summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p05_matched_s3.json
blob: d2533dd6986980829be861c8eacdfb497634a406 (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": 0.05, "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_rho0p05_matched_s3", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.331878900527954, "cos_r_negg": [-0.015490960329771042, 0.01314460951834917, 0.005123969167470932], "cos_innovation_negg": [-0.014851665124297142, 0.013897672295570374, 0.00600943760946393], "cos_apical_negg": [-0.015490962192416191, 0.013144610449671745, 0.005123969167470932], "cos_Ac_negg": [-0.014851845800876617, 0.013897630386054516, 0.006009321194142103], "r_norm": [4.857795238494873, 4.751279354095459, 4.85173225402832], "ldr": {"dL_sdil": -0.0007169246673583984, "dL_bp": -0.22743678092956543, "ratio": 0.003152193169584219}}, {"step": 100, "epoch": 0, "train_loss": 0.33210813999176025, "cos_r_negg": [0.2446061372756958, 0.24338620901107788, 0.22636139392852783], "cos_innovation_negg": [0.48129013180732727, 0.44585877656936646, 0.3610212504863739], "cos_apical_negg": [0.2446061372756958, 0.24338620901107788, 0.22636139392852783], "cos_Ac_negg": [0.4819962978363037, 0.44557255506515503, 0.36056792736053467], "r_norm": [0.9718714952468872, 0.9574735164642334, 0.9659876227378845]}, {"step": 200, "epoch": 0, "train_loss": 0.35869932174682617, "cos_r_negg": [0.23496845364570618, 0.2489742934703827, 0.2390528917312622], "cos_innovation_negg": [0.5499919652938843, 0.5420595407485962, 0.4637102484703064], "cos_apical_negg": [0.23496845364570618, 0.24897430837154388, 0.2390528917312622], "cos_Ac_negg": [0.5499877333641052, 0.5419607162475586, 0.46368440985679626], "r_norm": [0.7770188450813293, 0.76595139503479, 0.7725645899772644]}, {"step": 300, "epoch": 0, "train_loss": 0.25705695152282715, "cos_r_negg": [0.2327309250831604, 0.23788033425807953, 0.25863051414489746], "cos_innovation_negg": [0.5484759211540222, 0.5564261674880981, 0.5521402359008789], "cos_apical_negg": [0.2327309250831604, 0.23788033425807953, 0.25863051414489746], "cos_Ac_negg": [0.5482003092765808, 0.5562231540679932, 0.5521334409713745], "r_norm": [0.6918046474456787, 0.6813380718231201, 0.6911088228225708]}, {"step": 400, "epoch": 0, "train_loss": 0.17998269200325012, "cos_r_negg": [0.2094043791294098, 0.2362065464258194, 0.2659427523612976], "cos_innovation_negg": [0.536241352558136, 0.5778297781944275, 0.5647197365760803], "cos_apical_negg": [0.2094043791294098, 0.2362065464258194, 0.2659427523612976], "cos_Ac_negg": [0.5361621379852295, 0.5775743722915649, 0.5645044445991516], "r_norm": [0.679763913154602, 0.6754298210144043, 0.6804885864257812]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9355, "test_loss": 0.21739756927490234}, {"step": 500, "epoch": 1, "train_loss": 0.2138676792383194, "cos_r_negg": [0.19318073987960815, 0.21462637186050415, 0.24055153131484985], "cos_innovation_negg": [0.5731021165847778, 0.5893297791481018, 0.6107022762298584], "cos_apical_negg": [0.19318073987960815, 0.21462638676166534, 0.24055154621601105], "cos_Ac_negg": [0.5742365121841431, 0.5893971920013428, 0.609838604927063], "r_norm": [0.5675994157791138, 0.561404287815094, 0.5678472518920898], "ldr": {"dL_sdil": -0.01185181736946106, "dL_bp": -0.010675176978111267, "ratio": 1.1102220969040995}}, {"step": 600, "epoch": 1, "train_loss": 0.09671124070882797, "cos_r_negg": [0.16785210371017456, 0.19913065433502197, 0.22777554392814636], "cos_innovation_negg": [0.568167507648468, 0.5897747278213501, 0.6196615695953369], "cos_apical_negg": [0.16785210371017456, 0.19913065433502197, 0.22777554392814636], "cos_Ac_negg": [0.5680513381958008, 0.5894980430603027, 0.6193525791168213], "r_norm": [0.45053526759147644, 0.4490148723125458, 0.4517337679862976]}, {"step": 700, "epoch": 1, "train_loss": 0.25853413343429565, "cos_r_negg": [0.1954101324081421, 0.2096310257911682, 0.2290453165769577], "cos_innovation_negg": [0.5455281734466553, 0.5825574398040771, 0.6364662647247314], "cos_apical_negg": [0.1954101324081421, 0.2096310257911682, 0.2290453165769577], "cos_Ac_negg": [0.5453530550003052, 0.5823148488998413, 0.6363263130187988], "r_norm": [0.452070415019989, 0.4455122947692871, 0.45027536153793335]}, {"step": 800, "epoch": 1, "train_loss": 0.23864632844924927, "cos_r_negg": [0.16706319153308868, 0.18102678656578064, 0.23181748390197754], "cos_innovation_negg": [0.564144492149353, 0.5755118131637573, 0.6405776739120483], "cos_apical_negg": [0.16706319153308868, 0.18102678656578064, 0.23181748390197754], "cos_Ac_negg": [0.5643078088760376, 0.5752726197242737, 0.6404033899307251], "r_norm": [0.42372632026672363, 0.4244443476200104, 0.4254371225833893]}, {"step": 900, "epoch": 1, "train_loss": 0.19433821737766266, "cos_r_negg": [0.19115428626537323, 0.2319030910730362, 0.2743412256240845], "cos_innovation_negg": [0.5572805404663086, 0.5882014036178589, 0.6656894683837891], "cos_apical_negg": [0.19115427136421204, 0.2319030910730362, 0.2743411958217621], "cos_Ac_negg": [0.5571479797363281, 0.5879946351051331, 0.6655135750770569], "r_norm": [0.4814557433128357, 0.4753068685531616, 0.48049071431159973]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9541, "test_loss": 0.16767139129638672}, {"step": 1000, "epoch": 2, "train_loss": 0.07323691993951797, "cos_r_negg": [0.18533377349376678, 0.20593209564685822, 0.22659942507743835], "cos_innovation_negg": [0.5686722993850708, 0.5983743667602539, 0.6596771478652954], "cos_apical_negg": [0.1853337585926056, 0.20593209564685822, 0.22659939527511597], "cos_Ac_negg": [0.5684210062026978, 0.5980938673019409, 0.6594594120979309], "r_norm": [0.48285090923309326, 0.4805031418800354, 0.48812124133110046], "ldr": {"dL_sdil": -0.01850762963294983, "dL_bp": -0.018236294388771057, "ratio": 1.0148788585221484}}, {"step": 1100, "epoch": 2, "train_loss": 0.18410810828208923, "cos_r_negg": [0.15007540583610535, 0.17071875929832458, 0.21554194390773773], "cos_innovation_negg": [0.5643891096115112, 0.614929735660553, 0.7009598016738892], "cos_apical_negg": [0.15007540583610535, 0.17071875929832458, 0.21554194390773773], "cos_Ac_negg": [0.5641417503356934, 0.6147982478141785, 0.7007851600646973], "r_norm": [0.39942309260368347, 0.3952123820781708, 0.399445116519928]}, {"step": 1200, "epoch": 2, "train_loss": 0.1506100744009018, "cos_r_negg": [0.16346396505832672, 0.18483838438987732, 0.21001099050045013], "cos_innovation_negg": [0.5539546012878418, 0.6020072102546692, 0.7045530080795288], "cos_apical_negg": [0.16346396505832672, 0.18483838438987732, 0.21001097559928894], "cos_Ac_negg": [0.5536083579063416, 0.6017486453056335, 0.704311192035675], "r_norm": [0.42719393968582153, 0.4216434955596924, 0.4261261224746704]}, {"step": 1300, "epoch": 2, "train_loss": 0.18525248765945435, "cos_r_negg": [0.1321549415588379, 0.16249878704547882, 0.20184743404388428], "cos_innovation_negg": [0.5685097575187683, 0.6059775948524475, 0.7228296399116516], "cos_apical_negg": [0.1321549415588379, 0.16249877214431763, 0.20184743404388428], "cos_Ac_negg": [0.5683749914169312, 0.6057950854301453, 0.722582221031189], "r_norm": [0.39254075288772583, 0.389163613319397, 0.3909877836704254]}, {"step": 1400, "epoch": 2, "train_loss": 0.15481902658939362, "cos_r_negg": [0.1297166794538498, 0.1661204695701599, 0.2534674406051636], "cos_innovation_negg": [0.5598414540290833, 0.6230447888374329, 0.6980350017547607], "cos_apical_negg": [0.1297166794538498, 0.1661204844713211, 0.25346747040748596], "cos_Ac_negg": [0.5596442222595215, 0.6228131055831909, 0.6977487802505493], "r_norm": [0.3701799511909485, 0.3658253848552704, 0.37039294838905334]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9607, "test_loss": 0.13900364608764648}, {"step": 1500, "epoch": 3, "train_loss": 0.110859215259552, "cos_r_negg": [0.14650633931159973, 0.16383367776870728, 0.23567108809947968], "cos_innovation_negg": [0.575775146484375, 0.6230165362358093, 0.7273306846618652], "cos_apical_negg": [0.14650633931159973, 0.16383367776870728, 0.23567108809947968], "cos_Ac_negg": [0.5751730799674988, 0.6225860714912415, 0.7268973588943481], "r_norm": [0.3419315218925476, 0.336915522813797, 0.340427041053772], "ldr": {"dL_sdil": -0.004863366484642029, "dL_bp": -0.005314543843269348, "ratio": 0.9151051582350351}}, {"step": 1600, "epoch": 3, "train_loss": 0.10772039741277695, "cos_r_negg": [0.14053931832313538, 0.1739019751548767, 0.22011321783065796], "cos_innovation_negg": [0.5879946947097778, 0.6072326302528381, 0.7198115587234497], "cos_apical_negg": [0.14053931832313538, 0.1739019751548767, 0.22011321783065796], "cos_Ac_negg": [0.5874791145324707, 0.6069145202636719, 0.7194622755050659], "r_norm": [0.356960654258728, 0.35426419973373413, 0.3570789694786072]}, {"step": 1700, "epoch": 3, "train_loss": 0.09103884547948837, "cos_r_negg": [0.14236924052238464, 0.1826709359884262, 0.21392950415611267], "cos_innovation_negg": [0.573638916015625, 0.6218717694282532, 0.7085980772972107], "cos_apical_negg": [0.14236922562122345, 0.1826709359884262, 0.21392948925495148], "cos_Ac_negg": [0.5734999179840088, 0.6216661930084229, 0.7083688974380493], "r_norm": [0.3555583357810974, 0.3530515134334564, 0.3562694787979126]}, {"step": 1800, "epoch": 3, "train_loss": 0.030617505311965942, "cos_r_negg": [0.131395623087883, 0.14451764523983002, 0.21224914491176605], "cos_innovation_negg": [0.5636694431304932, 0.6041983962059021, 0.6821848750114441], "cos_apical_negg": [0.131395623087883, 0.14451764523983002, 0.21224914491176605], "cos_Ac_negg": [0.5631300210952759, 0.6032777428627014, 0.6802993416786194], "r_norm": [0.45386093854904175, 0.44351261854171753, 0.45473772287368774]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9606, "test_loss": 0.1345690662384033}, {"step": 1900, "epoch": 4, "train_loss": 0.06669837236404419, "cos_r_negg": [0.12382270395755768, 0.155398428440094, 0.18213950097560883], "cos_innovation_negg": [0.5737964510917664, 0.6262027025222778, 0.7180590629577637], "cos_apical_negg": [0.12382270395755768, 0.155398428440094, 0.18213950097560883], "cos_Ac_negg": [0.5733422040939331, 0.6257920265197754, 0.7177380323410034], "r_norm": [0.2786588966846466, 0.27537527680397034, 0.2776572108268738]}, {"step": 2000, "epoch": 4, "train_loss": 0.11916625499725342, "cos_r_negg": [0.09751120954751968, 0.13077056407928467, 0.14181913435459137], "cos_innovation_negg": [0.5742237567901611, 0.6206357479095459, 0.7055182456970215], "cos_apical_negg": [0.09751120209693909, 0.13077056407928467, 0.14181913435459137], "cos_Ac_negg": [0.5737118124961853, 0.6201224327087402, 0.7049134969711304], "r_norm": [0.2879878282546997, 0.2822757959365845, 0.2883542776107788], "ldr": {"dL_sdil": -0.008462302386760712, "dL_bp": -0.010234296321868896, "ratio": 0.8268572768093744}}, {"step": 2100, "epoch": 4, "train_loss": 0.06574305891990662, "cos_r_negg": [0.1261293590068817, 0.165409117937088, 0.21647286415100098], "cos_innovation_negg": [0.587814211845398, 0.6350489854812622, 0.6933523416519165], "cos_apical_negg": [0.1261293590068817, 0.1654091477394104, 0.21647286415100098], "cos_Ac_negg": [0.5874064564704895, 0.6346099376678467, 0.6926973462104797], "r_norm": [0.3062635362148285, 0.3029603064060211, 0.3067421317100525]}, {"step": 2200, "epoch": 4, "train_loss": 0.050690002739429474, "cos_r_negg": [0.10052382946014404, 0.1354762464761734, 0.21772167086601257], "cos_innovation_negg": [0.5667058825492859, 0.6212449669837952, 0.6726551651954651], "cos_apical_negg": [0.10052382946014404, 0.1354762464761734, 0.21772167086601257], "cos_Ac_negg": [0.5662604570388794, 0.6210074424743652, 0.6722522974014282], "r_norm": [0.3140619993209839, 0.3138364851474762, 0.3157208561897278]}, {"step": 2300, "epoch": 4, "train_loss": 0.2856258749961853, "cos_r_negg": [0.12347837537527084, 0.13504227995872498, 0.1656659096479416], "cos_innovation_negg": [0.5976657867431641, 0.63372802734375, 0.7001543045043945], "cos_apical_negg": [0.12347836792469025, 0.13504226505756378, 0.1656659096479416], "cos_Ac_negg": [0.5968269109725952, 0.6332087516784668, 0.699736475944519], "r_norm": [0.30333665013313293, 0.3017101287841797, 0.3041290044784546]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9656, "test_loss": 0.12579462852478027}, {"step": 2400, "epoch": 5, "train_loss": 0.10922393947839737, "cos_r_negg": [0.0800085961818695, 0.1303296536207199, 0.17170783877372742], "cos_innovation_negg": [0.5716754198074341, 0.6151139140129089, 0.7041137218475342], "cos_apical_negg": [0.0800085961818695, 0.1303296536207199, 0.17170783877372742], "cos_Ac_negg": [0.5712298154830933, 0.6144628524780273, 0.70359206199646], "r_norm": [0.26329895853996277, 0.2613946199417114, 0.26422256231307983]}, {"step": 2500, "epoch": 5, "train_loss": 0.03882201761007309, "cos_r_negg": [0.10450796782970428, 0.14539843797683716, 0.14641329646110535], "cos_innovation_negg": [0.5877382755279541, 0.6312771439552307, 0.7209851741790771], "cos_apical_negg": [0.10450796782970428, 0.14539843797683716, 0.14641329646110535], "cos_Ac_negg": [0.5871776938438416, 0.630534291267395, 0.719475269317627], "r_norm": [0.300679475069046, 0.296906441450119, 0.30106282234191895], "ldr": {"dL_sdil": -0.010640151798725128, "dL_bp": -0.012677699327468872, "ratio": 0.8392809707729089}}, {"step": 2600, "epoch": 5, "train_loss": 0.08008192479610443, "cos_r_negg": [0.11345787346363068, 0.1607760190963745, 0.17814789712429047], "cos_innovation_negg": [0.5683139562606812, 0.615220308303833, 0.723012387752533], "cos_apical_negg": [0.11345787346363068, 0.1607760190963745, 0.17814789712429047], "cos_Ac_negg": [0.5678278207778931, 0.6144174337387085, 0.721819281578064], "r_norm": [0.30044424533843994, 0.29629045724868774, 0.3002690374851227]}, {"step": 2700, "epoch": 5, "train_loss": 0.060378801077604294, "cos_r_negg": [0.0910666212439537, 0.1561325490474701, 0.17633703351020813], "cos_innovation_negg": [0.5736572742462158, 0.6276485323905945, 0.7415227890014648], "cos_apical_negg": [0.0910666212439537, 0.1561325192451477, 0.17633703351020813], "cos_Ac_negg": [0.5730524063110352, 0.6272779703140259, 0.740976870059967], "r_norm": [0.2680075168609619, 0.2659075856208801, 0.2676078677177429]}, {"step": 2800, "epoch": 5, "train_loss": 0.13768145442008972, "cos_r_negg": [0.09511639177799225, 0.1233239695429802, 0.15861858427524567], "cos_innovation_negg": [0.5799553394317627, 0.6164218187332153, 0.7362614870071411], "cos_apical_negg": [0.09511638432741165, 0.12332397699356079, 0.15861856937408447], "cos_Ac_negg": [0.5797080993652344, 0.6170012950897217, 0.7352924346923828], "r_norm": [0.2905276119709015, 0.2830788493156433, 0.28721317648887634]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.965, "test_loss": 0.1283594997406006}, {"step": 2900, "epoch": 6, "train_loss": 0.05049640312790871, "cos_r_negg": [0.08567244559526443, 0.10055835545063019, 0.1446242332458496], "cos_innovation_negg": [0.5851818323135376, 0.640073299407959, 0.7310131788253784], "cos_apical_negg": [0.08567244559526443, 0.10055836290121078, 0.1446242481470108], "cos_Ac_negg": [0.5852477550506592, 0.6404622793197632, 0.731020987033844], "r_norm": [0.29810234904289246, 0.2922148108482361, 0.29742997884750366]}, {"step": 3000, "epoch": 6, "train_loss": 0.13612395524978638, "cos_r_negg": [0.06678971648216248, 0.09917480498552322, 0.10427999496459961], "cos_innovation_negg": [0.5583495497703552, 0.6143648624420166, 0.7150453925132751], "cos_apical_negg": [0.06678971648216248, 0.09917480498552322, 0.10427998751401901], "cos_Ac_negg": [0.5581167936325073, 0.6140143871307373, 0.7146580219268799], "r_norm": [0.233027845621109, 0.22977158427238464, 0.23344995081424713], "ldr": {"dL_sdil": -0.004542052745819092, "dL_bp": -0.006951414048671722, "ratio": 0.653399828296372}}, {"step": 3100, "epoch": 6, "train_loss": 0.031174927949905396, "cos_r_negg": [0.07337063550949097, 0.11348855495452881, 0.1354759931564331], "cos_innovation_negg": [0.5577149391174316, 0.6143739819526672, 0.719869077205658], "cos_apical_negg": [0.07337063550949097, 0.11348855495452881, 0.1354759931564331], "cos_Ac_negg": [0.5575193762779236, 0.6142213344573975, 0.7192609310150146], "r_norm": [0.26552560925483704, 0.26270338892936707, 0.2663534879684448]}, {"step": 3200, "epoch": 6, "train_loss": 0.058665502816438675, "cos_r_negg": [0.07464861869812012, 0.10516990721225739, 0.12432101368904114], "cos_innovation_negg": [0.5637915134429932, 0.6214838027954102, 0.7110974788665771], "cos_apical_negg": [0.07464861869812012, 0.10516990721225739, 0.12432101368904114], "cos_Ac_negg": [0.5637941360473633, 0.6208317279815674, 0.7090954780578613], "r_norm": [0.221273273229599, 0.21947352588176727, 0.22070924937725067]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9643, "test_loss": 0.13263772964477538}, {"step": 3300, "epoch": 7, "train_loss": 0.019306685775518417, "cos_r_negg": [0.07242114841938019, 0.09392867237329483, 0.08804640173912048], "cos_innovation_negg": [0.5655518770217896, 0.6146605014801025, 0.710368275642395], "cos_apical_negg": [0.07242114841938019, 0.09392867982387543, 0.08804639428853989], "cos_Ac_negg": [0.5656253695487976, 0.6150987148284912, 0.7095399498939514], "r_norm": [0.2598894238471985, 0.25448179244995117, 0.2578650116920471]}, {"step": 3400, "epoch": 7, "train_loss": 0.0447990819811821, "cos_r_negg": [0.0849866047501564, 0.0959339588880539, 0.12536758184432983], "cos_innovation_negg": [0.5790256857872009, 0.6278877258300781, 0.743285059928894], "cos_apical_negg": [0.0849865972995758, 0.0959339588880539, 0.12536758184432983], "cos_Ac_negg": [0.5787602663040161, 0.6278969049453735, 0.7430012226104736], "r_norm": [0.22801300883293152, 0.22419679164886475, 0.2248200923204422]}, {"step": 3500, "epoch": 7, "train_loss": 0.029470354318618774, "cos_r_negg": [0.0645921528339386, 0.08396442234516144, 0.07825030386447906], "cos_innovation_negg": [0.5549382567405701, 0.6131943464279175, 0.7136404514312744], "cos_apical_negg": [0.0645921528339386, 0.08396442234516144, 0.07825030386447906], "cos_Ac_negg": [0.5542522072792053, 0.6133655905723572, 0.712344229221344], "r_norm": [0.21154215931892395, 0.20973695814609528, 0.21197417378425598], "ldr": {"dL_sdil": -0.0032623186707496643, "dL_bp": -0.0049062371253967285, "ratio": 0.6649329389039398}}, {"step": 3600, "epoch": 7, "train_loss": 0.017343800514936447, "cos_r_negg": [0.07564590871334076, 0.09254085272550583, 0.09874577075242996], "cos_innovation_negg": [0.5643134117126465, 0.6200522184371948, 0.7193397283554077], "cos_apical_negg": [0.07564590871334076, 0.09254084527492523, 0.09874577820301056], "cos_Ac_negg": [0.5647028088569641, 0.6200195550918579, 0.717861533164978], "r_norm": [0.24469712376594543, 0.24186371266841888, 0.24495328962802887]}, {"step": 3700, "epoch": 7, "train_loss": 0.02560429461300373, "cos_r_negg": [0.0934041291475296, 0.10227875411510468, 0.11355741322040558], "cos_innovation_negg": [0.5772973299026489, 0.6290334463119507, 0.7404904961585999], "cos_apical_negg": [0.0934041291475296, 0.10227875411510468, 0.11355740576982498], "cos_Ac_negg": [0.5771619081497192, 0.6288613080978394, 0.7394621968269348], "r_norm": [0.235325887799263, 0.23235464096069336, 0.23532480001449585]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9669, "test_loss": 0.12723760547637938}, {"step": 3800, "epoch": 8, "train_loss": 0.02622942626476288, "cos_r_negg": [0.04437864199280739, 0.07326246052980423, 0.08763329684734344], "cos_innovation_negg": [0.561237633228302, 0.6289453506469727, 0.7156274318695068], "cos_apical_negg": [0.04437864571809769, 0.07326246052980423, 0.08763329684734344], "cos_Ac_negg": [0.5624828934669495, 0.6301823258399963, 0.7157561182975769], "r_norm": [0.21577385067939758, 0.21152213215827942, 0.21474748849868774]}, {"step": 3900, "epoch": 8, "train_loss": 0.11024263501167297, "cos_r_negg": [0.07163235545158386, 0.08981633186340332, 0.10607653856277466], "cos_innovation_negg": [0.5642953515052795, 0.6160035133361816, 0.6934971809387207], "cos_apical_negg": [0.07163235545158386, 0.08981633186340332, 0.10607653856277466], "cos_Ac_negg": [0.5641449689865112, 0.616386890411377, 0.692419171333313], "r_norm": [0.2235879898071289, 0.22061657905578613, 0.22167834639549255]}, {"step": 4000, "epoch": 8, "train_loss": 0.032469671219587326, "cos_r_negg": [0.05761587619781494, 0.08533866703510284, 0.10965381562709808], "cos_innovation_negg": [0.5626274347305298, 0.629766047000885, 0.7038819789886475], "cos_apical_negg": [0.05761587619781494, 0.08533865958452225, 0.10965380817651749], "cos_Ac_negg": [0.5637836456298828, 0.6311998963356018, 0.7026136517524719], "r_norm": [0.2095927894115448, 0.20795543491840363, 0.2076958417892456], "ldr": {"dL_sdil": -0.004674933850765228, "dL_bp": -0.00748114287853241, "ratio": 0.6248956779291347}}, {"step": 4100, "epoch": 8, "train_loss": 0.015821238979697227, "cos_r_negg": [0.06512697786092758, 0.06940074265003204, 0.10688220709562302], "cos_innovation_negg": [0.5558160543441772, 0.6281419992446899, 0.7058539390563965], "cos_apical_negg": [0.06512698531150818, 0.06940074265003204, 0.10688220709562302], "cos_Ac_negg": [0.5582858920097351, 0.6296072006225586, 0.7058401107788086], "r_norm": [0.23751769959926605, 0.23330527544021606, 0.23546218872070312]}, {"step": 4200, "epoch": 8, "train_loss": 0.0786832720041275, "cos_r_negg": [0.07665185630321503, 0.07518652081489563, 0.12023349851369858], "cos_innovation_negg": [0.5658077001571655, 0.6285502314567566, 0.7334243059158325], "cos_apical_negg": [0.07665185630321503, 0.07518652081489563, 0.12023349106311798], "cos_Ac_negg": [0.5689389705657959, 0.6302039623260498, 0.7326260209083557], "r_norm": [0.17267274856567383, 0.1692860722541809, 0.17201559245586395]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9699, "test_loss": 0.12432236766815186}, {"step": 4300, "epoch": 9, "train_loss": 0.03487960249185562, "cos_r_negg": [0.04885092377662659, 0.09549008309841156, 0.14046651124954224], "cos_innovation_negg": [0.5548266172409058, 0.6173526048660278, 0.7183548212051392], "cos_apical_negg": [0.04885092377662659, 0.09549008309841156, 0.14046651124954224], "cos_Ac_negg": [0.5621811747550964, 0.6217618584632874, 0.7210865020751953], "r_norm": [0.22707624733448029, 0.22213496267795563, 0.22637689113616943]}, {"step": 4400, "epoch": 9, "train_loss": 0.016691552475094795, "cos_r_negg": [0.04015784710645676, 0.09020178020000458, 0.09162594377994537], "cos_innovation_negg": [0.5598901510238647, 0.6221605539321899, 0.73450767993927], "cos_apical_negg": [0.04015784338116646, 0.09020177274942398, 0.09162594377994537], "cos_Ac_negg": [0.5633233785629272, 0.6249106526374817, 0.7337782382965088], "r_norm": [0.21741098165512085, 0.21283850073814392, 0.21458357572555542]}, {"step": 4500, "epoch": 9, "train_loss": 0.08103659003973007, "cos_r_negg": [0.07006952166557312, 0.09883368760347366, 0.10027097910642624], "cos_innovation_negg": [0.5810632705688477, 0.64183509349823, 0.7424862384796143], "cos_apical_negg": [0.07006952166557312, 0.09883368015289307, 0.10027097165584564], "cos_Ac_negg": [0.5830268859863281, 0.64426589012146, 0.7422841191291809], "r_norm": [0.23907096683979034, 0.23545652627944946, 0.23635035753250122], "ldr": {"dL_sdil": -0.0072174072265625, "dL_bp": -0.012745216488838196, "ratio": 0.5662836118070844}}, {"step": 4600, "epoch": 9, "train_loss": 0.01189473643898964, "cos_r_negg": [0.04925631359219551, 0.09685302525758743, 0.09273850172758102], "cos_innovation_negg": [0.5442144870758057, 0.6160503625869751, 0.7281934022903442], "cos_apical_negg": [0.04925630986690521, 0.09685303270816803, 0.09273849427700043], "cos_Ac_negg": [0.549444317817688, 0.6212999820709229, 0.7269050478935242], "r_norm": [0.22633762657642365, 0.2216828614473343, 0.22389601171016693]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9705, "test_loss": 0.1226830888748169}, {"step": 4700, "epoch": 10, "train_loss": 0.01758262701332569, "cos_r_negg": [0.049477897584438324, 0.05030638724565506, 0.06188071519136429], "cos_innovation_negg": [0.5504025220870972, 0.6181167960166931, 0.7068608999252319], "cos_apical_negg": [0.049477897584438324, 0.05030638724565506, 0.06188071891665459], "cos_Ac_negg": [0.5557631254196167, 0.6228522658348083, 0.7075021266937256], "r_norm": [0.20983056724071503, 0.20785710215568542, 0.2090214639902115]}, {"step": 4800, "epoch": 10, "train_loss": 0.06082163378596306, "cos_r_negg": [0.030915986746549606, 0.05864178016781807, 0.08276590704917908], "cos_innovation_negg": [0.5669792890548706, 0.6159775257110596, 0.7320576906204224], "cos_apical_negg": [0.030915983021259308, 0.05864178389310837, 0.08276590704917908], "cos_Ac_negg": [0.5709697604179382, 0.6198879480361938, 0.7328310012817383], "r_norm": [0.2022201120853424, 0.19934101402759552, 0.20066407322883606]}, {"step": 4900, "epoch": 10, "train_loss": 0.032953646034002304, "cos_r_negg": [0.05380989611148834, 0.08456790447235107, 0.08466912806034088], "cos_innovation_negg": [0.5723910927772522, 0.6243938207626343, 0.7248529195785522], "cos_apical_negg": [0.05380989611148834, 0.08456790447235107, 0.08466912806034088], "cos_Ac_negg": [0.5737979412078857, 0.627792477607727, 0.7235276699066162], "r_norm": [0.21145141124725342, 0.207720547914505, 0.20969641208648682]}, {"step": 5000, "epoch": 10, "train_loss": 0.008504677563905716, "cos_r_negg": [0.06354457139968872, 0.07587283104658127, 0.06342294812202454], "cos_innovation_negg": [0.5497069358825684, 0.6113346219062805, 0.7288207411766052], "cos_apical_negg": [0.06354457139968872, 0.07587282359600067, 0.06342294812202454], "cos_Ac_negg": [0.5550521016120911, 0.6166893243789673, 0.7290618419647217], "r_norm": [0.18436767160892487, 0.1815374195575714, 0.18342182040214539], "ldr": {"dL_sdil": -0.0031595155596733093, "dL_bp": -0.006094314157962799, "ratio": 0.5184366079233219}}, {"step": 5100, "epoch": 10, "train_loss": 0.051396749913692474, "cos_r_negg": [0.06187586486339569, 0.07489577680826187, 0.10692720860242844], "cos_innovation_negg": [0.5524444580078125, 0.6378144025802612, 0.7474163174629211], "cos_apical_negg": [0.06187586486339569, 0.07489577680826187, 0.10692720860242844], "cos_Ac_negg": [0.5572330355644226, 0.6413361430168152, 0.7495434284210205], "r_norm": [0.19959013164043427, 0.1974097490310669, 0.1995234191417694]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9692, "test_loss": 0.1310543426513672}, {"step": 5200, "epoch": 11, "train_loss": 0.007362754084169865, "cos_r_negg": [0.0649820864200592, 0.08390449732542038, 0.09689496457576752], "cos_innovation_negg": [0.5592031478881836, 0.6397548913955688, 0.7427685856819153], "cos_apical_negg": [0.06498207896947861, 0.08390449732542038, 0.09689496457576752], "cos_Ac_negg": [0.5643541812896729, 0.6436315178871155, 0.7445123195648193], "r_norm": [0.22553887963294983, 0.2213938981294632, 0.2233155220746994]}, {"step": 5300, "epoch": 11, "train_loss": 0.011183683760464191, "cos_r_negg": [0.043541137129068375, 0.06899329274892807, 0.10155381262302399], "cos_innovation_negg": [0.5573720932006836, 0.6282849311828613, 0.7575475573539734], "cos_apical_negg": [0.043541137129068375, 0.06899329274892807, 0.10155382007360458], "cos_Ac_negg": [0.5621886253356934, 0.6333240270614624, 0.7567877769470215], "r_norm": [0.18876740336418152, 0.18407520651817322, 0.18542222678661346]}, {"step": 5400, "epoch": 11, "train_loss": 0.011638451367616653, "cos_r_negg": [0.04954775795340538, 0.059115152806043625, 0.08713265508413315], "cos_innovation_negg": [0.5465665459632874, 0.6171325445175171, 0.7398940324783325], "cos_apical_negg": [0.04954775422811508, 0.059115152806043625, 0.08713264763355255], "cos_Ac_negg": [0.5519935488700867, 0.6228265762329102, 0.7411057353019714], "r_norm": [0.19537073373794556, 0.1906038224697113, 0.19320827722549438]}, {"step": 5500, "epoch": 11, "train_loss": 0.02806015871465206, "cos_r_negg": [0.06098166108131409, 0.062265340238809586, 0.09589625149965286], "cos_innovation_negg": [0.5488113760948181, 0.6229140758514404, 0.7256801724433899], "cos_apical_negg": [0.06098165735602379, 0.062265340238809586, 0.09589624404907227], "cos_Ac_negg": [0.5533926486968994, 0.6290234327316284, 0.7265867590904236], "r_norm": [0.20373515784740448, 0.20048436522483826, 0.20147669315338135], "ldr": {"dL_sdil": -0.004074700176715851, "dL_bp": -0.00852278620004654, "ratio": 0.4780948484538543}}, {"step": 5600, "epoch": 11, "train_loss": 0.010782569646835327, "cos_r_negg": [0.049422577023506165, 0.06817072629928589, 0.12336263805627823], "cos_innovation_negg": [0.5475457906723022, 0.6164689064025879, 0.7429060935974121], "cos_apical_negg": [0.049422577023506165, 0.06817072629928589, 0.12336264550685883], "cos_Ac_negg": [0.5579742193222046, 0.6262809634208679, 0.746960461139679], "r_norm": [0.2047860473394394, 0.20101822912693024, 0.20428185164928436]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9706, "test_loss": 0.12458508834838868}, {"step": 5700, "epoch": 12, "train_loss": 0.012705364264547825, "cos_r_negg": [0.04466864466667175, 0.07217530906200409, 0.12809297442436218], "cos_innovation_negg": [0.5514708757400513, 0.6251588463783264, 0.7614907622337341], "cos_apical_negg": [0.04466864466667175, 0.07217530906200409, 0.12809298932552338], "cos_Ac_negg": [0.5596020221710205, 0.6324319243431091, 0.7608262896537781], "r_norm": [0.22649411857128143, 0.22218579053878784, 0.2237890660762787]}, {"step": 5800, "epoch": 12, "train_loss": 0.007963050156831741, "cos_r_negg": [0.04596461355686188, 0.05485370382666588, 0.09964601695537567], "cos_innovation_negg": [0.5427035093307495, 0.6207987070083618, 0.7493276000022888], "cos_apical_negg": [0.04596461355686188, 0.05485369265079498, 0.09964600950479507], "cos_Ac_negg": [0.5507823824882507, 0.6290920376777649, 0.7533329725265503], "r_norm": [0.1967996209859848, 0.1934458613395691, 0.19505123794078827]}, {"step": 5900, "epoch": 12, "train_loss": 0.005631913430988789, "cos_r_negg": [0.02647482231259346, 0.05131883919239044, 0.11191203445196152], "cos_innovation_negg": [0.5246216058731079, 0.597536563873291, 0.7366989850997925], "cos_apical_negg": [0.02647482231259346, 0.05131884291768074, 0.11191204190254211], "cos_Ac_negg": [0.5407565832138062, 0.6185312867164612, 0.7418075799942017], "r_norm": [0.2164660096168518, 0.21117088198661804, 0.21352757513523102]}, {"step": 6000, "epoch": 12, "train_loss": 0.007956218905746937, "cos_r_negg": [0.03920429199934006, 0.0754752978682518, 0.12268101423978806], "cos_innovation_negg": [0.5336796045303345, 0.6201992630958557, 0.7478694915771484], "cos_apical_negg": [0.03920429199934006, 0.0754752904176712, 0.12268102914094925], "cos_Ac_negg": [0.5447767376899719, 0.6308369636535645, 0.7444664239883423], "r_norm": [0.21638329327106476, 0.21274390816688538, 0.21485933661460876], "ldr": {"dL_sdil": -0.007918410003185272, "dL_bp": -0.016881465911865234, "ratio": 0.46905938409174364}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.967, "test_loss": 0.13685971374511718}, {"step": 6100, "epoch": 13, "train_loss": 0.02517504058778286, "cos_r_negg": [0.056664321571588516, 0.06747549027204514, 0.09756827354431152], "cos_innovation_negg": [0.5624748468399048, 0.6267976760864258, 0.765487790107727], "cos_apical_negg": [0.056664325296878815, 0.06747548282146454, 0.09756826609373093], "cos_Ac_negg": [0.5680994987487793, 0.6344419717788696, 0.7677033543586731], "r_norm": [0.19518572092056274, 0.19355317950248718, 0.19456177949905396]}, {"step": 6200, "epoch": 13, "train_loss": 0.012008452787995338, "cos_r_negg": [0.02989443950355053, 0.059632353484630585, 0.09686162322759628], "cos_innovation_negg": [0.5549407601356506, 0.6254200339317322, 0.7544149160385132], "cos_apical_negg": [0.029894443228840828, 0.059632353484630585, 0.09686162322759628], "cos_Ac_negg": [0.5630207657814026, 0.6364709138870239, 0.7583058476448059], "r_norm": [0.18568243086338043, 0.18361473083496094, 0.18413721024990082]}, {"step": 6300, "epoch": 13, "train_loss": 0.00748961279168725, "cos_r_negg": [0.04999706149101257, 0.07078620791435242, 0.10934516042470932], "cos_innovation_negg": [0.5585170984268188, 0.6274136900901794, 0.7738465070724487], "cos_apical_negg": [0.04999706149101257, 0.07078620791435242, 0.10934515297412872], "cos_Ac_negg": [0.5641648173332214, 0.6356494426727295, 0.7718241214752197], "r_norm": [0.2185528725385666, 0.21450120210647583, 0.21682244539260864]}, {"step": 6400, "epoch": 13, "train_loss": 0.0045163496397435665, "cos_r_negg": [0.045995961874723434, 0.06127827614545822, 0.10519485920667648], "cos_innovation_negg": [0.5285505652427673, 0.619375467300415, 0.7469182014465332], "cos_apical_negg": [0.04599596560001373, 0.06127827614545822, 0.10519486665725708], "cos_Ac_negg": [0.5412546396255493, 0.6326603889465332, 0.7492698431015015], "r_norm": [0.2200552523136139, 0.2154013067483902, 0.21812433004379272]}, {"step": 6500, "epoch": 13, "train_loss": 0.003596516326069832, "cos_r_negg": [0.04465258866548538, 0.07837079465389252, 0.0702044665813446], "cos_innovation_negg": [0.5400137901306152, 0.6217684745788574, 0.7461681365966797], "cos_apical_negg": [0.04465258866548538, 0.07837080955505371, 0.0702044665813446], "cos_Ac_negg": [0.5481486320495605, 0.6315551996231079, 0.7488528490066528], "r_norm": [0.19117380678653717, 0.18793326616287231, 0.1886856108903885], "ldr": {"dL_sdil": -0.004521757364273071, "dL_bp": -0.011233538389205933, "ratio": 0.402522981415894}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9708, "test_loss": 0.1253721086502075}, {"step": 6600, "epoch": 14, "train_loss": 0.0057099200785160065, "cos_r_negg": [0.047673728317022324, 0.07365186512470245, 0.11604772508144379], "cos_innovation_negg": [0.5477915406227112, 0.6239545941352844, 0.7583589553833008], "cos_apical_negg": [0.047673724591732025, 0.07365186512470245, 0.11604771018028259], "cos_Ac_negg": [0.5555763840675354, 0.6343176364898682, 0.756290853023529], "r_norm": [0.19733330607414246, 0.1941097229719162, 0.19474002718925476]}, {"step": 6700, "epoch": 14, "train_loss": 0.009178523905575275, "cos_r_negg": [0.040871068835258484, 0.06444497406482697, 0.12502259016036987], "cos_innovation_negg": [0.5270074605941772, 0.6034623384475708, 0.7457575798034668], "cos_apical_negg": [0.040871068835258484, 0.06444497406482697, 0.12502259016036987], "cos_Ac_negg": [0.5431610345840454, 0.623917281627655, 0.7431623339653015], "r_norm": [0.21842065453529358, 0.21405012905597687, 0.21624219417572021]}, {"step": 6800, "epoch": 14, "train_loss": 0.024552559480071068, "cos_r_negg": [0.058451589196920395, 0.07176758348941803, 0.09812598675489426], "cos_innovation_negg": [0.5318054556846619, 0.603507936000824, 0.7431598901748657], "cos_apical_negg": [0.058451589196920395, 0.07176758348941803, 0.09812598675489426], "cos_Ac_negg": [0.5433749556541443, 0.6183546781539917, 0.7466729879379272], "r_norm": [0.2067817896604538, 0.20290452241897583, 0.20498932898044586]}, {"step": 6900, "epoch": 14, "train_loss": 0.0077741569839417934, "cos_r_negg": [0.04254547506570816, 0.06389213353395462, 0.08488097786903381], "cos_innovation_negg": [0.5364390015602112, 0.597833514213562, 0.7362339496612549], "cos_apical_negg": [0.04254548251628876, 0.06389213353395462, 0.08488097786903381], "cos_Ac_negg": [0.5489690899848938, 0.6135779023170471, 0.7435045838356018], "r_norm": [0.20058926939964294, 0.19730912148952484, 0.19768646359443665]}, {"step": 7000, "epoch": 14, "train_loss": 0.006662845145910978, "cos_r_negg": [0.05107412114739418, 0.06778308749198914, 0.10127559304237366], "cos_innovation_negg": [0.5472577810287476, 0.6209309697151184, 0.7657016515731812], "cos_apical_negg": [0.05107412114739418, 0.06778308004140854, 0.10127559304237366], "cos_Ac_negg": [0.5561801791191101, 0.634848952293396, 0.7695335149765015], "r_norm": [0.24305014312267303, 0.23800106346607208, 0.24165913462638855], "ldr": {"dL_sdil": -0.006982788443565369, "dL_bp": -0.01557321846485138, "ratio": 0.4483844145207018}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9705, "test_loss": 0.13234205627441406}], "final": {"test_acc": 0.9705, "test_loss": 0.13234205627441406, "wall_s": 25.417935609817505, "cos_r_negg": [0.046488724648952484, 0.06853098422288895, 0.12159444391727448], "cos_innovation_negg": [0.5427753925323486, 0.6186004877090454, 0.7642629146575928], "cos_apical_negg": [0.046488720923662186, 0.06853098422288895, 0.12159444391727448], "cos_Ac_negg": [0.5519708395004272, 0.6351718306541443, 0.771507978439331]}}