dancher00's picture
Add files using upload-large-folder tool
4a410db verified
Raw History Blame Contribute Delete
25 kB
{"step": 1, "samples": 32, "loss": 97.23249435424805, "elapsed_s": 0.806594289955683, "gpu_peak_bytes": 1764845568}
{"step": 50, "samples": 1600, "loss": 4.44867730140686, "elapsed_s": 5.9837520299479365, "gpu_peak_bytes": 2190078464}
{"step": 100, "samples": 3200, "loss": 3.2360103130340576, "elapsed_s": 11.27319775801152, "gpu_peak_bytes": 2190078464}
{"step": 150, "samples": 4800, "loss": 2.649272322654724, "elapsed_s": 16.570440987008624, "gpu_peak_bytes": 2190078464}
{"step": 200, "samples": 6400, "loss": 2.382279396057129, "elapsed_s": 21.868673268938437, "gpu_peak_bytes": 2190078464}
{"step": 250, "samples": 8000, "loss": 2.331014394760132, "elapsed_s": 27.167194544919766, "gpu_peak_bytes": 2190078464}
{"step": 300, "samples": 9600, "loss": 2.1956160068511963, "elapsed_s": 32.46378080593422, "gpu_peak_bytes": 2190078464}
{"step": 350, "samples": 11200, "loss": 2.0285145044326782, "elapsed_s": 37.76752786792349, "gpu_peak_bytes": 2190078464}
{"step": 400, "samples": 12800, "loss": 1.9089829921722412, "elapsed_s": 43.08087970595807, "gpu_peak_bytes": 2190078464}
{"step": 450, "samples": 14400, "loss": 1.7331058979034424, "elapsed_s": 48.391218436998315, "gpu_peak_bytes": 2190078464}
{"step": 500, "samples": 16000, "loss": 1.6335176229476929, "elapsed_s": 53.69843208999373, "gpu_peak_bytes": 2190078464}
{"step": 550, "samples": 17600, "loss": 1.7034125328063965, "elapsed_s": 59.00773289601784, "gpu_peak_bytes": 2190078464}
{"step": 600, "samples": 19200, "loss": 1.5263854265213013, "elapsed_s": 64.32847883494105, "gpu_peak_bytes": 2190078464}
{"step": 650, "samples": 20800, "loss": 1.506377637386322, "elapsed_s": 69.63279770396184, "gpu_peak_bytes": 2190078464}
{"step": 700, "samples": 22400, "loss": 1.400745153427124, "elapsed_s": 74.94283087993972, "gpu_peak_bytes": 2190078464}
{"step": 750, "samples": 24000, "loss": 1.3519147038459778, "elapsed_s": 80.2589223489631, "gpu_peak_bytes": 2190078464}
{"step": 800, "samples": 25600, "loss": 1.3910820484161377, "elapsed_s": 85.57112260803115, "gpu_peak_bytes": 2190078464}
{"step": 850, "samples": 27200, "loss": 1.2487383484840393, "elapsed_s": 90.886372442008, "gpu_peak_bytes": 2190078464}
{"step": 900, "samples": 28800, "loss": 1.258819043636322, "elapsed_s": 96.20722900901455, "gpu_peak_bytes": 2190078464}
{"step": 950, "samples": 30400, "loss": 1.1732364892959595, "elapsed_s": 101.54274050798267, "gpu_peak_bytes": 2190078464}
{"step": 1000, "samples": 32000, "loss": 1.0459273755550385, "elapsed_s": 106.86471679399256, "gpu_peak_bytes": 2190078464}
{"step": 1050, "samples": 33600, "loss": 1.0042082071304321, "elapsed_s": 113.31849187496118, "gpu_peak_bytes": 2190078464}
{"step": 1100, "samples": 35200, "loss": 0.9996316432952881, "elapsed_s": 118.6266358710127, "gpu_peak_bytes": 2190078464}
{"step": 1150, "samples": 36800, "loss": 0.9639615714550018, "elapsed_s": 123.94723792898003, "gpu_peak_bytes": 2190078464}
{"step": 1200, "samples": 38400, "loss": 0.9665570855140686, "elapsed_s": 129.25313811795786, "gpu_peak_bytes": 2190078464}
{"step": 1250, "samples": 40000, "loss": 0.8356319665908813, "elapsed_s": 134.55315800698008, "gpu_peak_bytes": 2190078464}
{"step": 1300, "samples": 41600, "loss": 0.8291738033294678, "elapsed_s": 139.86301810701843, "gpu_peak_bytes": 2190078464}
{"step": 1350, "samples": 43200, "loss": 0.821506530046463, "elapsed_s": 145.14620219694916, "gpu_peak_bytes": 2190078464}
{"step": 1400, "samples": 44800, "loss": 0.8076695501804352, "elapsed_s": 150.44713914499152, "gpu_peak_bytes": 2190078464}
{"step": 1450, "samples": 46400, "loss": 0.7354048192501068, "elapsed_s": 155.7552924589254, "gpu_peak_bytes": 2190078464}
{"step": 1500, "samples": 48000, "loss": 0.6757569313049316, "elapsed_s": 161.0499951710226, "gpu_peak_bytes": 2190078464}
{"step": 1550, "samples": 49600, "loss": 0.702389657497406, "elapsed_s": 166.34741384000517, "gpu_peak_bytes": 2190078464}
{"step": 1600, "samples": 51200, "loss": 0.6318093836307526, "elapsed_s": 171.64478915696964, "gpu_peak_bytes": 2190078464}
{"step": 1650, "samples": 52800, "loss": 0.5465146005153656, "elapsed_s": 176.94133304792922, "gpu_peak_bytes": 2190078464}
{"step": 1700, "samples": 54400, "loss": 0.5511115491390228, "elapsed_s": 182.23264627403114, "gpu_peak_bytes": 2190078464}
{"step": 1750, "samples": 56000, "loss": 0.5472948849201202, "elapsed_s": 187.52178706694394, "gpu_peak_bytes": 2190078464}
{"step": 1800, "samples": 57600, "loss": 0.48824234306812286, "elapsed_s": 192.81827336992137, "gpu_peak_bytes": 2190078464}
{"step": 1850, "samples": 59200, "loss": 0.5246657729148865, "elapsed_s": 198.1105284099467, "gpu_peak_bytes": 2190078464}
{"step": 1900, "samples": 60800, "loss": 0.4738021194934845, "elapsed_s": 203.4022906019818, "gpu_peak_bytes": 2190078464}
{"step": 1950, "samples": 62400, "loss": 0.4458756148815155, "elapsed_s": 208.7014016089961, "gpu_peak_bytes": 2190078464}
{"step": 2000, "samples": 64000, "loss": 0.44699013233184814, "elapsed_s": 213.96828605001792, "gpu_peak_bytes": 2190078464}
{"step": 2050, "samples": 65600, "loss": 0.3983282744884491, "elapsed_s": 220.24350490095094, "gpu_peak_bytes": 2190078464}
{"step": 2100, "samples": 67200, "loss": 0.41660840809345245, "elapsed_s": 225.5357730659889, "gpu_peak_bytes": 2190078464}
{"step": 2150, "samples": 68800, "loss": 0.37194445729255676, "elapsed_s": 230.83341287099756, "gpu_peak_bytes": 2190078464}
{"step": 2200, "samples": 70400, "loss": 0.38820940256118774, "elapsed_s": 236.13108226296026, "gpu_peak_bytes": 2190078464}
{"step": 2250, "samples": 72000, "loss": 0.39105044305324554, "elapsed_s": 241.43496479000896, "gpu_peak_bytes": 2190078464}
{"step": 2300, "samples": 73600, "loss": 0.3411318212747574, "elapsed_s": 246.73331418295857, "gpu_peak_bytes": 2190078464}
{"step": 2350, "samples": 75200, "loss": 0.3237869590520859, "elapsed_s": 252.0316479660105, "gpu_peak_bytes": 2190078464}
{"step": 2400, "samples": 76800, "loss": 0.33254915475845337, "elapsed_s": 257.3242571540177, "gpu_peak_bytes": 2190078464}
{"step": 2450, "samples": 78400, "loss": 0.29191169142723083, "elapsed_s": 262.62473928392865, "gpu_peak_bytes": 2190078464}
{"step": 2500, "samples": 80000, "loss": 0.2735885828733444, "elapsed_s": 267.9208213549573, "gpu_peak_bytes": 2190078464}
{"step": 2550, "samples": 81600, "loss": 0.3056785613298416, "elapsed_s": 273.2196267469553, "gpu_peak_bytes": 2190078464}
{"step": 2600, "samples": 83200, "loss": 0.2563309520483017, "elapsed_s": 278.52281048800796, "gpu_peak_bytes": 2190078464}
{"step": 2650, "samples": 84800, "loss": 0.25254184752702713, "elapsed_s": 283.82141215598676, "gpu_peak_bytes": 2190078464}
{"step": 2700, "samples": 86400, "loss": 0.25799547135829926, "elapsed_s": 289.12476700893603, "gpu_peak_bytes": 2190078464}
{"step": 2750, "samples": 88000, "loss": 0.2530265748500824, "elapsed_s": 294.42559395893477, "gpu_peak_bytes": 2190078464}
{"step": 2800, "samples": 89600, "loss": 0.2617863267660141, "elapsed_s": 299.72376939398237, "gpu_peak_bytes": 2190078464}
{"step": 2850, "samples": 91200, "loss": 0.25425951182842255, "elapsed_s": 305.0172295749653, "gpu_peak_bytes": 2190078464}
{"step": 2900, "samples": 92800, "loss": 0.2172660455107689, "elapsed_s": 310.3186826349702, "gpu_peak_bytes": 2190078464}
{"step": 2950, "samples": 94400, "loss": 0.2123677134513855, "elapsed_s": 315.61749423795845, "gpu_peak_bytes": 2190078464}
{"step": 3000, "samples": 96000, "loss": 0.23210866749286652, "elapsed_s": 320.9030905739637, "gpu_peak_bytes": 2190078464}
{"step": 3050, "samples": 97600, "loss": 0.19584345817565918, "elapsed_s": 326.948113531922, "gpu_peak_bytes": 2190078464}
{"step": 3100, "samples": 99200, "loss": 0.18952026218175888, "elapsed_s": 332.25051777402405, "gpu_peak_bytes": 2190078464}
{"step": 3150, "samples": 100800, "loss": 0.20799300074577332, "elapsed_s": 337.56107959500514, "gpu_peak_bytes": 2190078464}
{"step": 3200, "samples": 102400, "loss": 0.1595018431544304, "elapsed_s": 342.8682704459643, "gpu_peak_bytes": 2190078464}
{"step": 3250, "samples": 104000, "loss": 0.17172850668430328, "elapsed_s": 348.17548751598224, "gpu_peak_bytes": 2190078464}
{"step": 3300, "samples": 105600, "loss": 0.17706073075532913, "elapsed_s": 353.4834777120268, "gpu_peak_bytes": 2190078464}
{"step": 3350, "samples": 107200, "loss": 0.1807897835969925, "elapsed_s": 358.793772961013, "gpu_peak_bytes": 2190078464}
{"step": 3400, "samples": 108800, "loss": 0.17567245662212372, "elapsed_s": 364.09125465899706, "gpu_peak_bytes": 2190078464}
{"step": 3450, "samples": 110400, "loss": 0.19487307965755463, "elapsed_s": 369.40482032997534, "gpu_peak_bytes": 2190078464}
{"step": 3500, "samples": 112000, "loss": 0.16127637773752213, "elapsed_s": 374.71603208594024, "gpu_peak_bytes": 2190078464}
{"step": 3550, "samples": 113600, "loss": 0.14200159907341003, "elapsed_s": 380.0147680869559, "gpu_peak_bytes": 2190078464}
{"step": 3600, "samples": 115200, "loss": 0.1628316566348076, "elapsed_s": 385.3226073589176, "gpu_peak_bytes": 2190078464}
{"step": 3650, "samples": 116800, "loss": 0.15180087089538574, "elapsed_s": 390.6282536489889, "gpu_peak_bytes": 2190078464}
{"step": 3700, "samples": 118400, "loss": 0.14517439901828766, "elapsed_s": 395.9341571339173, "gpu_peak_bytes": 2190078464}
{"step": 3750, "samples": 120000, "loss": 0.1650959998369217, "elapsed_s": 401.2361310250126, "gpu_peak_bytes": 2190078464}
{"step": 3800, "samples": 121600, "loss": 0.15026217699050903, "elapsed_s": 406.53317699197214, "gpu_peak_bytes": 2190078464}
{"step": 3850, "samples": 123200, "loss": 0.14495865255594254, "elapsed_s": 411.809477073024, "gpu_peak_bytes": 2190078464}
{"step": 3900, "samples": 124800, "loss": 0.13295932114124298, "elapsed_s": 417.11374453501776, "gpu_peak_bytes": 2190078464}
{"step": 3950, "samples": 126400, "loss": 0.14281868189573288, "elapsed_s": 422.4180177989183, "gpu_peak_bytes": 2190078464}
{"step": 4000, "samples": 128000, "loss": 0.14245427399873734, "elapsed_s": 427.632534315926, "gpu_peak_bytes": 2190078464}
{"step": 4050, "samples": 129600, "loss": 0.12835073843598366, "elapsed_s": 433.52824529097416, "gpu_peak_bytes": 2190078464}
{"step": 4100, "samples": 131200, "loss": 0.12172762304544449, "elapsed_s": 438.8361934289569, "gpu_peak_bytes": 2190078464}
{"step": 4150, "samples": 132800, "loss": 0.11965402960777283, "elapsed_s": 444.13499818893615, "gpu_peak_bytes": 2190078464}
{"step": 4200, "samples": 134400, "loss": 0.12617888674139977, "elapsed_s": 449.4373353049159, "gpu_peak_bytes": 2190078464}
{"step": 4250, "samples": 136000, "loss": 0.14665665477514267, "elapsed_s": 454.73447815398686, "gpu_peak_bytes": 2190078464}
{"step": 4300, "samples": 137600, "loss": 0.14132171869277954, "elapsed_s": 460.0370387909934, "gpu_peak_bytes": 2190078464}
{"step": 4350, "samples": 139200, "loss": 0.13767974823713303, "elapsed_s": 465.3373211149592, "gpu_peak_bytes": 2190078464}
{"step": 4400, "samples": 140800, "loss": 0.1149781234562397, "elapsed_s": 470.6410619060043, "gpu_peak_bytes": 2190078464}
{"step": 4450, "samples": 142400, "loss": 0.11354921758174896, "elapsed_s": 475.94074636301957, "gpu_peak_bytes": 2190078464}
{"step": 4500, "samples": 144000, "loss": 0.10432625189423561, "elapsed_s": 481.2489328099182, "gpu_peak_bytes": 2190078464}
{"step": 4550, "samples": 145600, "loss": 0.10389555618166924, "elapsed_s": 486.5537482550135, "gpu_peak_bytes": 2190078464}
{"step": 4600, "samples": 147200, "loss": 0.13567029684782028, "elapsed_s": 491.8546780089382, "gpu_peak_bytes": 2190078464}
{"step": 4650, "samples": 148800, "loss": 0.10692187771201134, "elapsed_s": 497.1564589869231, "gpu_peak_bytes": 2190078464}
{"step": 4700, "samples": 150400, "loss": 0.1122741736471653, "elapsed_s": 502.4594789890107, "gpu_peak_bytes": 2190078464}
{"step": 4750, "samples": 152000, "loss": 0.09723108261823654, "elapsed_s": 507.7509268489666, "gpu_peak_bytes": 2190078464}
{"step": 4800, "samples": 153600, "loss": 0.11166052520275116, "elapsed_s": 513.0510302960174, "gpu_peak_bytes": 2190078464}
{"step": 4850, "samples": 155200, "loss": 0.11397334933280945, "elapsed_s": 518.3495396899525, "gpu_peak_bytes": 2190078464}
{"step": 4900, "samples": 156800, "loss": 0.09655984863638878, "elapsed_s": 523.6494906099979, "gpu_peak_bytes": 2190078464}
{"step": 4950, "samples": 158400, "loss": 0.11744865402579308, "elapsed_s": 528.9492970679421, "gpu_peak_bytes": 2190078464}
{"step": 5000, "samples": 160000, "loss": 0.09831105917692184, "elapsed_s": 534.01437084598, "gpu_peak_bytes": 2190078464}
{"step": 5050, "samples": 161600, "loss": 0.09115675091743469, "elapsed_s": 539.9205669459188, "gpu_peak_bytes": 2190078464}
{"step": 5100, "samples": 163200, "loss": 0.11284533515572548, "elapsed_s": 545.2226077939849, "gpu_peak_bytes": 2190078464}
{"step": 5150, "samples": 164800, "loss": 0.10486491397023201, "elapsed_s": 550.5186616409337, "gpu_peak_bytes": 2190078464}
{"step": 5200, "samples": 166400, "loss": 0.08540752157568932, "elapsed_s": 555.8173181830207, "gpu_peak_bytes": 2190078464}
{"step": 5250, "samples": 168000, "loss": 0.08265968412160873, "elapsed_s": 561.1071955530206, "gpu_peak_bytes": 2190078464}
{"step": 5300, "samples": 169600, "loss": 0.0852733924984932, "elapsed_s": 566.4190898350207, "gpu_peak_bytes": 2190078464}
{"step": 5350, "samples": 171200, "loss": 0.10322801768779755, "elapsed_s": 571.7117984809447, "gpu_peak_bytes": 2190078464}
{"step": 5400, "samples": 172800, "loss": 0.10043694078922272, "elapsed_s": 577.0247545529855, "gpu_peak_bytes": 2190078464}
{"step": 5450, "samples": 174400, "loss": 0.08857046812772751, "elapsed_s": 582.3250326790148, "gpu_peak_bytes": 2190078464}
{"step": 5500, "samples": 176000, "loss": 0.0911763422191143, "elapsed_s": 587.6202087629354, "gpu_peak_bytes": 2190078464}
{"step": 5550, "samples": 177600, "loss": 0.10865067690610886, "elapsed_s": 592.9077906690072, "gpu_peak_bytes": 2190078464}
{"step": 5600, "samples": 179200, "loss": 0.08388276770710945, "elapsed_s": 598.0528621770209, "gpu_peak_bytes": 2190078464}
{"step": 5650, "samples": 180800, "loss": 0.08962345495820045, "elapsed_s": 603.7647069590166, "gpu_peak_bytes": 2190078464}
{"step": 5700, "samples": 182400, "loss": 0.09007006138563156, "elapsed_s": 609.0387843620265, "gpu_peak_bytes": 2190078464}
{"step": 5750, "samples": 184000, "loss": 0.08023739606142044, "elapsed_s": 614.3134286720306, "gpu_peak_bytes": 2190078464}
{"step": 5800, "samples": 185600, "loss": 0.07455030456185341, "elapsed_s": 619.5932406599168, "gpu_peak_bytes": 2190078464}
{"step": 5850, "samples": 187200, "loss": 0.0812087133526802, "elapsed_s": 624.8750578549225, "gpu_peak_bytes": 2190078464}
{"step": 5900, "samples": 188800, "loss": 0.0797996036708355, "elapsed_s": 630.1582108959556, "gpu_peak_bytes": 2190078464}
{"step": 5950, "samples": 190400, "loss": 0.07794725522398949, "elapsed_s": 635.440338886925, "gpu_peak_bytes": 2190078464}
{"step": 6000, "samples": 192000, "loss": 0.07880613952875137, "elapsed_s": 640.5693386449711, "gpu_peak_bytes": 2190078464}
{"step": 6050, "samples": 193600, "loss": 0.07678492367267609, "elapsed_s": 646.2362048039213, "gpu_peak_bytes": 2190078464}
{"step": 6100, "samples": 195200, "loss": 0.08299105986952782, "elapsed_s": 651.5170563899446, "gpu_peak_bytes": 2190078464}
{"step": 6150, "samples": 196800, "loss": 0.0789850689470768, "elapsed_s": 656.8045526449569, "gpu_peak_bytes": 2190078464}
{"step": 6200, "samples": 198400, "loss": 0.06674584001302719, "elapsed_s": 662.0934377559461, "gpu_peak_bytes": 2190078464}
{"step": 6250, "samples": 200000, "loss": 0.0754854679107666, "elapsed_s": 667.3800009029219, "gpu_peak_bytes": 2190078464}
{"step": 6300, "samples": 201600, "loss": 0.08739304542541504, "elapsed_s": 672.6741184720304, "gpu_peak_bytes": 2190078464}
{"step": 6350, "samples": 203200, "loss": 0.07863003760576248, "elapsed_s": 677.9716675769305, "gpu_peak_bytes": 2190078464}
{"step": 6400, "samples": 204800, "loss": 0.07405796274542809, "elapsed_s": 683.2632473849226, "gpu_peak_bytes": 2190078464}
{"step": 6450, "samples": 206400, "loss": 0.06385591253638268, "elapsed_s": 688.5596935319481, "gpu_peak_bytes": 2190078464}
{"step": 6500, "samples": 208000, "loss": 0.07706351950764656, "elapsed_s": 693.8498667199165, "gpu_peak_bytes": 2190078464}
{"step": 6550, "samples": 209600, "loss": 0.0949971005320549, "elapsed_s": 699.1534362379462, "gpu_peak_bytes": 2190078464}
{"step": 6600, "samples": 211200, "loss": 0.07463876903057098, "elapsed_s": 704.4570527540054, "gpu_peak_bytes": 2190078464}
{"step": 6650, "samples": 212800, "loss": 0.07724759727716446, "elapsed_s": 709.7623565789545, "gpu_peak_bytes": 2190078464}
{"step": 6700, "samples": 214400, "loss": 0.07673726230859756, "elapsed_s": 715.0676578689599, "gpu_peak_bytes": 2190078464}
{"step": 6750, "samples": 216000, "loss": 0.07798149064183235, "elapsed_s": 720.3702855219599, "gpu_peak_bytes": 2190078464}
{"step": 6800, "samples": 217600, "loss": 0.09271221607923508, "elapsed_s": 725.6752624609508, "gpu_peak_bytes": 2190078464}
{"step": 6850, "samples": 219200, "loss": 0.053850844502449036, "elapsed_s": 730.9773093799595, "gpu_peak_bytes": 2190078464}
{"step": 6900, "samples": 220800, "loss": 0.08091879263520241, "elapsed_s": 736.2783574669156, "gpu_peak_bytes": 2190078464}
{"step": 6950, "samples": 222400, "loss": 0.06589839234948158, "elapsed_s": 741.573085877928, "gpu_peak_bytes": 2190078464}
{"step": 7000, "samples": 224000, "loss": 0.06575567275285721, "elapsed_s": 746.6866996380268, "gpu_peak_bytes": 2190078464}
{"step": 7050, "samples": 225600, "loss": 0.08105223253369331, "elapsed_s": 752.3583228399511, "gpu_peak_bytes": 2190078464}
{"step": 7100, "samples": 227200, "loss": 0.06573525257408619, "elapsed_s": 757.6603146749549, "gpu_peak_bytes": 2190078464}
{"step": 7150, "samples": 228800, "loss": 0.06553252786397934, "elapsed_s": 762.9588967359159, "gpu_peak_bytes": 2190078464}
{"step": 7200, "samples": 230400, "loss": 0.07314055413007736, "elapsed_s": 768.2560510570183, "gpu_peak_bytes": 2190078464}
{"step": 7250, "samples": 232000, "loss": 0.06899807788431644, "elapsed_s": 773.5592623390257, "gpu_peak_bytes": 2190078464}
{"step": 7300, "samples": 233600, "loss": 0.07506545633077621, "elapsed_s": 778.8630028689513, "gpu_peak_bytes": 2190078464}
{"step": 7350, "samples": 235200, "loss": 0.04956291615962982, "elapsed_s": 784.1579028229462, "gpu_peak_bytes": 2190078464}
{"step": 7400, "samples": 236800, "loss": 0.06835357658565044, "elapsed_s": 789.4552695639431, "gpu_peak_bytes": 2190078464}
{"step": 7450, "samples": 238400, "loss": 0.05683075822889805, "elapsed_s": 794.7537380509311, "gpu_peak_bytes": 2190078464}
{"step": 7500, "samples": 240000, "loss": 0.06077123433351517, "elapsed_s": 800.0550153800286, "gpu_peak_bytes": 2190078464}
{"step": 7550, "samples": 241600, "loss": 0.05875101871788502, "elapsed_s": 805.349426663015, "gpu_peak_bytes": 2190078464}
{"step": 7600, "samples": 243200, "loss": 0.054195838049054146, "elapsed_s": 810.6510692599695, "gpu_peak_bytes": 2190078464}
{"step": 7650, "samples": 244800, "loss": 0.08166589587926865, "elapsed_s": 815.9486721659778, "gpu_peak_bytes": 2190078464}
{"step": 7700, "samples": 246400, "loss": 0.061700401827692986, "elapsed_s": 821.2464913500007, "gpu_peak_bytes": 2190078464}
{"step": 7750, "samples": 248000, "loss": 0.08171028271317482, "elapsed_s": 826.545734415995, "gpu_peak_bytes": 2190078464}
{"step": 7800, "samples": 249600, "loss": 0.07927760109305382, "elapsed_s": 831.8435447629308, "gpu_peak_bytes": 2190078464}
{"step": 7850, "samples": 251200, "loss": 0.05411558412015438, "elapsed_s": 837.1508793840185, "gpu_peak_bytes": 2190078464}
{"step": 7900, "samples": 252800, "loss": 0.07342825084924698, "elapsed_s": 842.450374195003, "gpu_peak_bytes": 2190078464}
{"step": 7950, "samples": 254400, "loss": 0.06561771407723427, "elapsed_s": 847.7479039740283, "gpu_peak_bytes": 2190078464}
{"step": 8000, "samples": 256000, "loss": 0.06295766495168209, "elapsed_s": 852.8428210580023, "gpu_peak_bytes": 2190078464}
{"step": 8050, "samples": 257600, "loss": 0.05916598439216614, "elapsed_s": 858.5076983399922, "gpu_peak_bytes": 2190078464}
{"step": 8100, "samples": 259200, "loss": 0.05610601790249348, "elapsed_s": 863.7975550019182, "gpu_peak_bytes": 2190078464}
{"step": 8150, "samples": 260800, "loss": 0.05285186879336834, "elapsed_s": 869.094424586976, "gpu_peak_bytes": 2190078464}
{"step": 8200, "samples": 262400, "loss": 0.07677793502807617, "elapsed_s": 874.3972710319795, "gpu_peak_bytes": 2190078464}
{"step": 8250, "samples": 264000, "loss": 0.05513322725892067, "elapsed_s": 879.7199465659214, "gpu_peak_bytes": 2190078464}
{"step": 8300, "samples": 265600, "loss": 0.0633285716176033, "elapsed_s": 885.0212917729514, "gpu_peak_bytes": 2190078464}
{"step": 8350, "samples": 267200, "loss": 0.08028282225131989, "elapsed_s": 890.3130337470211, "gpu_peak_bytes": 2190078464}
{"step": 8400, "samples": 268800, "loss": 0.07577314972877502, "elapsed_s": 895.6166184230242, "gpu_peak_bytes": 2190078464}
{"step": 8450, "samples": 270400, "loss": 0.05280552990734577, "elapsed_s": 900.923837927985, "gpu_peak_bytes": 2190078464}
{"step": 8500, "samples": 272000, "loss": 0.04760557599365711, "elapsed_s": 906.2173836400034, "gpu_peak_bytes": 2190078464}
{"step": 8550, "samples": 273600, "loss": 0.055924782529473305, "elapsed_s": 911.517907239031, "gpu_peak_bytes": 2190078464}
{"step": 8600, "samples": 275200, "loss": 0.06880510225892067, "elapsed_s": 916.8191737469751, "gpu_peak_bytes": 2190078464}
{"step": 8650, "samples": 276800, "loss": 0.05963336117565632, "elapsed_s": 922.1204665049445, "gpu_peak_bytes": 2190078464}
{"step": 8700, "samples": 278400, "loss": 0.07123910263180733, "elapsed_s": 927.4307237929897, "gpu_peak_bytes": 2190078464}
{"step": 8750, "samples": 280000, "loss": 0.05914436839520931, "elapsed_s": 932.7282071139198, "gpu_peak_bytes": 2190078464}
{"step": 8800, "samples": 281600, "loss": 0.05580343306064606, "elapsed_s": 938.0344093119493, "gpu_peak_bytes": 2190078464}
{"step": 8850, "samples": 283200, "loss": 0.0711045190691948, "elapsed_s": 943.3380101380171, "gpu_peak_bytes": 2190078464}
{"step": 8900, "samples": 284800, "loss": 0.06393722631037235, "elapsed_s": 948.6525514869718, "gpu_peak_bytes": 2190078464}
{"step": 8950, "samples": 286400, "loss": 0.0754445493221283, "elapsed_s": 953.9612621309934, "gpu_peak_bytes": 2190078464}
{"step": 9000, "samples": 288000, "loss": 0.06553420051932335, "elapsed_s": 959.0653455660213, "gpu_peak_bytes": 2190078464}
{"step": 9050, "samples": 289600, "loss": 0.05917549505829811, "elapsed_s": 964.6968880229397, "gpu_peak_bytes": 2190078464}
{"step": 9100, "samples": 291200, "loss": 0.061731498688459396, "elapsed_s": 970.0018244059756, "gpu_peak_bytes": 2190078464}
{"step": 9150, "samples": 292800, "loss": 0.05147015489637852, "elapsed_s": 975.3048329669982, "gpu_peak_bytes": 2190078464}
{"step": 9200, "samples": 294400, "loss": 0.039514150470495224, "elapsed_s": 980.6062755769817, "gpu_peak_bytes": 2190078464}
{"step": 9250, "samples": 296000, "loss": 0.052239976823329926, "elapsed_s": 985.9095305219525, "gpu_peak_bytes": 2190078464}
{"step": 9300, "samples": 297600, "loss": 0.0823059231042862, "elapsed_s": 991.2104810549645, "gpu_peak_bytes": 2190078464}
{"step": 9350, "samples": 299200, "loss": 0.07022574543952942, "elapsed_s": 996.5114414669806, "gpu_peak_bytes": 2190078464}
{"step": 9400, "samples": 300800, "loss": 0.0612080879509449, "elapsed_s": 1001.8123968989821, "gpu_peak_bytes": 2190078464}
{"step": 9450, "samples": 302400, "loss": 0.05824749916791916, "elapsed_s": 1007.1175941040274, "gpu_peak_bytes": 2190078464}
{"step": 9500, "samples": 304000, "loss": 0.06950371526181698, "elapsed_s": 1012.4239548899932, "gpu_peak_bytes": 2190078464}
{"step": 9550, "samples": 305600, "loss": 0.058449091389775276, "elapsed_s": 1017.7294642319903, "gpu_peak_bytes": 2190078464}
{"step": 9600, "samples": 307200, "loss": 0.06072451174259186, "elapsed_s": 1023.0350917970063, "gpu_peak_bytes": 2190078464}
{"step": 9650, "samples": 308800, "loss": 0.064135005697608, "elapsed_s": 1028.34071906202, "gpu_peak_bytes": 2190078464}
{"step": 9700, "samples": 310400, "loss": 0.05579986982047558, "elapsed_s": 1033.6427948049968, "gpu_peak_bytes": 2190078464}
{"step": 9750, "samples": 312000, "loss": 0.053283385932445526, "elapsed_s": 1038.9470368199982, "gpu_peak_bytes": 2190078464}
{"step": 9800, "samples": 313600, "loss": 0.04938000440597534, "elapsed_s": 1044.248222435941, "gpu_peak_bytes": 2190078464}
{"step": 9850, "samples": 315200, "loss": 0.055233752354979515, "elapsed_s": 1049.5537422179477, "gpu_peak_bytes": 2190078464}
{"step": 9900, "samples": 316800, "loss": 0.05494244769215584, "elapsed_s": 1054.8562554349191, "gpu_peak_bytes": 2190078464}
{"step": 9950, "samples": 318400, "loss": 0.048886049538850784, "elapsed_s": 1060.15756371303, "gpu_peak_bytes": 2190078464}
{"step": 10000, "samples": 320000, "loss": 0.05215058475732803, "elapsed_s": 1064.4412331409985, "gpu_peak_bytes": 2190078464}