summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p5_raw_s3.json
blob: 86b899e0392379db4d2ccda3b4ddde0b43a27555 (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": 0, "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_raw_s3", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.331878900527954, "cos_r_negg": [-0.026506898924708366, 0.012330188415944576, -0.0019659455865621567], "cos_innovation_negg": [-0.029736431315541267, 0.019409991800785065, 0.008285761810839176], "cos_apical_negg": [-0.026506898924708366, 0.012330188415944576, -0.0019659455865621567], "cos_Ac_negg": [-0.02973800152540207, 0.019408803433179855, 0.008284433744847775], "r_norm": [8.17677116394043, 6.836240768432617, 6.476951599121094], "ldr": {"dL_sdil": 0.04908013343811035, "dL_bp": -0.21805977821350098, "ratio": -0.2250765081034628}}, {"step": 100, "epoch": 0, "train_loss": 12.267818450927734, "cos_r_negg": [-0.023786088451743126, -0.11871932446956635, -0.14084212481975555], "cos_innovation_negg": [-0.04167438670992851, -0.06051561236381531, 0.26192307472229004], "cos_apical_negg": [-0.023786088451743126, -0.11871932446956635, -0.14084212481975555], "cos_Ac_negg": [-0.041508134454488754, -0.06033655256032944, 0.26192396879196167], "r_norm": [10.945213317871094, 10.342334747314453, 9.944719314575195]}, {"step": 200, "epoch": 0, "train_loss": 11.568981170654297, "cos_r_negg": [0.05637676268815994, 0.16992853581905365, -0.49123942852020264], "cos_innovation_negg": [0.10289576649665833, 0.298967570066452, -0.0014939913526177406], "cos_apical_negg": [0.05637676268815994, 0.16992853581905365, -0.49123942852020264], "cos_Ac_negg": [0.10278133302927017, 0.29901963472366333, -0.0011754860170185566], "r_norm": [10.425918579101562, 9.996030807495117, 9.698972702026367]}, {"step": 300, "epoch": 0, "train_loss": 7.396692276000977, "cos_r_negg": [-0.027777038514614105, 0.0007287114858627319, -0.11273282766342163], "cos_innovation_negg": [-0.034390345215797424, 0.19405455887317657, 0.07537731528282166], "cos_apical_negg": [-0.027777038514614105, 0.0007287114858627319, -0.11273282766342163], "cos_Ac_negg": [-0.03439591825008392, 0.1940525770187378, 0.07536330819129944], "r_norm": [9.88321304321289, 9.525853157043457, 9.523195266723633]}, {"step": 400, "epoch": 0, "train_loss": 9.533720016479492, "cos_r_negg": [0.03617967665195465, 0.23808139562606812, -0.6518124938011169], "cos_innovation_negg": [0.03402910754084587, 0.2270115315914154, -0.4757543206214905], "cos_apical_negg": [0.03617967665195465, 0.23808139562606812, -0.6518124938011169], "cos_Ac_negg": [0.03401627391576767, 0.22700953483581543, -0.4757598042488098], "r_norm": [9.756996154785156, 9.257879257202148, 9.494220733642578]}, {"epoch_end": 0, "step": 469, "test_acc": 0.0982, "test_loss": 9.74996513671875}, {"step": 500, "epoch": 1, "train_loss": 16.520923614501953, "cos_r_negg": [-0.03965315967798233, 0.007813815027475357, -0.5233477354049683], "cos_innovation_negg": [-0.0692405104637146, -0.02429916523396969, -0.025975346565246582], "cos_apical_negg": [-0.03965315967798233, 0.007813815027475357, -0.5233477354049683], "cos_Ac_negg": [-0.06923192739486694, -0.02420998178422451, -0.025978829711675644], "r_norm": [9.435873031616211, 9.374765396118164, 9.238445281982422], "ldr": {"dL_sdil": -2.2517032623291016, "dL_bp": -2.2517032623291016, "ratio": 1.0}}, {"step": 600, "epoch": 1, "train_loss": 12.30771255493164, "cos_r_negg": [-0.015822438523173332, -0.1540403962135315, 0.08863641321659088], "cos_innovation_negg": [-0.01979876682162285, 0.10042668133974075, 0.3308795094490051], "cos_apical_negg": [-0.015822438523173332, -0.1540403962135315, 0.08863641321659088], "cos_Ac_negg": [-0.019798077642917633, 0.10043680667877197, 0.3308820128440857], "r_norm": [9.034561157226562, 8.978337287902832, 9.016508102416992]}, {"step": 700, "epoch": 1, "train_loss": 11.132923126220703, "cos_r_negg": [0.05001148581504822, -0.01987697370350361, 0.31429994106292725], "cos_innovation_negg": [0.05103549733757973, 0.1299055814743042, -0.21260008215904236], "cos_apical_negg": [0.05001148581504822, -0.01987697370350361, 0.31429994106292725], "cos_Ac_negg": [0.05102431774139404, 0.129903182387352, -0.21259984374046326], "r_norm": [9.006699562072754, 9.06875991821289, 9.216055870056152]}, {"step": 800, "epoch": 1, "train_loss": 8.47889518737793, "cos_r_negg": [0.005555157549679279, 0.047896090894937515, -0.29187458753585815], "cos_innovation_negg": [0.031460344791412354, 0.23891372978687286, -0.12157592177391052], "cos_apical_negg": [0.005555157549679279, 