diff options
Diffstat (limited to 'results/fa_depth_scan_d512_L8/results_cifar10.json')
| -rw-r--r-- | results/fa_depth_scan_d512_L8/results_cifar10.json | 455 |
1 files changed, 455 insertions, 0 deletions
diff --git a/results/fa_depth_scan_d512_L8/results_cifar10.json b/results/fa_depth_scan_d512_L8/results_cifar10.json new file mode 100644 index 0000000..ae9217c --- /dev/null +++ b/results/fa_depth_scan_d512_L8/results_cifar10.json @@ -0,0 +1,455 @@ +{ + "42": { + "fa": { + "log": { + "train_loss": [ + 2.043062113418579, + 1.95162189743042, + 1.920884204940796, + 1.9046970273590087, + 1.894838911705017, + 1.8837308880233765, + 1.8795231484603883, + 1.870696762161255, + 1.864269206237793, + 1.8619007619857788, + 1.8598490927505493, + 1.8526468254852295, + 1.8434244635009767, + 1.8405422570037842, + 1.8366851897811889, + 1.8321461069107057, + 1.8310893895721436, + 1.825379607810974, + 1.8238727295303345, + 1.8163351691055298, + 1.8104810161590577, + 1.8127811385726929, + 1.8086815365982056, + 1.804211870803833, + 1.8026433480453492, + 1.7963054836654664, + 1.7984822222518921, + 1.7911431057739258, + 1.7928030535888673, + 1.7889471031951905, + 1.7860312490081787, + 1.779016375427246, + 1.782450693588257, + 1.7774018466186523, + 1.7763366507339478, + 1.7730076089859008, + 1.7679658282852173, + 1.7679697444915772, + 1.7660620833969116, + 1.7650900997543335, + 1.7642925763320922, + 1.7597538912582398, + 1.7603528467178344, + 1.7629493494415283, + 1.7623634843826295, + 1.756051732444763, + 1.7566206322860718, + 1.7543605234146118, + 1.7578279922485351, + 1.7509302628326415, + 1.7494868576812743, + 1.751580044555664, + 1.7507692102813721, + 1.748738152809143, + 1.7474167670059204, + 1.7463805029678345, + 1.7446185354995727, + 1.742856813430786, + 1.7411311776351928, + 1.7384925448989867, + 1.7379512426376342, + 1.7373969628524781, + 1.7368676550674438, + 1.735899489364624, + 1.7348208691787719, + 1.7361816341400147, + 1.7328850330352783, + 1.7339306282806397, + 1.7350596377563476, + 1.729668268814087, + 1.7330098389053346, + 1.7316270180892945, + 1.7296399521255492, + 1.7288719012451172, + 1.7293606524658203, + 1.7286542098999023, + 1.7271976955795287, + 1.726504995765686, + 1.7264521031951905, + 1.7255205670547484, + 1.7223241259384154, + 1.7238075354385376, + 1.7262169020843505, + 1.7254022949981689, + 1.7218294512176513, + 1.7234743248748778, + 1.7214548685455322, + 1.7229200372695923, + 1.722522060546875, + 1.7214176669311523, + 1.721013473739624, + 1.7213758332061768, + 1.7200873915863037, + 1.7215943475341797, + 1.7193219763183594, + 1.7175406676483154, + 1.7181979109954835, + 1.7207658834075927, + 1.7238012107086182, + 1.7181251257705688 + ], + "train_acc": [ + 0.2481, + 0.28956, + 0.29888, + 0.30732, + 0.31422, + 0.31622, + 0.32282, + 0.3247, + 0.3269, + 0.33028, + 0.33286, + 0.33738, + 0.3392, + 0.34038, + 0.3437, + 0.34256, + 0.34454, + 0.34842, + 0.34672, + 0.35004, + 0.3535, + 0.35304, + 0.35456, + 0.35594, + 0.35742, + 0.35726, + 0.35818, + 0.35816, + 0.35896, + 0.36116, + 0.36062, + 0.36406, + 0.36038, + 0.36366, + 0.36602, + 0.3664, + 0.36858, + 0.36588, + 0.3697, + 0.3681, + 0.3694, + 0.37272, + 0.37256, + 0.3704, + 0.36982, + 0.3713, + 0.37258, + 0.37364, + 0.37452, + 0.37376, + 0.37334, + 0.37602, + 0.37386, + 0.37334, + 0.37724, + 0.37578, + 0.37558, + 0.3779, + 0.37994, + 0.37834, + 0.37772, + 0.38086, + 0.38184, + 0.38024, + 0.38166, + 0.38102, + 0.38296, + 0.38192, + 0.38094, + 0.3819, + 0.38064, + 0.38246, + 0.3825, + 0.38418, + 0.3848, + 0.3841, + 0.38336, + 0.38264, + 0.3835, + 0.3861, + 0.38616, + 0.38576, + 0.38372, + 0.38684, + 0.3885, + 0.3868, + 0.3877, + 0.38486, + 0.38744, + 0.38718, + 0.3855, + 0.38922, + 0.3886, + 0.38746, + 0.388, + 0.39044, + 0.38896, + 0.38804, + 0.3862, + 0.38788 + ], + "test_acc": [ + 0.296, + 0.3104, + 0.3372, + 0.3454, + 0.3476, + 0.3503, + 0.3516, + 0.3529, + 0.3615, + 0.3574, + 0.3669, + 0.3575, + 0.3654, + 0.3694, + 0.3666, + 0.3687, + 0.3662, + 0.3673, + 0.3741, + 0.3617, + 0.368, + 0.3787, + 0.3763, + 0.3839, + 0.3868, + 0.378, + 0.3857, + 0.3626, + 0.3829, + 0.3795, + 0.3808, + 0.3767, + 0.3809, + 0.3887, + 0.3878, + 0.3792, + 0.3945, + 0.384, + 0.3959, + 0.3955, + 0.3925, + 0.3985, + 0.3845, + 0.3945, + 0.3938, + 0.3921, + 0.3958, + 0.4002, + 0.3951, + 0.4002, + 0.4059, + 0.3946, + 0.3994, + 0.398, + 0.3924, + 0.3941, + 0.3965, + 0.3964, + 0.405, + 0.406, + 0.398, + 0.402, + 0.4056, + 0.3978, + 0.4048, + 0.4017, + 0.4051, + 0.4054, + 0.4028, + 0.4046, + 0.4031, + 0.4043, + 0.4074, + 0.4097, + 0.4072, + 0.4063, + 0.4068, + 0.4068, + 0.4075, + 0.4064, + 0.408, + 0.4072, + 0.4083, + 0.4043, + 0.4071, + 0.4075, + 0.407, + 0.4105, + 0.4073, + 0.4076, + 0.4071, + 0.4095, + 0.409, + 0.4099, + 0.4093, + 0.409, + 0.4087, + 0.4092, + 0.4093, + 0.4094 + ] + }, + "diagnostics": { + "bp_cosine": [ + 0.026475638151168823, + 0.0839557871222496, + -0.0336647555232048, + -0.057554133236408234, + -0.06538982689380646, + -0.06987135112285614, + -0.051756590604782104, + 0.9981189966201782 + ], + "perturbation_rho": [ + 0.021994909271597862, + -0.047078944742679596, + 0.022503621876239777, + 0.025055162608623505, + -0.025491345673799515, + -0.03935955837368965, + -0.002435870934277773, + 0.047669265419244766 + ], + "nudging": { + "0.001": [ + -2.411194145679474e-06, + -3.704335540533066e-07, + 4.307366907596588e-08, + 9.359791874885559e-08, + 1.1106021702289581e-07, + 7.031485438346863e-08, + 1.210719347000122e-08, + -1.080334186553955e-06 + ], + "0.003": [ + -7.076392648741603e-06, + -1.1655502021312714e-06, + 2.3515895009040833e-07, + 3.2177194952964783e-07, + 3.494787961244583e-07, + 2.796296030282974e-07, + 1.310836523771286e-07, + -4.032859578728676e-06 + ], + "0.01": [ + -2.3676970158703625e-05, + -4.229601472616196e-06, + 3.962777554988861e-07, + 9.336508810520172e-07, + 1.050299033522606e-06, + 1.1418014764785767e-06, + 7.725320756435394e-07, + -1.4404766261577606e-05 + ] + }, + "hidden_norms_per_layer": [ + 7406.1591796875, + 81019.6640625, + 737389.0, + 1286861.0, + 1535224.625, + 1916653.125, + 2282567.25, + 2428105.25, + 1218042.25 + ], + "bp_grad_norms_per_layer": [ + 3.065524288103916e-05, + 1.910160335683031e-06, + 6.406741590581078e-07, + 6.244783321562863e-07, + 6.195934929564828e-07, + 6.173827955535671e-07, + 6.186029395394144e-07, + 6.207166052263347e-07, + 6.07166782629065e-07 + ] + }, + "drift": { + "embed.weight": 47.55569593801599, + "embed.bias": 14.17447035910836, + "blocks.0.ln.weight": 1.0983379558592676, + "blocks.0.w1.weight": 16.250524799209163, + "blocks.0.w1.bias": 12.62261139589463, + "blocks.0.w2.weight": 61.59978351635402, + "blocks.1.ln.weight": 1.1319422288079952, + "blocks.1.w1.weight": 22.880541075681524, + "blocks.1.w1.bias": 17.383779962472858, + "blocks.1.w2.weight": 44.96085798621816, + "blocks.2.ln.weight": 0.8177694150936128, + "blocks.2.w1.weight": 22.134352628330692, + "blocks.2.w1.bias": 21.07281150651575, + "blocks.2.w2.weight": 35.61435775314368, + "blocks.3.ln.weight": 0.7311969207849126, + "blocks.3.w1.weight": 20.785325587610842, + "blocks.3.w1.bias": 19.03573291074914, + "blocks.3.w2.weight": 35.05588574995981, + "blocks.4.ln.weight": 0.6490724785036317, + "blocks.4.w1.weight": 21.354390651446167, + "blocks.4.w1.bias": 21.159821383377235, + "blocks.4.w2.weight": 28.795764202368506, + "blocks.5.ln.weight": 0.6339903970932835, + "blocks.5.w1.weight": 21.59073387917402, + "blocks.5.w1.bias": 21.94406712381077, + "blocks.5.w2.weight": 32.095039396227314, + "blocks.6.ln.weight": 0.6243319829763818, + "blocks.6.w1.weight": 19.49885333502707, + "blocks.6.w1.bias": 20.57317316179931, + "blocks.6.w2.weight": 32.88519516715278, + "blocks.7.ln.weight": 0.7128901839398082, + "blocks.7.w1.weight": 20.944827961922154, + "blocks.7.w1.bias": 21.447201189692905, + "blocks.7.w2.weight": 42.87255169680327, + "out_ln.weight": 0.37541884485095733, + "out_head.weight": 6.641485211520903, + "out_head.bias": 1.0438900292204085 + } + } + }, + "config": { + "dataset": "cifar10", + "d_hidden": 512, + "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": [ + 42 + ], + "gpu": 0, + "output_dir": "results/fa_depth_scan_d512_L8", + "methods": [ + "fa" + ], + "random_targets": false, + "penalty_lam": 0.0, + "num_classes": 10 + } +}
\ No newline at end of file |
