{ "3": { "dfa": { "log": { "train_loss": [ 2.068615023956299, 2.047176646156311, 2.045687096939087, 2.041560323257446, 2.039340658111572, 2.037013464202881, 2.035925048980713, 2.0311173275756835, 2.032121988372803, 2.028200259399414, 2.0268639810943605, 2.0260205685043333, 2.023872131958008, 2.022557463645935, 2.019715512161255, 2.0188200274276733, 2.0172054164886473, 2.017824410171509, 2.017126524734497, 2.0166522624969483, 2.0137256134414674, 2.0104475395965578, 2.014379795913696, 2.014344210357666, 2.0130453625488283, 2.0132340604400634, 2.015250590057373, 2.0126567953872683, 2.011817781829834, 2.011758652381897, 2.012901442489624, 2.010367298049927, 2.01117289024353, 2.0130296683502196, 2.011817899551392, 2.0138639054107665, 2.0104484258270263, 2.01166593460083, 2.0109901863861084, 2.009072921295166, 2.0114363181304933, 2.0125079582214354, 2.0117907460784914, 2.0110322050476075, 2.0118529551696778, 2.0105218141174315, 2.0101245944213866, 2.011210913467407, 2.00997705657959, 2.0106901919555664, 2.0133068067932127, 2.010482346534729, 2.01014770980835, 2.013678374786377, 2.0105705118560793, 2.009467424964905, 2.0119786975097655, 2.012015775146484, 2.0094030991363527, 2.012004408950806, 2.0111507415008547, 2.0106979748535156, 2.008728541183472, 2.010878988723755, 2.009884360809326, 2.0100272506713868, 2.010221890258789, 2.0115498251342774, 2.012461339263916, 2.010008108291626, 2.009727705307007, 2.0100395127105712, 2.009380127105713, 2.0105259854888917, 2.0112087518310546, 2.0097919299316405, 2.009896908798218, 2.0096029611968995, 2.0109228817749023, 2.0104407540130613, 2.008896068954468, 2.010570963516235, 2.0119053901672364, 2.010287576522827, 2.008419459609985, 2.0072363785171508, 2.00801055557251, 2.008693112564087, 2.0080592823791505, 2.0085273458862303, 2.009861908569336, 2.0077927374267577, 2.009273677444458, 2.008448128814697, 2.00934712600708, 2.0098364579772947, 2.0096740563964843, 2.009671881637573, 2.0100610285949707, 2.0068935653305053 ], "train_acc": [ 0.23866, 0.24262, 0.24576, 0.25012, 0.25268, 0.25266, 0.25262, 0.25804, 0.2571, 0.25908, 0.26094, 0.26028, 0.26044, 0.26054, 0.262, 0.26448, 0.26688, 0.26672, 0.266, 0.26744, 0.26834, 0.26974, 0.26904, 0.2705, 0.26832, 0.26774, 0.26668, 0.26762, 0.267, 0.27284, 0.27026, 0.27038, 0.27058, 0.27046, 0.27236, 0.26896, 0.27174, 0.27068, 0.27064, 0.2734, 0.27182, 0.27226, 0.27142, 0.2746, 0.27316, 0.27024, 0.27572, 0.27118, 0.2743, 0.27594, 0.27086, 0.27004, 0.27266, 0.26996, 0.27428, 0.27528, 0.27046, 0.27342, 0.27552, 0.27356, 0.27532, 0.27676, 0.27258, 0.27636, 0.27536, 0.27252, 0.2765, 0.27326, 0.2722, 0.27532, 0.27574, 0.2746, 0.27538, 0.27316, 0.2756, 0.27282, 0.27362, 0.27488, 0.2711, 0.27402, 0.27544, 0.27228, 0.27592, 0.2756, 0.27716, 0.27642, 0.27756, 0.27424, 0.27854, 0.27518, 0.27258, 0.2778, 0.27602, 0.277, 0.27468, 0.27454, 0.27424, 0.27666, 0.27484, 0.27578 ], "test_acc": [ 0.2605, 0.2721, 0.2761, 0.2552, 0.2751, 0.2671, 0.2772, 0.279, 0.2708, 0.2664, 0.2807, 0.2803, 0.2787, 0.2911, 0.2852, 0.2856, 0.2844, 0.2747, 0.2802, 0.2843, 0.294, 0.2912, 0.2858, 0.276, 0.2842, 0.2825, 0.2921, 0.2849, 0.289, 0.2758, 0.2964, 0.2852, 0.2834, 0.2877, 0.2999, 0.2849, 0.293, 0.2873, 0.2934, 0.2883, 0.293, 0.2895, 0.2914, 0.292, 0.3002, 0.2865, 0.2926, 0.2945, 0.2989, 0.2977, 0.2916, 0.2894, 0.2977, 0.2938, 0.3004, 0.2967, 0.3009, 0.2818, 0.2924, 0.2934, 0.2852, 0.2866, 0.2939, 0.2929, 0.2948, 0.296, 0.2999, 0.2936, 0.2911, 0.2955, 0.2922, 0.2948, 0.292, 0.2922, 0.2954, 0.2909, 0.2971, 0.2964, 0.2964, 0.2975, 0.2926, 0.2936, 0.2895, 0.2968, 0.295, 0.2928, 0.2962, 0.2962, 0.2981, 0.295, 0.2952, 0.2962, 0.2954, 0.2942, 0.294, 0.2946, 0.2946, 0.2945, 0.2946, 0.2946 ] }, "diagnostics": { "bp_cosine": [ 0.39051711559295654, 0.00047808888484723866, 0.00286676874384284, -0.0025043508503586054, 0.000322053674608469, -0.0008587267366237938, -0.0018792838091030717, 0.0018272181041538715 ], "perturbation_rho": [ 0.005894219968467951, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0 ], "nudging": { "0.001": [ -2.915039658546448e-07, -9.313225746154785e-10, 9.313225746154785e-10, 0.0, 0.0, 0.0, 0.0, 0.0 ], "0.003": [ -1.0211952030658722e-06, -1.862645149230957e-09, 9.313225746154785e-10, 0.0, 0.0, 0.0, 9.313225746154785e-10, 0.0 ], "0.01": [ -3.0533410608768463e-06, -3.725290298461914e-09, 9.313225746154785e-10, 3.725290298461914e-09, 0.0, 0.0, -9.313225746154785e-10, 0.0 ] }, "hidden_norms_per_layer": [ 40316.796875, 307326208.0, 913832192.0, 988009024.0, 1508025088.0, 1806359168.0, 1849629568.0, 2188687616.0, 2485296896.0 ], "bp_grad_norms_per_layer": [ 3.1247913057086407e-07, 6.341530633946491e-10, 6.226453796998044e-10, 6.2297295100322e-10, 6.209277536584068e-10, 6.210325587119314e-10, 6.210361669367614e-10, 6.206388181162481e-10, 6.205727043351317e-10 ] }, "drift": { "embed.weight": 346.8648177248426, "embed.bias": 294.36548366042695, "blocks.0.ln.weight": 11.012950897216797, "blocks.0.w1.weight": 218.82511475631335, "blocks.0.w1.bias": 208.27393479202811, "blocks.0.w2.weight": 491.86978296944665, "blocks.1.ln.weight": 10.164929389953613, "blocks.1.w1.weight": 277.9108744662706, "blocks.1.w1.bias": 255.17942982496433, "blocks.1.w2.weight": 416.2935804854679, "blocks.2.ln.weight": 9.155071258544922, "blocks.2.w1.weight": 230.32353546806806, "blocks.2.w1.bias": 225.3918473151377, "blocks.2.w2.weight": 335.1338975810203, "blocks.3.ln.weight": 10.646790504455566, "blocks.3.w1.weight": 305.38387277965444, "blocks.3.w1.bias": 272.0924461450336, "blocks.3.w2.weight": 449.71231243467605, "blocks.4.ln.weight": 10.511990547180176, "blocks.4.w1.weight": 301.02475276314414, "blocks.4.w1.bias": 283.9163308436812, "blocks.4.w2.weight": 425.48006497712345, "blocks.5.ln.weight": 7.359297275543213, "blocks.5.w1.weight": 197.05490869212008, "blocks.5.w1.bias": 175.3607301882127, "blocks.5.w2.weight": 279.18290459751876, "blocks.6.ln.weight": 11.279300689697266, "blocks.6.w1.weight": 321.6075481759436, "blocks.6.w1.bias": 299.55963820199565, "blocks.6.w2.weight": 437.1964645534032, "blocks.7.ln.weight": 11.766033172607422, "blocks.7.w1.weight": 330.68186483881703, "blocks.7.w1.bias": 322.85133764510664, "blocks.7.w2.weight": 455.9640356105679, "out_ln.weight": 0.8627486824989319, "out_head.weight": 8.827849511384864, "out_head.bias": 1.0810052204688796 } }, "fa": { "log": { "train_loss": [ 2.058981375579834, 1.9834817496490478, 1.9641429858398438, 1.9436818535614013, 1.9326764168930053, 1.9252216508483886, 1.9228894757080077, 1.9135075526428222, 1.908241348800659, 1.898921445579529, 1.8944130271148683, 1.8912477577972413, 1.8830120736312865, 1.881072061805725, 1.8768733514022826, 1.8708937075805665, 1.8682633953094483, 1.867644786605835, 1.862151336288452, 1.8556909922027587, 1.8531654512786866, 1.8516737128448486, 1.856813112487793, 1.8525701689147949, 1.845026340560913, 1.839306372718811, 1.8322341234970092, 1.8263341422271728, 1.8188115433502197, 1.8107766916275025, 1.8092199557113648, 1.8012385620880127, 1.7975863833999635, 1.7922526489257813, 1.7880666454315186, 1.7886801773071288, 1.7807807125091553, 1.779724262084961, 1.775158731918335, 1.7758396688842772, 1.7778481522369385, 1.774957112350464, 1.7728896438598634, 1.7728447552490234, 1.7720583478546144, 1.7720315837860108, 1.7645318777084351, 1.7661468465423584, 1.7653337018203736, 1.7627277339935303, 1.7612493091583252, 1.7608684145736695, 1.7618030507659912, 1.7599208697128297, 1.7579380053710938, 1.7533478388214112, 1.7546381701278686, 1.7539206622314454, 1.74741821723938, 1.750472043838501, 1.7494391228866577, 1.748632820663452, 1.7438091558837892, 1.7478703462982177, 1.7456060763549806, 1.7437335445785522, 1.7426370120239258, 1.7418463362884522, 1.7448427515029907, 1.7404597217559814, 1.743501665992737, 1.7376666475677491, 1.7360430716705322, 1.7362179106903075, 1.7341359579086304, 1.733851330947876, 1.7334849172210693, 1.7326775600814819, 1.7341204782104491, 1.7362167639541626, 1.7309519817352295, 1.7323176555633546, 1.733476494140625, 1.7358508117294311, 1.7305281336212157, 1.7303411197662353, 1.73070291847229, 1.7317729217147828, 1.727743486557007, 1.7267417626571655, 1.7302013341522218, 1.7299649307250977, 1.7291040048217774, 1.7297229943847656, 1.7304346767807006, 1.7294611694335937, 1.731980471458435, 1.7267779105377197, 1.728856870651245, 1.7232196910858155 ], "train_acc": [ 0.24234, 0.2742, 0.29098, 0.29598, 0.29902, 0.29986, 0.3002, 0.30472, 0.30692, 0.31152, 0.31082, 0.31398, 0.3191, 0.31702, 0.32336, 0.32422, 0.32658, 0.32714, 0.32938, 0.33382, 0.33342, 0.33472, 0.33338, 0.33292, 0.33888, 0.3423, 0.34322, 0.3461, 0.34668, 0.35166, 0.35128, 0.35728, 0.3546, 0.35978, 0.361, 0.3614, 0.36304, 0.36228, 0.36516, 0.36602, 0.36758, 0.36578, 0.36746, 0.3672, 0.36552, 0.3666, 0.37044, 0.3676, 0.37066, 0.37068, 0.37034, 0.37252, 0.37222, 0.37374, 0.372, 0.37456, 0.37316, 0.37352, 0.37646, 0.37608, 0.37806, 0.37702, 0.37736, 0.3746, 0.37672, 0.37986, 0.37954, 0.38086, 0.3791, 0.37724, 0.3785, 0.38128, 0.37962, 0.3815, 0.38328, 0.38234, 0.38302, 0.38246, 0.38054, 0.37982, 0.38474, 0.38062, 0.382, 0.3824, 0.38212, 0.38174, 0.38348, 0.38414, 0.38298, 0.38414, 0.38316, 0.38198, 0.38374, 0.3837, 0.38268, 0.38232, 0.38248, 0.38382, 0.38298, 0.38342 ], "test_acc": [ 0.2801, 0.3073, 0.3193, 0.3203, 0.3257, 0.3225, 0.3347, 0.3351, 0.3319, 0.3249, 0.337, 0.3477, 0.3485, 0.3531, 0.3504, 0.3551, 0.3542, 0.3603, 0.3587, 0.3676, 0.3669, 0.3616, 0.3612, 0.3648, 0.3724, 0.3675, 0.3764, 0.3806, 0.3759, 0.3844, 0.38, 0.3841, 0.3848, 0.3885, 0.3898, 0.3917, 0.3918, 0.3932, 0.3872, 0.3948, 0.3953, 0.3905, 0.3913, 0.3936, 0.3956, 0.3946, 0.3982, 0.3893, 0.4001, 0.3968, 0.3968, 0.3963, 0.3964, 0.4, 0.3958, 0.3986, 0.397, 0.398, 0.3948, 0.3973, 0.3975, 0.401, 0.4026, 0.3986, 0.3986, 0.3997, 0.4009, 0.3998, 0.3987, 0.4021, 0.4003, 0.4008, 0.4017, 0.4014, 0.4009, 0.4024, 0.3995, 0.4047, 0.4028, 0.4026, 0.4041, 0.4028, 0.4013, 0.4016, 0.4038, 0.404, 0.404, 0.405, 0.4042, 0.4014, 0.4029, 0.404, 0.4044, 0.4035, 0.4035, 0.4036, 0.4035, 0.4036, 0.4036, 0.4035 ] }, "diagnostics": { "bp_cosine": [ 0.015765948221087456, 0.07965049147605896, 0.01825195923447609, -0.06194503605365753, -0.06945411115884781, -0.027306973934173584, -0.06518878042697906, 0.9981832504272461 ], "perturbation_rho": [ -0.003869679756462574, 0.028149917721748352, -0.03722488507628441, 0.010893862694501877, -0.053424231708049774, 0.0007988980505615473, 0.014475717209279537, 0.039686620235443115 ], "nudging": { "0.001": [ 1.698499545454979e-07, -3.166496753692627e-07, -3.213062882423401e-08, 1.387670636177063e-07, 1.8265563994646072e-07, 7.741618901491165e-08, 1.4121178537607193e-07, -1.2140953913331032e-06 ], "0.003": [ 3.164168447256088e-07, -1.0326039046049118e-06, -5.611218512058258e-08, 2.748565748333931e-07, 3.5634730011224747e-07, 1.3550743460655212e-07, 3.384193405508995e-07, -4.2710453271865845e-06 ], "0.01": [ 1.0926742106676102e-06, -3.4980475902557373e-06, -2.4691689759492874e-07, 1.008971594274044e-06, 1.0706717148423195e-06, 4.578614607453346e-07, 1.044594682753086e-06, -1.4978926628828049e-05 ] }, "hidden_norms_per_layer": [ 5578.46875, 33612.796875, 156752.953125, 338066.0625, 628331.5, 779574.8125, 844256.6875, 888649.0, 425876.65625 ], "bp_grad_norms_per_layer": [ 2.167309503420256e-05, 2.129869471900747e-06, 8.520624419361411e-07, 8.023235977816512e-07, 8.024675253182068e-07, 8.062054348556558e-07, 8.076636390796921e-07, 8.067011663115409e-07, 8.030539788705937e-07 ] }, "drift": { "embed.weight": 54.49055679435881, "embed.bias": 20.29269603704931, "blocks.0.ln.weight": 1.6682279109954834, "blocks.0.w1.weight": 17.062151194987237, "blocks.0.w1.bias": 13.499766617268852, "blocks.0.w2.weight": 71.40829154639394, "blocks.1.ln.weight": 1.4543473720550537, "blocks.1.w1.weight": 19.719236452544127, "blocks.1.w1.bias": 9.828062345144708, "blocks.1.w2.weight": 58.72293185943862, "blocks.2.ln.weight": 1.0721585750579834, "blocks.2.w1.weight": 18.58227228825852, "blocks.2.w1.bias": 15.746003179119816, "blocks.2.w2.weight": 41.71235474869442, "blocks.3.ln.weight": 0.8726892471313477, "blocks.3.w1.weight": 20.471342330584488, "blocks.3.w1.bias": 19.512499778298427, "blocks.3.w2.weight": 36.10169832252701, "blocks.4.ln.weight": 0.7409772276878357, "blocks.4.w1.weight": 17.51931629450496, "blocks.4.w1.bias": 18.723236583560368, "blocks.4.w2.weight": 30.24901961387597, "blocks.5.ln.weight": 0.7194885015487671, "blocks.5.w1.weight": 16.539786302289123, "blocks.5.w1.bias": 16.75240225769585, "blocks.5.w2.weight": 28.9854050706238, "blocks.6.ln.weight": 0.7770423293113708, "blocks.6.w1.weight": 17.2275004107886, "blocks.6.w1.bias": 18.38493950420895, "blocks.6.w2.weight": 32.31440391348023, "blocks.7.ln.weight": 0.9231076240539551, "blocks.7.w1.weight": 19.50195743824471, "blocks.7.w1.bias": 21.20774557625887, "blocks.7.w2.weight": 46.39053908141515, "out_ln.weight": 0.307411789894104, "out_head.weight": 4.604781316777015, "out_head.bias": 1.258175899516633 } } }, "config": { "dataset": "cifar10", "d_hidden": 256, "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": [ 3 ], "gpu": 0, "output_dir": "results/fa_dfa_d256_L8_seed3", "methods": [ "fa", "dfa" ], "random_targets": false, "penalty_lam": 0.0, "num_classes": 10 } }