diff options
Diffstat (limited to 'results/fa_dfa_d512_L8_seed9/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L8_seed9/results_cifar10.json | 881 |
1 files changed, 881 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L8_seed9/results_cifar10.json b/results/fa_dfa_d512_L8_seed9/results_cifar10.json new file mode 100644 index 0000000..b70df2d --- /dev/null +++ b/results/fa_dfa_d512_L8_seed9/results_cifar10.json @@ -0,0 +1,881 @@ +{ + "9": { + "dfa": { + "log": { + "train_loss": [ + 2.0749315245056152, + 2.0436887923812868, + 2.0410750790023804, + 2.0344971646881103, + 2.030566175842285, + 2.0275797971343996, + 2.0244680599212646, + 2.024917350387573, + 2.027327563819885, + 2.018273809814453, + 2.0191563272094726, + 2.0152392296600343, + 2.0181246972274782, + 2.0139052614974977, + 2.011121483078003, + 2.007120831375122, + 2.009763352279663, + 2.010490125274658, + 2.0083817123413086, + 2.010725986175537, + 2.006365931854248, + 2.0057907654571534, + 2.006084675445557, + 2.0068956647491456, + 2.0047656631851196, + 2.005573451766968, + 2.003518327026367, + 2.003638141479492, + 2.0002300247192384, + 2.001433812713623, + 2.004064089279175, + 2.001738834075928, + 1.999301480255127, + 2.0016421214294433, + 1.9995300130462645, + 1.9971350913238526, + 2.002013059692383, + 2.000174626083374, + 1.9979997787475585, + 1.9982461312103272, + 1.9966425020599365, + 1.9979894361114503, + 1.9972480834579467, + 1.9982498080444335, + 1.9934808911514281, + 1.9956281592559815, + 1.99603251953125, + 1.9947702381134034, + 1.9941040051269532, + 1.9968071872329711, + 1.9977976748657227, + 1.9950814849853515, + 1.996515994567871, + 1.9940743815612794, + 1.9959130297470093, + 1.9953608598327637, + 1.995676156616211, + 1.9918495462036132, + 1.9927745639801024, + 1.9944643795013428, + 1.9957545782852173, + 1.99323220954895, + 1.9948831502914428, + 1.99567852684021, + 1.9960307669830322, + 1.993249538116455, + 1.992333110961914, + 1.9930713208007813, + 1.9917521324157714, + 1.9922201361846923, + 1.990846061668396, + 1.9919507978057862, + 1.993395569152832, + 1.9929897798919678, + 1.9927783852386474, + 1.992465722427368, + 1.9913483853149414, + 1.9910357999420165, + 1.9905700182724, + 1.9907023773956298, + 1.9905259375, + 1.9902730661010741, + 1.9915288207244872, + 1.9918687856292725, + 1.9918737398910522, + 1.9905504457092285, + 1.991807127685547, + 1.991159518737793, + 1.9909059362030028, + 1.987746954574585, + 1.9886708137512208, + 1.991118684539795, + 1.9877248714447022, + 1.9914885107040405, + 1.9913817571258545, + 1.990808463897705, + 1.9888677963638306, + 1.9901370908355713, + 1.9902113021850585, + 1.9896661295318603 + ], + "train_acc": [ + 0.233, + 0.24974, + 0.24902, + 0.2555, + 0.2535, + 0.25314, + 0.2584, + 0.25894, + 0.25562, + 0.25858, + 0.25952, + 0.26262, + 0.2592, + 0.26194, + 0.26504, + 0.26518, + 0.26234, + 0.26346, + 0.2654, + 0.2614, + 0.2644, + 0.26158, + 0.26868, + 0.26646, + 0.26696, + 0.26778, + 0.26766, + 0.26614, + 0.2709, + 0.26686, + 0.26754, + 0.26956, + 0.27002, + 0.27106, + 0.27018, + 0.27268, + 0.26956, + 0.26954, + 0.27276, + 0.2717, + 0.27314, + 0.27052, + 0.27232, + 0.2731, + 0.27476, + 0.27392, + 0.27266, + 0.27468, + 0.27238, + 0.27302, + 0.27042, + 0.27416, + 0.27348, + 0.27274, + 0.27358, + 0.27494, + 0.27386, + 0.27458, + 0.2768, + 0.27558, + 0.27496, + 0.27546, + 0.27332, + 0.27632, + 0.27632, + 0.27756, + 0.27488, + 0.27564, + 0.27622, + 0.27424, + 0.27614, + 0.27482, + 0.27548, + 0.2777, + 0.27614, + 0.27582, + 0.27502, + 0.27412, + 0.27556, + 0.2774, + 0.27578, + 0.27622, + 0.27676, + 0.27774, + 0.27778, + 0.27504, + 0.276, + 0.27678, + 0.2769, + 0.27788, + 0.2765, + 0.27636, + 0.27804, + 0.27786, + 0.27472, + 0.2759, + 0.2799, + 0.27662, + 0.27648, + 0.27812 + ], + "test_acc": [ + 0.2539, + 0.2722, + 0.2611, + 0.265, + 0.2773, + 0.2768, + 0.2858, + 0.2962, + 0.2833, + 0.2576, + 0.2732, + 0.2985, + 0.2776, + 0.2927, + 0.2865, + 0.2838, + 0.2825, + 0.283, + 0.2827, + 0.2834, + 0.2795, + 0.2866, + 0.2944, + 0.2847, + 0.297, + 0.2895, + 0.2903, + 0.2906, + 0.2904, + 0.2849, + 0.3028, + 0.2955, + 0.2932, + 0.2846, + 0.2898, + 0.2915, + 0.2827, + 0.3025, + 0.2708, + 0.2993, + 0.2941, + 0.2879, + 0.3013, + 0.2873, + 0.3017, + 0.3012, + 0.2932, + 0.3037, + 0.2939, + 0.2959, + 0.2899, + 0.2903, + 0.291, + 0.2999, + 0.292, + 0.3001, + 0.3011, + 0.299, + 0.292, + 0.3016, + 0.2875, + 0.301, + 0.2948, + 0.2952, + 0.2906, + 0.2961, + 0.2975, + 0.2937, + 0.2969, + 0.2938, + 0.3015, + 0.2993, + 0.3007, + 0.3012, + 0.3012, + 0.2974, + 0.2929, + 0.3003, + 0.2973, + 0.2988, + 0.301, + 0.2941, + 0.2997, + 0.2995, + 0.2979, + 0.2993, + 0.2998, + 0.2963, + 0.3029, + 0.2969, + 0.2976, + 0.2979, + 0.2958, + 0.2979, + 0.2971, + 0.2974, + 0.2977, + 0.2977, + 0.2977, + 0.2977 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3850422501564026, + 0.0012538384180516005, + 0.0006158786127343774, + -0.0003950851387344301, + 0.000671692076139152, + 0.00011835923214675859, + -0.00013412139378488064, + -0.0009948844090104103 + ], + "perturbation_rho": [ + -0.009509662166237831, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.688037395477295e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.0244548320770264e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -1.862645149230957e-09 + ], + "0.01": [ + -3.604218363761902e-06, + -8.381903171539307e-09, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + -1.862645149230957e-09 + ] + }, + "hidden_norms_per_layer": [ + 57815.328125, + 1306904576.0, + 3130353408.0, + 6042526720.0, + 6570734080.0, + 7140251648.0, + 10449125376.0, + 10382031872.0, + 10819989504.0 + ], + "bp_grad_norms_per_layer": [ + 2.387568827089126e-07, + 2.7390667511895117e-10, + 2.7443888828138086e-10, + 2.747598537578e-10, + 2.7503455068966787e-10, + 2.749568628335197e-10, + 2.749303285032312e-10, + 2.749313832151046e-10, + 2.7492974563614325e-10 + ] + }, + "drift": { + "embed.weight": 338.964341619299, + "embed.bias": 255.6470029656094, + "blocks.0.ln.weight": 9.7660440200238, + "blocks.0.w1.weight": 293.6860368082436, + "blocks.0.w1.bias": 259.7815916326078, + "blocks.0.w2.weight": 507.1631892613405, + "blocks.1.ln.weight": 8.6057770011893, + "blocks.1.w1.weight": 327.85104096601293, + "blocks.1.w1.bias": 315.96204358536954, + "blocks.1.w2.weight": 340.58083302512796, + "blocks.2.ln.weight": 9.649643863338673, + "blocks.2.w1.weight": 416.26890827062306, + "blocks.2.w1.bias": 387.4960514327406, + "blocks.2.w2.weight": 412.7077111053307, + "blocks.3.ln.weight": 8.443154226363815, + "blocks.3.w1.weight": 329.2601915023256, + "blocks.3.w1.bias": 308.6467481709907, + "blocks.3.w2.weight": 307.62522052468177, + "blocks.4.ln.weight": 8.653036768995564, + "blocks.4.w1.weight": 354.6654001843918, + "blocks.4.w1.bias": 335.86010162934014, + "blocks.4.w2.weight": 337.73623750534483, + "blocks.5.ln.weight": 11.320001722963196, + "blocks.5.w1.weight": 468.93827725593786, + "blocks.5.w1.bias": 433.59224246544227, + "blocks.5.w2.weight": 463.6712206215392, + "blocks.6.ln.weight": 8.063403696936168, + "blocks.6.w1.weight": 318.91137867276336, + "blocks.6.w1.bias": 306.65686506595944, + "blocks.6.w2.weight": 299.5449959675926, + "blocks.7.ln.weight": 8.29383434585519, + "blocks.7.w1.weight": 333.33394053359984, + "blocks.7.w1.bias": 309.27674432467256, + "blocks.7.w2.weight": 305.2383919884082, + "out_ln.weight": 0.6921153697505151, + "out_head.weight": 9.920257520136026, + "out_head.bias": 0.6756283909065531 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.054712998123169, + 1.9743254754638673, + 1.9496866632080079, + 1.9313411603164672, + 1.916562248878479, + 1.9087221280288695, + 1.8928792087173463, + 1.8860099752426147, + 1.8823522333526612, + 1.8718870676040649, + 1.8658955471038818, + 1.8600263873672485, + 1.861494415512085, + 1.8535768915176392, + 1.848890178527832, + 1.8419178400421143, + 1.843526672897339, + 1.8426634002685547, + 1.836815809020996, + 1.8399362145233153, + 1.8320682555770873, + 1.8253674346923827, + 1.8266462970733643, + 1.8268505213165283, + 1.819463009338379, + 1.815872594833374, + 1.8166155709075928, + 1.8089701477432252, + 1.8076338037490844, + 1.8046819745635987, + 1.8099775692749023, + 1.804258670349121, + 1.800314033203125, + 1.802857616043091, + 1.795948698425293, + 1.7952545282745362, + 1.7960923547363281, + 1.7939829620361327, + 1.7962147301483153, + 1.7921875751113892, + 1.7896775894927979, + 1.7867036238861085, + 1.7867207586669922, + 1.7872415407562257, + 1.782658879699707, + 1.780779080429077, + 1.7818151058959961, + 1.7821691263580322, + 1.7803733910369872, + 1.7798878842163086, + 1.7819366843032838, + 1.779062672958374, + 1.7781674340438842, + 1.7732916201019286, + 1.7719713204574585, + 1.7729518437957763, + 1.7732273389053346, + 1.7703840225982665, + 1.768649151878357, + 1.7698098583221435, + 1.7702393863296508, + 1.7676599294662476, + 1.7648498352813722, + 1.767660079307556, + 1.7671956900024415, + 1.7631733521270752, + 1.7601489026260375, + 1.7607763108062744, + 1.7581446547698976, + 1.7573072035980224, + 1.7607083935546874, + 1.7579230298233033, + 1.7577186632537842, + 1.7563656659317017, + 1.7547048263931275, + 1.7598694692230226, + 1.757342479248047, + 1.7562967572021484, + 1.7538646789169312, + 