diff options
Diffstat (limited to 'results/fa_dfa_d512_L8_seed0/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L8_seed0/results_cifar10.json | 881 |
1 files changed, 881 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L8_seed0/results_cifar10.json b/results/fa_dfa_d512_L8_seed0/results_cifar10.json new file mode 100644 index 0000000..de5d08f --- /dev/null +++ b/results/fa_dfa_d512_L8_seed0/results_cifar10.json @@ -0,0 +1,881 @@ +{ + "0": { + "dfa": { + "log": { + "train_loss": [ + 2.068570284347534, + 2.0450995418930056, + 2.040309367141724, + 2.0384985535430906, + 2.0337540267944334, + 2.028315601158142, + 2.0283870655059815, + 2.0261440406036377, + 2.0277783988952636, + 2.0236783444976805, + 2.0177216738128663, + 2.020857338027954, + 2.0178855153656006, + 2.0204481651306154, + 2.0144878174209593, + 2.012771276702881, + 2.0133113939666747, + 2.0119590814208985, + 2.009890116882324, + 2.008774749298096, + 2.0083382148742674, + 2.0076066046142578, + 2.0096171319580076, + 2.0058218814849855, + 2.0037864183044434, + 2.005599175491333, + 2.0061829089355467, + 2.0027509896087645, + 2.0058307901763914, + 2.005169650306702, + 2.001157734375, + 2.0021650346374513, + 2.001039823226929, + 2.0046913961029054, + 2.0023456330871583, + 2.00130324256897, + 2.0025460794067382, + 2.001611663970947, + 2.0031372194671633, + 2.0014712918090822, + 2.0003627224731444, + 1.9975626316070556, + 1.9982662202453614, + 1.9976352671813964, + 1.9969971923828125, + 2.0001233780288694, + 1.997716463356018, + 1.9996397301483155, + 1.9975465998840332, + 1.9973710316467286, + 1.9996610871887206, + 1.9990007359313964, + 1.9958738163757324, + 1.9966175972747802, + 2.0009265493774415, + 1.9975428602600098, + 1.9962558139038087, + 1.9977977807998657, + 1.9964274404144287, + 1.9977582873535156, + 1.9967647107696533, + 1.9964400707626342, + 1.9988198545074463, + 1.9962505130386352, + 1.9980460896301269, + 1.9944342567443847, + 1.9969512873077393, + 1.9951286602783203, + 1.9959452591705322, + 1.9949243884658814, + 1.997910827407837, + 1.9961381217956542, + 1.9937760234832764, + 1.9977886120605468, + 1.9961791613006592, + 1.99590330657959, + 1.9961589616394042, + 1.996615062561035, + 1.9954028228759766, + 1.994140139312744, + 1.9964727613067628, + 1.995709580001831, + 1.9935905380630494, + 1.9950566864013672, + 1.9948614299011231, + 1.9930699211120606, + 1.9950269234466553, + 1.996148171005249, + 1.9932368953704833, + 1.9957756786346434, + 1.9943457048034667, + 1.9930462799072266, + 1.9908040158081055, + 1.9929836280059814, + 1.995184083518982, + 1.9943841827392579, + 1.99466581741333, + 1.9947607112121581, + 1.9959496617889405, + 1.9939526372909546 + ], + "train_acc": [ + 0.23544, + 0.24512, + 0.24644, + 0.25124, + 0.25002, + 0.25482, + 0.25544, + 0.25706, + 0.25856, + 0.2579, + 0.26, + 0.25966, + 0.26192, + 0.26044, + 0.26436, + 0.2634, + 0.26574, + 0.26464, + 0.26664, + 0.26776, + 0.26856, + 0.26684, + 0.269, + 0.26998, + 0.272, + 0.27086, + 0.269, + 0.27068, + 0.27142, + 0.27146, + 0.27248, + 0.2718, + 0.27584, + 0.27106, + 0.2725, + 0.27434, + 0.27146, + 0.27324, + 0.27368, + 0.2727, + 0.27496, + 0.2771, + 0.27492, + 0.27596, + 0.27874, + 0.2729, + 0.27634, + 0.2747, + 0.27376, + 0.277, + 0.275, + 0.2776, + 0.27882, + 0.27748, + 0.27616, + 0.2745, + 0.2764, + 0.2782, + 0.27698, + 0.276, + 0.27666, + 0.27568, + 0.27892, + 0.2781, + 0.27582, + 0.2795, + 0.27838, + 0.27616, + 0.2783, + 0.2789, + 0.2776, + 0.27716, + 0.28014, + 0.27904, + 0.2768, + 0.2776, + 0.27682, + 0.279, + 0.28004, + 0.27846, + 0.28068, + 0.27798, + 0.28202, + 0.27836, + 0.28026, + 0.28116, + 0.27894, + 0.28088, + 0.28236, + 0.27878, + 0.2797, + 0.28002, + 0.28094, + 0.27994, + 0.27928, + 0.27954, + 0.27828, + 0.28064, + 0.27654, + 0.2811 + ], + "test_acc": [ + 0.2555, + 0.2658, + 0.2338, + 0.2589, + 0.2645, + 0.2829, + 0.2836, + 0.2386, + 0.2797, + 0.2828, + 0.2825, + 0.2743, + 0.2832, + 0.2778, + 0.277, + 0.2861, + 0.2887, + 0.2843, + 0.2986, + 0.3013, + 0.2914, + 0.2909, + 0.2788, + 0.2839, + 0.301, + 0.3034, + 0.295, + 0.2851, + 0.3031, + 0.2935, + 0.3072, + 0.2842, + 0.2977, + 0.3087, + 0.2878, + 0.2992, + 0.2958, + 0.2776, + 0.3095, + 0.302, + 0.3019, + 0.3096, + 0.3102, + 0.2911, + 0.2998, + 0.2978, + 0.2993, + 0.3104, + 0.2967, + 0.289, + 0.3004, + 0.3059, + 0.3001, + 0.2963, + 0.3022, + 0.2988, + 0.3028, + 0.2962, + 0.3041, + 0.3057, + 0.2973, + 0.305, + 0.3004, + 0.3098, + 0.2968, + 0.3054, + 0.3037, + 0.2995, + 0.3053, + 0.3065, + 0.3013, + 0.3067, + 0.3097, + 0.2996, + 0.3024, + 0.3038, + 0.2982, + 0.3071, + 0.3011, + 0.3049, + 0.3004, + 0.3033, + 0.3033, + 0.3029, + 0.3018, + 0.3025, + 0.3059, + 0.3053, + 0.3051, + 0.3062, + 0.3074, + 0.3024, + 0.3045, + 0.3056, + 0.3046, + 0.3048, + 0.3052, + 0.3054, + 0.3055, + 0.3055 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3844696283340454, + -0.0003293692716397345, + 0.00036326167173683643, + -7.58874011808075e-06, + -0.000848759722430259, + -0.0005374888423830271, + -0.00015908177010715008, + 0.00029008230194449425 + ], + "perturbation_rho": [ + 0.009705127216875553, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -5.657784640789032e-07, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.2135133147239685e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -3.6587007343769073e-06, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ] + }, + "hidden_norms_per_layer": [ + 51559.4296875, + 1329486592.0, + 4660075520.0, + 5993569792.0, + 6888088576.0, + 7015652864.0, + 8532604416.0, + 10583718912.0, + 12350084096.0 + ], + "bp_grad_norms_per_layer": [ + 2.5801111291912093e-07, + 2.506394813917012e-10, + 2.495726680873389e-10, + 2.497514417498792e-10, + 2.4978316637280784e-10, + 2.497747564333963e-10, + 2.497407836088428e-10, + 2.497392292966083e-10, + 2.5016502758212766e-10 + ] + }, + "drift": { + "embed.weight": 331.3816715738323, + "embed.bias": 280.1972794762848, + "blocks.0.ln.weight": 10.206860276367006, + "blocks.0.w1.weight": 293.0648280946435, + "blocks.0.w1.bias": 287.8102020652532, + "blocks.0.w2.weight": 480.05673936823246, + "blocks.1.ln.weight": 9.536572296218509, + "blocks.1.w1.weight": 384.85005936378496, + "blocks.1.w1.bias": 372.907458869564, + "blocks.1.w2.weight": 395.5604232032487, + "blocks.2.ln.weight": 9.583575811181442, + "blocks.2.w1.weight": 391.35917488090115, + "blocks.2.w1.bias": 358.7794005832253, + "blocks.2.w2.weight": 359.6693258728706, + "blocks.3.ln.weight": 9.98809250356752, + "blocks.3.w1.weight": 372.15298556466314, + "blocks.3.w1.bias": 341.69408442244566, + "blocks.3.w2.weight": 332.4997523924133, + "blocks.4.ln.weight": 7.10483666283608, + "blocks.4.w1.weight": 274.0268087874156, + "blocks.4.w1.bias": 252.79226182282545, + "blocks.4.w2.weight": 253.45621037491264, + "blocks.5.ln.weight": 10.243956247187224, + "blocks.5.w1.weight": 408.60736547082587, + "blocks.5.w1.bias": 378.8848879049959, + "blocks.5.w2.weight": 385.0032037459413, + "blocks.6.ln.weight": 11.012226548081168, + "blocks.6.w1.weight": 445.2115675036214, + "blocks.6.w1.bias": 406.90392571558743, + "blocks.6.w2.weight": 422.7651604516687, + "blocks.7.ln.weight": 10.678682452403773, + "blocks.7.w1.weight": 427.86871943047703, + "blocks.7.w1.bias": 421.26775677167313, + "blocks.7.w2.weight": 424.8177101015246, + "out_ln.weight": 0.7424186036866861, + "out_head.weight": 9.83199206815889, + "out_head.bias": 0.4556331945087151 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.031117898826599, + 1.9475624210357667, + 1.920816502456665, + 1.909251974143982, + 1.8920108963775635, + 1.8767158478546142, + 1.8721334871673585, + 1.863781995162964, + 1.8590689016342163, + 1.8527897924041747, + 1.8401800704956055, + 1.846341034011841, + 1.8388335055160523, + 1.8359079233169555, + 1.8243437384414674, + 1.8181520553970336, + 1.8202268659210206, + 1.808049402732849, + 1.8043210766601563, + 1.804519454650879, + 1.805484197998047, + 1.8011124628448487, + 1.8000532245635987, + 1.793024960975647, + 1.7795401358413696, + 1.780609351196289, + 1.777570998802185, + 1.768387048072815, + 1.7707503821563721, + 1.7622215390396119, + 1.7559357610321045, + 1.7557656387710572, + 1.7530760443115234, + 1.7580088625717163, + 1.746208298110962, + 1.7444442225646972, + 1.7456918414306641, + 1.742843345336914, + 1.7410870483398437, + 1.7379502767181396, + 1.7418859932708741, + 1.7290691452789306, + 1.7275336447906493, + 1.7247933518218994, + 1.7241094170761109, + 1.723441693687439, + 1.7219038082122802, + 1.7236041189575195, + 1.7184861129379272, + 1.7169820540618896, + 1.7187074862289429, + 1.7181550318145753, + 1.7141529751205444, + 1.7122508407592774, + 1.7160141869735719, + 1.712083070716858, + 1.709077308998108, + 1.7120566661834717, + 1.7138131452178955, + 1.7056159759902954, + 1.706560867576599, + 1.7041735248565675, + 1.7031678924942018, + 1.7053270459747314, + 1.7001814685440064, + 1.698059520263672, + 1.700954571838379, + 1.6969071603012085, + 1.6964463064575195, + 1.6973033560562134, + 1.697128847579956, + 1.6953863137054443, + 1.6925591326141358, + 1.6960328076171876, + 1.694064889831543, + 1.6925205464935302, + 1.691927748451233, + 1.6897403769683839, + 1.6869727429580688, + 1.6902123848724364, + 1.692222232322693, + 1.6904909725952149, + 1.6880250762557982, + 1.6899029476165772, + 1.6866879183959962, + 1.6873126499176025, + 1.689984903640747, + 1.6877681387329102, + 1.6844988162994385, + 1.6888493407821654, + 1.6853452449798585, + 1.6879416900253297, + 1.6831971020889283, + 1.681332094078064, + 1.681463975868225, + 1.6855041525268555, + 1.6846018813323975, + 1.688128840942383, + 1.6876728596115111, + 1.6832955141830444 + ], + "train_acc": [ + 0.25122, + 0.29002, + 0.30074, + 0.3078, + 0.31202, + 0.32116, + 0.32302, + 0.32474, + 0.32998, + 0.3327, + 0.33662, + 0.3378, + 0.34062, + 0.3423, + 0.3468, + 0.3493, + 0.34718, + 0.35114, + 0.35658, + 0.35406, + 0.35438, + 0.356, + 0.35496, + 0.35712, + 0.36442, + 0.36318, + 0.3635, + 0.36856, + 0.36718, + 0.37118, + 0.37414, + 0.37068, + 0.37418, + 0.36996, + 0.3753, + 0.37762, + 0.3755, + 0.37454, + 0.37586, + 0.37612, + 0.37538, + 0.38164, + 0.3825, + 0.37992, + 0.38294, + 0.38242, + 0.38152, + 0.38306, + 0.3826, + 0.38608, + 0.38346, + 0.38492, + 0.38532, + 0.38858, + 0.38614, + 0.38406, + 0.38632, + 0.38606, + 0.38602, + 0.39134, + 0.39088, + 0.39142, + 0.39166, + 0.39208, + 0.39016, + 0.39156, + 0.3904, + 0.39498, + 0.39106, + 0.39302, + 0.39394, + 0.39172, + 0.39746, + 0.39332, + 0.39366, + 0.39408, + 0.39414, + 0.39544, + 0.3967, + 0.39562, + 0.39768, + 0.39624, + 0.397, + 0.39732, + 0.39628, + 0.39652, + 0.39712, + 0.39694, + 0.39822, + 0.39396, + 0.39708, + 0.39744, + 0.39826, + 0.39852, + 0.39892, + 0.39592, + 0.3976, + 0.39714, + 0.39696, + 0.40022 + ], + "test_acc": [ + 0.2945, + 0.3266, + 0.3121, + 0.33, + 0.3343, + 0.3566, + 0.3538, + 0.3381, + 0.3562, + 0.3602, + 0.3694, + 0.3632, + 0.3618, + 0.3697, + 0.3812, + 0.3821, + 0.3828, + 0.3851, + 0.3757, + 0.3917, + 0.389, + 0.3894, + 0.3834, + 0.3912, + 0.3926, + 0.3955, + 0.3907, + 0.3903, + 0.3941, + 0.384, + 0.3972, + 0.3875, + 0.399, + 0.4015, + 0.3924, + 0.3957, + 0.4009, + 0.3991, + 0.3996, + 0.4008, + 0.4036, + 0.4005, + 0.409, + 0.4097, + 0.4021, + 0.4009, + 0.4126, + 0.4071, + 0.4122, + 0.4089, + 0.4131, + 0.4113, + 0.4146, + 0.4059, + 0.4163, + 0.4114, + 0.4068, + 0.4117, + 0.4182, + 0.4115, + 0.4082, + 0.4193, + 0.4177, + 0.4224, + 0.416, + 0.4171, + 0.4137, + 0.4155, + 0.4184, + 0.4188, + 0.4182, + 0.4169, + 0.4182, + 0.414, + 0.4182, + 0.412, + 0.4183, + 0.4217, + 0.4169, + 0.4204, + 0.4189, + 0.4154, + 0.4163, + 0.4166, + 0.4175, + 0.4195, + 0.418, + 0.4198, + 0.4191, + 0.4194, + 0.4182, + 0.4198, + 0.4177, + 0.4183, + 0.4178, + 0.4181, + 0.4188, + 0.4187, + 0.4191, + 0.4191 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.04550348222255707, + 0.04952041804790497, + 0.012564106844365597, + -0.048967309296131134, + -0.018994076177477837, + -0.08921081572771072, + -0.09729112684726715, + 0.9970568418502808 + ], + "perturbation_rho": [ + -0.007405851036310196, + -0.01778452657163143, + 0.03651121258735657, + -0.004411454312503338, + -0.017335545271635056, + 0.024473480880260468, + 0.033633388578891754, + -0.0013661051634699106 + ], + "nudging": { + "0.001": [ + -5.257490556687117e-06, + -3.859749995172024e-07, + -9.487848728895187e-08, + 9.802170097827911e-08, + 2.2584572434425354e-08, + 1.0256189852952957e-07, + 7.171183824539185e-08, + -1.4471588656306267e-06 + ], + "0.003": [ + -1.5992671251296997e-05, + -1.0106596164405346e-06, + -1.9604340195655823e-07, + 2.4406472221016884e-07, + 1.200241968035698e-07, + 4.641478881239891e-07, + 5.108886398375034e-07, + -5.204754415899515e-06 + ], + "0.01": [ + -5.306815728545189e-05, + -3.1923409551382065e-06, + -3.415043465793133e-07, + 9.238137863576412e-07, + 3.079185262322426e-07, + 1.4764373190701008e-06, + 1.8483842723071575e-06, + -1.84740056283772e-05 + ] + }, + "hidden_norms_per_layer": [ + 5801.0615234375, + 58545.69140625, + 278396.125, + 541219.0625, + 842750.8125, + 1009016.125, + 1188189.0, + 1390662.625, + 689550.6875 + ], + "bp_grad_norms_per_layer": [ + 3.488941365503706e-05, + 2.940359536296455e-06, + 8.656722911837278e-07, + 7.742645493635791e-07, + 7.646057156307506e-07, + 7.609253316331888e-07, + 7.622682005603565e-07, + 7.812644753357745e-07, + 7.56423219172575e-07 + ] + }, + "drift": { + "embed.weight": 40.85701563091728, + "embed.bias": 16.257987544795743, + "blocks.0.ln.weight": 1.0155844718531961, + "blocks.0.w1.weight": 14.308616002505659, + "blocks.0.w1.bias": 11.682066059725996, + "blocks.0.w2.weight": 47.88448262527169, + "blocks.1.ln.weight": 0.9800293898409066, + "blocks.1.w1.weight": 18.22850941673304, + "blocks.1.w1.bias": 10.248089109744662, + "blocks.1.w2.weight": 46.5074943877367, + "blocks.2.ln.weight": 0.6484944766132613, + "blocks.2.w1.weight": 18.04882879127194, + "blocks.2.w1.bias": 15.21848342753742, + "blocks.2.w2.weight": 28.777360387947045, + "blocks.3.ln.weight": 0.5971288470484941, + "blocks.3.w1.weight": 17.899823557577733, + "blocks.3.w1.bias": 18.289110041973974, + "blocks.3.w2.weight": 26.844107923153718, + "blocks.4.ln.weight": 0.5827124751217686, + "blocks.4.w1.weight": 16.72204012093031, + "blocks.4.w1.bias": 17.062801665778103, + "blocks.4.w2.weight": 33.8783390116795, + "blocks.5.ln.weight": 0.4109250365181375, + "blocks.5.w1.weight": 17.501541376807705, + "blocks.5.w1.bias": 19.653202019435096, + "blocks.5.w2.weight": 21.067728941101397, + "blocks.6.ln.weight": 0.5111244501205948, + "blocks.6.w1.weight": 18.004625850717634, + "blocks.6.w1.bias": 18.9556455723282, + "blocks.6.w2.weight": 26.773409027964927, + "blocks.7.ln.weight": 0.6137697718245447, + "blocks.7.w1.weight": 19.644875839051267, + "blocks.7.w1.bias": 20.525336825132296, + "blocks.7.w2.weight": 43.29822559398887, + "out_ln.weight": 0.2984864729265921, + "out_head.weight": 5.62935512400422, + "out_head.bias": 0.701363478107578 + } + } + }, + "config": { + "dataset": "cifar10", + "d_hidden": 512, + "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": [ + 0 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L8_seed0", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |
