Upload exp_phase5_stage_a_v7_reasoning_t2_r128_FIXED_20260505_193938/log.jsonl with huggingface_hub

exp_phase5_stage_a_v7_reasoning_t2_r128_FIXED_20260505_193938/log.jsonl

@@ -11,3 +11,140 @@
 {"step": 220, "loss": 0.9342570304870605, "elapsed_s": 63.89808368682861}
 {"step": 240, "loss": 1.1639739274978638, "elapsed_s": 69.15105271339417}
 {"step": 260, "loss": 0.9906618595123291, "elapsed_s": 74.28221917152405}
+{"step": 280, "loss": 0.9217384457588196, "elapsed_s": 79.52862787246704}
+{"step": 300, "loss": 1.0134793519973755, "elapsed_s": 84.66254329681396}
+{"step": 320, "loss": 0.9703929424285889, "elapsed_s": 89.68356895446777}
+{"step": 340, "loss": 0.934944748878479, "elapsed_s": 94.81912326812744}
+{"step": 360, "loss": 1.0074201822280884, "elapsed_s": 99.89717555046082}
+{"step": 380, "loss": 0.6672135591506958, "elapsed_s": 104.99657034873962}
+{"step": 400, "loss": 1.0285619497299194, "elapsed_s": 110.19235754013062}
+{"step": 420, "loss": 0.7391230463981628, "elapsed_s": 115.4782772064209}
+{"step": 440, "loss": 0.841978907585144, "elapsed_s": 120.51270604133606}
+{"step": 460, "loss": 1.0560815334320068, "elapsed_s": 125.65765833854675}
+{"step": 480, "loss": 0.6530693173408508, "elapsed_s": 130.65622186660767}
+{"step": 500, "loss": 0.7006113529205322, "elapsed_s": 135.74453568458557}
+{"step": 520, "loss": 0.8912713527679443, "elapsed_s": 140.99209880828857}
+{"step": 540, "loss": 0.8923031687736511, "elapsed_s": 146.01203083992004}
+{"step": 560, "loss": 0.7651110291481018, "elapsed_s": 151.10123348236084}
+{"step": 580, "loss": 0.8456388711929321, "elapsed_s": 156.1442632675171}
+{"step": 600, "loss": 0.5726004838943481, "elapsed_s": 161.09750866889954}
+{"step": 620, "loss": 0.7869890928268433, "elapsed_s": 166.1252555847168}
+{"step": 640, "loss": 0.6021872162818909, "elapsed_s": 171.1144254207611}
+{"step": 660, "loss": 0.7394794821739197, "elapsed_s": 176.17365884780884}
+{"step": 680, "loss": 0.6975713968276978, "elapsed_s": 181.4919183254242}
+{"step": 700, "loss": 0.6365028023719788, "elapsed_s": 186.3664162158966}
+{"step": 720, "loss": 0.9330570101737976, "elapsed_s": 191.32875537872314}
+{"step": 740, "loss": 0.8374877572059631, "elapsed_s": 196.4091992378235}
+{"step": 760, "loss": 0.7793827056884766, "elapsed_s": 201.38521718978882}
+{"step": 780, "loss": 0.604881227016449, "elapsed_s": 206.31816792488098}
+{"step": 800, "loss": 0.886054515838623, "elapsed_s": 211.34451532363892}
+{"step": 820, "loss": 0.7571917176246643, "elapsed_s": 216.3378665447235}
+{"step": 840, "loss": 0.8188043832778931, "elapsed_s": 221.21401405334473}
+{"step": 860, "loss": 0.5729079842567444, "elapsed_s": 226.2276954650879}
+{"step": 880, "loss": 0.7704076170921326, "elapsed_s": 231.11284494400024}
+{"step": 900, "loss": 0.5692376494407654, "elapsed_s": 235.97811269760132}
+{"step": 920, "loss": 0.902776300907135, "elapsed_s": 240.82952046394348}
+{"step": 940, "loss": 0.6443384289741516, "elapsed_s": 245.81543111801147}
+{"step": 960, "loss": 0.6292765736579895, "elapsed_s": 250.68277168273926}
+{"step": 980, "loss": 0.6460203528404236, "elapsed_s": 255.5410816669464}
+{"step": 1000, "loss": 0.690719723701477, "elapsed_s": 260.4375681877136}
+{"step": 1020, "loss": 0.8257105350494385, "elapsed_s": 265.3406116962433}
+{"step": 1040, "loss": 0.8443910479545593, "elapsed_s": 270.2501883506775}
+{"step": 1060, "loss": 0.9427888989448547, "elapsed_s": 274.97527503967285}
+{"step": 1080, "loss": 0.6492099165916443, "elapsed_s": 279.78276920318604}
+{"step": 1100, "loss": 0.7547662854194641, "elapsed_s": 284.7539110183716}
+{"step": 1120, "loss": 0.7964253425598145, "elapsed_s": 289.51826453208923}
+{"step": 1140, "loss": 0.5524964928627014, "elapsed_s": 294.37365674972534}
+{"step": 1160, "loss": 0.7119346261024475, "elapsed_s": 299.2078764438629}
+{"step": 1180, "loss": 0.6448314189910889, "elapsed_s": 304.03602147102356}
+{"step": 1200, "loss": 0.9455552697181702, "elapsed_s": 308.9580409526825}
+{"step": 1220, "loss": 0.514221727848053, "elapsed_s": 313.68317890167236}
+{"step": 1240, "loss": 0.8191384077072144, "elapsed_s": 318.56613516807556}
+{"step": 1260, "loss": 0.7116103768348694, "elapsed_s": 323.33849143981934}
+{"step": 1280, "loss": 0.3450673222541809, "elapsed_s": 328.32604789733887}
+{"step": 1300, "loss": 0.646104633808136, "elapsed_s": 333.17261838912964}
+{"step": 1320, "loss": 0.7678859233856201, "elapsed_s": 337.98112058639526}
+{"step": 1340, "loss": 0.7468214631080627, "elapsed_s": 342.98217272758484}
+{"step": 1360, "loss": 0.6010539531707764, "elapsed_s": 347.7629635334015}
+{"step": 1380, "loss": 0.42219147086143494, "elapsed_s": 352.5723373889923}
+{"step": 1400, "loss": 0.7000728249549866, "elapsed_s": 357.35963773727417}
+{"step": 1420, "loss": 0.44703811407089233, "elapsed_s": 362.2910363674164}
+{"step": 1440, "loss": 0.7499022483825684, "elapsed_s": 367.2216486930847}
+{"step": 1460, "loss": 0.4872989356517792, "elapsed_s": 372.018043756485}
+{"step": 1480, "loss": 0.785200834274292, "elapsed_s": 376.82967615127563}
+{"step": 1500, "loss": 0.7409285306930542, "elapsed_s": 381.59858679771423}
+{"step": 1520, "loss": 0.5835148096084595, "elapsed_s": 386.50145411491394}
+{"step": 1540, "loss": 0.5633326768875122, "elapsed_s": 391.41010093688965}
+{"step": 1560, "loss": 0.5728590488433838, "elapsed_s": 396.1072106361389}
+{"step": 1580, "loss": 0.7403851747512817, "elapsed_s": 400.9511921405792}
+{"step": 1600, "loss": 0.6663960218429565, "elapsed_s": 405.6753566265106}
+{"step": 1620, "loss": 0.6360340714454651, "elapsed_s": 410.7599458694458}
+{"step": 1640, "loss": 0.8170230984687805, "elapsed_s": 415.5339388847351}
+{"step": 1660, "loss": 0.47957849502563477, "elapsed_s": 420.41765427589417}
+{"step": 1680, "loss": 0.5263388156890869, "elapsed_s": 425.08281326293945}
+{"step": 1700, "loss": 0.7186670899391174, "elapsed_s": 429.9818768501282}
+{"step": 1720, "loss": 0.6396377086639404, "elapsed_s": 434.89931654930115}
+{"step": 1740, "loss": 0.7117732167243958, "elapsed_s": 439.65157318115234}
+{"step": 1760, "loss": 0.6532768607139587, "elapsed_s": 444.43253207206726}
+{"step": 1780, "loss": 0.5249560475349426, "elapsed_s": 449.24021005630493}
+{"step": 1800, "loss": 0.776771068572998, "elapsed_s": 454.2873365879059}
+{"step": 1820, "loss": 0.698784589767456, "elapsed_s": 459.029944896698}
+{"step": 1840, "loss": 0.6735374927520752, "elapsed_s": 463.77927827835083}
+{"step": 1860, "loss": 0.6430699229240417, "elapsed_s": 468.546804189682}
+{"step": 1880, "loss": 0.5950252413749695, "elapsed_s": 473.2620642185211}
+{"step": 1900, "loss": 0.8065794110298157, "elapsed_s": 478.14286160469055}
+{"step": 1920, "loss": 0.6855514049530029, "elapsed_s": 482.9271614551544}
+{"step": 1940, "loss": 0.5631692409515381, "elapsed_s": 487.5945448875427}
+{"step": 1960, "loss": 0.7875005602836609, "elapsed_s": 492.35640025138855}
+{"step": 1980, "loss": 0.690204918384552, "elapsed_s": 497.15804719924927}
+{"step": 2000, "loss": 0.6040239930152893, "elapsed_s": 501.9295856952667}
+{"step": 2020, "loss": 0.6380578875541687, "elapsed_s": 506.74664974212646}
+{"step": 2040, "loss": 0.5394318103790283, "elapsed_s": 511.5181906223297}
+{"step": 2060, "loss": 0.7548812031745911, "elapsed_s": 516.2892076969147}
+{"step": 2080, "loss": 0.842311441898346, "elapsed_s": 521.0991258621216}
+{"step": 2100, "loss": 0.664849042892456, "elapsed_s": 525.896852016449}
+{"step": 2120, "loss": 0.7084640264511108, "elapsed_s": 530.6799962520599}
+{"step": 2140, "loss": 0.5340833067893982, "elapsed_s": 535.4615280628204}
+{"step": 2160, "loss": 0.5146570205688477, "elapsed_s": 540.1990118026733}
+{"step": 2180, "loss": 0.6764290928840637, "elapsed_s": 544.9808127880096}
+{"step": 2200, "loss": 0.7196254134178162, "elapsed_s": 549.7997944355011}
+{"step": 2220, "loss": 0.663967490196228, "elapsed_s": 554.6405785083771}
+{"step": 2240, "loss": 0.5817448496818542, "elapsed_s": 559.393562078476}
+{"step": 2260, "loss": 0.6752156019210815, "elapsed_s": 564.1076827049255}
+{"step": 2280, "loss": 0.7756117582321167, "elapsed_s": 568.8703010082245}
+{"step": 2300, "loss": 0.8056526184082031, "elapsed_s": 573.7152421474457}
+{"step": 2320, "loss": 0.7155594229698181, "elapsed_s": 578.5602447986603}
+{"step": 2340, "loss": 0.5922985076904297, "elapsed_s": 583.4288778305054}
+{"step": 2360, "loss": 0.6393532752990723, "elapsed_s": 588.172794342041}
+{"step": 2380, "loss": 0.680176854133606, "elapsed_s": 592.9839241504669}
+{"step": 2400, "loss": 0.48348119854927063, "elapsed_s": 597.7881705760956}
+{"step": 2420, "loss": 0.7194271087646484, "elapsed_s": 602.5470266342163}
+{"step": 2440, "loss": 0.3756081759929657, "elapsed_s": 607.312563419342}
+{"step": 2460, "loss": 0.6777275204658508, "elapsed_s": 612.092348575592}
+{"step": 2480, "loss": 0.7009856104850769, "elapsed_s": 616.8879475593567}
+{"step": 2500, "loss": 0.6018471121788025, "elapsed_s": 621.657110452652}
+{"step": 2520, "loss": 0.6428532004356384, "elapsed_s": 626.4403429031372}
+{"step": 2540, "loss": 0.5736479163169861, "elapsed_s": 631.2488265037537}
+{"step": 2560, "loss": 0.5846017599105835, "elapsed_s": 635.997076511383}
+{"step": 2580, "loss": 0.521944522857666, "elapsed_s": 640.8334317207336}
+{"step": 2600, "loss": 0.381876677274704, "elapsed_s": 645.5900771617889}
+{"step": 2620, "loss": 0.4446439743041992, "elapsed_s": 650.3797936439514}
+{"step": 2640, "loss": 0.5909258127212524, "elapsed_s": 655.1421356201172}
+{"step": 2660, "loss": 0.7309444546699524, "elapsed_s": 659.8624119758606}
+{"step": 2680, "loss": 0.6599928140640259, "elapsed_s": 664.6286828517914}
+{"step": 2700, "loss": 0.6313130855560303, "elapsed_s": 669.4205410480499}
+{"step": 2720, "loss": 0.5966728329658508, "elapsed_s": 674.1117067337036}
+{"step": 2740, "loss": 0.5969704985618591, "elapsed_s": 678.9011998176575}
+{"step": 2760, "loss": 0.5524131655693054, "elapsed_s": 683.6077523231506}
+{"step": 2780, "loss": 0.656955361366272, "elapsed_s": 688.2720732688904}
+{"step": 2800, "loss": 0.5542805194854736, "elapsed_s": 693.0075232982635}
+{"step": 2820, "loss": 0.7230316400527954, "elapsed_s": 697.7881722450256}
+{"step": 2840, "loss": 0.6691194176673889, "elapsed_s": 702.5299088954926}
+{"step": 2860, "loss": 0.37612462043762207, "elapsed_s": 707.2653095722198}
+{"step": 2880, "loss": 0.6314190626144409, "elapsed_s": 712.0567135810852}
+{"step": 2900, "loss": 0.5167427062988281, "elapsed_s": 716.8392879962921}
+{"step": 2920, "loss": 0.5582223534584045, "elapsed_s": 721.4655737876892}
+{"step": 2940, "loss": 0.6079427003860474, "elapsed_s": 726.1486926078796}
+{"step": 2960, "loss": 0.5880207419395447, "elapsed_s": 730.9058756828308}
+{"step": 2980, "loss": 0.47934576869010925, "elapsed_s": 735.6948935985565}
+{"step": 3000, "loss": 0.6230853796005249, "elapsed_s": 740.4467792510986}