{"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": 0.2, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p2_raw_s0", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3603768348693848, "cos_r_negg": [-0.016473568975925446, -0.026327233761548996, 0.029903199523687363], "cos_innovation_negg": [-0.016580017283558846, -0.026751887053251266, 0.033074796199798584], "cos_apical_negg": [-0.016473568975925446, -0.026327233761548996, 0.029903199523687363], "cos_Ac_negg": [-0.01658003404736519, -0.026751989498734474, 0.03307466208934784], "r_norm": [5.579668998718262, 5.17504358291626, 5.082968711853027], "ldr": {"dL_sdil": -0.0021271705627441406, "dL_bp": -0.21677780151367188, "ratio": 0.009812677072518344}}, {"step": 100, "epoch": 0, "train_loss": 0.3388369083404541, "cos_r_negg": [0.1721920520067215, 0.2719985246658325, 0.6371455192565918], "cos_innovation_negg": [0.4169691801071167, 0.49493300914764404, 0.7356089949607849], "cos_apical_negg": [0.1721920520067215, 0.2719985246658325, 0.6371455192565918], "cos_Ac_negg": [0.4166533648967743, 0.49484872817993164, 0.7353556156158447], "r_norm": [3.8753223419189453, 3.8276116847991943, 3.784452438354492]}, {"step": 200, "epoch": 0, "train_loss": 0.3989391326904297, "cos_r_negg": [0.14291749894618988, 0.2217380702495575, 0.6675006151199341], "cos_innovation_negg": [0.44804346561431885, 0.5322386026382446, 0.7960344552993774], "cos_apical_negg": [0.14291749894618988, 0.2217380702495575, 0.6675006151199341], "cos_Ac_negg": [0.4471680223941803, 0.5317703485488892, 0.7958534955978394], "r_norm": [3.729452610015869, 3.7228593826293945, 3.7083444595336914]}, {"step": 300, "epoch": 0, "train_loss": 0.4919128119945526, "cos_r_negg": [0.14903941750526428, 0.3034444749355316, 0.6772742867469788], "cos_innovation_negg": [0.4774431586265564, 0.5283323526382446, 0.8345513343811035], "cos_apical_negg": [0.14903941750526428, 0.3034444749355316, 0.6772742867469788], "cos_Ac_negg": [0.4773305356502533, 0.5281742811203003, 0.8342525362968445], "r_norm": [3.6902332305908203, 3.693300485610962, 3.678870439529419]}, {"step": 400, "epoch": 0, "train_loss": 0.24398618936538696, "cos_r_negg": [0.11366908252239227, 0.21099181473255157, 0.6483669281005859], "cos_innovation_negg": [0.45127472281455994, 0.5458073616027832, 0.8159490823745728], "cos_apical_negg": [0.11366908252239227, 0.21099181473255157, 0.6483669281005859], "cos_Ac_negg": [0.4512435793876648, 0.5458665490150452, 0.815899670124054], "r_norm": [3.672726631164551, 3.6565113067626953, 3.6241979598999023]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9177, "test_loss": 0.3047351959228516}, {"step": 500, "epoch": 1, "train_loss": 0.32535645365715027, "cos_r_negg": [0.08938203752040863, 0.24729885160923004, 0.6637201309204102], "cos_innovation_negg": [0.42945319414138794, 0.5236873626708984, 0.8064945936203003], "cos_apical_negg": [0.08938203752040863, 0.24729885160923004, 0.6637201309204102], "cos_Ac_negg": [0.428948312997818, 0.5231894254684448, 0.8059635758399963], "r_norm": [3.6340513229370117, 3.626866102218628, 3.603440046310425], "ldr": {"dL_sdil": -0.02090245485305786, "dL_bp": -0.02135103940963745, "ratio": 0.9789900365985411}}, {"step": 600, "epoch": 1, "train_loss": 0.2030535787343979, "cos_r_negg": [0.11638487130403519, 0.18938882648944855, 0.6376956701278687], "cos_innovation_negg": [0.39706772565841675, 0.4187910258769989, 0.7993124723434448], "cos_apical_negg": [0.11638487130403519, 0.18938882648944855, 0.6376956701278687], "cos_Ac_negg": [0.3970112204551697, 0.41888427734375, 0.7987611889839172], "r_norm": [3.6063642501831055, 3.613574981689453, 3.5962107181549072]}, {"step": 700, "epoch": 1, "train_loss": 0.23081550002098083, "cos_r_negg": [0.10733923316001892, 0.20635592937469482, 0.6265333890914917], "cos_innovation_negg": [0.4361274838447571, 0.48120707273483276, 0.8315417766571045], "cos_apical_negg": [0.10733923316001892, 0.20635592937469482, 0.6265333890914917], "cos_Ac_negg": [0.4358212649822235, 0.4807732105255127, 0.8313336372375488], "r_norm": [3.590412139892578, 3.5937178134918213, 3.5744543075561523]}, {"step": 800, "epoch": 1, "train_loss": 0.1793380081653595, "cos_r_negg": [0.10001850128173828, 0.20235097408294678, 0.6001237630844116], "cos_innovation_negg": [0.4340983033180237, 0.5832151174545288, 0.7938954830169678], "cos_apical_negg": [0.10001850128173828, 0.20235097408294678, 0.6001237630844116], "cos_Ac_negg": [0.43403226137161255, 0.5829353928565979, 0.793729305267334], "r_norm": [3.5731160640716553, 3.5769083499908447, 3.5541131496429443]}, {"step": 900, "epoch": 1, "train_loss": 0.3070145547389984, "cos_r_negg": [0.08809584379196167, 0.2878977954387665, 0.5420974493026733], "cos_innovation_negg": [0.4349724054336548, 0.5595442652702332, 0.8187193274497986], "cos_apical_negg": [0.08809584379196167, 0.2878977954387665, 0.5420974493026733], "cos_Ac_negg": [0.43500420451164246, 0.559375524520874, 0.8182679414749146], "r_norm": [3.564882516860962, 3.5766515731811523, 3.5528924465179443]}, {"epoch_end": 1, "step": 938, "test_acc": 0.941, "test_loss": 0.2362076156616211}, {"step": 1000, "epoch": 2, "train_loss": 0.10118691623210907, "cos_r_negg": [0.10788147151470184, 0.2473321557044983, 0.5856626033782959], "cos_innovation_negg": [0.4178251624107361, 0.5001124143600464, 0.8181846737861633], "cos_apical_negg": [0.10788147151470184, 0.2473321557044983, 0.5856626033782959], "cos_Ac_negg": [0.4178936183452606, 0.5000171661376953, 0.8181976675987244], "r_norm": [3.5756051540374756, 3.576016664505005, 3.5485241413116455], "ldr": {"dL_sdil": -0.016332685947418213, "dL_bp": -0.015833169221878052, "ratio": 1.031548751771688}}, {"step": 1100, "epoch": 2, "train_loss": 0.34849005937576294, "cos_r_negg": [0.08034752309322357, 0.2810874581336975, 0.6226122379302979], "cos_innovation_negg": [0.4257814288139343, 0.5177045464515686, 0.8171130418777466], "cos_apical_negg": [0.08034752309322357, 0.2810874581336975, 0.6226122379302979], "cos_Ac_negg": [0.4257396459579468, 0.5172585248947144, 0.8169691562652588], "r_norm": [3.540311574935913, 3.5426650047302246, 3.5222136974334717]}, {"step": 1200, "epoch": 2, "train_loss": 0.2266823947429657, "cos_r_negg": [0.06467395275831223, 0.21835751831531525, 0.6337820291519165], "cos_innovation_negg": [0.44710350036621094, 0.5692906379699707, 0.793481707572937], "cos_apical_negg": [0.06467395275831223, 0.21835751831531525, 0.6337820291519165], "cos_Ac_negg": [0.44699984788894653, 0.569176435470581, 0.7932552099227905], "r_norm": [3.534238815307617, 3.537778615951538, 3.5188679695129395]}, {"step": 1300, "epoch": 2, "train_loss": 0.24251389503479004, "cos_r_negg": [0.07839789986610413, 0.2222185879945755, 0.6220861077308655], "cos_innovation_negg": [0.39006054401397705, 0.4479979872703552, 0.7673890590667725], "cos_apical_negg": [0.07839789986610413, 0.2222185879945755, 0.6220861077308655], "cos_Ac_negg": [0.3897953927516937, 0.4478038251399994, 0.766154408454895], "r_norm": [3.557648181915283, 3.5613315105438232, 3.5341551303863525]}, {"step": 1400, "epoch": 2, "train_loss": 0.2785467803478241, "cos_r_negg": [0.051284730434417725, 0.18493859469890594, 0.5170494318008423], "cos_innovation_negg": [0.46572354435920715, 0.5562729835510254, 0.7977768182754517], "cos_apical_negg": [0.051284730434417725, 0.18493859469890594, 0.5170494318008423], "cos_Ac_negg": [0.46563971042633057, 0.5561054944992065, 0.7977519035339355], "r_norm": [3.5240864753723145, 3.526305675506592, 3.508495330810547]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9465, "test_loss": 0.21352959671020508}, {"step": 1500, "epoch": 3, "train_loss": 0.18744012713432312, "cos_r_negg": [0.055619604885578156, 0.24060164391994476, 0.5720882415771484], "cos_innovation_negg": [0.4105251431465149, 0.5300585031509399, 0.7873919010162354], "cos_apical_negg": [0.055619604885578156, 0.24060164391994476, 0.5720882415771484], "cos_Ac_negg": [0.41038763523101807, 0.5297716856002808, 0.787345290184021], "r_norm": [3.528489589691162, 3.531886100769043, 3.507558822631836], "ldr": {"dL_sdil": -0.009793221950531006, "dL_bp": -0.010277211666107178, "ratio": 0.9529065147921101}}, {"step": 1600, "epoch": 3, "train_loss": 0.17836226522922516, "cos_r_negg": [0.06417367607355118, 0.2679503560066223, 0.6074199080467224], "cos_innovation_negg": [0.4860021471977234, 0.58686363697052, 0.8405827283859253], "cos_apical_negg": [0.06417367607355118, 0.2679503560066223, 0.6074199080467224], "cos_Ac_negg": [0.48599350452423096, 0.5863144397735596, 0.8405444025993347], "r_norm": [3.5293936729431152, 3.530726909637451, 3.510951280593872]}, {"step": 1700, "epoch": 3, "train_loss": 0.18673688173294067, "cos_r_negg": [0.10869573056697845, 0.25152212381362915, 0.6026225090026855], "cos_innovation_negg": [0.4714401960372925, 0.6265134811401367, 0.7920746803283691], "cos_apical_negg": [0.10869573056697845, 0.25152212381362915, 0.6026225090026855], "cos_Ac_negg": [0.4713318347930908, 0.626278817653656, 0.791936993598938], "r_norm": [3.511098623275757, 3.5152230262756348, 3.495692729949951]}, {"step": 1800, "epoch": 3, "train_loss": 0.055265218019485474, "cos_r_negg": [0.06471377611160278, 0.22349515557289124, 0.5369325876235962], "cos_innovation_negg": [0.3954787850379944, 0.5510791540145874, 0.7634647488594055], "cos_apical_negg": [0.06471377611160278, 0.22349515557289124, 0.5369325876235962], "cos_Ac_negg": [0.3954269289970398, 0.5504909753799438, 0.7636757493019104], "r_norm": [3.547030448913574, 3.549400568008423, 3.5224273204803467]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9502, "test_loss": 0.21128672790527345}, {"step": 1900, "epoch": 4, "train_loss": 0.12737774848937988, "cos_r_negg": [0.06768354028463364, 0.26480770111083984, 0.43547239899635315], "cos_innovation_negg": [0.3673062324523926, 0.5113232135772705, 0.7954840064048767], "cos_apical_negg": [0.06768354028463364, 0.26480770111083984, 0.43547239899635315], "cos_Ac_negg": [0.3674268126487732, 0.5118414759635925, 0.79532790184021], "r_norm": [3.495820999145508, 3.5033693313598633, 3.4865591526031494]}, {"step": 2000, "epoch": 4, "train_loss": 0.1944933980703354, "cos_r_negg": [0.04760921373963356, 0.19205546379089355, 0.6536286473274231], "cos_innovation_negg": [0.40347981452941895, 0.5031334161758423, 0.8259584903717041], "cos_apical_negg": [0.04760921373963356, 0.19205546379089355, 0.6536286473274231], "cos_Ac_negg": [0.40325045585632324, 0.5030980110168457, 0.825656533241272], "r_norm": [3.5149765014648438, 3.5217597484588623, 3.5062880516052246], "ldr": {"dL_sdil": -0.00924140214920044, "dL_bp": -0.013545364141464233, "ratio": 0.6822557188338133}}, {"step": 2100, "epoch": 4, "train_loss": 0.12359166890382767, "cos_r_negg": [0.09322690218687057, 0.21033260226249695, 0.598766565322876], "cos_innovation_negg": [0.4374707043170929, 0.544856071472168, 0.8425313234329224], "cos_apical_negg": [0.09322690218687057, 0.21033260226249695, 0.598766565322876], "cos_Ac_negg": [0.4371655583381653, 0.5446057319641113, 0.842496395111084], "r_norm": [3.509075164794922, 3.521883010864258, 3.499817371368408]}, {"step": 2200, "epoch": 4, "train_loss": 0.089406318962574, "cos_r_negg": [0.05137008801102638, 0.233104407787323, 0.5743218660354614], "cos_innovation_negg": [0.3753688931465149, 0.512610137462616, 0.8219261169433594], "cos_apical_negg": [0.05137008801102638, 0.233104407787323, 0.5743218660354614], "cos_Ac_negg": [0.3752927780151367, 0.5121101140975952, 0.8224152326583862], "r_norm": [3.5120458602905273, 3.51446270942688, 3.493922472000122]}, {"step": 2300, "epoch": 4, "train_loss": 0.27675482630729675, "cos_r_negg": [0.02435573935508728, 0.19956326484680176, 0.5856771469116211], "cos_innovation_negg": [0.4101802706718445, 0.5675158500671387, 0.7972996234893799], "cos_apical_negg": [0.02435573935508728, 0.19956326484680176, 0.5856771469116211], "cos_Ac_negg": [0.4101858139038086, 0.5674529075622559, 0.7969062924385071], "r_norm": [3.4907426834106445, 3.4916887283325195, 3.4779436588287354]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9517, "test_loss": 0.2051895065307617}, {"step": 2400, "epoch": 5, "train_loss": 0.21961703896522522, "cos_r_negg": [0.07271640747785568, 0.22103393077850342, 0.5571286678314209], "cos_innovation_negg": [0.42161825299263, 0.5564560294151306, 0.7148803472518921], "cos_apical_negg": [0.07271640747785568, 0.22103393077850342, 0.5571286678314209], "cos_Ac_negg": [0.42144575715065, 0.554915726184845, 0.7144851684570312], "r_norm": [3.495201587677002, 3.502061128616333, 3.486966609954834]}, {"step": 2500, "epoch": 5, "train_loss": 0.08197547495365143, "cos_r_negg": [0.06409181654453278, 0.23346653580665588, 0.5467738509178162], "cos_innovation_negg": [0.4133896231651306, 0.5755782723426819, 0.8129197955131531], "cos_apical_negg": [0.06409181654453278, 0.23346653580665588, 0.5467738509178162], "cos_Ac_negg": [0.41320452094078064, 0.5753757953643799, 0.8126294612884521], "r_norm": [3.4914605617523193, 3.4903128147125244, 3.4726197719573975], "ldr": {"dL_sdil": -0.008657664060592651, "dL_bp": -0.010447308421134949, "ratio": 0.8286980446636534}}, {"step": 2600, "epoch": 5, "train_loss": 0.1034022867679596, "cos_r_negg": [0.05353444442152977, 0.1933932900428772, 0.6284575462341309], "cos_innovation_negg": [0.34818780422210693, 0.5147819519042969, 0.8150479197502136], "cos_apical_negg": [0.05353444442152977, 0.1933932900428772, 0.6284575462341309], "cos_Ac_negg": [0.3481002449989319, 0.5146735906600952, 0.8149440288543701], "r_norm": [3.501924514770508, 3.4994311332702637, 3.478313446044922]}, {"step": 2700, "epoch": 5, "train_loss": 0.14125940203666687, "cos_r_negg": [0.08057860285043716, 0.22372595965862274, 0.5369303822517395], "cos_innovation_negg": [0.4004718065261841, 0.5381162762641907, 0.8511413335800171], "cos_apical_negg": [0.08057860285043716, 0.22372595965862274, 0.5369303822517395], "cos_Ac_negg": [0.4004393517971039, 0.5380550026893616, 0.851192831993103], "r_norm": [3.4954171180725098, 3.4984278678894043, 3.483983039855957]}, {"step": 2800, "epoch": 5, "train_loss": 0.21846488118171692, "cos_r_negg": [0.06009357422590256, 0.24833229184150696, 0.5907409191131592], "cos_innovation_negg": [0.41767460107803345, 0.5634260773658752, 0.7554038763046265], "cos_apical_negg": [0.06009357422590256, 0.24833229184150696, 0.5907409191131592], "cos_Ac_negg": [0.4173479378223419, 0.5631324052810669, 0.7553071975708008], "r_norm": [3.4656124114990234, 3.469135046005249, 3.4514381885528564]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9521, "test_loss": 0.20345081100463866}, {"step": 2900, "epoch": 6, "train_loss": 0.10953014343976974, "cos_r_negg": [0.06787798553705215, 0.2287384271621704, 0.468828022480011], "cos_innovation_negg": [0.3793896436691284, 0.4630575478076935, 0.718155026435852], "cos_apical_negg": [0.06787798553705215, 0.2287384271621704, 0.468828022480011], "cos_Ac_negg": [0.37901097536087036, 0.46289223432540894, 0.7178455591201782], "r_norm": [3.484696388244629, 3.4866385459899902, 3.4722611904144287]}, {"step": 3000, "epoch": 6, "train_loss": 0.46337395906448364, "cos_r_negg": [0.06830329447984695, 0.1616249680519104, 0.3647420108318329], "cos_innovation_negg": [0.35110774636268616, 0.5082316994667053, 0.6819912195205688], "cos_apical_negg": [0.06830329447984695, 0.1616249680519104, 0.3647420108318329], "cos_Ac_negg": [0.35069969296455383, 0.5088880658149719, 0.6835923194885254], "r_norm": [3.469820499420166, 3.4758381843566895, 3.461033344268799], "ldr": {"dL_sdil": -0.004351168870925903, "dL_bp": -0.005708053708076477, "ratio": 0.7622859022453343}}, {"step": 3100, "epoch": 6, "train_loss": 0.1685410588979721, "cos_r_negg": [0.04595254734158516, 0.20454484224319458, 0.5188210606575012], "cos_innovation_negg": [0.3580610752105713, 0.47345858812332153, 0.728188157081604], "cos_apical_negg": [0.04595254734158516, 0.20454484224319458, 0.5188210606575012], "cos_Ac_negg": [0.3577389121055603, 0.47334158420562744, 0.7288005352020264], "r_norm": [3.477811813354492, 3.478243350982666, 3.4599335193634033]}, {"step": 3200, "epoch": 6, "train_loss": 0.17329390347003937, "cos_r_negg": [0.09172523021697998, 0.22572630643844604, 0.6188867092132568], "cos_innovation_negg": [0.40740275382995605, 0.5726735591888428, 0.8382785320281982], "cos_apical_negg": [0.09172523021697998, 0.22572630643844604, 0.6188867092132568], "cos_Ac_negg": [0.40736135840415955, 0.5720812678337097, 0.8378410935401917], "r_norm": [3.483748435974121, 3.4868130683898926, 3.4683408737182617]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9531, "test_loss": 0.19152970275878906}, {"step": 3300, "epoch": 7, "train_loss": 0.12548884749412537, "cos_r_negg": [0.03371473774313927, 0.19116787612438202, 0.5536788702011108], "cos_innovation_negg": [0.347528338432312, 0.5024522542953491, 0.8131461143493652], "cos_apical_negg": [0.03371473774313927, 0.19116787612438202, 0.5536788702011108], "cos_Ac_negg": [0.3474498987197876, 0.5019901990890503, 0.8122908473014832], "r_norm": [3.476494789123535, 3.4794178009033203, 3.4578332901000977]}, {"step": 3400, "epoch": 7, "train_loss": 0.2698051631450653, "cos_r_negg": [0.06270653009414673, 0.22756868600845337, 0.592083215713501], "cos_innovation_negg": [0.4125613570213318, 0.5473350882530212, 0.784447431564331], "cos_apical_negg": [0.06270653009414673, 0.22756868600845337, 0.592083215713501], "cos_Ac_negg": [0.4124089479446411, 0.5468129515647888, 0.784945011138916], "r_norm": [3.473717212677002, 3.4742119312286377, 3.45658278465271]}, {"step": 3500, "epoch": 7, "train_loss": 0.11865478754043579, "cos_r_negg": [0.033839091658592224, 0.1918744444847107, 0.40691661834716797], "cos_innovation_negg": [0.3008725643157959, 0.43055564165115356, 0.6733378171920776], "cos_apical_negg": [0.033839091658592224, 0.1918744444847107, 0.40691661834716797], "cos_Ac_negg": [0.3010721802711487, 0.4305219352245331, 0.6731608510017395], "r_norm": [3.475013256072998, 3.4807357788085938, 3.4610140323638916], "ldr": {"dL_sdil": -0.005170732736587524, "dL_bp": -0.007225096225738525, "ratio": 0.7156628195715122}}, {"step": 3600, "epoch": 7, "train_loss": 0.08846645057201385, "cos_r_negg": [0.07484865933656693, 0.20826342701911926, 0.42368367314338684], "cos_innovation_negg": [0.39259037375450134, 0.5593224763870239, 0.7920131087303162], "cos_apical_negg": [0.07484865933656693, 0.20826342701911926, 0.42368367314338684], "cos_Ac_negg": [0.3920590877532959, 0.5587590932846069, 0.7919071316719055], "r_norm": [3.4648241996765137, 3.469045639038086, 3.4543657302856445]}, {"step": 3700, "epoch": 7, "train_loss": 0.10054798424243927, "cos_r_negg": [0.0349927693605423, 0.18674367666244507, 0.3883827030658722], "cos_innovation_negg": [0.3381352424621582, 0.4853806793689728, 0.7522209882736206], "cos_apical_negg": [0.0349927693605423, 0.18674367666244507, 0.3883827030658722], "cos_Ac_negg": [0.3382090926170349, 0.4852631092071533, 0.7521555423736572], "r_norm": [3.4578139781951904, 3.464902400970459, 3.449568510055542]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9514, "test_loss": 0.19811919784545898}, {"step": 3800, "epoch": 8, "train_loss": 0.14756694436073303, "cos_r_negg": [0.0530828982591629, 0.16706420481204987, 0.46107226610183716], "cos_innovation_negg": [0.31439846754074097, 0.48531389236450195, 0.7160123586654663], "cos_apical_negg": [0.0530828982591629, 0.16706420481204987, 0.46107226610183716], "cos_Ac_negg": [0.3142688572406769, 0.48524773120880127, 0.7158101797103882], "r_norm": [3.449390411376953, 3.4537947177886963, 3.4377968311309814]}, {"step": 3900, "epoch": 8, "train_loss": 0.21295592188835144, "cos_r_negg": [0.04609540104866028, 0.1801377236843109, 0.5433958172798157], "cos_innovation_negg": [0.38611119985580444, 0.516525149345398, 0.8119463324546814], "cos_apical_negg": [0.04609540104866028, 0.1801377236843109, 0.5433958172798157], "cos_Ac_negg": [0.38675862550735474, 0.5163048505783081, 0.8117234706878662], "r_norm": [3.4833412170410156, 3.4833476543426514, 3.465080738067627]}, {"step": 4000, "epoch": 8, "train_loss": 0.18036073446273804, "cos_r_negg": [0.05877310782670975, 0.17528653144836426, 0.5795111656188965], "cos_innovation_negg": [0.3740137219429016, 0.5436674952507019, 0.8108839988708496], "cos_apical_negg": [0.05877310782670975, 0.17528653144836426, 0.5795111656188965], "cos_Ac_negg": [0.3739895224571228, 0.5435382723808289, 0.8106233477592468], "r_norm": [3.4689111709594727, 3.4703192710876465, 3.451960563659668], "ldr": {"dL_sdil": -0.007080137729644775, "dL_bp": -0.010071009397506714, "ratio": 0.7030216585238824}}, {"step": 4100, "epoch": 8, "train_loss": 0.0936855748295784, "cos_r_negg": [0.04878559708595276, 0.14551624655723572, 0.41049301624298096], "cos_innovation_negg": [0.31169065833091736, 0.47075045108795166, 0.7122496366500854], "cos_apical_negg": [0.04878559708595276, 0.14551624655723572, 0.41049301624298096], "cos_Ac_negg": [0.31158238649368286, 0.47061651945114136, 0.713679313659668], "r_norm": [3.465526580810547, 3.4687132835388184, 3.4495201110839844]}, {"step": 4200, "epoch": 8, "train_loss": 0.23818635940551758, "cos_r_negg": [0.016306698322296143, 0.21776893734931946, 0.4163423180580139], "cos_innovation_negg": [0.3794938623905182, 0.49791502952575684, 0.7926942110061646], "cos_apical_negg": [0.016306698322296143, 0.21776893734931946, 0.4163423180580139], "cos_Ac_negg": [0.37872251868247986, 0.4976539611816406, 0.7927689552307129], "r_norm": [3.458850145339966, 3.460315704345703, 3.4439399242401123]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.955, "test_loss": 0.18651178131103516}, {"step": 4300, "epoch": 9, "train_loss": 0.08469755947589874, "cos_r_negg": [0.022022685036063194, 0.22184401750564575, 0.5056993961334229], "cos_innovation_negg": [0.35894662141799927, 0.5069069862365723, 0.8342896699905396], "cos_apical_negg": [0.022022685036063194, 0.22184401750564575, 0.5056993961334229], "cos_Ac_negg": [0.35921746492385864, 0.5069506764411926, 0.8342188596725464], "r_norm": [3.465068817138672, 3.4675374031066895, 3.448363780975342]}, {"step": 4400, "epoch": 9, "train_loss": 0.08441644161939621, "cos_r_negg": [0.028903231024742126, 0.11586825549602509, 0.43870049715042114], "cos_innovation_negg": [0.34993427991867065, 0.44487541913986206, 0.7806928157806396], "cos_apical_negg": [0.028903231024742126, 0.11586825549602509, 0.43870049715042114], "cos_Ac_negg": [0.34975385665893555, 0.44477254152297974, 0.7801551818847656], "r_norm": [3.4610204696655273, 3.4679622650146484, 3.44842529296875]}, {"step": 4500, "epoch": 9, "train_loss": 0.23763245344161987, "cos_r_negg": [0.04483743757009506, 0.16346581280231476, 0.538713812828064], "cos_innovation_negg": [0.3578297197818756, 0.5398879051208496, 0.7825108766555786], "cos_apical_negg": [0.04483743757009506, 0.16346581280231476, 0.538713812828064], "cos_Ac_negg": [0.3577312231063843, 0.5398834943771362, 0.782530665397644], "r_norm": [3.4615845680236816, 3.4629898071289062, 3.4443883895874023], "ldr": {"dL_sdil": -0.005442783236503601, "dL_bp": -0.008815303444862366, "ratio": 0.6174243768858236}}, {"step": 4600, "epoch": 9, "train_loss": 0.13584284484386444, "cos_r_negg": [0.04618809372186661, 0.12648281455039978, 0.5143966674804688], "cos_innovation_negg": [0.35457372665405273, 0.488802433013916, 0.8005614280700684], "cos_apical_negg": [0.04618809372186661, 0.12648281455039978, 0.5143966674804688], "cos_Ac_negg": [0.3547212481498718, 0.4885520339012146, 0.8008222579956055], "r_norm": [3.471250534057617, 3.472142219543457, 3.4547955989837646]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9567, "test_loss": 0.19370537109375}, {"step": 4700, "epoch": 10, "train_loss": 0.12388288974761963, "cos_r_negg": [0.04427836090326309, 0.15790385007858276, 0.5705488324165344], "cos_innovation_negg": [0.40977948904037476, 0.5688140392303467, 0.7962357401847839], "cos_apical_negg": [0.04427836090326309, 0.15790385007858276, 0.5705488324165344], "cos_Ac_negg": [0.40972256660461426, 0.5687934160232544, 0.7956339120864868], "r_norm": [3.4553651809692383, 3.461555004119873, 3.4469351768493652]}, {"step": 4800, "epoch": 10, "train_loss": 0.0798642635345459, "cos_r_negg": [0.014005042612552643, 0.1993107795715332, 0.37496769428253174], "cos_innovation_negg": [0.30328428745269775, 0.4553671181201935, 0.7214771509170532], "cos_apical_negg": [0.014005042612552643, 0.1993107795715332, 0.37496769428253174], "cos_Ac_negg": [0.30473095178604126, 0.45609545707702637, 0.7215150594711304], "r_norm": [3.4687094688415527, 3.4758286476135254, 3.4603168964385986]}, {"step": 4900, "epoch": 10, "train_loss": 0.11851084232330322, "cos_r_negg": [0.05125885456800461, 0.18183550238609314, 0.5410358309745789], "cos_innovation_negg": [0.4377947449684143, 0.5575900077819824, 0.783425509929657], "cos_apical_negg": [0.05125885456800461, 0.18183550238609314, 0.5410358309745789], "cos_Ac_negg": [0.4377279281616211, 0.5575608015060425, 0.7831692695617676], "r_norm": [3.451565742492676, 3.453493118286133, 3.4385886192321777]}, {"step": 5000, "epoch": 10, "train_loss": 0.03343769535422325, "cos_r_negg": [0.046992167830467224, 0.14809100329875946, 0.40192416310310364], "cos_innovation_negg": [0.35171037912368774, 0.46964967250823975, 0.7679435014724731], "cos_apical_negg": [0.046992167830467224, 0.14809100329875946, 0.40192416310310364], "cos_Ac_negg": [0.3516887128353119, 0.46960732340812683, 0.7675899267196655], "r_norm": [3.464869976043701, 3.468376874923706, 3.4489669799804688], "ldr": {"dL_sdil": -0.0051101744174957275, "dL_bp": -0.0077325403690338135, "ratio": 0.6608661802737212}}, {"step": 5100, "epoch": 10, "train_loss": 0.25048214197158813, "cos_r_negg": [0.02246563509106636, 0.19471031427383423, 0.32953083515167236], "cos_innovation_negg": [0.3488203287124634, 0.5030003190040588, 0.7331562042236328], "cos_apical_negg": [0.02246563509106636, 0.19471031427383423, 0.32953083515167236], "cos_Ac_negg": [0.348846435546875, 0.5025877952575684, 0.7333074808120728], "r_norm": [3.4506001472473145, 3.4524917602539062, 3.4326670169830322]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9553, "test_loss": 0.1973849281311035}, {"step": 5200, "epoch": 11, "train_loss": 0.029526151716709137, "cos_r_negg": [0.06901518255472183, 0.1713683009147644, 0.4000300168991089], "cos_innovation_negg": [0.4098756015300751, 0.48163580894470215, 0.7498794794082642], "cos_apical_negg": [0.06901518255472183, 0.1713683009147644, 0.4000300168991089], "cos_Ac_negg": [0.41022875905036926, 0.4820902645587921, 0.7499216198921204], "r_norm": [3.441467761993408, 3.444317579269409, 3.4281182289123535]}, {"step": 5300, "epoch": 11, "train_loss": 0.14333248138427734, "cos_r_negg": [0.08245712518692017, 0.18008121848106384, 0.4830939471721649], "cos_innovation_negg": [0.3427019715309143, 0.4861484467983246, 0.8593589067459106], "cos_apical_negg": [0.08245712518692017, 0.18008121848106384, 0.4830939471721649], "cos_Ac_negg": [0.34280669689178467, 0.4862815737724304, 0.8594378232955933], "r_norm": [3.440645694732666, 3.4473037719726562, 3.4331471920013428]}, {"step": 5400, "epoch": 11, "train_loss": 0.25774702429771423, "cos_r_negg": [0.06123093515634537, 0.16761237382888794, 0.5060168504714966], "cos_innovation_negg": [0.3641701340675354, 0.5245863199234009, 0.7539292573928833], "cos_apical_negg": [0.06123093515634537, 0.16761237382888794, 0.5060168504714966], "cos_Ac_negg": [0.36482760310173035, 0.5246273279190063, 0.7533466219902039], "r_norm": [3.471930980682373, 3.475895881652832, 3.4576027393341064]}, {"step": 5500, "epoch": 11, "train_loss": 0.20137831568717957, "cos_r_negg": [0.014790154062211514, 0.2128017097711563, 0.46241673827171326], "cos_innovation_negg": [0.3637382388114929, 0.4812740087509155, 0.7801693677902222], "cos_apical_negg": [0.014790154062211514, 0.2128017097711563, 0.46241673827171326], "cos_Ac_negg": [0.3644803762435913, 0.4828115701675415, 0.7800079584121704], "r_norm": [3.447646141052246, 3.4518251419067383, 3.4343323707580566], "ldr": {"dL_sdil": -0.004327937960624695, "dL_bp": -0.005821391940116882, "ratio": 0.7434541438104575}}, {"step": 5600, "epoch": 11, "train_loss": 0.08066187798976898, "cos_r_negg": [0.03613896667957306, 0.19648058712482452, 0.49575281143188477], "cos_innovation_negg": [0.3911544680595398, 0.5481551885604858, 0.8146194219589233], "cos_apical_negg": [0.03613896667957306, 0.19648058712482452, 0.49575281143188477], "cos_Ac_negg": [0.3913358449935913, 0.5480250120162964, 0.8147847652435303], "r_norm": [3.458461284637451, 3.4564781188964844, 3.438340425491333]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9547, "test_loss": 0.20010665740966796}, {"step": 5700, "epoch": 12, "train_loss": 0.10322320461273193, "cos_r_negg": [0.027585340663790703, 0.1562771499156952, 0.42901039123535156], "cos_innovation_negg": [0.3205551505088806, 0.47852081060409546, 0.7474812865257263], "cos_apical_negg": [0.027585340663790703, 0.1562771499156952, 0.42901039123535156], "cos_Ac_negg": [0.3221622705459595, 0.4781225025653839, 0.7490335702896118], "r_norm": [3.4514660835266113, 3.4558229446411133, 3.437520742416382]}, {"step": 5800, "epoch": 12, "train_loss": 0.09904471784830093, "cos_r_negg": [-0.002126248786225915, 0.15922358632087708, 0.377025842666626], "cos_innovation_negg": [0.27672380208969116, 0.44702157378196716, 0.8337773084640503], "cos_apical_negg": [-0.002126248786225915, 0.15922358632087708, 0.377025842666626], "cos_Ac_negg": [0.2766464352607727, 0.44679200649261475, 0.8340064287185669], "r_norm": [3.4628639221191406, 3.4620978832244873, 3.4439620971679688]}, {"step": 5900, "epoch": 12, "train_loss": 0.07734407484531403, "cos_r_negg": [0.014376225881278515, 0.16064172983169556, 0.3750764727592468], "cos_innovation_negg": [0.32075750827789307, 0.4464177191257477, 0.7651004791259766], "cos_apical_negg": [0.014376225881278515, 0.16064172983169556, 0.3750764727592468], "cos_Ac_negg": [0.3215935528278351, 0.44536978006362915, 0.7645524740219116], "r_norm": [3.461233139038086, 3.4652247428894043, 3.447413444519043]}, {"step": 6000, "epoch": 12, "train_loss": 0.024701859802007675, "cos_r_negg": [0.010952739045023918, 0.13153177499771118, 0.43485701084136963], "cos_innovation_negg": [0.3408902883529663, 0.38688892126083374, 0.7951912879943848], "cos_apical_negg": [0.010952739045023918, 0.13153177499771118, 0.43485701084136963], "cos_Ac_negg": [0.3406761884689331, 0.3871164917945862, 0.795509934425354], "r_norm": [3.4502010345458984, 3.4490461349487305, 3.4353888034820557], "ldr": {"dL_sdil": -0.00697992742061615, "dL_bp": -0.013710647821426392, "ratio": 0.5090880833295294}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9567, "test_loss": 0.18746873474121092}, {"step": 6100, "epoch": 13, "train_loss": 0.05140694975852966, "cos_r_negg": [0.0009655402973294258, 0.1851300299167633, 0.39914822578430176], "cos_innovation_negg": [0.3021148145198822, 0.4084289073944092, 0.7972220182418823], "cos_apical_negg": [0.0009655402973294258, 0.1851300299167633, 0.39914822578430176], "cos_Ac_negg": [0.3022276759147644, 0.40826839208602905, 0.7963982820510864], "r_norm": [3.453662395477295, 3.456664800643921, 3.4398272037506104]}, {"step": 6200, "epoch": 13, "train_loss": 0.024884842336177826, "cos_r_negg": [0.02299400046467781, 0.14123308658599854, 0.49263402819633484], "cos_innovation_negg": [0.4263499975204468, 0.5489314794540405, 0.7903797626495361], "cos_apical_negg": [0.02299400046467781, 0.14123308658599854, 0.49263402819633484], "cos_Ac_negg": [0.42628806829452515, 0.5488638877868652, 0.7889189720153809], "r_norm": [3.4657716751098633, 3.463202476501465, 3.4446065425872803]}, {"step": 6300, "epoch": 13, "train_loss": 0.05869191884994507, "cos_r_negg": [0.01883682608604431, 0.14145004749298096, 0.5869549512863159], "cos_innovation_negg": [0.3940189480781555, 0.5031102895736694, 0.790341854095459], "cos_apical_negg": [0.01883682608604431, 0.14145004749298096, 0.5869549512863159], "cos_Ac_negg": [0.39499062299728394, 0.5030839443206787, 0.7890118956565857], "r_norm": [3.4605984687805176, 3.462883234024048, 3.4468939304351807]}, {"step": 6400, "epoch": 13, "train_loss": 0.06937310844659805, "cos_r_negg": [0.02568833716213703, 0.1709422469139099, 0.507933497428894], "cos_innovation_negg": [0.433333158493042, 0.526401162147522, 0.7905400991439819], "cos_apical_negg": [0.02568833716213703, 0.1709422469139099, 0.507933497428894], "cos_Ac_negg": [0.43355828523635864, 0.5258115530014038, 0.79041588306427], "r_norm": [3.451402187347412, 3.4500906467437744, 3.4321341514587402]}, {"step": 6500, "epoch": 13, "train_loss": 0.023130763322114944, "cos_r_negg": [0.034315600991249084, 0.18895575404167175, 0.49127936363220215], "cos_innovation_negg": [0.385489821434021, 0.5866650342941284, 0.7990735173225403], "cos_apical_negg": [0.034315600991249084, 0.18895575404167175, 0.49127936363220215], "cos_Ac_negg": [0.3861708641052246, 0.5882511138916016, 0.7986502647399902], "r_norm": [3.461559295654297, 3.4570980072021484, 3.4403011798858643], "ldr": {"dL_sdil": -0.010107576847076416, "dL_bp": -0.013823136687278748, "ratio": 0.731207183705149}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9585, "test_loss": 0.19585581665039062}, {"step": 6600, "epoch": 14, "train_loss": 0.04518500342965126, "cos_r_negg": [0.03899315744638443, 0.18686360120773315, 0.3514447510242462], "cos_innovation_negg": [0.3174074590206146, 0.4429108500480652, 0.7336504459381104], "cos_apical_negg": [0.03899315744638443, 0.18686360120773315, 0.3514447510242462], "cos_Ac_negg": [0.3177051246166229, 0.44369471073150635, 0.7339938879013062], "r_norm": [3.4517478942871094, 3.4519762992858887, 3.4337713718414307]}, {"step": 6700, "epoch": 14, "train_loss": 0.11906260251998901, "cos_r_negg": [0.002237762324512005, 0.21442167460918427, 0.47415995597839355], "cos_innovation_negg": [0.3493881821632385, 0.4919668734073639, 0.7755817174911499], "cos_apical_negg": [0.002237762324512005, 0.21442167460918427, 0.47415995597839355], "cos_Ac_negg": [0.36083731055259705, 0.49203193187713623, 0.7772910594940186], "r_norm": [3.441812515258789, 3.4441897869110107, 3.4262266159057617]}, {"step": 6800, "epoch": 14, "train_loss": 0.18277990818023682, "cos_r_negg": [0.040904827415943146, 0.15774965286254883, 0.5693460702896118], "cos_innovation_negg": [0.38608092069625854, 0.5170440077781677, 0.8491572141647339], "cos_apical_negg": [0.040904827415943146, 0.15774965286254883, 0.5693460702896118], "cos_Ac_negg": [0.3857008218765259, 0.5173830389976501, 0.8529412150382996], "r_norm": [3.4625468254089355, 3.461876392364502, 3.4417924880981445]}, {"step": 6900, "epoch": 14, "train_loss": 0.054661303758621216, "cos_r_negg": [0.014178114011883736, 0.15414831042289734, 0.4815225601196289], "cos_innovation_negg": [0.4010230004787445, 0.5195152759552002, 0.7063506841659546], "cos_apical_negg": [0.014178114011883736, 0.15414831042289734, 0.4815225601196289], "cos_Ac_negg": [0.4010884165763855, 0.5229536294937134, 0.7053012847900391], "r_norm": [3.4492599964141846, 3.450108289718628, 3.432631492614746]}, {"step": 7000, "epoch": 14, "train_loss": 0.12123630195856094, "cos_r_negg": [0.02383439615368843, 0.19232985377311707, 0.5521015524864197], "cos_innovation_negg": [0.31642913818359375, 0.5420209169387817, 0.8157789707183838], "cos_apical_negg": [0.02383439615368843, 0.19232985377311707, 0.5521015524864197], "cos_Ac_negg": [0.31504759192466736, 0.5415788292884827, 0.8173043727874756], "r_norm": [3.4482507705688477, 3.452746868133545, 3.437047004699707], "ldr": {"dL_sdil": -0.0059343427419662476, "dL_bp": -0.01396980881690979, "ratio": 0.42479770623510665}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9556, "test_loss": 0.20303017807006835}], "final": {"test_acc": 0.9556, "test_loss": 0.20303017807006835, "wall_s": 22.80018925666809, "cos_r_negg": [0.0006068069487810135, 0.14546656608581543, 0.5014218091964722], "cos_innovation_negg": [0.3084047734737396, 0.5018429756164551, 0.8085402250289917], "cos_apical_negg": [0.0006068069487810135, 0.14546656608581543, 0.5014218091964722], "cos_Ac_negg": [0.30887991189956665, 0.5017359852790833, 0.807351291179657]}}