loss": 0.2119,
      "step": 12850
    },
    {
      "epoch": 0.18075760770257923,
      "grad_norm": 0.39005452394485474,
      "learning_rate": 0.00019079915678145385,
      "loss": 0.214,
      "step": 12860
    },
    {
      "epoch": 0.18089816571790007,
      "grad_norm": 0.39662966132164,
      "learning_rate": 0.00019077967146742926,
      "loss": 0.2117,
      "step": 12870
    },
    {
      "epoch": 0.18103872373322089,
      "grad_norm": 0.48629820346832275,
      "learning_rate": 0.00019076016653959282,
      "loss": 0.1954,
      "step": 12880
    },
    {
      "epoch": 0.1811792817485417,
      "grad_norm": 0.5073099732398987,
      "learning_rate": 0.00019074064200215877,
      "loss": 0.2155,
      "step": 12890
    },
    {
      "epoch": 0.18131983976386254,
      "grad_norm": 0.46670103073120117,
      "learning_rate": 0.00019072109785934558,
      "loss": 0.2193,
      "step": 12900
    },
    {
      "epoch": 0.18146039777918335,
      "grad_norm": 0.5066492557525635,
      "learning_rate": 0.0001907015341153759,
      "loss": 0.2282,
      "step": 12910
    },
    {
      "epoch": 0.1816009557945042,
      "grad_norm": 0.39050865173339844,
      "learning_rate": 0.00019068195077447673,
      "loss": 0.1896,
      "step": 12920
    },
    {
      "epoch": 0.181741513809825,
      "grad_norm": 0.5603247880935669,
      "learning_rate": 0.0001906623478408792,
      "loss": 0.2103,
      "step": 12930
    },
    {
      "epoch": 0.18188207182514582,
      "grad_norm": 0.6061719655990601,
      "learning_rate": 0.0001906427253188187,
      "loss": 0.2125,
      "step": 12940
    },
    {
      "epoch": 0.18202262984046666,
      "grad_norm": 0.3703446090221405,
      "learning_rate": 0.00019062308321253485,
      "loss": 0.2129,
      "step": 12950
    },
    {
      "epoch": 0.18216318785578747,
      "grad_norm": 0.5037427544593811,
      "learning_rate": 0.0001906034215262716,
      "loss": 0.2336,
      "step": 12960
    },
    {
      "epoch": 0.1823037458711083,
      "grad_norm": 0.6292206048965454,
      "learning_rate": 0.0001905837402642769,
      "loss": 0.2211,
      "step": 12970
    },
    {
      "epoch": 0.18244430388642913,
      "grad_norm": 0.44565314054489136,
      "learning_rate": 0.00019056403943080321,
      "loss": 0.2081,
      "step": 12980
    },
    {
      "epoch": 0.18258486190174994,
      "grad_norm": 0.6111086010932922,
      "learning_rate": 0.000190544319030107,
      "loss": 0.2155,
      "step": 12990
    },
    {
      "epoch": 0.18272541991707078,
      "grad_norm": 0.45473793148994446,
      "learning_rate": 0.0001905245790664491,
      "loss": 0.245,
      "step": 13000
    },
    {
      "epoch": 0.18272541991707078,
      "eval_chrf": 81.71859402430363,
      "eval_loss": 0.4482495188713074,
      "eval_runtime": 258.9033,
      "eval_samples_per_second": 0.386,
      "eval_steps_per_second": 0.015,
      "step": 13000
    },
    {
      "epoch": 0.1828659779323916,
      "grad_norm": 0.46908169984817505,
      "learning_rate": 0.0001905048195440945,
      "loss": 0.1984,
      "step": 13010
    },
    {
      "epoch": 0.1830065359477124,
      "grad_norm": 0.4461404085159302,
      "learning_rate": 0.00019048504046731242,
      "loss": 0.1747,
      "step": 13020
    },
    {
      "epoch": 0.18314709396303325,
      "grad_norm": 0.5931270718574524,
      "learning_rate": 0.00019046524184037632,
      "loss": 0.1915,
      "step": 13030
    },
    {
      "epoch": 0.18328765197835406,
      "grad_norm": 0.4598511755466461,
      "learning_rate": 0.0001904454236675639,
      "loss": 0.2267,
      "step": 13040
    },
    {
      "epoch": 0.1834282099936749,
      "grad_norm": 0.451461523771286,
      "learning_rate": 0.00019042558595315707,
      "loss": 0.2103,
      "step": 13050
    },
    {
      "epoch": 0.18356876800899571,
      "grad_norm": 0.3634202778339386,
      "learning_rate": 0.00019040572870144192,
      "loss": 0.2564,
      "step": 13060
    },
    {
      "epoch": 0.18370932602431653,
      "grad_norm": 0.43997520208358765,
      "learning_rate": 0.0001903858519167088,
      "loss": 0.2137,
      "step": 13070
    },
    {
      "epoch": 0.18384988403963737,
      "grad_norm": 0.3918921947479248,
      "learning_rate": 0.00019036595560325236,
      "loss": 0.2077,
      "step": 13080
    },
    {
      "epoch": 0.18399044205495818,
      "grad_norm": 0.43988847732543945,
      "learning_rate": 0.00019034603976537126,
      "loss": 0.2317,
      "step": 13090
    },
    {
      "epoch": 0.184131000070279,
      "grad_norm": 0.39643704891204834,
      "learning_rate": 0.00019032610440736862,
      "loss": 0.2075,
      "step": 13100
    },
    {
      "epoch": 0.18427155808559983,
      "grad_norm": 0.38452738523483276,
      "learning_rate": 0.0001903061495335516,
      "loss": 0.2103,
      "step": 13110
    },
    {
      "epoch": 0.18441211610092065,
      "grad_norm": 0.37879735231399536,
      "learning_rate": 0.00019028617514823163,
      "loss": 0.2307,
      "step": 13120
    },
    {
      "epoch": 0.1845526741162415,
      "grad_norm": 0.469174861907959,
      "learning_rate": 0.0001902661812557244,
      "loss": 0.1975,
      "step": 13130
    },
    {
      "epoch": 0.1846932321315623,
      "grad_norm": 0.4302513301372528,
      "learning_rate": 0.0001902461678603498,
      "loss": 0.2394,
      "step": 13140
    },
    {
      "epoch": 0.18483379014688311,
      "grad_norm": 0.5000349879264832,
      "learning_rate": 0.00019022613496643187,
      "loss": 0.1926,
      "step": 13150
    },
    {
      "epoch": 0.18497434816220396,
      "grad_norm": 0.4314963221549988,
      "learning_rate": 0.0001902060825782989,
      "loss": 0.2047,
      "step": 13160
    },
    {
      "epoch": 0.18511490617752477,
      "grad_norm": 0.33490025997161865,
      "learning_rate": 0.00019018601070028344,
      "loss": 0.2165,
      "step": 13170
    },
    {
      "epoch": 0.1852554641928456,
      "grad_norm": 0.5817905068397522,
      "learning_rate": 0.00019016591933672223,
      "loss": 0.2041,
      "step": 13180
    },
    {
      "epoch": 0.18539602220816642,
      "grad_norm": 0.47698774933815,
      "learning_rate": 0.0001901458084919561,
      "loss": 0.203,
      "step": 13190
    },
    {
      "epoch": 0.18553658022348724,
      "grad_norm": 0.4562699794769287,
      "learning_rate": 0.0001901256781703303,
      "loss": 0.2205,
      "step": 13200
    },
    {
      "epoch": 0.18567713823880808,
      "grad_norm": 0.4335058629512787,
      "learning_rate": 0.00019010552837619413,
      "loss": 0.2257,
      "step": 13210
    },
    {
      "epoch": 0.1858176962541289,
      "grad_norm": 0.6168200373649597,
      "learning_rate": 0.00019008535911390113,
      "loss": 0.1967,
      "step": 13220
    },
    {
      "epoch": 0.1859582542694497,
      "grad_norm": 0.5537468194961548,
      "learning_rate": 0.0001900651703878091,
      "loss": 0.2206,
      "step": 13230
    },
    {
      "epoch": 0.18609881228477054,
      "grad_norm": 0.37052398920059204,
      "learning_rate": 0.00019004496220227996,
      "loss": 0.2306,
      "step": 13240
    },
    {
      "epoch": 0.18623937030009136,
      "grad_norm": 0.4688161313533783,
      "learning_rate": 0.0001900247345616799,
      "loss": 0.2218,
      "step": 13250
    },
    {
      "epoch": 0.1863799283154122,
      "grad_norm": 0.41777825355529785,
      "learning_rate": 0.0001900044874703793,
      "loss": 0.2035,
      "step": 13260
    },
    {
      "epoch": 0.186520486330733,
      "grad_norm": 0.3905136287212372,
      "learning_rate": 0.00018998422093275275,
      "loss": 0.2016,
      "step": 13270
    },
    {
      "epoch": 0.18666104434605382,
      "grad_norm": 0.43584224581718445,
      "learning_rate": 0.000189963934953179,
      "loss": 0.21,
      "step": 13280
    },
    {
      "epoch": 0.18680160236137466,
      "grad_norm": 0.40059593319892883,
      "learning_rate": 0.00018994362953604106,
      "loss": 0.2297,
      "step": 13290
    },
    {
      "epoch": 0.18694216037669548,
      "grad_norm": 0.5512275099754333,
      "learning_rate": 0.00018992330468572605,
      "loss": 0.2083,
      "step": 13300
    },
    {
      "epoch": 0.18708271839201632,
      "grad_norm": 0.49532201886177063,
      "learning_rate": 0.00018990296040662544,
      "loss": 0.2327,
      "step": 13310
    },
    {
      "epoch": 0.18722327640733713,
      "grad_norm": 0.4382390081882477,
      "learning_rate": 0.0001898825967031347,
      "loss": 0.2257,
      "step": 13320
    },
    {
      "epoch": 0.18736383442265794,
      "grad_norm": 0.6387760043144226,
      "learning_rate": 0.0001898622135796537,
      "loss": 0.1984,
      "step": 13330
    },
    {
      "epoch": 0.18750439243797878,
      "grad_norm": 0.35923030972480774,
      "learning_rate": 0.00018984181104058636,
      "loss": 0.2124,
      "step": 13340
    },
    {
      "epoch": 0.1876449504532996,
      "grad_norm": 0.4097476303577423,
      "learning_rate": 0.00018982138909034082,
      "loss": 0.2065,
      "step": 13350
    },
    {
      "epoch": 0.1877855084686204,
      "grad_norm": 0.4327114224433899,
      "learning_rate": 0.00018980094773332948,
      "loss": 0.2198,
      "step": 13360
    },
    {
      "epoch": 0.18792606648394125,
      "grad_norm": 0.41230443120002747,
      "learning_rate": 0.00018978048697396887,
      "loss": 0.2367,
      "step": 13370
    },
    {
      "epoch": 0.18806662449926206,
      "grad_norm": 0.4218708276748657,
      "learning_rate": 0.00018976000681667973,
      "loss": 0.1985,
      "step": 13380
    },
    {
      "epoch": 0.1882071825145829,
      "grad_norm": 0.6109094619750977,
      "learning_rate": 0.00018973950726588698,
      "loss": 0.2591,
      "step": 13390
    },
    {
      "epoch": 0.18834774052990372,
      "grad_norm": 0.43765881657600403,
      "learning_rate": 0.0001897189883260198,
      "loss": 0.2243,
      "step": 13400
    },
    {
      "epoch": 0.18848829854522453,
      "grad_norm": 0.44101229310035706,
      "learning_rate": 0.0001896984500015114,
      "loss": 0.2173,
      "step": 13410
    },
    {
      "epoch": 0.18862885656054537,
      "grad_norm": 0.6721788048744202,
      "learning_rate": 0.0001896778922967994,
      "loss": 0.1979,
      "step": 13420
    },
    {
      "epoch": 0.18876941457586618,
      "grad_norm": 0.5215440392494202,
      "learning_rate": 0.0001896573152163254,
      "loss": 0.2283,
      "step": 13430
    },
    {
      "epoch": 0.18890997259118703,
      "grad_norm": 0.5103262662887573,
      "learning_rate": 0.00018963671876453533,
      "loss": 0.2215,
      "step": 13440
    },
    {
      "epoch": 0.18905053060650784,
      "grad_norm": 0.5467560887336731,
      "learning_rate": 0.0001896161029458792,
      "loss": 0.2221,
      "step": 13450
    },
    {
      "epoch": 0.18919108862182865,
      "grad_norm": 0.4276997745037079,
      "learning_rate": 0.0001895954677648113,
      "loss": 0.213,
      "step": 13460
    },
    {
      "epoch": 0.1893316466371495,
      "grad_norm": 0.41185396909713745,
      "learning_rate": 0.00018957481322579,
      "loss": 0.2052,
      "step": 13470
    },
    {
      "epoch": 0.1894722046524703,
      "grad_norm": 0.4776594936847687,
      "learning_rate": 0.00018955413933327797,
      "loss": 0.1955,
      "step": 13480
    },
    {
      "epoch": 0.18961276266779112,
      "grad_norm": 0.4993971288204193,
      "learning_rate": 0.00018953344609174197,
      "loss": 0.2146,
      "step": 13490
    },
    {
      "epoch": 0.18975332068311196,
      "grad_norm": 0.4730333685874939,
      "learning_rate": 0.00018951273350565297,
      "loss": 0.2338,
      "step": 13500
    },
    {
      "epoch": 0.18989387869843277,
      "grad_norm": 0.44033026695251465,
      "learning_rate": 0.00018949200157948614,
      "loss": 0.2186,
      "step": 13510
    },
    {
      "epoch": 0.1900344367137536,
      "grad_norm": 0.35666805505752563,
      "learning_rate": 0.00018947125031772076,
      "loss": 0.1963,
      "step": 13520
    },
    {
      "epoch": 0.19017499472907443,
      "grad_norm": 0.5002583861351013,
      "learning_rate": 0.00018945047972484042,
      "loss": 0.2188,
      "step": 13530
    },
    {
      "epoch": 0.19031555274439524,
      "grad_norm": 0.5445455312728882,
      "learning_rate": 0.00018942968980533275,
      "loss": 0.2194,
      "step": 13540
    },
    {
      "epoch": 0.19045611075971608,
      "grad_norm": 0.47202199697494507,
      "learning_rate": 0.00018940888056368963,
      "loss": 0.232,
      "step": 13550
    },
    {
      "epoch": 0.1905966687750369,
      "grad_norm": 0.430093914270401,
      "learning_rate": 0.00018938805200440707,
      "loss": 0.1965,
      "step": 13560
    },
    {
      "epoch": 0.19073722679035773,
      "grad_norm": 0.42162808775901794,
      "learning_rate": 0.0001893672041319853,
      "loss": 0.2359,
      "step": 13570
    },
    {
      "epoch": 0.19087778480567855,
      "grad_norm": 0.37718167901039124,
      "learning_rate": 0.0001893463369509287,
      "loss": 0.2112,
      "step": 13580
    },
    {
      "epoch": 0.19101834282099936,
      "grad_norm": 0.5031275153160095,
      "learning_rate": 0.0001893254504657458,
      "loss": 0.2085,
      "step": 13590
    },
    {
      "epoch": 0.1911589008363202,
      "grad_norm": 0.5971906781196594,
      "learning_rate": 0.00018930454468094933,
      "loss": 0.2166,
      "step": 13600
    },
    {
      "epoch": 0.191299458851641,
      "grad_norm": 0.39828595519065857,
      "learning_rate": 0.00018928361960105622,
      "loss": 0.2174,
      "step": 13610
    },
    {
      "epoch": 0.19144001686696183,
      "grad_norm": 0.4507116377353668,
      "learning_rate": 0.0001892626752305875,
      "loss": 0.2281,
      "step": 13620
    },
    {
      "epoch": 0.19158057488228267,
      "grad_norm": 0.5327116847038269,
      "learning_rate": 0.00018924171157406843,
      "loss": 0.1867,
      "step": 13630
    },
    {
      "epoch": 0.19172113289760348,
      "grad_norm": 0.4895416498184204,
      "learning_rate": 0.00018922072863602837,
      "loss": 0.2094,
      "step": 13640
    },
    {
      "epoch": 0.19186169091292432,
      "grad_norm": 0.5132678747177124,
      "learning_rate": 0.00018919972642100086,
      "loss": 0.1946,
      "step": 13650
    },
    {
      "epoch": 0.19200224892824513,
      "grad_norm": 0.4152938723564148,
      "learning_rate": 0.00018917870493352371,
      "loss": 0.2102,
      "step": 13660
    },
    {
      "epoch": 0.19214280694356595,
      "grad_norm": 0.33532798290252686,
      "learning_rate": 0.00018915766417813875,
      "loss": 0.2144,
      "step": 13670
    },
    {
      "epoch": 0.1922833649588868,
      "grad_norm": 0.5051287412643433,
      "learning_rate": 0.00018913660415939207,
      "loss": 0.2253,
      "step": 13680
    },
    {
      "epoch": 0.1924239229742076,
      "grad_norm": 0.481604665517807,
      "learning_rate": 0.0001891155248818339,
      "loss": 0.2151,
      "step": 13690
    },
    {
      "epoch": 0.19256448098952844,
      "grad_norm": 0.528761088848114,
      "learning_rate": 0.00018909442635001855,
      "loss": 0.2135,
      "step": 13700
    },
    {
      "epoch": 0.19270503900484925,
      "grad_norm": 0.5373625159263611,
      "learning_rate": 0.0001890733085685046,
      "loss": 0.2001,
      "step": 13710
    },
    {
      "epoch": 0.19284559702017007,
      "grad_norm": 0.4210106432437897,
      "learning_rate": 0.00018905217154185474,
      "loss": 0.2519,
      "step": 13720
    },
    {
      "epoch": 0.1929861550354909,
      "grad_norm": 0.43053504824638367,
      "learning_rate": 0.00018903101527463587,
      "loss": 0.2049,
      "step": 13730
    },
    {
      "epoch": 0.19312671305081172,
      "grad_norm": 0.6001156568527222,
      "learning_rate": 0.00018900983977141893,
      "loss": 0.2079,
      "step": 13740
    },
    {
      "epoch": 0.19326727106613253,
      "grad_norm": 0.4350476562976837,
      "learning_rate": 0.00018898864503677913,
      "loss": 0.211,
      "step": 13750
    },
    {
      "epoch": 0.19340782908145338,
      "grad_norm": 0.43925341963768005,
      "learning_rate": 0.00018896743107529583,
      "loss": 0.2195,
      "step": 13760
    },
    {
      "epoch": 0.1935483870967742,
      "grad_norm": 0.5062753558158875,
      "learning_rate": 0.0001889461978915524,
      "loss": 0.2326,
      "step": 13770
    },
    {
      "epoch": 0.19368894511209503,
      "grad_norm": 0.47204822301864624,
      "learning_rate": 0.0001889249454901366,
      "loss": 0.2176,
      "step": 13780
    },
    {
      "epoch": 0.19382950312741584,
      "grad_norm": 0.41110143065452576,
      "learning_rate": 0.00018890367387564013,
      "loss": 0.1917,
      "step": 13790
    },
    {
      "epoch": 0.19397006114273666,
      "grad_norm": 0.4885755777359009,
      "learning_rate": 0.00018888238305265893,
      "loss": 0.2354,
      "step": 13800
    },
    {
      "epoch": 0.1941106191580575,
      "grad_norm": 0.392553448677063,
      "learning_rate": 0.00018886107302579314,
      "loss": 0.2213,
      "step": 13810
    },
    {
      "epoch": 0.1942511771733783,
      "grad_norm": 0.40750473737716675,
      "learning_rate": 0.00018883974379964693,
      "loss": 0.2022,
      "step": 13820
    },
    {
      "epoch": 0.19439173518869912,
      "grad_norm": 0.37037041783332825,
      "learning_rate": 0.0001888183953788287,
      "loss": 0.2058,
      "step": 13830
    },
    {
      "epoch": 0.19453229320401996,
      "grad_norm": 0.5609288215637207,
      "learning_rate": 0.00018879702776795099,
      "loss": 0.22,
      "step": 13840
    },
    {
      "epoch": 0.19467285121934078,
      "grad_norm": 0.6449435949325562,
      "learning_rate": 0.0001887756409716305,
      "loss": 0.2034,
      "step": 13850
    },
    {
      "epoch": 0.19481340923466162,
      "grad_norm": 0.4815952777862549,
      "learning_rate": 0.00018875423499448802,
      "loss": 0.2233,
      "step": 13860
    },
    {
      "epoch": 0.19495396724998243,
      "grad_norm": 0.5087131857872009,
      "learning_rate": 0.00018873280984114848,
      "loss": 0.1971,
      "step": 13870
    },
    {
      "epoch": 0.19509452526530324,
      "grad_norm": 0.5035146474838257,
      "learning_rate": 0.00018871136551624104,
      "loss": 0.2373,
      "step": 13880
    },
    {
      "epoch": 0.19523508328062408,
      "grad_norm": 0.357378214597702,
      "learning_rate": 0.000188689902024399,
      "loss": 0.2291,
      "step": 13890
    },
    {
      "epoch": 0.1953756412959449,
      "grad_norm": 0.40452587604522705,
      "learning_rate": 0.00018866841937025962,
      "loss": 0.2147,
      "step": 13900
    },
    {
      "epoch": 0.19551619931126574,
      "grad_norm": 0.4429807662963867,
      "learning_rate": 0.0001886469175584645,
      "loss": 0.1939,
      "step": 13910
    },
    {
      "epoch": 0.19565675732658655,
      "grad_norm": 0.5234241485595703,
      "learning_rate": 0.00018862539659365935,
      "loss": 0.201,
      "step": 13920
    },
    {
      "epoch": 0.19579731534190736,
      "grad_norm": 0.3796229958534241,
      "learning_rate": 0.00018860385648049392,
      "loss": 0.2417,
      "step": 13930
    },
    {
      "epoch": 0.1959378733572282,
      "grad_norm": 0.3806133568286896,
      "learning_rate": 0.0001885822972236222,
      "loss": 0.2303,
      "step": 13940
    },
    {
      "epoch": 0.19607843137254902,
      "grad_norm": 0.49205514788627625,
      "learning_rate": 0.00018856071882770222,
      "loss": 0.2181,
      "step": 13950
    },
    {
      "epoch": 0.19621898938786983,
      "grad_norm": 0.4757485091686249,
      "learning_rate": 0.00018853912129739625,
      "loss": 0.2091,
      "step": 13960
    },
    {
      "epoch": 0.19635954740319067,
      "grad_norm": 0.5125443935394287,
      "learning_rate": 0.00018851750463737063,
      "loss": 0.2024,
      "step": 13970
    },
    {
      "epoch": 0.19650010541851148,
      "grad_norm": 0.49263033270835876,
      "learning_rate": 0.0001884958688522958,
      "loss": 0.2257,
      "step": 13980
    },
    {
      "epoch": 0.19664066343383232,
      "grad_norm": 0.5992262959480286,
      "learning_rate": 0.00018847421394684643,
      "loss": 0.2042,
      "step": 13990
    },
    {
      "epoch": 0.19678122144915314,
      "grad_norm": 0.4547590911388397,
      "learning_rate": 0.00018845253992570124,
      "loss": 0.24,
      "step": 14000
    },
    {
      "epoch": 0.19678122144915314,
      "eval_chrf": 82.73228602148815,
      "eval_loss": 0.4420166313648224,
      "eval_runtime": 190.7174,
      "eval_samples_per_second": 0.524,
      "eval_steps_per_second": 0.021,
      "step": 14000
    },
    {
      "epoch": 0.19692177946447395,
      "grad_norm": 0.287767618894577,
      "learning_rate": 0.00018843084679354315,
      "loss": 0.2096,
      "step": 14010
    },
    {
      "epoch": 0.1970623374797948,
      "grad_norm": 0.4257965087890625,
      "learning_rate": 0.0001884091345550591,
      "loss": 0.2235,
      "step": 14020
    },
    {
      "epoch": 0.1972028954951156,
      "grad_norm": 0.5454546213150024,
      "learning_rate": 0.0001883874032149403,
      "loss": 0.1984,
      "step": 14030
    },
    {
      "epoch": 0.19734345351043645,
      "grad_norm": 0.4185550808906555,
      "learning_rate": 0.00018836565277788195,
      "loss": 0.246,
      "step": 14040
    },
    {
      "epoch": 0.19748401152575726,
      "grad_norm": 0.41897377371788025,
      "learning_rate": 0.00018834388324858348,
      "loss": 0.2287,
      "step": 14050
    },
    {
      "epoch": 0.19762456954107807,
      "grad_norm": 0.42247578501701355,
      "learning_rate": 0.00018832209463174836,
      "loss": 0.186,
      "step": 14060
    },
    {
      "epoch": 0.1977651275563989,
      "grad_norm": 0.4365829527378082,
      "learning_rate": 0.0001883002869320843,
      "loss": 0.2034,
      "step": 14070
    },
    {
      "epoch": 0.19790568557171972,
      "grad_norm": 0.7284541130065918,
      "learning_rate": 0.00018827846015430298,
      "loss": 0.1982,
      "step": 14080
    },
    {
      "epoch": 0.19804624358704054,
      "grad_norm": 0.3755764365196228,
      "learning_rate": 0.00018825661430312037,
      "loss": 0.2262,
      "step": 14090
    },
    {
      "epoch": 0.19818680160236138,
      "grad_norm": 0.4915720224380493,
      "learning_rate": 0.0001882347493832564,
      "loss": 0.1925,
      "step": 14100
    },
    {
      "epoch": 0.1983273596176822,
      "grad_norm": 0.41482189297676086,
      "learning_rate": 0.00018821286539943518,
      "loss": 0.2127,
      "step": 14110
    },
    {
      "epoch": 0.19846791763300303,
      "grad_norm": 0.44070643186569214,
      "learning_rate": 0.00018819096235638504,
      "loss": 0.2281,
      "step": 14120
    },
    {
      "epoch": 0.19860847564832385,
      "grad_norm": 0.5023027658462524,
      "learning_rate": 0.00018816904025883825,
      "loss": 0.2305,
      "step": 14130
    },
    {
      "epoch": 0.19874903366364466,
      "grad_norm": 0.5836760997772217,
      "learning_rate": 0.00018814709911153137,
      "loss": 0.2253,
      "step": 14140
    },
    {
      "epoch": 0.1988895916789655,
      "grad_norm": 0.40505480766296387,
      "learning_rate": 0.00018812513891920495,
      "loss": 0.2162,
      "step": 14150
    },
    {
      "epoch": 0.1990301496942863,
      "grad_norm": 0.5243901014328003,
      "learning_rate": 0.0001881031596866037,
      "loss": 0.2209,
      "step": 14160
    },
    {
      "epoch": 0.19917070770960715,
      "grad_norm": 0.36334285140037537,
      "learning_rate": 0.00018808116141847642,
      "loss": 0.2311,
      "step": 14170
    },
    {
      "epoch": 0.19931126572492797,
      "grad_norm": 0.4597943425178528,
      "learning_rate": 0.00018805914411957608,
      "loss": 0.1936,
      "step": 14180
    },
    {
      "epoch": 0.19945182374024878,
      "grad_norm": 0.44531357288360596,
      "learning_rate": 0.00018803710779465973,
      "loss": 0.1955,
      "step": 14190
    },
    {
      "epoch": 0.19959238175556962,
      "grad_norm": 0.47967594861984253,
      "learning_rate": 0.0001880150524484885,
      "loss": 0.235,
      "step": 14200
    },
    {
      "epoch": 0.19973293977089043,
      "grad_norm": 0.39149224758148193,
      "learning_rate": 0.0001879929780858277,
      "loss": 0.2078,
      "step": 14210
    },
    {
      "epoch": 0.19987349778621125,
      "grad_norm": 0.5905307531356812,
      "learning_rate": 0.00018797088471144666,
      "loss": 0.1982,
      "step": 14220
    },
    {
      "epoch": 0.2000140558015321,
      "grad_norm": 0.3699857294559479,
      "learning_rate": 0.00018794877233011887,
      "loss": 0.2067,
      "step": 14230
    },
    {
      "epoch": 0.2001546138168529,
      "grad_norm": 0.3734479546546936,
      "learning_rate": 0.00018792664094662197,
      "loss": 0.2056,
      "step": 14240
    },
    {
      "epoch": 0.20029517183217374,
      "grad_norm": 0.40720364451408386,
      "learning_rate": 0.00018790449056573758,
      "loss": 0.2302,
      "step": 14250
    },
    {
      "epoch": 0.20043572984749455,
      "grad_norm": 0.5361537933349609,
      "learning_rate": 0.00018788232119225153,
      "loss": 0.2193,
      "step": 14260
    },
    {
      "epoch": 0.20057628786281537,
      "grad_norm": 0.46045634150505066,
      "learning_rate": 0.00018786013283095376,
      "loss": 0.185,
      "step": 14270
    },
    {
      "epoch": 0.2007168458781362,
      "grad_norm": 0.569119393825531,
      "learning_rate": 0.00018783792548663825,
      "loss": 0.1936,
      "step": 14280
    },
    {
      "epoch": 0.20085740389345702,
      "grad_norm": 0.4964165985584259,
      "learning_rate": 0.00018781569916410308,
      "loss": 0.2168,
      "step": 14290
    },
    {
      "epoch": 0.20099796190877786,
      "grad_norm": 0.6262520551681519,
      "learning_rate": 0.00018779345386815054,
      "loss": 0.2686,
      "step": 14300
    },
    {
      "epoch": 0.20113851992409867,
      "grad_norm": 0.8814814686775208,
      "learning_rate": 0.00018777118960358683,
      "loss": 0.2103,
      "step": 14310
    },
    {
      "epoch": 0.2012790779394195,
      "grad_norm": 0.376753568649292,
      "learning_rate": 0.0001877489063752224,
      "loss": 0.2288,
      "step": 14320
    },
    {
      "epoch": 0.20141963595474033,
      "grad_norm": 0.38907644152641296,
      "learning_rate": 0.00018772660418787184,
      "loss": 0.1988,
      "step": 14330
    },
    {
      "epoch": 0.20156019397006114,
      "grad_norm": 0.4345405399799347,
      "learning_rate": 0.00018770428304635363,
      "loss": 0.2709,
      "step": 14340
    },
    {
      "epoch": 0.20170075198538195,
      "grad_norm": 0.4775392413139343,
      "learning_rate": 0.0001876819429554905,
      "loss": 0.2192,
      "step": 14350
    },
    {
      "epoch": 0.2018413100007028,
      "grad_norm": 0.5034828186035156,
      "learning_rate": 0.00018765958392010925,
      "loss": 0.2587,
      "step": 14360
    },
    {
      "epoch": 0.2019818680160236,
      "grad_norm": 0.4657890498638153,
      "learning_rate": 0.00018763720594504077,
      "loss": 0.2169,
      "step": 14370
    },
    {
      "epoch": 0.20212242603134445,
      "grad_norm": 0.4603046774864197,
      "learning_rate": 0.00018761480903512002,
      "loss": 0.1639,
      "step": 14380
    },
    {
      "epoch": 0.20226298404666526,
      "grad_norm": 0.4143376648426056,
      "learning_rate": 0.00018759239319518603,
      "loss": 0.2072,
      "step": 14390
    },
    {
      "epoch": 0.20240354206198607,
      "grad_norm": 0.45824283361434937,
      "learning_rate": 0.00018756995843008208,
      "loss": 0.2259,
      "step": 14400
    },
    {
      "epoch": 0.20254410007730692,
      "grad_norm": 0.4499847888946533,
      "learning_rate": 0.0001875475047446553,
      "loss": 0.1921,
      "step": 14410
    },
    {
      "epoch": 0.20268465809262773,
      "grad_norm": 0.46914368867874146,
      "learning_rate": 0.000187525032143757,
      "loss": 0.1884,
      "step": 14420
    },
    {
      "epoch": 0.20282521610794857,
      "grad_norm": 0.48389866948127747,
      "learning_rate": 0.00018750254063224268,
      "loss": 0.1902,
      "step": 14430
    },
    {
      "epoch": 0.20296577412326938,
      "grad_norm": 0.4360426962375641,
      "learning_rate": 0.00018748003021497184,
      "loss": 0.2002,
      "step": 14440
    },
    {
      "epoch": 0.2031063321385902,
      "grad_norm": 0.4696538746356964,
      "learning_rate": 0.00018745750089680803,
      "loss": 0.21,
      "step": 14450
    },
    {
      "epoch": 0.20324689015391104,
      "grad_norm": 0.7128496766090393,
      "learning_rate": 0.00018743495268261896,
      "loss": 0.2014,
      "step": 14460
    },
    {
      "epoch": 0.20338744816923185,
      "grad_norm": 0.5214452147483826,
      "learning_rate": 0.00018741238557727635,
      "loss": 0.243,
      "step": 14470
    },
    {
      "epoch": 0.20352800618455266,
      "grad_norm": 0.5012695789337158,
      "learning_rate": 0.00018738979958565603,
      "loss": 0.2158,
      "step": 14480
    },
    {
      "epoch": 0.2036685641998735,
      "grad_norm": 0.5507377982139587,
      "learning_rate": 0.000187367194712638,
      "loss": 0.2138,
      "step": 14490
    },
    {
      "epoch": 0.20380912221519432,
      "grad_norm": 0.3723544180393219,
      "learning_rate": 0.00018734457096310613,
      "loss": 0.2079,
      "step": 14500
    },
    {
      "epoch": 0.20394968023051516,
      "grad_norm": 0.40772634744644165,
      "learning_rate": 0.00018732192834194862,
      "loss": 0.2123,
      "step": 14510
    },
    {
      "epoch": 0.20409023824583597,
      "grad_norm": 0.5124629735946655,
      "learning_rate": 0.00018729926685405755,
      "loss": 0.21,
      "step": 14520
    },
    {
      "epoch": 0.20423079626115678,
      "grad_norm": 0.45871278643608093,
      "learning_rate": 0.00018727658650432918,
      "loss": 0.2142,
      "step": 14530
    },
    {
      "epoch": 0.20437135427647762,
      "grad_norm": 0.41411224007606506,
      "learning_rate": 0.0001872538872976638,
      "loss": 0.2149,
      "step": 14540
    },
    {
      "epoch": 0.20451191229179844,
      "grad_norm": 0.554733157157898,
      "learning_rate": 0.0001872311692389658,
      "loss": 0.2243,
      "step": 14550
    },
    {
      "epoch": 0.20465247030711925,
      "grad_norm": 0.46813228726387024,
      "learning_rate": 0.00018720843233314364,
      "loss": 0.2046,
      "step": 14560
    },
    {
      "epoch": 0.2047930283224401,
      "grad_norm": 0.43941813707351685,
      "learning_rate": 0.0001871856765851098,
      "loss": 0.2083,
      "step": 14570
    },
    {
      "epoch": 0.2049335863377609,
      "grad_norm": 0.47301948070526123,
      "learning_rate": 0.00018716290199978093,
      "loss": 0.206,
      "step": 14580
    },
    {
      "epoch": 0.20507414435308174,
      "grad_norm": 0.39562416076660156,
      "learning_rate": 0.00018714010858207768,
      "loss": 0.2112,
      "step": 14590
    },
    {
      "epoch": 0.20521470236840256,
      "grad_norm": 0.4848250150680542,
      "learning_rate": 0.0001871172963369248,
      "loss": 0.2064,
      "step": 14600
    },
    {
      "epoch": 0.20535526038372337,
      "grad_norm": 0.3889874517917633,
      "learning_rate": 0.00018709446526925103,
      "loss": 0.2082,
      "step": 14610
    },
    {
      "epoch": 0.2054958183990442,
      "grad_norm": 0.3820645213127136,
      "learning_rate": 0.0001870716153839893,
      "loss": 0.2239,
      "step": 14620
    },
    {
      "epoch": 0.20563637641436502,
      "grad_norm": 0.4439220428466797,
      "learning_rate": 0.00018704874668607654,
      "loss": 0.2037,
      "step": 14630
    },
    {
      "epoch": 0.20577693442968586,
      "grad_norm": 0.47588831186294556,
      "learning_rate": 0.00018702585918045373,
      "loss": 0.2323,
      "step": 14640
    },
    {
      "epoch": 0.20591749244500668,
      "grad_norm": 0.5247348546981812,
      "learning_rate": 0.00018700295287206592,
      "loss": 0.1991,
      "step": 14650
    },
    {
      "epoch": 0.2060580504603275,
      "grad_norm": 0.4903160333633423,
      "learning_rate": 0.00018698002776586226,
      "loss": 0.2093,
      "step": 14660
    },
    {
      "epoch": 0.20619860847564833,
      "grad_norm": 0.43988537788391113,
      "learning_rate": 0.00018695708386679596,
      "loss": 0.2057,
      "step": 14670
    },
    {
      "epoch": 0.20633916649096914,
      "grad_norm": 0.5525961518287659,
      "learning_rate": 0.00018693412117982421,
      "loss": 0.2137,
      "step": 14680
    },
    {
      "epoch": 0.20647972450628996,
      "grad_norm": 0.4375472664833069,
      "learning_rate": 0.00018691113970990836,
      "loss": 0.1852,
      "step": 14690
    },
    {
      "epoch": 0.2066202825216108,
      "grad_norm": 0.3917044699192047,
      "learning_rate": 0.00018688813946201376,
      "loss": 0.2279,
      "step": 14700
    },
    {
      "epoch": 0.2067608405369316,
      "grad_norm": 0.40974360704421997,
      "learning_rate": 0.00018686512044110987,
      "loss": 0.2278,
      "step": 14710
    },
    {
      "epoch": 0.20690139855225245,
      "grad_norm": 0.352774441242218,
      "learning_rate": 0.0001868420826521701,
      "loss": 0.1905,
      "step": 14720
    },
    {
      "epoch": 0.20704195656757327,
      "grad_norm": 0.42764684557914734,
      "learning_rate": 0.000186819026100172,
      "loss": 0.1961,
      "step": 14730
    },
    {
      "epoch": 0.20718251458289408,
      "grad_norm": 0.39628422260284424,
      "learning_rate": 0.00018679595079009722,
      "loss": 0.1657,
      "step": 14740
    },
    {
      "epoch": 0.20732307259821492,
      "grad_norm": 0.44212454557418823,
      "learning_rate": 0.00018677285672693135,
      "loss": 0.2166,
      "step": 14750
    },
    {
      "epoch": 0.20746363061353573,
      "grad_norm": 0.391118586063385,
      "learning_rate": 0.00018674974391566407,
      "loss": 0.1821,
      "step": 14760
    },
    {
      "epoch": 0.20760418862885657,
      "grad_norm": 0.513839840888977,
      "learning_rate": 0.00018672661236128913,
      "loss": 0.2207,
      "step": 14770
    },
    {
      "epoch": 0.20774474664417739,
      "grad_norm": 0.3673436939716339,
      "learning_rate": 0.00018670346206880437,
      "loss": 0.2181,
      "step": 14780
    },
    {
      "epoch": 0.2078853046594982,
      "grad_norm": 0.5244107842445374,
      "learning_rate": 0.00018668029304321157,
      "loss": 0.2183,
      "step": 14790
    },
    {
      "epoch": 0.20802586267481904,
      "grad_norm": 0.41967636346817017,
      "learning_rate": 0.00018665710528951662,
      "loss": 0.2092,
      "step": 14800
    },
    {
      "epoch": 0.20816642069013985,
      "grad_norm": 0.4290749132633209,
      "learning_rate": 0.0001866338988127295,
      "loss": 0.1935,
      "step": 14810
    },
    {
      "epoch": 0.20830697870546067,
      "grad_norm": 0.4170161485671997,
      "learning_rate": 0.00018661067361786418,
      "loss": 0.1976,
      "step": 14820
    },
    {
      "epoch": 0.2084475367207815,
      "grad_norm": 0.574755847454071,
      "learning_rate": 0.0001865874297099387,
      "loss": 0.217,
      "step": 14830
    },
    {
      "epoch": 0.20858809473610232,
      "grad_norm": 0.4211540222167969,
      "learning_rate": 0.000186564167093975,
      "loss": 0.1967,
      "step": 14840
    },
    {
      "epoch": 0.20872865275142316,
      "grad_norm": 0.43254420161247253,
      "learning_rate": 0.00018654088577499937,
      "loss": 0.2107,
      "step": 14850
    },
    {
      "epoch": 0.20886921076674397,
      "grad_norm": 0.4957050383090973,
      "learning_rate": 0.00018651758575804187,
      "loss": 0.1828,
      "step": 14860
    },
    {
      "epoch": 0.2090097687820648,
      "grad_norm": 0.34140756726264954,
      "learning_rate": 0.00018649426704813672,
      "loss": 0.2137,
      "step": 14870
    },
    {
      "epoch": 0.20915032679738563,
      "grad_norm": 0.5176404118537903,
      "learning_rate": 0.00018647092965032208,
      "loss": 0.1977,
      "step": 14880
    },
    {
      "epoch": 0.20929088481270644,
      "grad_norm": 0.4792444109916687,
      "learning_rate": 0.00018644757356964025,
      "loss": 0.2029,
      "step": 14890
    },
    {
      "epoch": 0.20943144282802728,
      "grad_norm": 0.3671039640903473,
      "learning_rate": 0.0001864241988111376,
      "loss": 0.2126,
      "step": 14900
    },
    {
      "epoch": 0.2095720008433481,
      "grad_norm": 0.393860399723053,
      "learning_rate": 0.00018640080537986442,
      "loss": 0.198,
      "step": 14910
    },
    {
      "epoch": 0.2097125588586689,
      "grad_norm": 0.4475869834423065,
      "learning_rate": 0.00018637739328087506,
      "loss": 0.1871,
      "step": 14920
    },
    {
      "epoch": 0.20985311687398975,
      "grad_norm": 0.45307090878486633,
      "learning_rate": 0.00018635396251922797,
      "loss": 0.2012,
      "step": 14930
    },
    {
      "epoch": 0.20999367488931056,
      "grad_norm": 0.41060614585876465,
      "learning_rate": 0.00018633051309998555,
      "loss": 0.2149,
      "step": 14940
    },
    {
      "epoch": 0.21013423290463137,
      "grad_norm": 0.8010419607162476,
      "learning_rate": 0.0001863070450282143,
      "loss": 0.2274,
      "step": 14950
    },
    {
      "epoch": 0.21027479091995221,
      "grad_norm": 0.37445002794265747,
      "learning_rate": 0.00018628355830898466,
      "loss": 0.2408,
      "step": 14960
    },
    {
      "epoch": 0.21041534893527303,
      "grad_norm": 0.3989226520061493,
      "learning_rate": 0.00018626005294737125,
      "loss": 0.2151,
      "step": 14970
    },
    {
      "epoch": 0.21055590695059387,
      "grad_norm": 0.37011614441871643,
      "learning_rate": 0.00018623652894845258,
      "loss": 0.2236,
      "step": 14980
    },
    {
      "epoch": 0.21069646496591468,
      "grad_norm": 0.44250568747520447,
      "learning_rate": 0.00018621298631731124,
      "loss": 0.2028,
      "step": 14990
    },
    {
      "epoch": 0.2108370229812355,
      "grad_norm": 0.4633479416370392,
      "learning_rate": 0.00018618942505903383,
      "loss": 0.2012,
      "step": 15000
    },
    {
      "epoch": 0.2108370229812355,
      "eval_chrf": 82.82051205341394,
      "eval_loss": 0.4326775372028351,
      "eval_runtime": 257.9815,
      "eval_samples_per_second": 0.388,
      "eval_steps_per_second": 0.016,
      "step": 15000
    },
    {
      "epoch": 0.21097758099655634,
      "grad_norm": 0.4391336441040039,
      "learning_rate": 0.000186165845178711,
      "loss": 0.2285,
      "step": 15010
    },
    {
      "epoch": 0.21111813901187715,
      "grad_norm": 0.4783604145050049,
      "learning_rate": 0.0001861422466814374,
      "loss": 0.2139,
      "step": 15020
    },
    {
      "epoch": 0.211258697027198,
      "grad_norm": 0.3801127076148987,
      "learning_rate": 0.0001861186295723117,
      "loss": 0.2095,
      "step": 15030
    },
    {
      "epoch": 0.2113992550425188,
      "grad_norm": 0.43613746762275696,
      "learning_rate": 0.0001860949938564366,
      "loss": 0.2007,
      "step": 15040
    },
    {
      "epoch": 0.21153981305783962,
      "grad_norm": 0.39664286375045776,
      "learning_rate": 0.00018607133953891887,
      "loss": 0.2161,
      "step": 15050
    },
    {
      "epoch": 0.21168037107316046,
      "grad_norm": 0.47274744510650635,
      "learning_rate": 0.00018604766662486916,
      "loss": 0.1867,
      "step": 15060
    },
    {
      "epoch": 0.21182092908848127,
      "grad_norm": 0.45525887608528137,
      "learning_rate": 0.0001860239751194023,
      "loss": 0.227,
      "step": 15070
    },
    {
      "epoch": 0.21196148710380208,
      "grad_norm": 0.42642149329185486,
      "learning_rate": 0.00018600026502763707,
      "loss": 0.2166,
      "step": 15080
    },
    {
      "epoch": 0.21210204511912292,
      "grad_norm": 0.41591188311576843,
      "learning_rate": 0.00018597653635469622,
      "loss": 0.2226,
      "step": 15090
    },
    {
      "epoch": 0.21224260313444374,
      "grad_norm": 0.44854825735092163,
      "learning_rate": 0.00018595278910570658,
      "loss": 0.2025,
      "step": 15100
    },
    {
      "epoch": 0.21238316114976458,
      "grad_norm": 0.4680345058441162,
      "learning_rate": 0.00018592902328579895,
      "loss": 0.2058,
      "step": 15110
    },
    {
      "epoch": 0.2125237191650854,
      "grad_norm": 0.34579533338546753,
      "learning_rate": 0.00018590523890010818,
      "loss": 0.2152,
      "step": 15120
    },
    {
      "epoch": 0.2126642771804062,
      "grad_norm": 0.36790943145751953,
      "learning_rate": 0.0001858814359537731,
      "loss": 0.2128,
      "step": 15130
    },
    {
      "epoch": 0.21280483519572704,
      "grad_norm": 0.4481298327445984,
      "learning_rate": 0.00018585761445193656,
      "loss": 0.2156,
      "step": 15140
    },
    {
      "epoch": 0.21294539321104786,
      "grad_norm": 0.562809944152832,
      "learning_rate": 0.00018583377439974544,
      "loss": 0.2246,
      "step": 15150
    },
    {
      "epoch": 0.2130859512263687,
      "grad_norm": 0.5986753702163696,
      "learning_rate": 0.0001858099158023506,
      "loss": 0.2006,
      "step": 15160
    },
    {
      "epoch": 0.2132265092416895,
      "grad_norm": 0.4528370201587677,
      "learning_rate": 0.0001857860386649069,
      "loss": 0.1915,
      "step": 15170
    },
    {
      "epoch": 0.21336706725701032,
      "grad_norm": 0.5109283328056335,
      "learning_rate": 0.00018576214299257325,
      "loss": 0.2116,
      "step": 15180
    },
    {
      "epoch": 0.21350762527233116,
      "grad_norm": 0.36227118968963623,
      "learning_rate": 0.00018573822879051254,
      "loss": 0.1796,
      "step": 15190
    },
    {
      "epoch": 0.21364818328765198,
      "grad_norm": 0.5447039008140564,
      "learning_rate": 0.00018571429606389166,
      "loss": 0.1986,
      "step": 15200
    },
    {
      "epoch": 0.2137887413029728,
      "grad_norm": 0.5264976024627686,
      "learning_rate": 0.00018569034481788146,
      "loss": 0.2082,
      "step": 15210
    },
    {
      "epoch": 0.21392929931829363,
      "grad_norm": 0.48652032017707825,
      "learning_rate": 0.00018566637505765688,
      "loss": 0.199,
      "step": 15220
    },
    {
      "epoch": 0.21406985733361444,
      "grad_norm": 0.6124891638755798,
      "learning_rate": 0.0001856423867883968,
      "loss": 0.2012,
      "step": 15230
    },
    {
      "epoch": 0.21421041534893528,
      "grad_norm": 0.5273460745811462,
      "learning_rate": 0.0001856183800152841,
      "loss": 0.2263,
      "step": 15240
    },
    {
      "epoch": 0.2143509733642561,
      "grad_norm": 0.4930509030818939,
      "learning_rate": 0.00018559435474350573,
      "loss": 0.207,
      "step": 15250
    },
    {
      "epoch": 0.2144915313795769,
      "grad_norm": 0.46386903524398804,
      "learning_rate": 0.0001855703109782525,
      "loss": 0.2058,
      "step": 15260
    },
    {
      "epoch": 0.21463208939489775,
      "grad_norm": 0.45673784613609314,
      "learning_rate": 0.00018554624872471934,
      "loss": 0.1887,
      "step": 15270
    },
    {
      "epoch": 0.21477264741021856,
      "grad_norm": 0.46073588728904724,
      "learning_rate": 0.00018552216798810515,
      "loss": 0.2107,
      "step": 15280
    },
    {
      "epoch": 0.21491320542553938,
      "grad_norm": 0.4474230110645294,
      "learning_rate": 0.0001854980687736127,
      "loss": 0.2315,
      "step": 15290
    },
    {
      "epoch": 0.21505376344086022,
      "grad_norm": 0.38105547428131104,
      "learning_rate": 0.00018547395108644898,
      "loss": 0.2084,
      "step": 15300
    },
    {
      "epoch": 0.21519432145618103,
      "grad_norm": 0.4035305380821228,
      "learning_rate": 0.00018544981493182481,
      "loss": 0.2395,
      "step": 15310
    },
    {
      "epoch": 0.21533487947150187,
      "grad_norm": 0.532655656337738,
      "learning_rate": 0.00018542566031495497,
      "loss": 0.2164,
      "step": 15320
    },
    {
      "epoch": 0.21547543748682269,
      "grad_norm": 0.4325355589389801,
      "learning_rate": 0.00018540148724105834,
      "loss": 0.222,
      "step": 15330
    },
    {
      "epoch": 0.2156159955021435,
      "grad_norm": 0.5806542038917542,
      "learning_rate": 0.00018537729571535777,
      "loss": 0.1897,
      "step": 15340
    },
    {
      "epoch": 0.21575655351746434,
      "grad_norm": 0.437415212392807,
      "learning_rate": 0.00018535308574308003,
      "loss": 0.2089,
      "step": 15350
    },
    {
      "epoch": 0.21589711153278515,
      "grad_norm": 0.460148423910141,
      "learning_rate": 0.00018532885732945595,
      "loss": 0.2275,
      "step": 15360
    },
    {
      "epoch": 0.216037669548106,
      "grad_norm": 0.47994258999824524,
      "learning_rate": 0.00018530461047972026,
      "loss": 0.2063,
      "step": 15370
    },
    {
      "epoch": 0.2161782275634268,
      "grad_norm": 0.4585625231266022,
      "learning_rate": 0.00018528034519911177,
      "loss": 0.2147,
      "step": 15380
    },
    {
      "epoch": 0.21631878557874762,
      "grad_norm": 0.42683348059654236,
      "learning_rate": 0.0001852560614928732,
      "loss": 0.2218,
      "step": 15390
    },
    {
      "epoch": 0.21645934359406846,
      "grad_norm": 0.3197411000728607,
      "learning_rate": 0.00018523175936625128,
      "loss": 0.231,
      "step": 15400
    },
    {
      "epoch": 0.21659990160938927,
      "grad_norm": 0.4453927278518677,
      "learning_rate": 0.0001852074388244967,
      "loss": 0.2244,
      "step": 15410
    },
    {
      "epoch": 0.21674045962471009,
      "grad_norm": 0.47477197647094727,
      "learning_rate": 0.0001851830998728642,
      "loss": 0.2029,
      "step": 15420
    },
    {
      "epoch": 0.21688101764003093,
      "grad_norm": 0.3015683591365814,
      "learning_rate": 0.0001851587425166124,
      "loss": 0.1895,
      "step": 15430
    },
    {
      "epoch": 0.21702157565535174,
      "grad_norm": 0.3970557451248169,
      "learning_rate": 0.0001851343667610039,
      "loss": 0.2238,
      "step": 15440
    },
    {
      "epoch": 0.21716213367067258,
      "grad_norm": 0.48441988229751587,
      "learning_rate": 0.00018510997261130538,
      "loss": 0.2008,
      "step": 15450
    },
    {
      "epoch": 0.2173026916859934,
      "grad_norm": 0.6000593304634094,
      "learning_rate": 0.00018508556007278744,
      "loss": 0.1975,
      "step": 15460
    },
    {
      "epoch": 0.2174432497013142,
      "grad_norm": 0.320560097694397,
      "learning_rate": 0.00018506112915072455,
      "loss": 0.1964,
      "step": 15470
    },
    {
      "epoch": 0.21758380771663505,
      "grad_norm": 0.5494293570518494,
      "learning_rate": 0.00018503667985039532,
      "loss": 0.2143,
      "step": 15480
    },
    {
      "epoch": 0.21772436573195586,
      "grad_norm": 0.40263715386390686,
      "learning_rate": 0.00018501221217708224,
      "loss": 0.2493,
      "step": 15490
    },
    {
      "epoch": 0.2178649237472767,
      "grad_norm": 0.5201196670532227,
      "learning_rate": 0.00018498772613607177,
      "loss": 0.2009,
      "step": 15500
    },
    {
      "epoch": 0.2180054817625975,
      "grad_norm": 0.5006442070007324,
      "learning_rate": 0.00018496322173265436,
      "loss": 0.2501,
      "step": 15510
    },
    {
      "epoch": 0.21814603977791833,
      "grad_norm": 0.46168252825737,
      "learning_rate": 0.00018493869897212443,
      "loss": 0.2442,
      "step": 15520
    },
    {
      "epoch": 0.21828659779323917,
      "grad_norm": 0.43364977836608887,
      "learning_rate": 0.0001849141578597803,
      "loss": 0.2114,
      "step": 15530
    },
    {
      "epoch": 0.21842715580855998,
      "grad_norm": 0.4662631154060364,
      "learning_rate": 0.00018488959840092442,
      "loss": 0.2014,
      "step": 15540
    },
    {
      "epoch": 0.2185677138238808,
      "grad_norm": 0.44242143630981445,
      "learning_rate": 0.000184865020600863,
      "loss": 0.2131,
      "step": 15550
    },
    {
      "epoch": 0.21870827183920163,
      "grad_norm": 0.6606244444847107,
      "learning_rate": 0.00018484042446490634,
      "loss": 0.2399,
      "step": 15560
    },
    {
      "epoch": 0.21884882985452245,
      "grad_norm": 0.5681745409965515,
      "learning_rate": 0.00018481580999836866,
      "loss": 0.2157,
      "step": 15570
    },
    {
      "epoch": 0.2189893878698433,
      "grad_norm": 0.45000481605529785,
      "learning_rate": 0.00018479117720656814,
      "loss": 0.2016,
      "step": 15580
    },
    {
      "epoch": 0.2191299458851641,
      "grad_norm": 0.4646105170249939,
      "learning_rate": 0.00018476652609482694,
      "loss": 0.2096,
      "step": 15590
    },
    {
      "epoch": 0.21927050390048491,
      "grad_norm": 0.38280490040779114,
      "learning_rate": 0.00018474185666847123,
      "loss": 0.197,
      "step": 15600
    },
    {
      "epoch": 0.21941106191580576,
      "grad_norm": 0.35731178522109985,
      "learning_rate": 0.00018471716893283095,
      "loss": 0.1794,
      "step": 15610
    },
    {
      "epoch": 0.21955161993112657,
      "grad_norm": 0.4666403830051422,
      "learning_rate": 0.00018469246289324023,
      "loss": 0.2042,
      "step": 15620
    },
    {
      "epoch": 0.2196921779464474,
      "grad_norm": 0.41200706362724304,
      "learning_rate": 0.00018466773855503702,
      "loss": 0.1793,
      "step": 15630
    },
    {
      "epoch": 0.21983273596176822,
      "grad_norm": 0.41489627957344055,
      "learning_rate": 0.00018464299592356318,
      "loss": 0.2233,
      "step": 15640
    },
    {
      "epoch": 0.21997329397708903,
      "grad_norm": 0.46858951449394226,
      "learning_rate": 0.00018461823500416467,
      "loss": 0.2311,
      "step": 15650
    },
    {
      "epoch": 0.22011385199240988,
      "grad_norm": 0.3232119381427765,
      "learning_rate": 0.00018459345580219128,
      "loss": 0.1869,
      "step": 15660
    },
    {
      "epoch": 0.2202544100077307,
      "grad_norm": 0.48415064811706543,
      "learning_rate": 0.0001845686583229968,
      "loss": 0.224,
      "step": 15670
    },
    {
      "epoch": 0.2203949680230515,
      "grad_norm": 0.35237863659858704,
      "learning_rate": 0.00018454384257193896,
      "loss": 0.2156,
      "step": 15680
    },
    {
      "epoch": 0.22053552603837234,
      "grad_norm": 0.4336957335472107,
      "learning_rate": 0.0001845190085543795,
      "loss": 0.2382,
      "step": 15690
    },
    {
      "epoch": 0.22067608405369316,
      "grad_norm": 0.6433353424072266,
      "learning_rate": 0.00018449415627568393,
      "loss": 0.2264,
      "step": 15700
    },
    {
      "epoch": 0.220816642069014,
      "grad_norm": 0.5027011036872864,
      "learning_rate": 0.00018446928574122193,
      "loss": 0.2008,
      "step": 15710
    },
    {
      "epoch": 0.2209572000843348,
      "grad_norm": 0.3565092384815216,
      "learning_rate": 0.00018444439695636694,
      "loss": 0.2275,
      "step": 15720
    },
    {
      "epoch": 0.22109775809965562,
      "grad_norm": 0.4112890362739563,
      "learning_rate": 0.00018441948992649646,
      "loss": 0.2175,
      "step": 15730
    },
    {
      "epoch": 0.22123831611497646,
      "grad_norm": 0.48462048172950745,
      "learning_rate": 0.00018439456465699187,
      "loss": 0.2027,
      "step": 15740
    },
    {
      "epoch": 0.22137887413029728,
      "grad_norm": 0.49463990330696106,
      "learning_rate": 0.00018436962115323855,
      "loss": 0.2039,
      "step": 15750
    },
    {
      "epoch": 0.22151943214561812,
      "grad_norm": 0.33819833397865295,
      "learning_rate": 0.00018434465942062574,
      "loss": 0.21,
      "step": 15760
    },
    {
      "epoch": 0.22165999016093893,
      "grad_norm": 0.3969721794128418,
      "learning_rate": 0.00018431967946454665,
      "loss": 0.2409,
      "step": 15770
    },
    {
      "epoch": 0.22180054817625974,
      "grad_norm": 0.37428468465805054,
      "learning_rate": 0.00018429468129039847,
      "loss": 0.2053,
      "step": 15780
    },
    {
      "epoch": 0.22194110619158058,
      "grad_norm": 0.48097750544548035,
      "learning_rate": 0.0001842696649035823,
      "loss": 0.2133,
      "step": 15790
    },
    {
      "epoch": 0.2220816642069014,
      "grad_norm": 0.4189276695251465,
      "learning_rate": 0.00018424463030950312,
      "loss": 0.2064,
      "step": 15800
    },
    {
      "epoch": 0.2222222222222222,
      "grad_norm": 0.5002239942550659,
      "learning_rate": 0.00018421957751356993,
      "loss": 0.2125,
      "step": 15810
    },
    {
      "epoch": 0.22236278023754305,
      "grad_norm": 0.4362868070602417,
      "learning_rate": 0.00018419450652119564,
      "loss": 0.2095,
      "step": 15820
    },
    {
      "epoch": 0.22250333825286386,
      "grad_norm": 0.5985012054443359,
      "learning_rate": 0.00018416941733779703,
      "loss": 0.2297,
      "step": 15830
    },
    {
      "epoch": 0.2226438962681847,
      "grad_norm": 0.4958445727825165,
      "learning_rate": 0.00018414430996879487,
      "loss": 0.1994,
      "step": 15840
    },
    {
      "epoch": 0.22278445428350552,
      "grad_norm": 0.5681561231613159,
      "learning_rate": 0.0001841191844196139,
      "loss": 0.2125,
      "step": 15850
    },
    {
      "epoch": 0.22292501229882633,
      "grad_norm": 0.4579131305217743,
      "learning_rate": 0.00018409404069568265,
      "loss": 0.1977,
      "step": 15860
    },
    {
      "epoch": 0.22306557031414717,
      "grad_norm": 0.46691954135894775,
      "learning_rate": 0.00018406887880243372,
      "loss": 0.1974,
      "step": 15870
    },
    {
      "epoch": 0.22320612832946798,
      "grad_norm": 0.5419637560844421,
      "learning_rate": 0.0001840436987453036,
      "loss": 0.2065,
      "step": 15880
    },
    {
      "epoch": 0.22334668634478883,
      "grad_norm": 0.4614500105381012,
      "learning_rate": 0.00018401850052973258,
      "loss": 0.2051,
      "step": 15890
    },
    {
      "epoch": 0.22348724436010964,
      "grad_norm": 0.44594377279281616,
      "learning_rate": 0.00018399328416116503,
      "loss": 0.2126,
      "step": 15900
    },
    {
      "epoch": 0.22362780237543045,
      "grad_norm": 0.47176501154899597,
      "learning_rate": 0.00018396804964504923,
      "loss": 0.2117,
      "step": 15910
    },
    {
      "epoch": 0.2237683603907513,
      "grad_norm": 0.4225786030292511,
      "learning_rate": 0.00018394279698683727,
      "loss": 0.1908,
      "step": 15920
    },
    {
      "epoch": 0.2239089184060721,
      "grad_norm": 0.38729235529899597,
      "learning_rate": 0.00018391752619198527,
      "loss": 0.2097,
      "step": 15930
    },
    {
      "epoch": 0.22404947642139292,
      "grad_norm": 0.5475749373435974,
      "learning_rate": 0.0001838922372659532,
      "loss": 0.2019,
      "step": 15940
    },
    {
      "epoch": 0.22419003443671376,
      "grad_norm": 0.5189427137374878,
      "learning_rate": 0.000183866930214205,
      "loss": 0.1956,
      "step": 15950
    },
    {
      "epoch": 0.22433059245203457,
      "grad_norm": 0.4810304045677185,
      "learning_rate": 0.00018384160504220846,
      "loss": 0.1958,
      "step": 15960
    },
    {
      "epoch": 0.2244711504673554,
      "grad_norm": 0.5092229843139648,
      "learning_rate": 0.00018381626175543536,
      "loss": 0.1996,
      "step": 15970
    },
    {
      "epoch": 0.22461170848267623,
      "grad_norm": 0.36356088519096375,
      "learning_rate": 0.00018379090035936134,
      "loss": 0.1989,
      "step": 15980
    },
    {
      "epoch": 0.22475226649799704,
      "grad_norm": 0.34988951683044434,
      "learning_rate": 0.00018376552085946598,
      "loss": 0.1811,
      "step": 15990
    },
    {
      "epoch": 0.22489282451331788,
      "grad_norm": 0.611853301525116,
      "learning_rate": 0.00018374012326123276,
      "loss": 0.2104,
      "step": 16000
    },
    {
      "epoch": 0.22489282451331788,
      "eval_chrf": 82.88572916646845,
      "eval_loss": 0.4360054135322571,
      "eval_runtime": 240.7154,
      "eval_samples_per_second": 0.415,
      "eval_steps_per_second": 0.017,
      "step": 16000
    },
    {
      "epoch": 0.2250333825286387,
      "grad_norm": 0.4638526737689972,
      "learning_rate": 0.0001837147075701491,
      "loss": 0.2037,
      "step": 16010
    },
    {
      "epoch": 0.2251739405439595,
      "grad_norm": 0.5316300392150879,
      "learning_rate": 0.00018368927379170622,
      "loss": 0.1983,
      "step": 16020
    },
    {
      "epoch": 0.22531449855928035,
      "grad_norm": 0.4769590198993683,
      "learning_rate": 0.00018366382193139939,
      "loss": 0.2131,
      "step": 16030
    },
    {
      "epoch": 0.22545505657460116,
      "grad_norm": 0.39321017265319824,
      "learning_rate": 0.00018363835199472774,
      "loss": 0.2069,
      "step": 16040
    },
    {
      "epoch": 0.225595614589922,
      "grad_norm": 0.4696575999259949,
      "learning_rate": 0.00018361286398719428,
      "loss": 0.2193,
      "step": 16050
    },
    {
      "epoch": 0.2257361726052428,
      "grad_norm": 0.49544069170951843,
      "learning_rate": 0.00018358735791430592,
      "loss": 0.2088,
      "step": 16060
    },
    {
      "epoch": 0.22587673062056363,
      "grad_norm": 0.47814834117889404,
      "learning_rate": 0.0001835618337815735,
      "loss": 0.2276,
      "step": 16070
    },
    {
      "epoch": 0.22601728863588447,
      "grad_norm": 0.4744305908679962,
      "learning_rate": 0.00018353629159451176,
      "loss": 0.2162,
      "step": 16080
    },
    {
      "epoch": 0.22615784665120528,
      "grad_norm": 0.455851286649704,
      "learning_rate": 0.00018351073135863927,
      "loss": 0.2195,
      "step": 16090
    },
    {
      "epoch": 0.22629840466652612,
      "grad_norm": 0.64469313621521,
      "learning_rate": 0.00018348515307947867,
      "loss": 0.2198,
      "step": 16100
    },
    {
      "epoch": 0.22643896268184693,
      "grad_norm": 0.43515893816947937,
      "learning_rate": 0.00018345955676255633,
      "loss": 0.2141,
      "step": 16110
    },
    {
      "epoch": 0.22657952069716775,
      "grad_norm": 0.45102620124816895,
      "learning_rate": 0.00018343394241340257,
      "loss": 0.21,
      "step": 16120
    },
    {
      "epoch": 0.2267200787124886,
      "grad_norm": 0.5091341733932495,
      "learning_rate": 0.00018340831003755162,
      "loss": 0.2125,
      "step": 16130
    },
    {
      "epoch": 0.2268606367278094,
      "grad_norm": 0.3054172992706299,
      "learning_rate": 0.00018338265964054163,
      "loss": 0.1855,
      "step": 16140
    },
    {
      "epoch": 0.2270011947431302,
      "grad_norm": 0.4756179451942444,
      "learning_rate": 0.00018335699122791455,
      "loss": 0.2135,
      "step": 16150
    },
    {
      "epoch": 0.22714175275845105,
      "grad_norm": 0.40579459071159363,
      "learning_rate": 0.00018333130480521637,
      "loss": 0.2245,
      "step": 16160
    },
    {
      "epoch": 0.22728231077377187,
      "grad_norm": 0.46695980429649353,
      "learning_rate": 0.0001833056003779968,
      "loss": 0.2166,
      "step": 16170
    },
    {
      "epoch": 0.2274228687890927,
      "grad_norm": 0.32243990898132324,
      "learning_rate": 0.0001832798779518096,
      "loss": 0.1676,
      "step": 16180
    },
    {
      "epoch": 0.22756342680441352,
      "grad_norm": 0.45454373955726624,
      "learning_rate": 0.00018325413753221228,
      "loss": 0.2095,
      "step": 16190
    },
    {
      "epoch": 0.22770398481973433,
      "grad_norm": 0.43410179018974304,
      "learning_rate": 0.00018322837912476635,
      "loss": 0.1961,
      "step": 16200
    },
    {
      "epoch": 0.22784454283505517,
      "grad_norm": 0.44975170493125916,
      "learning_rate": 0.00018320260273503713,
      "loss": 0.2127,
      "step": 16210
    },
    {
      "epoch": 0.227985100850376,
      "grad_norm": 0.49282968044281006,
      "learning_rate": 0.00018317680836859386,
      "loss": 0.2208,
      "step": 16220
    },
    {
      "epoch": 0.22812565886569683,
      "grad_norm": 0.46582046151161194,
      "learning_rate": 0.00018315099603100968,
      "loss": 0.227,
      "step": 16230
    },
    {
      "epoch": 0.22826621688101764,
      "grad_norm": 0.42106932401657104,
      "learning_rate": 0.00018312516572786158,
      "loss": 0.2068,
      "step": 16240
    },
    {
      "epoch": 0.22840677489633845,
      "grad_norm": 0.43135741353034973,
      "learning_rate": 0.0001830993174647304,
      "loss": 0.2116,
      "step": 16250
    },
    {
      "epoch": 0.2285473329116593,
      "grad_norm": 0.603646993637085,
      "learning_rate": 0.000183073451247201,
      "loss": 0.2234,
      "step": 16260
    },
    {
      "epoch": 0.2286878909269801,
      "grad_norm": 0.40896886587142944,
      "learning_rate": 0.00018304756708086193,
      "loss": 0.1962,
      "step": 16270
    },
    {
      "epoch": 0.22882844894230092,
      "grad_norm": 0.5486946702003479,
      "learning_rate": 0.00018302166497130576,
      "loss": 0.2184,
      "step": 16280
    },
    {
      "epoch": 0.22896900695762176,
      "grad_norm": 0.4316113591194153,
      "learning_rate": 0.00018299574492412884,
      "loss": 0.1846,
      "step": 16290
    },
    {
      "epoch": 0.22910956497294258,
      "grad_norm": 0.6302161812782288,
      "learning_rate": 0.0001829698069449315,
      "loss": 0.2188,
      "step": 16300
    },
    {
      "epoch": 0.22925012298826342,
      "grad_norm": 0.4198704957962036,
      "learning_rate": 0.00018294385103931785,
      "loss": 0.2137,
      "step": 16310
    },
    {
      "epoch": 0.22939068100358423,
      "grad_norm": 0.41124218702316284,
      "learning_rate": 0.00018291787721289596,
      "loss": 0.2481,
      "step": 16320
    },
    {
      "epoch": 0.22953123901890504,
      "grad_norm": 0.4199219048023224,
      "learning_rate": 0.00018289188547127765,
      "loss": 0.2094,
      "step": 16330
    },
    {
      "epoch": 0.22967179703422588,
      "grad_norm": 0.48190534114837646,
      "learning_rate": 0.0001828658758200787,
      "loss": 0.2291,
      "step": 16340
    },
    {
      "epoch": 0.2298123550495467,
      "grad_norm": 0.424686998128891,
      "learning_rate": 0.00018283984826491877,
      "loss": 0.2012,
      "step": 16350
    },
    {
      "epoch": 0.22995291306486754,
      "grad_norm": 0.40577951073646545,
      "learning_rate": 0.00018281380281142139,
      "loss": 0.2089,
      "step": 16360
    },
    {
      "epoch": 0.23009347108018835,
      "grad_norm": 0.5267784595489502,
      "learning_rate": 0.00018278773946521387,
      "loss": 0.2284,
      "step": 16370
    },
    {
      "epoch": 0.23023402909550916,
      "grad_norm": 0.5176638960838318,
      "learning_rate": 0.00018276165823192744,
      "loss": 0.171,
      "step": 16380
    },
    {
      "epoch": 0.23037458711083,
      "grad_norm": 0.378432959318161,
      "learning_rate": 0.00018273555911719723,
      "loss": 0.192,
      "step": 16390
    },
    {
      "epoch": 0.23051514512615082,
      "grad_norm": 0.5014709234237671,
      "learning_rate": 0.00018270944212666221,
      "loss": 0.2117,
      "step": 16400
    },
    {
      "epoch": 0.23065570314147163,
      "grad_norm": 0.4741896390914917,
      "learning_rate": 0.0001826833072659652,
      "loss": 0.2264,
      "step": 16410
    },
    {
      "epoch": 0.23079626115679247,
      "grad_norm": 0.45438021421432495,
      "learning_rate": 0.00018265715454075283,
      "loss": 0.2268,
      "step": 16420
    },
    {
      "epoch": 0.23093681917211328,
      "grad_norm": 0.49692875146865845,
      "learning_rate": 0.00018263098395667572,
      "loss": 0.2201,
      "step": 16430
    },
    {
      "epoch": 0.23107737718743412,
      "grad_norm": 0.5172484517097473,
      "learning_rate": 0.00018260479551938824,
      "loss": 0.2068,
      "step": 16440
    },
    {
      "epoch": 0.23121793520275494,
      "grad_norm": 0.4847676455974579,
      "learning_rate": 0.00018257858923454863,
      "loss": 0.194,
      "step": 16450
    },
    {
      "epoch": 0.23135849321807575,
      "grad_norm": 0.4178107976913452,
      "learning_rate": 0.00018255236510781906,
      "loss": 0.2018,
      "step": 16460
    },
    {
      "epoch": 0.2314990512333966,
      "grad_norm": 0.43576788902282715,
      "learning_rate": 0.00018252612314486545,
      "loss": 0.175,
      "step": 16470
    },
    {
      "epoch": 0.2316396092487174,
      "grad_norm": 0.43119123578071594,
      "learning_rate": 0.00018249986335135767,
      "loss": 0.1909,
      "step": 16480
    },
    {
      "epoch": 0.23178016726403824,
      "grad_norm": 0.5131032466888428,
      "learning_rate": 0.00018247358573296934,
      "loss": 0.2193,
      "step": 16490
    },
    {
      "epoch": 0.23192072527935906,
      "grad_norm": 0.411156564950943,
      "learning_rate": 0.0001824472902953781,
      "loss": 0.2003,
      "step": 16500
    },
    {
      "epoch": 0.23206128329467987,
      "grad_norm": 0.45583435893058777,
      "learning_rate": 0.00018242097704426516,
      "loss": 0.2122,
      "step": 16510
    },
    {
      "epoch": 0.2322018413100007,
      "grad_norm": 0.3981030285358429,
      "learning_rate": 0.00018239464598531591,
      "loss": 0.2043,
      "step": 16520
    },
    {
      "epoch": 0.23234239932532152,
      "grad_norm": 0.6014482975006104,
      "learning_rate": 0.00018236829712421933,
      "loss": 0.208,
      "step": 16530
    },
    {
      "epoch": 0.23248295734064234,
      "grad_norm": 0.45501312613487244,
      "learning_rate": 0.00018234193046666838,
      "loss": 0.211,
      "step": 16540
    },
    {
      "epoch": 0.23262351535596318,
      "grad_norm": 0.4550839066505432,
      "learning_rate": 0.00018231554601835982,
      "loss": 0.2037,
      "step": 16550
    },
    {
      "epoch": 0.232764073371284,
      "grad_norm": 0.44017985463142395,
      "learning_rate": 0.00018228914378499428,
      "loss": 0.201,
      "step": 16560
    },
    {
      "epoch": 0.23290463138660483,
      "grad_norm": 0.5057845115661621,
      "learning_rate": 0.00018226272377227617,
      "loss": 0.1819,
      "step": 16570
    },
    {
      "epoch": 0.23304518940192565,
      "grad_norm": 0.3775175213813782,
      "learning_rate": 0.00018223628598591383,
      "loss": 0.2052,
      "step": 16580
    },
    {
      "epoch": 0.23318574741724646,
      "grad_norm": 0.37587618827819824,
      "learning_rate": 0.00018220983043161937,
      "loss": 0.201,
      "step": 16590
    },
    {
      "epoch": 0.2333263054325673,
      "grad_norm": 0.4216788411140442,
      "learning_rate": 0.00018218335711510874,
      "loss": 0.2156,
      "step": 16600
    },
    {
      "epoch": 0.2334668634478881,
      "grad_norm": 0.4030705392360687,
      "learning_rate": 0.0001821568660421018,
      "loss": 0.2056,
      "step": 16610
    },
    {
      "epoch": 0.23360742146320895,
      "grad_norm": 0.346451073884964,
      "learning_rate": 0.0001821303572183222,
      "loss": 0.2017,
      "step": 16620
    },
    {
      "epoch": 0.23374797947852977,
      "grad_norm": 0.49614450335502625,
      "learning_rate": 0.00018210383064949737,
      "loss": 0.2038,
      "step": 16630
    },
    {
      "epoch": 0.23388853749385058,
      "grad_norm": 0.45596131682395935,
      "learning_rate": 0.0001820772863413587,
      "loss": 0.219,
      "step": 16640
    },
    {
      "epoch": 0.23402909550917142,
      "grad_norm": 0.49397969245910645,
      "learning_rate": 0.00018205072429964124,
      "loss": 0.2031,
      "step": 16650
    },
    {
      "epoch": 0.23416965352449223,
      "grad_norm": 0.43771892786026,
      "learning_rate": 0.00018202414453008409,
      "loss": 0.1984,
      "step": 16660
    },
    {
      "epoch": 0.23431021153981305,
      "grad_norm": 0.39425256848335266,
      "learning_rate": 0.00018199754703843,
      "loss": 0.2244,
      "step": 16670
    },
    {
      "epoch": 0.2344507695551339,
      "grad_norm": 0.5792267322540283,
      "learning_rate": 0.0001819709318304256,
      "loss": 0.2039,
      "step": 16680
    },
    {
      "epoch": 0.2345913275704547,
      "grad_norm": 0.417339026927948,
      "learning_rate": 0.00018194429891182137,
      "loss": 0.182,
      "step": 16690
    },
    {
      "epoch": 0.23473188558577554,
      "grad_norm": 0.5354083776473999,
      "learning_rate": 0.00018191764828837158,
      "loss": 0.2253,
      "step": 16700
    },
    {
      "epoch": 0.23487244360109635,
      "grad_norm": 0.4503895044326782,
      "learning_rate": 0.00018189097996583443,
      "loss": 0.2231,
      "step": 16710
    },
    {
      "epoch": 0.23501300161641717,
      "grad_norm": 0.5055163502693176,
      "learning_rate": 0.00018186429394997182,
      "loss": 0.2395,
      "step": 16720
    },
    {
      "epoch": 0.235153559631738,
      "grad_norm": 0.4840635359287262,
      "learning_rate": 0.00018183759024654949,
      "loss": 0.2135,
      "step": 16730
    },
    {
      "epoch": 0.23529411764705882,
      "grad_norm": 0.3905552625656128,
      "learning_rate": 0.00018181086886133704,
      "loss": 0.2024,
      "step": 16740
    },
    {
      "epoch": 0.23543467566237966,
      "grad_norm": 0.36550042033195496,
      "learning_rate": 0.00018178412980010794,
      "loss": 0.2198,
      "step": 16750
    },
    {
      "epoch": 0.23557523367770047,
      "grad_norm": 0.43158572912216187,
      "learning_rate": 0.00018175737306863932,
      "loss": 0.1993,
      "step": 16760
    },
    {
      "epoch": 0.2357157916930213,
      "grad_norm": 0.3679793179035187,
      "learning_rate": 0.00018173059867271233,
      "loss": 0.2072,
      "step": 16770
    },
    {
      "epoch": 0.23585634970834213,
      "grad_norm": 0.5061398148536682,
      "learning_rate": 0.00018170380661811173,
      "loss": 0.202,
      "step": 16780
    },
    {
      "epoch": 0.23599690772366294,
      "grad_norm": 0.5475767850875854,
      "learning_rate": 0.00018167699691062626,
      "loss": 0.2039,
      "step": 16790
    },
    {
      "epoch": 0.23613746573898375,
      "grad_norm": 0.6202900409698486,
      "learning_rate": 0.00018165016955604843,
      "loss": 0.2007,
      "step": 16800
    },
    {
      "epoch": 0.2362780237543046,
      "grad_norm": 0.4951488971710205,
      "learning_rate": 0.0001816233245601745,
      "loss": 0.2619,
      "step": 16810
    },
    {
      "epoch": 0.2364185817696254,
      "grad_norm": 0.4565191864967346,
      "learning_rate": 0.0001815964619288046,
      "loss": 0.2022,
      "step": 16820
    },
    {
      "epoch": 0.23655913978494625,
      "grad_norm": 0.3810023069381714,
      "learning_rate": 0.00018156958166774267,
      "loss": 0.1996,
      "step": 16830
    },
    {
      "epoch": 0.23669969780026706,
      "grad_norm": 0.5086340308189392,
      "learning_rate": 0.00018154268378279645,
      "loss": 0.2292,
      "step": 16840
    },
    {
      "epoch": 0.23684025581558787,
      "grad_norm": 0.4447498321533203,
      "learning_rate": 0.00018151576827977743,
      "loss": 0.2026,
      "step": 16850
    },
    {
      "epoch": 0.23698081383090872,
      "grad_norm": 0.44191133975982666,
      "learning_rate": 0.00018148883516450103,
      "loss": 0.2195,
      "step": 16860
    },
    {
      "epoch": 0.23712137184622953,
      "grad_norm": 0.5270133018493652,
      "learning_rate": 0.00018146188444278635,
      "loss": 0.216,
      "step": 16870
    },
    {
      "epoch": 0.23726192986155034,
      "grad_norm": 0.5399002432823181,
      "learning_rate": 0.0001814349161204564,
      "loss": 0.1924,
      "step": 16880
    },
    {
      "epoch": 0.23740248787687118,
      "grad_norm": 0.46844422817230225,
      "learning_rate": 0.00018140793020333793,
      "loss": 0.2233,
      "step": 16890
    },
    {
      "epoch": 0.237543045892192,
      "grad_norm": 0.4430544376373291,
      "learning_rate": 0.0001813809266972615,
      "loss": 0.2369,
      "step": 16900
    },
    {
      "epoch": 0.23768360390751284,
      "grad_norm": 0.4523754119873047,
      "learning_rate": 0.00018135390560806142,
      "loss": 0.1982,
      "step": 16910
    },
    {
      "epoch": 0.23782416192283365,
      "grad_norm": 0.5143480896949768,
      "learning_rate": 0.00018132686694157592,
      "loss": 0.2017,
      "step": 16920
    },
    {
      "epoch": 0.23796471993815446,
      "grad_norm": 0.4497261643409729,
      "learning_rate": 0.00018129981070364698,
      "loss": 0.208,
      "step": 16930
    },
    {
      "epoch": 0.2381052779534753,
      "grad_norm": 0.5222679972648621,
      "learning_rate": 0.00018127273690012028,
      "loss": 0.1978,
      "step": 16940
    },
    {
      "epoch": 0.23824583596879612,
      "grad_norm": 0.5843683481216431,
      "learning_rate": 0.00018124564553684543,
      "loss": 0.2123,
      "step": 16950
    },
    {
      "epoch": 0.23838639398411696,
      "grad_norm": 0.5215798616409302,
      "learning_rate": 0.00018121853661967578,
      "loss": 0.2211,
      "step": 16960
    },
    {
      "epoch": 0.23852695199943777,
      "grad_norm": 0.3762476444244385,
      "learning_rate": 0.00018119141015446842,
      "loss": 0.2345,
      "step": 16970
    },
    {
      "epoch": 0.23866751001475858,
      "grad_norm": 0.4703845977783203,
      "learning_rate": 0.0001811642661470843,
      "loss": 0.1882,
      "step": 16980
    },
    {
      "epoch": 0.23880806803007942,
      "grad_norm": 0.3904546797275543,
      "learning_rate": 0.0001811371046033882,
      "loss": 0.2094,
      "step": 16990
    },
    {
      "epoch": 0.23894862604540024,
      "grad_norm": 0.3003694713115692,
      "learning_rate": 0.00018110992552924855,
      "loss": 0.1755,
      "step": 17000
    },
    {
      "epoch": 0.23894862604540024,
      "eval_chrf": 83.42682500619117,
      "eval_loss": 0.433625727891922,
      "eval_runtime": 194.2793,
      "eval_samples_per_second": 0.515,
      "eval_steps_per_second": 0.021,
      "step": 17000
    },
    {
      "epoch": 0.23908918406072105,
      "grad_norm": 0.4176693856716156,
      "learning_rate": 0.00018108272893053767,
      "loss": 0.2057,
      "step": 17010
    },
    {
      "epoch": 0.2392297420760419,
      "grad_norm": 0.4226667881011963,
      "learning_rate": 0.00018105551481313167,
      "loss": 0.1982,
      "step": 17020
    },
    {
      "epoch": 0.2393703000913627,
      "grad_norm": 0.5105968713760376,
      "learning_rate": 0.00018102828318291038,
      "loss": 0.2114,
      "step": 17030
    },
    {
      "epoch": 0.23951085810668354,
      "grad_norm": 0.3624613583087921,
      "learning_rate": 0.0001810010340457575,
      "loss": 0.2005,
      "step": 17040
    },
    {
      "epoch": 0.23965141612200436,
      "grad_norm": 0.4732239544391632,
      "learning_rate": 0.00018097376740756045,
      "loss": 0.1839,
      "step": 17050
    },
    {
      "epoch": 0.23979197413732517,
      "grad_norm": 0.4396938383579254,
      "learning_rate": 0.0001809464832742104,
      "loss": 0.2279,
      "step": 17060
    },
    {
      "epoch": 0.239932532152646,
      "grad_norm": 0.4598935842514038,
      "learning_rate": 0.0001809191816516024,
      "loss": 0.2389,
      "step": 17070
    },
    {
      "epoch": 0.24007309016796682,
      "grad_norm": 0.43191003799438477,
      "learning_rate": 0.00018089186254563523,
      "loss": 0.1953,
      "step": 17080
    },
    {
      "epoch": 0.24021364818328766,
      "grad_norm": 0.29088646173477173,
      "learning_rate": 0.0001808645259622114,
      "loss": 0.1809,
      "step": 17090
    },
    {
      "epoch": 0.24035420619860848,
      "grad_norm": 0.5946388244628906,
      "learning_rate": 0.00018083717190723727,
      "loss": 0.1966,
      "step": 17100
    },
    {
      "epoch": 0.2404947642139293,
      "grad_norm": 0.4110700488090515,
      "learning_rate": 0.00018080980038662293,
      "loss": 0.1832,
      "step": 17110
    },
    {
      "epoch": 0.24063532222925013,
      "grad_norm": 0.44298121333122253,
      "learning_rate": 0.00018078241140628224,
      "loss": 0.2172,
      "step": 17120
    },
    {
      "epoch": 0.24077588024457094,
      "grad_norm": 0.5646606683731079,
      "learning_rate": 0.00018075500497213295,
      "loss": 0.2133,
      "step": 17130
    },
    {
      "epoch": 0.24091643825989176,
      "grad_norm": 0.40852683782577515,
      "learning_rate": 0.0001807275810900963,
      "loss": 0.2284,
      "step": 17140
    },
    {
      "epoch": 0.2410569962752126,
      "grad_norm": 0.38603800535202026,
      "learning_rate": 0.00018070013976609765,
      "loss": 0.2188,
      "step": 17150
    },
    {
      "epoch": 0.2411975542905334,
      "grad_norm": 0.4328840374946594,
      "learning_rate": 0.0001806726810060659,
      "loss": 0.2019,
      "step": 17160
    },
    {
      "epoch": 0.24133811230585425,
      "grad_norm": 0.44296500086784363,
      "learning_rate": 0.00018064520481593372,
      "loss": 0.1915,
      "step": 17170
    },
    {
      "epoch": 0.24147867032117507,
      "grad_norm": 0.3564743101596832,
      "learning_rate": 0.00018061771120163767,
      "loss": 0.222,
      "step": 17180
    },
    {
      "epoch": 0.24161922833649588,
      "grad_norm": 0.367158442735672,
      "learning_rate": 0.00018059020016911802,
      "loss": 0.1837,
      "step": 17190
    },
    {
      "epoch": 0.24175978635181672,
      "grad_norm": 0.40433263778686523,
      "learning_rate": 0.00018056267172431874,
      "loss": 0.2111,
      "step": 17200
    },
    {
      "epoch": 0.24190034436713753,
      "grad_norm": 0.4994758367538452,
      "learning_rate": 0.00018053512587318758,
      "loss": 0.2068,
      "step": 17210
    },
    {
      "epoch": 0.24204090238245837,
      "grad_norm": 0.5511106848716736,
      "learning_rate": 0.00018050756262167622,
      "loss": 0.2285,
      "step": 17220
    },
    {
      "epoch": 0.24218146039777919,
      "grad_norm": 0.4459092319011688,
      "learning_rate": 0.0001804799819757398,
      "loss": 0.221,
      "step": 17230
    },
    {
      "epoch": 0.2423220184131,
      "grad_norm": 0.38065147399902344,
      "learning_rate": 0.00018045238394133752,
      "loss": 0.2076,
      "step": 17240
    },
    {
      "epoch": 0.24246257642842084,
      "grad_norm": 0.4815782606601715,
      "learning_rate": 0.0001804247685244321,
      "loss": 0.2036,
      "step": 17250
    },
    {
      "epoch": 0.24260313444374165,
      "grad_norm": 0.4660325348377228,
      "learning_rate": 0.00018039713573099017,
      "loss": 0.2204,
      "step": 17260
    },
    {
      "epoch": 0.24274369245906247,
      "grad_norm": 0.4304184913635254,
      "learning_rate": 0.00018036948556698204,
      "loss": 0.214,
      "step": 17270
    },
    {
      "epoch": 0.2428842504743833,
      "grad_norm": 0.49734094738960266,
      "learning_rate": 0.00018034181803838178,
      "loss": 0.212,
      "step": 17280
    },
    {
      "epoch": 0.24302480848970412,
      "grad_norm": 0.48818838596343994,
      "learning_rate": 0.00018031413315116722,
      "loss": 0.2311,
      "step": 17290
    },
    {
      "epoch": 0.24316536650502496,
      "grad_norm": 0.43453922867774963,
      "learning_rate": 0.00018028643091132,
      "loss": 0.2106,
      "step": 17300
    },
    {
      "epoch": 0.24330592452034577,
      "grad_norm": 0.3406917452812195,
      "learning_rate": 0.00018025871132482536,
      "loss": 0.2251,
      "step": 17310
    },
    {
      "epoch": 0.2434464825356666,
      "grad_norm": 0.43452584743499756,
      "learning_rate": 0.00018023097439767244,
      "loss": 0.2188,
      "step": 17320
    },
    {
      "epoch": 0.24358704055098743,
      "grad_norm": 0.5142020583152771,
      "learning_rate": 0.00018020322013585408,
      "loss": 0.2188,
      "step": 17330
    },
    {
      "epoch": 0.24372759856630824,
      "grad_norm": 0.5353051424026489,
      "learning_rate": 0.00018017544854536679,
      "loss": 0.1949,
      "step": 17340
    },
    {
      "epoch": 0.24386815658162908,
      "grad_norm": 0.49185070395469666,
      "learning_rate": 0.0001801476596322109,
      "loss": 0.2284,
      "step": 17350
    },
    {
      "epoch": 0.2440087145969499,
      "grad_norm": 0.40892067551612854,
      "learning_rate": 0.00018011985340239054,
      "loss": 0.2017,
      "step": 17360
    },
    {
      "epoch": 0.2441492726122707,
      "grad_norm": 0.4842357933521271,
      "learning_rate": 0.00018009202986191348,
      "loss": 0.2098,
      "step": 17370
    },
    {
      "epoch": 0.24428983062759155,
      "grad_norm": 0.5735198855400085,
      "learning_rate": 0.0001800641890167912,
      "loss": 0.2444,
      "step": 17380
    },
    {
      "epoch": 0.24443038864291236,
      "grad_norm": 0.39232829213142395,
      "learning_rate": 0.00018003633087303907,
      "loss": 0.1893,
      "step": 17390
    },
    {
      "epoch": 0.24457094665823317,
      "grad_norm": 0.4202956557273865,
      "learning_rate": 0.000180008455436676,
      "loss": 0.209,
      "step": 17400
    },
    {
      "epoch": 0.24471150467355401,
      "grad_norm": 0.41963180899620056,
      "learning_rate": 0.00017998056271372484,
      "loss": 0.2169,
      "step": 17410
    },
    {
      "epoch": 0.24485206268887483,
      "grad_norm": 0.44267401099205017,
      "learning_rate": 0.00017995265271021207,
      "loss": 0.1852,
      "step": 17420
    },
    {
      "epoch": 0.24499262070419567,
      "grad_norm": 0.4419970214366913,
      "learning_rate": 0.00017992472543216785,
      "loss": 0.1821,
      "step": 17430
    },
    {
      "epoch": 0.24513317871951648,
      "grad_norm": 0.4276140630245209,
      "learning_rate": 0.00017989678088562618,
      "loss": 0.2028,
      "step": 17440
    },
    {
      "epoch": 0.2452737367348373,
      "grad_norm": 0.5623214840888977,
      "learning_rate": 0.00017986881907662471,
      "loss": 0.1916,
      "step": 17450
    },
    {
      "epoch": 0.24541429475015814,
      "grad_norm": 0.36114174127578735,
      "learning_rate": 0.00017984084001120492,
      "loss": 0.2202,
      "step": 17460
    },
    {
      "epoch": 0.24555485276547895,
      "grad_norm": 0.5440938472747803,
      "learning_rate": 0.00017981284369541188,
      "loss": 0.221,
      "step": 17470
    },
    {
      "epoch": 0.2456954107807998,
      "grad_norm": 0.40802478790283203,
      "learning_rate": 0.00017978483013529452,
      "loss": 0.2104,
      "step": 17480
    },
    {
      "epoch": 0.2458359687961206,
      "grad_norm": 0.5030210614204407,
      "learning_rate": 0.00017975679933690536,
      "loss": 0.2122,
      "step": 17490
    },
    {
      "epoch": 0.24597652681144141,
      "grad_norm": 0.34485700726509094,
      "learning_rate": 0.0001797287513063008,
      "loss": 0.2088,
      "step": 17500
    },
    {
      "epoch": 0.24611708482676226,
      "grad_norm": 0.42414161562919617,
      "learning_rate": 0.00017970068604954086,
      "loss": 0.2085,
      "step": 17510
    },
    {
      "epoch": 0.24625764284208307,
      "grad_norm": 0.41926831007003784,
      "learning_rate": 0.00017967260357268928,
      "loss": 0.1908,
      "step": 17520
    },
    {
      "epoch": 0.24639820085740388,
      "grad_norm": 0.4200794994831085,
      "learning_rate": 0.00017964450388181357,
      "loss": 0.1899,
      "step": 17530
    },
    {
      "epoch": 0.24653875887272472,
      "grad_norm": 0.4162304997444153,
      "learning_rate": 0.00017961638698298492,
      "loss": 0.2034,
      "step": 17540
    },
    {
      "epoch": 0.24667931688804554,
      "grad_norm": 0.48807257413864136,
      "learning_rate": 0.00017958825288227827,
      "loss": 0.2171,
      "step": 17550
    },
    {
      "epoch": 0.24681987490336638,
      "grad_norm": 0.47989019751548767,
      "learning_rate": 0.0001795601015857723,
      "loss": 0.2188,
      "step": 17560
    },
    {
      "epoch": 0.2469604329186872,
      "grad_norm": 0.3965669870376587,
      "learning_rate": 0.00017953193309954926,
      "loss": 0.2018,
      "step": 17570
    },
    {
      "epoch": 0.247100990934008,
      "grad_norm": 0.49545612931251526,
      "learning_rate": 0.0001795037474296953,
      "loss": 0.2242,
      "step": 17580
    },
    {
      "epoch": 0.24724154894932884,
      "grad_norm": 0.35966184735298157,
      "learning_rate": 0.0001794755445823002,
      "loss": 0.213,
      "step": 17590
    },
    {
      "epoch": 0.24738210696464966,
      "grad_norm": 0.4222733676433563,
      "learning_rate": 0.00017944732456345741,
      "loss": 0.1917,
      "step": 17600
    },
    {
      "epoch": 0.24752266497997047,
      "grad_norm": 0.4538722038269043,
      "learning_rate": 0.0001794190873792642,
      "loss": 0.1971,
      "step": 17610
    },
    {
      "epoch": 0.2476632229952913,
      "grad_norm": 0.4126892387866974,
      "learning_rate": 0.00017939083303582137,
      "loss": 0.2106,
      "step": 17620
    },
    {
      "epoch": 0.24780378101061212,
      "grad_norm": 0.5503072142601013,
      "learning_rate": 0.00017936256153923367,
      "loss": 0.1967,
      "step": 17630
    },
    {
      "epoch": 0.24794433902593296,
      "grad_norm": 0.44349250197410583,
      "learning_rate": 0.00017933427289560934,
      "loss": 0.225,
      "step": 17640
    },
    {
      "epoch": 0.24808489704125378,
      "grad_norm": 0.5387433767318726,
      "learning_rate": 0.00017930596711106044,
      "loss": 0.2118,
      "step": 17650
    },
    {
      "epoch": 0.2482254550565746,
      "grad_norm": 0.4662763774394989,
      "learning_rate": 0.0001792776441917027,
      "loss": 0.192,
      "step": 17660
    },
    {
      "epoch": 0.24836601307189543,
      "grad_norm": 0.640113890171051,
      "learning_rate": 0.00017924930414365558,
      "loss": 0.1827,
      "step": 17670
    },
    {
      "epoch": 0.24850657108721624,
      "grad_norm": 0.4936950206756592,
      "learning_rate": 0.00017922094697304217,
      "loss": 0.2297,
      "step": 17680
    },
    {
      "epoch": 0.24864712910253708,
      "grad_norm": 0.41726192831993103,
      "learning_rate": 0.00017919257268598932,
      "loss": 0.2046,
      "step": 17690
    },
    {
      "epoch": 0.2487876871178579,
      "grad_norm": 0.42578253149986267,
      "learning_rate": 0.0001791641812886276,
      "loss": 0.1857,
      "step": 17700
    },
    {
      "epoch": 0.2489282451331787,
      "grad_norm": 0.39796364307403564,
      "learning_rate": 0.0001791357727870912,
      "loss": 0.2083,
      "step": 17710
    },
    {
      "epoch": 0.24906880314849955,
      "grad_norm": 0.5429249405860901,
      "learning_rate": 0.0001791073471875181,
      "loss": 0.2113,
      "step": 17720
    },
    {
      "epoch": 0.24920936116382036,
      "grad_norm": 0.42298436164855957,
      "learning_rate": 0.00017907890449604986,
      "loss": 0.2002,
      "step": 17730
    },
    {
      "epoch": 0.24934991917914118,
      "grad_norm": 0.4462754428386688,
      "learning_rate": 0.00017905044471883184,
      "loss": 0.2039,
      "step": 17740
    },
    {
      "epoch": 0.24949047719446202,
      "grad_norm": 0.4810706377029419,
      "learning_rate": 0.000179021967862013,
      "loss": 0.1958,
      "step": 17750
    },
    {
      "epoch": 0.24963103520978283,
      "grad_norm": 0.45998165011405945,
      "learning_rate": 0.00017899347393174605,
      "loss": 0.1908,
      "step": 17760
    },
    {
      "epoch": 0.24977159322510367,
      "grad_norm": 0.44851624965667725,
      "learning_rate": 0.0001789649629341874,
      "loss": 0.201,
      "step": 17770
    },
    {
      "epoch": 0.24991215124042448,
      "grad_norm": 0.4333913028240204,
      "learning_rate": 0.0001789364348754971,
      "loss": 0.194,
      "step": 17780
    },
    {
      "epoch": 0.2500527092557453,
      "grad_norm": 0.40873411297798157,
      "learning_rate": 0.00017890788976183895,
      "loss": 0.1963,
      "step": 17790
    },
    {
      "epoch": 0.2501932672710661,
      "grad_norm": 0.5309141874313354,
      "learning_rate": 0.00017887932759938036,
      "loss": 0.2043,
      "step": 17800
    },
    {
      "epoch": 0.250333825286387,
      "grad_norm": 0.4489333927631378,
      "learning_rate": 0.00017885074839429246,
      "loss": 0.2397,
      "step": 17810
    },
    {
      "epoch": 0.2504743833017078,
      "grad_norm": 0.700109601020813,
      "learning_rate": 0.00017882215215275,
      "loss": 0.2205,
      "step": 17820
    },
    {
      "epoch": 0.2506149413170286,
      "grad_norm": 0.4139418303966522,
      "learning_rate": 0.00017879353888093158,
      "loss": 0.2263,
      "step": 17830
    },
    {
      "epoch": 0.2507554993323494,
      "grad_norm": 0.411004900932312,
      "learning_rate": 0.0001787649085850193,
      "loss": 0.1978,
      "step": 17840
    },
    {
      "epoch": 0.25089605734767023,
      "grad_norm": 0.5109737515449524,
      "learning_rate": 0.00017873626127119902,
      "loss": 0.2025,
      "step": 17850
    },
    {
      "epoch": 0.2510366153629911,
      "grad_norm": 0.6595651507377625,
      "learning_rate": 0.00017870759694566028,
      "loss": 0.2063,
      "step": 17860
    },
    {
      "epoch": 0.2511771733783119,
      "grad_norm": 0.5950677990913391,
      "learning_rate": 0.00017867891561459625,
      "loss": 0.2064,
      "step": 17870
    },
    {
      "epoch": 0.2513177313936327,
      "grad_norm": 0.4480992555618286,
      "learning_rate": 0.00017865021728420385,
      "loss": 0.221,
      "step": 17880
    },
    {
      "epoch": 0.25145828940895354,
      "grad_norm": 0.44384530186653137,
      "learning_rate": 0.00017862150196068359,
      "loss": 0.211,
      "step": 17890
    },
    {
      "epoch": 0.25159884742427435,
      "grad_norm": 0.46428948640823364,
      "learning_rate": 0.0001785927696502397,
      "loss": 0.1983,
      "step": 17900
    },
    {
      "epoch": 0.25173940543959517,
      "grad_norm": 0.3470834791660309,
      "learning_rate": 0.00017856402035908007,
      "loss": 0.1977,
      "step": 17910
    },
    {
      "epoch": 0.25187996345491603,
      "grad_norm": 0.48576465249061584,
      "learning_rate": 0.00017853525409341624,
      "loss": 0.2215,
      "step": 17920
    },
    {
      "epoch": 0.25202052147023685,
      "grad_norm": 0.44239556789398193,
      "learning_rate": 0.00017850647085946347,
      "loss": 0.2219,
      "step": 17930
    },
    {
      "epoch": 0.25216107948555766,
      "grad_norm": 0.3892723619937897,
      "learning_rate": 0.00017847767066344065,
      "loss": 0.1916,
      "step": 17940
    },
    {
      "epoch": 0.2523016375008785,
      "grad_norm": 0.4856000542640686,
      "learning_rate": 0.0001784488535115703,
      "loss": 0.2231,
      "step": 17950
    },
    {
      "epoch": 0.2524421955161993,
      "grad_norm": 0.49842894077301025,
      "learning_rate": 0.00017842001941007868,
      "loss": 0.2161,
      "step": 17960
    },
    {
      "epoch": 0.25258275353152015,
      "grad_norm": 0.5050910711288452,
      "learning_rate": 0.00017839116836519564,
      "loss": 0.2183,
      "step": 17970
    },
    {
      "epoch": 0.25272331154684097,
      "grad_norm": 0.42533183097839355,
      "learning_rate": 0.0001783623003831547,
      "loss": 0.2169,
      "step": 17980
    },
    {
      "epoch": 0.2528638695621618,
      "grad_norm": 0.42304936051368713,
      "learning_rate": 0.00017833341547019316,
      "loss": 0.195,
      "step": 17990
    },
    {
      "epoch": 0.2530044275774826,
      "grad_norm": 0.4381530284881592,
      "learning_rate": 0.00017830451363255178,
      "loss": 0.2159,
      "step": 18000
    },
    {
      "epoch": 0.2530044275774826,
      "eval_chrf": 83.43002044720006,
      "eval_loss": 0.4113892614841461,
      "eval_runtime": 138.5753,
      "eval_samples_per_second": 0.722,
      "eval_steps_per_second": 0.029,
      "step": 18000
    },
    {
      "epoch": 0.2531449855928034,
      "grad_norm": 0.40219688415527344,
      "learning_rate": 0.00017827559487647513,
      "loss": 0.2029,
      "step": 18010
    },
    {
      "epoch": 0.2532855436081243,
      "grad_norm": 0.38504263758659363,
      "learning_rate": 0.00017824665920821136,
      "loss": 0.1915,
      "step": 18020
    },
    {
      "epoch": 0.2534261016234451,
      "grad_norm": 0.5165150761604309,
      "learning_rate": 0.0001782177066340123,
      "loss": 0.1825,
      "step": 18030
    },
    {
      "epoch": 0.2535666596387659,
      "grad_norm": 0.3051179349422455,
      "learning_rate": 0.00017818873716013344,
      "loss": 0.1767,
      "step": 18040
    },
    {
      "epoch": 0.2537072176540867,
      "grad_norm": 0.3943491578102112,
      "learning_rate": 0.0001781597507928339,
      "loss": 0.2013,
      "step": 18050
    },
    {
      "epoch": 0.2538477756694075,
      "grad_norm": 0.5130552649497986,
      "learning_rate": 0.00017813074753837647,
      "loss": 0.1949,
      "step": 18060
    },
    {
      "epoch": 0.2539883336847284,
      "grad_norm": 0.34325119853019714,
      "learning_rate": 0.00017810172740302758,
      "loss": 0.1749,
      "step": 18070
    },
    {
      "epoch": 0.2541288917000492,
      "grad_norm": 0.44568413496017456,
      "learning_rate": 0.00017807269039305727,
      "loss": 0.193,
      "step": 18080
    },
    {
      "epoch": 0.25426944971537,
      "grad_norm": 0.3938472867012024,
      "learning_rate": 0.00017804363651473932,
      "loss": 0.1897,
      "step": 18090
    },
    {
      "epoch": 0.25441000773069083,
      "grad_norm": 0.3924921154975891,
      "learning_rate": 0.00017801456577435105,
      "loss": 0.2201,
      "step": 18100
    },
    {
      "epoch": 0.25455056574601165,
      "grad_norm": 0.3777751922607422,
      "learning_rate": 0.0001779854781781735,
      "loss": 0.2211,
      "step": 18110
    },
    {
      "epoch": 0.2546911237613325,
      "grad_norm": 0.46796715259552,
      "learning_rate": 0.0001779563737324913,
      "loss": 0.1866,
      "step": 18120
    },
    {
      "epoch": 0.25483168177665333,
      "grad_norm": 0.3461954891681671,
      "learning_rate": 0.00017792725244359276,
      "loss": 0.2112,
      "step": 18130
    },
    {
      "epoch": 0.25497223979197414,
      "grad_norm": 0.40790891647338867,
      "learning_rate": 0.0001778981143177698,
      "loss": 0.1958,
      "step": 18140
    },
    {
      "epoch": 0.25511279780729496,
      "grad_norm": 0.3641470968723297,
      "learning_rate": 0.000177868959361318,
      "loss": 0.1938,
      "step": 18150
    },
    {
      "epoch": 0.25525335582261577,
      "grad_norm": 0.43044883012771606,
      "learning_rate": 0.00017783978758053657,
      "loss": 0.1978,
      "step": 18160
    },
    {
      "epoch": 0.2553939138379366,
      "grad_norm": 0.39675578474998474,
      "learning_rate": 0.00017781059898172835,
      "loss": 0.2101,
      "step": 18170
    },
    {
      "epoch": 0.25553447185325745,
      "grad_norm": 0.4233925938606262,
      "learning_rate": 0.00017778139357119976,
      "loss": 0.2083,
      "step": 18180
    },
    {
      "epoch": 0.25567502986857826,
      "grad_norm": 0.38321974873542786,
      "learning_rate": 0.00017775217135526097,
      "loss": 0.2095,
      "step": 18190
    },
    {
      "epoch": 0.2558155878838991,
      "grad_norm": 0.4945431351661682,
      "learning_rate": 0.00017772293234022572,
      "loss": 0.1863,
      "step": 18200
    },
    {
      "epoch": 0.2559561458992199,
      "grad_norm": 0.518644392490387,
      "learning_rate": 0.00017769367653241134,
      "loss": 0.2165,
      "step": 18210
    },
    {
      "epoch": 0.2560967039145407,
      "grad_norm": 0.36446699500083923,
      "learning_rate": 0.00017766440393813884,
      "loss": 0.1928,
      "step": 18220
    },
    {
      "epoch": 0.25623726192986157,
      "grad_norm": 0.3861456513404846,
      "learning_rate": 0.00017763511456373284,
      "loss": 0.2266,
      "step": 18230
    },
    {
      "epoch": 0.2563778199451824,
      "grad_norm": 0.38828468322753906,
      "learning_rate": 0.00017760580841552162,
      "loss": 0.2068,
      "step": 18240
    },
    {
      "epoch": 0.2565183779605032,
      "grad_norm": 0.3888741135597229,
      "learning_rate": 0.000177576485499837,
      "loss": 0.2199,
      "step": 18250
    },
    {
      "epoch": 0.256658935975824,
      "grad_norm": 0.5249620676040649,
      "learning_rate": 0.0001775471458230145,
      "loss": 0.2339,
      "step": 18260
    },
    {
      "epoch": 0.2567994939911448,
      "grad_norm": 0.33323395252227783,
      "learning_rate": 0.00017751778939139326,
      "loss": 0.204,
      "step": 18270
    },
    {
      "epoch": 0.2569400520064657,
      "grad_norm": 0.548209547996521,
      "learning_rate": 0.000177488416211316,
      "loss": 0.2021,
      "step": 18280
    },
    {
      "epoch": 0.2570806100217865,
      "grad_norm": 0.4949204623699188,
      "learning_rate": 0.00017745902628912903,
      "loss": 0.2204,
      "step": 18290
    },
    {
      "epoch": 0.2572211680371073,
      "grad_norm": 0.3896910548210144,
      "learning_rate": 0.00017742961963118238,
      "loss": 0.1932,
      "step": 18300
    },
    {
      "epoch": 0.25736172605242813,
      "grad_norm": 0.4428505599498749,
      "learning_rate": 0.00017740019624382962,
      "loss": 0.1915,
      "step": 18310
    },
    {
      "epoch": 0.25750228406774894,
      "grad_norm": 0.5322151184082031,
      "learning_rate": 0.00017737075613342794,
      "loss": 0.2097,
      "step": 18320
    },
    {
      "epoch": 0.2576428420830698,
      "grad_norm": 0.3616546094417572,
      "learning_rate": 0.00017734129930633817,
      "loss": 0.1985,
      "step": 18330
    },
    {
      "epoch": 0.2577834000983906,
      "grad_norm": 0.6213672161102295,
      "learning_rate": 0.00017731182576892477,
      "loss": 0.2052,
      "step": 18340
    },
    {
      "epoch": 0.25792395811371144,
      "grad_norm": 0.48710718750953674,
      "learning_rate": 0.0001772823355275557,
      "loss": 0.188,
      "step": 18350
    },
    {
      "epoch": 0.25806451612903225,
      "grad_norm": 0.45230305194854736,
      "learning_rate": 0.00017725282858860268,
      "loss": 0.205,
      "step": 18360
    },
    {
      "epoch": 0.25820507414435306,
      "grad_norm": 0.46916815638542175,
      "learning_rate": 0.00017722330495844094,
      "loss": 0.2174,
      "step": 18370
    },
    {
      "epoch": 0.25834563215967393,
      "grad_norm": 0.7887526750564575,
      "learning_rate": 0.0001771937646434493,
      "loss": 0.2108,
      "step": 18380
    },
    {
      "epoch": 0.25848619017499475,
      "grad_norm": 0.38442105054855347,
      "learning_rate": 0.0001771642076500103,
      "loss": 0.1842,
      "step": 18390
    },
    {
      "epoch": 0.25862674819031556,
      "grad_norm": 0.5363273024559021,
      "learning_rate": 0.00017713463398450995,
      "loss": 0.2144,
      "step": 18400
    },
    {
      "epoch": 0.25876730620563637,
      "grad_norm": 0.48678267002105713,
      "learning_rate": 0.00017710504365333796,
      "loss": 0.2134,
      "step": 18410
    },
    {
      "epoch": 0.2589078642209572,
      "grad_norm": 0.5642648935317993,
      "learning_rate": 0.0001770754366628876,
      "loss": 0.2127,
      "step": 18420
    },
    {
      "epoch": 0.259048422236278,
      "grad_norm": 0.4107931852340698,
      "learning_rate": 0.0001770458130195557,
      "loss": 0.2001,
      "step": 18430
    },
    {
      "epoch": 0.25918898025159887,
      "grad_norm": 0.4847404658794403,
      "learning_rate": 0.00017701617272974275,
      "loss": 0.1919,
      "step": 18440
    },
    {
      "epoch": 0.2593295382669197,
      "grad_norm": 0.44709765911102295,
      "learning_rate": 0.00017698651579985285,
      "loss": 0.2095,
      "step": 18450
    },
    {
      "epoch": 0.2594700962822405,
      "grad_norm": 0.4158313572406769,
      "learning_rate": 0.00017695684223629365,
      "loss": 0.1873,
      "step": 18460
    },
    {
      "epoch": 0.2596106542975613,
      "grad_norm": 0.4474424719810486,
      "learning_rate": 0.0001769271520454764,
      "loss": 0.2023,
      "step": 18470
    },
    {
      "epoch": 0.2597512123128821,
      "grad_norm": 0.5589454174041748,
      "learning_rate": 0.0001768974452338159,
      "loss": 0.2115,
      "step": 18480
    },
    {
      "epoch": 0.259891770328203,
      "grad_norm": 0.3891916573047638,
      "learning_rate": 0.00017686772180773065,
      "loss": 0.2203,
      "step": 18490
    },
    {
      "epoch": 0.2600323283435238,
      "grad_norm": 0.43455711007118225,
      "learning_rate": 0.00017683798177364266,
      "loss": 0.2092,
      "step": 18500
    },
    {
      "epoch": 0.2601728863588446,
      "grad_norm": 0.4177962839603424,
      "learning_rate": 0.00017680822513797757,
      "loss": 0.2134,
      "step": 18510
    },
    {
      "epoch": 0.2603134443741654,
      "grad_norm": 0.4519466757774353,
      "learning_rate": 0.00017677845190716454,
      "loss": 0.196,
      "step": 18520
    },
    {
      "epoch": 0.26045400238948624,
      "grad_norm": 0.47306278347969055,
      "learning_rate": 0.00017674866208763638,
      "loss": 0.2026,
      "step": 18530
    },
    {
      "epoch": 0.2605945604048071,
      "grad_norm": 0.4423769414424896,
      "learning_rate": 0.00017671885568582946,
      "loss": 0.2158,
      "step": 18540
    },
    {
      "epoch": 0.2607351184201279,
      "grad_norm": 0.4966670274734497,
      "learning_rate": 0.00017668903270818377,
      "loss": 0.2113,
      "step": 18550
    },
    {
      "epoch": 0.26087567643544873,
      "grad_norm": 0.40313148498535156,
      "learning_rate": 0.0001766591931611428,
      "loss": 0.2062,
      "step": 18560
    },
    {
      "epoch": 0.26101623445076955,
      "grad_norm": 0.37123391032218933,
      "learning_rate": 0.0001766293370511537,
      "loss": 0.2014,
      "step": 18570
    },
    {
      "epoch": 0.26115679246609036,
      "grad_norm": 0.5386476516723633,
      "learning_rate": 0.00017659946438466714,
      "loss": 0.202,
      "step": 18580
    },
    {
      "epoch": 0.26129735048141123,
      "grad_norm": 0.5218870639801025,
      "learning_rate": 0.00017656957516813744,
      "loss": 0.211,
      "step": 18590
    },
    {
      "epoch": 0.26143790849673204,
      "grad_norm": 0.4854200482368469,
      "learning_rate": 0.0001765396694080224,
      "loss": 0.1978,
      "step": 18600
    },
    {
      "epoch": 0.26157846651205285,
      "grad_norm": 0.4300956428050995,
      "learning_rate": 0.00017650974711078346,
      "loss": 0.1931,
      "step": 18610
    },
    {
      "epoch": 0.26171902452737367,
      "grad_norm": 0.5700703859329224,
      "learning_rate": 0.0001764798082828856,
      "loss": 0.2024,
      "step": 18620
    },
    {
      "epoch": 0.2618595825426945,
      "grad_norm": 0.4762032926082611,
      "learning_rate": 0.00017644985293079746,
      "loss": 0.2197,
      "step": 18630
    },
    {
      "epoch": 0.2620001405580153,
      "grad_norm": 0.35937386751174927,
      "learning_rate": 0.00017641988106099107,
      "loss": 0.212,
      "step": 18640
    },
    {
      "epoch": 0.26214069857333616,
      "grad_norm": 0.3749754726886749,
      "learning_rate": 0.00017638989267994225,
      "loss": 0.1959,
      "step": 18650
    },
    {
      "epoch": 0.262281256588657,
      "grad_norm": 0.3686469793319702,
      "learning_rate": 0.00017635988779413019,
      "loss": 0.1866,
      "step": 18660
    },
    {
      "epoch": 0.2624218146039778,
      "grad_norm": 0.5276985764503479,
      "learning_rate": 0.00017632986641003776,
      "loss": 0.229,
      "step": 18670
    },
    {
      "epoch": 0.2625623726192986,
      "grad_norm": 0.3779345452785492,
      "learning_rate": 0.00017629982853415138,
      "loss": 0.1868,
      "step": 18680
    },
    {
      "epoch": 0.2627029306346194,
      "grad_norm": 0.5236541032791138,
      "learning_rate": 0.000176269774172961,
      "loss": 0.2268,
      "step": 18690
    },
    {
      "epoch": 0.2628434886499403,
      "grad_norm": 0.4989383816719055,
      "learning_rate": 0.00017623970333296017,
      "loss": 0.2125,
      "step": 18700
    },
    {
      "epoch": 0.2629840466652611,
      "grad_norm": 0.47551271319389343,
      "learning_rate": 0.00017620961602064593,
      "loss": 0.2148,
      "step": 18710
    },
    {
      "epoch": 0.2631246046805819,
      "grad_norm": 0.4368431568145752,
      "learning_rate": 0.000176179512242519,
      "loss": 0.2271,
      "step": 18720
    },
    {
      "epoch": 0.2632651626959027,
      "grad_norm": 0.6284667253494263,
      "learning_rate": 0.00017614939200508357,
      "loss": 0.214,
      "step": 18730
    },
    {
      "epoch": 0.26340572071122353,
      "grad_norm": 0.42767229676246643,
      "learning_rate": 0.00017611925531484736,
      "loss": 0.2123,
      "step": 18740
    },
    {
      "epoch": 0.2635462787265444,
      "grad_norm": 0.6016656160354614,
      "learning_rate": 0.0001760891021783217,
      "loss": 0.2052,
      "step": 18750
    },
    {
      "epoch": 0.2636868367418652,
      "grad_norm": 0.43791624903678894,
      "learning_rate": 0.00017605893260202153,
      "loss": 0.2119,
      "step": 18760
    },
    {
      "epoch": 0.26382739475718603,
      "grad_norm": 0.3640478849411011,
      "learning_rate": 0.0001760287465924652,
      "loss": 0.2138,
      "step": 18770
    },
    {
      "epoch": 0.26396795277250684,
      "grad_norm": 0.38699495792388916,
      "learning_rate": 0.00017599854415617473,
      "loss": 0.1983,
      "step": 18780
    },
    {
      "epoch": 0.26410851078782765,
      "grad_norm": 0.4709777235984802,
      "learning_rate": 0.0001759683252996756,
      "loss": 0.2169,
      "step": 18790
    },
    {
      "epoch": 0.2642490688031485,
      "grad_norm": 0.42550647258758545,
      "learning_rate": 0.0001759380900294969,
      "loss": 0.2114,
      "step": 18800
    },
    {
      "epoch": 0.26438962681846934,
      "grad_norm": 0.5002481341362,
      "learning_rate": 0.00017590783835217126,
      "loss": 0.2095,
      "step": 18810
    },
    {
      "epoch": 0.26453018483379015,
      "grad_norm": 0.4592345654964447,
      "learning_rate": 0.00017587757027423483,
      "loss": 0.1928,
      "step": 18820
    },
    {
      "epoch": 0.26467074284911096,
      "grad_norm": 0.3958444595336914,
      "learning_rate": 0.00017584728580222734,
      "loss": 0.228,
      "step": 18830
    },
    {
      "epoch": 0.2648113008644318,
      "grad_norm": 0.3654407858848572,
      "learning_rate": 0.00017581698494269202,
      "loss": 0.2025,
      "step": 18840
    },
    {
      "epoch": 0.26495185887975264,
      "grad_norm": 0.4766884446144104,
      "learning_rate": 0.00017578666770217563,
      "loss": 0.1977,
      "step": 18850
    },
    {
      "epoch": 0.26509241689507346,
      "grad_norm": 0.4469415545463562,
      "learning_rate": 0.00017575633408722853,
      "loss": 0.1858,
      "step": 18860
    },
    {
      "epoch": 0.26523297491039427,
      "grad_norm": 0.47276103496551514,
      "learning_rate": 0.0001757259841044046,
      "loss": 0.2349,
      "step": 18870
    },
    {
      "epoch": 0.2653735329257151,
      "grad_norm": 0.4530576765537262,
      "learning_rate": 0.00017569561776026123,
      "loss": 0.2035,
      "step": 18880
    },
    {
      "epoch": 0.2655140909410359,
      "grad_norm": 0.4598478376865387,
      "learning_rate": 0.00017566523506135935,
      "loss": 0.189,
      "step": 18890
    },
    {
      "epoch": 0.2656546489563567,
      "grad_norm": 0.5444883704185486,
      "learning_rate": 0.0001756348360142634,
      "loss": 0.1917,
      "step": 18900
    },
    {
      "epoch": 0.2657952069716776,
      "grad_norm": 0.5340624451637268,
      "learning_rate": 0.00017560442062554147,
      "loss": 0.2025,
      "step": 18910
    },
    {
      "epoch": 0.2659357649869984,
      "grad_norm": 0.5737981200218201,
      "learning_rate": 0.00017557398890176502,
      "loss": 0.1819,
      "step": 18920
    },
    {
      "epoch": 0.2660763230023192,
      "grad_norm": 0.5385230779647827,
      "learning_rate": 0.00017554354084950915,
      "loss": 0.2375,
      "step": 18930
    },
    {
      "epoch": 0.26621688101764,
      "grad_norm": 0.49425017833709717,
      "learning_rate": 0.0001755130764753524,
      "loss": 0.216,
      "step": 18940
    },
    {
      "epoch": 0.26635743903296083,
      "grad_norm": 0.44835686683654785,
      "learning_rate": 0.00017548259578587696,
      "loss": 0.216,
      "step": 18950
    },
    {
      "epoch": 0.2664979970482817,
      "grad_norm": 0.4047083258628845,
      "learning_rate": 0.0001754520987876684,
      "loss": 0.1913,
      "step": 18960
    },
    {
      "epoch": 0.2666385550636025,
      "grad_norm": 0.4027661383152008,
      "learning_rate": 0.00017542158548731598,
      "loss": 0.1955,
      "step": 18970
    },
    {
      "epoch": 0.2667791130789233,
      "grad_norm": 0.4459643065929413,
      "learning_rate": 0.0001753910558914123,
      "loss": 0.2197,
      "step": 18980
    },
    {
      "epoch": 0.26691967109424414,
      "grad_norm": 0.3401888906955719,
      "learning_rate": 0.0001753605100065536,
      "loss": 0.2196,
      "step": 18990
    },
    {
      "epoch": 0.26706022910956495,
      "grad_norm": 0.46558499336242676,
      "learning_rate": 0.00017532994783933962,
      "loss": 0.2281,
      "step": 19000
    },
    {
      "epoch": 0.26706022910956495,
      "eval_chrf": 76.47169301348917,
      "eval_loss": 0.4275648593902588,
      "eval_runtime": 326.6959,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 19000
    },
    {
      "epoch": 0.2672007871248858,
      "grad_norm": 0.4833649694919586,
      "learning_rate": 0.0001752993693963736,
      "loss": 0.182,
      "step": 19010
    },
    {
      "epoch": 0.26734134514020663,
      "grad_norm": 0.5820573568344116,
      "learning_rate": 0.0001752687746842623,
      "loss": 0.2045,
      "step": 19020
    },
    {
      "epoch": 0.26748190315552745,
      "grad_norm": 0.5024533271789551,
      "learning_rate": 0.00017523816370961602,
      "loss": 0.1888,
      "step": 19030
    },
    {
      "epoch": 0.26762246117084826,
      "grad_norm": 0.4784148037433624,
      "learning_rate": 0.0001752075364790485,
      "loss": 0.2106,
      "step": 19040
    },
    {
      "epoch": 0.26776301918616907,
      "grad_norm": 0.5217591524124146,
      "learning_rate": 0.0001751768929991771,
      "loss": 0.2359,
      "step": 19050
    },
    {
      "epoch": 0.26790357720148994,
      "grad_norm": 0.43953946232795715,
      "learning_rate": 0.0001751462332766226,
      "loss": 0.2068,
      "step": 19060
    },
    {
      "epoch": 0.26804413521681075,
      "grad_norm": 0.5105519890785217,
      "learning_rate": 0.00017511555731800936,
      "loss": 0.2008,
      "step": 19070
    },
    {
      "epoch": 0.26818469323213157,
      "grad_norm": 0.4403964877128601,
      "learning_rate": 0.00017508486512996516,
      "loss": 0.2134,
      "step": 19080
    },
    {
      "epoch": 0.2683252512474524,
      "grad_norm": 0.5443793535232544,
      "learning_rate": 0.00017505415671912139,
      "loss": 0.2092,
      "step": 19090
    },
    {
      "epoch": 0.2684658092627732,
      "grad_norm": 0.3778785765171051,
      "learning_rate": 0.00017502343209211288,
      "loss": 0.2164,
      "step": 19100
    },
    {
      "epoch": 0.26860636727809406,
      "grad_norm": 0.5645463466644287,
      "learning_rate": 0.000174992691255578,
      "loss": 0.201,
      "step": 19110
    },
    {
      "epoch": 0.2687469252934149,
      "grad_norm": 0.444047749042511,
      "learning_rate": 0.00017496193421615851,
      "loss": 0.1811,
      "step": 19120
    },
    {
      "epoch": 0.2688874833087357,
      "grad_norm": 0.4804798662662506,
      "learning_rate": 0.00017493116098049987,
      "loss": 0.2073,
      "step": 19130
    },
    {
      "epoch": 0.2690280413240565,
      "grad_norm": 0.417547345161438,
      "learning_rate": 0.00017490037155525086,
      "loss": 0.1938,
      "step": 19140
    },
    {
      "epoch": 0.2691685993393773,
      "grad_norm": 0.5173671841621399,
      "learning_rate": 0.00017486956594706387,
      "loss": 0.2288,
      "step": 19150
    },
    {
      "epoch": 0.2693091573546981,
      "grad_norm": 0.4034634232521057,
      "learning_rate": 0.00017483874416259471,
      "loss": 0.1969,
      "step": 19160
    },
    {
      "epoch": 0.269449715370019,
      "grad_norm": 0.3967057168483734,
      "learning_rate": 0.00017480790620850276,
      "loss": 0.2383,
      "step": 19170
    },
    {
      "epoch": 0.2695902733853398,
      "grad_norm": 0.4266200661659241,
      "learning_rate": 0.00017477705209145082,
      "loss": 0.2376,
      "step": 19180
    },
    {
      "epoch": 0.2697308314006606,
      "grad_norm": 0.3088657557964325,
      "learning_rate": 0.00017474618181810523,
      "loss": 0.2151,
      "step": 19190
    },
    {
      "epoch": 0.26987138941598143,
      "grad_norm": 0.5541336536407471,
      "learning_rate": 0.0001747152953951358,
      "loss": 0.2106,
      "step": 19200
    },
    {
      "epoch": 0.27001194743130225,
      "grad_norm": 0.4477527439594269,
      "learning_rate": 0.00017468439282921584,
      "loss": 0.2028,
      "step": 19210
    },
    {
      "epoch": 0.2701525054466231,
      "grad_norm": 0.4126959443092346,
      "learning_rate": 0.00017465347412702213,
      "loss": 0.1946,
      "step": 19220
    },
    {
      "epoch": 0.2702930634619439,
      "grad_norm": 0.3782075047492981,
      "learning_rate": 0.00017462253929523497,
      "loss": 0.1879,
      "step": 19230
    },
    {
      "epoch": 0.27043362147726474,
      "grad_norm": 0.3895922899246216,
      "learning_rate": 0.00017459158834053814,
      "loss": 0.2412,
      "step": 19240
    },
    {
      "epoch": 0.27057417949258555,
      "grad_norm": 0.45436742901802063,
      "learning_rate": 0.00017456062126961882,
      "loss": 0.2113,
      "step": 19250
    },
    {
      "epoch": 0.27071473750790637,
      "grad_norm": 0.41081178188323975,
      "learning_rate": 0.00017452963808916786,
      "loss": 0.1954,
      "step": 19260
    },
    {
      "epoch": 0.27085529552322724,
      "grad_norm": 0.46563029289245605,
      "learning_rate": 0.00017449863880587932,
      "loss": 0.1945,
      "step": 19270
    },
    {
      "epoch": 0.27099585353854805,
      "grad_norm": 0.44589686393737793,
      "learning_rate": 0.000174467623426451,
      "loss": 0.1926,
      "step": 19280
    },
    {
      "epoch": 0.27113641155386886,
      "grad_norm": 0.481802761554718,
      "learning_rate": 0.000174436591957584,
      "loss": 0.2063,
      "step": 19290
    },
    {
      "epoch": 0.2712769695691897,
      "grad_norm": 0.5556591749191284,
      "learning_rate": 0.00017440554440598306,
      "loss": 0.2105,
      "step": 19300
    },
    {
      "epoch": 0.2714175275845105,
      "grad_norm": 0.6346491575241089,
      "learning_rate": 0.00017437448077835618,
      "loss": 0.2069,
      "step": 19310
    },
    {
      "epoch": 0.27155808559983136,
      "grad_norm": 0.32252898812294006,
      "learning_rate": 0.00017434340108141505,
      "loss": 0.1964,
      "step": 19320
    },
    {
      "epoch": 0.27169864361515217,
      "grad_norm": 0.5045115947723389,
      "learning_rate": 0.00017431230532187467,
      "loss": 0.2325,
      "step": 19330
    },
    {
      "epoch": 0.271839201630473,
      "grad_norm": 0.8819959163665771,
      "learning_rate": 0.00017428119350645363,
      "loss": 0.2,
      "step": 19340
    },
    {
      "epoch": 0.2719797596457938,
      "grad_norm": 0.4297947883605957,
      "learning_rate": 0.00017425006564187386,
      "loss": 0.1934,
      "step": 19350
    },
    {
      "epoch": 0.2721203176611146,
      "grad_norm": 0.4750787913799286,
      "learning_rate": 0.0001742189217348609,
      "loss": 0.205,
      "step": 19360
    },
    {
      "epoch": 0.2722608756764354,
      "grad_norm": 0.3775157034397125,
      "learning_rate": 0.00017418776179214365,
      "loss": 0.1828,
      "step": 19370
    },
    {
      "epoch": 0.2724014336917563,
      "grad_norm": 0.5501800179481506,
      "learning_rate": 0.00017415658582045453,
      "loss": 0.1994,
      "step": 19380
    },
    {
      "epoch": 0.2725419917070771,
      "grad_norm": 0.6898669004440308,
      "learning_rate": 0.0001741253938265294,
      "loss": 0.1744,
      "step": 19390
    },
    {
      "epoch": 0.2726825497223979,
      "grad_norm": 0.49606406688690186,
      "learning_rate": 0.00017409418581710757,
      "loss": 0.1686,
      "step": 19400
    },
    {
      "epoch": 0.27282310773771873,
      "grad_norm": 0.44804930686950684,
      "learning_rate": 0.00017406296179893187,
      "loss": 0.2089,
      "step": 19410
    },
    {
      "epoch": 0.27296366575303954,
      "grad_norm": 0.38459184765815735,
      "learning_rate": 0.00017403172177874848,
      "loss": 0.1954,
      "step": 19420
    },
    {
      "epoch": 0.2731042237683604,
      "grad_norm": 0.48969027400016785,
      "learning_rate": 0.00017400046576330718,
      "loss": 0.2108,
      "step": 19430
    },
    {
      "epoch": 0.2732447817836812,
      "grad_norm": 0.42338326573371887,
      "learning_rate": 0.00017396919375936107,
      "loss": 0.2352,
      "step": 19440
    },
    {
      "epoch": 0.27338533979900204,
      "grad_norm": 0.45828261971473694,
      "learning_rate": 0.0001739379057736668,
      "loss": 0.1963,
      "step": 19450
    },
    {
      "epoch": 0.27352589781432285,
      "grad_norm": 0.42012134194374084,
      "learning_rate": 0.00017390660181298441,
      "loss": 0.1928,
      "step": 19460
    },
    {
      "epoch": 0.27366645582964366,
      "grad_norm": 0.548736572265625,
      "learning_rate": 0.00017387528188407745,
      "loss": 0.2258,
      "step": 19470
    },
    {
      "epoch": 0.27380701384496453,
      "grad_norm": 0.41380250453948975,
      "learning_rate": 0.00017384394599371286,
      "loss": 0.1968,
      "step": 19480
    },
    {
      "epoch": 0.27394757186028534,
      "grad_norm": 0.5325862765312195,
      "learning_rate": 0.00017381259414866107,
      "loss": 0.207,
      "step": 19490
    },
    {
      "epoch": 0.27408812987560616,
      "grad_norm": 0.39338448643684387,
      "learning_rate": 0.00017378122635569593,
      "loss": 0.1751,
      "step": 19500
    },
    {
      "epoch": 0.27422868789092697,
      "grad_norm": 0.4523838460445404,
      "learning_rate": 0.0001737498426215948,
      "loss": 0.1822,
      "step": 19510
    },
    {
      "epoch": 0.2743692459062478,
      "grad_norm": 0.6740010976791382,
      "learning_rate": 0.00017371844295313837,
      "loss": 0.206,
      "step": 19520
    },
    {
      "epoch": 0.27450980392156865,
      "grad_norm": 0.4458725154399872,
      "learning_rate": 0.00017368702735711086,
      "loss": 0.1859,
      "step": 19530
    },
    {
      "epoch": 0.27465036193688946,
      "grad_norm": 0.41209864616394043,
      "learning_rate": 0.00017365559584029994,
      "loss": 0.2018,
      "step": 19540
    },
    {
      "epoch": 0.2747909199522103,
      "grad_norm": 0.38490617275238037,
      "learning_rate": 0.00017362414840949662,
      "loss": 0.1948,
      "step": 19550
    },
    {
      "epoch": 0.2749314779675311,
      "grad_norm": 0.3892601728439331,
      "learning_rate": 0.00017359268507149552,
      "loss": 0.2101,
      "step": 19560
    },
    {
      "epoch": 0.2750720359828519,
      "grad_norm": 0.4824336767196655,
      "learning_rate": 0.00017356120583309447,
      "loss": 0.2029,
      "step": 19570
    },
    {
      "epoch": 0.27521259399817277,
      "grad_norm": 0.3767576813697815,
      "learning_rate": 0.00017352971070109494,
      "loss": 0.1963,
      "step": 19580
    },
    {
      "epoch": 0.2753531520134936,
      "grad_norm": 0.4037289321422577,
      "learning_rate": 0.00017349819968230174,
      "loss": 0.1774,
      "step": 19590
    },
    {
      "epoch": 0.2754937100288144,
      "grad_norm": 0.4326764643192291,
      "learning_rate": 0.00017346667278352312,
      "loss": 0.2109,
      "step": 19600
    },
    {
      "epoch": 0.2756342680441352,
      "grad_norm": 0.39291003346443176,
      "learning_rate": 0.00017343513001157073,
      "loss": 0.1853,
      "step": 19610
    },
    {
      "epoch": 0.275774826059456,
      "grad_norm": 0.4207342267036438,
      "learning_rate": 0.00017340357137325975,
      "loss": 0.2188,
      "step": 19620
    },
    {
      "epoch": 0.27591538407477684,
      "grad_norm": 0.5787175297737122,
      "learning_rate": 0.0001733719968754087,
      "loss": 0.2112,
      "step": 19630
    },
    {
      "epoch": 0.2760559420900977,
      "grad_norm": 0.35075294971466064,
      "learning_rate": 0.00017334040652483958,
      "loss": 0.1989,
      "step": 19640
    },
    {
      "epoch": 0.2761965001054185,
      "grad_norm": 0.5642573833465576,
      "learning_rate": 0.00017330880032837768,
      "loss": 0.2036,
      "step": 19650
    },
    {
      "epoch": 0.27633705812073933,
      "grad_norm": 0.5270206332206726,
      "learning_rate": 0.00017327717829285195,
      "loss": 0.2034,
      "step": 19660
    },
    {
      "epoch": 0.27647761613606014,
      "grad_norm": 0.5013365745544434,
      "learning_rate": 0.00017324554042509455,
      "loss": 0.1999,
      "step": 19670
    },
    {
      "epoch": 0.27661817415138096,
      "grad_norm": 0.41403472423553467,
      "learning_rate": 0.00017321388673194115,
      "loss": 0.2034,
      "step": 19680
    },
    {
      "epoch": 0.2767587321667018,
      "grad_norm": 0.4933346211910248,
      "learning_rate": 0.00017318221722023087,
      "loss": 0.2119,
      "step": 19690
    },
    {
      "epoch": 0.27689929018202264,
      "grad_norm": 0.4709338843822479,
      "learning_rate": 0.00017315053189680618,
      "loss": 0.1887,
      "step": 19700
    },
    {
      "epoch": 0.27703984819734345,
      "grad_norm": 0.5111008882522583,
      "learning_rate": 0.000173118830768513,
      "loss": 0.2051,
      "step": 19710
    },
    {
      "epoch": 0.27718040621266427,
      "grad_norm": 0.5342327952384949,
      "learning_rate": 0.00017308711384220066,
      "loss": 0.2095,
      "step": 19720
    },
    {
      "epoch": 0.2773209642279851,
      "grad_norm": 0.4055479168891907,
      "learning_rate": 0.00017305538112472193,
      "loss": 0.2163,
      "step": 19730
    },
    {
      "epoch": 0.27746152224330595,
      "grad_norm": 0.4838210344314575,
      "learning_rate": 0.00017302363262293292,
      "loss": 0.1924,
      "step": 19740
    },
    {
      "epoch": 0.27760208025862676,
      "grad_norm": 0.4406696557998657,
      "learning_rate": 0.0001729918683436932,
      "loss": 0.2102,
      "step": 19750
    },
    {
      "epoch": 0.2777426382739476,
      "grad_norm": 0.4246480464935303,
      "learning_rate": 0.0001729600882938658,
      "loss": 0.1846,
      "step": 19760
    },
    {
      "epoch": 0.2778831962892684,
      "grad_norm": 0.35406747460365295,
      "learning_rate": 0.00017292829248031705,
      "loss": 0.1803,
      "step": 19770
    },
    {
      "epoch": 0.2780237543045892,
      "grad_norm": 0.4278491735458374,
      "learning_rate": 0.00017289648090991673,
      "loss": 0.2352,
      "step": 19780
    },
    {
      "epoch": 0.27816431231991007,
      "grad_norm": 0.3947303295135498,
      "learning_rate": 0.0001728646535895381,
      "loss": 0.2102,
      "step": 19790
    },
    {
      "epoch": 0.2783048703352309,
      "grad_norm": 0.4705360233783722,
      "learning_rate": 0.00017283281052605766,
      "loss": 0.1874,
      "step": 19800
    },
    {
      "epoch": 0.2784454283505517,
      "grad_norm": 0.48299962282180786,
      "learning_rate": 0.0001728009517263555,
      "loss": 0.19,
      "step": 19810
    },
    {
      "epoch": 0.2785859863658725,
      "grad_norm": 0.5594393014907837,
      "learning_rate": 0.00017276907719731495,
      "loss": 0.2374,
      "step": 19820
    },
    {
      "epoch": 0.2787265443811933,
      "grad_norm": 0.351291686296463,
      "learning_rate": 0.00017273718694582285,
      "loss": 0.2106,
      "step": 19830
    },
    {
      "epoch": 0.2788671023965142,
      "grad_norm": 0.44059497117996216,
      "learning_rate": 0.00017270528097876935,
      "loss": 0.2039,
      "step": 19840
    },
    {
      "epoch": 0.279007660411835,
      "grad_norm": 0.35108813643455505,
      "learning_rate": 0.00017267335930304807,
      "loss": 0.179,
      "step": 19850
    },
    {
      "epoch": 0.2791482184271558,
      "grad_norm": 0.3934147357940674,
      "learning_rate": 0.00017264142192555598,
      "loss": 0.2089,
      "step": 19860
    },
    {
      "epoch": 0.2792887764424766,
      "grad_norm": 0.4623779356479645,
      "learning_rate": 0.00017260946885319344,
      "loss": 0.2022,
      "step": 19870
    },
    {
      "epoch": 0.27942933445779744,
      "grad_norm": 0.9234285354614258,
      "learning_rate": 0.00017257750009286427,
      "loss": 0.1883,
      "step": 19880
    },
    {
      "epoch": 0.27956989247311825,
      "grad_norm": 0.47531360387802124,
      "learning_rate": 0.00017254551565147556,
      "loss": 0.2205,
      "step": 19890
    },
    {
      "epoch": 0.2797104504884391,
      "grad_norm": 0.3452872335910797,
      "learning_rate": 0.0001725135155359379,
      "loss": 0.1775,
      "step": 19900
    },
    {
      "epoch": 0.27985100850375993,
      "grad_norm": 0.4100421369075775,
      "learning_rate": 0.00017248149975316517,
      "loss": 0.2196,
      "step": 19910
    },
    {
      "epoch": 0.27999156651908075,
      "grad_norm": 0.38270118832588196,
      "learning_rate": 0.00017244946831007472,
      "loss": 0.2085,
      "step": 19920
    },
    {
      "epoch": 0.28013212453440156,
      "grad_norm": 0.46484166383743286,
      "learning_rate": 0.00017241742121358727,
      "loss": 0.1854,
      "step": 19930
    },
    {
      "epoch": 0.2802726825497224,
      "grad_norm": 0.3831994831562042,
      "learning_rate": 0.00017238535847062682,
      "loss": 0.2175,
      "step": 19940
    },
    {
      "epoch": 0.28041324056504324,
      "grad_norm": 0.5021612644195557,
      "learning_rate": 0.00017235328008812092,
      "loss": 0.2161,
      "step": 19950
    },
    {
      "epoch": 0.28055379858036406,
      "grad_norm": 0.5603237748146057,
      "learning_rate": 0.00017232118607300037,
      "loss": 0.2288,
      "step": 19960
    },
    {
      "epoch": 0.28069435659568487,
      "grad_norm": 0.3286333680152893,
      "learning_rate": 0.00017228907643219939,
      "loss": 0.1912,
      "step": 19970
    },
    {
      "epoch": 0.2808349146110057,
      "grad_norm": 0.36596253514289856,
      "learning_rate": 0.00017225695117265556,
      "loss": 0.1874,
      "step": 19980
    },
    {
      "epoch": 0.2809754726263265,
      "grad_norm": 0.42818084359169006,
      "learning_rate": 0.00017222481030130987,
      "loss": 0.1769,
      "step": 19990
    },
    {
      "epoch": 0.28111603064164736,
      "grad_norm": 0.6397153735160828,
      "learning_rate": 0.00017219265382510663,
      "loss": 0.2251,
      "step": 20000
    },
    {
      "epoch": 0.28111603064164736,
      "eval_chrf": 74.02033345888455,
      "eval_loss": 0.42471030354499817,
      "eval_runtime": 286.3507,
      "eval_samples_per_second": 0.349,
      "eval_steps_per_second": 0.014,
      "step": 20000
    },
    {
      "epoch": 0.2812565886569682,
      "grad_norm": 0.41157883405685425,
      "learning_rate": 0.00017216048175099357,
      "loss": 0.2134,
      "step": 20010
    },
    {
      "epoch": 0.281397146672289,
      "grad_norm": 0.5360423922538757,
      "learning_rate": 0.0001721282940859218,
      "loss": 0.2334,
      "step": 20020
    },
    {
      "epoch": 0.2815377046876098,
      "grad_norm": 0.45648783445358276,
      "learning_rate": 0.00017209609083684575,
      "loss": 0.1913,
      "step": 20030
    },
    {
      "epoch": 0.2816782627029306,
      "grad_norm": 0.32651346921920776,
      "learning_rate": 0.00017206387201072323,
      "loss": 0.1697,
      "step": 20040
    },
    {
      "epoch": 0.2818188207182515,
      "grad_norm": 0.4762553870677948,
      "learning_rate": 0.00017203163761451545,
      "loss": 0.2135,
      "step": 20050
    },
    {
      "epoch": 0.2819593787335723,
      "grad_norm": 0.5247132182121277,
      "learning_rate": 0.0001719993876551869,
      "loss": 0.2065,
      "step": 20060
    },
    {
      "epoch": 0.2820999367488931,
      "grad_norm": 0.5137999057769775,
      "learning_rate": 0.00017196712213970555,
      "loss": 0.1927,
      "step": 20070
    },
    {
      "epoch": 0.2822404947642139,
      "grad_norm": 0.5455237030982971,
      "learning_rate": 0.00017193484107504267,
      "loss": 0.2025,
      "step": 20080
    },
    {
      "epoch": 0.28238105277953474,
      "grad_norm": 0.37664520740509033,
      "learning_rate": 0.00017190254446817284,
      "loss": 0.2149,
      "step": 20090
    },
    {
      "epoch": 0.2825216107948556,
      "grad_norm": 0.35842660069465637,
      "learning_rate": 0.0001718702323260741,
      "loss": 0.2263,
      "step": 20100
    },
    {
      "epoch": 0.2826621688101764,
      "grad_norm": 0.42714473605155945,
      "learning_rate": 0.00017183790465572779,
      "loss": 0.1862,
      "step": 20110
    },
    {
      "epoch": 0.28280272682549723,
      "grad_norm": 0.5020290613174438,
      "learning_rate": 0.00017180556146411858,
      "loss": 0.1904,
      "step": 20120
    },
    {
      "epoch": 0.28294328484081804,
      "grad_norm": 0.6206112504005432,
      "learning_rate": 0.00017177320275823453,
      "loss": 0.1985,
      "step": 20130
    },
    {
      "epoch": 0.28308384285613886,
      "grad_norm": 0.4329967200756073,
      "learning_rate": 0.00017174082854506706,
      "loss": 0.1947,
      "step": 20140
    },
    {
      "epoch": 0.28322440087145967,
      "grad_norm": 0.3217017650604248,
      "learning_rate": 0.00017170843883161093,
      "loss": 0.1752,
      "step": 20150
    },
    {
      "epoch": 0.28336495888678054,
      "grad_norm": 0.5323647856712341,
      "learning_rate": 0.00017167603362486426,
      "loss": 0.1975,
      "step": 20160
    },
    {
      "epoch": 0.28350551690210135,
      "grad_norm": 0.39277899265289307,
      "learning_rate": 0.0001716436129318285,
      "loss": 0.1985,
      "step": 20170
    },
    {
      "epoch": 0.28364607491742216,
      "grad_norm": 0.3518845736980438,
      "learning_rate": 0.00017161117675950838,
      "loss": 0.2053,
      "step": 20180
    },
    {
      "epoch": 0.283786632932743,
      "grad_norm": 0.3981213867664337,
      "learning_rate": 0.00017157872511491213,
      "loss": 0.197,
      "step": 20190
    },
    {
      "epoch": 0.2839271909480638,
      "grad_norm": 0.41575250029563904,
      "learning_rate": 0.0001715462580050512,
      "loss": 0.2127,
      "step": 20200
    },
    {
      "epoch": 0.28406774896338466,
      "grad_norm": 0.5877156257629395,
      "learning_rate": 0.00017151377543694042,
      "loss": 0.212,
      "step": 20210
    },
    {
      "epoch": 0.28420830697870547,
      "grad_norm": 0.6077449917793274,
      "learning_rate": 0.00017148127741759797,
      "loss": 0.1971,
      "step": 20220
    },
    {
      "epoch": 0.2843488649940263,
      "grad_norm": 0.406627893447876,
      "learning_rate": 0.00017144876395404533,
      "loss": 0.1944,
      "step": 20230
    },
    {
      "epoch": 0.2844894230093471,
      "grad_norm": 0.4879082143306732,
      "learning_rate": 0.00017141623505330738,
      "loss": 0.2024,
      "step": 20240
    },
    {
      "epoch": 0.2846299810246679,
      "grad_norm": 0.3847009539604187,
      "learning_rate": 0.00017138369072241226,
      "loss": 0.2069,
      "step": 20250
    },
    {
      "epoch": 0.2847705390399888,
      "grad_norm": 0.597709596157074,
      "learning_rate": 0.0001713511309683915,
      "loss": 0.1977,
      "step": 20260
    },
    {
      "epoch": 0.2849110970553096,
      "grad_norm": 0.5190564393997192,
      "learning_rate": 0.00017131855579827993,
      "loss": 0.1811,
      "step": 20270
    },
    {
      "epoch": 0.2850516550706304,
      "grad_norm": 0.43286004662513733,
      "learning_rate": 0.00017128596521911579,
      "loss": 0.2033,
      "step": 20280
    },
    {
      "epoch": 0.2851922130859512,
      "grad_norm": 0.380369633436203,
      "learning_rate": 0.00017125335923794048,
      "loss": 0.2236,
      "step": 20290
    },
    {
      "epoch": 0.28533277110127203,
      "grad_norm": 0.46421048045158386,
      "learning_rate": 0.00017122073786179893,
      "loss": 0.2127,
      "step": 20300
    },
    {
      "epoch": 0.2854733291165929,
      "grad_norm": 0.38199180364608765,
      "learning_rate": 0.00017118810109773924,
      "loss": 0.2146,
      "step": 20310
    },
    {
      "epoch": 0.2856138871319137,
      "grad_norm": 0.434123694896698,
      "learning_rate": 0.00017115544895281292,
      "loss": 0.2098,
      "step": 20320
    },
    {
      "epoch": 0.2857544451472345,
      "grad_norm": 0.6215119957923889,
      "learning_rate": 0.00017112278143407473,
      "loss": 0.2239,
      "step": 20330
    },
    {
      "epoch": 0.28589500316255534,
      "grad_norm": 0.3718737065792084,
      "learning_rate": 0.00017109009854858286,
      "loss": 0.1964,
      "step": 20340
    },
    {
      "epoch": 0.28603556117787615,
      "grad_norm": 0.4760577976703644,
      "learning_rate": 0.00017105740030339876,
      "loss": 0.1954,
      "step": 20350
    },
    {
      "epoch": 0.28617611919319697,
      "grad_norm": 0.48852115869522095,
      "learning_rate": 0.00017102468670558712,
      "loss": 0.2071,
      "step": 20360
    },
    {
      "epoch": 0.28631667720851783,
      "grad_norm": 0.31590837240219116,
      "learning_rate": 0.00017099195776221614,
      "loss": 0.1875,
      "step": 20370
    },
    {
      "epoch": 0.28645723522383865,
      "grad_norm": 0.4684780538082123,
      "learning_rate": 0.00017095921348035708,
      "loss": 0.1818,
      "step": 20380
    },
    {
      "epoch": 0.28659779323915946,
      "grad_norm": 0.4114950895309448,
      "learning_rate": 0.00017092645386708477,
      "loss": 0.1767,
      "step": 20390
    },
    {
      "epoch": 0.2867383512544803,
      "grad_norm": 0.5756374001502991,
      "learning_rate": 0.00017089367892947718,
      "loss": 0.2108,
      "step": 20400
    },
    {
      "epoch": 0.2868789092698011,
      "grad_norm": 0.45275378227233887,
      "learning_rate": 0.0001708608886746157,
      "loss": 0.2035,
      "step": 20410
    },
    {
      "epoch": 0.28701946728512195,
      "grad_norm": 0.6170960664749146,
      "learning_rate": 0.0001708280831095849,
      "loss": 0.2236,
      "step": 20420
    },
    {
      "epoch": 0.28716002530044277,
      "grad_norm": 0.5411795377731323,
      "learning_rate": 0.00017079526224147277,
      "loss": 0.2219,
      "step": 20430
    },
    {
      "epoch": 0.2873005833157636,
      "grad_norm": 0.5155619382858276,
      "learning_rate": 0.00017076242607737057,
      "loss": 0.2233,
      "step": 20440
    },
    {
      "epoch": 0.2874411413310844,
      "grad_norm": 0.4209650754928589,
      "learning_rate": 0.0001707295746243729,
      "loss": 0.1939,
      "step": 20450
    },
    {
      "epoch": 0.2875816993464052,
      "grad_norm": 0.5280047655105591,
      "learning_rate": 0.00017069670788957757,
      "loss": 0.2113,
      "step": 20460
    },
    {
      "epoch": 0.2877222573617261,
      "grad_norm": 0.43174266815185547,
      "learning_rate": 0.00017066382588008577,
      "loss": 0.2134,
      "step": 20470
    },
    {
      "epoch": 0.2878628153770469,
      "grad_norm": 0.3827521502971649,
      "learning_rate": 0.000170630928603002,
      "loss": 0.2027,
      "step": 20480
    },
    {
      "epoch": 0.2880033733923677,
      "grad_norm": 0.41761866211891174,
      "learning_rate": 0.000170598016065434,
      "loss": 0.2044,
      "step": 20490
    },
    {
      "epoch": 0.2881439314076885,
      "grad_norm": 0.5618655681610107,
      "learning_rate": 0.00017056508827449288,
      "loss": 0.1894,
      "step": 20500
    },
    {
      "epoch": 0.2882844894230093,
      "grad_norm": 0.564956784248352,
      "learning_rate": 0.0001705321452372929,
      "loss": 0.2218,
      "step": 20510
    },
    {
      "epoch": 0.2884250474383302,
      "grad_norm": 0.5125460624694824,
      "learning_rate": 0.00017049918696095183,
      "loss": 0.1988,
      "step": 20520
    },
    {
      "epoch": 0.288565605453651,
      "grad_norm": 0.40226978063583374,
      "learning_rate": 0.00017046621345259057,
      "loss": 0.2005,
      "step": 20530
    },
    {
      "epoch": 0.2887061634689718,
      "grad_norm": 0.46868157386779785,
      "learning_rate": 0.0001704332247193334,
      "loss": 0.1879,
      "step": 20540
    },
    {
      "epoch": 0.28884672148429263,
      "grad_norm": 0.42704060673713684,
      "learning_rate": 0.00017040022076830774,
      "loss": 0.2064,
      "step": 20550
    },
    {
      "epoch": 0.28898727949961345,
      "grad_norm": 0.4820135235786438,
      "learning_rate": 0.0001703672016066445,
      "loss": 0.1926,
      "step": 20560
    },
    {
      "epoch": 0.2891278375149343,
      "grad_norm": 0.3230167627334595,
      "learning_rate": 0.0001703341672414778,
      "loss": 0.1874,
      "step": 20570
    },
    {
      "epoch": 0.28926839553025513,
      "grad_norm": 0.49147534370422363,
      "learning_rate": 0.00017030111767994498,
      "loss": 0.2336,
      "step": 20580
    },
    {
      "epoch": 0.28940895354557594,
      "grad_norm": 0.5110415816307068,
      "learning_rate": 0.00017026805292918675,
      "loss": 0.1897,
      "step": 20590
    },
    {
      "epoch": 0.28954951156089676,
      "grad_norm": 0.36359795928001404,
      "learning_rate": 0.000170234972996347,
      "loss": 0.1973,
      "step": 20600
    },
    {
      "epoch": 0.28969006957621757,
      "grad_norm": 0.38895362615585327,
      "learning_rate": 0.00017020187788857304,
      "loss": 0.1973,
      "step": 20610
    },
    {
      "epoch": 0.2898306275915384,
      "grad_norm": 0.5448920726776123,
      "learning_rate": 0.00017016876761301533,
      "loss": 0.2058,
      "step": 20620
    },
    {
      "epoch": 0.28997118560685925,
      "grad_norm": 0.39889147877693176,
      "learning_rate": 0.0001701356421768277,
      "loss": 0.1964,
      "step": 20630
    },
    {
      "epoch": 0.29011174362218006,
      "grad_norm": 0.3628638684749603,
      "learning_rate": 0.00017010250158716717,
      "loss": 0.1901,
      "step": 20640
    },
    {
      "epoch": 0.2902523016375009,
      "grad_norm": 0.3719097375869751,
      "learning_rate": 0.0001700693458511941,
      "loss": 0.2037,
      "step": 20650
    },
    {
      "epoch": 0.2903928596528217,
      "grad_norm": 0.39121580123901367,
      "learning_rate": 0.00017003617497607216,
      "loss": 0.2085,
      "step": 20660
    },
    {
      "epoch": 0.2905334176681425,
      "grad_norm": 0.3391832411289215,
      "learning_rate": 0.00017000298896896812,
      "loss": 0.1819,
      "step": 20670
    },
    {
      "epoch": 0.29067397568346337,
      "grad_norm": 0.3420446813106537,
      "learning_rate": 0.00016996978783705223,
      "loss": 0.167,
      "step": 20680
    },
    {
      "epoch": 0.2908145336987842,
      "grad_norm": 0.6026040315628052,
      "learning_rate": 0.00016993657158749783,
      "loss": 0.2142,
      "step": 20690
    },
    {
      "epoch": 0.290955091714105,
      "grad_norm": 0.5469983816146851,
      "learning_rate": 0.0001699033402274817,
      "loss": 0.2045,
      "step": 20700
    },
    {
      "epoch": 0.2910956497294258,
      "grad_norm": 0.41889193654060364,
      "learning_rate": 0.00016987009376418373,
      "loss": 0.2001,
      "step": 20710
    },
    {
      "epoch": 0.2912362077447466,
      "grad_norm": 0.45113030076026917,
      "learning_rate": 0.0001698368322047871,
      "loss": 0.1858,
      "step": 20720
    },
    {
      "epoch": 0.2913767657600675,
      "grad_norm": 0.32660970091819763,
      "learning_rate": 0.0001698035555564784,
      "loss": 0.1993,
      "step": 20730
    },
    {
      "epoch": 0.2915173237753883,
      "grad_norm": 0.4603225588798523,
      "learning_rate": 0.00016977026382644726,
      "loss": 0.2065,
      "step": 20740
    },
    {
      "epoch": 0.2916578817907091,
      "grad_norm": 0.43970394134521484,
      "learning_rate": 0.0001697369570218867,
      "loss": 0.22,
      "step": 20750
    },
    {
      "epoch": 0.29179843980602993,
      "grad_norm": 0.36348602175712585,
      "learning_rate": 0.000169703635149993,
      "loss": 0.1755,
      "step": 20760
    },
    {
      "epoch": 0.29193899782135074,
      "grad_norm": 0.4576697051525116,
      "learning_rate": 0.00016967029821796567,
      "loss": 0.198,
      "step": 20770
    },
    {
      "epoch": 0.2920795558366716,
      "grad_norm": 0.4353881776332855,
      "learning_rate": 0.00016963694623300745,
      "loss": 0.1958,
      "step": 20780
    },
    {
      "epoch": 0.2922201138519924,
      "grad_norm": 0.5406979322433472,
      "learning_rate": 0.00016960357920232434,
      "loss": 0.1949,
      "step": 20790
    },
    {
      "epoch": 0.29236067186731324,
      "grad_norm": 0.4171971082687378,
      "learning_rate": 0.00016957019713312565,
      "loss": 0.1827,
      "step": 20800
    },
    {
      "epoch": 0.29250122988263405,
      "grad_norm": 0.3972976803779602,
      "learning_rate": 0.00016953680003262383,
      "loss": 0.1887,
      "step": 20810
    },
    {
      "epoch": 0.29264178789795486,
      "grad_norm": 0.34832844138145447,
      "learning_rate": 0.00016950338790803473,
      "loss": 0.209,
      "step": 20820
    },
    {
      "epoch": 0.29278234591327573,
      "grad_norm": 0.46556970477104187,
      "learning_rate": 0.00016946996076657727,
      "loss": 0.2036,
      "step": 20830
    },
    {
      "epoch": 0.29292290392859655,
      "grad_norm": 0.3853689432144165,
      "learning_rate": 0.00016943651861547375,
      "loss": 0.1986,
      "step": 20840
    },
    {
      "epoch": 0.29306346194391736,
      "grad_norm": 0.371705561876297,
      "learning_rate": 0.00016940306146194964,
      "loss": 0.1914,
      "step": 20850
    },
    {
      "epoch": 0.29320401995923817,
      "grad_norm": 0.3775120973587036,
      "learning_rate": 0.0001693695893132337,
      "loss": 0.2176,
      "step": 20860
    },
    {
      "epoch": 0.293344577974559,
      "grad_norm": 0.5206791162490845,
      "learning_rate": 0.0001693361021765579,
      "loss": 0.2141,
      "step": 20870
    },
    {
      "epoch": 0.2934851359898798,
      "grad_norm": 0.38626107573509216,
      "learning_rate": 0.00016930260005915742,
      "loss": 0.1928,
      "step": 20880
    },
    {
      "epoch": 0.29362569400520067,
      "grad_norm": 0.45110735297203064,
      "learning_rate": 0.00016926908296827078,
      "loss": 0.2122,
      "step": 20890
    },
    {
      "epoch": 0.2937662520205215,
      "grad_norm": 0.36143726110458374,
      "learning_rate": 0.00016923555091113957,
      "loss": 0.1837,
      "step": 20900
    },
    {
      "epoch": 0.2939068100358423,
      "grad_norm": 0.39780905842781067,
      "learning_rate": 0.00016920200389500885,
      "loss": 0.2166,
      "step": 20910
    },
    {
      "epoch": 0.2940473680511631,
      "grad_norm": 0.38996177911758423,
      "learning_rate": 0.00016916844192712662,
      "loss": 0.1919,
      "step": 20920
    },
    {
      "epoch": 0.2941879260664839,
      "grad_norm": 0.4880051910877228,
      "learning_rate": 0.00016913486501474436,
      "loss": 0.1985,
      "step": 20930
    },
    {
      "epoch": 0.2943284840818048,
      "grad_norm": 0.4125623106956482,
      "learning_rate": 0.00016910127316511664,
      "loss": 0.1913,
      "step": 20940
    },
    {
      "epoch": 0.2944690420971256,
      "grad_norm": 0.4753319025039673,
      "learning_rate": 0.00016906766638550134,
      "loss": 0.1957,
      "step": 20950
    },
    {
      "epoch": 0.2946096001124464,
      "grad_norm": 0.5684584975242615,
      "learning_rate": 0.0001690340446831595,
      "loss": 0.2181,
      "step": 20960
    },
    {
      "epoch": 0.2947501581277672,
      "grad_norm": 0.4074452221393585,
      "learning_rate": 0.00016900040806535537,
      "loss": 0.199,
      "step": 20970
    },
    {
      "epoch": 0.29489071614308804,
      "grad_norm": 0.42722028493881226,
      "learning_rate": 0.00016896675653935652,
      "loss": 0.2169,
      "step": 20980
    },
    {
      "epoch": 0.2950312741584089,
      "grad_norm": 0.5126949548721313,
      "learning_rate": 0.00016893309011243365,
      "loss": 0.1987,
      "step": 20990
    },
    {
      "epoch": 0.2951718321737297,
      "grad_norm": 0.6361300945281982,
      "learning_rate": 0.00016889940879186074,
      "loss": 0.2296,
      "step": 21000
    },
    {
      "epoch": 0.2951718321737297,
      "eval_chrf": 81.46533911299353,
      "eval_loss": 0.42531388998031616,
      "eval_runtime": 210.6293,
      "eval_samples_per_second": 0.475,
      "eval_steps_per_second": 0.019,
      "step": 21000
    },
    {
      "epoch": 0.29531239018905053,
      "grad_norm": 0.32635369896888733,
      "learning_rate": 0.0001688657125849149,
      "loss": 0.1957,
      "step": 21010
    },
    {
      "epoch": 0.29545294820437135,
      "grad_norm": 0.4398081600666046,
      "learning_rate": 0.0001688320014988766,
      "loss": 0.2096,
      "step": 21020
    },
    {
      "epoch": 0.29559350621969216,
      "grad_norm": 0.3880372941493988,
      "learning_rate": 0.00016879827554102937,
      "loss": 0.2124,
      "step": 21030
    },
    {
      "epoch": 0.295734064235013,
      "grad_norm": 0.3721957802772522,
      "learning_rate": 0.00016876453471866007,
      "loss": 0.2062,
      "step": 21040
    },
    {
      "epoch": 0.29587462225033384,
      "grad_norm": 0.4164586365222931,
      "learning_rate": 0.0001687307790390587,
      "loss": 0.1824,
      "step": 21050
    },
    {
      "epoch": 0.29601518026565465,
      "grad_norm": 0.4159533977508545,
      "learning_rate": 0.00016869700850951852,
      "loss": 0.1961,
      "step": 21060
    },
    {
      "epoch": 0.29615573828097547,
      "grad_norm": 0.368132084608078,
      "learning_rate": 0.00016866322313733594,
      "loss": 0.2022,
      "step": 21070
    },
    {
      "epoch": 0.2962962962962963,
      "grad_norm": 0.44499659538269043,
      "learning_rate": 0.0001686294229298106,
      "loss": 0.2088,
      "step": 21080
    },
    {
      "epoch": 0.2964368543116171,
      "grad_norm": 0.392949640750885,
      "learning_rate": 0.00016859560789424545,
      "loss": 0.1717,
      "step": 21090
    },
    {
      "epoch": 0.29657741232693796,
      "grad_norm": 0.33532652258872986,
      "learning_rate": 0.00016856177803794645,
      "loss": 0.2105,
      "step": 21100
    },
    {
      "epoch": 0.2967179703422588,
      "grad_norm": 0.425544798374176,
      "learning_rate": 0.0001685279333682229,
      "loss": 0.1839,
      "step": 21110
    },
    {
      "epoch": 0.2968585283575796,
      "grad_norm": 0.4223870038986206,
      "learning_rate": 0.00016849407389238726,
      "loss": 0.2251,
      "step": 21120
    },
    {
      "epoch": 0.2969990863729004,
      "grad_norm": 0.38955533504486084,
      "learning_rate": 0.00016846019961775518,
      "loss": 0.2222,
      "step": 21130
    },
    {
      "epoch": 0.2971396443882212,
      "grad_norm": 0.39352136850357056,
      "learning_rate": 0.00016842631055164557,
      "loss": 0.2088,
      "step": 21140
    },
    {
      "epoch": 0.2972802024035421,
      "grad_norm": 0.3265032172203064,
      "learning_rate": 0.0001683924067013804,
      "loss": 0.2031,
      "step": 21150
    },
    {
      "epoch": 0.2974207604188629,
      "grad_norm": 0.5114708542823792,
      "learning_rate": 0.000168358488074285,
      "loss": 0.2096,
      "step": 21160
    },
    {
      "epoch": 0.2975613184341837,
      "grad_norm": 0.3312976658344269,
      "learning_rate": 0.00016832455467768778,
      "loss": 0.1852,
      "step": 21170
    },
    {
      "epoch": 0.2977018764495045,
      "grad_norm": 0.3760843276977539,
      "learning_rate": 0.00016829060651892035,
      "loss": 0.2057,
      "step": 21180
    },
    {
      "epoch": 0.29784243446482533,
      "grad_norm": 0.4861770570278168,
      "learning_rate": 0.0001682566436053176,
      "loss": 0.1934,
      "step": 21190
    },
    {
      "epoch": 0.2979829924801462,
      "grad_norm": 0.44579216837882996,
      "learning_rate": 0.00016822266594421746,
      "loss": 0.2367,
      "step": 21200
    },
    {
      "epoch": 0.298123550495467,
      "grad_norm": 0.4577611982822418,
      "learning_rate": 0.00016818867354296115,
      "loss": 0.2103,
      "step": 21210
    },
    {
      "epoch": 0.29826410851078783,
      "grad_norm": 0.4669945538043976,
      "learning_rate": 0.00016815466640889308,
      "loss": 0.2096,
      "step": 21220
    },
    {
      "epoch": 0.29840466652610864,
      "grad_norm": 0.4052143096923828,
      "learning_rate": 0.0001681206445493608,
      "loss": 0.207,
      "step": 21230
    },
    {
      "epoch": 0.29854522454142945,
      "grad_norm": 0.4139108955860138,
      "learning_rate": 0.00016808660797171506,
      "loss": 0.1877,
      "step": 21240
    },
    {
      "epoch": 0.2986857825567503,
      "grad_norm": 0.441862016916275,
      "learning_rate": 0.00016805255668330977,
      "loss": 0.212,
      "step": 21250
    },
    {
      "epoch": 0.29882634057207114,
      "grad_norm": 0.5393145084381104,
      "learning_rate": 0.00016801849069150206,
      "loss": 0.2141,
      "step": 21260
    },
    {
      "epoch": 0.29896689858739195,
      "grad_norm": 0.7432650923728943,
      "learning_rate": 0.0001679844100036522,
      "loss": 0.2141,
      "step": 21270
    },
    {
      "epoch": 0.29910745660271276,
      "grad_norm": 0.38837116956710815,
      "learning_rate": 0.00016795031462712366,
      "loss": 0.1961,
      "step": 21280
    },
    {
      "epoch": 0.2992480146180336,
      "grad_norm": 0.4575008451938629,
      "learning_rate": 0.00016791620456928303,
      "loss": 0.2002,
      "step": 21290
    },
    {
      "epoch": 0.29938857263335444,
      "grad_norm": 0.5123000741004944,
      "learning_rate": 0.00016788207983750014,
      "loss": 0.2003,
      "step": 21300
    },
    {
      "epoch": 0.29952913064867526,
      "grad_norm": 0.4672884941101074,
      "learning_rate": 0.00016784794043914796,
      "loss": 0.217,
      "step": 21310
    },
    {
      "epoch": 0.29966968866399607,
      "grad_norm": 0.3814888894557953,
      "learning_rate": 0.00016781378638160266,
      "loss": 0.1916,
      "step": 21320
    },
    {
      "epoch": 0.2998102466793169,
      "grad_norm": 0.3043074905872345,
      "learning_rate": 0.0001677796176722435,
      "loss": 0.1816,
      "step": 21330
    },
    {
      "epoch": 0.2999508046946377,
      "grad_norm": 0.5070516467094421,
      "learning_rate": 0.000167745434318453,
      "loss": 0.2197,
      "step": 21340
    },
    {
      "epoch": 0.3000913627099585,
      "grad_norm": 0.39147183299064636,
      "learning_rate": 0.00016771123632761676,
      "loss": 0.1823,
      "step": 21350
    },
    {
      "epoch": 0.3002319207252794,
      "grad_norm": 0.3833473026752472,
      "learning_rate": 0.00016767702370712359,
      "loss": 0.206,
      "step": 21360
    },
    {
      "epoch": 0.3003724787406002,
      "grad_norm": 0.5475829243659973,
      "learning_rate": 0.00016764279646436548,
      "loss": 0.2121,
      "step": 21370
    },
    {
      "epoch": 0.300513036755921,
      "grad_norm": 0.6117001175880432,
      "learning_rate": 0.0001676085546067375,
      "loss": 0.195,
      "step": 21380
    },
    {
      "epoch": 0.3006535947712418,
      "grad_norm": 0.45678287744522095,
      "learning_rate": 0.00016757429814163798,
      "loss": 0.1876,
      "step": 21390
    },
    {
      "epoch": 0.30079415278656263,
      "grad_norm": 0.4021288752555847,
      "learning_rate": 0.00016754002707646839,
      "loss": 0.2076,
      "step": 21400
    },
    {
      "epoch": 0.3009347108018835,
      "grad_norm": 0.267534464597702,
      "learning_rate": 0.0001675057414186332,
      "loss": 0.211,
      "step": 21410
    },
    {
      "epoch": 0.3010752688172043,
      "grad_norm": 0.33864277601242065,
      "learning_rate": 0.00016747144117554023,
      "loss": 0.2011,
      "step": 21420
    },
    {
      "epoch": 0.3012158268325251,
      "grad_norm": 0.4673684239387512,
      "learning_rate": 0.00016743712635460036,
      "loss": 0.1935,
      "step": 21430
    },
    {
      "epoch": 0.30135638484784594,
      "grad_norm": 0.5295349359512329,
      "learning_rate": 0.00016740279696322765,
      "loss": 0.1794,
      "step": 21440
    },
    {
      "epoch": 0.30149694286316675,
      "grad_norm": 0.4514896273612976,
      "learning_rate": 0.00016736845300883925,
      "loss": 0.1848,
      "step": 21450
    },
    {
      "epoch": 0.3016375008784876,
      "grad_norm": 0.44729170203208923,
      "learning_rate": 0.00016733409449885557,
      "loss": 0.1871,
      "step": 21460
    },
    {
      "epoch": 0.30177805889380843,
      "grad_norm": 0.46859967708587646,
      "learning_rate": 0.00016729972144070003,
      "loss": 0.2038,
      "step": 21470
    },
    {
      "epoch": 0.30191861690912924,
      "grad_norm": 0.5687006711959839,
      "learning_rate": 0.00016726533384179925,
      "loss": 0.1793,
      "step": 21480
    },
    {
      "epoch": 0.30205917492445006,
      "grad_norm": 0.4455181062221527,
      "learning_rate": 0.00016723093170958303,
      "loss": 0.1949,
      "step": 21490
    },
    {
      "epoch": 0.30219973293977087,
      "grad_norm": 0.4142560362815857,
      "learning_rate": 0.0001671965150514843,
      "loss": 0.1794,
      "step": 21500
    },
    {
      "epoch": 0.30234029095509174,
      "grad_norm": 0.5188212990760803,
      "learning_rate": 0.00016716208387493908,
      "loss": 0.1994,
      "step": 21510
    },
    {
      "epoch": 0.30248084897041255,
      "grad_norm": 0.44499433040618896,
      "learning_rate": 0.00016712763818738652,
      "loss": 0.2142,
      "step": 21520
    },
    {
      "epoch": 0.30262140698573337,
      "grad_norm": 0.34323349595069885,
      "learning_rate": 0.00016709317799626898,
      "loss": 0.2045,
      "step": 21530
    },
    {
      "epoch": 0.3027619650010542,
      "grad_norm": 0.43302005529403687,
      "learning_rate": 0.0001670587033090319,
      "loss": 0.175,
      "step": 21540
    },
    {
      "epoch": 0.302902523016375,
      "grad_norm": 0.45775189995765686,
      "learning_rate": 0.00016702421413312388,
      "loss": 0.2127,
      "step": 21550
    },
    {
      "epoch": 0.30304308103169586,
      "grad_norm": 0.5039337873458862,
      "learning_rate": 0.00016698971047599662,
      "loss": 0.1861,
      "step": 21560
    },
    {
      "epoch": 0.3031836390470167,
      "grad_norm": 0.45097997784614563,
      "learning_rate": 0.00016695519234510492,
      "loss": 0.2329,
      "step": 21570
    },
    {
      "epoch": 0.3033241970623375,
      "grad_norm": 0.41815900802612305,
      "learning_rate": 0.00016692065974790684,
      "loss": 0.1934,
      "step": 21580
    },
    {
      "epoch": 0.3034647550776583,
      "grad_norm": 0.42974695563316345,
      "learning_rate": 0.00016688611269186342,
      "loss": 0.1921,
      "step": 21590
    },
    {
      "epoch": 0.3036053130929791,
      "grad_norm": 0.4625963866710663,
      "learning_rate": 0.0001668515511844389,
      "loss": 0.2079,
      "step": 21600
    },
    {
      "epoch": 0.3037458711082999,
      "grad_norm": 0.3194527328014374,
      "learning_rate": 0.0001668169752331006,
      "loss": 0.2104,
      "step": 21610
    },
    {
      "epoch": 0.3038864291236208,
      "grad_norm": 0.4360635280609131,
      "learning_rate": 0.000166782384845319,
      "loss": 0.1933,
      "step": 21620
    },
    {
      "epoch": 0.3040269871389416,
      "grad_norm": 0.2982389032840729,
      "learning_rate": 0.0001667477800285677,
      "loss": 0.1846,
      "step": 21630
    },
    {
      "epoch": 0.3041675451542624,
      "grad_norm": 0.43925002217292786,
      "learning_rate": 0.00016671316079032336,
      "loss": 0.2029,
      "step": 21640
    },
    {
      "epoch": 0.30430810316958323,
      "grad_norm": 0.41043299436569214,
      "learning_rate": 0.0001666785271380658,
      "loss": 0.1902,
      "step": 21650
    },
    {
      "epoch": 0.30444866118490405,
      "grad_norm": 0.4733089506626129,
      "learning_rate": 0.000166643879079278,
      "loss": 0.1974,
      "step": 21660
    },
    {
      "epoch": 0.3045892192002249,
      "grad_norm": 0.4703904092311859,
      "learning_rate": 0.00016660921662144595,
      "loss": 0.2195,
      "step": 21670
    },
    {
      "epoch": 0.3047297772155457,
      "grad_norm": 0.4680962860584259,
      "learning_rate": 0.0001665745397720588,
      "loss": 0.2057,
      "step": 21680
    },
    {
      "epoch": 0.30487033523086654,
      "grad_norm": 0.37620243430137634,
      "learning_rate": 0.00016653984853860887,
      "loss": 0.2152,
      "step": 21690
    },
    {
      "epoch": 0.30501089324618735,
      "grad_norm": 0.5280198454856873,
      "learning_rate": 0.00016650514292859147,
      "loss": 0.1821,
      "step": 21700
    },
    {
      "epoch": 0.30515145126150817,
      "grad_norm": 0.5200461745262146,
      "learning_rate": 0.00016647042294950511,
      "loss": 0.1995,
      "step": 21710
    },
    {
      "epoch": 0.30529200927682903,
      "grad_norm": 0.5603528618812561,
      "learning_rate": 0.00016643568860885138,
      "loss": 0.2014,
      "step": 21720
    },
    {
      "epoch": 0.30543256729214985,
      "grad_norm": 0.4702509045600891,
      "learning_rate": 0.00016640093991413495,
      "loss": 0.2164,
      "step": 21730
    },
    {
      "epoch": 0.30557312530747066,
      "grad_norm": 0.4287772476673126,
      "learning_rate": 0.0001663661768728636,
      "loss": 0.2112,
      "step": 21740
    },
    {
      "epoch": 0.3057136833227915,
      "grad_norm": 0.3885875642299652,
      "learning_rate": 0.0001663313994925482,
      "loss": 0.2374,
      "step": 21750
    },
    {
      "epoch": 0.3058542413381123,
      "grad_norm": 0.5625565648078918,
      "learning_rate": 0.0001662966077807028,
      "loss": 0.2035,
      "step": 21760
    },
    {
      "epoch": 0.30599479935343316,
      "grad_norm": 0.5335406064987183,
      "learning_rate": 0.00016626180174484444,
      "loss": 0.2292,
      "step": 21770
    },
    {
      "epoch": 0.30613535736875397,
      "grad_norm": 0.5790080428123474,
      "learning_rate": 0.00016622698139249324,
      "loss": 0.189,
      "step": 21780
    },
    {
      "epoch": 0.3062759153840748,
      "grad_norm": 0.4857686758041382,
      "learning_rate": 0.00016619214673117252,
      "loss": 0.1889,
      "step": 21790
    },
    {
      "epoch": 0.3064164733993956,
      "grad_norm": 0.34772664308547974,
      "learning_rate": 0.0001661572977684087,
      "loss": 0.1984,
      "step": 21800
    },
    {
      "epoch": 0.3065570314147164,
      "grad_norm": 0.493201345205307,
      "learning_rate": 0.00016612243451173114,
      "loss": 0.1846,
      "step": 21810
    },
    {
      "epoch": 0.3066975894300372,
      "grad_norm": 0.5512494444847107,
      "learning_rate": 0.0001660875569686724,
      "loss": 0.1978,
      "step": 21820
    },
    {
      "epoch": 0.3068381474453581,
      "grad_norm": 0.5157384276390076,
      "learning_rate": 0.00016605266514676812,
      "loss": 0.1975,
      "step": 21830
    },
    {
      "epoch": 0.3069787054606789,
      "grad_norm": 0.4055367708206177,
      "learning_rate": 0.00016601775905355698,
      "loss": 0.1948,
      "step": 21840
    },
    {
      "epoch": 0.3071192634759997,
      "grad_norm": 0.43613380193710327,
      "learning_rate": 0.00016598283869658084,
      "loss": 0.2258,
      "step": 21850
    },
    {
      "epoch": 0.30725982149132053,
      "grad_norm": 0.4905271828174591,
      "learning_rate": 0.00016594790408338452,
      "loss": 0.1999,
      "step": 21860
    },
    {
      "epoch": 0.30740037950664134,
      "grad_norm": 0.5527341365814209,
      "learning_rate": 0.00016591295522151597,
      "loss": 0.1974,
      "step": 21870
    },
    {
      "epoch": 0.3075409375219622,
      "grad_norm": 0.47549405694007874,
      "learning_rate": 0.0001658779921185263,
      "loss": 0.2088,
      "step": 21880
    },
    {
      "epoch": 0.307681495537283,
      "grad_norm": 0.46096235513687134,
      "learning_rate": 0.00016584301478196948,
      "loss": 0.2046,
      "step": 21890
    },
    {
      "epoch": 0.30782205355260384,
      "grad_norm": 0.44452401995658875,
      "learning_rate": 0.00016580802321940284,
      "loss": 0.1835,
      "step": 21900
    },
    {
      "epoch": 0.30796261156792465,
      "grad_norm": 0.3749406039714813,
      "learning_rate": 0.00016577301743838656,
      "loss": 0.2006,
      "step": 21910
    },
    {
      "epoch": 0.30810316958324546,
      "grad_norm": 0.45477259159088135,
      "learning_rate": 0.00016573799744648397,
      "loss": 0.2271,
      "step": 21920
    },
    {
      "epoch": 0.30824372759856633,
      "grad_norm": 0.34558719396591187,
      "learning_rate": 0.00016570296325126154,
      "loss": 0.1862,
      "step": 21930
    },
    {
      "epoch": 0.30838428561388714,
      "grad_norm": 0.4179009795188904,
      "learning_rate": 0.00016566791486028867,
      "loss": 0.2325,
      "step": 21940
    },
    {
      "epoch": 0.30852484362920796,
      "grad_norm": 0.38069191575050354,
      "learning_rate": 0.00016563285228113792,
      "loss": 0.1955,
      "step": 21950
    },
    {
      "epoch": 0.30866540164452877,
      "grad_norm": 0.4004470705986023,
      "learning_rate": 0.0001655977755213849,
      "loss": 0.2304,
      "step": 21960
    },
    {
      "epoch": 0.3088059596598496,
      "grad_norm": 0.3554079532623291,
      "learning_rate": 0.00016556268458860828,
      "loss": 0.1909,
      "step": 21970
    },
    {
      "epoch": 0.30894651767517045,
      "grad_norm": 0.5444430708885193,
      "learning_rate": 0.00016552757949038979,
      "loss": 0.2086,
      "step": 21980
    },
    {
      "epoch": 0.30908707569049126,
      "grad_norm": 0.3903445899486542,
      "learning_rate": 0.00016549246023431422,
      "loss": 0.18,
      "step": 21990
    },
    {
      "epoch": 0.3092276337058121,
      "grad_norm": 0.5591972470283508,
      "learning_rate": 0.0001654573268279694,
      "loss": 0.196,
      "step": 22000
    },
    {
      "epoch": 0.3092276337058121,
      "eval_chrf": 76.0635875128844,
      "eval_loss": 0.42267727851867676,
      "eval_runtime": 326.6993,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 22000
    },
    {
      "epoch": 0.3093681917211329,
      "grad_norm": 0.46631136536598206,
      "learning_rate": 0.00016542217927894626,
      "loss": 0.2193,
      "step": 22010
    },
    {
      "epoch": 0.3095087497364537,
      "grad_norm": 0.5263903737068176,
      "learning_rate": 0.0001653870175948388,
      "loss": 0.1821,
      "step": 22020
    },
    {
      "epoch": 0.30964930775177457,
      "grad_norm": 0.28334668278694153,
      "learning_rate": 0.00016535184178324394,
      "loss": 0.2041,
      "step": 22030
    },
    {
      "epoch": 0.3097898657670954,
      "grad_norm": 0.3982601463794708,
      "learning_rate": 0.0001653166518517618,
      "loss": 0.1877,
      "step": 22040
    },
    {
      "epoch": 0.3099304237824162,
      "grad_norm": 0.5098924040794373,
      "learning_rate": 0.0001652814478079955,
      "loss": 0.1948,
      "step": 22050
    },
    {
      "epoch": 0.310070981797737,
      "grad_norm": 0.47173330187797546,
      "learning_rate": 0.00016524622965955122,
      "loss": 0.1939,
      "step": 22060
    },
    {
      "epoch": 0.3102115398130578,
      "grad_norm": 0.41165921092033386,
      "learning_rate": 0.00016521099741403822,
      "loss": 0.1911,
      "step": 22070
    },
    {
      "epoch": 0.31035209782837864,
      "grad_norm": 0.43187540769577026,
      "learning_rate": 0.0001651757510790686,
      "loss": 0.1961,
      "step": 22080
    },
    {
      "epoch": 0.3104926558436995,
      "grad_norm": 0.4152243435382843,
      "learning_rate": 0.00016514049066225785,
      "loss": 0.187,
      "step": 22090
    },
    {
      "epoch": 0.3106332138590203,
      "grad_norm": 0.616346001625061,
      "learning_rate": 0.0001651052161712242,
      "loss": 0.21,
      "step": 22100
    },
    {
      "epoch": 0.31077377187434113,
      "grad_norm": 0.4170876145362854,
      "learning_rate": 0.0001650699276135891,
      "loss": 0.1979,
      "step": 22110
    },
    {
      "epoch": 0.31091432988966194,
      "grad_norm": 0.39913642406463623,
      "learning_rate": 0.00016503462499697696,
      "loss": 0.1885,
      "step": 22120
    },
    {
      "epoch": 0.31105488790498276,
      "grad_norm": 0.43161845207214355,
      "learning_rate": 0.00016499930832901524,
      "loss": 0.1961,
      "step": 22130
    },
    {
      "epoch": 0.3111954459203036,
      "grad_norm": 0.5345866084098816,
      "learning_rate": 0.00016496397761733443,
      "loss": 0.2122,
      "step": 22140
    },
    {
      "epoch": 0.31133600393562444,
      "grad_norm": 0.45740818977355957,
      "learning_rate": 0.00016492863286956812,
      "loss": 0.2093,
      "step": 22150
    },
    {
      "epoch": 0.31147656195094525,
      "grad_norm": 0.4239158630371094,
      "learning_rate": 0.00016489327409335283,
      "loss": 0.1962,
      "step": 22160
    },
    {
      "epoch": 0.31161711996626607,
      "grad_norm": 0.4945349097251892,
      "learning_rate": 0.00016485790129632818,
      "loss": 0.1856,
      "step": 22170
    },
    {
      "epoch": 0.3117576779815869,
      "grad_norm": 0.33270493149757385,
      "learning_rate": 0.0001648225144861368,
      "loss": 0.2224,
      "step": 22180
    },
    {
      "epoch": 0.31189823599690775,
      "grad_norm": 0.5921101570129395,
      "learning_rate": 0.0001647871136704243,
      "loss": 0.2017,
      "step": 22190
    },
    {
      "epoch": 0.31203879401222856,
      "grad_norm": 0.6349021196365356,
      "learning_rate": 0.00016475169885683943,
      "loss": 0.1913,
      "step": 22200
    },
    {
      "epoch": 0.3121793520275494,
      "grad_norm": 0.294806569814682,
      "learning_rate": 0.00016471627005303386,
      "loss": 0.1838,
      "step": 22210
    },
    {
      "epoch": 0.3123199100428702,
      "grad_norm": 0.4655342996120453,
      "learning_rate": 0.00016468082726666235,
      "loss": 0.202,
      "step": 22220
    },
    {
      "epoch": 0.312460468058191,
      "grad_norm": 0.40321406722068787,
      "learning_rate": 0.0001646453705053826,
      "loss": 0.183,
      "step": 22230
    },
    {
      "epoch": 0.31260102607351187,
      "grad_norm": 0.421733558177948,
      "learning_rate": 0.00016460989977685545,
      "loss": 0.2133,
      "step": 22240
    },
    {
      "epoch": 0.3127415840888327,
      "grad_norm": 0.4640405774116516,
      "learning_rate": 0.00016457441508874464,
      "loss": 0.2006,
      "step": 22250
    },
    {
      "epoch": 0.3128821421041535,
      "grad_norm": 0.4528983235359192,
      "learning_rate": 0.00016453891644871695,
      "loss": 0.1965,
      "step": 22260
    },
    {
      "epoch": 0.3130227001194743,
      "grad_norm": 0.39258962869644165,
      "learning_rate": 0.00016450340386444228,
      "loss": 0.1737,
      "step": 22270
    },
    {
      "epoch": 0.3131632581347951,
      "grad_norm": 0.3652934432029724,
      "learning_rate": 0.0001644678773435934,
      "loss": 0.1846,
      "step": 22280
    },
    {
      "epoch": 0.313303816150116,
      "grad_norm": 0.5782024264335632,
      "learning_rate": 0.00016443233689384615,
      "loss": 0.1864,
      "step": 22290
    },
    {
      "epoch": 0.3134443741654368,
      "grad_norm": 0.3768521249294281,
      "learning_rate": 0.00016439678252287942,
      "loss": 0.2079,
      "step": 22300
    },
    {
      "epoch": 0.3135849321807576,
      "grad_norm": 0.4075048863887787,
      "learning_rate": 0.00016436121423837505,
      "loss": 0.2033,
      "step": 22310
    },
    {
      "epoch": 0.3137254901960784,
      "grad_norm": 0.38486969470977783,
      "learning_rate": 0.0001643256320480179,
      "loss": 0.1918,
      "step": 22320
    },
    {
      "epoch": 0.31386604821139924,
      "grad_norm": 0.43265506625175476,
      "learning_rate": 0.0001642900359594959,
      "loss": 0.1837,
      "step": 22330
    },
    {
      "epoch": 0.31400660622672005,
      "grad_norm": 0.5715003609657288,
      "learning_rate": 0.0001642544259804998,
      "loss": 0.1797,
      "step": 22340
    },
    {
      "epoch": 0.3141471642420409,
      "grad_norm": 0.49120792746543884,
      "learning_rate": 0.00016421880211872357,
      "loss": 0.2218,
      "step": 22350
    },
    {
      "epoch": 0.31428772225736173,
      "grad_norm": 0.4198523461818695,
      "learning_rate": 0.00016418316438186406,
      "loss": 0.1843,
      "step": 22360
    },
    {
      "epoch": 0.31442828027268255,
      "grad_norm": 0.4386689066886902,
      "learning_rate": 0.00016414751277762116,
      "loss": 0.1927,
      "step": 22370
    },
    {
      "epoch": 0.31456883828800336,
      "grad_norm": 0.4620615839958191,
      "learning_rate": 0.0001641118473136977,
      "loss": 0.1942,
      "step": 22380
    },
    {
      "epoch": 0.3147093963033242,
      "grad_norm": 0.2823697626590729,
      "learning_rate": 0.0001640761679977996,
      "loss": 0.1989,
      "step": 22390
    },
    {
      "epoch": 0.31484995431864504,
      "grad_norm": 0.510749876499176,
      "learning_rate": 0.00016404047483763564,
      "loss": 0.2058,
      "step": 22400
    },
    {
      "epoch": 0.31499051233396586,
      "grad_norm": 0.4054532051086426,
      "learning_rate": 0.00016400476784091773,
      "loss": 0.1961,
      "step": 22410
    },
    {
      "epoch": 0.31513107034928667,
      "grad_norm": 0.36367928981781006,
      "learning_rate": 0.0001639690470153607,
      "loss": 0.196,
      "step": 22420
    },
    {
      "epoch": 0.3152716283646075,
      "grad_norm": 0.41342687606811523,
      "learning_rate": 0.00016393331236868233,
      "loss": 0.1961,
      "step": 22430
    },
    {
      "epoch": 0.3154121863799283,
      "grad_norm": 0.4696900546550751,
      "learning_rate": 0.00016389756390860347,
      "loss": 0.2123,
      "step": 22440
    },
    {
      "epoch": 0.31555274439524916,
      "grad_norm": 0.5743334293365479,
      "learning_rate": 0.0001638618016428479,
      "loss": 0.212,
      "step": 22450
    },
    {
      "epoch": 0.31569330241057,
      "grad_norm": 0.3811054527759552,
      "learning_rate": 0.0001638260255791424,
      "loss": 0.1841,
      "step": 22460
    },
    {
      "epoch": 0.3158338604258908,
      "grad_norm": 0.3848494589328766,
      "learning_rate": 0.00016379023572521675,
      "loss": 0.2067,
      "step": 22470
    },
    {
      "epoch": 0.3159744184412116,
      "grad_norm": 0.4317486882209778,
      "learning_rate": 0.00016375443208880362,
      "loss": 0.2067,
      "step": 22480
    },
    {
      "epoch": 0.3161149764565324,
      "grad_norm": 0.40094897150993347,
      "learning_rate": 0.00016371861467763882,
      "loss": 0.2024,
      "step": 22490
    },
    {
      "epoch": 0.3162555344718533,
      "grad_norm": 0.4053865671157837,
      "learning_rate": 0.00016368278349946098,
      "loss": 0.19,
      "step": 22500
    },
    {
      "epoch": 0.3163960924871741,
      "grad_norm": 0.37913885712623596,
      "learning_rate": 0.0001636469385620118,
      "loss": 0.1952,
      "step": 22510
    },
    {
      "epoch": 0.3165366505024949,
      "grad_norm": 0.48451685905456543,
      "learning_rate": 0.00016361107987303587,
      "loss": 0.2111,
      "step": 22520
    },
    {
      "epoch": 0.3166772085178157,
      "grad_norm": 0.3487234115600586,
      "learning_rate": 0.0001635752074402809,
      "loss": 0.1847,
      "step": 22530
    },
    {
      "epoch": 0.31681776653313654,
      "grad_norm": 0.5989207625389099,
      "learning_rate": 0.00016353932127149734,
      "loss": 0.2311,
      "step": 22540
    },
    {
      "epoch": 0.31695832454845735,
      "grad_norm": 0.3807673752307892,
      "learning_rate": 0.00016350342137443882,
      "loss": 0.1984,
      "step": 22550
    },
    {
      "epoch": 0.3170988825637782,
      "grad_norm": 0.3877153694629669,
      "learning_rate": 0.00016346750775686185,
      "loss": 0.1846,
      "step": 22560
    },
    {
      "epoch": 0.31723944057909903,
      "grad_norm": 0.42989611625671387,
      "learning_rate": 0.00016343158042652592,
      "loss": 0.214,
      "step": 22570
    },
    {
      "epoch": 0.31737999859441984,
      "grad_norm": 0.3693827986717224,
      "learning_rate": 0.0001633956393911934,
      "loss": 0.23,
      "step": 22580
    },
    {
      "epoch": 0.31752055660974066,
      "grad_norm": 0.6093205213546753,
      "learning_rate": 0.00016335968465862978,
      "loss": 0.2042,
      "step": 22590
    },
    {
      "epoch": 0.31766111462506147,
      "grad_norm": 0.3517461121082306,
      "learning_rate": 0.0001633237162366034,
      "loss": 0.2055,
      "step": 22600
    },
    {
      "epoch": 0.31780167264038234,
      "grad_norm": 0.3896740972995758,
      "learning_rate": 0.00016328773413288553,
      "loss": 0.1931,
      "step": 22610
    },
    {
      "epoch": 0.31794223065570315,
      "grad_norm": 0.4707591235637665,
      "learning_rate": 0.00016325173835525045,
      "loss": 0.1996,
      "step": 22620
    },
    {
      "epoch": 0.31808278867102396,
      "grad_norm": 0.44279807806015015,
      "learning_rate": 0.00016321572891147548,
      "loss": 0.191,
      "step": 22630
    },
    {
      "epoch": 0.3182233466863448,
      "grad_norm": 0.5250387787818909,
      "learning_rate": 0.0001631797058093407,
      "loss": 0.2373,
      "step": 22640
    },
    {
      "epoch": 0.3183639047016656,
      "grad_norm": 0.5072187185287476,
      "learning_rate": 0.0001631436690566293,
      "loss": 0.199,
      "step": 22650
    },
    {
      "epoch": 0.31850446271698646,
      "grad_norm": 0.5126092433929443,
      "learning_rate": 0.00016310761866112733,
      "loss": 0.1883,
      "step": 22660
    },
    {
      "epoch": 0.31864502073230727,
      "grad_norm": 0.5698773264884949,
      "learning_rate": 0.00016307155463062388,
      "loss": 0.1965,
      "step": 22670
    },
    {
      "epoch": 0.3187855787476281,
      "grad_norm": 0.4435856342315674,
      "learning_rate": 0.0001630354769729109,
      "loss": 0.2086,
      "step": 22680
    },
    {
      "epoch": 0.3189261367629489,
      "grad_norm": 0.3472186326980591,
      "learning_rate": 0.00016299938569578327,
      "loss": 0.1974,
      "step": 22690
    },
    {
      "epoch": 0.3190666947782697,
      "grad_norm": 0.3711005449295044,
      "learning_rate": 0.00016296328080703885,
      "loss": 0.1728,
      "step": 22700
    },
    {
      "epoch": 0.3192072527935906,
      "grad_norm": 0.37461093068122864,
      "learning_rate": 0.00016292716231447853,
      "loss": 0.212,
      "step": 22710
    },
    {
      "epoch": 0.3193478108089114,
      "grad_norm": 0.47322550415992737,
      "learning_rate": 0.000162891030225906,
      "loss": 0.205,
      "step": 22720
    },
    {
      "epoch": 0.3194883688242322,
      "grad_norm": 0.3658617436885834,
      "learning_rate": 0.0001628548845491279,
      "loss": 0.1726,
      "step": 22730
    },
    {
      "epoch": 0.319628926839553,
      "grad_norm": 0.44577306509017944,
      "learning_rate": 0.00016281872529195392,
      "loss": 0.2265,
      "step": 22740
    },
    {
      "epoch": 0.31976948485487383,
      "grad_norm": 0.4455224871635437,
      "learning_rate": 0.0001627825524621966,
      "loss": 0.2256,
      "step": 22750
    },
    {
      "epoch": 0.3199100428701947,
      "grad_norm": 0.5204285979270935,
      "learning_rate": 0.0001627463660676714,
      "loss": 0.1902,
      "step": 22760
    },
    {
      "epoch": 0.3200506008855155,
      "grad_norm": 0.387105256319046,
      "learning_rate": 0.00016271016611619672,
      "loss": 0.1894,
      "step": 22770
    },
    {
      "epoch": 0.3201911589008363,
      "grad_norm": 0.4587792158126831,
      "learning_rate": 0.00016267395261559394,
      "loss": 0.1799,
      "step": 22780
    },
    {
      "epoch": 0.32033171691615714,
      "grad_norm": 0.400437593460083,
      "learning_rate": 0.00016263772557368732,
      "loss": 0.1872,
      "step": 22790
    },
    {
      "epoch": 0.32047227493147795,
      "grad_norm": 0.501432478427887,
      "learning_rate": 0.00016260148499830403,
      "loss": 0.1977,
      "step": 22800
    },
    {
      "epoch": 0.32061283294679876,
      "grad_norm": 0.5811870694160461,
      "learning_rate": 0.00016256523089727421,
      "loss": 0.2002,
      "step": 22810
    },
    {
      "epoch": 0.32075339096211963,
      "grad_norm": 0.46564650535583496,
      "learning_rate": 0.00016252896327843095,
      "loss": 0.1865,
      "step": 22820
    },
    {
      "epoch": 0.32089394897744045,
      "grad_norm": 0.4727911949157715,
      "learning_rate": 0.00016249268214961018,
      "loss": 0.1926,
      "step": 22830
    },
    {
      "epoch": 0.32103450699276126,
      "grad_norm": 0.3758562505245209,
      "learning_rate": 0.00016245638751865075,
      "loss": 0.2167,
      "step": 22840
    },
    {
      "epoch": 0.3211750650080821,
      "grad_norm": 0.4082559049129486,
      "learning_rate": 0.00016242007939339445,
      "loss": 0.195,
      "step": 22850
    },
    {
      "epoch": 0.3213156230234029,
      "grad_norm": 0.466960608959198,
      "learning_rate": 0.00016238375778168605,
      "loss": 0.2011,
      "step": 22860
    },
    {
      "epoch": 0.32145618103872375,
      "grad_norm": 0.4250938296318054,
      "learning_rate": 0.00016234742269137322,
      "loss": 0.1948,
      "step": 22870
    },
    {
      "epoch": 0.32159673905404457,
      "grad_norm": 0.4879426658153534,
      "learning_rate": 0.00016231107413030638,
      "loss": 0.1942,
      "step": 22880
    },
    {
      "epoch": 0.3217372970693654,
      "grad_norm": 0.3468765914440155,
      "learning_rate": 0.00016227471210633908,
      "loss": 0.1897,
      "step": 22890
    },
    {
      "epoch": 0.3218778550846862,
      "grad_norm": 0.40255075693130493,
      "learning_rate": 0.00016223833662732764,
      "loss": 0.2037,
      "step": 22900
    },
    {
      "epoch": 0.322018413100007,
      "grad_norm": 0.4981006979942322,
      "learning_rate": 0.00016220194770113133,
      "loss": 0.1837,
      "step": 22910
    },
    {
      "epoch": 0.3221589711153279,
      "grad_norm": 0.4926389753818512,
      "learning_rate": 0.0001621655453356123,
      "loss": 0.2107,
      "step": 22920
    },
    {
      "epoch": 0.3222995291306487,
      "grad_norm": 0.45715492963790894,
      "learning_rate": 0.0001621291295386357,
      "loss": 0.1927,
      "step": 22930
    },
    {
      "epoch": 0.3224400871459695,
      "grad_norm": 0.5912772417068481,
      "learning_rate": 0.00016209270031806948,
      "loss": 0.1956,
      "step": 22940
    },
    {
      "epoch": 0.3225806451612903,
      "grad_norm": 0.3169713020324707,
      "learning_rate": 0.00016205625768178446,
      "loss": 0.17,
      "step": 22950
    },
    {
      "epoch": 0.3227212031766111,
      "grad_norm": 0.5209740400314331,
      "learning_rate": 0.0001620198016376545,
      "loss": 0.1947,
      "step": 22960
    },
    {
      "epoch": 0.322861761191932,
      "grad_norm": 0.5017517805099487,
      "learning_rate": 0.00016198333219355626,
      "loss": 0.1828,
      "step": 22970
    },
    {
      "epoch": 0.3230023192072528,
      "grad_norm": 0.4529948830604553,
      "learning_rate": 0.00016194684935736927,
      "loss": 0.1842,
      "step": 22980
    },
    {
      "epoch": 0.3231428772225736,
      "grad_norm": 0.526214063167572,
      "learning_rate": 0.000161910353136976,
      "loss": 0.2203,
      "step": 22990
    },
    {
      "epoch": 0.32328343523789443,
      "grad_norm": 0.42212504148483276,
      "learning_rate": 0.0001618738435402618,
      "loss": 0.2013,
      "step": 23000
    },
    {
      "epoch": 0.32328343523789443,
      "eval_chrf": 82.48321766359608,
      "eval_loss": 0.3963998854160309,
      "eval_runtime": 196.6304,
      "eval_samples_per_second": 0.509,
      "eval_steps_per_second": 0.02,
      "step": 23000
    },
    {
      "epoch": 0.32342399325321525,
      "grad_norm": 0.5202959179878235,
      "learning_rate": 0.00016183732057511497,
      "loss": 0.1968,
      "step": 23010
    },
    {
      "epoch": 0.3235645512685361,
      "grad_norm": 0.5392760634422302,
      "learning_rate": 0.0001618007842494266,
      "loss": 0.2024,
      "step": 23020
    },
    {
      "epoch": 0.32370510928385693,
      "grad_norm": 0.4830007553100586,
      "learning_rate": 0.00016176423457109072,
      "loss": 0.1902,
      "step": 23030
    },
    {
      "epoch": 0.32384566729917774,
      "grad_norm": 0.5701043009757996,
      "learning_rate": 0.00016172767154800424,
      "loss": 0.1675,
      "step": 23040
    },
    {
      "epoch": 0.32398622531449855,
      "grad_norm": 0.48150068521499634,
      "learning_rate": 0.00016169109518806692,
      "loss": 0.2048,
      "step": 23050
    },
    {
      "epoch": 0.32412678332981937,
      "grad_norm": 0.47341838479042053,
      "learning_rate": 0.0001616545054991815,
      "loss": 0.2221,
      "step": 23060
    },
    {
      "epoch": 0.3242673413451402,
      "grad_norm": 0.5987292528152466,
      "learning_rate": 0.0001616179024892535,
      "loss": 0.2002,
      "step": 23070
    },
    {
      "epoch": 0.32440789936046105,
      "grad_norm": 0.46618038415908813,
      "learning_rate": 0.0001615812861661913,
      "loss": 0.2115,
      "step": 23080
    },
    {
      "epoch": 0.32454845737578186,
      "grad_norm": 0.3018644452095032,
      "learning_rate": 0.00016154465653790625,
      "loss": 0.2034,
      "step": 23090
    },
    {
      "epoch": 0.3246890153911027,
      "grad_norm": 0.3897569477558136,
      "learning_rate": 0.00016150801361231252,
      "loss": 0.1973,
      "step": 23100
    },
    {
      "epoch": 0.3248295734064235,
      "grad_norm": 0.3934481739997864,
      "learning_rate": 0.00016147135739732715,
      "loss": 0.2161,
      "step": 23110
    },
    {
      "epoch": 0.3249701314217443,
      "grad_norm": 0.4347870349884033,
      "learning_rate": 0.0001614346879008701,
      "loss": 0.1901,
      "step": 23120
    },
    {
      "epoch": 0.32511068943706517,
      "grad_norm": 0.5074248909950256,
      "learning_rate": 0.00016139800513086412,
      "loss": 0.2003,
      "step": 23130
    },
    {
      "epoch": 0.325251247452386,
      "grad_norm": 0.4701947867870331,
      "learning_rate": 0.0001613613090952349,
      "loss": 0.1855,
      "step": 23140
    },
    {
      "epoch": 0.3253918054677068,
      "grad_norm": 0.38716921210289,
      "learning_rate": 0.00016132459980191098,
      "loss": 0.1892,
      "step": 23150
    },
    {
      "epoch": 0.3255323634830276,
      "grad_norm": 0.4173943102359772,
      "learning_rate": 0.0001612878772588237,
      "loss": 0.1923,
      "step": 23160
    },
    {
      "epoch": 0.3256729214983484,
      "grad_norm": 0.43605056405067444,
      "learning_rate": 0.00016125114147390733,
      "loss": 0.1913,
      "step": 23170
    },
    {
      "epoch": 0.3258134795136693,
      "grad_norm": 0.5023015141487122,
      "learning_rate": 0.00016121439245509906,
      "loss": 0.211,
      "step": 23180
    },
    {
      "epoch": 0.3259540375289901,
      "grad_norm": 0.4328177571296692,
      "learning_rate": 0.00016117763021033875,
      "loss": 0.2187,
      "step": 23190
    },
    {
      "epoch": 0.3260945955443109,
      "grad_norm": 0.3364429473876953,
      "learning_rate": 0.00016114085474756932,
      "loss": 0.1864,
      "step": 23200
    },
    {
      "epoch": 0.32623515355963173,
      "grad_norm": 0.5123487114906311,
      "learning_rate": 0.00016110406607473644,
      "loss": 0.2139,
      "step": 23210
    },
    {
      "epoch": 0.32637571157495254,
      "grad_norm": 0.40395841002464294,
      "learning_rate": 0.00016106726419978866,
      "loss": 0.2131,
      "step": 23220
    },
    {
      "epoch": 0.3265162695902734,
      "grad_norm": 0.7314327359199524,
      "learning_rate": 0.00016103044913067733,
      "loss": 0.1898,
      "step": 23230
    },
    {
      "epoch": 0.3266568276055942,
      "grad_norm": 0.45684218406677246,
      "learning_rate": 0.00016099362087535672,
      "loss": 0.1983,
      "step": 23240
    },
    {
      "epoch": 0.32679738562091504,
      "grad_norm": 0.43260887265205383,
      "learning_rate": 0.00016095677944178392,
      "loss": 0.1891,
      "step": 23250
    },
    {
      "epoch": 0.32693794363623585,
      "grad_norm": 0.5483604073524475,
      "learning_rate": 0.0001609199248379189,
      "loss": 0.218,
      "step": 23260
    },
    {
      "epoch": 0.32707850165155666,
      "grad_norm": 0.34057939052581787,
      "learning_rate": 0.00016088305707172442,
      "loss": 0.1746,
      "step": 23270
    },
    {
      "epoch": 0.3272190596668775,
      "grad_norm": 0.3196142911911011,
      "learning_rate": 0.00016084617615116613,
      "loss": 0.1825,
      "step": 23280
    },
    {
      "epoch": 0.32735961768219835,
      "grad_norm": 0.40307173132896423,
      "learning_rate": 0.00016080928208421247,
      "loss": 0.2045,
      "step": 23290
    },
    {
      "epoch": 0.32750017569751916,
      "grad_norm": 0.470195472240448,
      "learning_rate": 0.00016077237487883481,
      "loss": 0.1935,
      "step": 23300
    },
    {
      "epoch": 0.32764073371283997,
      "grad_norm": 0.34664785861968994,
      "learning_rate": 0.00016073545454300728,
      "loss": 0.1741,
      "step": 23310
    },
    {
      "epoch": 0.3277812917281608,
      "grad_norm": 0.3692355155944824,
      "learning_rate": 0.00016069852108470682,
      "loss": 0.2,
      "step": 23320
    },
    {
      "epoch": 0.3279218497434816,
      "grad_norm": 0.4574620723724365,
      "learning_rate": 0.00016066157451191332,
      "loss": 0.1738,
      "step": 23330
    },
    {
      "epoch": 0.32806240775880247,
      "grad_norm": 0.44873225688934326,
      "learning_rate": 0.0001606246148326094,
      "loss": 0.1918,
      "step": 23340
    },
    {
      "epoch": 0.3282029657741233,
      "grad_norm": 0.4090384244918823,
      "learning_rate": 0.00016058764205478058,
      "loss": 0.1681,
      "step": 23350
    },
    {
      "epoch": 0.3283435237894441,
      "grad_norm": 0.4433750510215759,
      "learning_rate": 0.00016055065618641515,
      "loss": 0.1936,
      "step": 23360
    },
    {
      "epoch": 0.3284840818047649,
      "grad_norm": 0.29744261503219604,
      "learning_rate": 0.0001605136572355043,
      "loss": 0.1776,
      "step": 23370
    },
    {
      "epoch": 0.3286246398200857,
      "grad_norm": 0.4242916703224182,
      "learning_rate": 0.00016047664521004194,
      "loss": 0.1839,
      "step": 23380
    },
    {
      "epoch": 0.3287651978354066,
      "grad_norm": 0.5000902414321899,
      "learning_rate": 0.00016043962011802495,
      "loss": 0.2317,
      "step": 23390
    },
    {
      "epoch": 0.3289057558507274,
      "grad_norm": 0.4689231216907501,
      "learning_rate": 0.00016040258196745293,
      "loss": 0.2173,
      "step": 23400
    },
    {
      "epoch": 0.3290463138660482,
      "grad_norm": 0.5616615414619446,
      "learning_rate": 0.00016036553076632828,
      "loss": 0.2117,
      "step": 23410
    },
    {
      "epoch": 0.329186871881369,
      "grad_norm": 0.43278253078460693,
      "learning_rate": 0.00016032846652265628,
      "loss": 0.23,
      "step": 23420
    },
    {
      "epoch": 0.32932742989668984,
      "grad_norm": 0.46418872475624084,
      "learning_rate": 0.00016029138924444504,
      "loss": 0.1797,
      "step": 23430
    },
    {
      "epoch": 0.3294679879120107,
      "grad_norm": 0.5142697691917419,
      "learning_rate": 0.00016025429893970548,
      "loss": 0.2135,
      "step": 23440
    },
    {
      "epoch": 0.3296085459273315,
      "grad_norm": 0.456232488155365,
      "learning_rate": 0.00016021719561645122,
      "loss": 0.2199,
      "step": 23450
    },
    {
      "epoch": 0.32974910394265233,
      "grad_norm": 0.4834028482437134,
      "learning_rate": 0.00016018007928269887,
      "loss": 0.2194,
      "step": 23460
    },
    {
      "epoch": 0.32988966195797315,
      "grad_norm": 0.343547523021698,
      "learning_rate": 0.00016014294994646773,
      "loss": 0.2129,
      "step": 23470
    },
    {
      "epoch": 0.33003021997329396,
      "grad_norm": 0.5281736254692078,
      "learning_rate": 0.00016010580761577998,
      "loss": 0.2313,
      "step": 23480
    },
    {
      "epoch": 0.3301707779886148,
      "grad_norm": 0.4694022834300995,
      "learning_rate": 0.00016006865229866053,
      "loss": 0.1975,
      "step": 23490
    },
    {
      "epoch": 0.33031133600393564,
      "grad_norm": 0.449082612991333,
      "learning_rate": 0.00016003148400313717,
      "loss": 0.1962,
      "step": 23500
    },
    {
      "epoch": 0.33045189401925645,
      "grad_norm": 0.5389946699142456,
      "learning_rate": 0.00015999430273724044,
      "loss": 0.1958,
      "step": 23510
    },
    {
      "epoch": 0.33059245203457727,
      "grad_norm": 0.38287481665611267,
      "learning_rate": 0.00015995710850900374,
      "loss": 0.1983,
      "step": 23520
    },
    {
      "epoch": 0.3307330100498981,
      "grad_norm": 0.42514464259147644,
      "learning_rate": 0.0001599199013264632,
      "loss": 0.1976,
      "step": 23530
    },
    {
      "epoch": 0.3308735680652189,
      "grad_norm": 0.3916068375110626,
      "learning_rate": 0.00015988268119765782,
      "loss": 0.1936,
      "step": 23540
    },
    {
      "epoch": 0.33101412608053976,
      "grad_norm": 0.5027921795845032,
      "learning_rate": 0.00015984544813062935,
      "loss": 0.2461,
      "step": 23550
    },
    {
      "epoch": 0.3311546840958606,
      "grad_norm": 0.47344645857810974,
      "learning_rate": 0.00015980820213342236,
      "loss": 0.194,
      "step": 23560
    },
    {
      "epoch": 0.3312952421111814,
      "grad_norm": 0.5741089582443237,
      "learning_rate": 0.00015977094321408414,
      "loss": 0.1883,
      "step": 23570
    },
    {
      "epoch": 0.3314358001265022,
      "grad_norm": 0.526816189289093,
      "learning_rate": 0.00015973367138066494,
      "loss": 0.1959,
      "step": 23580
    },
    {
      "epoch": 0.331576358141823,
      "grad_norm": 0.37461012601852417,
      "learning_rate": 0.0001596963866412176,
      "loss": 0.1948,
      "step": 23590
    },
    {
      "epoch": 0.3317169161571439,
      "grad_norm": 0.4073886573314667,
      "learning_rate": 0.00015965908900379789,
      "loss": 0.1879,
      "step": 23600
    },
    {
      "epoch": 0.3318574741724647,
      "grad_norm": 0.4429916441440582,
      "learning_rate": 0.0001596217784764643,
      "loss": 0.2059,
      "step": 23610
    },
    {
      "epoch": 0.3319980321877855,
      "grad_norm": 0.4292476773262024,
      "learning_rate": 0.00015958445506727813,
      "loss": 0.1914,
      "step": 23620
    },
    {
      "epoch": 0.3321385902031063,
      "grad_norm": 0.5539444088935852,
      "learning_rate": 0.00015954711878430347,
      "loss": 0.1895,
      "step": 23630
    },
    {
      "epoch": 0.33227914821842713,
      "grad_norm": 0.39998406171798706,
      "learning_rate": 0.00015950976963560718,
      "loss": 0.2033,
      "step": 23640
    },
    {
      "epoch": 0.332419706233748,
      "grad_norm": 0.3286818563938141,
      "learning_rate": 0.00015947240762925885,
      "loss": 0.217,
      "step": 23650
    },
    {
      "epoch": 0.3325602642490688,
      "grad_norm": 0.34060147404670715,
      "learning_rate": 0.00015943503277333099,
      "loss": 0.1989,
      "step": 23660
    },
    {
      "epoch": 0.33270082226438963,
      "grad_norm": 0.36114442348480225,
      "learning_rate": 0.0001593976450758987,
      "loss": 0.1967,
      "step": 23670
    },
    {
      "epoch": 0.33284138027971044,
      "grad_norm": 0.37886273860931396,
      "learning_rate": 0.00015936024454504,
      "loss": 0.2272,
      "step": 23680
    },
    {
      "epoch": 0.33298193829503125,
      "grad_norm": 0.3490985333919525,
      "learning_rate": 0.00015932283118883563,
      "loss": 0.1915,
      "step": 23690
    },
    {
      "epoch": 0.3331224963103521,
      "grad_norm": 0.4528276026248932,
      "learning_rate": 0.00015928540501536907,
      "loss": 0.1972,
      "step": 23700
    },
    {
      "epoch": 0.33326305432567294,
      "grad_norm": 0.3115861415863037,
      "learning_rate": 0.00015924796603272663,
      "loss": 0.1785,
      "step": 23710
    },
    {
      "epoch": 0.33340361234099375,
      "grad_norm": 0.42265215516090393,
      "learning_rate": 0.00015921051424899732,
      "loss": 0.2127,
      "step": 23720
    },
    {
      "epoch": 0.33354417035631456,
      "grad_norm": 0.4673636853694916,
      "learning_rate": 0.00015917304967227302,
      "loss": 0.2088,
      "step": 23730
    },
    {
      "epoch": 0.3336847283716354,
      "grad_norm": 0.5135847926139832,
      "learning_rate": 0.00015913557231064823,
      "loss": 0.22,
      "step": 23740
    },
    {
      "epoch": 0.33382528638695624,
      "grad_norm": 0.37671685218811035,
      "learning_rate": 0.0001590980821722204,
      "loss": 0.1973,
      "step": 23750
    },
    {
      "epoch": 0.33396584440227706,
      "grad_norm": 0.33548229932785034,
      "learning_rate": 0.00015906057926508952,
      "loss": 0.1891,
      "step": 23760
    },
    {
      "epoch": 0.33410640241759787,
      "grad_norm": 0.37380048632621765,
      "learning_rate": 0.0001590230635973585,
      "loss": 0.2028,
      "step": 23770
    },
    {
      "epoch": 0.3342469604329187,
      "grad_norm": 0.5409539937973022,
      "learning_rate": 0.00015898553517713294,
      "loss": 0.1939,
      "step": 23780
    },
    {
      "epoch": 0.3343875184482395,
      "grad_norm": 0.32086920738220215,
      "learning_rate": 0.00015894799401252124,
      "loss": 0.187,
      "step": 23790
    },
    {
      "epoch": 0.3345280764635603,
      "grad_norm": 0.4053663909435272,
      "learning_rate": 0.00015891044011163448,
      "loss": 0.1905,
      "step": 23800
    },
    {
      "epoch": 0.3346686344788812,
      "grad_norm": 0.5001649856567383,
      "learning_rate": 0.00015887287348258656,
      "loss": 0.1967,
      "step": 23810
    },
    {
      "epoch": 0.334809192494202,
      "grad_norm": 0.3653586804866791,
      "learning_rate": 0.00015883529413349417,
      "loss": 0.2088,
      "step": 23820
    },
    {
      "epoch": 0.3349497505095228,
      "grad_norm": 0.4791796803474426,
      "learning_rate": 0.00015879770207247656,
      "loss": 0.1996,
      "step": 23830
    },
    {
      "epoch": 0.3350903085248436,
      "grad_norm": 0.3865216374397278,
      "learning_rate": 0.00015876009730765594,
      "loss": 0.1929,
      "step": 23840
    },
    {
      "epoch": 0.33523086654016443,
      "grad_norm": 0.4039939045906067,
      "learning_rate": 0.0001587224798471572,
      "loss": 0.1838,
      "step": 23850
    },
    {
      "epoch": 0.3353714245554853,
      "grad_norm": 0.5174484848976135,
      "learning_rate": 0.0001586848496991079,
      "loss": 0.1938,
      "step": 23860
    },
    {
      "epoch": 0.3355119825708061,
      "grad_norm": 0.48277246952056885,
      "learning_rate": 0.00015864720687163835,
      "loss": 0.1937,
      "step": 23870
    },
    {
      "epoch": 0.3356525405861269,
      "grad_norm": 0.4065185785293579,
      "learning_rate": 0.00015860955137288172,
      "loss": 0.2079,
      "step": 23880
    },
    {
      "epoch": 0.33579309860144774,
      "grad_norm": 0.5326710343360901,
      "learning_rate": 0.00015857188321097382,
      "loss": 0.1969,
      "step": 23890
    },
    {
      "epoch": 0.33593365661676855,
      "grad_norm": 0.4714176058769226,
      "learning_rate": 0.0001585342023940532,
      "loss": 0.1886,
      "step": 23900
    },
    {
      "epoch": 0.3360742146320894,
      "grad_norm": 0.4024302661418915,
      "learning_rate": 0.00015849650893026116,
      "loss": 0.193,
      "step": 23910
    },
    {
      "epoch": 0.33621477264741023,
      "grad_norm": 0.521590530872345,
      "learning_rate": 0.00015845880282774175,
      "loss": 0.2082,
      "step": 23920
    },
    {
      "epoch": 0.33635533066273104,
      "grad_norm": 0.3961153030395508,
      "learning_rate": 0.0001584210840946417,
      "loss": 0.1984,
      "step": 23930
    },
    {
      "epoch": 0.33649588867805186,
      "grad_norm": 0.4426018297672272,
      "learning_rate": 0.00015838335273911052,
      "loss": 0.2076,
      "step": 23940
    },
    {
      "epoch": 0.33663644669337267,
      "grad_norm": 0.38262760639190674,
      "learning_rate": 0.00015834560876930043,
      "loss": 0.1729,
      "step": 23950
    },
    {
      "epoch": 0.33677700470869354,
      "grad_norm": 0.4940842092037201,
      "learning_rate": 0.00015830785219336638,
      "loss": 0.206,
      "step": 23960
    },
    {
      "epoch": 0.33691756272401435,
      "grad_norm": 0.4233616888523102,
      "learning_rate": 0.00015827008301946602,
      "loss": 0.1832,
      "step": 23970
    },
    {
      "epoch": 0.33705812073933517,
      "grad_norm": 0.4347827434539795,
      "learning_rate": 0.00015823230125575977,
      "loss": 0.2158,
      "step": 23980
    },
    {
      "epoch": 0.337198678754656,
      "grad_norm": 0.39352595806121826,
      "learning_rate": 0.0001581945069104107,
      "loss": 0.2111,
      "step": 23990
    },
    {
      "epoch": 0.3373392367699768,
      "grad_norm": 0.3998149633407593,
      "learning_rate": 0.00015815669999158467,
      "loss": 0.2143,
      "step": 24000
    },
    {
      "epoch": 0.3373392367699768,
      "eval_chrf": 80.06792337784833,
      "eval_loss": 0.4090176224708557,
      "eval_runtime": 326.915,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 24000
    },
    {
      "epoch": 0.3374797947852976,
      "grad_norm": 0.39046767354011536,
      "learning_rate": 0.0001581188805074502,
      "loss": 0.1933,
      "step": 24010
    },
    {
      "epoch": 0.3376203528006185,
      "grad_norm": 0.3566581904888153,
      "learning_rate": 0.00015808104846617855,
      "loss": 0.1996,
      "step": 24020
    },
    {
      "epoch": 0.3377609108159393,
      "grad_norm": 0.3693186938762665,
      "learning_rate": 0.0001580432038759437,
      "loss": 0.2069,
      "step": 24030
    },
    {
      "epoch": 0.3379014688312601,
      "grad_norm": 0.4311753213405609,
      "learning_rate": 0.00015800534674492237,
      "loss": 0.2165,
      "step": 24040
    },
    {
      "epoch": 0.3380420268465809,
      "grad_norm": 0.42389604449272156,
      "learning_rate": 0.00015796747708129386,
      "loss": 0.2079,
      "step": 24050
    },
    {
      "epoch": 0.3381825848619017,
      "grad_norm": 0.42120248079299927,
      "learning_rate": 0.0001579295948932404,
      "loss": 0.1859,
      "step": 24060
    },
    {
      "epoch": 0.3383231428772226,
      "grad_norm": 0.35434579849243164,
      "learning_rate": 0.0001578917001889467,
      "loss": 0.1832,
      "step": 24070
    },
    {
      "epoch": 0.3384637008925434,
      "grad_norm": 0.45823779702186584,
      "learning_rate": 0.00015785379297660027,
      "loss": 0.1804,
      "step": 24080
    },
    {
      "epoch": 0.3386042589078642,
      "grad_norm": 0.41358405351638794,
      "learning_rate": 0.0001578158732643914,
      "loss": 0.2048,
      "step": 24090
    },
    {
      "epoch": 0.33874481692318503,
      "grad_norm": 0.45774754881858826,
      "learning_rate": 0.00015777794106051297,
      "loss": 0.1947,
      "step": 24100
    },
    {
      "epoch": 0.33888537493850585,
      "grad_norm": 0.593467652797699,
      "learning_rate": 0.00015773999637316052,
      "loss": 0.212,
      "step": 24110
    },
    {
      "epoch": 0.3390259329538267,
      "grad_norm": 0.5359540581703186,
      "learning_rate": 0.00015770203921053246,
      "loss": 0.1842,
      "step": 24120
    },
    {
      "epoch": 0.3391664909691475,
      "grad_norm": 0.32489413022994995,
      "learning_rate": 0.00015766406958082975,
      "loss": 0.2051,
      "step": 24130
    },
    {
      "epoch": 0.33930704898446834,
      "grad_norm": 0.45257705450057983,
      "learning_rate": 0.0001576260874922561,
      "loss": 0.1957,
      "step": 24140
    },
    {
      "epoch": 0.33944760699978915,
      "grad_norm": 0.34853655099868774,
      "learning_rate": 0.00015758809295301788,
      "loss": 0.1677,
      "step": 24150
    },
    {
      "epoch": 0.33958816501510997,
      "grad_norm": 0.49031469225883484,
      "learning_rate": 0.00015755008597132422,
      "loss": 0.2104,
      "step": 24160
    },
    {
      "epoch": 0.33972872303043083,
      "grad_norm": 0.47045207023620605,
      "learning_rate": 0.00015751206655538686,
      "loss": 0.1913,
      "step": 24170
    },
    {
      "epoch": 0.33986928104575165,
      "grad_norm": 0.5129997730255127,
      "learning_rate": 0.00015747403471342027,
      "loss": 0.1976,
      "step": 24180
    },
    {
      "epoch": 0.34000983906107246,
      "grad_norm": 0.4752309322357178,
      "learning_rate": 0.00015743599045364153,
      "loss": 0.2212,
      "step": 24190
    },
    {
      "epoch": 0.3401503970763933,
      "grad_norm": 0.41055867075920105,
      "learning_rate": 0.00015739793378427057,
      "loss": 0.1898,
      "step": 24200
    },
    {
      "epoch": 0.3402909550917141,
      "grad_norm": 0.3874673545360565,
      "learning_rate": 0.00015735986471352982,
      "loss": 0.1825,
      "step": 24210
    },
    {
      "epoch": 0.34043151310703496,
      "grad_norm": 0.335651695728302,
      "learning_rate": 0.00015732178324964448,
      "loss": 0.2047,
      "step": 24220
    },
    {
      "epoch": 0.34057207112235577,
      "grad_norm": 0.4309629201889038,
      "learning_rate": 0.0001572836894008424,
      "loss": 0.1568,
      "step": 24230
    },
    {
      "epoch": 0.3407126291376766,
      "grad_norm": 0.4961472749710083,
      "learning_rate": 0.00015724558317535417,
      "loss": 0.207,
      "step": 24240
    },
    {
      "epoch": 0.3408531871529974,
      "grad_norm": 0.3872990310192108,
      "learning_rate": 0.00015720746458141295,
      "loss": 0.1961,
      "step": 24250
    },
    {
      "epoch": 0.3409937451683182,
      "grad_norm": 0.46306172013282776,
      "learning_rate": 0.00015716933362725465,
      "loss": 0.2034,
      "step": 24260
    },
    {
      "epoch": 0.341134303183639,
      "grad_norm": 0.4858575463294983,
      "learning_rate": 0.0001571311903211178,
      "loss": 0.2087,
      "step": 24270
    },
    {
      "epoch": 0.3412748611989599,
      "grad_norm": 0.30993708968162537,
      "learning_rate": 0.00015709303467124368,
      "loss": 0.2131,
      "step": 24280
    },
    {
      "epoch": 0.3414154192142807,
      "grad_norm": 0.3692980706691742,
      "learning_rate": 0.00015705486668587612,
      "loss": 0.1943,
      "step": 24290
    },
    {
      "epoch": 0.3415559772296015,
      "grad_norm": 0.42944425344467163,
      "learning_rate": 0.0001570166863732617,
      "loss": 0.2096,
      "step": 24300
    },
    {
      "epoch": 0.34169653524492233,
      "grad_norm": 0.6496607661247253,
      "learning_rate": 0.00015697849374164964,
      "loss": 0.2066,
      "step": 24310
    },
    {
      "epoch": 0.34183709326024314,
      "grad_norm": 0.4161585569381714,
      "learning_rate": 0.00015694028879929185,
      "loss": 0.1847,
      "step": 24320
    },
    {
      "epoch": 0.341977651275564,
      "grad_norm": 0.3806317448616028,
      "learning_rate": 0.00015690207155444284,
      "loss": 0.1714,
      "step": 24330
    },
    {
      "epoch": 0.3421182092908848,
      "grad_norm": 0.4975564777851105,
      "learning_rate": 0.00015686384201535979,
      "loss": 0.201,
      "step": 24340
    },
    {
      "epoch": 0.34225876730620564,
      "grad_norm": 0.4811922013759613,
      "learning_rate": 0.00015682560019030257,
      "loss": 0.1861,
      "step": 24350
    },
    {
      "epoch": 0.34239932532152645,
      "grad_norm": 0.3213375508785248,
      "learning_rate": 0.00015678734608753371,
      "loss": 0.1893,
      "step": 24360
    },
    {
      "epoch": 0.34253988333684726,
      "grad_norm": 0.4221675992012024,
      "learning_rate": 0.00015674907971531838,
      "loss": 0.2132,
      "step": 24370
    },
    {
      "epoch": 0.34268044135216813,
      "grad_norm": 0.48913589119911194,
      "learning_rate": 0.00015671080108192436,
      "loss": 0.1798,
      "step": 24380
    },
    {
      "epoch": 0.34282099936748894,
      "grad_norm": 0.3994086980819702,
      "learning_rate": 0.0001566725101956221,
      "loss": 0.1863,
      "step": 24390
    },
    {
      "epoch": 0.34296155738280976,
      "grad_norm": 0.40696194767951965,
      "learning_rate": 0.00015663420706468477,
      "loss": 0.206,
      "step": 24400
    },
    {
      "epoch": 0.34310211539813057,
      "grad_norm": 0.44838500022888184,
      "learning_rate": 0.00015659589169738812,
      "loss": 0.1901,
      "step": 24410
    },
    {
      "epoch": 0.3432426734134514,
      "grad_norm": 0.4780958294868469,
      "learning_rate": 0.0001565575641020105,
      "loss": 0.2107,
      "step": 24420
    },
    {
      "epoch": 0.34338323142877225,
      "grad_norm": 0.4724385440349579,
      "learning_rate": 0.00015651922428683296,
      "loss": 0.1899,
      "step": 24430
    },
    {
      "epoch": 0.34352378944409306,
      "grad_norm": 0.5628231763839722,
      "learning_rate": 0.00015648087226013925,
      "loss": 0.1875,
      "step": 24440
    },
    {
      "epoch": 0.3436643474594139,
      "grad_norm": 0.31662383675575256,
      "learning_rate": 0.0001564425080302156,
      "loss": 0.1964,
      "step": 24450
    },
    {
      "epoch": 0.3438049054747347,
      "grad_norm": 0.4724717438220978,
      "learning_rate": 0.00015640413160535098,
      "loss": 0.2064,
      "step": 24460
    },
    {
      "epoch": 0.3439454634900555,
      "grad_norm": 0.43273380398750305,
      "learning_rate": 0.00015636574299383703,
      "loss": 0.1795,
      "step": 24470
    },
    {
      "epoch": 0.34408602150537637,
      "grad_norm": 0.4508928060531616,
      "learning_rate": 0.00015632734220396796,
      "loss": 0.2112,
      "step": 24480
    },
    {
      "epoch": 0.3442265795206972,
      "grad_norm": 0.4900144934654236,
      "learning_rate": 0.00015628892924404063,
      "loss": 0.2006,
      "step": 24490
    },
    {
      "epoch": 0.344367137536018,
      "grad_norm": 0.440208375453949,
      "learning_rate": 0.0001562505041223545,
      "loss": 0.1994,
      "step": 24500
    },
    {
      "epoch": 0.3445076955513388,
      "grad_norm": 0.46662935614585876,
      "learning_rate": 0.0001562120668472117,
      "loss": 0.1965,
      "step": 24510
    },
    {
      "epoch": 0.3446482535666596,
      "grad_norm": 0.5307490229606628,
      "learning_rate": 0.00015617361742691695,
      "loss": 0.2195,
      "step": 24520
    },
    {
      "epoch": 0.34478881158198044,
      "grad_norm": 0.6571158170700073,
      "learning_rate": 0.00015613515586977762,
      "loss": 0.1994,
      "step": 24530
    },
    {
      "epoch": 0.3449293695973013,
      "grad_norm": 0.2975288927555084,
      "learning_rate": 0.00015609668218410374,
      "loss": 0.1942,
      "step": 24540
    },
    {
      "epoch": 0.3450699276126221,
      "grad_norm": 0.4509091079235077,
      "learning_rate": 0.00015605819637820787,
      "loss": 0.1865,
      "step": 24550
    },
    {
      "epoch": 0.34521048562794293,
      "grad_norm": 0.342122346162796,
      "learning_rate": 0.00015601969846040523,
      "loss": 0.1913,
      "step": 24560
    },
    {
      "epoch": 0.34535104364326374,
      "grad_norm": 0.3601060211658478,
      "learning_rate": 0.00015598118843901368,
      "loss": 0.2014,
      "step": 24570
    },
    {
      "epoch": 0.34549160165858456,
      "grad_norm": 0.41186758875846863,
      "learning_rate": 0.0001559426663223537,
      "loss": 0.1966,
      "step": 24580
    },
    {
      "epoch": 0.3456321596739054,
      "grad_norm": 0.4192986488342285,
      "learning_rate": 0.0001559041321187483,
      "loss": 0.1782,
      "step": 24590
    },
    {
      "epoch": 0.34577271768922624,
      "grad_norm": 0.4782000184059143,
      "learning_rate": 0.00015586558583652319,
      "loss": 0.1998,
      "step": 24600
    },
    {
      "epoch": 0.34591327570454705,
      "grad_norm": 0.42880335450172424,
      "learning_rate": 0.0001558270274840067,
      "loss": 0.2009,
      "step": 24610
    },
    {
      "epoch": 0.34605383371986786,
      "grad_norm": 0.3918968141078949,
      "learning_rate": 0.00015578845706952965,
      "loss": 0.2082,
      "step": 24620
    },
    {
      "epoch": 0.3461943917351887,
      "grad_norm": 0.3660811483860016,
      "learning_rate": 0.00015574987460142563,
      "loss": 0.1895,
      "step": 24630
    },
    {
      "epoch": 0.34633494975050955,
      "grad_norm": 0.6555930376052856,
      "learning_rate": 0.00015571128008803068,
      "loss": 0.2044,
      "step": 24640
    },
    {
      "epoch": 0.34647550776583036,
      "grad_norm": 0.34414178133010864,
      "learning_rate": 0.0001556726735376836,
      "loss": 0.1901,
      "step": 24650
    },
    {
      "epoch": 0.3466160657811512,
      "grad_norm": 0.5629868507385254,
      "learning_rate": 0.00015563405495872557,
      "loss": 0.2279,
      "step": 24660
    },
    {
      "epoch": 0.346756623796472,
      "grad_norm": 0.4871220886707306,
      "learning_rate": 0.00015559542435950057,
      "loss": 0.2243,
      "step": 24670
    },
    {
      "epoch": 0.3468971818117928,
      "grad_norm": 0.3737644553184509,
      "learning_rate": 0.00015555678174835515,
      "loss": 0.1799,
      "step": 24680
    },
    {
      "epoch": 0.34703773982711367,
      "grad_norm": 0.3960535228252411,
      "learning_rate": 0.00015551812713363839,
      "loss": 0.2316,
      "step": 24690
    },
    {
      "epoch": 0.3471782978424345,
      "grad_norm": 0.32365134358406067,
      "learning_rate": 0.00015547946052370194,
      "loss": 0.168,
      "step": 24700
    },
    {
      "epoch": 0.3473188558577553,
      "grad_norm": 0.44706541299819946,
      "learning_rate": 0.00015544078192690014,
      "loss": 0.1731,
      "step": 24710
    },
    {
      "epoch": 0.3474594138730761,
      "grad_norm": 0.34758463501930237,
      "learning_rate": 0.00015540209135158985,
      "loss": 0.1784,
      "step": 24720
    },
    {
      "epoch": 0.3475999718883969,
      "grad_norm": 0.4847332239151001,
      "learning_rate": 0.00015536338880613054,
      "loss": 0.1908,
      "step": 24730
    },
    {
      "epoch": 0.34774052990371773,
      "grad_norm": 0.3961634635925293,
      "learning_rate": 0.00015532467429888426,
      "loss": 0.2029,
      "step": 24740
    },
    {
      "epoch": 0.3478810879190386,
      "grad_norm": 0.48429301381111145,
      "learning_rate": 0.00015528594783821564,
      "loss": 0.1909,
      "step": 24750
    },
    {
      "epoch": 0.3480216459343594,
      "grad_norm": 0.3571798503398895,
      "learning_rate": 0.00015524720943249192,
      "loss": 0.1983,
      "step": 24760
    },
    {
      "epoch": 0.3481622039496802,
      "grad_norm": 0.42743244767189026,
      "learning_rate": 0.0001552084590900829,
      "loss": 0.2104,
      "step": 24770
    },
    {
      "epoch": 0.34830276196500104,
      "grad_norm": 0.4831909239292145,
      "learning_rate": 0.00015516969681936095,
      "loss": 0.2138,
      "step": 24780
    },
    {
      "epoch": 0.34844331998032185,
      "grad_norm": 0.4320811927318573,
      "learning_rate": 0.00015513092262870105,
      "loss": 0.1998,
      "step": 24790
    },
    {
      "epoch": 0.3485838779956427,
      "grad_norm": 0.3804016709327698,
      "learning_rate": 0.00015509213652648067,
      "loss": 0.1848,
      "step": 24800
    },
    {
      "epoch": 0.34872443601096353,
      "grad_norm": 0.33573248982429504,
      "learning_rate": 0.00015505333852107998,
      "loss": 0.2102,
      "step": 24810
    },
    {
      "epoch": 0.34886499402628435,
      "grad_norm": 0.4067336916923523,
      "learning_rate": 0.00015501452862088164,
      "loss": 0.1986,
      "step": 24820
    },
    {
      "epoch": 0.34900555204160516,
      "grad_norm": 0.609912097454071,
      "learning_rate": 0.00015497570683427087,
      "loss": 0.1976,
      "step": 24830
    },
    {
      "epoch": 0.349146110056926,
      "grad_norm": 0.39099013805389404,
      "learning_rate": 0.00015493687316963555,
      "loss": 0.2116,
      "step": 24840
    },
    {
      "epoch": 0.34928666807224684,
      "grad_norm": 0.38108426332473755,
      "learning_rate": 0.000154898027635366,
      "loss": 0.2155,
      "step": 24850
    },
    {
      "epoch": 0.34942722608756766,
      "grad_norm": 0.382459431886673,
      "learning_rate": 0.0001548591702398552,
      "loss": 0.1883,
      "step": 24860
    },
    {
      "epoch": 0.34956778410288847,
      "grad_norm": 0.40003153681755066,
      "learning_rate": 0.00015482030099149864,
      "loss": 0.2062,
      "step": 24870
    },
    {
      "epoch": 0.3497083421182093,
      "grad_norm": 0.48427441716194153,
      "learning_rate": 0.00015478141989869437,
      "loss": 0.2063,
      "step": 24880
    },
    {
      "epoch": 0.3498489001335301,
      "grad_norm": 0.43394699692726135,
      "learning_rate": 0.0001547425269698431,
      "loss": 0.1831,
      "step": 24890
    },
    {
      "epoch": 0.34998945814885096,
      "grad_norm": 0.3585866391658783,
      "learning_rate": 0.00015470362221334794,
      "loss": 0.2122,
      "step": 24900
    },
    {
      "epoch": 0.3501300161641718,
      "grad_norm": 0.655471920967102,
      "learning_rate": 0.00015466470563761471,
      "loss": 0.1693,
      "step": 24910
    },
    {
      "epoch": 0.3502705741794926,
      "grad_norm": 0.4609263837337494,
      "learning_rate": 0.00015462577725105164,
      "loss": 0.1979,
      "step": 24920
    },
    {
      "epoch": 0.3504111321948134,
      "grad_norm": 0.43604934215545654,
      "learning_rate": 0.00015458683706206957,
      "loss": 0.1821,
      "step": 24930
    },
    {
      "epoch": 0.3505516902101342,
      "grad_norm": 0.3810408413410187,
      "learning_rate": 0.00015454788507908196,
      "loss": 0.1994,
      "step": 24940
    },
    {
      "epoch": 0.3506922482254551,
      "grad_norm": 0.39820948243141174,
      "learning_rate": 0.0001545089213105047,
      "loss": 0.2092,
      "step": 24950
    },
    {
      "epoch": 0.3508328062407759,
      "grad_norm": 0.2860207259654999,
      "learning_rate": 0.00015446994576475632,
      "loss": 0.1709,
      "step": 24960
    },
    {
      "epoch": 0.3509733642560967,
      "grad_norm": 0.36918723583221436,
      "learning_rate": 0.00015443095845025782,
      "loss": 0.2007,
      "step": 24970
    },
    {
      "epoch": 0.3511139222714175,
      "grad_norm": 0.46982356905937195,
      "learning_rate": 0.00015439195937543284,
      "loss": 0.2095,
      "step": 24980
    },
    {
      "epoch": 0.35125448028673834,
      "grad_norm": 0.35570284724235535,
      "learning_rate": 0.00015435294854870742,
      "loss": 0.1673,
      "step": 24990
    },
    {
      "epoch": 0.35139503830205915,
      "grad_norm": 0.4363248348236084,
      "learning_rate": 0.00015431392597851028,
      "loss": 0.1936,
      "step": 25000
    },
    {
      "epoch": 0.35139503830205915,
      "eval_chrf": 81.15394096300686,
      "eval_loss": 0.4088630974292755,
      "eval_runtime": 262.4779,
      "eval_samples_per_second": 0.381,
      "eval_steps_per_second": 0.015,
      "step": 25000
    },
    {
      "epoch": 0.35153559631738,
      "grad_norm": 0.4828396141529083,
      "learning_rate": 0.00015427489167327263,
      "loss": 0.2159,
      "step": 25010
    },
    {
      "epoch": 0.35167615433270083,
      "grad_norm": 0.30296817421913147,
      "learning_rate": 0.00015423584564142816,
      "loss": 0.1785,
      "step": 25020
    },
    {
      "epoch": 0.35181671234802164,
      "grad_norm": 0.4212421476840973,
      "learning_rate": 0.00015419678789141314,
      "loss": 0.1845,
      "step": 25030
    },
    {
      "epoch": 0.35195727036334246,
      "grad_norm": 0.3856677711009979,
      "learning_rate": 0.0001541577184316664,
      "loss": 0.202,
      "step": 25040
    },
    {
      "epoch": 0.35209782837866327,
      "grad_norm": 0.45877838134765625,
      "learning_rate": 0.00015411863727062924,
      "loss": 0.1786,
      "step": 25050
    },
    {
      "epoch": 0.35223838639398414,
      "grad_norm": 0.4523278474807739,
      "learning_rate": 0.00015407954441674552,
      "loss": 0.1806,
      "step": 25060
    },
    {
      "epoch": 0.35237894440930495,
      "grad_norm": 0.4404580593109131,
      "learning_rate": 0.00015404043987846163,
      "loss": 0.2065,
      "step": 25070
    },
    {
      "epoch": 0.35251950242462576,
      "grad_norm": 0.45593929290771484,
      "learning_rate": 0.00015400132366422652,
      "loss": 0.1979,
      "step": 25080
    },
    {
      "epoch": 0.3526600604399466,
      "grad_norm": 0.3615800440311432,
      "learning_rate": 0.0001539621957824915,
      "loss": 0.1534,
      "step": 25090
    },
    {
      "epoch": 0.3528006184552674,
      "grad_norm": 0.37038901448249817,
      "learning_rate": 0.00015392305624171065,
      "loss": 0.1878,
      "step": 25100
    },
    {
      "epoch": 0.35294117647058826,
      "grad_norm": 0.3618924915790558,
      "learning_rate": 0.0001538839050503404,
      "loss": 0.1869,
      "step": 25110
    },
    {
      "epoch": 0.35308173448590907,
      "grad_norm": 0.45690590143203735,
      "learning_rate": 0.00015384474221683968,
      "loss": 0.1916,
      "step": 25120
    },
    {
      "epoch": 0.3532222925012299,
      "grad_norm": 1.2872811555862427,
      "learning_rate": 0.00015380556774967008,
      "loss": 0.1642,
      "step": 25130
    },
    {
      "epoch": 0.3533628505165507,
      "grad_norm": 0.4172723591327667,
      "learning_rate": 0.00015376638165729553,
      "loss": 0.1918,
      "step": 25140
    },
    {
      "epoch": 0.3535034085318715,
      "grad_norm": 0.4361129105091095,
      "learning_rate": 0.00015372718394818263,
      "loss": 0.2063,
      "step": 25150
    },
    {
      "epoch": 0.3536439665471924,
      "grad_norm": 0.45274701714515686,
      "learning_rate": 0.0001536879746308004,
      "loss": 0.2131,
      "step": 25160
    },
    {
      "epoch": 0.3537845245625132,
      "grad_norm": 0.4044634997844696,
      "learning_rate": 0.00015364875371362037,
      "loss": 0.209,
      "step": 25170
    },
    {
      "epoch": 0.353925082577834,
      "grad_norm": 0.4239368438720703,
      "learning_rate": 0.0001536095212051166,
      "loss": 0.2017,
      "step": 25180
    },
    {
      "epoch": 0.3540656405931548,
      "grad_norm": 0.3916068375110626,
      "learning_rate": 0.00015357027711376567,
      "loss": 0.1812,
      "step": 25190
    },
    {
      "epoch": 0.35420619860847563,
      "grad_norm": 0.47555193305015564,
      "learning_rate": 0.00015353102144804656,
      "loss": 0.1949,
      "step": 25200
    },
    {
      "epoch": 0.3543467566237965,
      "grad_norm": 0.4671885371208191,
      "learning_rate": 0.0001534917542164409,
      "loss": 0.2321,
      "step": 25210
    },
    {
      "epoch": 0.3544873146391173,
      "grad_norm": 0.3739650249481201,
      "learning_rate": 0.00015345247542743277,
      "loss": 0.2,
      "step": 25220
    },
    {
      "epoch": 0.3546278726544381,
      "grad_norm": 0.3284544348716736,
      "learning_rate": 0.0001534131850895087,
      "loss": 0.1709,
      "step": 25230
    },
    {
      "epoch": 0.35476843066975894,
      "grad_norm": 0.42038440704345703,
      "learning_rate": 0.00015337388321115766,
      "loss": 0.219,
      "step": 25240
    },
    {
      "epoch": 0.35490898868507975,
      "grad_norm": 0.5038439035415649,
      "learning_rate": 0.00015333456980087134,
      "loss": 0.2001,
      "step": 25250
    },
    {
      "epoch": 0.35504954670040056,
      "grad_norm": 0.40093642473220825,
      "learning_rate": 0.0001532952448671437,
      "loss": 0.2016,
      "step": 25260
    },
    {
      "epoch": 0.35519010471572143,
      "grad_norm": 0.4847545325756073,
      "learning_rate": 0.0001532559084184712,
      "loss": 0.1905,
      "step": 25270
    },
    {
      "epoch": 0.35533066273104225,
      "grad_norm": 0.4720471501350403,
      "learning_rate": 0.000153216560463353,
      "loss": 0.1891,
      "step": 25280
    },
    {
      "epoch": 0.35547122074636306,
      "grad_norm": 0.5366537570953369,
      "learning_rate": 0.0001531772010102905,
      "loss": 0.2147,
      "step": 25290
    },
    {
      "epoch": 0.35561177876168387,
      "grad_norm": 0.35445621609687805,
      "learning_rate": 0.0001531378300677877,
      "loss": 0.2055,
      "step": 25300
    },
    {
      "epoch": 0.3557523367770047,
      "grad_norm": 0.39145347476005554,
      "learning_rate": 0.00015309844764435106,
      "loss": 0.2144,
      "step": 25310
    },
    {
      "epoch": 0.35589289479232555,
      "grad_norm": 0.5245220065116882,
      "learning_rate": 0.00015305905374848956,
      "loss": 0.1929,
      "step": 25320
    },
    {
      "epoch": 0.35603345280764637,
      "grad_norm": 0.5469962358474731,
      "learning_rate": 0.0001530196483887146,
      "loss": 0.2135,
      "step": 25330
    },
    {
      "epoch": 0.3561740108229672,
      "grad_norm": 0.4332871735095978,
      "learning_rate": 0.00015298023157354012,
      "loss": 0.1983,
      "step": 25340
    },
    {
      "epoch": 0.356314568838288,
      "grad_norm": 0.40701428055763245,
      "learning_rate": 0.00015294080331148245,
      "loss": 0.2051,
      "step": 25350
    },
    {
      "epoch": 0.3564551268536088,
      "grad_norm": 0.5671848058700562,
      "learning_rate": 0.00015290136361106047,
      "loss": 0.2128,
      "step": 25360
    },
    {
      "epoch": 0.3565956848689297,
      "grad_norm": 0.39455220103263855,
      "learning_rate": 0.00015286191248079547,
      "loss": 0.2067,
      "step": 25370
    },
    {
      "epoch": 0.3567362428842505,
      "grad_norm": 0.3148670792579651,
      "learning_rate": 0.0001528224499292113,
      "loss": 0.1832,
      "step": 25380
    },
    {
      "epoch": 0.3568768008995713,
      "grad_norm": 0.4024641215801239,
      "learning_rate": 0.00015278297596483417,
      "loss": 0.1805,
      "step": 25390
    },
    {
      "epoch": 0.3570173589148921,
      "grad_norm": 0.5334517955780029,
      "learning_rate": 0.0001527434905961928,
      "loss": 0.208,
      "step": 25400
    },
    {
      "epoch": 0.3571579169302129,
      "grad_norm": 0.4131854176521301,
      "learning_rate": 0.00015270399383181843,
      "loss": 0.1718,
      "step": 25410
    },
    {
      "epoch": 0.3572984749455338,
      "grad_norm": 0.448175847530365,
      "learning_rate": 0.00015266448568024464,
      "loss": 0.1713,
      "step": 25420
    },
    {
      "epoch": 0.3574390329608546,
      "grad_norm": 0.402320921421051,
      "learning_rate": 0.00015262496615000762,
      "loss": 0.1946,
      "step": 25430
    },
    {
      "epoch": 0.3575795909761754,
      "grad_norm": 0.4061627686023712,
      "learning_rate": 0.0001525854352496459,
      "loss": 0.2063,
      "step": 25440
    },
    {
      "epoch": 0.35772014899149623,
      "grad_norm": 0.6434004902839661,
      "learning_rate": 0.00015254589298770052,
      "loss": 0.2101,
      "step": 25450
    },
    {
      "epoch": 0.35786070700681705,
      "grad_norm": 0.38769012689590454,
      "learning_rate": 0.00015250633937271494,
      "loss": 0.1921,
      "step": 25460
    },
    {
      "epoch": 0.35800126502213786,
      "grad_norm": 0.4754445552825928,
      "learning_rate": 0.00015246677441323512,
      "loss": 0.2021,
      "step": 25470
    },
    {
      "epoch": 0.35814182303745873,
      "grad_norm": 0.4753167927265167,
      "learning_rate": 0.0001524271981178094,
      "loss": 0.2043,
      "step": 25480
    },
    {
      "epoch": 0.35828238105277954,
      "grad_norm": 0.4295427203178406,
      "learning_rate": 0.0001523876104949887,
      "loss": 0.1838,
      "step": 25490
    },
    {
      "epoch": 0.35842293906810035,
      "grad_norm": 0.4452112913131714,
      "learning_rate": 0.00015234801155332623,
      "loss": 0.1948,
      "step": 25500
    },
    {
      "epoch": 0.35856349708342117,
      "grad_norm": 0.4044558107852936,
      "learning_rate": 0.00015230840130137777,
      "loss": 0.1915,
      "step": 25510
    },
    {
      "epoch": 0.358704055098742,
      "grad_norm": 0.46111053228378296,
      "learning_rate": 0.0001522687797477014,
      "loss": 0.1995,
      "step": 25520
    },
    {
      "epoch": 0.35884461311406285,
      "grad_norm": 0.33259961009025574,
      "learning_rate": 0.00015222914690085783,
      "loss": 0.2093,
      "step": 25530
    },
    {
      "epoch": 0.35898517112938366,
      "grad_norm": 0.3852463960647583,
      "learning_rate": 0.00015218950276941012,
      "loss": 0.1865,
      "step": 25540
    },
    {
      "epoch": 0.3591257291447045,
      "grad_norm": 0.437765896320343,
      "learning_rate": 0.00015214984736192365,
      "loss": 0.2064,
      "step": 25550
    },
    {
      "epoch": 0.3592662871600253,
      "grad_norm": 0.44627997279167175,
      "learning_rate": 0.00015211018068696648,
      "loss": 0.1807,
      "step": 25560
    },
    {
      "epoch": 0.3594068451753461,
      "grad_norm": 0.4240309000015259,
      "learning_rate": 0.00015207050275310886,
      "loss": 0.1795,
      "step": 25570
    },
    {
      "epoch": 0.35954740319066697,
      "grad_norm": 0.4682501256465912,
      "learning_rate": 0.00015203081356892368,
      "loss": 0.2183,
      "step": 25580
    },
    {
      "epoch": 0.3596879612059878,
      "grad_norm": 0.5094911456108093,
      "learning_rate": 0.0001519911131429861,
      "loss": 0.19,
      "step": 25590
    },
    {
      "epoch": 0.3598285192213086,
      "grad_norm": 0.467891663312912,
      "learning_rate": 0.0001519514014838738,
      "loss": 0.1903,
      "step": 25600
    },
    {
      "epoch": 0.3599690772366294,
      "grad_norm": 0.4371752440929413,
      "learning_rate": 0.00015191167860016688,
      "loss": 0.2062,
      "step": 25610
    },
    {
      "epoch": 0.3601096352519502,
      "grad_norm": 0.3752915561199188,
      "learning_rate": 0.0001518719445004478,
      "loss": 0.1883,
      "step": 25620
    },
    {
      "epoch": 0.3602501932672711,
      "grad_norm": 0.40677115321159363,
      "learning_rate": 0.00015183219919330155,
      "loss": 0.2387,
      "step": 25630
    },
    {
      "epoch": 0.3603907512825919,
      "grad_norm": 0.406812846660614,
      "learning_rate": 0.00015179244268731542,
      "loss": 0.1941,
      "step": 25640
    },
    {
      "epoch": 0.3605313092979127,
      "grad_norm": 0.4294555187225342,
      "learning_rate": 0.00015175267499107923,
      "loss": 0.1938,
      "step": 25650
    },
    {
      "epoch": 0.36067186731323353,
      "grad_norm": 0.5252370238304138,
      "learning_rate": 0.00015171289611318515,
      "loss": 0.2067,
      "step": 25660
    },
    {
      "epoch": 0.36081242532855434,
      "grad_norm": 0.36849528551101685,
      "learning_rate": 0.00015167310606222778,
      "loss": 0.1794,
      "step": 25670
    },
    {
      "epoch": 0.3609529833438752,
      "grad_norm": 0.3125917613506317,
      "learning_rate": 0.00015163330484680413,
      "loss": 0.225,
      "step": 25680
    },
    {
      "epoch": 0.361093541359196,
      "grad_norm": 0.4415869116783142,
      "learning_rate": 0.00015159349247551365,
      "loss": 0.1617,
      "step": 25690
    },
    {
      "epoch": 0.36123409937451684,
      "grad_norm": 0.5339673161506653,
      "learning_rate": 0.0001515536689569582,
      "loss": 0.2319,
      "step": 25700
    },
    {
      "epoch": 0.36137465738983765,
      "grad_norm": 0.4298225939273834,
      "learning_rate": 0.000151513834299742,
      "loss": 0.1851,
      "step": 25710
    },
    {
      "epoch": 0.36151521540515846,
      "grad_norm": 0.45103615522384644,
      "learning_rate": 0.00015147398851247173,
      "loss": 0.18,
      "step": 25720
    },
    {
      "epoch": 0.3616557734204793,
      "grad_norm": 0.35722455382347107,
      "learning_rate": 0.0001514341316037564,
      "loss": 0.2086,
      "step": 25730
    },
    {
      "epoch": 0.36179633143580014,
      "grad_norm": 0.45123592019081116,
      "learning_rate": 0.00015139426358220753,
      "loss": 0.1824,
      "step": 25740
    },
    {
      "epoch": 0.36193688945112096,
      "grad_norm": 0.35224390029907227,
      "learning_rate": 0.00015135438445643894,
      "loss": 0.1852,
      "step": 25750
    },
    {
      "epoch": 0.36207744746644177,
      "grad_norm": 0.41913264989852905,
      "learning_rate": 0.00015131449423506693,
      "loss": 0.1771,
      "step": 25760
    },
    {
      "epoch": 0.3622180054817626,
      "grad_norm": 0.390919953584671,
      "learning_rate": 0.0001512745929267102,
      "loss": 0.1847,
      "step": 25770
    },
    {
      "epoch": 0.3623585634970834,
      "grad_norm": 0.5339962840080261,
      "learning_rate": 0.00015123468053998971,
      "loss": 0.1906,
      "step": 25780
    },
    {
      "epoch": 0.36249912151240427,
      "grad_norm": 0.4494784474372864,
      "learning_rate": 0.00015119475708352898,
      "loss": 0.1651,
      "step": 25790
    },
    {
      "epoch": 0.3626396795277251,
      "grad_norm": 0.3535462021827698,
      "learning_rate": 0.00015115482256595382,
      "loss": 0.1948,
      "step": 25800
    },
    {
      "epoch": 0.3627802375430459,
      "grad_norm": 0.35224857926368713,
      "learning_rate": 0.00015111487699589248,
      "loss": 0.1704,
      "step": 25810
    },
    {
      "epoch": 0.3629207955583667,
      "grad_norm": 0.32310375571250916,
      "learning_rate": 0.0001510749203819756,
      "loss": 0.2059,
      "step": 25820
    },
    {
      "epoch": 0.3630613535736875,
      "grad_norm": 0.31112274527549744,
      "learning_rate": 0.00015103495273283612,
      "loss": 0.1818,
      "step": 25830
    },
    {
      "epoch": 0.3632019115890084,
      "grad_norm": 0.47570037841796875,
      "learning_rate": 0.0001509949740571095,
      "loss": 0.1964,
      "step": 25840
    },
    {
      "epoch": 0.3633424696043292,
      "grad_norm": 0.5536921620368958,
      "learning_rate": 0.00015095498436343348,
      "loss": 0.2223,
      "step": 25850
    },
    {
      "epoch": 0.36348302761965,
      "grad_norm": 0.47299832105636597,
      "learning_rate": 0.00015091498366044822,
      "loss": 0.1902,
      "step": 25860
    },
    {
      "epoch": 0.3636235856349708,
      "grad_norm": 0.36159366369247437,
      "learning_rate": 0.00015087497195679625,
      "loss": 0.1644,
      "step": 25870
    },
    {
      "epoch": 0.36376414365029164,
      "grad_norm": 0.44765231013298035,
      "learning_rate": 0.00015083494926112248,
      "loss": 0.2018,
      "step": 25880
    },
    {
      "epoch": 0.3639047016656125,
      "grad_norm": 0.3738773465156555,
      "learning_rate": 0.0001507949155820742,
      "loss": 0.2001,
      "step": 25890
    },
    {
      "epoch": 0.3640452596809333,
      "grad_norm": 0.3980732262134552,
      "learning_rate": 0.00015075487092830106,
      "loss": 0.188,
      "step": 25900
    },
    {
      "epoch": 0.36418581769625413,
      "grad_norm": 0.4188894033432007,
      "learning_rate": 0.00015071481530845511,
      "loss": 0.1737,
      "step": 25910
    },
    {
      "epoch": 0.36432637571157495,
      "grad_norm": 0.3625176250934601,
      "learning_rate": 0.0001506747487311907,
      "loss": 0.2304,
      "step": 25920
    },
    {
      "epoch": 0.36446693372689576,
      "grad_norm": 0.3531343638896942,
      "learning_rate": 0.00015063467120516469,
      "loss": 0.2039,
      "step": 25930
    },
    {
      "epoch": 0.3646074917422166,
      "grad_norm": 0.3657441735267639,
      "learning_rate": 0.00015059458273903608,
      "loss": 0.175,
      "step": 25940
    },
    {
      "epoch": 0.36474804975753744,
      "grad_norm": 0.3509301245212555,
      "learning_rate": 0.00015055448334146644,
      "loss": 0.1954,
      "step": 25950
    },
    {
      "epoch": 0.36488860777285825,
      "grad_norm": 0.33539968729019165,
      "learning_rate": 0.00015051437302111964,
      "loss": 0.1806,
      "step": 25960
    },
    {
      "epoch": 0.36502916578817907,
      "grad_norm": 0.6233363151550293,
      "learning_rate": 0.00015047425178666188,
      "loss": 0.1923,
      "step": 25970
    },
    {
      "epoch": 0.3651697238034999,
      "grad_norm": 0.3398738205432892,
      "learning_rate": 0.0001504341196467617,
      "loss": 0.1756,
      "step": 25980
    },
    {
      "epoch": 0.3653102818188207,
      "grad_norm": 0.47035306692123413,
      "learning_rate": 0.0001503939766100901,
      "loss": 0.1843,
      "step": 25990
    },
    {
      "epoch": 0.36545083983414156,
      "grad_norm": 0.3709024488925934,
      "learning_rate": 0.00015035382268532027,
      "loss": 0.2017,
      "step": 26000
    },
    {
      "epoch": 0.36545083983414156,
      "eval_chrf": 80.16979195018159,
      "eval_loss": 0.4062730669975281,
      "eval_runtime": 326.8393,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 26000
    },
    {
      "epoch": 0.3655913978494624,
      "grad_norm": 0.4761894643306732,
      "learning_rate": 0.00015031365788112795,
      "loss": 0.1971,
      "step": 26010
    },
    {
      "epoch": 0.3657319558647832,
      "grad_norm": 0.4762408435344696,
      "learning_rate": 0.00015027348220619105,
      "loss": 0.2146,
      "step": 26020
    },
    {
      "epoch": 0.365872513880104,
      "grad_norm": 0.421764999628067,
      "learning_rate": 0.00015023329566918995,
      "loss": 0.2101,
      "step": 26030
    },
    {
      "epoch": 0.3660130718954248,
      "grad_norm": 0.46174123883247375,
      "learning_rate": 0.0001501930982788073,
      "loss": 0.2025,
      "step": 26040
    },
    {
      "epoch": 0.3661536299107457,
      "grad_norm": 0.37564560770988464,
      "learning_rate": 0.00015015289004372817,
      "loss": 0.1912,
      "step": 26050
    },
    {
      "epoch": 0.3662941879260665,
      "grad_norm": 0.4434491991996765,
      "learning_rate": 0.00015011267097263993,
      "loss": 0.1862,
      "step": 26060
    },
    {
      "epoch": 0.3664347459413873,
      "grad_norm": 0.3763035535812378,
      "learning_rate": 0.00015007244107423224,
      "loss": 0.2299,
      "step": 26070
    },
    {
      "epoch": 0.3665753039567081,
      "grad_norm": 0.4968195855617523,
      "learning_rate": 0.00015003220035719717,
      "loss": 0.1949,
      "step": 26080
    },
    {
      "epoch": 0.36671586197202893,
      "grad_norm": 0.44107377529144287,
      "learning_rate": 0.00014999194883022918,
      "loss": 0.1881,
      "step": 26090
    },
    {
      "epoch": 0.3668564199873498,
      "grad_norm": 0.36147475242614746,
      "learning_rate": 0.0001499516865020249,
      "loss": 0.1979,
      "step": 26100
    },
    {
      "epoch": 0.3669969780026706,
      "grad_norm": 0.31467700004577637,
      "learning_rate": 0.00014991141338128346,
      "loss": 0.1858,
      "step": 26110
    },
    {
      "epoch": 0.36713753601799143,
      "grad_norm": 0.4247540533542633,
      "learning_rate": 0.0001498711294767062,
      "loss": 0.1927,
      "step": 26120
    },
    {
      "epoch": 0.36727809403331224,
      "grad_norm": 0.5584858059883118,
      "learning_rate": 0.00014983083479699688,
      "loss": 0.1917,
      "step": 26130
    },
    {
      "epoch": 0.36741865204863305,
      "grad_norm": 0.4436037242412567,
      "learning_rate": 0.00014979052935086153,
      "loss": 0.1789,
      "step": 26140
    },
    {
      "epoch": 0.3675592100639539,
      "grad_norm": 0.366242915391922,
      "learning_rate": 0.00014975021314700851,
      "loss": 0.1944,
      "step": 26150
    },
    {
      "epoch": 0.36769976807927474,
      "grad_norm": 0.3136445879936218,
      "learning_rate": 0.00014970988619414857,
      "loss": 0.1893,
      "step": 26160
    },
    {
      "epoch": 0.36784032609459555,
      "grad_norm": 0.46261394023895264,
      "learning_rate": 0.00014966954850099466,
      "loss": 0.1855,
      "step": 26170
    },
    {
      "epoch": 0.36798088410991636,
      "grad_norm": 0.4286184310913086,
      "learning_rate": 0.00014962920007626214,
      "loss": 0.193,
      "step": 26180
    },
    {
      "epoch": 0.3681214421252372,
      "grad_norm": 0.3819252550601959,
      "learning_rate": 0.00014958884092866877,
      "loss": 0.1972,
      "step": 26190
    },
    {
      "epoch": 0.368262000140558,
      "grad_norm": 0.32119134068489075,
      "learning_rate": 0.0001495484710669344,
      "loss": 0.2079,
      "step": 26200
    },
    {
      "epoch": 0.36840255815587886,
      "grad_norm": 0.6470380425453186,
      "learning_rate": 0.00014950809049978135,
      "loss": 0.2185,
      "step": 26210
    },
    {
      "epoch": 0.36854311617119967,
      "grad_norm": 0.5115725994110107,
      "learning_rate": 0.0001494676992359343,
      "loss": 0.1962,
      "step": 26220
    },
    {
      "epoch": 0.3686836741865205,
      "grad_norm": 0.4752556383609772,
      "learning_rate": 0.00014942729728412008,
      "loss": 0.1734,
      "step": 26230
    },
    {
      "epoch": 0.3688242322018413,
      "grad_norm": 0.3664419949054718,
      "learning_rate": 0.00014938688465306797,
      "loss": 0.2078,
      "step": 26240
    },
    {
      "epoch": 0.3689647902171621,
      "grad_norm": 0.3666921555995941,
      "learning_rate": 0.00014934646135150947,
      "loss": 0.1887,
      "step": 26250
    },
    {
      "epoch": 0.369105348232483,
      "grad_norm": 0.40116775035858154,
      "learning_rate": 0.00014930602738817843,
      "loss": 0.1955,
      "step": 26260
    },
    {
      "epoch": 0.3692459062478038,
      "grad_norm": 0.4006282389163971,
      "learning_rate": 0.00014926558277181102,
      "loss": 0.1845,
      "step": 26270
    },
    {
      "epoch": 0.3693864642631246,
      "grad_norm": 0.36226609349250793,
      "learning_rate": 0.00014922512751114564,
      "loss": 0.1893,
      "step": 26280
    },
    {
      "epoch": 0.3695270222784454,
      "grad_norm": 0.42465442419052124,
      "learning_rate": 0.00014918466161492307,
      "loss": 0.1977,
      "step": 26290
    },
    {
      "epoch": 0.36966758029376623,
      "grad_norm": 0.4116226136684418,
      "learning_rate": 0.0001491441850918863,
      "loss": 0.1887,
      "step": 26300
    },
    {
      "epoch": 0.3698081383090871,
      "grad_norm": 0.4674515426158905,
      "learning_rate": 0.00014910369795078073,
      "loss": 0.1741,
      "step": 26310
    },
    {
      "epoch": 0.3699486963244079,
      "grad_norm": 0.534906268119812,
      "learning_rate": 0.00014906320020035397,
      "loss": 0.1715,
      "step": 26320
    },
    {
      "epoch": 0.3700892543397287,
      "grad_norm": 0.4622785747051239,
      "learning_rate": 0.00014902269184935595,
      "loss": 0.206,
      "step": 26330
    },
    {
      "epoch": 0.37022981235504954,
      "grad_norm": 0.4936352074146271,
      "learning_rate": 0.00014898217290653886,
      "loss": 0.1968,
      "step": 26340
    },
    {
      "epoch": 0.37037037037037035,
      "grad_norm": 0.4342193901538849,
      "learning_rate": 0.00014894164338065722,
      "loss": 0.2211,
      "step": 26350
    },
    {
      "epoch": 0.3705109283856912,
      "grad_norm": 0.3889023959636688,
      "learning_rate": 0.0001489011032804678,
      "loss": 0.2049,
      "step": 26360
    },
    {
      "epoch": 0.37065148640101203,
      "grad_norm": 0.43299949169158936,
      "learning_rate": 0.00014886055261472973,
      "loss": 0.1946,
      "step": 26370
    },
    {
      "epoch": 0.37079204441633284,
      "grad_norm": 0.41790804266929626,
      "learning_rate": 0.00014881999139220432,
      "loss": 0.2017,
      "step": 26380
    },
    {
      "epoch": 0.37093260243165366,
      "grad_norm": 0.48441413044929504,
      "learning_rate": 0.0001487794196216552,
      "loss": 0.2093,
      "step": 26390
    },
    {
      "epoch": 0.37107316044697447,
      "grad_norm": 0.48560383915901184,
      "learning_rate": 0.0001487388373118483,
      "loss": 0.2107,
      "step": 26400
    },
    {
      "epoch": 0.37121371846229534,
      "grad_norm": 0.402222216129303,
      "learning_rate": 0.00014869824447155184,
      "loss": 0.2212,
      "step": 26410
    },
    {
      "epoch": 0.37135427647761615,
      "grad_norm": 0.32909220457077026,
      "learning_rate": 0.0001486576411095363,
      "loss": 0.1932,
      "step": 26420
    },
    {
      "epoch": 0.37149483449293697,
      "grad_norm": 0.40750375390052795,
      "learning_rate": 0.00014861702723457433,
      "loss": 0.2119,
      "step": 26430
    },
    {
      "epoch": 0.3716353925082578,
      "grad_norm": 0.6265875101089478,
      "learning_rate": 0.00014857640285544105,
      "loss": 0.2163,
      "step": 26440
    },
    {
      "epoch": 0.3717759505235786,
      "grad_norm": 0.4019491374492645,
      "learning_rate": 0.00014853576798091369,
      "loss": 0.1855,
      "step": 26450
    },
    {
      "epoch": 0.3719165085388994,
      "grad_norm": 0.38702765107154846,
      "learning_rate": 0.0001484951226197718,
      "loss": 0.1856,
      "step": 26460
    },
    {
      "epoch": 0.3720570665542203,
      "grad_norm": 0.4920539855957031,
      "learning_rate": 0.0001484544667807972,
      "loss": 0.2007,
      "step": 26470
    },
    {
      "epoch": 0.3721976245695411,
      "grad_norm": 0.33784493803977966,
      "learning_rate": 0.00014841380047277397,
      "loss": 0.2243,
      "step": 26480
    },
    {
      "epoch": 0.3723381825848619,
      "grad_norm": 0.4236350357532501,
      "learning_rate": 0.00014837312370448848,
      "loss": 0.1936,
      "step": 26490
    },
    {
      "epoch": 0.3724787406001827,
      "grad_norm": 0.40542203187942505,
      "learning_rate": 0.00014833243648472932,
      "loss": 0.2115,
      "step": 26500
    },
    {
      "epoch": 0.3726192986155035,
      "grad_norm": 0.3526892364025116,
      "learning_rate": 0.00014829173882228736,
      "loss": 0.1929,
      "step": 26510
    },
    {
      "epoch": 0.3727598566308244,
      "grad_norm": 0.30464527010917664,
      "learning_rate": 0.00014825103072595565,
      "loss": 0.1994,
      "step": 26520
    },
    {
      "epoch": 0.3729004146461452,
      "grad_norm": 0.5723264217376709,
      "learning_rate": 0.00014821031220452966,
      "loss": 0.1867,
      "step": 26530
    },
    {
      "epoch": 0.373040972661466,
      "grad_norm": 0.4225793778896332,
      "learning_rate": 0.00014816958326680695,
      "loss": 0.1961,
      "step": 26540
    },
    {
      "epoch": 0.37318153067678683,
      "grad_norm": 0.3909800350666046,
      "learning_rate": 0.0001481288439215874,
      "loss": 0.1813,
      "step": 26550
    },
    {
      "epoch": 0.37332208869210765,
      "grad_norm": 0.6097882986068726,
      "learning_rate": 0.00014808809417767317,
      "loss": 0.2281,
      "step": 26560
    },
    {
      "epoch": 0.3734626467074285,
      "grad_norm": 0.34307539463043213,
      "learning_rate": 0.0001480473340438686,
      "loss": 0.2191,
      "step": 26570
    },
    {
      "epoch": 0.3736032047227493,
      "grad_norm": 0.35505470633506775,
      "learning_rate": 0.0001480065635289803,
      "loss": 0.1825,
      "step": 26580
    },
    {
      "epoch": 0.37374376273807014,
      "grad_norm": 0.334445059299469,
      "learning_rate": 0.00014796578264181714,
      "loss": 0.1772,
      "step": 26590
    },
    {
      "epoch": 0.37388432075339095,
      "grad_norm": 0.3102055788040161,
      "learning_rate": 0.0001479249913911902,
      "loss": 0.17,
      "step": 26600
    },
    {
      "epoch": 0.37402487876871177,
      "grad_norm": 0.373024046421051,
      "learning_rate": 0.00014788418978591288,
      "loss": 0.1569,
      "step": 26610
    },
    {
      "epoch": 0.37416543678403263,
      "grad_norm": 0.47594505548477173,
      "learning_rate": 0.00014784337783480068,
      "loss": 0.1946,
      "step": 26620
    },
    {
      "epoch": 0.37430599479935345,
      "grad_norm": 0.36843621730804443,
      "learning_rate": 0.00014780255554667143,
      "loss": 0.1964,
      "step": 26630
    },
    {
      "epoch": 0.37444655281467426,
      "grad_norm": 0.44522184133529663,
      "learning_rate": 0.00014776172293034522,
      "loss": 0.1773,
      "step": 26640
    },
    {
      "epoch": 0.3745871108299951,
      "grad_norm": 0.48647576570510864,
      "learning_rate": 0.00014772087999464427,
      "loss": 0.1863,
      "step": 26650
    },
    {
      "epoch": 0.3747276688453159,
      "grad_norm": 0.45451873540878296,
      "learning_rate": 0.0001476800267483931,
      "loss": 0.1915,
      "step": 26660
    },
    {
      "epoch": 0.37486822686063676,
      "grad_norm": 0.4005815088748932,
      "learning_rate": 0.00014763916320041843,
      "loss": 0.1648,
      "step": 26670
    },
    {
      "epoch": 0.37500878487595757,
      "grad_norm": 0.4150063395500183,
      "learning_rate": 0.00014759828935954927,
      "loss": 0.1978,
      "step": 26680
    },
    {
      "epoch": 0.3751493428912784,
      "grad_norm": 0.4720441401004791,
      "learning_rate": 0.00014755740523461673,
      "loss": 0.1898,
      "step": 26690
    },
    {
      "epoch": 0.3752899009065992,
      "grad_norm": 0.4665713608264923,
      "learning_rate": 0.0001475165108344543,
      "loss": 0.183,
      "step": 26700
    },
    {
      "epoch": 0.37543045892192,
      "grad_norm": 0.4463876187801361,
      "learning_rate": 0.0001474756061678975,
      "loss": 0.1882,
      "step": 26710
    },
    {
      "epoch": 0.3755710169372408,
      "grad_norm": 0.3561939299106598,
      "learning_rate": 0.0001474346912437843,
      "loss": 0.2042,
      "step": 26720
    },
    {
      "epoch": 0.3757115749525617,
      "grad_norm": 0.4115943908691406,
      "learning_rate": 0.00014739376607095461,
      "loss": 0.217,
      "step": 26730
    },
    {
      "epoch": 0.3758521329678825,
      "grad_norm": 0.5582150220870972,
      "learning_rate": 0.00014735283065825082,
      "loss": 0.1848,
      "step": 26740
    },
    {
      "epoch": 0.3759926909832033,
      "grad_norm": 0.4477035403251648,
      "learning_rate": 0.00014731188501451739,
      "loss": 0.1843,
      "step": 26750
    },
    {
      "epoch": 0.37613324899852413,
      "grad_norm": 0.4303182363510132,
      "learning_rate": 0.00014727092914860096,
      "loss": 0.208,
      "step": 26760
    },
    {
      "epoch": 0.37627380701384494,
      "grad_norm": 0.4030746519565582,
      "learning_rate": 0.0001472299630693505,
      "loss": 0.1797,
      "step": 26770
    },
    {
      "epoch": 0.3764143650291658,
      "grad_norm": 0.37013038992881775,
      "learning_rate": 0.0001471889867856171,
      "loss": 0.1991,
      "step": 26780
    },
    {
      "epoch": 0.3765549230444866,
      "grad_norm": 0.38665181398391724,
      "learning_rate": 0.00014714800030625407,
      "loss": 0.1945,
      "step": 26790
    },
    {
      "epoch": 0.37669548105980744,
      "grad_norm": 0.5154008865356445,
      "learning_rate": 0.0001471070036401169,
      "loss": 0.1846,
      "step": 26800
    },
    {
      "epoch": 0.37683603907512825,
      "grad_norm": 0.5148115754127502,
      "learning_rate": 0.00014706599679606338,
      "loss": 0.1935,
      "step": 26810
    },
    {
      "epoch": 0.37697659709044906,
      "grad_norm": 0.33753859996795654,
      "learning_rate": 0.00014702497978295339,
      "loss": 0.1942,
      "step": 26820
    },
    {
      "epoch": 0.37711715510576993,
      "grad_norm": 0.38848909735679626,
      "learning_rate": 0.00014698395260964904,
      "loss": 0.1692,
      "step": 26830
    },
    {
      "epoch": 0.37725771312109074,
      "grad_norm": 0.43191730976104736,
      "learning_rate": 0.00014694291528501464,
      "loss": 0.1856,
      "step": 26840
    },
    {
      "epoch": 0.37739827113641156,
      "grad_norm": 0.384564608335495,
      "learning_rate": 0.00014690186781791674,
      "loss": 0.1801,
      "step": 26850
    },
    {
      "epoch": 0.37753882915173237,
      "grad_norm": 0.3669491112232208,
      "learning_rate": 0.00014686081021722392,
      "loss": 0.1794,
      "step": 26860
    },
    {
      "epoch": 0.3776793871670532,
      "grad_norm": 0.4367958903312683,
      "learning_rate": 0.0001468197424918072,
      "loss": 0.1858,
      "step": 26870
    },
    {
      "epoch": 0.37781994518237405,
      "grad_norm": 0.388022243976593,
      "learning_rate": 0.0001467786646505396,
      "loss": 0.168,
      "step": 26880
    },
    {
      "epoch": 0.37796050319769486,
      "grad_norm": 0.4472343623638153,
      "learning_rate": 0.00014673757670229635,
      "loss": 0.1801,
      "step": 26890
    },
    {
      "epoch": 0.3781010612130157,
      "grad_norm": 0.36750322580337524,
      "learning_rate": 0.00014669647865595494,
      "loss": 0.1877,
      "step": 26900
    },
    {
      "epoch": 0.3782416192283365,
      "grad_norm": 0.3891570568084717,
      "learning_rate": 0.00014665537052039498,
      "loss": 0.1732,
      "step": 26910
    },
    {
      "epoch": 0.3783821772436573,
      "grad_norm": 0.41716599464416504,
      "learning_rate": 0.00014661425230449825,
      "loss": 0.1946,
      "step": 26920
    },
    {
      "epoch": 0.3785227352589781,
      "grad_norm": 0.37398362159729004,
      "learning_rate": 0.00014657312401714873,
      "loss": 0.2045,
      "step": 26930
    },
    {
      "epoch": 0.378663293274299,
      "grad_norm": 0.3927588164806366,
      "learning_rate": 0.0001465319856672326,
      "loss": 0.1902,
      "step": 26940
    },
    {
      "epoch": 0.3788038512896198,
      "grad_norm": 0.5926169157028198,
      "learning_rate": 0.00014649083726363818,
      "loss": 0.1974,
      "step": 26950
    },
    {
      "epoch": 0.3789444093049406,
      "grad_norm": 0.3898915946483612,
      "learning_rate": 0.000146449678815256,
      "loss": 0.2059,
      "step": 26960
    },
    {
      "epoch": 0.3790849673202614,
      "grad_norm": 0.4634039103984833,
      "learning_rate": 0.00014640851033097873,
      "loss": 0.2182,
      "step": 26970
    },
    {
      "epoch": 0.37922552533558224,
      "grad_norm": 0.42778122425079346,
      "learning_rate": 0.00014636733181970117,
      "loss": 0.2049,
      "step": 26980
    },
    {
      "epoch": 0.3793660833509031,
      "grad_norm": 0.42831602692604065,
      "learning_rate": 0.00014632614329032037,
      "loss": 0.1908,
      "step": 26990
    },
    {
      "epoch": 0.3795066413662239,
      "grad_norm": 0.600955605506897,
      "learning_rate": 0.0001462849447517355,
      "loss": 0.1944,
      "step": 27000
    },
    {
      "epoch": 0.3795066413662239,
      "eval_chrf": 81.85665205492397,
      "eval_loss": 0.4122803807258606,
      "eval_runtime": 268.0612,
      "eval_samples_per_second": 0.373,
      "eval_steps_per_second": 0.015,
      "step": 27000
    },
    {
      "epoch": 0.37964719938154473,
      "grad_norm": 0.45613276958465576,
      "learning_rate": 0.00014624373621284785,
      "loss": 0.1937,
      "step": 27010
    },
    {
      "epoch": 0.37978775739686554,
      "grad_norm": 0.4588741362094879,
      "learning_rate": 0.000146202517682561,
      "loss": 0.1972,
      "step": 27020
    },
    {
      "epoch": 0.37992831541218636,
      "grad_norm": 0.605856716632843,
      "learning_rate": 0.00014616128916978053,
      "loss": 0.1955,
      "step": 27030
    },
    {
      "epoch": 0.3800688734275072,
      "grad_norm": 0.5468184351921082,
      "learning_rate": 0.00014612005068341427,
      "loss": 0.2204,
      "step": 27040
    },
    {
      "epoch": 0.38020943144282804,
      "grad_norm": 0.37999919056892395,
      "learning_rate": 0.00014607880223237223,
      "loss": 0.18,
      "step": 27050
    },
    {
      "epoch": 0.38034998945814885,
      "grad_norm": 0.37661004066467285,
      "learning_rate": 0.00014603754382556649,
      "loss": 0.1853,
      "step": 27060
    },
    {
      "epoch": 0.38049054747346966,
      "grad_norm": 0.4400460720062256,
      "learning_rate": 0.00014599627547191133,
      "loss": 0.2043,
      "step": 27070
    },
    {
      "epoch": 0.3806311054887905,
      "grad_norm": 0.3990391790866852,
      "learning_rate": 0.00014595499718032318,
      "loss": 0.1889,
      "step": 27080
    },
    {
      "epoch": 0.38077166350411135,
      "grad_norm": 0.34186676144599915,
      "learning_rate": 0.00014591370895972058,
      "loss": 0.2308,
      "step": 27090
    },
    {
      "epoch": 0.38091222151943216,
      "grad_norm": 0.42443957924842834,
      "learning_rate": 0.00014587241081902427,
      "loss": 0.1986,
      "step": 27100
    },
    {
      "epoch": 0.381052779534753,
      "grad_norm": 0.4864816665649414,
      "learning_rate": 0.00014583110276715714,
      "loss": 0.178,
      "step": 27110
    },
    {
      "epoch": 0.3811933375500738,
      "grad_norm": 0.3732251524925232,
      "learning_rate": 0.0001457897848130441,
      "loss": 0.1863,
      "step": 27120
    },
    {
      "epoch": 0.3813338955653946,
      "grad_norm": 0.33789321780204773,
      "learning_rate": 0.00014574845696561234,
      "loss": 0.203,
      "step": 27130
    },
    {
      "epoch": 0.38147445358071547,
      "grad_norm": 0.3282896876335144,
      "learning_rate": 0.00014570711923379115,
      "loss": 0.1823,
      "step": 27140
    },
    {
      "epoch": 0.3816150115960363,
      "grad_norm": 0.3856814205646515,
      "learning_rate": 0.00014566577162651188,
      "loss": 0.173,
      "step": 27150
    },
    {
      "epoch": 0.3817555696113571,
      "grad_norm": 0.2681775391101837,
      "learning_rate": 0.00014562441415270813,
      "loss": 0.1774,
      "step": 27160
    },
    {
      "epoch": 0.3818961276266779,
      "grad_norm": 0.4621484875679016,
      "learning_rate": 0.00014558304682131556,
      "loss": 0.2096,
      "step": 27170
    },
    {
      "epoch": 0.3820366856419987,
      "grad_norm": 0.4743730425834656,
      "learning_rate": 0.000145541669641272,
      "loss": 0.189,
      "step": 27180
    },
    {
      "epoch": 0.38217724365731953,
      "grad_norm": 0.4920254349708557,
      "learning_rate": 0.00014550028262151732,
      "loss": 0.2073,
      "step": 27190
    },
    {
      "epoch": 0.3823178016726404,
      "grad_norm": 0.3450794219970703,
      "learning_rate": 0.00014545888577099364,
      "loss": 0.1805,
      "step": 27200
    },
    {
      "epoch": 0.3824583596879612,
      "grad_norm": 0.4046187698841095,
      "learning_rate": 0.0001454174790986451,
      "loss": 0.1976,
      "step": 27210
    },
    {
      "epoch": 0.382598917703282,
      "grad_norm": 0.38482755422592163,
      "learning_rate": 0.00014537606261341802,
      "loss": 0.1744,
      "step": 27220
    },
    {
      "epoch": 0.38273947571860284,
      "grad_norm": 0.4715237021446228,
      "learning_rate": 0.00014533463632426086,
      "loss": 0.1769,
      "step": 27230
    },
    {
      "epoch": 0.38288003373392365,
      "grad_norm": 0.3997655212879181,
      "learning_rate": 0.0001452932002401241,
      "loss": 0.1845,
      "step": 27240
    },
    {
      "epoch": 0.3830205917492445,
      "grad_norm": 0.5396472811698914,
      "learning_rate": 0.00014525175436996048,
      "loss": 0.216,
      "step": 27250
    },
    {
      "epoch": 0.38316114976456533,
      "grad_norm": 0.4229860305786133,
      "learning_rate": 0.00014521029872272473,
      "loss": 0.1819,
      "step": 27260
    },
    {
      "epoch": 0.38330170777988615,
      "grad_norm": 0.5465221405029297,
      "learning_rate": 0.0001451688333073737,
      "loss": 0.2071,
      "step": 27270
    },
    {
      "epoch": 0.38344226579520696,
      "grad_norm": 0.46100810170173645,
      "learning_rate": 0.00014512735813286643,
      "loss": 0.1994,
      "step": 27280
    },
    {
      "epoch": 0.3835828238105278,
      "grad_norm": 0.5227595567703247,
      "learning_rate": 0.00014508587320816404,
      "loss": 0.1996,
      "step": 27290
    },
    {
      "epoch": 0.38372338182584864,
      "grad_norm": 0.36178404092788696,
      "learning_rate": 0.00014504437854222975,
      "loss": 0.1898,
      "step": 27300
    },
    {
      "epoch": 0.38386393984116945,
      "grad_norm": 0.371266633272171,
      "learning_rate": 0.0001450028741440288,
      "loss": 0.1782,
      "step": 27310
    },
    {
      "epoch": 0.38400449785649027,
      "grad_norm": 0.3904975652694702,
      "learning_rate": 0.0001449613600225287,
      "loss": 0.1673,
      "step": 27320
    },
    {
      "epoch": 0.3841450558718111,
      "grad_norm": 0.3872131109237671,
      "learning_rate": 0.00014491983618669888,
      "loss": 0.1851,
      "step": 27330
    },
    {
      "epoch": 0.3842856138871319,
      "grad_norm": 0.4446090757846832,
      "learning_rate": 0.0001448783026455111,
      "loss": 0.1924,
      "step": 27340
    },
    {
      "epoch": 0.38442617190245276,
      "grad_norm": 0.3272359073162079,
      "learning_rate": 0.0001448367594079389,
      "loss": 0.1792,
      "step": 27350
    },
    {
      "epoch": 0.3845667299177736,
      "grad_norm": 0.4467935860157013,
      "learning_rate": 0.0001447952064829582,
      "loss": 0.2026,
      "step": 27360
    },
    {
      "epoch": 0.3847072879330944,
      "grad_norm": 0.40729010105133057,
      "learning_rate": 0.00014475364387954692,
      "loss": 0.2163,
      "step": 27370
    },
    {
      "epoch": 0.3848478459484152,
      "grad_norm": 0.42603397369384766,
      "learning_rate": 0.000144712071606685,
      "loss": 0.1802,
      "step": 27380
    },
    {
      "epoch": 0.384988403963736,
      "grad_norm": 0.34360334277153015,
      "learning_rate": 0.00014467048967335458,
      "loss": 0.1806,
      "step": 27390
    },
    {
      "epoch": 0.3851289619790569,
      "grad_norm": 0.5067285895347595,
      "learning_rate": 0.00014462889808853973,
      "loss": 0.1905,
      "step": 27400
    },
    {
      "epoch": 0.3852695199943777,
      "grad_norm": 0.42074596881866455,
      "learning_rate": 0.00014458729686122682,
      "loss": 0.1928,
      "step": 27410
    },
    {
      "epoch": 0.3854100780096985,
      "grad_norm": 0.4252513647079468,
      "learning_rate": 0.00014454568600040418,
      "loss": 0.179,
      "step": 27420
    },
    {
      "epoch": 0.3855506360250193,
      "grad_norm": 0.36288729310035706,
      "learning_rate": 0.00014450406551506215,
      "loss": 0.1901,
      "step": 27430
    },
    {
      "epoch": 0.38569119404034014,
      "grad_norm": 0.4799077808856964,
      "learning_rate": 0.0001444624354141933,
      "loss": 0.1747,
      "step": 27440
    },
    {
      "epoch": 0.38583175205566095,
      "grad_norm": 0.3890087902545929,
      "learning_rate": 0.0001444207957067922,
      "loss": 0.1786,
      "step": 27450
    },
    {
      "epoch": 0.3859723100709818,
      "grad_norm": 0.3378077745437622,
      "learning_rate": 0.0001443791464018555,
      "loss": 0.1662,
      "step": 27460
    },
    {
      "epoch": 0.38611286808630263,
      "grad_norm": 0.41123804450035095,
      "learning_rate": 0.00014433748750838191,
      "loss": 0.1993,
      "step": 27470
    },
    {
      "epoch": 0.38625342610162344,
      "grad_norm": 0.34629175066947937,
      "learning_rate": 0.00014429581903537225,
      "loss": 0.1914,
      "step": 27480
    },
    {
      "epoch": 0.38639398411694426,
      "grad_norm": 0.5317151546478271,
      "learning_rate": 0.0001442541409918294,
      "loss": 0.1862,
      "step": 27490
    },
    {
      "epoch": 0.38653454213226507,
      "grad_norm": 0.4020622968673706,
      "learning_rate": 0.00014421245338675824,
      "loss": 0.2017,
      "step": 27500
    },
    {
      "epoch": 0.38667510014758594,
      "grad_norm": 0.4146219491958618,
      "learning_rate": 0.00014417075622916585,
      "loss": 0.1717,
      "step": 27510
    },
    {
      "epoch": 0.38681565816290675,
      "grad_norm": 0.45714467763900757,
      "learning_rate": 0.00014412904952806126,
      "loss": 0.2013,
      "step": 27520
    },
    {
      "epoch": 0.38695621617822756,
      "grad_norm": 0.3680949807167053,
      "learning_rate": 0.0001440873332924556,
      "loss": 0.2321,
      "step": 27530
    },
    {
      "epoch": 0.3870967741935484,
      "grad_norm": 0.5289217829704285,
      "learning_rate": 0.00014404560753136209,
      "loss": 0.1907,
      "step": 27540
    },
    {
      "epoch": 0.3872373322088692,
      "grad_norm": 0.4522726833820343,
      "learning_rate": 0.00014400387225379594,
      "loss": 0.1902,
      "step": 27550
    },
    {
      "epoch": 0.38737789022419006,
      "grad_norm": 0.37127548456192017,
      "learning_rate": 0.00014396212746877444,
      "loss": 0.1961,
      "step": 27560
    },
    {
      "epoch": 0.38751844823951087,
      "grad_norm": 0.4859474301338196,
      "learning_rate": 0.000143920373185317,
      "loss": 0.209,
      "step": 27570
    },
    {
      "epoch": 0.3876590062548317,
      "grad_norm": 0.4222281873226166,
      "learning_rate": 0.000143878609412445,
      "loss": 0.156,
      "step": 27580
    },
    {
      "epoch": 0.3877995642701525,
      "grad_norm": 0.34376925230026245,
      "learning_rate": 0.00014383683615918188,
      "loss": 0.1735,
      "step": 27590
    },
    {
      "epoch": 0.3879401222854733,
      "grad_norm": 0.44502338767051697,
      "learning_rate": 0.00014379505343455322,
      "loss": 0.1869,
      "step": 27600
    },
    {
      "epoch": 0.3880806803007942,
      "grad_norm": 0.36206844449043274,
      "learning_rate": 0.0001437532612475865,
      "loss": 0.1988,
      "step": 27610
    },
    {
      "epoch": 0.388221238316115,
      "grad_norm": 0.46434277296066284,
      "learning_rate": 0.00014371145960731136,
      "loss": 0.1987,
      "step": 27620
    },
    {
      "epoch": 0.3883617963314358,
      "grad_norm": 0.41770943999290466,
      "learning_rate": 0.00014366964852275942,
      "loss": 0.1728,
      "step": 27630
    },
    {
      "epoch": 0.3885023543467566,
      "grad_norm": 0.4312964081764221,
      "learning_rate": 0.00014362782800296438,
      "loss": 0.2,
      "step": 27640
    },
    {
      "epoch": 0.38864291236207743,
      "grad_norm": 0.508865475654602,
      "learning_rate": 0.00014358599805696195,
      "loss": 0.1996,
      "step": 27650
    },
    {
      "epoch": 0.38878347037739824,
      "grad_norm": 0.4049730896949768,
      "learning_rate": 0.0001435441586937899,
      "loss": 0.2066,
      "step": 27660
    },
    {
      "epoch": 0.3889240283927191,
      "grad_norm": 0.509647011756897,
      "learning_rate": 0.000143502309922488,
      "loss": 0.207,
      "step": 27670
    },
    {
      "epoch": 0.3890645864080399,
      "grad_norm": 0.3966078758239746,
      "learning_rate": 0.0001434604517520981,
      "loss": 0.1784,
      "step": 27680
    },
    {
      "epoch": 0.38920514442336074,
      "grad_norm": 0.4309219419956207,
      "learning_rate": 0.00014341858419166407,
      "loss": 0.2181,
      "step": 27690
    },
    {
      "epoch": 0.38934570243868155,
      "grad_norm": 0.4796231687068939,
      "learning_rate": 0.00014337670725023173,
      "loss": 0.2018,
      "step": 27700
    },
    {
      "epoch": 0.38948626045400236,
      "grad_norm": 0.40402162075042725,
      "learning_rate": 0.00014333482093684908,
      "loss": 0.2026,
      "step": 27710
    },
    {
      "epoch": 0.38962681846932323,
      "grad_norm": 0.4409323036670685,
      "learning_rate": 0.00014329292526056597,
      "loss": 0.1912,
      "step": 27720
    },
    {
      "epoch": 0.38976737648464405,
      "grad_norm": 0.4390415549278259,
      "learning_rate": 0.0001432510202304344,
      "loss": 0.2037,
      "step": 27730
    },
    {
      "epoch": 0.38990793449996486,
      "grad_norm": 0.49357882142066956,
      "learning_rate": 0.00014320910585550838,
      "loss": 0.1859,
      "step": 27740
    },
    {
      "epoch": 0.39004849251528567,
      "grad_norm": 0.414014995098114,
      "learning_rate": 0.00014316718214484384,
      "loss": 0.2,
      "step": 27750
    },
    {
      "epoch": 0.3901890505306065,
      "grad_norm": 0.45711779594421387,
      "learning_rate": 0.00014312524910749884,
      "loss": 0.1848,
      "step": 27760
    },
    {
      "epoch": 0.39032960854592735,
      "grad_norm": 0.5342926979064941,
      "learning_rate": 0.00014308330675253346,
      "loss": 0.1993,
      "step": 27770
    },
    {
      "epoch": 0.39047016656124817,
      "grad_norm": 0.4213629961013794,
      "learning_rate": 0.00014304135508900963,
      "loss": 0.1966,
      "step": 27780
    },
    {
      "epoch": 0.390610724576569,
      "grad_norm": 0.4413548409938812,
      "learning_rate": 0.0001429993941259915,
      "loss": 0.1938,
      "step": 27790
    },
    {
      "epoch": 0.3907512825918898,
      "grad_norm": 0.3546472191810608,
      "learning_rate": 0.00014295742387254512,
      "loss": 0.1861,
      "step": 27800
    },
    {
      "epoch": 0.3908918406072106,
      "grad_norm": 0.4804078936576843,
      "learning_rate": 0.0001429154443377385,
      "loss": 0.2145,
      "step": 27810
    },
    {
      "epoch": 0.3910323986225315,
      "grad_norm": 0.4604095220565796,
      "learning_rate": 0.00014287345553064178,
      "loss": 0.1848,
      "step": 27820
    },
    {
      "epoch": 0.3911729566378523,
      "grad_norm": 0.36869752407073975,
      "learning_rate": 0.00014283145746032705,
      "loss": 0.1816,
      "step": 27830
    },
    {
      "epoch": 0.3913135146531731,
      "grad_norm": 0.34594419598579407,
      "learning_rate": 0.00014278945013586835,
      "loss": 0.1638,
      "step": 27840
    },
    {
      "epoch": 0.3914540726684939,
      "grad_norm": 0.41775190830230713,
      "learning_rate": 0.0001427474335663418,
      "loss": 0.22,
      "step": 27850
    },
    {
      "epoch": 0.3915946306838147,
      "grad_norm": 0.3673890233039856,
      "learning_rate": 0.00014270540776082543,
      "loss": 0.2051,
      "step": 27860
    },
    {
      "epoch": 0.3917351886991356,
      "grad_norm": 0.379302978515625,
      "learning_rate": 0.00014266337272839935,
      "loss": 0.1975,
      "step": 27870
    },
    {
      "epoch": 0.3918757467144564,
      "grad_norm": 0.4624926447868347,
      "learning_rate": 0.00014262132847814565,
      "loss": 0.1911,
      "step": 27880
    },
    {
      "epoch": 0.3920163047297772,
      "grad_norm": 0.3992692530155182,
      "learning_rate": 0.00014257927501914834,
      "loss": 0.1886,
      "step": 27890
    },
    {
      "epoch": 0.39215686274509803,
      "grad_norm": 0.3637385964393616,
      "learning_rate": 0.0001425372123604935,
      "loss": 0.1959,
      "step": 27900
    },
    {
      "epoch": 0.39229742076041885,
      "grad_norm": 0.44345909357070923,
      "learning_rate": 0.00014249514051126917,
      "loss": 0.1772,
      "step": 27910
    },
    {
      "epoch": 0.39243797877573966,
      "grad_norm": 0.44268521666526794,
      "learning_rate": 0.00014245305948056534,
      "loss": 0.187,
      "step": 27920
    },
    {
      "epoch": 0.39257853679106053,
      "grad_norm": 0.39628198742866516,
      "learning_rate": 0.00014241096927747408,
      "loss": 0.1715,
      "step": 27930
    },
    {
      "epoch": 0.39271909480638134,
      "grad_norm": 0.4096054434776306,
      "learning_rate": 0.00014236886991108932,
      "loss": 0.1934,
      "step": 27940
    },
    {
      "epoch": 0.39285965282170215,
      "grad_norm": 0.5102948546409607,
      "learning_rate": 0.00014232676139050706,
      "loss": 0.2156,
      "step": 27950
    },
    {
      "epoch": 0.39300021083702297,
      "grad_norm": 0.485239714384079,
      "learning_rate": 0.00014228464372482524,
      "loss": 0.1939,
      "step": 27960
    },
    {
      "epoch": 0.3931407688523438,
      "grad_norm": 0.44243955612182617,
      "learning_rate": 0.00014224251692314377,
      "loss": 0.2131,
      "step": 27970
    },
    {
      "epoch": 0.39328132686766465,
      "grad_norm": 0.47501078248023987,
      "learning_rate": 0.00014220038099456455,
      "loss": 0.1921,
      "step": 27980
    },
    {
      "epoch": 0.39342188488298546,
      "grad_norm": 0.5273157358169556,
      "learning_rate": 0.00014215823594819146,
      "loss": 0.2171,
      "step": 27990
    },
    {
      "epoch": 0.3935624428983063,
      "grad_norm": 0.6149890422821045,
      "learning_rate": 0.00014211608179313034,
      "loss": 0.1935,
      "step": 28000
    },
    {
      "epoch": 0.3935624428983063,
      "eval_chrf": 77.61258146719794,
      "eval_loss": 0.41459283232688904,
      "eval_runtime": 260.8751,
      "eval_samples_per_second": 0.383,
      "eval_steps_per_second": 0.015,
      "step": 28000
    },
    {
      "epoch": 0.3937030009136271,
      "grad_norm": 0.3866507411003113,
      "learning_rate": 0.000142073918538489,
      "loss": 0.1956,
      "step": 28010
    },
    {
      "epoch": 0.3938435589289479,
      "grad_norm": 0.4531294107437134,
      "learning_rate": 0.00014203174619337714,
      "loss": 0.1747,
      "step": 28020
    },
    {
      "epoch": 0.39398411694426877,
      "grad_norm": 0.5326635241508484,
      "learning_rate": 0.00014198956476690663,
      "loss": 0.19,
      "step": 28030
    },
    {
      "epoch": 0.3941246749595896,
      "grad_norm": 0.3673665523529053,
      "learning_rate": 0.00014194737426819108,
      "loss": 0.1926,
      "step": 28040
    },
    {
      "epoch": 0.3942652329749104,
      "grad_norm": 0.39492812752723694,
      "learning_rate": 0.00014190517470634616,
      "loss": 0.1817,
      "step": 28050
    },
    {
      "epoch": 0.3944057909902312,
      "grad_norm": 0.41755416989326477,
      "learning_rate": 0.00014186296609048946,
      "loss": 0.1698,
      "step": 28060
    },
    {
      "epoch": 0.394546349005552,
      "grad_norm": 0.44416022300720215,
      "learning_rate": 0.0001418207484297406,
      "loss": 0.1725,
      "step": 28070
    },
    {
      "epoch": 0.3946869070208729,
      "grad_norm": 0.44826391339302063,
      "learning_rate": 0.00014177852173322112,
      "loss": 0.1837,
      "step": 28080
    },
    {
      "epoch": 0.3948274650361937,
      "grad_norm": 0.4329129755496979,
      "learning_rate": 0.00014173628601005445,
      "loss": 0.1838,
      "step": 28090
    },
    {
      "epoch": 0.3949680230515145,
      "grad_norm": 0.4439648389816284,
      "learning_rate": 0.00014169404126936603,
      "loss": 0.1995,
      "step": 28100
    },
    {
      "epoch": 0.39510858106683533,
      "grad_norm": 0.3365096151828766,
      "learning_rate": 0.00014165178752028327,
      "loss": 0.2013,
      "step": 28110
    },
    {
      "epoch": 0.39524913908215614,
      "grad_norm": 0.4657164216041565,
      "learning_rate": 0.00014160952477193547,
      "loss": 0.2057,
      "step": 28120
    },
    {
      "epoch": 0.395389697097477,
      "grad_norm": 0.36999014019966125,
      "learning_rate": 0.0001415672530334539,
      "loss": 0.175,
      "step": 28130
    },
    {
      "epoch": 0.3955302551127978,
      "grad_norm": 0.3687359094619751,
      "learning_rate": 0.00014152497231397176,
      "loss": 0.1886,
      "step": 28140
    },
    {
      "epoch": 0.39567081312811864,
      "grad_norm": 0.5541471242904663,
      "learning_rate": 0.00014148268262262426,
      "loss": 0.192,
      "step": 28150
    },
    {
      "epoch": 0.39581137114343945,
      "grad_norm": 0.5069437623023987,
      "learning_rate": 0.00014144038396854842,
      "loss": 0.2007,
      "step": 28160
    },
    {
      "epoch": 0.39595192915876026,
      "grad_norm": 0.4327186942100525,
      "learning_rate": 0.0001413980763608833,
      "loss": 0.1938,
      "step": 28170
    },
    {
      "epoch": 0.3960924871740811,
      "grad_norm": 0.37074199318885803,
      "learning_rate": 0.00014135575980876982,
      "loss": 0.1907,
      "step": 28180
    },
    {
      "epoch": 0.39623304518940194,
      "grad_norm": 0.3386900722980499,
      "learning_rate": 0.00014131343432135095,
      "loss": 0.1834,
      "step": 28190
    },
    {
      "epoch": 0.39637360320472276,
      "grad_norm": 0.42331430315971375,
      "learning_rate": 0.00014127109990777145,
      "loss": 0.1741,
      "step": 28200
    },
    {
      "epoch": 0.39651416122004357,
      "grad_norm": 0.4919547736644745,
      "learning_rate": 0.0001412287565771781,
      "loss": 0.1972,
      "step": 28210
    },
    {
      "epoch": 0.3966547192353644,
      "grad_norm": 0.4163661301136017,
      "learning_rate": 0.00014118640433871956,
      "loss": 0.204,
      "step": 28220
    },
    {
      "epoch": 0.3967952772506852,
      "grad_norm": 0.4148716628551483,
      "learning_rate": 0.00014114404320154648,
      "loss": 0.1884,
      "step": 28230
    },
    {
      "epoch": 0.39693583526600607,
      "grad_norm": 0.3614480197429657,
      "learning_rate": 0.00014110167317481133,
      "loss": 0.1894,
      "step": 28240
    },
    {
      "epoch": 0.3970763932813269,
      "grad_norm": 0.5240013003349304,
      "learning_rate": 0.00014105929426766863,
      "loss": 0.2107,
      "step": 28250
    },
    {
      "epoch": 0.3972169512966477,
      "grad_norm": 0.44796907901763916,
      "learning_rate": 0.00014101690648927464,
      "loss": 0.1985,
      "step": 28260
    },
    {
      "epoch": 0.3973575093119685,
      "grad_norm": 0.451960951089859,
      "learning_rate": 0.00014097450984878774,
      "loss": 0.1878,
      "step": 28270
    },
    {
      "epoch": 0.3974980673272893,
      "grad_norm": 0.5274572372436523,
      "learning_rate": 0.00014093210435536807,
      "loss": 0.2039,
      "step": 28280
    },
    {
      "epoch": 0.3976386253426102,
      "grad_norm": 0.5012400150299072,
      "learning_rate": 0.00014088969001817777,
      "loss": 0.2114,
      "step": 28290
    },
    {
      "epoch": 0.397779183357931,
      "grad_norm": 0.398480087518692,
      "learning_rate": 0.00014084726684638086,
      "loss": 0.2074,
      "step": 28300
    },
    {
      "epoch": 0.3979197413732518,
      "grad_norm": 0.5972045660018921,
      "learning_rate": 0.00014080483484914323,
      "loss": 0.2121,
      "step": 28310
    },
    {
      "epoch": 0.3980602993885726,
      "grad_norm": 0.37806838750839233,
      "learning_rate": 0.00014076239403563277,
      "loss": 0.1786,
      "step": 28320
    },
    {
      "epoch": 0.39820085740389344,
      "grad_norm": 0.31259986758232117,
      "learning_rate": 0.0001407199444150192,
      "loss": 0.1867,
      "step": 28330
    },
    {
      "epoch": 0.3983414154192143,
      "grad_norm": 0.47804269194602966,
      "learning_rate": 0.00014067748599647418,
      "loss": 0.1821,
      "step": 28340
    },
    {
      "epoch": 0.3984819734345351,
      "grad_norm": 0.30588817596435547,
      "learning_rate": 0.0001406350187891712,
      "loss": 0.1988,
      "step": 28350
    },
    {
      "epoch": 0.39862253144985593,
      "grad_norm": 0.4533221423625946,
      "learning_rate": 0.00014059254280228574,
      "loss": 0.1784,
      "step": 28360
    },
    {
      "epoch": 0.39876308946517675,
      "grad_norm": 0.41111546754837036,
      "learning_rate": 0.0001405500580449952,
      "loss": 0.2089,
      "step": 28370
    },
    {
      "epoch": 0.39890364748049756,
      "grad_norm": 0.3993707597255707,
      "learning_rate": 0.00014050756452647873,
      "loss": 0.1705,
      "step": 28380
    },
    {
      "epoch": 0.39904420549581837,
      "grad_norm": 0.398714542388916,
      "learning_rate": 0.00014046506225591748,
      "loss": 0.1802,
      "step": 28390
    },
    {
      "epoch": 0.39918476351113924,
      "grad_norm": 0.4545983076095581,
      "learning_rate": 0.0001404225512424945,
      "loss": 0.2177,
      "step": 28400
    },
    {
      "epoch": 0.39932532152646005,
      "grad_norm": 0.438765287399292,
      "learning_rate": 0.00014038003149539468,
      "loss": 0.2054,
      "step": 28410
    },
    {
      "epoch": 0.39946587954178087,
      "grad_norm": 0.49160799384117126,
      "learning_rate": 0.0001403375030238048,
      "loss": 0.1798,
      "step": 28420
    },
    {
      "epoch": 0.3996064375571017,
      "grad_norm": 0.41309601068496704,
      "learning_rate": 0.00014029496583691358,
      "loss": 0.2224,
      "step": 28430
    },
    {
      "epoch": 0.3997469955724225,
      "grad_norm": 0.43208959698677063,
      "learning_rate": 0.0001402524199439115,
      "loss": 0.2046,
      "step": 28440
    },
    {
      "epoch": 0.39988755358774336,
      "grad_norm": 0.3021363914012909,
      "learning_rate": 0.00014020986535399113,
      "loss": 0.1961,
      "step": 28450
    },
    {
      "epoch": 0.4000281116030642,
      "grad_norm": 0.4002769887447357,
      "learning_rate": 0.00014016730207634673,
      "loss": 0.1969,
      "step": 28460
    },
    {
      "epoch": 0.400168669618385,
      "grad_norm": 0.3687337040901184,
      "learning_rate": 0.00014012473012017446,
      "loss": 0.2353,
      "step": 28470
    },
    {
      "epoch": 0.4003092276337058,
      "grad_norm": 0.41458526253700256,
      "learning_rate": 0.00014008214949467248,
      "loss": 0.1902,
      "step": 28480
    },
    {
      "epoch": 0.4004497856490266,
      "grad_norm": 0.3621925413608551,
      "learning_rate": 0.00014003956020904065,
      "loss": 0.1678,
      "step": 28490
    },
    {
      "epoch": 0.4005903436643475,
      "grad_norm": 0.6078202724456787,
      "learning_rate": 0.00013999696227248083,
      "loss": 0.1908,
      "step": 28500
    },
    {
      "epoch": 0.4007309016796683,
      "grad_norm": 0.4965280592441559,
      "learning_rate": 0.00013995435569419678,
      "loss": 0.1933,
      "step": 28510
    },
    {
      "epoch": 0.4008714596949891,
      "grad_norm": 0.29492101073265076,
      "learning_rate": 0.00013991174048339392,
      "loss": 0.1781,
      "step": 28520
    },
    {
      "epoch": 0.4010120177103099,
      "grad_norm": 0.3656838536262512,
      "learning_rate": 0.0001398691166492798,
      "loss": 0.1667,
      "step": 28530
    },
    {
      "epoch": 0.40115257572563073,
      "grad_norm": 0.42332929372787476,
      "learning_rate": 0.00013982648420106362,
      "loss": 0.1977,
      "step": 28540
    },
    {
      "epoch": 0.4012931337409516,
      "grad_norm": 0.5157352685928345,
      "learning_rate": 0.00013978384314795655,
      "loss": 0.1909,
      "step": 28550
    },
    {
      "epoch": 0.4014336917562724,
      "grad_norm": 0.4120025038719177,
      "learning_rate": 0.00013974119349917163,
      "loss": 0.173,
      "step": 28560
    },
    {
      "epoch": 0.40157424977159323,
      "grad_norm": 0.3541117012500763,
      "learning_rate": 0.00013969853526392368,
      "loss": 0.1934,
      "step": 28570
    },
    {
      "epoch": 0.40171480778691404,
      "grad_norm": 0.5453256368637085,
      "learning_rate": 0.00013965586845142943,
      "loss": 0.1927,
      "step": 28580
    },
    {
      "epoch": 0.40185536580223485,
      "grad_norm": 0.43507325649261475,
      "learning_rate": 0.00013961319307090743,
      "loss": 0.1848,
      "step": 28590
    },
    {
      "epoch": 0.4019959238175557,
      "grad_norm": 0.37308594584465027,
      "learning_rate": 0.00013957050913157814,
      "loss": 0.1835,
      "step": 28600
    },
    {
      "epoch": 0.40213648183287654,
      "grad_norm": 0.4535294473171234,
      "learning_rate": 0.0001395278166426638,
      "loss": 0.2153,
      "step": 28610
    },
    {
      "epoch": 0.40227703984819735,
      "grad_norm": 0.4658803939819336,
      "learning_rate": 0.00013948511561338858,
      "loss": 0.1854,
      "step": 28620
    },
    {
      "epoch": 0.40241759786351816,
      "grad_norm": 0.4867015480995178,
      "learning_rate": 0.0001394424060529784,
      "loss": 0.2026,
      "step": 28630
    },
    {
      "epoch": 0.402558155878839,
      "grad_norm": 0.4584827125072479,
      "learning_rate": 0.00013939968797066103,
      "loss": 0.1915,
      "step": 28640
    },
    {
      "epoch": 0.4026987138941598,
      "grad_norm": 0.37534889578819275,
      "learning_rate": 0.0001393569613756662,
      "loss": 0.1783,
      "step": 28650
    },
    {
      "epoch": 0.40283927190948066,
      "grad_norm": 0.3874412477016449,
      "learning_rate": 0.00013931422627722534,
      "loss": 0.1889,
      "step": 28660
    },
    {
      "epoch": 0.40297982992480147,
      "grad_norm": 0.5786199569702148,
      "learning_rate": 0.0001392714826845718,
      "loss": 0.2024,
      "step": 28670
    },
    {
      "epoch": 0.4031203879401223,
      "grad_norm": 0.4794784188270569,
      "learning_rate": 0.00013922873060694075,
      "loss": 0.1832,
      "step": 28680
    },
    {
      "epoch": 0.4032609459554431,
      "grad_norm": 0.6809529066085815,
      "learning_rate": 0.00013918597005356916,
      "loss": 0.2182,
      "step": 28690
    },
    {
      "epoch": 0.4034015039707639,
      "grad_norm": 0.40321481227874756,
      "learning_rate": 0.00013914320103369587,
      "loss": 0.1956,
      "step": 28700
    },
    {
      "epoch": 0.4035420619860848,
      "grad_norm": 0.3870124816894531,
      "learning_rate": 0.0001391004235565615,
      "loss": 0.1873,
      "step": 28710
    },
    {
      "epoch": 0.4036826200014056,
      "grad_norm": 0.3738168179988861,
      "learning_rate": 0.00013905763763140858,
      "loss": 0.1874,
      "step": 28720
    },
    {
      "epoch": 0.4038231780167264,
      "grad_norm": 0.38982343673706055,
      "learning_rate": 0.00013901484326748136,
      "loss": 0.2103,
      "step": 28730
    },
    {
      "epoch": 0.4039637360320472,
      "grad_norm": 0.4845069348812103,
      "learning_rate": 0.00013897204047402605,
      "loss": 0.1945,
      "step": 28740
    },
    {
      "epoch": 0.40410429404736803,
      "grad_norm": 0.5125102996826172,
      "learning_rate": 0.00013892922926029053,
      "loss": 0.1922,
      "step": 28750
    },
    {
      "epoch": 0.4042448520626889,
      "grad_norm": 0.4411031901836395,
      "learning_rate": 0.0001388864096355246,
      "loss": 0.2086,
      "step": 28760
    },
    {
      "epoch": 0.4043854100780097,
      "grad_norm": 0.43539199233055115,
      "learning_rate": 0.00013884358160897982,
      "loss": 0.2,
      "step": 28770
    },
    {
      "epoch": 0.4045259680933305,
      "grad_norm": 0.388676255941391,
      "learning_rate": 0.00013880074518990963,
      "loss": 0.1872,
      "step": 28780
    },
    {
      "epoch": 0.40466652610865134,
      "grad_norm": 0.4598192572593689,
      "learning_rate": 0.00013875790038756923,
      "loss": 0.1959,
      "step": 28790
    },
    {
      "epoch": 0.40480708412397215,
      "grad_norm": 0.36540326476097107,
      "learning_rate": 0.00013871504721121565,
      "loss": 0.19,
      "step": 28800
    },
    {
      "epoch": 0.404947642139293,
      "grad_norm": 0.3289833962917328,
      "learning_rate": 0.0001386721856701077,
      "loss": 0.1977,
      "step": 28810
    },
    {
      "epoch": 0.40508820015461383,
      "grad_norm": 0.3701980710029602,
      "learning_rate": 0.00013862931577350606,
      "loss": 0.1801,
      "step": 28820
    },
    {
      "epoch": 0.40522875816993464,
      "grad_norm": 0.3710598647594452,
      "learning_rate": 0.00013858643753067318,
      "loss": 0.1626,
      "step": 28830
    },
    {
      "epoch": 0.40536931618525546,
      "grad_norm": 0.4302135109901428,
      "learning_rate": 0.00013854355095087328,
      "loss": 0.1876,
      "step": 28840
    },
    {
      "epoch": 0.40550987420057627,
      "grad_norm": 0.4434810280799866,
      "learning_rate": 0.00013850065604337244,
      "loss": 0.1891,
      "step": 28850
    },
    {
      "epoch": 0.40565043221589714,
      "grad_norm": 0.4210662841796875,
      "learning_rate": 0.0001384577528174385,
      "loss": 0.1802,
      "step": 28860
    },
    {
      "epoch": 0.40579099023121795,
      "grad_norm": 0.5197582244873047,
      "learning_rate": 0.00013841484128234113,
      "loss": 0.2159,
      "step": 28870
    },
    {
      "epoch": 0.40593154824653876,
      "grad_norm": 0.4137897193431854,
      "learning_rate": 0.00013837192144735175,
      "loss": 0.181,
      "step": 28880
    },
    {
      "epoch": 0.4060721062618596,
      "grad_norm": 0.5064562559127808,
      "learning_rate": 0.00013832899332174358,
      "loss": 0.2087,
      "step": 28890
    },
    {
      "epoch": 0.4062126642771804,
      "grad_norm": 0.4158742427825928,
      "learning_rate": 0.0001382860569147917,
      "loss": 0.2004,
      "step": 28900
    },
    {
      "epoch": 0.4063532222925012,
      "grad_norm": 0.4387105405330658,
      "learning_rate": 0.00013824311223577292,
      "loss": 0.197,
      "step": 28910
    },
    {
      "epoch": 0.4064937803078221,
      "grad_norm": 0.3631752133369446,
      "learning_rate": 0.00013820015929396585,
      "loss": 0.2,
      "step": 28920
    },
    {
      "epoch": 0.4066343383231429,
      "grad_norm": 0.456225723028183,
      "learning_rate": 0.00013815719809865082,
      "loss": 0.2216,
      "step": 28930
    },
    {
      "epoch": 0.4067748963384637,
      "grad_norm": 0.36056461930274963,
      "learning_rate": 0.0001381142286591101,
      "loss": 0.1935,
      "step": 28940
    },
    {
      "epoch": 0.4069154543537845,
      "grad_norm": 0.43140852451324463,
      "learning_rate": 0.00013807125098462756,
      "loss": 0.169,
      "step": 28950
    },
    {
      "epoch": 0.4070560123691053,
      "grad_norm": 0.29992830753326416,
      "learning_rate": 0.00013802826508448897,
      "loss": 0.1701,
      "step": 28960
    },
    {
      "epoch": 0.4071965703844262,
      "grad_norm": 0.3830731213092804,
      "learning_rate": 0.00013798527096798189,
      "loss": 0.1916,
      "step": 28970
    },
    {
      "epoch": 0.407337128399747,
      "grad_norm": 0.3855833113193512,
      "learning_rate": 0.00013794226864439555,
      "loss": 0.1848,
      "step": 28980
    },
    {
      "epoch": 0.4074776864150678,
      "grad_norm": 0.4350045621395111,
      "learning_rate": 0.000137899258123021,
      "loss": 0.1908,
      "step": 28990
    },
    {
      "epoch": 0.40761824443038863,
      "grad_norm": 0.4594937562942505,
      "learning_rate": 0.00013785623941315112,
      "loss": 0.2323,
      "step": 29000
    },
    {
      "epoch": 0.40761824443038863,
      "eval_chrf": 81.69022545145947,
      "eval_loss": 0.40677130222320557,
      "eval_runtime": 275.1454,
      "eval_samples_per_second": 0.363,
      "eval_steps_per_second": 0.015,
      "step": 29000
    },
    {
      "epoch": 0.40775880244570945,
      "grad_norm": 0.34250855445861816,
      "learning_rate": 0.00013781321252408048,
      "loss": 0.2033,
      "step": 29010
    },
    {
      "epoch": 0.4078993604610303,
      "grad_norm": 0.436296284198761,
      "learning_rate": 0.00013777017746510547,
      "loss": 0.2063,
      "step": 29020
    },
    {
      "epoch": 0.4080399184763511,
      "grad_norm": 0.4227631390094757,
      "learning_rate": 0.0001377271342455242,
      "loss": 0.1932,
      "step": 29030
    },
    {
      "epoch": 0.40818047649167194,
      "grad_norm": 0.35152482986450195,
      "learning_rate": 0.00013768408287463657,
      "loss": 0.1758,
      "step": 29040
    },
    {
      "epoch": 0.40832103450699275,
      "grad_norm": 0.4631098508834839,
      "learning_rate": 0.00013764102336174427,
      "loss": 0.2096,
      "step": 29050
    },
    {
      "epoch": 0.40846159252231357,
      "grad_norm": 0.4547802209854126,
      "learning_rate": 0.0001375979557161507,
      "loss": 0.2019,
      "step": 29060
    },
    {
      "epoch": 0.40860215053763443,
      "grad_norm": 0.3912622332572937,
      "learning_rate": 0.000137554879947161,
      "loss": 0.188,
      "step": 29070
    },
    {
      "epoch": 0.40874270855295525,
      "grad_norm": 0.43728306889533997,
      "learning_rate": 0.00013751179606408216,
      "loss": 0.2023,
      "step": 29080
    },
    {
      "epoch": 0.40888326656827606,
      "grad_norm": 0.43088382482528687,
      "learning_rate": 0.0001374687040762228,
      "loss": 0.1858,
      "step": 29090
    },
    {
      "epoch": 0.4090238245835969,
      "grad_norm": 0.33864808082580566,
      "learning_rate": 0.00013742560399289343,
      "loss": 0.1832,
      "step": 29100
    },
    {
      "epoch": 0.4091643825989177,
      "grad_norm": 0.38373056054115295,
      "learning_rate": 0.00013738249582340617,
      "loss": 0.1974,
      "step": 29110
    },
    {
      "epoch": 0.4093049406142385,
      "grad_norm": 0.4404429495334625,
      "learning_rate": 0.00013733937957707498,
      "loss": 0.1761,
      "step": 29120
    },
    {
      "epoch": 0.40944549862955937,
      "grad_norm": 0.3572157621383667,
      "learning_rate": 0.00013729625526321553,
      "loss": 0.1828,
      "step": 29130
    },
    {
      "epoch": 0.4095860566448802,
      "grad_norm": 0.4174441993236542,
      "learning_rate": 0.0001372531228911452,
      "loss": 0.1845,
      "step": 29140
    },
    {
      "epoch": 0.409726614660201,
      "grad_norm": 0.5497799515724182,
      "learning_rate": 0.0001372099824701832,
      "loss": 0.208,
      "step": 29150
    },
    {
      "epoch": 0.4098671726755218,
      "grad_norm": 0.42890477180480957,
      "learning_rate": 0.0001371668340096504,
      "loss": 0.185,
      "step": 29160
    },
    {
      "epoch": 0.4100077306908426,
      "grad_norm": 0.4199981689453125,
      "learning_rate": 0.0001371236775188695,
      "loss": 0.1879,
      "step": 29170
    },
    {
      "epoch": 0.4101482887061635,
      "grad_norm": 0.4099089205265045,
      "learning_rate": 0.00013708051300716478,
      "loss": 0.196,
      "step": 29180
    },
    {
      "epoch": 0.4102888467214843,
      "grad_norm": 0.5033042430877686,
      "learning_rate": 0.0001370373404838624,
      "loss": 0.1872,
      "step": 29190
    },
    {
      "epoch": 0.4104294047368051,
      "grad_norm": 0.36574870347976685,
      "learning_rate": 0.0001369941599582901,
      "loss": 0.19,
      "step": 29200
    },
    {
      "epoch": 0.4105699627521259,
      "grad_norm": 0.4706093668937683,
      "learning_rate": 0.0001369509714397776,
      "loss": 0.192,
      "step": 29210
    },
    {
      "epoch": 0.41071052076744674,
      "grad_norm": 0.4324891269207001,
      "learning_rate": 0.0001369077749376561,
      "loss": 0.2115,
      "step": 29220
    },
    {
      "epoch": 0.4108510787827676,
      "grad_norm": 0.38830357789993286,
      "learning_rate": 0.0001368645704612586,
      "loss": 0.1933,
      "step": 29230
    },
    {
      "epoch": 0.4109916367980884,
      "grad_norm": 0.44283023476600647,
      "learning_rate": 0.0001368213580199199,
      "loss": 0.1897,
      "step": 29240
    },
    {
      "epoch": 0.41113219481340924,
      "grad_norm": 0.3675142228603363,
      "learning_rate": 0.00013677813762297638,
      "loss": 0.2035,
      "step": 29250
    },
    {
      "epoch": 0.41127275282873005,
      "grad_norm": 0.4156498312950134,
      "learning_rate": 0.00013673490927976626,
      "loss": 0.1758,
      "step": 29260
    },
    {
      "epoch": 0.41141331084405086,
      "grad_norm": 0.43260544538497925,
      "learning_rate": 0.00013669167299962945,
      "loss": 0.177,
      "step": 29270
    },
    {
      "epoch": 0.41155386885937173,
      "grad_norm": 0.38720545172691345,
      "learning_rate": 0.00013664842879190756,
      "loss": 0.1986,
      "step": 29280
    },
    {
      "epoch": 0.41169442687469254,
      "grad_norm": 0.453714519739151,
      "learning_rate": 0.00013660517666594387,
      "loss": 0.1853,
      "step": 29290
    },
    {
      "epoch": 0.41183498489001336,
      "grad_norm": 0.3589514493942261,
      "learning_rate": 0.00013656191663108345,
      "loss": 0.2112,
      "step": 29300
    },
    {
      "epoch": 0.41197554290533417,
      "grad_norm": 0.2884405553340912,
      "learning_rate": 0.00013651864869667306,
      "loss": 0.1964,
      "step": 29310
    },
    {
      "epoch": 0.412116100920655,
      "grad_norm": 0.4473627507686615,
      "learning_rate": 0.00013647537287206108,
      "loss": 0.1715,
      "step": 29320
    },
    {
      "epoch": 0.41225665893597585,
      "grad_norm": 0.37632375955581665,
      "learning_rate": 0.00013643208916659774,
      "loss": 0.1693,
      "step": 29330
    },
    {
      "epoch": 0.41239721695129666,
      "grad_norm": 0.5020673274993896,
      "learning_rate": 0.00013638879758963486,
      "loss": 0.1828,
      "step": 29340
    },
    {
      "epoch": 0.4125377749666175,
      "grad_norm": 0.33215266466140747,
      "learning_rate": 0.00013634549815052597,
      "loss": 0.1789,
      "step": 29350
    },
    {
      "epoch": 0.4126783329819383,
      "grad_norm": 0.4781617224216461,
      "learning_rate": 0.00013630219085862639,
      "loss": 0.1757,
      "step": 29360
    },
    {
      "epoch": 0.4128188909972591,
      "grad_norm": 0.5313750505447388,
      "learning_rate": 0.000136258875723293,
      "loss": 0.1863,
      "step": 29370
    },
    {
      "epoch": 0.4129594490125799,
      "grad_norm": 0.45688632130622864,
      "learning_rate": 0.00013621555275388452,
      "loss": 0.1863,
      "step": 29380
    },
    {
      "epoch": 0.4131000070279008,
      "grad_norm": 0.38106387853622437,
      "learning_rate": 0.00013617222195976125,
      "loss": 0.1726,
      "step": 29390
    },
    {
      "epoch": 0.4132405650432216,
      "grad_norm": 0.3599858582019806,
      "learning_rate": 0.0001361288833502852,
      "loss": 0.1556,
      "step": 29400
    },
    {
      "epoch": 0.4133811230585424,
      "grad_norm": 0.5172914862632751,
      "learning_rate": 0.00013608553693482013,
      "loss": 0.1913,
      "step": 29410
    },
    {
      "epoch": 0.4135216810738632,
      "grad_norm": 0.38247478008270264,
      "learning_rate": 0.00013604218272273143,
      "loss": 0.1804,
      "step": 29420
    },
    {
      "epoch": 0.41366223908918404,
      "grad_norm": 0.3799438178539276,
      "learning_rate": 0.0001359988207233862,
      "loss": 0.1891,
      "step": 29430
    },
    {
      "epoch": 0.4138027971045049,
      "grad_norm": 0.5469621419906616,
      "learning_rate": 0.0001359554509461532,
      "loss": 0.1884,
      "step": 29440
    },
    {
      "epoch": 0.4139433551198257,
      "grad_norm": 0.40987929701805115,
      "learning_rate": 0.00013591207340040286,
      "loss": 0.182,
      "step": 29450
    },
    {
      "epoch": 0.41408391313514653,
      "grad_norm": 0.40848323702812195,
      "learning_rate": 0.00013586868809550737,
      "loss": 0.2087,
      "step": 29460
    },
    {
      "epoch": 0.41422447115046734,
      "grad_norm": 0.3570212423801422,
      "learning_rate": 0.00013582529504084048,
      "loss": 0.2001,
      "step": 29470
    },
    {
      "epoch": 0.41436502916578816,
      "grad_norm": 0.40303704142570496,
      "learning_rate": 0.0001357818942457777,
      "loss": 0.1646,
      "step": 29480
    },
    {
      "epoch": 0.414505587181109,
      "grad_norm": 0.4557267725467682,
      "learning_rate": 0.0001357384857196962,
      "loss": 0.196,
      "step": 29490
    },
    {
      "epoch": 0.41464614519642984,
      "grad_norm": 0.41009917855262756,
      "learning_rate": 0.00013569506947197477,
      "loss": 0.2005,
      "step": 29500
    },
    {
      "epoch": 0.41478670321175065,
      "grad_norm": 0.37826165556907654,
      "learning_rate": 0.0001356516455119939,
      "loss": 0.2175,
      "step": 29510
    },
    {
      "epoch": 0.41492726122707146,
      "grad_norm": 0.2804080843925476,
      "learning_rate": 0.00013560821384913583,
      "loss": 0.1681,
      "step": 29520
    },
    {
      "epoch": 0.4150678192423923,
      "grad_norm": 0.45264995098114014,
      "learning_rate": 0.00013556477449278427,
      "loss": 0.2006,
      "step": 29530
    },
    {
      "epoch": 0.41520837725771315,
      "grad_norm": 0.42463797330856323,
      "learning_rate": 0.00013552132745232477,
      "loss": 0.1771,
      "step": 29540
    },
    {
      "epoch": 0.41534893527303396,
      "grad_norm": 0.5739278793334961,
      "learning_rate": 0.0001354778727371445,
      "loss": 0.2045,
      "step": 29550
    },
    {
      "epoch": 0.41548949328835477,
      "grad_norm": 0.3454574644565582,
      "learning_rate": 0.0001354344103566322,
      "loss": 0.1861,
      "step": 29560
    },
    {
      "epoch": 0.4156300513036756,
      "grad_norm": 0.5129108428955078,
      "learning_rate": 0.00013539094032017837,
      "loss": 0.1747,
      "step": 29570
    },
    {
      "epoch": 0.4157706093189964,
      "grad_norm": 0.37665149569511414,
      "learning_rate": 0.00013534746263717511,
      "loss": 0.1862,
      "step": 29580
    },
    {
      "epoch": 0.41591116733431727,
      "grad_norm": 0.45705080032348633,
      "learning_rate": 0.00013530397731701622,
      "loss": 0.1956,
      "step": 29590
    },
    {
      "epoch": 0.4160517253496381,
      "grad_norm": 0.40519487857818604,
      "learning_rate": 0.00013526048436909705,
      "loss": 0.2066,
      "step": 29600
    },
    {
      "epoch": 0.4161922833649589,
      "grad_norm": 0.32703614234924316,
      "learning_rate": 0.00013521698380281473,
      "loss": 0.1903,
      "step": 29610
    },
    {
      "epoch": 0.4163328413802797,
      "grad_norm": 0.6017855405807495,
      "learning_rate": 0.00013517347562756794,
      "loss": 0.2394,
      "step": 29620
    },
    {
      "epoch": 0.4164733993956005,
      "grad_norm": 0.43627041578292847,
      "learning_rate": 0.00013512995985275706,
      "loss": 0.1829,
      "step": 29630
    },
    {
      "epoch": 0.41661395741092133,
      "grad_norm": 0.30025333166122437,
      "learning_rate": 0.00013508643648778407,
      "loss": 0.2177,
      "step": 29640
    },
    {
      "epoch": 0.4167545154262422,
      "grad_norm": 0.4151211977005005,
      "learning_rate": 0.00013504290554205262,
      "loss": 0.177,
      "step": 29650
    },
    {
      "epoch": 0.416895073441563,
      "grad_norm": 0.3238172233104706,
      "learning_rate": 0.00013499936702496794,
      "loss": 0.1727,
      "step": 29660
    },
    {
      "epoch": 0.4170356314568838,
      "grad_norm": 0.45996254682540894,
      "learning_rate": 0.000134955820945937,
      "loss": 0.1613,
      "step": 29670
    },
    {
      "epoch": 0.41717618947220464,
      "grad_norm": 0.4581054449081421,
      "learning_rate": 0.0001349122673143683,
      "loss": 0.1908,
      "step": 29680
    },
    {
      "epoch": 0.41731674748752545,
      "grad_norm": 0.6247484087944031,
      "learning_rate": 0.00013486870613967203,
      "loss": 0.2026,
      "step": 29690
    },
    {
      "epoch": 0.4174573055028463,
      "grad_norm": 0.3672480881214142,
      "learning_rate": 0.00013482513743126004,
      "loss": 0.2064,
      "step": 29700
    },
    {
      "epoch": 0.41759786351816713,
      "grad_norm": 0.4779052734375,
      "learning_rate": 0.0001347815611985457,
      "loss": 0.1943,
      "step": 29710
    },
    {
      "epoch": 0.41773842153348795,
      "grad_norm": 0.5254116654396057,
      "learning_rate": 0.00013473797745094409,
      "loss": 0.1954,
      "step": 29720
    },
    {
      "epoch": 0.41787897954880876,
      "grad_norm": 0.35492080450057983,
      "learning_rate": 0.00013469438619787192,
      "loss": 0.1879,
      "step": 29730
    },
    {
      "epoch": 0.4180195375641296,
      "grad_norm": 0.3733363747596741,
      "learning_rate": 0.00013465078744874745,
      "loss": 0.1801,
      "step": 29740
    },
    {
      "epoch": 0.41816009557945044,
      "grad_norm": 0.4023160934448242,
      "learning_rate": 0.00013460718121299062,
      "loss": 0.202,
      "step": 29750
    },
    {
      "epoch": 0.41830065359477125,
      "grad_norm": 0.4042220115661621,
      "learning_rate": 0.000134563567500023,
      "loss": 0.1929,
      "step": 29760
    },
    {
      "epoch": 0.41844121161009207,
      "grad_norm": 0.40776363015174866,
      "learning_rate": 0.00013451994631926773,
      "loss": 0.201,
      "step": 29770
    },
    {
      "epoch": 0.4185817696254129,
      "grad_norm": 0.5823947191238403,
      "learning_rate": 0.0001344763176801496,
      "loss": 0.192,
      "step": 29780
    },
    {
      "epoch": 0.4187223276407337,
      "grad_norm": 0.3720506429672241,
      "learning_rate": 0.00013443268159209491,
      "loss": 0.1861,
      "step": 29790
    },
    {
      "epoch": 0.41886288565605456,
      "grad_norm": 0.5004929304122925,
      "learning_rate": 0.0001343890380645318,
      "loss": 0.1978,
      "step": 29800
    },
    {
      "epoch": 0.4190034436713754,
      "grad_norm": 0.49414458870887756,
      "learning_rate": 0.00013434538710688972,
      "loss": 0.1811,
      "step": 29810
    },
    {
      "epoch": 0.4191440016866962,
      "grad_norm": 0.46427121758461,
      "learning_rate": 0.00013430172872859995,
      "loss": 0.1868,
      "step": 29820
    },
    {
      "epoch": 0.419284559702017,
      "grad_norm": 0.37025830149650574,
      "learning_rate": 0.00013425806293909532,
      "loss": 0.1786,
      "step": 29830
    },
    {
      "epoch": 0.4194251177173378,
      "grad_norm": 0.39013585448265076,
      "learning_rate": 0.00013421438974781018,
      "loss": 0.1801,
      "step": 29840
    },
    {
      "epoch": 0.4195656757326586,
      "grad_norm": 0.40397846698760986,
      "learning_rate": 0.00013417070916418056,
      "loss": 0.1753,
      "step": 29850
    },
    {
      "epoch": 0.4197062337479795,
      "grad_norm": 0.4552091956138611,
      "learning_rate": 0.00013412702119764406,
      "loss": 0.1885,
      "step": 29860
    },
    {
      "epoch": 0.4198467917633003,
      "grad_norm": 0.40092238783836365,
      "learning_rate": 0.00013408332585763993,
      "loss": 0.1877,
      "step": 29870
    },
    {
      "epoch": 0.4199873497786211,
      "grad_norm": 0.40552565455436707,
      "learning_rate": 0.00013403962315360887,
      "loss": 0.2058,
      "step": 29880
    },
    {
      "epoch": 0.42012790779394193,
      "grad_norm": 0.4105009138584137,
      "learning_rate": 0.00013399591309499334,
      "loss": 0.1835,
      "step": 29890
    },
    {
      "epoch": 0.42026846580926275,
      "grad_norm": 0.5223738551139832,
      "learning_rate": 0.0001339521956912373,
      "loss": 0.176,
      "step": 29900
    },
    {
      "epoch": 0.4204090238245836,
      "grad_norm": 0.29498058557510376,
      "learning_rate": 0.00013390847095178632,
      "loss": 0.1914,
      "step": 29910
    },
    {
      "epoch": 0.42054958183990443,
      "grad_norm": 0.41524139046669006,
      "learning_rate": 0.00013386473888608752,
      "loss": 0.1673,
      "step": 29920
    },
    {
      "epoch": 0.42069013985522524,
      "grad_norm": 0.36371296644210815,
      "learning_rate": 0.00013382099950358964,
      "loss": 0.1881,
      "step": 29930
    },
    {
      "epoch": 0.42083069787054606,
      "grad_norm": 0.34990018606185913,
      "learning_rate": 0.000133777252813743,
      "loss": 0.1753,
      "step": 29940
    },
    {
      "epoch": 0.42097125588586687,
      "grad_norm": 0.43816953897476196,
      "learning_rate": 0.00013373349882599947,
      "loss": 0.2146,
      "step": 29950
    },
    {
      "epoch": 0.42111181390118774,
      "grad_norm": 0.42917823791503906,
      "learning_rate": 0.0001336897375498125,
      "loss": 0.2063,
      "step": 29960
    },
    {
      "epoch": 0.42125237191650855,
      "grad_norm": 0.3981551229953766,
      "learning_rate": 0.00013364596899463719,
      "loss": 0.1914,
      "step": 29970
    },
    {
      "epoch": 0.42139292993182936,
      "grad_norm": 0.3675701320171356,
      "learning_rate": 0.0001336021931699301,
      "loss": 0.1757,
      "step": 29980
    },
    {
      "epoch": 0.4215334879471502,
      "grad_norm": 0.4614124596118927,
      "learning_rate": 0.00013355841008514944,
      "loss": 0.214,
      "step": 29990
    },
    {
      "epoch": 0.421674045962471,
      "grad_norm": 0.3702334761619568,
      "learning_rate": 0.00013351461974975496,
      "loss": 0.2035,
      "step": 30000
    },
    {
      "epoch": 0.421674045962471,
      "eval_chrf": 78.626866083246,
      "eval_loss": 0.392253577709198,
      "eval_runtime": 325.0765,
      "eval_samples_per_second": 0.308,
      "eval_steps_per_second": 0.012,
      "step": 30000
    },
    {
      "epoch": 0.42181460397779186,
      "grad_norm": 0.31596478819847107,
      "learning_rate": 0.00013347082217320794,
      "loss": 0.1744,
      "step": 30010
    },
    {
      "epoch": 0.42195516199311267,
      "grad_norm": 0.4133164584636688,
      "learning_rate": 0.0001334270173649713,
      "loss": 0.218,
      "step": 30020
    },
    {
      "epoch": 0.4220957200084335,
      "grad_norm": 0.40303972363471985,
      "learning_rate": 0.0001333832053345095,
      "loss": 0.1817,
      "step": 30030
    },
    {
      "epoch": 0.4222362780237543,
      "grad_norm": 0.47535112500190735,
      "learning_rate": 0.0001333393860912885,
      "loss": 0.1762,
      "step": 30040
    },
    {
      "epoch": 0.4223768360390751,
      "grad_norm": 0.4388129413127899,
      "learning_rate": 0.00013329555964477593,
      "loss": 0.1811,
      "step": 30050
    },
    {
      "epoch": 0.422517394054396,
      "grad_norm": 0.4499322772026062,
      "learning_rate": 0.00013325172600444084,
      "loss": 0.1955,
      "step": 30060
    },
    {
      "epoch": 0.4226579520697168,
      "grad_norm": 0.5237435102462769,
      "learning_rate": 0.00013320788517975395,
      "loss": 0.1679,
      "step": 30070
    },
    {
      "epoch": 0.4227985100850376,
      "grad_norm": 0.5499116778373718,
      "learning_rate": 0.0001331640371801875,
      "loss": 0.1894,
      "step": 30080
    },
    {
      "epoch": 0.4229390681003584,
      "grad_norm": 0.3641197681427002,
      "learning_rate": 0.00013312018201521518,
      "loss": 0.1929,
      "step": 30090
    },
    {
      "epoch": 0.42307962611567923,
      "grad_norm": 0.4277847707271576,
      "learning_rate": 0.00013307631969431243,
      "loss": 0.1748,
      "step": 30100
    },
    {
      "epoch": 0.42322018413100004,
      "grad_norm": 0.522125244140625,
      "learning_rate": 0.00013303245022695603,
      "loss": 0.197,
      "step": 30110
    },
    {
      "epoch": 0.4233607421463209,
      "grad_norm": 0.4268653392791748,
      "learning_rate": 0.00013298857362262445,
      "loss": 0.1718,
      "step": 30120
    },
    {
      "epoch": 0.4235013001616417,
      "grad_norm": 0.5319456458091736,
      "learning_rate": 0.00013294468989079764,
      "loss": 0.2118,
      "step": 30130
    },
    {
      "epoch": 0.42364185817696254,
      "grad_norm": 0.4997817575931549,
      "learning_rate": 0.0001329007990409571,
      "loss": 0.1686,
      "step": 30140
    },
    {
      "epoch": 0.42378241619228335,
      "grad_norm": 0.3909251391887665,
      "learning_rate": 0.00013285690108258584,
      "loss": 0.1824,
      "step": 30150
    },
    {
      "epoch": 0.42392297420760416,
      "grad_norm": 0.405780553817749,
      "learning_rate": 0.00013281299602516843,
      "loss": 0.2133,
      "step": 30160
    },
    {
      "epoch": 0.42406353222292503,
      "grad_norm": 0.37109053134918213,
      "learning_rate": 0.00013276908387819102,
      "loss": 0.2054,
      "step": 30170
    },
    {
      "epoch": 0.42420409023824585,
      "grad_norm": 0.3621840476989746,
      "learning_rate": 0.0001327251646511412,
      "loss": 0.1623,
      "step": 30180
    },
    {
      "epoch": 0.42434464825356666,
      "grad_norm": 0.367295503616333,
      "learning_rate": 0.00013268123835350818,
      "loss": 0.1961,
      "step": 30190
    },
    {
      "epoch": 0.42448520626888747,
      "grad_norm": 0.4074755609035492,
      "learning_rate": 0.00013263730499478261,
      "loss": 0.2072,
      "step": 30200
    },
    {
      "epoch": 0.4246257642842083,
      "grad_norm": 0.5425959825515747,
      "learning_rate": 0.00013259336458445678,
      "loss": 0.1978,
      "step": 30210
    },
    {
      "epoch": 0.42476632229952915,
      "grad_norm": 0.3877510726451874,
      "learning_rate": 0.00013254941713202432,
      "loss": 0.2074,
      "step": 30220
    },
    {
      "epoch": 0.42490688031484997,
      "grad_norm": 0.3716188669204712,
      "learning_rate": 0.00013250546264698061,
      "loss": 0.1942,
      "step": 30230
    },
    {
      "epoch": 0.4250474383301708,
      "grad_norm": 0.3302082121372223,
      "learning_rate": 0.00013246150113882234,
      "loss": 0.1865,
      "step": 30240
    },
    {
      "epoch": 0.4251879963454916,
      "grad_norm": 0.4123994708061218,
      "learning_rate": 0.0001324175326170479,
      "loss": 0.2006,
      "step": 30250
    },
    {
      "epoch": 0.4253285543608124,
      "grad_norm": 0.37305039167404175,
      "learning_rate": 0.00013237355709115706,
      "loss": 0.1696,
      "step": 30260
    },
    {
      "epoch": 0.4254691123761333,
      "grad_norm": 0.4247036576271057,
      "learning_rate": 0.00013232957457065113,
      "loss": 0.1904,
      "step": 30270
    },
    {
      "epoch": 0.4256096703914541,
      "grad_norm": 0.34812501072883606,
      "learning_rate": 0.000132285585065033,
      "loss": 0.1844,
      "step": 30280
    },
    {
      "epoch": 0.4257502284067749,
      "grad_norm": 0.39941826462745667,
      "learning_rate": 0.000132241588583807,
      "loss": 0.1913,
      "step": 30290
    },
    {
      "epoch": 0.4258907864220957,
      "grad_norm": 0.28224509954452515,
      "learning_rate": 0.000132197585136479,
      "loss": 0.1586,
      "step": 30300
    },
    {
      "epoch": 0.4260313444374165,
      "grad_norm": 0.4895475506782532,
      "learning_rate": 0.0001321535747325563,
      "loss": 0.173,
      "step": 30310
    },
    {
      "epoch": 0.4261719024527374,
      "grad_norm": 0.41064637899398804,
      "learning_rate": 0.0001321095573815479,
      "loss": 0.2005,
      "step": 30320
    },
    {
      "epoch": 0.4263124604680582,
      "grad_norm": 0.44213464856147766,
      "learning_rate": 0.000132065533092964,
      "loss": 0.1621,
      "step": 30330
    },
    {
      "epoch": 0.426453018483379,
      "grad_norm": 0.4305702745914459,
      "learning_rate": 0.00013202150187631662,
      "loss": 0.1737,
      "step": 30340
    },
    {
      "epoch": 0.42659357649869983,
      "grad_norm": 0.5681144595146179,
      "learning_rate": 0.00013197746374111902,
      "loss": 0.1883,
      "step": 30350
    },
    {
      "epoch": 0.42673413451402065,
      "grad_norm": 0.40257877111434937,
      "learning_rate": 0.0001319334186968861,
      "loss": 0.2054,
      "step": 30360
    },
    {
      "epoch": 0.42687469252934146,
      "grad_norm": 0.4283764958381653,
      "learning_rate": 0.0001318893667531342,
      "loss": 0.1691,
      "step": 30370
    },
    {
      "epoch": 0.42701525054466233,
      "grad_norm": 0.5085741281509399,
      "learning_rate": 0.0001318453079193812,
      "loss": 0.2176,
      "step": 30380
    },
    {
      "epoch": 0.42715580855998314,
      "grad_norm": 0.44593504071235657,
      "learning_rate": 0.00013180124220514637,
      "loss": 0.1949,
      "step": 30390
    },
    {
      "epoch": 0.42729636657530395,
      "grad_norm": 0.42691755294799805,
      "learning_rate": 0.00013175716961995052,
      "loss": 0.1924,
      "step": 30400
    },
    {
      "epoch": 0.42743692459062477,
      "grad_norm": 0.4168946444988251,
      "learning_rate": 0.00013171309017331604,
      "loss": 0.1677,
      "step": 30410
    },
    {
      "epoch": 0.4275774826059456,
      "grad_norm": 0.3309684693813324,
      "learning_rate": 0.00013166900387476664,
      "loss": 0.2184,
      "step": 30420
    },
    {
      "epoch": 0.42771804062126645,
      "grad_norm": 0.4581981897354126,
      "learning_rate": 0.0001316249107338276,
      "loss": 0.1948,
      "step": 30430
    },
    {
      "epoch": 0.42785859863658726,
      "grad_norm": 0.4287571907043457,
      "learning_rate": 0.00013158081076002567,
      "loss": 0.2007,
      "step": 30440
    },
    {
      "epoch": 0.4279991566519081,
      "grad_norm": 0.41077181696891785,
      "learning_rate": 0.00013153670396288904,
      "loss": 0.1747,
      "step": 30450
    },
    {
      "epoch": 0.4281397146672289,
      "grad_norm": 0.40592631697654724,
      "learning_rate": 0.00013149259035194746,
      "loss": 0.1802,
      "step": 30460
    },
    {
      "epoch": 0.4282802726825497,
      "grad_norm": 0.3631041646003723,
      "learning_rate": 0.00013144846993673207,
      "loss": 0.1799,
      "step": 30470
    },
    {
      "epoch": 0.42842083069787057,
      "grad_norm": 0.401694655418396,
      "learning_rate": 0.00013140434272677548,
      "loss": 0.1762,
      "step": 30480
    },
    {
      "epoch": 0.4285613887131914,
      "grad_norm": 0.4939548969268799,
      "learning_rate": 0.00013136020873161182,
      "loss": 0.1884,
      "step": 30490
    },
    {
      "epoch": 0.4287019467285122,
      "grad_norm": 0.5443631410598755,
      "learning_rate": 0.00013131606796077666,
      "loss": 0.1854,
      "step": 30500
    },
    {
      "epoch": 0.428842504743833,
      "grad_norm": 0.4527170658111572,
      "learning_rate": 0.000131271920423807,
      "loss": 0.2076,
      "step": 30510
    },
    {
      "epoch": 0.4289830627591538,
      "grad_norm": 0.42109552025794983,
      "learning_rate": 0.00013122776613024137,
      "loss": 0.2124,
      "step": 30520
    },
    {
      "epoch": 0.4291236207744747,
      "grad_norm": 0.4336777925491333,
      "learning_rate": 0.0001311836050896197,
      "loss": 0.1754,
      "step": 30530
    },
    {
      "epoch": 0.4292641787897955,
      "grad_norm": 0.2628873884677887,
      "learning_rate": 0.00013113943731148347,
      "loss": 0.1917,
      "step": 30540
    },
    {
      "epoch": 0.4294047368051163,
      "grad_norm": 0.40117138624191284,
      "learning_rate": 0.00013109526280537546,
      "loss": 0.193,
      "step": 30550
    },
    {
      "epoch": 0.42954529482043713,
      "grad_norm": 0.36322981119155884,
      "learning_rate": 0.00013105108158084004,
      "loss": 0.1799,
      "step": 30560
    },
    {
      "epoch": 0.42968585283575794,
      "grad_norm": 0.34170740842819214,
      "learning_rate": 0.00013100689364742296,
      "loss": 0.1863,
      "step": 30570
    },
    {
      "epoch": 0.42982641085107876,
      "grad_norm": 0.612261176109314,
      "learning_rate": 0.00013096269901467146,
      "loss": 0.2018,
      "step": 30580
    },
    {
      "epoch": 0.4299669688663996,
      "grad_norm": 0.42685508728027344,
      "learning_rate": 0.00013091849769213422,
      "loss": 0.1789,
      "step": 30590
    },
    {
      "epoch": 0.43010752688172044,
      "grad_norm": 0.3997404873371124,
      "learning_rate": 0.00013087428968936137,
      "loss": 0.1894,
      "step": 30600
    },
    {
      "epoch": 0.43024808489704125,
      "grad_norm": 0.38838812708854675,
      "learning_rate": 0.0001308300750159044,
      "loss": 0.1699,
      "step": 30610
    },
    {
      "epoch": 0.43038864291236206,
      "grad_norm": 0.43439820408821106,
      "learning_rate": 0.0001307858536813164,
      "loss": 0.1808,
      "step": 30620
    },
    {
      "epoch": 0.4305292009276829,
      "grad_norm": 0.3681730628013611,
      "learning_rate": 0.00013074162569515172,
      "loss": 0.1899,
      "step": 30630
    },
    {
      "epoch": 0.43066975894300374,
      "grad_norm": 0.46958011388778687,
      "learning_rate": 0.0001306973910669663,
      "loss": 0.1861,
      "step": 30640
    },
    {
      "epoch": 0.43081031695832456,
      "grad_norm": 0.39444705843925476,
      "learning_rate": 0.00013065314980631747,
      "loss": 0.1974,
      "step": 30650
    },
    {
      "epoch": 0.43095087497364537,
      "grad_norm": 0.4495857059955597,
      "learning_rate": 0.00013060890192276393,
      "loss": 0.1887,
      "step": 30660
    },
    {
      "epoch": 0.4310914329889662,
      "grad_norm": 0.417550653219223,
      "learning_rate": 0.00013056464742586583,
      "loss": 0.1893,
      "step": 30670
    },
    {
      "epoch": 0.431231991004287,
      "grad_norm": 0.4482434391975403,
      "learning_rate": 0.0001305203863251848,
      "loss": 0.1882,
      "step": 30680
    },
    {
      "epoch": 0.43137254901960786,
      "grad_norm": 0.44060251116752625,
      "learning_rate": 0.00013047611863028396,
      "loss": 0.1964,
      "step": 30690
    },
    {
      "epoch": 0.4315131070349287,
      "grad_norm": 0.3529464900493622,
      "learning_rate": 0.00013043184435072763,
      "loss": 0.18,
      "step": 30700
    },
    {
      "epoch": 0.4316536650502495,
      "grad_norm": 0.4485524296760559,
      "learning_rate": 0.00013038756349608176,
      "loss": 0.1757,
      "step": 30710
    },
    {
      "epoch": 0.4317942230655703,
      "grad_norm": 0.4147986173629761,
      "learning_rate": 0.00013034327607591364,
      "loss": 0.1702,
      "step": 30720
    },
    {
      "epoch": 0.4319347810808911,
      "grad_norm": 0.46421390771865845,
      "learning_rate": 0.000130298982099792,
      "loss": 0.1745,
      "step": 30730
    },
    {
      "epoch": 0.432075339096212,
      "grad_norm": 0.4629431962966919,
      "learning_rate": 0.00013025468157728696,
      "loss": 0.2061,
      "step": 30740
    },
    {
      "epoch": 0.4322158971115328,
      "grad_norm": 0.34252962470054626,
      "learning_rate": 0.0001302103745179701,
      "loss": 0.1708,
      "step": 30750
    },
    {
      "epoch": 0.4323564551268536,
      "grad_norm": 0.3832494020462036,
      "learning_rate": 0.00013016606093141432,
      "loss": 0.1782,
      "step": 30760
    },
    {
      "epoch": 0.4324970131421744,
      "grad_norm": 0.4208075702190399,
      "learning_rate": 0.00013012174082719405,
      "loss": 0.2094,
      "step": 30770
    },
    {
      "epoch": 0.43263757115749524,
      "grad_norm": 0.3315330445766449,
      "learning_rate": 0.00013007741421488503,
      "loss": 0.2001,
      "step": 30780
    },
    {
      "epoch": 0.4327781291728161,
      "grad_norm": 0.42039304971694946,
      "learning_rate": 0.00013003308110406446,
      "loss": 0.1908,
      "step": 30790
    },
    {
      "epoch": 0.4329186871881369,
      "grad_norm": 0.5534359216690063,
      "learning_rate": 0.00012998874150431099,
      "loss": 0.1977,
      "step": 30800
    },
    {
      "epoch": 0.43305924520345773,
      "grad_norm": 0.3723289668560028,
      "learning_rate": 0.00012994439542520452,
      "loss": 0.191,
      "step": 30810
    },
    {
      "epoch": 0.43319980321877855,
      "grad_norm": 0.5126309394836426,
      "learning_rate": 0.00012990004287632647,
      "loss": 0.1958,
      "step": 30820
    },
    {
      "epoch": 0.43334036123409936,
      "grad_norm": 0.4108158349990845,
      "learning_rate": 0.00012985568386725962,
      "loss": 0.177,
      "step": 30830
    },
    {
      "epoch": 0.43348091924942017,
      "grad_norm": 0.440837562084198,
      "learning_rate": 0.0001298113184075882,
      "loss": 0.1865,
      "step": 30840
    },
    {
      "epoch": 0.43362147726474104,
      "grad_norm": 0.43276897072792053,
      "learning_rate": 0.00012976694650689778,
      "loss": 0.1736,
      "step": 30850
    },
    {
      "epoch": 0.43376203528006185,
      "grad_norm": 0.3253236413002014,
      "learning_rate": 0.0001297225681747753,
      "loss": 0.1736,
      "step": 30860
    },
    {
      "epoch": 0.43390259329538267,
      "grad_norm": 0.4109024703502655,
      "learning_rate": 0.00012967818342080913,
      "loss": 0.1978,
      "step": 30870
    },
    {
      "epoch": 0.4340431513107035,
      "grad_norm": 0.5015186071395874,
      "learning_rate": 0.00012963379225458904,
      "loss": 0.1845,
      "step": 30880
    },
    {
      "epoch": 0.4341837093260243,
      "grad_norm": 0.35976794362068176,
      "learning_rate": 0.00012958939468570615,
      "loss": 0.1916,
      "step": 30890
    },
    {
      "epoch": 0.43432426734134516,
      "grad_norm": 0.39229390025138855,
      "learning_rate": 0.00012954499072375297,
      "loss": 0.2056,
      "step": 30900
    },
    {
      "epoch": 0.434464825356666,
      "grad_norm": 0.469139039516449,
      "learning_rate": 0.0001295005803783234,
      "loss": 0.2059,
      "step": 30910
    },
    {
      "epoch": 0.4346053833719868,
      "grad_norm": 0.45978787541389465,
      "learning_rate": 0.0001294561636590127,
      "loss": 0.2012,
      "step": 30920
    },
    {
      "epoch": 0.4347459413873076,
      "grad_norm": 0.3263724148273468,
      "learning_rate": 0.0001294117405754176,
      "loss": 0.1946,
      "step": 30930
    },
    {
      "epoch": 0.4348864994026284,
      "grad_norm": 0.4245821237564087,
      "learning_rate": 0.00012936731113713607,
      "loss": 0.1807,
      "step": 30940
    },
    {
      "epoch": 0.4350270574179493,
      "grad_norm": 0.4069032669067383,
      "learning_rate": 0.0001293228753537675,
      "loss": 0.1967,
      "step": 30950
    },
    {
      "epoch": 0.4351676154332701,
      "grad_norm": 0.3639621436595917,
      "learning_rate": 0.00012927843323491268,
      "loss": 0.1872,
      "step": 30960
    },
    {
      "epoch": 0.4353081734485909,
      "grad_norm": 0.394876629114151,
      "learning_rate": 0.00012923398479017378,
      "loss": 0.2026,
      "step": 30970
    },
    {
      "epoch": 0.4354487314639117,
      "grad_norm": 0.5878793597221375,
      "learning_rate": 0.0001291895300291543,
      "loss": 0.1807,
      "step": 30980
    },
    {
      "epoch": 0.43558928947923253,
      "grad_norm": 0.33681783080101013,
      "learning_rate": 0.00012914506896145907,
      "loss": 0.1824,
      "step": 30990
    },
    {
      "epoch": 0.4357298474945534,
      "grad_norm": 0.41459086537361145,
      "learning_rate": 0.00012910060159669436,
      "loss": 0.1984,
      "step": 31000
    },
    {
      "epoch": 0.4357298474945534,
      "eval_chrf": 77.59793143605809,
      "eval_loss": 0.3922773599624634,
      "eval_runtime": 325.7326,
      "eval_samples_per_second": 0.307,
      "eval_steps_per_second": 0.012,
      "step": 31000
    },
    {
      "epoch": 0.4358704055098742,
      "grad_norm": 0.29785963892936707,
      "learning_rate": 0.00012905612794446774,
      "loss": 0.202,
      "step": 31010
    },
    {
      "epoch": 0.436010963525195,
      "grad_norm": 0.40135979652404785,
      "learning_rate": 0.00012901164801438822,
      "loss": 0.177,
      "step": 31020
    },
    {
      "epoch": 0.43615152154051584,
      "grad_norm": 0.32746079564094543,
      "learning_rate": 0.00012896716181606612,
      "loss": 0.1813,
      "step": 31030
    },
    {
      "epoch": 0.43629207955583665,
      "grad_norm": 0.44277051091194153,
      "learning_rate": 0.00012892266935911301,
      "loss": 0.1676,
      "step": 31040
    },
    {
      "epoch": 0.4364326375711575,
      "grad_norm": 0.4225252866744995,
      "learning_rate": 0.000128878170653142,
      "loss": 0.175,
      "step": 31050
    },
    {
      "epoch": 0.43657319558647834,
      "grad_norm": 0.39410722255706787,
      "learning_rate": 0.0001288336657077674,
      "loss": 0.1797,
      "step": 31060
    },
    {
      "epoch": 0.43671375360179915,
      "grad_norm": 0.5188403129577637,
      "learning_rate": 0.000128789154532605,
      "loss": 0.18,
      "step": 31070
    },
    {
      "epoch": 0.43685431161711996,
      "grad_norm": 0.5571266412734985,
      "learning_rate": 0.0001287446371372718,
      "loss": 0.1728,
      "step": 31080
    },
    {
      "epoch": 0.4369948696324408,
      "grad_norm": 0.4622716009616852,
      "learning_rate": 0.0001287001135313862,
      "loss": 0.1775,
      "step": 31090
    },
    {
      "epoch": 0.4371354276477616,
      "grad_norm": 0.35365378856658936,
      "learning_rate": 0.00012865558372456798,
      "loss": 0.185,
      "step": 31100
    },
    {
      "epoch": 0.43727598566308246,
      "grad_norm": 0.41484931111335754,
      "learning_rate": 0.00012861104772643825,
      "loss": 0.1842,
      "step": 31110
    },
    {
      "epoch": 0.43741654367840327,
      "grad_norm": 0.4718504846096039,
      "learning_rate": 0.00012856650554661937,
      "loss": 0.1982,
      "step": 31120
    },
    {
      "epoch": 0.4375571016937241,
      "grad_norm": 0.31714683771133423,
      "learning_rate": 0.00012852195719473516,
      "loss": 0.1905,
      "step": 31130
    },
    {
      "epoch": 0.4376976597090449,
      "grad_norm": 0.3640306890010834,
      "learning_rate": 0.00012847740268041071,
      "loss": 0.2141,
      "step": 31140
    },
    {
      "epoch": 0.4378382177243657,
      "grad_norm": 0.36697953939437866,
      "learning_rate": 0.0001284328420132724,
      "loss": 0.1718,
      "step": 31150
    },
    {
      "epoch": 0.4379787757396866,
      "grad_norm": 0.4646686911582947,
      "learning_rate": 0.00012838827520294803,
      "loss": 0.1716,
      "step": 31160
    },
    {
      "epoch": 0.4381193337550074,
      "grad_norm": 0.3565073311328888,
      "learning_rate": 0.00012834370225906668,
      "loss": 0.1978,
      "step": 31170
    },
    {
      "epoch": 0.4382598917703282,
      "grad_norm": 0.3856450617313385,
      "learning_rate": 0.00012829912319125875,
      "loss": 0.1956,
      "step": 31180
    },
    {
      "epoch": 0.438400449785649,
      "grad_norm": 0.5122458934783936,
      "learning_rate": 0.00012825453800915595,
      "loss": 0.1863,
      "step": 31190
    },
    {
      "epoch": 0.43854100780096983,
      "grad_norm": 0.3763080835342407,
      "learning_rate": 0.00012820994672239135,
      "loss": 0.2008,
      "step": 31200
    },
    {
      "epoch": 0.4386815658162907,
      "grad_norm": 0.3721507489681244,
      "learning_rate": 0.00012816534934059933,
      "loss": 0.1913,
      "step": 31210
    },
    {
      "epoch": 0.4388221238316115,
      "grad_norm": 0.30980393290519714,
      "learning_rate": 0.00012812074587341562,
      "loss": 0.1676,
      "step": 31220
    },
    {
      "epoch": 0.4389626818469323,
      "grad_norm": 0.3242766559123993,
      "learning_rate": 0.00012807613633047712,
      "loss": 0.1652,
      "step": 31230
    },
    {
      "epoch": 0.43910323986225314,
      "grad_norm": 0.49657100439071655,
      "learning_rate": 0.00012803152072142222,
      "loss": 0.2039,
      "step": 31240
    },
    {
      "epoch": 0.43924379787757395,
      "grad_norm": 0.3150328993797302,
      "learning_rate": 0.00012798689905589052,
      "loss": 0.1749,
      "step": 31250
    },
    {
      "epoch": 0.4393843558928948,
      "grad_norm": 0.3681991398334503,
      "learning_rate": 0.00012794227134352298,
      "loss": 0.1781,
      "step": 31260
    },
    {
      "epoch": 0.43952491390821563,
      "grad_norm": 0.4745456576347351,
      "learning_rate": 0.00012789763759396184,
      "loss": 0.185,
      "step": 31270
    },
    {
      "epoch": 0.43966547192353644,
      "grad_norm": 0.4351559281349182,
      "learning_rate": 0.00012785299781685065,
      "loss": 0.1967,
      "step": 31280
    },
    {
      "epoch": 0.43980602993885726,
      "grad_norm": 0.3964342474937439,
      "learning_rate": 0.00012780835202183422,
      "loss": 0.1856,
      "step": 31290
    },
    {
      "epoch": 0.43994658795417807,
      "grad_norm": 0.5241504311561584,
      "learning_rate": 0.00012776370021855873,
      "loss": 0.2079,
      "step": 31300
    },
    {
      "epoch": 0.4400871459694989,
      "grad_norm": 0.45348280668258667,
      "learning_rate": 0.00012771904241667162,
      "loss": 0.1838,
      "step": 31310
    },
    {
      "epoch": 0.44022770398481975,
      "grad_norm": 0.3943793475627899,
      "learning_rate": 0.00012767437862582162,
      "loss": 0.1795,
      "step": 31320
    },
    {
      "epoch": 0.44036826200014056,
      "grad_norm": 0.3890320658683777,
      "learning_rate": 0.00012762970885565878,
      "loss": 0.1705,
      "step": 31330
    },
    {
      "epoch": 0.4405088200154614,
      "grad_norm": 0.34188640117645264,
      "learning_rate": 0.00012758503311583442,
      "loss": 0.1688,
      "step": 31340
    },
    {
      "epoch": 0.4406493780307822,
      "grad_norm": 0.5416295528411865,
      "learning_rate": 0.00012754035141600119,
      "loss": 0.1899,
      "step": 31350
    },
    {
      "epoch": 0.440789936046103,
      "grad_norm": 0.3821467161178589,
      "learning_rate": 0.00012749566376581294,
      "loss": 0.1727,
      "step": 31360
    },
    {
      "epoch": 0.44093049406142387,
      "grad_norm": 0.39596933126449585,
      "learning_rate": 0.00012745097017492494,
      "loss": 0.1685,
      "step": 31370
    },
    {
      "epoch": 0.4410710520767447,
      "grad_norm": 0.34711122512817383,
      "learning_rate": 0.0001274062706529936,
      "loss": 0.2023,
      "step": 31380
    },
    {
      "epoch": 0.4412116100920655,
      "grad_norm": 0.3162592053413391,
      "learning_rate": 0.0001273615652096767,
      "loss": 0.1958,
      "step": 31390
    },
    {
      "epoch": 0.4413521681073863,
      "grad_norm": 0.48744603991508484,
      "learning_rate": 0.00012731685385463328,
      "loss": 0.1977,
      "step": 31400
    },
    {
      "epoch": 0.4414927261227071,
      "grad_norm": 0.4292646050453186,
      "learning_rate": 0.00012727213659752363,
      "loss": 0.2317,
      "step": 31410
    },
    {
      "epoch": 0.441633284138028,
      "grad_norm": 0.40441256761550903,
      "learning_rate": 0.00012722741344800936,
      "loss": 0.1794,
      "step": 31420
    },
    {
      "epoch": 0.4417738421533488,
      "grad_norm": 0.48188072443008423,
      "learning_rate": 0.00012718268441575336,
      "loss": 0.2171,
      "step": 31430
    },
    {
      "epoch": 0.4419144001686696,
      "grad_norm": 0.46570271253585815,
      "learning_rate": 0.0001271379495104197,
      "loss": 0.1781,
      "step": 31440
    },
    {
      "epoch": 0.44205495818399043,
      "grad_norm": 0.32672974467277527,
      "learning_rate": 0.00012709320874167382,
      "loss": 0.176,
      "step": 31450
    },
    {
      "epoch": 0.44219551619931124,
      "grad_norm": 0.3220718801021576,
      "learning_rate": 0.0001270484621191824,
      "loss": 0.1883,
      "step": 31460
    },
    {
      "epoch": 0.4423360742146321,
      "grad_norm": 0.3279297947883606,
      "learning_rate": 0.00012700370965261336,
      "loss": 0.1882,
      "step": 31470
    },
    {
      "epoch": 0.4424766322299529,
      "grad_norm": 0.48345816135406494,
      "learning_rate": 0.00012695895135163586,
      "loss": 0.2158,
      "step": 31480
    },
    {
      "epoch": 0.44261719024527374,
      "grad_norm": 0.4876115620136261,
      "learning_rate": 0.00012691418722592042,
      "loss": 0.1797,
      "step": 31490
    },
    {
      "epoch": 0.44275774826059455,
      "grad_norm": 0.3962550461292267,
      "learning_rate": 0.00012686941728513873,
      "loss": 0.1706,
      "step": 31500
    },
    {
      "epoch": 0.44289830627591537,
      "grad_norm": 0.3843517005443573,
      "learning_rate": 0.00012682464153896373,
      "loss": 0.1948,
      "step": 31510
    },
    {
      "epoch": 0.44303886429123623,
      "grad_norm": 0.4024946391582489,
      "learning_rate": 0.00012677985999706967,
      "loss": 0.1807,
      "step": 31520
    },
    {
      "epoch": 0.44317942230655705,
      "grad_norm": 0.35814380645751953,
      "learning_rate": 0.00012673507266913205,
      "loss": 0.2129,
      "step": 31530
    },
    {
      "epoch": 0.44331998032187786,
      "grad_norm": 0.46133631467819214,
      "learning_rate": 0.00012669027956482758,
      "loss": 0.1855,
      "step": 31540
    },
    {
      "epoch": 0.4434605383371987,
      "grad_norm": 0.5115548968315125,
      "learning_rate": 0.00012664548069383423,
      "loss": 0.1896,
      "step": 31550
    },
    {
      "epoch": 0.4436010963525195,
      "grad_norm": 0.423533171415329,
      "learning_rate": 0.00012660067606583123,
      "loss": 0.2351,
      "step": 31560
    },
    {
      "epoch": 0.4437416543678403,
      "grad_norm": 0.35322490334510803,
      "learning_rate": 0.00012655586569049903,
      "loss": 0.1905,
      "step": 31570
    },
    {
      "epoch": 0.44388221238316117,
      "grad_norm": 0.4697679579257965,
      "learning_rate": 0.00012651104957751939,
      "loss": 0.173,
      "step": 31580
    },
    {
      "epoch": 0.444022770398482,
      "grad_norm": 0.36890703439712524,
      "learning_rate": 0.00012646622773657517,
      "loss": 0.1718,
      "step": 31590
    },
    {
      "epoch": 0.4441633284138028,
      "grad_norm": 0.37209486961364746,
      "learning_rate": 0.00012642140017735064,
      "loss": 0.2068,
      "step": 31600
    },
    {
      "epoch": 0.4443038864291236,
      "grad_norm": 0.46700626611709595,
      "learning_rate": 0.00012637656690953116,
      "loss": 0.2146,
      "step": 31610
    },
    {
      "epoch": 0.4444444444444444,
      "grad_norm": 0.40611961483955383,
      "learning_rate": 0.00012633172794280343,
      "loss": 0.1943,
      "step": 31620
    },
    {
      "epoch": 0.4445850024597653,
      "grad_norm": 0.3542831540107727,
      "learning_rate": 0.0001262868832868553,
      "loss": 0.1833,
      "step": 31630
    },
    {
      "epoch": 0.4447255604750861,
      "grad_norm": 0.4369039535522461,
      "learning_rate": 0.00012624203295137594,
      "loss": 0.178,
      "step": 31640
    },
    {
      "epoch": 0.4448661184904069,
      "grad_norm": 0.3356902003288269,
      "learning_rate": 0.0001261971769460556,
      "loss": 0.1878,
      "step": 31650
    },
    {
      "epoch": 0.4450066765057277,
      "grad_norm": 0.3247499167919159,
      "learning_rate": 0.00012615231528058593,
      "loss": 0.19,
      "step": 31660
    },
    {
      "epoch": 0.44514723452104854,
      "grad_norm": 0.3748258650302887,
      "learning_rate": 0.0001261074479646597,
      "loss": 0.1887,
      "step": 31670
    },
    {
      "epoch": 0.4452877925363694,
      "grad_norm": 0.4689333438873291,
      "learning_rate": 0.00012606257500797087,
      "loss": 0.1858,
      "step": 31680
    },
    {
      "epoch": 0.4454283505516902,
      "grad_norm": 0.5092548131942749,
      "learning_rate": 0.00012601769642021476,
      "loss": 0.1985,
      "step": 31690
    },
    {
      "epoch": 0.44556890856701103,
      "grad_norm": 0.35228753089904785,
      "learning_rate": 0.00012597281221108769,
      "loss": 0.1773,
      "step": 31700
    },
    {
      "epoch": 0.44570946658233185,
      "grad_norm": 0.5632361769676208,
      "learning_rate": 0.00012592792239028746,
      "loss": 0.1812,
      "step": 31710
    },
    {
      "epoch": 0.44585002459765266,
      "grad_norm": 0.4932362139225006,
      "learning_rate": 0.00012588302696751286,
      "loss": 0.2045,
      "step": 31720
    },
    {
      "epoch": 0.44599058261297353,
      "grad_norm": 0.41713523864746094,
      "learning_rate": 0.00012583812595246398,
      "loss": 0.1805,
      "step": 31730
    },
    {
      "epoch": 0.44613114062829434,
      "grad_norm": 0.4142841398715973,
      "learning_rate": 0.00012579321935484212,
      "loss": 0.1726,
      "step": 31740
    },
    {
      "epoch": 0.44627169864361516,
      "grad_norm": 0.4331299066543579,
      "learning_rate": 0.0001257483071843498,
      "loss": 0.1894,
      "step": 31750
    },
    {
      "epoch": 0.44641225665893597,
      "grad_norm": 0.3925873339176178,
      "learning_rate": 0.00012570338945069068,
      "loss": 0.1845,
      "step": 31760
    },
    {
      "epoch": 0.4465528146742568,
      "grad_norm": 0.4228861629962921,
      "learning_rate": 0.00012565846616356972,
      "loss": 0.2181,
      "step": 31770
    },
    {
      "epoch": 0.44669337268957765,
      "grad_norm": 0.4462876319885254,
      "learning_rate": 0.00012561353733269294,
      "loss": 0.1942,
      "step": 31780
    },
    {
      "epoch": 0.44683393070489846,
      "grad_norm": 0.3477708101272583,
      "learning_rate": 0.00012556860296776773,
      "loss": 0.2002,
      "step": 31790
    },
    {
      "epoch": 0.4469744887202193,
      "grad_norm": 0.36163148283958435,
      "learning_rate": 0.00012552366307850253,
      "loss": 0.1849,
      "step": 31800
    },
    {
      "epoch": 0.4471150467355401,
      "grad_norm": 0.4544931650161743,
      "learning_rate": 0.00012547871767460707,
      "loss": 0.1779,
      "step": 31810
    },
    {
      "epoch": 0.4472556047508609,
      "grad_norm": 0.45102718472480774,
      "learning_rate": 0.0001254337667657922,
      "loss": 0.1745,
      "step": 31820
    },
    {
      "epoch": 0.4473961627661817,
      "grad_norm": 0.4464181065559387,
      "learning_rate": 0.00012538881036177002,
      "loss": 0.2091,
      "step": 31830
    },
    {
      "epoch": 0.4475367207815026,
      "grad_norm": 0.45269858837127686,
      "learning_rate": 0.00012534384847225373,
      "loss": 0.217,
      "step": 31840
    },
    {
      "epoch": 0.4476772787968234,
      "grad_norm": 0.5486025214195251,
      "learning_rate": 0.00012529888110695785,
      "loss": 0.1976,
      "step": 31850
    },
    {
      "epoch": 0.4478178368121442,
      "grad_norm": 0.3856346309185028,
      "learning_rate": 0.00012525390827559796,
      "loss": 0.1894,
      "step": 31860
    },
    {
      "epoch": 0.447958394827465,
      "grad_norm": 0.5907379388809204,
      "learning_rate": 0.00012520892998789087,
      "loss": 0.1867,
      "step": 31870
    },
    {
      "epoch": 0.44809895284278584,
      "grad_norm": 0.5895049571990967,
      "learning_rate": 0.0001251639462535546,
      "loss": 0.1682,
      "step": 31880
    },
    {
      "epoch": 0.4482395108581067,
      "grad_norm": 0.3159005045890808,
      "learning_rate": 0.00012511895708230828,
      "loss": 0.1803,
      "step": 31890
    },
    {
      "epoch": 0.4483800688734275,
      "grad_norm": 0.4468821883201599,
      "learning_rate": 0.00012507396248387224,
      "loss": 0.1849,
      "step": 31900
    },
    {
      "epoch": 0.44852062688874833,
      "grad_norm": 0.35855212807655334,
      "learning_rate": 0.00012502896246796802,
      "loss": 0.1737,
      "step": 31910
    },
    {
      "epoch": 0.44866118490406914,
      "grad_norm": 0.44083160161972046,
      "learning_rate": 0.0001249839570443183,
      "loss": 0.1771,
      "step": 31920
    },
    {
      "epoch": 0.44880174291938996,
      "grad_norm": 0.37042614817619324,
      "learning_rate": 0.00012493894622264695,
      "loss": 0.1706,
      "step": 31930
    },
    {
      "epoch": 0.4489423009347108,
      "grad_norm": 0.9062139987945557,
      "learning_rate": 0.00012489393001267893,
      "loss": 0.1964,
      "step": 31940
    },
    {
      "epoch": 0.44908285895003164,
      "grad_norm": 0.27651646733283997,
      "learning_rate": 0.00012484890842414046,
      "loss": 0.1701,
      "step": 31950
    },
    {
      "epoch": 0.44922341696535245,
      "grad_norm": 0.5589680671691895,
      "learning_rate": 0.00012480388146675884,
      "loss": 0.199,
      "step": 31960
    },
    {
      "epoch": 0.44936397498067326,
      "grad_norm": 0.502112627029419,
      "learning_rate": 0.00012475884915026265,
      "loss": 0.2058,
      "step": 31970
    },
    {
      "epoch": 0.4495045329959941,
      "grad_norm": 0.38898777961730957,
      "learning_rate": 0.0001247138114843815,
      "loss": 0.1892,
      "step": 31980
    },
    {
      "epoch": 0.44964509101131495,
      "grad_norm": 0.5574616193771362,
      "learning_rate": 0.00012466876847884618,
      "loss": 0.1802,
      "step": 31990
    },
    {
      "epoch": 0.44978564902663576,
      "grad_norm": 0.36505675315856934,
      "learning_rate": 0.00012462372014338872,
      "loss": 0.1796,
      "step": 32000
    },
    {
      "epoch": 0.44978564902663576,
      "eval_chrf": 77.43063803877338,
      "eval_loss": 0.39680206775665283,
      "eval_runtime": 326.2019,
      "eval_samples_per_second": 0.307,
      "eval_steps_per_second": 0.012,
      "step": 32000
    },
    {
      "epoch": 0.44992620704195657,
      "grad_norm": 0.31887194514274597,
      "learning_rate": 0.00012457866648774222,
      "loss": 0.1848,
      "step": 32010
    },
    {
      "epoch": 0.4500667650572774,
      "grad_norm": 0.391029953956604,
      "learning_rate": 0.00012453360752164093,
      "loss": 0.172,
      "step": 32020
    },
    {
      "epoch": 0.4502073230725982,
      "grad_norm": 0.4374319911003113,
      "learning_rate": 0.0001244885432548203,
      "loss": 0.1822,
      "step": 32030
    },
    {
      "epoch": 0.450347881087919,
      "grad_norm": 0.31356281042099,
      "learning_rate": 0.00012444347369701682,
      "loss": 0.1942,
      "step": 32040
    },
    {
      "epoch": 0.4504884391032399,
      "grad_norm": 0.41861143708229065,
      "learning_rate": 0.0001243983988579683,
      "loss": 0.1839,
      "step": 32050
    },
    {
      "epoch": 0.4506289971185607,
      "grad_norm": 0.3170023560523987,
      "learning_rate": 0.00012435331874741356,
      "loss": 0.1956,
      "step": 32060
    },
    {
      "epoch": 0.4507695551338815,
      "grad_norm": 0.3818584680557251,
      "learning_rate": 0.00012430823337509253,
      "loss": 0.1619,
      "step": 32070
    },
    {
      "epoch": 0.4509101131492023,
      "grad_norm": 0.4189601540565491,
      "learning_rate": 0.0001242631427507464,
      "loss": 0.1769,
      "step": 32080
    },
    {
      "epoch": 0.45105067116452313,
      "grad_norm": 0.3937597870826721,
      "learning_rate": 0.00012421804688411738,
      "loss": 0.1737,
      "step": 32090
    },
    {
      "epoch": 0.451191229179844,
      "grad_norm": 0.49900174140930176,
      "learning_rate": 0.0001241729457849489,
      "loss": 0.2054,
      "step": 32100
    },
    {
      "epoch": 0.4513317871951648,
      "grad_norm": 0.5906989574432373,
      "learning_rate": 0.00012412783946298542,
      "loss": 0.1727,
      "step": 32110
    },
    {
      "epoch": 0.4514723452104856,
      "grad_norm": 0.38759422302246094,
      "learning_rate": 0.00012408272792797267,
      "loss": 0.1829,
      "step": 32120
    },
    {
      "epoch": 0.45161290322580644,
      "grad_norm": 0.4419613182544708,
      "learning_rate": 0.0001240376111896574,
      "loss": 0.1961,
      "step": 32130
    },
    {
      "epoch": 0.45175346124112725,
      "grad_norm": 0.4294855296611786,
      "learning_rate": 0.00012399248925778746,
      "loss": 0.1987,
      "step": 32140
    },
    {
      "epoch": 0.4518940192564481,
      "grad_norm": 0.5556253790855408,
      "learning_rate": 0.00012394736214211196,
      "loss": 0.174,
      "step": 32150
    },
    {
      "epoch": 0.45203457727176893,
      "grad_norm": 0.3596500754356384,
      "learning_rate": 0.00012390222985238095,
      "loss": 0.194,
      "step": 32160
    },
    {
      "epoch": 0.45217513528708975,
      "grad_norm": 0.4233664572238922,
      "learning_rate": 0.0001238570923983458,
      "loss": 0.2017,
      "step": 32170
    },
    {
      "epoch": 0.45231569330241056,
      "grad_norm": 0.33549150824546814,
      "learning_rate": 0.00012381194978975876,
      "loss": 0.15,
      "step": 32180
    },
    {
      "epoch": 0.4524562513177314,
      "grad_norm": 0.47828513383865356,
      "learning_rate": 0.0001237668020363734,
      "loss": 0.1949,
      "step": 32190
    },
    {
      "epoch": 0.45259680933305224,
      "grad_norm": 0.3744308054447174,
      "learning_rate": 0.0001237216491479443,
      "loss": 0.1867,
      "step": 32200
    },
    {
      "epoch": 0.45273736734837305,
      "grad_norm": 0.4806154668331146,
      "learning_rate": 0.0001236764911342272,
      "loss": 0.1859,
      "step": 32210
    },
    {
      "epoch": 0.45287792536369387,
      "grad_norm": 0.4275056719779968,
      "learning_rate": 0.0001236313280049789,
      "loss": 0.1814,
      "step": 32220
    },
    {
      "epoch": 0.4530184833790147,
      "grad_norm": 0.5100582838058472,
      "learning_rate": 0.00012358615976995727,
      "loss": 0.1869,
      "step": 32230
    },
    {
      "epoch": 0.4531590413943355,
      "grad_norm": 0.4076481759548187,
      "learning_rate": 0.0001235409864389214,
      "loss": 0.2026,
      "step": 32240
    },
    {
      "epoch": 0.45329959940965636,
      "grad_norm": 0.4245436489582062,
      "learning_rate": 0.0001234958080216314,
      "loss": 0.2016,
      "step": 32250
    },
    {
      "epoch": 0.4534401574249772,
      "grad_norm": 0.3572922348976135,
      "learning_rate": 0.00012345062452784854,
      "loss": 0.1825,
      "step": 32260
    },
    {
      "epoch": 0.453580715440298,
      "grad_norm": 0.37890714406967163,
      "learning_rate": 0.00012340543596733508,
      "loss": 0.1654,
      "step": 32270
    },
    {
      "epoch": 0.4537212734556188,
      "grad_norm": 0.3865712285041809,
      "learning_rate": 0.00012336024234985444,
      "loss": 0.2064,
      "step": 32280
    },
    {
      "epoch": 0.4538618314709396,
      "grad_norm": 0.4462614357471466,
      "learning_rate": 0.00012331504368517118,
      "loss": 0.195,
      "step": 32290
    },
    {
      "epoch": 0.4540023894862604,
      "grad_norm": 0.43747371435165405,
      "learning_rate": 0.00012326983998305085,
      "loss": 0.1932,
      "step": 32300
    },
    {
      "epoch": 0.4541429475015813,
      "grad_norm": 0.44260385632514954,
      "learning_rate": 0.0001232246312532602,
      "loss": 0.1591,
      "step": 32310
    },
    {
      "epoch": 0.4542835055169021,
      "grad_norm": 0.3652852475643158,
      "learning_rate": 0.00012317941750556694,
      "loss": 0.195,
      "step": 32320
    },
    {
      "epoch": 0.4544240635322229,
      "grad_norm": 0.3209255635738373,
      "learning_rate": 0.00012313419874974,
      "loss": 0.1853,
      "step": 32330
    },
    {
      "epoch": 0.45456462154754373,
      "grad_norm": 0.2987498641014099,
      "learning_rate": 0.00012308897499554923,
      "loss": 0.196,
      "step": 32340
    },
    {
      "epoch": 0.45470517956286455,
      "grad_norm": 0.37293121218681335,
      "learning_rate": 0.00012304374625276575,
      "loss": 0.1855,
      "step": 32350
    },
    {
      "epoch": 0.4548457375781854,
      "grad_norm": 0.3847970962524414,
      "learning_rate": 0.0001229985125311616,
      "loss": 0.1982,
      "step": 32360
    },
    {
      "epoch": 0.45498629559350623,
      "grad_norm": 0.39437276124954224,
      "learning_rate": 0.00012295327384050997,
      "loss": 0.197,
      "step": 32370
    },
    {
      "epoch": 0.45512685360882704,
      "grad_norm": 0.44739240407943726,
      "learning_rate": 0.0001229080301905851,
      "loss": 0.1794,
      "step": 32380
    },
    {
      "epoch": 0.45526741162414786,
      "grad_norm": 0.4419465661048889,
      "learning_rate": 0.00012286278159116237,
      "loss": 0.2031,
      "step": 32390
    },
    {
      "epoch": 0.45540796963946867,
      "grad_norm": 0.3559831976890564,
      "learning_rate": 0.0001228175280520181,
      "loss": 0.1791,
      "step": 32400
    },
    {
      "epoch": 0.45554852765478954,
      "grad_norm": 0.45665082335472107,
      "learning_rate": 0.0001227722695829298,
      "loss": 0.1759,
      "step": 32410
    },
    {
      "epoch": 0.45568908567011035,
      "grad_norm": 0.30191051959991455,
      "learning_rate": 0.0001227270061936759,
      "loss": 0.1737,
      "step": 32420
    },
    {
      "epoch": 0.45582964368543116,
      "grad_norm": 0.392424076795578,
      "learning_rate": 0.0001226817378940361,
      "loss": 0.1998,
      "step": 32430
    },
    {
      "epoch": 0.455970201700752,
      "grad_norm": 0.41360557079315186,
      "learning_rate": 0.00012263646469379098,
      "loss": 0.1944,
      "step": 32440
    },
    {
      "epoch": 0.4561107597160728,
      "grad_norm": 0.4178352355957031,
      "learning_rate": 0.00012259118660272225,
      "loss": 0.1905,
      "step": 32450
    },
    {
      "epoch": 0.45625131773139366,
      "grad_norm": 0.3767535984516144,
      "learning_rate": 0.0001225459036306127,
      "loss": 0.162,
      "step": 32460
    },
    {
      "epoch": 0.45639187574671447,
      "grad_norm": 0.5038419961929321,
      "learning_rate": 0.00012250061578724614,
      "loss": 0.2058,
      "step": 32470
    },
    {
      "epoch": 0.4565324337620353,
      "grad_norm": 0.48213034868240356,
      "learning_rate": 0.00012245532308240742,
      "loss": 0.219,
      "step": 32480
    },
    {
      "epoch": 0.4566729917773561,
      "grad_norm": 0.3669455945491791,
      "learning_rate": 0.00012241002552588248,
      "loss": 0.1882,
      "step": 32490
    },
    {
      "epoch": 0.4568135497926769,
      "grad_norm": 0.32814040780067444,
      "learning_rate": 0.00012236472312745827,
      "loss": 0.1834,
      "step": 32500
    },
    {
      "epoch": 0.4569541078079978,
      "grad_norm": 0.6646811366081238,
      "learning_rate": 0.00012231941589692285,
      "loss": 0.2112,
      "step": 32510
    },
    {
      "epoch": 0.4570946658233186,
      "grad_norm": 0.3640930652618408,
      "learning_rate": 0.0001222741038440652,
      "loss": 0.1853,
      "step": 32520
    },
    {
      "epoch": 0.4572352238386394,
      "grad_norm": 0.34037432074546814,
      "learning_rate": 0.00012222878697867547,
      "loss": 0.2059,
      "step": 32530
    },
    {
      "epoch": 0.4573757818539602,
      "grad_norm": 0.6122976541519165,
      "learning_rate": 0.0001221834653105448,
      "loss": 0.1895,
      "step": 32540
    },
    {
      "epoch": 0.45751633986928103,
      "grad_norm": 0.4294646084308624,
      "learning_rate": 0.00012213813884946536,
      "loss": 0.1903,
      "step": 32550
    },
    {
      "epoch": 0.45765689788460184,
      "grad_norm": 0.46151578426361084,
      "learning_rate": 0.00012209280760523035,
      "loss": 0.1875,
      "step": 32560
    },
    {
      "epoch": 0.4577974558999227,
      "grad_norm": 0.4326536953449249,
      "learning_rate": 0.00012204747158763406,
      "loss": 0.1583,
      "step": 32570
    },
    {
      "epoch": 0.4579380139152435,
      "grad_norm": 0.4379238188266754,
      "learning_rate": 0.00012200213080647174,
      "loss": 0.1881,
      "step": 32580
    },
    {
      "epoch": 0.45807857193056434,
      "grad_norm": 0.4234411418437958,
      "learning_rate": 0.00012195678527153967,
      "loss": 0.17,
      "step": 32590
    },
    {
      "epoch": 0.45821912994588515,
      "grad_norm": 0.3968019485473633,
      "learning_rate": 0.00012191143499263522,
      "loss": 0.1975,
      "step": 32600
    },
    {
      "epoch": 0.45835968796120596,
      "grad_norm": 0.36948105692863464,
      "learning_rate": 0.00012186607997955675,
      "loss": 0.1973,
      "step": 32610
    },
    {
      "epoch": 0.45850024597652683,
      "grad_norm": 0.43759265542030334,
      "learning_rate": 0.0001218207202421036,
      "loss": 0.1862,
      "step": 32620
    },
    {
      "epoch": 0.45864080399184765,
      "grad_norm": 0.45381754636764526,
      "learning_rate": 0.00012177535579007621,
      "loss": 0.171,
      "step": 32630
    },
    {
      "epoch": 0.45878136200716846,
      "grad_norm": 0.33697906136512756,
      "learning_rate": 0.00012172998663327604,
      "loss": 0.2045,
      "step": 32640
    },
    {
      "epoch": 0.45892192002248927,
      "grad_norm": 0.39810094237327576,
      "learning_rate": 0.00012168461278150545,
      "loss": 0.1777,
      "step": 32650
    },
    {
      "epoch": 0.4590624780378101,
      "grad_norm": 0.4088006019592285,
      "learning_rate": 0.00012163923424456793,
      "loss": 0.1849,
      "step": 32660
    },
    {
      "epoch": 0.45920303605313095,
      "grad_norm": 0.3532857596874237,
      "learning_rate": 0.00012159385103226793,
      "loss": 0.1818,
      "step": 32670
    },
    {
      "epoch": 0.45934359406845177,
      "grad_norm": 0.44034630060195923,
      "learning_rate": 0.00012154846315441095,
      "loss": 0.1774,
      "step": 32680
    },
    {
      "epoch": 0.4594841520837726,
      "grad_norm": 0.43760716915130615,
      "learning_rate": 0.00012150307062080345,
      "loss": 0.1635,
      "step": 32690
    },
    {
      "epoch": 0.4596247100990934,
      "grad_norm": 0.28532445430755615,
      "learning_rate": 0.00012145767344125294,
      "loss": 0.1867,
      "step": 32700
    },
    {
      "epoch": 0.4597652681144142,
      "grad_norm": 0.3794820308685303,
      "learning_rate": 0.0001214122716255679,
      "loss": 0.1779,
      "step": 32710
    },
    {
      "epoch": 0.4599058261297351,
      "grad_norm": 0.39578428864479065,
      "learning_rate": 0.00012136686518355782,
      "loss": 0.1737,
      "step": 32720
    },
    {
      "epoch": 0.4600463841450559,
      "grad_norm": 0.4092610776424408,
      "learning_rate": 0.00012132145412503319,
      "loss": 0.2062,
      "step": 32730
    },
    {
      "epoch": 0.4601869421603767,
      "grad_norm": 0.374067485332489,
      "learning_rate": 0.00012127603845980553,
      "loss": 0.1978,
      "step": 32740
    },
    {
      "epoch": 0.4603275001756975,
      "grad_norm": 0.42492353916168213,
      "learning_rate": 0.00012123061819768729,
      "loss": 0.1772,
      "step": 32750
    },
    {
      "epoch": 0.4604680581910183,
      "grad_norm": 0.40157759189605713,
      "learning_rate": 0.00012118519334849196,
      "loss": 0.197,
      "step": 32760
    },
    {
      "epoch": 0.4606086162063392,
      "grad_norm": 0.4197980463504791,
      "learning_rate": 0.00012113976392203404,
      "loss": 0.177,
      "step": 32770
    },
    {
      "epoch": 0.46074917422166,
      "grad_norm": 0.2869431972503662,
      "learning_rate": 0.00012109432992812894,
      "loss": 0.1828,
      "step": 32780
    },
    {
      "epoch": 0.4608897322369808,
      "grad_norm": 0.45081043243408203,
      "learning_rate": 0.00012104889137659317,
      "loss": 0.1928,
      "step": 32790
    },
    {
      "epoch": 0.46103029025230163,
      "grad_norm": 0.2808906137943268,
      "learning_rate": 0.00012100344827724411,
      "loss": 0.1725,
      "step": 32800
    },
    {
      "epoch": 0.46117084826762245,
      "grad_norm": 0.4212302565574646,
      "learning_rate": 0.00012095800063990018,
      "loss": 0.1993,
      "step": 32810
    },
    {
      "epoch": 0.46131140628294326,
      "grad_norm": 0.36177191138267517,
      "learning_rate": 0.00012091254847438078,
      "loss": 0.1844,
      "step": 32820
    },
    {
      "epoch": 0.46145196429826413,
      "grad_norm": 0.3451072871685028,
      "learning_rate": 0.0001208670917905063,
      "loss": 0.1936,
      "step": 32830
    },
    {
      "epoch": 0.46159252231358494,
      "grad_norm": 0.3625692129135132,
      "learning_rate": 0.00012082163059809807,
      "loss": 0.1739,
      "step": 32840
    },
    {
      "epoch": 0.46173308032890575,
      "grad_norm": 0.4096881151199341,
      "learning_rate": 0.00012077616490697842,
      "loss": 0.1796,
      "step": 32850
    },
    {
      "epoch": 0.46187363834422657,
      "grad_norm": 0.342422753572464,
      "learning_rate": 0.00012073069472697063,
      "loss": 0.1936,
      "step": 32860
    },
    {
      "epoch": 0.4620141963595474,
      "grad_norm": 0.39686256647109985,
      "learning_rate": 0.00012068522006789899,
      "loss": 0.1709,
      "step": 32870
    },
    {
      "epoch": 0.46215475437486825,
      "grad_norm": 0.2695378065109253,
      "learning_rate": 0.00012063974093958874,
      "loss": 0.1746,
      "step": 32880
    },
    {
      "epoch": 0.46229531239018906,
      "grad_norm": 0.4101853370666504,
      "learning_rate": 0.00012059425735186605,
      "loss": 0.1949,
      "step": 32890
    },
    {
      "epoch": 0.4624358704055099,
      "grad_norm": 0.3779439330101013,
      "learning_rate": 0.00012054876931455807,
      "loss": 0.1822,
      "step": 32900
    },
    {
      "epoch": 0.4625764284208307,
      "grad_norm": 0.438286691904068,
      "learning_rate": 0.00012050327683749296,
      "loss": 0.1837,
      "step": 32910
    },
    {
      "epoch": 0.4627169864361515,
      "grad_norm": 0.45026054978370667,
      "learning_rate": 0.00012045777993049978,
      "loss": 0.1652,
      "step": 32920
    },
    {
      "epoch": 0.46285754445147237,
      "grad_norm": 0.3140639662742615,
      "learning_rate": 0.0001204122786034086,
      "loss": 0.1658,
      "step": 32930
    },
    {
      "epoch": 0.4629981024667932,
      "grad_norm": 0.3761960566043854,
      "learning_rate": 0.00012036677286605038,
      "loss": 0.1733,
      "step": 32940
    },
    {
      "epoch": 0.463138660482114,
      "grad_norm": 0.4367549419403076,
      "learning_rate": 0.00012032126272825708,
      "loss": 0.192,
      "step": 32950
    },
    {
      "epoch": 0.4632792184974348,
      "grad_norm": 0.5385770201683044,
      "learning_rate": 0.00012027574819986159,
      "loss": 0.17,
      "step": 32960
    },
    {
      "epoch": 0.4634197765127556,
      "grad_norm": 0.40825560688972473,
      "learning_rate": 0.00012023022929069778,
      "loss": 0.1941,
      "step": 32970
    },
    {
      "epoch": 0.4635603345280765,
      "grad_norm": 0.3881412744522095,
      "learning_rate": 0.00012018470601060042,
      "loss": 0.1755,
      "step": 32980
    },
    {
      "epoch": 0.4637008925433973,
      "grad_norm": 0.38460129499435425,
      "learning_rate": 0.00012013917836940525,
      "loss": 0.1712,
      "step": 32990
    },
    {
      "epoch": 0.4638414505587181,
      "grad_norm": 0.3805694580078125,
      "learning_rate": 0.00012009364637694897,
      "loss": 0.1568,
      "step": 33000
    },
    {
      "epoch": 0.4638414505587181,
      "eval_chrf": 80.88196750831037,
      "eval_loss": 0.40006187558174133,
      "eval_runtime": 237.8765,
      "eval_samples_per_second": 0.42,
      "eval_steps_per_second": 0.017,
      "step": 33000
    },
    {
      "epoch": 0.46398200857403893,
      "grad_norm": 0.6292216777801514,
      "learning_rate": 0.00012004811004306917,
      "loss": 0.1922,
      "step": 33010
    },
    {
      "epoch": 0.46412256658935974,
      "grad_norm": 0.33237358927726746,
      "learning_rate": 0.00012000256937760445,
      "loss": 0.169,
      "step": 33020
    },
    {
      "epoch": 0.46426312460468055,
      "grad_norm": 0.38910937309265137,
      "learning_rate": 0.00011995702439039426,
      "loss": 0.1917,
      "step": 33030
    },
    {
      "epoch": 0.4644036826200014,
      "grad_norm": 0.31341931223869324,
      "learning_rate": 0.00011991147509127907,
      "loss": 0.1707,
      "step": 33040
    },
    {
      "epoch": 0.46454424063532224,
      "grad_norm": 0.3623914122581482,
      "learning_rate": 0.00011986592149010019,
      "loss": 0.2166,
      "step": 33050
    },
    {
      "epoch": 0.46468479865064305,
      "grad_norm": 0.40682142972946167,
      "learning_rate": 0.00011982036359669995,
      "loss": 0.194,
      "step": 33060
    },
    {
      "epoch": 0.46482535666596386,
      "grad_norm": 0.3650200664997101,
      "learning_rate": 0.00011977480142092155,
      "loss": 0.1685,
      "step": 33070
    },
    {
      "epoch": 0.4649659146812847,
      "grad_norm": 0.48915278911590576,
      "learning_rate": 0.00011972923497260912,
      "loss": 0.1666,
      "step": 33080
    },
    {
      "epoch": 0.46510647269660554,
      "grad_norm": 0.3498603403568268,
      "learning_rate": 0.00011968366426160778,
      "loss": 0.2049,
      "step": 33090
    },
    {
      "epoch": 0.46524703071192636,
      "grad_norm": 0.5278698801994324,
      "learning_rate": 0.00011963808929776343,
      "loss": 0.1731,
      "step": 33100
    },
    {
      "epoch": 0.46538758872724717,
      "grad_norm": 0.3445676267147064,
      "learning_rate": 0.00011959251009092302,
      "loss": 0.1904,
      "step": 33110
    },
    {
      "epoch": 0.465528146742568,
      "grad_norm": 0.3266010880470276,
      "learning_rate": 0.00011954692665093441,
      "loss": 0.1839,
      "step": 33120
    },
    {
      "epoch": 0.4656687047578888,
      "grad_norm": 0.3357848525047302,
      "learning_rate": 0.00011950133898764626,
      "loss": 0.1787,
      "step": 33130
    },
    {
      "epoch": 0.46580926277320966,
      "grad_norm": 0.3921014964580536,
      "learning_rate": 0.00011945574711090828,
      "loss": 0.1854,
      "step": 33140
    },
    {
      "epoch": 0.4659498207885305,
      "grad_norm": 0.5705327987670898,
      "learning_rate": 0.00011941015103057098,
      "loss": 0.1831,
      "step": 33150
    },
    {
      "epoch": 0.4660903788038513,
      "grad_norm": 0.3367246687412262,
      "learning_rate": 0.0001193645507564859,
      "loss": 0.1946,
      "step": 33160
    },
    {
      "epoch": 0.4662309368191721,
      "grad_norm": 0.4650691747665405,
      "learning_rate": 0.0001193189462985053,
      "loss": 0.2061,
      "step": 33170
    },
    {
      "epoch": 0.4663714948344929,
      "grad_norm": 0.44068074226379395,
      "learning_rate": 0.00011927333766648254,
      "loss": 0.1625,
      "step": 33180
    },
    {
      "epoch": 0.4665120528498138,
      "grad_norm": 0.463360071182251,
      "learning_rate": 0.0001192277248702718,
      "loss": 0.1731,
      "step": 33190
    },
    {
      "epoch": 0.4666526108651346,
      "grad_norm": 0.669105589389801,
      "learning_rate": 0.00011918210791972812,
      "loss": 0.182,
      "step": 33200
    },
    {
      "epoch": 0.4667931688804554,
      "grad_norm": 0.33570921421051025,
      "learning_rate": 0.0001191364868247075,
      "loss": 0.1923,
      "step": 33210
    },
    {
      "epoch": 0.4669337268957762,
      "grad_norm": 0.37507128715515137,
      "learning_rate": 0.00011909086159506679,
      "loss": 0.1831,
      "step": 33220
    },
    {
      "epoch": 0.46707428491109704,
      "grad_norm": 0.5032942295074463,
      "learning_rate": 0.00011904523224066379,
      "loss": 0.1956,
      "step": 33230
    },
    {
      "epoch": 0.4672148429264179,
      "grad_norm": 0.4646669924259186,
      "learning_rate": 0.00011899959877135714,
      "loss": 0.1766,
      "step": 33240
    },
    {
      "epoch": 0.4673554009417387,
      "grad_norm": 0.41880184412002563,
      "learning_rate": 0.00011895396119700637,
      "loss": 0.1866,
      "step": 33250
    },
    {
      "epoch": 0.46749595895705953,
      "grad_norm": 0.46342238783836365,
      "learning_rate": 0.00011890831952747193,
      "loss": 0.1801,
      "step": 33260
    },
    {
      "epoch": 0.46763651697238034,
      "grad_norm": 0.36706215143203735,
      "learning_rate": 0.00011886267377261514,
      "loss": 0.1787,
      "step": 33270
    },
    {
      "epoch": 0.46777707498770116,
      "grad_norm": 0.4679487347602844,
      "learning_rate": 0.00011881702394229819,
      "loss": 0.1764,
      "step": 33280
    },
    {
      "epoch": 0.46791763300302197,
      "grad_norm": 0.3672434389591217,
      "learning_rate": 0.00011877137004638415,
      "loss": 0.1868,
      "step": 33290
    },
    {
      "epoch": 0.46805819101834284,
      "grad_norm": 0.3122919201850891,
      "learning_rate": 0.00011872571209473699,
      "loss": 0.1562,
      "step": 33300
    },
    {
      "epoch": 0.46819874903366365,
      "grad_norm": 0.5655485987663269,
      "learning_rate": 0.00011868005009722156,
      "loss": 0.1907,
      "step": 33310
    },
    {
      "epoch": 0.46833930704898447,
      "grad_norm": 0.4060775935649872,
      "learning_rate": 0.00011863438406370355,
      "loss": 0.1876,
      "step": 33320
    },
    {
      "epoch": 0.4684798650643053,
      "grad_norm": 0.41255131363868713,
      "learning_rate": 0.00011858871400404954,
      "loss": 0.1886,
      "step": 33330
    },
    {
      "epoch": 0.4686204230796261,
      "grad_norm": 0.312167763710022,
      "learning_rate": 0.00011854303992812699,
      "loss": 0.1772,
      "step": 33340
    },
    {
      "epoch": 0.46876098109494696,
      "grad_norm": 0.33613449335098267,
      "learning_rate": 0.0001184973618458042,
      "loss": 0.1711,
      "step": 33350
    },
    {
      "epoch": 0.4689015391102678,
      "grad_norm": 0.39489543437957764,
      "learning_rate": 0.00011845167976695039,
      "loss": 0.1929,
      "step": 33360
    },
    {
      "epoch": 0.4690420971255886,
      "grad_norm": 0.3719608783721924,
      "learning_rate": 0.00011840599370143559,
      "loss": 0.1497,
      "step": 33370
    },
    {
      "epoch": 0.4691826551409094,
      "grad_norm": 0.5516514778137207,
      "learning_rate": 0.00011836030365913069,
      "loss": 0.1848,
      "step": 33380
    },
    {
      "epoch": 0.4693232131562302,
      "grad_norm": 0.378025084733963,
      "learning_rate": 0.00011831460964990748,
      "loss": 0.1702,
      "step": 33390
    },
    {
      "epoch": 0.4694637711715511,
      "grad_norm": 0.37435686588287354,
      "learning_rate": 0.00011826891168363859,
      "loss": 0.1629,
      "step": 33400
    },
    {
      "epoch": 0.4696043291868719,
      "grad_norm": 0.432295024394989,
      "learning_rate": 0.00011822320977019748,
      "loss": 0.1856,
      "step": 33410
    },
    {
      "epoch": 0.4697448872021927,
      "grad_norm": 0.34483829140663147,
      "learning_rate": 0.00011817750391945849,
      "loss": 0.1782,
      "step": 33420
    },
    {
      "epoch": 0.4698854452175135,
      "grad_norm": 0.5057793855667114,
      "learning_rate": 0.00011813179414129683,
      "loss": 0.1911,
      "step": 33430
    },
    {
      "epoch": 0.47002600323283433,
      "grad_norm": 0.3595276176929474,
      "learning_rate": 0.00011808608044558848,
      "loss": 0.1798,
      "step": 33440
    },
    {
      "epoch": 0.4701665612481552,
      "grad_norm": 0.38613057136535645,
      "learning_rate": 0.00011804036284221038,
      "loss": 0.1955,
      "step": 33450
    },
    {
      "epoch": 0.470307119263476,
      "grad_norm": 0.4427388310432434,
      "learning_rate": 0.00011799464134104021,
      "loss": 0.1849,
      "step": 33460
    },
    {
      "epoch": 0.4704476772787968,
      "grad_norm": 0.4762212634086609,
      "learning_rate": 0.00011794891595195654,
      "loss": 0.1781,
      "step": 33470
    },
    {
      "epoch": 0.47058823529411764,
      "grad_norm": 0.495022714138031,
      "learning_rate": 0.0001179031866848388,
      "loss": 0.1846,
      "step": 33480
    },
    {
      "epoch": 0.47072879330943845,
      "grad_norm": 0.3033413589000702,
      "learning_rate": 0.00011785745354956721,
      "loss": 0.1617,
      "step": 33490
    },
    {
      "epoch": 0.4708693513247593,
      "grad_norm": 0.5457707643508911,
      "learning_rate": 0.00011781171655602285,
      "loss": 0.1808,
      "step": 33500
    },
    {
      "epoch": 0.47100990934008014,
      "grad_norm": 0.4955344498157501,
      "learning_rate": 0.00011776597571408765,
      "loss": 0.1851,
      "step": 33510
    },
    {
      "epoch": 0.47115046735540095,
      "grad_norm": 0.44962772727012634,
      "learning_rate": 0.00011772023103364434,
      "loss": 0.1811,
      "step": 33520
    },
    {
      "epoch": 0.47129102537072176,
      "grad_norm": 0.33938634395599365,
      "learning_rate": 0.00011767448252457649,
      "loss": 0.1868,
      "step": 33530
    },
    {
      "epoch": 0.4714315833860426,
      "grad_norm": 0.42045858502388,
      "learning_rate": 0.0001176287301967685,
      "loss": 0.1927,
      "step": 33540
    },
    {
      "epoch": 0.4715721414013634,
      "grad_norm": 0.5053248405456543,
      "learning_rate": 0.00011758297406010563,
      "loss": 0.1832,
      "step": 33550
    },
    {
      "epoch": 0.47171269941668426,
      "grad_norm": 0.4107218384742737,
      "learning_rate": 0.00011753721412447387,
      "loss": 0.1985,
      "step": 33560
    },
    {
      "epoch": 0.47185325743200507,
      "grad_norm": 0.3895529508590698,
      "learning_rate": 0.00011749145039976015,
      "loss": 0.1695,
      "step": 33570
    },
    {
      "epoch": 0.4719938154473259,
      "grad_norm": 0.45933929085731506,
      "learning_rate": 0.00011744568289585211,
      "loss": 0.1757,
      "step": 33580
    },
    {
      "epoch": 0.4721343734626467,
      "grad_norm": 0.3263451159000397,
      "learning_rate": 0.00011739991162263828,
      "loss": 0.1695,
      "step": 33590
    },
    {
      "epoch": 0.4722749314779675,
      "grad_norm": 0.4005551338195801,
      "learning_rate": 0.00011735413659000798,
      "loss": 0.1895,
      "step": 33600
    },
    {
      "epoch": 0.4724154894932884,
      "grad_norm": 0.42519721388816833,
      "learning_rate": 0.00011730835780785132,
      "loss": 0.1876,
      "step": 33610
    },
    {
      "epoch": 0.4725560475086092,
      "grad_norm": 0.4660035967826843,
      "learning_rate": 0.00011726257528605926,
      "loss": 0.1904,
      "step": 33620
    },
    {
      "epoch": 0.47269660552393,
      "grad_norm": 0.34425896406173706,
      "learning_rate": 0.00011721678903452356,
      "loss": 0.184,
      "step": 33630
    },
    {
      "epoch": 0.4728371635392508,
      "grad_norm": 0.46487653255462646,
      "learning_rate": 0.00011717099906313675,
      "loss": 0.1991,
      "step": 33640
    },
    {
      "epoch": 0.47297772155457163,
      "grad_norm": 0.3566063642501831,
      "learning_rate": 0.00011712520538179221,
      "loss": 0.1717,
      "step": 33650
    },
    {
      "epoch": 0.4731182795698925,
      "grad_norm": 0.41969212889671326,
      "learning_rate": 0.00011707940800038408,
      "loss": 0.1789,
      "step": 33660
    },
    {
      "epoch": 0.4732588375852133,
      "grad_norm": 0.41644376516342163,
      "learning_rate": 0.00011703360692880734,
      "loss": 0.1804,
      "step": 33670
    },
    {
      "epoch": 0.4733993956005341,
      "grad_norm": 0.3613830506801605,
      "learning_rate": 0.00011698780217695771,
      "loss": 0.1883,
      "step": 33680
    },
    {
      "epoch": 0.47353995361585494,
      "grad_norm": 0.38608357310295105,
      "learning_rate": 0.0001169419937547318,
      "loss": 0.1838,
      "step": 33690
    },
    {
      "epoch": 0.47368051163117575,
      "grad_norm": 0.44402819871902466,
      "learning_rate": 0.00011689618167202692,
      "loss": 0.2057,
      "step": 33700
    },
    {
      "epoch": 0.4738210696464966,
      "grad_norm": 0.3911150097846985,
      "learning_rate": 0.00011685036593874122,
      "loss": 0.1801,
      "step": 33710
    },
    {
      "epoch": 0.47396162766181743,
      "grad_norm": 0.3988668620586395,
      "learning_rate": 0.0001168045465647736,
      "loss": 0.1965,
      "step": 33720
    },
    {
      "epoch": 0.47410218567713824,
      "grad_norm": 0.39634349942207336,
      "learning_rate": 0.00011675872356002377,
      "loss": 0.1985,
      "step": 33730
    },
    {
      "epoch": 0.47424274369245906,
      "grad_norm": 0.2985493838787079,
      "learning_rate": 0.0001167128969343923,
      "loss": 0.1831,
      "step": 33740
    },
    {
      "epoch": 0.47438330170777987,
      "grad_norm": 0.41120585799217224,
      "learning_rate": 0.00011666706669778038,
      "loss": 0.1944,
      "step": 33750
    },
    {
      "epoch": 0.4745238597231007,
      "grad_norm": 0.5060003399848938,
      "learning_rate": 0.0001166212328600901,
      "loss": 0.1971,
      "step": 33760
    },
    {
      "epoch": 0.47466441773842155,
      "grad_norm": 0.44410645961761475,
      "learning_rate": 0.0001165753954312243,
      "loss": 0.193,
      "step": 33770
    },
    {
      "epoch": 0.47480497575374236,
      "grad_norm": 0.3515392541885376,
      "learning_rate": 0.00011652955442108662,
      "loss": 0.1792,
      "step": 33780
    },
    {
      "epoch": 0.4749455337690632,
      "grad_norm": 0.3999476432800293,
      "learning_rate": 0.00011648370983958139,
      "loss": 0.1837,
      "step": 33790
    },
    {
      "epoch": 0.475086091784384,
      "grad_norm": 0.40338027477264404,
      "learning_rate": 0.00011643786169661379,
      "loss": 0.2021,
      "step": 33800
    },
    {
      "epoch": 0.4752266497997048,
      "grad_norm": 0.27882617712020874,
      "learning_rate": 0.0001163920100020898,
      "loss": 0.1635,
      "step": 33810
    },
    {
      "epoch": 0.47536720781502567,
      "grad_norm": 0.5009968876838684,
      "learning_rate": 0.00011634615476591601,
      "loss": 0.1913,
      "step": 33820
    },
    {
      "epoch": 0.4755077658303465,
      "grad_norm": 0.4280058741569519,
      "learning_rate": 0.000116300295998,
      "loss": 0.2067,
      "step": 33830
    },
    {
      "epoch": 0.4756483238456673,
      "grad_norm": 0.43252554535865784,
      "learning_rate": 0.00011625443370824989,
      "loss": 0.1769,
      "step": 33840
    },
    {
      "epoch": 0.4757888818609881,
      "grad_norm": 0.4498821198940277,
      "learning_rate": 0.00011620856790657472,
      "loss": 0.1958,
      "step": 33850
    },
    {
      "epoch": 0.4759294398763089,
      "grad_norm": 0.4288261830806732,
      "learning_rate": 0.00011616269860288422,
      "loss": 0.1859,
      "step": 33860
    },
    {
      "epoch": 0.4760699978916298,
      "grad_norm": 0.3644183576107025,
      "learning_rate": 0.00011611682580708889,
      "loss": 0.1715,
      "step": 33870
    },
    {
      "epoch": 0.4762105559069506,
      "grad_norm": 0.29036372900009155,
      "learning_rate": 0.00011607094952909998,
      "loss": 0.1746,
      "step": 33880
    },
    {
      "epoch": 0.4763511139222714,
      "grad_norm": 0.4704986810684204,
      "learning_rate": 0.0001160250697788295,
      "loss": 0.1794,
      "step": 33890
    },
    {
      "epoch": 0.47649167193759223,
      "grad_norm": 0.31704598665237427,
      "learning_rate": 0.0001159791865661902,
      "loss": 0.2009,
      "step": 33900
    },
    {
      "epoch": 0.47663222995291304,
      "grad_norm": 0.39006519317626953,
      "learning_rate": 0.00011593329990109558,
      "loss": 0.1608,
      "step": 33910
    },
    {
      "epoch": 0.4767727879682339,
      "grad_norm": 0.3246588706970215,
      "learning_rate": 0.00011588740979345993,
      "loss": 0.1874,
      "step": 33920
    },
    {
      "epoch": 0.4769133459835547,
      "grad_norm": 0.5718732476234436,
      "learning_rate": 0.0001158415162531982,
      "loss": 0.1862,
      "step": 33930
    },
    {
      "epoch": 0.47705390399887554,
      "grad_norm": 0.34852126240730286,
      "learning_rate": 0.00011579561929022613,
      "loss": 0.1918,
      "step": 33940
    },
    {
      "epoch": 0.47719446201419635,
      "grad_norm": 0.3804478645324707,
      "learning_rate": 0.00011574971891446022,
      "loss": 0.1837,
      "step": 33950
    },
    {
      "epoch": 0.47733502002951717,
      "grad_norm": 0.29604998230934143,
      "learning_rate": 0.00011570381513581768,
      "loss": 0.1955,
      "step": 33960
    },
    {
      "epoch": 0.47747557804483803,
      "grad_norm": 0.6176908016204834,
      "learning_rate": 0.00011565790796421645,
      "loss": 0.1702,
      "step": 33970
    },
    {
      "epoch": 0.47761613606015885,
      "grad_norm": 0.4599247872829437,
      "learning_rate": 0.00011561199740957521,
      "loss": 0.1996,
      "step": 33980
    },
    {
      "epoch": 0.47775669407547966,
      "grad_norm": 0.3114668130874634,
      "learning_rate": 0.00011556608348181335,
      "loss": 0.1835,
      "step": 33990
    },
    {
      "epoch": 0.4778972520908005,
      "grad_norm": 0.3145674765110016,
      "learning_rate": 0.00011552016619085104,
      "loss": 0.1832,
      "step": 34000
    },
    {
      "epoch": 0.4778972520908005,
      "eval_chrf": 82.71781475485383,
      "eval_loss": 0.3998299837112427,
      "eval_runtime": 202.4047,
      "eval_samples_per_second": 0.494,
      "eval_steps_per_second": 0.02,
      "step": 34000
    },
    {
      "epoch": 0.4780378101061213,
      "grad_norm": 0.4320267140865326,
      "learning_rate": 0.00011547424554660915,
      "loss": 0.188,
      "step": 34010
    },
    {
      "epoch": 0.4781783681214421,
      "grad_norm": 0.4681837558746338,
      "learning_rate": 0.00011542832155900924,
      "loss": 0.195,
      "step": 34020
    },
    {
      "epoch": 0.47831892613676297,
      "grad_norm": 0.32885053753852844,
      "learning_rate": 0.00011538239423797367,
      "loss": 0.1598,
      "step": 34030
    },
    {
      "epoch": 0.4784594841520838,
      "grad_norm": 0.4490990936756134,
      "learning_rate": 0.00011533646359342544,
      "loss": 0.1798,
      "step": 34040
    },
    {
      "epoch": 0.4786000421674046,
      "grad_norm": 0.3614675998687744,
      "learning_rate": 0.00011529052963528835,
      "loss": 0.1913,
      "step": 34050
    },
    {
      "epoch": 0.4787406001827254,
      "grad_norm": 0.5846191048622131,
      "learning_rate": 0.0001152445923734868,
      "loss": 0.1761,
      "step": 34060
    },
    {
      "epoch": 0.4788811581980462,
      "grad_norm": 0.3790275454521179,
      "learning_rate": 0.000115198651817946,
      "loss": 0.1933,
      "step": 34070
    },
    {
      "epoch": 0.4790217162133671,
      "grad_norm": 0.33133602142333984,
      "learning_rate": 0.00011515270797859187,
      "loss": 0.1782,
      "step": 34080
    },
    {
      "epoch": 0.4791622742286879,
      "grad_norm": 0.3211809992790222,
      "learning_rate": 0.000115106760865351,
      "loss": 0.1747,
      "step": 34090
    },
    {
      "epoch": 0.4793028322440087,
      "grad_norm": 0.624617874622345,
      "learning_rate": 0.00011506081048815067,
      "loss": 0.2045,
      "step": 34100
    },
    {
      "epoch": 0.4794433902593295,
      "grad_norm": 0.5131089687347412,
      "learning_rate": 0.00011501485685691893,
      "loss": 0.1975,
      "step": 34110
    },
    {
      "epoch": 0.47958394827465034,
      "grad_norm": 0.64466792345047,
      "learning_rate": 0.00011496889998158451,
      "loss": 0.2255,
      "step": 34120
    },
    {
      "epoch": 0.4797245062899712,
      "grad_norm": 0.4620768427848816,
      "learning_rate": 0.00011492293987207678,
      "loss": 0.2072,
      "step": 34130
    },
    {
      "epoch": 0.479865064305292,
      "grad_norm": 0.5024957060813904,
      "learning_rate": 0.00011487697653832591,
      "loss": 0.1834,
      "step": 34140
    },
    {
      "epoch": 0.48000562232061283,
      "grad_norm": 0.3953956663608551,
      "learning_rate": 0.00011483100999026268,
      "loss": 0.1889,
      "step": 34150
    },
    {
      "epoch": 0.48014618033593365,
      "grad_norm": 0.44761160016059875,
      "learning_rate": 0.00011478504023781864,
      "loss": 0.1971,
      "step": 34160
    },
    {
      "epoch": 0.48028673835125446,
      "grad_norm": 0.4793362319469452,
      "learning_rate": 0.00011473906729092596,
      "loss": 0.1936,
      "step": 34170
    },
    {
      "epoch": 0.48042729636657533,
      "grad_norm": 0.5443357825279236,
      "learning_rate": 0.00011469309115951754,
      "loss": 0.203,
      "step": 34180
    },
    {
      "epoch": 0.48056785438189614,
      "grad_norm": 0.5695284008979797,
      "learning_rate": 0.00011464711185352697,
      "loss": 0.1706,
      "step": 34190
    },
    {
      "epoch": 0.48070841239721696,
      "grad_norm": 0.4707348048686981,
      "learning_rate": 0.0001146011293828885,
      "loss": 0.1784,
      "step": 34200
    },
    {
      "epoch": 0.48084897041253777,
      "grad_norm": 0.43354344367980957,
      "learning_rate": 0.00011455514375753706,
      "loss": 0.1865,
      "step": 34210
    },
    {
      "epoch": 0.4809895284278586,
      "grad_norm": 0.6231280565261841,
      "learning_rate": 0.00011450915498740835,
      "loss": 0.1621,
      "step": 34220
    },
    {
      "epoch": 0.48113008644317945,
      "grad_norm": 0.36138397455215454,
      "learning_rate": 0.0001144631630824386,
      "loss": 0.1945,
      "step": 34230
    },
    {
      "epoch": 0.48127064445850026,
      "grad_norm": 0.4865313768386841,
      "learning_rate": 0.00011441716805256488,
      "loss": 0.1698,
      "step": 34240
    },
    {
      "epoch": 0.4814112024738211,
      "grad_norm": 0.35231879353523254,
      "learning_rate": 0.0001143711699077248,
      "loss": 0.204,
      "step": 34250
    },
    {
      "epoch": 0.4815517604891419,
      "grad_norm": 0.5219005942344666,
      "learning_rate": 0.00011432516865785672,
      "loss": 0.1762,
      "step": 34260
    },
    {
      "epoch": 0.4816923185044627,
      "grad_norm": 0.39501020312309265,
      "learning_rate": 0.00011427916431289963,
      "loss": 0.1715,
      "step": 34270
    },
    {
      "epoch": 0.4818328765197835,
      "grad_norm": 0.4571538269519806,
      "learning_rate": 0.00011423315688279319,
      "loss": 0.1695,
      "step": 34280
    },
    {
      "epoch": 0.4819734345351044,
      "grad_norm": 0.29014673829078674,
      "learning_rate": 0.0001141871463774778,
      "loss": 0.1737,
      "step": 34290
    },
    {
      "epoch": 0.4821139925504252,
      "grad_norm": 0.38009896874427795,
      "learning_rate": 0.00011414113280689442,
      "loss": 0.1883,
      "step": 34300
    },
    {
      "epoch": 0.482254550565746,
      "grad_norm": 0.44037485122680664,
      "learning_rate": 0.00011409511618098477,
      "loss": 0.1773,
      "step": 34310
    },
    {
      "epoch": 0.4823951085810668,
      "grad_norm": 0.30522382259368896,
      "learning_rate": 0.00011404909650969111,
      "loss": 0.1687,
      "step": 34320
    },
    {
      "epoch": 0.48253566659638764,
      "grad_norm": 0.345973402261734,
      "learning_rate": 0.00011400307380295648,
      "loss": 0.1884,
      "step": 34330
    },
    {
      "epoch": 0.4826762246117085,
      "grad_norm": 0.46845465898513794,
      "learning_rate": 0.00011395704807072449,
      "loss": 0.1864,
      "step": 34340
    },
    {
      "epoch": 0.4828167826270293,
      "grad_norm": 0.5156724452972412,
      "learning_rate": 0.00011391101932293946,
      "loss": 0.1744,
      "step": 34350
    },
    {
      "epoch": 0.48295734064235013,
      "grad_norm": 0.27931588888168335,
      "learning_rate": 0.00011386498756954633,
      "loss": 0.1949,
      "step": 34360
    },
    {
      "epoch": 0.48309789865767094,
      "grad_norm": 0.38670992851257324,
      "learning_rate": 0.0001138189528204907,
      "loss": 0.1696,
      "step": 34370
    },
    {
      "epoch": 0.48323845667299176,
      "grad_norm": 0.4227330982685089,
      "learning_rate": 0.00011377291508571884,
      "loss": 0.1929,
      "step": 34380
    },
    {
      "epoch": 0.4833790146883126,
      "grad_norm": 0.4957590401172638,
      "learning_rate": 0.00011372687437517761,
      "loss": 0.1895,
      "step": 34390
    },
    {
      "epoch": 0.48351957270363344,
      "grad_norm": 0.3703496754169464,
      "learning_rate": 0.00011368083069881454,
      "loss": 0.1713,
      "step": 34400
    },
    {
      "epoch": 0.48366013071895425,
      "grad_norm": 0.41321736574172974,
      "learning_rate": 0.00011363478406657782,
      "loss": 0.1755,
      "step": 34410
    },
    {
      "epoch": 0.48380068873427506,
      "grad_norm": 0.35261067748069763,
      "learning_rate": 0.00011358873448841627,
      "loss": 0.1685,
      "step": 34420
    },
    {
      "epoch": 0.4839412467495959,
      "grad_norm": 0.44572609663009644,
      "learning_rate": 0.00011354268197427931,
      "loss": 0.2073,
      "step": 34430
    },
    {
      "epoch": 0.48408180476491675,
      "grad_norm": 0.4484621286392212,
      "learning_rate": 0.00011349662653411705,
      "loss": 0.2035,
      "step": 34440
    },
    {
      "epoch": 0.48422236278023756,
      "grad_norm": 0.32330384850502014,
      "learning_rate": 0.00011345056817788023,
      "loss": 0.1734,
      "step": 34450
    },
    {
      "epoch": 0.48436292079555837,
      "grad_norm": 0.45375800132751465,
      "learning_rate": 0.00011340450691552011,
      "loss": 0.163,
      "step": 34460
    },
    {
      "epoch": 0.4845034788108792,
      "grad_norm": 0.36283940076828003,
      "learning_rate": 0.00011335844275698874,
      "loss": 0.1856,
      "step": 34470
    },
    {
      "epoch": 0.4846440368262,
      "grad_norm": 0.3621930181980133,
      "learning_rate": 0.00011331237571223872,
      "loss": 0.2008,
      "step": 34480
    },
    {
      "epoch": 0.4847845948415208,
      "grad_norm": 0.35948091745376587,
      "learning_rate": 0.00011326630579122323,
      "loss": 0.1699,
      "step": 34490
    },
    {
      "epoch": 0.4849251528568417,
      "grad_norm": 0.31326571106910706,
      "learning_rate": 0.00011322023300389617,
      "loss": 0.1733,
      "step": 34500
    },
    {
      "epoch": 0.4850657108721625,
      "grad_norm": 0.41531944274902344,
      "learning_rate": 0.00011317415736021195,
      "loss": 0.1616,
      "step": 34510
    },
    {
      "epoch": 0.4852062688874833,
      "grad_norm": 0.39085519313812256,
      "learning_rate": 0.00011312807887012566,
      "loss": 0.1741,
      "step": 34520
    },
    {
      "epoch": 0.4853468269028041,
      "grad_norm": 0.44744712114334106,
      "learning_rate": 0.00011308199754359303,
      "loss": 0.2084,
      "step": 34530
    },
    {
      "epoch": 0.48548738491812493,
      "grad_norm": 0.4303813576698303,
      "learning_rate": 0.00011303591339057034,
      "loss": 0.1712,
      "step": 34540
    },
    {
      "epoch": 0.4856279429334458,
      "grad_norm": 0.3569890260696411,
      "learning_rate": 0.0001129898264210145,
      "loss": 0.1827,
      "step": 34550
    },
    {
      "epoch": 0.4857685009487666,
      "grad_norm": 0.40748974680900574,
      "learning_rate": 0.00011294373664488307,
      "loss": 0.191,
      "step": 34560
    },
    {
      "epoch": 0.4859090589640874,
      "grad_norm": 0.3596547544002533,
      "learning_rate": 0.00011289764407213417,
      "loss": 0.1876,
      "step": 34570
    },
    {
      "epoch": 0.48604961697940824,
      "grad_norm": 0.4161217212677002,
      "learning_rate": 0.00011285154871272652,
      "loss": 0.1968,
      "step": 34580
    },
    {
      "epoch": 0.48619017499472905,
      "grad_norm": 0.3585597574710846,
      "learning_rate": 0.00011280545057661951,
      "loss": 0.1848,
      "step": 34590
    },
    {
      "epoch": 0.4863307330100499,
      "grad_norm": 0.4169002175331116,
      "learning_rate": 0.00011275934967377305,
      "loss": 0.1901,
      "step": 34600
    },
    {
      "epoch": 0.48647129102537073,
      "grad_norm": 0.39894920587539673,
      "learning_rate": 0.00011271324601414766,
      "loss": 0.1787,
      "step": 34610
    },
    {
      "epoch": 0.48661184904069155,
      "grad_norm": 0.46989011764526367,
      "learning_rate": 0.0001126671396077045,
      "loss": 0.2102,
      "step": 34620
    },
    {
      "epoch": 0.48675240705601236,
      "grad_norm": 0.3442949950695038,
      "learning_rate": 0.00011262103046440528,
      "loss": 0.1696,
      "step": 34630
    },
    {
      "epoch": 0.4868929650713332,
      "grad_norm": 0.35811588168144226,
      "learning_rate": 0.00011257491859421234,
      "loss": 0.1651,
      "step": 34640
    },
    {
      "epoch": 0.48703352308665404,
      "grad_norm": 0.40460866689682007,
      "learning_rate": 0.00011252880400708861,
      "loss": 0.1604,
      "step": 34650
    },
    {
      "epoch": 0.48717408110197485,
      "grad_norm": 0.3593714237213135,
      "learning_rate": 0.00011248268671299751,
      "loss": 0.2043,
      "step": 34660
    },
    {
      "epoch": 0.48731463911729567,
      "grad_norm": 0.38151001930236816,
      "learning_rate": 0.00011243656672190319,
      "loss": 0.1869,
      "step": 34670
    },
    {
      "epoch": 0.4874551971326165,
      "grad_norm": 0.3958202004432678,
      "learning_rate": 0.00011239044404377029,
      "loss": 0.17,
      "step": 34680
    },
    {
      "epoch": 0.4875957551479373,
      "grad_norm": 0.5403009653091431,
      "learning_rate": 0.00011234431868856406,
      "loss": 0.1945,
      "step": 34690
    },
    {
      "epoch": 0.48773631316325816,
      "grad_norm": 0.34916040301322937,
      "learning_rate": 0.00011229819066625029,
      "loss": 0.1911,
      "step": 34700
    },
    {
      "epoch": 0.487876871178579,
      "grad_norm": 0.40562403202056885,
      "learning_rate": 0.00011225205998679539,
      "loss": 0.1918,
      "step": 34710
    },
    {
      "epoch": 0.4880174291938998,
      "grad_norm": 0.47156044840812683,
      "learning_rate": 0.00011220592666016638,
      "loss": 0.1842,
      "step": 34720
    },
    {
      "epoch": 0.4881579872092206,
      "grad_norm": 0.3460146486759186,
      "learning_rate": 0.00011215979069633075,
      "loss": 0.1727,
      "step": 34730
    },
    {
      "epoch": 0.4882985452245414,
      "grad_norm": 0.447365403175354,
      "learning_rate": 0.00011211365210525664,
      "loss": 0.2014,
      "step": 34740
    },
    {
      "epoch": 0.4884391032398622,
      "grad_norm": 0.3855605125427246,
      "learning_rate": 0.00011206751089691272,
      "loss": 0.1682,
      "step": 34750
    },
    {
      "epoch": 0.4885796612551831,
      "grad_norm": 0.3858765959739685,
      "learning_rate": 0.00011202136708126825,
      "loss": 0.1727,
      "step": 34760
    },
    {
      "epoch": 0.4887202192705039,
      "grad_norm": 0.4090711176395416,
      "learning_rate": 0.00011197522066829303,
      "loss": 0.1872,
      "step": 34770
    },
    {
      "epoch": 0.4888607772858247,
      "grad_norm": 0.3905404806137085,
      "learning_rate": 0.00011192907166795741,
      "loss": 0.1525,
      "step": 34780
    },
    {
      "epoch": 0.48900133530114553,
      "grad_norm": 0.3775665760040283,
      "learning_rate": 0.00011188292009023235,
      "loss": 0.1803,
      "step": 34790
    },
    {
      "epoch": 0.48914189331646635,
      "grad_norm": 0.44145843386650085,
      "learning_rate": 0.00011183676594508937,
      "loss": 0.1932,
      "step": 34800
    },
    {
      "epoch": 0.4892824513317872,
      "grad_norm": 0.4132518470287323,
      "learning_rate": 0.00011179060924250043,
      "loss": 0.1701,
      "step": 34810
    },
    {
      "epoch": 0.48942300934710803,
      "grad_norm": 0.33350783586502075,
      "learning_rate": 0.00011174444999243819,
      "loss": 0.1734,
      "step": 34820
    },
    {
      "epoch": 0.48956356736242884,
      "grad_norm": 0.41543808579444885,
      "learning_rate": 0.00011169828820487575,
      "loss": 0.1971,
      "step": 34830
    },
    {
      "epoch": 0.48970412537774966,
      "grad_norm": 0.39029350876808167,
      "learning_rate": 0.00011165212388978685,
      "loss": 0.1684,
      "step": 34840
    },
    {
      "epoch": 0.48984468339307047,
      "grad_norm": 0.5022748708724976,
      "learning_rate": 0.00011160595705714569,
      "loss": 0.176,
      "step": 34850
    },
    {
      "epoch": 0.48998524140839134,
      "grad_norm": 0.3474188446998596,
      "learning_rate": 0.0001115597877169271,
      "loss": 0.1958,
      "step": 34860
    },
    {
      "epoch": 0.49012579942371215,
      "grad_norm": 0.3608366549015045,
      "learning_rate": 0.00011151361587910635,
      "loss": 0.204,
      "step": 34870
    },
    {
      "epoch": 0.49026635743903296,
      "grad_norm": 0.3608495593070984,
      "learning_rate": 0.00011146744155365934,
      "loss": 0.1718,
      "step": 34880
    },
    {
      "epoch": 0.4904069154543538,
      "grad_norm": 0.358041912317276,
      "learning_rate": 0.00011142126475056245,
      "loss": 0.1927,
      "step": 34890
    },
    {
      "epoch": 0.4905474734696746,
      "grad_norm": 0.4198744595050812,
      "learning_rate": 0.00011137508547979265,
      "loss": 0.1851,
      "step": 34900
    },
    {
      "epoch": 0.49068803148499546,
      "grad_norm": 0.36275073885917664,
      "learning_rate": 0.00011132890375132736,
      "loss": 0.1778,
      "step": 34910
    },
    {
      "epoch": 0.49082858950031627,
      "grad_norm": 0.42901909351348877,
      "learning_rate": 0.00011128271957514465,
      "loss": 0.1908,
      "step": 34920
    },
    {
      "epoch": 0.4909691475156371,
      "grad_norm": 0.5087999105453491,
      "learning_rate": 0.00011123653296122298,
      "loss": 0.1899,
      "step": 34930
    },
    {
      "epoch": 0.4911097055309579,
      "grad_norm": 0.3971782624721527,
      "learning_rate": 0.00011119034391954146,
      "loss": 0.1712,
      "step": 34940
    },
    {
      "epoch": 0.4912502635462787,
      "grad_norm": 0.4209510385990143,
      "learning_rate": 0.00011114415246007968,
      "loss": 0.1814,
      "step": 34950
    },
    {
      "epoch": 0.4913908215615996,
      "grad_norm": 0.32000836730003357,
      "learning_rate": 0.00011109795859281767,
      "loss": 0.1681,
      "step": 34960
    },
    {
      "epoch": 0.4915313795769204,
      "grad_norm": 0.43289345502853394,
      "learning_rate": 0.00011105176232773612,
      "loss": 0.175,
      "step": 34970
    },
    {
      "epoch": 0.4916719375922412,
      "grad_norm": 0.39267903566360474,
      "learning_rate": 0.00011100556367481616,
      "loss": 0.178,
      "step": 34980
    },
    {
      "epoch": 0.491812495607562,
      "grad_norm": 0.5739196538925171,
      "learning_rate": 0.00011095936264403945,
      "loss": 0.1908,
      "step": 34990
    },
    {
      "epoch": 0.49195305362288283,
      "grad_norm": 0.3983236849308014,
      "learning_rate": 0.00011091315924538814,
      "loss": 0.169,
      "step": 35000
    },
    {
      "epoch": 0.49195305362288283,
      "eval_chrf": 83.79861109162626,
      "eval_loss": 0.39429256319999695,
      "eval_runtime": 139.9694,
      "eval_samples_per_second": 0.714,
      "eval_steps_per_second": 0.029,
      "step": 35000
    },
    {
      "epoch": 0.49209361163820364,
      "grad_norm": 0.4108808636665344,
      "learning_rate": 0.00011086695348884493,
      "loss": 0.1764,
      "step": 35010
    },
    {
      "epoch": 0.4922341696535245,
      "grad_norm": 0.5308189392089844,
      "learning_rate": 0.00011082074538439302,
      "loss": 0.1863,
      "step": 35020
    },
    {
      "epoch": 0.4923747276688453,
      "grad_norm": 0.39313215017318726,
      "learning_rate": 0.0001107745349420161,
      "loss": 0.1595,
      "step": 35030
    },
    {
      "epoch": 0.49251528568416614,
      "grad_norm": 0.3336087167263031,
      "learning_rate": 0.00011072832217169839,
      "loss": 0.1688,
      "step": 35040
    },
    {
      "epoch": 0.49265584369948695,
      "grad_norm": 0.41592684388160706,
      "learning_rate": 0.00011068210708342459,
      "loss": 0.1841,
      "step": 35050
    },
    {
      "epoch": 0.49279640171480776,
      "grad_norm": 0.48767977952957153,
      "learning_rate": 0.00011063588968717989,
      "loss": 0.1711,
      "step": 35060
    },
    {
      "epoch": 0.49293695973012863,
      "grad_norm": 0.4140823185443878,
      "learning_rate": 0.00011058966999295002,
      "loss": 0.1836,
      "step": 35070
    },
    {
      "epoch": 0.49307751774544945,
      "grad_norm": 0.4304581880569458,
      "learning_rate": 0.00011054344801072119,
      "loss": 0.179,
      "step": 35080
    },
    {
      "epoch": 0.49321807576077026,
      "grad_norm": 0.43712833523750305,
      "learning_rate": 0.00011049722375048008,
      "loss": 0.1792,
      "step": 35090
    },
    {
      "epoch": 0.49335863377609107,
      "grad_norm": 0.4752121567726135,
      "learning_rate": 0.0001104509972222139,
      "loss": 0.1902,
      "step": 35100
    },
    {
      "epoch": 0.4934991917914119,
      "grad_norm": 0.4985567331314087,
      "learning_rate": 0.00011040476843591032,
      "loss": 0.1821,
      "step": 35110
    },
    {
      "epoch": 0.49363974980673275,
      "grad_norm": 0.31013724207878113,
      "learning_rate": 0.0001103585374015575,
      "loss": 0.1438,
      "step": 35120
    },
    {
      "epoch": 0.49378030782205357,
      "grad_norm": 0.46870386600494385,
      "learning_rate": 0.00011031230412914413,
      "loss": 0.1936,
      "step": 35130
    },
    {
      "epoch": 0.4939208658373744,
      "grad_norm": 0.35234829783439636,
      "learning_rate": 0.00011026606862865932,
      "loss": 0.1676,
      "step": 35140
    },
    {
      "epoch": 0.4940614238526952,
      "grad_norm": 0.34757283329963684,
      "learning_rate": 0.0001102198309100927,
      "loss": 0.1774,
      "step": 35150
    },
    {
      "epoch": 0.494201981868016,
      "grad_norm": 0.6265631914138794,
      "learning_rate": 0.00011017359098343434,
      "loss": 0.1761,
      "step": 35160
    },
    {
      "epoch": 0.4943425398833369,
      "grad_norm": 0.4004918336868286,
      "learning_rate": 0.00011012734885867486,
      "loss": 0.2101,
      "step": 35170
    },
    {
      "epoch": 0.4944830978986577,
      "grad_norm": 0.4307462275028229,
      "learning_rate": 0.00011008110454580529,
      "loss": 0.1699,
      "step": 35180
    },
    {
      "epoch": 0.4946236559139785,
      "grad_norm": 0.37476179003715515,
      "learning_rate": 0.00011003485805481715,
      "loss": 0.1985,
      "step": 35190
    },
    {
      "epoch": 0.4947642139292993,
      "grad_norm": 0.38108786940574646,
      "learning_rate": 0.00010998860939570243,
      "loss": 0.1686,
      "step": 35200
    },
    {
      "epoch": 0.4949047719446201,
      "grad_norm": 0.4117715358734131,
      "learning_rate": 0.0001099423585784536,
      "loss": 0.1774,
      "step": 35210
    },
    {
      "epoch": 0.49504532995994094,
      "grad_norm": 0.37168529629707336,
      "learning_rate": 0.00010989610561306363,
      "loss": 0.1572,
      "step": 35220
    },
    {
      "epoch": 0.4951858879752618,
      "grad_norm": 0.3484684228897095,
      "learning_rate": 0.00010984985050952583,
      "loss": 0.1696,
      "step": 35230
    },
    {
      "epoch": 0.4953264459905826,
      "grad_norm": 0.5094234347343445,
      "learning_rate": 0.00010980359327783412,
      "loss": 0.1819,
      "step": 35240
    },
    {
      "epoch": 0.49546700400590343,
      "grad_norm": 0.5020202398300171,
      "learning_rate": 0.0001097573339279828,
      "loss": 0.2098,
      "step": 35250
    },
    {
      "epoch": 0.49560756202122425,
      "grad_norm": 0.44014009833335876,
      "learning_rate": 0.00010971107246996664,
      "loss": 0.2071,
      "step": 35260
    },
    {
      "epoch": 0.49574812003654506,
      "grad_norm": 0.25457870960235596,
      "learning_rate": 0.00010966480891378083,
      "loss": 0.1948,
      "step": 35270
    },
    {
      "epoch": 0.4958886780518659,
      "grad_norm": 0.34224721789360046,
      "learning_rate": 0.00010961854326942114,
      "loss": 0.1739,
      "step": 35280
    },
    {
      "epoch": 0.49602923606718674,
      "grad_norm": 0.2786223888397217,
      "learning_rate": 0.00010957227554688361,
      "loss": 0.1844,
      "step": 35290
    },
    {
      "epoch": 0.49616979408250755,
      "grad_norm": 0.461157888174057,
      "learning_rate": 0.00010952600575616487,
      "loss": 0.1579,
      "step": 35300
    },
    {
      "epoch": 0.49631035209782837,
      "grad_norm": 0.3331812620162964,
      "learning_rate": 0.00010947973390726196,
      "loss": 0.1923,
      "step": 35310
    },
    {
      "epoch": 0.4964509101131492,
      "grad_norm": 0.45609837770462036,
      "learning_rate": 0.00010943346001017232,
      "loss": 0.1975,
      "step": 35320
    },
    {
      "epoch": 0.49659146812847005,
      "grad_norm": 0.4462905526161194,
      "learning_rate": 0.00010938718407489388,
      "loss": 0.1738,
      "step": 35330
    },
    {
      "epoch": 0.49673202614379086,
      "grad_norm": 0.3662757873535156,
      "learning_rate": 0.00010934090611142501,
      "loss": 0.1827,
      "step": 35340
    },
    {
      "epoch": 0.4968725841591117,
      "grad_norm": 0.37068015336990356,
      "learning_rate": 0.00010929462612976446,
      "loss": 0.1705,
      "step": 35350
    },
    {
      "epoch": 0.4970131421744325,
      "grad_norm": 0.4905817210674286,
      "learning_rate": 0.00010924834413991152,
      "loss": 0.2016,
      "step": 35360
    },
    {
      "epoch": 0.4971537001897533,
      "grad_norm": 0.4112740457057953,
      "learning_rate": 0.0001092020601518658,
      "loss": 0.1686,
      "step": 35370
    },
    {
      "epoch": 0.49729425820507417,
      "grad_norm": 0.31365418434143066,
      "learning_rate": 0.00010915577417562743,
      "loss": 0.1827,
      "step": 35380
    },
    {
      "epoch": 0.497434816220395,
      "grad_norm": 0.32228732109069824,
      "learning_rate": 0.00010910948622119695,
      "loss": 0.1683,
      "step": 35390
    },
    {
      "epoch": 0.4975753742357158,
      "grad_norm": 0.3350965082645416,
      "learning_rate": 0.00010906319629857524,
      "loss": 0.1751,
      "step": 35400
    },
    {
      "epoch": 0.4977159322510366,
      "grad_norm": 0.3961012065410614,
      "learning_rate": 0.00010901690441776377,
      "loss": 0.1752,
      "step": 35410
    },
    {
      "epoch": 0.4978564902663574,
      "grad_norm": 0.4828808307647705,
      "learning_rate": 0.00010897061058876428,
      "loss": 0.1969,
      "step": 35420
    },
    {
      "epoch": 0.4979970482816783,
      "grad_norm": 0.39320072531700134,
      "learning_rate": 0.00010892431482157903,
      "loss": 0.1851,
      "step": 35430
    },
    {
      "epoch": 0.4981376062969991,
      "grad_norm": 0.403592586517334,
      "learning_rate": 0.00010887801712621062,
      "loss": 0.1842,
      "step": 35440
    },
    {
      "epoch": 0.4982781643123199,
      "grad_norm": 0.5812364816665649,
      "learning_rate": 0.00010883171751266215,
      "loss": 0.1895,
      "step": 35450
    },
    {
      "epoch": 0.49841872232764073,
      "grad_norm": 0.48084115982055664,
      "learning_rate": 0.00010878541599093706,
      "loss": 0.2011,
      "step": 35460
    },
    {
      "epoch": 0.49855928034296154,
      "grad_norm": 0.5177582502365112,
      "learning_rate": 0.00010873911257103926,
      "loss": 0.1797,
      "step": 35470
    },
    {
      "epoch": 0.49869983835828235,
      "grad_norm": 0.5985971689224243,
      "learning_rate": 0.00010869280726297302,
      "loss": 0.1998,
      "step": 35480
    },
    {
      "epoch": 0.4988403963736032,
      "grad_norm": 0.48021918535232544,
      "learning_rate": 0.00010864650007674304,
      "loss": 0.1723,
      "step": 35490
    },
    {
      "epoch": 0.49898095438892404,
      "grad_norm": 0.35807913541793823,
      "learning_rate": 0.00010860019102235445,
      "loss": 0.165,
      "step": 35500
    },
    {
      "epoch": 0.49912151240424485,
      "grad_norm": 0.5980018377304077,
      "learning_rate": 0.0001085538801098128,
      "loss": 0.1654,
      "step": 35510
    },
    {
      "epoch": 0.49926207041956566,
      "grad_norm": 0.46984145045280457,
      "learning_rate": 0.00010850756734912391,
      "loss": 0.2066,
      "step": 35520
    },
    {
      "epoch": 0.4994026284348865,
      "grad_norm": 0.3486853241920471,
      "learning_rate": 0.00010846125275029415,
      "loss": 0.1677,
      "step": 35530
    },
    {
      "epoch": 0.49954318645020734,
      "grad_norm": 0.5842266082763672,
      "learning_rate": 0.00010841493632333021,
      "loss": 0.1851,
      "step": 35540
    },
    {
      "epoch": 0.49968374446552816,
      "grad_norm": 0.30013322830200195,
      "learning_rate": 0.00010836861807823921,
      "loss": 0.1776,
      "step": 35550
    },
    {
      "epoch": 0.49982430248084897,
      "grad_norm": 0.3671824634075165,
      "learning_rate": 0.00010832229802502864,
      "loss": 0.1957,
      "step": 35560
    },
    {
      "epoch": 0.4999648604961698,
      "grad_norm": 0.42613282799720764,
      "learning_rate": 0.00010827597617370639,
      "loss": 0.1783,
      "step": 35570
    },
    {
      "epoch": 0.5001054185114906,
      "grad_norm": 0.3783090114593506,
      "learning_rate": 0.00010822965253428074,
      "loss": 0.2099,
      "step": 35580
    },
    {
      "epoch": 0.5002459765268115,
      "grad_norm": 0.3716335892677307,
      "learning_rate": 0.00010818332711676036,
      "loss": 0.1741,
      "step": 35590
    },
    {
      "epoch": 0.5003865345421322,
      "grad_norm": 0.336635023355484,
      "learning_rate": 0.00010813699993115427,
      "loss": 0.189,
      "step": 35600
    },
    {
      "epoch": 0.5005270925574531,
      "grad_norm": 0.46448060870170593,
      "learning_rate": 0.00010809067098747191,
      "loss": 0.1969,
      "step": 35610
    },
    {
      "epoch": 0.500667650572774,
      "grad_norm": 0.41883373260498047,
      "learning_rate": 0.00010804434029572313,
      "loss": 0.1931,
      "step": 35620
    },
    {
      "epoch": 0.5008082085880947,
      "grad_norm": 0.3746285140514374,
      "learning_rate": 0.00010799800786591805,
      "loss": 0.2023,
      "step": 35630
    },
    {
      "epoch": 0.5009487666034156,
      "grad_norm": 0.5401817560195923,
      "learning_rate": 0.00010795167370806726,
      "loss": 0.1662,
      "step": 35640
    },
    {
      "epoch": 0.5010893246187363,
      "grad_norm": 0.34712448716163635,
      "learning_rate": 0.00010790533783218174,
      "loss": 0.1881,
      "step": 35650
    },
    {
      "epoch": 0.5012298826340572,
      "grad_norm": 0.3793562650680542,
      "learning_rate": 0.00010785900024827271,
      "loss": 0.1591,
      "step": 35660
    },
    {
      "epoch": 0.5013704406493781,
      "grad_norm": 0.3823815882205963,
      "learning_rate": 0.00010781266096635191,
      "loss": 0.1899,
      "step": 35670
    },
    {
      "epoch": 0.5015109986646988,
      "grad_norm": 0.30518558621406555,
      "learning_rate": 0.00010776631999643138,
      "loss": 0.1503,
      "step": 35680
    },
    {
      "epoch": 0.5016515566800197,
      "grad_norm": 0.4431426227092743,
      "learning_rate": 0.0001077199773485235,
      "loss": 0.1786,
      "step": 35690
    },
    {
      "epoch": 0.5017921146953405,
      "grad_norm": 0.490865021944046,
      "learning_rate": 0.00010767363303264106,
      "loss": 0.1917,
      "step": 35700
    },
    {
      "epoch": 0.5019326727106613,
      "grad_norm": 0.4799773693084717,
      "learning_rate": 0.00010762728705879719,
      "loss": 0.1869,
      "step": 35710
    },
    {
      "epoch": 0.5020732307259822,
      "grad_norm": 0.3862857520580292,
      "learning_rate": 0.00010758093943700538,
      "loss": 0.1859,
      "step": 35720
    },
    {
      "epoch": 0.502213788741303,
      "grad_norm": 0.43539267778396606,
      "learning_rate": 0.00010753459017727945,
      "loss": 0.2054,
      "step": 35730
    },
    {
      "epoch": 0.5023543467566238,
      "grad_norm": 0.43648383021354675,
      "learning_rate": 0.00010748823928963361,
      "loss": 0.1594,
      "step": 35740
    },
    {
      "epoch": 0.5024949047719446,
      "grad_norm": 0.44934573769569397,
      "learning_rate": 0.00010744188678408242,
      "loss": 0.1726,
      "step": 35750
    },
    {
      "epoch": 0.5026354627872655,
      "grad_norm": 0.4945816993713379,
      "learning_rate": 0.0001073955326706408,
      "loss": 0.1557,
      "step": 35760
    },
    {
      "epoch": 0.5027760208025863,
      "grad_norm": 0.38272902369499207,
      "learning_rate": 0.00010734917695932394,
      "loss": 0.1952,
      "step": 35770
    },
    {
      "epoch": 0.5029165788179071,
      "grad_norm": 0.4691140949726105,
      "learning_rate": 0.00010730281966014748,
      "loss": 0.1731,
      "step": 35780
    },
    {
      "epoch": 0.503057136833228,
      "grad_norm": 0.4082811176776886,
      "learning_rate": 0.00010725646078312734,
      "loss": 0.1875,
      "step": 35790
    },
    {
      "epoch": 0.5031976948485487,
      "grad_norm": 0.5158867239952087,
      "learning_rate": 0.00010721010033827981,
      "loss": 0.183,
      "step": 35800
    },
    {
      "epoch": 0.5033382528638696,
      "grad_norm": 0.399926096200943,
      "learning_rate": 0.00010716373833562147,
      "loss": 0.2157,
      "step": 35810
    },
    {
      "epoch": 0.5034788108791903,
      "grad_norm": 0.3821859359741211,
      "learning_rate": 0.00010711737478516933,
      "loss": 0.1705,
      "step": 35820
    },
    {
      "epoch": 0.5036193688945112,
      "grad_norm": 0.5483487844467163,
      "learning_rate": 0.00010707100969694062,
      "loss": 0.1743,
      "step": 35830
    },
    {
      "epoch": 0.5037599269098321,
      "grad_norm": 0.35209783911705017,
      "learning_rate": 0.00010702464308095297,
      "loss": 0.1829,
      "step": 35840
    },
    {
      "epoch": 0.5039004849251528,
      "grad_norm": 0.4123036861419678,
      "learning_rate": 0.00010697827494722435,
      "loss": 0.1919,
      "step": 35850
    },
    {
      "epoch": 0.5040410429404737,
      "grad_norm": 0.548042893409729,
      "learning_rate": 0.00010693190530577301,
      "loss": 0.206,
      "step": 35860
    },
    {
      "epoch": 0.5041816009557945,
      "grad_norm": 0.40403738617897034,
      "learning_rate": 0.0001068855341666176,
      "loss": 0.1972,
      "step": 35870
    },
    {
      "epoch": 0.5043221589711153,
      "grad_norm": 0.3454335927963257,
      "learning_rate": 0.000106839161539777,
      "loss": 0.1972,
      "step": 35880
    },
    {
      "epoch": 0.5044627169864362,
      "grad_norm": 0.2906029522418976,
      "learning_rate": 0.00010679278743527046,
      "loss": 0.1753,
      "step": 35890
    },
    {
      "epoch": 0.504603275001757,
      "grad_norm": 0.46441200375556946,
      "learning_rate": 0.00010674641186311756,
      "loss": 0.1804,
      "step": 35900
    },
    {
      "epoch": 0.5047438330170778,
      "grad_norm": 0.4945662319660187,
      "learning_rate": 0.00010670003483333819,
      "loss": 0.1874,
      "step": 35910
    },
    {
      "epoch": 0.5048843910323986,
      "grad_norm": 0.44752687215805054,
      "learning_rate": 0.00010665365635595252,
      "loss": 0.191,
      "step": 35920
    },
    {
      "epoch": 0.5050249490477194,
      "grad_norm": 0.3905443549156189,
      "learning_rate": 0.0001066072764409811,
      "loss": 0.166,
      "step": 35930
    },
    {
      "epoch": 0.5051655070630403,
      "grad_norm": 0.547477662563324,
      "learning_rate": 0.00010656089509844474,
      "loss": 0.1905,
      "step": 35940
    },
    {
      "epoch": 0.5053060650783611,
      "grad_norm": 0.3551482558250427,
      "learning_rate": 0.00010651451233836454,
      "loss": 0.1978,
      "step": 35950
    },
    {
      "epoch": 0.5054466230936819,
      "grad_norm": 0.5016312003135681,
      "learning_rate": 0.00010646812817076197,
      "loss": 0.2035,
      "step": 35960
    },
    {
      "epoch": 0.5055871811090027,
      "grad_norm": 0.33267927169799805,
      "learning_rate": 0.00010642174260565876,
      "loss": 0.1741,
      "step": 35970
    },
    {
      "epoch": 0.5057277391243236,
      "grad_norm": 0.41453447937965393,
      "learning_rate": 0.00010637535565307695,
      "loss": 0.1701,
      "step": 35980
    },
    {
      "epoch": 0.5058682971396444,
      "grad_norm": 0.4634517729282379,
      "learning_rate": 0.00010632896732303889,
      "loss": 0.1914,
      "step": 35990
    },
    {
      "epoch": 0.5060088551549652,
      "grad_norm": 0.3329765498638153,
      "learning_rate": 0.00010628257762556722,
      "loss": 0.1665,
      "step": 36000
    },
    {
      "epoch": 0.5060088551549652,
      "eval_chrf": 79.90633488812688,
      "eval_loss": 0.3829784393310547,
      "eval_runtime": 284.4675,
      "eval_samples_per_second": 0.352,
      "eval_steps_per_second": 0.014,
      "step": 36000
    },
    {
      "epoch": 0.506149413170286,
      "grad_norm": 0.432464599609375,
      "learning_rate": 0.00010623618657068486,
      "loss": 0.1979,
      "step": 36010
    },
    {
      "epoch": 0.5062899711856068,
      "grad_norm": 0.3946160078048706,
      "learning_rate": 0.00010618979416841505,
      "loss": 0.1917,
      "step": 36020
    },
    {
      "epoch": 0.5064305292009277,
      "grad_norm": 0.39581990242004395,
      "learning_rate": 0.00010614340042878132,
      "loss": 0.1966,
      "step": 36030
    },
    {
      "epoch": 0.5065710872162486,
      "grad_norm": 0.4072232246398926,
      "learning_rate": 0.00010609700536180749,
      "loss": 0.183,
      "step": 36040
    },
    {
      "epoch": 0.5067116452315693,
      "grad_norm": 0.38204532861709595,
      "learning_rate": 0.0001060506089775176,
      "loss": 0.1692,
      "step": 36050
    },
    {
      "epoch": 0.5068522032468902,
      "grad_norm": 0.553989827632904,
      "learning_rate": 0.00010600421128593608,
      "loss": 0.1958,
      "step": 36060
    },
    {
      "epoch": 0.5069927612622109,
      "grad_norm": 0.4289276599884033,
      "learning_rate": 0.00010595781229708756,
      "loss": 0.1753,
      "step": 36070
    },
    {
      "epoch": 0.5071333192775318,
      "grad_norm": 0.41937553882598877,
      "learning_rate": 0.00010591141202099704,
      "loss": 0.1925,
      "step": 36080
    },
    {
      "epoch": 0.5072738772928527,
      "grad_norm": 0.39642074704170227,
      "learning_rate": 0.00010586501046768967,
      "loss": 0.1653,
      "step": 36090
    },
    {
      "epoch": 0.5074144353081734,
      "grad_norm": 0.36946389079093933,
      "learning_rate": 0.00010581860764719098,
      "loss": 0.2015,
      "step": 36100
    },
    {
      "epoch": 0.5075549933234943,
      "grad_norm": 0.4972825348377228,
      "learning_rate": 0.00010577220356952676,
      "loss": 0.1818,
      "step": 36110
    },
    {
      "epoch": 0.507695551338815,
      "grad_norm": 0.36207687854766846,
      "learning_rate": 0.00010572579824472302,
      "loss": 0.1725,
      "step": 36120
    },
    {
      "epoch": 0.5078361093541359,
      "grad_norm": 0.4180412292480469,
      "learning_rate": 0.00010567939168280608,
      "loss": 0.1728,
      "step": 36130
    },
    {
      "epoch": 0.5079766673694568,
      "grad_norm": 0.3445020616054535,
      "learning_rate": 0.00010563298389380251,
      "loss": 0.1779,
      "step": 36140
    },
    {
      "epoch": 0.5081172253847775,
      "grad_norm": 0.41422396898269653,
      "learning_rate": 0.0001055865748877392,
      "loss": 0.1967,
      "step": 36150
    },
    {
      "epoch": 0.5082577834000984,
      "grad_norm": 0.34044012427330017,
      "learning_rate": 0.00010554016467464321,
      "loss": 0.1912,
      "step": 36160
    },
    {
      "epoch": 0.5083983414154192,
      "grad_norm": 0.30843567848205566,
      "learning_rate": 0.00010549375326454194,
      "loss": 0.1828,
      "step": 36170
    },
    {
      "epoch": 0.50853889943074,
      "grad_norm": 0.40784013271331787,
      "learning_rate": 0.000105447340667463,
      "loss": 0.1861,
      "step": 36180
    },
    {
      "epoch": 0.5086794574460609,
      "grad_norm": 0.49464085698127747,
      "learning_rate": 0.00010540092689343427,
      "loss": 0.1882,
      "step": 36190
    },
    {
      "epoch": 0.5088200154613817,
      "grad_norm": 0.4711960554122925,
      "learning_rate": 0.00010535451195248387,
      "loss": 0.184,
      "step": 36200
    },
    {
      "epoch": 0.5089605734767025,
      "grad_norm": 0.34780821204185486,
      "learning_rate": 0.00010530809585464023,
      "loss": 0.1787,
      "step": 36210
    },
    {
      "epoch": 0.5091011314920233,
      "grad_norm": 0.6330233216285706,
      "learning_rate": 0.00010526167860993197,
      "loss": 0.2015,
      "step": 36220
    },
    {
      "epoch": 0.5092416895073442,
      "grad_norm": 0.36124908924102783,
      "learning_rate": 0.00010521526022838797,
      "loss": 0.167,
      "step": 36230
    },
    {
      "epoch": 0.509382247522665,
      "grad_norm": 0.6212900876998901,
      "learning_rate": 0.00010516884072003739,
      "loss": 0.1863,
      "step": 36240
    },
    {
      "epoch": 0.5095228055379858,
      "grad_norm": 0.3643467128276825,
      "learning_rate": 0.00010512242009490956,
      "loss": 0.1832,
      "step": 36250
    },
    {
      "epoch": 0.5096633635533067,
      "grad_norm": 0.34473875164985657,
      "learning_rate": 0.00010507599836303415,
      "loss": 0.174,
      "step": 36260
    },
    {
      "epoch": 0.5098039215686274,
      "grad_norm": 0.4868568480014801,
      "learning_rate": 0.00010502957553444097,
      "loss": 0.1681,
      "step": 36270
    },
    {
      "epoch": 0.5099444795839483,
      "grad_norm": 0.3094613254070282,
      "learning_rate": 0.00010498315161916015,
      "loss": 0.167,
      "step": 36280
    },
    {
      "epoch": 0.5100850375992692,
      "grad_norm": 0.47234678268432617,
      "learning_rate": 0.000104936726627222,
      "loss": 0.1805,
      "step": 36290
    },
    {
      "epoch": 0.5102255956145899,
      "grad_norm": 0.3965536057949066,
      "learning_rate": 0.0001048903005686571,
      "loss": 0.1735,
      "step": 36300
    },
    {
      "epoch": 0.5103661536299108,
      "grad_norm": 0.4829418957233429,
      "learning_rate": 0.0001048438734534962,
      "loss": 0.1909,
      "step": 36310
    },
    {
      "epoch": 0.5105067116452315,
      "grad_norm": 0.44158634543418884,
      "learning_rate": 0.00010479744529177038,
      "loss": 0.1855,
      "step": 36320
    },
    {
      "epoch": 0.5106472696605524,
      "grad_norm": 1.4402402639389038,
      "learning_rate": 0.00010475101609351083,
      "loss": 0.1744,
      "step": 36330
    },
    {
      "epoch": 0.5107878276758732,
      "grad_norm": 0.455005019903183,
      "learning_rate": 0.00010470458586874906,
      "loss": 0.1945,
      "step": 36340
    },
    {
      "epoch": 0.510928385691194,
      "grad_norm": 0.31440451741218567,
      "learning_rate": 0.00010465815462751674,
      "loss": 0.1733,
      "step": 36350
    },
    {
      "epoch": 0.5110689437065149,
      "grad_norm": 0.40663766860961914,
      "learning_rate": 0.00010461172237984578,
      "loss": 0.1881,
      "step": 36360
    },
    {
      "epoch": 0.5112095017218357,
      "grad_norm": 0.36541232466697693,
      "learning_rate": 0.00010456528913576833,
      "loss": 0.1849,
      "step": 36370
    },
    {
      "epoch": 0.5113500597371565,
      "grad_norm": 0.28649237751960754,
      "learning_rate": 0.00010451885490531672,
      "loss": 0.17,
      "step": 36380
    },
    {
      "epoch": 0.5114906177524773,
      "grad_norm": 0.3698398768901825,
      "learning_rate": 0.0001044724196985235,
      "loss": 0.178,
      "step": 36390
    },
    {
      "epoch": 0.5116311757677982,
      "grad_norm": 0.45252224802970886,
      "learning_rate": 0.00010442598352542143,
      "loss": 0.1966,
      "step": 36400
    },
    {
      "epoch": 0.511771733783119,
      "grad_norm": 0.38496294617652893,
      "learning_rate": 0.00010437954639604352,
      "loss": 0.1808,
      "step": 36410
    },
    {
      "epoch": 0.5119122917984398,
      "grad_norm": 0.4141913056373596,
      "learning_rate": 0.00010433310832042296,
      "loss": 0.1911,
      "step": 36420
    },
    {
      "epoch": 0.5120528498137606,
      "grad_norm": 0.4155624806880951,
      "learning_rate": 0.0001042866693085931,
      "loss": 0.1737,
      "step": 36430
    },
    {
      "epoch": 0.5121934078290814,
      "grad_norm": 0.3937687575817108,
      "learning_rate": 0.00010424022937058752,
      "loss": 0.1819,
      "step": 36440
    },
    {
      "epoch": 0.5123339658444023,
      "grad_norm": 0.40030568838119507,
      "learning_rate": 0.00010419378851644006,
      "loss": 0.1696,
      "step": 36450
    },
    {
      "epoch": 0.5124745238597231,
      "grad_norm": 0.42675095796585083,
      "learning_rate": 0.0001041473467561847,
      "loss": 0.1785,
      "step": 36460
    },
    {
      "epoch": 0.5126150818750439,
      "grad_norm": 0.5587252378463745,
      "learning_rate": 0.00010410090409985558,
      "loss": 0.1895,
      "step": 36470
    },
    {
      "epoch": 0.5127556398903648,
      "grad_norm": 0.3191625475883484,
      "learning_rate": 0.00010405446055748712,
      "loss": 0.1687,
      "step": 36480
    },
    {
      "epoch": 0.5128961979056855,
      "grad_norm": 0.393315851688385,
      "learning_rate": 0.00010400801613911386,
      "loss": 0.1804,
      "step": 36490
    },
    {
      "epoch": 0.5130367559210064,
      "grad_norm": 0.3991680443286896,
      "learning_rate": 0.00010396157085477059,
      "loss": 0.1911,
      "step": 36500
    },
    {
      "epoch": 0.5131773139363273,
      "grad_norm": 0.45736780762672424,
      "learning_rate": 0.00010391512471449223,
      "loss": 0.1984,
      "step": 36510
    },
    {
      "epoch": 0.513317871951648,
      "grad_norm": 0.392051100730896,
      "learning_rate": 0.00010386867772831394,
      "loss": 0.1757,
      "step": 36520
    },
    {
      "epoch": 0.5134584299669689,
      "grad_norm": 0.3728495240211487,
      "learning_rate": 0.00010382222990627102,
      "loss": 0.1999,
      "step": 36530
    },
    {
      "epoch": 0.5135989879822896,
      "grad_norm": 0.41388094425201416,
      "learning_rate": 0.00010377578125839893,
      "loss": 0.1693,
      "step": 36540
    },
    {
      "epoch": 0.5137395459976105,
      "grad_norm": 0.35319089889526367,
      "learning_rate": 0.00010372933179473338,
      "loss": 0.1824,
      "step": 36550
    },
    {
      "epoch": 0.5138801040129314,
      "grad_norm": 0.5678279399871826,
      "learning_rate": 0.0001036828815253102,
      "loss": 0.1639,
      "step": 36560
    },
    {
      "epoch": 0.5140206620282521,
      "grad_norm": 0.40534478425979614,
      "learning_rate": 0.00010363643046016545,
      "loss": 0.1565,
      "step": 36570
    },
    {
      "epoch": 0.514161220043573,
      "grad_norm": 0.40824031829833984,
      "learning_rate": 0.00010358997860933528,
      "loss": 0.1728,
      "step": 36580
    },
    {
      "epoch": 0.5143017780588938,
      "grad_norm": 0.45494553446769714,
      "learning_rate": 0.00010354352598285606,
      "loss": 0.2081,
      "step": 36590
    },
    {
      "epoch": 0.5144423360742146,
      "grad_norm": 0.4255465269088745,
      "learning_rate": 0.00010349707259076433,
      "loss": 0.1785,
      "step": 36600
    },
    {
      "epoch": 0.5145828940895355,
      "grad_norm": 0.4779079854488373,
      "learning_rate": 0.0001034506184430968,
      "loss": 0.1776,
      "step": 36610
    },
    {
      "epoch": 0.5147234521048563,
      "grad_norm": 0.40576520562171936,
      "learning_rate": 0.00010340416354989033,
      "loss": 0.1777,
      "step": 36620
    },
    {
      "epoch": 0.5148640101201771,
      "grad_norm": 0.38735461235046387,
      "learning_rate": 0.00010335770792118191,
      "loss": 0.1616,
      "step": 36630
    },
    {
      "epoch": 0.5150045681354979,
      "grad_norm": 0.32425543665885925,
      "learning_rate": 0.00010331125156700876,
      "loss": 0.1661,
      "step": 36640
    },
    {
      "epoch": 0.5151451261508188,
      "grad_norm": 0.4369176924228668,
      "learning_rate": 0.0001032647944974082,
      "loss": 0.1912,
      "step": 36650
    },
    {
      "epoch": 0.5152856841661396,
      "grad_norm": 0.3734516501426697,
      "learning_rate": 0.00010321833672241772,
      "loss": 0.1793,
      "step": 36660
    },
    {
      "epoch": 0.5154262421814604,
      "grad_norm": 0.45591041445732117,
      "learning_rate": 0.00010317187825207497,
      "loss": 0.1922,
      "step": 36670
    },
    {
      "epoch": 0.5155668001967812,
      "grad_norm": 0.4752703905105591,
      "learning_rate": 0.00010312541909641773,
      "loss": 0.1755,
      "step": 36680
    },
    {
      "epoch": 0.515707358212102,
      "grad_norm": 0.41179245710372925,
      "learning_rate": 0.00010307895926548396,
      "loss": 0.1888,
      "step": 36690
    },
    {
      "epoch": 0.5158479162274229,
      "grad_norm": 0.5543412566184998,
      "learning_rate": 0.00010303249876931174,
      "loss": 0.1806,
      "step": 36700
    },
    {
      "epoch": 0.5159884742427437,
      "grad_norm": 0.40941378474235535,
      "learning_rate": 0.0001029860376179393,
      "loss": 0.1757,
      "step": 36710
    },
    {
      "epoch": 0.5161290322580645,
      "grad_norm": 0.4833003580570221,
      "learning_rate": 0.00010293957582140503,
      "loss": 0.197,
      "step": 36720
    },
    {
      "epoch": 0.5162695902733854,
      "grad_norm": 0.40841785073280334,
      "learning_rate": 0.00010289311338974743,
      "loss": 0.1865,
      "step": 36730
    },
    {
      "epoch": 0.5164101482887061,
      "grad_norm": 0.5991966128349304,
      "learning_rate": 0.00010284665033300515,
      "loss": 0.1688,
      "step": 36740
    },
    {
      "epoch": 0.516550706304027,
      "grad_norm": 0.31613221764564514,
      "learning_rate": 0.00010280018666121698,
      "loss": 0.1724,
      "step": 36750
    },
    {
      "epoch": 0.5166912643193479,
      "grad_norm": 0.36570432782173157,
      "learning_rate": 0.00010275372238442185,
      "loss": 0.1873,
      "step": 36760
    },
    {
      "epoch": 0.5168318223346686,
      "grad_norm": 0.2841842770576477,
      "learning_rate": 0.00010270725751265877,
      "loss": 0.1729,
      "step": 36770
    },
    {
      "epoch": 0.5169723803499895,
      "grad_norm": 0.5042742490768433,
      "learning_rate": 0.00010266079205596695,
      "loss": 0.1893,
      "step": 36780
    },
    {
      "epoch": 0.5171129383653102,
      "grad_norm": 0.3679254353046417,
      "learning_rate": 0.0001026143260243857,
      "loss": 0.1759,
      "step": 36790
    },
    {
      "epoch": 0.5172534963806311,
      "grad_norm": 0.3465472459793091,
      "learning_rate": 0.00010256785942795442,
      "loss": 0.1682,
      "step": 36800
    },
    {
      "epoch": 0.5173940543959519,
      "grad_norm": 0.39342233538627625,
      "learning_rate": 0.00010252139227671269,
      "loss": 0.1794,
      "step": 36810
    },
    {
      "epoch": 0.5175346124112727,
      "grad_norm": 0.4358043074607849,
      "learning_rate": 0.00010247492458070016,
      "loss": 0.1927,
      "step": 36820
    },
    {
      "epoch": 0.5176751704265936,
      "grad_norm": 0.3802052140235901,
      "learning_rate": 0.00010242845634995662,
      "loss": 0.1719,
      "step": 36830
    },
    {
      "epoch": 0.5178157284419144,
      "grad_norm": 0.4845336377620697,
      "learning_rate": 0.00010238198759452197,
      "loss": 0.1841,
      "step": 36840
    },
    {
      "epoch": 0.5179562864572352,
      "grad_norm": 0.5214284062385559,
      "learning_rate": 0.00010233551832443626,
      "loss": 0.1831,
      "step": 36850
    },
    {
      "epoch": 0.518096844472556,
      "grad_norm": 0.3969631493091583,
      "learning_rate": 0.00010228904854973957,
      "loss": 0.1836,
      "step": 36860
    },
    {
      "epoch": 0.5182374024878769,
      "grad_norm": 0.3800993859767914,
      "learning_rate": 0.00010224257828047219,
      "loss": 0.1623,
      "step": 36870
    },
    {
      "epoch": 0.5183779605031977,
      "grad_norm": 0.3893860876560211,
      "learning_rate": 0.0001021961075266744,
      "loss": 0.173,
      "step": 36880
    },
    {
      "epoch": 0.5185185185185185,
      "grad_norm": 0.4365367591381073,
      "learning_rate": 0.0001021496362983867,
      "loss": 0.1498,
      "step": 36890
    },
    {
      "epoch": 0.5186590765338394,
      "grad_norm": 0.44684091210365295,
      "learning_rate": 0.0001021031646056496,
      "loss": 0.1872,
      "step": 36900
    },
    {
      "epoch": 0.5187996345491601,
      "grad_norm": 0.3404183089733124,
      "learning_rate": 0.00010205669245850377,
      "loss": 0.1881,
      "step": 36910
    },
    {
      "epoch": 0.518940192564481,
      "grad_norm": 0.4390861988067627,
      "learning_rate": 0.00010201021986698999,
      "loss": 0.198,
      "step": 36920
    },
    {
      "epoch": 0.5190807505798019,
      "grad_norm": 0.6471704840660095,
      "learning_rate": 0.00010196374684114906,
      "loss": 0.1732,
      "step": 36930
    },
    {
      "epoch": 0.5192213085951226,
      "grad_norm": 0.330267995595932,
      "learning_rate": 0.00010191727339102197,
      "loss": 0.1813,
      "step": 36940
    },
    {
      "epoch": 0.5193618666104435,
      "grad_norm": 0.5875914692878723,
      "learning_rate": 0.00010187079952664968,
      "loss": 0.1752,
      "step": 36950
    },
    {
      "epoch": 0.5195024246257642,
      "grad_norm": 0.28342390060424805,
      "learning_rate": 0.00010182432525807337,
      "loss": 0.1759,
      "step": 36960
    },
    {
      "epoch": 0.5196429826410851,
      "grad_norm": 0.448543518781662,
      "learning_rate": 0.00010177785059533425,
      "loss": 0.1778,
      "step": 36970
    },
    {
      "epoch": 0.519783540656406,
      "grad_norm": 0.4849998950958252,
      "learning_rate": 0.00010173137554847356,
      "loss": 0.1858,
      "step": 36980
    },
    {
      "epoch": 0.5199240986717267,
      "grad_norm": 0.4241492450237274,
      "learning_rate": 0.00010168490012753272,
      "loss": 0.1918,
      "step": 36990
    },
    {
      "epoch": 0.5200646566870476,
      "grad_norm": 0.3861071765422821,
      "learning_rate": 0.00010163842434255318,
      "loss": 0.1836,
      "step": 37000
    },
    {
      "epoch": 0.5200646566870476,
      "eval_chrf": 80.22303827352079,
      "eval_loss": 0.38697686791419983,
      "eval_runtime": 326.8138,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 37000
    },
    {
      "epoch": 0.5202052147023684,
      "grad_norm": 0.49819403886795044,
      "learning_rate": 0.00010159194820357644,
      "loss": 0.1747,
      "step": 37010
    },
    {
      "epoch": 0.5203457727176892,
      "grad_norm": 0.34560924768447876,
      "learning_rate": 0.00010154547172064417,
      "loss": 0.1958,
      "step": 37020
    },
    {
      "epoch": 0.5204863307330101,
      "grad_norm": 0.373997300863266,
      "learning_rate": 0.00010149899490379801,
      "loss": 0.1941,
      "step": 37030
    },
    {
      "epoch": 0.5206268887483309,
      "grad_norm": 0.3896332085132599,
      "learning_rate": 0.00010145251776307976,
      "loss": 0.1825,
      "step": 37040
    },
    {
      "epoch": 0.5207674467636517,
      "grad_norm": 0.6348997354507446,
      "learning_rate": 0.00010140604030853121,
      "loss": 0.1778,
      "step": 37050
    },
    {
      "epoch": 0.5209080047789725,
      "grad_norm": 0.3897026777267456,
      "learning_rate": 0.00010135956255019427,
      "loss": 0.1704,
      "step": 37060
    },
    {
      "epoch": 0.5210485627942933,
      "grad_norm": 0.3688303232192993,
      "learning_rate": 0.00010131308449811093,
      "loss": 0.1983,
      "step": 37070
    },
    {
      "epoch": 0.5211891208096142,
      "grad_norm": 0.36505571007728577,
      "learning_rate": 0.00010126660616232315,
      "loss": 0.1873,
      "step": 37080
    },
    {
      "epoch": 0.521329678824935,
      "grad_norm": 0.3224770426750183,
      "learning_rate": 0.00010122012755287309,
      "loss": 0.1918,
      "step": 37090
    },
    {
      "epoch": 0.5214702368402558,
      "grad_norm": 0.4346891939640045,
      "learning_rate": 0.00010117364867980284,
      "loss": 0.191,
      "step": 37100
    },
    {
      "epoch": 0.5216107948555766,
      "grad_norm": 0.3521254062652588,
      "learning_rate": 0.00010112716955315465,
      "loss": 0.1638,
      "step": 37110
    },
    {
      "epoch": 0.5217513528708975,
      "grad_norm": 0.3422907292842865,
      "learning_rate": 0.00010108069018297073,
      "loss": 0.1667,
      "step": 37120
    },
    {
      "epoch": 0.5218919108862183,
      "grad_norm": 0.4173455834388733,
      "learning_rate": 0.0001010342105792934,
      "loss": 0.1714,
      "step": 37130
    },
    {
      "epoch": 0.5220324689015391,
      "grad_norm": 0.4705955386161804,
      "learning_rate": 0.00010098773075216505,
      "loss": 0.1802,
      "step": 37140
    },
    {
      "epoch": 0.52217302691686,
      "grad_norm": 0.5038504004478455,
      "learning_rate": 0.00010094125071162807,
      "loss": 0.2096,
      "step": 37150
    },
    {
      "epoch": 0.5223135849321807,
      "grad_norm": 0.4828181564807892,
      "learning_rate": 0.00010089477046772493,
      "loss": 0.1928,
      "step": 37160
    },
    {
      "epoch": 0.5224541429475016,
      "grad_norm": 0.5034831166267395,
      "learning_rate": 0.00010084829003049812,
      "loss": 0.1799,
      "step": 37170
    },
    {
      "epoch": 0.5225947009628225,
      "grad_norm": 0.40919286012649536,
      "learning_rate": 0.00010080180940999015,
      "loss": 0.1881,
      "step": 37180
    },
    {
      "epoch": 0.5227352589781432,
      "grad_norm": 0.45640450716018677,
      "learning_rate": 0.00010075532861624364,
      "loss": 0.168,
      "step": 37190
    },
    {
      "epoch": 0.5228758169934641,
      "grad_norm": 0.47407418489456177,
      "learning_rate": 0.0001007088476593012,
      "loss": 0.1887,
      "step": 37200
    },
    {
      "epoch": 0.5230163750087848,
      "grad_norm": 0.32496556639671326,
      "learning_rate": 0.00010066236654920546,
      "loss": 0.1781,
      "step": 37210
    },
    {
      "epoch": 0.5231569330241057,
      "grad_norm": 0.4471747577190399,
      "learning_rate": 0.00010061588529599915,
      "loss": 0.1884,
      "step": 37220
    },
    {
      "epoch": 0.5232974910394266,
      "grad_norm": 0.45092710852622986,
      "learning_rate": 0.00010056940390972495,
      "loss": 0.187,
      "step": 37230
    },
    {
      "epoch": 0.5234380490547473,
      "grad_norm": 0.5467546582221985,
      "learning_rate": 0.00010052292240042562,
      "loss": 0.1982,
      "step": 37240
    },
    {
      "epoch": 0.5235786070700682,
      "grad_norm": 0.3501761257648468,
      "learning_rate": 0.00010047644077814393,
      "loss": 0.1884,
      "step": 37250
    },
    {
      "epoch": 0.523719165085389,
      "grad_norm": 0.5036628842353821,
      "learning_rate": 0.00010042995905292267,
      "loss": 0.2039,
      "step": 37260
    },
    {
      "epoch": 0.5238597231007098,
      "grad_norm": 0.440433531999588,
      "learning_rate": 0.00010038347723480466,
      "loss": 0.1892,
      "step": 37270
    },
    {
      "epoch": 0.5240002811160306,
      "grad_norm": 0.4935232400894165,
      "learning_rate": 0.00010033699533383277,
      "loss": 0.1596,
      "step": 37280
    },
    {
      "epoch": 0.5241408391313515,
      "grad_norm": 0.38276660442352295,
      "learning_rate": 0.0001002905133600498,
      "loss": 0.163,
      "step": 37290
    },
    {
      "epoch": 0.5242813971466723,
      "grad_norm": 0.3699820339679718,
      "learning_rate": 0.00010024403132349867,
      "loss": 0.1836,
      "step": 37300
    },
    {
      "epoch": 0.5244219551619931,
      "grad_norm": 0.2769719064235687,
      "learning_rate": 0.00010019754923422225,
      "loss": 0.1545,
      "step": 37310
    },
    {
      "epoch": 0.524562513177314,
      "grad_norm": 0.3033786118030548,
      "learning_rate": 0.00010015106710226343,
      "loss": 0.1662,
      "step": 37320
    },
    {
      "epoch": 0.5247030711926347,
      "grad_norm": 0.49226874113082886,
      "learning_rate": 0.00010010458493766512,
      "loss": 0.2252,
      "step": 37330
    },
    {
      "epoch": 0.5248436292079556,
      "grad_norm": 0.3806842863559723,
      "learning_rate": 0.00010005810275047025,
      "loss": 0.1894,
      "step": 37340
    },
    {
      "epoch": 0.5249841872232764,
      "grad_norm": 0.38338780403137207,
      "learning_rate": 0.00010001162055072173,
      "loss": 0.1613,
      "step": 37350
    },
    {
      "epoch": 0.5251247452385972,
      "grad_norm": 0.36801716685295105,
      "learning_rate": 9.996513834846251e-05,
      "loss": 0.1724,
      "step": 37360
    },
    {
      "epoch": 0.5252653032539181,
      "grad_norm": 0.30908024311065674,
      "learning_rate": 9.991865615373545e-05,
      "loss": 0.1683,
      "step": 37370
    },
    {
      "epoch": 0.5254058612692388,
      "grad_norm": 0.4597463309764862,
      "learning_rate": 9.987217397658349e-05,
      "loss": 0.1851,
      "step": 37380
    },
    {
      "epoch": 0.5255464192845597,
      "grad_norm": 0.2701462209224701,
      "learning_rate": 9.982569182704959e-05,
      "loss": 0.1782,
      "step": 37390
    },
    {
      "epoch": 0.5256869772998806,
      "grad_norm": 0.3483943045139313,
      "learning_rate": 9.97792097151766e-05,
      "loss": 0.1924,
      "step": 37400
    },
    {
      "epoch": 0.5258275353152013,
      "grad_norm": 0.42426466941833496,
      "learning_rate": 9.973272765100747e-05,
      "loss": 0.185,
      "step": 37410
    },
    {
      "epoch": 0.5259680933305222,
      "grad_norm": 0.40943610668182373,
      "learning_rate": 9.968624564458506e-05,
      "loss": 0.1922,
      "step": 37420
    },
    {
      "epoch": 0.526108651345843,
      "grad_norm": 0.3415329158306122,
      "learning_rate": 9.963976370595226e-05,
      "loss": 0.1665,
      "step": 37430
    },
    {
      "epoch": 0.5262492093611638,
      "grad_norm": 0.4101947546005249,
      "learning_rate": 9.959328184515195e-05,
      "loss": 0.169,
      "step": 37440
    },
    {
      "epoch": 0.5263897673764847,
      "grad_norm": 0.33659279346466064,
      "learning_rate": 9.954680007222696e-05,
      "loss": 0.1846,
      "step": 37450
    },
    {
      "epoch": 0.5265303253918054,
      "grad_norm": 0.43769797682762146,
      "learning_rate": 9.950031839722014e-05,
      "loss": 0.1691,
      "step": 37460
    },
    {
      "epoch": 0.5266708834071263,
      "grad_norm": 0.34851711988449097,
      "learning_rate": 9.945383683017427e-05,
      "loss": 0.1611,
      "step": 37470
    },
    {
      "epoch": 0.5268114414224471,
      "grad_norm": 0.6938095688819885,
      "learning_rate": 9.940735538113216e-05,
      "loss": 0.2046,
      "step": 37480
    },
    {
      "epoch": 0.5269519994377679,
      "grad_norm": 0.4683268964290619,
      "learning_rate": 9.936087406013656e-05,
      "loss": 0.1661,
      "step": 37490
    },
    {
      "epoch": 0.5270925574530888,
      "grad_norm": 0.4278859496116638,
      "learning_rate": 9.93143928772302e-05,
      "loss": 0.1783,
      "step": 37500
    },
    {
      "epoch": 0.5272331154684096,
      "grad_norm": 0.42725682258605957,
      "learning_rate": 9.926791184245579e-05,
      "loss": 0.1731,
      "step": 37510
    },
    {
      "epoch": 0.5273736734837304,
      "grad_norm": 0.4099690914154053,
      "learning_rate": 9.922143096585598e-05,
      "loss": 0.1459,
      "step": 37520
    },
    {
      "epoch": 0.5275142314990512,
      "grad_norm": 0.4662937521934509,
      "learning_rate": 9.917495025747343e-05,
      "loss": 0.1759,
      "step": 37530
    },
    {
      "epoch": 0.5276547895143721,
      "grad_norm": 0.2619777321815491,
      "learning_rate": 9.91284697273507e-05,
      "loss": 0.1808,
      "step": 37540
    },
    {
      "epoch": 0.5277953475296929,
      "grad_norm": 0.4805005192756653,
      "learning_rate": 9.908198938553041e-05,
      "loss": 0.1749,
      "step": 37550
    },
    {
      "epoch": 0.5279359055450137,
      "grad_norm": 0.5037651062011719,
      "learning_rate": 9.903550924205504e-05,
      "loss": 0.1955,
      "step": 37560
    },
    {
      "epoch": 0.5280764635603346,
      "grad_norm": 0.41545727849006653,
      "learning_rate": 9.898902930696706e-05,
      "loss": 0.1804,
      "step": 37570
    },
    {
      "epoch": 0.5282170215756553,
      "grad_norm": 0.36044058203697205,
      "learning_rate": 9.894254959030895e-05,
      "loss": 0.1941,
      "step": 37580
    },
    {
      "epoch": 0.5283575795909762,
      "grad_norm": 0.44440537691116333,
      "learning_rate": 9.889607010212302e-05,
      "loss": 0.161,
      "step": 37590
    },
    {
      "epoch": 0.528498137606297,
      "grad_norm": 0.6077246069908142,
      "learning_rate": 9.884959085245168e-05,
      "loss": 0.183,
      "step": 37600
    },
    {
      "epoch": 0.5286386956216178,
      "grad_norm": 0.3023645877838135,
      "learning_rate": 9.880311185133719e-05,
      "loss": 0.1788,
      "step": 37610
    },
    {
      "epoch": 0.5287792536369387,
      "grad_norm": 0.3990058898925781,
      "learning_rate": 9.875663310882174e-05,
      "loss": 0.2035,
      "step": 37620
    },
    {
      "epoch": 0.5289198116522594,
      "grad_norm": 0.30556347966194153,
      "learning_rate": 9.871015463494757e-05,
      "loss": 0.1794,
      "step": 37630
    },
    {
      "epoch": 0.5290603696675803,
      "grad_norm": 0.3057769536972046,
      "learning_rate": 9.866367643975676e-05,
      "loss": 0.1723,
      "step": 37640
    },
    {
      "epoch": 0.5292009276829012,
      "grad_norm": 0.4252437651157379,
      "learning_rate": 9.861719853329135e-05,
      "loss": 0.182,
      "step": 37650
    },
    {
      "epoch": 0.5293414856982219,
      "grad_norm": 0.3990003764629364,
      "learning_rate": 9.857072092559338e-05,
      "loss": 0.1683,
      "step": 37660
    },
    {
      "epoch": 0.5294820437135428,
      "grad_norm": 0.444238543510437,
      "learning_rate": 9.852424362670474e-05,
      "loss": 0.1886,
      "step": 37670
    },
    {
      "epoch": 0.5296226017288636,
      "grad_norm": 0.37116536498069763,
      "learning_rate": 9.847776664666732e-05,
      "loss": 0.1867,
      "step": 37680
    },
    {
      "epoch": 0.5297631597441844,
      "grad_norm": 0.4291340112686157,
      "learning_rate": 9.84312899955229e-05,
      "loss": 0.182,
      "step": 37690
    },
    {
      "epoch": 0.5299037177595053,
      "grad_norm": 0.3295157551765442,
      "learning_rate": 9.838481368331318e-05,
      "loss": 0.1853,
      "step": 37700
    },
    {
      "epoch": 0.530044275774826,
      "grad_norm": 0.5010133385658264,
      "learning_rate": 9.833833772007985e-05,
      "loss": 0.173,
      "step": 37710
    },
    {
      "epoch": 0.5301848337901469,
      "grad_norm": 0.4085811376571655,
      "learning_rate": 9.829186211586447e-05,
      "loss": 0.1828,
      "step": 37720
    },
    {
      "epoch": 0.5303253918054677,
      "grad_norm": 0.2573600113391876,
      "learning_rate": 9.824538688070852e-05,
      "loss": 0.1752,
      "step": 37730
    },
    {
      "epoch": 0.5304659498207885,
      "grad_norm": 0.4763016104698181,
      "learning_rate": 9.819891202465344e-05,
      "loss": 0.1962,
      "step": 37740
    },
    {
      "epoch": 0.5306065078361094,
      "grad_norm": 0.47767651081085205,
      "learning_rate": 9.815243755774056e-05,
      "loss": 0.1811,
      "step": 37750
    },
    {
      "epoch": 0.5307470658514302,
      "grad_norm": 0.417579710483551,
      "learning_rate": 9.810596349001113e-05,
      "loss": 0.1826,
      "step": 37760
    },
    {
      "epoch": 0.530887623866751,
      "grad_norm": 0.3614961504936218,
      "learning_rate": 9.80594898315063e-05,
      "loss": 0.1622,
      "step": 37770
    },
    {
      "epoch": 0.5310281818820718,
      "grad_norm": 0.3221801817417145,
      "learning_rate": 9.801301659226716e-05,
      "loss": 0.1447,
      "step": 37780
    },
    {
      "epoch": 0.5311687398973927,
      "grad_norm": 0.4179418981075287,
      "learning_rate": 9.79665437823347e-05,
      "loss": 0.1655,
      "step": 37790
    },
    {
      "epoch": 0.5313092979127134,
      "grad_norm": 0.3602157235145569,
      "learning_rate": 9.792007141174979e-05,
      "loss": 0.1818,
      "step": 37800
    },
    {
      "epoch": 0.5314498559280343,
      "grad_norm": 0.30040112137794495,
      "learning_rate": 9.787359949055326e-05,
      "loss": 0.1932,
      "step": 37810
    },
    {
      "epoch": 0.5315904139433552,
      "grad_norm": 0.37365633249282837,
      "learning_rate": 9.782712802878577e-05,
      "loss": 0.1536,
      "step": 37820
    },
    {
      "epoch": 0.5317309719586759,
      "grad_norm": 0.4313228130340576,
      "learning_rate": 9.778065703648798e-05,
      "loss": 0.1934,
      "step": 37830
    },
    {
      "epoch": 0.5318715299739968,
      "grad_norm": 0.36040905117988586,
      "learning_rate": 9.773418652370033e-05,
      "loss": 0.2154,
      "step": 37840
    },
    {
      "epoch": 0.5320120879893175,
      "grad_norm": 0.4079858064651489,
      "learning_rate": 9.768771650046325e-05,
      "loss": 0.1841,
      "step": 37850
    },
    {
      "epoch": 0.5321526460046384,
      "grad_norm": 0.3194751739501953,
      "learning_rate": 9.764124697681702e-05,
      "loss": 0.1886,
      "step": 37860
    },
    {
      "epoch": 0.5322932040199593,
      "grad_norm": 0.2546814978122711,
      "learning_rate": 9.759477796280183e-05,
      "loss": 0.1865,
      "step": 37870
    },
    {
      "epoch": 0.53243376203528,
      "grad_norm": 0.4091358482837677,
      "learning_rate": 9.754830946845775e-05,
      "loss": 0.1794,
      "step": 37880
    },
    {
      "epoch": 0.5325743200506009,
      "grad_norm": 0.580310583114624,
      "learning_rate": 9.750184150382473e-05,
      "loss": 0.1813,
      "step": 37890
    },
    {
      "epoch": 0.5327148780659217,
      "grad_norm": 0.3971152901649475,
      "learning_rate": 9.745537407894263e-05,
      "loss": 0.1806,
      "step": 37900
    },
    {
      "epoch": 0.5328554360812425,
      "grad_norm": 0.34374722838401794,
      "learning_rate": 9.740890720385117e-05,
      "loss": 0.1755,
      "step": 37910
    },
    {
      "epoch": 0.5329959940965634,
      "grad_norm": 0.43604105710983276,
      "learning_rate": 9.736244088858997e-05,
      "loss": 0.1956,
      "step": 37920
    },
    {
      "epoch": 0.5331365521118842,
      "grad_norm": 0.3789943754673004,
      "learning_rate": 9.731597514319852e-05,
      "loss": 0.1944,
      "step": 37930
    },
    {
      "epoch": 0.533277110127205,
      "grad_norm": 0.26952505111694336,
      "learning_rate": 9.726950997771616e-05,
      "loss": 0.1775,
      "step": 37940
    },
    {
      "epoch": 0.5334176681425258,
      "grad_norm": 0.3544701039791107,
      "learning_rate": 9.722304540218216e-05,
      "loss": 0.1693,
      "step": 37950
    },
    {
      "epoch": 0.5335582261578466,
      "grad_norm": 0.34584537148475647,
      "learning_rate": 9.717658142663564e-05,
      "loss": 0.1771,
      "step": 37960
    },
    {
      "epoch": 0.5336987841731675,
      "grad_norm": 0.33332979679107666,
      "learning_rate": 9.713011806111554e-05,
      "loss": 0.1749,
      "step": 37970
    },
    {
      "epoch": 0.5338393421884883,
      "grad_norm": 0.3988155126571655,
      "learning_rate": 9.708365531566078e-05,
      "loss": 0.1798,
      "step": 37980
    },
    {
      "epoch": 0.5339799002038091,
      "grad_norm": 0.38039711117744446,
      "learning_rate": 9.703719320031001e-05,
      "loss": 0.1661,
      "step": 37990
    },
    {
      "epoch": 0.5341204582191299,
      "grad_norm": 0.3397109806537628,
      "learning_rate": 9.699073172510186e-05,
      "loss": 0.167,
      "step": 38000
    },
    {
      "epoch": 0.5341204582191299,
      "eval_chrf": 71.61878239659443,
      "eval_loss": 0.36813652515411377,
      "eval_runtime": 326.4744,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 38000
    },
    {
      "epoch": 0.5342610162344508,
      "grad_norm": 0.3394010663032532,
      "learning_rate": 9.694427090007474e-05,
      "loss": 0.1764,
      "step": 38010
    },
    {
      "epoch": 0.5344015742497716,
      "grad_norm": 0.24723602831363678,
      "learning_rate": 9.689781073526698e-05,
      "loss": 0.2023,
      "step": 38020
    },
    {
      "epoch": 0.5345421322650924,
      "grad_norm": 0.47091975808143616,
      "learning_rate": 9.685135124071675e-05,
      "loss": 0.1811,
      "step": 38030
    },
    {
      "epoch": 0.5346826902804133,
      "grad_norm": 0.5207058191299438,
      "learning_rate": 9.680489242646201e-05,
      "loss": 0.176,
      "step": 38040
    },
    {
      "epoch": 0.534823248295734,
      "grad_norm": 0.4883480966091156,
      "learning_rate": 9.675843430254068e-05,
      "loss": 0.1744,
      "step": 38050
    },
    {
      "epoch": 0.5349638063110549,
      "grad_norm": 0.450432151556015,
      "learning_rate": 9.671197687899046e-05,
      "loss": 0.1801,
      "step": 38060
    },
    {
      "epoch": 0.5351043643263758,
      "grad_norm": 0.3998246490955353,
      "learning_rate": 9.666552016584894e-05,
      "loss": 0.1863,
      "step": 38070
    },
    {
      "epoch": 0.5352449223416965,
      "grad_norm": 0.4299420416355133,
      "learning_rate": 9.661906417315349e-05,
      "loss": 0.1832,
      "step": 38080
    },
    {
      "epoch": 0.5353854803570174,
      "grad_norm": 0.3945608139038086,
      "learning_rate": 9.657260891094141e-05,
      "loss": 0.179,
      "step": 38090
    },
    {
      "epoch": 0.5355260383723381,
      "grad_norm": 0.409795880317688,
      "learning_rate": 9.652615438924979e-05,
      "loss": 0.1784,
      "step": 38100
    },
    {
      "epoch": 0.535666596387659,
      "grad_norm": 0.38981491327285767,
      "learning_rate": 9.647970061811556e-05,
      "loss": 0.1796,
      "step": 38110
    },
    {
      "epoch": 0.5358071544029799,
      "grad_norm": 0.47124508023262024,
      "learning_rate": 9.643324760757552e-05,
      "loss": 0.1847,
      "step": 38120
    },
    {
      "epoch": 0.5359477124183006,
      "grad_norm": 0.3736439049243927,
      "learning_rate": 9.638679536766627e-05,
      "loss": 0.19,
      "step": 38130
    },
    {
      "epoch": 0.5360882704336215,
      "grad_norm": 0.40727341175079346,
      "learning_rate": 9.634034390842426e-05,
      "loss": 0.193,
      "step": 38140
    },
    {
      "epoch": 0.5362288284489423,
      "grad_norm": 0.5049578547477722,
      "learning_rate": 9.629389323988578e-05,
      "loss": 0.1657,
      "step": 38150
    },
    {
      "epoch": 0.5363693864642631,
      "grad_norm": 0.4338602125644684,
      "learning_rate": 9.624744337208692e-05,
      "loss": 0.1883,
      "step": 38160
    },
    {
      "epoch": 0.536509944479584,
      "grad_norm": 0.3599756062030792,
      "learning_rate": 9.620099431506364e-05,
      "loss": 0.1853,
      "step": 38170
    },
    {
      "epoch": 0.5366505024949048,
      "grad_norm": 0.4722822308540344,
      "learning_rate": 9.615454607885168e-05,
      "loss": 0.1575,
      "step": 38180
    },
    {
      "epoch": 0.5367910605102256,
      "grad_norm": 0.5411520004272461,
      "learning_rate": 9.610809867348664e-05,
      "loss": 0.2003,
      "step": 38190
    },
    {
      "epoch": 0.5369316185255464,
      "grad_norm": 0.5211664438247681,
      "learning_rate": 9.606165210900391e-05,
      "loss": 0.1781,
      "step": 38200
    },
    {
      "epoch": 0.5370721765408673,
      "grad_norm": 0.4710027575492859,
      "learning_rate": 9.601520639543871e-05,
      "loss": 0.1752,
      "step": 38210
    },
    {
      "epoch": 0.5372127345561881,
      "grad_norm": 0.387617826461792,
      "learning_rate": 9.596876154282609e-05,
      "loss": 0.1564,
      "step": 38220
    },
    {
      "epoch": 0.5373532925715089,
      "grad_norm": 0.40515100955963135,
      "learning_rate": 9.592231756120091e-05,
      "loss": 0.1681,
      "step": 38230
    },
    {
      "epoch": 0.5374938505868297,
      "grad_norm": 0.49223166704177856,
      "learning_rate": 9.58758744605978e-05,
      "loss": 0.1587,
      "step": 38240
    },
    {
      "epoch": 0.5376344086021505,
      "grad_norm": 0.4113706946372986,
      "learning_rate": 9.582943225105127e-05,
      "loss": 0.1711,
      "step": 38250
    },
    {
      "epoch": 0.5377749666174714,
      "grad_norm": 0.429145485162735,
      "learning_rate": 9.57829909425956e-05,
      "loss": 0.1673,
      "step": 38260
    },
    {
      "epoch": 0.5379155246327921,
      "grad_norm": 0.5879194736480713,
      "learning_rate": 9.573655054526485e-05,
      "loss": 0.1806,
      "step": 38270
    },
    {
      "epoch": 0.538056082648113,
      "grad_norm": 0.37690460681915283,
      "learning_rate": 9.569011106909293e-05,
      "loss": 0.1814,
      "step": 38280
    },
    {
      "epoch": 0.5381966406634339,
      "grad_norm": 0.2968907356262207,
      "learning_rate": 9.564367252411351e-05,
      "loss": 0.1504,
      "step": 38290
    },
    {
      "epoch": 0.5383371986787546,
      "grad_norm": 0.44435423612594604,
      "learning_rate": 9.559723492036011e-05,
      "loss": 0.185,
      "step": 38300
    },
    {
      "epoch": 0.5384777566940755,
      "grad_norm": 0.3620116412639618,
      "learning_rate": 9.5550798267866e-05,
      "loss": 0.1562,
      "step": 38310
    },
    {
      "epoch": 0.5386183147093963,
      "grad_norm": 0.42804157733917236,
      "learning_rate": 9.550436257666425e-05,
      "loss": 0.1995,
      "step": 38320
    },
    {
      "epoch": 0.5387588727247171,
      "grad_norm": 0.4588015675544739,
      "learning_rate": 9.545792785678776e-05,
      "loss": 0.166,
      "step": 38330
    },
    {
      "epoch": 0.538899430740038,
      "grad_norm": 0.44554048776626587,
      "learning_rate": 9.541149411826917e-05,
      "loss": 0.1716,
      "step": 38340
    },
    {
      "epoch": 0.5390399887553587,
      "grad_norm": 0.5267353057861328,
      "learning_rate": 9.536506137114092e-05,
      "loss": 0.1567,
      "step": 38350
    },
    {
      "epoch": 0.5391805467706796,
      "grad_norm": 0.4399483799934387,
      "learning_rate": 9.531862962543528e-05,
      "loss": 0.1775,
      "step": 38360
    },
    {
      "epoch": 0.5393211047860004,
      "grad_norm": 0.4195297658443451,
      "learning_rate": 9.527219889118426e-05,
      "loss": 0.161,
      "step": 38370
    },
    {
      "epoch": 0.5394616628013212,
      "grad_norm": 0.4252000153064728,
      "learning_rate": 9.522576917841966e-05,
      "loss": 0.1813,
      "step": 38380
    },
    {
      "epoch": 0.5396022208166421,
      "grad_norm": 0.3439667224884033,
      "learning_rate": 9.517934049717304e-05,
      "loss": 0.1732,
      "step": 38390
    },
    {
      "epoch": 0.5397427788319629,
      "grad_norm": 0.4506164789199829,
      "learning_rate": 9.51329128574758e-05,
      "loss": 0.193,
      "step": 38400
    },
    {
      "epoch": 0.5398833368472837,
      "grad_norm": 0.54718416929245,
      "learning_rate": 9.508648626935903e-05,
      "loss": 0.2013,
      "step": 38410
    },
    {
      "epoch": 0.5400238948626045,
      "grad_norm": 0.4310196340084076,
      "learning_rate": 9.504006074285367e-05,
      "loss": 0.1659,
      "step": 38420
    },
    {
      "epoch": 0.5401644528779254,
      "grad_norm": 0.380362868309021,
      "learning_rate": 9.499363628799039e-05,
      "loss": 0.1869,
      "step": 38430
    },
    {
      "epoch": 0.5403050108932462,
      "grad_norm": 0.32745522260665894,
      "learning_rate": 9.494721291479961e-05,
      "loss": 0.1787,
      "step": 38440
    },
    {
      "epoch": 0.540445568908567,
      "grad_norm": 0.43625086545944214,
      "learning_rate": 9.490079063331156e-05,
      "loss": 0.1669,
      "step": 38450
    },
    {
      "epoch": 0.5405861269238879,
      "grad_norm": 0.4483635723590851,
      "learning_rate": 9.485436945355623e-05,
      "loss": 0.1531,
      "step": 38460
    },
    {
      "epoch": 0.5407266849392086,
      "grad_norm": 0.45638909935951233,
      "learning_rate": 9.480794938556334e-05,
      "loss": 0.1668,
      "step": 38470
    },
    {
      "epoch": 0.5408672429545295,
      "grad_norm": 0.3815574645996094,
      "learning_rate": 9.476153043936239e-05,
      "loss": 0.1885,
      "step": 38480
    },
    {
      "epoch": 0.5410078009698503,
      "grad_norm": 0.3492136299610138,
      "learning_rate": 9.471511262498263e-05,
      "loss": 0.1687,
      "step": 38490
    },
    {
      "epoch": 0.5411483589851711,
      "grad_norm": 0.3178686499595642,
      "learning_rate": 9.466869595245308e-05,
      "loss": 0.1733,
      "step": 38500
    },
    {
      "epoch": 0.541288917000492,
      "grad_norm": 0.4572628140449524,
      "learning_rate": 9.46222804318025e-05,
      "loss": 0.1669,
      "step": 38510
    },
    {
      "epoch": 0.5414294750158127,
      "grad_norm": 0.34754234552383423,
      "learning_rate": 9.45758660730594e-05,
      "loss": 0.185,
      "step": 38520
    },
    {
      "epoch": 0.5415700330311336,
      "grad_norm": 0.43663638830184937,
      "learning_rate": 9.452945288625206e-05,
      "loss": 0.1672,
      "step": 38530
    },
    {
      "epoch": 0.5417105910464545,
      "grad_norm": 0.3496999740600586,
      "learning_rate": 9.448304088140847e-05,
      "loss": 0.1825,
      "step": 38540
    },
    {
      "epoch": 0.5418511490617752,
      "grad_norm": 0.40601423382759094,
      "learning_rate": 9.44366300685564e-05,
      "loss": 0.1899,
      "step": 38550
    },
    {
      "epoch": 0.5419917070770961,
      "grad_norm": 0.4218822419643402,
      "learning_rate": 9.439022045772335e-05,
      "loss": 0.1852,
      "step": 38560
    },
    {
      "epoch": 0.5421322650924169,
      "grad_norm": 0.3982556462287903,
      "learning_rate": 9.434381205893653e-05,
      "loss": 0.169,
      "step": 38570
    },
    {
      "epoch": 0.5422728231077377,
      "grad_norm": 0.5806414484977722,
      "learning_rate": 9.429740488222295e-05,
      "loss": 0.1743,
      "step": 38580
    },
    {
      "epoch": 0.5424133811230586,
      "grad_norm": 0.4485076069831848,
      "learning_rate": 9.425099893760931e-05,
      "loss": 0.1708,
      "step": 38590
    },
    {
      "epoch": 0.5425539391383793,
      "grad_norm": 0.39736509323120117,
      "learning_rate": 9.420459423512204e-05,
      "loss": 0.163,
      "step": 38600
    },
    {
      "epoch": 0.5426944971537002,
      "grad_norm": 0.34437084197998047,
      "learning_rate": 9.415819078478736e-05,
      "loss": 0.1705,
      "step": 38610
    },
    {
      "epoch": 0.542835055169021,
      "grad_norm": 0.3532562553882599,
      "learning_rate": 9.411178859663112e-05,
      "loss": 0.1522,
      "step": 38620
    },
    {
      "epoch": 0.5429756131843418,
      "grad_norm": 0.38678795099258423,
      "learning_rate": 9.406538768067901e-05,
      "loss": 0.188,
      "step": 38630
    },
    {
      "epoch": 0.5431161711996627,
      "grad_norm": 0.456844687461853,
      "learning_rate": 9.401898804695635e-05,
      "loss": 0.1839,
      "step": 38640
    },
    {
      "epoch": 0.5432567292149835,
      "grad_norm": 0.3993634581565857,
      "learning_rate": 9.397258970548823e-05,
      "loss": 0.1612,
      "step": 38650
    },
    {
      "epoch": 0.5433972872303043,
      "grad_norm": 0.42853468656539917,
      "learning_rate": 9.392619266629947e-05,
      "loss": 0.2004,
      "step": 38660
    },
    {
      "epoch": 0.5435378452456251,
      "grad_norm": 0.4220224618911743,
      "learning_rate": 9.387979693941457e-05,
      "loss": 0.1661,
      "step": 38670
    },
    {
      "epoch": 0.543678403260946,
      "grad_norm": 0.3359472453594208,
      "learning_rate": 9.383340253485779e-05,
      "loss": 0.1551,
      "step": 38680
    },
    {
      "epoch": 0.5438189612762668,
      "grad_norm": 0.4105564057826996,
      "learning_rate": 9.378700946265305e-05,
      "loss": 0.1752,
      "step": 38690
    },
    {
      "epoch": 0.5439595192915876,
      "grad_norm": 0.4355560541152954,
      "learning_rate": 9.37406177328241e-05,
      "loss": 0.1838,
      "step": 38700
    },
    {
      "epoch": 0.5441000773069085,
      "grad_norm": 0.5953741669654846,
      "learning_rate": 9.369422735539421e-05,
      "loss": 0.207,
      "step": 38710
    },
    {
      "epoch": 0.5442406353222292,
      "grad_norm": 0.42537426948547363,
      "learning_rate": 9.364783834038648e-05,
      "loss": 0.1863,
      "step": 38720
    },
    {
      "epoch": 0.5443811933375501,
      "grad_norm": 0.36153459548950195,
      "learning_rate": 9.360145069782373e-05,
      "loss": 0.1701,
      "step": 38730
    },
    {
      "epoch": 0.5445217513528708,
      "grad_norm": 0.36473673582077026,
      "learning_rate": 9.355506443772846e-05,
      "loss": 0.1773,
      "step": 38740
    },
    {
      "epoch": 0.5446623093681917,
      "grad_norm": 0.32480740547180176,
      "learning_rate": 9.350867957012283e-05,
      "loss": 0.1631,
      "step": 38750
    },
    {
      "epoch": 0.5448028673835126,
      "grad_norm": 0.6103132367134094,
      "learning_rate": 9.346229610502874e-05,
      "loss": 0.1778,
      "step": 38760
    },
    {
      "epoch": 0.5449434253988333,
      "grad_norm": 0.4113089144229889,
      "learning_rate": 9.341591405246782e-05,
      "loss": 0.1755,
      "step": 38770
    },
    {
      "epoch": 0.5450839834141542,
      "grad_norm": 0.3067460358142853,
      "learning_rate": 9.336953342246129e-05,
      "loss": 0.174,
      "step": 38780
    },
    {
      "epoch": 0.545224541429475,
      "grad_norm": 0.6269018650054932,
      "learning_rate": 9.332315422503017e-05,
      "loss": 0.197,
      "step": 38790
    },
    {
      "epoch": 0.5453650994447958,
      "grad_norm": 0.38305601477622986,
      "learning_rate": 9.32767764701951e-05,
      "loss": 0.1696,
      "step": 38800
    },
    {
      "epoch": 0.5455056574601167,
      "grad_norm": 0.3979334831237793,
      "learning_rate": 9.323040016797646e-05,
      "loss": 0.1744,
      "step": 38810
    },
    {
      "epoch": 0.5456462154754375,
      "grad_norm": 0.46121445298194885,
      "learning_rate": 9.318402532839427e-05,
      "loss": 0.1849,
      "step": 38820
    },
    {
      "epoch": 0.5457867734907583,
      "grad_norm": 0.40946972370147705,
      "learning_rate": 9.313765196146829e-05,
      "loss": 0.1805,
      "step": 38830
    },
    {
      "epoch": 0.5459273315060791,
      "grad_norm": 0.522844135761261,
      "learning_rate": 9.309128007721788e-05,
      "loss": 0.179,
      "step": 38840
    },
    {
      "epoch": 0.5460678895214,
      "grad_norm": 0.3552819788455963,
      "learning_rate": 9.304490968566216e-05,
      "loss": 0.1784,
      "step": 38850
    },
    {
      "epoch": 0.5462084475367208,
      "grad_norm": 0.34201109409332275,
      "learning_rate": 9.299854079681989e-05,
      "loss": 0.1754,
      "step": 38860
    },
    {
      "epoch": 0.5463490055520416,
      "grad_norm": 0.48454296588897705,
      "learning_rate": 9.295217342070949e-05,
      "loss": 0.1848,
      "step": 38870
    },
    {
      "epoch": 0.5464895635673624,
      "grad_norm": 0.3617215156555176,
      "learning_rate": 9.290580756734909e-05,
      "loss": 0.1616,
      "step": 38880
    },
    {
      "epoch": 0.5466301215826832,
      "grad_norm": 0.45073843002319336,
      "learning_rate": 9.285944324675648e-05,
      "loss": 0.2035,
      "step": 38890
    },
    {
      "epoch": 0.5467706795980041,
      "grad_norm": 0.37417346239089966,
      "learning_rate": 9.28130804689491e-05,
      "loss": 0.1646,
      "step": 38900
    },
    {
      "epoch": 0.5469112376133249,
      "grad_norm": 0.4643031656742096,
      "learning_rate": 9.276671924394407e-05,
      "loss": 0.1408,
      "step": 38910
    },
    {
      "epoch": 0.5470517956286457,
      "grad_norm": 0.34645259380340576,
      "learning_rate": 9.272035958175816e-05,
      "loss": 0.1635,
      "step": 38920
    },
    {
      "epoch": 0.5471923536439666,
      "grad_norm": 0.39230334758758545,
      "learning_rate": 9.267400149240786e-05,
      "loss": 0.1831,
      "step": 38930
    },
    {
      "epoch": 0.5473329116592873,
      "grad_norm": 0.3278043270111084,
      "learning_rate": 9.262764498590924e-05,
      "loss": 0.17,
      "step": 38940
    },
    {
      "epoch": 0.5474734696746082,
      "grad_norm": 0.3736383318901062,
      "learning_rate": 9.258129007227807e-05,
      "loss": 0.1762,
      "step": 38950
    },
    {
      "epoch": 0.5476140276899291,
      "grad_norm": 0.3995749056339264,
      "learning_rate": 9.253493676152976e-05,
      "loss": 0.1861,
      "step": 38960
    },
    {
      "epoch": 0.5477545857052498,
      "grad_norm": 0.3844054341316223,
      "learning_rate": 9.248858506367943e-05,
      "loss": 0.177,
      "step": 38970
    },
    {
      "epoch": 0.5478951437205707,
      "grad_norm": 0.3187495470046997,
      "learning_rate": 9.244223498874174e-05,
      "loss": 0.1734,
      "step": 38980
    },
    {
      "epoch": 0.5480357017358914,
      "grad_norm": 0.3380083441734314,
      "learning_rate": 9.239588654673111e-05,
      "loss": 0.1782,
      "step": 38990
    },
    {
      "epoch": 0.5481762597512123,
      "grad_norm": 0.3558598458766937,
      "learning_rate": 9.234953974766156e-05,
      "loss": 0.1777,
      "step": 39000
    },
    {
      "epoch": 0.5481762597512123,
      "eval_chrf": 84.03459253857896,
      "eval_loss": 0.3780422806739807,
      "eval_runtime": 173.4269,
      "eval_samples_per_second": 0.577,
      "eval_steps_per_second": 0.023,
      "step": 39000
    },
    {
      "epoch": 0.5483168177665332,
      "grad_norm": 0.3646630048751831,
      "learning_rate": 9.230319460154672e-05,
      "loss": 0.1859,
      "step": 39010
    },
    {
      "epoch": 0.5484573757818539,
      "grad_norm": 0.3137366473674774,
      "learning_rate": 9.225685111839995e-05,
      "loss": 0.1758,
      "step": 39020
    },
    {
      "epoch": 0.5485979337971748,
      "grad_norm": 0.4131186008453369,
      "learning_rate": 9.221050930823418e-05,
      "loss": 0.1904,
      "step": 39030
    },
    {
      "epoch": 0.5487384918124956,
      "grad_norm": 0.42204728722572327,
      "learning_rate": 9.216416918106198e-05,
      "loss": 0.1864,
      "step": 39040
    },
    {
      "epoch": 0.5488790498278164,
      "grad_norm": 0.44460195302963257,
      "learning_rate": 9.211783074689559e-05,
      "loss": 0.185,
      "step": 39050
    },
    {
      "epoch": 0.5490196078431373,
      "grad_norm": 0.3441852331161499,
      "learning_rate": 9.207149401574688e-05,
      "loss": 0.175,
      "step": 39060
    },
    {
      "epoch": 0.5491601658584581,
      "grad_norm": 0.2591150104999542,
      "learning_rate": 9.202515899762733e-05,
      "loss": 0.1828,
      "step": 39070
    },
    {
      "epoch": 0.5493007238737789,
      "grad_norm": 0.47425776720046997,
      "learning_rate": 9.197882570254806e-05,
      "loss": 0.1655,
      "step": 39080
    },
    {
      "epoch": 0.5494412818890997,
      "grad_norm": 0.4549548029899597,
      "learning_rate": 9.193249414051983e-05,
      "loss": 0.1806,
      "step": 39090
    },
    {
      "epoch": 0.5495818399044206,
      "grad_norm": 0.3400246500968933,
      "learning_rate": 9.188616432155301e-05,
      "loss": 0.145,
      "step": 39100
    },
    {
      "epoch": 0.5497223979197414,
      "grad_norm": 0.39775407314300537,
      "learning_rate": 9.183983625565759e-05,
      "loss": 0.1753,
      "step": 39110
    },
    {
      "epoch": 0.5498629559350622,
      "grad_norm": 0.32858631014823914,
      "learning_rate": 9.17935099528432e-05,
      "loss": 0.1635,
      "step": 39120
    },
    {
      "epoch": 0.550003513950383,
      "grad_norm": 0.4243503212928772,
      "learning_rate": 9.174718542311907e-05,
      "loss": 0.1822,
      "step": 39130
    },
    {
      "epoch": 0.5501440719657038,
      "grad_norm": 0.421855092048645,
      "learning_rate": 9.170086267649406e-05,
      "loss": 0.1888,
      "step": 39140
    },
    {
      "epoch": 0.5502846299810247,
      "grad_norm": 0.38638654351234436,
      "learning_rate": 9.165454172297664e-05,
      "loss": 0.1737,
      "step": 39150
    },
    {
      "epoch": 0.5504251879963455,
      "grad_norm": 0.43157637119293213,
      "learning_rate": 9.16082225725749e-05,
      "loss": 0.1757,
      "step": 39160
    },
    {
      "epoch": 0.5505657460116663,
      "grad_norm": 0.3671276271343231,
      "learning_rate": 9.156190523529651e-05,
      "loss": 0.1755,
      "step": 39170
    },
    {
      "epoch": 0.5507063040269872,
      "grad_norm": 0.3911401033401489,
      "learning_rate": 9.151558972114879e-05,
      "loss": 0.1654,
      "step": 39180
    },
    {
      "epoch": 0.5508468620423079,
      "grad_norm": 0.3790930509567261,
      "learning_rate": 9.146927604013866e-05,
      "loss": 0.1944,
      "step": 39190
    },
    {
      "epoch": 0.5509874200576288,
      "grad_norm": 0.41435009241104126,
      "learning_rate": 9.142296420227261e-05,
      "loss": 0.1716,
      "step": 39200
    },
    {
      "epoch": 0.5511279780729497,
      "grad_norm": 0.35916468501091003,
      "learning_rate": 9.137665421755677e-05,
      "loss": 0.1821,
      "step": 39210
    },
    {
      "epoch": 0.5512685360882704,
      "grad_norm": 0.3221915066242218,
      "learning_rate": 9.133034609599683e-05,
      "loss": 0.1985,
      "step": 39220
    },
    {
      "epoch": 0.5514090941035913,
      "grad_norm": 0.417850524187088,
      "learning_rate": 9.128403984759811e-05,
      "loss": 0.2018,
      "step": 39230
    },
    {
      "epoch": 0.551549652118912,
      "grad_norm": 0.46724241971969604,
      "learning_rate": 9.123773548236553e-05,
      "loss": 0.172,
      "step": 39240
    },
    {
      "epoch": 0.5516902101342329,
      "grad_norm": 0.3086174726486206,
      "learning_rate": 9.119143301030356e-05,
      "loss": 0.1723,
      "step": 39250
    },
    {
      "epoch": 0.5518307681495537,
      "grad_norm": 0.3772790729999542,
      "learning_rate": 9.114513244141632e-05,
      "loss": 0.1934,
      "step": 39260
    },
    {
      "epoch": 0.5519713261648745,
      "grad_norm": 0.2803363502025604,
      "learning_rate": 9.109883378570747e-05,
      "loss": 0.1863,
      "step": 39270
    },
    {
      "epoch": 0.5521118841801954,
      "grad_norm": 0.3337017595767975,
      "learning_rate": 9.105253705318028e-05,
      "loss": 0.1671,
      "step": 39280
    },
    {
      "epoch": 0.5522524421955162,
      "grad_norm": 0.3834514617919922,
      "learning_rate": 9.100624225383761e-05,
      "loss": 0.177,
      "step": 39290
    },
    {
      "epoch": 0.552393000210837,
      "grad_norm": 0.37610965967178345,
      "learning_rate": 9.095994939768187e-05,
      "loss": 0.1801,
      "step": 39300
    },
    {
      "epoch": 0.5525335582261578,
      "grad_norm": 0.3597879111766815,
      "learning_rate": 9.09136584947151e-05,
      "loss": 0.1818,
      "step": 39310
    },
    {
      "epoch": 0.5526741162414787,
      "grad_norm": 0.499024897813797,
      "learning_rate": 9.086736955493887e-05,
      "loss": 0.2017,
      "step": 39320
    },
    {
      "epoch": 0.5528146742567995,
      "grad_norm": 0.2908949553966522,
      "learning_rate": 9.082108258835433e-05,
      "loss": 0.158,
      "step": 39330
    },
    {
      "epoch": 0.5529552322721203,
      "grad_norm": 0.31243860721588135,
      "learning_rate": 9.077479760496226e-05,
      "loss": 0.1815,
      "step": 39340
    },
    {
      "epoch": 0.5530957902874412,
      "grad_norm": 0.44429317116737366,
      "learning_rate": 9.072851461476296e-05,
      "loss": 0.2251,
      "step": 39350
    },
    {
      "epoch": 0.5532363483027619,
      "grad_norm": 0.373526006937027,
      "learning_rate": 9.068223362775627e-05,
      "loss": 0.1629,
      "step": 39360
    },
    {
      "epoch": 0.5533769063180828,
      "grad_norm": 0.3967413604259491,
      "learning_rate": 9.063595465394172e-05,
      "loss": 0.1601,
      "step": 39370
    },
    {
      "epoch": 0.5535174643334037,
      "grad_norm": 0.41019967198371887,
      "learning_rate": 9.058967770331822e-05,
      "loss": 0.1644,
      "step": 39380
    },
    {
      "epoch": 0.5536580223487244,
      "grad_norm": 0.4954618811607361,
      "learning_rate": 9.054340278588443e-05,
      "loss": 0.1681,
      "step": 39390
    },
    {
      "epoch": 0.5537985803640453,
      "grad_norm": 0.44948747754096985,
      "learning_rate": 9.049712991163842e-05,
      "loss": 0.1729,
      "step": 39400
    },
    {
      "epoch": 0.553939138379366,
      "grad_norm": 0.37606337666511536,
      "learning_rate": 9.045085909057795e-05,
      "loss": 0.179,
      "step": 39410
    },
    {
      "epoch": 0.5540796963946869,
      "grad_norm": 0.3375130891799927,
      "learning_rate": 9.040459033270021e-05,
      "loss": 0.2018,
      "step": 39420
    },
    {
      "epoch": 0.5542202544100078,
      "grad_norm": 0.4176686704158783,
      "learning_rate": 9.035832364800205e-05,
      "loss": 0.1786,
      "step": 39430
    },
    {
      "epoch": 0.5543608124253285,
      "grad_norm": 0.33619213104248047,
      "learning_rate": 9.031205904647981e-05,
      "loss": 0.1892,
      "step": 39440
    },
    {
      "epoch": 0.5545013704406494,
      "grad_norm": 0.3469635248184204,
      "learning_rate": 9.02657965381294e-05,
      "loss": 0.1484,
      "step": 39450
    },
    {
      "epoch": 0.5546419284559702,
      "grad_norm": 0.4018300771713257,
      "learning_rate": 9.021953613294628e-05,
      "loss": 0.1497,
      "step": 39460
    },
    {
      "epoch": 0.554782486471291,
      "grad_norm": 0.42926666140556335,
      "learning_rate": 9.017327784092544e-05,
      "loss": 0.1942,
      "step": 39470
    },
    {
      "epoch": 0.5549230444866119,
      "grad_norm": 0.37829723954200745,
      "learning_rate": 9.012702167206144e-05,
      "loss": 0.1916,
      "step": 39480
    },
    {
      "epoch": 0.5550636025019327,
      "grad_norm": 0.37087535858154297,
      "learning_rate": 9.008076763634834e-05,
      "loss": 0.1876,
      "step": 39490
    },
    {
      "epoch": 0.5552041605172535,
      "grad_norm": 0.3491499722003937,
      "learning_rate": 9.00345157437798e-05,
      "loss": 0.1668,
      "step": 39500
    },
    {
      "epoch": 0.5553447185325743,
      "grad_norm": 0.37351173162460327,
      "learning_rate": 8.998826600434893e-05,
      "loss": 0.183,
      "step": 39510
    },
    {
      "epoch": 0.5554852765478951,
      "grad_norm": 0.469526082277298,
      "learning_rate": 8.994201842804848e-05,
      "loss": 0.207,
      "step": 39520
    },
    {
      "epoch": 0.555625834563216,
      "grad_norm": 0.4910909831523895,
      "learning_rate": 8.989577302487066e-05,
      "loss": 0.1552,
      "step": 39530
    },
    {
      "epoch": 0.5557663925785368,
      "grad_norm": 0.42745739221572876,
      "learning_rate": 8.98495298048072e-05,
      "loss": 0.1805,
      "step": 39540
    },
    {
      "epoch": 0.5559069505938576,
      "grad_norm": 0.493329256772995,
      "learning_rate": 8.980328877784943e-05,
      "loss": 0.1839,
      "step": 39550
    },
    {
      "epoch": 0.5560475086091784,
      "grad_norm": 0.48328346014022827,
      "learning_rate": 8.975704995398816e-05,
      "loss": 0.1947,
      "step": 39560
    },
    {
      "epoch": 0.5561880666244993,
      "grad_norm": 0.4481086730957031,
      "learning_rate": 8.97108133432137e-05,
      "loss": 0.1766,
      "step": 39570
    },
    {
      "epoch": 0.5563286246398201,
      "grad_norm": 0.38591402769088745,
      "learning_rate": 8.966457895551592e-05,
      "loss": 0.1893,
      "step": 39580
    },
    {
      "epoch": 0.5564691826551409,
      "grad_norm": 0.43421241641044617,
      "learning_rate": 8.961834680088423e-05,
      "loss": 0.1892,
      "step": 39590
    },
    {
      "epoch": 0.5566097406704618,
      "grad_norm": 0.371267169713974,
      "learning_rate": 8.957211688930748e-05,
      "loss": 0.1862,
      "step": 39600
    },
    {
      "epoch": 0.5567502986857825,
      "grad_norm": 0.4000300467014313,
      "learning_rate": 8.952588923077409e-05,
      "loss": 0.1742,
      "step": 39610
    },
    {
      "epoch": 0.5568908567011034,
      "grad_norm": 0.47955048084259033,
      "learning_rate": 8.947966383527203e-05,
      "loss": 0.1993,
      "step": 39620
    },
    {
      "epoch": 0.5570314147164243,
      "grad_norm": 0.3886945843696594,
      "learning_rate": 8.943344071278869e-05,
      "loss": 0.1999,
      "step": 39630
    },
    {
      "epoch": 0.557171972731745,
      "grad_norm": 0.3787805438041687,
      "learning_rate": 8.938721987331104e-05,
      "loss": 0.1777,
      "step": 39640
    },
    {
      "epoch": 0.5573125307470659,
      "grad_norm": 0.36745595932006836,
      "learning_rate": 8.934100132682551e-05,
      "loss": 0.1904,
      "step": 39650
    },
    {
      "epoch": 0.5574530887623866,
      "grad_norm": 0.3356489837169647,
      "learning_rate": 8.929478508331807e-05,
      "loss": 0.164,
      "step": 39660
    },
    {
      "epoch": 0.5575936467777075,
      "grad_norm": 0.4322313666343689,
      "learning_rate": 8.92485711527742e-05,
      "loss": 0.1652,
      "step": 39670
    },
    {
      "epoch": 0.5577342047930284,
      "grad_norm": 0.4178856611251831,
      "learning_rate": 8.920235954517882e-05,
      "loss": 0.194,
      "step": 39680
    },
    {
      "epoch": 0.5578747628083491,
      "grad_norm": 0.3862941563129425,
      "learning_rate": 8.915615027051642e-05,
      "loss": 0.1811,
      "step": 39690
    },
    {
      "epoch": 0.55801532082367,
      "grad_norm": 0.4267183840274811,
      "learning_rate": 8.910994333877091e-05,
      "loss": 0.2013,
      "step": 39700
    },
    {
      "epoch": 0.5581558788389908,
      "grad_norm": 0.32371628284454346,
      "learning_rate": 8.90637387599258e-05,
      "loss": 0.2189,
      "step": 39710
    },
    {
      "epoch": 0.5582964368543116,
      "grad_norm": 0.3845933973789215,
      "learning_rate": 8.901753654396398e-05,
      "loss": 0.1729,
      "step": 39720
    },
    {
      "epoch": 0.5584369948696324,
      "grad_norm": 0.34567537903785706,
      "learning_rate": 8.89713367008679e-05,
      "loss": 0.19,
      "step": 39730
    },
    {
      "epoch": 0.5585775528849533,
      "grad_norm": 0.34996098279953003,
      "learning_rate": 8.892513924061948e-05,
      "loss": 0.177,
      "step": 39740
    },
    {
      "epoch": 0.5587181109002741,
      "grad_norm": 0.44245201349258423,
      "learning_rate": 8.88789441732001e-05,
      "loss": 0.181,
      "step": 39750
    },
    {
      "epoch": 0.5588586689155949,
      "grad_norm": 0.45229363441467285,
      "learning_rate": 8.883275150859065e-05,
      "loss": 0.1667,
      "step": 39760
    },
    {
      "epoch": 0.5589992269309157,
      "grad_norm": 0.41712188720703125,
      "learning_rate": 8.87865612567715e-05,
      "loss": 0.1905,
      "step": 39770
    },
    {
      "epoch": 0.5591397849462365,
      "grad_norm": 0.3251127600669861,
      "learning_rate": 8.87403734277225e-05,
      "loss": 0.1748,
      "step": 39780
    },
    {
      "epoch": 0.5592803429615574,
      "grad_norm": 0.4091569781303406,
      "learning_rate": 8.869418803142297e-05,
      "loss": 0.2082,
      "step": 39790
    },
    {
      "epoch": 0.5594209009768782,
      "grad_norm": 0.5211314558982849,
      "learning_rate": 8.864800507785168e-05,
      "loss": 0.1671,
      "step": 39800
    },
    {
      "epoch": 0.559561458992199,
      "grad_norm": 0.5784714818000793,
      "learning_rate": 8.860182457698692e-05,
      "loss": 0.1808,
      "step": 39810
    },
    {
      "epoch": 0.5597020170075199,
      "grad_norm": 0.31065332889556885,
      "learning_rate": 8.855564653880644e-05,
      "loss": 0.1769,
      "step": 39820
    },
    {
      "epoch": 0.5598425750228406,
      "grad_norm": 0.39571863412857056,
      "learning_rate": 8.85094709732874e-05,
      "loss": 0.2061,
      "step": 39830
    },
    {
      "epoch": 0.5599831330381615,
      "grad_norm": 0.5271860361099243,
      "learning_rate": 8.84632978904065e-05,
      "loss": 0.1812,
      "step": 39840
    },
    {
      "epoch": 0.5601236910534824,
      "grad_norm": 0.3915952742099762,
      "learning_rate": 8.841712730013988e-05,
      "loss": 0.1686,
      "step": 39850
    },
    {
      "epoch": 0.5602642490688031,
      "grad_norm": 0.35233601927757263,
      "learning_rate": 8.837095921246312e-05,
      "loss": 0.1561,
      "step": 39860
    },
    {
      "epoch": 0.560404807084124,
      "grad_norm": 0.36162757873535156,
      "learning_rate": 8.832479363735128e-05,
      "loss": 0.1867,
      "step": 39870
    },
    {
      "epoch": 0.5605453650994447,
      "grad_norm": 0.35949790477752686,
      "learning_rate": 8.827863058477886e-05,
      "loss": 0.1749,
      "step": 39880
    },
    {
      "epoch": 0.5606859231147656,
      "grad_norm": 0.45237404108047485,
      "learning_rate": 8.823247006471985e-05,
      "loss": 0.1818,
      "step": 39890
    },
    {
      "epoch": 0.5608264811300865,
      "grad_norm": 0.39162522554397583,
      "learning_rate": 8.818631208714765e-05,
      "loss": 0.1728,
      "step": 39900
    },
    {
      "epoch": 0.5609670391454072,
      "grad_norm": 0.28499725461006165,
      "learning_rate": 8.814015666203514e-05,
      "loss": 0.1625,
      "step": 39910
    },
    {
      "epoch": 0.5611075971607281,
      "grad_norm": 0.36940038204193115,
      "learning_rate": 8.809400379935466e-05,
      "loss": 0.1648,
      "step": 39920
    },
    {
      "epoch": 0.5612481551760489,
      "grad_norm": 0.38717323541641235,
      "learning_rate": 8.804785350907791e-05,
      "loss": 0.1416,
      "step": 39930
    },
    {
      "epoch": 0.5613887131913697,
      "grad_norm": 0.3635973036289215,
      "learning_rate": 8.800170580117617e-05,
      "loss": 0.1642,
      "step": 39940
    },
    {
      "epoch": 0.5615292712066906,
      "grad_norm": 0.49187415838241577,
      "learning_rate": 8.795556068562005e-05,
      "loss": 0.1688,
      "step": 39950
    },
    {
      "epoch": 0.5616698292220114,
      "grad_norm": 0.362467497587204,
      "learning_rate": 8.790941817237966e-05,
      "loss": 0.1737,
      "step": 39960
    },
    {
      "epoch": 0.5618103872373322,
      "grad_norm": 0.3785790205001831,
      "learning_rate": 8.786327827142453e-05,
      "loss": 0.1499,
      "step": 39970
    },
    {
      "epoch": 0.561950945252653,
      "grad_norm": 0.39637821912765503,
      "learning_rate": 8.781714099272362e-05,
      "loss": 0.1574,
      "step": 39980
    },
    {
      "epoch": 0.5620915032679739,
      "grad_norm": 0.47537532448768616,
      "learning_rate": 8.77710063462453e-05,
      "loss": 0.1842,
      "step": 39990
    },
    {
      "epoch": 0.5622320612832947,
      "grad_norm": 0.42900022864341736,
      "learning_rate": 8.772487434195745e-05,
      "loss": 0.1754,
      "step": 40000
    },
    {
      "epoch": 0.5622320612832947,
      "eval_chrf": 64.4424962345778,
      "eval_loss": 0.3777381181716919,
      "eval_runtime": 326.6983,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 40000
    },
    {
      "epoch": 0.5623726192986155,
      "grad_norm": 0.4063546657562256,
      "learning_rate": 8.767874498982728e-05,
      "loss": 0.1794,
      "step": 40010
    },
    {
      "epoch": 0.5625131773139364,
      "grad_norm": 0.35100147128105164,
      "learning_rate": 8.763261829982153e-05,
      "loss": 0.1845,
      "step": 40020
    },
    {
      "epoch": 0.5626537353292571,
      "grad_norm": 0.3833320736885071,
      "learning_rate": 8.758649428190626e-05,
      "loss": 0.1744,
      "step": 40030
    },
    {
      "epoch": 0.562794293344578,
      "grad_norm": 0.3445558547973633,
      "learning_rate": 8.754037294604707e-05,
      "loss": 0.1708,
      "step": 40040
    },
    {
      "epoch": 0.5629348513598988,
      "grad_norm": 0.47707948088645935,
      "learning_rate": 8.749425430220882e-05,
      "loss": 0.1587,
      "step": 40050
    },
    {
      "epoch": 0.5630754093752196,
      "grad_norm": 0.4903023838996887,
      "learning_rate": 8.744813836035593e-05,
      "loss": 0.174,
      "step": 40060
    },
    {
      "epoch": 0.5632159673905405,
      "grad_norm": 0.36455291509628296,
      "learning_rate": 8.740202513045219e-05,
      "loss": 0.1656,
      "step": 40070
    },
    {
      "epoch": 0.5633565254058612,
      "grad_norm": 0.4470309913158417,
      "learning_rate": 8.735591462246078e-05,
      "loss": 0.1922,
      "step": 40080
    },
    {
      "epoch": 0.5634970834211821,
      "grad_norm": 0.488297700881958,
      "learning_rate": 8.730980684634435e-05,
      "loss": 0.1852,
      "step": 40090
    },
    {
      "epoch": 0.563637641436503,
      "grad_norm": 0.5019252896308899,
      "learning_rate": 8.726370181206492e-05,
      "loss": 0.1811,
      "step": 40100
    },
    {
      "epoch": 0.5637781994518237,
      "grad_norm": 0.3459615111351013,
      "learning_rate": 8.721759952958388e-05,
      "loss": 0.1762,
      "step": 40110
    },
    {
      "epoch": 0.5639187574671446,
      "grad_norm": 0.3986132740974426,
      "learning_rate": 8.71715000088621e-05,
      "loss": 0.1808,
      "step": 40120
    },
    {
      "epoch": 0.5640593154824654,
      "grad_norm": 0.6364266872406006,
      "learning_rate": 8.712540325985982e-05,
      "loss": 0.1837,
      "step": 40130
    },
    {
      "epoch": 0.5641998734977862,
      "grad_norm": 0.3040315508842468,
      "learning_rate": 8.707930929253668e-05,
      "loss": 0.1474,
      "step": 40140
    },
    {
      "epoch": 0.5643404315131071,
      "grad_norm": 0.39072078466415405,
      "learning_rate": 8.703321811685169e-05,
      "loss": 0.2028,
      "step": 40150
    },
    {
      "epoch": 0.5644809895284278,
      "grad_norm": 0.5064664483070374,
      "learning_rate": 8.698712974276333e-05,
      "loss": 0.1691,
      "step": 40160
    },
    {
      "epoch": 0.5646215475437487,
      "grad_norm": 0.32676252722740173,
      "learning_rate": 8.694104418022941e-05,
      "loss": 0.1771,
      "step": 40170
    },
    {
      "epoch": 0.5647621055590695,
      "grad_norm": 0.3531136214733124,
      "learning_rate": 8.689496143920716e-05,
      "loss": 0.1792,
      "step": 40180
    },
    {
      "epoch": 0.5649026635743903,
      "grad_norm": 0.38171055912971497,
      "learning_rate": 8.684888152965321e-05,
      "loss": 0.186,
      "step": 40190
    },
    {
      "epoch": 0.5650432215897112,
      "grad_norm": 0.4591136574745178,
      "learning_rate": 8.680280446152352e-05,
      "loss": 0.1693,
      "step": 40200
    },
    {
      "epoch": 0.565183779605032,
      "grad_norm": 0.3574647605419159,
      "learning_rate": 8.675673024477353e-05,
      "loss": 0.1526,
      "step": 40210
    },
    {
      "epoch": 0.5653243376203528,
      "grad_norm": 0.3667439818382263,
      "learning_rate": 8.671065888935797e-05,
      "loss": 0.158,
      "step": 40220
    },
    {
      "epoch": 0.5654648956356736,
      "grad_norm": 0.3794707655906677,
      "learning_rate": 8.666459040523102e-05,
      "loss": 0.1534,
      "step": 40230
    },
    {
      "epoch": 0.5656054536509945,
      "grad_norm": 0.3461852967739105,
      "learning_rate": 8.66185248023462e-05,
      "loss": 0.1937,
      "step": 40240
    },
    {
      "epoch": 0.5657460116663152,
      "grad_norm": 0.32846784591674805,
      "learning_rate": 8.657246209065644e-05,
      "loss": 0.185,
      "step": 40250
    },
    {
      "epoch": 0.5658865696816361,
      "grad_norm": 0.4151070713996887,
      "learning_rate": 8.652640228011401e-05,
      "loss": 0.1666,
      "step": 40260
    },
    {
      "epoch": 0.566027127696957,
      "grad_norm": 0.43426674604415894,
      "learning_rate": 8.648034538067058e-05,
      "loss": 0.1796,
      "step": 40270
    },
    {
      "epoch": 0.5661676857122777,
      "grad_norm": 0.3492185175418854,
      "learning_rate": 8.643429140227718e-05,
      "loss": 0.1485,
      "step": 40280
    },
    {
      "epoch": 0.5663082437275986,
      "grad_norm": 0.5607163906097412,
      "learning_rate": 8.63882403548842e-05,
      "loss": 0.1965,
      "step": 40290
    },
    {
      "epoch": 0.5664488017429193,
      "grad_norm": 0.42045024037361145,
      "learning_rate": 8.634219224844142e-05,
      "loss": 0.165,
      "step": 40300
    },
    {
      "epoch": 0.5665893597582402,
      "grad_norm": 0.315655380487442,
      "learning_rate": 8.629614709289798e-05,
      "loss": 0.1908,
      "step": 40310
    },
    {
      "epoch": 0.5667299177735611,
      "grad_norm": 0.47120293974876404,
      "learning_rate": 8.625010489820234e-05,
      "loss": 0.1913,
      "step": 40320
    },
    {
      "epoch": 0.5668704757888818,
      "grad_norm": 0.5474007725715637,
      "learning_rate": 8.62040656743024e-05,
      "loss": 0.1632,
      "step": 40330
    },
    {
      "epoch": 0.5670110338042027,
      "grad_norm": 0.35841497778892517,
      "learning_rate": 8.615802943114533e-05,
      "loss": 0.16,
      "step": 40340
    },
    {
      "epoch": 0.5671515918195235,
      "grad_norm": 0.507109522819519,
      "learning_rate": 8.611199617867772e-05,
      "loss": 0.1864,
      "step": 40350
    },
    {
      "epoch": 0.5672921498348443,
      "grad_norm": 0.43633168935775757,
      "learning_rate": 8.606596592684548e-05,
      "loss": 0.1734,
      "step": 40360
    },
    {
      "epoch": 0.5674327078501652,
      "grad_norm": 0.4774210751056671,
      "learning_rate": 8.60199386855939e-05,
      "loss": 0.1757,
      "step": 40370
    },
    {
      "epoch": 0.567573265865486,
      "grad_norm": 0.4148516058921814,
      "learning_rate": 8.597391446486758e-05,
      "loss": 0.1808,
      "step": 40380
    },
    {
      "epoch": 0.5677138238808068,
      "grad_norm": 0.33463791012763977,
      "learning_rate": 8.592789327461052e-05,
      "loss": 0.1974,
      "step": 40390
    },
    {
      "epoch": 0.5678543818961276,
      "grad_norm": 0.4078878164291382,
      "learning_rate": 8.588187512476603e-05,
      "loss": 0.1585,
      "step": 40400
    },
    {
      "epoch": 0.5679949399114484,
      "grad_norm": 0.383430540561676,
      "learning_rate": 8.583586002527676e-05,
      "loss": 0.1604,
      "step": 40410
    },
    {
      "epoch": 0.5681354979267693,
      "grad_norm": 0.4105486571788788,
      "learning_rate": 8.57898479860847e-05,
      "loss": 0.1751,
      "step": 40420
    },
    {
      "epoch": 0.5682760559420901,
      "grad_norm": 0.5492495894432068,
      "learning_rate": 8.574383901713121e-05,
      "loss": 0.1764,
      "step": 40430
    },
    {
      "epoch": 0.5684166139574109,
      "grad_norm": 0.4135695993900299,
      "learning_rate": 8.569783312835697e-05,
      "loss": 0.1687,
      "step": 40440
    },
    {
      "epoch": 0.5685571719727317,
      "grad_norm": 0.3660704493522644,
      "learning_rate": 8.565183032970195e-05,
      "loss": 0.1766,
      "step": 40450
    },
    {
      "epoch": 0.5686977299880526,
      "grad_norm": 0.5003263354301453,
      "learning_rate": 8.560583063110553e-05,
      "loss": 0.192,
      "step": 40460
    },
    {
      "epoch": 0.5688382880033734,
      "grad_norm": 0.37117767333984375,
      "learning_rate": 8.555983404250638e-05,
      "loss": 0.1699,
      "step": 40470
    },
    {
      "epoch": 0.5689788460186942,
      "grad_norm": 0.4867563843727112,
      "learning_rate": 8.551384057384247e-05,
      "loss": 0.1613,
      "step": 40480
    },
    {
      "epoch": 0.5691194040340151,
      "grad_norm": 0.36081835627555847,
      "learning_rate": 8.546785023505117e-05,
      "loss": 0.1786,
      "step": 40490
    },
    {
      "epoch": 0.5692599620493358,
      "grad_norm": 0.46900445222854614,
      "learning_rate": 8.542186303606909e-05,
      "loss": 0.1884,
      "step": 40500
    },
    {
      "epoch": 0.5694005200646567,
      "grad_norm": 0.5062784552574158,
      "learning_rate": 8.537587898683224e-05,
      "loss": 0.195,
      "step": 40510
    },
    {
      "epoch": 0.5695410780799776,
      "grad_norm": 0.3653715252876282,
      "learning_rate": 8.532989809727588e-05,
      "loss": 0.1945,
      "step": 40520
    },
    {
      "epoch": 0.5696816360952983,
      "grad_norm": 0.2945951521396637,
      "learning_rate": 8.528392037733464e-05,
      "loss": 0.1703,
      "step": 40530
    },
    {
      "epoch": 0.5698221941106192,
      "grad_norm": 0.33782869577407837,
      "learning_rate": 8.523794583694243e-05,
      "loss": 0.1915,
      "step": 40540
    },
    {
      "epoch": 0.5699627521259399,
      "grad_norm": 0.3650352954864502,
      "learning_rate": 8.51919744860325e-05,
      "loss": 0.1746,
      "step": 40550
    },
    {
      "epoch": 0.5701033101412608,
      "grad_norm": 0.3990078866481781,
      "learning_rate": 8.514600633453737e-05,
      "loss": 0.1675,
      "step": 40560
    },
    {
      "epoch": 0.5702438681565817,
      "grad_norm": 0.43076345324516296,
      "learning_rate": 8.510004139238893e-05,
      "loss": 0.1861,
      "step": 40570
    },
    {
      "epoch": 0.5703844261719024,
      "grad_norm": 0.39436662197113037,
      "learning_rate": 8.505407966951833e-05,
      "loss": 0.172,
      "step": 40580
    },
    {
      "epoch": 0.5705249841872233,
      "grad_norm": 0.34969374537467957,
      "learning_rate": 8.500812117585602e-05,
      "loss": 0.1758,
      "step": 40590
    },
    {
      "epoch": 0.5706655422025441,
      "grad_norm": 0.3330826759338379,
      "learning_rate": 8.496216592133181e-05,
      "loss": 0.1647,
      "step": 40600
    },
    {
      "epoch": 0.5708061002178649,
      "grad_norm": 0.5729359984397888,
      "learning_rate": 8.491621391587473e-05,
      "loss": 0.2057,
      "step": 40610
    },
    {
      "epoch": 0.5709466582331858,
      "grad_norm": 0.3391762375831604,
      "learning_rate": 8.487026516941318e-05,
      "loss": 0.1642,
      "step": 40620
    },
    {
      "epoch": 0.5710872162485066,
      "grad_norm": 0.4327235221862793,
      "learning_rate": 8.482431969187479e-05,
      "loss": 0.1756,
      "step": 40630
    },
    {
      "epoch": 0.5712277742638274,
      "grad_norm": 0.4161527454853058,
      "learning_rate": 8.477837749318655e-05,
      "loss": 0.1593,
      "step": 40640
    },
    {
      "epoch": 0.5713683322791482,
      "grad_norm": 0.3668680489063263,
      "learning_rate": 8.473243858327469e-05,
      "loss": 0.1821,
      "step": 40650
    },
    {
      "epoch": 0.571508890294469,
      "grad_norm": 0.43897321820259094,
      "learning_rate": 8.468650297206476e-05,
      "loss": 0.1796,
      "step": 40660
    },
    {
      "epoch": 0.5716494483097899,
      "grad_norm": 0.44199275970458984,
      "learning_rate": 8.464057066948158e-05,
      "loss": 0.1754,
      "step": 40670
    },
    {
      "epoch": 0.5717900063251107,
      "grad_norm": 0.4629787802696228,
      "learning_rate": 8.459464168544927e-05,
      "loss": 0.1847,
      "step": 40680
    },
    {
      "epoch": 0.5719305643404315,
      "grad_norm": 0.3117261826992035,
      "learning_rate": 8.454871602989119e-05,
      "loss": 0.1559,
      "step": 40690
    },
    {
      "epoch": 0.5720711223557523,
      "grad_norm": 0.3957635760307312,
      "learning_rate": 8.450279371273006e-05,
      "loss": 0.1875,
      "step": 40700
    },
    {
      "epoch": 0.5722116803710732,
      "grad_norm": 0.31855207681655884,
      "learning_rate": 8.44568747438878e-05,
      "loss": 0.1644,
      "step": 40710
    },
    {
      "epoch": 0.5723522383863939,
      "grad_norm": 0.38155150413513184,
      "learning_rate": 8.441095913328568e-05,
      "loss": 0.1728,
      "step": 40720
    },
    {
      "epoch": 0.5724927964017148,
      "grad_norm": 0.4814320504665375,
      "learning_rate": 8.436504689084416e-05,
      "loss": 0.1698,
      "step": 40730
    },
    {
      "epoch": 0.5726333544170357,
      "grad_norm": 0.5537896752357483,
      "learning_rate": 8.431913802648304e-05,
      "loss": 0.2047,
      "step": 40740
    },
    {
      "epoch": 0.5727739124323564,
      "grad_norm": 0.32907259464263916,
      "learning_rate": 8.427323255012136e-05,
      "loss": 0.1641,
      "step": 40750
    },
    {
      "epoch": 0.5729144704476773,
      "grad_norm": 0.48211991786956787,
      "learning_rate": 8.422733047167743e-05,
      "loss": 0.2014,
      "step": 40760
    },
    {
      "epoch": 0.573055028462998,
      "grad_norm": 0.41999366879463196,
      "learning_rate": 8.418143180106883e-05,
      "loss": 0.1858,
      "step": 40770
    },
    {
      "epoch": 0.5731955864783189,
      "grad_norm": 0.3663288354873657,
      "learning_rate": 8.413553654821243e-05,
      "loss": 0.1854,
      "step": 40780
    },
    {
      "epoch": 0.5733361444936398,
      "grad_norm": 0.3534975051879883,
      "learning_rate": 8.408964472302432e-05,
      "loss": 0.1972,
      "step": 40790
    },
    {
      "epoch": 0.5734767025089605,
      "grad_norm": 0.43506941199302673,
      "learning_rate": 8.404375633541985e-05,
      "loss": 0.1823,
      "step": 40800
    },
    {
      "epoch": 0.5736172605242814,
      "grad_norm": 0.3940921723842621,
      "learning_rate": 8.399787139531367e-05,
      "loss": 0.1743,
      "step": 40810
    },
    {
      "epoch": 0.5737578185396022,
      "grad_norm": 0.34778329730033875,
      "learning_rate": 8.395198991261963e-05,
      "loss": 0.163,
      "step": 40820
    },
    {
      "epoch": 0.573898376554923,
      "grad_norm": 0.691506028175354,
      "learning_rate": 8.390611189725088e-05,
      "loss": 0.1815,
      "step": 40830
    },
    {
      "epoch": 0.5740389345702439,
      "grad_norm": 0.37461793422698975,
      "learning_rate": 8.386023735911981e-05,
      "loss": 0.2127,
      "step": 40840
    },
    {
      "epoch": 0.5741794925855647,
      "grad_norm": 0.4207349121570587,
      "learning_rate": 8.381436630813803e-05,
      "loss": 0.1821,
      "step": 40850
    },
    {
      "epoch": 0.5743200506008855,
      "grad_norm": 0.36333221197128296,
      "learning_rate": 8.376849875421642e-05,
      "loss": 0.1753,
      "step": 40860
    },
    {
      "epoch": 0.5744606086162063,
      "grad_norm": 0.4035714566707611,
      "learning_rate": 8.372263470726512e-05,
      "loss": 0.1559,
      "step": 40870
    },
    {
      "epoch": 0.5746011666315272,
      "grad_norm": 0.3609062135219574,
      "learning_rate": 8.367677417719348e-05,
      "loss": 0.1753,
      "step": 40880
    },
    {
      "epoch": 0.574741724646848,
      "grad_norm": 0.4326991140842438,
      "learning_rate": 8.36309171739101e-05,
      "loss": 0.1652,
      "step": 40890
    },
    {
      "epoch": 0.5748822826621688,
      "grad_norm": 0.4023519456386566,
      "learning_rate": 8.358506370732282e-05,
      "loss": 0.1817,
      "step": 40900
    },
    {
      "epoch": 0.5750228406774897,
      "grad_norm": 0.35921162366867065,
      "learning_rate": 8.353921378733873e-05,
      "loss": 0.1642,
      "step": 40910
    },
    {
      "epoch": 0.5751633986928104,
      "grad_norm": 0.3350284695625305,
      "learning_rate": 8.349336742386413e-05,
      "loss": 0.1715,
      "step": 40920
    },
    {
      "epoch": 0.5753039567081313,
      "grad_norm": 0.4562940001487732,
      "learning_rate": 8.344752462680458e-05,
      "loss": 0.1768,
      "step": 40930
    },
    {
      "epoch": 0.5754445147234521,
      "grad_norm": 0.3363032341003418,
      "learning_rate": 8.340168540606483e-05,
      "loss": 0.1793,
      "step": 40940
    },
    {
      "epoch": 0.5755850727387729,
      "grad_norm": 0.29547351598739624,
      "learning_rate": 8.33558497715489e-05,
      "loss": 0.1704,
      "step": 40950
    },
    {
      "epoch": 0.5757256307540938,
      "grad_norm": 0.31880930066108704,
      "learning_rate": 8.331001773315998e-05,
      "loss": 0.185,
      "step": 40960
    },
    {
      "epoch": 0.5758661887694145,
      "grad_norm": 0.4245101809501648,
      "learning_rate": 8.326418930080057e-05,
      "loss": 0.1812,
      "step": 40970
    },
    {
      "epoch": 0.5760067467847354,
      "grad_norm": 0.42613980174064636,
      "learning_rate": 8.321836448437232e-05,
      "loss": 0.179,
      "step": 40980
    },
    {
      "epoch": 0.5761473048000563,
      "grad_norm": 0.5401775240898132,
      "learning_rate": 8.317254329377608e-05,
      "loss": 0.1773,
      "step": 40990
    },
    {
      "epoch": 0.576287862815377,
      "grad_norm": 0.3493726849555969,
      "learning_rate": 8.312672573891202e-05,
      "loss": 0.1827,
      "step": 41000
    },
    {
      "epoch": 0.576287862815377,
      "eval_chrf": 70.05708677995977,
      "eval_loss": 0.37228840589523315,
      "eval_runtime": 326.7373,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 41000
    },
    {
      "epoch": 0.5764284208306979,
      "grad_norm": 0.3397007882595062,
      "learning_rate": 8.30809118296794e-05,
      "loss": 0.1789,
      "step": 41010
    },
    {
      "epoch": 0.5765689788460187,
      "grad_norm": 0.3283292055130005,
      "learning_rate": 8.303510157597678e-05,
      "loss": 0.1847,
      "step": 41020
    },
    {
      "epoch": 0.5767095368613395,
      "grad_norm": 0.3661724925041199,
      "learning_rate": 8.29892949877019e-05,
      "loss": 0.203,
      "step": 41030
    },
    {
      "epoch": 0.5768500948766604,
      "grad_norm": 0.39149966835975647,
      "learning_rate": 8.29434920747517e-05,
      "loss": 0.1894,
      "step": 41040
    },
    {
      "epoch": 0.5769906528919811,
      "grad_norm": 0.41224539279937744,
      "learning_rate": 8.289769284702234e-05,
      "loss": 0.1874,
      "step": 41050
    },
    {
      "epoch": 0.577131210907302,
      "grad_norm": 0.33227643370628357,
      "learning_rate": 8.285189731440919e-05,
      "loss": 0.1756,
      "step": 41060
    },
    {
      "epoch": 0.5772717689226228,
      "grad_norm": 0.40387648344039917,
      "learning_rate": 8.280610548680678e-05,
      "loss": 0.1678,
      "step": 41070
    },
    {
      "epoch": 0.5774123269379436,
      "grad_norm": 0.3667382001876831,
      "learning_rate": 8.27603173741089e-05,
      "loss": 0.1744,
      "step": 41080
    },
    {
      "epoch": 0.5775528849532645,
      "grad_norm": 0.2799052894115448,
      "learning_rate": 8.271453298620851e-05,
      "loss": 0.1808,
      "step": 41090
    },
    {
      "epoch": 0.5776934429685853,
      "grad_norm": 0.4208914041519165,
      "learning_rate": 8.266875233299773e-05,
      "loss": 0.1794,
      "step": 41100
    },
    {
      "epoch": 0.5778340009839061,
      "grad_norm": 0.31498298048973083,
      "learning_rate": 8.262297542436793e-05,
      "loss": 0.1706,
      "step": 41110
    },
    {
      "epoch": 0.5779745589992269,
      "grad_norm": 0.4450007379055023,
      "learning_rate": 8.257720227020965e-05,
      "loss": 0.1936,
      "step": 41120
    },
    {
      "epoch": 0.5781151170145478,
      "grad_norm": 0.3582266867160797,
      "learning_rate": 8.253143288041261e-05,
      "loss": 0.1942,
      "step": 41130
    },
    {
      "epoch": 0.5782556750298686,
      "grad_norm": 0.4047289788722992,
      "learning_rate": 8.248566726486571e-05,
      "loss": 0.1622,
      "step": 41140
    },
    {
      "epoch": 0.5783962330451894,
      "grad_norm": 0.3665313422679901,
      "learning_rate": 8.243990543345707e-05,
      "loss": 0.1777,
      "step": 41150
    },
    {
      "epoch": 0.5785367910605103,
      "grad_norm": 0.3635713458061218,
      "learning_rate": 8.239414739607394e-05,
      "loss": 0.1694,
      "step": 41160
    },
    {
      "epoch": 0.578677349075831,
      "grad_norm": 0.3851402997970581,
      "learning_rate": 8.23483931626028e-05,
      "loss": 0.1754,
      "step": 41170
    },
    {
      "epoch": 0.5788179070911519,
      "grad_norm": 0.36724674701690674,
      "learning_rate": 8.230264274292929e-05,
      "loss": 0.1605,
      "step": 41180
    },
    {
      "epoch": 0.5789584651064726,
      "grad_norm": 0.39243537187576294,
      "learning_rate": 8.225689614693821e-05,
      "loss": 0.169,
      "step": 41190
    },
    {
      "epoch": 0.5790990231217935,
      "grad_norm": 0.4184189736843109,
      "learning_rate": 8.221115338451355e-05,
      "loss": 0.1934,
      "step": 41200
    },
    {
      "epoch": 0.5792395811371144,
      "grad_norm": 0.37976041436195374,
      "learning_rate": 8.216541446553848e-05,
      "loss": 0.1907,
      "step": 41210
    },
    {
      "epoch": 0.5793801391524351,
      "grad_norm": 0.4090244472026825,
      "learning_rate": 8.211967939989531e-05,
      "loss": 0.1817,
      "step": 41220
    },
    {
      "epoch": 0.579520697167756,
      "grad_norm": 0.4144601821899414,
      "learning_rate": 8.207394819746558e-05,
      "loss": 0.1642,
      "step": 41230
    },
    {
      "epoch": 0.5796612551830768,
      "grad_norm": 0.3041282594203949,
      "learning_rate": 8.202822086812989e-05,
      "loss": 0.1777,
      "step": 41240
    },
    {
      "epoch": 0.5798018131983976,
      "grad_norm": 0.34671542048454285,
      "learning_rate": 8.19824974217681e-05,
      "loss": 0.1616,
      "step": 41250
    },
    {
      "epoch": 0.5799423712137185,
      "grad_norm": 0.39715781807899475,
      "learning_rate": 8.19367778682592e-05,
      "loss": 0.1927,
      "step": 41260
    },
    {
      "epoch": 0.5800829292290393,
      "grad_norm": 0.43676459789276123,
      "learning_rate": 8.189106221748131e-05,
      "loss": 0.1739,
      "step": 41270
    },
    {
      "epoch": 0.5802234872443601,
      "grad_norm": 0.4754163324832916,
      "learning_rate": 8.184535047931175e-05,
      "loss": 0.1809,
      "step": 41280
    },
    {
      "epoch": 0.5803640452596809,
      "grad_norm": 0.3058910667896271,
      "learning_rate": 8.179964266362697e-05,
      "loss": 0.1597,
      "step": 41290
    },
    {
      "epoch": 0.5805046032750018,
      "grad_norm": 0.30070939660072327,
      "learning_rate": 8.175393878030257e-05,
      "loss": 0.1652,
      "step": 41300
    },
    {
      "epoch": 0.5806451612903226,
      "grad_norm": 0.4284345507621765,
      "learning_rate": 8.170823883921335e-05,
      "loss": 0.1975,
      "step": 41310
    },
    {
      "epoch": 0.5807857193056434,
      "grad_norm": 0.4294290542602539,
      "learning_rate": 8.166254285023315e-05,
      "loss": 0.1803,
      "step": 41320
    },
    {
      "epoch": 0.5809262773209642,
      "grad_norm": 0.4316389560699463,
      "learning_rate": 8.161685082323508e-05,
      "loss": 0.1586,
      "step": 41330
    },
    {
      "epoch": 0.581066835336285,
      "grad_norm": 0.35033249855041504,
      "learning_rate": 8.157116276809132e-05,
      "loss": 0.1688,
      "step": 41340
    },
    {
      "epoch": 0.5812073933516059,
      "grad_norm": 0.2849494218826294,
      "learning_rate": 8.152547869467319e-05,
      "loss": 0.1864,
      "step": 41350
    },
    {
      "epoch": 0.5813479513669267,
      "grad_norm": 0.4948350489139557,
      "learning_rate": 8.14797986128512e-05,
      "loss": 0.171,
      "step": 41360
    },
    {
      "epoch": 0.5814885093822475,
      "grad_norm": 0.3462991714477539,
      "learning_rate": 8.143412253249494e-05,
      "loss": 0.1831,
      "step": 41370
    },
    {
      "epoch": 0.5816290673975684,
      "grad_norm": 0.539348840713501,
      "learning_rate": 8.138845046347316e-05,
      "loss": 0.1776,
      "step": 41380
    },
    {
      "epoch": 0.5817696254128891,
      "grad_norm": 0.32116562128067017,
      "learning_rate": 8.134278241565383e-05,
      "loss": 0.1731,
      "step": 41390
    },
    {
      "epoch": 0.58191018342821,
      "grad_norm": 0.4358501732349396,
      "learning_rate": 8.129711839890383e-05,
      "loss": 0.1815,
      "step": 41400
    },
    {
      "epoch": 0.5820507414435309,
      "grad_norm": 0.3794027864933014,
      "learning_rate": 8.125145842308938e-05,
      "loss": 0.1849,
      "step": 41410
    },
    {
      "epoch": 0.5821912994588516,
      "grad_norm": 0.4954479932785034,
      "learning_rate": 8.120580249807573e-05,
      "loss": 0.1944,
      "step": 41420
    },
    {
      "epoch": 0.5823318574741725,
      "grad_norm": 0.40640848875045776,
      "learning_rate": 8.11601506337273e-05,
      "loss": 0.1725,
      "step": 41430
    },
    {
      "epoch": 0.5824724154894932,
      "grad_norm": 0.4593553841114044,
      "learning_rate": 8.11145028399076e-05,
      "loss": 0.1695,
      "step": 41440
    },
    {
      "epoch": 0.5826129735048141,
      "grad_norm": 0.47762438654899597,
      "learning_rate": 8.106885912647926e-05,
      "loss": 0.1697,
      "step": 41450
    },
    {
      "epoch": 0.582753531520135,
      "grad_norm": 0.3857489228248596,
      "learning_rate": 8.102321950330404e-05,
      "loss": 0.1781,
      "step": 41460
    },
    {
      "epoch": 0.5828940895354557,
      "grad_norm": 0.2961033582687378,
      "learning_rate": 8.097758398024283e-05,
      "loss": 0.147,
      "step": 41470
    },
    {
      "epoch": 0.5830346475507766,
      "grad_norm": 0.42926403880119324,
      "learning_rate": 8.093195256715561e-05,
      "loss": 0.1745,
      "step": 41480
    },
    {
      "epoch": 0.5831752055660974,
      "grad_norm": 0.345276415348053,
      "learning_rate": 8.088632527390148e-05,
      "loss": 0.1595,
      "step": 41490
    },
    {
      "epoch": 0.5833157635814182,
      "grad_norm": 0.33365002274513245,
      "learning_rate": 8.084070211033866e-05,
      "loss": 0.1625,
      "step": 41500
    },
    {
      "epoch": 0.5834563215967391,
      "grad_norm": 0.393058717250824,
      "learning_rate": 8.079508308632447e-05,
      "loss": 0.1745,
      "step": 41510
    },
    {
      "epoch": 0.5835968796120599,
      "grad_norm": 0.299457311630249,
      "learning_rate": 8.07494682117153e-05,
      "loss": 0.1485,
      "step": 41520
    },
    {
      "epoch": 0.5837374376273807,
      "grad_norm": 0.4067396819591522,
      "learning_rate": 8.070385749636672e-05,
      "loss": 0.1664,
      "step": 41530
    },
    {
      "epoch": 0.5838779956427015,
      "grad_norm": 0.4003329873085022,
      "learning_rate": 8.065825095013333e-05,
      "loss": 0.1511,
      "step": 41540
    },
    {
      "epoch": 0.5840185536580224,
      "grad_norm": 0.4504907727241516,
      "learning_rate": 8.061264858286888e-05,
      "loss": 0.174,
      "step": 41550
    },
    {
      "epoch": 0.5841591116733432,
      "grad_norm": 0.3335106074810028,
      "learning_rate": 8.056705040442619e-05,
      "loss": 0.1903,
      "step": 41560
    },
    {
      "epoch": 0.584299669688664,
      "grad_norm": 0.40294119715690613,
      "learning_rate": 8.052145642465717e-05,
      "loss": 0.1739,
      "step": 41570
    },
    {
      "epoch": 0.5844402277039848,
      "grad_norm": 0.31343063712120056,
      "learning_rate": 8.047586665341284e-05,
      "loss": 0.1546,
      "step": 41580
    },
    {
      "epoch": 0.5845807857193056,
      "grad_norm": 0.3426019847393036,
      "learning_rate": 8.043028110054327e-05,
      "loss": 0.1707,
      "step": 41590
    },
    {
      "epoch": 0.5847213437346265,
      "grad_norm": 0.44034767150878906,
      "learning_rate": 8.038469977589773e-05,
      "loss": 0.1704,
      "step": 41600
    },
    {
      "epoch": 0.5848619017499473,
      "grad_norm": 0.4241983890533447,
      "learning_rate": 8.033912268932446e-05,
      "loss": 0.1471,
      "step": 41610
    },
    {
      "epoch": 0.5850024597652681,
      "grad_norm": 0.5022419691085815,
      "learning_rate": 8.029354985067078e-05,
      "loss": 0.1903,
      "step": 41620
    },
    {
      "epoch": 0.585143017780589,
      "grad_norm": 0.4251757860183716,
      "learning_rate": 8.024798126978321e-05,
      "loss": 0.193,
      "step": 41630
    },
    {
      "epoch": 0.5852835757959097,
      "grad_norm": 0.37948426604270935,
      "learning_rate": 8.020241695650723e-05,
      "loss": 0.1887,
      "step": 41640
    },
    {
      "epoch": 0.5854241338112306,
      "grad_norm": 0.3809700608253479,
      "learning_rate": 8.015685692068744e-05,
      "loss": 0.1804,
      "step": 41650
    },
    {
      "epoch": 0.5855646918265515,
      "grad_norm": 0.43913763761520386,
      "learning_rate": 8.011130117216755e-05,
      "loss": 0.1549,
      "step": 41660
    },
    {
      "epoch": 0.5857052498418722,
      "grad_norm": 0.25458937883377075,
      "learning_rate": 8.006574972079028e-05,
      "loss": 0.1705,
      "step": 41670
    },
    {
      "epoch": 0.5858458078571931,
      "grad_norm": 0.29581257700920105,
      "learning_rate": 8.002020257639748e-05,
      "loss": 0.1491,
      "step": 41680
    },
    {
      "epoch": 0.5859863658725138,
      "grad_norm": 0.43577784299850464,
      "learning_rate": 7.997465974883003e-05,
      "loss": 0.1953,
      "step": 41690
    },
    {
      "epoch": 0.5861269238878347,
      "grad_norm": 0.23517319560050964,
      "learning_rate": 7.992912124792789e-05,
      "loss": 0.1517,
      "step": 41700
    },
    {
      "epoch": 0.5862674819031555,
      "grad_norm": 0.43467965722084045,
      "learning_rate": 7.988358708353009e-05,
      "loss": 0.1815,
      "step": 41710
    },
    {
      "epoch": 0.5864080399184763,
      "grad_norm": 0.5543309450149536,
      "learning_rate": 7.983805726547471e-05,
      "loss": 0.1531,
      "step": 41720
    },
    {
      "epoch": 0.5865485979337972,
      "grad_norm": 0.27722403407096863,
      "learning_rate": 7.979253180359893e-05,
      "loss": 0.1457,
      "step": 41730
    },
    {
      "epoch": 0.586689155949118,
      "grad_norm": 0.41455915570259094,
      "learning_rate": 7.974701070773893e-05,
      "loss": 0.174,
      "step": 41740
    },
    {
      "epoch": 0.5868297139644388,
      "grad_norm": 0.30396610498428345,
      "learning_rate": 7.970149398773001e-05,
      "loss": 0.1733,
      "step": 41750
    },
    {
      "epoch": 0.5869702719797596,
      "grad_norm": 0.41698095202445984,
      "learning_rate": 7.965598165340644e-05,
      "loss": 0.1778,
      "step": 41760
    },
    {
      "epoch": 0.5871108299950805,
      "grad_norm": 0.504375696182251,
      "learning_rate": 7.961047371460162e-05,
      "loss": 0.1687,
      "step": 41770
    },
    {
      "epoch": 0.5872513880104013,
      "grad_norm": 0.43992507457733154,
      "learning_rate": 7.956497018114798e-05,
      "loss": 0.1843,
      "step": 41780
    },
    {
      "epoch": 0.5873919460257221,
      "grad_norm": 0.35274389386177063,
      "learning_rate": 7.951947106287699e-05,
      "loss": 0.1744,
      "step": 41790
    },
    {
      "epoch": 0.587532504041043,
      "grad_norm": 0.38785210251808167,
      "learning_rate": 7.947397636961915e-05,
      "loss": 0.1676,
      "step": 41800
    },
    {
      "epoch": 0.5876730620563637,
      "grad_norm": 0.44539543986320496,
      "learning_rate": 7.942848611120404e-05,
      "loss": 0.1705,
      "step": 41810
    },
    {
      "epoch": 0.5878136200716846,
      "grad_norm": 0.39283066987991333,
      "learning_rate": 7.938300029746025e-05,
      "loss": 0.1576,
      "step": 41820
    },
    {
      "epoch": 0.5879541780870055,
      "grad_norm": 0.39538851380348206,
      "learning_rate": 7.933751893821544e-05,
      "loss": 0.1693,
      "step": 41830
    },
    {
      "epoch": 0.5880947361023262,
      "grad_norm": 0.3279722034931183,
      "learning_rate": 7.929204204329627e-05,
      "loss": 0.1867,
      "step": 41840
    },
    {
      "epoch": 0.5882352941176471,
      "grad_norm": 0.3832647502422333,
      "learning_rate": 7.924656962252846e-05,
      "loss": 0.1659,
      "step": 41850
    },
    {
      "epoch": 0.5883758521329678,
      "grad_norm": 0.41260239481925964,
      "learning_rate": 7.920110168573678e-05,
      "loss": 0.1773,
      "step": 41860
    },
    {
      "epoch": 0.5885164101482887,
      "grad_norm": 0.32211098074913025,
      "learning_rate": 7.915563824274497e-05,
      "loss": 0.1869,
      "step": 41870
    },
    {
      "epoch": 0.5886569681636096,
      "grad_norm": 0.33297044038772583,
      "learning_rate": 7.911017930337589e-05,
      "loss": 0.1659,
      "step": 41880
    },
    {
      "epoch": 0.5887975261789303,
      "grad_norm": 0.372994989156723,
      "learning_rate": 7.906472487745133e-05,
      "loss": 0.1707,
      "step": 41890
    },
    {
      "epoch": 0.5889380841942512,
      "grad_norm": 0.5005825757980347,
      "learning_rate": 7.901927497479219e-05,
      "loss": 0.1782,
      "step": 41900
    },
    {
      "epoch": 0.589078642209572,
      "grad_norm": 0.5057122707366943,
      "learning_rate": 7.897382960521833e-05,
      "loss": 0.178,
      "step": 41910
    },
    {
      "epoch": 0.5892192002248928,
      "grad_norm": 0.466781884431839,
      "learning_rate": 7.892838877854866e-05,
      "loss": 0.1734,
      "step": 41920
    },
    {
      "epoch": 0.5893597582402137,
      "grad_norm": 0.45633605122566223,
      "learning_rate": 7.888295250460111e-05,
      "loss": 0.1942,
      "step": 41930
    },
    {
      "epoch": 0.5895003162555345,
      "grad_norm": 0.5042839050292969,
      "learning_rate": 7.883752079319259e-05,
      "loss": 0.1671,
      "step": 41940
    },
    {
      "epoch": 0.5896408742708553,
      "grad_norm": 0.4372533857822418,
      "learning_rate": 7.87920936541391e-05,
      "loss": 0.1834,
      "step": 41950
    },
    {
      "epoch": 0.5897814322861761,
      "grad_norm": 0.4715641438961029,
      "learning_rate": 7.874667109725557e-05,
      "loss": 0.161,
      "step": 41960
    },
    {
      "epoch": 0.589921990301497,
      "grad_norm": 0.49999579787254333,
      "learning_rate": 7.8701253132356e-05,
      "loss": 0.1812,
      "step": 41970
    },
    {
      "epoch": 0.5900625483168178,
      "grad_norm": 0.38548150658607483,
      "learning_rate": 7.865583976925336e-05,
      "loss": 0.2009,
      "step": 41980
    },
    {
      "epoch": 0.5902031063321386,
      "grad_norm": 0.5082098245620728,
      "learning_rate": 7.861043101775963e-05,
      "loss": 0.1661,
      "step": 41990
    },
    {
      "epoch": 0.5903436643474594,
      "grad_norm": 0.5039867162704468,
      "learning_rate": 7.856502688768582e-05,
      "loss": 0.2032,
      "step": 42000
    },
    {
      "epoch": 0.5903436643474594,
      "eval_chrf": 69.27115948936047,
      "eval_loss": 0.363955020904541,
      "eval_runtime": 324.8691,
      "eval_samples_per_second": 0.308,
      "eval_steps_per_second": 0.012,
      "step": 42000
    },
    {
      "epoch": 0.5904842223627802,
      "grad_norm": 0.45851778984069824,
      "learning_rate": 7.851962738884191e-05,
      "loss": 0.1649,
      "step": 42010
    },
    {
      "epoch": 0.5906247803781011,
      "grad_norm": 0.422050803899765,
      "learning_rate": 7.847423253103691e-05,
      "loss": 0.1759,
      "step": 42020
    },
    {
      "epoch": 0.5907653383934219,
      "grad_norm": 0.4894850254058838,
      "learning_rate": 7.84288423240788e-05,
      "loss": 0.1693,
      "step": 42030
    },
    {
      "epoch": 0.5909058964087427,
      "grad_norm": 0.33519265055656433,
      "learning_rate": 7.838345677777457e-05,
      "loss": 0.1721,
      "step": 42040
    },
    {
      "epoch": 0.5910464544240636,
      "grad_norm": 0.45269614458084106,
      "learning_rate": 7.83380759019302e-05,
      "loss": 0.1921,
      "step": 42050
    },
    {
      "epoch": 0.5911870124393843,
      "grad_norm": 0.4886782765388489,
      "learning_rate": 7.82926997063507e-05,
      "loss": 0.1732,
      "step": 42060
    },
    {
      "epoch": 0.5913275704547052,
      "grad_norm": 0.39376261830329895,
      "learning_rate": 7.824732820083993e-05,
      "loss": 0.1578,
      "step": 42070
    },
    {
      "epoch": 0.591468128470026,
      "grad_norm": 0.39648663997650146,
      "learning_rate": 7.82019613952009e-05,
      "loss": 0.1759,
      "step": 42080
    },
    {
      "epoch": 0.5916086864853468,
      "grad_norm": 0.4441344141960144,
      "learning_rate": 7.815659929923554e-05,
      "loss": 0.1838,
      "step": 42090
    },
    {
      "epoch": 0.5917492445006677,
      "grad_norm": 0.3926589787006378,
      "learning_rate": 7.811124192274473e-05,
      "loss": 0.1902,
      "step": 42100
    },
    {
      "epoch": 0.5918898025159884,
      "grad_norm": 0.43851348757743835,
      "learning_rate": 7.806588927552841e-05,
      "loss": 0.1802,
      "step": 42110
    },
    {
      "epoch": 0.5920303605313093,
      "grad_norm": 0.39181485772132874,
      "learning_rate": 7.802054136738544e-05,
      "loss": 0.1659,
      "step": 42120
    },
    {
      "epoch": 0.5921709185466302,
      "grad_norm": 0.5086449384689331,
      "learning_rate": 7.797519820811363e-05,
      "loss": 0.1996,
      "step": 42130
    },
    {
      "epoch": 0.5923114765619509,
      "grad_norm": 0.26584893465042114,
      "learning_rate": 7.792985980750984e-05,
      "loss": 0.1756,
      "step": 42140
    },
    {
      "epoch": 0.5924520345772718,
      "grad_norm": 0.3105525076389313,
      "learning_rate": 7.788452617536984e-05,
      "loss": 0.1816,
      "step": 42150
    },
    {
      "epoch": 0.5925925925925926,
      "grad_norm": 0.37008169293403625,
      "learning_rate": 7.783919732148841e-05,
      "loss": 0.161,
      "step": 42160
    },
    {
      "epoch": 0.5927331506079134,
      "grad_norm": 0.36043819785118103,
      "learning_rate": 7.779387325565926e-05,
      "loss": 0.191,
      "step": 42170
    },
    {
      "epoch": 0.5928737086232342,
      "grad_norm": 0.39846110343933105,
      "learning_rate": 7.774855398767511e-05,
      "loss": 0.1795,
      "step": 42180
    },
    {
      "epoch": 0.593014266638555,
      "grad_norm": 0.3892119526863098,
      "learning_rate": 7.77032395273276e-05,
      "loss": 0.1741,
      "step": 42190
    },
    {
      "epoch": 0.5931548246538759,
      "grad_norm": 0.3905502259731293,
      "learning_rate": 7.765792988440737e-05,
      "loss": 0.1506,
      "step": 42200
    },
    {
      "epoch": 0.5932953826691967,
      "grad_norm": 0.5239992141723633,
      "learning_rate": 7.761262506870397e-05,
      "loss": 0.201,
      "step": 42210
    },
    {
      "epoch": 0.5934359406845175,
      "grad_norm": 0.34727463126182556,
      "learning_rate": 7.756732509000597e-05,
      "loss": 0.1549,
      "step": 42220
    },
    {
      "epoch": 0.5935764986998383,
      "grad_norm": 0.3456290662288666,
      "learning_rate": 7.752202995810083e-05,
      "loss": 0.1571,
      "step": 42230
    },
    {
      "epoch": 0.5937170567151592,
      "grad_norm": 0.4618767499923706,
      "learning_rate": 7.7476739682775e-05,
      "loss": 0.1798,
      "step": 42240
    },
    {
      "epoch": 0.59385761473048,
      "grad_norm": 0.3260291814804077,
      "learning_rate": 7.74314542738139e-05,
      "loss": 0.1549,
      "step": 42250
    },
    {
      "epoch": 0.5939981727458008,
      "grad_norm": 0.28016695380210876,
      "learning_rate": 7.738617374100187e-05,
      "loss": 0.1613,
      "step": 42260
    },
    {
      "epoch": 0.5941387307611217,
      "grad_norm": 0.4371799826622009,
      "learning_rate": 7.734089809412217e-05,
      "loss": 0.1774,
      "step": 42270
    },
    {
      "epoch": 0.5942792887764424,
      "grad_norm": 0.4408225417137146,
      "learning_rate": 7.729562734295705e-05,
      "loss": 0.1873,
      "step": 42280
    },
    {
      "epoch": 0.5944198467917633,
      "grad_norm": 0.33245933055877686,
      "learning_rate": 7.72503614972877e-05,
      "loss": 0.1552,
      "step": 42290
    },
    {
      "epoch": 0.5945604048070842,
      "grad_norm": 0.38343554735183716,
      "learning_rate": 7.720510056689423e-05,
      "loss": 0.1553,
      "step": 42300
    },
    {
      "epoch": 0.5947009628224049,
      "grad_norm": 0.43015557527542114,
      "learning_rate": 7.715984456155567e-05,
      "loss": 0.1812,
      "step": 42310
    },
    {
      "epoch": 0.5948415208377258,
      "grad_norm": 0.2760184407234192,
      "learning_rate": 7.711459349105004e-05,
      "loss": 0.1672,
      "step": 42320
    },
    {
      "epoch": 0.5949820788530465,
      "grad_norm": 0.48811301589012146,
      "learning_rate": 7.706934736515425e-05,
      "loss": 0.166,
      "step": 42330
    },
    {
      "epoch": 0.5951226368683674,
      "grad_norm": 0.4176845848560333,
      "learning_rate": 7.702410619364417e-05,
      "loss": 0.1685,
      "step": 42340
    },
    {
      "epoch": 0.5952631948836883,
      "grad_norm": 0.33408379554748535,
      "learning_rate": 7.697886998629457e-05,
      "loss": 0.1521,
      "step": 42350
    },
    {
      "epoch": 0.595403752899009,
      "grad_norm": 0.4298063814640045,
      "learning_rate": 7.693363875287918e-05,
      "loss": 0.1593,
      "step": 42360
    },
    {
      "epoch": 0.5955443109143299,
      "grad_norm": 0.4347199499607086,
      "learning_rate": 7.68884125031706e-05,
      "loss": 0.1955,
      "step": 42370
    },
    {
      "epoch": 0.5956848689296507,
      "grad_norm": 0.39815792441368103,
      "learning_rate": 7.684319124694045e-05,
      "loss": 0.1786,
      "step": 42380
    },
    {
      "epoch": 0.5958254269449715,
      "grad_norm": 0.274598628282547,
      "learning_rate": 7.679797499395916e-05,
      "loss": 0.1785,
      "step": 42390
    },
    {
      "epoch": 0.5959659849602924,
      "grad_norm": 0.3358634114265442,
      "learning_rate": 7.675276375399618e-05,
      "loss": 0.1747,
      "step": 42400
    },
    {
      "epoch": 0.5961065429756132,
      "grad_norm": 0.3847486674785614,
      "learning_rate": 7.670755753681977e-05,
      "loss": 0.1581,
      "step": 42410
    },
    {
      "epoch": 0.596247100990934,
      "grad_norm": 0.4903138279914856,
      "learning_rate": 7.666235635219721e-05,
      "loss": 0.1654,
      "step": 42420
    },
    {
      "epoch": 0.5963876590062548,
      "grad_norm": 0.4884909391403198,
      "learning_rate": 7.661716020989464e-05,
      "loss": 0.1845,
      "step": 42430
    },
    {
      "epoch": 0.5965282170215757,
      "grad_norm": 0.4085238575935364,
      "learning_rate": 7.65719691196771e-05,
      "loss": 0.1645,
      "step": 42440
    },
    {
      "epoch": 0.5966687750368965,
      "grad_norm": 0.3783973157405853,
      "learning_rate": 7.652678309130857e-05,
      "loss": 0.1779,
      "step": 42450
    },
    {
      "epoch": 0.5968093330522173,
      "grad_norm": 0.3126957416534424,
      "learning_rate": 7.648160213455191e-05,
      "loss": 0.1647,
      "step": 42460
    },
    {
      "epoch": 0.5969498910675382,
      "grad_norm": 0.3112151324748993,
      "learning_rate": 7.64364262591689e-05,
      "loss": 0.1688,
      "step": 42470
    },
    {
      "epoch": 0.5970904490828589,
      "grad_norm": 0.27285337448120117,
      "learning_rate": 7.639125547492023e-05,
      "loss": 0.1805,
      "step": 42480
    },
    {
      "epoch": 0.5972310070981798,
      "grad_norm": 0.44554203748703003,
      "learning_rate": 7.634608979156545e-05,
      "loss": 0.1633,
      "step": 42490
    },
    {
      "epoch": 0.5973715651135006,
      "grad_norm": 0.33010488748550415,
      "learning_rate": 7.630092921886305e-05,
      "loss": 0.1624,
      "step": 42500
    },
    {
      "epoch": 0.5975121231288214,
      "grad_norm": 0.29994094371795654,
      "learning_rate": 7.625577376657042e-05,
      "loss": 0.1679,
      "step": 42510
    },
    {
      "epoch": 0.5976526811441423,
      "grad_norm": 0.4038458466529846,
      "learning_rate": 7.62106234444438e-05,
      "loss": 0.1776,
      "step": 42520
    },
    {
      "epoch": 0.597793239159463,
      "grad_norm": 0.3720175623893738,
      "learning_rate": 7.616547826223835e-05,
      "loss": 0.1886,
      "step": 42530
    },
    {
      "epoch": 0.5979337971747839,
      "grad_norm": 0.39369431138038635,
      "learning_rate": 7.612033822970813e-05,
      "loss": 0.1736,
      "step": 42540
    },
    {
      "epoch": 0.5980743551901048,
      "grad_norm": 0.28870657086372375,
      "learning_rate": 7.607520335660606e-05,
      "loss": 0.1575,
      "step": 42550
    },
    {
      "epoch": 0.5982149132054255,
      "grad_norm": 0.5239557027816772,
      "learning_rate": 7.603007365268397e-05,
      "loss": 0.1843,
      "step": 42560
    },
    {
      "epoch": 0.5983554712207464,
      "grad_norm": 0.40606215596199036,
      "learning_rate": 7.598494912769255e-05,
      "loss": 0.1622,
      "step": 42570
    },
    {
      "epoch": 0.5984960292360672,
      "grad_norm": 0.33013060688972473,
      "learning_rate": 7.593982979138139e-05,
      "loss": 0.1497,
      "step": 42580
    },
    {
      "epoch": 0.598636587251388,
      "grad_norm": 0.4070757031440735,
      "learning_rate": 7.589471565349895e-05,
      "loss": 0.1736,
      "step": 42590
    },
    {
      "epoch": 0.5987771452667089,
      "grad_norm": 0.4686710834503174,
      "learning_rate": 7.584960672379257e-05,
      "loss": 0.1712,
      "step": 42600
    },
    {
      "epoch": 0.5989177032820296,
      "grad_norm": 0.34591057896614075,
      "learning_rate": 7.580450301200845e-05,
      "loss": 0.1806,
      "step": 42610
    },
    {
      "epoch": 0.5990582612973505,
      "grad_norm": 0.46998921036720276,
      "learning_rate": 7.575940452789171e-05,
      "loss": 0.1556,
      "step": 42620
    },
    {
      "epoch": 0.5991988193126713,
      "grad_norm": 0.37321600317955017,
      "learning_rate": 7.571431128118627e-05,
      "loss": 0.1842,
      "step": 42630
    },
    {
      "epoch": 0.5993393773279921,
      "grad_norm": 0.34320640563964844,
      "learning_rate": 7.566922328163499e-05,
      "loss": 0.1618,
      "step": 42640
    },
    {
      "epoch": 0.5994799353433129,
      "grad_norm": 0.5494560599327087,
      "learning_rate": 7.562414053897952e-05,
      "loss": 0.2126,
      "step": 42650
    },
    {
      "epoch": 0.5996204933586338,
      "grad_norm": 0.4479246139526367,
      "learning_rate": 7.557906306296046e-05,
      "loss": 0.1835,
      "step": 42660
    },
    {
      "epoch": 0.5997610513739546,
      "grad_norm": 0.451515793800354,
      "learning_rate": 7.553399086331719e-05,
      "loss": 0.2045,
      "step": 42670
    },
    {
      "epoch": 0.5999016093892754,
      "grad_norm": 0.6131281852722168,
      "learning_rate": 7.5488923949788e-05,
      "loss": 0.1606,
      "step": 42680
    },
    {
      "epoch": 0.6000421674045963,
      "grad_norm": 0.36279651522636414,
      "learning_rate": 7.544386233211003e-05,
      "loss": 0.1586,
      "step": 42690
    },
    {
      "epoch": 0.600182725419917,
      "grad_norm": 0.3585605025291443,
      "learning_rate": 7.539880602001927e-05,
      "loss": 0.165,
      "step": 42700
    },
    {
      "epoch": 0.6003232834352379,
      "grad_norm": 0.3063223361968994,
      "learning_rate": 7.535375502325057e-05,
      "loss": 0.1623,
      "step": 42710
    },
    {
      "epoch": 0.6004638414505588,
      "grad_norm": 0.3708120882511139,
      "learning_rate": 7.530870935153762e-05,
      "loss": 0.1526,
      "step": 42720
    },
    {
      "epoch": 0.6006043994658795,
      "grad_norm": 0.34393540024757385,
      "learning_rate": 7.5263669014613e-05,
      "loss": 0.1814,
      "step": 42730
    },
    {
      "epoch": 0.6007449574812004,
      "grad_norm": 0.5064597129821777,
      "learning_rate": 7.521863402220802e-05,
      "loss": 0.1827,
      "step": 42740
    },
    {
      "epoch": 0.6008855154965211,
      "grad_norm": 0.41531017422676086,
      "learning_rate": 7.517360438405298e-05,
      "loss": 0.1799,
      "step": 42750
    },
    {
      "epoch": 0.601026073511842,
      "grad_norm": 0.35172322392463684,
      "learning_rate": 7.512858010987697e-05,
      "loss": 0.1679,
      "step": 42760
    },
    {
      "epoch": 0.6011666315271629,
      "grad_norm": 0.27190056443214417,
      "learning_rate": 7.508356120940787e-05,
      "loss": 0.1644,
      "step": 42770
    },
    {
      "epoch": 0.6013071895424836,
      "grad_norm": 0.2985529601573944,
      "learning_rate": 7.503854769237249e-05,
      "loss": 0.1759,
      "step": 42780
    },
    {
      "epoch": 0.6014477475578045,
      "grad_norm": 0.3518924415111542,
      "learning_rate": 7.499353956849639e-05,
      "loss": 0.19,
      "step": 42790
    },
    {
      "epoch": 0.6015883055731253,
      "grad_norm": 0.4358084797859192,
      "learning_rate": 7.494853684750402e-05,
      "loss": 0.1802,
      "step": 42800
    },
    {
      "epoch": 0.6017288635884461,
      "grad_norm": 0.4318670630455017,
      "learning_rate": 7.490353953911863e-05,
      "loss": 0.1691,
      "step": 42810
    },
    {
      "epoch": 0.601869421603767,
      "grad_norm": 0.33650708198547363,
      "learning_rate": 7.485854765306235e-05,
      "loss": 0.184,
      "step": 42820
    },
    {
      "epoch": 0.6020099796190878,
      "grad_norm": 0.3853902518749237,
      "learning_rate": 7.481356119905607e-05,
      "loss": 0.1895,
      "step": 42830
    },
    {
      "epoch": 0.6021505376344086,
      "grad_norm": 0.32648152112960815,
      "learning_rate": 7.476858018681957e-05,
      "loss": 0.1454,
      "step": 42840
    },
    {
      "epoch": 0.6022910956497294,
      "grad_norm": 0.3527396023273468,
      "learning_rate": 7.472360462607141e-05,
      "loss": 0.1506,
      "step": 42850
    },
    {
      "epoch": 0.6024316536650502,
      "grad_norm": 0.38165199756622314,
      "learning_rate": 7.4678634526529e-05,
      "loss": 0.1669,
      "step": 42860
    },
    {
      "epoch": 0.6025722116803711,
      "grad_norm": 0.3815620243549347,
      "learning_rate": 7.463366989790852e-05,
      "loss": 0.1751,
      "step": 42870
    },
    {
      "epoch": 0.6027127696956919,
      "grad_norm": 0.4380703270435333,
      "learning_rate": 7.458871074992506e-05,
      "loss": 0.1818,
      "step": 42880
    },
    {
      "epoch": 0.6028533277110127,
      "grad_norm": 0.33199450373649597,
      "learning_rate": 7.454375709229244e-05,
      "loss": 0.1681,
      "step": 42890
    },
    {
      "epoch": 0.6029938857263335,
      "grad_norm": 0.24013389647006989,
      "learning_rate": 7.449880893472333e-05,
      "loss": 0.1786,
      "step": 42900
    },
    {
      "epoch": 0.6031344437416544,
      "grad_norm": 0.3959920406341553,
      "learning_rate": 7.445386628692921e-05,
      "loss": 0.1861,
      "step": 42910
    },
    {
      "epoch": 0.6032750017569752,
      "grad_norm": 0.41576969623565674,
      "learning_rate": 7.440892915862036e-05,
      "loss": 0.1805,
      "step": 42920
    },
    {
      "epoch": 0.603415559772296,
      "grad_norm": 0.47591328620910645,
      "learning_rate": 7.43639975595059e-05,
      "loss": 0.1849,
      "step": 42930
    },
    {
      "epoch": 0.6035561177876169,
      "grad_norm": 0.43059951066970825,
      "learning_rate": 7.431907149929369e-05,
      "loss": 0.1777,
      "step": 42940
    },
    {
      "epoch": 0.6036966758029376,
      "grad_norm": 0.3965088725090027,
      "learning_rate": 7.427415098769045e-05,
      "loss": 0.1807,
      "step": 42950
    },
    {
      "epoch": 0.6038372338182585,
      "grad_norm": 0.34018439054489136,
      "learning_rate": 7.422923603440168e-05,
      "loss": 0.1672,
      "step": 42960
    },
    {
      "epoch": 0.6039777918335794,
      "grad_norm": 0.2599017024040222,
      "learning_rate": 7.418432664913171e-05,
      "loss": 0.171,
      "step": 42970
    },
    {
      "epoch": 0.6041183498489001,
      "grad_norm": 0.43904367089271545,
      "learning_rate": 7.413942284158362e-05,
      "loss": 0.1837,
      "step": 42980
    },
    {
      "epoch": 0.604258907864221,
      "grad_norm": 0.4200136363506317,
      "learning_rate": 7.40945246214593e-05,
      "loss": 0.1575,
      "step": 42990
    },
    {
      "epoch": 0.6043994658795417,
      "grad_norm": 0.4445408582687378,
      "learning_rate": 7.404963199845942e-05,
      "loss": 0.1594,
      "step": 43000
    },
    {
      "epoch": 0.6043994658795417,
      "eval_chrf": 83.57044035800844,
      "eval_loss": 0.36899685859680176,
      "eval_runtime": 164.275,
      "eval_samples_per_second": 0.609,
      "eval_steps_per_second": 0.024,
      "step": 43000
    },
    {
      "epoch": 0.6045400238948626,
      "grad_norm": 0.3644404113292694,
      "learning_rate": 7.400474498228351e-05,
      "loss": 0.1529,
      "step": 43010
    },
    {
      "epoch": 0.6046805819101835,
      "grad_norm": 0.3766258955001831,
      "learning_rate": 7.395986358262978e-05,
      "loss": 0.1922,
      "step": 43020
    },
    {
      "epoch": 0.6048211399255042,
      "grad_norm": 0.36604171991348267,
      "learning_rate": 7.391498780919532e-05,
      "loss": 0.1608,
      "step": 43030
    },
    {
      "epoch": 0.6049616979408251,
      "grad_norm": 0.38928326964378357,
      "learning_rate": 7.387011767167597e-05,
      "loss": 0.1642,
      "step": 43040
    },
    {
      "epoch": 0.6051022559561459,
      "grad_norm": 0.24062176048755646,
      "learning_rate": 7.382525317976634e-05,
      "loss": 0.161,
      "step": 43050
    },
    {
      "epoch": 0.6052428139714667,
      "grad_norm": 0.33095383644104004,
      "learning_rate": 7.378039434315983e-05,
      "loss": 0.1712,
      "step": 43060
    },
    {
      "epoch": 0.6053833719867876,
      "grad_norm": 0.3314269781112671,
      "learning_rate": 7.37355411715486e-05,
      "loss": 0.1719,
      "step": 43070
    },
    {
      "epoch": 0.6055239300021084,
      "grad_norm": 0.4529399573802948,
      "learning_rate": 7.369069367462362e-05,
      "loss": 0.1719,
      "step": 43080
    },
    {
      "epoch": 0.6056644880174292,
      "grad_norm": 0.25623270869255066,
      "learning_rate": 7.364585186207464e-05,
      "loss": 0.1728,
      "step": 43090
    },
    {
      "epoch": 0.60580504603275,
      "grad_norm": 0.4020390808582306,
      "learning_rate": 7.360101574359014e-05,
      "loss": 0.1715,
      "step": 43100
    },
    {
      "epoch": 0.6059456040480709,
      "grad_norm": 0.3411688506603241,
      "learning_rate": 7.355618532885738e-05,
      "loss": 0.1634,
      "step": 43110
    },
    {
      "epoch": 0.6060861620633917,
      "grad_norm": 0.34853067994117737,
      "learning_rate": 7.351136062756241e-05,
      "loss": 0.1778,
      "step": 43120
    },
    {
      "epoch": 0.6062267200787125,
      "grad_norm": 0.4465995132923126,
      "learning_rate": 7.346654164939002e-05,
      "loss": 0.1649,
      "step": 43130
    },
    {
      "epoch": 0.6063672780940333,
      "grad_norm": 0.37591084837913513,
      "learning_rate": 7.342172840402381e-05,
      "loss": 0.164,
      "step": 43140
    },
    {
      "epoch": 0.6065078361093541,
      "grad_norm": 0.4197112023830414,
      "learning_rate": 7.337692090114607e-05,
      "loss": 0.2018,
      "step": 43150
    },
    {
      "epoch": 0.606648394124675,
      "grad_norm": 0.27792155742645264,
      "learning_rate": 7.333211915043791e-05,
      "loss": 0.1351,
      "step": 43160
    },
    {
      "epoch": 0.6067889521399957,
      "grad_norm": 0.2809070646762848,
      "learning_rate": 7.328732316157918e-05,
      "loss": 0.1657,
      "step": 43170
    },
    {
      "epoch": 0.6069295101553166,
      "grad_norm": 0.3960578739643097,
      "learning_rate": 7.324253294424845e-05,
      "loss": 0.1856,
      "step": 43180
    },
    {
      "epoch": 0.6070700681706375,
      "grad_norm": 0.39426571130752563,
      "learning_rate": 7.319774850812312e-05,
      "loss": 0.1759,
      "step": 43190
    },
    {
      "epoch": 0.6072106261859582,
      "grad_norm": 0.39615365862846375,
      "learning_rate": 7.315296986287922e-05,
      "loss": 0.1794,
      "step": 43200
    },
    {
      "epoch": 0.6073511842012791,
      "grad_norm": 0.42621809244155884,
      "learning_rate": 7.310819701819168e-05,
      "loss": 0.1744,
      "step": 43210
    },
    {
      "epoch": 0.6074917422165999,
      "grad_norm": 0.5943285822868347,
      "learning_rate": 7.306342998373406e-05,
      "loss": 0.178,
      "step": 43220
    },
    {
      "epoch": 0.6076323002319207,
      "grad_norm": 0.4223845601081848,
      "learning_rate": 7.30186687691787e-05,
      "loss": 0.1743,
      "step": 43230
    },
    {
      "epoch": 0.6077728582472416,
      "grad_norm": 0.3097108006477356,
      "learning_rate": 7.297391338419671e-05,
      "loss": 0.1815,
      "step": 43240
    },
    {
      "epoch": 0.6079134162625623,
      "grad_norm": 0.3980135917663574,
      "learning_rate": 7.292916383845789e-05,
      "loss": 0.1806,
      "step": 43250
    },
    {
      "epoch": 0.6080539742778832,
      "grad_norm": 0.44541212916374207,
      "learning_rate": 7.288442014163084e-05,
      "loss": 0.1992,
      "step": 43260
    },
    {
      "epoch": 0.608194532293204,
      "grad_norm": 0.3464401066303253,
      "learning_rate": 7.283968230338282e-05,
      "loss": 0.1724,
      "step": 43270
    },
    {
      "epoch": 0.6083350903085248,
      "grad_norm": 0.44827163219451904,
      "learning_rate": 7.279495033337988e-05,
      "loss": 0.1543,
      "step": 43280
    },
    {
      "epoch": 0.6084756483238457,
      "grad_norm": 0.4815289378166199,
      "learning_rate": 7.275022424128681e-05,
      "loss": 0.1797,
      "step": 43290
    },
    {
      "epoch": 0.6086162063391665,
      "grad_norm": 0.35351064801216125,
      "learning_rate": 7.270550403676708e-05,
      "loss": 0.1675,
      "step": 43300
    },
    {
      "epoch": 0.6087567643544873,
      "grad_norm": 0.46254482865333557,
      "learning_rate": 7.266078972948292e-05,
      "loss": 0.1866,
      "step": 43310
    },
    {
      "epoch": 0.6088973223698081,
      "grad_norm": 0.5490948557853699,
      "learning_rate": 7.26160813290953e-05,
      "loss": 0.1682,
      "step": 43320
    },
    {
      "epoch": 0.609037880385129,
      "grad_norm": 0.31148210167884827,
      "learning_rate": 7.257137884526388e-05,
      "loss": 0.1824,
      "step": 43330
    },
    {
      "epoch": 0.6091784384004498,
      "grad_norm": 0.47011902928352356,
      "learning_rate": 7.252668228764708e-05,
      "loss": 0.179,
      "step": 43340
    },
    {
      "epoch": 0.6093189964157706,
      "grad_norm": 0.3612685799598694,
      "learning_rate": 7.248199166590197e-05,
      "loss": 0.1729,
      "step": 43350
    },
    {
      "epoch": 0.6094595544310915,
      "grad_norm": 0.36253541707992554,
      "learning_rate": 7.24373069896844e-05,
      "loss": 0.1797,
      "step": 43360
    },
    {
      "epoch": 0.6096001124464122,
      "grad_norm": 0.41139015555381775,
      "learning_rate": 7.239262826864896e-05,
      "loss": 0.171,
      "step": 43370
    },
    {
      "epoch": 0.6097406704617331,
      "grad_norm": 0.39285191893577576,
      "learning_rate": 7.234795551244888e-05,
      "loss": 0.1687,
      "step": 43380
    },
    {
      "epoch": 0.609881228477054,
      "grad_norm": 0.3595815598964691,
      "learning_rate": 7.230328873073613e-05,
      "loss": 0.1706,
      "step": 43390
    },
    {
      "epoch": 0.6100217864923747,
      "grad_norm": 0.3913259506225586,
      "learning_rate": 7.225862793316144e-05,
      "loss": 0.1642,
      "step": 43400
    },
    {
      "epoch": 0.6101623445076956,
      "grad_norm": 0.39592108130455017,
      "learning_rate": 7.221397312937413e-05,
      "loss": 0.2039,
      "step": 43410
    },
    {
      "epoch": 0.6103029025230163,
      "grad_norm": 0.4641571640968323,
      "learning_rate": 7.21693243290223e-05,
      "loss": 0.1731,
      "step": 43420
    },
    {
      "epoch": 0.6104434605383372,
      "grad_norm": 0.29540741443634033,
      "learning_rate": 7.21246815417528e-05,
      "loss": 0.1505,
      "step": 43430
    },
    {
      "epoch": 0.6105840185536581,
      "grad_norm": 0.4817897379398346,
      "learning_rate": 7.20800447772111e-05,
      "loss": 0.1769,
      "step": 43440
    },
    {
      "epoch": 0.6107245765689788,
      "grad_norm": 0.4189295768737793,
      "learning_rate": 7.20354140450414e-05,
      "loss": 0.1536,
      "step": 43450
    },
    {
      "epoch": 0.6108651345842997,
      "grad_norm": 0.49806585907936096,
      "learning_rate": 7.199078935488658e-05,
      "loss": 0.1635,
      "step": 43460
    },
    {
      "epoch": 0.6110056925996205,
      "grad_norm": 0.3175036609172821,
      "learning_rate": 7.194617071638827e-05,
      "loss": 0.1414,
      "step": 43470
    },
    {
      "epoch": 0.6111462506149413,
      "grad_norm": 0.3762895464897156,
      "learning_rate": 7.190155813918671e-05,
      "loss": 0.1642,
      "step": 43480
    },
    {
      "epoch": 0.6112868086302622,
      "grad_norm": 0.410321444272995,
      "learning_rate": 7.185695163292089e-05,
      "loss": 0.161,
      "step": 43490
    },
    {
      "epoch": 0.611427366645583,
      "grad_norm": 0.4362190067768097,
      "learning_rate": 7.181235120722847e-05,
      "loss": 0.1781,
      "step": 43500
    },
    {
      "epoch": 0.6115679246609038,
      "grad_norm": 0.4020962417125702,
      "learning_rate": 7.17677568717458e-05,
      "loss": 0.1691,
      "step": 43510
    },
    {
      "epoch": 0.6117084826762246,
      "grad_norm": 0.4284510314464569,
      "learning_rate": 7.172316863610789e-05,
      "loss": 0.1645,
      "step": 43520
    },
    {
      "epoch": 0.6118490406915454,
      "grad_norm": 0.346312940120697,
      "learning_rate": 7.167858650994848e-05,
      "loss": 0.1721,
      "step": 43530
    },
    {
      "epoch": 0.6119895987068663,
      "grad_norm": 0.4400968551635742,
      "learning_rate": 7.163401050289994e-05,
      "loss": 0.1648,
      "step": 43540
    },
    {
      "epoch": 0.6121301567221871,
      "grad_norm": 0.34509405493736267,
      "learning_rate": 7.158944062459335e-05,
      "loss": 0.1639,
      "step": 43550
    },
    {
      "epoch": 0.6122707147375079,
      "grad_norm": 0.4468303322792053,
      "learning_rate": 7.154487688465848e-05,
      "loss": 0.1668,
      "step": 43560
    },
    {
      "epoch": 0.6124112727528287,
      "grad_norm": 0.36385270953178406,
      "learning_rate": 7.150031929272369e-05,
      "loss": 0.1767,
      "step": 43570
    },
    {
      "epoch": 0.6125518307681496,
      "grad_norm": 0.48590368032455444,
      "learning_rate": 7.145576785841613e-05,
      "loss": 0.1792,
      "step": 43580
    },
    {
      "epoch": 0.6126923887834704,
      "grad_norm": 0.3635001480579376,
      "learning_rate": 7.141122259136153e-05,
      "loss": 0.1671,
      "step": 43590
    },
    {
      "epoch": 0.6128329467987912,
      "grad_norm": 0.3440009653568268,
      "learning_rate": 7.136668350118434e-05,
      "loss": 0.1436,
      "step": 43600
    },
    {
      "epoch": 0.6129735048141121,
      "grad_norm": 0.48527681827545166,
      "learning_rate": 7.132215059750764e-05,
      "loss": 0.1812,
      "step": 43610
    },
    {
      "epoch": 0.6131140628294328,
      "grad_norm": 0.33638450503349304,
      "learning_rate": 7.12776238899532e-05,
      "loss": 0.182,
      "step": 43620
    },
    {
      "epoch": 0.6132546208447537,
      "grad_norm": 0.3818179965019226,
      "learning_rate": 7.12331033881414e-05,
      "loss": 0.1757,
      "step": 43630
    },
    {
      "epoch": 0.6133951788600744,
      "grad_norm": 0.4085322916507721,
      "learning_rate": 7.118858910169138e-05,
      "loss": 0.1645,
      "step": 43640
    },
    {
      "epoch": 0.6135357368753953,
      "grad_norm": 0.3118310272693634,
      "learning_rate": 7.114408104022082e-05,
      "loss": 0.1631,
      "step": 43650
    },
    {
      "epoch": 0.6136762948907162,
      "grad_norm": 0.3122931718826294,
      "learning_rate": 7.109957921334614e-05,
      "loss": 0.1626,
      "step": 43660
    },
    {
      "epoch": 0.6138168529060369,
      "grad_norm": 0.34794265031814575,
      "learning_rate": 7.105508363068238e-05,
      "loss": 0.1772,
      "step": 43670
    },
    {
      "epoch": 0.6139574109213578,
      "grad_norm": 0.3584219515323639,
      "learning_rate": 7.101059430184322e-05,
      "loss": 0.1818,
      "step": 43680
    },
    {
      "epoch": 0.6140979689366786,
      "grad_norm": 0.3252063989639282,
      "learning_rate": 7.096611123644104e-05,
      "loss": 0.1755,
      "step": 43690
    },
    {
      "epoch": 0.6142385269519994,
      "grad_norm": 0.4239749610424042,
      "learning_rate": 7.092163444408677e-05,
      "loss": 0.1489,
      "step": 43700
    },
    {
      "epoch": 0.6143790849673203,
      "grad_norm": 0.3646966218948364,
      "learning_rate": 7.087716393439009e-05,
      "loss": 0.1814,
      "step": 43710
    },
    {
      "epoch": 0.6145196429826411,
      "grad_norm": 0.3831283748149872,
      "learning_rate": 7.083269971695926e-05,
      "loss": 0.1858,
      "step": 43720
    },
    {
      "epoch": 0.6146602009979619,
      "grad_norm": 0.342099130153656,
      "learning_rate": 7.07882418014012e-05,
      "loss": 0.1801,
      "step": 43730
    },
    {
      "epoch": 0.6148007590132827,
      "grad_norm": 0.4391648471355438,
      "learning_rate": 7.074379019732147e-05,
      "loss": 0.165,
      "step": 43740
    },
    {
      "epoch": 0.6149413170286036,
      "grad_norm": 0.3105049133300781,
      "learning_rate": 7.069934491432426e-05,
      "loss": 0.1685,
      "step": 43750
    },
    {
      "epoch": 0.6150818750439244,
      "grad_norm": 0.3759618103504181,
      "learning_rate": 7.065490596201239e-05,
      "loss": 0.1519,
      "step": 43760
    },
    {
      "epoch": 0.6152224330592452,
      "grad_norm": 0.4365192651748657,
      "learning_rate": 7.061047334998733e-05,
      "loss": 0.1562,
      "step": 43770
    },
    {
      "epoch": 0.615362991074566,
      "grad_norm": 0.4824438691139221,
      "learning_rate": 7.056604708784915e-05,
      "loss": 0.1674,
      "step": 43780
    },
    {
      "epoch": 0.6155035490898868,
      "grad_norm": 0.40726444125175476,
      "learning_rate": 7.05216271851966e-05,
      "loss": 0.1879,
      "step": 43790
    },
    {
      "epoch": 0.6156441071052077,
      "grad_norm": 0.42368054389953613,
      "learning_rate": 7.047721365162699e-05,
      "loss": 0.1888,
      "step": 43800
    },
    {
      "epoch": 0.6157846651205285,
      "grad_norm": 0.4291306734085083,
      "learning_rate": 7.043280649673628e-05,
      "loss": 0.1765,
      "step": 43810
    },
    {
      "epoch": 0.6159252231358493,
      "grad_norm": 0.2989252805709839,
      "learning_rate": 7.038840573011909e-05,
      "loss": 0.1806,
      "step": 43820
    },
    {
      "epoch": 0.6160657811511702,
      "grad_norm": 0.43362295627593994,
      "learning_rate": 7.034401136136861e-05,
      "loss": 0.1802,
      "step": 43830
    },
    {
      "epoch": 0.6162063391664909,
      "grad_norm": 0.33333566784858704,
      "learning_rate": 7.029962340007669e-05,
      "loss": 0.1729,
      "step": 43840
    },
    {
      "epoch": 0.6163468971818118,
      "grad_norm": 0.4206397235393524,
      "learning_rate": 7.025524185583372e-05,
      "loss": 0.164,
      "step": 43850
    },
    {
      "epoch": 0.6164874551971327,
      "grad_norm": 0.28481581807136536,
      "learning_rate": 7.02108667382288e-05,
      "loss": 0.1713,
      "step": 43860
    },
    {
      "epoch": 0.6166280132124534,
      "grad_norm": 0.43364596366882324,
      "learning_rate": 7.01664980568496e-05,
      "loss": 0.1958,
      "step": 43870
    },
    {
      "epoch": 0.6167685712277743,
      "grad_norm": 0.4091656804084778,
      "learning_rate": 7.012213582128237e-05,
      "loss": 0.1409,
      "step": 43880
    },
    {
      "epoch": 0.616909129243095,
      "grad_norm": 0.4062401354312897,
      "learning_rate": 7.007778004111198e-05,
      "loss": 0.1775,
      "step": 43890
    },
    {
      "epoch": 0.6170496872584159,
      "grad_norm": 0.39458316564559937,
      "learning_rate": 7.003343072592197e-05,
      "loss": 0.1877,
      "step": 43900
    },
    {
      "epoch": 0.6171902452737368,
      "grad_norm": 0.5522444248199463,
      "learning_rate": 6.998908788529439e-05,
      "loss": 0.1857,
      "step": 43910
    },
    {
      "epoch": 0.6173308032890575,
      "grad_norm": 0.4108028709888458,
      "learning_rate": 6.994475152880993e-05,
      "loss": 0.1999,
      "step": 43920
    },
    {
      "epoch": 0.6174713613043784,
      "grad_norm": 0.3741578459739685,
      "learning_rate": 6.990042166604791e-05,
      "loss": 0.1678,
      "step": 43930
    },
    {
      "epoch": 0.6176119193196992,
      "grad_norm": 0.2947121262550354,
      "learning_rate": 6.985609830658621e-05,
      "loss": 0.1519,
      "step": 43940
    },
    {
      "epoch": 0.61775247733502,
      "grad_norm": 0.4122995138168335,
      "learning_rate": 6.98117814600013e-05,
      "loss": 0.1519,
      "step": 43950
    },
    {
      "epoch": 0.6178930353503409,
      "grad_norm": 0.3719595670700073,
      "learning_rate": 6.976747113586827e-05,
      "loss": 0.1843,
      "step": 43960
    },
    {
      "epoch": 0.6180335933656617,
      "grad_norm": 0.2966946065425873,
      "learning_rate": 6.972316734376078e-05,
      "loss": 0.1717,
      "step": 43970
    },
    {
      "epoch": 0.6181741513809825,
      "grad_norm": 0.3694175183773041,
      "learning_rate": 6.967887009325108e-05,
      "loss": 0.1816,
      "step": 43980
    },
    {
      "epoch": 0.6183147093963033,
      "grad_norm": 0.4243808388710022,
      "learning_rate": 6.963457939391002e-05,
      "loss": 0.1524,
      "step": 43990
    },
    {
      "epoch": 0.6184552674116242,
      "grad_norm": 0.3591419756412506,
      "learning_rate": 6.959029525530704e-05,
      "loss": 0.1861,
      "step": 44000
    },
    {
      "epoch": 0.6184552674116242,
      "eval_chrf": 77.63238475835136,
      "eval_loss": 0.3709178864955902,
      "eval_runtime": 327.2596,
      "eval_samples_per_second": 0.306,
      "eval_steps_per_second": 0.012,
      "step": 44000
    },
    {
      "epoch": 0.618595825426945,
      "grad_norm": 0.3396461009979248,
      "learning_rate": 6.954601768701013e-05,
      "loss": 0.2045,
      "step": 44010
    },
    {
      "epoch": 0.6187363834422658,
      "grad_norm": 0.430901437997818,
      "learning_rate": 6.95017466985859e-05,
      "loss": 0.1812,
      "step": 44020
    },
    {
      "epoch": 0.6188769414575866,
      "grad_norm": 0.56734699010849,
      "learning_rate": 6.945748229959948e-05,
      "loss": 0.1798,
      "step": 44030
    },
    {
      "epoch": 0.6190174994729074,
      "grad_norm": 0.426201730966568,
      "learning_rate": 6.941322449961465e-05,
      "loss": 0.1879,
      "step": 44040
    },
    {
      "epoch": 0.6191580574882283,
      "grad_norm": 0.27214428782463074,
      "learning_rate": 6.936897330819372e-05,
      "loss": 0.1441,
      "step": 44050
    },
    {
      "epoch": 0.6192986155035491,
      "grad_norm": 0.550829291343689,
      "learning_rate": 6.932472873489758e-05,
      "loss": 0.1682,
      "step": 44060
    },
    {
      "epoch": 0.6194391735188699,
      "grad_norm": 0.3202691376209259,
      "learning_rate": 6.928049078928569e-05,
      "loss": 0.1624,
      "step": 44070
    },
    {
      "epoch": 0.6195797315341908,
      "grad_norm": 0.30396291613578796,
      "learning_rate": 6.92362594809161e-05,
      "loss": 0.1768,
      "step": 44080
    },
    {
      "epoch": 0.6197202895495115,
      "grad_norm": 0.2941432297229767,
      "learning_rate": 6.919203481934536e-05,
      "loss": 0.1744,
      "step": 44090
    },
    {
      "epoch": 0.6198608475648324,
      "grad_norm": 0.49695345759391785,
      "learning_rate": 6.914781681412866e-05,
      "loss": 0.1689,
      "step": 44100
    },
    {
      "epoch": 0.6200014055801532,
      "grad_norm": 0.35379379987716675,
      "learning_rate": 6.910360547481971e-05,
      "loss": 0.1465,
      "step": 44110
    },
    {
      "epoch": 0.620141963595474,
      "grad_norm": 0.43449699878692627,
      "learning_rate": 6.90594008109708e-05,
      "loss": 0.1933,
      "step": 44120
    },
    {
      "epoch": 0.6202825216107949,
      "grad_norm": 0.3357935845851898,
      "learning_rate": 6.901520283213277e-05,
      "loss": 0.1513,
      "step": 44130
    },
    {
      "epoch": 0.6204230796261156,
      "grad_norm": 0.3763403296470642,
      "learning_rate": 6.897101154785499e-05,
      "loss": 0.1661,
      "step": 44140
    },
    {
      "epoch": 0.6205636376414365,
      "grad_norm": 0.40166908502578735,
      "learning_rate": 6.892682696768545e-05,
      "loss": 0.175,
      "step": 44150
    },
    {
      "epoch": 0.6207041956567573,
      "grad_norm": 0.42627114057540894,
      "learning_rate": 6.88826491011706e-05,
      "loss": 0.2007,
      "step": 44160
    },
    {
      "epoch": 0.6208447536720781,
      "grad_norm": 0.3611266016960144,
      "learning_rate": 6.883847795785552e-05,
      "loss": 0.1616,
      "step": 44170
    },
    {
      "epoch": 0.620985311687399,
      "grad_norm": 0.5528517365455627,
      "learning_rate": 6.879431354728378e-05,
      "loss": 0.1873,
      "step": 44180
    },
    {
      "epoch": 0.6211258697027198,
      "grad_norm": 0.37437355518341064,
      "learning_rate": 6.875015587899755e-05,
      "loss": 0.1667,
      "step": 44190
    },
    {
      "epoch": 0.6212664277180406,
      "grad_norm": 0.27261611819267273,
      "learning_rate": 6.870600496253752e-05,
      "loss": 0.1907,
      "step": 44200
    },
    {
      "epoch": 0.6214069857333614,
      "grad_norm": 0.30701377987861633,
      "learning_rate": 6.86618608074429e-05,
      "loss": 0.1683,
      "step": 44210
    },
    {
      "epoch": 0.6215475437486823,
      "grad_norm": 0.4434259235858917,
      "learning_rate": 6.861772342325144e-05,
      "loss": 0.1506,
      "step": 44220
    },
    {
      "epoch": 0.6216881017640031,
      "grad_norm": 0.48307085037231445,
      "learning_rate": 6.857359281949946e-05,
      "loss": 0.1743,
      "step": 44230
    },
    {
      "epoch": 0.6218286597793239,
      "grad_norm": 0.43294084072113037,
      "learning_rate": 6.852946900572179e-05,
      "loss": 0.1671,
      "step": 44240
    },
    {
      "epoch": 0.6219692177946448,
      "grad_norm": 0.49979066848754883,
      "learning_rate": 6.848535199145179e-05,
      "loss": 0.21,
      "step": 44250
    },
    {
      "epoch": 0.6221097758099655,
      "grad_norm": 0.4027746319770813,
      "learning_rate": 6.84412417862214e-05,
      "loss": 0.1741,
      "step": 44260
    },
    {
      "epoch": 0.6222503338252864,
      "grad_norm": 0.5052631497383118,
      "learning_rate": 6.8397138399561e-05,
      "loss": 0.1953,
      "step": 44270
    },
    {
      "epoch": 0.6223908918406073,
      "grad_norm": 0.34749501943588257,
      "learning_rate": 6.835304184099958e-05,
      "loss": 0.1759,
      "step": 44280
    },
    {
      "epoch": 0.622531449855928,
      "grad_norm": 0.5017284750938416,
      "learning_rate": 6.83089521200646e-05,
      "loss": 0.1775,
      "step": 44290
    },
    {
      "epoch": 0.6226720078712489,
      "grad_norm": 0.40785008668899536,
      "learning_rate": 6.826486924628209e-05,
      "loss": 0.1786,
      "step": 44300
    },
    {
      "epoch": 0.6228125658865696,
      "grad_norm": 0.46408775448799133,
      "learning_rate": 6.822079322917656e-05,
      "loss": 0.1796,
      "step": 44310
    },
    {
      "epoch": 0.6229531239018905,
      "grad_norm": 0.3966285288333893,
      "learning_rate": 6.817672407827103e-05,
      "loss": 0.1831,
      "step": 44320
    },
    {
      "epoch": 0.6230936819172114,
      "grad_norm": 0.45978012681007385,
      "learning_rate": 6.813266180308711e-05,
      "loss": 0.1756,
      "step": 44330
    },
    {
      "epoch": 0.6232342399325321,
      "grad_norm": 0.4069959223270416,
      "learning_rate": 6.808860641314481e-05,
      "loss": 0.163,
      "step": 44340
    },
    {
      "epoch": 0.623374797947853,
      "grad_norm": 0.4341197609901428,
      "learning_rate": 6.804455791796277e-05,
      "loss": 0.1828,
      "step": 44350
    },
    {
      "epoch": 0.6235153559631738,
      "grad_norm": 0.406414657831192,
      "learning_rate": 6.800051632705807e-05,
      "loss": 0.1813,
      "step": 44360
    },
    {
      "epoch": 0.6236559139784946,
      "grad_norm": 0.3163243532180786,
      "learning_rate": 6.79564816499463e-05,
      "loss": 0.1713,
      "step": 44370
    },
    {
      "epoch": 0.6237964719938155,
      "grad_norm": 0.3498212397098541,
      "learning_rate": 6.791245389614159e-05,
      "loss": 0.158,
      "step": 44380
    },
    {
      "epoch": 0.6239370300091363,
      "grad_norm": 0.48263484239578247,
      "learning_rate": 6.786843307515655e-05,
      "loss": 0.1524,
      "step": 44390
    },
    {
      "epoch": 0.6240775880244571,
      "grad_norm": 0.4155692756175995,
      "learning_rate": 6.782441919650229e-05,
      "loss": 0.1712,
      "step": 44400
    },
    {
      "epoch": 0.6242181460397779,
      "grad_norm": 0.35613593459129333,
      "learning_rate": 6.778041226968843e-05,
      "loss": 0.1769,
      "step": 44410
    },
    {
      "epoch": 0.6243587040550987,
      "grad_norm": 0.434958815574646,
      "learning_rate": 6.77364123042231e-05,
      "loss": 0.177,
      "step": 44420
    },
    {
      "epoch": 0.6244992620704196,
      "grad_norm": 0.44713306427001953,
      "learning_rate": 6.76924193096129e-05,
      "loss": 0.1778,
      "step": 44430
    },
    {
      "epoch": 0.6246398200857404,
      "grad_norm": 0.3982578217983246,
      "learning_rate": 6.764843329536293e-05,
      "loss": 0.1426,
      "step": 44440
    },
    {
      "epoch": 0.6247803781010612,
      "grad_norm": 0.3782222867012024,
      "learning_rate": 6.760445427097678e-05,
      "loss": 0.1883,
      "step": 44450
    },
    {
      "epoch": 0.624920936116382,
      "grad_norm": 0.3814140856266022,
      "learning_rate": 6.756048224595656e-05,
      "loss": 0.1851,
      "step": 44460
    },
    {
      "epoch": 0.6250614941317029,
      "grad_norm": 0.3756048381328583,
      "learning_rate": 6.751651722980283e-05,
      "loss": 0.1625,
      "step": 44470
    },
    {
      "epoch": 0.6252020521470237,
      "grad_norm": 0.3470192551612854,
      "learning_rate": 6.747255923201465e-05,
      "loss": 0.1604,
      "step": 44480
    },
    {
      "epoch": 0.6253426101623445,
      "grad_norm": 0.44632065296173096,
      "learning_rate": 6.742860826208956e-05,
      "loss": 0.202,
      "step": 44490
    },
    {
      "epoch": 0.6254831681776654,
      "grad_norm": 0.24534133076667786,
      "learning_rate": 6.73846643295236e-05,
      "loss": 0.1708,
      "step": 44500
    },
    {
      "epoch": 0.6256237261929861,
      "grad_norm": 0.40814921259880066,
      "learning_rate": 6.734072744381129e-05,
      "loss": 0.1739,
      "step": 44510
    },
    {
      "epoch": 0.625764284208307,
      "grad_norm": 0.4645785987377167,
      "learning_rate": 6.729679761444553e-05,
      "loss": 0.1732,
      "step": 44520
    },
    {
      "epoch": 0.6259048422236279,
      "grad_norm": 0.4125882685184479,
      "learning_rate": 6.725287485091789e-05,
      "loss": 0.1726,
      "step": 44530
    },
    {
      "epoch": 0.6260454002389486,
      "grad_norm": 0.3475102186203003,
      "learning_rate": 6.720895916271821e-05,
      "loss": 0.1671,
      "step": 44540
    },
    {
      "epoch": 0.6261859582542695,
      "grad_norm": 0.4807623028755188,
      "learning_rate": 6.716505055933496e-05,
      "loss": 0.1769,
      "step": 44550
    },
    {
      "epoch": 0.6263265162695902,
      "grad_norm": 0.46677112579345703,
      "learning_rate": 6.712114905025498e-05,
      "loss": 0.1957,
      "step": 44560
    },
    {
      "epoch": 0.6264670742849111,
      "grad_norm": 0.33543965220451355,
      "learning_rate": 6.707725464496358e-05,
      "loss": 0.1889,
      "step": 44570
    },
    {
      "epoch": 0.626607632300232,
      "grad_norm": 0.30811792612075806,
      "learning_rate": 6.70333673529446e-05,
      "loss": 0.1725,
      "step": 44580
    },
    {
      "epoch": 0.6267481903155527,
      "grad_norm": 0.43859946727752686,
      "learning_rate": 6.698948718368033e-05,
      "loss": 0.1575,
      "step": 44590
    },
    {
      "epoch": 0.6268887483308736,
      "grad_norm": 0.3518104553222656,
      "learning_rate": 6.694561414665145e-05,
      "loss": 0.1745,
      "step": 44600
    },
    {
      "epoch": 0.6270293063461944,
      "grad_norm": 0.30487966537475586,
      "learning_rate": 6.690174825133717e-05,
      "loss": 0.1695,
      "step": 44610
    },
    {
      "epoch": 0.6271698643615152,
      "grad_norm": 0.49385231733322144,
      "learning_rate": 6.685788950721512e-05,
      "loss": 0.1716,
      "step": 44620
    },
    {
      "epoch": 0.627310422376836,
      "grad_norm": 0.4758278727531433,
      "learning_rate": 6.681403792376142e-05,
      "loss": 0.1658,
      "step": 44630
    },
    {
      "epoch": 0.6274509803921569,
      "grad_norm": 0.4662957489490509,
      "learning_rate": 6.677019351045061e-05,
      "loss": 0.1829,
      "step": 44640
    },
    {
      "epoch": 0.6275915384074777,
      "grad_norm": 0.41147729754447937,
      "learning_rate": 6.67263562767557e-05,
      "loss": 0.1669,
      "step": 44650
    },
    {
      "epoch": 0.6277320964227985,
      "grad_norm": 0.36372533440589905,
      "learning_rate": 6.668252623214812e-05,
      "loss": 0.1452,
      "step": 44660
    },
    {
      "epoch": 0.6278726544381193,
      "grad_norm": 0.41888707876205444,
      "learning_rate": 6.663870338609779e-05,
      "loss": 0.1693,
      "step": 44670
    },
    {
      "epoch": 0.6280132124534401,
      "grad_norm": 0.35983574390411377,
      "learning_rate": 6.659488774807304e-05,
      "loss": 0.1639,
      "step": 44680
    },
    {
      "epoch": 0.628153770468761,
      "grad_norm": 0.3690883219242096,
      "learning_rate": 6.655107932754068e-05,
      "loss": 0.1577,
      "step": 44690
    },
    {
      "epoch": 0.6282943284840818,
      "grad_norm": 0.4388582408428192,
      "learning_rate": 6.65072781339659e-05,
      "loss": 0.1529,
      "step": 44700
    },
    {
      "epoch": 0.6284348864994026,
      "grad_norm": 0.42833447456359863,
      "learning_rate": 6.646348417681239e-05,
      "loss": 0.1642,
      "step": 44710
    },
    {
      "epoch": 0.6285754445147235,
      "grad_norm": 0.2447792887687683,
      "learning_rate": 6.641969746554225e-05,
      "loss": 0.1692,
      "step": 44720
    },
    {
      "epoch": 0.6287160025300442,
      "grad_norm": 0.33198803663253784,
      "learning_rate": 6.637591800961598e-05,
      "loss": 0.1718,
      "step": 44730
    },
    {
      "epoch": 0.6288565605453651,
      "grad_norm": 0.2678699195384979,
      "learning_rate": 6.63321458184926e-05,
      "loss": 0.1855,
      "step": 44740
    },
    {
      "epoch": 0.628997118560686,
      "grad_norm": 0.32302653789520264,
      "learning_rate": 6.628838090162955e-05,
      "loss": 0.1678,
      "step": 44750
    },
    {
      "epoch": 0.6291376765760067,
      "grad_norm": 0.35951146483421326,
      "learning_rate": 6.624462326848252e-05,
      "loss": 0.1565,
      "step": 44760
    },
    {
      "epoch": 0.6292782345913276,
      "grad_norm": 0.3611389100551605,
      "learning_rate": 6.620087292850584e-05,
      "loss": 0.1743,
      "step": 44770
    },
    {
      "epoch": 0.6294187926066483,
      "grad_norm": 0.41236069798469543,
      "learning_rate": 6.615712989115219e-05,
      "loss": 0.1638,
      "step": 44780
    },
    {
      "epoch": 0.6295593506219692,
      "grad_norm": 0.30151498317718506,
      "learning_rate": 6.611339416587265e-05,
      "loss": 0.1568,
      "step": 44790
    },
    {
      "epoch": 0.6296999086372901,
      "grad_norm": 0.35278046131134033,
      "learning_rate": 6.606966576211678e-05,
      "loss": 0.1721,
      "step": 44800
    },
    {
      "epoch": 0.6298404666526108,
      "grad_norm": 0.5340271592140198,
      "learning_rate": 6.602594468933245e-05,
      "loss": 0.1837,
      "step": 44810
    },
    {
      "epoch": 0.6299810246679317,
      "grad_norm": 0.24998606741428375,
      "learning_rate": 6.598223095696608e-05,
      "loss": 0.1609,
      "step": 44820
    },
    {
      "epoch": 0.6301215826832525,
      "grad_norm": 0.3351782262325287,
      "learning_rate": 6.593852457446242e-05,
      "loss": 0.1543,
      "step": 44830
    },
    {
      "epoch": 0.6302621406985733,
      "grad_norm": 0.3094533681869507,
      "learning_rate": 6.589482555126461e-05,
      "loss": 0.1765,
      "step": 44840
    },
    {
      "epoch": 0.6304026987138942,
      "grad_norm": 0.4538188576698303,
      "learning_rate": 6.58511338968143e-05,
      "loss": 0.1815,
      "step": 44850
    },
    {
      "epoch": 0.630543256729215,
      "grad_norm": 0.47430428862571716,
      "learning_rate": 6.580744962055145e-05,
      "loss": 0.1754,
      "step": 44860
    },
    {
      "epoch": 0.6306838147445358,
      "grad_norm": 0.3575490117073059,
      "learning_rate": 6.576377273191447e-05,
      "loss": 0.1666,
      "step": 44870
    },
    {
      "epoch": 0.6308243727598566,
      "grad_norm": 0.4652651250362396,
      "learning_rate": 6.572010324034016e-05,
      "loss": 0.18,
      "step": 44880
    },
    {
      "epoch": 0.6309649307751775,
      "grad_norm": 0.25143569707870483,
      "learning_rate": 6.567644115526374e-05,
      "loss": 0.1638,
      "step": 44890
    },
    {
      "epoch": 0.6311054887904983,
      "grad_norm": 0.40408599376678467,
      "learning_rate": 6.563278648611884e-05,
      "loss": 0.1695,
      "step": 44900
    },
    {
      "epoch": 0.6312460468058191,
      "grad_norm": 0.5499407052993774,
      "learning_rate": 6.558913924233743e-05,
      "loss": 0.1978,
      "step": 44910
    },
    {
      "epoch": 0.63138660482114,
      "grad_norm": 0.38111111521720886,
      "learning_rate": 6.554549943334992e-05,
      "loss": 0.1811,
      "step": 44920
    },
    {
      "epoch": 0.6315271628364607,
      "grad_norm": 0.3579551875591278,
      "learning_rate": 6.550186706858513e-05,
      "loss": 0.172,
      "step": 44930
    },
    {
      "epoch": 0.6316677208517816,
      "grad_norm": 0.34631818532943726,
      "learning_rate": 6.54582421574702e-05,
      "loss": 0.1627,
      "step": 44940
    },
    {
      "epoch": 0.6318082788671024,
      "grad_norm": 0.4256238043308258,
      "learning_rate": 6.541462470943076e-05,
      "loss": 0.1887,
      "step": 44950
    },
    {
      "epoch": 0.6319488368824232,
      "grad_norm": 0.4322136640548706,
      "learning_rate": 6.537101473389075e-05,
      "loss": 0.1734,
      "step": 44960
    },
    {
      "epoch": 0.6320893948977441,
      "grad_norm": 0.3775302469730377,
      "learning_rate": 6.532741224027252e-05,
      "loss": 0.1997,
      "step": 44970
    },
    {
      "epoch": 0.6322299529130648,
      "grad_norm": 0.34045374393463135,
      "learning_rate": 6.528381723799682e-05,
      "loss": 0.1798,
      "step": 44980
    },
    {
      "epoch": 0.6323705109283857,
      "grad_norm": 0.30288296937942505,
      "learning_rate": 6.524022973648273e-05,
      "loss": 0.1743,
      "step": 44990
    },
    {
      "epoch": 0.6325110689437066,
      "grad_norm": 0.35040807723999023,
      "learning_rate": 6.519664974514778e-05,
      "loss": 0.1677,
      "step": 45000
    },
    {
      "epoch": 0.6325110689437066,
      "eval_chrf": 80.21349446857397,
      "eval_loss": 0.36095374822616577,
      "eval_runtime": 256.1043,
      "eval_samples_per_second": 0.39,
      "eval_steps_per_second": 0.016,
      "step": 45000
    },
    {
      "epoch": 0.6326516269590273,
      "grad_norm": 0.31671202182769775,
      "learning_rate": 6.515307727340782e-05,
      "loss": 0.1726,
      "step": 45010
    },
    {
      "epoch": 0.6327921849743482,
      "grad_norm": 0.2751798629760742,
      "learning_rate": 6.510951233067711e-05,
      "loss": 0.1577,
      "step": 45020
    },
    {
      "epoch": 0.632932742989669,
      "grad_norm": 0.34134364128112793,
      "learning_rate": 6.506595492636827e-05,
      "loss": 0.1356,
      "step": 45030
    },
    {
      "epoch": 0.6330733010049898,
      "grad_norm": 0.37104201316833496,
      "learning_rate": 6.502240506989228e-05,
      "loss": 0.1797,
      "step": 45040
    },
    {
      "epoch": 0.6332138590203107,
      "grad_norm": 0.3534786105155945,
      "learning_rate": 6.49788627706585e-05,
      "loss": 0.1464,
      "step": 45050
    },
    {
      "epoch": 0.6333544170356314,
      "grad_norm": 0.3754238784313202,
      "learning_rate": 6.493532803807468e-05,
      "loss": 0.1658,
      "step": 45060
    },
    {
      "epoch": 0.6334949750509523,
      "grad_norm": 0.3704669177532196,
      "learning_rate": 6.48918008815469e-05,
      "loss": 0.18,
      "step": 45070
    },
    {
      "epoch": 0.6336355330662731,
      "grad_norm": 0.5426111817359924,
      "learning_rate": 6.48482813104796e-05,
      "loss": 0.1762,
      "step": 45080
    },
    {
      "epoch": 0.6337760910815939,
      "grad_norm": 0.39849957823753357,
      "learning_rate": 6.480476933427562e-05,
      "loss": 0.166,
      "step": 45090
    },
    {
      "epoch": 0.6339166490969147,
      "grad_norm": 0.4700013995170593,
      "learning_rate": 6.476126496233611e-05,
      "loss": 0.1613,
      "step": 45100
    },
    {
      "epoch": 0.6340572071122356,
      "grad_norm": 0.30370020866394043,
      "learning_rate": 6.471776820406066e-05,
      "loss": 0.1578,
      "step": 45110
    },
    {
      "epoch": 0.6341977651275564,
      "grad_norm": 0.38362917304039,
      "learning_rate": 6.467427906884709e-05,
      "loss": 0.1761,
      "step": 45120
    },
    {
      "epoch": 0.6343383231428772,
      "grad_norm": 0.3319981098175049,
      "learning_rate": 6.46307975660917e-05,
      "loss": 0.1836,
      "step": 45130
    },
    {
      "epoch": 0.6344788811581981,
      "grad_norm": 0.3726891875267029,
      "learning_rate": 6.458732370518903e-05,
      "loss": 0.1803,
      "step": 45140
    },
    {
      "epoch": 0.6346194391735188,
      "grad_norm": 0.49176207184791565,
      "learning_rate": 6.454385749553205e-05,
      "loss": 0.1562,
      "step": 45150
    },
    {
      "epoch": 0.6347599971888397,
      "grad_norm": 0.4453577697277069,
      "learning_rate": 6.450039894651205e-05,
      "loss": 0.1788,
      "step": 45160
    },
    {
      "epoch": 0.6349005552041606,
      "grad_norm": 0.3720984160900116,
      "learning_rate": 6.445694806751868e-05,
      "loss": 0.1886,
      "step": 45170
    },
    {
      "epoch": 0.6350411132194813,
      "grad_norm": 0.37454113364219666,
      "learning_rate": 6.441350486793986e-05,
      "loss": 0.1628,
      "step": 45180
    },
    {
      "epoch": 0.6351816712348022,
      "grad_norm": 0.3713572025299072,
      "learning_rate": 6.437006935716195e-05,
      "loss": 0.1683,
      "step": 45190
    },
    {
      "epoch": 0.6353222292501229,
      "grad_norm": 0.3631324768066406,
      "learning_rate": 6.43266415445696e-05,
      "loss": 0.1829,
      "step": 45200
    },
    {
      "epoch": 0.6354627872654438,
      "grad_norm": 0.39969420433044434,
      "learning_rate": 6.42832214395458e-05,
      "loss": 0.1791,
      "step": 45210
    },
    {
      "epoch": 0.6356033452807647,
      "grad_norm": 0.35384640097618103,
      "learning_rate": 6.423980905147187e-05,
      "loss": 0.1793,
      "step": 45220
    },
    {
      "epoch": 0.6357439032960854,
      "grad_norm": 0.3424532115459442,
      "learning_rate": 6.419640438972748e-05,
      "loss": 0.1981,
      "step": 45230
    },
    {
      "epoch": 0.6358844613114063,
      "grad_norm": 0.3682054579257965,
      "learning_rate": 6.415300746369062e-05,
      "loss": 0.1606,
      "step": 45240
    },
    {
      "epoch": 0.6360250193267271,
      "grad_norm": 0.44392141699790955,
      "learning_rate": 6.410961828273758e-05,
      "loss": 0.1649,
      "step": 45250
    },
    {
      "epoch": 0.6361655773420479,
      "grad_norm": 0.43493548035621643,
      "learning_rate": 6.406623685624304e-05,
      "loss": 0.1514,
      "step": 45260
    },
    {
      "epoch": 0.6363061353573688,
      "grad_norm": 0.37985971570014954,
      "learning_rate": 6.402286319357997e-05,
      "loss": 0.1608,
      "step": 45270
    },
    {
      "epoch": 0.6364466933726896,
      "grad_norm": 0.3957103490829468,
      "learning_rate": 6.397949730411966e-05,
      "loss": 0.1817,
      "step": 45280
    },
    {
      "epoch": 0.6365872513880104,
      "grad_norm": 0.38979092240333557,
      "learning_rate": 6.393613919723168e-05,
      "loss": 0.188,
      "step": 45290
    },
    {
      "epoch": 0.6367278094033312,
      "grad_norm": 0.41796258091926575,
      "learning_rate": 6.389278888228402e-05,
      "loss": 0.1813,
      "step": 45300
    },
    {
      "epoch": 0.636868367418652,
      "grad_norm": 0.48477429151535034,
      "learning_rate": 6.38494463686429e-05,
      "loss": 0.1702,
      "step": 45310
    },
    {
      "epoch": 0.6370089254339729,
      "grad_norm": 0.45257049798965454,
      "learning_rate": 6.380611166567288e-05,
      "loss": 0.1675,
      "step": 45320
    },
    {
      "epoch": 0.6371494834492937,
      "grad_norm": 0.41892340779304504,
      "learning_rate": 6.376278478273685e-05,
      "loss": 0.1669,
      "step": 45330
    },
    {
      "epoch": 0.6372900414646145,
      "grad_norm": 0.36784976720809937,
      "learning_rate": 6.371946572919598e-05,
      "loss": 0.1787,
      "step": 45340
    },
    {
      "epoch": 0.6374305994799353,
      "grad_norm": 0.4263574779033661,
      "learning_rate": 6.367615451440976e-05,
      "loss": 0.1565,
      "step": 45350
    },
    {
      "epoch": 0.6375711574952562,
      "grad_norm": 0.6635227203369141,
      "learning_rate": 6.3632851147736e-05,
      "loss": 0.1909,
      "step": 45360
    },
    {
      "epoch": 0.637711715510577,
      "grad_norm": 0.5651838183403015,
      "learning_rate": 6.358955563853082e-05,
      "loss": 0.1652,
      "step": 45370
    },
    {
      "epoch": 0.6378522735258978,
      "grad_norm": 0.3701779544353485,
      "learning_rate": 6.35462679961486e-05,
      "loss": 0.1761,
      "step": 45380
    },
    {
      "epoch": 0.6379928315412187,
      "grad_norm": 0.30128878355026245,
      "learning_rate": 6.350298822994207e-05,
      "loss": 0.1413,
      "step": 45390
    },
    {
      "epoch": 0.6381333895565394,
      "grad_norm": 0.44980573654174805,
      "learning_rate": 6.345971634926222e-05,
      "loss": 0.1779,
      "step": 45400
    },
    {
      "epoch": 0.6382739475718603,
      "grad_norm": 0.3776741325855255,
      "learning_rate": 6.341645236345836e-05,
      "loss": 0.1827,
      "step": 45410
    },
    {
      "epoch": 0.6384145055871812,
      "grad_norm": 0.4075419306755066,
      "learning_rate": 6.337319628187813e-05,
      "loss": 0.1579,
      "step": 45420
    },
    {
      "epoch": 0.6385550636025019,
      "grad_norm": 0.45170167088508606,
      "learning_rate": 6.33299481138673e-05,
      "loss": 0.1862,
      "step": 45430
    },
    {
      "epoch": 0.6386956216178228,
      "grad_norm": 0.42756417393684387,
      "learning_rate": 6.328670786877013e-05,
      "loss": 0.1575,
      "step": 45440
    },
    {
      "epoch": 0.6388361796331435,
      "grad_norm": 0.4477543830871582,
      "learning_rate": 6.324347555592909e-05,
      "loss": 0.1777,
      "step": 45450
    },
    {
      "epoch": 0.6389767376484644,
      "grad_norm": 0.34204620122909546,
      "learning_rate": 6.320025118468491e-05,
      "loss": 0.166,
      "step": 45460
    },
    {
      "epoch": 0.6391172956637853,
      "grad_norm": 0.4834466874599457,
      "learning_rate": 6.315703476437662e-05,
      "loss": 0.1582,
      "step": 45470
    },
    {
      "epoch": 0.639257853679106,
      "grad_norm": 0.399301141500473,
      "learning_rate": 6.311382630434158e-05,
      "loss": 0.157,
      "step": 45480
    },
    {
      "epoch": 0.6393984116944269,
      "grad_norm": 0.43653443455696106,
      "learning_rate": 6.307062581391533e-05,
      "loss": 0.1949,
      "step": 45490
    },
    {
      "epoch": 0.6395389697097477,
      "grad_norm": 0.3309311866760254,
      "learning_rate": 6.30274333024318e-05,
      "loss": 0.1794,
      "step": 45500
    },
    {
      "epoch": 0.6396795277250685,
      "grad_norm": 0.3255656659603119,
      "learning_rate": 6.298424877922311e-05,
      "loss": 0.1794,
      "step": 45510
    },
    {
      "epoch": 0.6398200857403894,
      "grad_norm": 0.3199480473995209,
      "learning_rate": 6.294107225361969e-05,
      "loss": 0.1477,
      "step": 45520
    },
    {
      "epoch": 0.6399606437557102,
      "grad_norm": 0.33728545904159546,
      "learning_rate": 6.289790373495027e-05,
      "loss": 0.1704,
      "step": 45530
    },
    {
      "epoch": 0.640101201771031,
      "grad_norm": 0.41263148188591003,
      "learning_rate": 6.285474323254179e-05,
      "loss": 0.1498,
      "step": 45540
    },
    {
      "epoch": 0.6402417597863518,
      "grad_norm": 0.45552393794059753,
      "learning_rate": 6.281159075571948e-05,
      "loss": 0.1771,
      "step": 45550
    },
    {
      "epoch": 0.6403823178016727,
      "grad_norm": 0.3347470462322235,
      "learning_rate": 6.276844631380687e-05,
      "loss": 0.1627,
      "step": 45560
    },
    {
      "epoch": 0.6405228758169934,
      "grad_norm": 0.4019238352775574,
      "learning_rate": 6.27253099161257e-05,
      "loss": 0.1787,
      "step": 45570
    },
    {
      "epoch": 0.6406634338323143,
      "grad_norm": 0.40482115745544434,
      "learning_rate": 6.268218157199604e-05,
      "loss": 0.1616,
      "step": 45580
    },
    {
      "epoch": 0.6408039918476351,
      "grad_norm": 0.4995248019695282,
      "learning_rate": 6.263906129073614e-05,
      "loss": 0.1627,
      "step": 45590
    },
    {
      "epoch": 0.6409445498629559,
      "grad_norm": 0.471030592918396,
      "learning_rate": 6.259594908166255e-05,
      "loss": 0.1831,
      "step": 45600
    },
    {
      "epoch": 0.6410851078782768,
      "grad_norm": 0.32379043102264404,
      "learning_rate": 6.25528449540901e-05,
      "loss": 0.1687,
      "step": 45610
    },
    {
      "epoch": 0.6412256658935975,
      "grad_norm": 0.36522582173347473,
      "learning_rate": 6.250974891733182e-05,
      "loss": 0.1673,
      "step": 45620
    },
    {
      "epoch": 0.6413662239089184,
      "grad_norm": 0.2776927649974823,
      "learning_rate": 6.246666098069904e-05,
      "loss": 0.1723,
      "step": 45630
    },
    {
      "epoch": 0.6415067819242393,
      "grad_norm": 0.43431001901626587,
      "learning_rate": 6.242358115350128e-05,
      "loss": 0.1905,
      "step": 45640
    },
    {
      "epoch": 0.64164733993956,
      "grad_norm": 0.38113686442375183,
      "learning_rate": 6.23805094450464e-05,
      "loss": 0.1844,
      "step": 45650
    },
    {
      "epoch": 0.6417878979548809,
      "grad_norm": 0.369980126619339,
      "learning_rate": 6.233744586464042e-05,
      "loss": 0.1753,
      "step": 45660
    },
    {
      "epoch": 0.6419284559702017,
      "grad_norm": 0.4596186578273773,
      "learning_rate": 6.229439042158765e-05,
      "loss": 0.1886,
      "step": 45670
    },
    {
      "epoch": 0.6420690139855225,
      "grad_norm": 0.40157437324523926,
      "learning_rate": 6.225134312519061e-05,
      "loss": 0.1618,
      "step": 45680
    },
    {
      "epoch": 0.6422095720008434,
      "grad_norm": 0.4568968713283539,
      "learning_rate": 6.220830398475009e-05,
      "loss": 0.1746,
      "step": 45690
    },
    {
      "epoch": 0.6423501300161641,
      "grad_norm": 0.4889320731163025,
      "learning_rate": 6.216527300956512e-05,
      "loss": 0.1655,
      "step": 45700
    },
    {
      "epoch": 0.642490688031485,
      "grad_norm": 0.34276077151298523,
      "learning_rate": 6.212225020893292e-05,
      "loss": 0.1794,
      "step": 45710
    },
    {
      "epoch": 0.6426312460468058,
      "grad_norm": 0.35472655296325684,
      "learning_rate": 6.207923559214899e-05,
      "loss": 0.146,
      "step": 45720
    },
    {
      "epoch": 0.6427718040621266,
      "grad_norm": 0.36485761404037476,
      "learning_rate": 6.203622916850706e-05,
      "loss": 0.1783,
      "step": 45730
    },
    {
      "epoch": 0.6429123620774475,
      "grad_norm": 0.46038782596588135,
      "learning_rate": 6.199323094729905e-05,
      "loss": 0.1774,
      "step": 45740
    },
    {
      "epoch": 0.6430529200927683,
      "grad_norm": 0.3016781508922577,
      "learning_rate": 6.195024093781518e-05,
      "loss": 0.1374,
      "step": 45750
    },
    {
      "epoch": 0.6431934781080891,
      "grad_norm": 0.4437497556209564,
      "learning_rate": 6.19072591493438e-05,
      "loss": 0.165,
      "step": 45760
    },
    {
      "epoch": 0.6433340361234099,
      "grad_norm": 0.5395594239234924,
      "learning_rate": 6.186428559117156e-05,
      "loss": 0.1723,
      "step": 45770
    },
    {
      "epoch": 0.6434745941387308,
      "grad_norm": 0.3817726969718933,
      "learning_rate": 6.18213202725833e-05,
      "loss": 0.1703,
      "step": 45780
    },
    {
      "epoch": 0.6436151521540516,
      "grad_norm": 0.36339154839515686,
      "learning_rate": 6.17783632028621e-05,
      "loss": 0.1683,
      "step": 45790
    },
    {
      "epoch": 0.6437557101693724,
      "grad_norm": 0.4736917018890381,
      "learning_rate": 6.173541439128923e-05,
      "loss": 0.1922,
      "step": 45800
    },
    {
      "epoch": 0.6438962681846933,
      "grad_norm": 0.4476756453514099,
      "learning_rate": 6.169247384714417e-05,
      "loss": 0.1716,
      "step": 45810
    },
    {
      "epoch": 0.644036826200014,
      "grad_norm": 0.4236280918121338,
      "learning_rate": 6.164954157970467e-05,
      "loss": 0.1825,
      "step": 45820
    },
    {
      "epoch": 0.6441773842153349,
      "grad_norm": 0.4264964163303375,
      "learning_rate": 6.160661759824664e-05,
      "loss": 0.1683,
      "step": 45830
    },
    {
      "epoch": 0.6443179422306557,
      "grad_norm": 0.33900779485702515,
      "learning_rate": 6.156370191204421e-05,
      "loss": 0.1681,
      "step": 45840
    },
    {
      "epoch": 0.6444585002459765,
      "grad_norm": 0.3525594472885132,
      "learning_rate": 6.152079453036974e-05,
      "loss": 0.1855,
      "step": 45850
    },
    {
      "epoch": 0.6445990582612974,
      "grad_norm": 0.4360409379005432,
      "learning_rate": 6.147789546249375e-05,
      "loss": 0.1769,
      "step": 45860
    },
    {
      "epoch": 0.6447396162766181,
      "grad_norm": 0.45855650305747986,
      "learning_rate": 6.143500471768502e-05,
      "loss": 0.1655,
      "step": 45870
    },
    {
      "epoch": 0.644880174291939,
      "grad_norm": 0.4100463390350342,
      "learning_rate": 6.139212230521047e-05,
      "loss": 0.1807,
      "step": 45880
    },
    {
      "epoch": 0.6450207323072599,
      "grad_norm": 0.4619031548500061,
      "learning_rate": 6.13492482343353e-05,
      "loss": 0.1843,
      "step": 45890
    },
    {
      "epoch": 0.6451612903225806,
      "grad_norm": 0.3914530873298645,
      "learning_rate": 6.130638251432281e-05,
      "loss": 0.1784,
      "step": 45900
    },
    {
      "epoch": 0.6453018483379015,
      "grad_norm": 0.3291589021682739,
      "learning_rate": 6.126352515443457e-05,
      "loss": 0.171,
      "step": 45910
    },
    {
      "epoch": 0.6454424063532223,
      "grad_norm": 0.3202427327632904,
      "learning_rate": 6.122067616393035e-05,
      "loss": 0.1779,
      "step": 45920
    },
    {
      "epoch": 0.6455829643685431,
      "grad_norm": 0.37621304392814636,
      "learning_rate": 6.117783555206802e-05,
      "loss": 0.1576,
      "step": 45930
    },
    {
      "epoch": 0.645723522383864,
      "grad_norm": 0.3314257860183716,
      "learning_rate": 6.113500332810376e-05,
      "loss": 0.2038,
      "step": 45940
    },
    {
      "epoch": 0.6458640803991847,
      "grad_norm": 0.3285437822341919,
      "learning_rate": 6.109217950129183e-05,
      "loss": 0.1762,
      "step": 45950
    },
    {
      "epoch": 0.6460046384145056,
      "grad_norm": 0.3959253430366516,
      "learning_rate": 6.104936408088478e-05,
      "loss": 0.1664,
      "step": 45960
    },
    {
      "epoch": 0.6461451964298264,
      "grad_norm": 0.46875542402267456,
      "learning_rate": 6.100655707613323e-05,
      "loss": 0.1462,
      "step": 45970
    },
    {
      "epoch": 0.6462857544451472,
      "grad_norm": 0.3809106647968292,
      "learning_rate": 6.096375849628608e-05,
      "loss": 0.1649,
      "step": 45980
    },
    {
      "epoch": 0.6464263124604681,
      "grad_norm": 0.3985312581062317,
      "learning_rate": 6.092096835059038e-05,
      "loss": 0.1776,
      "step": 45990
    },
    {
      "epoch": 0.6465668704757889,
      "grad_norm": 0.32413896918296814,
      "learning_rate": 6.08781866482913e-05,
      "loss": 0.1569,
      "step": 46000
    },
    {
      "epoch": 0.6465668704757889,
      "eval_chrf": 82.38197137761941,
      "eval_loss": 0.3631060719490051,
      "eval_runtime": 276.4599,
      "eval_samples_per_second": 0.362,
      "eval_steps_per_second": 0.014,
      "step": 46000
    },
    {
      "epoch": 0.6467074284911097,
      "grad_norm": 0.4690287113189697,
      "learning_rate": 6.083541339863228e-05,
      "loss": 0.1541,
      "step": 46010
    },
    {
      "epoch": 0.6468479865064305,
      "grad_norm": 0.40396568179130554,
      "learning_rate": 6.0792648610854875e-05,
      "loss": 0.1696,
      "step": 46020
    },
    {
      "epoch": 0.6469885445217514,
      "grad_norm": 0.43039193749427795,
      "learning_rate": 6.0749892294198806e-05,
      "loss": 0.1696,
      "step": 46030
    },
    {
      "epoch": 0.6471291025370722,
      "grad_norm": 0.33346179127693176,
      "learning_rate": 6.0707144457902e-05,
      "loss": 0.1659,
      "step": 46040
    },
    {
      "epoch": 0.647269660552393,
      "grad_norm": 0.2777289152145386,
      "learning_rate": 6.066440511120053e-05,
      "loss": 0.1637,
      "step": 46050
    },
    {
      "epoch": 0.6474102185677139,
      "grad_norm": 0.35530346632003784,
      "learning_rate": 6.062167426332863e-05,
      "loss": 0.1926,
      "step": 46060
    },
    {
      "epoch": 0.6475507765830346,
      "grad_norm": 0.33783388137817383,
      "learning_rate": 6.0578951923518725e-05,
      "loss": 0.1676,
      "step": 46070
    },
    {
      "epoch": 0.6476913345983555,
      "grad_norm": 0.2598033845424652,
      "learning_rate": 6.053623810100135e-05,
      "loss": 0.186,
      "step": 46080
    },
    {
      "epoch": 0.6478318926136762,
      "grad_norm": 0.3518681228160858,
      "learning_rate": 6.0493532805005294e-05,
      "loss": 0.1792,
      "step": 46090
    },
    {
      "epoch": 0.6479724506289971,
      "grad_norm": 0.4977501630783081,
      "learning_rate": 6.045083604475739e-05,
      "loss": 0.1836,
      "step": 46100
    },
    {
      "epoch": 0.648113008644318,
      "grad_norm": 0.5222976803779602,
      "learning_rate": 6.040814782948267e-05,
      "loss": 0.1719,
      "step": 46110
    },
    {
      "epoch": 0.6482535666596387,
      "grad_norm": 0.5117230415344238,
      "learning_rate": 6.036546816840438e-05,
      "loss": 0.1886,
      "step": 46120
    },
    {
      "epoch": 0.6483941246749596,
      "grad_norm": 0.41326743364334106,
      "learning_rate": 6.032279707074382e-05,
      "loss": 0.182,
      "step": 46130
    },
    {
      "epoch": 0.6485346826902804,
      "grad_norm": 0.3465782701969147,
      "learning_rate": 6.028013454572051e-05,
      "loss": 0.1791,
      "step": 46140
    },
    {
      "epoch": 0.6486752407056012,
      "grad_norm": 0.44033685326576233,
      "learning_rate": 6.0237480602552085e-05,
      "loss": 0.214,
      "step": 46150
    },
    {
      "epoch": 0.6488157987209221,
      "grad_norm": 0.3798646926879883,
      "learning_rate": 6.019483525045435e-05,
      "loss": 0.1723,
      "step": 46160
    },
    {
      "epoch": 0.6489563567362429,
      "grad_norm": 0.49568066000938416,
      "learning_rate": 6.015219849864122e-05,
      "loss": 0.1687,
      "step": 46170
    },
    {
      "epoch": 0.6490969147515637,
      "grad_norm": 0.28596997261047363,
      "learning_rate": 6.010957035632479e-05,
      "loss": 0.174,
      "step": 46180
    },
    {
      "epoch": 0.6492374727668845,
      "grad_norm": 0.3705233037471771,
      "learning_rate": 6.0066950832715254e-05,
      "loss": 0.1724,
      "step": 46190
    },
    {
      "epoch": 0.6493780307822054,
      "grad_norm": 0.392262727022171,
      "learning_rate": 6.002433993702099e-05,
      "loss": 0.1595,
      "step": 46200
    },
    {
      "epoch": 0.6495185887975262,
      "grad_norm": 0.3774138391017914,
      "learning_rate": 5.998173767844847e-05,
      "loss": 0.1484,
      "step": 46210
    },
    {
      "epoch": 0.649659146812847,
      "grad_norm": 0.3614377975463867,
      "learning_rate": 5.993914406620231e-05,
      "loss": 0.1693,
      "step": 46220
    },
    {
      "epoch": 0.6497997048281678,
      "grad_norm": 0.37364089488983154,
      "learning_rate": 5.98965591094853e-05,
      "loss": 0.1599,
      "step": 46230
    },
    {
      "epoch": 0.6499402628434886,
      "grad_norm": 0.4179651439189911,
      "learning_rate": 5.98539828174983e-05,
      "loss": 0.1796,
      "step": 46240
    },
    {
      "epoch": 0.6500808208588095,
      "grad_norm": 0.38695430755615234,
      "learning_rate": 5.981141519944031e-05,
      "loss": 0.1643,
      "step": 46250
    },
    {
      "epoch": 0.6502213788741303,
      "grad_norm": 0.5926967263221741,
      "learning_rate": 5.97688562645085e-05,
      "loss": 0.1841,
      "step": 46260
    },
    {
      "epoch": 0.6503619368894511,
      "grad_norm": 0.3806653320789337,
      "learning_rate": 5.9726306021898106e-05,
      "loss": 0.1817,
      "step": 46270
    },
    {
      "epoch": 0.650502494904772,
      "grad_norm": 0.3245481550693512,
      "learning_rate": 5.9683764480802526e-05,
      "loss": 0.1733,
      "step": 46280
    },
    {
      "epoch": 0.6506430529200927,
      "grad_norm": 0.33144858479499817,
      "learning_rate": 5.964123165041327e-05,
      "loss": 0.1667,
      "step": 46290
    },
    {
      "epoch": 0.6507836109354136,
      "grad_norm": 0.41733354330062866,
      "learning_rate": 5.959870753991995e-05,
      "loss": 0.1931,
      "step": 46300
    },
    {
      "epoch": 0.6509241689507345,
      "grad_norm": 0.45859894156455994,
      "learning_rate": 5.955619215851031e-05,
      "loss": 0.166,
      "step": 46310
    },
    {
      "epoch": 0.6510647269660552,
      "grad_norm": 0.39740046858787537,
      "learning_rate": 5.95136855153702e-05,
      "loss": 0.1539,
      "step": 46320
    },
    {
      "epoch": 0.6512052849813761,
      "grad_norm": 0.4183885455131531,
      "learning_rate": 5.947118761968359e-05,
      "loss": 0.1568,
      "step": 46330
    },
    {
      "epoch": 0.6513458429966968,
      "grad_norm": 0.369146466255188,
      "learning_rate": 5.942869848063255e-05,
      "loss": 0.1804,
      "step": 46340
    },
    {
      "epoch": 0.6514864010120177,
      "grad_norm": 0.3931564390659332,
      "learning_rate": 5.9386218107397266e-05,
      "loss": 0.1845,
      "step": 46350
    },
    {
      "epoch": 0.6516269590273386,
      "grad_norm": 0.43380603194236755,
      "learning_rate": 5.9343746509156016e-05,
      "loss": 0.1913,
      "step": 46360
    },
    {
      "epoch": 0.6517675170426593,
      "grad_norm": 0.3476782441139221,
      "learning_rate": 5.9301283695085205e-05,
      "loss": 0.169,
      "step": 46370
    },
    {
      "epoch": 0.6519080750579802,
      "grad_norm": 0.40904736518859863,
      "learning_rate": 5.925882967435932e-05,
      "loss": 0.1661,
      "step": 46380
    },
    {
      "epoch": 0.652048633073301,
      "grad_norm": 0.3820328712463379,
      "learning_rate": 5.9216384456150966e-05,
      "loss": 0.1722,
      "step": 46390
    },
    {
      "epoch": 0.6521891910886218,
      "grad_norm": 0.427290141582489,
      "learning_rate": 5.9173948049630836e-05,
      "loss": 0.1763,
      "step": 46400
    },
    {
      "epoch": 0.6523297491039427,
      "grad_norm": 0.3932560682296753,
      "learning_rate": 5.913152046396771e-05,
      "loss": 0.1875,
      "step": 46410
    },
    {
      "epoch": 0.6524703071192635,
      "grad_norm": 0.37783360481262207,
      "learning_rate": 5.908910170832846e-05,
      "loss": 0.1606,
      "step": 46420
    },
    {
      "epoch": 0.6526108651345843,
      "grad_norm": 0.35456210374832153,
      "learning_rate": 5.9046691791878106e-05,
      "loss": 0.1847,
      "step": 46430
    },
    {
      "epoch": 0.6527514231499051,
      "grad_norm": 0.4274060130119324,
      "learning_rate": 5.9004290723779665e-05,
      "loss": 0.1705,
      "step": 46440
    },
    {
      "epoch": 0.652891981165226,
      "grad_norm": 0.41579386591911316,
      "learning_rate": 5.8961898513194335e-05,
      "loss": 0.1713,
      "step": 46450
    },
    {
      "epoch": 0.6530325391805468,
      "grad_norm": 0.41511207818984985,
      "learning_rate": 5.8919515169281315e-05,
      "loss": 0.1773,
      "step": 46460
    },
    {
      "epoch": 0.6531730971958676,
      "grad_norm": 0.3858306109905243,
      "learning_rate": 5.887714070119794e-05,
      "loss": 0.1742,
      "step": 46470
    },
    {
      "epoch": 0.6533136552111884,
      "grad_norm": 0.3599134087562561,
      "learning_rate": 5.883477511809965e-05,
      "loss": 0.1605,
      "step": 46480
    },
    {
      "epoch": 0.6534542132265092,
      "grad_norm": 0.4521912932395935,
      "learning_rate": 5.87924184291399e-05,
      "loss": 0.16,
      "step": 46490
    },
    {
      "epoch": 0.6535947712418301,
      "grad_norm": 0.4453141689300537,
      "learning_rate": 5.8750070643470245e-05,
      "loss": 0.1583,
      "step": 46500
    },
    {
      "epoch": 0.6537353292571509,
      "grad_norm": 0.34592360258102417,
      "learning_rate": 5.870773177024036e-05,
      "loss": 0.1864,
      "step": 46510
    },
    {
      "epoch": 0.6538758872724717,
      "grad_norm": 0.32892781496047974,
      "learning_rate": 5.8665401818597955e-05,
      "loss": 0.1605,
      "step": 46520
    },
    {
      "epoch": 0.6540164452877926,
      "grad_norm": 0.3231342136859894,
      "learning_rate": 5.86230807976888e-05,
      "loss": 0.1699,
      "step": 46530
    },
    {
      "epoch": 0.6541570033031133,
      "grad_norm": 0.45281505584716797,
      "learning_rate": 5.858076871665676e-05,
      "loss": 0.1741,
      "step": 46540
    },
    {
      "epoch": 0.6542975613184342,
      "grad_norm": 0.47716307640075684,
      "learning_rate": 5.8538465584643776e-05,
      "loss": 0.1722,
      "step": 46550
    },
    {
      "epoch": 0.654438119333755,
      "grad_norm": 0.38653871417045593,
      "learning_rate": 5.849617141078985e-05,
      "loss": 0.1508,
      "step": 46560
    },
    {
      "epoch": 0.6545786773490758,
      "grad_norm": 0.3510001003742218,
      "learning_rate": 5.8453886204233e-05,
      "loss": 0.1739,
      "step": 46570
    },
    {
      "epoch": 0.6547192353643967,
      "grad_norm": 0.39627891778945923,
      "learning_rate": 5.841160997410937e-05,
      "loss": 0.1895,
      "step": 46580
    },
    {
      "epoch": 0.6548597933797174,
      "grad_norm": 0.5046212077140808,
      "learning_rate": 5.8369342729553176e-05,
      "loss": 0.149,
      "step": 46590
    },
    {
      "epoch": 0.6550003513950383,
      "grad_norm": 0.3888484537601471,
      "learning_rate": 5.8327084479696593e-05,
      "loss": 0.1654,
      "step": 46600
    },
    {
      "epoch": 0.6551409094103591,
      "grad_norm": 0.2727382779121399,
      "learning_rate": 5.828483523366997e-05,
      "loss": 0.1564,
      "step": 46610
    },
    {
      "epoch": 0.6552814674256799,
      "grad_norm": 0.3435800075531006,
      "learning_rate": 5.824259500060162e-05,
      "loss": 0.17,
      "step": 46620
    },
    {
      "epoch": 0.6554220254410008,
      "grad_norm": 0.33127716183662415,
      "learning_rate": 5.8200363789617996e-05,
      "loss": 0.1951,
      "step": 46630
    },
    {
      "epoch": 0.6555625834563216,
      "grad_norm": 0.4050454795360565,
      "learning_rate": 5.815814160984349e-05,
      "loss": 0.153,
      "step": 46640
    },
    {
      "epoch": 0.6557031414716424,
      "grad_norm": 0.3367047607898712,
      "learning_rate": 5.811592847040067e-05,
      "loss": 0.1765,
      "step": 46650
    },
    {
      "epoch": 0.6558436994869632,
      "grad_norm": 0.4014679193496704,
      "learning_rate": 5.807372438041003e-05,
      "loss": 0.157,
      "step": 46660
    },
    {
      "epoch": 0.6559842575022841,
      "grad_norm": 0.32407405972480774,
      "learning_rate": 5.803152934899022e-05,
      "loss": 0.1822,
      "step": 46670
    },
    {
      "epoch": 0.6561248155176049,
      "grad_norm": 0.5178097486495972,
      "learning_rate": 5.798934338525781e-05,
      "loss": 0.1719,
      "step": 46680
    },
    {
      "epoch": 0.6562653735329257,
      "grad_norm": 0.36699798703193665,
      "learning_rate": 5.794716649832756e-05,
      "loss": 0.1617,
      "step": 46690
    },
    {
      "epoch": 0.6564059315482466,
      "grad_norm": 0.3077496588230133,
      "learning_rate": 5.790499869731212e-05,
      "loss": 0.1518,
      "step": 46700
    },
    {
      "epoch": 0.6565464895635673,
      "grad_norm": 0.393049418926239,
      "learning_rate": 5.786283999132229e-05,
      "loss": 0.1715,
      "step": 46710
    },
    {
      "epoch": 0.6566870475788882,
      "grad_norm": 0.3637532889842987,
      "learning_rate": 5.782069038946681e-05,
      "loss": 0.1926,
      "step": 46720
    },
    {
      "epoch": 0.656827605594209,
      "grad_norm": 0.3915221393108368,
      "learning_rate": 5.777854990085258e-05,
      "loss": 0.1512,
      "step": 46730
    },
    {
      "epoch": 0.6569681636095298,
      "grad_norm": 0.27357760071754456,
      "learning_rate": 5.773641853458436e-05,
      "loss": 0.1722,
      "step": 46740
    },
    {
      "epoch": 0.6571087216248507,
      "grad_norm": 0.49531319737434387,
      "learning_rate": 5.7694296299765106e-05,
      "loss": 0.1715,
      "step": 46750
    },
    {
      "epoch": 0.6572492796401714,
      "grad_norm": 0.3616010248661041,
      "learning_rate": 5.7652183205495726e-05,
      "loss": 0.153,
      "step": 46760
    },
    {
      "epoch": 0.6573898376554923,
      "grad_norm": 0.3260374963283539,
      "learning_rate": 5.761007926087507e-05,
      "loss": 0.1732,
      "step": 46770
    },
    {
      "epoch": 0.6575303956708132,
      "grad_norm": 0.3030858039855957,
      "learning_rate": 5.7567984475000205e-05,
      "loss": 0.1587,
      "step": 46780
    },
    {
      "epoch": 0.6576709536861339,
      "grad_norm": 0.29215386509895325,
      "learning_rate": 5.752589885696602e-05,
      "loss": 0.1574,
      "step": 46790
    },
    {
      "epoch": 0.6578115117014548,
      "grad_norm": 0.4393480122089386,
      "learning_rate": 5.7483822415865584e-05,
      "loss": 0.1534,
      "step": 46800
    },
    {
      "epoch": 0.6579520697167756,
      "grad_norm": 0.3719373345375061,
      "learning_rate": 5.7441755160789845e-05,
      "loss": 0.1771,
      "step": 46810
    },
    {
      "epoch": 0.6580926277320964,
      "grad_norm": 0.3512132465839386,
      "learning_rate": 5.739969710082792e-05,
      "loss": 0.1745,
      "step": 46820
    },
    {
      "epoch": 0.6582331857474173,
      "grad_norm": 0.42287617921829224,
      "learning_rate": 5.7357648245066756e-05,
      "loss": 0.1796,
      "step": 46830
    },
    {
      "epoch": 0.658373743762738,
      "grad_norm": 0.39335954189300537,
      "learning_rate": 5.731560860259151e-05,
      "loss": 0.1852,
      "step": 46840
    },
    {
      "epoch": 0.6585143017780589,
      "grad_norm": 0.29379117488861084,
      "learning_rate": 5.727357818248513e-05,
      "loss": 0.1512,
      "step": 46850
    },
    {
      "epoch": 0.6586548597933797,
      "grad_norm": 0.34768614172935486,
      "learning_rate": 5.723155699382881e-05,
      "loss": 0.1621,
      "step": 46860
    },
    {
      "epoch": 0.6587954178087005,
      "grad_norm": 0.4134244918823242,
      "learning_rate": 5.718954504570152e-05,
      "loss": 0.1711,
      "step": 46870
    },
    {
      "epoch": 0.6589359758240214,
      "grad_norm": 0.4068963825702667,
      "learning_rate": 5.714754234718044e-05,
      "loss": 0.2003,
      "step": 46880
    },
    {
      "epoch": 0.6590765338393422,
      "grad_norm": 0.3430961072444916,
      "learning_rate": 5.710554890734054e-05,
      "loss": 0.1627,
      "step": 46890
    },
    {
      "epoch": 0.659217091854663,
      "grad_norm": 0.4104665517807007,
      "learning_rate": 5.7063564735255025e-05,
      "loss": 0.155,
      "step": 46900
    },
    {
      "epoch": 0.6593576498699838,
      "grad_norm": 0.464220255613327,
      "learning_rate": 5.702158983999488e-05,
      "loss": 0.1849,
      "step": 46910
    },
    {
      "epoch": 0.6594982078853047,
      "grad_norm": 0.37425777316093445,
      "learning_rate": 5.697962423062925e-05,
      "loss": 0.1815,
      "step": 46920
    },
    {
      "epoch": 0.6596387659006255,
      "grad_norm": 0.3134264051914215,
      "learning_rate": 5.693766791622512e-05,
      "loss": 0.1735,
      "step": 46930
    },
    {
      "epoch": 0.6597793239159463,
      "grad_norm": 0.3093084990978241,
      "learning_rate": 5.689572090584765e-05,
      "loss": 0.1631,
      "step": 46940
    },
    {
      "epoch": 0.6599198819312672,
      "grad_norm": 0.4233850836753845,
      "learning_rate": 5.685378320855981e-05,
      "loss": 0.1952,
      "step": 46950
    },
    {
      "epoch": 0.6600604399465879,
      "grad_norm": 0.3595212697982788,
      "learning_rate": 5.68118548334227e-05,
      "loss": 0.179,
      "step": 46960
    },
    {
      "epoch": 0.6602009979619088,
      "grad_norm": 0.3647650480270386,
      "learning_rate": 5.676993578949527e-05,
      "loss": 0.1656,
      "step": 46970
    },
    {
      "epoch": 0.6603415559772297,
      "grad_norm": 0.4111272394657135,
      "learning_rate": 5.6728026085834606e-05,
      "loss": 0.2024,
      "step": 46980
    },
    {
      "epoch": 0.6604821139925504,
      "grad_norm": 0.48057103157043457,
      "learning_rate": 5.668612573149564e-05,
      "loss": 0.1677,
      "step": 46990
    },
    {
      "epoch": 0.6606226720078713,
      "grad_norm": 0.37313178181648254,
      "learning_rate": 5.664423473553141e-05,
      "loss": 0.1945,
      "step": 47000
    },
    {
      "epoch": 0.6606226720078713,
      "eval_chrf": 78.54251639053005,
      "eval_loss": 0.37721508741378784,
      "eval_runtime": 258.2804,
      "eval_samples_per_second": 0.387,
      "eval_steps_per_second": 0.015,
      "step": 47000
    },
    {
      "epoch": 0.660763230023192,
      "grad_norm": 0.39650222659111023,
      "learning_rate": 5.660235310699278e-05,
      "loss": 0.1679,
      "step": 47010
    },
    {
      "epoch": 0.6609037880385129,
      "grad_norm": 0.3072584867477417,
      "learning_rate": 5.6560480854928745e-05,
      "loss": 0.1588,
      "step": 47020
    },
    {
      "epoch": 0.6610443460538337,
      "grad_norm": 0.40000805258750916,
      "learning_rate": 5.6518617988386135e-05,
      "loss": 0.1671,
      "step": 47030
    },
    {
      "epoch": 0.6611849040691545,
      "grad_norm": 0.3673750162124634,
      "learning_rate": 5.647676451640991e-05,
      "loss": 0.1754,
      "step": 47040
    },
    {
      "epoch": 0.6613254620844754,
      "grad_norm": 0.2628951370716095,
      "learning_rate": 5.643492044804282e-05,
      "loss": 0.1567,
      "step": 47050
    },
    {
      "epoch": 0.6614660200997962,
      "grad_norm": 0.4595640301704407,
      "learning_rate": 5.639308579232576e-05,
      "loss": 0.1686,
      "step": 47060
    },
    {
      "epoch": 0.661606578115117,
      "grad_norm": 0.5201691389083862,
      "learning_rate": 5.63512605582974e-05,
      "loss": 0.1877,
      "step": 47070
    },
    {
      "epoch": 0.6617471361304378,
      "grad_norm": 0.39629697799682617,
      "learning_rate": 5.63094447549946e-05,
      "loss": 0.1544,
      "step": 47080
    },
    {
      "epoch": 0.6618876941457587,
      "grad_norm": 0.44209417700767517,
      "learning_rate": 5.626763839145196e-05,
      "loss": 0.1511,
      "step": 47090
    },
    {
      "epoch": 0.6620282521610795,
      "grad_norm": 0.4663390815258026,
      "learning_rate": 5.6225841476702226e-05,
      "loss": 0.1534,
      "step": 47100
    },
    {
      "epoch": 0.6621688101764003,
      "grad_norm": 0.38166019320487976,
      "learning_rate": 5.618405401977593e-05,
      "loss": 0.175,
      "step": 47110
    },
    {
      "epoch": 0.6623093681917211,
      "grad_norm": 0.2989104986190796,
      "learning_rate": 5.6142276029701746e-05,
      "loss": 0.1669,
      "step": 47120
    },
    {
      "epoch": 0.6624499262070419,
      "grad_norm": 0.28844520449638367,
      "learning_rate": 5.6100507515506103e-05,
      "loss": 0.1688,
      "step": 47130
    },
    {
      "epoch": 0.6625904842223628,
      "grad_norm": 0.3244578242301941,
      "learning_rate": 5.60587484862136e-05,
      "loss": 0.1666,
      "step": 47140
    },
    {
      "epoch": 0.6627310422376836,
      "grad_norm": 0.27320772409439087,
      "learning_rate": 5.6016998950846576e-05,
      "loss": 0.1865,
      "step": 47150
    },
    {
      "epoch": 0.6628716002530044,
      "grad_norm": 0.3454894423484802,
      "learning_rate": 5.5975258918425476e-05,
      "loss": 0.1937,
      "step": 47160
    },
    {
      "epoch": 0.6630121582683253,
      "grad_norm": 0.31500983238220215,
      "learning_rate": 5.593352839796858e-05,
      "loss": 0.1714,
      "step": 47170
    },
    {
      "epoch": 0.663152716283646,
      "grad_norm": 0.43122321367263794,
      "learning_rate": 5.589180739849223e-05,
      "loss": 0.1715,
      "step": 47180
    },
    {
      "epoch": 0.6632932742989669,
      "grad_norm": 0.2885819375514984,
      "learning_rate": 5.5850095929010573e-05,
      "loss": 0.1891,
      "step": 47190
    },
    {
      "epoch": 0.6634338323142878,
      "grad_norm": 0.2869061529636383,
      "learning_rate": 5.5808393998535835e-05,
      "loss": 0.1577,
      "step": 47200
    },
    {
      "epoch": 0.6635743903296085,
      "grad_norm": 0.3633897602558136,
      "learning_rate": 5.576670161607804e-05,
      "loss": 0.1766,
      "step": 47210
    },
    {
      "epoch": 0.6637149483449294,
      "grad_norm": 0.40465348958969116,
      "learning_rate": 5.5725018790645336e-05,
      "loss": 0.1677,
      "step": 47220
    },
    {
      "epoch": 0.6638555063602501,
      "grad_norm": 0.3205110728740692,
      "learning_rate": 5.5683345531243546e-05,
      "loss": 0.1826,
      "step": 47230
    },
    {
      "epoch": 0.663996064375571,
      "grad_norm": 0.46258682012557983,
      "learning_rate": 5.564168184687673e-05,
      "loss": 0.1679,
      "step": 47240
    },
    {
      "epoch": 0.6641366223908919,
      "grad_norm": 0.40163615345954895,
      "learning_rate": 5.560002774654658e-05,
      "loss": 0.1627,
      "step": 47250
    },
    {
      "epoch": 0.6642771804062126,
      "grad_norm": 0.37548112869262695,
      "learning_rate": 5.5558383239252974e-05,
      "loss": 0.1596,
      "step": 47260
    },
    {
      "epoch": 0.6644177384215335,
      "grad_norm": 0.5590593218803406,
      "learning_rate": 5.551674833399353e-05,
      "loss": 0.1625,
      "step": 47270
    },
    {
      "epoch": 0.6645582964368543,
      "grad_norm": 0.4699946939945221,
      "learning_rate": 5.5475123039763924e-05,
      "loss": 0.2086,
      "step": 47280
    },
    {
      "epoch": 0.6646988544521751,
      "grad_norm": 0.32331383228302,
      "learning_rate": 5.5433507365557624e-05,
      "loss": 0.1633,
      "step": 47290
    },
    {
      "epoch": 0.664839412467496,
      "grad_norm": 0.3101024925708771,
      "learning_rate": 5.539190132036618e-05,
      "loss": 0.1909,
      "step": 47300
    },
    {
      "epoch": 0.6649799704828168,
      "grad_norm": 0.40582162141799927,
      "learning_rate": 5.5350304913178884e-05,
      "loss": 0.1667,
      "step": 47310
    },
    {
      "epoch": 0.6651205284981376,
      "grad_norm": 0.41470903158187866,
      "learning_rate": 5.530871815298313e-05,
      "loss": 0.1964,
      "step": 47320
    },
    {
      "epoch": 0.6652610865134584,
      "grad_norm": 0.36652326583862305,
      "learning_rate": 5.526714104876403e-05,
      "loss": 0.1833,
      "step": 47330
    },
    {
      "epoch": 0.6654016445287793,
      "grad_norm": 0.4002378284931183,
      "learning_rate": 5.522557360950482e-05,
      "loss": 0.215,
      "step": 47340
    },
    {
      "epoch": 0.6655422025441001,
      "grad_norm": 0.38589349389076233,
      "learning_rate": 5.518401584418644e-05,
      "loss": 0.1753,
      "step": 47350
    },
    {
      "epoch": 0.6656827605594209,
      "grad_norm": 0.395865797996521,
      "learning_rate": 5.514246776178793e-05,
      "loss": 0.1807,
      "step": 47360
    },
    {
      "epoch": 0.6658233185747418,
      "grad_norm": 0.3177359104156494,
      "learning_rate": 5.5100929371286056e-05,
      "loss": 0.1795,
      "step": 47370
    },
    {
      "epoch": 0.6659638765900625,
      "grad_norm": 0.3520699143409729,
      "learning_rate": 5.505940068165569e-05,
      "loss": 0.1686,
      "step": 47380
    },
    {
      "epoch": 0.6661044346053834,
      "grad_norm": 0.331185519695282,
      "learning_rate": 5.501788170186939e-05,
      "loss": 0.1641,
      "step": 47390
    },
    {
      "epoch": 0.6662449926207042,
      "grad_norm": 0.36551252007484436,
      "learning_rate": 5.497637244089784e-05,
      "loss": 0.1515,
      "step": 47400
    },
    {
      "epoch": 0.666385550636025,
      "grad_norm": 0.4092820882797241,
      "learning_rate": 5.493487290770939e-05,
      "loss": 0.1857,
      "step": 47410
    },
    {
      "epoch": 0.6665261086513459,
      "grad_norm": 0.34370318055152893,
      "learning_rate": 5.489338311127052e-05,
      "loss": 0.1575,
      "step": 47420
    },
    {
      "epoch": 0.6666666666666666,
      "grad_norm": 0.3167739808559418,
      "learning_rate": 5.48519030605454e-05,
      "loss": 0.184,
      "step": 47430
    },
    {
      "epoch": 0.6668072246819875,
      "grad_norm": 0.40686434507369995,
      "learning_rate": 5.481043276449632e-05,
      "loss": 0.1823,
      "step": 47440
    },
    {
      "epoch": 0.6669477826973084,
      "grad_norm": 0.3953606188297272,
      "learning_rate": 5.476897223208317e-05,
      "loss": 0.1836,
      "step": 47450
    },
    {
      "epoch": 0.6670883407126291,
      "grad_norm": 0.4484312832355499,
      "learning_rate": 5.472752147226401e-05,
      "loss": 0.1475,
      "step": 47460
    },
    {
      "epoch": 0.66722889872795,
      "grad_norm": 0.41706743836402893,
      "learning_rate": 5.468608049399457e-05,
      "loss": 0.1721,
      "step": 47470
    },
    {
      "epoch": 0.6673694567432708,
      "grad_norm": 0.31754666566848755,
      "learning_rate": 5.46446493062287e-05,
      "loss": 0.15,
      "step": 47480
    },
    {
      "epoch": 0.6675100147585916,
      "grad_norm": 0.4365607500076294,
      "learning_rate": 5.4603227917917875e-05,
      "loss": 0.1654,
      "step": 47490
    },
    {
      "epoch": 0.6676505727739125,
      "grad_norm": 0.3824831545352936,
      "learning_rate": 5.4561816338011676e-05,
      "loss": 0.1589,
      "step": 47500
    },
    {
      "epoch": 0.6677911307892332,
      "grad_norm": 0.3625919818878174,
      "learning_rate": 5.452041457545739e-05,
      "loss": 0.1783,
      "step": 47510
    },
    {
      "epoch": 0.6679316888045541,
      "grad_norm": 0.5666170716285706,
      "learning_rate": 5.447902263920034e-05,
      "loss": 0.1971,
      "step": 47520
    },
    {
      "epoch": 0.6680722468198749,
      "grad_norm": 0.37668928503990173,
      "learning_rate": 5.4437640538183564e-05,
      "loss": 0.188,
      "step": 47530
    },
    {
      "epoch": 0.6682128048351957,
      "grad_norm": 0.41840583086013794,
      "learning_rate": 5.439626828134816e-05,
      "loss": 0.1758,
      "step": 47540
    },
    {
      "epoch": 0.6683533628505165,
      "grad_norm": 0.41168487071990967,
      "learning_rate": 5.435490587763288e-05,
      "loss": 0.1548,
      "step": 47550
    },
    {
      "epoch": 0.6684939208658374,
      "grad_norm": 0.49045082926750183,
      "learning_rate": 5.431355333597459e-05,
      "loss": 0.173,
      "step": 47560
    },
    {
      "epoch": 0.6686344788811582,
      "grad_norm": 0.3818732798099518,
      "learning_rate": 5.427221066530777e-05,
      "loss": 0.1629,
      "step": 47570
    },
    {
      "epoch": 0.668775036896479,
      "grad_norm": 0.39010149240493774,
      "learning_rate": 5.423087787456502e-05,
      "loss": 0.1632,
      "step": 47580
    },
    {
      "epoch": 0.6689155949117999,
      "grad_norm": 0.3519304692745209,
      "learning_rate": 5.418955497267657e-05,
      "loss": 0.1769,
      "step": 47590
    },
    {
      "epoch": 0.6690561529271206,
      "grad_norm": 0.49452972412109375,
      "learning_rate": 5.414824196857073e-05,
      "loss": 0.1943,
      "step": 47600
    },
    {
      "epoch": 0.6691967109424415,
      "grad_norm": 0.34571000933647156,
      "learning_rate": 5.4106938871173494e-05,
      "loss": 0.1466,
      "step": 47610
    },
    {
      "epoch": 0.6693372689577624,
      "grad_norm": 0.41626954078674316,
      "learning_rate": 5.406564568940884e-05,
      "loss": 0.1544,
      "step": 47620
    },
    {
      "epoch": 0.6694778269730831,
      "grad_norm": 0.47072553634643555,
      "learning_rate": 5.402436243219851e-05,
      "loss": 0.1589,
      "step": 47630
    },
    {
      "epoch": 0.669618384988404,
      "grad_norm": 0.414405882358551,
      "learning_rate": 5.398308910846217e-05,
      "loss": 0.1486,
      "step": 47640
    },
    {
      "epoch": 0.6697589430037247,
      "grad_norm": 0.3934914171695709,
      "learning_rate": 5.394182572711728e-05,
      "loss": 0.1696,
      "step": 47650
    },
    {
      "epoch": 0.6698995010190456,
      "grad_norm": 0.4239861071109772,
      "learning_rate": 5.3900572297079256e-05,
      "loss": 0.1819,
      "step": 47660
    },
    {
      "epoch": 0.6700400590343665,
      "grad_norm": 0.41308853030204773,
      "learning_rate": 5.385932882726121e-05,
      "loss": 0.1623,
      "step": 47670
    },
    {
      "epoch": 0.6701806170496872,
      "grad_norm": 0.3246447741985321,
      "learning_rate": 5.381809532657427e-05,
      "loss": 0.1765,
      "step": 47680
    },
    {
      "epoch": 0.6703211750650081,
      "grad_norm": 0.3457357883453369,
      "learning_rate": 5.377687180392723e-05,
      "loss": 0.1595,
      "step": 47690
    },
    {
      "epoch": 0.6704617330803289,
      "grad_norm": 0.34478646516799927,
      "learning_rate": 5.373565826822693e-05,
      "loss": 0.1643,
      "step": 47700
    },
    {
      "epoch": 0.6706022910956497,
      "grad_norm": 0.3674975633621216,
      "learning_rate": 5.369445472837786e-05,
      "loss": 0.1793,
      "step": 47710
    },
    {
      "epoch": 0.6707428491109706,
      "grad_norm": 0.3989918828010559,
      "learning_rate": 5.36532611932825e-05,
      "loss": 0.1775,
      "step": 47720
    },
    {
      "epoch": 0.6708834071262914,
      "grad_norm": 0.3287481367588043,
      "learning_rate": 5.3612077671841047e-05,
      "loss": 0.1665,
      "step": 47730
    },
    {
      "epoch": 0.6710239651416122,
      "grad_norm": 0.27225741744041443,
      "learning_rate": 5.357090417295165e-05,
      "loss": 0.1644,
      "step": 47740
    },
    {
      "epoch": 0.671164523156933,
      "grad_norm": 0.353710412979126,
      "learning_rate": 5.352974070551018e-05,
      "loss": 0.1814,
      "step": 47750
    },
    {
      "epoch": 0.6713050811722538,
      "grad_norm": 0.44788897037506104,
      "learning_rate": 5.348858727841045e-05,
      "loss": 0.1553,
      "step": 47760
    },
    {
      "epoch": 0.6714456391875747,
      "grad_norm": 0.3965194523334503,
      "learning_rate": 5.344744390054399e-05,
      "loss": 0.1557,
      "step": 47770
    },
    {
      "epoch": 0.6715861972028955,
      "grad_norm": 0.28312981128692627,
      "learning_rate": 5.340631058080029e-05,
      "loss": 0.1725,
      "step": 47780
    },
    {
      "epoch": 0.6717267552182163,
      "grad_norm": 0.4658781588077545,
      "learning_rate": 5.336518732806651e-05,
      "loss": 0.1649,
      "step": 47790
    },
    {
      "epoch": 0.6718673132335371,
      "grad_norm": 0.3657032251358032,
      "learning_rate": 5.332407415122782e-05,
      "loss": 0.1773,
      "step": 47800
    },
    {
      "epoch": 0.672007871248858,
      "grad_norm": 0.3435717225074768,
      "learning_rate": 5.3282971059167005e-05,
      "loss": 0.1594,
      "step": 47810
    },
    {
      "epoch": 0.6721484292641788,
      "grad_norm": 0.3758085072040558,
      "learning_rate": 5.3241878060764886e-05,
      "loss": 0.1731,
      "step": 47820
    },
    {
      "epoch": 0.6722889872794996,
      "grad_norm": 0.3348824977874756,
      "learning_rate": 5.3200795164899887e-05,
      "loss": 0.1649,
      "step": 47830
    },
    {
      "epoch": 0.6724295452948205,
      "grad_norm": 0.2984980642795563,
      "learning_rate": 5.315972238044847e-05,
      "loss": 0.1716,
      "step": 47840
    },
    {
      "epoch": 0.6725701033101412,
      "grad_norm": 0.46859562397003174,
      "learning_rate": 5.31186597162847e-05,
      "loss": 0.1545,
      "step": 47850
    },
    {
      "epoch": 0.6727106613254621,
      "grad_norm": 0.35370686650276184,
      "learning_rate": 5.307760718128065e-05,
      "loss": 0.1857,
      "step": 47860
    },
    {
      "epoch": 0.672851219340783,
      "grad_norm": 0.4333016574382782,
      "learning_rate": 5.303656478430601e-05,
      "loss": 0.1738,
      "step": 47870
    },
    {
      "epoch": 0.6729917773561037,
      "grad_norm": 0.39497295022010803,
      "learning_rate": 5.299553253422849e-05,
      "loss": 0.1655,
      "step": 47880
    },
    {
      "epoch": 0.6731323353714246,
      "grad_norm": 0.2927321791648865,
      "learning_rate": 5.2954510439913395e-05,
      "loss": 0.1507,
      "step": 47890
    },
    {
      "epoch": 0.6732728933867453,
      "grad_norm": 0.4232181906700134,
      "learning_rate": 5.2913498510224027e-05,
      "loss": 0.168,
      "step": 47900
    },
    {
      "epoch": 0.6734134514020662,
      "grad_norm": 0.3505028784275055,
      "learning_rate": 5.2872496754021304e-05,
      "loss": 0.1633,
      "step": 47910
    },
    {
      "epoch": 0.6735540094173871,
      "grad_norm": 0.30001300573349,
      "learning_rate": 5.2831505180164177e-05,
      "loss": 0.1564,
      "step": 47920
    },
    {
      "epoch": 0.6736945674327078,
      "grad_norm": 0.45110681653022766,
      "learning_rate": 5.279052379750913e-05,
      "loss": 0.1703,
      "step": 47930
    },
    {
      "epoch": 0.6738351254480287,
      "grad_norm": 0.3818683922290802,
      "learning_rate": 5.274955261491068e-05,
      "loss": 0.1792,
      "step": 47940
    },
    {
      "epoch": 0.6739756834633495,
      "grad_norm": 0.5145854353904724,
      "learning_rate": 5.270859164122097e-05,
      "loss": 0.1965,
      "step": 47950
    },
    {
      "epoch": 0.6741162414786703,
      "grad_norm": 0.4271232783794403,
      "learning_rate": 5.266764088529009e-05,
      "loss": 0.1831,
      "step": 47960
    },
    {
      "epoch": 0.6742567994939912,
      "grad_norm": 0.40935248136520386,
      "learning_rate": 5.2626700355965754e-05,
      "loss": 0.1641,
      "step": 47970
    },
    {
      "epoch": 0.674397357509312,
      "grad_norm": 0.31136754155158997,
      "learning_rate": 5.258577006209362e-05,
      "loss": 0.1696,
      "step": 47980
    },
    {
      "epoch": 0.6745379155246328,
      "grad_norm": 0.4646071195602417,
      "learning_rate": 5.254485001251701e-05,
      "loss": 0.1812,
      "step": 47990
    },
    {
      "epoch": 0.6746784735399536,
      "grad_norm": 0.34271240234375,
      "learning_rate": 5.250394021607715e-05,
      "loss": 0.1756,
      "step": 48000
    },
    {
      "epoch": 0.6746784735399536,
      "eval_chrf": 73.59441090909657,
      "eval_loss": 0.3677200675010681,
      "eval_runtime": 324.8817,
      "eval_samples_per_second": 0.308,
      "eval_steps_per_second": 0.012,
      "step": 48000
    },
    {
      "epoch": 0.6748190315552745,
      "grad_norm": 0.3200497329235077,
      "learning_rate": 5.246304068161292e-05,
      "loss": 0.1691,
      "step": 48010
    },
    {
      "epoch": 0.6749595895705952,
      "grad_norm": 0.45317354798316956,
      "learning_rate": 5.2422151417961144e-05,
      "loss": 0.1373,
      "step": 48020
    },
    {
      "epoch": 0.6751001475859161,
      "grad_norm": 0.4153939187526703,
      "learning_rate": 5.238127243395624e-05,
      "loss": 0.2004,
      "step": 48030
    },
    {
      "epoch": 0.675240705601237,
      "grad_norm": 0.42276909947395325,
      "learning_rate": 5.234040373843059e-05,
      "loss": 0.163,
      "step": 48040
    },
    {
      "epoch": 0.6753812636165577,
      "grad_norm": 0.4025135040283203,
      "learning_rate": 5.229954534021417e-05,
      "loss": 0.1785,
      "step": 48050
    },
    {
      "epoch": 0.6755218216318786,
      "grad_norm": 0.2253536731004715,
      "learning_rate": 5.225869724813493e-05,
      "loss": 0.1644,
      "step": 48060
    },
    {
      "epoch": 0.6756623796471993,
      "grad_norm": 0.3587108254432678,
      "learning_rate": 5.22178594710184e-05,
      "loss": 0.1794,
      "step": 48070
    },
    {
      "epoch": 0.6758029376625202,
      "grad_norm": 0.413657546043396,
      "learning_rate": 5.2177032017688046e-05,
      "loss": 0.1876,
      "step": 48080
    },
    {
      "epoch": 0.6759434956778411,
      "grad_norm": 0.39810359477996826,
      "learning_rate": 5.213621489696494e-05,
      "loss": 0.1671,
      "step": 48090
    },
    {
      "epoch": 0.6760840536931618,
      "grad_norm": 0.3903466463088989,
      "learning_rate": 5.20954081176681e-05,
      "loss": 0.1659,
      "step": 48100
    },
    {
      "epoch": 0.6762246117084827,
      "grad_norm": 0.43574848771095276,
      "learning_rate": 5.205461168861418e-05,
      "loss": 0.1618,
      "step": 48110
    },
    {
      "epoch": 0.6763651697238034,
      "grad_norm": 0.2700211703777313,
      "learning_rate": 5.201382561861759e-05,
      "loss": 0.1716,
      "step": 48120
    },
    {
      "epoch": 0.6765057277391243,
      "grad_norm": 0.3783707320690155,
      "learning_rate": 5.197304991649064e-05,
      "loss": 0.1645,
      "step": 48130
    },
    {
      "epoch": 0.6766462857544452,
      "grad_norm": 0.3352111577987671,
      "learning_rate": 5.1932284591043226e-05,
      "loss": 0.1508,
      "step": 48140
    },
    {
      "epoch": 0.676786843769766,
      "grad_norm": 0.42214563488960266,
      "learning_rate": 5.1891529651083146e-05,
      "loss": 0.2041,
      "step": 48150
    },
    {
      "epoch": 0.6769274017850868,
      "grad_norm": 0.385493665933609,
      "learning_rate": 5.185078510541583e-05,
      "loss": 0.196,
      "step": 48160
    },
    {
      "epoch": 0.6770679598004076,
      "grad_norm": 0.5575045347213745,
      "learning_rate": 5.18100509628446e-05,
      "loss": 0.1541,
      "step": 48170
    },
    {
      "epoch": 0.6772085178157284,
      "grad_norm": 0.3971259593963623,
      "learning_rate": 5.176932723217038e-05,
      "loss": 0.1716,
      "step": 48180
    },
    {
      "epoch": 0.6773490758310493,
      "grad_norm": 0.5190660953521729,
      "learning_rate": 5.172861392219199e-05,
      "loss": 0.1586,
      "step": 48190
    },
    {
      "epoch": 0.6774896338463701,
      "grad_norm": 0.3985922932624817,
      "learning_rate": 5.168791104170586e-05,
      "loss": 0.1964,
      "step": 48200
    },
    {
      "epoch": 0.6776301918616909,
      "grad_norm": 0.42941612005233765,
      "learning_rate": 5.164721859950631e-05,
      "loss": 0.1662,
      "step": 48210
    },
    {
      "epoch": 0.6777707498770117,
      "grad_norm": 0.4269723892211914,
      "learning_rate": 5.1606536604385235e-05,
      "loss": 0.169,
      "step": 48220
    },
    {
      "epoch": 0.6779113078923326,
      "grad_norm": 0.373107373714447,
      "learning_rate": 5.156586506513248e-05,
      "loss": 0.1754,
      "step": 48230
    },
    {
      "epoch": 0.6780518659076534,
      "grad_norm": 0.37568169832229614,
      "learning_rate": 5.15252039905354e-05,
      "loss": 0.1838,
      "step": 48240
    },
    {
      "epoch": 0.6781924239229742,
      "grad_norm": 0.5764591097831726,
      "learning_rate": 5.148455338937931e-05,
      "loss": 0.1815,
      "step": 48250
    },
    {
      "epoch": 0.678332981938295,
      "grad_norm": 0.4133480191230774,
      "learning_rate": 5.1443913270447076e-05,
      "loss": 0.169,
      "step": 48260
    },
    {
      "epoch": 0.6784735399536158,
      "grad_norm": 0.5136486291885376,
      "learning_rate": 5.1403283642519454e-05,
      "loss": 0.169,
      "step": 48270
    },
    {
      "epoch": 0.6786140979689367,
      "grad_norm": 0.3314149081707001,
      "learning_rate": 5.1362664514374794e-05,
      "loss": 0.1474,
      "step": 48280
    },
    {
      "epoch": 0.6787546559842575,
      "grad_norm": 0.3599088191986084,
      "learning_rate": 5.132205589478931e-05,
      "loss": 0.159,
      "step": 48290
    },
    {
      "epoch": 0.6788952139995783,
      "grad_norm": 0.34585297107696533,
      "learning_rate": 5.1281457792536835e-05,
      "loss": 0.1665,
      "step": 48300
    },
    {
      "epoch": 0.6790357720148992,
      "grad_norm": 0.3417344093322754,
      "learning_rate": 5.1240870216389004e-05,
      "loss": 0.1624,
      "step": 48310
    },
    {
      "epoch": 0.6791763300302199,
      "grad_norm": 0.3887704312801361,
      "learning_rate": 5.1200293175115096e-05,
      "loss": 0.19,
      "step": 48320
    },
    {
      "epoch": 0.6793168880455408,
      "grad_norm": 0.3785271942615509,
      "learning_rate": 5.115972667748227e-05,
      "loss": 0.1849,
      "step": 48330
    },
    {
      "epoch": 0.6794574460608617,
      "grad_norm": 0.37237548828125,
      "learning_rate": 5.111917073225519e-05,
      "loss": 0.1611,
      "step": 48340
    },
    {
      "epoch": 0.6795980040761824,
      "grad_norm": 0.39942654967308044,
      "learning_rate": 5.107862534819643e-05,
      "loss": 0.2033,
      "step": 48350
    },
    {
      "epoch": 0.6797385620915033,
      "grad_norm": 0.34830865263938904,
      "learning_rate": 5.103809053406615e-05,
      "loss": 0.1488,
      "step": 48360
    },
    {
      "epoch": 0.679879120106824,
      "grad_norm": 0.2858423888683319,
      "learning_rate": 5.0997566298622355e-05,
      "loss": 0.1555,
      "step": 48370
    },
    {
      "epoch": 0.6800196781221449,
      "grad_norm": 0.3308117687702179,
      "learning_rate": 5.0957052650620605e-05,
      "loss": 0.1709,
      "step": 48380
    },
    {
      "epoch": 0.6801602361374658,
      "grad_norm": 0.33425357937812805,
      "learning_rate": 5.091654959881433e-05,
      "loss": 0.1811,
      "step": 48390
    },
    {
      "epoch": 0.6803007941527865,
      "grad_norm": 0.2872956097126007,
      "learning_rate": 5.087605715195456e-05,
      "loss": 0.1591,
      "step": 48400
    },
    {
      "epoch": 0.6804413521681074,
      "grad_norm": 0.35558682680130005,
      "learning_rate": 5.083557531879011e-05,
      "loss": 0.1506,
      "step": 48410
    },
    {
      "epoch": 0.6805819101834282,
      "grad_norm": 0.3197231590747833,
      "learning_rate": 5.0795104108067406e-05,
      "loss": 0.1683,
      "step": 48420
    },
    {
      "epoch": 0.680722468198749,
      "grad_norm": 0.2837541699409485,
      "learning_rate": 5.075464352853071e-05,
      "loss": 0.1529,
      "step": 48430
    },
    {
      "epoch": 0.6808630262140699,
      "grad_norm": 0.5361260175704956,
      "learning_rate": 5.0714193588921846e-05,
      "loss": 0.1495,
      "step": 48440
    },
    {
      "epoch": 0.6810035842293907,
      "grad_norm": 0.39931023120880127,
      "learning_rate": 5.067375429798047e-05,
      "loss": 0.1781,
      "step": 48450
    },
    {
      "epoch": 0.6811441422447115,
      "grad_norm": 0.4552576541900635,
      "learning_rate": 5.063332566444381e-05,
      "loss": 0.1786,
      "step": 48460
    },
    {
      "epoch": 0.6812847002600323,
      "grad_norm": 0.39349377155303955,
      "learning_rate": 5.059290769704694e-05,
      "loss": 0.1717,
      "step": 48470
    },
    {
      "epoch": 0.6814252582753532,
      "grad_norm": 0.4419980049133301,
      "learning_rate": 5.0552500404522465e-05,
      "loss": 0.1721,
      "step": 48480
    },
    {
      "epoch": 0.6815658162906739,
      "grad_norm": 0.3095633089542389,
      "learning_rate": 5.051210379560083e-05,
      "loss": 0.1397,
      "step": 48490
    },
    {
      "epoch": 0.6817063743059948,
      "grad_norm": 0.4042552709579468,
      "learning_rate": 5.047171787901004e-05,
      "loss": 0.1854,
      "step": 48500
    },
    {
      "epoch": 0.6818469323213157,
      "grad_norm": 0.3907172381877899,
      "learning_rate": 5.0431342663475935e-05,
      "loss": 0.2011,
      "step": 48510
    },
    {
      "epoch": 0.6819874903366364,
      "grad_norm": 0.3393981158733368,
      "learning_rate": 5.039097815772188e-05,
      "loss": 0.1706,
      "step": 48520
    },
    {
      "epoch": 0.6821280483519573,
      "grad_norm": 0.4830770194530487,
      "learning_rate": 5.035062437046909e-05,
      "loss": 0.1828,
      "step": 48530
    },
    {
      "epoch": 0.682268606367278,
      "grad_norm": 0.45952385663986206,
      "learning_rate": 5.031028131043632e-05,
      "loss": 0.1731,
      "step": 48540
    },
    {
      "epoch": 0.6824091643825989,
      "grad_norm": 0.511646032333374,
      "learning_rate": 5.026994898634014e-05,
      "loss": 0.1683,
      "step": 48550
    },
    {
      "epoch": 0.6825497223979198,
      "grad_norm": 0.2801165282726288,
      "learning_rate": 5.022962740689465e-05,
      "loss": 0.1728,
      "step": 48560
    },
    {
      "epoch": 0.6826902804132405,
      "grad_norm": 0.33137014508247375,
      "learning_rate": 5.018931658081179e-05,
      "loss": 0.1684,
      "step": 48570
    },
    {
      "epoch": 0.6828308384285614,
      "grad_norm": 0.30291131138801575,
      "learning_rate": 5.014901651680102e-05,
      "loss": 0.156,
      "step": 48580
    },
    {
      "epoch": 0.6829713964438822,
      "grad_norm": 0.4711996614933014,
      "learning_rate": 5.010872722356964e-05,
      "loss": 0.1823,
      "step": 48590
    },
    {
      "epoch": 0.683111954459203,
      "grad_norm": 0.29929518699645996,
      "learning_rate": 5.006844870982245e-05,
      "loss": 0.1776,
      "step": 48600
    },
    {
      "epoch": 0.6832525124745239,
      "grad_norm": 0.4039543569087982,
      "learning_rate": 5.002818098426207e-05,
      "loss": 0.1739,
      "step": 48610
    },
    {
      "epoch": 0.6833930704898447,
      "grad_norm": 0.3180190622806549,
      "learning_rate": 4.998792405558869e-05,
      "loss": 0.1673,
      "step": 48620
    },
    {
      "epoch": 0.6835336285051655,
      "grad_norm": 0.4465638995170593,
      "learning_rate": 4.9947677932500225e-05,
      "loss": 0.1812,
      "step": 48630
    },
    {
      "epoch": 0.6836741865204863,
      "grad_norm": 0.4132784605026245,
      "learning_rate": 4.990744262369219e-05,
      "loss": 0.1746,
      "step": 48640
    },
    {
      "epoch": 0.6838147445358072,
      "grad_norm": 0.4020783305168152,
      "learning_rate": 4.9867218137857884e-05,
      "loss": 0.1977,
      "step": 48650
    },
    {
      "epoch": 0.683955302551128,
      "grad_norm": 0.44012129306793213,
      "learning_rate": 4.98270044836881e-05,
      "loss": 0.1664,
      "step": 48660
    },
    {
      "epoch": 0.6840958605664488,
      "grad_norm": 0.3492875099182129,
      "learning_rate": 4.978680166987145e-05,
      "loss": 0.1712,
      "step": 48670
    },
    {
      "epoch": 0.6842364185817696,
      "grad_norm": 0.41620054841041565,
      "learning_rate": 4.9746609705094085e-05,
      "loss": 0.1763,
      "step": 48680
    },
    {
      "epoch": 0.6843769765970904,
      "grad_norm": 0.3826924264431,
      "learning_rate": 4.970642859803992e-05,
      "loss": 0.1799,
      "step": 48690
    },
    {
      "epoch": 0.6845175346124113,
      "grad_norm": 0.4146118760108948,
      "learning_rate": 4.966625835739038e-05,
      "loss": 0.1771,
      "step": 48700
    },
    {
      "epoch": 0.6846580926277321,
      "grad_norm": 0.44485095143318176,
      "learning_rate": 4.962609899182473e-05,
      "loss": 0.1632,
      "step": 48710
    },
    {
      "epoch": 0.6847986506430529,
      "grad_norm": 0.3078020513057709,
      "learning_rate": 4.95859505100197e-05,
      "loss": 0.1611,
      "step": 48720
    },
    {
      "epoch": 0.6849392086583738,
      "grad_norm": 0.3447209596633911,
      "learning_rate": 4.95458129206498e-05,
      "loss": 0.1548,
      "step": 48730
    },
    {
      "epoch": 0.6850797666736945,
      "grad_norm": 0.38146939873695374,
      "learning_rate": 4.950568623238709e-05,
      "loss": 0.1693,
      "step": 48740
    },
    {
      "epoch": 0.6852203246890154,
      "grad_norm": 0.3679945468902588,
      "learning_rate": 4.94655704539014e-05,
      "loss": 0.1837,
      "step": 48750
    },
    {
      "epoch": 0.6853608827043363,
      "grad_norm": 0.32762065529823303,
      "learning_rate": 4.942546559386003e-05,
      "loss": 0.1629,
      "step": 48760
    },
    {
      "epoch": 0.685501440719657,
      "grad_norm": 0.35465124249458313,
      "learning_rate": 4.93853716609281e-05,
      "loss": 0.1759,
      "step": 48770
    },
    {
      "epoch": 0.6856419987349779,
      "grad_norm": 0.41711315512657166,
      "learning_rate": 4.9345288663768266e-05,
      "loss": 0.161,
      "step": 48780
    },
    {
      "epoch": 0.6857825567502986,
      "grad_norm": 0.36827749013900757,
      "learning_rate": 4.930521661104083e-05,
      "loss": 0.1477,
      "step": 48790
    },
    {
      "epoch": 0.6859231147656195,
      "grad_norm": 0.39657124876976013,
      "learning_rate": 4.9265155511403685e-05,
      "loss": 0.1675,
      "step": 48800
    },
    {
      "epoch": 0.6860636727809404,
      "grad_norm": 0.3166278004646301,
      "learning_rate": 4.9225105373512505e-05,
      "loss": 0.1622,
      "step": 48810
    },
    {
      "epoch": 0.6862042307962611,
      "grad_norm": 0.28927573561668396,
      "learning_rate": 4.918506620602043e-05,
      "loss": 0.161,
      "step": 48820
    },
    {
      "epoch": 0.686344788811582,
      "grad_norm": 0.3811091482639313,
      "learning_rate": 4.914503801757836e-05,
      "loss": 0.1602,
      "step": 48830
    },
    {
      "epoch": 0.6864853468269028,
      "grad_norm": 0.364082008600235,
      "learning_rate": 4.910502081683471e-05,
      "loss": 0.185,
      "step": 48840
    },
    {
      "epoch": 0.6866259048422236,
      "grad_norm": 0.38235294818878174,
      "learning_rate": 4.9065014612435657e-05,
      "loss": 0.183,
      "step": 48850
    },
    {
      "epoch": 0.6867664628575445,
      "grad_norm": 0.40850192308425903,
      "learning_rate": 4.902501941302483e-05,
      "loss": 0.1853,
      "step": 48860
    },
    {
      "epoch": 0.6869070208728653,
      "grad_norm": 0.4095821976661682,
      "learning_rate": 4.898503522724366e-05,
      "loss": 0.1591,
      "step": 48870
    },
    {
      "epoch": 0.6870475788881861,
      "grad_norm": 0.45164576172828674,
      "learning_rate": 4.894506206373102e-05,
      "loss": 0.1602,
      "step": 48880
    },
    {
      "epoch": 0.6871881369035069,
      "grad_norm": 0.36443573236465454,
      "learning_rate": 4.890509993112358e-05,
      "loss": 0.1652,
      "step": 48890
    },
    {
      "epoch": 0.6873286949188278,
      "grad_norm": 0.3586047887802124,
      "learning_rate": 4.886514883805546e-05,
      "loss": 0.1623,
      "step": 48900
    },
    {
      "epoch": 0.6874692529341486,
      "grad_norm": 0.37332266569137573,
      "learning_rate": 4.882520879315856e-05,
      "loss": 0.1845,
      "step": 48910
    },
    {
      "epoch": 0.6876098109494694,
      "grad_norm": 0.3771437406539917,
      "learning_rate": 4.878527980506221e-05,
      "loss": 0.1746,
      "step": 48920
    },
    {
      "epoch": 0.6877503689647902,
      "grad_norm": 0.37698593735694885,
      "learning_rate": 4.874536188239354e-05,
      "loss": 0.1782,
      "step": 48930
    },
    {
      "epoch": 0.687890926980111,
      "grad_norm": 0.33789968490600586,
      "learning_rate": 4.870545503377711e-05,
      "loss": 0.1572,
      "step": 48940
    },
    {
      "epoch": 0.6880314849954319,
      "grad_norm": 0.3339237570762634,
      "learning_rate": 4.866555926783527e-05,
      "loss": 0.1485,
      "step": 48950
    },
    {
      "epoch": 0.6881720430107527,
      "grad_norm": 0.5001163482666016,
      "learning_rate": 4.862567459318778e-05,
      "loss": 0.157,
      "step": 48960
    },
    {
      "epoch": 0.6883126010260735,
      "grad_norm": 0.3400193750858307,
      "learning_rate": 4.8585801018452194e-05,
      "loss": 0.1636,
      "step": 48970
    },
    {
      "epoch": 0.6884531590413944,
      "grad_norm": 0.4310280382633209,
      "learning_rate": 4.8545938552243495e-05,
      "loss": 0.193,
      "step": 48980
    },
    {
      "epoch": 0.6885937170567151,
      "grad_norm": 0.36668428778648376,
      "learning_rate": 4.850608720317442e-05,
      "loss": 0.1676,
      "step": 48990
    },
    {
      "epoch": 0.688734275072036,
      "grad_norm": 0.44923144578933716,
      "learning_rate": 4.846624697985518e-05,
      "loss": 0.1674,
      "step": 49000
    },
    {
      "epoch": 0.688734275072036,
      "eval_chrf": 79.95379135798156,
      "eval_loss": 0.3630891442298889,
      "eval_runtime": 295.3513,
      "eval_samples_per_second": 0.339,
      "eval_steps_per_second": 0.014,
      "step": 49000
    },
    {
      "epoch": 0.6888748330873568,
      "grad_norm": 0.33405956625938416,
      "learning_rate": 4.842641789089368e-05,
      "loss": 0.1702,
      "step": 49010
    },
    {
      "epoch": 0.6890153911026776,
      "grad_norm": 0.33272942900657654,
      "learning_rate": 4.838659994489532e-05,
      "loss": 0.1571,
      "step": 49020
    },
    {
      "epoch": 0.6891559491179985,
      "grad_norm": 0.4708268642425537,
      "learning_rate": 4.8346793150463196e-05,
      "loss": 0.1676,
      "step": 49030
    },
    {
      "epoch": 0.6892965071333192,
      "grad_norm": 0.3452412486076355,
      "learning_rate": 4.830699751619789e-05,
      "loss": 0.1723,
      "step": 49040
    },
    {
      "epoch": 0.6894370651486401,
      "grad_norm": 0.30366259813308716,
      "learning_rate": 4.8267213050697714e-05,
      "loss": 0.1711,
      "step": 49050
    },
    {
      "epoch": 0.6895776231639609,
      "grad_norm": 0.42754900455474854,
      "learning_rate": 4.822743976255838e-05,
      "loss": 0.1847,
      "step": 49060
    },
    {
      "epoch": 0.6897181811792817,
      "grad_norm": 0.4515030086040497,
      "learning_rate": 4.8187677660373364e-05,
      "loss": 0.1846,
      "step": 49070
    },
    {
      "epoch": 0.6898587391946026,
      "grad_norm": 0.30266475677490234,
      "learning_rate": 4.814792675273359e-05,
      "loss": 0.1607,
      "step": 49080
    },
    {
      "epoch": 0.6899992972099234,
      "grad_norm": 0.340821772813797,
      "learning_rate": 4.810818704822767e-05,
      "loss": 0.1925,
      "step": 49090
    },
    {
      "epoch": 0.6901398552252442,
      "grad_norm": 0.28537577390670776,
      "learning_rate": 4.8068458555441687e-05,
      "loss": 0.1834,
      "step": 49100
    },
    {
      "epoch": 0.690280413240565,
      "grad_norm": 0.44687119126319885,
      "learning_rate": 4.8028741282959435e-05,
      "loss": 0.1698,
      "step": 49110
    },
    {
      "epoch": 0.6904209712558859,
      "grad_norm": 0.40084707736968994,
      "learning_rate": 4.7989035239362126e-05,
      "loss": 0.1839,
      "step": 49120
    },
    {
      "epoch": 0.6905615292712067,
      "grad_norm": 0.47335416078567505,
      "learning_rate": 4.79493404332287e-05,
      "loss": 0.1514,
      "step": 49130
    },
    {
      "epoch": 0.6907020872865275,
      "grad_norm": 0.4480488896369934,
      "learning_rate": 4.7909656873135534e-05,
      "loss": 0.195,
      "step": 49140
    },
    {
      "epoch": 0.6908426453018484,
      "grad_norm": 0.4829513430595398,
      "learning_rate": 4.786998456765671e-05,
      "loss": 0.155,
      "step": 49150
    },
    {
      "epoch": 0.6909832033171691,
      "grad_norm": 0.3073470890522003,
      "learning_rate": 4.783032352536373e-05,
      "loss": 0.1567,
      "step": 49160
    },
    {
      "epoch": 0.69112376133249,
      "grad_norm": 0.4633159041404724,
      "learning_rate": 4.7790673754825824e-05,
      "loss": 0.1781,
      "step": 49170
    },
    {
      "epoch": 0.6912643193478109,
      "grad_norm": 0.3136443495750427,
      "learning_rate": 4.775103526460961e-05,
      "loss": 0.1568,
      "step": 49180
    },
    {
      "epoch": 0.6914048773631316,
      "grad_norm": 0.370098352432251,
      "learning_rate": 4.7711408063279454e-05,
      "loss": 0.142,
      "step": 49190
    },
    {
      "epoch": 0.6915454353784525,
      "grad_norm": 0.3789938688278198,
      "learning_rate": 4.76717921593971e-05,
      "loss": 0.1763,
      "step": 49200
    },
    {
      "epoch": 0.6916859933937732,
      "grad_norm": 0.4118127226829529,
      "learning_rate": 4.763218756152201e-05,
      "loss": 0.1727,
      "step": 49210
    },
    {
      "epoch": 0.6918265514090941,
      "grad_norm": 0.4234118163585663,
      "learning_rate": 4.759259427821109e-05,
      "loss": 0.1654,
      "step": 49220
    },
    {
      "epoch": 0.691967109424415,
      "grad_norm": 0.7092957496643066,
      "learning_rate": 4.7553012318018896e-05,
      "loss": 0.1703,
      "step": 49230
    },
    {
      "epoch": 0.6921076674397357,
      "grad_norm": 0.49885180592536926,
      "learning_rate": 4.751344168949742e-05,
      "loss": 0.1686,
      "step": 49240
    },
    {
      "epoch": 0.6922482254550566,
      "grad_norm": 0.45480483770370483,
      "learning_rate": 4.7473882401196346e-05,
      "loss": 0.1517,
      "step": 49250
    },
    {
      "epoch": 0.6923887834703774,
      "grad_norm": 0.27201780676841736,
      "learning_rate": 4.743433446166275e-05,
      "loss": 0.146,
      "step": 49260
    },
    {
      "epoch": 0.6925293414856982,
      "grad_norm": 0.36466437578201294,
      "learning_rate": 4.739479787944143e-05,
      "loss": 0.1629,
      "step": 49270
    },
    {
      "epoch": 0.6926698995010191,
      "grad_norm": 0.2820543348789215,
      "learning_rate": 4.735527266307457e-05,
      "loss": 0.169,
      "step": 49280
    },
    {
      "epoch": 0.6928104575163399,
      "grad_norm": 0.3912178874015808,
      "learning_rate": 4.731575882110202e-05,
      "loss": 0.1655,
      "step": 49290
    },
    {
      "epoch": 0.6929510155316607,
      "grad_norm": 0.37822580337524414,
      "learning_rate": 4.7276256362061075e-05,
      "loss": 0.1568,
      "step": 49300
    },
    {
      "epoch": 0.6930915735469815,
      "grad_norm": 0.32967841625213623,
      "learning_rate": 4.723676529448668e-05,
      "loss": 0.1857,
      "step": 49310
    },
    {
      "epoch": 0.6932321315623023,
      "grad_norm": 0.816712498664856,
      "learning_rate": 4.7197285626911156e-05,
      "loss": 0.1935,
      "step": 49320
    },
    {
      "epoch": 0.6933726895776232,
      "grad_norm": 0.4365677833557129,
      "learning_rate": 4.715781736786459e-05,
      "loss": 0.1751,
      "step": 49330
    },
    {
      "epoch": 0.693513247592944,
      "grad_norm": 0.37101316452026367,
      "learning_rate": 4.711836052587435e-05,
      "loss": 0.1572,
      "step": 49340
    },
    {
      "epoch": 0.6936538056082648,
      "grad_norm": 0.41701507568359375,
      "learning_rate": 4.707891510946556e-05,
      "loss": 0.1634,
      "step": 49350
    },
    {
      "epoch": 0.6937943636235856,
      "grad_norm": 0.32426103949546814,
      "learning_rate": 4.70394811271607e-05,
      "loss": 0.1639,
      "step": 49360
    },
    {
      "epoch": 0.6939349216389065,
      "grad_norm": 0.3698156774044037,
      "learning_rate": 4.700005858747993e-05,
      "loss": 0.1431,
      "step": 49370
    },
    {
      "epoch": 0.6940754796542273,
      "grad_norm": 0.3618178367614746,
      "learning_rate": 4.69606474989408e-05,
      "loss": 0.1568,
      "step": 49380
    },
    {
      "epoch": 0.6942160376695481,
      "grad_norm": 0.29728826880455017,
      "learning_rate": 4.692124787005852e-05,
      "loss": 0.1765,
      "step": 49390
    },
    {
      "epoch": 0.694356595684869,
      "grad_norm": 0.3886052072048187,
      "learning_rate": 4.688185970934565e-05,
      "loss": 0.1636,
      "step": 49400
    },
    {
      "epoch": 0.6944971537001897,
      "grad_norm": 0.43441951274871826,
      "learning_rate": 4.6842483025312513e-05,
      "loss": 0.1608,
      "step": 49410
    },
    {
      "epoch": 0.6946377117155106,
      "grad_norm": 0.47952476143836975,
      "learning_rate": 4.680311782646668e-05,
      "loss": 0.1713,
      "step": 49420
    },
    {
      "epoch": 0.6947782697308315,
      "grad_norm": 0.33909785747528076,
      "learning_rate": 4.676376412131349e-05,
      "loss": 0.1552,
      "step": 49430
    },
    {
      "epoch": 0.6949188277461522,
      "grad_norm": 0.3531533181667328,
      "learning_rate": 4.67244219183556e-05,
      "loss": 0.1422,
      "step": 49440
    },
    {
      "epoch": 0.6950593857614731,
      "grad_norm": 0.39314037561416626,
      "learning_rate": 4.6685091226093334e-05,
      "loss": 0.187,
      "step": 49450
    },
    {
      "epoch": 0.6951999437767938,
      "grad_norm": 0.46536579728126526,
      "learning_rate": 4.664577205302444e-05,
      "loss": 0.1507,
      "step": 49460
    },
    {
      "epoch": 0.6953405017921147,
      "grad_norm": 0.27574050426483154,
      "learning_rate": 4.660646440764416e-05,
      "loss": 0.1515,
      "step": 49470
    },
    {
      "epoch": 0.6954810598074355,
      "grad_norm": 0.32661160826683044,
      "learning_rate": 4.656716829844534e-05,
      "loss": 0.1662,
      "step": 49480
    },
    {
      "epoch": 0.6956216178227563,
      "grad_norm": 0.356479674577713,
      "learning_rate": 4.652788373391822e-05,
      "loss": 0.1804,
      "step": 49490
    },
    {
      "epoch": 0.6957621758380772,
      "grad_norm": 0.349311888217926,
      "learning_rate": 4.648861072255068e-05,
      "loss": 0.1711,
      "step": 49500
    },
    {
      "epoch": 0.695902733853398,
      "grad_norm": 0.3784264028072357,
      "learning_rate": 4.644934927282795e-05,
      "loss": 0.1633,
      "step": 49510
    },
    {
      "epoch": 0.6960432918687188,
      "grad_norm": 0.463031142950058,
      "learning_rate": 4.6410099393232906e-05,
      "loss": 0.1661,
      "step": 49520
    },
    {
      "epoch": 0.6961838498840396,
      "grad_norm": 0.4308297634124756,
      "learning_rate": 4.63708610922458e-05,
      "loss": 0.169,
      "step": 49530
    },
    {
      "epoch": 0.6963244078993605,
      "grad_norm": 0.4579606056213379,
      "learning_rate": 4.63316343783445e-05,
      "loss": 0.1648,
      "step": 49540
    },
    {
      "epoch": 0.6964649659146813,
      "grad_norm": 0.3730986714363098,
      "learning_rate": 4.629241926000425e-05,
      "loss": 0.1761,
      "step": 49550
    },
    {
      "epoch": 0.6966055239300021,
      "grad_norm": 0.41617172956466675,
      "learning_rate": 4.625321574569791e-05,
      "loss": 0.1989,
      "step": 49560
    },
    {
      "epoch": 0.696746081945323,
      "grad_norm": 0.42932072281837463,
      "learning_rate": 4.6214023843895705e-05,
      "loss": 0.1723,
      "step": 49570
    },
    {
      "epoch": 0.6968866399606437,
      "grad_norm": 0.29537931084632874,
      "learning_rate": 4.617484356306549e-05,
      "loss": 0.1579,
      "step": 49580
    },
    {
      "epoch": 0.6970271979759646,
      "grad_norm": 0.40688666701316833,
      "learning_rate": 4.6135674911672466e-05,
      "loss": 0.1583,
      "step": 49590
    },
    {
      "epoch": 0.6971677559912854,
      "grad_norm": 0.33323273062705994,
      "learning_rate": 4.609651789817947e-05,
      "loss": 0.1639,
      "step": 49600
    },
    {
      "epoch": 0.6973083140066062,
      "grad_norm": 0.39615142345428467,
      "learning_rate": 4.605737253104667e-05,
      "loss": 0.1729,
      "step": 49610
    },
    {
      "epoch": 0.6974488720219271,
      "grad_norm": 0.5205367207527161,
      "learning_rate": 4.6018238818731865e-05,
      "loss": 0.1976,
      "step": 49620
    },
    {
      "epoch": 0.6975894300372478,
      "grad_norm": 0.5548632144927979,
      "learning_rate": 4.597911676969019e-05,
      "loss": 0.1675,
      "step": 49630
    },
    {
      "epoch": 0.6977299880525687,
      "grad_norm": 0.3631393313407898,
      "learning_rate": 4.594000639237443e-05,
      "loss": 0.1538,
      "step": 49640
    },
    {
      "epoch": 0.6978705460678896,
      "grad_norm": 0.3613353371620178,
      "learning_rate": 4.5900907695234654e-05,
      "loss": 0.1595,
      "step": 49650
    },
    {
      "epoch": 0.6980111040832103,
      "grad_norm": 0.4969308376312256,
      "learning_rate": 4.5861820686718596e-05,
      "loss": 0.1557,
      "step": 49660
    },
    {
      "epoch": 0.6981516620985312,
      "grad_norm": 0.2675575613975525,
      "learning_rate": 4.5822745375271294e-05,
      "loss": 0.1673,
      "step": 49670
    },
    {
      "epoch": 0.698292220113852,
      "grad_norm": 0.4694734811782837,
      "learning_rate": 4.578368176933543e-05,
      "loss": 0.1487,
      "step": 49680
    },
    {
      "epoch": 0.6984327781291728,
      "grad_norm": 0.32156407833099365,
      "learning_rate": 4.5744629877350975e-05,
      "loss": 0.1498,
      "step": 49690
    },
    {
      "epoch": 0.6985733361444937,
      "grad_norm": 0.3179536759853363,
      "learning_rate": 4.5705589707755545e-05,
      "loss": 0.1644,
      "step": 49700
    },
    {
      "epoch": 0.6987138941598144,
      "grad_norm": 0.46308329701423645,
      "learning_rate": 4.566656126898407e-05,
      "loss": 0.1727,
      "step": 49710
    },
    {
      "epoch": 0.6988544521751353,
      "grad_norm": 0.45487168431282043,
      "learning_rate": 4.5627544569469083e-05,
      "loss": 0.1752,
      "step": 49720
    },
    {
      "epoch": 0.6989950101904561,
      "grad_norm": 0.4241594076156616,
      "learning_rate": 4.5588539617640446e-05,
      "loss": 0.1655,
      "step": 49730
    },
    {
      "epoch": 0.6991355682057769,
      "grad_norm": 0.36288541555404663,
      "learning_rate": 4.554954642192562e-05,
      "loss": 0.1685,
      "step": 49740
    },
    {
      "epoch": 0.6992761262210978,
      "grad_norm": 0.324632465839386,
      "learning_rate": 4.551056499074937e-05,
      "loss": 0.1819,
      "step": 49750
    },
    {
      "epoch": 0.6994166842364186,
      "grad_norm": 0.2858855724334717,
      "learning_rate": 4.54715953325341e-05,
      "loss": 0.1636,
      "step": 49760
    },
    {
      "epoch": 0.6995572422517394,
      "grad_norm": 0.47081872820854187,
      "learning_rate": 4.543263745569949e-05,
      "loss": 0.1658,
      "step": 49770
    },
    {
      "epoch": 0.6996978002670602,
      "grad_norm": 0.34186944365501404,
      "learning_rate": 4.539369136866284e-05,
      "loss": 0.171,
      "step": 49780
    },
    {
      "epoch": 0.6998383582823811,
      "grad_norm": 0.40782099962234497,
      "learning_rate": 4.535475707983874e-05,
      "loss": 0.1792,
      "step": 49790
    },
    {
      "epoch": 0.6999789162977019,
      "grad_norm": 0.407845139503479,
      "learning_rate": 4.531583459763939e-05,
      "loss": 0.1389,
      "step": 49800
    },
    {
      "epoch": 0.7001194743130227,
      "grad_norm": 0.4455857276916504,
      "learning_rate": 4.527692393047429e-05,
      "loss": 0.1733,
      "step": 49810
    },
    {
      "epoch": 0.7002600323283436,
      "grad_norm": 0.31916528940200806,
      "learning_rate": 4.5238025086750544e-05,
      "loss": 0.1562,
      "step": 49820
    },
    {
      "epoch": 0.7004005903436643,
      "grad_norm": 0.34428226947784424,
      "learning_rate": 4.5199138074872525e-05,
      "loss": 0.1694,
      "step": 49830
    },
    {
      "epoch": 0.7005411483589852,
      "grad_norm": 0.384682834148407,
      "learning_rate": 4.5160262903242235e-05,
      "loss": 0.1586,
      "step": 49840
    },
    {
      "epoch": 0.700681706374306,
      "grad_norm": 0.4845108091831207,
      "learning_rate": 4.512139958025894e-05,
      "loss": 0.1692,
      "step": 49850
    },
    {
      "epoch": 0.7008222643896268,
      "grad_norm": 0.40447092056274414,
      "learning_rate": 4.50825481143195e-05,
      "loss": 0.1537,
      "step": 49860
    },
    {
      "epoch": 0.7009628224049477,
      "grad_norm": 0.42161500453948975,
      "learning_rate": 4.504370851381807e-05,
      "loss": 0.1729,
      "step": 49870
    },
    {
      "epoch": 0.7011033804202684,
      "grad_norm": 0.38627347350120544,
      "learning_rate": 4.500488078714641e-05,
      "loss": 0.1649,
      "step": 49880
    },
    {
      "epoch": 0.7012439384355893,
      "grad_norm": 0.5103524923324585,
      "learning_rate": 4.496606494269352e-05,
      "loss": 0.1773,
      "step": 49890
    },
    {
      "epoch": 0.7013844964509102,
      "grad_norm": 0.337223619222641,
      "learning_rate": 4.492726098884602e-05,
      "loss": 0.1598,
      "step": 49900
    },
    {
      "epoch": 0.7015250544662309,
      "grad_norm": 0.4297640025615692,
      "learning_rate": 4.4888468933987806e-05,
      "loss": 0.1828,
      "step": 49910
    },
    {
      "epoch": 0.7016656124815518,
      "grad_norm": 0.31979233026504517,
      "learning_rate": 4.484968878650032e-05,
      "loss": 0.1836,
      "step": 49920
    },
    {
      "epoch": 0.7018061704968726,
      "grad_norm": 0.28272655606269836,
      "learning_rate": 4.481092055476233e-05,
      "loss": 0.1734,
      "step": 49930
    },
    {
      "epoch": 0.7019467285121934,
      "grad_norm": 0.41349393129348755,
      "learning_rate": 4.477216424715015e-05,
      "loss": 0.166,
      "step": 49940
    },
    {
      "epoch": 0.7020872865275142,
      "grad_norm": 0.34932541847229004,
      "learning_rate": 4.473341987203737e-05,
      "loss": 0.1565,
      "step": 49950
    },
    {
      "epoch": 0.702227844542835,
      "grad_norm": 0.369492769241333,
      "learning_rate": 4.469468743779516e-05,
      "loss": 0.1618,
      "step": 49960
    },
    {
      "epoch": 0.7023684025581559,
      "grad_norm": 0.38204425573349,
      "learning_rate": 4.4655966952791964e-05,
      "loss": 0.1648,
      "step": 49970
    },
    {
      "epoch": 0.7025089605734767,
      "grad_norm": 0.40265777707099915,
      "learning_rate": 4.461725842539376e-05,
      "loss": 0.187,
      "step": 49980
    },
    {
      "epoch": 0.7026495185887975,
      "grad_norm": 0.38271403312683105,
      "learning_rate": 4.4578561863963856e-05,
      "loss": 0.1768,
      "step": 49990
    },
    {
      "epoch": 0.7027900766041183,
      "grad_norm": 0.39733636379241943,
      "learning_rate": 4.453987727686305e-05,
      "loss": 0.1746,
      "step": 50000
    },
    {
      "epoch": 0.7027900766041183,
      "eval_chrf": 76.08611862854853,
      "eval_loss": 0.3601585030555725,
      "eval_runtime": 325.1544,
      "eval_samples_per_second": 0.308,
      "eval_steps_per_second": 0.012,
      "step": 50000
    }
  ],
  "logging_steps": 10,
  "max_steps": 71145,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "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": 1.1762575999305253e+19,
  "train_batch_size": 16,
  "trial_name": null,
  "trial_params": null
}