{ "42": { "state_bridge": { "log": { "train_loss": [ 2.0474753201293945, 1.9117268241119385 ], "train_acc": [ 0.24662, 0.30954 ], "test_acc": [ 0.273, 0.3501 ], "state_pred_error": [ 0.24535170847892762, 0.13223054832935333 ] }, "diagnostics": { "bp_cosine": [ 0.3206905722618103, 0.2752090096473694, 0.24660587310791016, 0.22906462848186493 ], "perturbation_rho": [ 0.4074876010417938, 0.2618027925491333, 0.26648417115211487, 0.24533966183662415 ], "nudging": { "0.001": [ -0.00024993112310767174, -0.00019314652308821678, -0.00016400963068008423, -0.00015044421888887882 ], "0.003": [ -0.0007498480845242739, -0.0005794172175228596, -0.0004920377396047115, -0.00045136199332773685 ], "0.01": [ -0.0025002574548125267, -0.001930799102410674, -0.001639701658859849, -0.0015042978338897228 ] }, "hidden_norms_per_layer": [ 197.37428283691406, 208.8948211669922, 213.67898559570312, 225.69448852539062, 229.3214111328125 ], "bp_grad_norms_per_layer": [ 0.0002925087173935026, 0.00027807100559584796, 0.0002793201128952205, 0.0002787425764836371, 0.00027831399347633123 ] }, "drift": { "embed.weight": 3.0943938196002705, "embed.bias": 5.276486174790121, "blocks.0.ln.weight": 0.041422076523303986, "blocks.0.w1.weight": 1.2334240308956859, "blocks.0.w1.bias": 1.7498680169137257, "blocks.0.w2.weight": 4.629686685849745, "blocks.1.ln.weight": 0.0331297293305397, "blocks.1.w1.weight": 1.1419210278459624, "blocks.1.w1.bias": 1.3133909085501447, "blocks.1.w2.weight": 4.0167023654211755, "blocks.2.ln.weight": 0.036234382539987564, "blocks.2.w1.weight": 1.0949732338041316, "blocks.2.w1.bias": 1.347265776948355, "blocks.2.w2.weight": 3.8533434905704533, "blocks.3.ln.weight": 0.04701152443885803, "blocks.3.w1.weight": 1.0335121836652872, "blocks.3.w1.bias": 1.4470639304699096, "blocks.3.w2.weight": 3.5886411268756073, "out_ln.weight": 0.035340577363967896, "out_head.weight": 0.5555762293577677, "out_head.bias": 1.3793257927193587 } }, "credit_bridge": { "log": { "train_loss": [ 1.9963660776519776, 2.0451175025177 ], "train_acc": [ 0.2772, 0.26838 ], "test_acc": [ 0.3299, 0.2827 ], "value_loss": [ 0.666878916387558, 0.47583423060417174 ] }, "diagnostics": { "bp_cosine": [ 0.03980115056037903, 0.079774409532547, 0.1276342272758484, 0.15156477689743042 ], "perturbation_rho": [ 0.04505116492509842, 0.14404122531414032, 0.17026099562644958, 0.15786880254745483 ], "nudging": { "0.001": [ -4.916917532682419e-06, -1.4375429600477219e-05, -1.966068521142006e-05, -2.3062806576490402e-05 ], "0.003": [ -1.4716759324073792e-05, -4.31169755756855e-05, -5.895830690860748e-05, -6.919307634234428e-05 ], "0.01": [ -4.905415698885918e-05, -0.00014382274821400642, -0.00019651977345347404, -0.00023055588826537132 ] }, "hidden_norms_per_layer": [ 1009.140380859375, 1082.5697021484375, 1145.231201171875, 1182.99951171875, 1257.748046875 ], "bp_grad_norms_per_layer": [ 6.559405301231891e-05, 6.734106864314526e-05, 6.742271943949163e-05, 6.65746774757281e-05, 6.616810424020514e-05 ] }, "drift": { "embed.weight": 7.564059143620495, "embed.bias": 8.746969303910673, "blocks.0.ln.weight": 0.11126699298620224, "blocks.0.w1.weight": 1.9535038516346213, "blocks.0.w1.bias": 1.6108819536646128, "blocks.0.w2.weight": 7.201445067417667, "blocks.1.ln.weight": 0.11589338630437851, "blocks.1.w1.weight": 1.871920531155357, "blocks.1.w1.bias": 1.7519376093645112, "blocks.1.w2.weight": 6.716478346434812, "blocks.2.ln.weight": 0.11303362250328064, "blocks.2.w1.weight": 1.9497895886204863, "blocks.2.w1.bias": 1.6654855654120733, "blocks.2.w2.weight": 6.897888781603305, "blocks.3.ln.weight": 0.12764930725097656, "blocks.3.w1.weight": 2.1127271785838393, "blocks.3.w1.bias": 1.8058662035226984, "blocks.3.w2.weight": 7.331784125259354, "out_ln.weight": 0.041299086064100266, "out_head.weight": 0.9332516761578312, "out_head.bias": 1.5418626566795017 } } }, "config": { "dataset": "cifar10", "d_hidden": 256, "num_blocks": 4, "batch_size": 128, "epochs": 2, "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/round38_smoke_sbcb_pen", "methods": [ "state_bridge", "credit_bridge" ], "random_targets": false, "penalty_lam": 0.01, "num_classes": 10 } }