{ "42": { "fa": { "log": { "train_loss": [ 2.0300785620117185, 1.9446230431365967, 1.923573359298706, 1.9025392212677001, 1.8942920899963378, 1.8785306893157958, 1.8725497973251344, 1.8588989192962646, 1.8521492172241212, 1.8486995751571655, 1.8387074700546264, 1.8285302072906495, 1.8229916802597046, 1.8140362118911744, 1.812324379043579, 1.8051523593902588, 1.8015315892791748, 1.7916892041015624, 1.7894653066253663, 1.7864193521118163, 1.7835033431243896, 1.7791855911636352, 1.7703584003448487, 1.7708648685073853, 1.7641783823394774, 1.762984627342224, 1.7599577627182006, 1.7553836069107056, 1.75658689201355, 1.7510007577896118, 1.7528975827026367, 1.7468154891586303, 1.7496728462982178, 1.7414471523666382, 1.7459090908050536, 1.7412950765228272, 1.7402052612304688, 1.7362829974365235, 1.734298306350708, 1.7386894550323486, 1.735369085960388, 1.7248860055541992, 1.7300375979232787, 1.7265542502212525, 1.72896113155365, 1.724715986366272, 1.723635640487671, 1.7220598682022095, 1.717545898704529, 1.7209024740600587, 1.7202868030548095, 1.716822794113159, 1.7148029877090454, 1.7123187323760987, 1.7099079937744142, 1.7115138537597656, 1.7089375814056396, 1.710836911201477, 1.7087473178863526, 1.7052376666641236, 1.7051094373321534, 1.7055147225952147, 1.7014220635986328, 1.6999273017120362, 1.6991235815811156, 1.697499723777771, 1.6986251516342163, 1.6985239101409912, 1.6973113412857055, 1.694503396911621, 1.6924839485549927, 1.6907986538314819, 1.6948804891204834, 1.6908944356536866, 1.694915372390747, 1.692836050567627, 1.690400951499939, 1.68655061794281, 1.6889203528213501, 1.686811109046936, 1.688070924682617, 1.6837623129653931, 1.684536149520874, 1.6821202702713012, 1.6863232943344115, 1.6847787243652343, 1.6850631763076782, 1.6828089485931397, 1.6857323441314698, 1.683035474510193, 1.6798774777221679, 1.6814900134658815, 1.6816223468017577, 1.6811535089874268, 1.681296463279724, 1.6815302163696288, 1.6837011975097655, 1.6801897742843628, 1.6809225379180908, 1.6781013634490967 ], "train_acc": [ 0.25312, 0.29106, 0.3008, 0.31326, 0.31492, 0.3233, 0.32324, 0.33074, 0.33364, 0.33658, 0.3413, 0.34538, 0.3485, 0.34936, 0.3499, 0.3504, 0.35564, 0.35788, 0.35942, 0.36092, 0.36132, 0.36218, 0.36636, 0.36582, 0.3677, 0.3675, 0.36926, 0.37292, 0.37044, 0.37244, 0.37438, 0.37436, 0.37378, 0.37712, 0.37622, 0.37694, 0.37552, 0.37782, 0.37806, 0.37674, 0.3797, 0.38284, 0.3805, 0.38056, 0.3796, 0.38304, 0.38512, 0.38384, 0.38492, 0.3875, 0.38534, 0.38444, 0.38538, 0.38828, 0.38914, 0.3886, 0.38882, 0.38966, 0.39026, 0.39132, 0.39252, 0.38896, 0.3926, 0.39232, 0.39498, 0.394, 0.39308, 0.39394, 0.39548, 0.39628, 0.39624, 0.3982, 0.39524, 0.39714, 0.39504, 0.39736, 0.39672, 0.39934, 0.397, 0.40178, 0.3973, 0.4012, 0.39838, 0.40118, 0.40006, 0.40216, 0.4005, 0.39964, 0.4008, 0.40088, 0.40064, 0.40226, 0.40106, 0.40206, 0.40272, 0.4009, 0.40232, 0.40004, 0.403, 0.4021 ], "test_acc": [ 0.2917, 0.3201, 0.3234, 0.3265, 0.3415, 0.3452, 0.341, 0.3599, 0.3605, 0.3541, 0.3715, 0.3712, 0.3656, 0.3709, 0.3852, 0.3723, 0.3799, 0.3724, 0.3741, 0.3908, 0.3881, 0.3848, 0.3868, 0.3732, 0.3985, 0.3937, 0.3954, 0.3986, 0.3997, 0.399, 0.3937, 0.3925, 0.4031, 0.4036, 0.4049, 0.4084, 0.4026, 0.3988, 0.3939, 0.3946, 0.4049, 0.4057, 0.4006, 0.4033, 0.402, 0.4098, 0.4052, 0.4075, 0.3987, 0.4165, 0.401, 0.4107, 0.3986, 0.4164, 0.4144, 0.4063, 0.4156, 0.4166, 0.4158, 0.4173, 0.4165, 0.4143, 0.4139, 0.4156, 0.4148, 0.4128, 0.4153, 0.4186, 0.4181, 0.4137, 0.4163, 0.4148, 0.418, 0.421, 0.4217, 0.4192, 0.4199, 0.4217, 0.4208, 0.4219, 0.4241, 0.4226, 0.4209, 0.4193, 0.4221, 0.4205, 0.4229, 0.42, 0.4224, 0.425, 0.4253, 0.4246, 0.4224, 0.4247, 0.4235, 0.4238, 0.4241, 0.4242, 0.4245, 0.4244 ] }, "diagnostics": { "bp_cosine": [ 0.007479430641978979, 0.03817027807235718, -0.1650964617729187, 0.9953471422195435 ], "perturbation_rho": [ -0.0640043318271637, 0.045080721378326416, -5.970895290374756e-05, -0.07074473798274994 ], "nudging": { "0.001": [ 4.5995693653821945e-07, -2.60770320892334e-07, 3.650784492492676e-07, -2.395769115537405e-06 ], "0.003": [ 1.9447761587798595e-06, -8.136266842484474e-07, 1.4320830814540386e-06, -8.26612813398242e-06 ], "0.01": [ 6.193062290549278e-06, -3.1341915018856525e-06, 4.9326918087899685e-06, -2.9165181331336498e-05 ] }, "hidden_norms_per_layer": [ 6685.263671875, 69715.734375, 615023.8125, 1362937.75, 466747.5 ], "bp_grad_norms_per_layer": [ 3.627526166383177e-05, 3.065474629693199e-06, 1.1288288987998385e-06, 1.124225377679977e-06, 1.1187978543603094e-06 ] }, "drift": { "embed.weight": 41.806395160952334, "embed.bias": 22.65316128534114, "blocks.0.ln.weight": 1.0245623407955453, "blocks.0.w1.weight": 14.934223802543457, "blocks.0.w1.bias": 15.12866030040346, "blocks.0.w2.weight": 53.87746375738443, "blocks.1.ln.weight": 0.9157411148330525, "blocks.1.w1.weight": 20.47192274795864, "blocks.1.w1.bias": 17.83477261022895, "blocks.1.w2.weight": 42.46486129463301, "blocks.2.ln.weight": 0.609426201836372, "blocks.2.w1.weight": 21.9031621576148, "blocks.2.w1.bias": 25.87166178551529, "blocks.2.w2.weight": 26.77134315272843, "blocks.3.ln.weight": 0.6636105283840906, "blocks.3.w1.weight": 21.84827836339499, "blocks.3.w1.bias": 23.312782840496634, "blocks.3.w2.weight": 37.76159247144327, "out_ln.weight": 0.2876820859790667, "out_head.weight": 5.782440518290608, "out_head.bias": 1.2881523166930118 } } }, "config": { "dataset": "cifar10", "d_hidden": 512, "num_blocks": 4, "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_L4", "methods": [ "fa" ], "random_targets": false, "penalty_lam": 0.0, "num_classes": 10 } }