diff options
Diffstat (limited to 'results/fa_dfa_d512_L2_seed0/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L2_seed0/results_cifar10.json | 749 |
1 files changed, 749 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L2_seed0/results_cifar10.json b/results/fa_dfa_d512_L2_seed0/results_cifar10.json new file mode 100644 index 0000000..7d68e48 --- /dev/null +++ b/results/fa_dfa_d512_L2_seed0/results_cifar10.json @@ -0,0 +1,749 @@ +{ + "0": { + "dfa": { + "log": { + "train_loss": [ + 2.062894982070923, + 2.042903160247803, + 2.0362783416748047, + 2.0289835427093506, + 2.0305548177719115, + 2.027372160644531, + 2.0264537405395506, + 2.024415637550354, + 2.0239928167343137, + 2.0210681298828126, + 2.0223728131103518, + 2.0176786605072023, + 2.0183338179016115, + 2.015661160888672, + 2.017801855773926, + 2.0157006243896483, + 2.013551385803223, + 2.013131450958252, + 2.012682000961304, + 2.0142799925231936, + 2.010995005722046, + 2.01094291557312, + 2.010355429763794, + 2.00825265914917, + 2.0103073081207277, + 2.0115878075790405, + 2.00740104183197, + 2.008656568450928, + 2.007300491104126, + 2.0068678036499024, + 2.0060133220672607, + 2.005689826889038, + 2.0071374508666993, + 2.0041723249053955, + 2.007439960021973, + 2.004801420516968, + 2.007543896255493, + 2.008051019935608, + 2.0040309619140624, + 2.002984625930786, + 2.0036395249176024, + 2.0038933919906614, + 2.000672343521118, + 2.002527554702759, + 2.0040715547180175, + 2.002382646026611, + 2.002094281616211, + 2.0012509018707276, + 2.0035453020477294, + 2.0021117819213865, + 2.0006714540863038, + 2.0024239098739622, + 2.000547568016052, + 2.0004410552978515, + 2.000826226043701, + 1.9995253698730469, + 2.0001726056671143, + 2.0008310513687135, + 1.998274679107666, + 1.9978595404052735, + 1.9991289748382568, + 1.9976552139663697, + 1.9974908544540406, + 1.9972500681304932, + 1.9982564139556884, + 1.9990739729309082, + 1.9982671169281006, + 1.9968176255035401, + 1.9985010131072998, + 1.997684263381958, + 1.9971537530517578, + 1.9961954132843018, + 1.9972638996124268, + 1.9965271125793458, + 1.9960845895385741, + 1.9970510581970216, + 1.9964770040512085, + 1.9966202107238769, + 1.99665422416687, + 1.9977766342163086, + 1.9974106648254395, + 1.9950730167007447, + 1.9971654697418213, + 1.995940417022705, + 1.9962149201202393, + 1.9963797412109374, + 1.9944206412506102, + 1.9952468407440185, + 1.9964326373291015, + 1.99426823387146, + 1.9950427423477173, + 1.9937485892105102, + 1.995298335914612, + 1.9937747159194947, + 1.99329125541687, + 1.9960021032333375, + 1.9957950158309936, + 1.9929876937103272, + 1.994628695678711, + 1.9954896281433105 + ], + "train_acc": [ + 0.2387, + 0.24782, + 0.24834, + 0.255, + 0.2542, + 0.25346, + 0.2542, + 0.25464, + 0.25776, + 0.25884, + 0.25932, + 0.26184, + 0.26154, + 0.2615, + 0.26154, + 0.26214, + 0.26598, + 0.26548, + 0.26678, + 0.26396, + 0.2672, + 0.2662, + 0.26618, + 0.26934, + 0.2649, + 0.26488, + 0.26854, + 0.2669, + 0.26922, + 0.27032, + 0.2688, + 0.26942, + 0.26852, + 0.2698, + 0.2708, + 0.27092, + 0.27064, + 0.2683, + 0.27222, + 0.2735, + 0.27254, + 0.26794, + 0.27416, + 0.27154, + 0.2709, + 0.2735, + 0.27234, + 0.2735, + 0.2704, + 0.27326, + 0.27246, + 0.2728, + 0.27374, + 0.2747, + 0.27344, + 0.2739, + 0.2744, + 0.27318, + 0.27582, + 0.27816, + 0.27576, + 0.27492, + 0.27696, + 0.27608, + 0.27436, + 0.27578, + 0.27502, + 0.27806, + 0.274, + 0.27654, + 0.27598, + 0.27728, + 0.2783, + 0.27636, + 0.27394, + 0.27612, + 0.2788, + 0.27772, + 0.27614, + 0.27662, + 0.27458, + 0.27506, + 0.27614, + 0.2755, + 0.27668, + 0.27622, + 0.27808, + 0.27856, + 0.27796, + 0.27642, + 0.27808, + 0.28, + 0.27782, + 0.27638, + 0.27682, + 0.2774, + 0.27772, + 0.27642, + 0.27804, + 0.27828 + ], + "test_acc": [ + 0.2555, + 0.2749, + 0.2689, + 0.2689, + 0.2829, + 0.2478, + 0.2839, + 0.2723, + 0.2772, + 0.2879, + 0.2743, + 0.2663, + 0.2681, + 0.2905, + 0.2902, + 0.284, + 0.2788, + 0.2919, + 0.2851, + 0.2912, + 0.2889, + 0.2929, + 0.2868, + 0.2687, + 0.2855, + 0.304, + 0.292, + 0.2903, + 0.2882, + 0.2887, + 0.2923, + 0.2995, + 0.2973, + 0.2915, + 0.2914, + 0.2869, + 0.2737, + 0.2922, + 0.2898, + 0.2812, + 0.2924, + 0.3001, + 0.305, + 0.2996, + 0.2954, + 0.3002, + 0.2926, + 0.2858, + 0.3015, + 0.2967, + 0.2964, + 0.2897, + 0.3042, + 0.286, + 0.294, + 0.2986, + 0.2865, + 0.3013, + 0.2935, + 0.2824, + 0.2985, + 0.2908, + 0.3008, + 0.2949, + 0.3059, + 0.289, + 0.2914, + 0.305, + 0.2975, + 0.3007, + 0.3019, + 0.3009, + 0.2996, + 0.2976, + 0.2945, + 0.2986, + 0.3022, + 0.2991, + 0.2976, + 0.2988, + 0.2988, + 0.2979, + 0.2974, + 0.298, + 0.3014, + 0.2972, + 0.3003, + 0.2981, + 0.3005, + 0.3031, + 0.3026, + 0.3007, + 0.3004, + 0.2997, + 0.3016, + 0.3003, + 0.2993, + 0.2995, + 0.3002, + 0.3001 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.38934820890426636, + -0.0007739475695416331 + ], + "perturbation_rho": [ + -0.011460170149803162, + 0.0 + ], + "nudging": { + "0.001": [ + -4.163011908531189e-07, + 0.0 + ], + "0.003": [ + -1.0598450899124146e-06, + 0.0 + ], + "0.01": [ + -3.500375896692276e-06, + 0.0 + ] + }, + "hidden_norms_per_layer": [ + 52784.04296875, + 1514456960.0, + 4954951168.0 + ], + "bp_grad_norms_per_layer": [ + 2.3542017402178317e-07, + 3.12540243685433e-10, + 3.13137349383652e-10 + ] + }, + "drift": { + "embed.weight": 333.9178900630863, + "embed.bias": 277.5308018858926, + "blocks.0.ln.weight": 9.301640363801488, + "blocks.0.w1.weight": 311.12199748150937, + "blocks.0.w1.bias": 280.3822361773791, + "blocks.0.w2.weight": 500.10543177407226, + "blocks.1.ln.weight": 9.416271027168449, + "blocks.1.w1.weight": 397.745772813683, + "blocks.1.w1.bias": 381.1405085841875, + "blocks.1.w2.weight": 396.08258246162814, + "out_ln.weight": 0.5525577953421494, + "out_head.weight": 8.092380783169268, + "out_head.bias": 3.6707232448212266 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.074207492980957, + 1.9848071645736693, + 1.943194331741333, + 1.915086039428711, + 1.8987552841186524, + 1.8773444555282592, + 1.8620493222808838, + 1.8454281293106078, + 1.843613411216736, + 1.8376456963348389, + 1.8340519763946532, + 1.8251803881072999, + 1.8269455047607421, + 1.825891435623169, + 1.8311459658432008, + 1.8294941538238525, + 1.8284651235580445, + 1.8239954425811769, + 1.824907121887207, + 1.826982011680603, + 1.8236070413970946, + 1.8217376638412475, + 1.82049068069458, + 1.8151681272125244, + 1.8155477234649657, + 1.8134433920288087, + 1.806967512512207, + 1.8068813376235962, + 1.801505942955017, + 1.8042971012115478, + 1.800074999961853, + 1.7939591689300538, + 1.792923740272522, + 1.7907480084991456, + 1.7882287372589112, + 1.7897650888824463, + 1.7839135689926147, + 1.7830314935302733, + 1.7808781423187257, + 1.7737779104614257, + 1.7762451565933228, + 1.7733127856826782, + 1.7701874697875977, + 1.7670377206802368, + 1.7671197887802124, + 1.7665505773162842, + 1.7623907526016236, + 1.7633579796600343, + 1.7624771506500243, + 1.7620069681167603, + 1.7585726612091064, + 1.7578929253387452, + 1.7546437027740478, + 1.7530586669921875, + 1.7568175424575805, + 1.7472540600585937, + 1.7497945638656616, + 1.7486419037246703, + 1.7454212425994873, + 1.7467387356948854, + 1.742750499343872, + 1.7459785363006592, + 1.7462484993743896, + 1.7419252165985109, + 1.7439173141479491, + 1.7400725153350831, + 1.7417476744842528, + 1.7432426296234131, + 1.7410478344726563, + 1.735186776046753, + 1.7373164199447633, + 1.7379439109039307, + 1.7370698404312135, + 1.7362006621932984, + 1.7366024084091187, + 1.734513318862915, + 1.7316589307403565, + 1.734174518661499, + 1.7326682236480713, + 1.7339326565170288, + 1.734504469909668, + 1.728739596824646, + 1.7321637844848632, + 1.7324034854125976, + 1.7288938723754883, + 1.7308405535888671, + 1.7306951354980469, + 1.7300820154190064, + 1.7285541592788696, + 1.7268984225845336, + 1.727956399230957, + 1.724655009765625, + 1.7292626781845093, + 1.7271040807723999, + 1.7264983687973023, + 1.7272509225845336, + 1.7300393019104003, + 1.7276934866714477, + 1.7269879544067384, + 1.72983638671875 + ], + "train_acc": [ + 0.24412, + 0.28062, + 0.29426, + 0.30988, + 0.31292, + 0.32326, + 0.32948, + 0.33612, + 0.33972, + 0.34158, + 0.34404, + 0.34674, + 0.34434, + 0.34662, + 0.3424, + 0.3445, + 0.34358, + 0.34696, + 0.34462, + 0.3454, + 0.34674, + 0.3471, + 0.34642, + 0.34934, + 0.34898, + 0.35068, + 0.35258, + 0.3542, + 0.35448, + 0.35074, + 0.354, + 0.35526, + 0.35522, + 0.35664, + 0.36026, + 0.36056, + 0.36136, + 0.35988, + 0.3653, + 0.3664, + 0.36336, + 0.36712, + 0.36668, + 0.36956, + 0.36908, + 0.36934, + 0.3731, + 0.36976, + 0.36908, + 0.3699, + 0.37166, + 0.37488, + 0.3728, + 0.37648, + 0.37234, + 0.3769, + 0.3778, + 0.37956, + 0.37994, + 0.37854, + 0.3793, + 0.38058, + 0.3788, + 0.3813, + 0.3811, + 0.38082, + 0.3786, + 0.38008, + 0.37906, + 0.383, + 0.3824, + 0.38106, + 0.38238, + 0.38434, + 0.37922, + 0.3843, + 0.38626, + 0.38438, + 0.38332, + 0.38316, + 0.3814, + 0.3837, + 0.38628, + 0.3838, + 0.38572, + 0.38528, + 0.3855, + 0.38382, + 0.38534, + 0.38696, + 0.3866, + 0.38526, + 0.38498, + 0.38372, + 0.38534, + 0.38782, + 0.38812, + 0.38732, + 0.38564, + 0.38654 + ], + "test_acc": [ + 0.2891, + 0.3185, + 0.3279, + 0.3405, + 0.3535, + 0.3324, + 0.3578, + 0.3568, + 0.3611, + 0.368, + 0.3716, + 0.3657, + 0.3624, + 0.37, + 0.3642, + 0.3776, + 0.3742, + 0.3629, + 0.3623, + 0.3785, + 0.3689, + 0.3538, + 0.3593, + 0.3592, + 0.3713, + 0.3788, + 0.3722, + 0.3675, + 0.3747, + 0.3758, + 0.3671, + 0.3866, + 0.3809, + 0.3827, + 0.3781, + 0.3711, + 0.375, + 0.3801, + 0.3829, + 0.3886, + 0.3844, + 0.3859, + 0.3842, + 0.3854, + 0.3855, + 0.3871, + 0.3835, + 0.3874, + 0.3931, + 0.3825, + 0.3856, + 0.3901, + 0.3877, + 0.3854, + 0.3894, + 0.3873, + 0.3886, + 0.3982, + 0.3923, + 0.3881, + 0.3883, + 0.3901, + 0.3907, + 0.3926, + 0.3942, + 0.3933, + 0.3851, + 0.3912, + 0.393, + 0.3928, + 0.3961, + 0.3907, + 0.3876, + 0.3881, + 0.3875, + 0.3885, + 0.3909, + 0.3893, + 0.394, + 0.3959, + 0.3886, + 0.3945, + 0.3932, + 0.3919, + 0.3906, + 0.3944, + 0.3939, + 0.3946, + 0.3923, + 0.393, + 0.3929, + 0.3923, + 0.3915, + 0.3948, + 0.3934, + 0.3929, + 0.3932, + 0.3936, + 0.394, + 0.3938 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.017680617049336433, + 0.8765696287155151 + ], + "perturbation_rho": [ + -0.00557959359139204, + 0.06313759833574295 + ], + "nudging": { + "0.001": [ + -2.048211172223091e-06, + -3.955443389713764e-06 + ], + "0.003": [ + -6.195507012307644e-06, + -1.213036011904478e-05 + ], + "0.01": [ + -2.056185621768236e-05, + -4.089018329977989e-05 + ] + }, + "hidden_norms_per_layer": [ + 5668.7490234375, + 427742.1875, + 626308.1875 + ], + "bp_grad_norms_per_layer": [ + 2.2485837689600885e-05, + 1.0125075959876995e-06, + 7.618949666721164e-07 + ] + }, + "drift": { + "embed.weight": 43.36126831717888, + "embed.bias": 11.647761057592673, + "blocks.0.ln.weight": 1.7734721493198804, + "blocks.0.w1.weight": 28.788589686086883, + "blocks.0.w1.bias": 16.981347027161885, + "blocks.0.w2.weight": 61.91338151396999, + "blocks.1.ln.weight": 1.2975357664484612, + "blocks.1.w1.weight": 23.055782710684934, + "blocks.1.w1.bias": 17.973225870539384, + "blocks.1.w2.weight": 38.08861834310624, + "out_ln.weight": 0.38876131505924627, + "out_head.weight": 5.716507111968003, + "out_head.bias": 4.553337714995325 + } + } + }, + "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": [ + 0 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L2_seed0", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |