{ "42": { "fa": { "log": { "train_loss": [ 2.0374581846618653, 1.9505271494293213, 1.9189605682373048, 1.8992357767105104, 1.8916226636123656, 1.8799470928955078, 1.8715093560409546, 1.8660212090301513, 1.8614284765625, 1.8615510483169555, 1.858706633758545, 1.8505436419296264, 1.8469465209197997, 1.84686426486969, 1.8409993788909913, 1.8422894380950927, 1.8362944522094726, 1.838251561012268, 1.8342618326568603, 1.8288902439117432, 1.8288569342803955, 1.8237011749649048, 1.8190487328338623, 1.8216550624847412, 1.818442062072754, 1.8108094896697997, 1.8102704141998291, 1.8097241515350342, 1.8078967553710938, 1.799188058128357, 1.8009985346221924, 1.8004494039535524, 1.79110828830719, 1.7962453186035157, 1.7955271081924438, 1.7904337732696534, 1.788247091140747, 1.7867638687133789, 1.787053722076416, 1.7837253421020507, 1.7821894748687743, 1.7816637888336182, 1.7749454086685181, 1.7757588800811768, 1.7753548749542236, 1.7709813816070556, 1.7743957537078858, 1.7711823197174072, 1.7686557120132447, 1.7698634561157227, 1.76816160987854, 1.7640094296646118, 1.7590285652923583, 1.759483937072754, 1.7608595893096923, 1.759725392189026, 1.7576867937850953, 1.7575671384429932, 1.7513107781600952, 1.7550161592102052, 1.752431219062805, 1.7487841637802124, 1.7479721618652344, 1.7504890463638305, 1.7505805645751953, 1.7466501081085206, 1.7456124398422241, 1.7472877283096313, 1.740817502784729, 1.7439257540893556, 1.7465672635650635, 1.744754091758728, 1.7406636810684204, 1.741213535194397, 1.7423687073516845, 1.742182956199646, 1.742147672958374, 1.7392435836029052, 1.7430149713516236, 1.7413112850952148, 1.7365029236602783, 1.739822388305664, 1.7374979788208007, 1.7375597472763062, 1.7379079512786866, 1.735270802078247, 1.7385935315322876, 1.7365304137420654, 1.7351179248428346, 1.7382376040649414, 1.734423624343872, 1.7308770125579833, 1.7355197325897216, 1.7343845403671265, 1.734937107887268, 1.7350866909408569, 1.735432066116333, 1.734759259376526, 1.733669355697632, 1.73626565574646 ], "train_acc": [ 0.24744, 0.28938, 0.3027, 0.31002, 0.31422, 0.32022, 0.32548, 0.32622, 0.33078, 0.32936, 0.33574, 0.33636, 0.33724, 0.33784, 0.34, 0.3413, 0.34206, 0.33976, 0.3412, 0.34588, 0.34544, 0.34566, 0.34976, 0.34872, 0.35166, 0.35328, 0.3528, 0.35174, 0.35398, 0.35788, 0.35598, 0.35664, 0.35946, 0.3579, 0.35866, 0.35922, 0.36176, 0.36192, 0.36018, 0.36312, 0.36604, 0.36232, 0.36722, 0.36558, 0.36774, 0.36944, 0.367, 0.3692, 0.36964, 0.37104, 0.37188, 0.37134, 0.37254, 0.37084, 0.37208, 0.3718, 0.37112, 0.3728, 0.37378, 0.37414, 0.373, 0.3766, 0.37534, 0.37602, 0.37528, 0.37748, 0.37684, 0.37644, 0.3761, 0.37702, 0.37658, 0.3781, 0.37686, 0.38036, 0.3784, 0.37902, 0.3791, 0.37944, 0.37872, 0.37962, 0.3794, 0.37758, 0.37992, 0.37976, 0.38084, 0.37792, 0.37966, 0.3806, 0.38122, 0.37914, 0.38144, 0.3834, 0.38042, 0.3804, 0.38228, 0.38292, 0.38086, 0.38258, 0.38086, 0.37848 ], "test_acc": [ 0.2938, 0.3207, 0.3406, 0.3392, 0.3402, 0.349, 0.3619, 0.3561, 0.3742, 0.3541, 0.3632, 0.3742, 0.3679, 0.3696, 0.3734, 0.3754, 0.3854, 0.3709, 0.3747, 0.3711, 0.3773, 0.3787, 0.384, 0.3688, 0.386, 0.3897, 0.383, 0.3852, 0.3876, 0.3857, 0.3878, 0.3886, 0.3873, 0.3929, 0.3862, 0.3864, 0.3871, 0.3947, 0.3901, 0.3942, 0.3854, 0.3841, 0.3892, 0.3876, 0.3965, 0.3924, 0.382, 0.3953, 0.3896, 0.3921, 0.3975, 0.3964, 0.3916, 0.3991, 0.3928, 0.4014, 0.3993, 0.4035, 0.3844, 0.3975, 0.4034, 0.4017, 0.3952, 0.3992, 0.4025, 0.4019, 0.3993, 0.3959, 0.3993, 0.4058, 0.3945, 0.4016, 0.4055, 0.4015, 0.4036, 0.4034, 0.3976, 0.4013, 0.4024, 0.4014, 0.3967, 0.398, 0.3947, 0.4056, 0.4012, 0.4001, 0.402, 0.4004, 0.4031, 0.4008, 0.4025, 0.402, 0.4018, 0.4026, 0.4007, 0.4012, 0.4015, 0.4021, 0.4015, 0.4014 ] }, "diagnostics": { "bp_cosine": [ 0.018813904374837875, 0.08222152292728424, -0.07224605977535248, -0.09462258964776993, -0.09898842871189117, 0.9958882927894592 ], "perturbation_rho": [ -0.017015788704156876, -0.0317268893122673, -0.004754193127155304, -0.006329755764454603, 0.03243139386177063, -0.012378348037600517 ], "nudging": { "0.001": [ -1.5720142982900143e-06, -4.273606464266777e-07, 7.217749953269958e-08, 1.2828968465328217e-07, 1.2386590242385864e-07, -1.2737000361084938e-06 ], "0.003": [ -4.96965367347002e-06, -1.4296965673565865e-06, 4.01865690946579e-07, 4.814937710762024e-07, 5.081528797745705e-07, -4.469649866223335e-06 ], "0.01": [ -1.6447564121335745e-05, -4.997651558369398e-06, 1.1826632544398308e-06, 1.6157864592969418e-06, 1.566135324537754e-06, -1.586077269166708e-05 ] }, "hidden_norms_per_layer": [ 7316.48291015625, 78755.234375, 576686.625, 1329546.125, 1568668.375, 1854290.875, 1069435.5 ], "bp_grad_norms_per_layer": [ 2.8450938771129586e-05, 1.794567538127012e-06, 6.647851478192024e-07, 6.471778988270671e-07, 6.503790359602135e-07, 6.49630294446979e-07, 6.414796303033654e-07 ] }, "drift": { "embed.weight": 48.85389529931473, "embed.bias": 13.565561887190782, "blocks.0.ln.weight": 1.0624637379045918, "blocks.0.w1.weight": 16.554310656593483, "blocks.0.w1.bias": 11.372855905617644, "blocks.0.w2.weight": 53.87251555976206, "blocks.1.ln.weight": 1.0837239573564923, "blocks.1.w1.weight": 22.586400420670774, "blocks.1.w1.bias": 15.377758862989662, "blocks.1.w2.weight": 35.85773072344426, "blocks.2.ln.weight": 0.9078778614654982, "blocks.2.w1.weight": 23.522098095099405, "blocks.2.w1.bias": 22.904554985601767, "blocks.2.w2.weight": 42.07479881121452, "blocks.3.ln.weight": 0.6724399561234441, "blocks.3.w1.weight": 20.752863459155623, "blocks.3.w1.bias": 20.37186903428861, "blocks.3.w2.weight": 36.191664332701706, "blocks.4.ln.weight": 0.6795979662627959, "blocks.4.w1.weight": 20.607291873742884, "blocks.4.w1.bias": 20.91900548314815, "blocks.4.w2.weight": 42.61545291740059, "blocks.5.ln.weight": 0.6835153473115325, "blocks.5.w1.weight": 20.653168018914645, "blocks.5.w1.bias": 21.158310771641304, "blocks.5.w2.weight": 45.94200541763277, "out_ln.weight": 0.34596348666722354, "out_head.weight": 6.343616372534138, "out_head.bias": 0.7709641239019656 } } }, "config": { "dataset": "cifar10", "d_hidden": 512, "num_blocks": 6, "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": [ 42 ], "gpu": 0, "output_dir": "results/fa_depth_scan_d512_L6", "methods": [ "fa" ], "random_targets": false, "penalty_lam": 0.0, "num_classes": 10 } }