diff options
Diffstat (limited to 'results/fa_dfa_d512_L2_seed8/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L2_seed8/results_cifar10.json | 749 |
1 files changed, 749 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L2_seed8/results_cifar10.json b/results/fa_dfa_d512_L2_seed8/results_cifar10.json new file mode 100644 index 0000000..ad4f8bf --- /dev/null +++ b/results/fa_dfa_d512_L2_seed8/results_cifar10.json @@ -0,0 +1,749 @@ +{ + "8": { + "dfa": { + "log": { + "train_loss": [ + 2.0461414054870604, + 2.028953609237671, + 2.0224830960845948, + 2.0189640741348267, + 2.0186043047332762, + 2.011939255371094, + 2.00884051902771, + 2.006375752105713, + 2.0061196725463866, + 2.0044810569763185, + 2.000473762359619, + 1.999425143661499, + 1.9991976582336426, + 1.9983828188323975, + 1.9999516902160646, + 2.000131714401245, + 2.000538674583435, + 1.999087513999939, + 1.9965772060394287, + 1.9952695233154296, + 1.9958783276367187, + 1.9935351610565186, + 1.9946063667297362, + 1.9927719008636475, + 1.9930043865203857, + 1.992498685836792, + 1.9923327257537842, + 1.9919360898208618, + 1.9922407612609863, + 1.992235760192871, + 1.992523240623474, + 1.9913750788116455, + 1.9934618493652343, + 1.9923072975921632, + 1.991889149017334, + 1.9904881185913086, + 1.9916703072357178, + 1.9886171953582763, + 1.9906551904296874, + 1.9912475842285156, + 1.9906490882873535, + 1.9897464477920532, + 1.9902090267562866, + 1.98889538230896, + 1.9899030431365967, + 1.9914473651885987, + 1.987290951499939, + 1.9882332851409912, + 1.987876655883789, + 1.986627554321289, + 1.9874235430526734, + 1.985631577758789, + 1.9859224739074708, + 1.984819174156189, + 1.9877312922668458, + 1.9880256802749634, + 1.9883026748657227, + 1.9880133050918578, + 1.9872900528717041, + 1.9861614239501952, + 1.9879061701202392, + 1.986334129486084, + 1.9889874462509156, + 1.9848816363143922, + 1.985108938598633, + 1.9852813848114013, + 1.9847304062652589, + 1.9842171102142334, + 1.986599141769409, + 1.9858817081451416, + 1.9852649099349975, + 1.982235690460205, + 1.9838326638793946, + 1.9827596533966065, + 1.9836663906097411, + 1.9844815605163575, + 1.9842618350982666, + 1.984488549156189, + 1.9830066680908203, + 1.9835596384048462, + 1.9822921755981446, + 1.9845924890899658, + 1.9827096743392945, + 1.9821208544158935, + 1.983244455909729, + 1.9823134091567993, + 1.9837292085266114, + 1.9827219149017334, + 1.9828275063323975, + 1.9811715400695802, + 1.9821106326675415, + 1.9830516720199585, + 1.9831754531097412, + 1.9825017529296876, + 1.9844729486465453, + 1.9823845765686035, + 1.9825008573913574, + 1.9817371285247802, + 1.9809684440612794, + 1.981824859008789 + ], + "train_acc": [ + 0.24566, + 0.25216, + 0.25672, + 0.2583, + 0.25908, + 0.2611, + 0.26292, + 0.26504, + 0.26436, + 0.26372, + 0.26838, + 0.2684, + 0.26774, + 0.26886, + 0.26864, + 0.26928, + 0.26802, + 0.26984, + 0.27068, + 0.27056, + 0.27002, + 0.27018, + 0.27034, + 0.27138, + 0.27238, + 0.27266, + 0.27208, + 0.27464, + 0.27406, + 0.27306, + 0.27278, + 0.27354, + 0.27164, + 0.27506, + 0.27494, + 0.27682, + 0.27244, + 0.2749, + 0.27302, + 0.2745, + 0.2736, + 0.27382, + 0.27478, + 0.27344, + 0.27582, + 0.2732, + 0.27662, + 0.27616, + 0.2763, + 0.27742, + 0.27642, + 0.27984, + 0.2777, + 0.28112, + 0.27754, + 0.27344, + 0.2769, + 0.27664, + 0.27734, + 0.27754, + 0.27784, + 0.27642, + 0.27598, + 0.2796, + 0.27964, + 0.27868, + 0.27864, + 0.27868, + 0.27826, + 0.27986, + 0.27616, + 0.2817, + 0.28024, + 0.27688, + 0.28002, + 0.27942, + 0.27838, + 0.2787, + 0.27856, + 0.28034, + 0.27972, + 0.27722, + 0.27992, + 0.28172, + 0.27796, + 0.2775, + 0.28122, + 0.2803, + 0.27958, + 0.27936, + 0.28062, + 0.2791, + 0.2797, + 0.28202, + 0.27898, + 0.27894, + 0.27866, + 0.27832, + 0.28, + 0.28114 + ], + "test_acc": [ + 0.2641, + 0.2447, + 0.2423, + 0.2586, + 0.2666, + 0.2868, + 0.2779, + 0.27, + 0.2886, + 0.2704, + 0.2846, + 0.2825, + 0.2801, + 0.2912, + 0.277, + 0.2834, + 0.2838, + 0.2802, + 0.2793, + 0.2883, + 0.2871, + 0.2825, + 0.288, + 0.2913, + 0.2813, + 0.2931, + 0.2994, + 0.3001, + 0.2753, + 0.2966, + 0.2996, + 0.2965, + 0.2835, + 0.3002, + 0.2979, + 0.2783, + 0.3017, + 0.2848, + 0.2846, + 0.2928, + 0.2956, + 0.2705, + 0.2811, + 0.292, + 0.2763, + 0.2834, + 0.2837, + 0.2949, + 0.2998, + 0.2894, + 0.2853, + 0.2921, + 0.2986, + 0.2918, + 0.2933, + 0.28, + 0.2988, + 0.2925, + 0.2828, + 0.2875, + 0.2985, + 0.2927, + 0.2922, + 0.3016, + 0.2947, + 0.2952, + 0.2936, + 0.2923, + 0.2992, + 0.2968, + 0.2915, + 0.2992, + 0.2941, + 0.2969, + 0.2936, + 0.2972, + 0.2964, + 0.2928, + 0.2958, + 0.2973, + 0.2958, + 0.2971, + 0.2983, + 0.2983, + 0.2951, + 0.2963, + 0.2953, + 0.2964, + 0.2963, + 0.293, + 0.2951, + 0.2967, + 0.2967, + 0.297, + 0.2966, + 0.2967, + 0.2967, + 0.2966, + 0.2966, + 0.2967 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3839848041534424, + -0.0006596383173018694 + ], + "perturbation_rho": [ + 0.02412901259958744, + 0.0 + ], + "nudging": { + "0.001": [ + -4.284083843231201e-07, + 0.0 + ], + "0.003": [ + -1.3029202818870544e-06, + 0.0 + ], + "0.01": [ + -4.258938133716583e-06, + 0.0 + ] + }, + "hidden_norms_per_layer": [ + 53606.07421875, + 782741952.0, + 4561426432.0 + ], + "bp_grad_norms_per_layer": [ + 3.0959373020778003e-07, + 3.211692578553027e-10, + 3.2110258896267396e-10 + ] + }, + "drift": { + "embed.weight": 322.14716880304843, + "embed.bias": 253.7663922411994, + "blocks.0.ln.weight": 9.639264948834146, + "blocks.0.w1.weight": 278.73838749620353, + "blocks.0.w1.bias": 250.4325839565606, + "blocks.0.w2.weight": 488.7145731499842, + "blocks.1.ln.weight": 9.351844341907716, + "blocks.1.w1.weight": 376.1230199590697, + "blocks.1.w1.bias": 376.4920228948829, + "blocks.1.w2.weight": 403.36194320458816, + "out_ln.weight": 0.513534701987637, + "out_head.weight": 7.742896807554914, + "out_head.bias": 1.1408249446090957 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.0653204177856446, + 1.9681754495239259, + 1.9207439241790771, + 1.8943786935806275, + 1.8802625274276734, + 1.8628883916473389, + 1.8490794301605225, + 1.8400364615249634, + 1.8355530837249756, + 1.8266534114837647, + 1.8170614450454712, + 1.8170341843414306, + 1.809728816757202, + 1.80435710231781, + 1.8044258756256104, + 1.7987670265960694, + 1.7999644842147826, + 1.7953315704345703, + 1.7825802756118774, + 1.7845099126434327, + 1.7779603283691405, + 1.7730702184677125, + 1.7695367748260498, + 1.7647147689056397, + 1.7608037928009033, + 1.7535330474090576, + 1.75540680393219, + 1.7487119026947022, + 1.7451758492279053, + 1.745663233718872, + 1.7409377331924438, + 1.7372819427871704, + 1.7388353146743774, + 1.7410107398605346, + 1.7357459603881835, + 1.7290400104522705, + 1.7318879638290405, + 1.729545888900757, + 1.7291300962066651, + 1.733487999343872, + 1.7302352466583253, + 1.7331863732147217, + 1.7339300806045532, + 1.7359314554595948, + 1.7347373684310914, + 1.7375746230316162, + 1.7313235457611085, + 1.7353338851165772, + 1.73820286403656, + 1.7340232833862306, + 1.7341221398925781, + 1.7339955667114257, + 1.7339833393096924, + 1.7300079647064208, + 1.7303287143707275, + 1.7298947993469238, + 1.7274299069213868, + 1.7319763919067384, + 1.7264507590103149, + 1.7299312719726563, + 1.7262817291641235, + 1.7270497384262085, + 1.7246126723098756, + 1.7255360033798217, + 1.7196656997680664, + 1.7214281465911865, + 1.7198693132781981, + 1.7217327685546875, + 1.7179188094329834, + 1.7154542624664306, + 1.7195643279266357, + 1.7149108078384399, + 1.7153279098510743, + 1.7153327802276612, + 1.7094103066253663, + 1.7162176602554322, + 1.7122739435195924, + 1.7128758419418335, + 1.708295757408142, + 1.7086283514785767, + 1.7106998838043213, + 1.711903858909607, + 1.7090504042434693, + 1.7111234002304077, + 1.7090101749420166, + 1.7079448748779298, + 1.707335115890503, + 1.7080174974822997, + 1.705443286972046, + 1.7056840801239013, + 1.708050958175659, + 1.7072754711151124, + 1.7059114752960205, + 1.7063753726959228, + 1.7061457764434815, + 1.7009761544418336, + 1.707751445274353, + 1.7066948248291016, + 1.7041513320922852, + 1.7030947677612305 + ], + "train_acc": [ + 0.24824, + 0.29, + 0.30804, + 0.32064, + 0.32604, + 0.33114, + 0.33818, + 0.3373, + 0.34064, + 0.3462, + 0.34982, + 0.34634, + 0.35334, + 0.35346, + 0.35488, + 0.3554, + 0.35676, + 0.35738, + 0.36206, + 0.36144, + 0.36364, + 0.3671, + 0.36916, + 0.3698, + 0.37032, + 0.3709, + 0.37188, + 0.3757, + 0.37546, + 0.37376, + 0.37888, + 0.37908, + 0.37828, + 0.37822, + 0.37984, + 0.3791, + 0.38144, + 0.38286, + 0.37966, + 0.3797, + 0.38152, + 0.38084, + 0.3812, + 0.3809, + 0.38156, + 0.382, + 0.38494, + 0.38258, + 0.38066, + 0.3824, + 0.38058, + 0.38342, + 0.3823, + 0.38198, + 0.38534, + 0.38368, + 0.38524, + 0.38594, + 0.38536, + 0.38514, + 0.38676, + 0.38838, + 0.38778, + 0.38664, + 0.39118, + 0.38748, + 0.39154, + 0.3882, + 0.38852, + 0.39078, + 0.38918, + 0.39202, + 0.39086, + 0.39128, + 0.39242, + 0.3935, + 0.39422, + 0.39322, + 0.39388, + 0.39516, + 0.39496, + 0.39414, + 0.39428, + 0.39522, + 0.39426, + 0.3953, + 0.39628, + 0.39662, + 0.39662, + 0.39608, + 0.39646, + 0.39568, + 0.39606, + 0.39558, + 0.3961, + 0.3992, + 0.39438, + 0.39786, + 0.39626, + 0.39922 + ], + "test_acc": [ + 0.297, + 0.3142, + 0.3407, + 0.3409, + 0.3537, + 0.3628, + 0.3482, + 0.3626, + 0.3557, + 0.351, + 0.3614, + 0.3693, + 0.363, + 0.3681, + 0.374, + 0.3756, + 0.3778, + 0.3795, + 0.3757, + 0.3742, + 0.3801, + 0.3732, + 0.3782, + 0.3805, + 0.3718, + 0.3786, + 0.3914, + 0.3924, + 0.3797, + 0.3967, + 0.3891, + 0.3829, + 0.3811, + 0.3967, + 0.3876, + 0.3862, + 0.3945, + 0.3851, + 0.3807, + 0.3871, + 0.3924, + 0.3941, + 0.3823, + 0.392, + 0.3942, + 0.3908, + 0.3979, + 0.3954, + 0.4003, + 0.401, + 0.3938, + 0.3938, + 0.3997, + 0.3943, + 0.4069, + 0.4012, + 0.4056, + 0.408, + 0.4037, + 0.3996, + 0.4115, + 0.4098, + 0.4062, + 0.4083, + 0.4115, + 0.4094, + 0.4147, + 0.4109, + 0.4101, + 0.4084, + 0.4112, + 0.4114, + 0.4127, + 0.4113, + 0.407, + 0.4126, + 0.4167, + 0.4132, + 0.4156, + 0.4138, + 0.4122, + 0.4128, + 0.417, + 0.4138, + 0.4128, + 0.4136, + 0.4137, + 0.4132, + 0.4153, + 0.4136, + 0.4123, + 0.413, + 0.4134, + 0.4142, + 0.4155, + 0.4155, + 0.4142, + 0.4147, + 0.4149, + 0.4146 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.02496020682156086, + 0.9624444246292114 + ], + "perturbation_rho": [ + -0.018970193341374397, + 0.1267668902873993 + ], + "nudging": { + "0.001": [ + -2.1758460206910968e-06, + -7.471287972293794e-06 + ], + "0.003": [ + -6.601490895263851e-06, + -2.2840846213512123e-05 + ], + "0.01": [ + -2.1907704649493098e-05, + -7.614441710757092e-05 + ] + }, + "hidden_norms_per_layer": [ + 6420.08984375, + 205601.1875, + 92651.546875 + ], + "bp_grad_norms_per_layer": [ + 2.9750526664429344e-05, + 2.456108632031828e-06, + 2.075838210657821e-06 + ] + }, + "drift": { + "embed.weight": 37.110060620728056, + "embed.bias": 24.00213600327038, + "blocks.0.ln.weight": 1.3952106680756078, + "blocks.0.w1.weight": 19.29938322080228, + "blocks.0.w1.bias": 17.715456046599957, + "blocks.0.w2.weight": 54.87838247174745, + "blocks.1.ln.weight": 1.0669201771550085, + "blocks.1.w1.weight": 18.224214324200183, + "blocks.1.w1.bias": 17.474148202150594, + "blocks.1.w2.weight": 31.754151778532563, + "out_ln.weight": 0.42931707858608786, + "out_head.weight": 4.036471355678216, + "out_head.bias": 7.088871139191845 + } + } + }, + "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": [ + 8 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L2_seed8", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |