diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-04-26 09:31:30 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-04-26 09:31:30 -0500 |
| commit | a501c1c84b6ac4ff7dbf2e4b92cebd3122eb7abe (patch) | |
| tree | 25a83479302e211359bd4f49df44b2bf69d0aaee /results | |
| parent | 9751e97dd190b8667c337215dcb70e0cab8f92ff (diff) | |
BP+EP audit for d=512 L=2 qualifying seeds + CIFAR-100 support
BP results for qualifying seeds (1, 2, 5) on d=512 L=2:
BP s1: 0.606, s2: 0.608, s5: 0.607 (all above frozen 0.349)
FA s1: 0.347, s2: 0.346, s5: 0.341 (all below frozen, cos +0.47-0.49)
DFA s1: 0.298, s2: 0.297, s5: 0.296 (all below frozen, cos +0.18-0.21)
EP did not save (likely architecture compatibility issue at d=512 L=2).
Also: added CIFAR-100 dataset support to both cifar_resmlp.py and
resmlp_frozen_blocks_baseline.py for the harder-task scan.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Diffstat (limited to 'results')
| -rw-r--r-- | results/audit_d512_L2_bpep.log | 69 | ||||
| -rw-r--r-- | results/audit_d512_L2_seed1/results_cifar10.json | 390 | ||||
| -rw-r--r-- | results/audit_d512_L2_seed2/results_cifar10.json | 390 | ||||
| -rw-r--r-- | results/audit_d512_L2_seed5/results_cifar10.json | 390 |
4 files changed, 1239 insertions, 0 deletions
diff --git a/results/audit_d512_L2_bpep.log b/results/audit_d512_L2_bpep.log new file mode 100644 index 0000000..189ca40 --- /dev/null +++ b/results/audit_d512_L2_bpep.log @@ -0,0 +1,69 @@ +=== BP+EP ONLY: d=512 L=2, seeds 1,2,5 === +Start: Sun Apr 26 09:03:57 AM CDT 2026 + seed=1 (Sun Apr 26 09:03:57 AM CDT 2026) +Using device: cuda:0 + +============================================================ +Seed 1 +============================================================ + +--- BP --- + [BP] Epoch 1: loss=1.9296, train=0.3129, test=0.3920 + [BP] Epoch 10: loss=1.3859, train=0.4995, test=0.5210 + [BP] Epoch 20: loss=1.2614, train=0.5473, test=0.5560 + [BP] Epoch 30: loss=1.1744, train=0.5783, test=0.5708 + [BP] Epoch 40: loss=1.1065, train=0.6047, test=0.5824 + [BP] Epoch 50: loss=1.0347, train=0.6289, test=0.5947 + [BP] Epoch 60: loss=0.9762, train=0.6505, test=0.5996 + [BP] Epoch 70: loss=0.9339, train=0.6657, test=0.6027 + [BP] Epoch 80: loss=0.8948, train=0.6800, test=0.6057 + [BP] Epoch 90: loss=0.8688, train=0.6898, test=0.6049 + [BP] Epoch 100: loss=0.8640, train=0.6923, test=0.6061 + Final test acc: 0.6061 + +All results saved to results/audit_d512_L2_seed1/results_cifar10.json + seed=2 (Sun Apr 26 09:10:20 AM CDT 2026) +Using device: cuda:0 + +============================================================ +Seed 2 +============================================================ + +--- BP --- + [BP] Epoch 1: loss=1.9221, train=0.3144, test=0.3766 + [BP] Epoch 10: loss=1.3836, train=0.5035, test=0.5250 + [BP] Epoch 20: loss=1.2625, train=0.5460, test=0.5533 + [BP] Epoch 30: loss=1.1729, train=0.5789, test=0.5744 + [BP] Epoch 40: loss=1.1072, train=0.6043, test=0.5873 + [BP] Epoch 50: loss=1.0421, train=0.6267, test=0.5965 + [BP] Epoch 60: loss=0.9809, train=0.6478, test=0.5991 + [BP] Epoch 70: loss=0.9360, train=0.6651, test=0.6015 + [BP] Epoch 80: loss=0.8951, train=0.6780, test=0.6060 + [BP] Epoch 90: loss=0.8717, train=0.6886, test=0.6066 + [BP] Epoch 100: loss=0.8696, train=0.6891, test=0.6076 + Final test acc: 0.6076 + +All results saved to results/audit_d512_L2_seed2/results_cifar10.json + seed=5 (Sun Apr 26 09:17:16 AM CDT 2026) +Using device: cuda:0 + +============================================================ +Seed 5 +============================================================ + +--- BP --- + [BP] Epoch 1: loss=1.9317, train=0.3139, test=0.3962 + [BP] Epoch 10: loss=1.3844, train=0.5014, test=0.5149 + [BP] Epoch 20: loss=1.2584, train=0.5467, test=0.5460 + [BP] Epoch 30: loss=1.1757, train=0.5749, test=0.5686 + [BP] Epoch 40: loss=1.1029, train=0.6043, test=0.5863 + [BP] Epoch 50: loss=1.0424, train=0.6253, test=0.5893 + [BP] Epoch 60: loss=0.9865, train=0.6452, test=0.5900 + [BP] Epoch 70: loss=0.9359, train=0.6647, test=0.6023 + [BP] Epoch 80: loss=0.8963, train=0.6780, test=0.6028 + [BP] Epoch 90: loss=0.8742, train=0.6873, test=0.6052 + [BP] Epoch 100: loss=0.8637, train=0.6916, test=0.6065 + Final test acc: 0.6065 + +All results saved to results/audit_d512_L2_seed5/results_cifar10.json +=== BP+EP DONE (Sun Apr 26 09:24:15 AM CDT 2026) === diff --git a/results/audit_d512_L2_seed1/results_cifar10.json b/results/audit_d512_L2_seed1/results_cifar10.json new file mode 100644 index 0000000..3d10a87 --- /dev/null +++ b/results/audit_d512_L2_seed1/results_cifar10.json @@ -0,0 +1,390 @@ +{ + "1": { + "bp": { + "log": { + "train_loss": [ + 1.9296325939178467, + 1.6955299231719971, + 1.6138011435317994, + 1.55534116355896, + 1.5123624341583253, + 1.4784945290756226, + 1.4516578591156006, + 1.4285693119049072, + 1.405858750038147, + 1.3859224660491944, + 1.3668672243499755, + 1.3519928998565673, + 1.3397786053848266, + 1.3254009517669678, + 1.313598297805786, + 1.2993963008880616, + 1.289289883041382, + 1.284569970626831, + 1.2686502683258056, + 1.2613813147735595, + 1.249849778366089, + 1.2403198554992676, + 1.2297505572891236, + 1.221708407974243, + 1.2131192414093017, + 1.2066168117523193, + 1.1957441337966919, + 1.1862484833526612, + 1.1765419130706787, + 1.1744289278411866, + 1.167908534889221, + 1.1589642641448974, + 1.153614723968506, + 1.139004853553772, + 1.1321262197113038, + 1.1283520357131958, + 1.122227259197235, + 1.1184809755325318, + 1.1113620584487915, + 1.1065167045211792, + 1.0947601670074463, + 1.093278687286377, + 1.0809579884338378, + 1.0786378560256957, + 1.0698642824935913, + 1.0579045901679993, + 1.0570202745437622, + 1.0537335617065429, + 1.0405854816818236, + 1.0347318133354186, + 1.0302755058288575, + 1.0242040428924561, + 1.025987678565979, + 1.016097176322937, + 1.0128265772819518, + 1.006155388031006, + 0.997192684135437, + 0.9936126647567749, + 0.9894703555488586, + 0.976179165840149, + 0.9738626984977722, + 0.9731567725181579, + 0.965264437637329, + 0.9619330932426453, + 0.9586215613174438, + 0.9522200344657898, + 0.9470485179901124, + 0.9462749715423584, + 0.9391794966125488, + 0.9339024310302735, + 0.927083604259491, + 0.9226054348373413, + 0.9200397926712036, + 0.9103769140243531, + 0.9085988710212708, + 0.9063865534591675, + 0.901920437450409, + 0.9022768329238892, + 0.8965523072433471, + 0.8947684677314758, + 0.8879585018920898, + 0.8842870276451111, + 0.8808215922546386, + 0.8800564193344116, + 0.8839287296867371, + 0.8729074939155579, + 0.8754293099594116, + 0.8725126181030274, + 0.8723966160202027, + 0.8688337619400024, + 0.8690940658760071, + 0.8679663080215454, + 0.8710675184631348, + 0.8609974372291564, + 0.8644896974182129, + 0.863099695854187, + 0.8652363301277161, + 0.8637897633361816, + 0.8604615099716186, + 0.864042571220398 + ], + "train_acc": [ + 0.31294, + 0.38728, + 0.41466, + 0.4372, + 0.45268, + 0.4665, + 0.47696, + 0.4846, + 0.49608, + 0.49946, + 0.50698, + 0.513, + 0.5166, + 0.5226, + 0.52834, + 0.53034, + 0.53718, + 0.53996, + 0.54494, + 0.54734, + 0.54966, + 0.55176, + 0.5573, + 0.55872, + 0.56454, + 0.56696, + 0.57056, + 0.57534, + 0.58012, + 0.57834, + 0.58108, + 0.58338, + 0.58752, + 0.59352, + 0.59306, + 0.59424, + 0.59922, + 0.59846, + 0.60114, + 0.60474, + 0.60902, + 0.6088, + 0.61266, + 0.61336, + 0.6186, + 0.62132, + 0.62238, + 0.62048, + 0.62696, + 0.6289, + 0.63132, + 0.63454, + 0.63286, + 0.63724, + 0.63726, + 0.64324, + 0.64212, + 0.64406, + 0.64948, + 0.65046, + 0.65014, + 0.65088, + 0.65372, + 0.65564, + 0.6583, + 0.66002, + 0.66078, + 0.66154, + 0.66326, + 0.6657, + 0.66906, + 0.6686, + 0.6717, + 0.67242, + 0.67378, + 0.67736, + 0.67734, + 0.67662, + 0.67948, + 0.67996, + 0.68162, + 0.68254, + 0.68498, + 0.68376, + 0.6843, + 0.6891, + 0.68676, + 0.68826, + 0.68864, + 0.68982, + 0.6884, + 0.6911, + 0.68942, + 0.69414, + 0.69016, + 0.69146, + 0.69174, + 0.6907, + 0.69126, + 0.69228 + ], + "test_acc": [ + 0.392, + 0.4208, + 0.4512, + 0.4558, + 0.4744, + 0.485, + 0.4941, + 0.4976, + 0.51, + 0.521, + 0.51, + 0.5209, + 0.5297, + 0.5277, + 0.5278, + 0.5274, + 0.5407, + 0.5339, + 0.5493, + 0.556, + 0.5448, + 0.5543, + 0.555, + 0.5604, + 0.5579, + 0.5581, + 0.5641, + 0.5703, + 0.566, + 0.5708, + 0.571, + 0.571, + 0.5764, + 0.5813, + 0.5807, + 0.5781, + 0.5801, + 0.5823, + 0.5807, + 0.5824, + 0.5848, + 0.5926, + 0.5847, + 0.5875, + 0.5875, + 0.5879, + 0.589, + 0.5912, + 0.5893, + 0.5947, + 0.5862, + 0.5891, + 0.5946, + 0.5941, + 0.5954, + 0.5984, + 0.5969, + 0.605, + 0.5959, + 0.5996, + 0.5992, + 0.5982, + 0.5977, + 0.6037, + 0.6017, + 0.5998, + 0.6022, + 0.6015, + 0.6026, + 0.6027, + 0.6021, + 0.6094, + 0.6067, + 0.6076, + 0.6091, + 0.6042, + 0.6088, + 0.6039, + 0.6046, + 0.6057, + 0.6063, + 0.6047, + 0.6066, + 0.6049, + 0.6044, + 0.6055, + 0.6058, + 0.6064, + 0.6044, + 0.6049, + 0.6061, + 0.605, + 0.6057, + 0.6053, + 0.6057, + 0.6054, + 0.6064, + 0.606, + 0.6063, + 0.6061 + ] + }, + "diagnostics": { + "bp_cosine": [ + 1.0, + 1.0 + ], + "perturbation_rho": [ + 0.9848940372467041, + 0.9808741807937622 + ], + "nudging": { + "0.001": [ + -0.0009475597180426121, + -0.0008203912293538451 + ], + "0.003": [ + -0.002840994391590357, + -0.0024601160548627377 + ], + "0.01": [ + -0.009449327364563942, + -0.00818631798028946 + ] + }, + "hidden_norms_per_layer": [ + 391.6769714355469, + 444.2131652832031, + 513.0332641601562 + ], + "bp_grad_norms_per_layer": [ + 0.00027059661806561053, + 0.00025608466239646077, + 0.00019367740605957806 + ] + }, + "drift": { + "embed.weight": 8.915427741857473, + "embed.bias": 27.156905442864737, + "blocks.0.ln.weight": 0.19595621212104608, + "blocks.0.w1.weight": 4.640129799275547, + "blocks.0.w1.bias": 7.056336038538968, + "blocks.0.w2.weight": 12.58791155857812, + "blocks.1.ln.weight": 0.16250931550735154, + "blocks.1.w1.weight": 4.857328676677314, + "blocks.1.w1.bias": 3.996940086789265, + "blocks.1.w2.weight": 15.480760953262367, + "out_ln.weight": 0.42853373724625804, + "out_head.weight": 2.3755790989575662, + "out_head.bias": 1.4741172964683136 + } + } + }, + "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": [ + 1 + ], + "gpu": 0, + "output_dir": "results/audit_d512_L2_seed1", + "methods": [ + "bp", + "ep" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file diff --git a/results/audit_d512_L2_seed2/results_cifar10.json b/results/audit_d512_L2_seed2/results_cifar10.json new file mode 100644 index 0000000..91b0017 --- /dev/null +++ b/results/audit_d512_L2_seed2/results_cifar10.json @@ -0,0 +1,390 @@ +{ + "2": { + "bp": { + "log": { + "train_loss": [ + 1.922061587524414, + 1.692084983215332, + 1.607172614402771, + 1.5482525830841065, + 1.5021197470855714, + 1.4748955045700074, + 1.4428388889312744, + 1.4217174727630615, + 1.4051225070953368, + 1.3835643965911866, + 1.3667922228622436, + 1.354358706626892, + 1.339921778755188, + 1.3257393606567383, + 1.311626997642517, + 1.300900167274475, + 1.2898498025894165, + 1.2796314553833008, + 1.2673308673858643, + 1.2625081433868408, + 1.2486013053512572, + 1.241865634498596, + 1.2300066821670532, + 1.224858923110962, + 1.213783438873291, + 1.2022000830459594, + 1.1973355312728882, + 1.1863385457611084, + 1.1818072442245484, + 1.1728536146354676, + 1.1690682736206055, + 1.1595733835983277, + 1.1506435432815552, + 1.1464643268585204, + 1.137304499015808, + 1.1314265491104125, + 1.1229438304519654, + 1.1183279276275635, + 1.112445654220581, + 1.1071891917228698, + 1.097363534412384, + 1.0876824979400634, + 1.0859120985031128, + 1.0791477709197999, + 1.0662284672546387, + 1.0634171083831787, + 1.0585565639495849, + 1.0517011973381043, + 1.0471893041229248, + 1.0421191468048097, + 1.036344855041504, + 1.0279344216156006, + 1.0175055844116212, + 1.016998133125305, + 1.0116913210296632, + 1.006302006034851, + 1.0035799601745605, + 0.9983466986274719, + 0.9875040967941284, + 0.9809433363342285, + 0.9765851749420166, + 0.968288851776123, + 0.9666447104072571, + 0.9670352198982238, + 0.9577340194511413, + 0.9536547498893738, + 0.9494867282104492, + 0.9404494221878051, + 0.9403044367790222, + 0.9359520991897583, + 0.9269838360595704, + 0.9228393173217774, + 0.9189790390014648, + 0.9184281276702881, + 0.9100610099029541, + 0.9085202851676941, + 0.9071778277778626, + 0.9026333428573609, + 0.8975262915420532, + 0.8950850497817994, + 0.8901209000396728, + 0.8907435465621948, + 0.8859056187629699, + 0.8801224571228027, + 0.8822859317779541, + 0.8792905470657348, + 0.8799502360916138, + 0.8722799802780151, + 0.8758245003128051, + 0.8717239134597778, + 0.8706532048034668, + 0.8694537221908569, + 0.8709142148017883, + 0.8691516464614868, + 0.863338586769104, + 0.8660436923408509, + 0.8676521118927002, + 0.8690781581497192, + 0.8637138714408874, + 0.8695979451179504 + ], + "train_acc": [ + 0.31444, + 0.3864, + 0.41868, + 0.44042, + 0.45938, + 0.46832, + 0.4789, + 0.48926, + 0.49212, + 0.50354, + 0.50704, + 0.51316, + 0.51776, + 0.52244, + 0.52614, + 0.53234, + 0.53746, + 0.5378, + 0.54706, + 0.54602, + 0.54994, + 0.55234, + 0.55876, + 0.55914, + 0.56334, + 0.56898, + 0.57144, + 0.5739, + 0.57522, + 0.57892, + 0.58238, + 0.58474, + 0.58796, + 0.58802, + 0.59284, + 0.59466, + 0.59596, + 0.59806, + 0.60328, + 0.60434, + 0.60496, + 0.61006, + 0.61062, + 0.61354, + 0.61664, + 0.61956, + 0.6218, + 0.62308, + 0.62402, + 0.6267, + 0.62784, + 0.62872, + 0.63306, + 0.63584, + 0.63564, + 0.63842, + 0.64122, + 0.644, + 0.64672, + 0.64776, + 0.65148, + 0.65424, + 0.65212, + 0.65378, + 0.65734, + 0.65862, + 0.65962, + 0.66406, + 0.6627, + 0.6651, + 0.66764, + 0.67064, + 0.66892, + 0.67046, + 0.67454, + 0.6734, + 0.6747, + 0.6744, + 0.67754, + 0.67802, + 0.67966, + 0.68162, + 0.68238, + 0.68536, + 0.68618, + 0.6842, + 0.68406, + 0.68778, + 0.68746, + 0.6886, + 0.68716, + 0.68938, + 0.68766, + 0.68822, + 0.69044, + 0.69216, + 0.68952, + 0.68688, + 0.69096, + 0.68908 + ], + "test_acc": [ + 0.3766, + 0.4216, + 0.4409, + 0.467, + 0.4843, + 0.5011, + 0.489, + 0.5101, + 0.5192, + 0.525, + 0.5196, + 0.5258, + 0.5316, + 0.538, + 0.5347, + 0.543, + 0.5438, + 0.5466, + 0.5462, + 0.5533, + 0.5515, + 0.561, + 0.5646, + 0.5695, + 0.5649, + 0.56, + 0.5663, + 0.5705, + 0.5739, + 0.5744, + 0.5731, + 0.5777, + 0.5781, + 0.5754, + 0.5825, + 0.5832, + 0.5809, + 0.5825, + 0.5867, + 0.5873, + 0.587, + 0.5849, + 0.5924, + 0.5908, + 0.5855, + 0.5936, + 0.5925, + 0.5932, + 0.5947, + 0.5965, + 0.5947, + 0.5933, + 0.5966, + 0.601, + 0.5971, + 0.6016, + 0.5995, + 0.6003, + 0.6008, + 0.5991, + 0.6075, + 0.6019, + 0.5998, + 0.6009, + 0.6048, + 0.6026, + 0.6049, + 0.6055, + 0.6041, + 0.6015, + 0.6033, + 0.6023, + 0.6084, + 0.6072, + 0.6092, + 0.6077, + 0.6074, + 0.6089, + 0.6085, + 0.606, + 0.6073, + 0.6082, + 0.6049, + 0.6055, + 0.6061, + 0.6079, + 0.6054, + 0.6046, + 0.607, + 0.6066, + 0.607, + 0.6087, + 0.6082, + 0.6077, + 0.6083, + 0.6085, + 0.6081, + 0.6081, + 0.6075, + 0.6076 + ] + }, + "diagnostics": { + "bp_cosine": [ + 1.0, + 1.0 + ], + "perturbation_rho": [ + 0.9815833568572998, + 0.9817954301834106 + ], + "nudging": { + "0.001": [ + -0.0009584517683833838, + -0.0008271471597254276 + ], + "0.003": [ + -0.002873521763831377, + -0.0024800291284918785 + ], + "0.01": [ + -0.009556342847645283, + -0.008249775506556034 + ] + }, + "hidden_norms_per_layer": [ + 390.58270263671875, + 431.9788818359375, + 519.6278076171875 + ], + "bp_grad_norms_per_layer": [ + 0.00027044795569963753, + 0.0002459954412188381, + 0.00019575131591409445 + ] + }, + "drift": { + "embed.weight": 8.898083083083813, + "embed.bias": 26.347694310803384, + "blocks.0.ln.weight": 0.20274328784030526, + "blocks.0.w1.weight": 4.621360453224136, + "blocks.0.w1.bias": 7.5095563814792134, + "blocks.0.w2.weight": 12.461606875777026, + "blocks.1.ln.weight": 0.15771156053514024, + "blocks.1.w1.weight": 4.827172758362064, + "blocks.1.w1.bias": 4.123634659408102, + "blocks.1.w2.weight": 15.461087487652549, + "out_ln.weight": 0.4248704174046931, + "out_head.weight": 2.3659908988944958, + "out_head.bias": 1.435738306727874 + } + } + }, + "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": [ + 2 + ], + "gpu": 0, + "output_dir": "results/audit_d512_L2_seed2", + "methods": [ + "bp", + "ep" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file diff --git a/results/audit_d512_L2_seed5/results_cifar10.json b/results/audit_d512_L2_seed5/results_cifar10.json new file mode 100644 index 0000000..ee5f2db --- /dev/null +++ b/results/audit_d512_L2_seed5/results_cifar10.json @@ -0,0 +1,390 @@ +{ + "5": { + "bp": { + "log": { + "train_loss": [ + 1.931713076095581, + 1.693591755027771, + 1.6130402114868163, + 1.552940613937378, + 1.5053284896087646, + 1.4791135555648804, + 1.4480172431945801, + 1.4265927730560304, + 1.4027651021194458, + 1.384369655380249, + 1.367165181541443, + 1.3568550646591186, + 1.338783321838379, + 1.3342096368789673, + 1.3139104537200927, + 1.3003809491729736, + 1.2941038347625733, + 1.2822353719329833, + 1.2721263457489014, + 1.2584363819122315, + 1.2488371838378907, + 1.2416769229125977, + 1.2323745751190185, + 1.221822734489441, + 1.212864588394165, + 1.2087859251785278, + 1.1994400387954711, + 1.192341411972046, + 1.1844210750198365, + 1.1757035552978516, + 1.1669261846542358, + 1.160658155593872, + 1.1519462855911256, + 1.146460373954773, + 1.1364786016082764, + 1.1330338333511352, + 1.1193102619171142, + 1.1164188197135925, + 1.1119497132492064, + 1.102858406047821, + 1.0976424259567261, + 1.0894710400009155, + 1.0833782506942748, + 1.0755352562141418, + 1.069308932685852, + 1.0644779945755005, + 1.0603491143989563, + 1.053652259902954, + 1.0516060263442992, + 1.0424115861129761, + 1.0367504576873778, + 1.0326291426086425, + 1.02485561460495, + 1.0223284980392455, + 1.012132692604065, + 1.0019124921798706, + 1.0020748730659486, + 1.0001373498725892, + 0.9894566692924499, + 0.9865088276290893, + 0.9775360752868653, + 0.9728040076446534, + 0.9685138183212281, + 0.9657968924713135, + 0.9552892886161805, + 0.9537780396270752, + 0.9462199269485474, + 0.9408095097160339, + 0.9418458258247375, + 0.9358901969718934, + 0.9352027550888061, + 0.9245894254684448, + 0.923262604751587, + 0.9145718508911133, + 0.9137569423294067, + 0.9098806952667237, + 0.9072191628837586, + 0.902167465801239, + 0.8989853531646729, + 0.8962803745651245, + 0.8937453430366517, + 0.8901589241027832, + 0.8856565335273743, + 0.8899172011184693, + 0.881110893688202, + 0.8785582168960572, + 0.8790584596443176, + 0.8775734127426148, + 0.8746353070831299, + 0.8741918186378479, + 0.8740821181869507, + 0.8737027023506164, + 0.8711176183319091, + 0.868740560760498, + 0.875171669178009, + 0.8707077650833129, + 0.870856398601532, + 0.8660038369369507, + 0.8711395135307312, + 0.8637448191642761 + ], + "train_acc": [ + 0.31394, + 0.38464, + 0.41358, + 0.4385, + 0.45666, + 0.46498, + 0.4762, + 0.48486, + 0.4938, + 0.50142, + 0.50914, + 0.5129, + 0.51718, + 0.5171, + 0.52996, + 0.53262, + 0.53478, + 0.53948, + 0.54284, + 0.54666, + 0.55344, + 0.553, + 0.558, + 0.56114, + 0.56418, + 0.56542, + 0.5681, + 0.57326, + 0.57598, + 0.57486, + 0.58204, + 0.58382, + 0.58714, + 0.58948, + 0.59282, + 0.59114, + 0.59722, + 0.59984, + 0.59884, + 0.60428, + 0.60908, + 0.60954, + 0.61242, + 0.61476, + 0.61634, + 0.61874, + 0.6212, + 0.62136, + 0.62356, + 0.6253, + 0.629, + 0.63092, + 0.63226, + 0.63538, + 0.63828, + 0.64056, + 0.64176, + 0.64034, + 0.6472, + 0.64522, + 0.65022, + 0.64948, + 0.65394, + 0.6535, + 0.657, + 0.65866, + 0.6591, + 0.66512, + 0.66134, + 0.66468, + 0.66632, + 0.66738, + 0.6686, + 0.67304, + 0.67236, + 0.67272, + 0.67638, + 0.67832, + 0.67808, + 0.678, + 0.6805, + 0.681, + 0.68454, + 0.68108, + 0.6839, + 0.687, + 0.68758, + 0.68748, + 0.68672, + 0.68726, + 0.68704, + 0.68612, + 0.6872, + 0.69248, + 0.68708, + 0.68654, + 0.68832, + 0.69, + 0.6882, + 0.69156 + ], + "test_acc": [ + 0.3962, + 0.4116, + 0.4541, + 0.459, + 0.4677, + 0.4872, + 0.4915, + 0.5064, + 0.504, + 0.5149, + 0.5217, + 0.5335, + 0.5216, + 0.5324, + 0.5385, + 0.5413, + 0.5437, + 0.5501, + 0.5456, + 0.546, + 0.5523, + 0.5557, + 0.5483, + 0.557, + 0.5628, + 0.5662, + 0.5653, + 0.5666, + 0.5743, + 0.5686, + 0.5722, + 0.5762, + 0.5698, + 0.5672, + 0.5782, + 0.5779, + 0.5796, + 0.5848, + 0.5813, + 0.5863, + 0.5924, + 0.5902, + 0.5891, + 0.5895, + 0.5954, + 0.5913, + 0.5953, + 0.5926, + 0.5869, + 0.5893, + 0.595, + 0.5928, + 0.5928, + 0.5971, + 0.5952, + 0.5922, + 0.5935, + 0.5967, + 0.5936, + 0.59, + 0.5996, + 0.6027, + 0.5997, + 0.5962, + 0.5979, + 0.5996, + 0.6008, + 0.5973, + 0.602, + 0.6023, + 0.6046, + 0.604, + 0.605, + 0.603, + 0.6032, + 0.6031, + 0.6038, + 0.606, + 0.6057, + 0.6028, + 0.603, + 0.6063, + 0.6045, + 0.6022, + 0.6035, + 0.6058, + 0.6043, + 0.6064, + 0.6056, + 0.6052, + 0.6063, + 0.6057, + 0.6066, + 0.607, + 0.6062, + 0.6068, + 0.6067, + 0.6068, + 0.6064, + 0.6065 + ] + }, + "diagnostics": { + "bp_cosine": [ + 1.0, + 1.0 + ], + "perturbation_rho": [ + 0.9835590720176697, + 0.9854168891906738 + ], + "nudging": { + "0.001": [ + -0.001010372070595622, + -0.0008467765292152762 + ], + "0.003": [ + -0.003029399085789919, + -0.002539127366617322 + ], + "0.01": [ + -0.010077598504722118, + -0.008449133485555649 + ] + }, + "hidden_norms_per_layer": [ + 391.5282897949219, + 434.1443176269531, + 505.1020812988281 + ], + "bp_grad_norms_per_layer": [ + 0.0002652886905707419, + 0.00023903341207187623, + 0.00018291105516254902 + ] + }, + "drift": { + "embed.weight": 8.855401746520494, + "embed.bias": 26.608965363709814, + "blocks.0.ln.weight": 0.2061336436128178, + "blocks.0.w1.weight": 4.634322016280042, + "blocks.0.w1.bias": 6.984453011136666, + "blocks.0.w2.weight": 12.580916503938784, + "blocks.1.ln.weight": 0.16124904042161012, + "blocks.1.w1.weight": 4.87348406828928, + "blocks.1.w1.bias": 4.00290878921701, + "blocks.1.w2.weight": 15.581969461256797, + "out_ln.weight": 0.44012627027442575, + "out_head.weight": 2.3259179585019845, + "out_head.bias": 1.324384388591369 + } + } + }, + "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": [ + 5 + ], + "gpu": 0, + "output_dir": "results/audit_d512_L2_seed5", + "methods": [ + "bp", + "ep" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |
