diff options
Diffstat (limited to 'results/fa_dfa_d256_L8_seed1/results_cifar10.json')
| -rw-r--r-- | results/fa_dfa_d256_L8_seed1/results_cifar10.json | 881 |
1 files changed, 881 insertions, 0 deletions
diff --git a/results/fa_dfa_d256_L8_seed1/results_cifar10.json b/results/fa_dfa_d256_L8_seed1/results_cifar10.json new file mode 100644 index 0000000..cd78ab9 --- /dev/null +++ b/results/fa_dfa_d256_L8_seed1/results_cifar10.json @@ -0,0 +1,881 @@ +{ + "1": { + "dfa": { + "log": { + "train_loss": [ + 2.073248834991455, + 2.0561216505432127, + 2.0487407315063475, + 2.0397916249084473, + 2.0339059049224852, + 2.031160478439331, + 2.0272507332992555, + 2.0234306407928466, + 2.0243458992767334, + 2.0212309561920168, + 2.0213267765045164, + 2.0173781090927125, + 2.015773401260376, + 2.016544706039429, + 2.0143174066925047, + 2.015236302452087, + 2.016163635253906, + 2.0134117998504637, + 2.0144200479125978, + 2.013800316390991, + 2.0096480722045897, + 2.0113335668945314, + 2.0122015085601808, + 2.0122210752868654, + 2.0115504538345337, + 2.007763221130371, + 2.009548883666992, + 2.0078995152282717, + 2.0087636469268797, + 2.008320099105835, + 2.008820813369751, + 2.006456883468628, + 2.0074588145446777, + 2.008500497131348, + 2.0084443688964844, + 2.0064106777954103, + 2.0071660092544557, + 2.0042392598724366, + 2.0080142588043213, + 2.0092139667510986, + 2.0066707161712647, + 2.0068252431869507, + 2.0088330657958986, + 2.00828530166626, + 2.0062371633148195, + 2.0070145973205564, + 2.006389614715576, + 2.008107543182373, + 2.0061909777069094, + 2.0069403369522094, + 2.0080043979644775, + 2.0067166803359986, + 2.0072613145446776, + 2.0074930825042725, + 2.0079901724243165, + 2.00686990776062, + 2.007503282737732, + 2.0056561468124388, + 2.0081290016937254, + 2.0048272270965577, + 2.006778687286377, + 2.0057738173675537, + 2.004890469207764, + 2.0073782594299314, + 2.0066935089874267, + 2.005761955795288, + 2.006959405899048, + 2.006399262161255, + 2.006497128829956, + 2.0050615126800535, + 2.006692824935913, + 2.0056320917510986, + 2.0048789993286134, + 2.005718688735962, + 2.0067832677459716, + 2.004570386314392, + 2.004494798736572, + 2.0044968630599977, + 2.0067117002105714, + 2.0063977341461183, + 2.0049595592880247, + 2.0041789445495604, + 2.0047897937774657, + 2.0072312893676756, + 2.004594275741577, + 2.004407970199585, + 2.0041916183471677, + 2.002536773300171, + 2.0047487148284913, + 2.0024493196105957, + 2.0030294651794436, + 2.0037170224761964, + 2.0036521311187743, + 2.002812971572876, + 2.00589838760376, + 2.0043148944854736, + 2.005489154701233, + 2.0043797174072266, + 2.0031765459442137, + 2.0041999047851564 + ], + "train_acc": [ + 0.23724, + 0.24082, + 0.24538, + 0.25098, + 0.25208, + 0.25444, + 0.25704, + 0.25828, + 0.25656, + 0.25916, + 0.26014, + 0.26036, + 0.26374, + 0.25984, + 0.26378, + 0.26424, + 0.26426, + 0.26496, + 0.26426, + 0.2631, + 0.2658, + 0.26536, + 0.2682, + 0.2644, + 0.26904, + 0.27, + 0.27042, + 0.26654, + 0.27128, + 0.2687, + 0.26822, + 0.27002, + 0.26908, + 0.27106, + 0.2686, + 0.27238, + 0.27072, + 0.27068, + 0.26996, + 0.26948, + 0.27084, + 0.2695, + 0.2713, + 0.27054, + 0.27256, + 0.26982, + 0.26974, + 0.27168, + 0.27108, + 0.27182, + 0.27238, + 0.27258, + 0.27288, + 0.27236, + 0.27066, + 0.27214, + 0.27302, + 0.27214, + 0.27026, + 0.2738, + 0.27384, + 0.2733, + 0.27542, + 0.27348, + 0.27172, + 0.27406, + 0.27318, + 0.2763, + 0.27302, + 0.27414, + 0.27384, + 0.27542, + 0.27476, + 0.27312, + 0.27358, + 0.275, + 0.27514, + 0.27526, + 0.27434, + 0.2745, + 0.27498, + 0.2763, + 0.27358, + 0.27466, + 0.27588, + 0.27678, + 0.27598, + 0.2766, + 0.27502, + 0.27544, + 0.27712, + 0.27632, + 0.27598, + 0.2788, + 0.27462, + 0.27336, + 0.27738, + 0.27462, + 0.27666, + 0.27568 + ], + "test_acc": [ + 0.2766, + 0.2536, + 0.2625, + 0.2542, + 0.2703, + 0.271, + 0.2674, + 0.2674, + 0.2539, + 0.2792, + 0.2744, + 0.2832, + 0.2755, + 0.2799, + 0.2729, + 0.2704, + 0.2736, + 0.2734, + 0.2651, + 0.2756, + 0.2709, + 0.2847, + 0.2691, + 0.2717, + 0.278, + 0.2623, + 0.2801, + 0.2833, + 0.2587, + 0.272, + 0.2868, + 0.2876, + 0.2743, + 0.2912, + 0.2824, + 0.2835, + 0.2831, + 0.2874, + 0.2945, + 0.2793, + 0.2786, + 0.2921, + 0.2922, + 0.2856, + 0.2997, + 0.28, + 0.2756, + 0.2834, + 0.29, + 0.2877, + 0.2853, + 0.279, + 0.2947, + 0.2922, + 0.284, + 0.2894, + 0.2943, + 0.2887, + 0.2787, + 0.2891, + 0.29, + 0.2823, + 0.2815, + 0.2846, + 0.2876, + 0.2899, + 0.2918, + 0.2927, + 0.2944, + 0.2979, + 0.2868, + 0.286, + 0.2894, + 0.293, + 0.2877, + 0.2914, + 0.2892, + 0.2938, + 0.2884, + 0.2871, + 0.2869, + 0.2841, + 0.2897, + 0.2911, + 0.2882, + 0.2856, + 0.2901, + 0.2928, + 0.2905, + 0.2922, + 0.2897, + 0.2898, + 0.2902, + 0.2899, + 0.291, + 0.2906, + 0.2903, + 0.2901, + 0.2899, + 0.29 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.39772462844848633, + -0.002370076021179557, + 0.0021920399740338326, + -0.0014136324170976877, + 0.001076650689356029, + 0.001179476035758853, + 0.0019159603398293257, + -0.00038467111880891025 + ], + "perturbation_rho": [ + -0.014354787766933441, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -3.3620744943618774e-07, + 0.0, + 2.7939677238464355e-09, + 1.862645149230957e-09, + 0.0, + 0.0, + -1.862645149230957e-09, + 0.0 + ], + "0.003": [ + -1.0924413800239563e-06, + 9.313225746154785e-10, + 2.7939677238464355e-09, + 1.862645149230957e-09, + 0.0, + 0.0, + -3.725290298461914e-09, + 0.0 + ], + "0.01": [ + -3.4980475902557373e-06, + -6.51925802230835e-09, + 1.862645149230957e-09, + 1.862645149230957e-09, + 0.0, + 9.313225746154785e-10, + -3.725290298461914e-09, + 1.862645149230957e-09 + ] + }, + "hidden_norms_per_layer": [ + 45727.35546875, + 263312528.0, + 537438144.0, + 664858176.0, + 845261312.0, + 1070211328.0, + 1238112640.0, + 1792965120.0, + 1877736448.0 + ], + "bp_grad_norms_per_layer": [ + 3.161038364396518e-07, + 7.485224107206534e-10, + 7.498183740572983e-10, + 7.446125938059822e-10, + 7.443420324548811e-10, + 7.443580751775869e-10, + 7.442703675586415e-10, + 7.441662841500829e-10, + 7.441969818167138e-10 + ] + }, + "drift": { + "embed.weight": 357.86461919115555, + "embed.bias": 312.95737180504705, + "blocks.0.ln.weight": 10.188159942626953, + "blocks.0.w1.weight": 211.37068163903083, + "blocks.0.w1.bias": 223.0519629100973, + "blocks.0.w2.weight": 475.30658612467033, + "blocks.1.ln.weight": 9.102700233459473, + "blocks.1.w1.weight": 234.88719578297977, + "blocks.1.w1.bias": 224.13271789865033, + "blocks.1.w2.weight": 330.0486356799854, + "blocks.2.ln.weight": 8.034659385681152, + "blocks.2.w1.weight": 218.84814555155938, + "blocks.2.w1.bias": 202.24920680355018, + "blocks.2.w2.weight": 322.67560949424507, + "blocks.3.ln.weight": 8.901957511901855, + "blocks.3.w1.weight": 255.31555530937166, + "blocks.3.w1.bias": 231.65373117308073, + "blocks.3.w2.weight": 349.8458716057918, + "blocks.4.ln.weight": 9.597921371459961, + "blocks.4.w1.weight": 268.9594747130486, + "blocks.4.w1.bias": 248.71219274081335, + "blocks.4.w2.weight": 348.3370635956996, + "blocks.5.ln.weight": 8.862635612487793, + "blocks.5.w1.weight": 246.04530310618395, + "blocks.5.w1.bias": 236.82548627938067, + "blocks.5.w2.weight": 334.66771305862346, + "blocks.6.ln.weight": 11.917081832885742, + "blocks.6.w1.weight": 333.54126147765857, + "blocks.6.w1.bias": 313.5525629552238, + "blocks.6.w2.weight": 456.098261674838, + "blocks.7.ln.weight": 8.582963943481445, + "blocks.7.w1.weight": 238.44273828859957, + "blocks.7.w1.bias": 233.1627077875165, + "blocks.7.w2.weight": 333.95054465825257, + "out_ln.weight": 0.8174178600311279, + "out_head.weight": 9.070668717195536, + "out_head.bias": 0.3922775460932503 + } + }, + "fa": { + "log": { + "train_loss": [ + 2.0616466424560547, + 2.0068159678649904, + 1.9794388083648682, + 1.9586813756561279, + 1.9426360791778565, + 1.9318589163589477, + 1.9205885326766967, + 1.9138255990600586, + 1.9087649211120605, + 1.9012188312149048, + 1.8993849408340455, + 1.889568119506836, + 1.8889932192230225, + 1.886743229675293, + 1.881684862098694, + 1.8811524909210204, + 1.8822594582366943, + 1.8778549640655517, + 1.8765599794769288, + 1.8783365441131592, + 1.8720827561950684, + 1.8732831771087646, + 1.8713359966659546, + 1.8697589547729492, + 1.8710370569610595, + 1.8677758638763429, + 1.8653087866973876, + 1.8623830352783204, + 1.860157603187561, + 1.8587954444885253, + 1.8555354092788696, + 1.851196873703003, + 1.8507367937850951, + 1.847343747253418, + 1.8439307402801515, + 1.8404593259048463, + 1.8425774618530273, + 1.834388339767456, + 1.838042152633667, + 1.8388391250610352, + 1.8344174416732788, + 1.8349179996109009, + 1.8346392430877685, + 1.834155149269104, + 1.830893579940796, + 1.83102765045166, + 1.8300759775543214, + 1.8317738976287843, + 1.8306509511947633, + 1.8280337787246703, + 1.8300961248779297, + 1.8254442840576173, + 1.827361393814087, + 1.8285963610839844, + 1.8233610538101197, + 1.8227714263534547, + 1.825165744857788, + 1.8187065335083008, + 1.8204853109359742, + 1.8197687387084962, + 1.8170435654449464, + 1.8141087036132812, + 1.8137724670410156, + 1.8160163790893555, + 1.8154231036376953, + 1.8116517877197265, + 1.81113690158844, + 1.8139338250732422, + 1.8109377099990844, + 1.8099997354125976, + 1.808261435623169, + 1.8079131549835206, + 1.8053869789886474, + 1.808599826927185, + 1.8066932723999023, + 1.8063899947738646, + 1.8004683576202392, + 1.803932279701233, + 1.8084900228500367, + 1.8048654077911377, + 1.8045777291107177, + 1.805527420272827, + 1.8058658325576782, + 1.8057460729980468, + 1.8027491494750976, + 1.7995332913589477, + 1.8038537915039063, + 1.8010695806884767, + 1.801180100402832, + 1.799683776512146, + 1.800211155319214, + 1.8025256408309938, + 1.801271418762207, + 1.8011763118743898, + 1.802209068031311, + 1.8032223471450806, + 1.8022766012191773, + 1.8044791673278808, + 1.799467188949585, + 1.8024462516021729 + ], + "train_acc": [ + 0.23552, + 0.25616, + 0.27266, + 0.2833, + 0.2882, + 0.29386, + 0.29772, + 0.30354, + 0.30432, + 0.3065, + 0.30664, + 0.31514, + 0.3148, + 0.31538, + 0.32214, + 0.31962, + 0.3197, + 0.32268, + 0.32306, + 0.32158, + 0.32436, + 0.32346, + 0.32862, + 0.32786, + 0.33004, + 0.33162, + 0.33204, + 0.33184, + 0.3347, + 0.33274, + 0.3368, + 0.33864, + 0.33874, + 0.33938, + 0.34288, + 0.34424, + 0.34244, + 0.34636, + 0.34196, + 0.34354, + 0.34362, + 0.34602, + 0.34536, + 0.3441, + 0.34752, + 0.34754, + 0.34766, + 0.34916, + 0.34726, + 0.34842, + 0.34642, + 0.34936, + 0.3475, + 0.34798, + 0.34848, + 0.35126, + 0.34852, + 0.3516, + 0.35024, + 0.34976, + 0.35364, + 0.3529, + 0.35252, + 0.35444, + 0.35268, + 0.35314, + 0.35424, + 0.3535, + 0.3522, + 0.35664, + 0.35702, + 0.35622, + 0.35608, + 0.35676, + 0.35834, + 0.35704, + 0.359, + 0.35584, + 0.35488, + 0.35586, + 0.35672, + 0.35682, + 0.3565, + 0.35698, + 0.3576, + 0.35854, + 0.35728, + 0.3583, + 0.35986, + 0.36106, + 0.35794, + 0.3601, + 0.35654, + 0.3593, + 0.35772, + 0.3568, + 0.35766, + 0.35846, + 0.3578, + 0.36002 + ], + "test_acc": [ + 0.2672, + 0.262, + 0.2993, + 0.3091, + 0.3147, + 0.3179, + 0.3358, + 0.33, + 0.3176, + 0.34, + 0.34, + 0.3502, + 0.3515, + 0.3545, + 0.3441, + 0.3566, + 0.3555, + 0.3536, + 0.3524, + 0.3667, + 0.3434, + 0.3625, + 0.3608, + 0.365, + 0.3656, + 0.3608, + 0.372, + 0.376, + 0.3609, + 0.3628, + 0.3688, + 0.3727, + 0.3679, + 0.3753, + 0.372, + 0.3693, + 0.3738, + 0.3656, + 0.3765, + 0.3747, + 0.3699, + 0.3733, + 0.3738, + 0.3738, + 0.382, + 0.3755, + 0.3781, + 0.3784, + 0.3822, + 0.3797, + 0.3736, + 0.379, + 0.3821, + 0.3765, + 0.3776, + 0.3751, + 0.3767, + 0.3844, + 0.3724, + 0.3812, + 0.3826, + 0.3806, + 0.3836, + 0.3805, + 0.3826, + 0.3838, + 0.382, + 0.3868, + 0.389, + 0.3827, + 0.3814, + 0.3811, + 0.3798, + 0.3848, + 0.3887, + 0.3859, + 0.3839, + 0.3853, + 0.3844, + 0.3848, + 0.382, + 0.3853, + 0.3875, + 0.39, + 0.3858, + 0.3851, + 0.3881, + 0.3862, + 0.3857, + 0.3878, + 0.3861, + 0.3847, + 0.386, + 0.3856, + 0.3861, + 0.3856, + 0.3853, + 0.3859, + 0.3863, + 0.3864 + ] + }, + "diagnostics": { + "bp_cosine": [ + -0.004850233439356089, + 0.02608177438378334, + -0.07089883089065552, + -0.018480747938156128, + -0.021952074021100998, + -0.0742325484752655, + 0.014702335000038147, + 0.9978023171424866 + ], + "perturbation_rho": [ + 0.006630884483456612, + 0.04272107779979706, + 0.011450880207121372, + -0.02322380244731903, + -0.023082010447978973, + 0.005496563855558634, + 0.01337945181876421, + 0.0004176963120698929 + ], + "nudging": { + "0.001": [ + 1.0048970580101013e-06, + -9.452924132347107e-08, + 7.194466888904572e-08, + 8.987262845039368e-08, + 1.6298145055770874e-09, + 1.0640360414981842e-07, + -1.5133991837501526e-08, + -1.0149087756872177e-06 + ], + "0.003": [ + 3.1320378184318542e-06, + -1.778826117515564e-07, + 2.0326115190982819e-07, + 6.426125764846802e-08, + 8.940696716308594e-08, + 2.9616057872772217e-07, + -5.820766091346741e-08, + -3.819353878498077e-06 + ], + "0.01": [ + 1.0303221642971039e-05, + -9.301584213972092e-07, + 9.851064532995224e-07, + 3.310851752758026e-07, + 2.6961788535118103e-07, + 1.0957010090351105e-06, + -2.2887252271175385e-07, + -1.3955170288681984e-05 + ] + }, + "hidden_norms_per_layer": [ + 8388.6279296875, + 78234.5703125, + 285077.09375, + 690526.125, + 807174.625, + 983118.0625, + 1164967.375, + 1221883.25, + 645722.0 + ], + "bp_grad_norms_per_layer": [ + 1.7088063032133505e-05, + 1.4101407259659027e-06, + 7.534777637374646e-07, + 7.644222819180868e-07, + 7.630256391166768e-07, + 7.645797950317501e-07, + 7.790439440213959e-07, + 7.754296120765503e-07, + 7.678858651161136e-07 + ] + }, + "drift": { + "embed.weight": 74.95033008101261, + "embed.bias": 26.33394267321031, + "blocks.0.ln.weight": 1.718515396118164, + "blocks.0.w1.weight": 19.27588868651768, + "blocks.0.w1.bias": 16.597455066774533, + "blocks.0.w2.weight": 78.15464599428759, + "blocks.1.ln.weight": 1.166561484336853, + "blocks.1.w1.weight": 22.94470041655997, + "blocks.1.w1.bias": 19.67519900380536, + "blocks.1.w2.weight": 48.26877331154166, + "blocks.2.ln.weight": 0.8080865740776062, + "blocks.2.w1.weight": 22.880080354832984, + "blocks.2.w1.bias": 24.117736850367468, + "blocks.2.w2.weight": 35.99124940422002, + "blocks.3.ln.weight": 0.7484873533248901, + "blocks.3.w1.weight": 19.064483285445874, + "blocks.3.w1.bias": 19.658006710932415, + "blocks.3.w2.weight": 33.07320462824024, + "blocks.4.ln.weight": 0.8579117059707642, + "blocks.4.w1.weight": 21.835583499123793, + "blocks.4.w1.bias": 22.344412303522038, + "blocks.4.w2.weight": 37.98679923879473, + "blocks.5.ln.weight": 0.8886668682098389, + "blocks.5.w1.weight": 21.7450742777726, + "blocks.5.w1.bias": 23.066791818306008, + "blocks.5.w2.weight": 39.601364504544236, + "blocks.6.ln.weight": 0.8415105938911438, + "blocks.6.w1.weight": 19.797554767242907, + "blocks.6.w1.bias": 20.68672501454032, + "blocks.6.w2.weight": 35.63045020577285, + "blocks.7.ln.weight": 1.1435216665267944, + "blocks.7.w1.weight": 23.68255893921491, + "blocks.7.w1.bias": 24.7022235520292, + "blocks.7.w2.weight": 50.913078873963286, + "out_ln.weight": 0.38547876477241516, + "out_head.weight": 5.732715293326757, + "out_head.bias": 0.752175082244237 + } + } + }, + "config": { + "dataset": "cifar10", + "d_hidden": 256, + "num_blocks": 8, + "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": [ + 1 + ], + "gpu": 0, + "output_dir": "results/fa_dfa_d256_L8_seed1", + "methods": [ + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |
