diff options
Diffstat (limited to 'results/fa_depth_scan_d512_L2/results_cifar10.json')
| -rw-r--r-- | results/fa_depth_scan_d512_L2/results_cifar10.json | 389 |
1 files changed, 389 insertions, 0 deletions
diff --git a/results/fa_depth_scan_d512_L2/results_cifar10.json b/results/fa_depth_scan_d512_L2/results_cifar10.json new file mode 100644 index 0000000..0c41380 --- /dev/null +++ b/results/fa_depth_scan_d512_L2/results_cifar10.json @@ -0,0 +1,389 @@ +{ + "42": { + "fa": { + "log": { + "train_loss": [ + 2.061217526626587, + 1.9565438193130493, + 1.9224825448608398, + 1.9001438703155518, + 1.8821109796524047, + 1.8704035317230225, + 1.857736103477478, + 1.844048505897522, + 1.8342980657958985, + 1.8290129190063475, + 1.8224646506118773, + 1.8214716808319091, + 1.8136491051483155, + 1.8157705508041382, + 1.8199064575195312, + 1.816836177368164, + 1.8065774325180053, + 1.8125089934921266, + 1.8090396573638916, + 1.81023629032135, + 1.8082425650787353, + 1.8127569119644165, + 1.808132509727478, + 1.8062178915405274, + 1.8039376613998412, + 1.8006342431259155, + 1.797272424888611, + 1.795366856956482, + 1.795199264755249, + 1.796318228149414, + 1.7931722381973267, + 1.790380922317505, + 1.787453678970337, + 1.7863555522918702, + 1.7848892670059204, + 1.782602568359375, + 1.7844260739135742, + 1.7822325122833251, + 1.7821675496673584, + 1.7775094485092162, + 1.775276083984375, + 1.772415744934082, + 1.771381389389038, + 1.7729799353790283, + 1.7665789907073974, + 1.7650687371063232, + 1.7673343083953856, + 1.7628559505844117, + 1.7635160286712646, + 1.7610479801177978, + 1.757714970741272, + 1.7587621269989013, + 1.75955526802063, + 1.7605536008453369, + 1.7632213006973267, + 1.7597938927459718, + 1.764720836868286, + 1.7602048022460937, + 1.7601288507843018, + 1.7592250678253174, + 1.763182290611267, + 1.7588830117034913, + 1.7620349166107179, + 1.762400091934204, + 1.7608643816375733, + 1.7610327990341186, + 1.7634911743164063, + 1.761160830154419, + 1.7584447234344482, + 1.758816735496521, + 1.7580244549560546, + 1.758498070678711, + 1.757904008178711, + 1.7605663097381592, + 1.7611833373260497, + 1.7573552610778809, + 1.7571686492538452, + 1.7577348608016967, + 1.7563239881134034, + 1.7563734949493408, + 1.757859500465393, + 1.7567143533325196, + 1.7577434346771241, + 1.75511939453125, + 1.7555543961334228, + 1.756235351524353, + 1.7555060604095458, + 1.7561510982894897, + 1.7538081964874268, + 1.7542751317977905, + 1.7549605862426758, + 1.7509819832611084, + 1.755103257446289, + 1.7511512590789795, + 1.7497140390396118, + 1.755158496170044, + 1.7556276675033569, + 1.7520254708862304, + 1.7512920989227294, + 1.7526899521255492 + ], + "train_acc": [ + 0.24758, + 0.29054, + 0.30476, + 0.31646, + 0.3238, + 0.33244, + 0.334, + 0.33794, + 0.34238, + 0.34354, + 0.3469, + 0.34778, + 0.34922, + 0.34738, + 0.34754, + 0.3477, + 0.3522, + 0.35128, + 0.3501, + 0.34894, + 0.3502, + 0.34714, + 0.35396, + 0.35312, + 0.35266, + 0.35432, + 0.35642, + 0.35256, + 0.35418, + 0.35464, + 0.35604, + 0.35646, + 0.35824, + 0.35708, + 0.35856, + 0.35934, + 0.35824, + 0.35668, + 0.36006, + 0.36048, + 0.36242, + 0.3658, + 0.36572, + 0.36512, + 0.36582, + 0.36634, + 0.36442, + 0.36774, + 0.3675, + 0.36846, + 0.36932, + 0.36868, + 0.3692, + 0.36584, + 0.36652, + 0.3708, + 0.36522, + 0.36784, + 0.36796, + 0.37036, + 0.36874, + 0.37152, + 0.37144, + 0.37078, + 0.36992, + 0.37114, + 0.3711, + 0.37028, + 0.37068, + 0.37472, + 0.3726, + 0.372, + 0.37436, + 0.37286, + 0.37184, + 0.37572, + 0.37322, + 0.37538, + 0.37558, + 0.37514, + 0.37416, + 0.37244, + 0.37488, + 0.3776, + 0.37426, + 0.37688, + 0.37648, + 0.3746, + 0.37672, + 0.37686, + 0.3762, + 0.37536, + 0.3752, + 0.37636, + 0.37886, + 0.37486, + 0.3733, + 0.37878, + 0.3779, + 0.37676 + ], + "test_acc": [ + 0.3028, + 0.3196, + 0.3402, + 0.3524, + 0.3584, + 0.3591, + 0.3566, + 0.3543, + 0.3712, + 0.3705, + 0.3579, + 0.3373, + 0.3627, + 0.3557, + 0.3414, + 0.3688, + 0.3572, + 0.3632, + 0.354, + 0.3634, + 0.362, + 0.3549, + 0.363, + 0.3531, + 0.345, + 0.3459, + 0.3517, + 0.3421, + 0.3294, + 0.3398, + 0.3294, + 0.3409, + 0.3476, + 0.3318, + 0.3477, + 0.3302, + 0.3283, + 0.3309, + 0.3337, + 0.3497, + 0.3314, + 0.3292, + 0.3374, + 0.3361, + 0.3335, + 0.3461, + 0.3254, + 0.3366, + 0.3353, + 0.3288, + 0.3467, + 0.3374, + 0.3445, + 0.3415, + 0.3358, + 0.3474, + 0.3382, + 0.333, + 0.3356, + 0.3376, + 0.3356, + 0.3418, + 0.3358, + 0.3446, + 0.35, + 0.3412, + 0.3452, + 0.3429, + 0.3438, + 0.3421, + 0.3477, + 0.3497, + 0.3499, + 0.3473, + 0.3451, + 0.3504, + 0.3474, + 0.3503, + 0.344, + 0.3497, + 0.3487, + 0.3488, + 0.3521, + 0.3509, + 0.3455, + 0.3511, + 0.3469, + 0.3474, + 0.353, + 0.3472, + 0.3472, + 0.3501, + 0.3493, + 0.3492, + 0.349, + 0.3496, + 0.3496, + 0.3494, + 0.3494, + 0.3495 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.03441809490323067, + 0.955886721611023 + ], + "perturbation_rho": [ + 0.04754525423049927, + 0.05055658146739006 + ], + "nudging": { + "0.001": [ + -5.270587280392647e-06, + -7.06920400261879e-06 + ], + "0.003": [ + -1.57852191478014e-05, + -2.1289335563778877e-05 + ], + "0.01": [ + -5.261087790131569e-05, + -7.100868970155716e-05 + ] + }, + "hidden_norms_per_layer": [ + 4686.95556640625, + 94784.5546875, + 138362.359375 + ], + "bp_grad_norms_per_layer": [ + 1.9998093193862587e-05, + 1.4634815670433454e-06, + 1.188113401440205e-06 + ] + }, + "drift": { + "embed.weight": 28.586715545463985, + "embed.bias": 19.716378864552336, + "blocks.0.ln.weight": 1.4738119287890208, + "blocks.0.w1.weight": 26.045392429453724, + "blocks.0.w1.bias": 15.939058582542359, + "blocks.0.w2.weight": 59.12906119267325, + "blocks.1.ln.weight": 1.2124303196678357, + "blocks.1.w1.weight": 16.696964117465345, + "blocks.1.w1.bias": 8.535990814749171, + "blocks.1.w2.weight": 37.85731940298285, + "out_ln.weight": 0.5125606883018807, + "out_head.weight": 3.721893823331011, + "out_head.bias": 11.592811594248388 + } + } + }, + "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": [ + 42 + ], + "gpu": 0, + "output_dir": "results/fa_depth_scan_d512_L2", + "methods": [ + "fa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |