summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p05_residual_s0.json
blob: 3c0a6ac17893f468f24fcbbcacb545868d461ce2 (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": 0.05, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p05_residual_s0", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3603768348693848, "cos_r_negg": [-0.01952279359102249, -0.025533504784107208, 0.033745452761650085], "cos_innovation_negg": [-0.01952279359102249, -0.025533504784107208, 0.033745452761650085], "cos_apical_negg": [-0.019518455490469933, -0.025709278881549835, 0.03309042379260063], "cos_Ac_negg": [-0.01952274888753891, -0.02553350292146206, 0.03374544531106949], "r_norm": [4.898251056671143, 4.826769828796387, 4.823906898498535], "ldr": {"dL_sdil": -0.009195327758789062, "dL_bp": -0.2406785488128662, "ratio": 0.03820584677838766}}, {"step": 100, "epoch": 0, "train_loss": 0.20553797483444214, "cos_r_negg": [0.4579622447490692, 0.38216572999954224, 0.3608860969543457], "cos_innovation_negg": [0.4579622447490692, 0.38216572999954224, 0.3608860969543457], "cos_apical_negg": [0.19300329685211182, 0.17585119605064392, 0.21503859758377075], "cos_Ac_negg": [0.4582347273826599, 0.3824673295021057, 0.3610897660255432], "r_norm": [0.8250458240509033, 0.8087376356124878, 0.8063828945159912]}, {"step": 200, "epoch": 0, "train_loss": 0.3347880244255066, "cos_r_negg": [0.484967976808548, 0.40824341773986816, 0.3901892900466919], "cos_innovation_negg": [0.484967976808548, 0.40824341773986816, 0.3901892900466919], "cos_apical_negg": [0.18476158380508423, 0.17060211300849915, 0.2136172205209732], "cos_Ac_negg": [0.48558053374290466, 0.40904223918914795, 0.39106833934783936], "r_norm": [0.739167332649231, 0.7221766710281372, 0.7204556465148926]}, {"step": 300, "epoch": 0, "train_loss": 0.1608467698097229, "cos_r_negg": [0.5106900334358215, 0.4493744671344757, 0.4224192500114441], "cos_innovation_negg": [0.5106900334358215, 0.4493744671344757, 0.4224192500114441], "cos_apical_negg": [0.1733049750328064, 0.1539825201034546, 0.1863439977169037], "cos_Ac_negg": [0.5108871459960938, 0.44945770502090454, 0.42250585556030273], "r_norm": [0.5548641681671143, 0.5450409054756165, 0.5431962013244629]}, {"step": 400, "epoch": 0, "train_loss": 0.18942078948020935, "cos_r_negg": [0.5131446123123169, 0.4551883339881897, 0.42917412519454956], "cos_innovation_negg": [0.5131446123123169, 0.4551883339881897, 0.42917412519454956], "cos_apical_negg": [0.15334802865982056, 0.15426218509674072, 0.18405623733997345], "cos_Ac_negg": [0.5133811235427856, 0.4552389979362488, 0.4292002320289612], "r_norm": [0.5758621096611023, 0.5641695857048035, 0.5628612041473389]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9464, "test_loss": 0.1857574264526367}, {"step": 500, "epoch": 1, "train_loss": 0.2046205848455429, "cos_r_negg": [0.523687481880188, 0.4869231581687927, 0.4766062796115875], "cos_innovation_negg": [0.523687481880188, 0.4869231581687927, 0.4766062796115875], "cos_apical_negg": [0.13296380639076233, 0.1439264565706253, 0.18738295137882233], "cos_Ac_negg": [0.5244132876396179, 0.4869997501373291, 0.4767279326915741], "r_norm": [0.5014256238937378, 0.4845983386039734, 0.4856627881526947], "ldr": {"dL_sdil": -0.012597888708114624, "dL_bp": -0.010928422212600708, "ratio": 1.1527637259099475}}, {"step": 600, "epoch": 1, "train_loss": 0.08640950918197632, "cos_r_negg": [0.5135207176208496, 0.48049765825271606, 0.4631809592247009], "cos_innovation_negg": [0.5135207176208496, 0.48049765825271606, 0.4631809592247009], "cos_apical_negg": [0.13356101512908936, 0.13810202479362488, 0.18871255218982697], "cos_Ac_negg": [0.5141767859458923, 0.48077037930488586, 0.46318358182907104], "r_norm": [0.4284065067768097, 0.4179823696613312, 0.4188630282878876]}, {"step": 700, "epoch": 1, "train_loss": 0.22193440794944763, "cos_r_negg": [0.5253515839576721, 0.49967077374458313, 0.47384780645370483], "cos_innovation_negg": [0.5253515839576721, 0.49967077374458313, 0.47384780645370483], "cos_apical_negg": [0.13406303524971008, 0.13647781312465668, 0.20093558728694916], "cos_Ac_negg": [0.5262846350669861, 0.49967145919799805, 0.47360819578170776], "r_norm": [0.4450308084487915, 0.4301424026489258, 0.43267372250556946]}, {"step": 800, "epoch": 1, "train_loss": 0.16923829913139343, "cos_r_negg": [0.5265134572982788, 0.5200668573379517, 0.48621031641960144], "cos_innovation_negg": [0.5265134572982788, 0.5200668573379517, 0.48621031641960144], "cos_apical_negg": [0.1261288970708847, 0.12328580021858215, 0.1760111153125763], "cos_Ac_negg": [0.5271130800247192, 0.5202675461769104, 0.4861285984516144], "r_norm": [0.3633059561252594, 0.35162240266799927, 0.35249751806259155]}, {"step": 900, "epoch": 1, "train_loss": 0.2054840624332428, "cos_r_negg": [0.5366386771202087, 0.5371055603027344, 0.4891264736652374], "cos_innovation_negg": [0.5366386771202087, 0.5371055603027344, 0.4891264736652374], "cos_apical_negg": [0.12613967061042786, 0.1435614675283432, 0.17565500736236572], "cos_Ac_negg": [0.536929726600647, 0.5370739102363586, 0.4889616072177887], "r_norm": [0.3984663486480713, 0.3892437219619751, 0.38930922746658325]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9569, "test_loss": 0.14992256050109865}, {"step": 1000, "epoch": 2, "train_loss": 0.03922855854034424, "cos_r_negg": [0.5354604721069336, 0.5387663841247559, 0.49675408005714417], "cos_innovation_negg": [0.5354604721069336, 0.5387663841247559, 0.49675408005714417], "cos_apical_negg": [0.10201052576303482, 0.11773139238357544, 0.16800633072853088], "cos_Ac_negg": [0.5359271764755249, 0.538806140422821, 0.49607735872268677], "r_norm": [0.33832865953445435, 0.32900702953338623, 0.3283848166465759], "ldr": {"dL_sdil": -0.00833958387374878, "dL_bp": -0.0074491947889328, "ratio": 1.1195282322511988}}, {"step": 1100, "epoch": 2, "train_loss": 0.1274726837873459, "cos_r_negg": [0.5607089400291443, 0.5610537528991699, 0.5137262344360352], "cos_innovation_negg": [0.5607089400291443, 0.5610537528991699, 0.5137262344360352], "cos_apical_negg": [0.10683763772249222, 0.12646016478538513, 0.16395971179008484], "cos_Ac_negg": [0.5610625743865967, 0.561252236366272, 0.5135065913200378], "r_norm": [0.3592675030231476, 0.34692731499671936, 0.3486380875110626]}, {"step": 1200, "epoch": 2, "train_loss": 0.15907587110996246, "cos_r_negg": [0.5624092817306519, 0.5603923797607422, 0.514489471912384], "cos_innovation_negg": [0.5624092817306519, 0.5603923797607422, 0.514489471912384], "cos_apical_negg": [0.10560047626495361, 0.1084093376994133, 0.14780128002166748], "cos_Ac_negg": [0.5627895593643188, 0.5604621767997742, 0.5141286253929138], "r_norm": [0.32979658246040344, 0.3189634084701538, 0.3202262818813324]}, {"step": 1300, "epoch": 2, "train_loss": 0.12210610508918762, "cos_r_negg": [0.5728862285614014, 0.5762075185775757, 0.532311201095581], "cos_innovation_negg": [0.5728862285614014, 0.5762075185775757, 0.532311201095581], "cos_apical_negg": [0.09865699708461761, 0.11101168394088745, 0.15804867446422577], "cos_Ac_negg": [0.573574423789978, 0.5762590765953064, 0.5317193269729614], "r_norm": [0.3354035019874573, 0.3264496326446533, 0.32440072298049927]}, {"step": 1400, "epoch": 2, "train_loss": 0.16060467064380646, "cos_r_negg": [0.5716283321380615, 0.577843427658081, 0.5309174060821533], "cos_innovation_negg": [0.5716283321380615, 0.577843427658081, 0.5309174060821533], "cos_apical_negg": [0.10109320282936096, 0.12631100416183472, 0.16680295765399933], "cos_Ac_negg": [0.5754722952842712, 0.5790044069290161, 0.5308045744895935], "r_norm": [0.3186105191707611, 0.3082095980644226, 0.3089616894721985]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9655, "test_loss": 0.12051229820251465}, {"step": 1500, "epoch": 3, "train_loss": 0.06386608630418777, "cos_r_negg": [0.5805710554122925, 0.5955780744552612, 0.5463205575942993], "cos_innovation_negg": [0.5805710554122925, 0.5955780744552612, 0.5463205575942993], "cos_apical_negg": [0.07950741797685623, 0.10170567780733109, 0.1538735181093216], "cos_Ac_negg": [0.581992506980896, 0.5966823101043701, 0.5457150936126709], "r_norm": [0.2611697316169739, 0.2530985474586487, 0.2531789541244507], "ldr": {"dL_sdil": -0.00545801967382431, "dL_bp": -0.005244113504886627, "ratio": 1.0407897671814996}}, {"step": 1600, "epoch": 3, "train_loss": 0.06963219493627548, "cos_r_negg": [0.5691196918487549, 0.5914679765701294, 0.5502804517745972], "cos_innovation_negg": [0.5691196918487549, 0.5914679765701294, 0.5502804517745972], "cos_apical_negg": [0.08091358840465546, 0.10967737436294556, 0.1509772539138794], "cos_Ac_negg": [0.571120023727417, 0.5928018093109131, 0.5498823523521423], "r_norm": [0.2819720506668091, 0.2724171280860901, 0.27314233779907227]}, {"step": 1700, "epoch": 3, "train_loss": 0.06539039313793182, "cos_r_negg": [0.5813878774642944, 0.5956884026527405, 0.547366738319397], "cos_innovation_negg": [0.5813878774642944, 0.5956884026527405, 0.547366738319397], "cos_apical_negg": [0.0911014974117279, 0.10569155216217041, 0.15141209959983826], "cos_Ac_negg": [0.5834836959838867, 0.5967861413955688, 0.5467633008956909], "r_norm": [0.25422245264053345, 0.24650141596794128, 0.24772970378398895]}, {"step": 1800, "epoch": 3, "train_loss": 0.016470560804009438, "cos_r_negg": [0.5626012086868286, 0.5824457406997681, 0.5532260537147522], "cos_innovation_negg": [0.5626012086868286, 0.5824457406997681, 0.5532260537147522], "cos_apical_negg": [0.0819510817527771, 0.09958356618881226, 0.1566336750984192], "cos_Ac_negg": [0.568524956703186, 0.5860669612884521, 0.5519437789916992], "r_norm": [0.248224139213562, 0.2410775125026703, 0.24029222130775452]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9652, "test_loss": 0.12335003890991211}, {"step": 1900, "epoch": 4, "train_loss": 0.026236163452267647, "cos_r_negg": [0.5891915559768677, 0.6053824424743652, 0.5565018057823181], "cos_innovation_negg": [0.5891915559768677, 0.6053824424743652, 0.5565018057823181], "cos_apical_negg": [0.09248057007789612, 0.1169896274805069, 0.16277912259101868], "cos_Ac_negg": [0.5900899171829224, 0.6065192222595215, 0.5559866428375244], "r_norm": [0.2781800627708435, 0.27161142230033875, 0.27034926414489746]}, {"step": 2000, "epoch": 4, "train_loss": 0.057471826672554016, "cos_r_negg": [0.569402813911438, 0.5875678062438965, 0.5663565397262573], "cos_innovation_negg": [0.569402813911438, 0.5875678062438965, 0.5663565397262573], "cos_apical_negg": [0.052954062819480896, 0.0836525708436966, 0.13660719990730286], "cos_Ac_negg": [0.5739847421646118, 0.5923200845718384, 0.5661741495132446], "r_norm": [0.21866536140441895, 0.21067006886005402, 0.21089935302734375], "ldr": {"dL_sdil": -0.004901431500911713, "dL_bp": -0.005205012857913971, "ratio": 0.9416751955683111}}, {"step": 2100, "epoch": 4, "train_loss": 0.018851609900593758, "cos_r_negg": [0.5779240131378174, 0.5937851667404175, 0.5687053799629211], "cos_innovation_negg": [0.5779240131378174, 0.5937851667404175, 0.5687053799629211], "cos_apical_negg": [0.060382746160030365, 0.09317074716091156, 0.14972719550132751], "cos_Ac_negg": [0.5829090476036072, 0.596392035484314, 0.5674302577972412], "r_norm": [0.25009214878082275, 0.24291518330574036, 0.24175116419792175]}, {"step": 2200, "epoch": 4, "train_loss": 0.02432655729353428, "cos_r_negg": [0.5698356628417969, 0.5986080169677734, 0.5691801905632019], "cos_innovation_negg": [0.5698356628417969, 0.5986080169677734, 0.5691801905632019], "cos_apical_negg": [0.04793553054332733, 0.09527935087680817, 0.14640530943870544], "cos_Ac_negg": [0.5759973526000977, 0.6028170585632324, 0.5682347416877747], "r_norm": [0.22581236064434052, 0.2182098776102066, 0.2178492397069931]}, {"step": 2300, "epoch": 4, "train_loss": 0.1269402801990509, "cos_r_negg": [0.5726011991500854, 0.5903053283691406, 0.5651850700378418], "cos_innovation_negg": [0.5726011991500854, 0.5903053283691406, 0.5651850700378418], "cos_apical_negg": [0.07685291767120361, 0.09186537563800812, 0.13838668167591095], "cos_Ac_negg": [0.5830402970314026, 0.6027272939682007, 0.5688943862915039], "r_norm": [0.2527035176753998, 0.24287879467010498, 0.2436361163854599]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9693, "test_loss": 0.10827527141571044}, {"step": 2400, "epoch": 5, "train_loss": 0.08427467942237854, "cos_r_negg": [0.5815878510475159, 0.5953119993209839, 0.5846785306930542], "cos_innovation_negg": [0.5815878510475159, 0.5953119993209839, 0.5846785306930542], "cos_apical_negg": [0.07297449558973312, 0.08918994665145874, 0.1430235356092453], "cos_Ac_negg": [0.5873320698738098, 0.6000885963439941, 0.5838549137115479], "r_norm": [0.21910353004932404, 0.21354863047599792, 0.21174027025699615]}, {"step": 2500, "epoch": 5, "train_loss": 0.03700476884841919, "cos_r_negg": [0.5740196704864502, 0.5969030857086182, 0.5744849443435669], "cos_innovation_negg": [0.5740196704864502, 0.5969030857086182, 0.5744849443435669], "cos_apical_negg": [0.04668359458446503, 0.08081283420324326, 0.1406819075345993], "cos_Ac_negg": [0.5860415101051331, 0.6097989082336426, 0.5790238380432129], "r_norm": [0.21384358406066895, 0.20827405154705048, 0.20768722891807556], "ldr": {"dL_sdil": -0.005959019064903259, "dL_bp": -0.007187977433204651, "ratio": 0.8290258449304175}}, {"step": 2600, "epoch": 5, "train_loss": 0.03783847391605377, "cos_r_negg": [0.5712009072303772, 0.5938788652420044, 0.5852274298667908], "cos_innovation_negg": [0.5712009072303772, 0.5938788652420044, 0.5852274298667908], "cos_apical_negg": [0.0588875412940979, 0.09509463608264923, 0.14293599128723145], "cos_Ac_negg": [0.5924674272537231, 0.6127462983131409, 0.5921125411987305], "r_norm": [0.23549771308898926, 0.23063766956329346, 0.22743842005729675]}, {"step": 2700, "epoch": 5, "train_loss": 0.025878164917230606, "cos_r_negg": [0.6008349061012268, 0.6140125393867493, 0.5990324020385742], "cos_innovation_negg": [0.6008349061012268, 0.6140125393867493, 0.5990324020385742], "cos_apical_negg": [0.04953507333993912, 0.07712547481060028, 0.13934241235256195], "cos_Ac_negg": [0.6149699091911316, 0.6274121999740601, 0.6048215627670288], "r_norm": [0.17445026338100433, 0.17008483409881592, 0.16803543269634247]}, {"step": 2800, "epoch": 5, "train_loss": 0.06025358662009239, "cos_r_negg": [0.5873588919639587, 0.6082230806350708, 0.5916528701782227], "cos_innovation_negg": [0.5873588919639587, 0.6082230806350708, 0.5916528701782227], "cos_apical_negg": [0.03817795217037201, 0.08858905732631683, 0.13129693269729614], "cos_Ac_negg": [0.6018168926239014, 0.6221031546592712, 0.6005264520645142], "r_norm": [0.21433258056640625, 0.20842435956001282, 0.2057434469461441]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9705, "test_loss": 0.11484633636474609}, {"step": 2900, "epoch": 6, "train_loss": 0.005533664021641016, "cos_r_negg": [0.5634152293205261, 0.5864613056182861, 0.5864205360412598], "cos_innovation_negg": [0.5634152293205261, 0.5864613056182861, 0.5864205360412598], "cos_apical_negg": [0.05011267215013504, 0.07456262409687042, 0.13208502531051636], "cos_Ac_negg": [0.5883058905601501, 0.6146793961524963, 0.6048269271850586], "r_norm": [0.21608708798885345, 0.2090061902999878, 0.20696677267551422]}, {"step": 3000, "epoch": 6, "train_loss": 0.04362904280424118, "cos_r_negg": [0.5723827481269836, 0.5900061130523682, 0.5797159075737], "cos_innovation_negg": [0.5723827481269836, 0.5900061130523682, 0.5797159075737], "cos_apical_negg": [0.05936134606599808, 0.06955280900001526, 0.13168014585971832], "cos_Ac_negg": [0.601454496383667, 0.6238418817520142, 0.6006356477737427], "r_norm": [0.18368461728096008, 0.17956972122192383, 0.1773810237646103], "ldr": {"dL_sdil": -0.005626410245895386, "dL_bp": -0.007024504244327545, "ratio": 0.8009690150644931}}, {"step": 3100, "epoch": 6, "train_loss": 0.015955157577991486, "cos_r_negg": [0.5732883214950562, 0.5980974435806274, 0.5891245603561401], "cos_innovation_negg": [0.5732883214950562, 0.5980974435806274, 0.5891245603561401], "cos_apical_negg": [0.041265539824962616, 0.07053113728761673, 0.12908563017845154], "cos_Ac_negg": [0.6080539226531982, 0.6350600719451904, 0.6158679723739624], "r_norm": [0.1764211803674698, 0.17159801721572876, 0.1705530285835266]}, {"step": 3200, "epoch": 6, "train_loss": 0.025097154080867767, "cos_r_negg": [0.5631382465362549, 0.5797849893569946, 0.5840267539024353], "cos_innovation_negg": [0.5631382465362549, 0.5797849893569946, 0.5840267539024353], "cos_apical_negg": [0.02611123025417328, 0.06582914292812347, 0.12043886631727219], "cos_Ac_negg": [0.6107301712036133, 0.6322115063667297, 0.6230251789093018], "r_norm": [0.1865541636943817, 0.1821870654821396, 0.18060094118118286]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9701, "test_loss": 0.12378705406188965}, {"step": 3300, "epoch": 7, "train_loss": 0.029362596571445465, "cos_r_negg": [0.5528621673583984, 0.5784894227981567, 0.577578067779541], "cos_innovation_negg": [0.5528621673583984, 0.5784894227981567, 0.577578067779541], "cos_apical_negg": [0.034285690635442734, 0.06607280671596527, 0.1244848370552063], "cos_Ac_negg": [0.6092706918716431, 0.6358908414840698, 0.6264286637306213], "r_norm": [0.19721956551074982, 0.19170287251472473, 0.19004429876804352]}, {"step": 3400, "epoch": 7, "train_loss": 0.014400985091924667, "cos_r_negg": [0.5379105806350708, 0.5579407215118408, 0.5657634139060974], "cos_innovation_negg": [0.5379105806350708, 0.5579407215118408, 0.5657634139060974], "cos_apical_negg": [0.036192864179611206, 0.0702303797006607, 0.12546120584011078], "cos_Ac_negg": [0.6015320420265198, 0.6361156702041626, 0.6272715330123901], "r_norm": [0.20590651035308838, 0.20184734463691711, 0.19857719540596008]}, {"step": 3500, "epoch": 7, "train_loss": 0.004858485423028469, "cos_r_negg": [0.5217562913894653, 0.5577101707458496, 0.5610089898109436], "cos_innovation_negg": [0.5217562913894653, 0.5577101707458496, 0.5610089898109436], "cos_apical_negg": [0.02481369487941265, 0.06895075738430023, 0.12296687066555023], "cos_Ac_negg": [0.602568507194519, 0.6409847736358643, 0.6339326500892639], "r_norm": [0.16456836462020874, 0.16079244017601013, 0.16020901501178741], "ldr": {"dL_sdil": -0.005762666463851929, "dL_bp": -0.008025065064430237, "ratio": 0.7180834569670952}}, {"step": 3600, "epoch": 7, "train_loss": 0.002303030574694276, "cos_r_negg": [0.5303589105606079, 0.5598253607749939, 0.5574865937232971], "cos_innovation_negg": [0.5303589105606079, 0.5598253607749939, 0.5574865937232971], "cos_apical_negg": [0.028562750667333603, 0.07188724726438522, 0.11762115359306335], "cos_Ac_negg": [0.6107902526855469, 0.646130383014679, 0.6324824094772339], "r_norm": [0.1928023099899292, 0.18809273838996887, 0.18613044917583466]}, {"step": 3700, "epoch": 7, "train_loss": 0.012182761915028095, "cos_r_negg": [0.5196666121482849, 0.5439647436141968, 0.5465419292449951], "cos_innovation_negg": [0.5196666121482849, 0.5439647436141968, 0.5465419292449951], "cos_apical_negg": [0.02880164422094822, 0.061837583780288696, 0.12163949757814407], "cos_Ac_negg": [0.6145820617675781, 0.6401582360267639, 0.629977822303772], "r_norm": [0.18259994685649872, 0.17761272192001343, 0.17525437474250793]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9723, "test_loss": 0.12099323158264161}, {"step": 3800, "epoch": 8, "train_loss": 0.006128543522208929, "cos_r_negg": [0.5074456930160522, 0.541096568107605, 0.5479250550270081], "cos_innovation_negg": [0.5074456930160522, 0.541096568107605, 0.5479250550270081], "cos_apical_negg": [0.03880160301923752, 0.06963154673576355, 0.12138818204402924], "cos_Ac_negg": [0.6084407567977905, 0.6394813060760498, 0.6389426589012146], "r_norm": [0.18895649909973145, 0.18485723435878754, 0.18205872178077698]}, {"step": 3900, "epoch": 8, "train_loss": 0.046840276569128036, "cos_r_negg": [0.4817640781402588, 0.512580394744873, 0.5277144312858582], "cos_innovation_negg": [0.4817640781402588, 0.512580394744873, 0.5277144312858582], "cos_apical_negg": [0.03252935782074928, 0.06808394938707352, 0.11480497568845749], "cos_Ac_negg": [0.5976804494857788, 0.6336661577224731, 0.6421864628791809], "r_norm": [0.1611471027135849, 0.15656572580337524, 0.1543990969657898]}, {"step": 4000, "epoch": 8, "train_loss": 0.010303596965968609, "cos_r_negg": [0.46032238006591797, 0.490315705537796, 0.5095285773277283], "cos_innovation_negg": [0.46032238006591797, 0.490315705537796, 0.5095285773277283], "cos_apical_negg": [0.02937757596373558, 0.06782662123441696, 0.1145288273692131], "cos_Ac_negg": [0.5953971147537231, 0.6307305693626404, 0.6428290605545044], "r_norm": [0.16530068218708038, 0.16076159477233887, 0.15909411013126373], "ldr": {"dL_sdil": -0.006056338548660278, "dL_bp": -0.009109683334827423, "ratio": 0.6648242673272914}}, {"step": 4100, "epoch": 8, "train_loss": 0.006154875736683607, "cos_r_negg": [0.4616425633430481, 0.4977800250053406, 0.517359733581543], "cos_innovation_negg": [0.4616425633430481, 0.4977800250053406, 0.517359733581543], "cos_apical_negg": [0.033913906663656235, 0.07474394142627716, 0.11754339933395386], "cos_Ac_negg": [0.605496883392334, 0.6388245820999146, 0.6508820056915283], "r_norm": [0.20106157660484314, 0.19499371945858002, 0.19201216101646423]}, {"step": 4200, "epoch": 8, "train_loss": 0.027423150837421417, "cos_r_negg": [0.4457210600376129, 0.475233256816864, 0.4958668053150177], "cos_innovation_negg": [0.4457210600376129, 0.475233256816864, 0.4958668053150177], "cos_apical_negg": [0.02756674960255623, 0.06067146733403206, 0.11335664987564087], "cos_Ac_negg": [0.6091917157173157, 0.6434727907180786, 0.6467041969299316], "r_norm": [0.17253616452217102, 0.16870473325252533, 0.16614750027656555]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.972, "test_loss": 0.13293666820526123}, {"step": 4300, "epoch": 9, "train_loss": 0.008598307147622108, "cos_r_negg": [0.43383729457855225, 0.4656917452812195, 0.48779743909835815], "cos_innovation_negg": [0.43383729457855225, 0.4656917452812195, 0.48779743909835815], "cos_apical_negg": [0.025223735719919205, 0.060846418142318726, 0.10914748162031174], "cos_Ac_negg": [0.6169029474258423, 0.646248459815979, 0.6523618102073669], "r_norm": [0.18382036685943604, 0.18095865845680237, 0.17769470810890198]}, {"step": 4400, "epoch": 9, "train_loss": 0.00196667923592031, "cos_r_negg": [0.41872847080230713, 0.452251136302948, 0.47875547409057617], "cos_innovation_negg": [0.41872847080230713, 0.452251136302948, 0.47875547409057617], "cos_apical_negg": [0.02673802338540554, 0.059494443237781525, 0.105801060795784], "cos_Ac_negg": [0.605011522769928, 0.6423466205596924, 0.6571727395057678], "r_norm": [0.17376244068145752, 0.17010605335235596, 0.16702912747859955]}, {"step": 4500, "epoch": 9, "train_loss": 0.025235939770936966, "cos_r_negg": [0.402266263961792, 0.44017136096954346, 0.4623959958553314], "cos_innovation_negg": [0.402266263961792, 0.44017136096954346, 0.4623959958553314], "cos_apical_negg": [0.024557732045650482, 0.05936041101813316, 0.10745266079902649], "cos_Ac_negg": [0.6089265942573547, 0.641944944858551, 0.6555761098861694], "r_norm": [0.16229242086410522, 0.1578385978937149, 0.15560832619667053], "ldr": {"dL_sdil": -0.008048959076404572, "dL_bp": -0.013988755643367767, "ratio": 0.5753877815587319}}, {"step": 4600, "epoch": 9, "train_loss": 0.0004674484662245959, "cos_r_negg": [0.41375449299812317, 0.44948911666870117, 0.46787434816360474], "cos_innovation_negg": [0.41375449299812317, 0.44948911666870117, 0.46787434816360474], "cos_apical_negg": [0.0270821712911129, 0.06542938202619553, 0.11439449340105057], "cos_Ac_negg": [0.6017415523529053, 0.6417720317840576, 0.6538240909576416], "r_norm": [0.18058077991008759, 0.17527592182159424, 0.17288991808891296]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.972, "test_loss": 0.1355734146118164}, {"step": 4700, "epoch": 10, "train_loss": 0.0031531297136098146, "cos_r_negg": [0.38859570026397705, 0.4213126301765442, 0.4452584981918335], "cos_innovation_negg": [0.38859570026397705, 0.4213126301765442, 0.4452584981918335], "cos_apical_negg": [0.022557012736797333, 0.06057509407401085, 0.10062366724014282], "cos_Ac_negg": [0.6080507040023804, 0.6411598324775696, 0.6576842069625854], "r_norm": [0.18241015076637268, 0.1783653050661087, 0.17502665519714355]}, {"step": 4800, "epoch": 10, "train_loss": 0.001424507237970829, "cos_r_negg": [0.37501198053359985, 0.409299373626709, 0.4334845542907715], "cos_innovation_negg": [0.37501198053359985, 0.409299373626709, 0.4334845542907715], "cos_apical_negg": [0.025177378207445145, 0.061661191284656525, 0.10165520012378693], "cos_Ac_negg": [0.5981765985488892, 0.6386995911598206, 0.6585326790809631], "r_norm": [0.18406125903129578, 0.17913511395454407, 0.17667512595653534]}, {"step": 4900, "epoch": 10, "train_loss": 0.002842707559466362, "cos_r_negg": [0.38084954023361206, 0.4200208783149719, 0.44367271661758423], "cos_innovation_negg": [0.38084954023361206, 0.4200208783149719, 0.44367271661758423], "cos_apical_negg": [0.025119103491306305, 0.061888497322797775, 0.10602675378322601], "cos_Ac_negg": [0.5990905165672302, 0.6428531408309937, 0.6620258688926697], "r_norm": [0.176680326461792, 0.17173907160758972, 0.16918231546878815]}, {"step": 5000, "epoch": 10, "train_loss": 0.0005123436567373574, "cos_r_negg": [0.3533342778682709, 0.3915327787399292, 0.41553783416748047], "cos_innovation_negg": [0.3533342778682709, 0.3915327787399292, 0.41553783416748047], "cos_apical_negg": [0.01720595546066761, 0.05739036202430725, 0.09578986465930939], "cos_Ac_negg": [0.5992639064788818, 0.6391037702560425, 0.6600940227508545], "r_norm": [0.19456928968429565, 0.19128933548927307, 0.1872979998588562], "ldr": {"dL_sdil": -0.010446697473526001, "dL_bp": -0.018925122916698456, "ratio": 0.5520015653007161}}, {"step": 5100, "epoch": 10, "train_loss": 0.0036637960001826286, "cos_r_negg": [0.3606067895889282, 0.39345288276672363, 0.4226773977279663], "cos_innovation_negg": [0.3606067895889282, 0.39345288276672363, 0.4226773977279663], "cos_apical_negg": [0.022676775231957436, 0.05807805061340332, 0.10057053714990616], "cos_Ac_negg": [0.6026651859283447, 0.6419999003410339, 0.6661624908447266], "r_norm": [0.17673218250274658, 0.17317934334278107, 0.16993945837020874]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9734, "test_loss": 0.1404675461769104}, {"step": 5200, "epoch": 11, "train_loss": 0.0005994847742840648, "cos_r_negg": [0.35273417830467224, 0.3868192434310913, 0.41649049520492554], "cos_innovation_negg": [0.35273417830467224, 0.3868192434310913, 0.41649049520492554], "cos_apical_negg": [0.01789838634431362, 0.05813433229923248, 0.0956256240606308], "cos_Ac_negg": [0.6013449430465698, 0.6412538290023804, 0.6646535992622375], "r_norm": [0.17028164863586426, 0.16714194416999817, 0.1640455722808838]}, {"step": 5300, "epoch": 11, "train_loss": 0.006956327706575394, "cos_r_negg": [0.3471260368824005, 0.38593626022338867, 0.41008174419403076], "cos_innovation_negg": [0.3471260368824005, 0.38593626022338867, 0.41008174419403076], "cos_apical_negg": [0.022661101073026657, 0.06154236942529678, 0.09650105237960815], "cos_Ac_negg": [0.5949249863624573, 0.6413847208023071, 0.6650874614715576], "r_norm": [0.18341776728630066, 0.17986181378364563, 0.17660996317863464]}, {"step": 5400, "epoch": 11, "train_loss": 0.001547288615256548, "cos_r_negg": [0.33000773191452026, 0.3681381344795227, 0.39361703395843506], "cos_innovation_negg": [0.33000773191452026, 0.3681381344795227, 0.39361703395843506], "cos_apical_negg": [0.0172126442193985, 0.05573862046003342, 0.09154460579156876], "cos_Ac_negg": [0.5914196968078613, 0.6341928839683533, 0.6608392000198364], "r_norm": [0.1795005202293396, 0.17573903501033783, 0.17266663908958435]}, {"step": 5500, "epoch": 11, "train_loss": 0.0010484072845429182, "cos_r_negg": [0.32509326934814453, 0.3683776259422302, 0.39212900400161743], "cos_innovation_negg": [0.32509326934814453, 0.3683776259422302, 0.39212900400161743], "cos_apical_negg": [0.0187190193682909, 0.06007537618279457, 0.09168079495429993], "cos_Ac_negg": [0.5898714661598206, 0.6387882232666016, 0.665573000907898], "r_norm": [0.18720915913581848, 0.18333867192268372, 0.18037936091423035], "ldr": {"dL_sdil": -0.010868288576602936, "dL_bp": -0.020501919090747833, "ratio": 0.530110792482232}}, {"step": 5600, "epoch": 11, "train_loss": 0.003246604697778821, "cos_r_negg": [0.3071094751358032, 0.3475777208805084, 0.3725828230381012], "cos_innovation_negg": [0.3071094751358032, 0.3475777208805084, 0.3725828230381012], "cos_apical_negg": [0.015700895339250565, 0.0529693067073822, 0.0874062031507492], "cos_Ac_negg": [0.5879708528518677, 0.6322216987609863, 0.6637140512466431], "r_norm": [0.17360228300094604, 0.1698247492313385, 0.16715270280838013]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9733, "test_loss": 0.14312982892990112}, {"step": 5700, "epoch": 12, "train_loss": 0.0006515533896163106, "cos_r_negg": [0.3087655305862427, 0.34727153182029724, 0.37167757749557495], "cos_innovation_negg": [0.3087655305862427, 0.34727153182029724, 0.37167757749557495], "cos_apical_negg": [0.0184798464179039, 0.05521545186638832, 0.08523423224687576], "cos_Ac_negg": [0.5891385078430176, 0.6350855827331543, 0.6646750569343567], "r_norm": [0.1833624541759491, 0.1796121597290039, 0.1768713742494583]}, {"step": 5800, "epoch": 12, "train_loss": 0.0003087636723648757, "cos_r_negg": [0.30396589636802673, 0.34521055221557617, 0.37020623683929443], "cos_innovation_negg": [0.30396589636802673, 0.34521055221557617, 0.37020623683929443], "cos_apical_negg": [0.016882814466953278, 0.05496159568428993, 0.08495524525642395], "cos_Ac_negg": [0.5892714858055115, 0.6374879479408264, 0.6660394072532654], "r_norm": [0.17786654829978943, 0.17419278621673584, 0.17145559191703796]}, {"step": 5900, "epoch": 12, "train_loss": 0.0011283843778073788, "cos_r_negg": [0.306083083152771, 0.3477693796157837, 0.3714534342288971], "cos_innovation_negg": [0.306083083152771, 0.3477693796157837, 0.3714534342288971], "cos_apical_negg": [0.018021147698163986, 0.05589523911476135, 0.08503801375627518], "cos_Ac_negg": [0.5883532166481018, 0.6385139226913452, 0.6678342819213867], "r_norm": [0.17690302431583405, 0.17339177429676056, 0.17059600353240967]}, {"step": 6000, "epoch": 12, "train_loss": 0.00048075226368382573, "cos_r_negg": [0.29833588004112244, 0.33743441104888916, 0.3617570400238037], "cos_innovation_negg": [0.29833588004112244, 0.33743441104888916, 0.3617570400238037], "cos_apical_negg": [0.019266171380877495, 0.05505426600575447, 0.08429230749607086], "cos_Ac_negg": [0.5897235870361328, 0.6371286511421204, 0.6667823791503906], "r_norm": [0.17418819665908813, 0.17051303386688232, 0.1679476797580719], "ldr": {"dL_sdil": -0.010284975171089172, "dL_bp": -0.02005515992641449, "ratio": 0.5128343632674259}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9726, "test_loss": 0.14664524269104004}, {"step": 6100, "epoch": 13, "train_loss": 0.0006032318924553692, "cos_r_negg": [0.2905581295490265, 0.3290700316429138, 0.35489749908447266], "cos_innovation_negg": [0.2905581295490265, 0.3290700316429138, 0.35489749908447266], "cos_apical_negg": [0.01673714816570282, 0.053014423698186874, 0.0824136883020401], "cos_Ac_negg": [0.5891910195350647, 0.6343379020690918, 0.6674453616142273], "r_norm": [0.17484882473945618, 0.17138946056365967, 0.16872015595436096]}, {"step": 6200, "epoch": 13, "train_loss": 0.0008048104937188327, "cos_r_negg": [0.28783735632896423, 0.3279610276222229, 0.35211002826690674], "cos_innovation_negg": [0.28783735632896423, 0.3279610276222229, 0.35211002826690674], "cos_apical_negg": [0.017266422510147095, 0.05395973473787308, 0.0814470648765564], "cos_Ac_negg": [0.5860611200332642, 0.6336036920547485, 0.6678301692008972], "r_norm": [0.17862695455551147, 0.17502078413963318, 0.17230205237865448]}, {"step": 6300, "epoch": 13, "train_loss": 0.001458723098039627, "cos_r_negg": [0.28095054626464844, 0.319520503282547, 0.3462580740451813], "cos_innovation_negg": [0.28095054626464844, 0.319520503282547, 0.3462580740451813], "cos_apical_negg": [0.015333285555243492, 0.05295852944254875, 0.08145970851182938], "cos_Ac_negg": [0.5885621905326843, 0.6347628235816956, 0.667636513710022], "r_norm": [0.17873218655586243, 0.17528963088989258, 0.17260552942752838]}, {"step": 6400, "epoch": 13, "train_loss": 0.0009553181007504463, "cos_r_negg": [0.28301292657852173, 0.3196304440498352, 0.3458683490753174], "cos_innovation_negg": [0.28301292657852173, 0.3196304440498352, 0.3458683490753174], "cos_apical_negg": [0.016829241067171097, 0.05420325696468353, 0.08162521570920944], "cos_Ac_negg": [0.5902488231658936, 0.6372597217559814, 0.66744065284729], "r_norm": [0.18012583255767822, 0.17678941786289215, 0.17403095960617065]}, {"step": 6500, "epoch": 13, "train_loss": 0.0001677384425420314, "cos_r_negg": [0.2745632529258728, 0.3147183656692505, 0.33949393033981323], "cos_innovation_negg": [0.2745632529258728, 0.3147183656692505, 0.33949393033981323], "cos_apical_negg": [0.016230184584856033, 0.053433649241924286, 0.08032813668251038], "cos_Ac_negg": [0.5861450433731079, 0.6342117786407471, 0.6667425632476807], "r_norm": [0.17428138852119446, 0.17074427008628845, 0.1682063192129135], "ldr": {"dL_sdil": -0.010724306106567383, "dL_bp": -0.02085144817829132, "ratio": 0.5143194858634605}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9725, "test_loss": 0.15084042530059816}, {"step": 6600, "epoch": 14, "train_loss": 0.0004299550782889128, "cos_r_negg": [0.273845911026001, 0.31272608041763306, 0.3383336663246155], "cos_innovation_negg": [0.273845911026001, 0.31272608041763306, 0.3383336663246155], "cos_apical_negg": [0.01675397902727127, 0.05444930121302605, 0.07982146739959717], "cos_Ac_negg": [0.5896521806716919, 0.6363834142684937, 0.667066216468811], "r_norm": [0.17700453102588654, 0.1736709177494049, 0.17096003890037537]}, {"step": 6700, "epoch": 14, "train_loss": 0.0005057700327597558, "cos_r_negg": [0.2730553150177002, 0.31345319747924805, 0.3370913565158844], "cos_innovation_negg": [0.2730553150177002, 0.31345319747924805, 0.3370913565158844], "cos_apical_negg": [0.016011910513043404, 0.055612288415431976, 0.08066101372241974], "cos_Ac_negg": [0.5874826312065125, 0.6363601684570312, 0.6681541800498962], "r_norm": [0.1798991858959198, 0.17647922039031982, 0.17376676201820374]}, {"step": 6800, "epoch": 14, "train_loss": 0.0006526593351736665, "cos_r_negg": [0.2665591239929199, 0.30510541796684265, 0.3292950987815857], "cos_innovation_negg": [0.2665591239929199, 0.30510541796684265, 0.3292950987815857], "cos_apical_negg": [0.01495925523340702, 0.05403262749314308, 0.07963677495718002], "cos_Ac_negg": [0.5866935849189758, 0.6354987621307373, 0.6675518751144409], "r_norm": [0.18175745010375977, 0.1781677007675171, 0.17535631358623505]}, {"step": 6900, "epoch": 14, "train_loss": 0.0005965807940810919, "cos_r_negg": [0.26459217071533203, 0.30316829681396484, 0.32873469591140747], "cos_innovation_negg": [0.26459217071533203, 0.30316829681396484, 0.32873469591140747], "cos_apical_negg": [0.01662355661392212, 0.054152961820364, 0.07870881259441376], "cos_Ac_negg": [0.5859885215759277, 0.6350167393684387, 0.6674662828445435], "r_norm": [0.17693497240543365, 0.17355331778526306, 0.17087683081626892]}, {"step": 7000, "epoch": 14, "train_loss": 0.0012752392794936895, "cos_r_negg": [0.26172104477882385, 0.3006412088871002, 0.3232826590538025], "cos_innovation_negg": [0.26172104477882385, 0.3006412088871002, 0.3232826590538025], "cos_apical_negg": [0.017376698553562164, 0.054388709366321564, 0.07631789892911911], "cos_Ac_negg": [0.5867894887924194, 0.6361160278320312, 0.6681196689605713], "r_norm": [0.17798875272274017, 0.17451174557209015, 0.17195779085159302], "ldr": {"dL_sdil": -0.011315062642097473, "dL_bp": -0.02217201143503189, "ratio": 0.5103309041334706}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9727, "test_loss": 0.1537931764602661}], "final": {"test_acc": 0.9727, "test_loss": 0.1537931764602661, "wall_s": 23.313454627990723, "cos_r_negg": [0.26067402958869934, 0.2989043891429901, 0.3237406611442566], "cos_innovation_negg": [0.26067402958869934, 0.2989043891429901, 0.3237406611442566], "cos_apical_negg": [0.01768970862030983, 0.053484197705984116, 0.07788942754268646], "cos_Ac_negg": [0.5858397483825684, 0.6353812217712402, 0.6680448055267334]}}