diff options
Diffstat (limited to 'results/fa_dfa_d512_L6_seed7/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L6_seed7/results_cifar10.json | 837 |
1 files changed, 837 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L6_seed7/results_cifar10.json b/results/fa_dfa_d512_L6_seed7/results_cifar10.json new file mode 100644 index 0000000..1518de5 --- /dev/null +++ b/results/fa_dfa_d512_L6_seed7/results_cifar10.json @@ -0,0 +1,837 @@ +{ + "7": { + "dfa": { + "log": { + "train_loss": [ + 2.0644660066223146, + 2.0393594007873537, + 2.0298658728027346, + 2.027508737716675, + 2.021048464202881, + 2.0180776037979125, + 2.016010649871826, + 2.0145508810424806, + 2.0100066609954834, + 2.006050896835327, + 2.007042580833435, + 2.0072416454315185, + 2.0022777138519285, + 2.003505378875732, + 2.0038011445617676, + 2.0008850508880616, + 1.9970258136749268, + 1.9960380639648438, + 1.9933054499816893, + 1.9941288608551024, + 1.99546089138031, + 1.9936006065368652, + 1.994488134994507, + 1.9934685417938232, + 1.9926122019195556, + 1.9915543088531493, + 1.9883341995239259, + 1.9901361241149902, + 1.988699857711792, + 1.988091276512146, + 1.9886641019439697, + 1.9861467250061036, + 1.9892435473251342, + 1.986944725379944, + 1.9878197789001464, + 1.9859363149261475, + 1.9881446804046632, + 1.98439239112854, + 1.9870981986999512, + 1.9842346769714356, + 1.9852807202529907, + 1.9832774765396117, + 1.9852560285949707, + 1.985099944229126, + 1.984175789642334, + 1.9816588903808594, + 1.9819236499404906, + 1.9812517197418213, + 1.9811725481414795, + 1.9830146681213379, + 1.9799175099945068, + 1.9814645065689087, + 1.9808358678817748, + 1.9797256357192994, + 1.9818459476470947, + 1.980439381980896, + 1.9798791062164307, + 1.9782852731323242, + 1.9786473375701905, + 1.9781282710266113, + 1.9775994555664063, + 1.9782113193511963, + 1.9787678694915771, + 1.9774463504028321, + 1.9799045008850098, + 1.9775051036834717, + 1.977933309020996, + 1.9787949396514892, + 1.9781452545547484, + 1.9794640335845948, + 1.976175922241211, + 1.9797193655776977, + 1.9766233335113526, + 1.9788971273422242, + 1.9771730890274049, + 1.9750644179916381, + 1.9780011644744873, + 1.978542350692749, + 1.9770074579238892, + 1.9780642249298095, + 1.9761155870819092, + 1.9755325689697265, + 1.976641792564392, + 1.976419013442993, + 1.9755548094177247, + 1.9737617702484131, + 1.9747001873397827, + 1.974268023376465, + 1.974451787185669, + 1.9748253519439698, + 1.9772220765686035, + 1.9760730503845214, + 1.9763396139526368, + 1.9743848707580567, + 1.9757111903381348, + 1.9740122735214234, + 1.9744470761489867, + 1.9736765104675293, + 1.9756489211654662, + 1.975818688316345 + ], + "train_acc": [ + 0.23738, + 0.24404, + 0.25038, + 0.25232, + 0.25616, + 0.25784, + 0.2552, + 0.2584, + 0.26018, + 0.2647, + 0.26414, + 0.26396, + 0.26708, + 0.26626, + 0.2656, + 0.26712, + 0.26824, + 0.2697, + 0.27158, + 0.26954, + 0.27018, + 0.27384, + 0.27396, + 0.2686, + 0.27122, + 0.27376, + 0.2733, + 0.2745, + 0.276, + 0.27458, + 0.27622, + 0.27596, + 0.27604, + 0.2743, + 0.27596, + 0.27608, + 0.2749, + 0.27588, + 0.27448, + 0.27836, + 0.27578, + 0.27956, + 0.27798, + 0.27912, + 0.27864, + 0.27886, + 0.27796, + 0.2803, + 0.27986, + 0.27928, + 0.28028, + 0.28038, + 0.27754, + 0.2806, + 0.28016, + 0.27942, + 0.28168, + 0.28164, + 0.28036, + 0.2808, + 0.28056, + 0.2808, + 0.28214, + 0.283, + 0.28266, + 0.28086, + 0.28166, + 0.28236, + 0.28182, + 0.28274, + 0.28322, + 0.28242, + 0.28414, + 0.28278, + 0.2849, + 0.28342, + 0.28298, + 0.2835, + 0.28288, + 0.28306, + 0.28192, + 0.284, + 0.28374, + 0.2818, + 0.284, + 0.28564, + 0.28466, + 0.2856, + 0.28366, + 0.28388, + 0.28242, + 0.28532, + 0.2828, + 0.28576, + 0.28524, + 0.28524, + 0.28596, + 0.28682, + 0.28318, + 0.28228 + ], + "test_acc": [ + 0.2747, + 0.2448, + 0.2761, + 0.275, + 0.2784, + 0.2789, + 0.2712, + 0.2923, + 0.2752, + 0.2898, + 0.2816, + 0.2913, + 0.2903, + 0.2968, + 0.2885, + 0.2852, + 0.2949, + 0.2793, + 0.2989, + 0.2934, + 0.2914, + 0.2923, + 0.2946, + 0.3013, + 0.3014, + 0.2906, + 0.3031, + 0.297, + 0.2962, + 0.2942, + 0.3013, + 0.3006, + 0.3016, + 0.3082, + 0.311, + 0.2978, + 0.2803, + 0.299, + 0.3028, + 0.2966, + 0.3076, + 0.2768, + 0.3014, + 0.2929, + 0.3045, + 0.2953, + 0.2998, + 0.3013, + 0.3052, + 0.2902, + 0.3055, + 0.2939, + 0.3027, + 0.2966, + 0.3039, + 0.3032, + 0.294, + 0.3002, + 0.2989, + 0.307, + 0.3064, + 0.3031, + 0.3103, + 0.3015, + 0.2935, + 0.3131, + 0.2974, + 0.2958, + 0.3098, + 0.308, + 0.301, + 0.3079, + 0.306, + 0.3089, + 0.3077, + 0.304, + 0.3059, + 0.302, + 0.3017, + 0.3008, + 0.3065, + 0.3071, + 0.3039, + 0.3027, + 0.3065, + 0.3079, + 0.3051, + 0.305, + 0.3068, + 0.3084, + 0.3043, + 0.3041, + 0.3055, + 0.3043, + 0.3066, + 0.3062, + 0.3059, + 0.3066, + 0.3066, + 0.3065 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.4079825282096863, + 0.00017449544975534081, + -0.0002950271009467542, + -0.00047933883615769446, + 0.0008355096215382218, + -0.00013302025035955012 + ], + "perturbation_rho": [ + 0.005086352117359638, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.762543201446533e-07, + -2.7939677238464355e-09, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.003": [ + -1.4076940715312958e-06, + -2.7939677238464355e-09, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "0.01": [ + -4.495494067668915e-06, + -9.313225746154785e-10, + 0.0, + 2.7939677238464355e-09, + 0.0, + 0.0 + ] + }, + "hidden_norms_per_layer": [ + 53469.4921875, + 722511616.0, + 2513154304.0, + 4196997888.0, + 4850079232.0, + 7170064896.0, + 10128478208.0 + ], + "bp_grad_norms_per_layer": [ + 3.0659566618851386e-07, + 2.1547569284408041e-10, + 2.1372952019316216e-10, + 2.1350407553022421e-10, + 2.1338901479150962e-10, + 2.1328670773979042e-10, + 2.1351549694959004e-10 + ] + }, + "drift": { + "embed.weight": 319.2060524889922, + "embed.bias": 168.91177351992346, + "blocks.0.ln.weight": 10.20084372934541, + "blocks.0.w1.weight": 259.3585707340092, + "blocks.0.w1.bias": 204.8860780612092, + "blocks.0.w2.weight": 474.58360990249685, + "blocks.1.ln.weight": 8.524169911482344, + "blocks.1.w1.weight": 317.8100676412464, + "blocks.1.w1.bias": 286.625861663325, + "blocks.1.w2.weight": 337.7969456452116, + "blocks.2.ln.weight": 8.931831095995939, + "blocks.2.w1.weight": 362.27088509151724, + "blocks.2.w1.bias": 323.3719624301686, + "blocks.2.w2.weight": 340.6308866629528, + "blocks.3.ln.weight": 7.651484669125739, + "blocks.3.w1.weight": 309.48260206501516, + "blocks.3.w1.bias": 279.96119268361133, + "blocks.3.w2.weight": 277.85914995873645, + "blocks.4.ln.weight": 10.210800501008027, + "blocks.4.w1.weight": 421.6457628353572, + "blocks.4.w1.bias": 399.4367334675059, + "blocks.4.w2.weight": 412.14310110719924, + "blocks.5.ln.weight": 11.351197134890906, + "blocks.5.w1.weight": 456.9953588488586, + "blocks.5.w1.bias": 441.75351526808663, + "blocks.5.w2.weight": 455.08343692703954, + "out_ln.weight": 0.6738960876175676, + "out_head.weight": 9.13993717128016, + "out_head.bias": 0.8636649912030224 