summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p5_residual_s3.json
blob: 99791e8f720c4356c0ca3c378948aca4374bf2ed (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": 0.5, "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_rho0p5_residual_s3", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.331878900527954, "cos_r_negg": [-0.014011913910508156, 0.013712333515286446, 0.00608993461355567], "cos_innovation_negg": [-0.014011913910508156, 0.013712333515286446, 0.00608993461355567], "cos_apical_negg": [-0.018168717622756958, 0.005241477862000465, -0.002439321018755436], "cos_Ac_negg": [-0.014013777486979961, 0.01371200755238533, 0.006088861729949713], "r_norm": [4.854389667510986, 4.749159336090088, 4.848528861999512], "ldr": {"dL_sdil": -0.005772829055786133, "dL_bp": -0.23062801361083984, "ratio": 0.025030910015672102}}, {"step": 100, "epoch": 0, "train_loss": 0.18872550129890442, "cos_r_negg": [0.42080679535865784, 0.3573850691318512, 0.31352412700653076], "cos_innovation_negg": [0.42080679535865784, 0.3573850691318512, 0.31352412700653076], "cos_apical_negg": [0.07740601152181625, 0.08406104147434235, 0.12225973606109619], "cos_Ac_negg": [0.4255574345588684, 0.35864710807800293, 0.312264084815979], "r_norm": [0.8661477565765381, 0.8538812398910522, 0.8630181550979614]}, {"step": 200, "epoch": 0, "train_loss": 0.23271331191062927, "cos_r_negg": [0.47490373253822327, 0.4291550815105438, 0.35083863139152527], "cos_innovation_negg": [0.47490373253822327, 0.4291550815105438, 0.35083863139152527], "cos_apical_negg": [0.06322602927684784, 0.08257284760475159, 0.11042001843452454], "cos_Ac_negg": [0.479240357875824, 0.43211889266967773, 0.3495291471481323], "r_norm": [0.7295346856117249, 0.7159911394119263, 0.7286012172698975]}, {"step": 300, "epoch": 0, "train_loss": 0.15973104536533356, "cos_r_negg": [0.4891892969608307, 0.4489712119102478, 0.387234091758728], "cos_innovation_negg": [0.4891892969608307, 0.4489712119102478, 0.387234091758728], "cos_apical_negg": [0.05702277645468712, 0.0757027268409729, 0.11901925504207611], "cos_Ac_negg": [0.4903077483177185, 0.45112019777297974, 0.3880354166030884], "r_norm": [0.6274819374084473, 0.6219789981842041, 0.626327395439148]}, {"step": 400, "epoch": 0, "train_loss": 0.15213412046432495, "cos_r_negg": [0.5123450756072998, 0.4869316816329956, 0.4288489818572998], "cos_innovation_negg": [0.5123450756072998, 0.4869316816329956, 0.4288489818572998], "cos_apical_negg": [0.04249223321676254, 0.0700196623802185, 0.09839313477277756], "cos_Ac_negg": [0.5162568092346191, 0.4874834716320038, 0.42815375328063965], "r_norm": [0.5795223712921143, 0.5721928477287292, 0.5810239315032959]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9473, "test_loss": 0.1803105567932129}, {"step": 500, "epoch": 1, "train_loss": 0.2707977890968323, "cos_r_negg": [0.4892926812171936, 0.47885721921920776, 0.4356996417045593], "cos_innovation_negg": [0.4892926812171936, 0.47885721921920776, 0.4356996417045593], "cos_apical_negg": [0.04015825316309929, 0.06950805336236954, 0.1100805476307869], "cos_Ac_negg": [0.49992287158966064, 0.48546868562698364, 0.43970683217048645], "r_norm": [0.5381065607070923, 0.5282456874847412, 0.5357717871665955], "ldr": {"dL_sdil": -0.018530339002609253, "dL_bp": -0.01635301113128662, "ratio": 1.1331453793948054}}, {"step": 600, "epoch": 1, "train_loss": 0.11664538830518723, "cos_r_negg": [0.49489182233810425, 0.5053346157073975, 0.443124383687973], "cos_innovation_negg": [0.49489182233810425, 0.5053346157073975, 0.443124383687973], "cos_apical_negg": [0.022730866447091103, 0.05235520377755165, 0.09304158389568329], "cos_Ac_negg": [0.504249632358551, 0.5136399865150452, 0.44650179147720337], "r_norm": [0.4068072438240051, 0.3990352749824524, 0.40562427043914795]}, {"step": 700, "epoch": 1, "train_loss": 0.17657727003097534, "cos_r_negg": [0.5248271226882935, 0.5245544910430908, 0.45468053221702576], "cos_innovation_negg": [0.5248271226882935, 0.5245544910430908, 0.45468053221702576], "cos_apical_negg": [0.03518783301115036, 0.06547833234071732, 0.10126242786645889], "cos_Ac_negg": [0.5350420475006104, 0.532680869102478, 0.4575701951980591], "r_norm": [0.4497723877429962, 0.44423526525497437, 0.4508831799030304]}, {"step": 800, "epoch": 1, "train_loss": 0.13849526643753052, "cos_r_negg": [0.5045038461685181, 0.5329877734184265, 0.47357553243637085], "cos_innovation_negg": [0.5045038461685181, 0.5329877734184265, 0.47357553243637085], "cos_apical_negg": [0.025899339467287064, 0.04956956207752228, 0.10712993144989014], "cos_Ac_negg": [0.5124516487121582, 0.5372099280357361, 0.4743672013282776], "r_norm": [0.3278277516365051, 0.32419031858444214, 0.327649861574173]}, {"step": 900, "epoch": 1, "train_loss": 0.1713041067123413, "cos_r_negg": [0.5401734709739685, 0.549518883228302, 0.47928205132484436], "cos_innovation_negg": [0.5401734709739685, 0.549518883228302, 0.47928205132484436], "cos_apical_negg": [0.0506848581135273, 0.059160977602005005, 0.09672336280345917], "cos_Ac_negg": [0.541490375995636, 0.5504716634750366, 0.4794657528400421], "r_norm": [0.4373318552970886, 0.430068701505661, 0.43439656496047974]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9561, "test_loss": 0.15255565452575684}, {"step": 1000, "epoch": 2, "train_loss": 0.02094106562435627, "cos_r_negg": [0.5435171127319336, 0.5617809891700745, 0.48627057671546936], "cos_innovation_negg": [0.5435171127319336, 0.5617809891700745, 0.48627057671546936], "cos_apical_negg": [0.03208060935139656, 0.06277690082788467, 0.09445619583129883], "cos_Ac_negg": [0.546578586101532, 0.5651637315750122, 0.4887245297431946], "r_norm": [0.31476879119873047, 0.31028246879577637, 0.3133445382118225], "ldr": {"dL_sdil": -0.006925120949745178, "dL_bp": -0.00616048276424408, "ratio": 1.1241198481942223}}, {"step": 1100, "epoch": 2, "train_loss": 0.09473104029893875, "cos_r_negg": [0.5680637955665588, 0.5742523670196533, 0.5016100406646729], "cos_innovation_negg": [0.5680637955665588, 0.5742523670196533, 0.5016100406646729], "cos_apical_negg": [0.037956446409225464, 0.05853698402643204, 0.09649095684289932], "cos_Ac_negg": [0.5749331712722778, 0.5808276534080505, 0.5028795003890991], "r_norm": [0.31742265820503235, 0.3098963499069214, 0.31269025802612305]}, {"step": 1200, "epoch": 2, "train_loss": 0.1336817443370819, "cos_r_negg": [0.5417118072509766, 0.5653090476989746, 0.5028981566429138], "cos_innovation_negg": [0.5417118072509766, 0.5653090476989746, 0.5028981566429138], "cos_apical_negg": [0.030298395082354546, 0.0590420663356781, 0.09857246279716492], "cos_Ac_negg": [0.5533450841903687, 0.5780097246170044, 0.5068104267120361], "r_norm": [0.3133755326271057, 0.30838990211486816, 0.3106110692024231]}, {"step": 1300, "epoch": 2, "train_loss": 0.115728460252285, "cos_r_negg": [0.5507748126983643, 0.5746300220489502, 0.5097249746322632], "cos_innovation_negg": [0.5507748126983643, 0.5746300220489502, 0.5097249746322632], "cos_apical_negg": [0.03266730159521103, 0.056138359010219574, 0.08230951428413391], "cos_Ac_negg": [0.5631198883056641, 0.5865797996520996, 0.5113161206245422], "r_norm": [0.30199965834617615, 0.29994842410087585, 0.3007659912109375]}, {"step": 1400, "epoch": 2, "train_loss": 0.16870275139808655, "cos_r_negg": [0.5444096326828003, 0.5728497505187988, 0.5018571019172668], "cos_innovation_negg": [0.5444096326828003, 0.5728497505187988, 0.5018571019172668], "cos_apical_negg": [0.025644447654485703, 0.05886274576187134, 0.07878376543521881], "cos_Ac_negg": [0.5628910064697266, 0.588857889175415, 0.5121197700500488], "r_norm": [0.27241218090057373, 0.2691349387168884, 0.27151989936828613]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9657, "test_loss": 0.11589976081848144}, {"step": 1500, "epoch": 3, "train_loss": 0.06594803929328918, "cos_r_negg": [0.5402629375457764, 0.5796506404876709, 0.5273761749267578], "cos_innovation_negg": [0.5402629375457764, 0.5796506404876709, 0.5273761749267578], "cos_apical_negg": [0.039710983633995056, 0.05154367908835411, 0.0969703197479248], "cos_Ac_negg": [0.5638954043388367, 0.601933479309082, 0.535617470741272], "r_norm": [0.27322760224342346, 0.26843664050102234, 0.26908040046691895], "ldr": {"dL_sdil": -0.0067085325717926025, "dL_bp": -0.006532587110996246, "ratio": 1.0269335039559118}}, {"step": 1600, "epoch": 3, "train_loss": 0.0769442692399025, "cos_r_negg": [0.5356593132019043, 0.5775347352027893, 0.5283666253089905], "cos_innovation_negg": [0.5356593132019043, 0.5775347352027893, 0.5283666253089905], "cos_apical_negg": [0.03112872503697872, 0.05700736492872238, 0.09994073212146759], "cos_Ac_negg": [0.5571012496948242, 0.6037145853042603, 0.5409438014030457], "r_norm": [0.26720550656318665, 0.2618980407714844, 0.263634592294693]}, {"step": 1700, "epoch": 3, "train_loss": 0.04415497928857803, "cos_r_negg": [0.5329152345657349, 0.5690578818321228, 0.519217848777771], "cos_innovation_negg": [0.5329152345657349, 0.5690578818321228, 0.519217848777771], "cos_apical_negg": [0.03167639672756195, 0.051312316209077835, 0.09680252522230148], "cos_Ac_negg": [0.570207953453064, 0.6094943284988403, 0.5474331378936768], "r_norm": [0.27063578367233276, 0.2675698697566986, 0.2672215402126312]}, {"step": 1800, "epoch": 3, "train_loss": 0.020682916045188904, "cos_r_negg": [0.543427586555481, 0.5755721926689148, 0.5407392382621765], "cos_innovation_negg": [0.543427586555481, 0.5755721926689148, 0.5407392382621765], "cos_apical_negg": [0.020985178649425507, 0.051615044474601746, 0.092034250497818], "cos_Ac_negg": [0.5796080827713013, 0.6094875335693359, 0.5618948936462402], "r_norm": [0.228158101439476, 0.22647729516029358, 0.22639253735542297]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9647, "test_loss": 0.12295613803863525}, {"step": 1900, "epoch": 4, "train_loss": 0.018825624138116837, "cos_r_negg": [0.528862476348877, 0.5797684192657471, 0.5388694405555725], "cos_innovation_negg": [0.528862476348877, 0.5797684192657471, 0.5388694405555725], "cos_apical_negg": [0.029006794095039368, 0.05604682117700577, 0.10635900497436523], "cos_Ac_negg": [0.5644512176513672, 0.6147205829620361, 0.5558013916015625], "r_norm": [0.28968456387519836, 0.28626900911331177, 0.2880341708660126]}, {"step": 2000, "epoch": 4, "train_loss": 0.12200286239385605, "cos_r_negg": [0.49291181564331055, 0.5381568670272827, 0.5108394026756287], "cos_innovation_negg": [0.49291181564331055, 0.5381568670272827, 0.5108394026756287], "cos_apical_negg": [0.036747612059116364, 0.0550658255815506, 0.10144195705652237], "cos_Ac_negg": [0.5494893789291382, 0.6038867235183716, 0.5505704879760742], "r_norm": [0.24793246388435364, 0.24428686499595642, 0.24390989542007446], "ldr": {"dL_sdil": -0.007415145635604858, "dL_bp": -0.008610181510448456, "ratio": 0.8612066570962039}}, {"step": 2100, "epoch": 4, "train_loss": 0.028551729395985603, "cos_r_negg": [0.5037931799888611, 0.5556021332740784, 0.5132607221603394], "cos_innovation_negg": [0.5037931799888611, 0.5556021332740784, 0.5132607221603394], "cos_apical_negg": [0.013839269056916237, 0.058557067066431046, 0.10582729429006577], "cos_Ac_negg": [0.5713272094726562, 0.6184545755386353, 0.5559008121490479], "r_norm": [0.24037976562976837, 0.23496440052986145, 0.23672522604465485]}, {"step": 2200, "epoch": 4, "train_loss": 0.009481847286224365, "cos_r_negg": [0.5175397992134094, 0.5628323554992676, 0.5348480939865112], "cos_innovation_negg": [0.5175397992134094, 0.5628323554992676, 0.5348480939865112], "cos_apical_negg": [0.0159706249833107, 0.04661000892519951, 0.09774476289749146], "cos_Ac_negg": [0.5720933675765991, 0.6177525520324707, 0.5666742324829102], "r_norm": [0.22913230955600739, 0.22418493032455444, 0.2254614233970642]}, {"step": 2300, "epoch": 4, "train_loss": 0.1396801620721817, "cos_r_negg": [0.4930618405342102, 0.5409141182899475, 0.5277037620544434], "cos_innovation_negg": [0.4930618405342102, 0.5409141182899475, 0.5277037620544434], "cos_apical_negg": [0.008156484924256802, 0.0498521588742733, 0.09864520281553268], "cos_Ac_negg": [0.571165144443512, 0.6128248572349548, 0.5722683668136597], "r_norm": [0.20367828011512756, 0.19950716197490692, 0.2002502977848053]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9707, "test_loss": 0.10883730421066284}, {"step": 2400, "epoch": 5, "train_loss": 0.06521674990653992, "cos_r_negg": [0.4649459421634674, 0.5170556306838989, 0.5078390836715698], "cos_innovation_negg": [0.4649459421634674, 0.5170556306838989, 0.5078390836715698], "cos_apical_negg": [0.0044569773599505424, 0.049284402281045914, 0.10828302800655365], "cos_Ac_negg": [0.579290509223938, 0.6281781792640686, 0.5828486680984497], "r_norm": [0.18944184482097626, 0.18613117933273315, 0.18644651770591736]}, {"step": 2500, "epoch": 5, "train_loss": 0.012442920356988907, "cos_r_negg": [0.44050684571266174, 0.490290105342865, 0.4901142716407776], "cos_innovation_negg": [0.44050684571266174, 0.490290105342865, 0.4901142716407776], "cos_apical_negg": [0.014281416311860085, 0.05564771592617035, 0.10965365171432495], "cos_Ac_negg": [0.5732725858688354, 0.6313501596450806, 0.582327127456665], "r_norm": [0.24482318758964539, 0.2420414686203003, 0.2414052039384842], "ldr": {"dL_sdil": -0.008126087486743927, "dL_bp": -0.008378878235816956, "ratio": 0.9698300008714256}}, {"step": 2600, "epoch": 5, "train_loss": 0.023398997262120247, "cos_r_negg": [0.4568743109703064, 0.5034669637680054, 0.5188430547714233], "cos_innovation_negg": [0.4568743109703064, 0.5034669637680054, 0.5188430547714233], "cos_apical_negg": [0.01007128320634365, 0.046921633183956146, 0.10567986965179443], "cos_Ac_negg": [0.586541473865509, 0.6357093453407288, 0.6017983555793762], "r_norm": [0.21272389590740204, 0.20777344703674316, 0.20944646000862122]}, {"step": 2700, "epoch": 5, "train_loss": 0.04132649302482605, "cos_r_negg": [0.4660347104072571, 0.5183552503585815, 0.5186158418655396], "cos_innovation_negg": [0.4660347104072571, 0.5183552503585815, 0.5186158418655396], "cos_apical_negg": [0.009593063965439796, 0.04411444813013077, 0.10364066809415817], "cos_Ac_negg": [0.5876105427742004, 0.6435339450836182, 0.6019076108932495], "r_norm": [0.17805656790733337, 0.17355357110500336, 0.17494475841522217]}, {"step": 2800, "epoch": 5, "train_loss": 0.05498193949460983, "cos_r_negg": [0.43562740087509155, 0.4989575743675232, 0.5196200609207153], "cos_innovation_negg": [0.43562740087509155, 0.4989575743675232, 0.5196200609207153], "cos_apical_negg": [0.005082856398075819, 0.04511084035038948, 0.1107931137084961], "cos_Ac_negg": [0.5642528533935547, 0.642410159111023, 0.6124656200408936], "r_norm": [0.20564180612564087, 0.20171013474464417, 0.20210757851600647]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9693, "test_loss": 0.11708884296417237}, {"step": 2900, "epoch": 6, "train_loss": 0.005203034728765488, "cos_r_negg": [0.4495949149131775, 0.5044668912887573, 0.5137101411819458], "cos_innovation_negg": [0.4495949149131775, 0.5044668912887573, 0.5137101411819458], "cos_apical_negg": [0.0029116824734956026, 0.04600415378808975, 0.10258268564939499], "cos_Ac_negg": [0.5837121605873108, 0.6465179920196533, 0.6148958206176758], "r_norm": [0.17525489628314972, 0.17214038968086243, 0.173087939620018]}, {"step": 3000, "epoch": 6, "train_loss": 0.08553176373243332, "cos_r_negg": [0.4206857979297638, 0.47899866104125977, 0.4925500452518463], "cos_innovation_negg": [0.4206857979297638, 0.47899866104125977, 0.4925500452518463], "cos_apical_negg": [-0.009011526592075825, 0.03955311328172684, 0.10614622384309769], "cos_Ac_negg": [0.5796300172805786, 0.6412211060523987, 0.6063554883003235], "r_norm": [0.17875617742538452, 0.1751418262720108, 0.17465780675411224], "ldr": {"dL_sdil": -0.004441358149051666, "dL_bp": -0.0056234002113342285, "ratio": 0.7897994064338333}}, {"step": 3100, "epoch": 6, "train_loss": 0.042554669082164764, "cos_r_negg": [0.40734750032424927, 0.4535464346408844, 0.4807383120059967], "cos_innovation_negg": [0.40734750032424927, 0.4535464346408844, 0.4807383120059967], "cos_apical_negg": [-0.0014351322315633297, 0.03926960751414299, 0.10535650700330734], "cos_Ac_negg": [0.589215099811554, 0.6471856832504272, 0.6189682483673096], "r_norm": [0.186112180352211, 0.18383631110191345, 0.18407349288463593]}, {"step": 3200, "epoch": 6, "train_loss": 0.047677651047706604, "cos_r_negg": [0.40638357400894165, 0.4533464312553406, 0.4813411831855774], "cos_innovation_negg": [0.40638357400894165, 0.4533464312553406, 0.4813411831855774], "cos_apical_negg": [0.00492829829454422, 0.0322842076420784, 0.10239633172750473], "cos_Ac_negg": [0.5803347229957581, 0.6529821157455444, 0.6237295866012573], "r_norm": [0.1556628942489624, 0.15382838249206543, 0.15468579530715942]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9713, "test_loss": 0.11475176219940185}, {"step": 3300, "epoch": 7, "train_loss": 0.035091057419776917, "cos_r_negg": [0.388364315032959, 0.44012635946273804, 0.4620133638381958], "cos_innovation_negg": [0.388364315032959, 0.44012635946273804, 0.4620133638381958], "cos_apical_negg": [-0.00514238839969039, 0.03361807018518448, 0.10036838054656982], "cos_Ac_negg": [0.5815119743347168, 0.6519482135772705, 0.6196298599243164], "r_norm": [0.15613043308258057, 0.15476371347904205, 0.155074805021286]}, {"step": 3400, "epoch": 7, "train_loss": 0.018217364326119423, "cos_r_negg": [0.350436806678772, 0.4000297784805298, 0.42972588539123535], "cos_innovation_negg": [0.350436806678772, 0.4000297784805298, 0.42972588539123535], "cos_apical_negg": [1.7241574823856354e-05, 0.03882444649934769, 0.10841445624828339], "cos_Ac_negg": [0.5855284333229065, 0.6549354195594788, 0.6250629425048828], "r_norm": [0.15655122697353363, 0.1532667875289917, 0.15430009365081787]}, {"step": 3500, "epoch": 7, "train_loss": 0.006485956255346537, "cos_r_negg": [0.34565138816833496, 0.39090925455093384, 0.42413222789764404], "cos_innovation_negg": [0.34565138816833496, 0.39090925455093384, 0.42413222789764404], "cos_apical_negg": [0.00023910123854875565, 0.03469627723097801, 0.1001327633857727], "cos_Ac_negg": [0.5923053622245789, 0.6582446098327637, 0.6333015561103821], "r_norm": [0.12530985474586487, 0.12307082861661911, 0.1236409991979599], "ldr": {"dL_sdil": -0.0033531971275806427, "dL_bp": -0.004545159637928009, "ratio": 0.7377512331138399}}, {"step": 3600, "epoch": 7, "train_loss": 0.0025435287971049547, "cos_r_negg": [0.361345112323761, 0.41499537229537964, 0.44833433628082275], "cos_innovation_negg": [0.361345112323761, 0.41499537229537964, 0.44833433628082275], "cos_apical_negg": [0.008182194083929062, 0.04040120542049408, 0.10462453961372375], "cos_Ac_negg": [0.583110511302948, 0.660904586315155, 0.6398817300796509], "r_norm": [0.14871007204055786, 0.14299443364143372, 0.14446698129177094]}, {"step": 3700, "epoch": 7, "train_loss": 0.022246941924095154, "cos_r_negg": [0.36678579449653625, 0.4257878065109253, 0.46085524559020996], "cos_innovation_negg": [0.36678579449653625, 0.4257878065109253, 0.46085524559020996], "cos_apical_negg": [0.010286729782819748, 0.0447201170027256, 0.10164155811071396], "cos_Ac_negg": [0.588095486164093, 0.6620072722434998, 0.6441003084182739], "r_norm": [0.17962360382080078, 0.17733833193778992, 0.17752781510353088]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9726, "test_loss": 0.11701411352157592}, {"step": 3800, "epoch": 8, "train_loss": 0.0020683319307863712, "cos_r_negg": [0.31483468413352966, 0.36443912982940674, 0.41338908672332764], "cos_innovation_negg": [0.31483468413352966, 0.36443912982940674, 0.41338908672332764], "cos_apical_negg": [-0.005256372503936291, 0.03418785333633423, 0.09913021326065063], "cos_Ac_negg": [0.5751336216926575, 0.6563472747802734, 0.6401858329772949], "r_norm": [0.1426611691713333, 0.1388171762228012, 0.13953682780265808]}, {"step": 3900, "epoch": 8, "train_loss": 0.003269444452598691, "cos_r_negg": [0.30942264199256897, 0.3527880609035492, 0.4018915593624115], "cos_innovation_negg": [0.30942264199256897, 0.3527880609035492, 0.4018915593624115], "cos_apical_negg": [0.0003192746080458164, 0.03188473731279373, 0.10008475184440613], "cos_Ac_negg": [0.5807934999465942, 0.656278133392334, 0.6394872665405273], "r_norm": [0.13429084420204163, 0.13071230053901672, 0.1315767467021942]}, {"step": 4000, "epoch": 8, "train_loss": 0.004132644273340702, "cos_r_negg": [0.29055094718933105, 0.33638814091682434, 0.3855329155921936], "cos_innovation_negg": [0.29055094718933105, 0.33638814091682434, 0.3855329155921936], "cos_apical_negg": [-0.00855519250035286, 0.033741869032382965, 0.09871513396501541], "cos_Ac_negg": [0.58148592710495, 0.6534676551818848, 0.644256055355072], "r_norm": [0.1311567723751068, 0.12831135094165802, 0.12942379713058472], "ldr": {"dL_sdil": -0.004478953778743744, "dL_bp": -0.006740003824234009, "ratio": 0.6645328245422428}}, {"step": 4100, "epoch": 8, "train_loss": 0.029480300843715668, "cos_r_negg": [0.32144200801849365, 0.35947489738464355, 0.40849611163139343], "cos_innovation_negg": [0.32144200801849365, 0.35947489738464355, 0.40849611163139343], "cos_apical_negg": [0.003794255666434765, 0.032217852771282196, 0.09655633568763733], "cos_Ac_negg": [0.5934809446334839, 0.6684763431549072, 0.6491529941558838], "r_norm": [0.13236254453659058, 0.12935969233512878, 0.1306191086769104]}, {"step": 4200, "epoch": 8, "train_loss": 0.02441762574017048, "cos_r_negg": [0.31035399436950684, 0.3590101897716522, 0.4074493646621704], "cos_innovation_negg": [0.31035399436950684, 0.3590101897716522, 0.4074493646621704], "cos_apical_negg": [-0.011325972154736519, 0.03502793610095978, 0.10148602724075317], "cos_Ac_negg": [0.5924097299575806, 0.6642954349517822, 0.6504828929901123], "r_norm": [0.13863743841648102, 0.13571245968341827, 0.13672906160354614]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.972, "test_loss": 0.12872028541564942}, {"step": 4300, "epoch": 9, "train_loss": 0.004121402744203806, "cos_r_negg": [0.27059096097946167, 0.31552380323410034, 0.37013643980026245], "cos_innovation_negg": [0.27059096097946167, 0.31552380323410034, 0.37013643980026245], "cos_apical_negg": [-0.008882439695298672, 0.037799298763275146, 0.099274642765522], "cos_Ac_negg": [0.5838530659675598, 0.6564987897872925, 0.6464356184005737], "r_norm": [0.1464133858680725, 0.14511489868164062, 0.14471206068992615]}, {"step": 4400, "epoch": 9, "train_loss": 0.0022535943426191807, "cos_r_negg": [0.2896331548690796, 0.3270096778869629, 0.37870895862579346], "cos_innovation_negg": [0.2896331548690796, 0.3270096778869629, 0.37870895862579346], "cos_apical_negg": [-0.01004791259765625, 0.03820240497589111, 0.09683576226234436], "cos_Ac_negg": [0.5843517780303955, 0.6612729430198669, 0.6517888903617859], "r_norm": [0.14669322967529297, 0.14385291934013367, 0.14458304643630981]}, {"step": 4500, "epoch": 9, "train_loss": 0.03937598317861557, "cos_r_negg": [0.25708484649658203, 0.2989466190338135, 0.3525906801223755], "cos_innovation_negg": [0.25708484649658203, 0.2989466190338135, 0.3525906801223755], "cos_apical_negg": [-0.014484313316643238, 0.03541833162307739, 0.09813150763511658], "cos_Ac_negg": [0.5815752148628235, 0.656388521194458, 0.6512470245361328], "r_norm": [0.14379392564296722, 0.14149987697601318, 0.14191201329231262], "ldr": {"dL_sdil": -0.005707666277885437, "dL_bp": -0.008829593658447266, "ratio": 0.6464245693146838}}, {"step": 4600, "epoch": 9, "train_loss": 0.00029099496896378696, "cos_r_negg": [0.26509222388267517, 0.3011626601219177, 0.35187166929244995], "cos_innovation_negg": [0.26509222388267517, 0.3011626601219177, 0.35187166929244995], "cos_apical_negg": [-0.009409554302692413, 0.02760307490825653, 0.08916454017162323], "cos_Ac_negg": [0.5758092403411865, 0.6568939685821533, 0.6504249572753906], "r_norm": [0.12511903047561646, 0.12319760769605637, 0.12391689419746399]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9741, "test_loss": 0.1214785472869873}, {"step": 4700, "epoch": 10, "train_loss": 0.003846376668661833, "cos_r_negg": [0.24793319404125214, 0.2752615809440613, 0.32855188846588135], "cos_innovation_negg": [0.24793319404125214, 0.2752615809440613, 0.32855188846588135], "cos_apical_negg": [-0.0168413445353508, 0.03185483440756798, 0.08513405919075012], "cos_Ac_negg": [0.5857955813407898, 0.6577849984169006, 0.6549846529960632], "r_norm": [0.1392510086297989, 0.1374475359916687, 0.1374175101518631]}, {"step": 4800, "epoch": 10, "train_loss": 0.003788778791204095, "cos_r_negg": [0.23982921242713928, 0.26741862297058105, 0.32012176513671875], "cos_innovation_negg": [0.23982921242713928, 0.26741862297058105, 0.32012176513671875], "cos_apical_negg": [-0.020968560129404068, 0.028218358755111694, 0.08401869237422943], "cos_Ac_negg": [0.5756700038909912, 0.6568929553031921, 0.653887152671814], "r_norm": [0.1297290325164795, 0.1271601915359497, 0.1279851645231247]}, {"step": 4900, "epoch": 10, "train_loss": 0.0023530491162091494, "cos_r_negg": [0.22722691297531128, 0.26289501786231995, 0.3097301125526428], "cos_innovation_negg": [0.22722691297531128, 0.26289501786231995, 0.3097301125526428], "cos_apical_negg": [-0.018490899354219437, 0.03279675915837288, 0.08060262352228165], "cos_Ac_negg": [0.577909529209137, 0.6588432788848877, 0.6562414169311523], "r_norm": [0.13232503831386566, 0.13036315143108368, 0.13081206381320953]}, {"step": 5000, "epoch": 10, "train_loss": 0.0003272384637966752, "cos_r_negg": [0.23629596829414368, 0.264797568321228, 0.31595954298973083], "cos_innovation_negg": [0.23629596829414368, 0.264797568321228, 0.31595954298973083], "cos_apical_negg": [-0.018870677798986435, 0.028028279542922974, 0.078565314412117], "cos_Ac_negg": [0.5826868414878845, 0.6588236093521118, 0.657131016254425], "r_norm": [0.12348625063896179, 0.12130117416381836, 0.12188145518302917], "ldr": {"dL_sdil": -0.004463203251361847, "dL_bp": -0.007074527442455292, "ratio": 0.6308835873018882}}, {"step": 5100, "epoch": 10, "train_loss": 0.001783785060979426, "cos_r_negg": [0.22899088263511658, 0.2561159133911133, 0.30505847930908203], "cos_innovation_negg": [0.22899088263511658, 0.2561159133911133, 0.30505847930908203], "cos_apical_negg": [-0.014518175274133682, 0.029250551015138626, 0.0742192268371582], "cos_Ac_negg": [0.5863211154937744, 0.6628580689430237, 0.6601808071136475], "r_norm": [0.13324525952339172, 0.1307089626789093, 0.13108068704605103]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9742, "test_loss": 0.12863891973495484}, {"step": 5200, "epoch": 11, "train_loss": 0.0016999029321596026, "cos_r_negg": [0.22935858368873596, 0.25430065393447876, 0.3051876425743103], "cos_innovation_negg": [0.22935858368873596, 0.25430065393447876, 0.3051876425743103], "cos_apical_negg": [-0.01597149670124054, 0.027901211753487587, 0.07903474569320679], "cos_Ac_negg": [0.5843941569328308, 0.6619303226470947, 0.6641751527786255], "r_norm": [0.13848471641540527, 0.13639108836650848, 0.13692373037338257]}, {"step": 5300, "epoch": 11, "train_loss": 0.00149227911606431, "cos_r_negg": [0.21886737644672394, 0.24021688103675842, 0.2908511161804199], "cos_innovation_negg": [0.21886737644672394, 0.24021688103675842, 0.2908511161804199], "cos_apical_negg": [-0.01306607574224472, 0.0243108831346035, 0.07685942947864532], "cos_Ac_negg": [0.5760036110877991, 0.6580331921577454, 0.6615997552871704], "r_norm": [0.12689343094825745, 0.12522368133068085, 0.1255747675895691]}, {"step": 5400, "epoch": 11, "train_loss": 0.003502150531858206, "cos_r_negg": [0.21088576316833496, 0.23419539630413055, 0.2831454277038574], "cos_innovation_negg": [0.21088576316833496, 0.23419539630413055, 0.2831454277038574], "cos_apical_negg": [-0.014293482527136803, 0.02643945813179016, 0.07471708953380585], "cos_Ac_negg": [0.5800578594207764, 0.661342978477478, 0.6615606546401978], "r_norm": [0.12902571260929108, 0.12745392322540283, 0.1277874857187271]}, {"step": 5500, "epoch": 11, "train_loss": 0.0016093183076009154, "cos_r_negg": [0.2041584849357605, 0.22600063681602478, 0.27503910660743713], "cos_innovation_negg": [0.2041584849357605, 0.22600063681602478, 0.27503910660743713], "cos_apical_negg": [-0.013327575288712978, 0.02545972913503647, 0.07442834228277206], "cos_Ac_negg": [0.5758709907531738, 0.6590155959129333, 0.662075400352478], "r_norm": [0.12717200815677643, 0.1250791996717453, 0.12573327124118805], "ldr": {"dL_sdil": -0.005176179111003876, "dL_bp": -0.00866156816482544, "ratio": 0.5976029989608924}}, {"step": 5600, "epoch": 11, "train_loss": 0.001172092743217945, "cos_r_negg": [0.20233036577701569, 0.2277071177959442, 0.27275943756103516], "cos_innovation_negg": [0.20233036577701569, 0.2277071177959442, 0.27275943756103516], "cos_apical_negg": [-0.018805058673024178, 0.026067454367876053, 0.07017955183982849], "cos_Ac_negg": [0.5761364698410034, 0.6579396724700928, 0.6598615646362305], "r_norm": [0.1238504946231842, 0.12152472138404846, 0.1221206784248352]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9745, "test_loss": 0.1301448366165161}, {"step": 5700, "epoch": 12, "train_loss": 0.00037419566069729626, "cos_r_negg": [0.19927281141281128, 0.22440919280052185, 0.27297529578208923], "cos_innovation_negg": [0.19927281141281128, 0.22440919280052185, 0.27297529578208923], "cos_apical_negg": [-0.017042972147464752, 0.027279961854219437, 0.07182110846042633], "cos_Ac_negg": [0.5766111612319946, 0.6589412093162537, 0.6626189947128296], "r_norm": [0.1261036992073059, 0.12395627796649933, 0.12456410378217697]}, {"step": 5800, "epoch": 12, "train_loss": 0.0006990813417360187, "cos_r_negg": [0.19310584664344788, 0.21785509586334229, 0.26709112524986267], "cos_innovation_negg": [0.19310584664344788, 0.21785509586334229, 0.26709112524986267], "cos_apical_negg": [-0.01848934218287468, 0.02607933059334755, 0.07075376808643341], "cos_Ac_negg": [0.5727724432945251, 0.6582148671150208, 0.6595107316970825], "r_norm": [0.12188367545604706, 0.11990620195865631, 0.12063635885715485]}, {"step": 5900, "epoch": 12, "train_loss": 0.002051290823146701, "cos_r_negg": [0.21225331723690033, 0.231919527053833, 0.28019171953201294], "cos_innovation_negg": [0.21225331723690033, 0.231919527053833, 0.28019171953201294], "cos_apical_negg": [-0.015212031081318855, 0.02723746933043003, 0.07102951407432556], "cos_Ac_negg": [0.5792114734649658, 0.6632776260375977, 0.66792231798172], "r_norm": [0.13065604865550995, 0.12856750190258026, 0.12947514653205872]}, {"step": 6000, "epoch": 12, "train_loss": 0.00021082983585074544, "cos_r_negg": [0.19337938725948334, 0.2173120081424713, 0.26181572675704956], "cos_innovation_negg": [0.19337938725948334, 0.2173120081424713, 0.26181572675704956], "cos_apical_negg": [-0.02074435167014599, 0.026510538533329964, 0.06835349649190903], "cos_Ac_negg": [0.5737708806991577, 0.6608791351318359, 0.6627746820449829], "r_norm": [0.12671895325183868, 0.1245783269405365, 0.12536340951919556], "ldr": {"dL_sdil": -0.005825906991958618, "dL_bp": -0.010343685746192932, "ratio": 0.5632331777000171}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9745, "test_loss": 0.13520586853027344}, {"step": 6100, "epoch": 13, "train_loss": 0.00114873715210706, "cos_r_negg": [0.18851301074028015, 0.21388106048107147, 0.2559266984462738], "cos_innovation_negg": [0.18851301074028015, 0.21388106048107147, 0.2559266984462738], "cos_apical_negg": [-0.023839516565203667, 0.025731373578310013, 0.06313656270503998], "cos_Ac_negg": [0.5762029886245728, 0.6611396670341492, 0.6640243530273438], "r_norm": [0.12193048000335693, 0.1200208067893982, 0.12074489146471024]}, {"step": 6200, "epoch": 13, "train_loss": 0.0012708634603768587, "cos_r_negg": [0.1813572645187378, 0.2073110044002533, 0.24960248172283173], "cos_innovation_negg": [0.1813572645187378, 0.2073110044002533, 0.24960248172283173], "cos_apical_negg": [-0.023270772770047188, 0.025720421224832535, 0.06304657459259033], "cos_Ac_negg": [0.5741591453552246, 0.6597561240196228, 0.6618976593017578], "r_norm": [0.12066633254289627, 0.11884524673223495, 0.11952510476112366]}, {"step": 6300, "epoch": 13, "train_loss": 0.0007790725794620812, "cos_r_negg": [0.17786607146263123, 0.20206767320632935, 0.24386122822761536], "cos_innovation_negg": [0.17786607146263123, 0.20206767320632935, 0.24386122822761536], "cos_apical_negg": [-0.021800057962536812, 0.02558799646794796, 0.06129763275384903], "cos_Ac_negg": [0.5735625624656677, 0.6590381860733032, 0.659139096736908], "r_norm": [0.11918114870786667, 0.11713533848524094, 0.11786467581987381]}, {"step": 6400, "epoch": 13, "train_loss": 0.0002151842345483601, "cos_r_negg": [0.1802089810371399, 0.2039659321308136, 0.24680691957473755], "cos_innovation_negg": [0.1802089810371399, 0.2039659321308136, 0.24680691957473755], "cos_apical_negg": [-0.019582778215408325, 0.026868212968111038, 0.06381715834140778], "cos_Ac_negg": [0.5758476853370667, 0.660313069820404, 0.6641278266906738], "r_norm": [0.12172451615333557, 0.12006828933954239, 0.12067775428295135]}, {"step": 6500, "epoch": 13, "train_loss": 0.0001270119973924011, "cos_r_negg": [0.17689760029315948, 0.1998426914215088, 0.24155038595199585], "cos_innovation_negg": [0.17689760029315948, 0.1998426914215088, 0.24155038595199585], "cos_apical_negg": [-0.02078068070113659, 0.0257948599755764, 0.06032883748412132], "cos_Ac_negg": [0.5716795325279236, 0.6583441495895386, 0.6622625589370728], "r_norm": [0.12071338295936584, 0.11919794976711273, 0.1198136955499649], "ldr": {"dL_sdil": -0.00588168203830719, "dL_bp": -0.010370306670665741, "ratio": 0.5671656803500879}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9749, "test_loss": 0.13689604864120483}, {"step": 6600, "epoch": 14, "train_loss": 0.00015675066970288754, "cos_r_negg": [0.1775672882795334, 0.1971694529056549, 0.2404470443725586], "cos_innovation_negg": [0.1775672882795334, 0.1971694529056549, 0.2404470443725586], "cos_apical_negg": [-0.01920282281935215, 0.025180485099554062, 0.062167614698410034], "cos_Ac_negg": [0.5715060234069824, 0.6585206985473633, 0.6648013591766357], "r_norm": [0.11828052252531052, 0.11654089391231537, 0.11733348667621613]}, {"step": 6700, "epoch": 14, "train_loss": 0.0003808314504567534, "cos_r_negg": [0.17531812191009521, 0.19601497054100037, 0.23779110610485077], "cos_innovation_negg": [0.17531812191009521, 0.19601497054100037, 0.23779110610485077], "cos_apical_negg": [-0.0205191969871521, 0.025503646582365036, 0.06019933521747589], "cos_Ac_negg": [0.5700347423553467, 0.6578106880187988, 0.6626570820808411], "r_norm": [0.11736148595809937, 0.11555007100105286, 0.1163320392370224]}, {"step": 6800, "epoch": 14, "train_loss": 0.0005349187995307148, "cos_r_negg": [0.1725764274597168, 0.1920589953660965, 0.23308731615543365], "cos_innovation_negg": [0.1725764274597168, 0.1920589953660965, 0.23308731615543365], "cos_apical_negg": [-0.019598782062530518, 0.025701753795146942, 0.05938730016350746], "cos_Ac_negg": [0.570386528968811, 0.6572756767272949, 0.6646796464920044], "r_norm": [0.11974599957466125, 0.11808669567108154, 0.11874199658632278]}, {"step": 6900, "epoch": 14, "train_loss": 0.00042516947723925114, "cos_r_negg": [0.16775819659233093, 0.18516004085540771, 0.22745344042778015], "cos_innovation_negg": [0.16775819659233093, 0.18516004085540771, 0.22745344042778015], "cos_apical_negg": [-0.016404155641794205, 0.02493401989340782, 0.05891628563404083], "cos_Ac_negg": [0.569895327091217, 0.6560081243515015, 0.661963701248169], "r_norm": [0.11935018748044968, 0.11756415665149689, 0.11828391253948212]}, {"step": 7000, "epoch": 14, "train_loss": 0.0005088113248348236, "cos_r_negg": [0.1683044284582138, 0.18657205998897552, 0.22744154930114746], "cos_innovation_negg": [0.1683044284582138, 0.18657205998897552, 0.22744154930114746], "cos_apical_negg": [-0.017778633162379265, 0.025177061557769775, 0.05926932394504547], "cos_Ac_negg": [0.5718478560447693, 0.6575223207473755, 0.6637944579124451], "r_norm": [0.11985176056623459, 0.118133045732975, 0.1188424825668335], "ldr": {"dL_sdil": -0.005673840641975403, "dL_bp": -0.010133080184459686, "ratio": 0.5599324725246849}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9749, "test_loss": 0.13971249618530274}], "final": {"test_acc": 0.9749, "test_loss": 0.13971249618530274, "wall_s": 22.210185527801514, "cos_r_negg": [0.16554266214370728, 0.185797318816185, 0.22594398260116577], "cos_innovation_negg": [0.16554266214370728, 0.185797318816185, 0.22594398260116577], "cos_apical_negg": [-0.019418207928538322, 0.025267083197832108, 0.05846957862377167], "cos_Ac_negg": [0.5687137246131897, 0.6561079621315002, 0.6624050736427307]}}