{"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": 0.05, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p05_residual_s4", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4633917808532715, "cos_r_negg": [-0.012779329903423786, 0.029934005811810493, 0.00944238156080246], "cos_innovation_negg": [-0.012779329903423786, 0.029934005811810493, 0.00944238156080246], "cos_apical_negg": [-0.013300376012921333, 0.029353521764278412, 0.008677935227751732], "cos_Ac_negg": [-0.012779482640326023, 0.029933910816907883, 0.009442325681447983], "r_norm": [4.848395347595215, 4.855302810668945, 4.948003768920898], "ldr": {"dL_sdil": -0.0033469200134277344, "dL_bp": -0.2672088146209717, "ratio": 0.01252548505248694}}, {"step": 100, "epoch": 0, "train_loss": 0.19363561272621155, "cos_r_negg": [0.459269642829895, 0.3797130286693573, 0.31844401359558105], "cos_innovation_negg": [0.459269642829895, 0.3797130286693573, 0.31844401359558105], "cos_apical_negg": [0.1869918406009674, 0.18621602654457092, 0.17861896753311157], "cos_Ac_negg": [0.45944228768348694, 0.3798229694366455, 0.31835034489631653], "r_norm": [0.9167285561561584, 0.9265720844268799, 0.9415656924247742]}, {"step": 200, "epoch": 0, "train_loss": 0.25281593203544617, "cos_r_negg": [0.5016197562217712, 0.4379856586456299, 0.3615146577358246], "cos_innovation_negg": [0.5016197562217712, 0.4379856586456299, 0.3615146577358246], "cos_apical_negg": [0.19982387125492096, 0.19391566514968872, 0.18751326203346252], "cos_Ac_negg": [0.5017473101615906, 0.4381555914878845, 0.3611891567707062], "r_norm": [0.7675690650939941, 0.7717165946960449, 0.7897419929504395]}, {"step": 300, "epoch": 0, "train_loss": 0.24676823616027832, "cos_r_negg": [0.534898042678833, 0.46356427669525146, 0.3760718107223511], "cos_innovation_negg": [0.534898042678833, 0.46356427669525146, 0.3760718107223511], "cos_apical_negg": [0.17926794290542603, 0.18904438614845276, 0.1845228672027588], "cos_Ac_negg": [0.5352267026901245, 0.46348699927330017, 0.37584298849105835], "r_norm": [0.5503602027893066, 0.5532886385917664, 0.5686604976654053]}, {"step": 400, "epoch": 0, "train_loss": 0.15302714705467224, "cos_r_negg": [0.5326367020606995, 0.47829481959342957, 0.4119530916213989], "cos_innovation_negg": [0.5326367020606995, 0.47829481959342957, 0.4119530916213989], "cos_apical_negg": [0.16547957062721252, 0.17010702192783356, 0.17529134452342987], "cos_Ac_negg": [0.5327945947647095, 0.4782412350177765, 0.41187041997909546], "r_norm": [0.524154782295227, 0.5277701020240784, 0.5397800207138062]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9447, "test_loss": 0.18155341796875}, {"step": 500, "epoch": 1, "train_loss": 0.17874321341514587, "cos_r_negg": [0.5227363109588623, 0.5101392269134521, 0.4588015377521515], "cos_innovation_negg": [0.5227363109588623, 0.5101392269134521, 0.4588015377521515], "cos_apical_negg": [0.1373881697654724, 0.15714529156684875, 0.17961546778678894], "cos_Ac_negg": [0.522687554359436, 0.5101361274719238, 0.45864057540893555], "r_norm": [0.5000540018081665, 0.5062152147293091, 0.5183804035186768], "ldr": {"dL_sdil": -0.012512341141700745, "dL_bp": -0.01118496060371399, "ratio": 1.1186754772784802}}, {"step": 600, "epoch": 1, "train_loss": 0.06491975486278534, "cos_r_negg": [0.5474497675895691, 0.5192220211029053, 0.4468640387058258], "cos_innovation_negg": [0.5474497675895691, 0.5192220211029053, 0.4468640387058258], "cos_apical_negg": [0.1271270364522934, 0.15750424563884735, 0.1773270070552826], "cos_Ac_negg": [0.5476289391517639, 0.5195400714874268, 0.44661420583724976], "r_norm": [0.4415815472602844, 0.44529294967651367, 0.45522838830947876]}, {"step": 700, "epoch": 1, "train_loss": 0.198802649974823, "cos_r_negg": [0.5695801973342896, 0.5482631921768188, 0.4604000151157379], "cos_innovation_negg": [0.5695801973342896, 0.5482631921768188, 0.4604000151157379], "cos_apical_negg": [0.147748202085495, 0.16263194382190704, 0.17765286564826965], "cos_Ac_negg": [0.5699089765548706, 0.5480311512947083, 0.4600382149219513], "r_norm": [0.47886714339256287, 0.4833182990550995, 0.4965375065803528]}, {"step": 800, "epoch": 1, "train_loss": 0.15646713972091675, "cos_r_negg": [0.567284107208252, 0.5482035279273987, 0.46796345710754395], "cos_innovation_negg": [0.567284107208252, 0.5482035279273987, 0.46796345710754395], "cos_apical_negg": [0.12772102653980255, 0.159915030002594, 0.17331650853157043], "cos_Ac_negg": [0.5676268339157104, 0.5480576753616333, 0.46787577867507935], "r_norm": [0.3809143900871277, 0.3821704685688019, 0.39236980676651]}, {"step": 900, "epoch": 1, "train_loss": 0.18561142683029175, "cos_r_negg": [0.5868179798126221, 0.5653269290924072, 0.481930673122406], "cos_innovation_negg": [0.5868179798126221, 0.5653269290924072, 0.481930673122406], "cos_apical_negg": [0.11873948574066162, 0.16003139317035675, 0.17157092690467834], "cos_Ac_negg": [0.5868556499481201, 0.5651793479919434, 0.4816763401031494], "r_norm": [0.43814408779144287, 0.4417032301425934, 0.451862096786499]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9536, "test_loss": 0.15419194259643554}, {"step": 1000, "epoch": 2, "train_loss": 0.02181781455874443, "cos_r_negg": [0.5739259719848633, 0.5635170340538025, 0.48731517791748047], "cos_innovation_negg": [0.5739259719848633, 0.5635170340538025, 0.48731517791748047], "cos_apical_negg": [0.11029858887195587, 0.13616223633289337, 0.1503383219242096], "cos_Ac_negg": [0.5741055607795715, 0.563295841217041, 0.4872198700904846], "r_norm": [0.36802178621292114, 0.3702920079231262, 0.3779096007347107], "ldr": {"dL_sdil": -0.011937856674194336, "dL_bp": -0.010014474391937256, "ratio": 1.1920602327173169}}, {"step": 1100, "epoch": 2, "train_loss": 0.11275234818458557, "cos_r_negg": [0.5853288769721985, 0.5786828994750977, 0.502781867980957], "cos_innovation_negg": [0.5853288769721985, 0.5786828994750977, 0.502781867980957], "cos_apical_negg": [0.11679544299840927, 0.14709730446338654, 0.16490384936332703], "cos_Ac_negg": [0.5859564542770386, 0.5788313746452332, 0.5025363564491272], "r_norm": [0.3950566053390503, 0.3970186412334442, 0.40572869777679443]}, {"step": 1200, "epoch": 2, "train_loss": 0.1583356261253357, "cos_r_negg": [0.5809619426727295, 0.5775206089019775, 0.49697285890579224], "cos_innovation_negg": [0.5809619426727295, 0.5775206089019775, 0.49697285890579224], "cos_apical_negg": [0.09338440001010895, 0.13124938309192657, 0.1401596963405609], "cos_Ac_negg": [0.5822113752365112, 0.5778946280479431, 0.49724453687667847], "r_norm": [0.3299587368965149, 0.3310512900352478, 0.33843719959259033]}, {"step": 1300, "epoch": 2, "train_loss": 0.12771108746528625, "cos_r_negg": [0.6014107465744019, 0.6022679209709167, 0.5082502365112305], "cos_innovation_negg": [0.6014107465744019, 0.6022679209709167, 0.5082502365112305], "cos_apical_negg": [0.10280664265155792, 0.13716718554496765, 0.1483091562986374], "cos_Ac_negg": [0.6017036437988281, 0.6021094918251038, 0.5082778334617615], "r_norm": [0.35670411586761475, 0.3595981299877167, 0.368849515914917]}, {"step": 1400, "epoch": 2, "train_loss": 0.10537157952785492, "cos_r_negg": [0.5921329259872437, 0.5912002325057983, 0.5061360001564026], "cos_innovation_negg": [0.5921329259872437, 0.5912002325057983, 0.5061360001564026], "cos_apical_negg": [0.09046350419521332, 0.13055309653282166, 0.15138307213783264], "cos_Ac_negg": [0.5946567058563232, 0.5925124883651733, 0.506708562374115], "r_norm": [0.31271618604660034, 0.3149046301841736, 0.3246137499809265]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9631, "test_loss": 0.12175595703125}, {"step": 1500, "epoch": 3, "train_loss": 0.035476528108119965, "cos_r_negg": [0.5890349745750427, 0.6021519303321838, 0.5385955572128296], "cos_innovation_negg": [0.5890349745750427, 0.6021519303321838, 0.5385955572128296], "cos_apical_negg": [0.06733257323503494, 0.10770899802446365, 0.14317253232002258], "cos_Ac_negg": [0.5905673503875732, 0.6032276153564453, 0.5383012294769287], "r_norm": [0.2837359309196472, 0.28518205881118774, 0.2939888834953308], "ldr": {"dL_sdil": -0.007902510464191437, "dL_bp": -0.007275938987731934, "ratio": 1.086115548455804}}, {"step": 1600, "epoch": 3, "train_loss": 0.05675691366195679, "cos_r_negg": [0.5812368392944336, 0.6006582379341125, 0.5353538990020752], "cos_innovation_negg": [0.5812368392944336, 0.6006582379341125, 0.5353538990020752], "cos_apical_negg": [0.08494053781032562, 0.11340245604515076, 0.1384579837322235], "cos_Ac_negg": [0.5837069153785706, 0.6030481457710266, 0.5360864996910095], "r_norm": [0.28139248490333557, 0.2813023626804352, 0.2908490300178528]}, {"step": 1700, "epoch": 3, "train_loss": 0.03819672018289566, "cos_r_negg": [0.576766848564148, 0.5974439382553101, 0.5257833003997803], "cos_innovation_negg": [0.576766848564148, 0.5974439382553101, 0.5257833003997803], "cos_apical_negg": [0.07903841137886047, 0.11245925724506378, 0.1422579288482666], "cos_Ac_negg": [0.5808981657028198, 0.6002588272094727, 0.5256128311157227], "r_norm": [0.22938254475593567, 0.228070467710495, 0.23500370979309082]}, {"step": 1800, "epoch": 3, "train_loss": 0.012577962130308151, "cos_r_negg": [0.5960943698883057, 0.6038258671760559, 0.5261780619621277], "cos_innovation_negg": [0.5960943698883057, 0.6038258671760559, 0.5261780619621277], "cos_apical_negg": [0.07420073449611664, 0.11327838152647018, 0.12981820106506348], "cos_Ac_negg": [0.598205029964447, 0.6051718592643738, 0.5251649022102356], "r_norm": [0.23552271723747253, 0.23420119285583496, 0.24097025394439697]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9653, "test_loss": 0.1188214527130127}, {"step": 1900, "epoch": 4, "train_loss": 0.033263400197029114, "cos_r_negg": [0.5949555039405823, 0.6183372735977173, 0.5297161936759949], "cos_innovation_negg": [0.5949555039405823, 0.6183372735977173, 0.5297161936759949], "cos_apical_negg": [0.0582868829369545, 0.11855240166187286, 0.1389888972043991], "cos_Ac_negg": [0.5992276668548584, 0.6210401654243469, 0.5292882919311523], "r_norm": [0.25039803981781006, 0.2509022355079651, 0.25720369815826416]}, {"step": 2000, "epoch": 4, "train_loss": 0.07706649601459503, "cos_r_negg": [0.5977005362510681, 0.609539270401001, 0.5329980850219727], "cos_innovation_negg": [0.5977005362510681, 0.609539270401001, 0.5329980850219727], "cos_apical_negg": [0.0513223260641098, 0.10410377383232117, 0.1242242157459259], "cos_Ac_negg": [0.6040879487991333, 0.6144353151321411, 0.5341240167617798], "r_norm": [0.2246806025505066, 0.223198801279068, 0.23086339235305786], "ldr": {"dL_sdil": -0.005203373730182648, "dL_bp": -0.005424492061138153, "ratio": 0.9592370440470124}}, {"step": 2100, "epoch": 4, "train_loss": 0.027847032994031906, "cos_r_negg": [0.6074799299240112, 0.6233466267585754, 0.5568576455116272], "cos_innovation_negg": [0.6074799299240112, 0.6233466267585754, 0.5568576455116272], "cos_apical_negg": [0.06025214493274689, 0.10350442677736282, 0.1337289661169052], "cos_Ac_negg": [0.6121597290039062, 0.626561164855957, 0.5566967725753784], "r_norm": [0.24921327829360962, 0.24901887774467468, 0.25541913509368896]}, {"step": 2200, "epoch": 4, "train_loss": 0.018447523936629295, "cos_r_negg": [0.6021316051483154, 0.6244376301765442, 0.5686969757080078], "cos_innovation_negg": [0.6021316051483154, 0.6244376301765442, 0.5686969757080078], "cos_apical_negg": [0.057232726365327835, 0.09396456182003021, 0.13490691781044006], "cos_Ac_negg": [0.609283447265625, 0.6303319931030273, 0.5689502954483032], "r_norm": [0.2358730435371399, 0.23480519652366638, 0.24117043614387512]}, {"step": 2300, "epoch": 4, "train_loss": 0.17033937573432922, "cos_r_negg": [0.5807777643203735, 0.6194117069244385, 0.5654565095901489], "cos_innovation_negg": [0.5807777643203735, 0.6194117069244385, 0.5654565095901489], "cos_apical_negg": [0.06197572499513626, 0.11007210612297058, 0.1384003460407257], "cos_Ac_negg": [0.5930091738700867, 0.6267299056053162, 0.5684207677841187], "r_norm": [0.23120008409023285, 0.2296319454908371, 0.23754820227622986]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9679, "test_loss": 0.11396435432434082}, {"step": 2400, "epoch": 5, "train_loss": 0.07778006047010422, "cos_r_negg": [0.5961868762969971, 0.628020703792572, 0.5711195468902588], "cos_innovation_negg": [0.5961868762969971, 0.628020703792572, 0.5711195468902588], "cos_apical_negg": [0.05230245739221573, 0.08951733261346817, 0.12939926981925964], "cos_Ac_negg": [0.613296389579773, 0.6407097578048706, 0.5756195187568665], "r_norm": [0.24527998268604279, 0.24481268227100372, 0.25092095136642456]}, {"step": 2500, "epoch": 5, "train_loss": 0.03131214156746864, "cos_r_negg": [0.5886510610580444, 0.6179237365722656, 0.5504201054573059], "cos_innovation_negg": [0.5886510610580444, 0.6179237365722656, 0.5504201054573059], "cos_apical_negg": [0.053261227905750275, 0.09723541885614395, 0.13756129145622253], "cos_Ac_negg": [0.6131516695022583, 0.6393392086029053, 0.5581721067428589], "r_norm": [0.203288272023201, 0.20195026695728302, 0.20764616131782532], "ldr": {"dL_sdil": -0.005027107894420624, "dL_bp": -0.00556723028421402, "ratio": 0.9029818487435444}}, {"step": 2600, "epoch": 5, "train_loss": 0.05235353484749794, "cos_r_negg": [0.5919328927993774, 0.6215847730636597, 0.5753085613250732], "cos_innovation_negg": [0.5919328927993774, 0.6215847730636597, 0.5753085613250732], "cos_apical_negg": [0.03845927119255066, 0.09658487141132355, 0.1456366777420044], "cos_Ac_negg": [0.6172924041748047, 0.6442195773124695, 0.5827828645706177], "r_norm": [0.21641728281974792, 0.21532754600048065, 0.22124408185482025]}, {"step": 2700, "epoch": 5, "train_loss": 0.04534776508808136, "cos_r_negg": [0.5973405838012695, 0.6376732587814331, 0.5805200934410095], "cos_innovation_negg": [0.5973405838012695, 0.6376732587814331, 0.5805200934410095], "cos_apical_negg": [0.05731664597988129, 0.09314532577991486, 0.13641613721847534], "cos_Ac_negg": [0.6209036111831665, 0.6600388288497925, 0.5867773294448853], "r_norm": [0.2010490745306015, 0.199209526181221, 0.2040197104215622]}, {"step": 2800, "epoch": 5, "train_loss": 0.04737924411892891, "cos_r_negg": [0.587195873260498, 0.629308819770813, 0.5862987637519836], "cos_innovation_negg": [0.587195873260498, 0.629308819770813, 0.5862987637519836], "cos_apical_negg": [0.04214123636484146, 0.091066375374794, 0.13282956182956696], "cos_Ac_negg": [0.6158110499382019, 0.6564095616340637, 0.5958699584007263], "r_norm": [0.21488341689109802, 0.21427884697914124, 0.2195425182580948]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9702, "test_loss": 0.11145480041503907}, {"step": 2900, "epoch": 6, "train_loss": 0.02309112623333931, "cos_r_negg": [0.565677285194397, 0.6131430864334106, 0.5850650072097778], "cos_innovation_negg": [0.565677285194397, 0.6131430864334106, 0.5850650072097778], "cos_apical_negg": [0.050060540437698364, 0.08606230467557907, 0.13248473405838013], "cos_Ac_negg": [0.6040457487106323, 0.6533260345458984, 0.6031286120414734], "r_norm": [0.20840919017791748, 0.20701190829277039, 0.21399539709091187]}, {"step": 3000, "epoch": 6, "train_loss": 0.07823412120342255, "cos_r_negg": [0.5583109259605408, 0.6086455583572388, 0.5694288015365601], "cos_innovation_negg": [0.5583109259605408, 0.6086455583572388, 0.5694288015365601], "cos_apical_negg": [0.036664627492427826, 0.08008675277233124, 0.13559503853321075], "cos_Ac_negg": [0.6049589514732361, 0.6520075798034668, 0.5900802612304688], "r_norm": [0.19045023620128632, 0.18850593268871307, 0.19622398912906647], "ldr": {"dL_sdil": -0.005181856453418732, "dL_bp": -0.006142161786556244, "ratio": 0.8436535268023392}}, {"step": 3100, "epoch": 6, "train_loss": 0.013891670852899551, "cos_r_negg": [0.5546495914459229, 0.6070494651794434, 0.5782874822616577], "cos_innovation_negg": [0.5546495914459229, 0.6070494651794434, 0.5782874822616577], "cos_apical_negg": [0.03110378235578537, 0.07507039606571198, 0.12861861288547516], "cos_Ac_negg": [0.6110647916793823, 0.6585196256637573, 0.6041256189346313], "r_norm": [0.18141630291938782, 0.180147185921669, 0.18428057432174683]}, {"step": 3200, "epoch": 6, "train_loss": 0.022837044671177864, "cos_r_negg": [0.5567091703414917, 0.5884954333305359, 0.5748907327651978], "cos_innovation_negg": [0.5567091703414917, 0.5884954333305359, 0.5748907327651978], "cos_apical_negg": [0.03017181158065796, 0.07870163023471832, 0.1239771842956543], "cos_Ac_negg": [0.6206838488578796, 0.6521150469779968, 0.614775538444519], "r_norm": [0.2074543535709381, 0.2065643072128296, 0.21191434562206268]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9699, "test_loss": 0.12311956329345704}, {"step": 3300, "epoch": 7, "train_loss": 0.0072054145857691765, "cos_r_negg": [0.5515697002410889, 0.5856539011001587, 0.5689003467559814], "cos_innovation_negg": [0.5515697002410889, 0.5856539011001587, 0.5689003467559814], "cos_apical_negg": [0.0335095189511776, 0.07867659628391266, 0.12692901492118835], "cos_Ac_negg": [0.6189903616905212, 0.6502972841262817, 0.6095094084739685], "r_norm": [0.199436753988266, 0.1982872486114502, 0.20319654047489166]}, {"step": 3400, "epoch": 7, "train_loss": 0.023439813405275345, "cos_r_negg": [0.5239970684051514, 0.5703652501106262, 0.5639129877090454], "cos_innovation_negg": [0.5239970684051514, 0.5703652501106262, 0.5639129877090454], "cos_apical_negg": [0.03912169858813286, 0.08066577464342117, 0.13582327961921692], "cos_Ac_negg": [0.6044108271598816, 0.6493125557899475, 0.6172916293144226], "r_norm": [0.21733565628528595, 0.21749639511108398, 0.2214263528585434]}, {"step": 3500, "epoch": 7, "train_loss": 0.0055810222402215, "cos_r_negg": [0.506786048412323, 0.5491162538528442, 0.5527008175849915], "cos_innovation_negg": [0.506786048412323, 0.5491162538528442, 0.5527008175849915], "cos_apical_negg": [0.023315371945500374, 0.08084549754858017, 0.13188579678535461], "cos_Ac_negg": [0.6118475794792175, 0.6495540142059326, 0.6248893737792969], "r_norm": [0.17132240533828735, 0.16975370049476624, 0.17366693913936615], "ldr": {"dL_sdil": -0.0044611915946006775, "dL_bp": -0.005936257541179657, "ratio": 0.7515158437203091}}, {"step": 3600, "epoch": 7, "train_loss": 0.011079233139753342, "cos_r_negg": [0.503638505935669, 0.5509056448936462, 0.5584039688110352], "cos_innovation_negg": [0.503638505935669, 0.5509056448936462, 0.5584039688110352], "cos_apical_negg": [0.032642610371112823, 0.08205906301736832, 0.13160008192062378], "cos_Ac_negg": [0.6165440678596497, 0.6597976684570312, 0.6356155872344971], "r_norm": [0.1889066845178604, 0.18782877922058105, 0.19196265935897827]}, {"step": 3700, "epoch": 7, "train_loss": 0.019731123000383377, "cos_r_negg": [0.5026959776878357, 0.5492781400680542, 0.5546693801879883], "cos_innovation_negg": [0.5026959776878357, 0.5492781400680542, 0.5546693801879883], "cos_apical_negg": [0.028425056487321854, 0.07734984159469604, 0.12169761210680008], "cos_Ac_negg": [0.6090456247329712, 0.6525936126708984, 0.6372885704040527], "r_norm": [0.20464704930782318, 0.20316657423973083, 0.20876076817512512]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9714, "test_loss": 0.12079893531799317}, {"step": 3800, "epoch": 8, "train_loss": 0.005900415126234293, "cos_r_negg": [0.5039244294166565, 0.5404571294784546, 0.5490973591804504], "cos_innovation_negg": [0.5039244294166565, 0.5404571294784546, 0.5490973591804504], "cos_apical_negg": [0.03376132994890213, 0.07072678208351135, 0.12185116112232208], "cos_Ac_negg": [0.6261590719223022, 0.6571127772331238, 0.6360654830932617], "r_norm": [0.17702902853488922, 0.17459113895893097, 0.18020063638687134]}, {"step": 3900, "epoch": 8, "train_loss": 0.009754760190844536, "cos_r_negg": [0.4679676294326782, 0.5141180753707886, 0.524848222732544], "cos_innovation_negg": [0.4679676294326782, 0.5141180753707886, 0.524848222732544], "cos_apical_negg": [0.027421411126852036, 0.06862695515155792, 0.1141679435968399], "cos_Ac_negg": [0.6102022528648376, 0.6468222141265869, 0.63695228099823], "r_norm": [0.18532976508140564, 0.18423914909362793, 0.18915408849716187]}, {"step": 4000, "epoch": 8, "train_loss": 0.007884769700467587, "cos_r_negg": [0.4535374045372009, 0.4969760775566101, 0.5088882446289062], "cos_innovation_negg": [0.4535374045372009, 0.4969760775566101, 0.5088882446289062], "cos_apical_negg": [0.018973732367157936, 0.06523087620735168, 0.1068141981959343], "cos_Ac_negg": [0.6104272603988647, 0.6508997678756714, 0.6331040859222412], "r_norm": [0.17583787441253662, 0.17401285469532013, 0.1786840558052063], "ldr": {"dL_sdil": -0.006151571869850159, "dL_bp": -0.009266041219234467, "ratio": 0.6638834993611634}}, {"step": 4100, "epoch": 8, "train_loss": 0.0251475740224123, "cos_r_negg": [0.4754554033279419, 0.5207161903381348, 0.5318887233734131], "cos_innovation_negg": [0.4754554033279419, 0.5207161903381348, 0.5318887233734131], "cos_apical_negg": [0.021943775936961174, 0.07283858954906464, 0.11243140697479248], "cos_Ac_negg": [0.6093577742576599, 0.6566318273544312, 0.6427503228187561], "r_norm": [0.19794800877571106, 0.19629013538360596, 0.2009926587343216]}, {"step": 4200, "epoch": 8, "train_loss": 0.021346043795347214, "cos_r_negg": [0.435648113489151, 0.48645347356796265, 0.49681100249290466], "cos_innovation_negg": [0.435648113489151, 0.48645347356796265, 0.49681100249290466], "cos_apical_negg": [0.02471092715859413, 0.06892858445644379, 0.10212395340204239], "cos_Ac_negg": [0.6034913063049316, 0.6531990766525269, 0.6370819211006165], "r_norm": [0.1873244047164917, 0.18534408509731293, 0.19004282355308533]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9714, "test_loss": 0.13042741432189942}, {"step": 4300, "epoch": 9, "train_loss": 0.00207694573327899, "cos_r_negg": [0.4229526221752167, 0.4700314402580261, 0.48948386311531067], "cos_innovation_negg": [0.4229526221752167, 0.4700314402580261, 0.48948386311531067], "cos_apical_negg": [0.022869721055030823, 0.0754602700471878, 0.11757481098175049], "cos_Ac_negg": [0.6045072078704834, 0.6531006097793579, 0.6414310932159424], "r_norm": [0.19610975682735443, 0.19450807571411133, 0.19950783252716064]}, {"step": 4400, "epoch": 9, "train_loss": 0.0006989761022850871, "cos_r_negg": [0.43930789828300476, 0.48184335231781006, 0.4982673227787018], "cos_innovation_negg": [0.43930789828300476, 0.48184335231781006, 0.4982673227787018], "cos_apical_negg": [0.02033047005534172, 0.06314921379089355, 0.1044493168592453], "cos_Ac_negg": [0.6156895160675049, 0.660896360874176, 0.6490445137023926], "r_norm": [0.188154935836792, 0.1873258352279663, 0.19170327484607697]}, {"step": 4500, "epoch": 9, "train_loss": 0.0670974925160408, "cos_r_negg": [0.40810462832450867, 0.45823216438293457, 0.47662705183029175], "cos_innovation_negg": [0.40810462832450867, 0.45823216438293457, 0.47662705183029175], "cos_apical_negg": [0.013115651905536652, 0.06181355565786362, 0.10247022658586502], "cos_Ac_negg": [0.6016995906829834, 0.6549803018569946, 0.6481812000274658], "r_norm": [0.16602489352226257, 0.16475547850131989, 0.17002645134925842], "ldr": {"dL_sdil": -0.005957663059234619, "dL_bp": -0.009596183896064758, "ratio": 0.6208366913099448}}, {"step": 4600, "epoch": 9, "train_loss": 0.003053926629945636, "cos_r_negg": [0.413961261510849, 0.4633459150791168, 0.48455673456192017], "cos_innovation_negg": [0.413961261510849, 0.4633459150791168, 0.48455673456192017], "cos_apical_negg": [0.02628561109304428, 0.06421220302581787, 0.10565322637557983], "cos_Ac_negg": [0.5981574058532715, 0.6555895805358887, 0.6470040082931519], "r_norm": [0.17618034780025482, 0.1753236949443817, 0.18065184354782104]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9723, "test_loss": 0.13113348865509034}, {"step": 4700, "epoch": 10, "train_loss": 0.001840092591010034, "cos_r_negg": [0.3906187415122986, 0.4429139196872711, 0.4664008617401123], "cos_innovation_negg": [0.3906187415122986, 0.4429139196872711, 0.4664008617401123], "cos_apical_negg": [0.02556668035686016, 0.06279965490102768, 0.10503296554088593], "cos_Ac_negg": [0.6065284013748169, 0.6604180335998535, 0.6510452032089233], "r_norm": [0.1806732714176178, 0.18004316091537476, 0.18403521180152893]}, {"step": 4800, "epoch": 10, "train_loss": 0.002465236699208617, "cos_r_negg": [0.3616454601287842, 0.4148251712322235, 0.4383332133293152], "cos_innovation_negg": [0.3616454601287842, 0.4148251712322235, 0.4383332133293152], "cos_apical_negg": [0.019552316516637802, 0.06104365363717079, 0.1010805070400238], "cos_Ac_negg": [0.6002198457717896, 0.6596728563308716, 0.6473992466926575], "r_norm": [0.16908667981624603, 0.16726604104042053, 0.1718483567237854]}, {"step": 4900, "epoch": 10, "train_loss": 0.0009035018738359213, "cos_r_negg": [0.3674590587615967, 0.41435733437538147, 0.44010066986083984], "cos_innovation_negg": [0.3674590587615967, 0.41435733437538147, 0.44010066986083984], "cos_apical_negg": [0.016901742666959763, 0.0587710440158844, 0.10026903450489044], "cos_Ac_negg": [0.598545491695404, 0.6587402820587158, 0.6462540030479431], "r_norm": [0.16530188918113708, 0.16370832920074463, 0.16769784688949585]}, {"step": 5000, "epoch": 10, "train_loss": 0.001396470470353961, "cos_r_negg": [0.35372424125671387, 0.40318048000335693, 0.42876148223876953], "cos_innovation_negg": [0.35372424125671387, 0.40318048000335693, 0.42876148223876953], "cos_apical_negg": [0.015281818807125092, 0.06484194099903107, 0.10090282559394836], "cos_Ac_negg": [0.5906784534454346, 0.6562011241912842, 0.6522396206855774], "r_norm": [0.16465550661087036, 0.16371382772922516, 0.1678890585899353], "ldr": {"dL_sdil": -0.006827443838119507, "dL_bp": -0.011931560933589935, "ratio": 0.5722171538259315}}, {"step": 5100, "epoch": 10, "train_loss": 0.0065423608757555485, "cos_r_negg": [0.3494974970817566, 0.4002249836921692, 0.4248924255371094], "cos_innovation_negg": [0.3494974970817566, 0.4002249836921692, 0.4248924255371094], "cos_apical_negg": [0.010526479221880436, 0.060380127280950546, 0.09512849152088165], "cos_Ac_negg": [0.6005849242210388, 0.6609558463096619, 0.6511019468307495], "r_norm": [0.17341387271881104, 0.1718728244304657, 0.1758197844028473]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9719, "test_loss": 0.13702383298873902}, {"step": 5200, "epoch": 11, "train_loss": 0.0007825077045708895, "cos_r_negg": [0.352170467376709, 0.4004124104976654, 0.42119207978248596], "cos_innovation_negg": [0.352170467376709, 0.4004124104976654, 0.42119207978248596], "cos_apical_negg": [0.009563511237502098, 0.05999604985117912, 0.0949019342660904], "cos_Ac_negg": [0.6010615825653076, 0.6613147854804993, 0.6512768864631653], "r_norm": [0.18210366368293762, 0.18123450875282288, 0.18493232131004333]}, {"step": 5300, "epoch": 11, "train_loss": 0.001030321465805173, "cos_r_negg": [0.3405805230140686, 0.39221519231796265, 0.4158976376056671], "cos_innovation_negg": [0.3405805230140686, 0.39221519231796265, 0.4158976376056671], "cos_apical_negg": [0.006938896141946316, 0.05888422578573227, 0.09161244332790375], "cos_Ac_negg": [0.5965147018432617, 0.6610060334205627, 0.6546416282653809], "r_norm": [0.17633305490016937, 0.17482371628284454, 0.178761824965477]}, {"step": 5400, "epoch": 11, "train_loss": 0.0042540449649095535, "cos_r_negg": [0.31929224729537964, 0.37281617522239685, 0.39695513248443604], "cos_innovation_negg": [0.31929224729537964, 0.37281617522239685, 0.39695513248443604], "cos_apical_negg": [0.007814566604793072, 0.055944036692380905, 0.08859697729349136], "cos_Ac_negg": [0.5906990766525269, 0.6589862108230591, 0.6522208452224731], "r_norm": [0.1626206636428833, 0.16141757369041443, 0.16495022177696228]}, {"step": 5500, "epoch": 11, "train_loss": 0.00109402637463063, "cos_r_negg": [0.3236708641052246, 0.37145695090293884, 0.3936476707458496], "cos_innovation_negg": [0.3236708641052246, 0.37145695090293884, 0.3936476707458496], "cos_apical_negg": [0.009464427828788757, 0.05492686852812767, 0.08791422843933105], "cos_Ac_negg": [0.5948855876922607, 0.6594706773757935, 0.6522601842880249], "r_norm": [0.16877172887325287, 0.16703590750694275, 0.17109626531600952], "ldr": {"dL_sdil": -0.0071260035037994385, "dL_bp": -0.01254795491695404, "ratio": 0.5679015864307269}}, {"step": 5600, "epoch": 11, "train_loss": 0.0003725747228600085, "cos_r_negg": [0.30669617652893066, 0.3494783639907837, 0.3695250451564789], "cos_innovation_negg": [0.30669617652893066, 0.3494783639907837, 0.3695250451564789], "cos_apical_negg": [0.008274903520941734, 0.05524459481239319, 0.08129556477069855], "cos_Ac_negg": [0.5895830988883972, 0.6564111113548279, 0.6514990329742432], "r_norm": [0.16499063372612, 0.16279491782188416, 0.1670016497373581]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9722, "test_loss": 0.13796027421951293}, {"step": 5700, "epoch": 12, "train_loss": 0.00023269107623491436, "cos_r_negg": [0.3141196668148041, 0.3611403703689575, 0.384514719247818], "cos_innovation_negg": [0.3141196668148041, 0.3611403703689575, 0.384514719247818], "cos_apical_negg": [0.006977427750825882, 0.05657617002725601, 0.08557741343975067], "cos_Ac_negg": [0.5944322943687439, 0.6608906984329224, 0.655097484588623], "r_norm": [0.16944685578346252, 0.16728347539901733, 0.17163588106632233]}, {"step": 5800, "epoch": 12, "train_loss": 0.00024193685385398567, "cos_r_negg": [0.3005027174949646, 0.3442278504371643, 0.36499232053756714], "cos_innovation_negg": [0.3005027174949646, 0.3442278504371643, 0.36499232053756714], "cos_apical_negg": [0.006468859501183033, 0.053578704595565796, 0.07757032662630081], "cos_Ac_negg": [0.584417462348938, 0.654517650604248, 0.6532986164093018], "r_norm": [0.16966773569583893, 0.1676851212978363, 0.17237864434719086]}, {"step": 5900, "epoch": 12, "train_loss": 0.0008451087051071227, "cos_r_negg": [0.2991112470626831, 0.3457573652267456, 0.36647093296051025], "cos_innovation_negg": [0.2991112470626831, 0.3457573652267456, 0.36647093296051025], "cos_apical_negg": [0.005139848217368126, 0.0564047209918499, 0.08021765947341919], "cos_Ac_negg": [0.5839756727218628, 0.655193567276001, 0.6558170318603516], "r_norm": [0.17006580531597137, 0.16849187016487122, 0.17286396026611328]}, {"step": 6000, "epoch": 12, "train_loss": 0.000429863459430635, "cos_r_negg": [0.2971387505531311, 0.34047025442123413, 0.3639088273048401], "cos_innovation_negg": [0.2971387505531311, 0.34047025442123413, 0.3639088273048401], "cos_apical_negg": [0.006260121241211891, 0.05526921898126602, 0.08228494971990585], "cos_Ac_negg": [0.5862312316894531, 0.6530728340148926, 0.6545469760894775], "r_norm": [0.16694281995296478, 0.16536740958690643, 0.16937360167503357], "ldr": {"dL_sdil": -0.007939636707305908, "dL_bp": -0.014572396874427795, "ratio": 0.5448408230796054}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9722, "test_loss": 0.1420075569152832}, {"step": 6100, "epoch": 13, "train_loss": 0.00045817336649633944, "cos_r_negg": [0.29320865869522095, 0.3298850655555725, 0.3504391610622406], "cos_innovation_negg": [0.29320865869522095, 0.3298850655555725, 0.3504391610622406], "cos_apical_negg": [0.007446726784110069, 0.05385761708021164, 0.07755032181739807], "cos_Ac_negg": [0.5862466096878052, 0.654208779335022, 0.6529350280761719], "r_norm": [0.1695239543914795, 0.16792729496955872, 0.17198199033737183]}, {"step": 6200, "epoch": 13, "train_loss": 0.0005318940966390073, "cos_r_negg": [0.28840914368629456, 0.32514163851737976, 0.3477312922477722], "cos_innovation_negg": [0.28840914368629456, 0.32514163851737976, 0.3477312922477722], "cos_apical_negg": [0.009004191495478153, 0.053208671510219574, 0.07626290619373322], "cos_Ac_negg": [0.5882088541984558, 0.6548550128936768, 0.6557705402374268], "r_norm": [0.17259481549263, 0.1705859899520874, 0.17510265111923218]}, {"step": 6300, "epoch": 13, "train_loss": 0.0020512929186224937, "cos_r_negg": [0.2850475311279297, 0.3188374638557434, 0.3398038148880005], "cos_innovation_negg": [0.2850475311279297, 0.3188374638557434, 0.3398038148880005], "cos_apical_negg": [0.008397301658987999, 0.05286361277103424, 0.07560718804597855], "cos_Ac_negg": [0.5849769115447998, 0.6537105441093445, 0.6546311974525452], "r_norm": [0.16976244747638702, 0.16809657216072083, 0.17217665910720825]}, {"step": 6400, "epoch": 13, "train_loss": 0.00027335615595802665, "cos_r_negg": [0.28622734546661377, 0.3215796947479248, 0.34399476647377014], "cos_innovation_negg": [0.28622734546661377, 0.3215796947479248, 0.34399476647377014], "cos_apical_negg": [0.007981441915035248, 0.053553562611341476, 0.0770130604505539], "cos_Ac_negg": [0.5863303542137146, 0.6551163196563721, 0.6560806035995483], "r_norm": [0.1701076626777649, 0.16836421191692352, 0.17273661494255066]}, {"step": 6500, "epoch": 13, "train_loss": 0.0001373695704387501, "cos_r_negg": [0.2824729084968567, 0.31745219230651855, 0.3403700590133667], "cos_innovation_negg": [0.2824729084968567, 0.31745219230651855, 0.3403700590133667], "cos_apical_negg": [0.009321839548647404, 0.05357018858194351, 0.07671408355236053], "cos_Ac_negg": [0.5849481821060181, 0.6539113521575928, 0.6556531190872192], "r_norm": [0.1614467054605484, 0.15989987552165985, 0.16377681493759155], "ldr": {"dL_sdil": -0.007981009781360626, "dL_bp": -0.01510775089263916, "ratio": 0.528272529648947}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9719, "test_loss": 0.14431319007873536}, {"step": 6600, "epoch": 14, "train_loss": 0.00011856538185384125, "cos_r_negg": [0.27665117383003235, 0.3105822801589966, 0.334550678730011], "cos_innovation_negg": [0.27665117383003235, 0.3105822801589966, 0.334550678730011], "cos_apical_negg": [0.007295699790120125, 0.05206109210848808, 0.07550030946731567], "cos_Ac_negg": [0.5844250917434692, 0.6526857614517212, 0.656484842300415], "r_norm": [0.16773851215839386, 0.16588163375854492, 0.16997557878494263]}, {"step": 6700, "epoch": 14, "train_loss": 0.0008524881559424102, "cos_r_negg": [0.2749953866004944, 0.31169018149375916, 0.33443671464920044], "cos_innovation_negg": [0.2749953866004944, 0.31169018149375916, 0.33443671464920044], "cos_apical_negg": [0.007936772890388966, 0.05364764854311943, 0.0748809427022934], "cos_Ac_negg": [0.579275369644165, 0.6512957811355591, 0.656309187412262], "r_norm": [0.16454872488975525, 0.16312849521636963, 0.16714921593666077]}, {"step": 6800, "epoch": 14, "train_loss": 0.0016292296350002289, "cos_r_negg": [0.26679256558418274, 0.3007053732872009, 0.3244709372520447], "cos_innovation_negg": [0.26679256558418274, 0.3007053732872009, 0.3244709372520447], "cos_apical_negg": [0.0077690621837973595, 0.05183498561382294, 0.07318760454654694], "cos_Ac_negg": [0.58195960521698, 0.6518175005912781, 0.6554392576217651], "r_norm": [0.16643011569976807, 0.16466417908668518, 0.16892865300178528]}, {"step": 6900, "epoch": 14, "train_loss": 0.0006684634136036038, "cos_r_negg": [0.26337456703186035, 0.2981535792350769, 0.32045748829841614], "cos_innovation_negg": [0.26337456703186035, 0.2981535792350769, 0.32045748829841614], "cos_apical_negg": [0.007672143168747425, 0.05288967117667198, 0.07224975526332855], "cos_Ac_negg": [0.5826812982559204, 0.6537188291549683, 0.656615138053894], "r_norm": [0.16910724341869354, 0.1671365648508072, 0.1714499294757843]}, {"step": 7000, "epoch": 14, "train_loss": 0.0012496009003371, "cos_r_negg": [0.2641991078853607, 0.2971504330635071, 0.3216572403907776], "cos_innovation_negg": [0.2641991078853607, 0.2971504330635071, 0.3216572403907776], "cos_apical_negg": [0.00854979082942009, 0.051634564995765686, 0.0736464112997055], "cos_Ac_negg": [0.5844832062721252, 0.6534990072250366, 0.6575264930725098], "r_norm": [0.16546377539634705, 0.16344332695007324, 0.16770461201667786], "ldr": {"dL_sdil": -0.008243657648563385, "dL_bp": -0.016118310391902924, "ratio": 0.5114467613618242}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9721, "test_loss": 0.14751017742156983}], "final": {"test_acc": 0.9721, "test_loss": 0.14751017742156983, "wall_s": 22.025368452072144, "cos_r_negg": [0.260580837726593, 0.29289305210113525, 0.31671878695487976], "cos_innovation_negg": [0.260580837726593, 0.29289305210113525, 0.31671878695487976], "cos_apical_negg": [0.009301045909523964, 0.0510302409529686, 0.07210469245910645], "cos_Ac_negg": [0.5810582041740417, 0.6503059267997742, 0.6564116477966309]}}