{ "3": { "dfa": { "log": { "train_loss": [ 2.049092872314453, 2.044887622756958, 2.062239527206421, 2.0690095069122316, 2.0704432483673094, 2.076322864227295, 2.0749057723999025, 2.0722323740005493, 2.0724080812835695, 2.072884361419678, 2.0764317562866212, 2.070321410064697, 2.0702889765930177, 2.067789299621582, 2.0646444761657716, 2.0636547829818728, 2.062572795333862, 2.0582885260772703, 2.059502869415283, 2.0571998697280884, 2.0527090773010253, 2.056031491851807, 2.0513366117095946, 2.0520858401870727, 2.0493150717926025, 2.0478450300598143, 2.0498407819366453, 2.043566691818237, 2.042290360069275, 2.039692493896484, 2.041093152770996, 2.0399529009246824, 2.039473579330444, 2.0396062004089357, 2.0359111277770996, 2.0347996918487548, 2.034452969818115, 2.033417662200928, 2.0335788008880615, 2.0308428141784667, 2.027684537124634, 2.0306087942504885, 2.0288818074798582, 2.028032119369507, 2.0284718214416504, 2.025789548034668, 2.0262292552948, 2.023198474197388, 2.024526604385376, 2.025226473312378, 2.0236859022521974, 2.021138525657654, 2.022560397567749, 2.0220771756744385, 2.0260347727203367, 2.022141723327637, 2.021563760910034, 2.018922806472778, 2.021806682510376, 2.0212029346466065, 2.01982150100708, 2.019622989349365, 2.0191136405944823, 2.0185832523345946, 2.0185110153961183, 2.0175517141342163, 2.0180020709991453, 2.0152184085845946, 2.0153978774261474, 2.017112913208008, 2.0173879592895507, 2.0179903555297853, 2.013354239501953, 2.016451063537598, 2.013973571510315, 2.018112847671509, 2.015232135925293, 2.014464662742615, 2.0156791037368773, 2.0115500025177, 2.014401368560791, 2.0142960264587404, 2.012123734397888, 2.0130788822937014, 2.0141370764923097, 2.014659231796265, 2.0124325815582274, 2.014389346160889, 2.0128879175567627, 2.0132610288238526, 2.014250590438843, 2.012836150970459, 2.0106287144470216, 2.014688469619751, 2.0122887326812746, 2.01329588142395, 2.0123015225601195, 2.010611874694824, 2.015006597442627, 2.0157135874176024 ], "train_acc": [ 0.25188, 0.25386, 0.24578, 0.2458, 0.24326, 0.2434, 0.23896, 0.24338, 0.24348, 0.24286, 0.24246, 0.24364, 0.24108, 0.24176, 0.24598, 0.24638, 0.24538, 0.24996, 0.24744, 0.24802, 0.25254, 0.2511, 0.2547, 0.2523, 0.25268, 0.25586, 0.2522, 0.25328, 0.25624, 0.25838, 0.25822, 0.25672, 0.25938, 0.25924, 0.25908, 0.259, 0.2602, 0.26116, 0.26152, 0.2633, 0.26614, 0.2616, 0.26606, 0.26386, 0.26708, 0.26482, 0.2682, 0.26386, 0.26484, 0.26862, 0.26774, 0.26798, 0.2692, 0.2687, 0.26776, 0.27036, 0.26802, 0.2715, 0.26714, 0.27108, 0.27032, 0.27288, 0.271, 0.27048, 0.26948, 0.27382, 0.27056, 0.27404, 0.27162, 0.27054, 0.27232, 0.27328, 0.27242, 0.26978, 0.2725, 0.27236, 0.27228, 0.27282, 0.27318, 0.2746, 0.27434, 0.27072, 0.27108, 0.2737, 0.273, 0.2735, 0.27328, 0.2745, 0.27388, 0.27212, 0.2744, 0.27352, 0.2723, 0.27264, 0.27482, 0.2727, 0.2734, 0.27328, 0.27136, 0.27322 ], "test_acc": [ 0.2855, 0.2551, 0.2775, 0.2743, 0.2562, 0.2696, 0.2672, 0.2538, 0.2753, 0.2372, 0.2459, 0.2719, 0.2666, 0.2665, 0.2673, 0.2632, 0.2686, 0.2624, 0.2753, 0.2718, 0.2735, 0.2621, 0.2539, 0.247, 0.2764, 0.2782, 0.2758, 0.2809, 0.2735, 0.2648, 0.2652, 0.2743, 0.2866, 0.2829, 0.2732, 0.2647, 0.282, 0.2671, 0.2715, 0.2861, 0.2807, 0.2831, 0.2899, 0.2743, 0.2842, 0.2796, 0.2721, 0.2816, 0.2874, 0.2873, 0.2878, 0.2746, 0.2911, 0.2938, 0.2945, 0.2906, 0.2921, 0.2921, 0.2867, 0.2754, 0.2951, 0.2874, 0.2816, 0.2894, 0.2686, 0.277, 0.2915, 0.295, 0.2958, 0.2885, 0.2861, 0.2907, 0.2913, 0.2959, 0.2897, 0.2872, 0.2932, 0.2882, 0.2933, 0.2942, 0.2939, 0.2946, 0.2902, 0.2888, 0.2957, 0.2944, 0.2917, 0.294, 0.2895, 0.2908, 0.2929, 0.2916, 0.2912, 0.2927, 0.2923, 0.2931, 0.2924, 0.2923, 0.2925, 0.2922 ] }, "diagnostics": { "bp_cosine": [ 0.3577328324317932, -0.002256808802485466 ], "perturbation_rho": [ 0.006564013194292784, 0.0245413389056921 ], "nudging": { "0.001": [ -3.3527612686157227e-07, 0.0 ], "0.003": [ -9.671784937381744e-07, 9.313225746154785e-10 ], "0.01": [ -3.1115487217903137e-06, 3.725290298461914e-09 ] }, "hidden_norms_per_layer": [ 61062.07421875, 1830187136.0, 2154652416.0 ], "bp_grad_norms_per_layer": [ 2.2415125044972228e-07, 2.5406721171350455e-10, 2.541881982676131e-10 ] }, "drift": { "embed.weight": 362.6998630270425, "embed.bias": 226.82991769893152, "blocks.0.ln.weight": 11.27242096475318, "blocks.0.w1.weight": 342.7047730991574, "blocks.0.w1.bias": 255.335818330469, "blocks.0.w2.weight": 571.0736810874553, "blocks.1.ln.weight": 6.709290843015378, "blocks.1.w1.weight": 234.88441196904554, "blocks.1.w1.bias": 208.6820413337303, "blocks.1.w2.weight": 259.6972453610996, "out_ln.weight": 0.47275142976863993, "out_head.weight": 5.14258998677984, "out_head.bias": 3.632529706269454 } }, "fa": { "log": { "train_loss": [ 2.062916795501709, 1.9544608277130127, 1.9338681225204468, 1.9168871939468384, 1.8961386334609986, 1.8842207279205323, 1.8664908053970337, 1.8596847631454467, 1.8500450038909912, 1.8418613220596314, 1.8393701876449584, 1.828501918258667, 1.8237099225234985, 1.8169598587799072, 1.8136948993682862, 1.8099042462539672, 1.8040387533569335, 1.7993673914337158, 1.801761696434021, 1.7964957051849366, 1.7922449420166016, 1.7918584114837646, 1.7928801208496095, 1.7959136298370362, 1.7932421494293214, 1.7945690915298462, 1.798325119857788, 1.7924271218490602, 1.795842142982483, 1.786767998275757, 1.7907304748153687, 1.7930365305328368, 1.7880332135009767, 1.7910264099884032, 1.786202894821167, 1.7834789249038696, 1.7853226938247682, 1.7844206839752197, 1.779783320388794, 1.777475042037964, 1.7763150534820558, 1.7789424829101563, 1.7754264488983154, 1.7715616805648804, 1.7733828824615478, 1.7673900942993164, 1.7679843439102172, 1.7700712796020508, 1.7681101821517944, 1.763485518951416, 1.7642558453369142, 1.7612942028427123, 1.7610017293548583, 1.7577664344024657, 1.7626592459869386, 1.7603577539443969, 1.754944856262207, 1.7521989233398438, 1.7558432748031616, 1.757179810218811, 1.7534725143051146, 1.7556944265365602, 1.7504646623992919, 1.7463257875823974, 1.7487027013397216, 1.7462505680084228, 1.7486760431289672, 1.7459170078277588, 1.740626848487854, 1.7471276846694945, 1.7445780047225952, 1.743365991783142, 1.7377847635650634, 1.7428057806396484, 1.7423382642364502, 1.739464909362793, 1.7394498482513427, 1.7388129167556763, 1.743233028869629, 1.7352628125, 1.740815905380249, 1.7376227493667602, 1.7356027941894532, 1.7353747326278686, 1.7357349237060546, 1.7323332436370849, 1.7330560940170288, 1.736462532081604, 1.7307056594085692, 1.7358304736709596, 1.7327582055282593, 1.7328367833709717, 1.730865809020996, 1.7343174131011962, 1.7336546128082275, 1.729335763282776, 1.731069825515747, 1.7309973764801025, 1.7332309117889404, 1.7309427404022217 ], "train_acc": [ 0.25124, 0.29078, 0.30174, 0.30854, 0.31558, 0.32094, 0.32596, 0.32986, 0.3375, 0.33924, 0.34048, 0.34462, 0.34806, 0.34818, 0.34936, 0.35036, 0.35452, 0.35762, 0.35606, 0.35678, 0.35724, 0.3556, 0.35836, 0.35558, 0.35872, 0.35734, 0.35428, 0.35594, 0.35682, 0.35978, 0.36006, 0.35966, 0.36212, 0.361, 0.3593, 0.36056, 0.36182, 0.36036, 0.3643, 0.36462, 0.36522, 0.3643, 0.3701, 0.37198, 0.36678, 0.36798, 0.37018, 0.3697, 0.3685, 0.37262, 0.36694, 0.3722, 0.37228, 0.37302, 0.37096, 0.37208, 0.3721, 0.37598, 0.37528, 0.37364, 0.37788, 0.37682, 0.37554, 0.37848, 0.37854, 0.3787, 0.37996, 0.37764, 0.3765, 0.37938, 0.37982, 0.38054, 0.383, 0.38022, 0.382, 0.3806, 0.38078, 0.38504, 0.38192, 0.38324, 0.38318, 0.38192, 0.38326, 0.38492, 0.38542, 0.38452, 0.3843, 0.38424, 0.38694, 0.38464, 0.38664, 0.3882, 0.38476, 0.3842, 0.38732, 0.38674, 0.38546, 0.38734, 0.38404, 0.38848 ], "test_acc": [ 0.2764, 0.3182, 0.3354, 0.3476, 0.3539, 0.339, 0.3478, 0.3737, 0.3595, 0.3642, 0.3624, 0.3739, 0.3717, 0.3809, 0.3776, 0.3763, 0.3824, 0.3807, 0.3754, 0.3832, 0.3864, 0.391, 0.3762, 0.3797, 0.3875, 0.3857, 0.3897, 0.3769, 0.3752, 0.3835, 0.3795, 0.3635, 0.3828, 0.3894, 0.3827, 0.3777, 0.3865, 0.3855, 0.3863, 0.3865, 0.3888, 0.3807, 0.3986, 0.3886, 0.389, 0.3891, 0.3782, 0.384, 0.3912, 0.3924, 0.3923, 0.3923, 0.3917, 0.3969, 0.3968, 0.3932, 0.3964, 0.3876, 0.3991, 0.3852, 0.4036, 0.3989, 0.3858, 0.3974, 0.3968, 0.4013, 0.403, 0.4009, 0.3967, 0.3979, 0.397, 0.3962, 0.3943, 0.4013, 0.4027, 0.3986, 0.4013, 0.3944, 0.4021, 0.3941, 0.3972, 0.3984, 0.4003, 0.3986, 0.3997, 0.3989, 0.3996, 0.4009, 0.4004, 0.3993, 0.4001, 0.3999, 0.4013, 0.4007, 0.4007, 0.4011, 0.401, 0.4007, 0.4011, 0.4012 ] }, "diagnostics": { "bp_cosine": [ 0.018671220168471336, 0.9215916395187378 ], "perturbation_rho": [ 0.05869613587856293, -0.006530101411044598 ], "nudging": { "0.001": [ -1.8319697119295597e-06, -3.306486178189516e-06 ], "0.003": [ -5.463924026116729e-06, -1.0309304343536496e-05 ], "0.01": [ -1.8201360944658518e-05, -3.473513061180711e-05 ] }, "hidden_norms_per_layer": [ 6274.154296875, 343870.53125, 375719.59375 ], "bp_grad_norms_per_layer": [ 2.1775536879431456e-05, 1.1024598052244983e-06, 8.687474064572598e-07 ] }, "drift": { "embed.weight": 47.037152107454325, "embed.bias": 16.388390598893867, "blocks.0.ln.weight": 1.5269469673782543, "blocks.0.w1.weight": 25.58456877637686, "blocks.0.w1.bias": 18.02406186908715, "blocks.0.w2.weight": 64.17096592056014, "blocks.1.ln.weight": 1.3407223784512263, "blocks.1.w1.weight": 21.242765670034206, "blocks.1.w1.bias": 17.953849785001406, "blocks.1.w2.weight": 23.00878057790312, "out_ln.weight": 0.3700761416548859, "out_head.weight": 5.133209234946238, "out_head.bias": 3.885876738755204 } } }, "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": [ 3 ], "gpu": 0, "output_dir": "results/fa_dfa_d512_L2_seed3", "methods": [ "fa", "dfa" ], "random_targets": false, "penalty_lam": 0.0, "num_classes": 10 } }