diff options
Diffstat (limited to 'results/cifar100_d256_L4_s123/results_cifar100.json')
| -rw-r--r-- | results/cifar100_d256_L4_s123/results_cifar100.json | 1175 |
1 files changed, 1175 insertions, 0 deletions
diff --git a/results/cifar100_d256_L4_s123/results_cifar100.json b/results/cifar100_d256_L4_s123/results_cifar100.json new file mode 100644 index 0000000..5678f3e --- /dev/null +++ b/results/cifar100_d256_L4_s123/results_cifar100.json @@ -0,0 +1,1175 @@ +{ + "123": { + "bp": { + "log": { + "train_loss": [ + 3.967920153274536, + 3.643245764923096, + 3.486677552947998, + 3.389170970993042, + 3.3176012518310545, + 3.2524516577148437, + 3.196259712524414, + 3.152468095245361, + 3.116606701889038, + 3.0753799269104003, + 3.0412558711242674, + 3.0103372289276122, + 2.9851040869140624, + 2.9536216430664064, + 2.919733713989258, + 2.8960993688964844, + 2.872693371734619, + 2.84370941696167, + 2.8186388218688965, + 2.802537234954834, + 2.775104519882202, + 2.752585594329834, + 2.7324084952545165, + 2.7153604712677004, + 2.6842776251220704, + 2.667392734527588, + 2.6493602058410644, + 2.6304399774932863, + 2.6058559870910645, + 2.5874294744110107, + 2.5774721023559573, + 2.560277436981201, + 2.5345041821289063, + 2.5185300437927247, + 2.5040577131652832, + 2.4822867348480226, + 2.4557650662994384, + 2.4456021406555175, + 2.4341802147674563, + 2.4113426206207276, + 2.401808072052002, + 2.3768489891052247, + 2.3607859144592287, + 2.3446667034912108, + 2.329873125, + 2.3144486641693116, + 2.297909416656494, + 2.2798938914489746, + 2.266936881713867, + 2.2467652272033694, + 2.2354507537841797, + 2.2205448300933837, + 2.2017585570526124, + 2.1915560935974123, + 2.1722356983947755, + 2.1580420698547362, + 2.138712609024048, + 2.1268783713531496, + 2.1147646588897704, + 2.1034324825286865, + 2.088131596069336, + 2.0740757128143312, + 2.049090367126465, + 2.0376090378570555, + 2.0276603718566895, + 2.0167199254989625, + 2.0046445736694336, + 1.9935844805145264, + 1.9779361727905274, + 1.9663786093521118, + 1.9583420275497436, + 1.94067984375, + 1.929230265197754, + 1.9221224881362915, + 1.912958982810974, + 1.9034557625579833, + 1.8937049017715455, + 1.8790358535003662, + 1.876692020301819, + 1.8658557400512696, + 1.8551803729629517, + 1.8528358583831788, + 1.8547788270568848, + 1.83579787399292, + 1.825839135055542, + 1.8247737094116212, + 1.8170814221191407, + 1.8188595218658448, + 1.815419641456604, + 1.8142681406402588, + 1.8094764706802369, + 1.7942750823593139, + 1.7980619449615478, + 1.7933076245498658, + 1.7883917538452148, + 1.7867490985870362, + 1.7903605484771727, + 1.79024905128479, + 1.7870432864379884, + 1.7758308526611328 + ], + "train_acc": [ + 0.0986, + 0.14448, + 0.17068, + 0.18832, + 0.20038, + 0.21186, + 0.22002, + 0.22894, + 0.23506, + 0.24404, + 0.24964, + 0.25374, + 0.25782, + 0.26606, + 0.2711, + 0.27704, + 0.28214, + 0.2849, + 0.29088, + 0.2938, + 0.29992, + 0.3054, + 0.3096, + 0.30988, + 0.3165, + 0.3225, + 0.32436, + 0.32574, + 0.33332, + 0.33762, + 0.33924, + 0.343, + 0.34878, + 0.3505, + 0.35344, + 0.35974, + 0.36384, + 0.36348, + 0.3704, + 0.36994, + 0.37362, + 0.38144, + 0.38266, + 0.3852, + 0.3905, + 0.3918, + 0.39726, + 0.40046, + 0.40264, + 0.40842, + 0.40998, + 0.41106, + 0.4151, + 0.4193, + 0.4234, + 0.42538, + 0.43214, + 0.43326, + 0.43584, + 0.4373, + 0.44392, + 0.4447, + 0.44954, + 0.45406, + 0.45662, + 0.45958, + 0.46094, + 0.46528, + 0.46588, + 0.47114, + 0.47258, + 0.47772, + 0.47956, + 0.48164, + 0.48602, + 0.48696, + 0.48908, + 0.49296, + 0.49344, + 0.49132, + 0.49576, + 0.49762, + 0.49882, + 0.49984, + 0.50594, + 0.5048, + 0.50622, + 0.50504, + 0.50664, + 0.50984, + 0.50942, + 0.5116, + 0.5117, + 0.51286, + 0.51336, + 0.51602, + 0.5143, + 0.51562, + 0.51416, + 0.51578 + ], + "test_acc": [ + 0.1439, + 0.1786, + 0.2007, + 0.208, + 0.2165, + 0.2315, + 0.2325, + 0.2385, + 0.2483, + 0.2501, + 0.254, + 0.2585, + 0.2634, + 0.2702, + 0.2749, + 0.2758, + 0.2809, + 0.2848, + 0.2814, + 0.2812, + 0.2857, + 0.2915, + 0.2921, + 0.2923, + 0.2939, + 0.295, + 0.302, + 0.2965, + 0.3039, + 0.3021, + 0.3035, + 0.3055, + 0.3095, + 0.3052, + 0.3129, + 0.3112, + 0.3097, + 0.308, + 0.3123, + 0.3104, + 0.3135, + 0.3145, + 0.3105, + 0.3173, + 0.3165, + 0.3154, + 0.315, + 0.3172, + 0.3154, + 0.316, + 0.3159, + 0.3184, + 0.316, + 0.3219, + 0.3177, + 0.3206, + 0.3164, + 0.3151, + 0.3233, + 0.3209, + 0.3224, + 0.3239, + 0.3212, + 0.318, + 0.3248, + 0.3255, + 0.324, + 0.3217, + 0.323, + 0.3212, + 0.323, + 0.3249, + 0.3196, + 0.3215, + 0.3217, + 0.3204, + 0.3211, + 0.3238, + 0.3237, + 0.3208, + 0.3243, + 0.3225, + 0.3216, + 0.3233, + 0.3208, + 0.3198, + 0.3185, + 0.319, + 0.3196, + 0.3201, + 0.3197, + 0.3214, + 0.3217, + 0.3219, + 0.3217, + 0.3221, + 0.3217, + 0.3217, + 0.3218, + 0.3218 + ] + }, + "diagnostics": { + "bp_cosine": [ + 1.0, + 1.0, + 1.0, + 1.0 + ], + "perturbation_rho": [ + 0.9976038932800293, + 0.9983073472976685, + 0.9986611604690552, + 0.9987543821334839 + ], + "nudging": { + "0.001": [ + -0.002336513716727495, + -0.0024050232023000717, + -0.002418666612356901, + -0.0022877217270433903 + ], + "0.003": [ + -0.007005014922469854, + -0.007211083546280861, + -0.007251948118209839, + -0.006859402172267437 + ], + "0.01": [ + -0.023301422595977783, + -0.02398984506726265, + -0.024125652387738228, + -0.022821901366114616 + ] + }, + "hidden_norms_per_layer": [ + 228.35731506347656, + 216.70431518554688, + 197.1549835205078, + 189.50289916992188, + 193.0804901123047 + ], + "bp_grad_norms_per_layer": [ + 0.0011750052217394114, + 0.0012150296242907643, + 0.00123639649245888, + 0.0011825052788481116, + 0.0009545879438519478 + ] + }, + "drift": { + "embed.weight": 7.3996956607158415, + "embed.bias": 13.278713687472328, + "blocks.0.ln.weight": 0.15684440732002258, + "blocks.0.w1.weight": 3.6258136635529388, + "blocks.0.w1.bias": 5.57490288506498, + "blocks.0.w2.weight": 13.211600071846687, + "blocks.1.ln.weight": 0.26902759075164795, + "blocks.1.w1.weight": 3.9457858088997937, + "blocks.1.w1.bias": 4.8656985397285295, + "blocks.1.w2.weight": 14.784374975769294, + "blocks.2.ln.weight": 0.37451720237731934, + "blocks.2.w1.weight": 4.11706460399454, + "blocks.2.w1.bias": 4.146142198756796, + "blocks.2.w2.weight": 15.840220385160595, + "blocks.3.ln.weight": 0.5431011319160461, + "blocks.3.w1.weight": 4.464837126822249, + "blocks.3.w1.bias": 3.6144859875754407, + "blocks.3.w2.weight": 17.545783297025896, + "out_ln.weight": 0.14564843475818634, + "out_head.weight": 2.8817944671871123, + "out_head.bias": 3.056171448565604 + } + }, + "dfa": { + "log": { + "train_loss": [ + 4.179048243713379, + 4.067725786132812, + 4.071110060424805, + 4.091065592956543, + 4.10388401260376, + 4.1050505090332035, + 4.10725640045166, + 4.107924006652832, + 4.103042182159424, + 4.101259585266114, + 4.0994670211791995, + 4.092466098937988, + 4.091690199432373, + 4.088890562286377, + 4.083471063919068, + 4.082734613113403, + 4.079294251174927, + 4.074109273834228, + 4.072821339111328, + 4.072047854766845, + 4.068868335876465, + 4.064674437255859, + 4.064519798278808, + 4.062208421783447, + 4.059947501296997, + 4.0568214564514165, + 4.0578119854736325, + 4.05678455657959, + 4.054043687744141, + 4.04934491607666, + 4.049620868835449, + 4.048787186126709, + 4.041555746765137, + 4.046662612380981, + 4.043773061065674, + 4.0444277174377445, + 4.043180988922119, + 4.040115342483521, + 4.041283304138184, + 4.040324337005615, + 4.03773109161377, + 4.037001483154297, + 4.036031613235473, + 4.035188417053223, + 4.035520594024658, + 4.031777966766358, + 4.036043049621582, + 4.0324571522521975, + 4.032010763549804, + 4.032135240783691, + 4.028632532043457, + 4.028738806915284, + 4.029276781768798, + 4.030177063598633, + 4.027558575515747, + 4.0295089608764645, + 4.02790505935669, + 4.0268916735076905, + 4.025881996612549, + 4.024340693664551, + 4.024784558410644, + 4.022587103195191, + 4.023770975189209, + 4.024697297668457, + 4.023757594070434, + 4.025282996749878, + 4.020828542404175, + 4.02296846206665, + 4.021536037902832, + 4.021342518310547, + 4.023781976318359, + 4.020269456634521, + 4.019074986038208, + 4.020880366287232, + 4.02205199508667, + 4.019182727050781, + 4.019642067718506, + 4.019440497131348, + 4.018397001342773, + 4.020665705871582, + 4.020597035064697, + 4.019291881713867, + 4.019497079620361, + 4.018186092681884, + 4.018166368408203, + 4.017542911529541, + 4.018283336334228, + 4.015659646606445, + 4.0151326177978515, + 4.017795926361084, + 4.016853940124512, + 4.018234666442871, + 4.015345659790039, + 4.017251117095947, + 4.018545875244141, + 4.017435375213623, + 4.015963760681152, + 4.015863068695069, + 4.015675266418457, + 4.018068486938477 + ], + "train_acc": [ + 0.06442, + 0.07788, + 0.07748, + 0.07344, + 0.07334, + 0.0722, + 0.07326, + 0.07414, + 0.07376, + 0.07382, + 0.07452, + 0.07688, + 0.07534, + 0.07782, + 0.08016, + 0.07796, + 0.08084, + 0.0805, + 0.0801, + 0.08082, + 0.08118, + 0.08206, + 0.08406, + 0.08258, + 0.08372, + 0.08414, + 0.08464, + 0.08358, + 0.0859, + 0.08652, + 0.08582, + 0.08718, + 0.08834, + 0.08784, + 0.08652, + 0.08678, + 0.08714, + 0.08734, + 0.0885, + 0.08662, + 0.08978, + 0.08958, + 0.08816, + 0.08956, + 0.08942, + 0.09074, + 0.08972, + 0.08778, + 0.08952, + 0.08974, + 0.09166, + 0.08992, + 0.0922, + 0.09098, + 0.09182, + 0.092, + 0.09068, + 0.09088, + 0.09264, + 0.09206, + 0.09184, + 0.09252, + 0.09102, + 0.09206, + 0.09348, + 0.09252, + 0.09214, + 0.09232, + 0.09276, + 0.09236, + 0.0915, + 0.09298, + 0.09236, + 0.0928, + 0.09326, + 0.09264, + 0.09282, + 0.0924, + 0.0927, + 0.09328, + 0.09474, + 0.09338, + 0.09336, + 0.09342, + 0.09358, + 0.09408, + 0.09404, + 0.09384, + 0.09388, + 0.09484, + 0.09448, + 0.09396, + 0.09364, + 0.09346, + 0.09286, + 0.09446, + 0.0944, + 0.094, + 0.0935, + 0.09318 + ], + "test_acc": [ + 0.0808, + 0.0776, + 0.0759, + 0.0747, + 0.0703, + 0.0777, + 0.0773, + 0.0701, + 0.0792, + 0.0764, + 0.0753, + 0.0743, + 0.0772, + 0.0824, + 0.0807, + 0.081, + 0.0784, + 0.0813, + 0.0774, + 0.0803, + 0.0826, + 0.0865, + 0.0793, + 0.0798, + 0.0821, + 0.082, + 0.0822, + 0.0812, + 0.0812, + 0.0845, + 0.0808, + 0.087, + 0.0849, + 0.0823, + 0.0861, + 0.0851, + 0.0873, + 0.0833, + 0.0826, + 0.0855, + 0.0843, + 0.0861, + 0.0904, + 0.0866, + 0.0882, + 0.0861, + 0.0875, + 0.0871, + 0.0843, + 0.0852, + 0.0874, + 0.0892, + 0.0868, + 0.0887, + 0.0882, + 0.0874, + 0.0816, + 0.0865, + 0.0861, + 0.0856, + 0.0874, + 0.0869, + 0.0905, + 0.0844, + 0.0838, + 0.0898, + 0.0881, + 0.0865, + 0.0891, + 0.0868, + 0.0861, + 0.0874, + 0.0909, + 0.0889, + 0.0856, + 0.0875, + 0.088, + 0.0864, + 0.0889, + 0.0867, + 0.0852, + 0.0869, + 0.0879, + 0.0875, + 0.0872, + 0.0883, + 0.0897, + 0.0879, + 0.0875, + 0.0875, + 0.0868, + 0.087, + 0.0869, + 0.0868, + 0.0875, + 0.0873, + 0.0871, + 0.0872, + 0.0872, + 0.0872 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.14216047525405884, + -0.000992744229733944, + -0.008401205763220787, + -0.018219897523522377 + ], + "perturbation_rho": [ + 0.053201157599687576, + 0.0, + 0.0, + 0.0 + ], + "nudging": { + "0.001": [ + -6.109476089477539e-07, + 3.725290298461914e-09, + 0.0, + 0.0 + ], + "0.003": [ + -1.6260892152786255e-06, + 3.725290298461914e-09, + 3.725290298461914e-09, + 0.0 + ], + "0.01": [ + -5.261041224002838e-06, + -1.4901161193847656e-08, + 5.587935447692871e-09, + 5.587935447692871e-09 + ] + }, + "hidden_norms_per_layer": [ + 19441.8984375, + 14779116.0, + 143599776.0, + 226312160.0, + 280974592.0 + ], + "bp_grad_norms_per_layer": [ + 1.4361885405378416e-06, + 8.184267485944474e-09, + 7.640669430486469e-09, + 7.643201627161034e-09, + 7.633985887878225e-09 + ] + }, + "drift": { + "embed.weight": 162.24755066692174, + "embed.bias": 87.51431369480697, + "blocks.0.ln.weight": 10.42286205291748, + "blocks.0.w1.weight": 122.86975821758014, + "blocks.0.w1.bias": 127.51968896872295, + "blocks.0.w2.weight": 144.60590878787062, + "blocks.1.ln.weight": 7.615029811859131, + "blocks.1.w1.weight": 201.6705191292493, + "blocks.1.w1.bias": 211.79976689727667, + "blocks.1.w2.weight": 125.73570047362797, + "blocks.2.ln.weight": 8.678234100341797, + "blocks.2.w1.weight": 226.59101485856874, + "blocks.2.w1.bias": 218.5035230014578, + "blocks.2.w2.weight": 132.07798132121874, + "blocks.3.ln.weight": 8.401058197021484, + "blocks.3.w1.weight": 221.76508002622688, + "blocks.3.w1.bias": 204.23780375800553, + "blocks.3.w2.weight": 123.66583688447245, + "out_ln.weight": 1.673746943473816, + "out_head.weight": 9.550233758399107, + "out_head.bias": 0.9096587370527971 + } + }, + "fa": { + "log": { + "train_loss": [ + 4.1970798970031735, + 4.170009743804932, + 4.130948940124512, + 4.094561860961914, + 4.050638759155273, + 4.024049156494141, + 4.013653193511963, + 4.014947728729248, + 4.032731681671143, + 4.047689598388672, + 4.044174816589355, + 4.0340768909454345, + 4.026809699783326, + 4.0191610552215575, + 4.009859062652588, + 4.004589066696167, + 4.000014752273559, + 3.9934563516235353, + 3.989729327392578, + 3.9867174030303953, + 3.980116393737793, + 3.979052162322998, + 3.976570917816162, + 3.972892049102783, + 3.9706854679870607, + 3.964727850341797, + 3.9669337309265136, + 3.962656950378418, + 3.959514749221802, + 3.950425816116333, + 3.9518183033752443, + 3.944525026168823, + 3.9370513851928712, + 3.9401658685302734, + 3.9340728694152833, + 3.933683126220703, + 3.929854642791748, + 3.9251790898132324, + 3.925471938934326, + 3.920428736419678, + 3.9194086865997315, + 3.916180799026489, + 3.910477230758667, + 3.9091271575164797, + 3.9079003392791747, + 3.9009140887451172, + 3.903854856109619, + 3.896608190765381, + 3.8983821444702147, + 3.8914879888153076, + 3.8876359399414064, + 3.8866743579101564, + 3.886013005371094, + 3.8823793061065675, + 3.8820696644592285, + 3.8807897300720215, + 3.8792161915588377, + 3.873366918258667, + 3.872225737609863, + 3.867956961669922, + 3.8688921507263183, + 3.864798822555542, + 3.8635835083770753, + 3.8616594306182863, + 3.86274955078125, + 3.8615631605529783, + 3.855813476715088, + 3.8575434510040285, + 3.8537498898315428, + 3.8516512804412844, + 3.853169044036865, + 3.848847638626099, + 3.8501204053497315, + 3.8475019567871094, + 3.8478130152130126, + 3.841899245147705, + 3.8445189476776123, + 3.8427941259002685, + 3.841190276489258, + 3.8432645402526857, + 3.840385712738037, + 3.841228730545044, + 3.8406887775421144, + 3.838410483932495, + 3.8366693685913087, + 3.8328694233703615, + 3.8348467766571046, + 3.8281507495880125, + 3.8312854344177247, + 3.8341920923614503, + 3.832209959793091, + 3.832971982574463, + 3.827496012649536, + 3.832459149017334, + 3.8320746698760986, + 3.833449503631592, + 3.8311014671325685, + 3.828391033859253, + 3.8309260621643064, + 3.832998709869385 + ], + "train_acc": [ + 0.06318, + 0.06262, + 0.06896, + 0.07402, + 0.08474, + 0.0894, + 0.09134, + 0.0919, + 0.08782, + 0.08538, + 0.08668, + 0.0868, + 0.08842, + 0.0918, + 0.09218, + 0.09394, + 0.09588, + 0.09596, + 0.09472, + 0.09676, + 0.09782, + 0.09964, + 0.10034, + 0.09964, + 0.09956, + 0.10158, + 0.10016, + 0.10198, + 0.10194, + 0.1036, + 0.10318, + 0.10394, + 0.1047, + 0.1039, + 0.10536, + 0.10528, + 0.10612, + 0.10692, + 0.10644, + 0.10704, + 0.10844, + 0.1082, + 0.10744, + 0.10862, + 0.10826, + 0.11176, + 0.10998, + 0.10964, + 0.11052, + 0.1107, + 0.11298, + 0.11352, + 0.11398, + 0.11322, + 0.11342, + 0.11344, + 0.11324, + 0.11226, + 0.11314, + 0.11472, + 0.11338, + 0.11524, + 0.11458, + 0.11598, + 0.11536, + 0.11664, + 0.11622, + 0.11714, + 0.1169, + 0.11656, + 0.11594, + 0.11838, + 0.1178, + 0.11898, + 0.1183, + 0.11852, + 0.11904, + 0.11914, + 0.1193, + 0.11876, + 0.11858, + 0.12086, + 0.11934, + 0.11944, + 0.12044, + 0.1214, + 0.12184, + 0.12008, + 0.12106, + 0.12018, + 0.12146, + 0.12004, + 0.12064, + 0.11986, + 0.12106, + 0.12248, + 0.12166, + 0.12014, + 0.12266, + 0.12278 + ], + "test_acc": [ + 0.0708, + 0.0659, + 0.0746, + 0.0752, + 0.0902, + 0.0946, + 0.0937, + 0.0824, + 0.0936, + 0.0847, + 0.095, + 0.084, + 0.0884, + 0.0959, + 0.0927, + 0.0944, + 0.101, + 0.0972, + 0.098, + 0.0997, + 0.095, + 0.1032, + 0.0977, + 0.0949, + 0.1003, + 0.1004, + 0.0969, + 0.101, + 0.1019, + 0.1037, + 0.0979, + 0.103, + 0.1047, + 0.1026, + 0.1092, + 0.1071, + 0.1062, + 0.1024, + 0.1051, + 0.1068, + 0.1088, + 0.105, + 0.108, + 0.1068, + 0.1092, + 0.1112, + 0.1096, + 0.1084, + 0.108, + 0.1091, + 0.1102, + 0.1141, + 0.1109, + 0.1138, + 0.1085, + 0.1143, + 0.1107, + 0.1135, + 0.1153, + 0.1135, + 0.1181, + 0.1159, + 0.1171, + 0.1126, + 0.1165, + 0.1193, + 0.1206, + 0.113, + 0.1139, + 0.1156, + 0.1152, + 0.1201, + 0.1239, + 0.1171, + 0.1187, + 0.1197, + 0.1187, + 0.1212, + 0.1187, + 0.1182, + 0.1187, + 0.1203, + 0.1195, + 0.1212, + 0.122, + 0.1207, + 0.1192, + 0.1205, + 0.1205, + 0.1215, + 0.12, + 0.1201, + 0.1203, + 0.1203, + 0.1213, + 0.1208, + 0.1209, + 0.1209, + 0.1209, + 0.1208 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.017588060349225998, + -0.016497433185577393, + -0.053346455097198486, + 0.9986427426338196 + ], + "perturbation_rho": [ + 0.02255980297923088, + 0.005564957857131958, + -0.049259480088949203, + -0.011541965417563915 + ], + "nudging": { + "0.001": [ + 1.280568540096283e-07, + 1.62515789270401e-07, + 7.82310962677002e-08, + -5.41098415851593e-07 + ], + "0.003": [ + 1.0710209608078003e-07, + 2.7334317564964294e-07, + 1.1641532182693481e-07, + -2.648215740919113e-06 + ], + "0.01": [ + 3.57162207365036e-07, + 2.039596438407898e-07, + 8.307397365570068e-07, + -1.1221971362829208e-05 + ] + }, + "hidden_norms_per_layer": [ + 19381.54296875, + 312615.90625, + 2321965.25, + 3640082.0, + 692179.875 + ], + "bp_grad_norms_per_layer": [ + 5.805031378258718e-06, + 6.129388907538669e-07, + 5.996270147079485e-07, + 6.020890737090667e-07, + 6.002287022965902e-07 + ] + }, + "drift": { + "embed.weight": 160.52406007787414, + "embed.bias": 24.62414456524661, + "blocks.0.ln.weight": 2.7208244800567627, + "blocks.0.w1.weight": 31.32931103808957, + "blocks.0.w1.bias": 24.578464288381767, + "blocks.0.w2.weight": 76.69986787657376, + "blocks.1.ln.weight": 2.1396796703338623, + "blocks.1.w1.weight": 41.923375948001016, + "blocks.1.w1.bias": 42.687481681423954, + "blocks.1.w2.weight": 41.924463722510936, + "blocks.2.ln.weight": 1.8985017538070679, + "blocks.2.w1.weight": 49.16152213344094, + "blocks.2.w1.bias": 49.20711111001434, + "blocks.2.w2.weight": 40.28985166850181, + "blocks.3.ln.weight": 1.8008719682693481, + "blocks.3.w1.weight": 48.175959456780525, + "blocks.3.w1.bias": 46.53396040604286, + "blocks.3.w2.weight": 50.22310178120013, + "out_ln.weight": 0.6654373407363892, + "out_head.weight": 4.396233728367674, + "out_head.bias": 1.1577362249288412 + } + } + }, + "config": { + "dataset": "cifar100", + "d_hidden": 256, + "num_blocks": 4, + "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": [ + 123 + ], + "gpu": 0, + "output_dir": "results/cifar100_d256_L4_s123", + "methods": [ + "bp", + "fa", + "dfa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 100 + } +}
\ No newline at end of file |