0.047896090894937515, -0.29187458753585815], "cos_Ac_negg": [0.03145395591855049, 0.23891803622245789, -0.12157764285802841], "r_norm": [8.915679931640625, 8.769262313842773, 9.054006576538086]}, {"step": 900, "epoch": 1, "train_loss": 14.730730056762695, "cos_r_negg": [-0.003456203965470195, -0.026968369260430336, -0.785609245300293], "cos_innovation_negg": [-0.047013331204652786, 0.16049107909202576, -0.6441376209259033], "cos_apical_negg": [-0.003456203965470195, -0.026968369260430336, -0.785609245300293], "cos_Ac_negg": [-0.047016218304634094, 0.16048306226730347, -0.6441439390182495], "r_norm": [9.154525756835938, 9.25345516204834, 9.408580780029297]}, {"epoch_end": 1, "step": 938, "test_acc": 0.1028, "test_loss": 12.37398525390625}, {"step": 1000, "epoch": 2, "train_loss": 14.343939781188965, "cos_r_negg": [0.0010967403650283813, -0.02762531489133835, -0.634750485420227], "cos_innovation_negg": [0.028234466910362244, 0.19616973400115967, 0.14262229204177856], "cos_apical_negg": [0.0010967403650283813, -0.02762531489133835, -0.634750485420227], "cos_Ac_negg": [0.028230516240000725, 0.1961737871170044, 0.14261820912361145], "r_norm": [8.693264961242676, 9.059747695922852, 8.956097602844238], "ldr": {"dL_sdil": -2.113520622253418, "dL_bp": -2.113520622253418, "ratio": 1.0}}, {"step": 1100, "epoch": 2, "train_loss": 18.238525390625, "cos_r_negg": [0.0033974298276007175, 0.09077562391757965, -0.5978332161903381], "cos_innovation_negg": [-0.06636692583560944, 0.3485768735408783, -0.6810895204544067], "cos_apical_negg": [0.0033974298276007175, 0.09077562391757965, -0.5978332161903381], "cos_Ac_negg": [-0.06580479443073273, 0.35291773080825806, -0.6802278757095337], "r_norm": [8.441755294799805, 8.380655288696289, 8.638420104980469]}, {"step": 1200, "epoch": 2, "train_loss": 6.251697540283203, "cos_r_negg": [-0.030002865940332413, 0.09644234925508499, -0.017023786902427673], "cos_innovation_negg": [-0.00677952915430069, -0.27618831396102905, -0.1790367215871811], "cos_apical_negg": [-0.030002865940332413, 0.09644234925508499, -0.017023786902427673], "cos_Ac_negg": [-0.006759162992238998, -0.2757236361503601, -0.17903634905815125], "r_norm": [8.473408699035645, 8.536704063415527, 8.422385215759277]}, {"step": 1300, "epoch": 2, "train_loss": 12.23982048034668, "cos_r_negg": [0.007115112617611885, 0.019046638160943985, 0.3740012049674988], "cos_innovation_negg": [0.116559237241745, 0.37029099464416504, 0.29361745715141296], "cos_apical_negg": [0.007115112617611885, 0.019046638160943985, 0.3740012049674988], "cos_Ac_negg": [0.11655343323945999, 0.3702782690525055, 0.29361504316329956], "r_norm": [8.059736251831055, 8.151721000671387, 8.136381149291992]}, {"step": 1400, "epoch": 2, "train_loss": 13.769638061523438, "cos_r_negg": [0.003345026634633541, -0.1665327250957489, -0.30864810943603516], "cos_innovation_negg": [-0.006110879592597485, -0.050864942371845245, -0.40291962027549744], "cos_apical_negg": [0.003345026634633541, -0.1665327250957489, -0.30864810943603516], "cos_Ac_negg": [-0.0032902751117944717, -0.03094194084405899, -0.4010933041572571], "r_norm": [8.41011905670166, 8.431804656982422, 8.314620971679688]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.0974, "test_loss": 10.6078318359375}, {"step": 1500, "epoch": 3, "train_loss": 13.389451026916504, "cos_r_negg": [0.03407767415046692, -0.03753257542848587, -0.7653929591178894], "cos_innovation_negg": [0.026950674131512642, -0.4486401081085205, 0.10614737868309021], "cos_apical_negg": [0.03407767415046692, -0.03753257542848587, -0.7653929591178894], "cos_Ac_negg": [0.026941735297441483, -0.44864434003829956, 0.10614706575870514], "r_norm": [8.351848602294922, 8.403018951416016, 8.268065452575684], "ldr": {"dL_sdil": -2.9230241775512695, "dL_bp": -2.9230241775512695, "ratio": 1.0}}, {"step": 1600, "epoch": 3, "train_loss": 13.374156951904297, "cos_r_negg": [-0.05378840118646622, -0.07085607945919037, 0.11260397732257843], "cos_innovation_negg": [-0.021177195012569427, -0.14570271968841553, 0.042436107993125916], "cos_apical_negg": [-0.05378840118646622, -0.07085607945919037, 0.11260397732257843], "cos_Ac_negg": [-0.021177995949983597, -0.14570161700248718, 0.04243840277194977], "r_norm": [8.2179594039917, 8.293608665466309, 8.363481521606445]}, {"step": 1700, "epoch": 3, "train_loss": 7.247921943664551, "cos_r_negg": [0.0029886290431022644, -0.4298208951950073, -0.6133275628089905], "cos_innovation_negg": [-0.0890326201915741, -0.04581644386053085, -0.14729204773902893], "cos_apical_negg": [0.0029886290431022644, -0.4298208951950073, -0.6133275628089905], "cos_Ac_negg": [-0.0890505313873291, -0.045822739601135254, -0.14729176461696625], "r_norm": [8.244298934936523, 8.475659370422363, 8.392572402954102]}, {"step": 1800, "epoch": 3, "train_loss": 11.349238395690918, "cos_r_negg": [0.026772530749440193, -0.4606841206550598, -0.45862460136413574], "cos_innovation_negg": [0.052183251827955246, -0.2358827143907547, -0.1804431974887848], "cos_apical_negg": [0.026772530749440193, -0.4606841206550598, -0.45862460136413574], "cos_Ac_negg": [0.05213166028261185, -0.23588411509990692, -0.18044695258140564], "r_norm": [8.370306015014648, 8.345922470092773, 8.283858299255371]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.1032, "test_loss": 10.46290751953125}, {"step": 1900, "epoch": 4, "train_loss": 11.21202278137207, "cos_r_negg": [0.0524648018181324, -0.48226457834243774, -0.457119882106781], "cos_innovation_negg": [0.07819177210330963, -0.024834981188178062, 0.22201880812644958], "cos_apical_negg": [0.0524648018181324, -0.48226457834243774, -0.457119882106781], "cos_Ac_negg": [0.07818964123725891, -0.02483757957816124, 0.2220175713300705], "r_norm": [8.23788070678711, 8.438673973083496, 8.403940200805664]}, {"step": 2000, "epoch": 4, "train_loss": 12.917704582214355, "cos_r_negg": [0.024774182587862015, -0.35545337200164795, -0.2560240924358368], "cos_innovation_negg": [0.1487734168767929, -0.35608553886413574, 0.31173187494277954], "cos_apical_negg": [0.024774182587862015, -0.35545337200164795, -0.2560240924358368], "cos_Ac_negg": [0.14858537912368774, -0.3560815751552582, 0.3117333948612213], "r_norm": [8.501142501831055, 8.462979316711426, 8.409832000732422], "ldr": {"dL_sdil": -2.218583106994629, "dL_bp": -2.2185840606689453, "ratio": 0.9999995701428072}}, {"step": 2100, "epoch": 4, "train_loss": 7.7389984130859375, "cos_r_negg": [-0.00706346845254302, -0.2522203028202057, -0.6268371343612671], "cos_innovation_negg": [0.14600668847560883, -0.01675611361861229, 0.6800406575202942], "cos_apical_negg": [-0.00706346845254302, -0.2522203028202057, -0.6268371343612671], "cos_Ac_negg": [0.14601609110832214, -0.016754742711782455, 0.6800392866134644], "r_norm": [7.930294036865234, 7.923968315124512, 7.821818828582764]}, {"step": 2200, "epoch": 4, "train_loss": 20.193058013916016, "cos_r_negg": [-0.00441304175183177, 0.27741676568984985, 0.1075911819934845], "cos_innovation_negg": [0.24156762659549713, -0.19210182130336761, 0.38424548506736755], "cos_apical_negg": [-0.00441304175183177, 0.27741676568984985, 0.1075911819934845], "cos_Ac_negg": [0.24169965088367462, -0.19211789965629578, 0.3842449188232422], "r_norm": [8.333871841430664, 8.394327163696289, 8.443845748901367]}, {"step": 2300, "epoch": 4, "train_loss": 13.372560501098633, "cos_r_negg": [-0.01916484721004963, -0.3653978705406189, -0.22849704325199127], "cos_innovation_negg": [-0.2404419183731079, -0.1987619400024414, 0.31692421436309814], "cos_apical_negg": [-0.01916484721004963, -0.3653978705406189, -0.22849704325199127], "cos_Ac_negg": [-0.24046073853969574, -0.1987646222114563, 0.31692367792129517], "r_norm": [8.675909996032715, 8.760309219360352, 8.730667114257812]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.101, "test_loss": 18.4356232421875}, {"step": 2400, "epoch": 5, "train_loss": 13.460979461669922, "cos_r_negg": [0.05073873698711395, -0.4613088369369507, -0.7670339345932007], "cos_innovation_negg": [-0.17681385576725006, -0.19277063012123108, -0.050577402114868164], "cos_apical_negg": [0.05073873698711395, -0.4613088369369507, -0.7670339345932007], "cos_Ac_negg": [-0.1768469512462616, -0.1927729845046997, -0.05057763308286667], "r_norm": [8.736212730407715, 8.846563339233398, 8.852428436279297]}, {"step": 2500, "epoch": 5, "train_loss": 14.70837688446045, "cos_r_negg": [-0.09024171531200409, -0.4535521864891052, -0.7785865664482117], "cos_innovation_negg": [-0.12258277088403702, -0.23267489671707153, -0.24728801846504211], "cos_apical_negg": [-0.09024171531200409, -0.4535521864891052, -0.7785865664482117], "cos_Ac_negg": [-0.1221993938088417, -0.23265603184700012, -0.24729013442993164], "r_norm": [8.82823371887207, 8.948509216308594, 9.067193984985352], "ldr": {"dL_sdil": -3.5159425735473633, "dL_bp": -3.5159425735473633, "ratio": 1.0}}, {"step": 2600, "epoch": 5, "train_loss": 11.377114295959473, "cos_r_negg": [0.04751364886760712, -0.4140342175960541, -0.279313862323761], "cos_innovation_negg": [-0.14037027955055237, 0.0013605942949652672, -0.02792779728770256], "cos_apical_negg": [0.04751364886760712, -0.4140342175960541, -0.279313862323761], "cos_Ac_negg": [-0.14030057191848755, 0.0013641472905874252, -0.027927536517381668], "r_norm": [8.877184867858887, 8.897836685180664, 9.025771141052246]}, {"step": 2700, "epoch": 5, "train_loss": 8.009690284729004, "cos_r_negg": [-0.0716434195637703, -0.49250704050064087, -0.27448514103889465], "cos_innovation_negg": [-0.1671718955039978, 0.33916231989860535, 0.3194931745529175], "cos_apical_negg": [-0.0716434195637703, -0.49250704050064087, -0.27448514103889465], "cos_Ac_negg": [-0.16718293726444244, 0.3391610383987427, 0.3194931745529175], "r_norm": [8.489503860473633, 8.43554973602295, 8.49609661102295]}, {"step": 2800, "epoch": 5, "train_loss": 11.573670387268066, "cos_r_negg": [0.036237310618162155, -0.4692964255809784, 0.0005364846438169479], "cos_innovation_negg": [-0.02590039372444153, 0.4003668427467346, 0.13431602716445923], "cos_apical_negg": [0.036237310618162155, -0.4692964255809784, 0.0005364846438169479], "cos_Ac_negg": [-0.025902483612298965, 0.40036606788635254, 0.1343161165714264], "r_norm": [8.71239185333252, 8.669532775878906, 8.771892547607422]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.1028, "test_loss": 6.336495654296875}, {"step": 2900, "epoch": 6, "train_loss": 8.879903793334961, "cos_r_negg": [0.015899918973445892, -0.15473672747612, 0.2709038257598877], "cos_innovation_negg": [0.13754628598690033, -0.036785535514354706, 0.735919177532196], "cos_apical_negg": [0.015899918973445892, -0.15473672747612, 0.2709038257598877], "cos_Ac_negg": [0.1375868171453476, -0.036786727607250214, 0.7359188199043274], "r_norm": [8.138741493225098, 8.189132690429688, 8.175435066223145]}, {"step": 3000, "epoch": 6, "train_loss": 9.869020462036133, "cos_r_negg": [-0.14766976237297058, 0.29564493894577026, -0.11502531170845032], "cos_innovation_negg": [-0.1497650444507599, -0.05764322727918625, 0.6482102870941162], "cos_apical_negg": [-0.14766976237297058, 0.29564493894577026, -0.11502531170845032], "cos_Ac_negg": [-0.14863978326320648, -0.05764535069465637, 0.6482106447219849], "r_norm": [8.702150344848633, 8.741813659667969, 8.820833206176758], "ldr": {"dL_sdil": -2.196687698364258, "dL_bp": -2.196687698364258, "ratio": 1.0}}, {"step": 3100, "epoch": 6, "train_loss": 14.6121244430542, "cos_r_negg": [0.10387705266475677, -0.4871651232242584, -0.34501147270202637], "cos_innovation_negg": [0.152975931763649, -0.02190149389207363, 0.4271950125694275], "cos_apical_negg": [0.10387705266475677, -0.4871651232242584, -0.34501147270202637], "cos_Ac_negg": [0.15296901762485504, -0.021896515041589737, 0.42719489336013794], "r_norm": [7.911396026611328, 7.8898491859436035, 7.8567962646484375]}, {"step": 3200, "epoch": 6, "train_loss": 17.088899612426758, "cos_r_negg": [0.04362340644001961, -0.35752764344215393, -0.15494117140769958], "cos_innovation_negg": [0.015425674617290497, -0.04150226712226868, 0.4982662796974182], "cos_apical_negg": [0.04362340644001961, -0.35752764344215393, -0.15494117140769958], "cos_Ac_negg": [0.015409242361783981, -0.04149831086397171, 0.4982661008834839], "r_norm": [7.96759557723999, 7.954899787902832, 7.929862022399902]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.098, "test_loss": 9.185275}, {"step": 3300, "epoch": 7, "train_loss": 7.184060573577881, "cos_r_negg": [-0.1503773182630539, -0.3704270124435425, -0.7313346862792969], "cos_innovation_negg": [0.06171245127916336, -0.06041629612445831, 0.04990876838564873], "cos_apical_negg": [-0.1503773182630539, -0.3704270124435425, -0.7313346862792969], "cos_Ac_negg": [0.061731692403554916, -0.06041664630174637, 0.04990893229842186], "r_norm": [8.458047866821289, 8.463258743286133, 8.480491638183594]}, {"step": 3400, "epoch": 7, "train_loss": 11.476707458496094, "cos_r_negg": [-0.18878662586212158, -0.2959175109863281, -0.1998378485441208], "cos_innovation_negg": [0.006415966432541609, 0.10584098100662231, 0.6804267764091492], "cos_apical_negg": [-0.18878662586212158, -0.2959175109863281, -0.1998378485441208], "cos_Ac_negg": [0.006680776830762625, 0.10584947466850281, 0.6804301142692566], "r_norm": [8.060136795043945, 8.106498718261719, 8.083683967590332]}, {"step": 3500, "epoch": 7, "train_loss": 14.253765106201172, "cos_r_negg": [-0.14449641108512878, -0.4677894711494446, -0.5634869337081909], "cos_innovation_negg": [-0.19055455923080444, -0.10068078339099884, 0.013241913169622421], "cos_apical_negg": [-0.14449641108512878, -0.4677894711494446, -0.5634869337081909], "cos_Ac_negg": [-0.19058990478515625, -0.10068131238222122, 0.013241893611848354], "r_norm": [8.630552291870117, 8.639185905456543, 8.669866561889648], "ldr": {"dL_sdil": -2.0724716186523438, "dL_bp": -2.0724716186523438, "ratio": 1.0}}, {"step": 3600, "epoch": 7, "train_loss": 7.054471969604492, "cos_r_negg": [-0.16243915259838104, -0.024647122249007225, -0.12824252247810364], "cos_innovation_negg": [0.06638328731060028, 0.022071314975619316, 0.4062129557132721], "cos_apical_negg": [-0.16243915259838104, -0.024647122249007225, -0.12824252247810364], "cos_Ac_negg": [0.06640221923589706, 0.022071365267038345, 0.40621307492256165], "r_norm": [7.919602394104004, 7.961474418640137, 7.932734489440918]}, {"step": 3700, "epoch": 7, "train_loss": 11.617191314697266, "cos_r_negg": [-0.06732095777988434, -0.3359779715538025, -0.6818439364433289], "cos_innovation_negg": [0.23047637939453125, 0.2817847728729248, 0.6891072392463684], "cos_apical_negg": [-0.06732095777988434, -0.3359779715538025, -0.6818439364433289], "cos_Ac_negg": [0.23048265278339386, 0.2817847728729248, 0.6891072988510132], "r_norm": [7.801712512969971, 7.833863258361816, 7.814785957336426]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.1135, "test_loss": 12.69340283203125}, {"step": 3800, "epoch": 8, "train_loss": 10.240337371826172, "cos_r_negg": [-0.17147889733314514, -0.5333845019340515, -0.3955014944076538], "cos_innovation_negg": [-0.2837035655975342, -0.08730654418468475, 0.22334203124046326], "cos_apical_negg": [-0.17147889733314514, -0.5333845019340515, -0.3955014944076538], "cos_Ac_negg": [-0.28373071551322937, -0.08730682730674744, 0.22334203124046326], "r_norm": [8.771690368652344, 8.787416458129883, 8.83869743347168]}, {"step": 3900, "epoch": 8, "train_loss": 10.28286361694336, "cos_r_negg": [-0.08871660381555557, -0.5282332897186279, -0.2800084948539734], "cos_innovation_negg": [0.01278554555028677, 0.2824883460998535, 0.6813057065010071], "cos_apical_negg": [-0.08871660381555557, -0.5282332897186279, -0.2800084948539734], "cos_Ac_negg": [0.012820485047996044, 0.2824826240539551, 0.6813047528266907], "r_norm": [7.617300033569336, 7.614985942840576, 7.571664333343506]}, {"step": 4000, "epoch": 8, "train_loss": 20.491256713867188, "cos_r_negg": [-0.28162068128585815, -0.35697758197784424, -0.21034105122089386], "cos_innovation_negg": [-0.03233714774250984, 0.25930458307266235, 0.7615891695022583], "cos_apical_negg": [-0.28162068128585815, -0.35697758197784424, -0.21034105122089386], "cos_Ac_negg": [-0.032187316566705704, 0.25930604338645935, 0.7615891695022583], "r_norm": [7.838986396789551, 7.877105236053467, 7.871551990509033], "ldr": {"dL_sdil": -2.564411163330078, "dL_bp": -2.564411163330078, "ratio": 1.0}}, {"step": 4100, "epoch": 8, "train_loss": 5.977367401123047, "cos_r_negg": [-0.17211070656776428, -0.4709746241569519, -0.6017636060714722], "cos_innovation_negg": [-0.22096621990203857, -0.29795733094215393, 0.06907035410404205], "cos_apical_negg": [-0.17211070656776428, -0.4709746241569519, -0.6017636060714722], "cos_Ac_negg": [-0.22096644341945648, -0.2979571521282196, 0.06906933337450027], "r_norm": [8.719335556030273, 8.767009735107422, 8.825902938842773]}, {"step": 4200, "epoch": 8, "train_loss": 12.024679183959961, "cos_r_negg": [-0.21996182203292847, -0.3851681351661682, -0.2526507079601288], "cos_innovation_negg": [0.17134536802768707, 0.2002130001783371, 0.8018349409103394], "cos_apical_negg": [-0.21996182203292847, -0.3851681351661682, -0.2526507079601288], "cos_Ac_negg": [0.17138904333114624, 0.20021295547485352, 0.8018350005149841], "r_norm": [8.059459686279297, 8.085622787475586, 8.084490776062012]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.098, "test_loss": 11.16186103515625}, {"step": 4300, "epoch": 9, "train_loss": 14.091529846191406, "cos_r_negg": [-0.19059602916240692, -0.3642621338367462, -0.7494282722473145], "cos_innovation_negg": [-0.11106379330158234, -0.33209705352783203, 0.9318732023239136], "cos_apical_negg": [-0.19059602916240692, -0.3642621338367462, -0.7494282722473145], "cos_Ac_negg": [-0.1656332015991211, -0.3321491479873657, 0.9318636059761047], "r_norm": [8.805402755737305, 8.811711311340332, 8.850028991699219]}, {"step": 4400, "epoch": 9, "train_loss": 7.883824348449707, "cos_r_negg": [-0.0642743781208992, -0.488760769367218, -0.5124220252037048], "cos_innovation_negg": [-0.0667971819639206, 0.3573368787765503, 0.37410908937454224], "cos_apical_negg": [-0.0642743781208992, -0.488760769367218, -0.5124220252037048], "cos_Ac_negg": [-0.06663763523101807, 0.35733675956726074, 0.3741092085838318], "r_norm": [7.717794418334961, 7.725761413574219, 7.704934597015381]}, {"step": 4500, "epoch": 9, "train_loss": 6.33135986328125, "cos_r_negg": [-0.23611389100551605, -0.47508981823921204, -0.5680159330368042], "cos_innovation_negg": [-0.2481411248445511, -0.4730244278907776, 0.14018017053604126], "cos_apical_negg": [-0.23611389100551605, -0.47508981823921204, -0.5680159330368042], "cos_Ac_negg": [-0.24856628477573395, -0.47302454710006714, 0.14015567302703857], "r_norm": [8.927231788635254, 8.9688138961792, 9.046774864196777], "ldr": {"dL_sdil": -4.615302085876465, "dL_bp": -4.615302085876465, "ratio": 1.0}}, {"step": 4600, "epoch": 9, "train_loss": 10.78182601928711, "cos_r_negg": [0.07771191745996475, -0.22075295448303223, 0.5076983571052551], "cos_innovation_negg": [-0.09848885983228683, -0.1406552791595459, 0.7602018117904663], "cos_apical_negg": [0.07771191745996475, -0.22075295448303223, 0.5076983571052551], "cos_Ac_negg": [-0.10116881877183914, -0.1406559944152832, 0.7602018117904663], "r_norm": [8.72419548034668, 8.749523162841797, 8.857084274291992]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.1135, "test_loss": 12.2586244140625}, {"step": 4700, "epoch": 10, "train_loss": 11.82200813293457, "cos_r_negg": [-0.0790974572300911, -0.5411538481712341, -0.17489510774612427], "cos_innovation_negg": [0.22131283581256866, 0.46562597155570984, 0.503681480884552], "cos_apical_negg": [-0.0790974572300911, -0.5411538481712341, -0.17489510774612427], "cos_Ac_negg": [0.2213067263364792, 0.4656198024749756, 0.5036664605140686], "r_norm": [8.224967956542969, 8.225711822509766, 8.285591125488281]}, {"step": 4800, "epoch": 10, "train_loss": 6.89193058013916, "cos_r_negg": [-0.0241834819316864, -0.1250014454126358, 0.565407395362854], "cos_innovation_negg": [0.04167366027832031, -0.07039815187454224, 0.8029267191886902], "cos_apical_negg": [-0.0241834819316864, -0.1250014454126358, 0.565407395362854], "cos_Ac_negg": [0.0416577123105526, -0.07039816677570343, 0.8029266595840454], "r_norm": [9.021595001220703, 9.044514656066895, 9.143796920776367]}, {"step": 4900, "epoch": 10, "train_loss": 14.364617347717285, "cos_r_negg": [-0.09531091153621674, -0.360135555267334, -0.3966771364212036], "cos_innovation_negg": [-0.07605627179145813, 0.12930813431739807, 0.842853307723999], "cos_apical_negg": [-0.09531091153621674, -0.360135555267334, -0.3966771364212036], "cos_Ac_negg": [-0.07411383092403412, 0.12930810451507568, 0.842853307723999], "r_norm": [7.924593925476074, 7.934759140014648, 7.925215721130371]}, {"step": 5000, "epoch": 10, "train_loss": 14.163548469543457, "cos_r_negg": [0.14270110428333282, 0.26727262139320374, -0.24067597091197968], "cos_innovation_negg": [0.2650975286960602, 0.09857301414012909, 0.46744969487190247], "cos_apical_negg": [0.14270110428333282, 0.26727262139320374, -0.24067597091197968], "cos_Ac_negg": [0.2659301459789276, 0.09856803715229034, 0.46744999289512634], "r_norm": [8.169264793395996, 8.181272506713867, 8.210319519042969], "ldr": {"dL_sdil": -3.5037050247192383, "dL_bp": -3.5037050247192383, "ratio": 1.0}}, {"step": 5100, "epoch": 10, "train_loss": 13.023917198181152, "cos_r_negg": [-0.35690081119537354, -0.47259896993637085, -0.6792689561843872], "cos_innovation_negg": [-0.41525083780288696, -0.3760499954223633, 0.219245046377182], "cos_apical_negg": [-0.35690081119537354, -0.47259896993637085, -0.6792689561843872], "cos_Ac_negg": [-0.41535335779190063, -0.37605011463165283, 0.219245046377182], "r_norm": [9.015456199645996, 9.048063278198242, 9.120501518249512]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.0958, "test_loss": 8.358147314453126}, {"step": 5200, "epoch": 11, "train_loss": 9.056848526000977, "cos_r_negg": [0.143497496843338, 0.29202643036842346, 0.40848076343536377], "cos_innovation_negg": [0.20552386343479156, 0.1962636113166809, 0.8469909429550171], "cos_apical_negg": [0.143497496843338, 0.29202643036842346, 0.40848076343536377], "cos_Ac_negg": [0.20551790297031403, 0.1962636113166809, 0.8469908237457275], "r_norm": [8.368833541870117, 8.382272720336914, 8.390267372131348]}, {"step": 5300, "epoch": 11, "train_loss": 10.967495918273926, "cos_r_negg": [-0.33753615617752075, -0.3109343647956848, -0.7291955947875977], "cos_innovation_negg": [-0.012439723126590252, -0.03944985568523407, 0.9048149585723877], "cos_apical_negg": [-0.33753615617752075, -0.3109343647956848, -0.7291955947875977], "cos_Ac_negg": [-0.012433229945600033, -0.039450086653232574, 0.9048150777816772], "r_norm": [8.316133499145508, 8.330072402954102, 8.333383560180664]}, {"step": 5400, "epoch": 11, "train_loss": 13.981063842773438, "cos_r_negg": [-0.15270976722240448, -0.41802653670310974, -0.41921016573905945], "cos_innovation_negg": [-0.000549065473023802, -0.3085049092769623, 0.4921051561832428], "cos_apical_negg": [-0.15270976722240448, -0.41802653670310974, -0.41921016573905945], "cos_Ac_negg": [-0.0007135692867450416, -0.30850499868392944, 0.4921052157878876], "r_norm": [8.57032585144043, 8.574323654174805, 8.580300331115723]}, {"step": 5500, "epoch": 11, "train_loss": 13.959887504577637, "cos_r_negg": [-0.27541258931159973, -0.34256115555763245, -0.7369647026062012], "cos_innovation_negg": [0.23313891887664795, 0.075493223965168, 0.9400489330291748], "cos_apical_negg": [-0.27541258931159973, -0.34256115555763245, -0.7369647026062012], "cos_Ac_negg": [0.23306874930858612, 0.07549314200878143, 0.9400489330291748], "r_norm": [8.232297897338867, 8.24512767791748, 8.244892120361328], "ldr": {"dL_sdil": -2.124760627746582, "dL_bp": -2.124760627746582, "ratio": 1.0}}, {"step": 5600, "epoch": 11, "train_loss": 9.787443161010742, "cos_r_negg": [-0.3273448944091797, -0.49620890617370605, -0.6517419219017029], "cos_innovation_negg": [-0.3757983446121216, -0.3121603727340698, 0.40261226892471313], "cos_apical_negg": [-0.3273448944091797, -0.49620890617370605, -0.6517419219017029], "cos_Ac_negg": [-0.37583214044570923, -0.3121604323387146, 0.40261223912239075], "r_norm": [8.811349868774414, 8.843202590942383, 8.900054931640625]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.098, "test_loss": 11.8320888671875}, {"step": 5700, "epoch": 12, "train_loss": 15.085618019104004, "cos_r_negg": [-0.25524723529815674, -0.3805316388607025, -0.2532828450202942], "cos_innovation_negg": [0.2877015769481659, 0.34255295991897583, 0.5990843176841736], "cos_apical_negg": [-0.25524723529815674, -0.3805316388607025, -0.2532828450202942], "cos_Ac_negg": [0.28769752383232117, 0.3425529897212982, 0.5990841388702393], "r_norm": [7.480499744415283, 7.491408348083496, 7.479680061340332]}, {"step": 5800, "epoch": 12, "train_loss": 7.651779651641846, "cos_r_negg": [-0.2340777963399887, -0.36308926343917847, -0.2652818560600281], "cos_innovation_negg": [0.22504860162734985, -0.03697636350989342, 0.6941032409667969], "cos_apical_negg": [-0.2340777963399887, -0.36308926343917847, -0.2652818560600281], "cos_Ac_negg": [0.22500714659690857, -0.03697629272937775, 0.6941028833389282], "r_norm": [8.390066146850586, 8.395925521850586, 8.41372299194336]}, {"step": 5900, "epoch": 12, "train_loss": 9.20200252532959, "cos_r_negg": [-0.19804447889328003, -0.2335311472415924, -0.022346118465065956], "cos_innovation_negg": [0.2170308530330658, 0.017645228654146194, 0.7609648704528809], "cos_apical_negg": [-0.19804447889328003, -0.2335311472415924, -0.022346118465065956], "cos_Ac_negg": [0.21703936159610748, 0.017645198851823807, 0.7609648704528809], "r_norm": [8.290702819824219, 8.303144454956055, 8.30451774597168]}, {"step": 6000, "epoch": 12, "train_loss": 8.42834186553955, "cos_r_negg": [-0.1310756653547287, -0.42125949263572693, -0.49664759635925293], "cos_innovation_negg": [0.05727000907063484, -0.1939908117055893, 0.49494606256484985], "cos_apical_negg": [-0.1310756653547287, -0.42125949263572693, -0.49664759635925293], "cos_Ac_negg": [0.05721993371844292, -0.1939908117055893, 0.4949462413787842], "r_norm": [8.714231491088867, 8.73741626739502, 8.776906967163086], "ldr": {"dL_sdil": -2.2645034790039062, "dL_bp": -2.2645034790039062, "ratio": 1.0}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.0982, "test_loss": 8.90008359375}, {"step": 6100, "epoch": 13, "train_loss": 10.190654754638672, "cos_r_negg": [-0.131789892911911, -0.34542152285575867, -0.577663004398346], "cos_innovation_negg": [0.32587963342666626, 0.2613799571990967, 0.9931835532188416], "cos_apical_negg": [-0.131789892911911, -0.34542152285575867, -0.577663004398346], "cos_Ac_negg": [0.3258972465991974, 0.26138001680374146, 0.9931835532188416], "r_norm": [8.087332725524902, 8.10606861114502, 8.103015899658203]}, {"step": 6200, "epoch": 13, "train_loss": 15.763067245483398, "cos_r_negg": [-0.22160448133945465, -0.44533249735832214, -0.5738877058029175], "cos_innovation_negg": [-0.34329837560653687, -0.5508235692977905, 0.3571058511734009], "cos_apical_negg": [-0.22160448133945465, -0.44533249735832214, -0.5738877058029175], "cos_Ac_negg": [-0.3432968854904175, -0.5508233308792114, 0.35710570216178894], "r_norm": [8.938228607177734, 8.966825485229492, 9.043449401855469]}, {"step": 6300, "epoch": 13, "train_loss": 7.8466410636901855, "cos_r_negg": [-0.29713305830955505, -0.3718841075897217, -0.7026219964027405], "cos_innovation_negg": [0.2911476492881775, 0.14079971611499786, 0.8674792051315308], "cos_apical_negg": [-0.29713305830955505, -0.3718841075897217, -0.7026219964027405], "cos_Ac_negg": [0.2911386489868164, 0.1407995969057083, 0.8674792051315308], "r_norm": [8.247562408447266, 8.26615047454834, 8.277181625366211]}, {"step": 6400, "epoch": 13, "train_loss": 14.184345245361328, "cos_r_negg": [-0.34605881571769714, -0.3860695958137512, -0.728797197341919], "cos_innovation_negg": [0.3147152364253998, 0.32621294260025024, 0.8966970443725586], "cos_apical_negg": [-0.34605881571769714, -0.3860695958137512, -0.728797197341919], "cos_Ac_negg": [0.31331801414489746, 0.3262106776237488, 0.8966970443725586], "r_norm": [8.174250602722168, 8.199581146240234, 8.235008239746094]}, {"step": 6500, "epoch": 13, "train_loss": 11.187944412231445, "cos_r_negg": [-0.0576140470802784, -0.3494236469268799, -0.27806684374809265], "cos_innovation_negg": [-0.009291226044297218, 0.36055776476860046, 0.7719247341156006], "cos_apical_negg": [-0.0576140470802784, -0.3494236469268799, -0.27806684374809265], "cos_Ac_negg": [-0.009204496629536152, 0.36056098341941833, 0.7719250917434692], "r_norm": [8.064088821411133, 8.076695442199707, 8.132064819335938], "ldr": {"dL_sdil": -6.145732879638672, "dL_bp": -6.145732879638672, "ratio": 1.0}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.0982, "test_loss": 12.07663271484375}, {"step": 6600, "epoch": 14, "train_loss": 7.594141006469727, "cos_r_negg": [-0.143095925450325, -0.2919423580169678, -0.12109537422657013], "cos_innovation_negg": [0.13657832145690918, 0.01721436157822609, 0.7777774930000305], "cos_apical_negg": [-0.143095925450325, -0.2919423580169678, -0.12109537422657013], "cos_Ac_negg": [0.13656969368457794, 0.01721443422138691, 0.7777774930000305], "r_norm": [8.260603904724121, 8.273170471191406, 8.27216911315918]}, {"step": 6700, "epoch": 14, "train_loss": 10.048295974731445, "cos_r_negg": [-0.31014254689216614, -0.4719916582107544, -0.7264832258224487], "cos_innovation_negg": [0.013995213434100151, 0.061046868562698364, 0.6761834621429443], "cos_apical_negg": [-0.31014254689216614, -0.4719916582107544, -0.7264832258224487], "cos_Ac_negg": [0.014052307233214378, 0.06104706972837448, 0.6761834621429443], "r_norm": [8.349325180053711, 8.3685302734375, 8.412313461303711]}, {"step": 6800, "epoch": 14, "train_loss": 8.724132537841797, "cos_r_negg": [-0.03416309133172035, -0.15668511390686035, 0.019666921347379684], "cos_innovation_negg": [-0.02465755119919777, 0.025783326476812363, 0.9365755915641785], "cos_apical_negg": [-0.03416309133172035, -0.15668511390686035, 0.019666921347379684], "cos_Ac_negg": [-0.02466672472655773, 0.0257832370698452, 0.9365756511688232], "r_norm": [8.282548904418945, 8.297465324401855, 8.315971374511719]}, {"step": 6900, "epoch": 14, "train_loss": 14.09990119934082, "cos_r_negg": [-0.255214661359787, -0.3579021096229553, -0.4531940817832947], "cos_innovation_negg": [0.328462690114975, 0.20817600190639496, 0.8455662727355957], "cos_apical_negg": [-0.255214661359787, -0.3579021096229553, -0.4531940817832947], "cos_Ac_negg": [0.3284487724304199, 0.20817597210407257, 0.8455663919448853], "r_norm": [8.243138313293457, 8.261075973510742, 8.279829025268555]}, {"step": 7000, "epoch": 14, "train_loss": 8.938194274902344, "cos_r_negg": [0.08056957274675369, -0.14365482330322266, 0.5462514758110046], "cos_innovation_negg": [0.21136142313480377, -0.13913632929325104, 0.8379552364349365], "cos_apical_negg": [0.08056957274675369, -0.14365482330322266, 0.5462514758110046], "cos_Ac_negg": [0.21127372980117798, -0.13913647830486298, 0.8379553556442261], "r_norm": [8.701251983642578, 8.723127365112305, 8.845736503601074], "ldr": {"dL_sdil": -2.529360771179199, "dL_bp": -2.529360771179199, "ratio": 1.0}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.1028, "test_loss": 8.322147998046875}], "final": {"test_acc": 0.1028, "test_loss": 8.322147998046875, "wall_s": 22.142099618911743, "cos_r_negg": [-0.2482898086309433, -0.316217303276062, -0.7253843545913696], "cos_innovation_negg": [0.15539732575416565, 0.10566452890634537, 0.8286952972412109], "cos_apical_negg": [-0.2482898086309433, -0.316217303276062, -0.7253843545913696], "cos_Ac_negg": [0.15538889169692993, 0.10566452145576477, 0.8286952972412109]}}