diff options
Diffstat (limited to 'results/fa_dfa_d512_L2_seed4/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d512_L2_seed4/results_cifar10.json | 749 |
1 files changed, 749 insertions, 0 deletions
diff --git a/results/fa_dfa_d512_L2_seed4/results_cifar10.json b/results/fa_dfa_d512_L2_seed4/results_cifar10.json new file mode 100644 index 0000000..9e35093 --- /dev/null +++ b/results/fa_dfa_d512_L2_seed4/results_cifar10.json @@ -0,0 +1,749 @@ +{ + "4": { + "dfa": { + "log": { + "train_loss": [ + 2.0538645137786866, + 2.0414622520446777, + 2.0403961332702636, + 2.0313826718139647, + 2.028870590591431, + 2.0276684378814696, + 2.025082506790161, + 2.027707338409424, + 2.024693886566162, + 2.0219140999603273, + 2.024347680053711, + 2.0226476731109617, + 2.0212312582397463, + 2.019639465713501, + 2.0200840225982666, + 2.022379661102295, + 2.0195740895843506, + 2.018866870689392, + 2.0154102825164797, + 2.019121079330444, + 2.0158904475402832, + 2.0182619565582276, + 2.0165906146240236, + 2.0169575183868407, + 2.0150413411712647, + 2.0142192585754395, + 2.0149837336730956, + 2.014163435897827, + 2.0160239248275755, + 2.012873958053589, + 2.014381752700806, + 2.0157296450805666, + 2.0139720638656615, + 2.0141250953674317, + 2.0129578493881226, + 2.0142643730926513, + 2.011523956451416, + 2.0130753877258303, + 2.0103190279388428, + 2.0131986405181883, + 2.011784356765747, + 2.0098198655700683, + 2.0134172485351565, + 2.011796088027954, + 2.009907284927368, + 2.0113371464538576, + 2.011871915245056, + 2.012780702667236, + 2.0106706674194337, + 2.0101452950286864, + 2.0108949478149416, + 2.0081811878967284, + 2.0113219532775877, + 2.0078719809722902, + 2.008186681213379, + 2.0087371883392335, + 2.010032041015625, + 2.0062248846054076, + 2.008138601531982, + 2.008908841209412, + 2.006502756500244, + 2.0063726428604127, + 2.006322613143921, + 2.0073862936401365, + 2.0092133827209473, + 2.0075592138671876, + 2.0070631226348876, + 2.0061116324615478, + 2.0080163690567017, + 2.0098242531585693, + 2.004972350997925, + 2.0045611180877687, + 2.0060609978485107, + 2.0061137674713136, + 2.0058565605163574, + 2.0072338876342775, + 2.0047287912750242, + 2.0041282120132444, + 2.007643541030884, + 2.0067062200546264, + 2.005947174911499, + 2.0044366609191893, + 2.0044241131591796, + 2.0034669429016114, + 2.0061908039855956, + 2.004900071258545, + 2.002522025909424, + 2.0041396823120117, + 2.0040297746276856, + 2.0036116275787355, + 2.003599824256897, + 2.003862293243408, + 2.002604514427185, + 2.003420566329956, + 2.0044958754730224, + 2.0024802682876586, + 2.001940184288025, + 2.004589928436279, + 2.00544753616333, + 2.004831633987427 + ], + "train_acc": [ + 0.24328, + 0.2486, + 0.2526, + 0.25192, + 0.25386, + 0.25942, + 0.2579, + 0.25428, + 0.2558, + 0.2587, + 0.25714, + 0.25432, + 0.2597, + 0.26176, + 0.25932, + 0.25756, + 0.26362, + 0.26154, + 0.26296, + 0.26458, + 0.2631, + 0.26162, + 0.2629, + 0.26272, + 0.26022, + 0.26332, + 0.26328, + 0.26642, + 0.26188, + 0.26562, + 0.26456, + 0.26682, + 0.2653, + 0.2635, + 0.26696, + 0.26618, + 0.26604, + 0.26422, + 0.26612, + 0.26754, + 0.26958, + 0.2674, + 0.26656, + 0.2681, + 0.2651, + 0.2682, + 0.266, + 0.26656, + 0.26712, + 0.26854, + 0.26726, + 0.26844, + 0.26866, + 0.26848, + 0.2684, + 0.26968, + 0.2694, + 0.27024, + 0.26914, + 0.2689, + 0.27068, + 0.26872, + 0.2698, + 0.2698, + 0.27038, + 0.27008, + 0.26956, + 0.26956, + 0.26738, + 0.2676, + 0.27036, + 0.27014, + 0.26994, + 0.27158, + 0.27192, + 0.27026, + 0.27058, + 0.27212, + 0.26966, + 0.26896, + 0.27094, + 0.27132, + 0.27096, + 0.27272, + 0.2715, + 0.2717, + 0.27104, + 0.27176, + 0.26956, + 0.26908, + 0.27146, + 0.2707, + 0.27088, + 0.26976, + 0.2708, + 0.27234, + 0.27182, + 0.27142, + 0.27088, + 0.27292 + ], + "test_acc": [ + 0.2691, + 0.2657, + 0.2489, + 0.2683, + 0.2735, + 0.2856, + 0.2869, + 0.2671, + 0.2775, + 0.2707, + 0.2668, + 0.2677, + 0.2711, + 0.284, + 0.28, + 0.2978, + 0.2649, + 0.2558, + 0.2813, + 0.2732, + 0.2875, + 0.2844, + 0.2664, + 0.2731, + 0.2948, + 0.2757, + 0.2818, + 0.2811, + 0.2842, + 0.2762, + 0.2852, + 0.2639, + 0.2834, + 0.2859, + 0.2804, + 0.2727, + 0.2794, + 0.2916, + 0.2746, + 0.2768, + 0.2903, + 0.2722, + 0.2896, + 0.2856, + 0.2906, + 0.2848, + 0.2776, + 0.29, + 0.2918, + 0.2703, + 0.2847, + 0.2838, + 0.2816, + 0.2894, + 0.2815, + 0.2783, + 0.2917, + 0.2712, + 0.285, + 0.2861, + 0.2844, + 0.2898, + 0.2839, + 0.2886, + 0.2809, + 0.2826, + 0.2792, + 0.2864, + 0.2978, + 0.2876, + 0.2855, + 0.2997, + 0.2912, + 0.2887, + 0.2867, + 0.2842, + 0.284, + 0.2781, + 0.2817, + 0.2911, + 0.2842, + 0.2834, + 0.2867, + 0.2861, + 0.2867, + 0.284, + 0.2887, + 0.2864, + 0.2875, + 0.2899, + 0.2864, + 0.287, + 0.2901, + 0.2856, + 0.2866, + 0.2863, + 0.2859, + 0.2861, + 0.2861, + 0.2861 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.3827235698699951, + -0.000831119017675519 + ], + "perturbation_rho": [ + -0.01562047004699707, + 0.0 + ], + "nudging": { + "0.001": [ + -2.812594175338745e-07, + 0.0 + ], + "0.003": [ + -9.383074939250946e-07, + 0.0 + ], + "0.01": [ + -3.2153911888599396e-06, + 0.0 + ] + }, + "hidden_norms_per_layer": [ + 55950.5390625, + 1669607168.0, + 3780465152.0 + ], + "bp_grad_norms_per_layer": [ + 2.2658105081063695e-07, + 2.3708951468748296e-10, + 2.3738841448128767e-10 + ] + }, + "drift": { + "embed.weight": 337.5136698692365, + "embed.bias": 273.5401823379648, + "blocks.0.ln.weight": 9.322317945209601, + "blocks.0.w1.weight": 322.0064838288751, + "blocks.0.w1.bias": 292.5825702677452, + "blocks.0.w2.weight": 505.89197780725414, + "blocks.1.ln.weight": 9.09606961243359, + "blocks.1.w1.weight": 353.39130820313096, + "blocks.1.w1.bias": 346.5478039461049, + "blocks.1.w2.weight": 349.3876744406949, + "out_ln.weight": 0.435883262668521, + "out_head.weight": 7.086345015239166, + "out_head.bias": 3.9170311167015974 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.083539034461975, + 1.9695448753356934, + 1.9397349642562867, + 1.9066173551177978, + 1.886961284866333, + 1.8710758205795288, + 1.8628956564331054, + 1.8567623680877685, + 1.8501425790405273, + 1.850063798828125, + 1.8536383200073243, + 1.851845763206482, + 1.8578751723480225, + 1.8505859604263306, + 1.8504567489242554, + 1.8505229975128175, + 1.8486209188079834, + 1.8476799687957763, + 1.8461452635192872, + 1.8506561833953858, + 1.8442412594985962, + 1.8458462058258056, + 1.842808702659607, + 1.8387870932769776, + 1.8379171523666382, + 1.828637089920044, + 1.824536803894043, + 1.822955154800415, + 1.8200513568878174, + 1.814583374671936, + 1.8130536968231201, + 1.8116273685073851, + 1.810110004234314, + 1.808945005569458, + 1.8023032517242432, + 1.8105073015975952, + 1.81240649143219, + 1.8092195032501222, + 1.8054952936172486, + 1.8071781116485595, + 1.8033202910614015, + 1.8002542707061768, + 1.7996792233276366, + 1.794723889541626, + 1.788533415184021, + 1.7873220053100587, + 1.7867383001327515, + 1.7845936785888672, + 1.7811818826675414, + 1.7822032886505126, + 1.7786470770263672, + 1.7798991229629517, + 1.7747682898712158, + 1.771372130050659, + 1.7732186395645142, + 1.767907095336914, + 1.7648547525405884, + 1.7638111727523804, + 1.7647753219223024, + 1.7633140285491944, + 1.758088468322754, + 1.7619236371612548, + 1.7579612891006469, + 1.7559792065811157, + 1.7589319384002686, + 1.7530268384170533, + 1.7540921353530883, + 1.7525135382080077, + 1.7494160869979858, + 1.7476271326446533, + 1.7485824377059938, + 1.7481532715225219, + 1.7493890603256226, + 1.753240061607361, + 1.7450246924209594, + 1.7479931283950805, + 1.7435640298461914, + 1.7442714694595336, + 1.7464028844833375, + 1.7440459408950806, + 1.7451819785308837, + 1.7424172634124755, + 1.7414722381591796, + 1.7440056212997437, + 1.7441547555160521, + 1.741508984413147, + 1.7379715420913697, + 1.743159959335327, + 1.737216314048767, + 1.73957718044281, + 1.7399621460342407, + 1.7403172652435304, + 1.7397853451919556, + 1.7367944290161133, + 1.7394116397476196, + 1.7405242685317994, + 1.7411947003173829, + 1.7401781003189087, + 1.7426191668319702, + 1.7414960692977905 + ], + "train_acc": [ + 0.24206, + 0.28608, + 0.30284, + 0.31378, + 0.32214, + 0.33222, + 0.33458, + 0.33384, + 0.339, + 0.3365, + 0.33642, + 0.33268, + 0.3321, + 0.3336, + 0.33528, + 0.33412, + 0.3342, + 0.33516, + 0.3349, + 0.3369, + 0.33338, + 0.33768, + 0.3359, + 0.3389, + 0.33848, + 0.33996, + 0.34268, + 0.34516, + 0.34514, + 0.34766, + 0.3483, + 0.35214, + 0.34858, + 0.34992, + 0.35344, + 0.35134, + 0.34842, + 0.35144, + 0.35262, + 0.35272, + 0.35618, + 0.35276, + 0.35816, + 0.3578, + 0.35874, + 0.3609, + 0.36208, + 0.3595, + 0.3629, + 0.36296, + 0.36438, + 0.36234, + 0.3654, + 0.36716, + 0.36458, + 0.36772, + 0.36752, + 0.36904, + 0.3677, + 0.3659, + 0.3724, + 0.36748, + 0.37008, + 0.37166, + 0.37228, + 0.37126, + 0.37246, + 0.3711, + 0.37376, + 0.3725, + 0.37384, + 0.37248, + 0.37364, + 0.37298, + 0.37408, + 0.37304, + 0.37604, + 0.37672, + 0.3764, + 0.37472, + 0.37532, + 0.3761, + 0.37432, + 0.37696, + 0.3772, + 0.37738, + 0.37684, + 0.37452, + 0.37578, + 0.37302, + 0.37482, + 0.3745, + 0.37608, + 0.37402, + 0.37522, + 0.37624, + 0.377, + 0.37676, + 0.37804, + 0.3765 + ], + "test_acc": [ + 0.2873, + 0.2968, + 0.3102, + 0.3278, + 0.3267, + 0.3394, + 0.3594, + 0.3508, + 0.348, + 0.3577, + 0.3237, + 0.3288, + 0.3256, + 0.3468, + 0.3314, + 0.3353, + 0.3342, + 0.3343, + 0.3363, + 0.3386, + 0.3359, + 0.3401, + 0.3353, + 0.3296, + 0.3361, + 0.3547, + 0.3367, + 0.342, + 0.3335, + 0.3402, + 0.3509, + 0.3259, + 0.3411, + 0.342, + 0.3459, + 0.3338, + 0.3447, + 0.3462, + 0.3413, + 0.3409, + 0.3561, + 0.3417, + 0.3429, + 0.3539, + 0.3554, + 0.3409, + 0.3557, + 0.3535, + 0.3559, + 0.3474, + 0.3477, + 0.3647, + 0.3483, + 0.337, + 0.3458, + 0.3461, + 0.3488, + 0.3399, + 0.351, + 0.3393, + 0.3521, + 0.3505, + 0.354, + 0.3489, + 0.3421, + 0.3435, + 0.3427, + 0.345, + 0.3509, + 0.3393, + 0.3513, + 0.3592, + 0.3435, + 0.3491, + 0.3462, + 0.3486, + 0.3435, + 0.3421, + 0.3463, + 0.3537, + 0.3507, + 0.3476, + 0.3523, + 0.346, + 0.3548, + 0.3469, + 0.3489, + 0.3457, + 0.3483, + 0.3495, + 0.3507, + 0.3498, + 0.3509, + 0.3498, + 0.3485, + 0.3503, + 0.3507, + 0.3502, + 0.3501, + 0.3501 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.03285929560661316, + 0.9476211071014404 + ], + "perturbation_rho": [ + 0.010396776720881462, + 0.21192286908626556 + ], + "nudging": { + "0.001": [ + -4.190136678516865e-06, + -7.935799658298492e-06 + ], + "0.003": [ + -1.2663658708333969e-05, + -2.3964676074683666e-05 + ], + "0.01": [ + -4.219170659780502e-05, + -8.006719872355461e-05 + ] + }, + "hidden_norms_per_layer": [ + 3884.70703125, + 284714.21875, + 219214.828125 + ], + "bp_grad_norms_per_layer": [ + 1.9074828742304817e-05, + 6.634672331529146e-07, + 6.522877811221406e-07 + ] + }, + "drift": { + "embed.weight": 27.016438956012585, + "embed.bias": 16.969540152339597, + "blocks.0.ln.weight": 1.5836171150936886, + "blocks.0.w1.weight": 25.706341628627694, + "blocks.0.w1.bias": 21.000244171066868, + "blocks.0.w2.weight": 64.31974159860198, + "blocks.1.ln.weight": 0.9270002241369442, + "blocks.1.w1.weight": 14.313388328940205, + "blocks.1.w1.bias": 6.9567229997624445, + "blocks.1.w2.weight": 31.78188983358227, + "out_ln.weight": 0.46949971605667673, + "out_head.weight": 3.9796491257565023, + "out_head.bias": 12.763196315235756 + } + } + }, + "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": [ + 4 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d512_L2_seed4", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |