{ "123": { "state_bridge": { "log": { "train_loss": [ 2.0394552979278564, 1.907211668624878, 1.8378597689056397, 1.8050992700958253, 1.782720252342224, 1.7607325444793702, 1.7526287609481812, 1.7301886476898194, 1.7114346005249024, 1.7017147624969482, 1.6891529846572877, 1.6735555848693848, 1.6678473961639404, 1.6606992713165283, 1.6562059018707276, 1.6442966329956055, 1.6339350330352784, 1.637065968284607, 1.6251955394744872, 1.6232860974502563, 1.62067741481781, 1.6153102271270752, 1.6100856274032593, 1.6128213213348388, 1.6058022644805907, 1.6015477989959717, 1.6040523609161377, 1.6024808090591431, 1.5986408697128296, 1.5955050112533569 ], "train_acc": [ 0.2546, 0.31232, 0.3415, 0.35616, 0.36698, 0.37548, 0.3775, 0.38916, 0.39518, 0.40052, 0.40288, 0.40992, 0.4117, 0.4142, 0.41554, 0.42002, 0.42486, 0.42292, 0.42918, 0.4291, 0.4284, 0.4346, 0.4362, 0.43588, 0.44048, 0.44034, 0.43952, 0.43918, 0.44066, 0.4431 ], "test_acc": [ 0.3128, 0.3548, 0.3591, 0.378, 0.3826, 0.3814, 0.4011, 0.4028, 0.404, 0.4254, 0.4276, 0.4332, 0.4345, 0.4356, 0.4316, 0.4384, 0.4423, 0.4425, 0.4369, 0.4439, 0.4447, 0.4442, 0.4483, 0.452, 0.453, 0.453, 0.4498, 0.4524, 0.4513, 0.4514 ], "state_pred_error": [ 0.2664960029602051, 0.17006848256111146, 0.1373988006401062, 0.13497620005130767, 0.1264824439716339, 0.11186170087099076, 0.10831025877714157, 0.10117165217876434, 0.0942336485338211, 0.09043296098470688, 0.08660082973480225, 0.08747579972743988, 0.07973947757720948, 0.07509488174915313, 0.07382626082420349, 0.07346862325191497, 0.07132561764001846, 0.07077178321242332, 0.06908800168275833, 0.06767414944648743, 0.06569002504706382, 0.062389732856750486, 0.05864977328658104, 0.05680723578810692, 0.053985064004659654, 0.05206361443519592, 0.0498943436563015, 0.047763368364572525, 0.04550520754933357, 0.043678019300699235 ] }, "diagnostics": { "bp_cosine": [ 0.42979174852371216, 0.37390586733818054, 0.32618582248687744, 0.28139200806617737 ], "perturbation_rho": [ 0.5135085582733154, 0.5129182934761047, 0.38893166184425354, 0.37062373757362366 ], "nudging": { "0.001": [ -0.0003920057788491249, -0.00025207665748894215, -0.00018884101882576942, -0.00014588376507163048 ], "0.003": [ -0.001175526063889265, -0.0007562476675957441, -0.000566466711461544, -0.00043755793012678623 ], "0.01": [ -0.003914414905011654, -0.002520129084587097, -0.0018878313712775707, -0.0014582867734134197 ] }, "hidden_norms_per_layer": [ 231.49375915527344, 238.22596740722656, 252.5166473388672, 278.00189208984375, 310.919189453125 ], "bp_grad_norms_per_layer": [ 0.000271676923148334, 0.00023371262068394572, 0.00020241527818143368, 0.000180525952600874, 0.00017415167530998588 ] }, "drift": { "embed.weight": 6.87082720778762, "embed.bias": 27.93003811703859, "blocks.0.ln.weight": 0.1618085354566574, "blocks.0.w1.weight": 3.191546927288011, "blocks.0.w1.bias": 6.668010298438182, "blocks.0.w2.weight": 15.615942220294206, "blocks.1.ln.weight": 0.12653203308582306, "blocks.1.w1.weight": 3.159400666127824, "blocks.1.w1.bias": 7.198150324919049, "blocks.1.w2.weight": 14.61006448474005, "blocks.2.ln.weight": 0.1170191541314125, "blocks.2.w1.weight": 3.2969716684544164, "blocks.2.w1.bias": 7.019388383996979, "blocks.2.w2.weight": 15.258183997524307, "blocks.3.ln.weight": 0.11643991619348526, "blocks.3.w1.weight": 3.5673772866278055, "blocks.3.w1.bias": 6.951775133615225, "blocks.3.w2.weight": 15.726364417935486, "out_ln.weight": 0.21612103283405304, "out_head.weight": 1.2186449227295404, "out_head.bias": 1.8007506807173053 } } }, "config": { "dataset": "cifar10", "d_hidden": 256, "num_blocks": 4, "batch_size": 128, "epochs": 30, "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/round38_sb_penalty_30ep_s123", "methods": [ "state_bridge" ], "random_targets": false, "penalty_lam": 0.01, "num_classes": 10 } }