summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p2_matched_s3.json
blob: 640517718d4820244e1d7f58a45793d61d3f443b (plain)
1
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 3, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p2_matched_s3", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.331878900527954, "cos_r_negg": [-0.02036810666322708, 0.011795876547694206, 0.0021237998735159636], "cos_innovation_negg": [-0.018753841519355774, 0.014596175402402878, 0.005968508310616016], "cos_apical_negg": [-0.02036810666322708, 0.011795875616371632, 0.0021237991750240326], "cos_Ac_negg": [-0.01875452697277069, 0.014595884829759598, 0.0059679653495550156], "r_norm": [4.867480278015137, 4.757315635681152, 4.860472202301025], "ldr": {"dL_sdil": 0.02511310577392578, "dL_bp": -0.2193007469177246, "ratio": -0.11451445618353276}}, {"step": 100, "epoch": 0, "train_loss": 0.4400538206100464, "cos_r_negg": [0.1756768822669983, 0.20653212070465088, 0.3905373811721802], "cos_innovation_negg": [0.31465643644332886, 0.3354240655899048, 0.7031124234199524], "cos_apical_negg": [0.1756768822669983, 0.20653212070465088, 0.3905373811721802], "cos_Ac_negg": [0.3145856261253357, 0.33537915349006653, 0.7031100392341614], "r_norm": [1.5759222507476807, 1.5576527118682861, 1.563110589981079]}, {"step": 200, "epoch": 0, "train_loss": 0.5143619775772095, "cos_r_negg": [0.15921232104301453, 0.17813655734062195, 0.24890244007110596], "cos_innovation_negg": [0.3653538227081299, 0.3882541060447693, 0.5765550136566162], "cos_apical_negg": [0.15921230614185333, 0.17813655734062195, 0.24890244007110596], "cos_Ac_negg": [0.36479735374450684, 0.38796794414520264, 0.5766723155975342], "r_norm": [1.2929868698120117, 1.2643277645111084, 1.292585849761963]}, {"step": 300, "epoch": 0, "train_loss": 0.47719043493270874, "cos_r_negg": [0.1595289707183838, 0.19067716598510742, 0.4023829400539398], "cos_innovation_negg": [0.32857978343963623, 0.40665096044540405, 0.7169294357299805], "cos_apical_negg": [0.1595289707183838, 0.19067716598510742, 0.4023829698562622], "cos_Ac_negg": [0.32851630449295044, 0.4066682457923889, 0.716888427734375], "r_norm": [1.0504348278045654, 1.0283509492874146, 1.0469398498535156]}, {"step": 400, "epoch": 0, "train_loss": 0.2309410572052002, "cos_r_negg": [0.13200967013835907, 0.22264397144317627, 0.35186904668807983], "cos_innovation_negg": [0.34929394721984863, 0.4233340620994568, 0.7414632439613342], "cos_apical_negg": [0.13200968503952026, 0.22264397144317627, 0.3518690764904022], "cos_Ac_negg": [0.3492230772972107, 0.4233376383781433, 0.7414527535438538], "r_norm": [0.9227346777915955, 0.9097601175308228, 0.9216660857200623]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9125, "test_loss": 0.34449742889404295}, {"step": 500, "epoch": 1, "train_loss": 0.3173186480998993, "cos_r_negg": [0.13482621312141418, 0.16621845960617065, 0.4622615575790405], "cos_innovation_negg": [0.342474102973938, 0.41030460596084595, 0.7856221199035645], "cos_apical_negg": [0.13482621312141418, 0.16621847450733185, 0.4622615575790405], "cos_Ac_negg": [0.3423904776573181, 0.4102475047111511, 0.7854601144790649], "r_norm": [0.8716844916343689, 0.8534141778945923, 0.8655869960784912], "ldr": {"dL_sdil": -0.03218638896942139, "dL_bp": -0.042296916246414185, "ratio": 0.7609630163558333}}, {"step": 600, "epoch": 1, "train_loss": 0.19914954900741577, "cos_r_negg": [0.10872647911310196, 0.15109533071517944, 0.38549095392227173], "cos_innovation_negg": [0.3493943512439728, 0.3854026794433594, 0.7403441071510315], "cos_apical_negg": [0.10872647166252136, 0.15109533071517944, 0.38549095392227173], "cos_Ac_negg": [0.3493119478225708, 0.38537198305130005, 0.7402387857437134], "r_norm": [0.7328153848648071, 0.725242555141449, 0.7366950511932373]}, {"step": 700, "epoch": 1, "train_loss": 0.26646754145622253, "cos_r_negg": [0.11888409405946732, 0.15967610478401184, 0.45279645919799805], "cos_innovation_negg": [0.32481685280799866, 0.4027654826641083, 0.7821754217147827], "cos_apical_negg": [0.11888409405946732, 0.15967610478401184, 0.45279648900032043], "cos_Ac_negg": [0.32472869753837585, 0.4027591347694397, 0.782071590423584], "r_norm": [0.8986352682113647, 0.8939275741577148, 0.9049227833747864]}, {"step": 800, "epoch": 1, "train_loss": 0.33199891448020935, "cos_r_negg": [0.08395060896873474, 0.11295019090175629, 0.07632268965244293], "cos_innovation_negg": [0.34351646900177, 0.3710211515426636, 0.6842052936553955], "cos_apical_negg": [0.08395060896873474, 0.11295019090175629, 0.07632268965244293], "cos_Ac_negg": [0.3434521555900574, 0.3709302246570587, 0.6840815544128418], "r_norm": [0.7469301223754883, 0.7383431196212769, 0.753665030002594]}, {"step": 900, "epoch": 1, "train_loss": 0.2762279808521271, "cos_r_negg": [0.12050512433052063, 0.13892970979213715, 0.03243357688188553], "cos_innovation_negg": [0.3857177495956421, 0.4192531108856201, 0.6353157162666321], "cos_apical_negg": [0.12050511687994003, 0.13892970979213715, 0.03243357688188553], "cos_Ac_negg": [0.38564231991767883, 0.4192121624946594, 0.6353316903114319], "r_norm": [0.7485752701759338, 0.7508242130279541, 0.760019838809967]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9348, "test_loss": 0.2599320869445801}, {"step": 1000, "epoch": 2, "train_loss": 0.14405804872512817, "cos_r_negg": [0.11138047277927399, 0.12647497653961182, 0.06470959633588791], "cos_innovation_negg": [0.3511101007461548, 0.3959304094314575, 0.6778622269630432], "cos_apical_negg": [0.11138047277927399, 0.12647497653961182, 0.06470959633588791], "cos_Ac_negg": [0.3510010838508606, 0.39592984318733215, 0.6778634786605835], "r_norm": [0.6931792497634888, 0.6907463073730469, 0.6990920305252075], "ldr": {"dL_sdil": -0.02318418025970459, "dL_bp": -0.031096190214157104, "ratio": 0.7455633664457575}}, {"step": 1100, "epoch": 2, "train_loss": 0.34493690729141235, "cos_r_negg": [0.08756481856107712, 0.11450334638357162, 0.29389017820358276], "cos_innovation_negg": [0.36921244859695435, 0.4279465973377228, 0.769111156463623], "cos_apical_negg": [0.08756481856107712, 0.11450335383415222, 0.29389017820358276], "cos_Ac_negg": [0.3691472113132477, 0.42785149812698364, 0.7690554857254028], "r_norm": [0.6190230846405029, 0.617702841758728, 0.6250365972518921]}, {"step": 1200, "epoch": 2, "train_loss": 0.29518693685531616, "cos_r_negg": [0.06379122287034988, 0.12758740782737732, 0.22313383221626282], "cos_innovation_negg": [0.378059983253479, 0.46550843119621277, 0.7720991373062134], "cos_apical_negg": [0.06379122287034988, 0.12758740782737732, 0.22313383221626282], "cos_Ac_negg": [0.37803614139556885, 0.46550166606903076, 0.7721033096313477], "r_norm": [0.5919125080108643, 0.5898166298866272, 0.5995679497718811]}, {"step": 1300, "epoch": 2, "train_loss": 0.274957537651062, "cos_r_negg": [0.08221554756164551, 0.10883848369121552, 0.296045184135437], "cos_innovation_negg": [0.3714882731437683, 0.4087797999382019, 0.6893191933631897], "cos_apical_negg": [0.08221554756164551, 0.10883848369121552, 0.296045184135437], "cos_Ac_negg": [0.3714684545993805, 0.4087407886981964, 0.6893448233604431], "r_norm": [0.7115012407302856, 0.7127417325973511, 0.7193174362182617]}, {"step": 1400, "epoch": 2, "train_loss": 0.26490557193756104, "cos_r_negg": [0.07762452960014343, 0.05591243505477905, 0.3536660075187683], "cos_innovation_negg": [0.37052151560783386, 0.43136483430862427, 0.7564950585365295], "cos_apical_negg": [0.07762452960014343, 0.05591243505477905, 0.3536660075187683], "cos_Ac_negg": [0.3696705400943756, 0.43082448840141296, 0.7561050653457642], "r_norm": [0.5983796715736389, 0.5932151675224304, 0.6052750945091248]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9375, "test_loss": 0.24186600036621095}, {"step": 1500, "epoch": 3, "train_loss": 0.28851747512817383, "cos_r_negg": [0.09779243171215057, 0.06952666491270065, 0.3150227963924408], "cos_innovation_negg": [0.37762948870658875, 0.4491156041622162, 0.7497493028640747], "cos_apical_negg": [0.09779242426156998, 0.06952666491270065, 0.3150227665901184], "cos_Ac_negg": [0.376889705657959, 0.4486525058746338, 0.7494707107543945], "r_norm": [0.5530264377593994, 0.5480079650878906, 0.5555835366249084], "ldr": {"dL_sdil": -0.014817297458648682, "dL_bp": -0.03878897428512573, "ratio": 0.38199766123567275}}, {"step": 1600, "epoch": 3, "train_loss": 0.3174806237220764, "cos_r_negg": [0.06319668143987656, 0.08834267407655716, 0.22097212076187134], "cos_innovation_negg": [0.3427238464355469, 0.4141557216644287, 0.7424756288528442], "cos_apical_negg": [0.06319668143987656, 0.08834266662597656, 0.22097212076187134], "cos_Ac_negg": [0.3426951766014099, 0.41411417722702026, 0.7424772381782532], "r_norm": [0.5841406583786011, 0.5867191553115845, 0.5884642601013184]}, {"step": 1700, "epoch": 3, "train_loss": 0.10985104739665985, "cos_r_negg": [0.08890071511268616, 0.10797759890556335, 0.3917781114578247], "cos_innovation_negg": [0.3356766104698181, 0.4258178472518921, 0.744598388671875], "cos_apical_negg": [0.08890072256326675, 0.10797760635614395, 0.3917781114578247], "cos_Ac_negg": [0.3357349634170532, 0.42589181661605835, 0.7445431351661682], "r_norm": [0.6230787634849548, 0.6204631328582764, 0.6286985874176025]}, {"step": 1800, "epoch": 3, "train_loss": 0.06924162060022354, "cos_r_negg": [0.11762216687202454, 0.12214531004428864, 0.47433561086654663], "cos_innovation_negg": [0.39117446541786194, 0.48938852548599243, 0.7135990262031555], "cos_apical_negg": [0.11762216687202454, 0.12214531004428864, 0.47433561086654663], "cos_Ac_negg": [0.3910622000694275, 0.4894335865974426, 0.7133069038391113], "r_norm": [0.7417702078819275, 0.7404168844223022, 0.7431856393814087]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9385, "test_loss": 0.25450943908691404}, {"step": 1900, "epoch": 4, "train_loss": 0.17704644799232483, "cos_r_negg": [0.08445397019386292, 0.08889074623584747, 0.3843984305858612], "cos_innovation_negg": [0.35447216033935547, 0.4475480914115906, 0.7383006811141968], "cos_apical_negg": [0.08445397019386292, 0.08889074623584747, 0.3843984603881836], "cos_Ac_negg": [0.3544359803199768, 0.4474848508834839, 0.7381622791290283], "r_norm": [0.6404048800468445, 0.6412316560745239, 0.6484839916229248]}, {"step": 2000, "epoch": 4, "train_loss": 0.21606126427650452, "cos_r_negg": [0.07945279777050018, 0.07109078019857407, 0.4097805917263031], "cos_innovation_negg": [0.33481740951538086, 0.4295274317264557, 0.7135540246963501], "cos_apical_negg": [0.07945279777050018, 0.07109078019857407, 0.4097805917263031], "cos_Ac_negg": [0.3350074887275696, 0.4299665093421936, 0.7135474681854248], "r_norm": [0.5689743757247925, 0.5581024885177612, 0.5706900358200073], "ldr": {"dL_sdil": -0.038610026240348816, "dL_bp": -0.08081431686878204, "ratio": 0.47776220521717455}}, {"step": 2100, "epoch": 4, "train_loss": 0.18260759115219116, "cos_r_negg": [0.07964015007019043, 0.09623073041439056, 0.34003013372421265], "cos_innovation_negg": [0.3523395359516144, 0.43928754329681396, 0.7320300340652466], "cos_apical_negg": [0.07964015007019043, 0.09623073041439056, 0.34003013372421265], "cos_Ac_negg": [0.3522557020187378, 0.4392588138580322, 0.7317688465118408], "r_norm": [0.5521249771118164, 0.544990062713623, 0.551296591758728]}, {"step": 2200, "epoch": 4, "train_loss": 0.11431927233934402, "cos_r_negg": [0.058552246540784836, 0.07483963668346405, 0.2631024122238159], "cos_innovation_negg": [0.3422366976737976, 0.45720887184143066, 0.7212303876876831], "cos_apical_negg": [0.05855224281549454, 0.07483963668346405, 0.2631024122238159], "cos_Ac_negg": [0.3421591818332672, 0.4572329521179199, 0.7209049463272095], "r_norm": [0.4327749013900757, 0.431601345539093, 0.4365268349647522]}, {"step": 2300, "epoch": 4, "train_loss": 0.3545019328594208, "cos_r_negg": [0.05145160108804703, 0.10118147730827332, 0.2988181412220001], "cos_innovation_negg": [0.37262725830078125, 0.5018264651298523, 0.7596726417541504], "cos_apical_negg": [0.05145159736275673, 0.10118147730827332, 0.2988181412220001], "cos_Ac_negg": [0.3725670576095581, 0.5017696022987366, 0.7596282362937927], "r_norm": [0.44180363416671753, 0.44251295924186707, 0.4432508945465088]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9426, "test_loss": 0.24033984069824219}, {"step": 2400, "epoch": 5, "train_loss": 0.13781969249248505, "cos_r_negg": [0.08198727667331696, 0.10721886157989502, 0.3741753101348877], "cos_innovation_negg": [0.32792022824287415, 0.4357667565345764, 0.8825811743736267], "cos_apical_negg": [0.08198726922273636, 0.10721886157989502, 0.3741753101348877], "cos_Ac_negg": [0.3278215527534485, 0.43569254875183105, 0.882398247718811], "r_norm": [0.568074107170105, 0.5682215690612793, 0.5739967823028564]}, {"step": 2500, "epoch": 5, "train_loss": 0.2105558216571808, "cos_r_negg": [0.07025723904371262, 0.05517081916332245, 0.37762144207954407], "cos_innovation_negg": [0.3441767692565918, 0.4478762149810791, 0.7895771265029907], "cos_apical_negg": [0.07025723904371262, 0.05517082288861275, 0.37762144207954407], "cos_Ac_negg": [0.34406840801239014, 0.4480290412902832, 0.7895047664642334], "r_norm": [0.4344387650489807, 0.429985910654068, 0.4351077973842621], "ldr": {"dL_sdil": -0.019099697470664978, "dL_bp": -0.04359477758407593, "ratio": 0.4381189337147029}}, {"step": 2600, "epoch": 5, "train_loss": 0.14259913563728333, "cos_r_negg": [0.06413869559764862, 0.08747546374797821, 0.3348381817340851], "cos_innovation_negg": [0.34231239557266235, 0.4611664414405823, 0.8132108449935913], "cos_apical_negg": [0.06413869559764862, 0.08747545629739761, 0.3348381817340851], "cos_Ac_negg": [0.3423386216163635, 0.4612843990325928, 0.8129971027374268], "r_norm": [0.46504828333854675, 0.4645615816116333, 0.46562665700912476]}, {"step": 2700, "epoch": 5, "train_loss": 0.14656184613704681, "cos_r_negg": [0.06414548307657242, 0.05336803197860718, 0.21878519654273987], "cos_innovation_negg": [0.3596385717391968, 0.41244661808013916, 0.7379372715950012], "cos_apical_negg": [0.06414547562599182, 0.05336803197860718, 0.21878519654273987], "cos_Ac_negg": [0.3595573902130127, 0.41238903999328613, 0.7376086711883545], "r_norm": [0.45414695143699646, 0.4527623951435089, 0.45553410053253174]}, {"step": 2800, "epoch": 5, "train_loss": 0.2994046211242676, "cos_r_negg": [0.05465333163738251, 0.07265643030405045, 0.2811175286769867], "cos_innovation_negg": [0.3468356132507324, 0.4211798310279846, 0.7703515887260437], "cos_apical_negg": [0.05465332791209221, 0.07265643775463104, 0.2811175286769867], "cos_Ac_negg": [0.3464857339859009, 0.42129141092300415, 0.7699546813964844], "r_norm": [0.449910968542099, 0.44745439291000366, 0.4500837028026581]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9482, "test_loss": 0.21917556610107422}, {"step": 2900, "epoch": 6, "train_loss": 0.13684412837028503, "cos_r_negg": [0.06713424623012543, 0.030654802918434143, 0.30837690830230713], "cos_innovation_negg": [0.39000511169433594, 0.4487501084804535, 0.8027445673942566], "cos_apical_negg": [0.06713424623012543, 0.03065480850636959, 0.30837690830230713], "cos_Ac_negg": [0.3899167776107788, 0.4486718773841858, 0.80238276720047], "r_norm": [0.4206772446632385, 0.42287734150886536, 0.42088812589645386]}, {"step": 3000, "epoch": 6, "train_loss": 0.32784920930862427, "cos_r_negg": [0.09284724295139313, 0.09827914834022522, 0.3178602457046509], "cos_innovation_negg": [0.3673955202102661, 0.49373292922973633, 0.7591451406478882], "cos_apical_negg": [0.09284724295139313, 0.09827914834022522, 0.3178602457046509], "cos_Ac_negg": [0.3669975697994232, 0.4936269223690033, 0.7590146064758301], "r_norm": [0.5538517832756042, 0.5461391806602478, 0.5484142303466797], "ldr": {"dL_sdil": -0.041023969650268555, "dL_bp": -0.07921682298183441, "ratio": 0.5178694134158341}}, {"step": 3100, "epoch": 6, "train_loss": 0.14292265474796295, "cos_r_negg": [0.06172025203704834, 0.05513174086809158, 0.3732379674911499], "cos_innovation_negg": [0.3735494613647461, 0.4711996912956238, 0.8251690864562988], "cos_apical_negg": [0.06172025203704834, 0.05513174831867218, 0.3732379674911499], "cos_Ac_negg": [0.3735364079475403, 0.471378356218338, 0.8250607848167419], "r_norm": [0.5026163458824158, 0.509120523929596, 0.5035586357116699]}, {"step": 3200, "epoch": 6, "train_loss": 0.1502373367547989, "cos_r_negg": [0.06354902684688568, 0.06215986981987953, 0.26972097158432007], "cos_innovation_negg": [0.31604406237602234, 0.3920234739780426, 0.8084297180175781], "cos_apical_negg": [0.06354902684688568, 0.06215986609458923, 0.26972097158432007], "cos_Ac_negg": [0.31607282161712646, 0.3919937014579773, 0.8079515695571899], "r_norm": [0.4039853811264038, 0.403599351644516, 0.4045742154121399]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9535, "test_loss": 0.2096079750061035}, {"step": 3300, "epoch": 7, "train_loss": 0.1335640698671341, "cos_r_negg": [0.06980779767036438, 0.09236127883195877, 0.15082693099975586], "cos_innovation_negg": [0.3417547345161438, 0.4821212589740753, 0.7918124794960022], "cos_apical_negg": [0.06980779767036438, 0.09236127883195877, 0.15082691609859467], "cos_Ac_negg": [0.34179019927978516, 0.4822649359703064, 0.791712760925293], "r_norm": [0.5176693201065063, 0.5247974395751953, 0.5229960680007935]}, {"step": 3400, "epoch": 7, "train_loss": 0.19798503816127777, "cos_r_negg": [0.07338063418865204, 0.06492847204208374, 0.113883838057518], "cos_innovation_negg": [0.3281344175338745, 0.4483921527862549, 0.8133120536804199], "cos_apical_negg": [0.07338063418865204, 0.06492847204208374, 0.113883838057518], "cos_Ac_negg": [0.3280527591705322, 0.4484575390815735, 0.8128377795219421], "r_norm": [0.4960020184516907, 0.4983512759208679, 0.4969604015350342]}, {"step": 3500, "epoch": 7, "train_loss": 0.15915080904960632, "cos_r_negg": [0.0944727435708046, 0.026904992759227753, 0.18848344683647156], "cos_innovation_negg": [0.3491503596305847, 0.44440147280693054, 0.7386237382888794], "cos_apical_negg": [0.0944727435708046, 0.026904992759227753, 0.18848344683647156], "cos_Ac_negg": [0.34957247972488403, 0.4445277154445648, 0.7377989888191223], "r_norm": [0.46674031019210815, 0.47091615200042725, 0.4683659076690674], "ldr": {"dL_sdil": -0.021384015679359436, "dL_bp": -0.0503864586353302, "ratio": 0.42440005228637556}}, {"step": 3600, "epoch": 7, "train_loss": 0.04387524724006653, "cos_r_negg": [0.046654507517814636, 0.010634372942149639, 0.11703355610370636], "cos_innovation_negg": [0.3120917081832886, 0.39470893144607544, 0.7540159225463867], "cos_apical_negg": [0.046654507517814636, 0.010634373873472214, 0.11703355610370636], "cos_Ac_negg": [0.3120144009590149, 0.3947693109512329, 0.75221848487854], "r_norm": [0.37566274404525757, 0.3760484457015991, 0.3757573366165161]}, {"step": 3700, "epoch": 7, "train_loss": 0.058575041592121124, "cos_r_negg": [0.04966064542531967, 0.04389099031686783, 0.22908291220664978], "cos_innovation_negg": [0.33947324752807617, 0.4016231298446655, 0.8086131811141968], "cos_apical_negg": [0.04966064915060997, 0.04389099404215813, 0.22908291220664978], "cos_Ac_negg": [0.3395770490169525, 0.4018130600452423, 0.8080916404724121], "r_norm": [0.3791559040546417, 0.37757930159568787, 0.38096868991851807]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9456, "test_loss": 0.24103100204467773}, {"step": 3800, "epoch": 8, "train_loss": 0.19025324285030365, "cos_r_negg": [0.08102741092443466, 0.01857561618089676, 0.3476852476596832], "cos_innovation_negg": [0.3338545858860016, 0.43275606632232666, 0.820645809173584], "cos_apical_negg": [0.08102741092443466, 0.01857561245560646, 0.34768521785736084], "cos_Ac_negg": [0.33342835307121277, 0.4327048361301422, 0.8198139667510986], "r_norm": [0.39218252897262573, 0.3957356810569763, 0.3993185758590698]}, {"step": 3900, "epoch": 8, "train_loss": 0.34581154584884644, "cos_r_negg": [0.054648131132125854, 0.05118422955274582, 0.26381927728652954], "cos_innovation_negg": [0.3540130853652954, 0.4399789571762085, 0.9029856324195862], "cos_apical_negg": [0.054648131132125854, 0.05118422955274582, 0.26381930708885193], "cos_Ac_negg": [0.3538229465484619, 0.44010695815086365, 0.9017641544342041], "r_norm": [0.4435705840587616, 0.4424494504928589, 0.4460024833679199]}, {"step": 4000, "epoch": 8, "train_loss": 0.11512741446495056, "cos_r_negg": [0.06321237981319427, 0.06220477819442749, 0.2510380744934082], "cos_innovation_negg": [0.360689640045166, 0.4856993854045868, 0.8238543272018433], "cos_apical_negg": [0.06321237981319427, 0.06220477446913719, 0.2510380744934082], "cos_Ac_negg": [0.3610338270664215, 0.4859244227409363, 0.8238083720207214], "r_norm": [0.3842754364013672, 0.3864317834377289, 0.3872900605201721], "ldr": {"dL_sdil": -0.03004211187362671, "dL_bp": -0.07631881535053253, "ratio": 0.39363965144955676}}, {"step": 4100, "epoch": 8, "train_loss": 0.0921652764081955, "cos_r_negg": [0.06788378953933716, 0.05737902969121933, 0.19586181640625], "cos_innovation_negg": [0.35160183906555176, 0.4368787109851837, 0.7348260879516602], "cos_apical_negg": [0.06788378953933716, 0.05737902969121933, 0.1958618015050888], "cos_Ac_negg": [0.35199347138404846, 0.43733251094818115, 0.7354164123535156], "r_norm": [0.4068611264228821, 0.4173135757446289, 0.4115045964717865]}, {"step": 4200, "epoch": 8, "train_loss": 0.21099771559238434, "cos_r_negg": [0.03902527317404747, 0.05180132016539574, 0.19548070430755615], "cos_innovation_negg": [0.38638246059417725, 0.4414784610271454, 0.7923601865768433], "cos_apical_negg": [0.03902527689933777, 0.05180132016539574, 0.19548071920871735], "cos_Ac_negg": [0.38578999042510986, 0.4416012167930603, 0.7907538414001465], "r_norm": [0.39080530405044556, 0.38810592889785767, 0.391198992729187]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9437, "test_loss": 0.24913544845581054}, {"step": 4300, "epoch": 9, "train_loss": 0.10257379710674286, "cos_r_negg": [0.0578710101544857, 0.06124154105782509, 0.16410021483898163], "cos_innovation_negg": [0.3881415128707886, 0.47438478469848633, 0.8227272033691406], "cos_apical_negg": [0.0578710101544857, 0.06124153360724449, 0.16410021483898163], "cos_Ac_negg": [0.3880169093608856, 0.47455883026123047, 0.8224486708641052], "r_norm": [0.43919825553894043, 0.43225860595703125, 0.4337231516838074]}, {"step": 4400, "epoch": 9, "train_loss": 0.10444098711013794, "cos_r_negg": [0.03795100003480911, 0.03195874020457268, 0.06112508475780487], "cos_innovation_negg": [0.3105248808860779, 0.38150686025619507, 0.8311172723770142], "cos_apical_negg": [0.03795100003480911, 0.03195874020457268, 0.06112508103251457], "cos_Ac_negg": [0.31050512194633484, 0.3816024959087372, 0.8309045433998108], "r_norm": [0.3962266445159912, 0.39936453104019165, 0.395593523979187]}, {"step": 4500, "epoch": 9, "train_loss": 0.2180986851453781, "cos_r_negg": [0.035004157572984695, 0.05222100764513016, 0.005726968869566917], "cos_innovation_negg": [0.3067225217819214, 0.40333491563796997, 0.8238222599029541], "cos_apical_negg": [0.0350041538476944, 0.052221011370420456, 0.005726972594857216], "cos_Ac_negg": [0.30659380555152893, 0.403276264667511, 0.8236179351806641], "r_norm": [0.32833749055862427, 0.3300885558128357, 0.3319564461708069], "ldr": {"dL_sdil": -0.007901772856712341, "dL_bp": -0.02477891743183136, "ratio": 0.31889096359640023}}, {"step": 4600, "epoch": 9, "train_loss": 0.09447373449802399, "cos_r_negg": [0.048628177493810654, 0.010213809087872505, 0.12252732366323471], "cos_innovation_negg": [0.32184451818466187, 0.4015880525112152, 0.7796953320503235], "cos_apical_negg": [0.048628177493810654, 0.010213812813162804, 0.12252731621265411], "cos_Ac_negg": [0.32202592492103577, 0.4013347625732422, 0.7796410322189331], "r_norm": [0.3619610667228699, 0.3635615408420563, 0.3631513714790344]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.957, "test_loss": 0.20237664337158204}, {"step": 4700, "epoch": 10, "train_loss": 0.12431761622428894, "cos_r_negg": [0.05357087776064873, 0.04383087903261185, 0.1457587331533432], "cos_innovation_negg": [0.3576279878616333, 0.4251352548599243, 0.831009030342102], "cos_apical_negg": [0.05357087403535843, 0.04383087903261185, 0.1457587331533432], "cos_Ac_negg": [0.3574097156524658, 0.4258958697319031, 0.8296326398849487], "r_norm": [0.29066699743270874, 0.2902194857597351, 0.2891004681587219]}, {"step": 4800, "epoch": 10, "train_loss": 0.16371572017669678, "cos_r_negg": [0.029036926105618477, 0.004269483499228954, 0.19963490962982178], "cos_innovation_negg": [0.32341933250427246, 0.4272913336753845, 0.8171665668487549], "cos_apical_negg": [0.029036924242973328, 0.004269478842616081, 0.19963490962982178], "cos_Ac_negg": [0.3231048882007599, 0.42749524116516113, 0.813264012336731], "r_norm": [0.285106360912323, 0.28631317615509033, 0.2873285114765167]}, {"step": 4900, "epoch": 10, "train_loss": 0.14902347326278687, "cos_r_negg": [0.04620438814163208, 0.02644183300435543, 0.15327578783035278], "cos_innovation_negg": [0.3307992219924927, 0.44949227571487427, 0.8080215454101562], "cos_apical_negg": [0.04620438814163208, 0.026441829279065132, 0.15327578783035278], "cos_Ac_negg": [0.3309200704097748, 0.4503331780433655, 0.8067384362220764], "r_norm": [0.3732837438583374, 0.3736773729324341, 0.37527549266815186]}, {"step": 5000, "epoch": 10, "train_loss": 0.09991974383592606, "cos_r_negg": [0.049699798226356506, 0.08670321106910706, 0.07492829859256744], "cos_innovation_negg": [0.3452371656894684, 0.4075286090373993, 0.813042402267456], "cos_apical_negg": [0.049699798226356506, 0.08670320361852646, 0.07492829859256744], "cos_Ac_negg": [0.34534674882888794, 0.40769147872924805, 0.8129086494445801], "r_norm": [0.3537364900112152, 0.3528369069099426, 0.3551481366157532], "ldr": {"dL_sdil": -0.017207548022270203, "dL_bp": -0.04944072663784027, "ratio": 0.3480439951523715}}, {"step": 5100, "epoch": 10, "train_loss": 0.12895873188972473, "cos_r_negg": [0.04770016297698021, 0.03193221986293793, 0.21643300354480743], "cos_innovation_negg": [0.3295574486255646, 0.42878806591033936, 0.8419816493988037], "cos_apical_negg": [0.04770016297698021, 0.031932227313518524, 0.21643300354480743], "cos_Ac_negg": [0.3295457363128662, 0.42891111969947815, 0.8411244750022888], "r_norm": [0.3373553156852722, 0.33803504705429077, 0.3393421769142151]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.951, "test_loss": 0.23327355194091798}, {"step": 5200, "epoch": 11, "train_loss": 0.08230389654636383, "cos_r_negg": [0.03477941453456879, 0.005066273733973503, 0.15518972277641296], "cos_innovation_negg": [0.310738742351532, 0.4165608882904053, 0.8202950954437256], "cos_apical_negg": [0.034779418259859085, 0.005066270008683205, 0.15518973767757416], "cos_Ac_negg": [0.31071263551712036, 0.4168030321598053, 0.8189512491226196], "r_norm": [0.3066258430480957, 0.30801162123680115, 0.3093065619468689]}, {"step": 5300, "epoch": 11, "train_loss": 0.14475546777248383, "cos_r_negg": [0.03744753822684288, 0.0545089952647686, 0.29369843006134033], "cos_innovation_negg": [0.3429303467273712, 0.4583066403865814, 0.8737197518348694], "cos_apical_negg": [0.03744753450155258, 0.054508984088897705, 0.29369840025901794], "cos_Ac_negg": [0.342940092086792, 0.4582170844078064, 0.8718173503875732], "r_norm": [0.35003840923309326, 0.35217857360839844, 0.3526148796081543]}, {"step": 5400, "epoch": 11, "train_loss": 0.2559114694595337, "cos_r_negg": [0.02582029439508915, 0.04114362597465515, 0.13393956422805786], "cos_innovation_negg": [0.30870330333709717, 0.4293370544910431, 0.8331000804901123], "cos_apical_negg": [0.025820299983024597, 0.04114362597465515, 0.13393956422805786], "cos_Ac_negg": [0.30875164270401, 0.429548978805542, 0.8332539796829224], "r_norm": [0.3590816557407379, 0.35885876417160034, 0.3555552363395691]}, {"step": 5500, "epoch": 11, "train_loss": 0.1681821644306183, "cos_r_negg": [0.04578680545091629, 0.05125431716442108, 0.0690905898809433], "cos_innovation_negg": [0.2915537357330322, 0.41783082485198975, 0.7715979218482971], "cos_apical_negg": [0.04578680545091629, 0.05125431716442108, 0.0690905898809433], "cos_Ac_negg": [0.29153886437416077, 0.41815367341041565, 0.7711772918701172], "r_norm": [0.3617756962776184, 0.36520957946777344, 0.35974299907684326], "ldr": {"dL_sdil": -0.015750110149383545, "dL_bp": -0.03815075755119324, "ratio": 0.41283872615763906}}, {"step": 5600, "epoch": 11, "train_loss": 0.10013377666473389, "cos_r_negg": [0.04685045778751373, 0.05785603076219559, 0.2090340405702591], "cos_innovation_negg": [0.35300010442733765, 0.4542841911315918, 0.7812929153442383], "cos_apical_negg": [0.04685045778751373, 0.05785602703690529, 0.2090340405702591], "cos_Ac_negg": [0.35341233015060425, 0.4559386968612671, 0.7789859771728516], "r_norm": [0.4355519711971283, 0.43699824810028076, 0.4287186861038208]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9507, "test_loss": 0.23671178817749022}, {"step": 5700, "epoch": 12, "train_loss": 0.036010392010211945, "cos_r_negg": [0.007591079920530319, -0.0011415127664804459, 0.048039890825748444], "cos_innovation_negg": [0.273660272359848, 0.4013635516166687, 0.8559930324554443], "cos_apical_negg": [0.007591081317514181, -0.0011415127664804459, 0.048039887100458145], "cos_Ac_negg": [0.2741205394268036, 0.4025787115097046, 0.855581521987915], "r_norm": [0.26295018196105957, 0.2653757929801941, 0.2609485685825348]}, {"step": 5800, "epoch": 12, "train_loss": 0.06087285280227661, "cos_r_negg": [0.030401382595300674, 0.002562267705798149, 0.24839895963668823], "cos_innovation_negg": [0.326499342918396, 0.4764612913131714, 0.8363528251647949], "cos_apical_negg": [0.030401380732655525, 0.002562267705798149, 0.24839895963668823], "cos_Ac_negg": [0.32592064142227173, 0.4774336516857147, 0.8326979875564575], "r_norm": [0.35251328349113464, 0.3532651960849762, 0.352918803691864]}, {"step": 5900, "epoch": 12, "train_loss": 0.0821932852268219, "cos_r_negg": [0.021442102268338203, 0.015266858972609043, 0.20375409722328186], "cos_innovation_negg": [0.3404744863510132, 0.4297502636909485, 0.763418436050415], "cos_apical_negg": [0.021442092955112457, 0.015266858041286469, 0.20375409722328186], "cos_Ac_negg": [0.34027743339538574, 0.4302564263343811, 0.7633066773414612], "r_norm": [0.3595448434352875, 0.3570975065231323, 0.3590535521507263]}, {"step": 6000, "epoch": 12, "train_loss": 0.008466094732284546, "cos_r_negg": [0.025640254840254784, 0.03812666982412338, 0.1619781255722046], "cos_innovation_negg": [0.34594208002090454, 0.43582791090011597, 0.7938475608825684], "cos_apical_negg": [0.025640252977609634, 0.038126666098833084, 0.1619781255722046], "cos_Ac_negg": [0.3454383611679077, 0.43601155281066895, 0.7889047861099243], "r_norm": [0.31336599588394165, 0.3176407814025879, 0.3096732497215271], "ldr": {"dL_sdil": -0.018425673246383667, "dL_bp": -0.049481213092803955, "ratio": 0.37237715275544264}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9525, "test_loss": 0.2193112091064453}, {"step": 6100, "epoch": 13, "train_loss": 0.11159200966358185, "cos_r_negg": [0.0439818799495697, 0.032109104096889496, 0.18376143276691437], "cos_innovation_negg": [0.32291585206985474, 0.48050880432128906, 0.8388372659683228], "cos_apical_negg": [0.0439818799495697, 0.032109104096889496, 0.18376144766807556], "cos_Ac_negg": [0.32420697808265686, 0.48160845041275024, 0.8389242887496948], "r_norm": [0.39198821783065796, 0.3899760842323303, 0.38981926441192627]}, {"step": 6200, "epoch": 13, "train_loss": 0.13688595592975616, "cos_r_negg": [0.03425388038158417, 0.0277108121663332, 0.1322983354330063], "cos_innovation_negg": [0.3180942237377167, 0.4551396667957306, 0.7792147397994995], "cos_apical_negg": [0.034253884106874466, 0.02771081030368805, 0.1322983354330063], "cos_Ac_negg": [0.31836777925491333, 0.4565110206604004, 0.7771570682525635], "r_norm": [0.3241232931613922, 0.3264669179916382, 0.32752513885498047]}, {"step": 6300, "epoch": 13, "train_loss": 0.0561244897544384, "cos_r_negg": [0.0288222748786211, 0.032034892588853836, 0.15862321853637695], "cos_innovation_negg": [0.2943158745765686, 0.3951963782310486, 0.8620610237121582], "cos_apical_negg": [0.02882227674126625, 0.032034896314144135, 0.15862321853637695], "cos_Ac_negg": [0.2938181161880493, 0.3969143331050873, 0.8599331974983215], "r_norm": [0.320553183555603, 0.3219207227230072, 0.3195515275001526]}, {"step": 6400, "epoch": 13, "train_loss": 0.04703892767429352, "cos_r_negg": [0.03496722877025604, 0.030867431312799454, 0.18513604998588562], "cos_innovation_negg": [0.3197341859340668, 0.43913203477859497, 0.8335174918174744], "cos_apical_negg": [0.03496722877025604, 0.030867429450154305, 0.18513604998588562], "cos_Ac_negg": [0.321494996547699, 0.44213366508483887, 0.8298313617706299], "r_norm": [0.3697455823421478, 0.3741699457168579, 0.36658036708831787]}, {"step": 6500, "epoch": 13, "train_loss": 0.016488533467054367, "cos_r_negg": [0.052222929894924164, 0.06285339593887329, 0.13343417644500732], "cos_innovation_negg": [0.38896992802619934, 0.49758970737457275, 0.8522046804428101], "cos_apical_negg": [0.052222929894924164, 0.06285339593887329, 0.13343417644500732], "cos_Ac_negg": [0.3891015648841858, 0.4979743957519531, 0.8522868156433105], "r_norm": [0.3370938301086426, 0.3465665280818939, 0.3368062376976013], "ldr": {"dL_sdil": -0.024978965520858765, "dL_bp": -0.07083387672901154, "ratio": 0.35264151383977677}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9596, "test_loss": 0.1998619110107422}, {"step": 6600, "epoch": 14, "train_loss": 0.04038822650909424, "cos_r_negg": [0.026838652789592743, 0.051512375473976135, 0.08026068657636642], "cos_innovation_negg": [0.35957109928131104, 0.4205498993396759, 0.8914711475372314], "cos_apical_negg": [0.02683865651488304, 0.05151237174868584, 0.08026069402694702], "cos_Ac_negg": [0.3585135340690613, 0.4217907786369324, 0.8914788961410522], "r_norm": [0.31473827362060547, 0.3170691728591919, 0.3125041723251343]}, {"step": 6700, "epoch": 14, "train_loss": 0.02604137733578682, "cos_r_negg": [0.036674678325653076, 0.07491622865200043, 0.1431065797805786], "cos_innovation_negg": [0.3476559519767761, 0.40931612253189087, 0.8611284494400024], "cos_apical_negg": [0.036674678325653076, 0.07491622120141983, 0.1431065797805786], "cos_Ac_negg": [0.3479863405227661, 0.41027823090553284, 0.8615411520004272], "r_norm": [0.3823716938495636, 0.3841034770011902, 0.3771008849143982]}, {"step": 6800, "epoch": 14, "train_loss": 0.08187447488307953, "cos_r_negg": [0.03891880437731743, 0.027264870703220367, 0.03262750431895256], "cos_innovation_negg": [0.30911949276924133, 0.38232144713401794, 0.85636305809021], "cos_apical_negg": [0.03891880065202713, 0.02726486697793007, 0.032627515494823456], "cos_Ac_negg": [0.3091888427734375, 0.3834376633167267, 0.8598622679710388], "r_norm": [0.3174425959587097, 0.3199201822280884, 0.31509917974472046]}, {"step": 6900, "epoch": 14, "train_loss": 0.07461591809988022, "cos_r_negg": [0.044822920113801956, 0.058030031621456146, 0.08663206547498703], "cos_innovation_negg": [0.3657809793949127, 0.42271316051483154, 0.8233124017715454], "cos_apical_negg": [0.044822920113801956, 0.058030035346746445, 0.08663207292556763], "cos_Ac_negg": [0.36638933420181274, 0.42316633462905884, 0.8230905532836914], "r_norm": [0.3332439064979553, 0.3347223103046417, 0.33284109830856323]}, {"step": 7000, "epoch": 14, "train_loss": 0.06315820664167404, "cos_r_negg": [0.04397159069776535, 0.02858033776283264, 0.04650462418794632], "cos_innovation_negg": [0.3464168310165405, 0.4357408881187439, 0.8297408819198608], "cos_apical_negg": [0.04397159069776535, 0.028580334037542343, 0.046504613012075424], "cos_Ac_negg": [0.34667283296585083, 0.4369245767593384, 0.8286241292953491], "r_norm": [0.3260006308555603, 0.33002299070358276, 0.32008862495422363], "ldr": {"dL_sdil": -0.014171004295349121, "dL_bp": -0.04084871709346771, "ratio": 0.3469143048709176}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9588, "test_loss": 0.20080369186401367}], "final": {"test_acc": 0.9588, "test_loss": 0.20080369186401367, "wall_s": 25.22831654548645, "cos_r_negg": [0.03533933311700821, 0.05207996815443039, 0.037533730268478394], "cos_innovation_negg": [0.33997786045074463, 0.41935911774635315, 0.8430595993995667], "cos_apical_negg": [0.03533933684229851, 0.052079975605010986, 0.03753373771905899], "cos_Ac_negg": [0.34118103981018066, 0.4210580885410309, 0.8421542644500732]}}