diff options
Diffstat (limited to 'results/fa_dfa_d512_L2_seed5/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L2_seed5/results_cifar10.json | 749 |
1 files changed, 749 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L2_seed5/results_cifar10.json b/results/fa_dfa_d512_L2_seed5/results_cifar10.json new file mode 100644 index 0000000..d5c8ea9 --- /dev/null +++ b/results/fa_dfa_d512_L2_seed5/results_cifar10.json @@ -0,0 +1,749 @@ +{ + "5": { + "dfa": { + "log": { + "train_loss": [ + 2.0501297537231444, + 2.0391713497924804, + 2.0405313035583497, + 2.0335817265319824, + 2.0266058160018923, + 2.0228387924957274, + 2.0232558026504517, + 2.0218732793426515, + 2.0191759132385254, + 2.0202307054901123, + 2.018844593505859, + 2.017033308792114, + 2.018027767868042, + 2.018668871307373, + 2.015074652633667, + 2.015589296875, + 2.016558995742798, + 2.019203759498596, + 2.0183322762298586, + 2.0155471825408937, + 2.018300429916382, + 2.0150776953125, + 2.0150013764953614, + 2.012721175842285, + 2.016366043663025, + 2.012075606918335, + 2.016505323944092, + 2.01135826461792, + 2.0136885620117186, + 2.013543639678955, + 2.013746788291931, + 2.0128848664093018, + 2.012771292114258, + 2.01191774017334, + 2.010462758102417, + 2.010521863632202, + 2.009425925369263, + 2.0124644644546508, + 2.0108147170639037, + 2.010228076095581, + 2.0088936639022825, + 2.0095657570648195, + 2.007657756729126, + 2.009475396652222, + 2.010636301422119, + 2.0098936741638185, + 2.0092102942276, + 2.009080667991638, + 2.007752434120178, + 2.0074822512054444, + 2.011261424026489, + 2.006237863006592, + 2.008023773574829, + 2.0073289957427978, + 2.006207457962036, + 2.0055681369781495, + 2.006566423034668, + 2.0086398764038087, + 2.005377918167114, + 2.008910454864502, + 2.0054182757568357, + 2.0053556463241575, + 2.005029055404663, + 2.0059510985565185, + 2.005951479187012, + 2.0057564875793457, + 2.0048374938964844, + 2.0036859961700437, + 2.005258397903442, + 2.005781553649902, + 2.004468630065918, + 2.0044502724456787, + 2.0023633781433103, + 2.002619049530029, + 2.003770675506592, + 2.0034406785583494, + 2.0038263823699953, + 2.004770489578247, + 2.0045603427505494, + 2.003471921005249, + 2.0053932760620117, + 2.002159836387634, + 2.0036436273574827, + 2.002813849029541, + 2.00232444770813, + 2.0031388371276857, + 2.003156742095947, + 2.0042506001281737, + 2.001820291137695, + 2.0023800246810914, + 2.0012053105926513, + 2.0006055866241454, + 2.0019854052352906, + 2.001209078979492, + 2.0008096754455567, + 2.0021328547668458, + 2.0031403130340575, + 1.9998649573516847, + 2.002096244430542, + 2.0019282162475585 + ], + "train_acc": [ + 0.25066, + 0.25648, + 0.25718, + 0.25752, + 0.26214, + 0.26184, + 0.26432, + 0.2653, + 0.26402, + 0.26416, + 0.26932, + 0.26444, + 0.26816, + 0.26564, + 0.2684, + 0.27134, + 0.26932, + 0.267, + 0.26686, + 0.26922, + 0.26814, + 0.26974, + 0.2686, + 0.26834, + 0.2661, + 0.26928, + 0.26724, + 0.26998, + 0.26744, + 0.26994, + 0.26896, + 0.2696, + 0.26804, + 0.2696, + 0.27204, + 0.26932, + 0.27152, + 0.26992, + 0.27178, + 0.27218, + 0.27082, + 0.27092, + 0.2737, + 0.27122, + 0.27138, + 0.27174, + 0.272, + 0.2722, + 0.2723, + 0.27346, + 0.26892, + 0.27434, + 0.27432, + 0.27238, + 0.27332, + 0.27454, + 0.27508, + 0.27256, + 0.27376, + 0.27296, + 0.27436, + 0.27456, + 0.27532, + 0.27364, + 0.2735, + 0.27452, + 0.27372, + 0.27588, + 0.27558, + 0.27532, + 0.27558, + 0.2755, + 0.27522, + 0.27524, + 0.2758, + 0.27718, + 0.27532, + 0.27618, + 0.27686, + 0.27688, + 0.27632, + 0.27328, + 0.27464, + 0.27712, + 0.27804, + 0.27666, + 0.2755, + 0.27784, + 0.27688, + 0.27586, + 0.27696, + 0.27672, + 0.27692, + 0.2786, + 0.27814, + 0.27484, + 0.27636, + 0.27772, + 0.27436, + 0.27908 + ], + "test_acc": [ + 0.2723, + 0.2812, + 0.2672, + 0.2819, + 0.2601, + 0.279, + 0.2898, + 0.289, + 0.2685, + 0.2896, + 0.2766, + 0.2838, + 0.2792, + 0.2863, + 0.3016, + 0.292, + 0.2951, + 0.2919, + 0.2929, + 0.2842, + 0.2837, + 0.2864, + 0.3007, + 0.2981, + 0.2983, + 0.2929, + 0.2962, + 0.3, + 0.2923, + 0.2989, + 0.2802, + 0.2939, + 0.269, + 0.2905, + 0.278, + 0.2981, + 0.3057, + 0.2984, + 0.3025, + 0.2843, + 0.3007, + 0.2824, + 0.2941, + 0.3064, + 0.284, + 0.2892, + 0.2981, + 0.2894, + 0.2983, + 0.2972, + 0.2969, + 0.2958, + 0.2892, + 0.299, + 0.2959, + 0.2899, + 0.2863, + 0.3044, + 0.2926, + 0.2916, + 0.2983, + 0.2914, + 0.3015, + 0.2956, + 0.2904, + 0.2972, + 0.2893, + 0.289, + 0.2961, + 0.2993, + 0.2911, + 0.289, + 0.2944, + 0.2942, + 0.2955, + 0.2939, + 0.2961, + 0.295, + 0.2949, + 0.2966, + 0.2977, + 0.2997, + 0.2982, + 0.2984, + 0.2965, + 0.2941, + 0.2911, + 0.2957, + 0.2962, + 0.2998, + 0.2916, + 0.2965, + 0.2952, + 0.2956, + 0.2959, + 0.2965, + 0.2957, + 0.296, + 0.2963, + 0.2963 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.38937193155288696, + -0.00133989576715976 + ], + "perturbation_rho": [ + -0.03419807553291321, + 0.0 + ], + "nudging": { + "0.001": [ + -3.7206336855888367e-07, + 0.0 + ], + "0.003": [ + -1.1147931218147278e-06, + 0.0 + ], + "0.01": [ + -3.5390257835388184e-06, + 3.725290298461914e-09 + ] + }, + "hidden_norms_per_layer": [ + 55290.671875, + 1164404992.0, + 1809793664.0 + ], + "bp_grad_norms_per_layer": [ + 2.527334572732798e-07, + 4.3922129822071554e-10, + 4.3930301063532795e-10 + ] + }, + "drift": { + "embed.weight": 334.45692579883826, + "embed.bias": 254.1500844207983, + "blocks.0.ln.weight": 10.378101238923492, + "blocks.0.w1.weight": 277.98485791253944, + "blocks.0.w1.bias": 241.48477586542305, + "blocks.0.w2.weight": 467.05002700094724, + "blocks.1.ln.weight": 6.909499166461763, + "blocks.1.w1.weight": 263.9133275584596, + "blocks.1.w1.bias": 244.87238078164592, + "blocks.1.w2.weight": 283.49051001246613, + "out_ln.weight": 0.43727802835016244, + "out_head.weight": 6.508111271175387, + "out_head.bias": 3.1430989188106846 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.057919206390381, + 1.9476089878845215, + 1.9109919176864625, + 1.8918181984710694, + 1.8734441751861572, + 1.859958696899414, + 1.8465128494644165, + 1.8378619164657592, + 1.8330042699813842, + 1.824237583885193, + 1.82030082321167, + 1.8168681536102296, + 1.8147564684677124, + 1.8174263799285888, + 1.8145158917236328, + 1.8134905625152589, + 1.816219719581604, + 1.8107114935684203, + 1.809884515991211, + 1.8064439716339111, + 1.8076383264160156, + 1.8070710248184203, + 1.799020498085022, + 1.799975090560913, + 1.7974333989715576, + 1.7963496450042724, + 1.8014785760498047, + 1.7986563251113892, + 1.7955734420776368, + 1.7972684210968017, + 1.7941396160125733, + 1.7950014336395264, + 1.7955820244979859, + 1.7959071813201903, + 1.79216160282135, + 1.7948664249038697, + 1.7919580102920531, + 1.7955779486465455, + 1.7922123107147216, + 1.7889366255950927, + 1.7819780150604247, + 1.7830965274429322, + 1.782749061050415, + 1.7800398288726806, + 1.780551498336792, + 1.7743108095932008, + 1.783085387916565, + 1.774484405479431, + 1.7695070098114014, + 1.7732080498504639, + 1.7756617493057252, + 1.7731031243133546, + 1.7697893057632446, + 1.7669785211181641, + 1.7650962362289428, + 1.763043028793335, + 1.7600419904708862, + 1.7632893228912354, + 1.7593432830429077, + 1.7613518405532838, + 1.754806079711914, + 1.7532699264144898, + 1.7546887551498414, + 1.7581820251083373, + 1.751631008605957, + 1.754752312927246, + 1.7504624893951417, + 1.7475749392318725, + 1.751591441116333, + 1.7496289820098876, + 1.748469584388733, + 1.747327193069458, + 1.7421584380722046, + 1.7428951406097413, + 1.7426397796249389, + 1.7432606916046143, + 1.7435048538589477, + 1.7410756462860106, + 1.743000368347168, + 1.7420079108047486, + 1.7423530182266236, + 1.7415100763320923, + 1.7408926866912842, + 1.7403043130874634, + 1.7404012395477295, + 1.7404220779418946, + 1.7388356212615967, + 1.7423903681182862, + 1.7363825454711914, + 1.7398050061035155, + 1.7384051816558839, + 1.7355061944198609, + 1.7398744315338135, + 1.7368783473968505, + 1.7394122046661378, + 1.7361857733154298, + 1.742128964920044, + 1.7403847741699219, + 1.737286774673462, + 1.7390742868041993 + ], + "train_acc": [ + 0.25062, + 0.29808, + 0.31302, + 0.32052, + 0.3262, + 0.33122, + 0.33844, + 0.3404, + 0.34186, + 0.34348, + 0.34654, + 0.34656, + 0.34834, + 0.34672, + 0.3507, + 0.34978, + 0.34784, + 0.35204, + 0.34968, + 0.35226, + 0.35198, + 0.35068, + 0.35748, + 0.35378, + 0.35458, + 0.35536, + 0.35374, + 0.3558, + 0.35606, + 0.35744, + 0.35864, + 0.3575, + 0.35448, + 0.35674, + 0.35792, + 0.3562, + 0.3585, + 0.35812, + 0.35828, + 0.3577, + 0.36192, + 0.36198, + 0.3611, + 0.36338, + 0.36406, + 0.3631, + 0.36, + 0.36512, + 0.36844, + 0.36508, + 0.3646, + 0.36468, + 0.36598, + 0.36518, + 0.3704, + 0.37062, + 0.3711, + 0.37078, + 0.36744, + 0.37034, + 0.37248, + 0.37374, + 0.3702, + 0.37384, + 0.37386, + 0.37256, + 0.3758, + 0.37518, + 0.3727, + 0.37388, + 0.37536, + 0.37548, + 0.37722, + 0.37788, + 0.37788, + 0.37554, + 0.37874, + 0.37764, + 0.37696, + 0.37702, + 0.37834, + 0.37902, + 0.37876, + 0.37774, + 0.37714, + 0.37816, + 0.38042, + 0.37544, + 0.38068, + 0.37858, + 0.38066, + 0.37826, + 0.38016, + 0.3802, + 0.37906, + 0.37884, + 0.37926, + 0.37942, + 0.37788, + 0.37972 + ], + "test_acc": [ + 0.2861, + 0.3325, + 0.3291, + 0.348, + 0.3452, + 0.3574, + 0.3412, + 0.3599, + 0.3485, + 0.3683, + 0.3663, + 0.3682, + 0.3616, + 0.367, + 0.3701, + 0.3702, + 0.3592, + 0.3731, + 0.3691, + 0.3588, + 0.3585, + 0.3754, + 0.3752, + 0.3647, + 0.3568, + 0.3678, + 0.3622, + 0.3687, + 0.3662, + 0.3697, + 0.3614, + 0.3669, + 0.3387, + 0.3594, + 0.3452, + 0.3532, + 0.3488, + 0.3581, + 0.3475, + 0.3599, + 0.3428, + 0.347, + 0.3621, + 0.3565, + 0.3451, + 0.3385, + 0.3365, + 0.3324, + 0.3551, + 0.3452, + 0.353, + 0.3504, + 0.3508, + 0.3448, + 0.3299, + 0.3413, + 0.3342, + 0.3506, + 0.3532, + 0.3428, + 0.3468, + 0.3395, + 0.3551, + 0.3435, + 0.334, + 0.3427, + 0.3352, + 0.3403, + 0.3419, + 0.3374, + 0.3358, + 0.3438, + 0.3401, + 0.3381, + 0.3437, + 0.3412, + 0.3362, + 0.3404, + 0.339, + 0.3406, + 0.3378, + 0.3481, + 0.3449, + 0.3412, + 0.334, + 0.3405, + 0.3431, + 0.3434, + 0.3401, + 0.3446, + 0.3386, + 0.3425, + 0.3413, + 0.3399, + 0.3425, + 0.3407, + 0.3402, + 0.3413, + 0.341, + 0.341 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.01863543502986431, + 0.966240406036377 + ], + "perturbation_rho": [ + 0.049932606518268585, + 0.11253532767295837 + ], + "nudging": { + "0.001": [ + -3.632623702287674e-06, + -7.5231073424220085e-06 + ], + "0.003": [ + -1.0822783224284649e-05, + -2.275872975587845e-05 + ], + "0.01": [ + -3.596034366637468e-05, + -7.59701943024993e-05 + ] + }, + "hidden_norms_per_layer": [ + 3561.231201171875, + 299735.8125, + 271834.53125 + ], + "bp_grad_norms_per_layer": [ + 2.061038867395837e-05, + 5.578897344094003e-07, + 5.395349376158265e-07 + ] + }, + "drift": { + "embed.weight": 25.27185200703667, + "embed.bias": 13.597802808023353, + "blocks.0.ln.weight": 1.7243371838975772, + "blocks.0.w1.weight": 26.803582221943408, + "blocks.0.w1.bias": 17.489010687530495, + "blocks.0.w2.weight": 62.11992968105335, + "blocks.1.ln.weight": 1.1537855920658724, + "blocks.1.w1.weight": 16.75993270124035, + "blocks.1.w1.bias": 9.810533469107403, + "blocks.1.w2.weight": 38.80660527636343, + "out_ln.weight": 0.460453695659119, + "out_head.weight": 4.155803649723169, + "out_head.bias": 13.238295370364474 + } + } + }, + "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": [ + 5 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L2_seed5", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |