{ "6": { "dfa": { "log": { "train_loss": [ 2.058094557647705, 2.0324147889709474, 2.0218327713394166, 2.0282914637756346, 2.021828155441284, 2.0190604560852052, 2.0181676456069946, 2.015587470474243, 2.0169739764022827, 2.012877261199951, 2.015299163284302, 2.015902668991089, 2.010847349205017, 2.0069280560302736, 2.008257807006836, 2.0082109774017334, 2.006445485191345, 2.0060603087615965, 2.0080129043579102, 2.0026243621444704, 2.0042781149291993, 2.0038208263397217, 2.00284217376709, 2.0032863708496094, 2.003503814620972, 2.0014031100082397, 2.00154658203125, 2.0027565859985352, 1.9988653813934327, 2.0048675049591065, 2.002612951965332, 1.9998337776184083, 2.0022170097351073, 2.0006915353012085, 1.9998700122451782, 2.001532240638733, 2.0009784855651858, 2.001896736717224, 1.997888755760193, 1.9997327941894532, 1.998836244468689, 1.9972671249008178, 1.9991955379867554, 1.996907271156311, 1.9989860800933839, 1.9979305069732667, 1.9979226346588135, 1.9973365603637696, 1.9991827986145019, 1.9979995524597167, 1.9976954571533203, 1.9975597812652588, 1.9963330318450927, 1.9988955184173585, 1.9985223587036134, 1.9940103855895996, 1.9969634999084473, 1.9979491520690917, 1.9940481422424317, 1.9955107457733154, 1.9963652723693848, 1.9967138528060913, 1.9949829993438721, 1.9970767431259155, 1.996754500579834, 1.9966884062194825, 1.9968673025131225, 1.9936358183288574, 1.994392073059082, 1.9950825340270997, 1.9949580252838135, 1.9944661192321778, 1.994148543624878, 1.992705733795166, 1.9950622162246705, 1.9951953311538697, 1.9943858066558837, 1.9932064359283448, 1.9938985696411133, 1.992524094696045, 1.9919383232116699, 1.9938886001586915, 1.992321823425293, 1.993851443786621, 1.99483976688385, 1.991985835800171, 1.995015989379883, 1.9928585873413085, 1.9930935015487672, 1.9924035341644286, 1.992943452758789, 1.992181587677002, 1.992508976135254, 1.9930797433471679, 1.9915329122924805, 1.9912273866271972, 1.9927945401000977, 1.991996759376526, 1.992920351409912, 1.99362177444458 ], "train_acc": [ 0.24312, 0.25066, 0.25642, 0.25462, 0.25868, 0.25902, 0.25752, 0.26142, 0.26284, 0.26284, 0.26276, 0.261, 0.26512, 0.2669, 0.26932, 0.2642, 0.26996, 0.26718, 0.26772, 0.26848, 0.2686, 0.27064, 0.26932, 0.26956, 0.2707, 0.27038, 0.27084, 0.27048, 0.27238, 0.26878, 0.27158, 0.27342, 0.27218, 0.27104, 0.2718, 0.27454, 0.27166, 0.27142, 0.2736, 0.27368, 0.27276, 0.2748, 0.2733, 0.2743, 0.27712, 0.2743, 0.27604, 0.2745, 0.27354, 0.27452, 0.27538, 0.27732, 0.27562, 0.27438, 0.27546, 0.27798, 0.27778, 0.27474, 0.27972, 0.27802, 0.27448, 0.27894, 0.27802, 0.27462, 0.27436, 0.27914, 0.27646, 0.27898, 0.27676, 0.27804, 0.27688, 0.2795, 0.27758, 0.27814, 0.27706, 0.27766, 0.27888, 0.2781, 0.27842, 0.27892, 0.2802, 0.27858, 0.2792, 0.2788, 0.27574, 0.28088, 0.27956, 0.27906, 0.27764, 0.27988, 0.27716, 0.27956, 0.2782, 0.2787, 0.2789, 0.2799, 0.27836, 0.28092, 0.27914, 0.27792 ], "test_acc": [ 0.2564, 0.2576, 0.2815, 0.2842, 0.2869, 0.2771, 0.2864, 0.2704, 0.2864, 0.2827, 0.3005, 0.2939, 0.2817, 0.2771, 0.3033, 0.2802, 0.2776, 0.3002, 0.3, 0.2773, 0.3036, 0.2858, 0.2917, 0.2964, 0.2956, 0.2996, 0.2972, 0.2926, 0.3109, 0.2826, 0.3046, 0.3077, 0.278, 0.2856, 0.2955, 0.2948, 0.3034, 0.3088, 0.2955, 0.2871, 0.303, 0.3009, 0.2893, 0.2922, 0.3096, 0.3061, 0.2974, 0.3028, 0.2996, 0.2988, 0.3022, 0.3003, 0.3068, 0.2961, 0.2908, 0.2971, 0.3027, 0.2998, 0.3008, 0.3071, 0.2999, 0.2988, 0.2959, 0.2982, 0.3025, 0.295, 0.2934, 0.2965, 0.3014, 0.2975, 0.2893, 0.2951, 0.3003, 0.3061, 0.297, 0.2987, 0.3, 0.3016, 0.2974, 0.303, 0.3003, 0.3043, 0.3029, 0.3004, 0.2992, 0.2983, 0.3017, 0.299, 0.3005, 0.2977, 0.3004, 0.2986, 0.3, 0.2993, 0.2988, 0.2994, 0.2998, 0.2994, 0.2993, 0.2994 ] }, "diagnostics": { "bp_cosine": [ 0.3915177881717682, 0.00031804549507796764, -0.0005461883265525103, 0.00020462644170038402, -0.0005664663622155786, -0.0008316519670188427 ], "perturbation_rho": [ -0.011084532365202904, 0.0, 0.0, 0.0, 0.0, 0.0 ], "nudging": { "0.001": [ -4.0885061025619507e-07, 0.0, 0.0, 0.0, 0.0, 0.0 ], "0.003": [ -1.2046657502651215e-06, 0.0, 9.313225746154785e-10, 0.0, 0.0, 0.0 ], "0.01": [ -4.0549784898757935e-06, 0.0, 9.313225746154785e-10, 0.0, 9.313225746154785e-10, 0.0 ] }, "hidden_norms_per_layer": [ 52265.20703125, 1156494592.0, 2602998784.0, 3437189888.0, 3901714432.0, 6607706112.0, 6800756736.0 ], "bp_grad_norms_per_layer": [ 2.745953793237277e-07, 2.664374554317561e-10, 2.662880471682172e-10, 2.6685945120341614e-10, 2.6688468102165075e-10, 2.6681562514951906e-10, 2.668691656548816e-10 ] }, "drift": { "embed.weight": 323.8438777567112, "embed.bias": 236.32127317549046, "blocks.0.ln.weight": 9.81678478040185, "blocks.0.w1.weight": 283.8033448856046, "blocks.0.w1.bias": 254.7986676902803, "blocks.0.w2.weight": 459.2312703971922, "blocks.1.ln.weight": 7.9902441845145225, "blocks.1.w1.weight": 303.8032457038149, "blocks.1.w1.bias": 290.75689909603165, "blocks.1.w2.weight": 302.7987642593431, "blocks.2.ln.weight": 7.8083033170497576, "blocks.2.w1.weight": 313.18221523970163, "blocks.2.w1.bias": 288.235969181505, "blocks.2.w2.weight": 301.0694759577726, "blocks.3.ln.weight": 7.457261198270327, "blocks.3.w1.weight": 299.37309231794677, "blocks.3.w1.bias": 276.9324398819875, "blocks.3.w2.weight": 279.5917777002267, "blocks.4.ln.weight": 10.517843332237796, "blocks.4.w1.weight": 436.93554951444736, "blocks.4.w1.bias": 404.54404533704337, "blocks.4.w2.weight": 386.63677119691914, "blocks.5.ln.weight": 7.187391888063363, "blocks.5.w1.weight": 278.782404520272, "blocks.5.w1.bias": 266.36329153767053, "blocks.5.w2.weight": 249.56060346351742, "out_ln.weight": 0.5731338871373126, "out_head.weight": 8.11049488578885, "out_head.bias": 1.1188503504473455 } }, "fa": { "log": { "train_loss": [ 2.0280703507995606, 1.9438096570587158, 1.9075839419174194, 1.891309684791565, 1.8722532345199585, 1.8634102462768554, 1.8554693552017212, 1.8489897765350343, 1.846228702659607, 1.8386287690734864, 1.8406133193206786, 1.8387405684661866, 1.8276975162506104, 1.8232349169921875, 1.8223698708724976, 1.821195379638672, 1.8214911682510375, 1.8160597175598145, 1.8229843518066406, 1.813365035133362, 1.8151773623275758, 1.8169539303207398, 1.8105776592254639, 1.8114565746307374, 1.807515641746521, 1.806776870765686, 1.8001547024154663, 1.8010753986358643, 1.7956665130233764, 1.8011931986236571, 1.7943034625625611, 1.791777073059082, 1.791583038711548, 1.7873410034179686, 1.7855097869873047, 1.7860526992416381, 1.7824570555877686, 1.7813435136795044, 1.7782593814086913, 1.77326356174469, 1.7763885403060913, 1.7732242642974854, 1.7696506113052368, 1.7713894509887695, 1.7732664206695556, 1.7675420838165283, 1.763011986732483, 1.7635349313354491, 1.7613904175186157, 1.7604380879974366, 1.7606450772476196, 1.7594652207183838, 1.7573485291290283, 1.7578170404434204, 1.754877228012085, 1.7504150357437134, 1.7527595220565797, 1.7520960317230225, 1.7485238824081422, 1.7481630941772461, 1.7490973169326782, 1.7442448379516602, 1.7454284634399415, 1.7441554889678954, 1.744502698326111, 1.7424594699478149, 1.7411711275482178, 1.737814305076599, 1.7386144577789306, 1.7368740311431885, 1.7391422924804687, 1.737129613571167, 1.7340563317108155, 1.733810231666565, 1.7361049320220947, 1.7318184774398804, 1.7328989402008057, 1.731393734779358, 1.7313611135482787, 1.7314666943740844, 1.7308115518188476, 1.7282487900543213, 1.7280186172485352, 1.7302542147445679, 1.729862018966675, 1.7284494402313233, 1.7274188668060302, 1.72759757938385, 1.728316011276245, 1.7254533053588867, 1.7282261031341553, 1.7274837893676758, 1.7255484241104126, 1.7243229049682618, 1.7256911779022217, 1.7249952478790282, 1.7246788320541382, 1.724236849632263, 1.7243867401885986, 1.7224459433746337 ], "train_acc": [ 0.26004, 0.2935, 0.3105, 0.31688, 0.32726, 0.3268, 0.33294, 0.33576, 0.33912, 0.34126, 0.3408, 0.33982, 0.34284, 0.34334, 0.34554, 0.3478, 0.34762, 0.34808, 0.34944, 0.35122, 0.35068, 0.35214, 0.35338, 0.35102, 0.35642, 0.35468, 0.35298, 0.35718, 0.35686, 0.3546, 0.3602, 0.36098, 0.35922, 0.35992, 0.36128, 0.36012, 0.36176, 0.3639, 0.36318, 0.36768, 0.36394, 0.36712, 0.36688, 0.36672, 0.36676, 0.36582, 0.37102, 0.36946, 0.37328, 0.37212, 0.36806, 0.37476, 0.3764, 0.37204, 0.37148, 0.37532, 0.37424, 0.37422, 0.37674, 0.37706, 0.37652, 0.37612, 0.37884, 0.37688, 0.37654, 0.37978, 0.37676, 0.37814, 0.3802, 0.38226, 0.3787, 0.38016, 0.38294, 0.38174, 0.37992, 0.38192, 0.38114, 0.38394, 0.38458, 0.38406, 0.38478, 0.38358, 0.38324, 0.38232, 0.38468, 0.3854, 0.3863, 0.38538, 0.38582, 0.38406, 0.38552, 0.38434, 0.3848, 0.38542, 0.38594, 0.38728, 0.38536, 0.3858, 0.38608, 0.386 ], "test_acc": [ 0.3013, 0.3142, 0.3483, 0.3543, 0.3608, 0.3458, 0.3633, 0.3576, 0.3713, 0.3632, 0.3817, 0.3741, 0.3661, 0.3795, 0.3734, 0.3749, 0.3706, 0.3813, 0.3822, 0.3692, 0.3725, 0.3749, 0.3836, 0.3822, 0.389, 0.3821, 0.3846, 0.3777, 0.3861, 0.3809, 0.3857, 0.3855, 0.3765, 0.3955, 0.3914, 0.3948, 0.3997, 0.3994, 0.3935, 0.3989, 0.3987, 0.4008, 0.3925, 0.3922, 0.4023, 0.4055, 0.3953, 0.4008, 0.4003, 0.3971, 0.3987, 0.3931, 0.4076, 0.4019, 0.4046, 0.3981, 0.4081, 0.4046, 0.4043, 0.406, 0.4118, 0.4076, 0.4057, 0.414, 0.3979, 0.4079, 0.4044, 0.4087, 0.4065, 0.4024, 0.3997, 0.4078, 0.4117, 0.4092, 0.409, 0.406, 0.4132, 0.4133, 0.4066, 0.4131, 0.4083, 0.413, 0.4126, 0.4117, 0.4109, 0.4096, 0.4111, 0.4099, 0.4094, 0.4105, 0.4105, 0.4081, 0.4122, 0.4113, 0.412, 0.4095, 0.41, 0.4114, 0.4111, 0.4117 ] }, "diagnostics": { "bp_cosine": [ 0.03050209954380989, 0.07341398298740387, -0.030213337391614914, -0.1065329909324646, -0.112009696662426, 0.9969239234924316 ], "perturbation_rho": [ 0.060943812131881714, 0.02977900393307209, 0.030837206169962883, 0.03361157327890396, 0.003993029240518808, -0.012494717724621296 ], "nudging": { "0.001": [ -2.1707965061068535e-06, -2.555316314101219e-07, 1.9907020032405853e-08, 1.1094380170106888e-07, 1.1106021702289581e-07, -1.6409903764724731e-06 ], "0.003": [ -5.929847247898579e-06, -7.075723260641098e-07, 1.384178176522255e-07, 6.816117092967033e-07, 7.138587534427643e-07, -6.536138243973255e-06 ], "0.01": [ -2.0072446204721928e-05, -2.6080524548888206e-06, 4.3155159801244736e-07, 2.434244379401207e-06, 2.515967935323715e-06, -2.3723463527858257e-05 ] }, "hidden_norms_per_layer": [ 6712.94189453125, 241810.65625, 672440.6875, 1111939.0, 1518122.75, 1746060.875, 545338.8125 ], "bp_grad_norms_per_layer": [ 2.4450268028886057e-05, 1.2711601584669552e-06, 9.635764399718028e-07, 9.082662018045085e-07, 9.076145488506882e-07, 9.084363341571589e-07, 8.959328283708601e-07 ] }, "drift": { "embed.weight": 52.75751575367878, "embed.bias": 16.290300308868286, "blocks.0.ln.weight": 1.3122960145180722, "blocks.0.w1.weight": 18.35861512742778, "blocks.0.w1.bias": 15.033623290031253, "blocks.0.w2.weight": 59.867189750033205, "blocks.1.ln.weight": 0.9612442862570167, "blocks.1.w1.weight": 19.510654541967284, "blocks.1.w1.bias": 16.988315121645034, "blocks.1.w2.weight": 44.17488250656146, "blocks.2.ln.weight": 0.8653865881813659, "blocks.2.w1.weight": 21.845526237032836, "blocks.2.w1.bias": 20.228308191109985, "blocks.2.w2.weight": 54.8483328488635, "blocks.3.ln.weight": 0.687489668753582, "blocks.3.w1.weight": 21.235696780055214, "blocks.3.w1.bias": 22.153293813124844, "blocks.3.w2.weight": 37.488286386992314, "blocks.4.ln.weight": 0.6826937366325999, "blocks.4.w1.weight": 20.105546147824473, "blocks.4.w1.bias": 21.27470328328852, "blocks.4.w2.weight": 45.57345483069108, "blocks.5.ln.weight": 0.7529120733274255, "blocks.5.w1.weight": 23.26064002188114, "blocks.5.w1.bias": 24.637626686580436, "blocks.5.w2.weight": 40.48651407120801, "out_ln.weight": 0.28684521814071196, "out_head.weight": 5.505824885038432, "out_head.bias": 1.5561206526473979 } } }, "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": [ 6 ], "gpu": 0, "output_dir": "results/fa_dfa_d512_L6_seed6", "methods": [ "fa", "dfa" ], "random_targets": false, "penalty_lam": 0.0, "num_classes": 10 } }