diff options
Diffstat (limited to 'results/fa_dfa_d512_L2_seed2/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L2_seed2/results_cifar10.json | 749 |
1 files changed, 749 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L2_seed2/results_cifar10.json b/results/fa_dfa_d512_L2_seed2/results_cifar10.json new file mode 100644 index 0000000..35785f5 --- /dev/null +++ b/results/fa_dfa_d512_L2_seed2/results_cifar10.json @@ -0,0 +1,749 @@ +{ + "2": { + "dfa": { + "log": { + "train_loss": [ + 2.058412809791565, + 2.0388552247619627, + 2.0358241034317017, + 2.027553281555176, + 2.030367368774414, + 2.027717582397461, + 2.023588889923096, + 2.022263825721741, + 2.0241731433105468, + 2.0235239835357666, + 2.023893345298767, + 2.0222612105560303, + 2.028809658164978, + 2.023522481994629, + 2.023535331993103, + 2.022634300880432, + 2.0261124728393556, + 2.027209746055603, + 2.024631557235718, + 2.0272511472320556, + 2.0248798042297365, + 2.0252146803283693, + 2.021958327026367, + 2.0255450409317017, + 2.0234208207702635, + 2.0240782318496704, + 2.0233148418807985, + 2.022000389137268, + 2.024022118911743, + 2.023005671234131, + 2.0231490325927735, + 2.024285191497803, + 2.0222519898986815, + 2.023439755783081, + 2.023540186843872, + 2.0246322383117676, + 2.0237508728027342, + 2.024476735076904, + 2.0262484502410887, + 2.0244761968612672, + 2.023149300994873, + 2.0230725455093386, + 2.0237448512268066, + 2.0239054921722412, + 2.0238359395980834, + 2.0231575259399412, + 2.022854585723877, + 2.022712833328247, + 2.022943268432617, + 2.023613681335449, + 2.022585965042114, + 2.0229795806503295, + 2.0224396421051027, + 2.024438477630615, + 2.0226136851501466, + 2.0218412887191772, + 2.020350592918396, + 2.021553417892456, + 2.0207389210510254, + 2.022400696411133, + 2.0226713175201416, + 2.0205454586029052, + 2.022753714904785, + 2.0204339378356932, + 2.0214138391113283, + 2.019458531036377, + 2.018622496871948, + 2.0211455561828613, + 2.020830050392151, + 2.0200049672698976, + 2.0226647985839845, + 2.020451495361328, + 2.0213021927642822, + 2.018636092834473, + 2.0201395918273928, + 2.0197924830627443, + 2.0189465530395507, + 2.0179246043395995, + 2.0190702261352538, + 2.018002661895752, + 2.0188750094604493, + 2.0186908027648927, + 2.018660167617798, + 2.016986283912659, + 2.016896242828369, + 2.0198894203948976, + 2.0182291090393067, + 2.0168725957870484, + 2.0189685047912596, + 2.0170926274108885, + 2.0171643753814696, + 2.0170515134429934, + 2.017839207763672, + 2.0162586155700684, + 2.016942921524048, + 2.0172505876159668, + 2.016627454185486, + 2.01472389213562, + 2.016936295089722, + 2.0162249319458008 + ], + "train_acc": [ + 0.24042, + 0.24948, + 0.25212, + 0.25448, + 0.25656, + 0.25446, + 0.26134, + 0.2625, + 0.25946, + 0.26152, + 0.26086, + 0.26258, + 0.2591, + 0.25878, + 0.25972, + 0.2621, + 0.26092, + 0.25914, + 0.25866, + 0.25816, + 0.26052, + 0.25714, + 0.26258, + 0.25996, + 0.26152, + 0.2614, + 0.2641, + 0.26338, + 0.26254, + 0.26234, + 0.26476, + 0.26218, + 0.2652, + 0.26612, + 0.2645, + 0.26434, + 0.26412, + 0.26428, + 0.26476, + 0.26552, + 0.26592, + 0.2645, + 0.26798, + 0.26322, + 0.26514, + 0.2651, + 0.26806, + 0.26828, + 0.26474, + 0.26502, + 0.26398, + 0.26496, + 0.26396, + 0.265, + 0.26866, + 0.26958, + 0.2685, + 0.26846, + 0.26926, + 0.27102, + 0.26792, + 0.26856, + 0.26608, + 0.26824, + 0.27114, + 0.26984, + 0.2685, + 0.26856, + 0.26812, + 0.26898, + 0.26946, + 0.26888, + 0.26992, + 0.2687, + 0.2679, + 0.2708, + 0.26838, + 0.27112, + 0.27018, + 0.26964, + 0.27128, + 0.27114, + 0.27024, + 0.27024, + 0.27036, + 0.271, + 0.27062, + 0.27166, + 0.27026, + 0.27294, + 0.2727, + 0.27276, + 0.27036, + 0.27272, + 0.26992, + 0.26996, + 0.2709, + 0.27316, + 0.2732, + 0.27336 + ], + "test_acc": [ + 0.278, + 0.2738, + 0.2583, + 0.2773, + 0.2658, + 0.2783, + 0.2681, + 0.2957, + 0.2671, + 0.2647, + 0.287, + 0.2846, + 0.284, + 0.287, + 0.2724, + 0.2925, + 0.2763, + 0.2711, + 0.2614, + 0.277, + 0.2817, + 0.2791, + 0.2834, + 0.2821, + 0.2641, + 0.2764, + 0.2963, + 0.2882, + 0.3005, + 0.2923, + 0.2579, + 0.2889, + 0.2999, + 0.2958, + 0.2739, + 0.2984, + 0.286, + 0.2643, + 0.2829, + 0.2919, + 0.292, + 0.2875, + 0.2932, + 0.2778, + 0.2753, + 0.291, + 0.2955, + 0.3014, + 0.2887, + 0.2856, + 0.2937, + 0.2946, + 0.298, + 0.2932, + 0.3005, + 0.2989, + 0.2924, + 0.2977, + 0.2922, + 0.2876, + 0.2943, + 0.2844, + 0.3003, + 0.2952, + 0.2881, + 0.2821, + 0.2934, + 0.28, + 0.2958, + 0.2933, + 0.2923, + 0.2941, + 0.2872, + 0.2946, + 0.2966, + 0.2962, + 0.2915, + 0.2969, + 0.2988, + 0.2968, + 0.2991, + 0.2972, + 0.297, + 0.2955, + 0.2973, + 0.2994, + 0.2959, + 0.2962, + 0.3, + 0.2971, + 0.2977, + 0.2962, + 0.2963, + 0.2955, + 0.2959, + 0.2964, + 0.2962, + 0.2963, + 0.2968, + 0.2968 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3571905195713043, + -2.4819419195409864e-05 + ], + "perturbation_rho": [ + -0.020394116640090942, + 0.0 + ], + "nudging": { + "0.001": [ + -3.2084062695503235e-07, + 0.0 + ], + "0.003": [ + -8.596107363700867e-07, + 0.0 + ], + "0.01": [ + -2.905726432800293e-06, + 0.0 + ] + }, + "hidden_norms_per_layer": [ + 53198.71875, + 2227773184.0, + 5250976256.0 + ], + "bp_grad_norms_per_layer": [ + 2.0440073456029495e-07, + 3.7151820508896094e-10, + 3.7148767395578375e-10 + ] + }, + "drift": { + "embed.weight": 342.0501862953785, + "embed.bias": 318.0357815319326, + "blocks.0.ln.weight": 9.910735164962347, + "blocks.0.w1.weight": 324.42259888737226, + "blocks.0.w1.bias": 352.6779805020607, + "blocks.0.w2.weight": 492.2710513712036, + "blocks.1.ln.weight": 9.723967303148715, + "blocks.1.w1.weight": 403.31369239097415, + "blocks.1.w1.bias": 386.0984329662907, + "blocks.1.w2.weight": 397.9343379416526, + "out_ln.weight": 0.5056100406607513, + "out_head.weight": 8.24905668320801, + "out_head.bias": 2.0933992602399067 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.063060122451782, + 1.9739086376571655, + 1.9282753707504272, + 1.8921499670410156, + 1.8799907299804688, + 1.8701481190490723, + 1.8541997326660156, + 1.8432505453109742, + 1.8392359512710572, + 1.8295656701278686, + 1.8286965405273437, + 1.8275740008163452, + 1.82835955909729, + 1.8276023639297485, + 1.8295053287506104, + 1.8279134408569335, + 1.8347502783966065, + 1.8358594959259034, + 1.832546812095642, + 1.8375971967315674, + 1.8337256093978882, + 1.8388268453598022, + 1.8368015246200562, + 1.8367653689956664, + 1.837699118347168, + 1.8374284832763672, + 1.8332337552642821, + 1.8305227688217163, + 1.8352981665802002, + 1.8384290933227538, + 1.8319619774627685, + 1.8331132510375976, + 1.8284437879180908, + 1.8323823754882813, + 1.8337605920791626, + 1.8296470736312866, + 1.824372737045288, + 1.8253158220672607, + 1.8254631029891968, + 1.8240845959472656, + 1.819489220275879, + 1.8196685235214234, + 1.8173167004013062, + 1.8210680514526367, + 1.812099649734497, + 1.8139883127212524, + 1.810353461265564, + 1.8109969634246825, + 1.808874561843872, + 1.8115904225158692, + 1.8089021768188476, + 1.8070925519561767, + 1.807039188232422, + 1.8048222934341431, + 1.8031300540924071, + 1.8039923257446289, + 1.8042795065307617, + 1.8037835544586183, + 1.8002930462646485, + 1.8007701531982423, + 1.8004488678359984, + 1.8029598141098022, + 1.8020868547821045, + 1.7990351900482178, + 1.8022524103546143, + 1.7987030548477172, + 1.7969628913116455, + 1.7972235089874267, + 1.7918564282226563, + 1.7940529309082032, + 1.7972945267105103, + 1.7940796615219117, + 1.7942549993133545, + 1.7947112902450562, + 1.7927878363037109, + 1.7915658866119384, + 1.7944716805267333, + 1.790445680809021, + 1.7891779144668578, + 1.789405579185486, + 1.7867662616729736, + 1.7887766037368775, + 1.7875445496368407, + 1.7851270193862916, + 1.7886884448623657, + 1.786267046775818, + 1.785640559425354, + 1.7853610308074952, + 1.7856942657470702, + 1.7871524771499634, + 1.7883876535797119, + 1.782026519203186, + 1.786798097305298, + 1.7790444551849365, + 1.7840377599334716, + 1.787256519126892, + 1.7822379702758788, + 1.7822947565460205, + 1.7829025159072875, + 1.7819519442749023 + ], + "train_acc": [ + 0.24228, + 0.28314, + 0.30214, + 0.31828, + 0.32302, + 0.3277, + 0.33338, + 0.33888, + 0.34066, + 0.34536, + 0.34814, + 0.34648, + 0.34716, + 0.34526, + 0.3462, + 0.3422, + 0.343, + 0.34162, + 0.3438, + 0.34388, + 0.34354, + 0.33854, + 0.34312, + 0.34176, + 0.34514, + 0.34304, + 0.34458, + 0.34162, + 0.34348, + 0.34098, + 0.34374, + 0.34488, + 0.34328, + 0.34504, + 0.34414, + 0.34548, + 0.34552, + 0.34848, + 0.34836, + 0.34802, + 0.34918, + 0.346, + 0.34986, + 0.3473, + 0.3515, + 0.3527, + 0.3532, + 0.35212, + 0.35068, + 0.34998, + 0.3505, + 0.3565, + 0.35392, + 0.35556, + 0.35556, + 0.35462, + 0.3548, + 0.35674, + 0.35552, + 0.35848, + 0.35704, + 0.35594, + 0.35598, + 0.35716, + 0.35818, + 0.35842, + 0.35966, + 0.35968, + 0.36022, + 0.36082, + 0.3571, + 0.3612, + 0.3604, + 0.3607, + 0.3602, + 0.36134, + 0.35734, + 0.36262, + 0.36352, + 0.35916, + 0.36462, + 0.3617, + 0.3608, + 0.3617, + 0.36106, + 0.36262, + 0.36296, + 0.36354, + 0.36138, + 0.36132, + 0.36446, + 0.36452, + 0.36388, + 0.3645, + 0.36378, + 0.36342, + 0.36496, + 0.36556, + 0.36274, + 0.36546 + ], + "test_acc": [ + 0.2954, + 0.3137, + 0.3295, + 0.3474, + 0.3551, + 0.3544, + 0.3515, + 0.3694, + 0.3633, + 0.3479, + 0.3744, + 0.3726, + 0.371, + 0.3618, + 0.3649, + 0.3685, + 0.3658, + 0.3704, + 0.3504, + 0.3729, + 0.3578, + 0.3685, + 0.3676, + 0.3643, + 0.3562, + 0.3486, + 0.3732, + 0.3624, + 0.3693, + 0.341, + 0.3467, + 0.3544, + 0.3679, + 0.3526, + 0.3593, + 0.3647, + 0.3604, + 0.3662, + 0.3632, + 0.3644, + 0.3628, + 0.3598, + 0.3564, + 0.3633, + 0.3531, + 0.3607, + 0.3616, + 0.3636, + 0.3514, + 0.3488, + 0.3408, + 0.3522, + 0.3651, + 0.3408, + 0.3548, + 0.3504, + 0.3469, + 0.3507, + 0.3515, + 0.3434, + 0.3548, + 0.351, + 0.351, + 0.3545, + 0.3487, + 0.3467, + 0.3492, + 0.3439, + 0.346, + 0.3357, + 0.332, + 0.3387, + 0.3546, + 0.3415, + 0.3458, + 0.3451, + 0.3477, + 0.342, + 0.3446, + 0.3427, + 0.3426, + 0.3423, + 0.3481, + 0.3467, + 0.3443, + 0.3437, + 0.3466, + 0.3459, + 0.3448, + 0.3458, + 0.3433, + 0.3461, + 0.3479, + 0.3447, + 0.3452, + 0.3467, + 0.3461, + 0.3467, + 0.3464, + 0.3464 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.01565447635948658, + 0.9284564256668091 + ], + "perturbation_rho": [ + 0.08889118582010269, + 0.1751367598772049 + ], + "nudging": { + "0.001": [ + -8.567003533244133e-07, + -7.14592169970274e-06 + ], + "0.003": [ + -2.5028130039572716e-06, + -2.1490384824573994e-05 + ], + "0.01": [ + -8.566654287278652e-06, + -7.176969666033983e-05 + ] + }, + "hidden_norms_per_layer": [ + 5015.7705078125, + 205831.875, + 281989.59375 + ], + "bp_grad_norms_per_layer": [ + 1.882538890640717e-05, + 8.274267884189612e-07, + 8.043497246035258e-07 + ] + }, + "drift": { + "embed.weight": 31.381319271594045, + "embed.bias": 16.70881646410694, + "blocks.0.ln.weight": 1.5550725713952875, + "blocks.0.w1.weight": 29.555012389084762, + "blocks.0.w1.bias": 17.153011809958922, + "blocks.0.w2.weight": 62.82184507015774, + "blocks.1.ln.weight": 1.2734634249220895, + "blocks.1.w1.weight": 21.575349693350912, + "blocks.1.w1.bias": 10.895115401942027, + "blocks.1.w2.weight": 35.67159460439069, + "out_ln.weight": 0.4183849023916904, + "out_head.weight": 4.1274562698411295, + "out_head.bias": 12.873937810986282 + } + } + }, + "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": [ + 2 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L2_seed2", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |