summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p2_residual_s3.json
blob: d0f7850e81567f8f152e3cd9f9aee63964bf82b0 (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": 1, "raw_scale_control": "none", "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_residual_s3", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.331878900527954, "cos_r_negg": [-0.01401346456259489, 0.01371360756456852, 0.006088063586503267], "cos_innovation_negg": [-0.01401346456259489, 0.01371360756456852, 0.006088063586503267], "cos_apical_negg": [-0.01633278653025627, 0.010381842963397503, 0.002567625604569912], "cos_Ac_negg": [-0.014014208689332008, 0.013713478110730648, 0.006087634712457657], "r_norm": [4.854386329650879, 4.749161720275879, 4.848539352416992], "ldr": {"dL_sdil": -0.00576019287109375, "dL_bp": -0.23061609268188477, "ratio": 0.02497741074400842}}, {"step": 100, "epoch": 0, "train_loss": 0.18873512744903564, "cos_r_negg": [0.42551085352897644, 0.3603532612323761, 0.31459999084472656], "cos_innovation_negg": [0.42551085352897644, 0.3603532612323761, 0.31459999084472656], "cos_apical_negg": [0.11449956893920898, 0.11613072454929352, 0.14878784120082855], "cos_Ac_negg": [0.42719539999961853, 0.36063358187675476, 0.3139951825141907], "r_norm": [0.8626118898391724, 0.8500277996063232, 0.8591506481170654]}, {"step": 200, "epoch": 0, "train_loss": 0.2459070384502411, "cos_r_negg": [0.48045146465301514, 0.43604618310928345, 0.35125911235809326], "cos_innovation_negg": [0.48045146465301514, 0.43604618310928345, 0.35125911235809326], "cos_apical_negg": [0.10844692587852478, 0.11874005943536758, 0.1324051320552826], "cos_Ac_negg": [0.4816785156726837, 0.4371064305305481, 0.3505333960056305], "r_norm": [0.7453992366790771, 0.7316138744354248, 0.7445143461227417]}, {"step": 300, "epoch": 0, "train_loss": 0.1282867193222046, "cos_r_negg": [0.49072495102882385, 0.45970141887664795, 0.38517892360687256], "cos_innovation_negg": [0.49072495102882385, 0.45970141887664795, 0.38517892360687256], "cos_apical_negg": [0.09590651839971542, 0.11035208404064178, 0.13731591403484344], "cos_Ac_negg": [0.4915056526660919, 0.4599386751651764, 0.3853808343410492], "r_norm": [0.5990716218948364, 0.5922057628631592, 0.598983645439148]}, {"step": 400, "epoch": 0, "train_loss": 0.1547357439994812, "cos_r_negg": [0.5252184867858887, 0.49495863914489746, 0.41902077198028564], "cos_innovation_negg": [0.5252184867858887, 0.49495863914489746, 0.41902077198028564], "cos_apical_negg": [0.08515290915966034, 0.1074976921081543, 0.13469132781028748], "cos_Ac_negg": [0.5259838104248047, 0.49456217885017395, 0.41876617074012756], "r_norm": [0.5662006735801697, 0.5600329637527466, 0.5683958530426025]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9467, "test_loss": 0.18566118240356444}, {"step": 500, "epoch": 1, "train_loss": 0.2022823989391327, "cos_r_negg": [0.5014853477478027, 0.49486565589904785, 0.4415689706802368], "cos_innovation_negg": [0.5014853477478027, 0.49486565589904785, 0.4415689706802368], "cos_apical_negg": [0.07622790336608887, 0.09425497055053711, 0.14676497876644135], "cos_Ac_negg": [0.5032318830490112, 0.4970094859600067, 0.4418157935142517], "r_norm": [0.5341676473617554, 0.5255112648010254, 0.5302309393882751], "ldr": {"dL_sdil": -0.015126332640647888, "dL_bp": -0.013028159737586975, "ratio": 1.1610490618262506}}, {"step": 600, "epoch": 1, "train_loss": 0.09608648717403412, "cos_r_negg": [0.5118353962898254, 0.5069502592086792, 0.4366770386695862], "cos_innovation_negg": [0.5118353962898254, 0.5069502592086792, 0.4366770386695862], "cos_apical_negg": [0.06319546699523926, 0.08845227211713791, 0.12924151122570038], "cos_Ac_negg": [0.5148888826370239, 0.5091942548751831, 0.4363383948802948], "r_norm": [0.4012887477874756, 0.39506545662879944, 0.4009026885032654]}, {"step": 700, "epoch": 1, "train_loss": 0.1620636284351349, "cos_r_negg": [0.5468978881835938, 0.5389772057533264, 0.4663362205028534], "cos_innovation_negg": [0.5468978881835938, 0.5389772057533264, 0.4663362205028534], "cos_apical_negg": [0.06697755306959152, 0.09641847759485245, 0.13771817088127136], "cos_Ac_negg": [0.5502389669418335, 0.5403789281845093, 0.4665992856025696], "r_norm": [0.4150056838989258, 0.40969038009643555, 0.4158512055873871]}, {"step": 800, "epoch": 1, "train_loss": 0.14280110597610474, "cos_r_negg": [0.5250078439712524, 0.5242472887039185, 0.4713127613067627], "cos_innovation_negg": [0.5250078439712524, 0.5242472887039185, 0.4713127613067627], "cos_apical_negg": [0.050785697996616364, 0.08504807949066162, 0.1308087259531021], "cos_Ac_negg": [0.5275696516036987, 0.5260343551635742, 0.4711673855781555], "r_norm": [0.34762388467788696, 0.34460484981536865, 0.3470115661621094]}, {"step": 900, "epoch": 1, "train_loss": 0.22257229685783386, "cos_r_negg": [0.542015552520752, 0.5338290333747864, 0.47017067670822144], "cos_innovation_negg": [0.542015552520752, 0.5338290333747864, 0.47017067670822144], "cos_apical_negg": [0.08236466348171234, 0.08128423243761063, 0.1286977231502533], "cos_Ac_negg": [0.5421379804611206, 0.5341194868087769, 0.469769686460495], "r_norm": [0.42938604950904846, 0.4241929054260254, 0.42841291427612305]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9578, "test_loss": 0.15420068130493164}, {"step": 1000, "epoch": 2, "train_loss": 0.036834731698036194, "cos_r_negg": [0.5252987146377563, 0.5434536337852478, 0.4803529679775238], "cos_innovation_negg": [0.5252987146377563, 0.5434536337852478, 0.4803529679775238], "cos_apical_negg": [0.055854231119155884, 0.08249439299106598, 0.11783044785261154], "cos_Ac_negg": [0.5273569822311401, 0.544925332069397, 0.4794422686100006], "r_norm": [0.33082473278045654, 0.3307965397834778, 0.3310631215572357], "ldr": {"dL_sdil": -0.006736010313034058, "dL_bp": -0.006280556321144104, "ratio": 1.0725180969011652}}, {"step": 1100, "epoch": 2, "train_loss": 0.1352086365222931, "cos_r_negg": [0.5510923862457275, 0.5639913082122803, 0.5086607933044434], "cos_innovation_negg": [0.5510923862457275, 0.5639913082122803, 0.5086607933044434], "cos_apical_negg": [0.06886047124862671, 0.08101245760917664, 0.12585797905921936], "cos_Ac_negg": [0.5541568994522095, 0.5669642090797424, 0.50771164894104], "r_norm": [0.3794521689414978, 0.37344086170196533, 0.37624090909957886]}, {"step": 1200, "epoch": 2, "train_loss": 0.1536949872970581, "cos_r_negg": [0.537842333316803, 0.560580313205719, 0.5249142646789551], "cos_innovation_negg": [0.537842333316803, 0.560580313205719, 0.5249142646789551], "cos_apical_negg": [0.05177416652441025, 0.06955734640359879, 0.13303402066230774], "cos_Ac_negg": [0.5428236126899719, 0.5652302503585815, 0.5233486890792847], "r_norm": [0.3386765718460083, 0.33176708221435547, 0.33597254753112793]}, {"step": 1300, "epoch": 2, "train_loss": 0.10706286132335663, "cos_r_negg": [0.5713360905647278, 0.5788295269012451, 0.5307925343513489], "cos_innovation_negg": [0.5713360905647278, 0.5788295269012451, 0.5307925343513489], "cos_apical_negg": [0.05037011206150055, 0.06893859058618546, 0.12395511567592621], "cos_Ac_negg": [0.5746093988418579, 0.581193745136261, 0.5294284820556641], "r_norm": [0.2783706784248352, 0.2770356237888336, 0.2780460715293884]}, {"step": 1400, "epoch": 2, "train_loss": 0.1364029049873352, "cos_r_negg": [0.5651035904884338, 0.5812767148017883, 0.523108720779419], "cos_innovation_negg": [0.5651035904884338, 0.5812767148017883, 0.523108720779419], "cos_apical_negg": [0.04892797768115997, 0.0780453309416771, 0.1301746368408203], "cos_Ac_negg": [0.5700987577438354, 0.5863732099533081, 0.5222785472869873], "r_norm": [0.28419798612594604, 0.28103306889533997, 0.2831452786922455]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9642, "test_loss": 0.11836729297637939}, {"step": 1500, "epoch": 3, "train_loss": 0.038324855268001556, "cos_r_negg": [0.5681567192077637, 0.5817520618438721, 0.5452873706817627], "cos_innovation_negg": [0.5681567192077637, 0.5817520618438721, 0.5452873706817627], "cos_apical_negg": [0.04089711979031563, 0.061245303601026535, 0.11822941154241562], "cos_Ac_negg": [0.5736209750175476, 0.5875518321990967, 0.545030951499939], "r_norm": [0.24270948767662048, 0.23855535686016083, 0.23999647796154022], "ldr": {"dL_sdil": -0.0048094242811203, "dL_bp": -0.004704371094703674, "ratio": 1.0223309735354207}}, {"step": 1600, "epoch": 3, "train_loss": 0.11071884632110596, "cos_r_negg": [0.573799192905426, 0.5846590995788574, 0.5462180376052856], "cos_innovation_negg": [0.573799192905426, 0.5846590995788574, 0.5462180376052856], "cos_apical_negg": [0.049994248896837234, 0.06502309441566467, 0.1271304488182068], "cos_Ac_negg": [0.5799444317817688, 0.5919026136398315, 0.5455365180969238], "r_norm": [0.27009037137031555, 0.26588764786720276, 0.26728346943855286]}, {"step": 1700, "epoch": 3, "train_loss": 0.08539792895317078, "cos_r_negg": [0.560289740562439, 0.5728726387023926, 0.5539113283157349], "cos_innovation_negg": [0.560289740562439, 0.5728726387023926, 0.5539113283157349], "cos_apical_negg": [0.0549091175198555, 0.06651626527309418, 0.1194741278886795], "cos_Ac_negg": [0.571294367313385, 0.5842162370681763, 0.5567351579666138], "r_norm": [0.26270005106925964, 0.2590261399745941, 0.2589811086654663]}, {"step": 1800, "epoch": 3, "train_loss": 0.04495394974946976, "cos_r_negg": [0.5494127869606018, 0.5575644373893738, 0.5537464618682861], "cos_innovation_negg": [0.5494127869606018, 0.5575644373893738, 0.5537464618682861], "cos_apical_negg": [0.03702382743358612, 0.05269481986761093, 0.12010857462882996], "cos_Ac_negg": [0.5647628307342529, 0.5746102929115295, 0.5641068816184998], "r_norm": [0.25814348459243774, 0.25500357151031494, 0.2557433247566223]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9646, "test_loss": 0.11824163208007812}, {"step": 1900, "epoch": 4, "train_loss": 0.04704385623335838, "cos_r_negg": [0.5639058947563171, 0.5820882320404053, 0.5742977857589722], "cos_innovation_negg": [0.5639058947563171, 0.5820882320404053, 0.5742977857589722], "cos_apical_negg": [0.04958593472838402, 0.07463567703962326, 0.12237516790628433], "cos_Ac_negg": [0.573023796081543, 0.5931601524353027, 0.5787421464920044], "r_norm": [0.3080936670303345, 0.30260777473449707, 0.3050350248813629]}, {"step": 2000, "epoch": 4, "train_loss": 0.1090206578373909, "cos_r_negg": [0.5349593758583069, 0.546436071395874, 0.5513330698013306], "cos_innovation_negg": [0.5349593758583069, 0.546436071395874, 0.5513330698013306], "cos_apical_negg": [0.035840194672346115, 0.05269690230488777, 0.11402727663516998], "cos_Ac_negg": [0.5707863569259644, 0.5776305198669434, 0.568793535232544], "r_norm": [0.22153672575950623, 0.21603810787200928, 0.21736770868301392], "ldr": {"dL_sdil": -0.006291545927524567, "dL_bp": -0.00635499507188797, "ratio": 0.9900158625387331}}, {"step": 2100, "epoch": 4, "train_loss": 0.023871950805187225, "cos_r_negg": [0.5592578649520874, 0.5849747061729431, 0.5727667808532715], "cos_innovation_negg": [0.5592578649520874, 0.5849747061729431, 0.5727667808532715], "cos_apical_negg": [0.026188362389802933, 0.060945652425289154, 0.12014678120613098], "cos_Ac_negg": [0.5772706270217896, 0.6038010120391846, 0.5796332359313965], "r_norm": [0.23918426036834717, 0.23436924815177917, 0.23573587834835052]}, {"step": 2200, "epoch": 4, "train_loss": 0.045306432992219925, "cos_r_negg": [0.5652357339859009, 0.5907092094421387, 0.5843456387519836], "cos_innovation_negg": [0.5652357339859009, 0.5907092094421387, 0.5843456387519836], "cos_apical_negg": [0.023994604125618935, 0.05470263212919235, 0.11344252526760101], "cos_Ac_negg": [0.5840235948562622, 0.6090487837791443, 0.5904650092124939], "r_norm": [0.22215567529201508, 0.21861065924167633, 0.21919593214988708]}, {"step": 2300, "epoch": 4, "train_loss": 0.18277393281459808, "cos_r_negg": [0.5552376508712769, 0.5786513090133667, 0.5839237570762634], "cos_innovation_negg": [0.5552376508712769, 0.5786513090133667, 0.5839237570762634], "cos_apical_negg": [0.020443785935640335, 0.041437696665525436, 0.11584413051605225], "cos_Ac_negg": [0.5809895992279053, 0.60596764087677, 0.5975843667984009], "r_norm": [0.19766755402088165, 0.1951572448015213, 0.19576910138130188]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9697, "test_loss": 0.11493425922393799}, {"step": 2400, "epoch": 5, "train_loss": 0.06708382815122604, "cos_r_negg": [0.5531058311462402, 0.5770075917243958, 0.5808913707733154], "cos_innovation_negg": [0.5531058311462402, 0.5770075917243958, 0.5808913707733154], "cos_apical_negg": [0.023905128240585327, 0.04676685109734535, 0.10950885713100433], "cos_Ac_negg": [0.5943174958229065, 0.6174683570861816, 0.6111822128295898], "r_norm": [0.181976318359375, 0.17854291200637817, 0.1797497570514679]}, {"step": 2500, "epoch": 5, "train_loss": 0.03424198925495148, "cos_r_negg": [0.5054612159729004, 0.5498422384262085, 0.5621560215950012], "cos_innovation_negg": [0.5054612159729004, 0.5498422384262085, 0.5621560215950012], "cos_apical_negg": [0.03350185975432396, 0.049942366778850555, 0.11881458014249802], "cos_Ac_negg": [0.5766274333000183, 0.6187105178833008, 0.605825662612915], "r_norm": [0.20842283964157104, 0.20446133613586426, 0.20585647225379944], "ldr": {"dL_sdil": -0.005982600152492523, "dL_bp": -0.006830863654613495, "ratio": 0.875818996687474}}, {"step": 2600, "epoch": 5, "train_loss": 0.057028595358133316, "cos_r_negg": [0.5284832715988159, 0.5672512054443359, 0.5801628828048706], "cos_innovation_negg": [0.5284832715988159, 0.5672512054443359, 0.5801628828048706], "cos_apical_negg": [0.032390762120485306, 0.04888254776597023, 0.11695471405982971], "cos_Ac_negg": [0.59566330909729, 0.6274288892745972, 0.6236974000930786], "r_norm": [0.20026922225952148, 0.19785061478614807, 0.19861774146556854]}, {"step": 2700, "epoch": 5, "train_loss": 0.04744546115398407, "cos_r_negg": [0.5307871699333191, 0.5803029537200928, 0.5932604074478149], "cos_innovation_negg": [0.5307871699333191, 0.5803029537200928, 0.5932604074478149], "cos_apical_negg": [0.018124477937817574, 0.053931158035993576, 0.12388523668050766], "cos_Ac_negg": [0.5922767519950867, 0.6326730251312256, 0.6278306245803833], "r_norm": [0.2167394459247589, 0.21486571431159973, 0.21571873128414154]}, {"step": 2800, "epoch": 5, "train_loss": 0.07268582284450531, "cos_r_negg": [0.5360569357872009, 0.5850205421447754, 0.5884784460067749], "cos_innovation_negg": [0.5360569357872009, 0.5850205421447754, 0.5884784460067749], "cos_apical_negg": [0.02157643996179104, 0.05472957715392113, 0.12207011133432388], "cos_Ac_negg": [0.5895681977272034, 0.6297746896743774, 0.6170710325241089], "r_norm": [0.2280716747045517, 0.2248237282037735, 0.22618234157562256]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9708, "test_loss": 0.11395362854003906}, {"step": 2900, "epoch": 6, "train_loss": 0.006686285138130188, "cos_r_negg": [0.51558518409729, 0.5628201961517334, 0.572902262210846], "cos_innovation_negg": [0.51558518409729, 0.5628201961517334, 0.572902262210846], "cos_apical_negg": [0.025216661393642426, 0.04977145791053772, 0.12181040644645691], "cos_Ac_negg": [0.5843631625175476, 0.6313927173614502, 0.6231183409690857], "r_norm": [0.23633728921413422, 0.23144465684890747, 0.23330065608024597]}, {"step": 3000, "epoch": 6, "train_loss": 0.0894983783364296, "cos_r_negg": [0.505083441734314, 0.5472036004066467, 0.5636588335037231], "cos_innovation_negg": [0.505083441734314, 0.5472036004066467, 0.5636588335037231], "cos_apical_negg": [0.016221128404140472, 0.03905455768108368, 0.11353820562362671], "cos_Ac_negg": [0.590485692024231, 0.6331952810287476, 0.6208517551422119], "r_norm": [0.19714771211147308, 0.19382169842720032, 0.19380110502243042], "ldr": {"dL_sdil": -0.005583271384239197, "dL_bp": -0.006370678544044495, "ratio": 0.8764013669248168}}, {"step": 3100, "epoch": 6, "train_loss": 0.024059748277068138, "cos_r_negg": [0.48229914903640747, 0.5265419483184814, 0.5552312135696411], "cos_innovation_negg": [0.48229914903640747, 0.5265419483184814, 0.5552312135696411], "cos_apical_negg": [0.01646444946527481, 0.04031222686171532, 0.11380480229854584], "cos_Ac_negg": [0.5978851914405823, 0.6291283965110779, 0.6335819959640503], "r_norm": [0.1687355637550354, 0.164513498544693, 0.16626185178756714]}, {"step": 3200, "epoch": 6, "train_loss": 0.06095162034034729, "cos_r_negg": [0.44403892755508423, 0.5013418197631836, 0.5414658784866333], "cos_innovation_negg": [0.44403892755508423, 0.5013418197631836, 0.5414658784866333], "cos_apical_negg": [0.005028644576668739, 0.05140520632266998, 0.1214417964220047], "cos_Ac_negg": [0.5862263441085815, 0.6361229419708252, 0.639459490776062], "r_norm": [0.1771746277809143, 0.17487776279449463, 0.1754063069820404]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9691, "test_loss": 0.12765268573760985}, {"step": 3300, "epoch": 7, "train_loss": 0.023883402347564697, "cos_r_negg": [0.4372710585594177, 0.49594300985336304, 0.5368139743804932], "cos_innovation_negg": [0.4372710585594177, 0.49594300985336304, 0.5368139743804932], "cos_apical_negg": [0.004016466438770294, 0.04841306433081627, 0.10837838053703308], "cos_Ac_negg": [0.5841434001922607, 0.6352051496505737, 0.6438031196594238], "r_norm": [0.17338460683822632, 0.17009533941745758, 0.17204077541828156]}, {"step": 3400, "epoch": 7, "train_loss": 0.011563774198293686, "cos_r_negg": [0.4032028019428253, 0.4476282000541687, 0.49372589588165283], "cos_innovation_negg": [0.4032028019428253, 0.4476282000541687, 0.49372589588165283], "cos_apical_negg": [0.02277015522122383, 0.0464412160217762, 0.11609035730361938], "cos_Ac_negg": [0.5841609239578247, 0.6342027187347412, 0.6419724225997925], "r_norm": [0.2020149677991867, 0.19642813503742218, 0.19873930513858795]}, {"step": 3500, "epoch": 7, "train_loss": 0.013702900148928165, "cos_r_negg": [0.41426321864128113, 0.454189658164978, 0.5024068355560303], "cos_innovation_negg": [0.41426321864128113, 0.454189658164978, 0.5024068355560303], "cos_apical_negg": [0.009891992434859276, 0.03601841628551483, 0.11002354323863983], "cos_Ac_negg": [0.6021798849105835, 0.6395878791809082, 0.6560165882110596], "r_norm": [0.17867809534072876, 0.17483720183372498, 0.17670764029026031], "ldr": {"dL_sdil": -0.006468772888183594, "dL_bp": -0.00890311598777771, "ratio": 0.7265740328514187}}, {"step": 3600, "epoch": 7, "train_loss": 0.0031365768518298864, "cos_r_negg": [0.40761950612068176, 0.4544501006603241, 0.49897146224975586], "cos_innovation_negg": [0.40761950612068176, 0.4544501006603241, 0.49897146224975586], "cos_apical_negg": [0.015406962484121323, 0.03500514477491379, 0.10255996882915497], "cos_Ac_negg": [0.5953332185745239, 0.6431121230125427, 0.6512629985809326], "r_norm": [0.16711248457431793, 0.16384246945381165, 0.165050208568573]}, {"step": 3700, "epoch": 7, "train_loss": 0.011196072213351727, "cos_r_negg": [0.43130093812942505, 0.4804968237876892, 0.5159401297569275], "cos_innovation_negg": [0.43130093812942505, 0.4804968237876892, 0.5159401297569275], "cos_apical_negg": [0.011967116966843605, 0.042917899787425995, 0.1083512157201767], "cos_Ac_negg": [0.6050165295600891, 0.6480522751808167, 0.6599148511886597], "r_norm": [0.16527262330055237, 0.16260609030723572, 0.1636393964290619]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9733, "test_loss": 0.11964113779067993}, {"step": 3800, "epoch": 8, "train_loss": 0.0023065307177603245, "cos_r_negg": [0.4035950005054474, 0.4397372603416443, 0.49234071373939514], "cos_innovation_negg": [0.4035950005054474, 0.4397372603416443, 0.49234071373939514], "cos_apical_negg": [0.013022277504205704, 0.03663323447108269, 0.11164499819278717], "cos_Ac_negg": [0.5987977981567383, 0.6396538019180298, 0.660892128944397], "r_norm": [0.16224171221256256, 0.16038155555725098, 0.16063547134399414]}, {"step": 3900, "epoch": 8, "train_loss": 0.017203664407134056, "cos_r_negg": [0.3728996217250824, 0.41909801959991455, 0.4681956171989441], "cos_innovation_negg": [0.3728996217250824, 0.41909801959991455, 0.4681956171989441], "cos_apical_negg": [0.013784561306238174, 0.037006475031375885, 0.10990840941667557], "cos_Ac_negg": [0.5922143459320068, 0.642852783203125, 0.6614621877670288], "r_norm": [0.15772667527198792, 0.1537609100341797, 0.15576636791229248]}, {"step": 4000, "epoch": 8, "train_loss": 0.00283118337392807, "cos_r_negg": [0.3320258855819702, 0.39198315143585205, 0.44947540760040283], "cos_innovation_negg": [0.3320258855819702, 0.39198315143585205, 0.44947540760040283], "cos_apical_negg": [0.002690427703782916, 0.040405161678791046, 0.1105620414018631], "cos_Ac_negg": [0.5902037620544434, 0.6408340930938721, 0.66427081823349], "r_norm": [0.14475294947624207, 0.1420484483242035, 0.1434231400489807], "ldr": {"dL_sdil": -0.0046197399497032166, "dL_bp": -0.006421804428100586, "ratio": 0.7193834694635233}}, {"step": 4100, "epoch": 8, "train_loss": 0.0056511638686060905, "cos_r_negg": [0.36505937576293945, 0.42459988594055176, 0.4812759757041931], "cos_innovation_negg": [0.36505937576293945, 0.42459988594055176, 0.4812759757041931], "cos_apical_negg": [0.004104168154299259, 0.04041241854429245, 0.11180675029754639], "cos_Ac_negg": [0.5889151096343994, 0.6453028321266174, 0.6706963181495667], "r_norm": [0.1615711897611618, 0.15812571346759796, 0.15983152389526367]}, {"step": 4200, "epoch": 8, "train_loss": 0.01308977510780096, "cos_r_negg": [0.3646843433380127, 0.4085808992385864, 0.45853450894355774], "cos_innovation_negg": [0.3646843433380127, 0.4085808992385864, 0.45853450894355774], "cos_apical_negg": [0.007420526351779699, 0.04285050928592682, 0.10662416368722916], "cos_Ac_negg": [0.6015160083770752, 0.650770902633667, 0.6693019866943359], "r_norm": [0.14816445112228394, 0.14529353380203247, 0.14685791730880737]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9734, "test_loss": 0.12385242824554443}, {"step": 4300, "epoch": 9, "train_loss": 0.009149294346570969, "cos_r_negg": [0.3273484408855438, 0.3603981137275696, 0.41702908277511597], "cos_innovation_negg": [0.3273484408855438, 0.3603981137275696, 0.41702908277511597], "cos_apical_negg": [0.011426137760281563, 0.03711375594139099, 0.10309694707393646], "cos_Ac_negg": [0.5904088020324707, 0.6431702375411987, 0.6659565567970276], "r_norm": [0.164593905210495, 0.16200922429561615, 0.16326048970222473]}, {"step": 4400, "epoch": 9, "train_loss": 0.0012302198447287083, "cos_r_negg": [0.34810036420822144, 0.3918251693248749, 0.44217023253440857], "cos_innovation_negg": [0.34810036420822144, 0.3918251693248749, 0.44217023253440857], "cos_apical_negg": [0.005247277207672596, 0.037915945053100586, 0.10195288062095642], "cos_Ac_negg": [0.5923473238945007, 0.6484144330024719, 0.6756606101989746], "r_norm": [0.13819624483585358, 0.13477015495300293, 0.13712504506111145]}, {"step": 4500, "epoch": 9, "train_loss": 0.03266465291380882, "cos_r_negg": [0.3250414729118347, 0.35153859853744507, 0.405872642993927], "cos_innovation_negg": [0.3250414729118347, 0.35153859853744507, 0.405872642993927], "cos_apical_negg": [0.00648121815174818, 0.03344428166747093, 0.09688430279493332], "cos_Ac_negg": [0.593754768371582, 0.637917160987854, 0.6645973920822144], "r_norm": [0.1476287692785263, 0.14393720030784607, 0.14630907773971558], "ldr": {"dL_sdil": -0.0060315802693367004, "dL_bp": -0.008732922375202179, "ratio": 0.6906714625637631}}, {"step": 4600, "epoch": 9, "train_loss": 0.0008049164898693562, "cos_r_negg": [0.33393365144729614, 0.37126344442367554, 0.4240191578865051], "cos_innovation_negg": [0.33393365144729614, 0.37126344442367554, 0.4240191578865051], "cos_apical_negg": [0.013297813013195992, 0.035311177372932434, 0.09767794609069824], "cos_Ac_negg": [0.5870927572250366, 0.6417601108551025, 0.6765743494033813], "r_norm": [0.15512298047542572, 0.15194562077522278, 0.1534360647201538]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9729, "test_loss": 0.12904755640029908}, {"step": 4700, "epoch": 10, "train_loss": 0.0050363666377961636, "cos_r_negg": [0.2971585690975189, 0.3426578938961029, 0.3923370838165283], "cos_innovation_negg": [0.2971585690975189, 0.3426578938961029, 0.3923370838165283], "cos_apical_negg": [0.006049447692930698, 0.0372016504406929, 0.09562559425830841], "cos_Ac_negg": [0.582510232925415, 0.6433707475662231, 0.6715540885925293], "r_norm": [0.14715415239334106, 0.14434099197387695, 0.14648661017417908]}, {"step": 4800, "epoch": 10, "train_loss": 0.005406802985817194, "cos_r_negg": [0.28977954387664795, 0.3338377773761749, 0.38440027832984924], "cos_innovation_negg": [0.28977954387664795, 0.3338377773761749, 0.38440027832984924], "cos_apical_negg": [0.00969766266644001, 0.037736676633358, 0.09605565667152405], "cos_Ac_negg": [0.5797987580299377, 0.6394490003585815, 0.6709378957748413], "r_norm": [0.14719560742378235, 0.14388829469680786, 0.14618919789791107]}, {"step": 4900, "epoch": 10, "train_loss": 0.0016784570179879665, "cos_r_negg": [0.29306820034980774, 0.33002734184265137, 0.3801448345184326], "cos_innovation_negg": [0.29306820034980774, 0.33002734184265137, 0.3801448345184326], "cos_apical_negg": [0.008315379731357098, 0.03247849643230438, 0.0929107517004013], "cos_Ac_negg": [0.5799634456634521, 0.637857973575592, 0.668197512626648], "r_norm": [0.13963955640792847, 0.13658124208450317, 0.1389416605234146]}, {"step": 5000, "epoch": 10, "train_loss": 0.0007381493342109025, "cos_r_negg": [0.27754515409469604, 0.31434905529022217, 0.3627849221229553], "cos_innovation_negg": [0.27754515409469604, 0.31434905529022217, 0.3627849221229553], "cos_apical_negg": [0.00608882075175643, 0.03207935392856598, 0.09047340601682663], "cos_Ac_negg": [0.576554536819458, 0.639495313167572, 0.6658531427383423], "r_norm": [0.14053618907928467, 0.13784439861774445, 0.14010469615459442], "ldr": {"dL_sdil": -0.006616033613681793, "dL_bp": -0.010654069483280182, "ratio": 0.6209865276422849}}, {"step": 5100, "epoch": 10, "train_loss": 0.0026385232340544462, "cos_r_negg": [0.28332141041755676, 0.318595290184021, 0.3708907663822174], "cos_innovation_negg": [0.28332141041755676, 0.318595290184021, 0.3708907663822174], "cos_apical_negg": [0.00385945662856102, 0.03137340396642685, 0.0917658880352974], "cos_Ac_negg": [0.5841526985168457, 0.6423323154449463, 0.6733631491661072], "r_norm": [0.14305317401885986, 0.14065806567668915, 0.14249151945114136]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9737, "test_loss": 0.1326816659927368}, {"step": 5200, "epoch": 11, "train_loss": 0.0011589385103434324, "cos_r_negg": [0.278288334608078, 0.3090996742248535, 0.36126548051834106], "cos_innovation_negg": [0.278288334608078, 0.3090996742248535, 0.36126548051834106], "cos_apical_negg": [0.008557699620723724, 0.03427043557167053, 0.08963680267333984], "cos_Ac_negg": [0.5796408653259277, 0.638587474822998, 0.6719264984130859], "r_norm": [0.134597510099411, 0.13201984763145447, 0.13406230509281158]}, {"step": 5300, "epoch": 11, "train_loss": 0.00224539777263999, "cos_r_negg": [0.26544421911239624, 0.30028676986694336, 0.34926825761795044], "cos_innovation_negg": [0.26544421911239624, 0.30028676986694336, 0.34926825761795044], "cos_apical_negg": [0.0069392225705087185, 0.030967628583312035, 0.0874195545911789], "cos_Ac_negg": [0.575717031955719, 0.6370882987976074, 0.6674982309341431], "r_norm": [0.13563451170921326, 0.13257160782814026, 0.1347551941871643]}, {"step": 5400, "epoch": 11, "train_loss": 0.002859623171389103, "cos_r_negg": [0.2713771462440491, 0.3025330901145935, 0.3500608205795288], "cos_innovation_negg": [0.2713771462440491, 0.3025330901145935, 0.3500608205795288], "cos_apical_negg": [0.007332724519073963, 0.029432091861963272, 0.08250143378973007], "cos_Ac_negg": [0.5809556245803833, 0.6392914652824402, 0.6695345640182495], "r_norm": [0.13454987108707428, 0.1319369077682495, 0.13394761085510254]}, {"step": 5500, "epoch": 11, "train_loss": 0.0011915442300960422, "cos_r_negg": [0.2577810287475586, 0.29153379797935486, 0.34306567907333374], "cos_innovation_negg": [0.2577810287475586, 0.29153379797935486, 0.34306567907333374], "cos_apical_negg": [0.004283071495592594, 0.03088618442416191, 0.0867825299501419], "cos_Ac_negg": [0.5771710872650146, 0.6397589445114136, 0.6721286773681641], "r_norm": [0.13539481163024902, 0.13255712389945984, 0.13476434350013733], "ldr": {"dL_sdil": -0.005841605365276337, "dL_bp": -0.009094223380088806, "ratio": 0.6423424102454027}}, {"step": 5600, "epoch": 11, "train_loss": 0.0009290726156905293, "cos_r_negg": [0.2523663640022278, 0.28946352005004883, 0.3401210308074951], "cos_innovation_negg": [0.2523663640022278, 0.28946352005004883, 0.3401210308074951], "cos_apical_negg": [0.0022307210601866245, 0.03210269659757614, 0.08586939424276352], "cos_Ac_negg": [0.5747398138046265, 0.6356382369995117, 0.671718955039978], "r_norm": [0.13824433088302612, 0.13527542352676392, 0.13748770952224731]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9738, "test_loss": 0.13491204013824462}, {"step": 5700, "epoch": 12, "train_loss": 0.001086224801838398, "cos_r_negg": [0.24984243512153625, 0.28309738636016846, 0.33308276534080505], "cos_innovation_negg": [0.24984243512153625, 0.28309738636016846, 0.33308276534080505], "cos_apical_negg": [0.007476150523871183, 0.03297806903719902, 0.08509723842144012], "cos_Ac_negg": [0.5744707584381104, 0.635699987411499, 0.6720552444458008], "r_norm": [0.1377202570438385, 0.13466627895832062, 0.13708040118217468]}, {"step": 5800, "epoch": 12, "train_loss": 0.00036634469870477915, "cos_r_negg": [0.24313117563724518, 0.27526184916496277, 0.3211178183555603], "cos_innovation_negg": [0.24313117563724518, 0.27526184916496277, 0.3211178183555603], "cos_apical_negg": [0.005894692614674568, 0.030842486768960953, 0.07910270243883133], "cos_Ac_negg": [0.5708178877830505, 0.6340307593345642, 0.6696828603744507], "r_norm": [0.13602876663208008, 0.1329473853111267, 0.13534313440322876]}, {"step": 5900, "epoch": 12, "train_loss": 0.002405446022748947, "cos_r_negg": [0.24503980576992035, 0.28094232082366943, 0.3298783302307129], "cos_innovation_negg": [0.24503980576992035, 0.28094232082366943, 0.3298783302307129], "cos_apical_negg": [0.00806405022740364, 0.03380434215068817, 0.08361522108316422], "cos_Ac_negg": [0.5700715184211731, 0.6392761468887329, 0.6756144762039185], "r_norm": [0.13945350050926208, 0.13656724989414215, 0.1388821005821228]}, {"step": 6000, "epoch": 12, "train_loss": 0.00017605547327548265, "cos_r_negg": [0.23001554608345032, 0.2646538019180298, 0.31029847264289856], "cos_innovation_negg": [0.23001554608345032, 0.2646538019180298, 0.31029847264289856], "cos_apical_negg": [0.00541045144200325, 0.032030005007982254, 0.07905453443527222], "cos_Ac_negg": [0.5708320140838623, 0.6355617046356201, 0.6697124242782593], "r_norm": [0.13937880098819733, 0.13642413914203644, 0.13876959681510925], "ldr": {"dL_sdil": -0.0074355751276016235, "dL_bp": -0.012880563735961914, "ratio": 0.5772709393799167}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9744, "test_loss": 0.13820682907104492}, {"step": 6100, "epoch": 13, "train_loss": 0.0009763767011463642, "cos_r_negg": [0.23343919217586517, 0.2614392042160034, 0.3084016740322113], "cos_innovation_negg": [0.23343919217586517, 0.2614392042160034, 0.3084016740322113], "cos_apical_negg": [0.006905930116772652, 0.02927277237176895, 0.07730467617511749], "cos_Ac_negg": [0.5668554902076721, 0.6321935653686523, 0.6682334542274475], "r_norm": [0.1409309208393097, 0.1382412612438202, 0.14066889882087708]}, {"step": 6200, "epoch": 13, "train_loss": 0.0009069643565453589, "cos_r_negg": [0.22708435356616974, 0.25632721185684204, 0.3010939657688141], "cos_innovation_negg": [0.22708435356616974, 0.25632721185684204, 0.3010939657688141], "cos_apical_negg": [0.008822782896459103, 0.029689308255910873, 0.07625748962163925], "cos_Ac_negg": [0.5691685676574707, 0.6327530145645142, 0.6673669815063477], "r_norm": [0.1352216601371765, 0.13269278407096863, 0.13488000631332397]}, {"step": 6300, "epoch": 13, "train_loss": 0.0006839922862127423, "cos_r_negg": [0.21666383743286133, 0.24478955566883087, 0.2880740165710449], "cos_innovation_negg": [0.21666383743286133, 0.24478955566883087, 0.2880740165710449], "cos_apical_negg": [0.006035208702087402, 0.029654499143362045, 0.07456348836421967], "cos_Ac_negg": [0.5658756494522095, 0.6283625364303589, 0.6640976667404175], "r_norm": [0.1393083930015564, 0.136541485786438, 0.13898012042045593]}, {"step": 6400, "epoch": 13, "train_loss": 0.0005505380686372519, "cos_r_negg": [0.21647170186042786, 0.24702751636505127, 0.29178813099861145], "cos_innovation_negg": [0.21647170186042786, 0.24702751636505127, 0.29178813099861145], "cos_apical_negg": [0.006189728155732155, 0.02983001060783863, 0.0744553953409195], "cos_Ac_negg": [0.5681161284446716, 0.6294869184494019, 0.6676573157310486], "r_norm": [0.14122754335403442, 0.13862696290016174, 0.14077913761138916]}, {"step": 6500, "epoch": 13, "train_loss": 0.00035615594242699444, "cos_r_negg": [0.21158862113952637, 0.2439221888780594, 0.28754645586013794], "cos_innovation_negg": [0.21158862113952637, 0.2439221888780594, 0.28754645586013794], "cos_apical_negg": [0.0036592865362763405, 0.029765024781227112, 0.0722854882478714], "cos_Ac_negg": [0.5633366107940674, 0.6278918981552124, 0.6642735600471497], "r_norm": [0.1363537609577179, 0.13358211517333984, 0.13598135113716125], "ldr": {"dL_sdil": -0.007500775158405304, "dL_bp": -0.013501659035682678, "ratio": 0.5555447029570204}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9739, "test_loss": 0.13985029439926147}, {"step": 6600, "epoch": 14, "train_loss": 0.0006743880803696811, "cos_r_negg": [0.2092241793870926, 0.24195091426372528, 0.28283411264419556], "cos_innovation_negg": [0.2092241793870926, 0.24195091426372528, 0.28283411264419556], "cos_apical_negg": [0.006467920728027821, 0.02924356423318386, 0.07268835604190826], "cos_Ac_negg": [0.5627449750900269, 0.6298940181732178, 0.6678542494773865], "r_norm": [0.13558074831962585, 0.13304200768470764, 0.1353493332862854]}, {"step": 6700, "epoch": 14, "train_loss": 0.00031959370244294405, "cos_r_negg": [0.20629185438156128, 0.23710137605667114, 0.2809441089630127], "cos_innovation_negg": [0.20629185438156128, 0.23710137605667114, 0.2809441089630127], "cos_apical_negg": [0.0054047489538788795, 0.029138857498764992, 0.07521852850914001], "cos_Ac_negg": [0.5613837242126465, 0.6291850805282593, 0.6665549278259277], "r_norm": [0.13533002138137817, 0.13286085426807404, 0.13514675199985504]}, {"step": 6800, "epoch": 14, "train_loss": 0.0004982365644536912, "cos_r_negg": [0.19535169005393982, 0.2248317003250122, 0.2654002904891968], "cos_innovation_negg": [0.19535169005393982, 0.2248317003250122, 0.2654002904891968], "cos_apical_negg": [0.005468869116157293, 0.028167756274342537, 0.07059791684150696], "cos_Ac_negg": [0.5609909296035767, 0.6257888078689575, 0.6647816896438599], "r_norm": [0.13461200892925262, 0.1320335566997528, 0.13426797091960907]}, {"step": 6900, "epoch": 14, "train_loss": 0.0008965390734374523, "cos_r_negg": [0.19515705108642578, 0.22602930665016174, 0.26685693860054016], "cos_innovation_negg": [0.19515705108642578, 0.22602930665016174, 0.26685693860054016], "cos_apical_negg": [0.00480653764680028, 0.0283636637032032, 0.07082416117191315], "cos_Ac_negg": [0.5626923441886902, 0.6275756359100342, 0.6644430160522461], "r_norm": [0.13189470767974854, 0.129350483417511, 0.13152959942817688]}, {"step": 7000, "epoch": 14, "train_loss": 0.00037491926923394203, "cos_r_negg": [0.19524270296096802, 0.2249695062637329, 0.26578110456466675], "cos_innovation_negg": [0.19524270296096802, 0.2249695062637329, 0.26578110456466675], "cos_apical_negg": [0.004228589590638876, 0.02768980711698532, 0.07010260224342346], "cos_Ac_negg": [0.56703782081604, 0.6305042505264282, 0.6680166721343994], "r_norm": [0.1324881613254547, 0.13006103038787842, 0.13218775391578674], "ldr": {"dL_sdil": -0.007371082901954651, "dL_bp": -0.013634532690048218, "ratio": 0.5406186680189465}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9741, "test_loss": 0.14238088626861573}], "final": {"test_acc": 0.9741, "test_loss": 0.14238088626861573, "wall_s": 23.89514470100403, "cos_r_negg": [0.19353516399860382, 0.2238001823425293, 0.2641659080982208], "cos_innovation_negg": [0.19353516399860382, 0.2238001823425293, 0.2641659080982208], "cos_apical_negg": [0.004093382507562637, 0.02720249630510807, 0.06915740668773651], "cos_Ac_negg": [0.5639503002166748, 0.6269025802612305, 0.6651663184165955]}}