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.035428274383545, + 1.9543501361465454, + 1.921069111099243, + 1.9030546938323976, + 1.8895003170013427, + 1.884778947982788, + 1.8827891104888916, + 1.8784568830108643, + 1.86878865776062, + 1.860254260787964, + 1.8565264916229247, + 1.8507944106674195, + 1.8448910778808594, + 1.8358412002563476, + 1.8325608110809326, + 1.8244803615570069, + 1.8170998212051392, + 1.8116112271499634, + 1.8029299909210206, + 1.798451124572754, + 1.796333960533142, + 1.7877936901092528, + 1.7880024145889282, + 1.783990569152832, + 1.773157628097534, + 1.7710246792984008, + 1.7636069051742553, + 1.7616489712142944, + 1.7549441582870484, + 1.754057723007202, + 1.7550610286712647, + 1.7466887496185304, + 1.7430288970565795, + 1.7387338048553467, + 1.7371670000839234, + 1.7317205599975587, + 1.733508980026245, + 1.7296288764190675, + 1.7255350440216064, + 1.7244919805908203, + 1.7210525772857665, + 1.7164554761123658, + 1.716478593711853, + 1.717660334854126, + 1.7179905541992186, + 1.7098763641738892, + 1.70570148979187, + 1.7075587792587281, + 1.7068590453720094, + 1.70389818359375, + 1.6986094250488282, + 1.6995252443695068, + 1.6985104891967773, + 1.6972654037094117, + 1.695931503982544, + 1.6940053173065186, + 1.6959078594589234, + 1.6945061222076416, + 1.6924745792388916, + 1.6908132946777343, + 1.6884509930038452, + 1.6876910009384156, + 1.684639096031189, + 1.6838580658721924, + 1.6875753708267212, + 1.6831429992294311, + 1.6840416891479493, + 1.6803303512573242, + 1.6848778852081299, + 1.6803966582870484, + 1.676663829689026, + 1.6796003302383422, + 1.677098304748535, + 1.678186548423767, + 1.675721337814331, + 1.671638519821167, + 1.6730699136734009, + 1.673575770187378, + 1.6733202564239502, + 1.671583014907837, + 1.6726769936370849, + 1.6701723910522461, + 1.6703911999893188, + 1.6733231628417968, + 1.6666682750701904, + 1.6670018822860717, + 1.6669657723999023, + 1.6693463718795776, + 1.6684244076156616, + 1.6675063357162476, + 1.668329490966797, + 1.6683633364486694, + 1.6668814519882202, + 1.6666266248321533, + 1.6659794528198242, + 1.6643282043075562, + 1.666822133102417, + 1.6650706377410889, + 1.6686715842056274, + 1.6671581310653687 + ], + "train_acc": [ + 0.252, + 0.28866, + 0.30228, + 0.31138, + 0.31636, + 0.32068, + 0.32168, + 0.32222, + 0.32708, + 0.32854, + 0.33246, + 0.33542, + 0.33984, + 0.34196, + 0.34244, + 0.34556, + 0.34844, + 0.35176, + 0.35512, + 0.35748, + 0.35646, + 0.3622, + 0.35886, + 0.36206, + 0.36288, + 0.36684, + 0.37276, + 0.37276, + 0.37378, + 0.37246, + 0.37274, + 0.37874, + 0.37508, + 0.37808, + 0.3804, + 0.38156, + 0.37756, + 0.37946, + 0.38278, + 0.38208, + 0.38226, + 0.3882, + 0.38378, + 0.38596, + 0.38588, + 0.38736, + 0.38724, + 0.38704, + 0.38846, + 0.387, + 0.39326, + 0.39342, + 0.3919, + 0.3942, + 0.39288, + 0.39274, + 0.3934, + 0.3913, + 0.39564, + 0.39482, + 0.39314, + 0.39526, + 0.39544, + 0.3977, + 0.39616, + 0.39674, + 0.39592, + 0.39896, + 0.39744, + 0.3976, + 0.39928, + 0.39962, + 0.39968, + 0.40054, + 0.39988, + 0.40192, + 0.40246, + 0.40078, + 0.40022, + 0.39994, + 0.40256, + 0.40522, + 0.40094, + 0.40278, + 0.40524, + 0.40424, + 0.4035, + 0.40222, + 0.40502, + 0.40334, + 0.40314, + 0.40426, + 0.40582, + 0.40816, + 0.40362, + 0.40492, + 0.40314, + 0.40624, + 0.40386, + 0.4031 + ], + "test_acc": [ + 0.2994, + 0.315, + 0.3222, + 0.3432, + 0.3472, + 0.3579, + 0.3389, + 0.357, + 0.3572, + 0.3638, + 0.3595, + 0.3655, + 0.3683, + 0.3749, + 0.3736, + 0.3761, + 0.379, + 0.3672, + 0.3847, + 0.3897, + 0.3818, + 0.3922, + 0.3878, + 0.3942, + 0.395, + 0.3922, + 0.3942, + 0.3994, + 0.393, + 0.3938, + 0.3987, + 0.3997, + 0.3985, + 0.4053, + 0.406, + 0.4058, + 0.4051, + 0.4097, + 0.4081, + 0.4058, + 0.4093, + 0.401, + 0.4153, + 0.4048, + 0.4131, + 0.4069, + 0.404, + 0.414, + 0.412, + 0.414, + 0.4144, + 0.4179, + 0.4122, + 0.4187, + 0.4136, + 0.4228, + 0.414, + 0.4186, + 0.418, + 0.4167, + 0.4195, + 0.4178, + 0.4212, + 0.4175, + 0.4217, + 0.4212, + 0.4164, + 0.4237, + 0.4199, + 0.4199, + 0.4209, + 0.4234, + 0.4238, + 0.4226, + 0.4199, + 0.4251, + 0.4239, + 0.4209, + 0.4222, + 0.4257, + 0.423, + 0.424, + 0.4253, + 0.4242, + 0.4253, + 0.4239, + 0.4255, + 0.4247, + 0.4241, + 0.4253, + 0.4256, + 0.4259, + 0.4276, + 0.426, + 0.4246, + 0.4258, + 0.4269, + 0.4251, + 0.4254, + 0.4253 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.05678346008062363, + 0.06278860569000244, + -0.06489825248718262, + -0.13550668954849243, + -0.030537210404872894, + 0.9970009326934814 + ], + "perturbation_rho": [ + 0.021435417234897614, + -0.006943210028111935, + 0.002432417357340455, + -0.022404946386814117, + 0.010333601385354996, + -0.011944804340600967 + ], + "nudging": { + "0.001": [ + -6.006448529660702e-06, + -6.604241207242012e-07, + 6.123445928096771e-08, + 1.4808028936386108e-07, + 1.4319084584712982e-08, + -1.7423881217837334e-06 + ], + "0.003": [ + -1.811189576983452e-05, + -1.56438909471035e-06, + 3.5529956221580505e-07, + 8.135102689266205e-07, + 1.6961712390184402e-07, + -6.420654244720936e-06 + ], + "0.01": [ + -6.019952706992626e-05, + -4.958710633218288e-06, + 1.4510005712509155e-06, + 2.9762741178274155e-06, + 6.504124030470848e-07, + -2.2383523173630238e-05 + ] + }, + "hidden_norms_per_layer": [ + 4713.947265625, + 59263.7578125, + 803386.0625, + 994639.875, + 1492731.0, + 1505302.75, + 597029.1875 + ], + "bp_grad_norms_per_layer": [ + 3.536270742188208e-05, + 2.4848127395671327e-06, + 9.235278071173525e-07, + 9.185976637127169e-07, + 9.248900596503518e-07, + 9.24963046600169e-07, + 9.076006222130673e-07 + ] + }, + "drift": { + "embed.weight": 37.15779008870908, + "embed.bias": 18.146058604345335, + "blocks.0.ln.weight": 1.0848731333476576, + "blocks.0.w1.weight": 14.826104871554262, + "blocks.0.w1.bias": 11.802494272115224, + "blocks.0.w2.weight": 46.31577951436631, + "blocks.1.ln.weight": 0.9864104228146627, + "blocks.1.w1.weight": 20.750805358036015, + "blocks.1.w1.bias": 18.84914683848346, + "blocks.1.w2.weight": 45.31226900575547, + "blocks.2.ln.weight": 0.6561709772918677, + "blocks.2.w1.weight": 17.738489694580174, + "blocks.2.w1.bias": 16.431735764543493, + "blocks.2.w2.weight": 39.339953277678816, + "blocks.3.ln.weight": 0.5804518445695007, + "blocks.3.w1.weight": 19.985841730818343, + "blocks.3.w1.bias": 21.371168400070296, + "blocks.3.w2.weight": 31.937812891223096, + "blocks.4.ln.weight": 0.634509772716372, + "blocks.4.w1.weight": 17.17337766921684, + "blocks.4.w1.bias": 15.654490632053776, + "blocks.4.w2.weight": 45.84886660282379, + "blocks.5.ln.weight": 0.6629717085902781, + "blocks.5.w1.weight": 20.18167535513518, + "blocks.5.w1.bias": 21.32865390385067, + "blocks.5.w2.weight": 40.56800371651544, + "out_ln.weight": 0.3071358282526009, + "out_head.weight": 5.834156740127778, + "out_head.bias": 1.541462699862782 + } + } + }, + "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": [ + 7 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L6_seed7", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |
