{ "456": { "state_bridge": { "log": { "train_loss": [ 2.0369125997924806, 1.9002794314575195, 1.839365246887207, 1.8127399161148072, 1.7898970299911499, 1.7758199447631835, 1.7543362093353272, 1.7370687340545654, 1.7275934963989257, 1.7085617017364503, 1.6948483248138428, 1.6850648455429078, 1.6752807913208008, 1.6712003810882567, 1.6620737231063842, 1.6546397353744506, 1.6401427822113037, 1.6391044388580323, 1.630859147644043, 1.6259963637542725, 1.6214555887603759, 1.6194580511856078, 1.613519859046936, 1.6123445831680299, 1.6089352087783813, 1.6038804431915283, 1.6018304259872436, 1.6036242258453368, 1.6006233197784423, 1.5980215132904052 ], "train_acc": [ 0.25408, 0.31238, 0.3414, 0.35388, 0.36162, 0.368, 0.37992, 0.38372, 0.38852, 0.39432, 0.40126, 0.40466, 0.40538, 0.40918, 0.4099, 0.41604, 0.4201, 0.42306, 0.42332, 0.42836, 0.4309, 0.4304, 0.43168, 0.43282, 0.43744, 0.43726, 0.4399, 0.43702, 0.43884, 0.44118 ], "test_acc": [ 0.3104, 0.3506, 0.3735, 0.3792, 0.3887, 0.3929, 0.3976, 0.3987, 0.3979, 0.4104, 0.4137, 0.4198, 0.4346, 0.4251, 0.419, 0.4391, 0.4347, 0.4468, 0.4455, 0.4459, 0.4411, 0.447, 0.4562, 0.455, 0.4521, 0.4514, 0.4522, 0.4531, 0.4519, 0.4509 ], "state_pred_error": [ 0.27555840564250944, 0.16246181419372557, 0.14625693091392516, 0.13436604994773865, 0.12426637276172638, 0.11355017890930176, 0.10647282010555267, 0.10073854411125183, 0.09217885914564132, 0.08755232047796249, 0.08124397647857666, 0.08034932177066803, 0.07723198721170425, 0.07435347493171691, 0.07414377628564835, 0.07049065297842026, 0.0675844569182396, 0.06546888866901397, 0.06477154892683029, 0.06301552970647811, 0.059929636999368666, 0.05725809707641601, 0.05675601554989815, 0.056774080897569655, 0.05566476465702057, 0.05161277118682861, 0.0487107616519928, 0.04572098425865173, 0.04381731255412102, 0.04239982622146606 ] }, "diagnostics": { "bp_cosine": [ 0.42256492376327515, 0.38477009534835815, 0.31862324476242065, 0.27332931756973267 ], "perturbation_rho": [ 0.5183612704277039, 0.45182135701179504, 0.3981747329235077, 0.32433199882507324 ], "nudging": { "0.001": [ -0.00040302483830600977, -0.0002813579048961401, -0.00019051437266170979, -0.00014453462790697813 ], "0.003": [ -0.001208781497552991, -0.0008438715012744069, -0.000571400043554604, -0.000433529494330287 ], "0.01": [ -0.004026009701192379, -0.0028109778650105, -0.001904117758385837, -0.0014447440626099706 ] }, "hidden_norms_per_layer": [ 234.2797088623047, 243.28529357910156, 250.8776092529297, 268.8158264160156, 291.8413391113281 ], "bp_grad_norms_per_layer": [ 0.0002783545642159879, 0.0002534560626372695, 0.0002214369596913457, 0.00020169034542050213, 0.00019217752560507506 ] }, "drift": { "embed.weight": 6.880160568400241, "embed.bias": 30.878601323854713, "blocks.0.ln.weight": 0.14636263251304626, "blocks.0.w1.weight": 3.0733067420757636, "blocks.0.w1.bias": 6.555792191121209, "blocks.0.w2.weight": 14.035427054326416, "blocks.1.ln.weight": 0.13672301173210144, "blocks.1.w1.weight": 3.124442427959728, "blocks.1.w1.bias": 7.090988357162414, "blocks.1.w2.weight": 14.513611759686587, "blocks.2.ln.weight": 0.13195334374904633, "blocks.2.w1.weight": 3.298519491819089, "blocks.2.w1.bias": 7.524602101872219, "blocks.2.w2.weight": 15.233467131609643, "blocks.3.ln.weight": 0.13041938841342926, "blocks.3.w1.weight": 3.472952946259165, "blocks.3.w1.bias": 7.358409158987916, "blocks.3.w2.weight": 14.931264123351346, "out_ln.weight": 0.2145024985074997, "out_head.weight": 1.1640670176605314, "out_head.bias": 1.760523224597371 } } }, "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": [ 456 ], "gpu": 0, "output_dir": "results/round38_sb_penalty_30ep_s456", "methods": [ "state_bridge" ], "random_targets": false, "penalty_lam": 0.01, "num_classes": 10 } }