{ "7": { "dfa": { "log": { "train_loss": [ 2.055191882095337, 2.043570478172302, 2.0389851570892334, 2.035236138343811, 2.034318865890503, 2.028933349571228, 2.027525556945801, 2.0267114724731443, 2.0244813998031614, 2.0217119938659667, 2.0195889541625975, 2.0210990812683107, 2.0200074877166747, 2.0166528286743164, 2.0172128536224365, 2.0116504791259766, 2.0116141784667967, 2.007682880592346, 2.0073163960266114, 2.0056607851791384, 2.0035111226654054, 2.0027661852264402, 2.001486244812012, 1.9986013162231446, 1.997822530975342, 1.994211967010498, 1.9924970539093017, 1.991006000213623, 1.9917142455291748, 1.9929195302581788, 1.991425333518982, 1.9895027528381348, 1.9874201156997682, 1.9884889245605468, 1.986476892967224, 1.9826913553237915, 1.9848502500152587, 1.982286188697815, 1.9824313144683838, 1.9826200205230713, 1.9815959204864502, 1.9807573030853272, 1.9788639946746827, 1.9796723455810548, 1.976197322998047, 1.9779787089538574, 1.9767974209594728, 1.9764958350372315, 1.9774945835876465, 1.9747930298614502, 1.9727936906433106, 1.9740131066131592, 1.9746413722991942, 1.9732583834838868, 1.9723580028533936, 1.9721147798919678, 1.972632116394043, 1.970795502243042, 1.9690511889648437, 1.9716948537826537, 1.9707828464508057, 1.9699502618026734, 1.969847806854248, 1.9696522138214112, 1.9674137978744506, 1.9683316724014281, 1.96831518825531, 1.9711441744613647, 1.9686481335830688, 1.9681067206573486, 1.9675092895507813, 1.9689227291107179, 1.9677108968353272, 1.9652254253387451, 1.9658482390594483, 1.9671778398895263, 1.9676923946762086, 1.9644599740219115, 1.9651101944351197, 1.9664753457260131, 1.9669549044799806, 1.9642051581573485, 1.9663040605163575, 1.9670575283432006, 1.9643160835266114, 1.9654435306167604, 1.964134133644104, 1.9641572988891602, 1.9649527727508544, 1.9648468869781495, 1.9629023349380492, 1.9663221743774415, 1.9669214375305175, 1.9649668531036377, 1.9656098468780518, 1.966679817123413, 1.9638590016174315, 1.9619427868270873, 1.9663854665756226, 1.966093027381897 ], "train_acc": [ 0.24264, 0.24694, 0.24798, 0.24876, 0.25246, 0.25632, 0.25708, 0.25698, 0.25824, 0.26182, 0.2633, 0.26056, 0.26094, 0.26186, 0.26334, 0.26518, 0.26768, 0.26548, 0.26774, 0.26878, 0.2695, 0.2733, 0.26882, 0.27458, 0.2744, 0.27714, 0.27356, 0.27912, 0.27626, 0.27554, 0.2775, 0.2796, 0.27856, 0.27752, 0.27988, 0.28044, 0.28096, 0.2833, 0.28072, 0.28224, 0.28246, 0.28386, 0.28302, 0.28622, 0.28438, 0.28368, 0.28372, 0.28316, 0.28436, 0.28636, 0.28652, 0.28768, 0.2859, 0.28832, 0.28462, 0.28838, 0.2877, 0.28634, 0.28904, 0.28846, 0.28746, 0.28738, 0.2889, 0.28872, 0.28988, 0.28838, 0.2881, 0.29144, 0.28876, 0.28998, 0.29006, 0.28796, 0.28852, 0.29134, 0.29044, 0.29064, 0.28968, 0.2908, 0.28898, 0.29126, 0.28764, 0.29144, 0.2911, 0.29084, 0.29148, 0.28904, 0.29322, 0.2902, 0.2903, 0.2903, 0.29346, 0.2899, 0.28924, 0.29046, 0.28966, 0.29182, 0.2899, 0.28986, 0.2916, 0.28824 ], "test_acc": [ 0.2524, 0.2673, 0.2815, 0.2506, 0.2816, 0.2711, 0.2809, 0.2844, 0.2929, 0.286, 0.2744, 0.289, 0.2842, 0.2845, 0.2813, 0.2816, 0.2741, 0.2919, 0.2911, 0.2705, 0.2945, 0.3034, 0.2878, 0.2798, 0.2791, 0.298, 0.2976, 0.293, 0.2996, 0.2971, 0.3057, 0.3087, 0.3011, 0.3061, 0.3061, 0.2919, 0.3051, 0.306, 0.3078, 0.3056, 0.3019, 0.3022, 0.3075, 0.3081, 0.3008, 0.2993, 0.2992, 0.3129, 0.3147, 0.2965, 0.31, 0.2998, 0.3082, 0.312, 0.3096, 0.3084, 0.306, 0.3072, 0.308, 0.3153, 0.3089, 0.3032, 0.3083, 0.3208, 0.3021, 0.3158, 0.3101, 0.3162, 0.3125, 0.3103, 0.3118, 0.309, 0.3117, 0.3098, 0.3162, 0.3136, 0.3107, 0.3074, 0.3122, 0.3141, 0.3127, 0.314, 0.3129, 0.3145, 0.3133, 0.3138, 0.3131, 0.3158, 0.3159, 0.3161, 0.3141, 0.3163, 0.3165, 0.3154, 0.3163, 0.3164, 0.3155, 0.3155, 0.3157, 0.3157 ] }, "diagnostics": { "bp_cosine": [ 0.4341934025287628, -0.0012890032958239317 ], "perturbation_rho": [ -0.016353363171219826, 0.0 ], "nudging": { "0.001": [ -6.253831088542938e-07, 0.0 ], "0.003": [ -1.7937272787094116e-06, 1.862645149230957e-09 ], "0.01": [ -6.007961928844452e-06, 5.587935447692871e-09 ] }, "hidden_norms_per_layer": [ 51618.765625, 586580544.0, 3527968768.0 ], "bp_grad_norms_per_layer": [ 3.575926825760689e-07, 3.2419492090873803e-10, 3.250963664935824e-10 ] }, "drift": { "embed.weight": 305.91109235091693, "embed.bias": 181.83596330336093, "blocks.0.ln.weight": 10.507483299521725, "blocks.0.w1.weight": 259.244642347064, "blocks.0.w1.bias": 195.88858302375624, "blocks.0.w2.weight": 487.57558684361754, "blocks.1.ln.weight": 9.741938719661329, "blocks.1.w1.weight": 370.35850714262523, "blocks.1.w1.bias": 288.30321808565367, "blocks.1.w2.weight": 411.72437143655895, "out_ln.weight": 0.4751365199666905, "out_head.weight": 6.958959988926819, "out_head.bias": 2.336896476076589 } }, "fa": { "log": { "train_loss": [ 2.065423551902771, 1.959441243019104, 1.922898962135315, 1.8985358092498779, 1.8807485336685181, 1.8632725805664063, 1.8535257900238038, 1.8472118298339844, 1.8383856839752197, 1.8354701901626587, 1.8370730047988892, 1.837422192993164, 1.8389883419036865, 1.8328349654388427, 1.8349895798110962, 1.8335126385116578, 1.8306356255722045, 1.8322726831817626, 1.8286183166885377, 1.8288718432617188, 1.8341219143295289, 1.8368035592269898, 1.836811799583435, 1.836639772872925, 1.8373656529998779, 1.836424264526367, 1.8352625201797486, 1.835800665588379, 1.832284771118164, 1.833700514755249, 1.8298065981674194, 1.8254384775543213, 1.821793406639099, 1.8201812942123412, 1.8183945245742799, 1.8144645150375367, 1.8105643741607667, 1.809894971961975, 1.804399468612671, 1.8044112042999267, 1.8014125130844116, 1.8024961044311523, 1.7954880487442018, 1.798938335647583, 1.790709694480896, 1.7935989169311524, 1.7944183544921875, 1.7913234217071534, 1.7892272469711303, 1.7870802127838135, 1.7874060748291016, 1.7849993677139282, 1.7821664669418336, 1.784739924659729, 1.7879353686523438, 1.786940495300293, 1.784608112564087, 1.7870150913238525, 1.7842396997070313, 1.7843707611465454, 1.7866798919296265, 1.7842874060440064, 1.7791111608886718, 1.7806335921859742, 1.7796328140640258, 1.7788975708770751, 1.7777235781097411, 1.7809623500823974, 1.7793159552764892, 1.7774666768646241, 1.7777468250274657, 1.7797447577667236, 1.7749399490356446, 1.7782198504257203, 1.7768975478744506, 1.7724751330184936, 1.7744919674682618, 1.771957764816284, 1.773150245323181, 1.7745548120880128, 1.7708911740112305, 1.7687720165634155, 1.7715272402191162, 1.7657094653320313, 1.7654507334899903, 1.7671412448120116, 1.7664664395904541, 1.768566664352417, 1.7679725888824462, 1.7632778255844117, 1.7664907150268554, 1.7670541214370727, 1.7653729449081421, 1.7616971383666993, 1.7609278664398194, 1.7637566931915283, 1.764719966506958, 1.7616608474349975, 1.7632052117538453, 1.7648505276107789 ], "train_acc": [ 0.24818, 0.2905, 0.3052, 0.31584, 0.32536, 0.33246, 0.33778, 0.33886, 0.33934, 0.34248, 0.3433, 0.34108, 0.33788, 0.34122, 0.3401, 0.34124, 0.3422, 0.3437, 0.34032, 0.34338, 0.33892, 0.33994, 0.3359, 0.33762, 0.33858, 0.33796, 0.33882, 0.34338, 0.34164, 0.34066, 0.34704, 0.3459, 0.34644, 0.34698, 0.34912, 0.34876, 0.3523, 0.35206, 0.35342, 0.35536, 0.35396, 0.3552, 0.35924, 0.35432, 0.35796, 0.35682, 0.35832, 0.35936, 0.36088, 0.36014, 0.36244, 0.3632, 0.36318, 0.36202, 0.36224, 0.36138, 0.36264, 0.35992, 0.3656, 0.36372, 0.3631, 0.3644, 0.3645, 0.36648, 0.36714, 0.36738, 0.3652, 0.36398, 0.36904, 0.3683, 0.36936, 0.3682, 0.36688, 0.36788, 0.3702, 0.36986, 0.36914, 0.3728, 0.3697, 0.37158, 0.37106, 0.37164, 0.37216, 0.3765, 0.37258, 0.3718, 0.37274, 0.37224, 0.3709, 0.372, 0.37362, 0.3726, 0.37316, 0.37384, 0.37496, 0.37558, 0.3719, 0.37734, 0.37454, 0.37464 ], "test_acc": [ 0.2781, 0.3155, 0.3352, 0.3255, 0.3405, 0.3545, 0.3621, 0.3461, 0.3528, 0.3543, 0.331, 0.3589, 0.3453, 0.349, 0.3481, 0.3479, 0.355, 0.3516, 0.347, 0.3387, 0.3431, 0.3418, 0.3297, 0.3231, 0.3156, 0.3425, 0.3359, 0.3325, 0.3627, 0.3371, 0.351, 0.3523, 0.3384, 0.3573, 0.3432, 0.3522, 0.3549, 0.3581, 0.361, 0.3577, 0.3477, 0.3498, 0.3374, 0.3495, 0.337, 0.3531, 0.3467, 0.3464, 0.3582, 0.3339, 0.3416, 0.3402, 0.3472, 0.3388, 0.3466, 0.3442, 0.3452, 0.3515, 0.3485, 0.3612, 0.3508, 0.3437, 0.3639, 0.3558, 0.3438, 0.3503, 0.3466, 0.3499, 0.3569, 0.3508, 0.3467, 0.3479, 0.3496, 0.352, 0.3543, 0.3569, 0.3474, 0.344, 0.3485, 0.3519, 0.3507, 0.352, 0.3534, 0.3545, 0.353, 0.3563, 0.3533, 0.355, 0.3552, 0.357, 0.354, 0.3546, 0.3545, 0.3529, 0.3537, 0.3533, 0.3535, 0.3541, 0.3539, 0.3537 ] }, "diagnostics": { "bp_cosine": [ 0.039823852479457855, 0.9718431234359741 ], "perturbation_rho": [ -0.031080063432455063, 0.03366325423121452 ], "nudging": { "0.001": [ -2.9830262064933777e-06, -5.972804501652718e-06 ], "0.003": [ -8.938834071159363e-06, -1.8277671188116074e-05 ], "0.01": [ -3.016670234501362e-05, -6.103678606450558e-05 ] }, "hidden_norms_per_layer": [ 5119.443359375, 277281.78125, 221633.09375 ], "bp_grad_norms_per_layer": [ 2.112029142153915e-05, 1.5214125141937984e-06, 1.2520613381639123e-06 ] }, "drift": { "embed.weight": 32.21887870916367, "embed.bias": 19.808534592837418, "blocks.0.ln.weight": 1.571417962464522, "blocks.0.w1.weight": 27.02065176792147, "blocks.0.w1.bias": 18.826762362229086, "blocks.0.w2.weight": 62.552517441341244, "blocks.1.ln.weight": 1.2403307752162702, "blocks.1.w1.weight": 19.856236759319962, "blocks.1.w1.bias": 14.207741981198222, "blocks.1.w2.weight": 38.602188965118806, "out_ln.weight": 0.442837113470192, "out_head.weight": 4.411774526619856, "out_head.bias": 11.444358576141303 } } }, "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": [ 7 ], "gpu": 0, "output_dir": "results/fa_dfa_d512_L2_seed7", "methods": [ "fa", "dfa" ], "random_targets": false, "penalty_lam": 0.0, "num_classes": 10 } }