Upload exp_phase5_stage_a_v7_reasoning_t1_r64_FIXED_20260502_122607/log.jsonl with huggingface_hub

exp_phase5_stage_a_v7_reasoning_t1_r64_FIXED_20260502_122607/log.jsonl

@@ -0,0 +1,150 @@
+{"step": 20, "loss": 1.973817229270935, "elapsed_s": 6.976860046386719}
+{"step": 40, "loss": 2.1782829761505127, "elapsed_s": 11.232503890991211}
+{"step": 60, "loss": 1.8894329071044922, "elapsed_s": 14.983447074890137}
+{"step": 80, "loss": 1.949980616569519, "elapsed_s": 18.760164976119995}
+{"step": 100, "loss": 1.2607388496398926, "elapsed_s": 22.497336864471436}
+{"step": 120, "loss": 1.6538296937942505, "elapsed_s": 26.311240196228027}
+{"step": 140, "loss": 1.4860097169876099, "elapsed_s": 29.978839874267578}
+{"step": 160, "loss": 1.148026466369629, "elapsed_s": 33.65741944313049}
+{"step": 180, "loss": 0.9114201068878174, "elapsed_s": 37.275312185287476}
+{"step": 200, "loss": 1.3209513425827026, "elapsed_s": 40.76411557197571}
+{"step": 220, "loss": 1.047371745109558, "elapsed_s": 44.33280086517334}
+{"step": 240, "loss": 1.3467183113098145, "elapsed_s": 47.86751103401184}
+{"step": 260, "loss": 1.1109836101531982, "elapsed_s": 51.42971634864807}
+{"step": 280, "loss": 0.9390279054641724, "elapsed_s": 54.865243673324585}
+{"step": 300, "loss": 0.8503736853599548, "elapsed_s": 58.34541702270508}
+{"step": 320, "loss": 1.1762611865997314, "elapsed_s": 61.95381546020508}
+{"step": 340, "loss": 0.9929231405258179, "elapsed_s": 65.46165442466736}
+{"step": 360, "loss": 1.1343711614608765, "elapsed_s": 68.88995385169983}
+{"step": 380, "loss": 0.8162987232208252, "elapsed_s": 72.40433645248413}
+{"step": 400, "loss": 1.2847820520401, "elapsed_s": 75.87081289291382}
+{"step": 420, "loss": 0.849372923374176, "elapsed_s": 79.38133645057678}
+{"step": 440, "loss": 0.8933641910552979, "elapsed_s": 82.85486340522766}
+{"step": 460, "loss": 0.975671112537384, "elapsed_s": 86.1804838180542}
+{"step": 480, "loss": 0.7396590113639832, "elapsed_s": 89.51736402511597}
+{"step": 500, "loss": 0.9198058247566223, "elapsed_s": 92.8671703338623}
+{"step": 520, "loss": 0.496670663356781, "elapsed_s": 96.50988125801086}
+{"step": 540, "loss": 0.9571793675422668, "elapsed_s": 99.76973581314087}
+{"step": 560, "loss": 0.9462494254112244, "elapsed_s": 103.14423441886902}
+{"step": 580, "loss": 1.0090746879577637, "elapsed_s": 106.56261539459229}
+{"step": 600, "loss": 0.8946813344955444, "elapsed_s": 109.89121794700623}
+{"step": 620, "loss": 0.9744301438331604, "elapsed_s": 113.38467979431152}
+{"step": 640, "loss": 0.7653281688690186, "elapsed_s": 116.6873209476471}
+{"step": 660, "loss": 0.5985884070396423, "elapsed_s": 120.06101536750793}
+{"step": 680, "loss": 0.9063829183578491, "elapsed_s": 123.59697771072388}
+{"step": 700, "loss": 0.7607925534248352, "elapsed_s": 127.08399605751038}
+{"step": 720, "loss": 0.6480311751365662, "elapsed_s": 130.33123183250427}
+{"step": 740, "loss": 0.8139599561691284, "elapsed_s": 133.86749386787415}
+{"step": 760, "loss": 0.6840896010398865, "elapsed_s": 137.25632739067078}
+{"step": 780, "loss": 0.9566132426261902, "elapsed_s": 140.47453570365906}
+{"step": 800, "loss": 0.9903825521469116, "elapsed_s": 143.8175015449524}
+{"step": 820, "loss": 0.845392644405365, "elapsed_s": 147.24301195144653}
+{"step": 840, "loss": 0.6428900361061096, "elapsed_s": 150.54200768470764}
+{"step": 860, "loss": 0.739020049571991, "elapsed_s": 153.88429808616638}
+{"step": 880, "loss": 0.8078552484512329, "elapsed_s": 157.09798431396484}
+{"step": 900, "loss": 1.1548645496368408, "elapsed_s": 160.33105206489563}
+{"step": 920, "loss": 0.8238468170166016, "elapsed_s": 163.69431447982788}
+{"step": 940, "loss": 0.9773339629173279, "elapsed_s": 167.1659197807312}
+{"step": 960, "loss": 0.7266267538070679, "elapsed_s": 170.52547907829285}
+{"step": 980, "loss": 0.7149640917778015, "elapsed_s": 173.75434112548828}
+{"step": 1000, "loss": 1.1688445806503296, "elapsed_s": 177.11836862564087}
+{"step": 1020, "loss": 1.0162376165390015, "elapsed_s": 180.4692165851593}
+{"step": 1040, "loss": 1.2105255126953125, "elapsed_s": 183.69739627838135}
+{"step": 1060, "loss": 0.9312613606452942, "elapsed_s": 187.06126165390015}
+{"step": 1080, "loss": 0.7988710999488831, "elapsed_s": 190.3137719631195}
+{"step": 1100, "loss": 0.9297041893005371, "elapsed_s": 193.52304100990295}
+{"step": 1120, "loss": 0.6455752849578857, "elapsed_s": 196.77700233459473}
+{"step": 1140, "loss": 0.7713839411735535, "elapsed_s": 200.04302883148193}
+{"step": 1160, "loss": 0.9001843929290771, "elapsed_s": 203.24907159805298}
+{"step": 1180, "loss": 0.6534145474433899, "elapsed_s": 206.54895305633545}
+{"step": 1200, "loss": 0.6055432558059692, "elapsed_s": 209.70283913612366}
+{"step": 1220, "loss": 0.8679221272468567, "elapsed_s": 212.87886333465576}
+{"step": 1240, "loss": 0.7732729315757751, "elapsed_s": 216.12783360481262}
+{"step": 1260, "loss": 0.47922825813293457, "elapsed_s": 219.34244632720947}
+{"step": 1280, "loss": 0.6272779703140259, "elapsed_s": 222.53251671791077}
+{"step": 1300, "loss": 0.635953426361084, "elapsed_s": 225.76652550697327}
+{"step": 1320, "loss": 0.912196159362793, "elapsed_s": 229.0540006160736}
+{"step": 1340, "loss": 0.730510413646698, "elapsed_s": 232.16071319580078}
+{"step": 1360, "loss": 1.0338305234909058, "elapsed_s": 235.37427234649658}
+{"step": 1380, "loss": 0.8748066425323486, "elapsed_s": 238.75236010551453}
+{"step": 1400, "loss": 0.5838791131973267, "elapsed_s": 241.93777704238892}
+{"step": 1420, "loss": 0.800075888633728, "elapsed_s": 245.16200470924377}
+{"step": 1440, "loss": 0.8158484101295471, "elapsed_s": 248.2453920841217}
+{"step": 1460, "loss": 0.7291438579559326, "elapsed_s": 251.54531908035278}
+{"step": 1480, "loss": 0.4816175699234009, "elapsed_s": 254.71285223960876}
+{"step": 1500, "loss": 0.9324489235877991, "elapsed_s": 258.0197603702545}
+{"step": 1520, "loss": 0.8991304636001587, "elapsed_s": 261.07587218284607}
+{"step": 1540, "loss": 0.9572902321815491, "elapsed_s": 264.2129952907562}
+{"step": 1560, "loss": 0.8675363063812256, "elapsed_s": 267.3243181705475}
+{"step": 1580, "loss": 0.8536784052848816, "elapsed_s": 270.5713620185852}
+{"step": 1600, "loss": 1.0429778099060059, "elapsed_s": 273.71581745147705}
+{"step": 1620, "loss": 0.8249198198318481, "elapsed_s": 277.0015707015991}
+{"step": 1640, "loss": 0.8385199308395386, "elapsed_s": 280.14160895347595}
+{"step": 1660, "loss": 0.5946246385574341, "elapsed_s": 283.46949791908264}
+{"step": 1680, "loss": 0.6088995933532715, "elapsed_s": 286.65251302719116}
+{"step": 1700, "loss": 0.6426693201065063, "elapsed_s": 289.9599962234497}
+{"step": 1720, "loss": 0.9973137974739075, "elapsed_s": 293.13199734687805}
+{"step": 1740, "loss": 0.8216977715492249, "elapsed_s": 296.2782726287842}
+{"step": 1760, "loss": 0.9425555467605591, "elapsed_s": 299.51996064186096}
+{"step": 1780, "loss": 0.4133487045764923, "elapsed_s": 302.699827671051}
+{"step": 1800, "loss": 0.7588715553283691, "elapsed_s": 305.733439207077}
+{"step": 1820, "loss": 0.5243424773216248, "elapsed_s": 309.09634017944336}
+{"step": 1840, "loss": 0.7482855916023254, "elapsed_s": 312.2362449169159}
+{"step": 1860, "loss": 0.7549587488174438, "elapsed_s": 315.38091015815735}
+{"step": 1880, "loss": 0.5681294798851013, "elapsed_s": 318.5859158039093}
+{"step": 1900, "loss": 0.5199101567268372, "elapsed_s": 321.7301607131958}
+{"step": 1920, "loss": 0.8424093127250671, "elapsed_s": 324.914128780365}
+{"step": 1940, "loss": 0.5798994302749634, "elapsed_s": 328.04201579093933}
+{"step": 1960, "loss": 0.8768720030784607, "elapsed_s": 331.2297728061676}
+{"step": 1980, "loss": 0.911787748336792, "elapsed_s": 334.49285554885864}
+{"step": 2000, "loss": 0.7323008179664612, "elapsed_s": 337.6248879432678}
+{"step": 2020, "loss": 0.8236690163612366, "elapsed_s": 340.7248103618622}
+{"step": 2040, "loss": 0.4660707712173462, "elapsed_s": 343.9435212612152}
+{"step": 2060, "loss": 0.6757634878158569, "elapsed_s": 347.10589027404785}
+{"step": 2080, "loss": 0.5912379026412964, "elapsed_s": 350.22826385498047}
+{"step": 2100, "loss": 0.5812445878982544, "elapsed_s": 353.37570905685425}
+{"step": 2120, "loss": 0.8579750061035156, "elapsed_s": 356.5699555873871}
+{"step": 2140, "loss": 0.9136027097702026, "elapsed_s": 359.7305145263672}
+{"step": 2160, "loss": 0.7357272505760193, "elapsed_s": 362.9609463214874}
+{"step": 2180, "loss": 0.7479361891746521, "elapsed_s": 366.06253695487976}
+{"step": 2200, "loss": 0.7563722133636475, "elapsed_s": 369.1480383872986}
+{"step": 2220, "loss": 0.7350270748138428, "elapsed_s": 372.2850465774536}
+{"step": 2240, "loss": 1.02490234375, "elapsed_s": 375.19779419898987}
+{"step": 2260, "loss": 0.6652093529701233, "elapsed_s": 377.36948561668396}
+{"step": 2280, "loss": 0.6055534482002258, "elapsed_s": 379.52470684051514}
+{"step": 2300, "loss": 0.8821242451667786, "elapsed_s": 381.8155083656311}
+{"step": 2320, "loss": 0.6045027375221252, "elapsed_s": 384.2314827442169}
+{"step": 2340, "loss": 0.9137253165245056, "elapsed_s": 386.5695881843567}
+{"step": 2360, "loss": 1.0844957828521729, "elapsed_s": 388.93946170806885}
+{"step": 2380, "loss": 0.7206993103027344, "elapsed_s": 391.26945543289185}
+{"step": 2400, "loss": 0.6361485719680786, "elapsed_s": 393.4761574268341}
+{"step": 2420, "loss": 0.7458500862121582, "elapsed_s": 395.737628698349}
+{"step": 2440, "loss": 0.8463431000709534, "elapsed_s": 398.084867477417}
+{"step": 2460, "loss": 0.6092759966850281, "elapsed_s": 400.4136896133423}
+{"step": 2480, "loss": 0.5300841331481934, "elapsed_s": 402.7135260105133}
+{"step": 2500, "loss": 0.5250701308250427, "elapsed_s": 404.96401619911194}
+{"step": 2520, "loss": 0.6566253900527954, "elapsed_s": 407.3171977996826}
+{"step": 2540, "loss": 0.7611895799636841, "elapsed_s": 409.62276220321655}
+{"step": 2560, "loss": 0.866986870765686, "elapsed_s": 411.8545308113098}
+{"step": 2580, "loss": 0.7364883422851562, "elapsed_s": 414.18982887268066}
+{"step": 2600, "loss": 0.9745838046073914, "elapsed_s": 416.3678665161133}
+{"step": 2620, "loss": 0.6199561357498169, "elapsed_s": 418.54227113723755}
+{"step": 2640, "loss": 0.4964369833469391, "elapsed_s": 420.8347907066345}
+{"step": 2660, "loss": 0.7471553087234497, "elapsed_s": 423.10990357398987}
+{"step": 2680, "loss": 0.6030955910682678, "elapsed_s": 425.4000108242035}
+{"step": 2700, "loss": 0.8481877446174622, "elapsed_s": 427.78687930107117}
+{"step": 2720, "loss": 0.7215229868888855, "elapsed_s": 430.05453991889954}
+{"step": 2740, "loss": 0.8614094853401184, "elapsed_s": 432.2688732147217}
+{"step": 2760, "loss": 0.5816600322723389, "elapsed_s": 434.46415519714355}
+{"step": 2780, "loss": 0.788153886795044, "elapsed_s": 436.6338400840759}
+{"step": 2800, "loss": 0.7373696565628052, "elapsed_s": 438.9320707321167}
+{"step": 2820, "loss": 0.692425012588501, "elapsed_s": 441.3242299556732}
+{"step": 2840, "loss": 0.6165370345115662, "elapsed_s": 443.6583354473114}
+{"step": 2860, "loss": 0.6183592081069946, "elapsed_s": 445.9972424507141}
+{"step": 2880, "loss": 0.6578185558319092, "elapsed_s": 448.29111409187317}
+{"step": 2900, "loss": 0.8337782025337219, "elapsed_s": 450.53372836112976}
+{"step": 2920, "loss": 0.858396589756012, "elapsed_s": 452.82757687568665}
+{"step": 2940, "loss": 0.8891561031341553, "elapsed_s": 455.08113646507263}
+{"step": 2960, "loss": 1.0607068538665771, "elapsed_s": 457.27222633361816}
+{"step": 2980, "loss": 0.5947519540786743, "elapsed_s": 459.46335768699646}
+{"step": 3000, "loss": 0.5911017060279846, "elapsed_s": 461.6908037662506}