1
|
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.05, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p05_matched_s1", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.327335834503174, "cos_r_negg": [0.014000799506902695, -0.013633744791150093, -0.0066742198541760445], "cos_innovation_negg": [0.014459304511547089, -0.013818614184856415, -0.0063325027003884315], "cos_apical_negg": [0.01400079857558012, -0.013633744791150093, -0.006674220319837332], "cos_Ac_negg": [0.014459235593676567, -0.013818634673953056, -0.0063325269147753716], "r_norm": [4.738112449645996, 4.797549247741699, 4.787727355957031], "ldr": {"dL_sdil": -0.042402029037475586, "dL_bp": -0.21435141563415527, "ratio": 0.1978154840360156}}, {"step": 100, "epoch": 0, "train_loss": 0.2781984210014343, "cos_r_negg": [0.26685237884521484, 0.21982751786708832, 0.22992366552352905], "cos_innovation_negg": [0.5157362222671509, 0.41939112544059753, 0.3292425274848938], "cos_apical_negg": [0.26685237884521484, 0.21982751786708832, 0.22992366552352905], "cos_Ac_negg": [0.5156465172767639, 0.4194236695766449, 0.3292240500450134], "r_norm": [1.0385547876358032, 1.0394731760025024, 1.056185007095337]}, {"step": 200, "epoch": 0, "train_loss": 0.3677998185157776, "cos_r_negg": [0.263663113117218, 0.2083938717842102, 0.22713875770568848], "cos_innovation_negg": [0.5560674071311951, 0.46555498242378235, 0.4336373507976532], "cos_apical_negg": [0.263663113117218, 0.2083938866853714, 0.22713875770568848], "cos_Ac_negg": [0.555426299571991, 0.46511727571487427, 0.4329182207584381], "r_norm": [0.8922678232192993, 0.8953602313995361, 0.9070045948028564]}, {"step": 300, "epoch": 0, "train_loss": 0.23251374065876007, "cos_r_negg": [0.23529763519763947, 0.21924659609794617, 0.2554989457130432], "cos_innovation_negg": [0.5555249452590942, 0.5305030941963196, 0.5329081416130066], "cos_apical_negg": [0.23529765009880066, 0.21924659609794617, 0.2554989457130432], "cos_Ac_negg": [0.5553916692733765, 0.5303660035133362, 0.5327898263931274], "r_norm": [0.6125134229660034, 0.6164205074310303, 0.6273356676101685]}, {"step": 400, "epoch": 0, "train_loss": 0.15909647941589355, "cos_r_negg": [0.22536510229110718, 0.22064456343650818, 0.24368901550769806], "cos_innovation_negg": [0.5665348768234253, 0.5570060610771179, 0.5425494909286499], "cos_apical_negg": [0.22536510229110718, 0.22064456343650818, 0.24368901550769806], "cos_Ac_negg": [0.5663563013076782, 0.5569068789482117, 0.542511522769928], "r_norm": [0.6293011903762817, 0.632056474685669, 0.6434175968170166]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9403, "test_loss": 0.20267537536621094}, {"step": 500, "epoch": 1, "train_loss": 0.18902839720249176, "cos_r_negg": [0.1790933609008789, 0.20257514715194702, 0.23208177089691162], "cos_innovation_negg": [0.5719155073165894, 0.5642278790473938, 0.5872927904129028], "cos_apical_negg": [0.1790933609008789, 0.20257514715194702, 0.23208177089691162], "cos_Ac_negg": [0.5718954801559448, 0.5642244815826416, 0.5871847867965698], "r_norm": [0.5600966811180115, 0.5626634359359741, 0.5669237971305847], "ldr": {"dL_sdil": -0.011338025331497192, "dL_bp": -0.01136857271194458, "ratio": 0.9973129977507825}}, {"step": 600, "epoch": 1, "train_loss": 0.11155551671981812, "cos_r_negg": [0.20599210262298584, 0.194766104221344, 0.20247206091880798], "cos_innovation_negg": [0.5785467624664307, 0.5494545698165894, 0.6112366914749146], "cos_apical_negg": [0.20599213242530823, 0.194766104221344, 0.20247206091880798], "cos_Ac_negg": [0.5783642530441284, 0.5492405891418457, 0.6110966205596924], "r_norm": [0.5204837322235107, 0.5207526683807373, 0.5278429985046387]}, {"step": 700, "epoch": 1, "train_loss": 0.19533555209636688, "cos_r_negg": [0.17774969339370728, 0.19593371450901031, 0.21875163912773132], "cos_innovation_negg": [0.6004726886749268, 0.5840119123458862, 0.6007018089294434], "cos_apical_negg": [0.17774969339370728, 0.1959337294101715, 0.21875163912773132], "cos_Ac_negg": [0.6002299189567566, 0.5839093923568726, 0.6005539894104004], "r_norm": [0.4677206873893738, 0.47016650438308716, 0.4738680422306061]}, {"step": 800, "epoch": 1, "train_loss": 0.15825039148330688, "cos_r_negg": [0.16074857115745544, 0.1626344472169876, 0.1874714493751526], "cos_innovation_negg": [0.5607525110244751, 0.5336446166038513, 0.5835872888565063], "cos_apical_negg": [0.16074855625629425, 0.1626344621181488, 0.1874714493751526], "cos_Ac_negg": [0.5605154037475586, 0.5334181785583496, 0.5833826065063477], "r_norm": [0.4187534749507904, 0.42067137360572815, 0.42354151606559753]}, {"step": 900, "epoch": 1, "train_loss": 0.18467900156974792, "cos_r_negg": [0.20914487540721893, 0.22298023104667664, 0.2523660659790039], "cos_innovation_negg": [0.5801753997802734, 0.562985897064209, 0.606001615524292], "cos_apical_negg": [0.20914486050605774, 0.22298023104667664, 0.2523660957813263], "cos_Ac_negg": [0.5800694227218628, 0.5628949403762817, 0.6058626174926758], "r_norm": [0.5190070867538452, 0.5219466686248779, 0.5291485786437988]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9506, "test_loss": 0.17010464248657226}, {"step": 1000, "epoch": 2, "train_loss": 0.04992591589689255, "cos_r_negg": [0.19572226703166962, 0.17794883251190186, 0.22146213054656982], "cos_innovation_negg": [0.5998021960258484, 0.5672405958175659, 0.6154801845550537], "cos_apical_negg": [0.19572226703166962, 0.17794883251190186, 0.22146213054656982], "cos_Ac_negg": [0.5996326804161072, 0.5670913457870483, 0.6152791976928711], "r_norm": [0.4730270802974701, 0.470811665058136, 0.4747304916381836], "ldr": {"dL_sdil": -0.015341967344284058, "dL_bp": -0.0163649320602417, "ratio": 0.9374904391786071}}, {"step": 1100, "epoch": 2, "train_loss": 0.20945867896080017, "cos_r_negg": [0.1694614589214325, 0.18095383048057556, 0.2116498351097107], "cos_innovation_negg": [0.5619527697563171, 0.5650385618209839, 0.6543017625808716], "cos_apical_negg": [0.1694614589214325, 0.18095383048057556, 0.2116498053073883], "cos_Ac_negg": [0.5618287324905396, 0.56494140625, 0.654178261756897], "r_norm": [0.4545907974243164, 0.4580495357513428, 0.4582388401031494]}, {"step": 1200, "epoch": 2, "train_loss": 0.1839890480041504, "cos_r_negg": [0.1600516140460968, 0.17260783910751343, 0.1995244026184082], "cos_innovation_negg": [0.5563066005706787, 0.5602741241455078, 0.6617388129234314], "cos_apical_negg": [0.1600516140460968, 0.17260785400867462, 0.1995244026184082], "cos_Ac_negg": [0.5561938881874084, 0.5601537227630615, 0.661475658416748], "r_norm": [0.40716421604156494, 0.4083688259124756, 0.4146285057067871]}, {"step": 1300, "epoch": 2, "train_loss": 0.15830054879188538, "cos_r_negg": [0.15883848071098328, 0.1917852759361267, 0.20369289815425873], "cos_innovation_negg": [0.555382490158081, 0.5663414597511292, 0.6880208849906921], "cos_apical_negg": [0.15883848071098328, 0.1917852759361267, 0.20369289815425873], "cos_Ac_negg": [0.5552250742912292, 0.5662208795547485, 0.6878046989440918], "r_norm": [0.43299487233161926, 0.4335848093032837, 0.4377847909927368]}, {"step": 1400, "epoch": 2, "train_loss": 0.19248934090137482, "cos_r_negg": [0.14802604913711548, 0.17009006440639496, 0.23110854625701904], "cos_innovation_negg": [0.559179425239563, 0.5863549113273621, 0.6957074403762817], "cos_apical_negg": [0.14802604913711548, 0.17009006440639496, 0.23110854625701904], "cos_Ac_negg": [0.5590461492538452, 0.586161196231842, 0.6953648924827576], "r_norm": [0.37540918588638306, 0.37148696184158325, 0.37575411796569824]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9596, "test_loss": 0.1394033809661865}, {"step": 1500, "epoch": 3, "train_loss": 0.08944995701313019, "cos_r_negg": [0.14514648914337158, 0.16608357429504395, 0.19010353088378906], "cos_innovation_negg": [0.5726072788238525, 0.5724003911018372, 0.7225419282913208], "cos_apical_negg": [0.14514648914337158, 0.16608357429504395, 0.19010353088378906], "cos_Ac_negg": [0.5723007321357727, 0.5721114873886108, 0.7221858501434326], "r_norm": [0.3024035096168518, 0.30209675431251526, 0.3078489899635315], "ldr": {"dL_sdil": -0.003830850124359131, "dL_bp": -0.004313953220844269, "ratio": 0.8880138305276775}}, {"step": 1600, "epoch": 3, "train_loss": 0.15920960903167725, "cos_r_negg": [0.1307712346315384, 0.1694299578666687, 0.19286535680294037], "cos_innovation_negg": [0.5470030307769775, 0.5656390190124512, 0.741963267326355], "cos_apical_negg": [0.1307712197303772, 0.1694299727678299, 0.19286534190177917], "cos_Ac_negg": [0.546863853931427, 0.565504789352417, 0.741804838180542], "r_norm": [0.37343645095825195, 0.3703274130821228, 0.3722206950187683]}, {"step": 1700, "epoch": 3, "train_loss": 0.10022221505641937, "cos_r_negg": [0.160159632563591, 0.15611574053764343, 0.1771431416273117], "cos_innovation_negg": [0.529837429523468, 0.5634852051734924, 0.6926331520080566], "cos_apical_negg": [0.160159632563591, 0.15611574053764343, 0.1771431565284729], "cos_Ac_negg": [0.5297232866287231, 0.5634281635284424, 0.6925235986709595], "r_norm": [0.37219926714897156, 0.36870720982551575, 0.36905521154403687]}, {"step": 1800, "epoch": 3, "train_loss": 0.0368051752448082, "cos_r_negg": [0.13545368611812592, 0.11821255087852478, 0.1634964942932129], "cos_innovation_negg": [0.5214605331420898, 0.5495294332504272, 0.6641377210617065], "cos_apical_negg": [0.13545368611812592, 0.11821255087852478, 0.1634964942932129], "cos_Ac_negg": [0.5212026238441467, 0.5492298603057861, 0.6633946895599365], "r_norm": [0.39454638957977295, 0.39656558632850647, 0.40040087699890137]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9582, "test_loss": 0.15445995407104493}, {"step": 1900, "epoch": 4, "train_loss": 0.06744775176048279, "cos_r_negg": [0.11355800181627274, 0.15126921236515045, 0.15317592024803162], "cos_innovation_negg": [0.5673056244850159, 0.5809839367866516, 0.7085917592048645], "cos_apical_negg": [0.11355799436569214, 0.15126919746398926, 0.15317592024803162], "cos_Ac_negg": [0.567139744758606, 0.5805246233940125, 0.7078638672828674], "r_norm": [0.3425997793674469, 0.3435506820678711, 0.3472229242324829]}, {"step": 2000, "epoch": 4, "train_loss": 0.133309468626976, "cos_r_negg": [0.0911845788359642, 0.11686769872903824, 0.12235550582408905], "cos_innovation_negg": [0.5420997142791748, 0.5717331767082214, 0.7403692007064819], "cos_apical_negg": [0.0911845788359642, 0.11686769872903824, 0.12235550582408905], "cos_Ac_negg": [0.5418422222137451, 0.5715316534042358, 0.740033745765686], "r_norm": [0.2965013384819031, 0.29893380403518677, 0.3011382818222046], "ldr": {"dL_sdil": -0.008002370595932007, "dL_bp": -0.009968265891075134, "ratio": 0.8027846250667081}}, {"step": 2100, "epoch": 4, "train_loss": 0.03637475147843361, "cos_r_negg": [0.11173398792743683, 0.13662609457969666, 0.1658332794904709], "cos_innovation_negg": [0.5483249425888062, 0.5740309357643127, 0.7501567006111145], "cos_apical_negg": [0.11173398792743683, 0.13662609457969666, 0.1658332794904709], "cos_Ac_negg": [0.5482739806175232, 0.5739814043045044, 0.7498853206634521], "r_norm": [0.2998882234096527, 0.2994261384010315, 0.3042312264442444]}, {"step": 2200, "epoch": 4, "train_loss": 0.022333752363920212, "cos_r_negg": [0.10447806119918823, 0.1357472538948059, 0.17529860138893127], "cos_innovation_negg": [0.5510979890823364, 0.5893969535827637, 0.7025282382965088], "cos_apical_negg": [0.10447806119918823, 0.1357472538948059, 0.17529860138893127], "cos_Ac_negg": [0.55082768201828, 0.5890421867370605, 0.7021288871765137], "r_norm": [0.33024102449417114, 0.3275192081928253, 0.3374108374118805]}, {"step": 2300, "epoch": 4, "train_loss": 0.18817636370658875, "cos_r_negg": [0.10440659523010254, 0.139730766415596, 0.15772543847560883], "cos_innovation_negg": [0.5324651002883911, 0.5545135736465454, 0.7314645648002625], "cos_apical_negg": [0.10440659523010254, 0.139730766415596, 0.15772542357444763], "cos_Ac_negg": [0.5320799350738525, 0.5541362762451172, 0.7308579683303833], "r_norm": [0.3156967759132385, 0.31805694103240967, 0.3205592930316925]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9652, "test_loss": 0.1349000961303711}, {"step": 2400, "epoch": 5, "train_loss": 0.13673873245716095, "cos_r_negg": [0.06086712330579758, 0.08893437683582306, 0.138218492269516], "cos_innovation_negg": [0.5715898275375366, 0.5935612320899963, 0.7480977773666382], "cos_apical_negg": [0.060867127031087875, 0.08893437683582306, 0.13821850717067719], "cos_Ac_negg": [0.5715224742889404, 0.5934250354766846, 0.74757319688797], "r_norm": [0.2423560917377472, 0.23958319425582886, 0.2430085837841034]}, {"step": 2500, "epoch": 5, "train_loss": 0.08846108615398407, "cos_r_negg": [0.06706833094358444, 0.10987864434719086, 0.15216797590255737], "cos_innovation_negg": [0.5385262966156006, 0.58486407995224, 0.7296252846717834], "cos_apical_negg": [0.06706832349300385, 0.10987864434719086, 0.15216797590255737], "cos_Ac_negg": [0.5377368330955505, 0.5837572813034058, 0.7292695045471191], "r_norm": [0.27576175332069397, 0.27765315771102905, 0.28112727403640747], "ldr": {"dL_sdil": -0.00995662808418274, "dL_bp": -0.013622581958770752, "ratio": 0.7308914062192353}}, {"step": 2600, "epoch": 5, "train_loss": 0.10714676976203918, "cos_r_negg": [0.11678977310657501, 0.12979930639266968, 0.18900683522224426], "cos_innovation_negg": [0.5647395849227905, 0.5982375144958496, 0.7303290367126465], "cos_apical_negg": [0.11678977310657501, 0.12979930639266968, 0.18900682032108307], "cos_Ac_negg": [0.564498782157898, 0.5980241298675537, 0.7301104068756104], "r_norm": [0.3022003769874573, 0.29887285828590393, 0.3037441074848175]}, {"step": 2700, "epoch": 5, "train_loss": 0.12422903627157211, "cos_r_negg": [0.0964074656367302, 0.1067756861448288, 0.15795418620109558], "cos_innovation_negg": [0.5652424097061157, 0.5902745723724365, 0.7420969605445862], "cos_apical_negg": [0.0964074656367302, 0.1067756861448288, 0.15795418620109558], "cos_Ac_negg": [0.5650827884674072, 0.5900545120239258, 0.7416421175003052], "r_norm": [0.28233474493026733, 0.2806481719017029, 0.2845459580421448]}, {"step": 2800, "epoch": 5, "train_loss": 0.19038572907447815, "cos_r_negg": [0.09882406890392303, 0.11376520991325378, 0.16026218235492706], "cos_innovation_negg": [0.5563746690750122, 0.568358302116394, 0.7109635472297668], "cos_apical_negg": [0.09882406890392303, 0.11376521736383438, 0.16026219725608826], "cos_Ac_negg": [0.5561993718147278, 0.5675990581512451, 0.7105340957641602], "r_norm": [0.30161112546920776, 0.29919740557670593, 0.30328208208084106]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.964, "test_loss": 0.12621053848266603}, {"step": 2900, "epoch": 6, "train_loss": 0.02114187926054001, "cos_r_negg": [0.07249077409505844, 0.09168776869773865, 0.15117180347442627], "cos_innovation_negg": [0.5390970706939697, 0.5703646540641785, 0.7006702423095703], "cos_apical_negg": [0.07249076664447784, 0.09168777614831924, 0.15117178857326508], "cos_Ac_negg": [0.5382399559020996, 0.5694425702095032, 0.6996619701385498], "r_norm": [0.2647293508052826, 0.2658502459526062, 0.2682751715183258]}, {"step": 3000, "epoch": 6, "train_loss": 0.20470091700553894, "cos_r_negg": [0.07848583906888962, 0.08388024568557739, 0.15430918335914612], "cos_innovation_negg": [0.5579773783683777, 0.5941951274871826, 0.7172616720199585], "cos_apical_negg": [0.07848583906888962, 0.08388024568557739, 0.15430918335914612], "cos_Ac_negg": [0.5577477216720581, 0.5935335755348206, 0.7163886427879333], "r_norm": [0.2652837336063385, 0.26477673649787903, 0.2703627049922943], "ldr": {"dL_sdil": -0.00884576141834259, "dL_bp": -0.011587589979171753, "ratio": 0.7633823283566735}}, {"step": 3100, "epoch": 6, "train_loss": 0.09332571923732758, "cos_r_negg": [0.08043253421783447, 0.10789922624826431, 0.15035396814346313], "cos_innovation_negg": [0.5523258447647095, 0.589907705783844, 0.7223434448242188], "cos_apical_negg": [0.08043253421783447, 0.10789922624826431, 0.15035396814346313], "cos_Ac_negg": [0.5521047711372375, 0.5893389582633972, 0.7221006155014038], "r_norm": [0.26571565866470337, 0.264682412147522, 0.26524415612220764]}, {"step": 3200, "epoch": 6, "train_loss": 0.08728523552417755, "cos_r_negg": [0.0591692179441452, 0.07672376185655594, 0.10397675633430481], "cos_innovation_negg": [0.5728427767753601, 0.5943782329559326, 0.7174142599105835], "cos_apical_negg": [0.059169214218854904, 0.07672376930713654, 0.10397675633430481], "cos_Ac_negg": [0.5727130770683289, 0.5937618613243103, 0.7167554497718811], "r_norm": [0.2354109287261963, 0.23463121056556702, 0.23798078298568726]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9652, "test_loss": 0.1265283432006836}, {"step": 3300, "epoch": 7, "train_loss": 0.049240585416555405, "cos_r_negg": [0.07296295464038849, 0.08288723230361938, 0.11177732050418854], "cos_innovation_negg": [0.552432656288147, 0.5862433314323425, 0.7291265726089478], "cos_apical_negg": [0.07296295464038849, 0.08288723975419998, 0.11177733540534973], "cos_Ac_negg": [0.5523402690887451, 0.5862059593200684, 0.7294195294380188], "r_norm": [0.24863889813423157, 0.2463444322347641, 0.25142961740493774]}, {"step": 3400, "epoch": 7, "train_loss": 0.036376114934682846, "cos_r_negg": [0.06327491998672485, 0.08526453375816345, 0.13728328049182892], "cos_innovation_negg": [0.5689361095428467, 0.5867960453033447, 0.7374023199081421], "cos_apical_negg": [0.06327491998672485, 0.08526453375816345, 0.13728328049182892], "cos_Ac_negg": [0.5688714981079102, 0.5868169069290161, 0.7381430864334106], "r_norm": [0.2511857748031616, 0.25147727131843567, 0.25480884313583374]}, {"step": 3500, "epoch": 7, "train_loss": 0.04425770416855812, "cos_r_negg": [0.07048217952251434, 0.09557612240314484, 0.11822102963924408], "cos_innovation_negg": [0.5470721125602722, 0.5822858810424805, 0.7139791250228882], "cos_apical_negg": [0.07048217952251434, 0.09557612240314484, 0.11822102963924408], "cos_Ac_negg": [0.5478157997131348, 0.5823948383331299, 0.7132679224014282], "r_norm": [0.21853691339492798, 0.21888524293899536, 0.2214718759059906], "ldr": {"dL_sdil": -0.00374777615070343, "dL_bp": -0.0060371458530426025, "ratio": 0.6207860869908626}}, {"step": 3600, "epoch": 7, "train_loss": 0.016083816066384315, "cos_r_negg": [0.07472144067287445, 0.0981871634721756, 0.14003993570804596], "cos_innovation_negg": [0.5394066572189331, 0.5919173359870911, 0.7581023573875427], "cos_apical_negg": [0.07472144067287445, 0.0981871634721756, 0.14003993570804596], "cos_Ac_negg": [0.5396754145622253, 0.5916072726249695, 0.7576492428779602], "r_norm": [0.27651557326316833, 0.27698010206222534, 0.2793763279914856]}, {"step": 3700, "epoch": 7, "train_loss": 0.02318321354687214, "cos_r_negg": [0.0632040798664093, 0.10064822435379028, 0.10517960786819458], "cos_innovation_negg": [0.5654623508453369, 0.5972422361373901, 0.7645934820175171], "cos_apical_negg": [0.0632040798664093, 0.10064822435379028, 0.10517960786819458], "cos_Ac_negg": [0.5651528835296631, 0.5971337556838989, 0.7645400762557983], "r_norm": [0.26289740204811096, 0.2633611559867859, 0.26702022552490234]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.966, "test_loss": 0.13126993370056153}, {"step": 3800, "epoch": 8, "train_loss": 0.033086568117141724, "cos_r_negg": [0.06775660812854767, 0.09610474109649658, 0.13352809846401215], "cos_innovation_negg": [0.5622141361236572, 0.5975745916366577, 0.7608423829078674], "cos_apical_negg": [0.06775660812854767, 0.09610474854707718, 0.13352811336517334], "cos_Ac_negg": [0.5616971254348755, 0.5979364514350891, 0.7606920003890991], "r_norm": [0.2271111011505127, 0.22552378475666046, 0.22792571783065796]}, {"step": 3900, "epoch": 8, "train_loss": 0.21299625933170319, "cos_r_negg": [0.070110023021698, 0.10455170273780823, 0.16838666796684265], "cos_innovation_negg": [0.549162745475769, 0.5947030186653137, 0.7463819980621338], "cos_apical_negg": [0.070110023021698, 0.10455170273780823, 0.16838666796684265], "cos_Ac_negg": [0.5506429076194763, 0.5961032509803772, 0.7443358898162842], "r_norm": [0.26119065284729004, 0.2603268623352051, 0.26267847418785095]}, {"step": 4000, "epoch": 8, "train_loss": 0.045660194009542465, "cos_r_negg": [0.0635378286242485, 0.061132095754146576, 0.09236867725849152], "cos_innovation_negg": [0.5339014530181885, 0.5708186030387878, 0.7236140966415405], "cos_apical_negg": [0.0635378286242485, 0.061132095754146576, 0.09236867725849152], "cos_Ac_negg": [0.5340219736099243, 0.5706844925880432, 0.7242632508277893], "r_norm": [0.24014504253864288, 0.2385360300540924, 0.24150380492210388], "ldr": {"dL_sdil": -0.0070352703332901, "dL_bp": -0.0109567791223526, "ratio": 0.6420929229957418}}, {"step": 4100, "epoch": 8, "train_loss": 0.02911919169127941, "cos_r_negg": [0.05804622545838356, 0.06820593774318695, 0.11597670614719391], "cos_innovation_negg": [0.5450683236122131, 0.5880045890808105, 0.7330162525177002], "cos_apical_negg": [0.05804622173309326, 0.06820593774318695, 0.1159767210483551], "cos_Ac_negg": [0.5449085831642151, 0.5885275602340698, 0.733569860458374], "r_norm": [0.2169014811515808, 0.21712006628513336, 0.22053925693035126]}, {"step": 4200, "epoch": 8, "train_loss": 0.0916910469532013, "cos_r_negg": [0.07937869429588318, 0.07797777652740479, 0.09943917393684387], "cos_innovation_negg": [0.5633472204208374, 0.6011552810668945, 0.7497129440307617], "cos_apical_negg": [0.07937869429588318, 0.07797776907682419, 0.09943918138742447], "cos_Ac_negg": [0.56303471326828, 0.6025720834732056, 0.7514515519142151], "r_norm": [0.2261510044336319, 0.2252989262342453, 0.22839811444282532]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9685, "test_loss": 0.12365431823730469}, {"step": 4300, "epoch": 9, "train_loss": 0.014758888632059097, "cos_r_negg": [0.0714099109172821, 0.06545557081699371, 0.1478579044342041], "cos_innovation_negg": [0.5366417169570923, 0.5749788284301758, 0.7624439001083374], "cos_apical_negg": [0.0714099109172821, 0.06545556336641312, 0.1478579193353653], "cos_Ac_negg": [0.5395289659500122, 0.5765002369880676, 0.761928141117096], "r_norm": [0.23815125226974487, 0.23755693435668945, 0.24205870926380157]}, {"step": 4400, "epoch": 9, "train_loss": 0.022416403517127037, "cos_r_negg": [0.06786151975393295, 0.08166593313217163, 0.16421091556549072], "cos_innovation_negg": [0.5520721673965454, 0.5914250016212463, 0.7781805992126465], "cos_apical_negg": [0.06786152720451355, 0.08166594058275223, 0.16421091556549072], "cos_Ac_negg": [0.5520327687263489, 0.5919884443283081, 0.778745174407959], "r_norm": [0.22056658565998077, 0.21906621754169464, 0.2203739881515503]}, {"step": 4500, "epoch": 9, "train_loss": 0.1037139743566513, "cos_r_negg": [0.0865529328584671, 0.09103463590145111, 0.14527535438537598], "cos_innovation_negg": [0.5638523101806641, 0.592156171798706, 0.7649362087249756], "cos_apical_negg": [0.0865529328584671, 0.09103463590145111, 0.14527535438537598], "cos_Ac_negg": [0.564079761505127, 0.5937734842300415, 0.7660852670669556], "r_norm": [0.22324925661087036, 0.22315490245819092, 0.22645336389541626], "ldr": {"dL_sdil": -0.008066743612289429, "dL_bp": -0.014248646795749664, "ratio": 0.5661410327537713}}, {"step": 4600, "epoch": 9, "train_loss": 0.06626339256763458, "cos_r_negg": [0.065360888838768, 0.060470327734947205, 0.12234711647033691], "cos_innovation_negg": [0.5448588132858276, 0.6093933582305908, 0.7503721714019775], "cos_apical_negg": [0.06536088138818741, 0.060470327734947205, 0.12234711647033691], "cos_Ac_negg": [0.5461468696594238, 0.610420823097229, 0.749968409538269], "r_norm": [0.22936847805976868, 0.22757463157176971, 0.22855454683303833]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9678, "test_loss": 0.12383741111755371}, {"step": 4700, "epoch": 10, "train_loss": 0.00938369520008564, "cos_r_negg": [0.037425555288791656, 0.06938180327415466, 0.11219051480293274], "cos_innovation_negg": [0.550470232963562, 0.5968233346939087, 0.7346566915512085], "cos_apical_negg": [0.03742556273937225, 0.06938180327415466, 0.11219051480293274], "cos_Ac_negg": [0.5507229566574097, 0.5979015827178955, 0.7344845533370972], "r_norm": [0.18589651584625244, 0.18497535586357117, 0.18600165843963623]}, {"step": 4800, "epoch": 10, "train_loss": 0.02552179805934429, "cos_r_negg": [0.06233280152082443, 0.07233235239982605, 0.1072329580783844], "cos_innovation_negg": [0.5499267578125, 0.5912210941314697, 0.7322781085968018], "cos_apical_negg": [0.06233280524611473, 0.07233234494924545, 0.1072329506278038], "cos_Ac_negg": [0.5506216287612915, 0.5924664735794067, 0.7314146757125854], "r_norm": [0.1988486349582672, 0.19887100160121918, 0.20252931118011475]}, {"step": 4900, "epoch": 10, "train_loss": 0.01581461913883686, "cos_r_negg": [0.05807516723871231, 0.06710566580295563, 0.12905558943748474], "cos_innovation_negg": [0.5676783323287964, 0.5948033332824707, 0.7567842602729797], "cos_apical_negg": [0.05807516723871231, 0.06710566580295563, 0.12905558943748474], "cos_Ac_negg": [0.5687493085861206, 0.5973098278045654, 0.7574092149734497], "r_norm": [0.2369706928730011, 0.23495718836784363, 0.23768897354602814]}, {"step": 5000, "epoch": 10, "train_loss": 0.014232782647013664, "cos_r_negg": [0.04947555810213089, 0.06722775101661682, 0.11993487179279327], "cos_innovation_negg": [0.5552632212638855, 0.6064639091491699, 0.7579385042190552], "cos_apical_negg": [0.04947555810213089, 0.06722775101661682, 0.11993487179279327], "cos_Ac_negg": [0.5565897226333618, 0.6067001819610596, 0.7580022811889648], "r_norm": [0.19451904296875, 0.19268527626991272, 0.1956268548965454], "ldr": {"dL_sdil": -0.007124781608581543, "dL_bp": -0.012223467230796814, "ratio": 0.5828773026552384}}, {"step": 5100, "epoch": 10, "train_loss": 0.03126565366983414, "cos_r_negg": [0.054448649287223816, 0.06965840607881546, 0.09593477845191956], "cos_innovation_negg": [0.5371192097663879, 0.5995650887489319, 0.7563377022743225], "cos_apical_negg": [0.054448649287223816, 0.06965841352939606, 0.09593477845191956], "cos_Ac_negg": [0.5426687002182007, 0.6089742183685303, 0.7673718929290771], "r_norm": [0.2243049293756485, 0.22551247477531433, 0.2285807728767395]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9682, "test_loss": 0.12618726348876952}, {"step": 5200, "epoch": 11, "train_loss": 0.013689623214304447, "cos_r_negg": [0.054590046405792236, 0.08569154888391495, 0.12170731276273727], "cos_innovation_negg": [0.5655243992805481, 0.6024428606033325, 0.7436208128929138], "cos_apical_negg": [0.054590046405792236, 0.08569155633449554, 0.12170730531215668], "cos_Ac_negg": [0.5670942068099976, 0.6042472124099731, 0.7451425790786743], "r_norm": [0.25497347116470337, 0.2538006901741028, 0.2551809847354889]}, {"step": 5300, "epoch": 11, "train_loss": 0.05269131064414978, "cos_r_negg": [0.05302569270133972, 0.06381505727767944, 0.13336099684238434], "cos_innovation_negg": [0.5120725631713867, 0.5659040808677673, 0.6957269906997681], "cos_apical_negg": [0.05302569270133972, 0.06381505727767944, 0.13336101174354553], "cos_Ac_negg": [0.51710045337677, 0.5723103880882263, 0.6966844797134399], "r_norm": [0.2222837507724762, 0.22246211767196655, 0.22576166689395905]}, {"step": 5400, "epoch": 11, "train_loss": 0.026194309815764427, "cos_r_negg": [0.052439652383327484, 0.06063498929142952, 0.10191938281059265], "cos_innovation_negg": [0.5543305277824402, 0.5927353501319885, 0.7473986148834229], "cos_apical_negg": [0.052439648658037186, 0.06063498929142952, 0.10191937536001205], "cos_Ac_negg": [0.5560851693153381, 0.5938558578491211, 0.7422469258308411], "r_norm": [0.19103989005088806, 0.19208963215351105, 0.19506944715976715]}, {"step": 5500, "epoch": 11, "train_loss": 0.03610093146562576, "cos_r_negg": [0.04882842302322388, 0.04626253619790077, 0.13743747770786285], "cos_innovation_negg": [0.525267481803894, 0.5838602781295776, 0.7465553283691406], "cos_apical_negg": [0.048828426748514175, 0.046262532472610474, 0.13743747770786285], "cos_Ac_negg": [0.5281760692596436, 0.5898793935775757, 0.7464447021484375], "r_norm": [0.19913125038146973, 0.19981563091278076, 0.2017751932144165], "ldr": {"dL_sdil": -0.0044263750314712524, "dL_bp": -0.00766851007938385, "ratio": 0.5772144765606024}}, {"step": 5600, "epoch": 11, "train_loss": 0.011690723709762096, "cos_r_negg": [0.04178202152252197, 0.04755251854658127, 0.13127601146697998], "cos_innovation_negg": [0.5479075312614441, 0.6049648523330688, 0.757026195526123], "cos_apical_negg": [0.041782017797231674, 0.04755251854658127, 0.13127601146697998], "cos_Ac_negg": [0.5566954016685486, 0.6144217848777771, 0.7610421180725098], "r_norm": [0.19590972363948822, 0.195388525724411, 0.19799646735191345]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9691, "test_loss": 0.13189056396484375}, {"step": 5700, "epoch": 12, "train_loss": 0.011434990912675858, "cos_r_negg": [0.04492928460240364, 0.04870538040995598, 0.11411378532648087], "cos_innovation_negg": [0.559892475605011, 0.614902138710022, 0.758762001991272], "cos_apical_negg": [0.04492928832769394, 0.04870537668466568, 0.11411378532648087], "cos_Ac_negg": [0.5645284652709961, 0.6230587959289551, 0.762065052986145], "r_norm": [0.19092899560928345, 0.18994152545928955, 0.19238710403442383]}, {"step": 5800, "epoch": 12, "train_loss": 0.01472583506256342, "cos_r_negg": [0.04657397419214249, 0.05546797066926956, 0.13012515008449554], "cos_innovation_negg": [0.5549585819244385, 0.6114252209663391, 0.7747702598571777], "cos_apical_negg": [0.04657397419214249, 0.05546797066926956, 0.13012516498565674], "cos_Ac_negg": [0.5605918765068054, 0.619053065776825, 0.7806684374809265], "r_norm": [0.21261551976203918, 0.21132779121398926, 0.2129901945590973]}, {"step": 5900, "epoch": 12, "train_loss": 0.027686260640621185, "cos_r_negg": [0.05407793074846268, 0.04876308888196945, 0.11269255727529526], "cos_innovation_negg": [0.5343650579452515, 0.5867140293121338, 0.7400655150413513], "cos_apical_negg": [0.05407793074846268, 0.04876308515667915, 0.11269255727529526], "cos_Ac_negg": [0.5467405319213867, 0.6034579277038574, 0.7492297887802124], "r_norm": [0.228031188249588, 0.22552496194839478, 0.22781668603420258]}, {"step": 6000, "epoch": 12, "train_loss": 0.006304215639829636, "cos_r_negg": [0.041400037705898285, 0.05253276973962784, 0.08973284810781479], "cos_innovation_negg": [0.5323338508605957, 0.5852477550506592, 0.7412900328636169], "cos_apical_negg": [0.04140004515647888, 0.05253276228904724, 0.08973284810781479], "cos_Ac_negg": [0.5374307632446289, 0.5913330316543579, 0.7445220351219177], "r_norm": [0.211392343044281, 0.20852413773536682, 0.20963749289512634], "ldr": {"dL_sdil": -0.008514508605003357, "dL_bp": -0.017385005950927734, "ratio": 0.489761615787597}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9696, "test_loss": 0.13102652072906495}, {"step": 6100, "epoch": 13, "train_loss": 0.007647733669728041, "cos_r_negg": [0.058543361723423004, 0.06763773411512375, 0.13031405210494995], "cos_innovation_negg": [0.5372574329376221, 0.5944938659667969, 0.7435393929481506], "cos_apical_negg": [0.058543361723423004, 0.06763774156570435, 0.13031405210494995], "cos_Ac_negg": [0.5454597473144531, 0.6029956340789795, 0.7495102882385254], "r_norm": [0.19791069626808167, 0.19725343585014343, 0.19918422400951385]}, {"step": 6200, "epoch": 13, "train_loss": 0.005339222028851509, "cos_r_negg": [0.03993219882249832, 0.0439935065805912, 0.12259697914123535], "cos_innovation_negg": [0.5299742817878723, 0.6072005033493042, 0.7573761940002441], "cos_apical_negg": [0.03993219509720802, 0.0439935028553009, 0.12259697169065475], "cos_Ac_negg": [0.5384821891784668, 0.6164162158966064, 0.7610653042793274], "r_norm": [0.2201082855463028, 0.21853888034820557, 0.22131897509098053]}, {"step": 6300, "epoch": 13, "train_loss": 0.020007651299238205, "cos_r_negg": [0.05164296180009842, 0.0499151386320591, 0.11171048134565353], "cos_innovation_negg": [0.5220210552215576, 0.578366219997406, 0.7537604570388794], "cos_apical_negg": [0.05164295434951782, 0.049915142357349396, 0.11171048879623413], "cos_Ac_negg": [0.5394889116287231, 0.5976953506469727, 0.764815092086792], "r_norm": [0.21189093589782715, 0.20991289615631104, 0.2125605046749115]}, {"step": 6400, "epoch": 13, "train_loss": 0.002438321942463517, "cos_r_negg": [0.05224587395787239, 0.05782228708267212, 0.13297517597675323], "cos_innovation_negg": [0.5443210601806641, 0.6096733212471008, 0.7620235681533813], "cos_apical_negg": [0.05224587023258209, 0.05782228708267212, 0.13297517597675323], "cos_Ac_negg": [0.5515585541725159, 0.6179340481758118, 0.7716501951217651], "r_norm": [0.19945526123046875, 0.19964176416397095, 0.2016235589981079]}, {"step": 6500, "epoch": 13, "train_loss": 0.002159678842872381, "cos_r_negg": [0.05180658400058746, 0.05799303948879242, 0.11578771471977234], "cos_innovation_negg": [0.5411338210105896, 0.5977074503898621, 0.7573133707046509], "cos_apical_negg": [0.05180658772587776, 0.05799303948879242, 0.11578771471977234], "cos_Ac_negg": [0.551083505153656, 0.6087433695793152, 0.765688419342041], "r_norm": [0.20487673580646515, 0.20408472418785095, 0.205448716878891], "ldr": {"dL_sdil": -0.007744446396827698, "dL_bp": -0.014109469950199127, "ratio": 0.5488828725786684}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9709, "test_loss": 0.12851683921813964}, {"step": 6600, "epoch": 14, "train_loss": 0.004894197452813387, "cos_r_negg": [0.047842200845479965, 0.04682867228984833, 0.10232914239168167], "cos_innovation_negg": [0.5222222805023193, 0.5879594087600708, 0.7287417650222778], "cos_apical_negg": [0.04784219712018967, 0.04682867228984833, 0.10232914239168167], "cos_Ac_negg": [0.5316355228424072, 0.599854588508606, 0.7331242561340332], "r_norm": [0.17443296313285828, 0.17286810278892517, 0.17643971741199493]}, {"step": 6700, "epoch": 14, "train_loss": 0.02553962916135788, "cos_r_negg": [0.0393855907022953, 0.05858352780342102, 0.1074909120798111], "cos_innovation_negg": [0.513884425163269, 0.570460319519043, 0.7442063689231873], "cos_apical_negg": [0.0393855907022953, 0.05858352780342102, 0.1074909120798111], "cos_Ac_negg": [0.5333433151245117, 0.5898236036300659, 0.7712637186050415], "r_norm": [0.19830980896949768, 0.19619552791118622, 0.19803929328918457]}, {"step": 6800, "epoch": 14, "train_loss": 0.004792890511453152, "cos_r_negg": [0.0455397292971611, 0.040460534393787384, 0.07618850469589233], "cos_innovation_negg": [0.535183310508728, 0.5900650024414062, 0.7600193619728088], "cos_apical_negg": [0.045539725571870804, 0.040460534393787384, 0.07618850469589233], "cos_Ac_negg": [0.5455622673034668, 0.603481650352478, 0.7685121297836304], "r_norm": [0.20605935156345367, 0.20373085141181946, 0.20592007040977478]}, {"step": 6900, "epoch": 14, "train_loss": 0.01101000141352415, "cos_r_negg": [0.051652900874614716, 0.0503389909863472, 0.08229105919599533], "cos_innovation_negg": [0.529188871383667, 0.5825726985931396, 0.75916987657547], "cos_apical_negg": [0.051652900874614716, 0.0503389909863472, 0.08229105919599533], "cos_Ac_negg": [0.5427979230880737, 0.5965732336044312, 0.7766227126121521], "r_norm": [0.21024341881275177, 0.20914150774478912, 0.21031688153743744]}, {"step": 7000, "epoch": 14, "train_loss": 0.03273762762546539, "cos_r_negg": [0.05334595963358879, 0.05724663287401199, 0.0838860496878624], "cos_innovation_negg": [0.5381057262420654, 0.5916114449501038, 0.7556002140045166], "cos_apical_negg": [0.05334595590829849, 0.057246625423431396, 0.083886057138443], "cos_Ac_negg": [0.5495507717132568, 0.6026450991630554, 0.7660986185073853], "r_norm": [0.21696007251739502, 0.2152629792690277, 0.21664190292358398], "ldr": {"dL_sdil": -0.006766654551029205, "dL_bp": -0.013199172914028168, "ratio": 0.5126574668809407}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.971, "test_loss": 0.13064061431884766}], "final": {"test_acc": 0.971, "test_loss": 0.13064061431884766, "wall_s": 25.36318016052246, "cos_r_negg": [0.059259649366140366, 0.056370340287685394, 0.07438581436872482], "cos_innovation_negg": [0.5242489576339722, 0.5776432752609253, 0.7376971244812012], "cos_apical_negg": [0.05925965681672096, 0.056370340287685394, 0.07438582181930542], "cos_Ac_negg": [0.5392985343933105, 0.5938428044319153, 0.7554187774658203]}}
|