1.7546308361816407, + 1.753781385269165, + 1.7569293072128296, + 1.7535972839736937, + 1.756253225708008, + 1.752462748413086, + 1.750845112876892, + 1.7530096230316161, + 1.753320961074829, + 1.751194098739624, + 1.7476333127593995, + 1.7507887158966065, + 1.75599822265625, + 1.7483607616424561, + 1.750292029685974, + 1.7498931618881226, + 1.7509470245742798, + 1.7501217209243773, + 1.7520614090728759, + 1.750200431213379, + 1.7495508447265624 + ], + "train_acc": [ + 0.23466, + 0.27712, + 0.2872, + 0.29944, + 0.30162, + 0.30732, + 0.31528, + 0.31818, + 0.31696, + 0.32156, + 0.32474, + 0.32968, + 0.32702, + 0.33288, + 0.33656, + 0.33372, + 0.33702, + 0.33554, + 0.33892, + 0.33848, + 0.34342, + 0.34664, + 0.34692, + 0.34766, + 0.34448, + 0.3486, + 0.346, + 0.35034, + 0.35144, + 0.35384, + 0.35184, + 0.35544, + 0.35718, + 0.35464, + 0.35864, + 0.358, + 0.35758, + 0.35774, + 0.35722, + 0.36054, + 0.35958, + 0.36134, + 0.36046, + 0.3609, + 0.3655, + 0.36438, + 0.36688, + 0.36438, + 0.36392, + 0.36454, + 0.36186, + 0.36844, + 0.36298, + 0.3657, + 0.36936, + 0.3674, + 0.36656, + 0.36678, + 0.3693, + 0.36684, + 0.36768, + 0.36934, + 0.36892, + 0.37064, + 0.37, + 0.3704, + 0.37166, + 0.37122, + 0.3729, + 0.37234, + 0.37368, + 0.37366, + 0.37094, + 0.3736, + 0.3747, + 0.37176, + 0.37262, + 0.37366, + 0.374, + 0.37594, + 0.37354, + 0.374, + 0.37318, + 0.3751, + 0.375, + 0.3755, + 0.3763, + 0.3747, + 0.37738, + 0.3774, + 0.37554, + 0.37554, + 0.37726, + 0.37876, + 0.3747, + 0.37554, + 0.37696, + 0.37494, + 0.37724, + 0.37714 + ], + "test_acc": [ + 0.2763, + 0.3104, + 0.3093, + 0.3052, + 0.3209, + 0.3254, + 0.3374, + 0.3255, + 0.3509, + 0.3324, + 0.3485, + 0.3621, + 0.3572, + 0.3522, + 0.3638, + 0.3522, + 0.3562, + 0.3678, + 0.3662, + 0.3637, + 0.3713, + 0.372, + 0.3708, + 0.3802, + 0.3793, + 0.3859, + 0.3685, + 0.3665, + 0.3765, + 0.3782, + 0.3917, + 0.3905, + 0.3833, + 0.3867, + 0.3788, + 0.3843, + 0.3776, + 0.3794, + 0.3847, + 0.383, + 0.3929, + 0.392, + 0.392, + 0.3834, + 0.3857, + 0.3926, + 0.3914, + 0.387, + 0.3899, + 0.3859, + 0.3952, + 0.3883, + 0.3936, + 0.3901, + 0.3938, + 0.3899, + 0.3887, + 0.3938, + 0.3882, + 0.395, + 0.3878, + 0.3907, + 0.3952, + 0.3954, + 0.393, + 0.3931, + 0.3952, + 0.3999, + 0.3926, + 0.3963, + 0.3966, + 0.3973, + 0.3991, + 0.4007, + 0.3957, + 0.399, + 0.3988, + 0.3977, + 0.3955, + 0.395, + 0.3975, + 0.3971, + 0.3999, + 0.4003, + 0.3973, + 0.3988, + 0.3966, + 0.3972, + 0.4, + 0.3995, + 0.4008, + 0.3991, + 0.3981, + 0.3983, + 0.4006, + 0.3998, + 0.4001, + 0.4001, + 0.3992, + 0.3992 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.009833792224526405, + 0.05025029927492142, + 0.011137298308312893, + -0.039807502180337906, + -0.012322505004703999, + -0.023192159831523895, + 0.01924338936805725, + 0.9853411912918091 + ], + "perturbation_rho": [ + 0.009103747084736824, + -0.019891876727342606, + -0.013953445479273796, + -0.03727542236447334, + 0.01967245526611805, + 0.006721112877130508, + 0.013294404372572899, + 0.0555168092250824 + ], + "nudging": { + "0.001": [ + -6.459013093262911e-07, + -2.127781044691801e-07, + -3.3585820347070694e-08, + 5.916808731853962e-08, + 5.3318217396736145e-08, + 1.6763806343078613e-08, + 2.9802322387695312e-08, + -1.0665098670870066e-06 + ], + "0.003": [ + -1.868669642135501e-06, + -6.879272405058146e-07, + -5.1280949264764786e-08, + 2.151064109057188e-07, + 3.4226104617118835e-08, + 5.404581315815449e-08, + -1.3181124813854694e-07, + -3.7869031075388193e-06 + ], + "0.01": [ + -6.252725142985582e-06, + -2.3539469111710787e-06, + -1.600128598511219e-07, + 5.455221980810165e-07, + 1.2578675523400307e-07, + 2.0262086763978004e-07, + -3.8879807107150555e-07, + -1.3263634173199534e-05 + ] + }, + "hidden_norms_per_layer": [ + 9282.8193359375, + 113090.3984375, + 1271631.0, + 1386270.5, + 1726514.25, + 1932736.5, + 2234588.0, + 2239280.5, + 1909485.875 + ], + "bp_grad_norms_per_layer": [ + 2.4564533305238e-05, + 1.5358415339505882e-06, + 5.894418677598878e-07, + 5.805990781482251e-07, + 5.811043592984788e-07, + 5.801387032988714e-07, + 5.800572466796439e-07, + 5.741365498579398e-07, + 5.41225858796679e-07 + ] + }, + "drift": { + "embed.weight": 57.80682284068776, + "embed.bias": 15.88198491033531, + "blocks.0.ln.weight": 1.158570355271808, + "blocks.0.w1.weight": 17.652543030292062, + "blocks.0.w1.bias": 13.648960422025777, + "blocks.0.w2.weight": 62.30914257814011, + "blocks.1.ln.weight": 0.9953869430578365, + "blocks.1.w1.weight": 25.201451590168976, + "blocks.1.w1.bias": 21.042828750553948, + "blocks.1.w2.weight": 46.84933855533669, + "blocks.2.ln.weight": 0.5242024488499516, + "blocks.2.w1.weight": 18.18564601499141, + "blocks.2.w1.bias": 16.795126625795433, + "blocks.2.w2.weight": 37.67242795528057, + "blocks.3.ln.weight": 0.6028075449266447, + "blocks.3.w1.weight": 22.217273085841715, + "blocks.3.w1.bias": 23.55749047158996, + "blocks.3.w2.weight": 31.067562938269838, + "blocks.4.ln.weight": 0.5882814647513678, + "blocks.4.w1.weight": 19.789776992865228, + "blocks.4.w1.bias": 19.617707942323612, + "blocks.4.w2.weight": 40.08965442413366, + "blocks.5.ln.weight": 0.6533636599582089, + "blocks.5.w1.weight": 22.778934917253743, + "blocks.5.w1.bias": 23.182695560236425, + "blocks.5.w2.weight": 28.95545561908214, + "blocks.6.ln.weight": 0.5059341920781962, + "blocks.6.w1.weight": 15.068172940394142, + "blocks.6.w1.bias": 10.870554507643952, + "blocks.6.w2.weight": 64.80351856105872, + "blocks.7.ln.weight": 0.5640754334192314, + "blocks.7.w1.weight": 17.02027959939073, + "blocks.7.w1.bias": 12.619662219928584, + "blocks.7.w2.weight": 75.96627478247885, + "out_ln.weight": 0.44409435398278874, + "out_head.weight": 7.673916065937935, + "out_head.bias": 0.9250628284434071 + } + } + }, + "config": { + "dataset": "cifar10", + "d_hidden": 512, + "num_blocks": 8, + "batch_size": 128, + "epochs": 100, + "lr": 0.001, + "lr_fb": 0.001, + "wd": 0.01, + "lam": 0.1, + "K": 4, + "sigma_bridge": 0.05, + "ema_momentum": 0.995, + "term_grad_weight": 1.0, + "seeds": [ + 9 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L8_seed9", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |
