{ "9": { "dfa": { "log": { "train_loss": [ 2.0386180068206787, 2.0063905084228515, 2.0115847008895873, 2.0072783072662355, 2.0072200788116454, 2.007624338607788, 2.004089120826721, 2.004089771194458, 1.9991737648010255, 1.9998447619247437, 1.9991543921661377, 2.0006776647186277, 1.9984499712371826, 1.9943631922149658, 1.994661587486267, 1.993870899658203, 1.9933709629821776, 1.9952729042816162, 1.9926185557556153, 1.9912919250488281, 1.9926570781707764, 1.9911886741638183, 1.9924334022521972, 1.9887267460632325, 1.9918579584121705, 1.990792886352539, 1.9901814585113526, 1.9899697443008424, 1.989364903640747, 1.9860043600463868, 1.9868051830291749, 1.9893270887756347, 1.9876267849349976, 1.9893067332458496, 1.9857772846221924, 1.9837282821655273, 1.9841915615081787, 1.9825071751403809, 1.9810244177246095, 1.9853471765136719, 1.9825159663391114, 1.9842312357330323, 1.9828273846817017, 1.9831989236450196, 1.9810442190551758, 1.9819502600479126, 1.981037528152466, 1.9779154906845093, 1.9809581425476075, 1.9797189464569092, 1.9827272113037109, 1.9808893032073975, 1.9813728839111329, 1.9781295204925538, 1.9786987835311889, 1.9786434796905517, 1.9790834701919555, 1.9781043656921387, 1.9786688150024414, 1.9782110293579103, 1.9773203885269166, 1.9758285034942626, 1.977617745285034, 1.97553291305542, 1.975182448425293, 1.9747401064682006, 1.9758236003875733, 1.9758562586212158, 1.9762433794403076, 1.9765984790420532, 1.9761512882232666, 1.9743927758407593, 1.9739378618621826, 1.9728027178192138, 1.9720366858673095, 1.9760308059692382, 1.9740920357513427, 1.9741222943115235, 1.9719459004974365, 1.9733767440032959, 1.9732320972061157, 1.9731594284057616, 1.9734650510025025, 1.9740613651275636, 1.970825253868103, 1.97225927734375, 1.9708090161132812, 1.972899548110962, 1.9718735347747802, 1.9716896154785157, 1.9735381398010254, 1.970859426651001, 1.970537699661255, 1.9716636752700805, 1.971408115081787, 1.97217986328125, 1.9708149477767944, 1.9725973287582397, 1.970340883255005, 1.970512617111206 ], "train_acc": [ 0.2506, 0.26582, 0.26488, 0.26812, 0.2673, 0.26874, 0.26924, 0.27152, 0.2706, 0.2732, 0.27076, 0.26982, 0.27112, 0.27538, 0.27152, 0.27576, 0.27424, 0.2737, 0.27522, 0.27654, 0.27548, 0.27776, 0.27694, 0.27644, 0.2755, 0.27554, 0.27802, 0.2774, 0.2763, 0.27872, 0.27982, 0.27564, 0.27854, 0.27758, 0.27964, 0.28212, 0.28362, 0.28318, 0.28196, 0.28094, 0.28112, 0.28024, 0.28322, 0.28244, 0.28272, 0.28128, 0.28152, 0.28524, 0.28126, 0.28276, 0.28214, 0.28506, 0.2806, 0.28196, 0.2851, 0.2821, 0.28354, 0.28526, 0.28296, 0.28452, 0.28764, 0.28686, 0.28512, 0.28712, 0.28588, 0.28462, 0.28596, 0.2847, 0.28546, 0.2833, 0.2868, 0.28714, 0.285, 0.2887, 0.28604, 0.28332, 0.28698, 0.28862, 0.2851, 0.2889, 0.29026, 0.28762, 0.28606, 0.2889, 0.28878, 0.29028, 0.2883, 0.28902, 0.2895, 0.28942, 0.29002, 0.29058, 0.28932, 0.28564, 0.286, 0.28974, 0.28982, 0.28878, 0.2885, 0.2891 ], "test_acc": [ 0.2683, 0.295, 0.2917, 0.2896, 0.2953, 0.2912, 0.311, 0.2938, 0.2849, 0.2891, 0.2916, 0.3049, 0.3017, 0.3203, 0.3102, 0.2822, 0.286, 0.3035, 0.3098, 0.3009, 0.3053, 0.3044, 0.302, 0.2901, 0.3166, 0.3038, 0.2909, 0.2941, 0.3023, 0.2937, 0.3143, 0.2817, 0.289, 0.3074, 0.3052, 0.3025, 0.3089, 0.3155, 0.3068, 0.3125, 0.3145, 0.3012, 0.3152, 0.315, 0.3013, 0.3054, 0.3112, 0.2904, 0.3169, 0.2959, 0.3056, 0.3098, 0.3101, 0.3139, 0.309, 0.2991, 0.3182, 0.3204, 0.2996, 0.3108, 0.3082, 0.3162, 0.3227, 0.3027, 0.3098, 0.2966, 0.309, 0.3138, 0.3095, 0.302, 0.3007, 0.3096, 0.3111, 0.3052, 0.3121, 0.3254, 0.3128, 0.3127, 0.3042, 0.3189, 0.3148, 0.3064, 0.3145, 0.3172, 0.3148, 0.3157, 0.3164, 0.3148, 0.3139, 0.3147, 0.315, 0.3137, 0.3139, 0.3151, 0.3161, 0.3159, 0.3157, 0.3154, 0.3157, 0.3156 ] }, "diagnostics": { "bp_cosine": [ 0.4173116087913513, -0.0009379963739775121 ], "perturbation_rho": [ 0.029962807893753052, 0.0 ], "nudging": { "0.001": [ -5.238689482212067e-07, 0.0 ], "0.003": [ -1.4784745872020721e-06, 1.862645149230957e-09 ], "0.01": [ -4.835892468690872e-06, -4.6566128730773926e-09 ] }, "hidden_norms_per_layer": [ 52101.48828125, 755819456.0, 2329725696.0 ], "bp_grad_norms_per_layer": [ 3.256158436215628e-07, 4.5803447146219867e-10, 4.584113644234833e-10 ] }, "drift": { "embed.weight": 314.4915524892561, "embed.bias": 242.71554547665664, "blocks.0.ln.weight": 9.316701961843108, "blocks.0.w1.weight": 260.9170515894993, "blocks.0.w1.bias": 217.2580939372968, "blocks.0.w2.weight": 445.91774813193825, "blocks.1.ln.weight": 8.403039866788982, "blocks.1.w1.weight": 293.10409601703697, "blocks.1.w1.bias": 296.0137857555933, "blocks.1.w2.weight": 322.56673718802114, "out_ln.weight": 0.438240662386626, "out_head.weight": 7.216506109788919, "out_head.bias": 4.092364299645094 } }, "fa": { "log": { "train_loss": [ 2.059632821121216, 1.969141392288208, 1.9324101000595093, 1.8976149897003174, 1.8789895468902589, 1.8676787176513672, 1.86150774559021, 1.8536200240707397, 1.8464182161712646, 1.843310608215332, 1.8476687601470947, 1.8471106829452515, 1.8417246197128296, 1.838934436569214, 1.8480940073013306, 1.841720523109436, 1.8359926760482788, 1.8341460582733153, 1.8318497580718993, 1.8298718184661866, 1.8337750116729736, 1.8297588416290282, 1.8286794381332399, 1.8261895336532592, 1.8255341546630859, 1.8234506717300416, 1.8259339986801149, 1.821055384902954, 1.8205125164794922, 1.8167257778549195, 1.8076189421081543, 1.8187879995727538, 1.8171783072662353, 1.8175717837524414, 1.8142739019775391, 1.8160600145721435, 1.8160681335449218, 1.8097417612075806, 1.8152825867080689, 1.8145546726226807, 1.814115087852478, 1.8139980951690673, 1.8110767394256593, 1.8108615531158447, 1.8082058542633057, 1.806761900253296, 1.8031038238143922, 1.8006043545150756, 1.8022277898406982, 1.8031743975067138, 1.8007837707138061, 1.8021138620758057, 1.8031328768920898, 1.7993617374038697, 1.7992449984359742, 1.7986480782699585, 1.7985403269195557, 1.7962297436141967, 1.7968010370254517, 1.795757841835022, 1.7965225546646117, 1.7958138821792602, 1.7978459465789796, 1.795165989151001, 1.79601362575531, 1.7981088857269287, 1.7986928964233397, 1.796389625015259, 1.7948159000015258, 1.795204453163147, 1.7975346917724608, 1.7945230379486083, 1.794725383377075, 1.7934243726348877, 1.7920495998382568, 1.7969142938995362, 1.7927454508209228, 1.7913766479110718, 1.792234444503784, 1.7971493310928344, 1.7925569116210938, 1.794579433517456, 1.7918265099716186, 1.7970994007110597, 1.7937700403213501, 1.79378436958313, 1.7918543418121338, 1.7927718152618408, 1.792527283859253, 1.7904213320922853, 1.791894981842041, 1.7942055084228516, 1.7854851037597657, 1.7892465375518798, 1.7894356133651734, 1.792232000427246, 1.7908443558120728, 1.7891317191314697, 1.7882942892456055, 1.7879403518295287 ], "train_acc": [ 0.24674, 0.28878, 0.30256, 0.31774, 0.32412, 0.32786, 0.33236, 0.33536, 0.33926, 0.34026, 0.33958, 0.33756, 0.34346, 0.34348, 0.33962, 0.34076, 0.34474, 0.34456, 0.34354, 0.34714, 0.3452, 0.347, 0.34758, 0.34994, 0.34704, 0.34678, 0.34878, 0.35086, 0.3489, 0.35086, 0.354, 0.35062, 0.35238, 0.35012, 0.35416, 0.3535, 0.35234, 0.3533, 0.35412, 0.35204, 0.35228, 0.35076, 0.35398, 0.35318, 0.35272, 0.3529, 0.3562, 0.35708, 0.35474, 0.35498, 0.35806, 0.35614, 0.3567, 0.359, 0.35654, 0.3578, 0.35942, 0.35754, 0.36092, 0.35994, 0.36174, 0.36086, 0.36242, 0.36584, 0.36354, 0.36262, 0.3625, 0.35902, 0.36222, 0.36288, 0.36128, 0.36226, 0.36288, 0.36458, 0.36242, 0.36178, 0.3654, 0.36442, 0.3654, 0.36446, 0.36594, 0.36486, 0.36312, 0.36548, 0.36586, 0.3661, 0.36682, 0.36846, 0.3665, 0.36726, 0.3648, 0.36732, 0.36906, 0.36904, 0.36968, 0.36782, 0.3685, 0.36574, 0.36768, 0.36662 ], "test_acc": [ 0.2753, 0.3239, 0.3396, 0.3381, 0.3497, 0.3518, 0.3631, 0.3416, 0.3506, 0.3507, 0.3488, 0.3577, 0.3492, 0.3608, 0.3625, 0.3672, 0.3648, 0.373, 0.3637, 0.3675, 0.3612, 0.3701, 0.3614, 0.3567, 0.3695, 0.3672, 0.3684, 0.3628, 0.3779, 0.3671, 0.3748, 0.3768, 0.3622, 0.3679, 0.3684, 0.3674, 0.365, 0.365, 0.3596, 0.3633, 0.3614, 0.3725, 0.3516, 0.3689, 0.3514, 0.356, 0.3549, 0.3407, 0.3627, 0.3489, 0.3519, 0.3524, 0.3578, 0.3532, 0.3396, 0.3532, 0.3527, 0.3599, 0.3532, 0.3571, 0.3569, 0.3491, 0.3603, 0.3561, 0.3594, 0.3522, 0.3517, 0.3523, 0.3618, 0.3583, 0.3522, 0.3556, 0.3617, 0.3617, 0.362, 0.3692, 0.3621, 0.3631, 0.3579, 0.3653, 0.3655, 0.3626, 0.3669, 0.3689, 0.3658, 0.3664, 0.3649, 0.3635, 0.3656, 0.3676, 0.3651, 0.3669, 0.3663, 0.3669, 0.3663, 0.3667, 0.3665, 0.3667, 0.3662, 0.3661 ] }, "diagnostics": { "bp_cosine": [ 0.005262219812721014, 0.9548712968826294 ], "perturbation_rho": [ 0.014450715854763985, 0.06202582269906998 ], "nudging": { "0.001": [ 1.9065337255597115e-06, -7.555587217211723e-06 ], "0.003": [ 5.672394763678312e-06, -2.293300349265337e-05 ], "0.01": [ 1.8914113752543926e-05, -7.650378393009305e-05 ] }, "hidden_norms_per_layer": [ 4985.0673828125, 267249.8125, 166942.46875 ], "bp_grad_norms_per_layer": [ 2.8030346584273502e-05, 1.5286594816643628e-06, 1.4554038898495492e-06 ] }, "drift": { "embed.weight": 30.89122351236108, "embed.bias": 29.388918157004504, "blocks.0.ln.weight": 1.5688009803019956, "blocks.0.w1.weight": 22.80179549488676, "blocks.0.w1.bias": 19.927274930467597, "blocks.0.w2.weight": 66.38185851388366, "blocks.1.ln.weight": 1.1343783166936168, "blocks.1.w1.weight": 17.613374844897635, "blocks.1.w1.bias": 9.480317921179704, "blocks.1.w2.weight": 41.63261375754297, "out_ln.weight": 0.4189237920059734, "out_head.weight": 4.914939575481723, "out_head.bias": 17.20838803283335 } } }, "config": { "dataset": "cifar10", "d_hidden": 512, "num_blocks": 2, "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_L2_seed9", "methods": [ "fa", "dfa" ], "random_targets": false, "penalty_lam": 0.0, "num_classes": 10 } }