summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p5_matched_s3.json
blob: 61f8f234cc038cbecf13495d8c6b18a3b6205b0a (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": "match_innovation_norm", "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_matched_s3", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.331878900527954, "cos_r_negg": [-0.023404492065310478, 0.009263666346669197, -0.003423936665058136], "cos_innovation_negg": [-0.023943275213241577, 0.015888750553131104, 0.006476129405200481], "cos_apical_negg": [-0.023404492065310478, 0.009263665415346622, -0.0034239361993968487], "cos_Ac_negg": [-0.023945096880197525, 0.015887551009655, 0.006474699825048447], "r_norm": [4.880455017089844, 4.766017913818359, 4.871740341186523], "ldr": {"dL_sdil": 0.05150461196899414, "dL_bp": -0.2127234935760498, "ratio": -0.24211999861021916}}, {"step": 100, "epoch": 0, "train_loss": 3.93803071975708, "cos_r_negg": [-0.05274544283747673, 0.18825885653495789, -0.44922441244125366], "cos_innovation_negg": [-0.09187984466552734, -0.00649038702249527, -0.33347389101982117], "cos_apical_negg": [-0.05274543911218643, 0.18825885653495789, -0.44922447204589844], "cos_Ac_negg": [-0.09188588708639145, -0.006491729989647865, -0.33348870277404785], "r_norm": [5.036410331726074, 4.923680305480957, 5.091097831726074]}, {"step": 200, "epoch": 0, "train_loss": 12.228022575378418, "cos_r_negg": [-0.017211202532052994, 0.05333784595131874, 0.36261531710624695], "cos_innovation_negg": [-0.04136255010962486, 0.09066534042358398, 0.14909055829048157], "cos_apical_negg": [-0.017211204394698143, 0.053337857127189636, 0.36261534690856934], "cos_Ac_negg": [-0.04134999215602875, 0.09060462564229965, 0.1490904688835144], "r_norm": [5.575984001159668, 5.387224197387695, 5.670280933380127]}, {"step": 300, "epoch": 0, "train_loss": 9.527034759521484, "cos_r_negg": [0.09905003011226654, 0.11598698794841766, -0.337249219417572], "cos_innovation_negg": [0.13977257907390594, 0.2395804226398468, -0.057061560451984406], "cos_apical_negg": [0.09905002266168594, 0.11598698794841766, -0.33724915981292725], "cos_Ac_negg": [0.13977400958538055, 0.2395714521408081, -0.05707992985844612], "r_norm": [5.27436637878418, 5.377788543701172, 5.334036350250244]}, {"step": 400, "epoch": 0, "train_loss": 10.18307876586914, "cos_r_negg": [-0.016085131093859673, -0.22224563360214233, -0.10172560811042786], "cos_innovation_negg": [-0.06050053983926773, -0.0628795176744461, 0.5877997875213623], "cos_apical_negg": [-0.016085131093859673, -0.22224563360214233, -0.10172560811042786], "cos_Ac_negg": [-0.06043291836977005, -0.06283138692378998, 0.5877983570098877], "r_norm": [5.364789009094238, 5.296049118041992, 5.427796840667725]}, {"epoch_end": 0, "step": 469, "test_acc": 0.098, "test_loss": 13.6578900390625}, {"step": 500, "epoch": 1, "train_loss": 11.668904304504395, "cos_r_negg": [-0.023274647071957588, 0.11952609568834305, -0.2331356406211853], "cos_innovation_negg": [-0.016597865149378777, 0.12962821125984192, -0.21243219077587128], "cos_apical_negg": [-0.023274652659893036, 0.11952609568834305, -0.23313559591770172], "cos_Ac_negg": [-0.016598738729953766, 0.12961795926094055, -0.2124209702014923], "r_norm": [4.465723991394043, 4.6360955238342285, 4.811782360076904], "ldr": {"dL_sdil": -1.660008430480957, "dL_bp": -1.660008430480957, "ratio": 1.0}}, {"step": 600, "epoch": 1, "train_loss": 9.913995742797852, "cos_r_negg": [0.03544580936431885, 0.010761044919490814, 0.11519358307123184], "cos_innovation_negg": [0.024819735437631607, 0.13745856285095215, 0.30716899037361145], "cos_apical_negg": [0.03544580191373825, 0.010761039331555367, 0.11519362777471542], "cos_Ac_negg": [0.024800240993499756, 0.13745294511318207, 0.3071695566177368], "r_norm": [4.528250694274902, 4.526756286621094, 4.732523441314697]}, {"step": 700, "epoch": 1, "train_loss": 11.596376419067383, "cos_r_negg": [-0.016515249386429787, -0.024830589070916176, -0.6743102073669434], "cos_innovation_negg": [-0.0237373486161232, 0.10925282537937164, -0.04644819349050522], "cos_apical_negg": [-0.016515253111720085, -0.024830587208271027, -0.6743102669715881], "cos_Ac_negg": [-0.024258635938167572, 0.10920343548059464, -0.04642816632986069], "r_norm": [4.360329627990723, 4.562563896179199, 4.7867584228515625]}, {"step": 800, "epoch": 1, "train_loss": 16.685710906982422, "cos_r_negg": [0.05001528561115265, -0.15051516890525818, -0.476337194442749], "cos_innovation_negg": [0.03804410994052887, 0.06662282347679138, 0.6450439691543579], "cos_apical_negg": [0.05001528188586235, -0.15051516890525818, -0.47633713483810425], "cos_Ac_negg": [0.03804052621126175, 0.06666028499603271, 0.6450420618057251], "r_norm": [4.281778335571289, 4.418721675872803, 4.574766159057617]}, {"step": 900, "epoch": 1, "train_loss": 8.289448738098145, "cos_r_negg": [-0.03426763042807579, -0.14169526100158691, -0.39391225576400757], "cos_innovation_negg": [-0.10632513463497162, 0.06073577329516411, -0.07448279857635498], "cos_apical_negg": [-0.03426763042807579, -0.14169526100158691, -0.3939122259616852], "cos_Ac_negg": [-0.10630496591329575, 0.0607910230755806, -0.07447711378335953], "r_norm": [3.7901594638824463, 4.016811847686768, 4.248052597045898]}, {"epoch_end": 1, "step": 938, "test_acc": 0.1028, "test_loss": 9.41012841796875}, {"step": 1000, "epoch": 2, "train_loss": 12.244726181030273, "cos_r_negg": [-0.004659376572817564, -0.005834559444338083, -0.5812898874282837], "cos_innovation_negg": [-0.014617500826716423, 0.18938259780406952, -0.0436372235417366], "cos_apical_negg": [-0.0046593728475272655, -0.005834552459418774, -0.5812898874282837], "cos_Ac_negg": [-0.014623701572418213, 0.18937918543815613, -0.04364272952079773], "r_norm": [3.4024252891540527, 3.7132411003112793, 3.841618061065674], "ldr": {"dL_sdil": -4.327396392822266, "dL_bp": -4.327396392822266, "ratio": 1.0}}, {"step": 1100, "epoch": 2, "train_loss": 13.012718200683594, "cos_r_negg": [0.035421840846538544, 0.07751062512397766, 0.29841554164886475], "cos_innovation_negg": [0.034294068813323975, 0.33310723304748535, 0.41063791513442993], "cos_apical_negg": [0.03542184457182884, 0.07751062512397766, 0.29841557145118713], "cos_Ac_negg": [0.033763423562049866, 0.3331449031829834, 0.4106253385543823], "r_norm": [3.697913646697998, 3.8519930839538574, 4.062531471252441]}, {"step": 1200, "epoch": 2, "train_loss": 13.053409576416016, "cos_r_negg": [0.004354442469775677, -0.117958202958107, -0.7837634086608887], "cos_innovation_negg": [-0.0614975206553936, 0.0449785478413105, -0.6589546799659729], "cos_apical_negg": [0.004354442469775677, -0.117958202958107, -0.7837634086608887], "cos_Ac_negg": [-0.05989518016576767, 0.023895755410194397, -0.6684364080429077], "r_norm": [3.6265907287597656, 4.030216217041016, 4.190539360046387]}, {"step": 1300, "epoch": 2, "train_loss": 5.807992458343506, "cos_r_negg": [0.021018171682953835, 0.05074962228536606, 0.2531043589115143], "cos_innovation_negg": [0.01744532585144043, 0.04455544427037239, 0.11088324338197708], "cos_apical_negg": [0.021018177270889282, 0.050749629735946655, 0.2531043291091919], "cos_Ac_negg": [0.017444536089897156, 0.044559065252542496, 0.11088646948337555], "r_norm": [2.9064719676971436, 3.20048451423645, 3.3757195472717285]}, {"step": 1400, "epoch": 2, "train_loss": 16.68470001220703, "cos_r_negg": [-0.02570568025112152, -0.21907366812229156, -0.7805821299552917], "cos_innovation_negg": [-0.0555734746158123, 0.1145012378692627, -0.4349041283130646], "cos_apical_negg": [-0.025705672800540924, -0.21907366812229156, -0.7805821299552917], "cos_Ac_negg": [-0.055576201528310776, 0.11466749012470245, -0.4349042475223541], "r_norm": [3.105128765106201, 3.472952365875244, 3.6450729370117188]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.0974, "test_loss": 16.10644501953125}, {"step": 1500, "epoch": 3, "train_loss": 10.99614429473877, "cos_r_negg": [-0.010981716215610504, -0.10529394447803497, -0.6657657623291016], "cos_innovation_negg": [-0.04375379532575607, -0.006582379341125488, 0.470498651266098], "cos_apical_negg": [-0.010981718078255653, -0.10529392957687378, -0.6657657623291016], "cos_Ac_negg": [-0.043721724301576614, -0.006521882489323616, 0.4704965651035309], "r_norm": [3.093276023864746, 3.3588638305664062, 3.678278923034668], "ldr": {"dL_sdil": -3.1172027587890625, "dL_bp": -3.1172027587890625, "ratio": 1.0}}, {"step": 1600, "epoch": 3, "train_loss": 13.572635650634766, "cos_r_negg": [0.032471850514411926, 0.09270533919334412, -0.5350372195243835], "cos_innovation_negg": [0.08556779474020004, 0.13869449496269226, 0.20785954594612122], "cos_apical_negg": [0.032471850514411926, 0.09270535409450531, -0.5350371599197388], "cos_Ac_negg": [0.08557653427124023, 0.13869059085845947, 0.20785681903362274], "r_norm": [2.8790063858032227, 3.2431540489196777, 3.545623779296875]}, {"step": 1700, "epoch": 3, "train_loss": 15.694596290588379, "cos_r_negg": [0.033936936408281326, -0.08957314491271973, -0.6843352317810059], "cos_innovation_negg": [-0.012293849140405655, 0.010994723066687584, 0.40581268072128296], "cos_apical_negg": [0.033936940133571625, -0.08957314491271973, -0.6843352913856506], "cos_Ac_negg": [-0.012302355840802193, 0.011027035303413868, 0.4058120548725128], "r_norm": [2.7849206924438477, 3.0635476112365723, 3.429154396057129]}, {"step": 1800, "epoch": 3, "train_loss": 16.105371475219727, "cos_r_negg": [0.021749984472990036, -0.14429184794425964, -0.6999454498291016], "cos_innovation_negg": [0.017163287848234177, 0.11140023171901703, 0.14725594222545624], "cos_apical_negg": [0.021749980747699738, -0.14429183304309845, -0.6999454498291016], "cos_Ac_negg": [0.01746225915849209, 0.11156095564365387, 0.14757612347602844], "r_norm": [2.761136531829834, 3.095857858657837, 3.5029616355895996]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.0982, "test_loss": 11.78546201171875}, {"step": 1900, "epoch": 4, "train_loss": 8.233247756958008, "cos_r_negg": [0.0118138138204813, -0.052501268684864044, 0.2882298231124878], "cos_innovation_negg": [-0.014771915040910244, 0.2622224986553192, 0.34787923097610474], "cos_apical_negg": [0.011813816614449024, -0.052501268684864044, 0.2882297933101654], "cos_Ac_negg": [-0.014770178124308586, 0.26222601532936096, 0.34787771105766296], "r_norm": [2.276346206665039, 2.6683669090270996, 2.9762845039367676]}, {"step": 2000, "epoch": 4, "train_loss": 10.540949821472168, "cos_r_negg": [-0.002448295708745718, 0.046116843819618225, -0.36156171560287476], "cos_innovation_negg": [0.020891379565000534, 0.04625238850712776, 0.1458940953016281], "cos_apical_negg": [-0.002448292449116707, 0.04611684009432793, -0.36156171560287476], "cos_Ac_negg": [0.02089184708893299, 0.04623724892735481, 0.1458939015865326], "r_norm": [2.3217287063598633, 2.6515660285949707, 3.0398008823394775], "ldr": {"dL_sdil": -1.352426528930664, "dL_bp": -1.352426528930664, "ratio": 1.0}}, {"step": 2100, "epoch": 4, "train_loss": 9.533159255981445, "cos_r_negg": [0.0063061583787202835, 0.07012288272380829, -0.20739030838012695], "cos_innovation_negg": [0.03759526461362839, 0.38612493872642517, 0.445768803358078], "cos_apical_negg": [0.0063061523251235485, 0.07012288272380829, -0.20739032328128815], "cos_Ac_negg": [0.037591226398944855, 0.3861226737499237, 0.4457683563232422], "r_norm": [2.1785190105438232, 2.5417747497558594, 2.8388586044311523]}, {"step": 2200, "epoch": 4, "train_loss": 7.513348579406738, "cos_r_negg": [-0.0007374449633061886, -0.00730521697551012, 0.04074469208717346], "cos_innovation_negg": [0.017737476155161858, 0.3992329239845276, 0.36693888902664185], "cos_apical_negg": [-0.0007374389097094536, -0.007305213250219822, 0.04074469208717346], "cos_Ac_negg": [0.017734572291374207, 0.3992356061935425, 0.36693844199180603], "r_norm": [2.142125368118286, 2.489021062850952, 2.871180534362793]}, {"step": 2300, "epoch": 4, "train_loss": 13.4251708984375, "cos_r_negg": [0.03304494544863701, 0.03516755998134613, -0.23744675517082214], "cos_innovation_negg": [0.0314888060092926, 0.23710165917873383, 0.3066062927246094], "cos_apical_negg": [0.03304494917392731, 0.03516756370663643, -0.23744671046733856], "cos_Ac_negg": [0.03148419409990311, 0.2371007353067398, 0.3066060245037079], "r_norm": [2.0966086387634277, 2.517566680908203, 2.866859197616577]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.1009, "test_loss": 9.70174755859375}, {"step": 2400, "epoch": 5, "train_loss": 12.111278533935547, "cos_r_negg": [0.04637152701616287, 0.03708937391638756, -0.44551995396614075], "cos_innovation_negg": [-0.011598706245422363, 0.29721200466156006, 0.18614904582500458], "cos_apical_negg": [0.046371523290872574, 0.03708937391638756, -0.44551992416381836], "cos_Ac_negg": [-0.01160514261573553, 0.2972118854522705, 0.18614903092384338], "r_norm": [1.9773354530334473, 2.295269012451172, 2.829575538635254]}, {"step": 2500, "epoch": 5, "train_loss": 12.332671165466309, "cos_r_negg": [0.041929714381694794, -0.06426175683736801, 0.5717096328735352], "cos_innovation_negg": [0.07432641088962555, 0.18175101280212402, 0.6384155750274658], "cos_apical_negg": [0.04192971810698509, -0.06426175683736801, 0.5717096328735352], "cos_Ac_negg": [0.0743139386177063, 0.18174266815185547, 0.6384155750274658], "r_norm": [2.136064291000366, 2.6100192070007324, 3.3647665977478027], "ldr": {"dL_sdil": -2.3579254150390625, "dL_bp": -2.3579254150390625, "ratio": 1.0}}, {"step": 2600, "epoch": 5, "train_loss": 14.47452163696289, "cos_r_negg": [-0.013248052448034286, 0.05644286423921585, -0.7643646001815796], "cos_innovation_negg": [-0.17318934202194214, 0.050919756293296814, -0.022107506170868874], "cos_apical_negg": [-0.013248054310679436, 0.05644286423921585, -0.7643646001815796], "cos_Ac_negg": [-0.1731943041086197, 0.050827860832214355, -0.02210751362144947], "r_norm": [2.2183806896209717, 2.5999670028686523, 3.0638198852539062]}, {"step": 2700, "epoch": 5, "train_loss": 7.529302597045898, "cos_r_negg": [0.06156715005636215, 0.09324220567941666, 0.5201743841171265], "cos_innovation_negg": [0.05897732824087143, 0.3124152421951294, -0.12993979454040527], "cos_apical_negg": [0.06156715005636215, 0.09324220567941666, 0.5201743841171265], "cos_Ac_negg": [0.05895711109042168, 0.3116137385368347, -0.1299402415752411], "r_norm": [2.144774913787842, 2.3828392028808594, 2.95363712310791]}, {"step": 2800, "epoch": 5, "train_loss": 11.245426177978516, "cos_r_negg": [0.021822623908519745, -0.07685096561908722, 0.1306837499141693], "cos_innovation_negg": [0.052651163190603256, 0.024222731590270996, 0.4682297110557556], "cos_apical_negg": [0.021822625771164894, -0.07685096561908722, 0.1306837499141693], "cos_Ac_negg": [0.05265333503484726, 0.024229511618614197, 0.46822965145111084], "r_norm": [1.837006688117981, 2.098672389984131, 2.7745492458343506]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.0982, "test_loss": 9.00151484375}, {"step": 2900, "epoch": 6, "train_loss": 10.699666023254395, "cos_r_negg": [0.003275001421570778, 0.01193692535161972, 0.5769363641738892], "cos_innovation_negg": [0.0802985355257988, 0.21270713210105896, 0.886192262172699], "cos_apical_negg": [0.003275001421570778, 0.011936930939555168, 0.5769364237785339], "cos_Ac_negg": [0.08028008043766022, 0.2127087414264679, 0.8861921429634094], "r_norm": [1.9166699647903442, 2.342777729034424, 3.2197136878967285]}, {"step": 3000, "epoch": 6, "train_loss": 8.966672897338867, "cos_r_negg": [0.012364748865365982, -0.06855656951665878, -0.5414963364601135], "cos_innovation_negg": [0.04855531081557274, 0.25342071056365967, 0.15319502353668213], "cos_apical_negg": [0.012364747002720833, -0.06855656951665878, -0.5414963364601135], "cos_Ac_negg": [0.04855132848024368, 0.25339818000793457, 0.15319499373435974], "r_norm": [1.7323641777038574, 1.9787015914916992, 2.7299466133117676], "ldr": {"dL_sdil": -1.3645620346069336, "dL_bp": -1.3645620346069336, "ratio": 1.0}}, {"step": 3100, "epoch": 6, "train_loss": 7.404589653015137, "cos_r_negg": [-0.00993923656642437, 0.03794018179178238, -0.2275383174419403], "cos_innovation_negg": [0.05834876000881195, 0.31114208698272705, 0.4983579218387604], "cos_apical_negg": [-0.009939240291714668, 0.03794018179178238, -0.2275383174419403], "cos_Ac_negg": [0.05834975466132164, 0.31113043427467346, 0.498357892036438], "r_norm": [1.839612603187561, 2.121492862701416, 2.895768404006958]}, {"step": 3200, "epoch": 6, "train_loss": 13.985771179199219, "cos_r_negg": [0.009052676148712635, 0.044186364859342575, -0.6484626531600952], "cos_innovation_negg": [0.007429224438965321, -0.08652340620756149, 0.35605454444885254], "cos_apical_negg": [0.00905267521739006, 0.044186364859342575, -0.6484626531600952], "cos_Ac_negg": [0.007442607544362545, -0.08654331415891647, 0.35605478286743164], "r_norm": [1.8577483892440796, 2.176766872406006, 3.1609139442443848]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.1032, "test_loss": 9.5078845703125}, {"step": 3300, "epoch": 7, "train_loss": 14.500346183776855, "cos_r_negg": [0.03023715317249298, 0.05987554416060448, 0.10631096363067627], "cos_innovation_negg": [0.01239866204559803, 0.21723376214504242, 0.3753962516784668], "cos_apical_negg": [0.03023715876042843, 0.05987554043531418, 0.10631096363067627], "cos_Ac_negg": [0.010563991032540798, 0.21948036551475525, 0.37538307905197144], "r_norm": [1.9232898950576782, 2.158068895339966, 2.9749834537506104]}, {"step": 3400, "epoch": 7, "train_loss": 8.561042785644531, "cos_r_negg": [-0.007763438392430544, 0.03267252817749977, -0.5633492469787598], "cos_innovation_negg": [0.005812717601656914, 0.08363096415996552, 0.49122267961502075], "cos_apical_negg": [-0.007763439789414406, 0.032672520726919174, -0.5633492469787598], "cos_Ac_negg": [0.005812252406030893, 0.0836193859577179, 0.491222620010376], "r_norm": [1.8085782527923584, 2.004695415496826, 2.943936824798584]}, {"step": 3500, "epoch": 7, "train_loss": 6.397796154022217, "cos_r_negg": [-0.009914370253682137, 0.10034030675888062, -0.46877723932266235], "cos_innovation_negg": [-0.003673017956316471, 0.16520525515079498, 0.35941267013549805], "cos_apical_negg": [-0.009914370253682137, 0.10034029930830002, -0.46877723932266235], "cos_Ac_negg": [-0.0036725131794810295, 0.16517749428749084, 0.35941267013549805], "r_norm": [1.6035305261611938, 1.8609508275985718, 2.7025599479675293], "ldr": {"dL_sdil": -1.8581633567810059, "dL_bp": -1.8581633567810059, "ratio": 1.0}}, {"step": 3600, "epoch": 7, "train_loss": 8.999161720275879, "cos_r_negg": [-0.02817051112651825, 0.03443991392850876, -0.6405396461486816], "cos_innovation_negg": [-0.10997366905212402, -0.2562880516052246, 0.07866506278514862], "cos_apical_negg": [-0.0281705092638731, 0.03443991392850876, -0.6405396461486816], "cos_Ac_negg": [-0.10997616499662399, -0.2562910318374634, 0.07866504788398743], "r_norm": [1.6904629468917847, 1.9610121250152588, 2.7385761737823486]}, {"step": 3700, "epoch": 7, "train_loss": 14.690881729125977, "cos_r_negg": [-0.008290690369904041, 0.07577173411846161, -0.6582216620445251], "cos_innovation_negg": [-0.06754305958747864, -0.10545620322227478, 0.6658633947372437], "cos_apical_negg": [-0.008290693163871765, 0.0757717415690422, -0.6582217216491699], "cos_Ac_negg": [-0.06750993430614471, -0.10544462502002716, 0.6658633947372437], "r_norm": [1.7294771671295166, 2.0196428298950195, 3.1941299438476562]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.1009, "test_loss": 12.96227607421875}, {"step": 3800, "epoch": 8, "train_loss": 10.057510375976562, "cos_r_negg": [-0.04070144519209862, 0.08962370455265045, -0.7099792957305908], "cos_innovation_negg": [-0.1700393259525299, -0.39376434683799744, 0.015467292629182339], "cos_apical_negg": [-0.04070143774151802, 0.08962368965148926, -0.7099792957305908], "cos_Ac_negg": [-0.16983087360858917, -0.3937705159187317, 0.015468060038983822], "r_norm": [1.814821720123291, 2.1310102939605713, 2.9847631454467773]}, {"step": 3900, "epoch": 8, "train_loss": 8.215349197387695, "cos_r_negg": [-0.0048723118379712105, 0.09534664452075958, -0.4127228260040283], "cos_innovation_negg": [-0.03572871536016464, -0.4212097227573395, 0.7570983171463013], "cos_apical_negg": [-0.004872307181358337, 0.09534664452075958, -0.4127228260040283], "cos_Ac_negg": [-0.035726048052310944, -0.42121100425720215, 0.7570985555648804], "r_norm": [1.6550533771514893, 1.9254438877105713, 3.1527276039123535]}, {"step": 4000, "epoch": 8, "train_loss": 10.831953048706055, "cos_r_negg": [-0.020632298663258553, 0.050789255648851395, -0.15508905053138733], "cos_innovation_negg": [-0.07258114218711853, -0.5072916150093079, 0.5620717406272888], "cos_apical_negg": [-0.020632293075323105, 0.050789255648851395, -0.15508908033370972], "cos_Ac_negg": [-0.07259061932563782, -0.5072973966598511, 0.5620717406272888], "r_norm": [1.445892095565796, 1.6738638877868652, 2.6616740226745605], "ldr": {"dL_sdil": -2.0770139694213867, "dL_bp": -2.0770139694213867, "ratio": 1.0}}, {"step": 4100, "epoch": 8, "train_loss": 13.534512519836426, "cos_r_negg": [0.03284500539302826, 0.1665118932723999, -0.3395427167415619], "cos_innovation_negg": [-0.11990010738372803, -0.18010328710079193, 0.40169036388397217], "cos_apical_negg": [0.03284499794244766, 0.1665118932723999, -0.3395427167415619], "cos_Ac_negg": [-0.11991329491138458, -0.1801161766052246, 0.40169036388397217], "r_norm": [1.5067338943481445, 1.6453357934951782, 2.785231590270996]}, {"step": 4200, "epoch": 8, "train_loss": 10.042614936828613, "cos_r_negg": [-0.009762363508343697, -0.010503213852643967, -0.6587464213371277], "cos_innovation_negg": [-0.1768125295639038, -0.3322966694831848, 0.32998669147491455], "cos_apical_negg": [-0.009762364439666271, -0.010503217577934265, -0.6587464809417725], "cos_Ac_negg": [-0.1767541468143463, -0.33232879638671875, 0.32998669147491455], "r_norm": [1.6771917343139648, 1.9635087251663208, 2.9597654342651367]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.0958, "test_loss": 6.23742001953125}, {"step": 4300, "epoch": 9, "train_loss": 10.101000785827637, "cos_r_negg": [0.008505766279995441, -0.045377250760793686, -0.6276191473007202], "cos_innovation_negg": [-0.12647788226604462, -0.24161115288734436, 0.3345617949962616], "cos_apical_negg": [0.008505766279995441, -0.045377250760793686, -0.6276191473007202], "cos_Ac_negg": [-0.1264556646347046, -0.24159809947013855, 0.3345620632171631], "r_norm": [1.6475998163223267, 1.9239721298217773, 2.967597484588623]}, {"step": 4400, "epoch": 9, "train_loss": 9.596513748168945, "cos_r_negg": [0.008562065660953522, 0.15001876652240753, -0.175533264875412], "cos_innovation_negg": [-0.04514005407691002, -0.48557454347610474, 0.6526634693145752], "cos_apical_negg": [0.008562060073018074, 0.15001878142356873, -0.17553327977657318], "cos_Ac_negg": [-0.045139163732528687, -0.48565950989723206, 0.6526634693145752], "r_norm": [1.4307036399841309, 1.6210697889328003, 2.8097445964813232]}, {"step": 4500, "epoch": 9, "train_loss": 16.366668701171875, "cos_r_negg": [-0.024215523153543472, 0.0523746982216835, -0.13722029328346252], "cos_innovation_negg": [-0.06498889625072479, -0.1888183355331421, 0.29898589849472046], "cos_apical_negg": [-0.02421552501618862, 0.0523746982216835, -0.13722029328346252], "cos_Ac_negg": [-0.06498885899782181, -0.18893036246299744, 0.29898592829704285], "r_norm": [1.4138031005859375, 1.6285908222198486, 2.8013172149658203], "ldr": {"dL_sdil": -3.148632049560547, "dL_bp": -3.148632049560547, "ratio": 1.0}}, {"step": 4600, "epoch": 9, "train_loss": 15.175790786743164, "cos_r_negg": [-0.001095588319003582, 0.017703622579574585, -0.6027648448944092], "cos_innovation_negg": [0.01033240370452404, -0.28939884901046753, 0.6939722299575806], "cos_apical_negg": [-0.0010955953039228916, 0.017703616991639137, -0.6027647852897644], "cos_Ac_negg": [0.01495253574103117, -0.37095627188682556, 0.6939723491668701], "r_norm": [1.5095683336257935, 1.7129693031311035, 3.209284543991089]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.0982, "test_loss": 10.176494140625}, {"step": 4700, "epoch": 10, "train_loss": 12.548572540283203, "cos_r_negg": [0.03662903606891632, 0.013323778286576271, 0.03179753199219704], "cos_innovation_negg": [-0.11625862121582031, -0.22014150023460388, 0.5986127853393555], "cos_apical_negg": [0.03662903606891632, 0.013323778286576271, 0.03179751709103584], "cos_Ac_negg": [-0.11626827716827393, -0.22014975547790527, 0.5986127853393555], "r_norm": [1.3283519744873047, 1.469618320465088, 2.7055137157440186]}, {"step": 4800, "epoch": 10, "train_loss": 10.963335037231445, "cos_r_negg": [-0.021155945956707, 0.3539727032184601, 0.5038083791732788], "cos_innovation_negg": [-0.15960907936096191, 0.0907750278711319, 0.729169487953186], "cos_apical_negg": [-0.02115594409406185, 0.3539727032184601, 0.5038083791732788], "cos_Ac_negg": [-0.1595846712589264, 0.09067634493112564, 0.729169487953186], "r_norm": [1.4317306280136108, 1.7174842357635498, 3.339275360107422]}, {"step": 4900, "epoch": 10, "train_loss": 10.232675552368164, "cos_r_negg": [0.07411853969097137, -0.1702471375465393, -0.48597821593284607], "cos_innovation_negg": [0.06541065871715546, 0.2892952859401703, 0.9796814322471619], "cos_apical_negg": [0.07411853969097137, -0.1702471524477005, -0.4859781861305237], "cos_Ac_negg": [0.06539630889892578, 0.2896279990673065, 0.9796813726425171], "r_norm": [1.5119335651397705, 1.5873453617095947, 3.180260419845581]}, {"step": 5000, "epoch": 10, "train_loss": 13.096416473388672, "cos_r_negg": [0.03610732778906822, 0.15713843703269958, -0.291078120470047], "cos_innovation_negg": [0.0072157131507992744, -0.3843618631362915, 0.8875001072883606], "cos_apical_negg": [0.036107324063777924, 0.15713843703269958, -0.291078120470047], "cos_Ac_negg": [0.007210638374090195, -0.38435596227645874, 0.8875000476837158], "r_norm": [1.2840259075164795, 1.4135812520980835, 2.7617764472961426], "ldr": {"dL_sdil": -2.350564956665039, "dL_bp": -2.350564956665039, "ratio": 1.0}}, {"step": 5100, "epoch": 10, "train_loss": 10.827922821044922, "cos_r_negg": [0.024950947612524033, 0.2864372730255127, 0.02290518581867218], "cos_innovation_negg": [-0.08479246497154236, -0.31673842668533325, 0.4500734210014343], "cos_apical_negg": [0.024950947612524033, 0.2864372134208679, 0.022905167192220688], "cos_Ac_negg": [-0.08480361104011536, -0.31676894426345825, 0.4500734210014343], "r_norm": [1.2995893955230713, 1.452526569366455, 2.8131027221679688]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.098, "test_loss": 14.10288837890625}, {"step": 5200, "epoch": 11, "train_loss": 9.503578186035156, "cos_r_negg": [0.013169646263122559, 0.08294951915740967, -0.6244248151779175], "cos_innovation_negg": [-0.11085792630910873, -0.37744754552841187, 0.45521920919418335], "cos_apical_negg": [0.01316964440047741, 0.08294951915740967, -0.6244248151779175], "cos_Ac_negg": [-0.11086191236972809, -0.37721192836761475, 0.4552191197872162], "r_norm": [1.2747329473495483, 1.3942285776138306, 2.767704963684082]}, {"step": 5300, "epoch": 11, "train_loss": 5.430373191833496, "cos_r_negg": [0.003571514505892992, -0.16799697279930115, -0.4963197708129883], "cos_innovation_negg": [0.0434497594833374, 0.3517071604728699, 0.8774371147155762], "cos_apical_negg": [0.003571515902876854, -0.16799697279930115, -0.4963197708129883], "cos_Ac_negg": [0.04343333840370178, 0.35168662667274475, 0.8774371147155762], "r_norm": [1.3673832416534424, 1.4287078380584717, 3.0254065990448]}, {"step": 5400, "epoch": 11, "train_loss": 14.83660888671875, "cos_r_negg": [0.005606683902442455, -0.10506697744131088, -0.5611649751663208], "cos_innovation_negg": [-0.0940738320350647, -0.2691645324230194, 0.7712353467941284], "cos_apical_negg": [0.005606682039797306, -0.10506697744131088, -0.5611649751663208], "cos_Ac_negg": [-0.09403727948665619, -0.2690330445766449, 0.7712351679801941], "r_norm": [1.351412057876587, 1.4972279071807861, 3.23873233795166]}, {"step": 5500, "epoch": 11, "train_loss": 11.84660530090332, "cos_r_negg": [0.04608001559972763, 0.0542776994407177, -0.14153742790222168], "cos_innovation_negg": [-0.16755807399749756, -0.1653653383255005, 0.32534727454185486], "cos_apical_negg": [0.046080008149147034, 0.0542776957154274, -0.1415373980998993], "cos_Ac_negg": [-0.1675722301006317, -0.16536720097064972, 0.32534730434417725], "r_norm": [1.2696797847747803, 1.4968675374984741, 3.008450984954834], "ldr": {"dL_sdil": -1.9025554656982422, "dL_bp": -1.9025554656982422, "ratio": 1.0}}, {"step": 5600, "epoch": 11, "train_loss": 11.761016845703125, "cos_r_negg": [-0.01287061721086502, -0.06735193729400635, -0.0385841503739357], "cos_innovation_negg": [-0.18042925000190735, -0.16590286791324615, 0.28066006302833557], "cos_apical_negg": [-0.012870614416897297, -0.06735193729400635, -0.03858416527509689], "cos_Ac_negg": [-0.18043065071105957, -0.16587179899215698, 0.28066006302833557], "r_norm": [1.2144157886505127, 1.294379472732544, 2.8880460262298584]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.1135, "test_loss": 10.54988564453125}, {"step": 5700, "epoch": 12, "train_loss": 4.504752159118652, "cos_r_negg": [0.033178504556417465, -0.10059517621994019, 0.26030296087265015], "cos_innovation_negg": [-0.08326803147792816, -0.07829194515943527, 0.7997688055038452], "cos_apical_negg": [0.03317850083112717, -0.10059518367052078, 0.26030290126800537], "cos_Ac_negg": [-0.0832659900188446, -0.07830281555652618, 0.7997688055038452], "r_norm": [1.1017403602600098, 1.2252583503723145, 2.6190295219421387]}, {"step": 5800, "epoch": 12, "train_loss": 12.402713775634766, "cos_r_negg": [-0.02120404690504074, 0.1870843470096588, -0.366138219833374], "cos_innovation_negg": [-0.2049187272787094, -0.3321942687034607, 0.4586130380630493], "cos_apical_negg": [-0.021204041317105293, 0.18708436191082, -0.366138219833374], "cos_Ac_negg": [-0.20494437217712402, -0.3322075605392456, 0.4586130678653717], "r_norm": [1.277613639831543, 1.384798288345337, 3.035029172897339]}, {"step": 5900, "epoch": 12, "train_loss": 5.6544508934021, "cos_r_negg": [0.021466072648763657, 0.08863453567028046, -0.3640625774860382], "cos_innovation_negg": [-0.0702090784907341, -0.3271505832672119, 0.8775585889816284], "cos_apical_negg": [0.021466074511408806, 0.08863453567028046, -0.3640625476837158], "cos_Ac_negg": [-0.07023029029369354, -0.327148973941803, 0.8775585889816284], "r_norm": [1.154801845550537, 1.253584623336792, 2.8464016914367676]}, {"step": 6000, "epoch": 12, "train_loss": 12.991623878479004, "cos_r_negg": [-0.0341978445649147, 0.0753202959895134, 0.034765347838401794], "cos_innovation_negg": [-0.08287002146244049, -0.7208178043365479, 0.8287156224250793], "cos_apical_negg": [-0.034197840839624405, 0.0753202885389328, 0.03476536646485329], "cos_Ac_negg": [-0.0829041600227356, -0.7208141088485718, 0.8287156820297241], "r_norm": [1.3042879104614258, 1.389836072921753, 3.172173500061035], "ldr": {"dL_sdil": -2.888543128967285, "dL_bp": -2.888543128967285, "ratio": 1.0}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.1028, "test_loss": 13.79490341796875}, {"step": 6100, "epoch": 13, "train_loss": 15.776175498962402, "cos_r_negg": [0.049888595938682556, -0.15614530444145203, -0.27222248911857605], "cos_innovation_negg": [-0.1472887396812439, -0.36704009771347046, 0.538937509059906], "cos_apical_negg": [0.04988858848810196, -0.15614530444145203, -0.27222245931625366], "cos_Ac_negg": [-0.14728501439094543, -0.3670456111431122, 0.538937509059906], "r_norm": [1.1628806591033936, 1.1766693592071533, 2.8416147232055664]}, {"step": 6200, "epoch": 13, "train_loss": 16.496044158935547, "cos_r_negg": [0.014117803424596786, -0.3773772716522217, -0.5219522714614868], "cos_innovation_negg": [0.032156918197870255, 0.0008837850764393806, 0.2601567804813385], "cos_apical_negg": [0.014117800630629063, -0.3773772716522217, -0.5219523906707764], "cos_Ac_negg": [0.03220418840646744, 0.0008832784369587898, 0.2601567804813385], "r_norm": [1.2400844097137451, 1.2198686599731445, 3.034775733947754]}, {"step": 6300, "epoch": 13, "train_loss": 9.05245590209961, "cos_r_negg": [0.03672745078802109, -0.35884979367256165, -0.47191503643989563], "cos_innovation_negg": [-0.0891181007027626, 0.08326084911823273, 0.5798735618591309], "cos_apical_negg": [0.036727454513311386, -0.35884979367256165, -0.47191503643989563], "cos_Ac_negg": [-0.0891268253326416, 0.08326170593500137, 0.5798735618591309], "r_norm": [1.1116385459899902, 1.2412306070327759, 2.753157615661621]}, {"step": 6400, "epoch": 13, "train_loss": 6.07943868637085, "cos_r_negg": [0.011679553426802158, -0.17425653338432312, 0.03628295287489891], "cos_innovation_negg": [-0.0526568777859211, 0.059432484209537506, 0.6518957614898682], "cos_apical_negg": [0.011679550632834435, -0.17425653338432312, 0.036282941699028015], "cos_Ac_negg": [-0.05268612504005432, 0.05943384766578674, 0.6518958806991577], "r_norm": [1.1780421733856201, 1.2166333198547363, 2.965238332748413]}, {"step": 6500, "epoch": 13, "train_loss": 17.919506072998047, "cos_r_negg": [-0.020011212676763535, 0.067601777613163, -0.15226654708385468], "cos_innovation_negg": [-0.030089104548096657, -0.05048910528421402, 0.7527004480361938], "cos_apical_negg": [-0.020011218264698982, 0.0676017701625824, -0.1522665023803711], "cos_Ac_negg": [-0.03696989268064499, -0.05045740306377411, 0.7526998519897461], "r_norm": [1.249541997909546, 1.3109828233718872, 3.2425920963287354], "ldr": {"dL_sdil": -5.552864074707031, "dL_bp": -5.552864074707031, "ratio": 1.0}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.1135, "test_loss": 11.9120390625}, {"step": 6600, "epoch": 14, "train_loss": 17.50206756591797, "cos_r_negg": [0.03442477434873581, -0.32724928855895996, -0.31365150213241577], "cos_innovation_negg": [0.10971882194280624, -0.27853628993034363, 0.16275182366371155], "cos_apical_negg": [0.03442477434873581, -0.32724928855895996, -0.3136514723300934], "cos_Ac_negg": [0.10979901254177094, -0.27853628993034363, 0.16275030374526978], "r_norm": [1.269162654876709, 1.2095606327056885, 3.23329496383667]}, {"step": 6700, "epoch": 14, "train_loss": 10.686664581298828, "cos_r_negg": [0.03364798054099083, -0.3928048014640808, -0.22049809992313385], "cos_innovation_negg": [0.044101160019636154, 0.1364651769399643, 0.586864709854126], "cos_apical_negg": [0.033647987991571426, -0.3928048014640808, -0.22049811482429504], "cos_Ac_negg": [0.044156402349472046, 0.1364658772945404, 0.5868647694587708], "r_norm": [1.05025053024292, 1.0628093481063843, 2.7642130851745605]}, {"step": 6800, "epoch": 14, "train_loss": 17.495454788208008, "cos_r_negg": [0.07686907052993774, -0.3829383850097656, -0.4261513352394104], "cos_innovation_negg": [0.1924108862876892, 0.23426871001720428, 0.9896490573883057], "cos_apical_negg": [0.07686907052993774, -0.3829383850097656, -0.426151305437088], "cos_Ac_negg": [0.21469765901565552, 0.23420634865760803, 0.9896491765975952], "r_norm": [1.2391469478607178, 1.2442371845245361, 3.37134051322937]}, {"step": 6900, "epoch": 14, "train_loss": 19.883670806884766, "cos_r_negg": [0.0815495103597641, -0.3799760937690735, -0.4372183084487915], "cos_innovation_negg": [0.18566444516181946, 0.2155233919620514, 0.9861960411071777], "cos_apical_negg": [0.0815495103597641, -0.3799760937690735, -0.4372183680534363], "cos_Ac_negg": [0.1859016865491867, 0.21552357077598572, 0.9861960411071777], "r_norm": [1.2189700603485107, 1.2186787128448486, 3.3501100540161133]}, {"step": 7000, "epoch": 14, "train_loss": 12.419652938842773, "cos_r_negg": [0.015219427645206451, -0.2697301208972931, -0.19048017263412476], "cos_innovation_negg": [0.1253231018781662, 0.008636455982923508, 0.5253019332885742], "cos_apical_negg": [0.015219430439174175, -0.2697300910949707, -0.19048017263412476], "cos_Ac_negg": [0.12531673908233643, 0.00863754004240036, 0.5253018736839294], "r_norm": [1.1569595336914062, 1.1897273063659668, 3.1856937408447266], "ldr": {"dL_sdil": -2.832364082336426, "dL_bp": -2.832364082336426, "ratio": 1.0}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.098, "test_loss": 14.7233546875}], "final": {"test_acc": 0.098, "test_loss": 14.7233546875, "wall_s": 24.9082932472229, "cos_r_negg": [0.04708272963762283, -0.17973271012306213, 0.5334709882736206], "cos_innovation_negg": [-0.01665298268198967, 0.03750469163060188, 0.8107142448425293], "cos_apical_negg": [0.04708272963762283, -0.17973271012306213, 0.5334709882736206], "cos_Ac_negg": [-0.014648519456386566, 0.03750882297754288, 0.8107141256332397]}}