diff options
Diffstat (limited to 'results/fa_dfa_d512_L2_seed6/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L2_seed6/results_cifar10.json | 749 |
1 files changed, 749 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L2_seed6/results_cifar10.json b/results/fa_dfa_d512_L2_seed6/results_cifar10.json new file mode 100644 index 0000000..a0da849 --- /dev/null +++ b/results/fa_dfa_d512_L2_seed6/results_cifar10.json @@ -0,0 +1,749 @@ +{ + "6": { + "dfa": { + "log": { + "train_loss": [ + 2.055892396621704, + 2.0292765501403807, + 2.023764776611328, + 2.020959409866333, + 2.019423359222412, + 2.0197315069580077, + 2.023359367675781, + 2.017872388153076, + 2.01921712928772, + 2.0194658319854737, + 2.021274960708618, + 2.016373544845581, + 2.0175488451766967, + 2.0179581134796143, + 2.0176988045501707, + 2.0177309605407716, + 2.0153829962921144, + 2.013690803070068, + 2.016160803833008, + 2.012104736251831, + 2.0164508349990844, + 2.0172074625396728, + 2.011329041824341, + 2.014233197860718, + 2.0119160061264036, + 2.0112428046417237, + 2.0135288095092774, + 2.0137764200592043, + 2.0114110033416748, + 2.011166734046936, + 2.01347174949646, + 2.0128866384887694, + 2.012690868988037, + 2.0142732004547117, + 2.013253126296997, + 2.012407520980835, + 2.0118109189605713, + 2.012256015167236, + 2.0096606398773194, + 2.010287212905884, + 2.0091940085601805, + 2.0101572361755373, + 2.0091212350845336, + 2.010151270904541, + 2.0109262997817994, + 2.0077846411514284, + 2.0088822582626342, + 2.0098557024383545, + 2.0110063369750977, + 2.0083373790740966, + 2.0075834423828125, + 2.0072854569244383, + 2.0076236277771, + 2.009549726829529, + 2.00914709564209, + 2.0072284757232666, + 2.0073054156112673, + 2.008326266860962, + 2.0065047761535646, + 2.0040750775527956, + 2.008966516571045, + 2.006659435195923, + 2.006130950393677, + 2.0071768866729736, + 2.004858699607849, + 2.005933465194702, + 2.0051698098754884, + 2.0040941329193114, + 2.004839565887451, + 2.004348826370239, + 2.0043614904785154, + 2.005439692611694, + 2.0047981397247314, + 2.0036350199127195, + 2.0039494177627564, + 2.002428635940552, + 2.002635501022339, + 2.005249846954346, + 2.0045922426605225, + 1.9999085303497315, + 2.0022091399383544, + 2.001223135147095, + 2.005691502914429, + 2.0016508586883544, + 2.0019130290222167, + 2.0011878201293944, + 2.002061458091736, + 2.001959662628174, + 2.0005823556518556, + 2.003897890167236, + 2.003386359901428, + 2.0019060565185547, + 2.0028662380981443, + 2.000936403427124, + 2.003748550567627, + 2.00183758392334, + 2.001099390411377, + 2.002016563873291, + 2.001500345993042, + 2.000833380355835 + ], + "train_acc": [ + 0.2468, + 0.25656, + 0.2583, + 0.25842, + 0.26294, + 0.26512, + 0.25982, + 0.26482, + 0.26072, + 0.26388, + 0.25908, + 0.26332, + 0.26268, + 0.26474, + 0.26364, + 0.26604, + 0.2647, + 0.26542, + 0.26462, + 0.26776, + 0.26296, + 0.26282, + 0.26768, + 0.2662, + 0.26584, + 0.26868, + 0.26712, + 0.2637, + 0.26642, + 0.2643, + 0.26304, + 0.2664, + 0.26832, + 0.2695, + 0.26582, + 0.26628, + 0.2661, + 0.26608, + 0.26796, + 0.26702, + 0.26834, + 0.26798, + 0.27016, + 0.26738, + 0.2666, + 0.26772, + 0.26924, + 0.2708, + 0.26744, + 0.26734, + 0.27232, + 0.26954, + 0.27186, + 0.2691, + 0.26756, + 0.27156, + 0.26944, + 0.26982, + 0.27, + 0.2734, + 0.27014, + 0.27428, + 0.2721, + 0.27192, + 0.27394, + 0.27134, + 0.2706, + 0.27214, + 0.2724, + 0.27452, + 0.27194, + 0.27366, + 0.273, + 0.27294, + 0.2731, + 0.27446, + 0.27392, + 0.2735, + 0.274, + 0.2742, + 0.2727, + 0.27306, + 0.27192, + 0.27572, + 0.27528, + 0.27414, + 0.2743, + 0.27318, + 0.27448, + 0.27334, + 0.27238, + 0.27438, + 0.27446, + 0.27388, + 0.27336, + 0.27486, + 0.27176, + 0.27446, + 0.2743, + 0.27204 + ], + "test_acc": [ + 0.2713, + 0.2844, + 0.2811, + 0.28, + 0.299, + 0.2643, + 0.2824, + 0.2917, + 0.2912, + 0.2738, + 0.2633, + 0.2873, + 0.285, + 0.2787, + 0.2918, + 0.2776, + 0.2763, + 0.278, + 0.2677, + 0.2808, + 0.2944, + 0.2953, + 0.2923, + 0.2954, + 0.297, + 0.2825, + 0.2932, + 0.2826, + 0.2908, + 0.285, + 0.2799, + 0.2848, + 0.3029, + 0.2813, + 0.2915, + 0.2709, + 0.3065, + 0.2788, + 0.3023, + 0.2961, + 0.2864, + 0.2937, + 0.297, + 0.3001, + 0.2957, + 0.282, + 0.2855, + 0.2917, + 0.2897, + 0.2983, + 0.2881, + 0.2843, + 0.2775, + 0.2932, + 0.2953, + 0.2891, + 0.3012, + 0.2843, + 0.2914, + 0.2863, + 0.2978, + 0.2948, + 0.2863, + 0.2944, + 0.2897, + 0.2924, + 0.2925, + 0.2939, + 0.2888, + 0.2956, + 0.2964, + 0.2878, + 0.2855, + 0.2904, + 0.2999, + 0.2861, + 0.3025, + 0.2935, + 0.2953, + 0.2878, + 0.2974, + 0.2959, + 0.2944, + 0.2924, + 0.2937, + 0.2953, + 0.2911, + 0.2938, + 0.2924, + 0.2938, + 0.2953, + 0.2933, + 0.2955, + 0.2954, + 0.2939, + 0.2937, + 0.2938, + 0.2936, + 0.2933, + 0.2933 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3805140256881714, + -0.00048614124534651637 + ], + "perturbation_rho": [ + 0.010780533775687218, + 0.0 + ], + "nudging": { + "0.001": [ + -3.3294782042503357e-07, + 0.0 + ], + "0.003": [ + -1.0631047189235687e-06, + 0.0 + ], + "0.01": [ + -3.623776137828827e-06, + 9.313225746154785e-10 + ] + }, + "hidden_norms_per_layer": [ + 54183.453125, + 1543298944.0, + 3004247040.0 + ], + "bp_grad_norms_per_layer": [ + 2.45587500558031e-07, + 3.692772754249063e-10, + 3.6931127600503544e-10 + ] + }, + "drift": { + "embed.weight": 334.1717985048247, + "embed.bias": 247.7536312263247, + "blocks.0.ln.weight": 9.54033246951961, + "blocks.0.w1.weight": 315.1784989948421, + "blocks.0.w1.bias": 269.25844995073356, + "blocks.0.w2.weight": 501.0628341707343, + "blocks.1.ln.weight": 8.010501646129491, + "blocks.1.w1.weight": 317.0441501524571, + "blocks.1.w1.bias": 304.79370018295725, + "blocks.1.w2.weight": 308.3494986335005, + "out_ln.weight": 0.4575791541009348, + "out_head.weight": 6.637383076518475, + "out_head.bias": 3.6615792171842636 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.065024294166565, + 1.9630456279754638, + 1.9233733417129517, + 1.9067592349243163, + 1.8861662637329102, + 1.8703109120941162, + 1.8611061660766601, + 1.8445787530899047, + 1.829258919067383, + 1.8199622161102296, + 1.8095393975448608, + 1.7991895624542236, + 1.79670791015625, + 1.7896576220321656, + 1.7787932138061524, + 1.777765901145935, + 1.768475433731079, + 1.7715507403945923, + 1.7699084145736694, + 1.7577085567474364, + 1.7492800643539428, + 1.7376883664703369, + 1.7303459451293945, + 1.7293071484375, + 1.7250730394744873, + 1.726899693222046, + 1.7227447848129271, + 1.7317372454452515, + 1.7304015692138672, + 1.7312020352172852, + 1.7351370712280274, + 1.7342145778656006, + 1.7334492336654663, + 1.731919374732971, + 1.7354538122940064, + 1.7322556600952148, + 1.7360534030914307, + 1.7412740663909911, + 1.7402722408676148, + 1.7424089098358155, + 1.7433867425918579, + 1.7484364804077148, + 1.7482806644058229, + 1.7441494891738891, + 1.7457399251937866, + 1.740269787902832, + 1.7429518072128296, + 1.7410654845428466, + 1.7409823212432862, + 1.7343881818008422, + 1.736907575340271, + 1.7340854137802124, + 1.734343325843811, + 1.7312401904296875, + 1.7296781457138062, + 1.7250156705474853, + 1.7303392902374268, + 1.7267954447174072, + 1.7268179355621338, + 1.7248962731552124, + 1.7265133374786377, + 1.725437378578186, + 1.7230116376495361, + 1.721951872406006, + 1.7212245053482056, + 1.7194156966781615, + 1.7177197817993164, + 1.7192226972198486, + 1.7195494704437255, + 1.7167850009536743, + 1.7179081740570068, + 1.7162539065551758, + 1.7143080168914795, + 1.7166276223373413, + 1.716199371986389, + 1.7125089855957032, + 1.7071118188858032, + 1.7140238445281983, + 1.711059020729065, + 1.7108189041137696, + 1.7138464722061157, + 1.712218090133667, + 1.7137628644561766, + 1.7120113299179076, + 1.7139518783950807, + 1.7121368927383422, + 1.7085959339523316, + 1.7080107479095459, + 1.7105122137069702, + 1.7120901413345337, + 1.7093008170318604, + 1.707778765335083, + 1.7079006521987916, + 1.7071793605804444, + 1.7064016004180909, + 1.7067470293426514, + 1.7087606598281861, + 1.7090342914581298, + 1.707086050338745, + 1.7054454077911376 + ], + "train_acc": [ + 0.24472, + 0.28828, + 0.3077, + 0.31704, + 0.32668, + 0.33122, + 0.3323, + 0.3406, + 0.34542, + 0.35004, + 0.35242, + 0.3556, + 0.35868, + 0.36042, + 0.36382, + 0.36434, + 0.3649, + 0.36774, + 0.36866, + 0.36968, + 0.37606, + 0.3764, + 0.3827, + 0.38044, + 0.38462, + 0.38274, + 0.38134, + 0.3805, + 0.38172, + 0.38162, + 0.37954, + 0.3812, + 0.38154, + 0.38246, + 0.38214, + 0.38156, + 0.38276, + 0.37756, + 0.37994, + 0.37818, + 0.37948, + 0.3745, + 0.3761, + 0.37776, + 0.37658, + 0.37964, + 0.3789, + 0.37938, + 0.37994, + 0.38366, + 0.38024, + 0.38026, + 0.38234, + 0.38264, + 0.38382, + 0.38478, + 0.3855, + 0.38412, + 0.38446, + 0.38598, + 0.38618, + 0.38722, + 0.38892, + 0.3878, + 0.38716, + 0.38748, + 0.38934, + 0.3868, + 0.391, + 0.39034, + 0.39048, + 0.39066, + 0.39448, + 0.3907, + 0.39146, + 0.39476, + 0.39424, + 0.39456, + 0.39384, + 0.3942, + 0.39406, + 0.39414, + 0.3935, + 0.39276, + 0.39352, + 0.3918, + 0.3955, + 0.39626, + 0.39398, + 0.39516, + 0.3936, + 0.39472, + 0.39608, + 0.39546, + 0.3953, + 0.3935, + 0.39494, + 0.39556, + 0.39782, + 0.39818 + ], + "test_acc": [ + 0.2947, + 0.3403, + 0.3462, + 0.3468, + 0.3593, + 0.3653, + 0.3723, + 0.3689, + 0.372, + 0.3699, + 0.3635, + 0.3648, + 0.3749, + 0.3716, + 0.3759, + 0.3936, + 0.3903, + 0.3867, + 0.3853, + 0.3955, + 0.3909, + 0.4035, + 0.3969, + 0.3965, + 0.3878, + 0.3929, + 0.3895, + 0.3756, + 0.3801, + 0.377, + 0.3725, + 0.3941, + 0.3851, + 0.3949, + 0.3941, + 0.3753, + 0.3733, + 0.3731, + 0.3846, + 0.3734, + 0.3695, + 0.3857, + 0.3871, + 0.3798, + 0.3841, + 0.3771, + 0.3762, + 0.3814, + 0.3893, + 0.3712, + 0.3673, + 0.3771, + 0.3882, + 0.3816, + 0.3894, + 0.3825, + 0.3915, + 0.369, + 0.3847, + 0.3726, + 0.3876, + 0.3789, + 0.3679, + 0.3797, + 0.3785, + 0.3897, + 0.3758, + 0.3883, + 0.3789, + 0.3812, + 0.375, + 0.3779, + 0.3876, + 0.3884, + 0.3853, + 0.3865, + 0.3874, + 0.3859, + 0.3907, + 0.3826, + 0.385, + 0.3929, + 0.3856, + 0.3809, + 0.385, + 0.3849, + 0.3824, + 0.3854, + 0.3838, + 0.3857, + 0.3835, + 0.3842, + 0.385, + 0.3866, + 0.3864, + 0.3868, + 0.3865, + 0.3866, + 0.3865, + 0.3865 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.017966898158192635, + 0.9611225724220276 + ], + "perturbation_rho": [ + -0.022423196583986282, + 0.05839487165212631 + ], + "nudging": { + "0.001": [ + -1.0927324183285236e-06, + -6.873218808323145e-06 + ], + "0.003": [ + -3.30101465806365e-06, + -2.0737992599606514e-05 + ], + "0.01": [ + -1.1010735761374235e-05, + -6.911164382472634e-05 + ] + }, + "hidden_norms_per_layer": [ + 5436.04931640625, + 220695.28125, + 94910.3203125 + ], + "bp_grad_norms_per_layer": [ + 2.4236895114881918e-05, + 2.0322845557529945e-06, + 1.8303358046978246e-06 + ] + }, + "drift": { + "embed.weight": 32.57084729521678, + "embed.bias": 26.337782289300907, + "blocks.0.ln.weight": 1.3633438568396286, + "blocks.0.w1.weight": 19.737032593158407, + "blocks.0.w1.bias": 13.272555669066517, + "blocks.0.w2.weight": 55.66342599708415, + "blocks.1.ln.weight": 1.0296837730331478, + "blocks.1.w1.weight": 17.104661462082266, + "blocks.1.w1.bias": 9.996125311409836, + "blocks.1.w2.weight": 40.18024027628224, + "out_ln.weight": 0.49660679931516427, + "out_head.weight": 3.626674560113674, + "out_head.bias": 8.095867414428609 + } + } + }, + "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": [ + 6 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L2_seed6", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |