diff options
Diffstat (limited to 'results/fa_dfa_d512_L2_seed1/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L2_seed1/results_cifar10.json | 749 |
1 files changed, 749 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L2_seed1/results_cifar10.json b/results/fa_dfa_d512_L2_seed1/results_cifar10.json new file mode 100644 index 0000000..0dd9e9b --- /dev/null +++ b/results/fa_dfa_d512_L2_seed1/results_cifar10.json @@ -0,0 +1,749 @@ +{ + "1": { + "dfa": { + "log": { + "train_loss": [ + 2.059094401855469, + 2.044447731781006, + 2.0331237268066404, + 2.029790968055725, + 2.0203696198272705, + 2.019836251373291, + 2.0172710887908933, + 2.0157542709350587, + 2.01279447265625, + 2.0095205695343017, + 2.0108920065689087, + 2.0079552867889405, + 2.0044708168792726, + 2.0045814642333983, + 2.008353828582764, + 2.0057097484970092, + 2.0044997080993654, + 2.008672186203003, + 2.007922875289917, + 2.005835166091919, + 2.0033903245544433, + 2.0044442585754396, + 2.002520341949463, + 2.00441702545166, + 2.0058324599456787, + 2.004086421432495, + 2.0051950196838377, + 2.0038982134246828, + 2.004509604110718, + 2.0052878368377685, + 2.002380726890564, + 2.004787947921753, + 2.0059314285278322, + 2.00248601272583, + 2.0011240525817873, + 2.0024265776062014, + 2.000450315246582, + 2.003341311035156, + 2.0023523516082764, + 2.0043818824768067, + 2.0009307250213624, + 1.9997900833129882, + 1.99952888671875, + 1.9998678870010376, + 2.000314803161621, + 2.0005521519470215, + 1.9993469972991944, + 2.0002477783966066, + 1.9992239441680908, + 2.000459006500244, + 1.999985199661255, + 1.9989831493377686, + 1.9985046770477295, + 1.999920511932373, + 2.001435203552246, + 1.999222763671875, + 1.9992314191436769, + 1.9978362256240845, + 1.9966827032470704, + 1.9963577353286743, + 1.9981027877807618, + 1.998446220703125, + 1.999375821762085, + 1.9966239959716796, + 1.9978697190856933, + 1.9981718766021728, + 1.9964544176864625, + 1.996339178161621, + 1.995297778892517, + 1.9981674702453613, + 1.9945542547988893, + 1.9964004328918457, + 1.9957070028686523, + 1.9959317153930665, + 1.996950538787842, + 1.9957062911987304, + 1.9937946883392335, + 1.9961785815429687, + 1.996840117225647, + 1.9966120117950439, + 1.9950081676483153, + 1.9959315377807618, + 1.9947433185195922, + 1.9953646588897704, + 1.9961627129745483, + 1.9956305498504638, + 1.9933074866104126, + 1.9939569332122802, + 1.9948224069213867, + 1.9938344969940185, + 1.9951542666625977, + 1.994460453224182, + 1.9951913018798828, + 1.992417700653076, + 1.9943852715682984, + 1.9950172023010253, + 1.9913645124816894, + 1.9934076538467407, + 1.993856616821289, + 1.9950498567962647 + ], + "train_acc": [ + 0.24538, + 0.2525, + 0.24984, + 0.25596, + 0.25824, + 0.26188, + 0.26018, + 0.26256, + 0.26788, + 0.26836, + 0.26572, + 0.26712, + 0.26946, + 0.2707, + 0.26872, + 0.26986, + 0.26942, + 0.26706, + 0.2688, + 0.26814, + 0.2714, + 0.26896, + 0.26966, + 0.27136, + 0.26994, + 0.27172, + 0.27024, + 0.2708, + 0.27018, + 0.27212, + 0.27326, + 0.2728, + 0.2694, + 0.27118, + 0.27332, + 0.2738, + 0.27492, + 0.27382, + 0.27348, + 0.27124, + 0.27338, + 0.27632, + 0.27474, + 0.27438, + 0.27608, + 0.27216, + 0.27472, + 0.27174, + 0.2764, + 0.27286, + 0.27426, + 0.27344, + 0.27646, + 0.27692, + 0.27348, + 0.27398, + 0.2745, + 0.27462, + 0.27702, + 0.2742, + 0.27524, + 0.27522, + 0.2742, + 0.27684, + 0.27522, + 0.2755, + 0.2759, + 0.27692, + 0.27698, + 0.27796, + 0.27768, + 0.27376, + 0.27822, + 0.2749, + 0.27752, + 0.2773, + 0.27838, + 0.27666, + 0.27632, + 0.27526, + 0.27792, + 0.27644, + 0.2755, + 0.27696, + 0.27616, + 0.27868, + 0.27898, + 0.27856, + 0.27766, + 0.27746, + 0.27564, + 0.27836, + 0.27742, + 0.27794, + 0.27654, + 0.277, + 0.28076, + 0.27934, + 0.27814, + 0.27694 + ], + "test_acc": [ + 0.2655, + 0.2561, + 0.2732, + 0.2647, + 0.2937, + 0.2959, + 0.2856, + 0.2952, + 0.2909, + 0.2903, + 0.2948, + 0.2937, + 0.2853, + 0.2983, + 0.2836, + 0.2904, + 0.2751, + 0.294, + 0.2826, + 0.2783, + 0.3006, + 0.2986, + 0.2981, + 0.2913, + 0.27, + 0.2959, + 0.2941, + 0.2893, + 0.2816, + 0.2762, + 0.2842, + 0.2888, + 0.294, + 0.3006, + 0.2761, + 0.2995, + 0.2824, + 0.2895, + 0.2918, + 0.2959, + 0.2896, + 0.2977, + 0.2867, + 0.2906, + 0.2955, + 0.2965, + 0.2819, + 0.2848, + 0.293, + 0.2901, + 0.2903, + 0.2893, + 0.2946, + 0.286, + 0.2811, + 0.2948, + 0.2884, + 0.2963, + 0.2914, + 0.2919, + 0.2882, + 0.2923, + 0.2964, + 0.3036, + 0.3028, + 0.2942, + 0.2974, + 0.286, + 0.2997, + 0.2913, + 0.2942, + 0.2914, + 0.295, + 0.2929, + 0.2945, + 0.2905, + 0.2979, + 0.3003, + 0.2997, + 0.295, + 0.2937, + 0.2884, + 0.2946, + 0.2959, + 0.2977, + 0.2996, + 0.2999, + 0.2952, + 0.297, + 0.2936, + 0.2981, + 0.2977, + 0.2982, + 0.2978, + 0.2979, + 0.2976, + 0.2983, + 0.2977, + 0.2978, + 0.2978 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.41370105743408203, + -0.001240116311237216 + ], + "perturbation_rho": [ + 0.015020761638879776, + 0.0 + ], + "nudging": { + "0.001": [ + -4.163011908531189e-07, + 0.0 + ], + "0.003": [ + -1.1897645890712738e-06, + 0.0 + ], + "0.01": [ + -3.778841346502304e-06, + 9.313225746154785e-10 + ] + }, + "hidden_norms_per_layer": [ + 56646.390625, + 1782275456.0, + 3970110464.0 + ], + "bp_grad_norms_per_layer": [ + 2.493850104201556e-07, + 3.667711412358443e-10, + 3.6680619652784685e-10 + ] + }, + "drift": { + "embed.weight": 340.80661711332743, + "embed.bias": 276.58178672612297, + "blocks.0.ln.weight": 9.889068312709053, + "blocks.0.w1.weight": 318.9410688527977, + "blocks.0.w1.bias": 284.37318853440576, + "blocks.0.w2.weight": 489.2728313372279, + "blocks.1.ln.weight": 9.298819559476074, + "blocks.1.w1.weight": 366.2278399546544, + "blocks.1.w1.bias": 340.3854434800174, + "blocks.1.w2.weight": 341.5363762004093, + "out_ln.weight": 0.5301057051408113, + "out_head.weight": 8.039116173269838, + "out_head.bias": 2.5371126762269562 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.05515945022583, + 1.9500655529785156, + 1.9110925933456422, + 1.8819315839004516, + 1.8608471837615967, + 1.8478346523666382, + 1.8411366244125367, + 1.8377560669326782, + 1.8311604034805298, + 1.8246206870269774, + 1.8166842291641236, + 1.8113972366714477, + 1.8071093799209594, + 1.7970440604400635, + 1.7988952663803102, + 1.7952887838363647, + 1.7926777132415772, + 1.790428493423462, + 1.7899820502090453, + 1.7887613116836547, + 1.7811174698257446, + 1.781726747894287, + 1.782828377418518, + 1.7825099970245362, + 1.7798745111846923, + 1.7748807119369507, + 1.769409514427185, + 1.7675959386444091, + 1.7701947622680665, + 1.7674276565551759, + 1.7609349981307982, + 1.7605391665267944, + 1.762060026626587, + 1.7580202764129638, + 1.7636039975357056, + 1.7673666812133788, + 1.7655566649627685, + 1.7669985010147096, + 1.7656920331192016, + 1.7639919567108153, + 1.7634542723846436, + 1.7601457201766968, + 1.761037275123596, + 1.7602271603775024, + 1.7624055517578125, + 1.7582754214477538, + 1.7596718951416015, + 1.7575918433380127, + 1.757328048439026, + 1.7560129935455322, + 1.7593553677749634, + 1.75662100856781, + 1.7546059392547608, + 1.7581290004730226, + 1.7570533187103272, + 1.7555389776992798, + 1.755943459587097, + 1.752056611404419, + 1.752125501060486, + 1.7542531232452392, + 1.7525837285614014, + 1.7537434762954711, + 1.7566586553573609, + 1.754645963783264, + 1.7561360279083251, + 1.7577947814178467, + 1.7517323531723021, + 1.7506846601104735, + 1.7523996893310547, + 1.7564484720611573, + 1.7542920526885986, + 1.7536059958648682, + 1.753201146583557, + 1.7531446646118165, + 1.7524780932617188, + 1.7523272793197633, + 1.752608459777832, + 1.7483984720611572, + 1.7505473443984985, + 1.750156759109497, + 1.7495093181991577, + 1.7508859223175048, + 1.7515791293716432, + 1.7466206330108642, + 1.751723793373108, + 1.750410927658081, + 1.7431821329116821, + 1.7460082646942139, + 1.7443738849639892, + 1.7485774746704101, + 1.747268639907837, + 1.7440908068847656, + 1.7475049467468262, + 1.7451262094497682, + 1.747353721961975, + 1.7474739736557008, + 1.7444730795669556, + 1.748691442642212, + 1.743288662033081, + 1.7470912045669555 + ], + "train_acc": [ + 0.25462, + 0.29498, + 0.31256, + 0.32432, + 0.33206, + 0.33724, + 0.33776, + 0.34132, + 0.34392, + 0.34576, + 0.35074, + 0.34868, + 0.35426, + 0.35624, + 0.35478, + 0.35702, + 0.35892, + 0.3572, + 0.35788, + 0.35796, + 0.3596, + 0.36274, + 0.3592, + 0.36092, + 0.36282, + 0.36558, + 0.36466, + 0.36602, + 0.36356, + 0.36714, + 0.36908, + 0.36888, + 0.36582, + 0.368, + 0.365, + 0.3677, + 0.3674, + 0.3658, + 0.36706, + 0.368, + 0.3688, + 0.3678, + 0.36976, + 0.36634, + 0.36654, + 0.3687, + 0.36986, + 0.36986, + 0.3703, + 0.37068, + 0.36806, + 0.37002, + 0.37306, + 0.3722, + 0.36826, + 0.37146, + 0.3704, + 0.37428, + 0.37186, + 0.37382, + 0.37398, + 0.3711, + 0.3723, + 0.37532, + 0.37158, + 0.37096, + 0.37172, + 0.37316, + 0.37314, + 0.37438, + 0.37196, + 0.3737, + 0.37458, + 0.37306, + 0.37588, + 0.37444, + 0.37408, + 0.37832, + 0.37416, + 0.37576, + 0.3746, + 0.375, + 0.37462, + 0.37456, + 0.37364, + 0.37744, + 0.37852, + 0.37832, + 0.3778, + 0.37706, + 0.37604, + 0.37654, + 0.37828, + 0.37762, + 0.37686, + 0.377, + 0.37886, + 0.37662, + 0.37522, + 0.37652 + ], + "test_acc": [ + 0.2978, + 0.3334, + 0.3399, + 0.3342, + 0.3634, + 0.3618, + 0.3505, + 0.3521, + 0.3696, + 0.3722, + 0.3679, + 0.3787, + 0.3663, + 0.3811, + 0.3611, + 0.365, + 0.3543, + 0.3666, + 0.3501, + 0.36, + 0.3572, + 0.3504, + 0.3467, + 0.3598, + 0.3353, + 0.3474, + 0.3382, + 0.3432, + 0.3617, + 0.3397, + 0.353, + 0.3288, + 0.3385, + 0.3414, + 0.3467, + 0.3389, + 0.3341, + 0.3434, + 0.3537, + 0.3496, + 0.3526, + 0.346, + 0.3603, + 0.3526, + 0.3342, + 0.346, + 0.3392, + 0.3277, + 0.3535, + 0.3419, + 0.3396, + 0.3404, + 0.3433, + 0.3468, + 0.339, + 0.3361, + 0.3421, + 0.3363, + 0.3544, + 0.3465, + 0.3454, + 0.3472, + 0.3401, + 0.3454, + 0.3511, + 0.343, + 0.3445, + 0.3368, + 0.3424, + 0.3425, + 0.3381, + 0.3435, + 0.3496, + 0.346, + 0.3411, + 0.3408, + 0.3512, + 0.3469, + 0.3473, + 0.3419, + 0.3554, + 0.3449, + 0.345, + 0.3431, + 0.3442, + 0.3501, + 0.3534, + 0.3469, + 0.3496, + 0.3436, + 0.3478, + 0.3501, + 0.3492, + 0.3481, + 0.3474, + 0.3466, + 0.3476, + 0.3468, + 0.3473, + 0.3471 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.007958251982927322, + 0.9600946307182312 + ], + "perturbation_rho": [ + 0.029625363647937775, + 0.18801619112491608 + ], + "nudging": { + "0.001": [ + 4.2957253754138947e-07, + -8.616363629698753e-06 + ], + "0.003": [ + 1.2052478268742561e-06, + -2.6152702048420906e-05 + ], + "0.01": [ + 4.0422892197966576e-06, + -8.745735976845026e-05 + ] + }, + "hidden_norms_per_layer": [ + 6253.97314453125, + 339687.71875, + 217173.15625 + ], + "bp_grad_norms_per_layer": [ + 1.8587075828691013e-05, + 8.973949547907978e-07, + 7.725411705905572e-07 + ] + }, + "drift": { + "embed.weight": 35.41278218973057, + "embed.bias": 29.83841191850244, + "blocks.0.ln.weight": 1.4875314025636917, + "blocks.0.w1.weight": 25.996758962578763, + "blocks.0.w1.bias": 15.439034204651744, + "blocks.0.w2.weight": 58.48910743468254, + "blocks.1.ln.weight": 1.2261067193625912, + "blocks.1.w1.weight": 24.34549540914942, + "blocks.1.w1.bias": 12.006957572444392, + "blocks.1.w2.weight": 29.62795958064609, + "out_ln.weight": 0.4822583681094011, + "out_head.weight": 3.604309031698496, + "out_head.bias": 10.930487632919764 + } + } + }, + "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": [ + 1 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L2_seed1", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |