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": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p5_residual_s1", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.327335834503174, "cos_r_negg": [0.011923439800739288, -0.013784728944301605, -0.005784478038549423], "cos_innovation_negg": [0.011923439800739288, -0.013784728944301605, -0.005784478038549423], "cos_apical_negg": [0.0065222978591918945, -0.010908648371696472, -0.008246229030191898], "cos_Ac_negg": [0.01192301046103239, -0.013784660957753658, -0.0057846191339194775], "r_norm": [4.739156723022461, 4.7983598709106445, 4.788495063781738], "ldr": {"dL_sdil": -0.04239368438720703, "dL_bp": -0.2199082374572754, "ratio": 0.19277897398201574}}, {"step": 100, "epoch": 0, "train_loss": 0.19819821417331696, "cos_r_negg": [0.4562336206436157, 0.3609468638896942, 0.3217041492462158], "cos_innovation_negg": [0.4562336206436157, 0.3609468638896942, 0.3217041492462158], "cos_apical_negg": [0.0924934595823288, 0.07498198002576828, 0.10977184027433395], "cos_Ac_negg": [0.4592011868953705, 0.3636699914932251, 0.32186806201934814], "r_norm": [0.9399021863937378, 0.9428062438964844, 0.9601663947105408]}, {"step": 200, "epoch": 0, "train_loss": 0.32231855392456055, "cos_r_negg": [0.4961671829223633, 0.42615118622779846, 0.3737158179283142], "cos_innovation_negg": [0.4961671829223633, 0.42615118622779846, 0.3737158179283142], "cos_apical_negg": [0.08375506103038788, 0.07275334745645523, 0.10912351310253143], "cos_Ac_negg": [0.5020463466644287, 0.4308222830295563, 0.3757030963897705], "r_norm": [0.7590075135231018, 0.7669145464897156, 0.7767611742019653]}, {"step": 300, "epoch": 0, "train_loss": 0.1517620086669922, "cos_r_negg": [0.4997817575931549, 0.45438945293426514, 0.418070912361145], "cos_innovation_negg": [0.4997817575931549, 0.45438945293426514, 0.418070912361145], "cos_apical_negg": [0.06755456328392029, 0.070278100669384, 0.11749660968780518], "cos_Ac_negg": [0.5018380880355835, 0.45461800694465637, 0.41634446382522583], "r_norm": [0.5646820068359375, 0.5667244791984558, 0.57851642370224]}, {"step": 400, "epoch": 0, "train_loss": 0.12696349620819092, "cos_r_negg": [0.5265330076217651, 0.47366011142730713, 0.4304295778274536], "cos_innovation_negg": [0.5265330076217651, 0.47366011142730713, 0.4304295778274536], "cos_apical_negg": [0.05604200437664986, 0.07228812575340271, 0.10936453193426132], "cos_Ac_negg": [0.5275131464004517, 0.4741481840610504, 0.42993783950805664], "r_norm": [0.5426884889602661, 0.5449879169464111, 0.5514365434646606]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9475, "test_loss": 0.18153308029174806}, {"step": 500, "epoch": 1, "train_loss": 0.1655046045780182, "cos_r_negg": [0.5191988945007324, 0.48418760299682617, 0.4616186022758484], "cos_innovation_negg": [0.5191988945007324, 0.48418760299682617, 0.4616186022758484], "cos_apical_negg": [0.06751049309968948, 0.07074972242116928, 0.125736802816391], "cos_Ac_negg": [0.5299745798110962, 0.4899352490901947, 0.46142110228538513], "r_norm": [0.4868793189525604, 0.48760986328125, 0.48948433995246887], "ldr": {"dL_sdil": -0.01239427924156189, "dL_bp": -0.010387584567070007, "ratio": 1.1931820300990246}}, {"step": 600, "epoch": 1, "train_loss": 0.10512755811214447, "cos_r_negg": [0.506693959236145, 0.47643205523490906, 0.45370903611183167], "cos_innovation_negg": [0.506693959236145, 0.47643205523490906, 0.45370903611183167], "cos_apical_negg": [0.06160319596529007, 0.0535128079354763, 0.1167837530374527], "cos_Ac_negg": [0.5124704837799072, 0.47943711280822754, 0.4548422396183014], "r_norm": [0.47535842657089233, 0.47875380516052246, 0.4810010492801666]}, {"step": 700, "epoch": 1, "train_loss": 0.18744073808193207, "cos_r_negg": [0.5276387929916382, 0.4997781217098236, 0.48154783248901367], "cos_innovation_negg": [0.5276387929916382, 0.4997781217098236, 0.48154783248901367], "cos_apical_negg": [0.05852305144071579, 0.05880728363990784, 0.09995196014642715], "cos_Ac_negg": [0.5367352962493896, 0.5038727521896362, 0.4854159355163574], "r_norm": [0.4378526508808136, 0.4416790306568146, 0.44054216146469116]}, {"step": 800, "epoch": 1, "train_loss": 0.14523780345916748, "cos_r_negg": [0.535984456539154, 0.5065294504165649, 0.47763127088546753], "cos_innovation_negg": [0.535984456539154, 0.5065294504165649, 0.47763127088546753], "cos_apical_negg": [0.043083205819129944, 0.060550667345523834, 0.11668789386749268], "cos_Ac_negg": [0.5489453673362732, 0.5140861868858337, 0.4792100787162781], "r_norm": [0.40186405181884766, 0.4020804166793823, 0.4056309759616852]}, {"step": 900, "epoch": 1, "train_loss": 0.2052808701992035, "cos_r_negg": [0.5632014274597168, 0.5317338705062866, 0.48478347063064575], "cos_innovation_negg": [0.5632014274597168, 0.5317338705062866, 0.48478347063064575], "cos_apical_negg": [0.0345851331949234, 0.0543665736913681, 0.11113052070140839], "cos_Ac_negg": [0.5638846158981323, 0.5336527824401855, 0.48559439182281494], "r_norm": [0.3906911313533783, 0.3913891017436981, 0.393975168466568]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9574, "test_loss": 0.14721737022399903}, {"step": 1000, "epoch": 2, "train_loss": 0.024760780856013298, "cos_r_negg": [0.5648545622825623, 0.5417273640632629, 0.496168315410614], "cos_innovation_negg": [0.5648545622825623, 0.5417273640632629, 0.496168315410614], "cos_apical_negg": [0.04227800667285919, 0.05153029412031174, 0.10890823602676392], "cos_Ac_negg": [0.571323812007904, 0.545744001865387, 0.4980310797691345], "r_norm": [0.3385966420173645, 0.3391200304031372, 0.34419602155685425], "ldr": {"dL_sdil": -0.0067184194922447205, "dL_bp": -0.005749911069869995, "ratio": 1.1684388524632647}}, {"step": 1100, "epoch": 2, "train_loss": 0.16186027228832245, "cos_r_negg": [0.5603777766227722, 0.531798779964447, 0.49178510904312134], "cos_innovation_negg": [0.5603777766227722, 0.531798779964447, 0.49178510904312134], "cos_apical_negg": [0.03889578580856323, 0.047157835215330124, 0.1025448739528656], "cos_Ac_negg": [0.5667866468429565, 0.5373232364654541, 0.4949599504470825], "r_norm": [0.35095852613449097, 0.35477763414382935, 0.35318225622177124]}, {"step": 1200, "epoch": 2, "train_loss": 0.1569242626428604, "cos_r_negg": [0.5550088882446289, 0.5344139337539673, 0.4981812834739685], "cos_innovation_negg": [0.5550088882446289, 0.5344139337539673, 0.4981812834739685], "cos_apical_negg": [0.02155444025993347, 0.040098030120134354, 0.10366980731487274], "cos_Ac_negg": [0.566131591796875, 0.5400733947753906, 0.5006672143936157], "r_norm": [0.2950282692909241, 0.29617446660995483, 0.29916858673095703]}, {"step": 1300, "epoch": 2, "train_loss": 0.1459813266992569, "cos_r_negg": [0.5581264495849609, 0.5500150918960571, 0.5197196006774902], "cos_innovation_negg": [0.5581264495849609, 0.5500150918960571, 0.5197196006774902], "cos_apical_negg": [0.024192284792661667, 0.0479893758893013, 0.116562619805336], "cos_Ac_negg": [0.5768191814422607, 0.5619202852249146, 0.5239880681037903], "r_norm": [0.3309983015060425, 0.32787635922431946, 0.32926201820373535]}, {"step": 1400, "epoch": 2, "train_loss": 0.10320207476615906, "cos_r_negg": [0.5416789054870605, 0.5423123836517334, 0.5176808834075928], "cos_innovation_negg": [0.5416789054870605, 0.5423123836517334, 0.5176808834075928], "cos_apical_negg": [0.014692528173327446, 0.04524938762187958, 0.10794402658939362], "cos_Ac_negg": [0.5685117244720459, 0.5620934963226318, 0.5263524651527405], "r_norm": [0.30710625648498535, 0.30455282330513, 0.30817869305610657]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9644, "test_loss": 0.1145642152786255}, {"step": 1500, "epoch": 3, "train_loss": 0.06334341317415237, "cos_r_negg": [0.5452038049697876, 0.5396445989608765, 0.5256383419036865], "cos_innovation_negg": [0.5452038049697876, 0.5396445989608765, 0.5256383419036865], "cos_apical_negg": [0.017272096127271652, 0.03823789209127426, 0.11339951306581497], "cos_Ac_negg": [0.5736880898475647, 0.5632601976394653, 0.5382489562034607], "r_norm": [0.2519826292991638, 0.2503807246685028, 0.2527422606945038], "ldr": {"dL_sdil": -0.005224883556365967, "dL_bp": -0.005164317786693573, "ratio": 1.0117277387205814}}, {"step": 1600, "epoch": 3, "train_loss": 0.08027355372905731, "cos_r_negg": [0.5536803007125854, 0.547955334186554, 0.5315324068069458], "cos_innovation_negg": [0.5536803007125854, 0.547955334186554, 0.5315324068069458], "cos_apical_negg": [0.016805946826934814, 0.04154978692531586, 0.11257649213075638], "cos_Ac_negg": [0.5790594816207886, 0.5665426850318909, 0.5419608354568481], "r_norm": [0.27155372500419617, 0.27073079347610474, 0.27323296666145325]}, {"step": 1700, "epoch": 3, "train_loss": 0.08025884628295898, "cos_r_negg": [0.5233545303344727, 0.5304712653160095, 0.5180680751800537], "cos_innovation_negg": [0.5233545303344727, 0.5304712653160095, 0.5180680751800537], "cos_apical_negg": [0.004102821461856365, 0.037973955273628235, 0.10750962793827057], "cos_Ac_negg": [0.5699038505554199, 0.5700200796127319, 0.5456762313842773], "r_norm": [0.25968384742736816, 0.2586820125579834, 0.2624744176864624]}, {"step": 1800, "epoch": 3, "train_loss": 0.04806072264909744, "cos_r_negg": [0.5499836802482605, 0.5529208183288574, 0.549586832523346], "cos_innovation_negg": [0.5499836802482605, 0.5529208183288574, 0.549586832523346], "cos_apical_negg": [0.003273674054071307, 0.036957550793886185, 0.10541849583387375], "cos_Ac_negg": [0.5759398937225342, 0.570105791091919, 0.55732661485672], "r_norm": [0.2637656331062317, 0.2634180188179016, 0.26708728075027466]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9673, "test_loss": 0.11477804679870605}, {"step": 1900, "epoch": 4, "train_loss": 0.02444826439023018, "cos_r_negg": [0.5487892627716064, 0.5355829000473022, 0.542591392993927], "cos_innovation_negg": [0.5487892627716064, 0.5355829000473022, 0.542591392993927], "cos_apical_negg": [0.022186705842614174, 0.04220614582300186, 0.10831473767757416], "cos_Ac_negg": [0.5984379649162292, 0.5697867274284363, 0.5640909671783447], "r_norm": [0.32423821091651917, 0.32326072454452515, 0.3276006579399109]}, {"step": 2000, "epoch": 4, "train_loss": 0.12103118747472763, "cos_r_negg": [0.5188897848129272, 0.5130773186683655, 0.5285385847091675], "cos_innovation_negg": [0.5188897848129272, 0.5130773186683655, 0.5285385847091675], "cos_apical_negg": [0.003930566366761923, 0.03765185922384262, 0.09973377734422684], "cos_Ac_negg": [0.5726943016052246, 0.561560869216919, 0.5648659467697144], "r_norm": [0.23365336656570435, 0.2309618890285492, 0.2362920045852661], "ldr": {"dL_sdil": -0.00528167188167572, "dL_bp": -0.005683548748493195, "ratio": 0.9292912079052689}}, {"step": 2100, "epoch": 4, "train_loss": 0.024333203211426735, "cos_r_negg": [0.532158613204956, 0.5350310802459717, 0.5475729703903198], "cos_innovation_negg": [0.532158613204956, 0.5350310802459717, 0.5475729703903198], "cos_apical_negg": [0.012790195643901825, 0.05143173038959503, 0.10263633728027344], "cos_Ac_negg": [0.5851036310195923, 0.5792784094810486, 0.5785484313964844], "r_norm": [0.3119562864303589, 0.30925729870796204, 0.3135855495929718]}, {"step": 2200, "epoch": 4, "train_loss": 0.007553962990641594, "cos_r_negg": [0.5498648881912231, 0.5508173108100891, 0.5553864240646362], "cos_innovation_negg": [0.5498648881912231, 0.5508173108100891, 0.5553864240646362], "cos_apical_negg": [0.00876094214618206, 0.040761008858680725, 0.1037021279335022], "cos_Ac_negg": [0.6020341515541077, 0.5934842824935913, 0.5816723704338074], "r_norm": [0.23748338222503662, 0.23687006533145905, 0.23978067934513092]}, {"step": 2300, "epoch": 4, "train_loss": 0.1032857894897461, "cos_r_negg": [0.5031359195709229, 0.5203165411949158, 0.537436306476593], "cos_innovation_negg": [0.5031359195709229, 0.5203165411949158, 0.537436306476593], "cos_apical_negg": [0.015608749352395535, 0.03436913341283798, 0.10463280230760574], "cos_Ac_negg": [0.5825303196907043, 0.5863634347915649, 0.5832058787345886], "r_norm": [0.23429690301418304, 0.2324206829071045, 0.23625370860099792]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9693, "test_loss": 0.11051450176239014}, {"step": 2400, "epoch": 5, "train_loss": 0.06244773417711258, "cos_r_negg": [0.5216194987297058, 0.532431960105896, 0.5434267520904541], "cos_innovation_negg": [0.5216194987297058, 0.532431960105896, 0.5434267520904541], "cos_apical_negg": [-0.010487285442650318, 0.0381862111389637, 0.09908141940832138], "cos_Ac_negg": [0.6037192344665527, 0.5965788960456848, 0.5930641889572144], "r_norm": [0.23071017861366272, 0.22901231050491333, 0.23396210372447968]}, {"step": 2500, "epoch": 5, "train_loss": 0.01840227097272873, "cos_r_negg": [0.4826870262622833, 0.501519501209259, 0.5196942090988159], "cos_innovation_negg": [0.4826870262622833, 0.501519501209259, 0.5196942090988159], "cos_apical_negg": [0.008614158257842064, 0.04547841101884842, 0.0974075198173523], "cos_Ac_negg": [0.5942498445510864, 0.5917768478393555, 0.599340558052063], "r_norm": [0.24934172630310059, 0.24767103791236877, 0.24996879696846008], "ldr": {"dL_sdil": -0.011170715093612671, "dL_bp": -0.01398564875125885, "ratio": 0.7987269873774854}}, {"step": 2600, "epoch": 5, "train_loss": 0.061060830950737, "cos_r_negg": [0.49484387040138245, 0.5180107355117798, 0.5349160432815552], "cos_innovation_negg": [0.49484387040138245, 0.5180107355117798, 0.5349160432815552], "cos_apical_negg": [0.007128430530428886, 0.04086736589670181, 0.09839516878128052], "cos_Ac_negg": [0.598257303237915, 0.6041316986083984, 0.5988324284553528], "r_norm": [0.21950285136699677, 0.21728940308094025, 0.22030556201934814]}, {"step": 2700, "epoch": 5, "train_loss": 0.012292544357478619, "cos_r_negg": [0.49846529960632324, 0.5182821750640869, 0.5381145477294922], "cos_innovation_negg": [0.49846529960632324, 0.5182821750640869, 0.5381145477294922], "cos_apical_negg": [0.010728912428021431, 0.03891203925013542, 0.09855789691209793], "cos_Ac_negg": [0.611120343208313, 0.6131026148796082, 0.6143841743469238], "r_norm": [0.21447685360908508, 0.2124914973974228, 0.21530643105506897]}, {"step": 2800, "epoch": 5, "train_loss": 0.08087609708309174, "cos_r_negg": [0.4886190891265869, 0.5083039402961731, 0.5329645872116089], "cos_innovation_negg": [0.4886190891265869, 0.5083039402961731, 0.5329645872116089], "cos_apical_negg": [0.014900158159434795, 0.035297930240631104, 0.10018932074308395], "cos_Ac_negg": [0.611050009727478, 0.608727216720581, 0.6146820783615112], "r_norm": [0.2476438730955124, 0.24570077657699585, 0.24911633133888245]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9683, "test_loss": 0.11789603614807129}, {"step": 2900, "epoch": 6, "train_loss": 0.017274226993322372, "cos_r_negg": [0.47784823179244995, 0.4935106039047241, 0.5223339200019836], "cos_innovation_negg": [0.47784823179244995, 0.4935106039047241, 0.5223339200019836], "cos_apical_negg": [0.020545009523630142, 0.03409266471862793, 0.09988909214735031], "cos_Ac_negg": [0.6047525405883789, 0.6058627367019653, 0.6138401627540588], "r_norm": [0.2350500077009201, 0.23234309256076813, 0.23486503958702087]}, {"step": 3000, "epoch": 6, "train_loss": 0.07949283719062805, "cos_r_negg": [0.45194587111473083, 0.4786764681339264, 0.5070891380310059], "cos_innovation_negg": [0.45194587111473083, 0.4786764681339264, 0.5070891380310059], "cos_apical_negg": [0.000577261671423912, 0.03616706654429436, 0.09920153021812439], "cos_Ac_negg": [0.596362829208374, 0.6020251512527466, 0.6144667863845825], "r_norm": [0.2332625836133957, 0.23132850229740143, 0.23530133068561554], "ldr": {"dL_sdil": -0.008470743894577026, "dL_bp": -0.012227363884449005, "ratio": 0.6927694288505047}}, {"step": 3100, "epoch": 6, "train_loss": 0.011230203323066235, "cos_r_negg": [0.43944498896598816, 0.4663470983505249, 0.5067719221115112], "cos_innovation_negg": [0.43944498896598816, 0.4663470983505249, 0.5067719221115112], "cos_apical_negg": [0.0059204502031207085, 0.0318872407078743, 0.11310937255620956], "cos_Ac_negg": [0.6001513600349426, 0.6119512319564819, 0.6258364915847778], "r_norm": [0.23979894816875458, 0.23688717186450958, 0.24084894359111786]}, {"step": 3200, "epoch": 6, "train_loss": 0.04621240496635437, "cos_r_negg": [0.4342973232269287, 0.46508535742759705, 0.5044590830802917], "cos_innovation_negg": [0.4342973232269287, 0.46508535742759705, 0.5044590830802917], "cos_apical_negg": [0.0012547691585496068, 0.03494802117347717, 0.10854874551296234], "cos_Ac_negg": [0.5974046587944031, 0.6151109337806702, 0.6297565698623657], "r_norm": [0.2083405703306198, 0.20680619776248932, 0.20945681631565094]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9686, "test_loss": 0.1256407434463501}, {"step": 3300, "epoch": 7, "train_loss": 0.014249805361032486, "cos_r_negg": [0.41731202602386475, 0.4580508768558502, 0.4931107461452484], "cos_innovation_negg": [0.41731202602386475, 0.4580508768558502, 0.4931107461452484], "cos_apical_negg": [0.009109636768698692, 0.03295160457491875, 0.10847292840480804], "cos_Ac_negg": [0.5967949628829956, 0.614556074142456, 0.6270172595977783], "r_norm": [0.2500160336494446, 0.24679669737815857, 0.2498304545879364]}, {"step": 3400, "epoch": 7, "train_loss": 0.01112859882414341, "cos_r_negg": [0.38858509063720703, 0.44096070528030396, 0.480546772480011], "cos_innovation_negg": [0.38858509063720703, 0.44096070528030396, 0.480546772480011], "cos_apical_negg": [-0.008255871012806892, 0.03822627291083336, 0.11014938354492188], "cos_Ac_negg": [0.6006128787994385, 0.6202166080474854, 0.6348124742507935], "r_norm": [0.2412058562040329, 0.23865504562854767, 0.2412518858909607]}, {"step": 3500, "epoch": 7, "train_loss": 0.008623255416750908, "cos_r_negg": [0.3743264675140381, 0.4105738401412964, 0.4597768485546112], "cos_innovation_negg": [0.3743264675140381, 0.4105738401412964, 0.4597768485546112], "cos_apical_negg": [0.0028895470313727856, 0.03591875731945038, 0.11270925402641296], "cos_Ac_negg": [0.6082568764686584, 0.6242395639419556, 0.6397649049758911], "r_norm": [0.21341930329799652, 0.21139098703861237, 0.21424329280853271], "ldr": {"dL_sdil": -0.009350374341011047, "dL_bp": -0.0139445960521698, "ratio": 0.6705374831963035}}, {"step": 3600, "epoch": 7, "train_loss": 0.001493149669840932, "cos_r_negg": [0.3852287530899048, 0.42796075344085693, 0.47940123081207275], "cos_innovation_negg": [0.3852287530899048, 0.42796075344085693, 0.47940123081207275], "cos_apical_negg": [-0.00488023366779089, 0.034974534064531326, 0.1126023679971695], "cos_Ac_negg": [0.6145345568656921, 0.6327401995658875, 0.649524986743927], "r_norm": [0.20902538299560547, 0.20602154731750488, 0.2091146558523178]}, {"step": 3700, "epoch": 7, "train_loss": 0.005869025364518166, "cos_r_negg": [0.38824576139450073, 0.4320777654647827, 0.4697416424751282], "cos_innovation_negg": [0.38824576139450073, 0.4320777654647827, 0.4697416424751282], "cos_apical_negg": [-0.006503776181489229, 0.028564482927322388, 0.10319067537784576], "cos_Ac_negg": [0.6190537214279175, 0.6311739087104797, 0.6462790369987488], "r_norm": [0.2119978666305542, 0.20814868807792664, 0.2123369574546814]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9701, "test_loss": 0.12263582782745361}, {"step": 3800, "epoch": 8, "train_loss": 0.0028725494630634785, "cos_r_negg": [0.3476846218109131, 0.3942021131515503, 0.4406149983406067], "cos_innovation_negg": [0.3476846218109131, 0.3942021131515503, 0.4406149983406067], "cos_apical_negg": [-0.003340412862598896, 0.032182492315769196, 0.10314054787158966], "cos_Ac_negg": [0.6167541742324829, 0.6279117465019226, 0.6491929292678833], "r_norm": [0.21939122676849365, 0.21579807996749878, 0.21942919492721558]}, {"step": 3900, "epoch": 8, "train_loss": 0.01645852066576481, "cos_r_negg": [0.3188874423503876, 0.3711095452308655, 0.41697824001312256], "cos_innovation_negg": [0.3188874423503876, 0.3711095452308655, 0.41697824001312256], "cos_apical_negg": [-0.0089386235922575, 0.030950363725423813, 0.10144755244255066], "cos_Ac_negg": [0.6055940985679626, 0.6227110624313354, 0.6511268615722656], "r_norm": [0.225270614027977, 0.22113573551177979, 0.2254658341407776]}, {"step": 4000, "epoch": 8, "train_loss": 0.01775703765451908, "cos_r_negg": [0.33081763982772827, 0.3760502338409424, 0.42436355352401733], "cos_innovation_negg": [0.33081763982772827, 0.3760502338409424, 0.42436355352401733], "cos_apical_negg": [0.004012702964246273, 0.03218439593911171, 0.10053519904613495], "cos_Ac_negg": [0.6183075904846191, 0.6272047758102417, 0.6522946357727051], "r_norm": [0.2212454229593277, 0.21894893050193787, 0.22263962030410767], "ldr": {"dL_sdil": -0.010140083730220795, "dL_bp": -0.017603881657123566, "ratio": 0.5760140818782158}}, {"step": 4100, "epoch": 8, "train_loss": 0.013629444874823093, "cos_r_negg": [0.3223866820335388, 0.37011513113975525, 0.4229600131511688], "cos_innovation_negg": [0.3223866820335388, 0.37011513113975525, 0.4229600131511688], "cos_apical_negg": [0.0003302227705717087, 0.03650972247123718, 0.10796025395393372], "cos_Ac_negg": [0.6143237352371216, 0.6272937059402466, 0.6533287763595581], "r_norm": [0.22584208846092224, 0.22285646200180054, 0.22673724591732025]}, {"step": 4200, "epoch": 8, "train_loss": 0.012107991613447666, "cos_r_negg": [0.28532353043556213, 0.34739142656326294, 0.39834463596343994], "cos_innovation_negg": [0.28532353043556213, 0.34739142656326294, 0.39834463596343994], "cos_apical_negg": [-0.008128422312438488, 0.037260595709085464, 0.10063967853784561], "cos_Ac_negg": [0.6082413196563721, 0.6286255121231079, 0.653980553150177], "r_norm": [0.21755939722061157, 0.21517139673233032, 0.21883633732795715]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9712, "test_loss": 0.13053375539779663}, {"step": 4300, "epoch": 9, "train_loss": 0.006148484535515308, "cos_r_negg": [0.3049679100513458, 0.35393524169921875, 0.4092366099357605], "cos_innovation_negg": [0.3049679100513458, 0.35393524169921875, 0.4092366099357605], "cos_apical_negg": [-0.0028543826192617416, 0.038646213710308075, 0.1020691990852356], "cos_Ac_negg": [0.6200163960456848, 0.6370707750320435, 0.6626262068748474], "r_norm": [0.22027118504047394, 0.21733739972114563, 0.22126659750938416]}, {"step": 4400, "epoch": 9, "train_loss": 0.0008114433148875833, "cos_r_negg": [0.2994518280029297, 0.3576602637767792, 0.4081227481365204], "cos_innovation_negg": [0.2994518280029297, 0.3576602637767792, 0.4081227481365204], "cos_apical_negg": [-0.008132841438055038, 0.039271071553230286, 0.10279826819896698], "cos_Ac_negg": [0.6209456324577332, 0.6397671103477478, 0.6645205616950989], "r_norm": [0.22258585691452026, 0.22009657323360443, 0.223506361246109]}, {"step": 4500, "epoch": 9, "train_loss": 0.022984560579061508, "cos_r_negg": [0.26498422026634216, 0.32743608951568604, 0.3742830753326416], "cos_innovation_negg": [0.26498422026634216, 0.32743608951568604, 0.3742830753326416], "cos_apical_negg": [-0.006952760741114616, 0.03886932134628296, 0.10099170356988907], "cos_Ac_negg": [0.6116693019866943, 0.6377890110015869, 0.6633337736129761], "r_norm": [0.23831143975257874, 0.23460842669010162, 0.23828981816768646], "ldr": {"dL_sdil": -0.011285528540611267, "dL_bp": -0.020764201879501343, "ratio": 0.5435089008526964}}, {"step": 4600, "epoch": 9, "train_loss": 0.0007069878047332168, "cos_r_negg": [0.2655528485774994, 0.3202352523803711, 0.3710695505142212], "cos_innovation_negg": [0.2655528485774994, 0.3202352523803711, 0.3710695505142212], "cos_apical_negg": [0.0013217898085713387, 0.03834256902337074, 0.10062165558338165], "cos_Ac_negg": [0.6067532300949097, 0.6318719387054443, 0.6621993780136108], "r_norm": [0.2354167103767395, 0.2314644306898117, 0.23491345345973969]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9727, "test_loss": 0.12967014846801758}, {"step": 4700, "epoch": 10, "train_loss": 0.0023708001244813204, "cos_r_negg": [0.24326619505882263, 0.2900390625, 0.34820228815078735], "cos_innovation_negg": [0.24326619505882263, 0.2900390625, 0.34820228815078735], "cos_apical_negg": [-3.511086106300354e-06, 0.032490044832229614, 0.0980047732591629], "cos_Ac_negg": [0.6065478324890137, 0.6289293169975281, 0.6626633405685425], "r_norm": [0.2016417682170868, 0.19869792461395264, 0.20257467031478882]}, {"step": 4800, "epoch": 10, "train_loss": 0.0020702225156128407, "cos_r_negg": [0.24571841955184937, 0.2944677174091339, 0.3514004349708557], "cos_innovation_negg": [0.24571841955184937, 0.2944677174091339, 0.3514004349708557], "cos_apical_negg": [-0.002526048105210066, 0.03226470202207565, 0.09737521409988403], "cos_Ac_negg": [0.6093655228614807, 0.6314992904663086, 0.6656678318977356], "r_norm": [0.20281454920768738, 0.19931988418102264, 0.20342528820037842]}, {"step": 4900, "epoch": 10, "train_loss": 0.0019158375216647983, "cos_r_negg": [0.24141237139701843, 0.28968027234077454, 0.34494662284851074], "cos_innovation_negg": [0.24141237139701843, 0.28968027234077454, 0.34494662284851074], "cos_apical_negg": [-0.0023834090679883957, 0.033177558332681656, 0.09725464880466461], "cos_Ac_negg": [0.6087672114372253, 0.6332275867462158, 0.6685798168182373], "r_norm": [0.20675504207611084, 0.20340441167354584, 0.20745927095413208]}, {"step": 5000, "epoch": 10, "train_loss": 0.0008548484183847904, "cos_r_negg": [0.23028664290905, 0.28492194414138794, 0.33767589926719666], "cos_innovation_negg": [0.23028664290905, 0.28492194414138794, 0.33767589926719666], "cos_apical_negg": [-0.0038913972675800323, 0.03447958081960678, 0.09226565808057785], "cos_Ac_negg": [0.6001114845275879, 0.6293758749961853, 0.6673866510391235], "r_norm": [0.20831169188022614, 0.2051803320646286, 0.20866748690605164], "ldr": {"dL_sdil": -0.012155190110206604, "dL_bp": -0.0229220911860466, "ratio": 0.5302827744444997}}, {"step": 5100, "epoch": 10, "train_loss": 0.0010202466510236263, "cos_r_negg": [0.2341373860836029, 0.28416383266448975, 0.340952068567276], "cos_innovation_negg": [0.2341373860836029, 0.28416383266448975, 0.340952068567276], "cos_apical_negg": [-0.00420326367020607, 0.033878736197948456, 0.09707412123680115], "cos_Ac_negg": [0.607232391834259, 0.6333211660385132, 0.6719170212745667], "r_norm": [0.20714673399925232, 0.2038562297821045, 0.2072623074054718]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9724, "test_loss": 0.13474912872314454}, {"step": 5200, "epoch": 11, "train_loss": 0.0006876580300740898, "cos_r_negg": [0.21897663176059723, 0.27664655447006226, 0.33035069704055786], "cos_innovation_negg": [0.21897663176059723, 0.27664655447006226, 0.33035069704055786], "cos_apical_negg": [-0.005986626259982586, 0.03328913077712059, 0.09508834779262543], "cos_Ac_negg": [0.5988229513168335, 0.6279399394989014, 0.672398567199707], "r_norm": [0.20531290769577026, 0.20248490571975708, 0.20603346824645996]}, {"step": 5300, "epoch": 11, "train_loss": 0.0012513920664787292, "cos_r_negg": [0.2129746377468109, 0.26975369453430176, 0.3232852816581726], "cos_innovation_negg": [0.2129746377468109, 0.26975369453430176, 0.3232852816581726], "cos_apical_negg": [-0.005739088170230389, 0.03182534873485565, 0.09561654925346375], "cos_Ac_negg": [0.5997483730316162, 0.6285946369171143, 0.6708213090896606], "r_norm": [0.20247995853424072, 0.19919392466545105, 0.20331595838069916]}, {"step": 5400, "epoch": 11, "train_loss": 0.002956363372504711, "cos_r_negg": [0.2116379737854004, 0.2628961205482483, 0.31698405742645264], "cos_innovation_negg": [0.2116379737854004, 0.2628961205482483, 0.31698405742645264], "cos_apical_negg": [-0.0015459386631846428, 0.03349905461072922, 0.09399515390396118], "cos_Ac_negg": [0.6028150916099548, 0.6296771168708801, 0.6726018190383911], "r_norm": [0.2075035572052002, 0.20453691482543945, 0.20845350623130798]}, {"step": 5500, "epoch": 11, "train_loss": 0.0026475482154637575, "cos_r_negg": [0.20966067910194397, 0.26422518491744995, 0.31851285696029663], "cos_innovation_negg": [0.20966067910194397, 0.26422518491744995, 0.31851285696029663], "cos_apical_negg": [-0.0028996989130973816, 0.03485501557588577, 0.09448803961277008], "cos_Ac_negg": [0.6050945520401001, 0.6289932727813721, 0.6731719970703125], "r_norm": [0.21088801324367523, 0.20764297246932983, 0.2111128866672516], "ldr": {"dL_sdil": -0.011994272470474243, "dL_bp": -0.023104965686798096, "ratio": 0.5191209817432288}}, {"step": 5600, "epoch": 11, "train_loss": 0.00218569696880877, "cos_r_negg": [0.19715338945388794, 0.25238174200057983, 0.30352944135665894], "cos_innovation_negg": [0.19715338945388794, 0.25238174200057983, 0.30352944135665894], "cos_apical_negg": [-0.004069292917847633, 0.033313825726509094, 0.09001317620277405], "cos_Ac_negg": [0.6001343727111816, 0.6257193088531494, 0.6720417141914368], "r_norm": [0.2028658092021942, 0.19982746243476868, 0.20375166833400726]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9724, "test_loss": 0.13916511611938476}, {"step": 5700, "epoch": 12, "train_loss": 0.0020794891752302647, "cos_r_negg": [0.20272225141525269, 0.2574833333492279, 0.307792991399765], "cos_innovation_negg": [0.20272225141525269, 0.2574833333492279, 0.307792991399765], "cos_apical_negg": [-0.002199282869696617, 0.03414363041520119, 0.09020665287971497], "cos_Ac_negg": [0.6050809621810913, 0.6289219260215759, 0.6739180684089661], "r_norm": [0.20607653260231018, 0.2031911015510559, 0.20651839673519135]}, {"step": 5800, "epoch": 12, "train_loss": 0.00020566982857417315, "cos_r_negg": [0.19398340582847595, 0.2503836154937744, 0.3005009889602661], "cos_innovation_negg": [0.19398340582847595, 0.2503836154937744, 0.3005009889602661], "cos_apical_negg": [-0.004616866819560528, 0.03416449576616287, 0.08954647183418274], "cos_Ac_negg": [0.6008589267730713, 0.6273083686828613, 0.6730796694755554], "r_norm": [0.20475776493549347, 0.20183506608009338, 0.20556774735450745]}, {"step": 5900, "epoch": 12, "train_loss": 0.0018817087402567267, "cos_r_negg": [0.19192993640899658, 0.24909460544586182, 0.29923221468925476], "cos_innovation_negg": [0.19192993640899658, 0.24909460544586182, 0.29923221468925476], "cos_apical_negg": [-0.004756544716656208, 0.0348251648247242, 0.08955187350511551], "cos_Ac_negg": [0.5998852252960205, 0.6275185346603394, 0.6741136312484741], "r_norm": [0.20769557356834412, 0.2047315537929535, 0.20835444331169128]}, {"step": 6000, "epoch": 12, "train_loss": 0.0005604384932667017, "cos_r_negg": [0.18731625378131866, 0.2441646009683609, 0.2948216199874878], "cos_innovation_negg": [0.18731625378131866, 0.2441646009683609, 0.2948216199874878], "cos_apical_negg": [-0.00470914039760828, 0.03447927534580231, 0.08883833885192871], "cos_Ac_negg": [0.599158525466919, 0.6261067390441895, 0.6752423048019409], "r_norm": [0.20639361441135406, 0.2033909112215042, 0.20716290175914764], "ldr": {"dL_sdil": -0.01247800886631012, "dL_bp": -0.02522388845682144, "ratio": 0.49469013818666885}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9723, "test_loss": 0.14235655269622802}, {"step": 6100, "epoch": 13, "train_loss": 0.0007953558815643191, "cos_r_negg": [0.18446385860443115, 0.2396831214427948, 0.290547251701355], "cos_innovation_negg": [0.18446385860443115, 0.2396831214427948, 0.290547251701355], "cos_apical_negg": [-0.002962573431432247, 0.033410150557756424, 0.08906933665275574], "cos_Ac_negg": [0.5952847599983215, 0.6237782835960388, 0.674699068069458], "r_norm": [0.20510908961296082, 0.20224229991436005, 0.2059614062309265]}, {"step": 6200, "epoch": 13, "train_loss": 0.0007006739033386111, "cos_r_negg": [0.17903070151805878, 0.2364688217639923, 0.28458109498023987], "cos_innovation_negg": [0.17903070151805878, 0.2364688217639923, 0.28458109498023987], "cos_apical_negg": [-0.0029544783756136894, 0.033886633813381195, 0.0861949771642685], "cos_Ac_negg": [0.5941613912582397, 0.6231882572174072, 0.6737856268882751], "r_norm": [0.20340999960899353, 0.20053435862064362, 0.20420712232589722]}, {"step": 6300, "epoch": 13, "train_loss": 0.0003973285492975265, "cos_r_negg": [0.17568711936473846, 0.231784388422966, 0.28029865026474], "cos_innovation_negg": [0.17568711936473846, 0.231784388422966, 0.28029865026474], "cos_apical_negg": [-0.002083425410091877, 0.03384298086166382, 0.08610855042934418], "cos_Ac_negg": [0.5944377183914185, 0.6225796937942505, 0.6730598211288452], "r_norm": [0.20646843314170837, 0.20356518030166626, 0.2072640061378479]}, {"step": 6400, "epoch": 13, "train_loss": 0.0006554600549861789, "cos_r_negg": [0.17507170140743256, 0.23097488284111023, 0.2792336940765381], "cos_innovation_negg": [0.17507170140743256, 0.23097488284111023, 0.2792336940765381], "cos_apical_negg": [-0.002629481256008148, 0.03345771133899689, 0.08591315150260925], "cos_Ac_negg": [0.5959250330924988, 0.6225063800811768, 0.6728956699371338], "r_norm": [0.20461906492710114, 0.2017722874879837, 0.20540666580200195]}, {"step": 6500, "epoch": 13, "train_loss": 0.0001446559908799827, "cos_r_negg": [0.17442870140075684, 0.22965818643569946, 0.27639836072921753], "cos_innovation_negg": [0.17442870140075684, 0.22965818643569946, 0.27639836072921753], "cos_apical_negg": [-0.003029278479516506, 0.03396303951740265, 0.08407631516456604], "cos_Ac_negg": [0.5945233106613159, 0.6232552528381348, 0.6756317615509033], "r_norm": [0.20908938348293304, 0.20641925930976868, 0.20985308289527893], "ldr": {"dL_sdil": -0.013210579752922058, "dL_bp": -0.027561113238334656, "ratio": 0.4793195267071981}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9726, "test_loss": 0.1455043478012085}, {"step": 6600, "epoch": 14, "train_loss": 0.0004824562347494066, "cos_r_negg": [0.1671731024980545, 0.22329702973365784, 0.2709178328514099], "cos_innovation_negg": [0.1671731024980545, 0.22329702973365784, 0.2709178328514099], "cos_apical_negg": [-0.004154294729232788, 0.03307810425758362, 0.08348521590232849], "cos_Ac_negg": [0.5958958268165588, 0.6226828098297119, 0.6753007173538208], "r_norm": [0.20208407938480377, 0.19935184717178345, 0.20300203561782837]}, {"step": 6700, "epoch": 14, "train_loss": 0.0008160572033375502, "cos_r_negg": [0.16498219966888428, 0.22207579016685486, 0.2698427140712738], "cos_innovation_negg": [0.16498219966888428, 0.22207579016685486, 0.2698427140712738], "cos_apical_negg": [-0.0052373334765434265, 0.03249061107635498, 0.08310656994581223], "cos_Ac_negg": [0.5944553017616272, 0.6222578883171082, 0.6761384606361389], "r_norm": [0.2044653594493866, 0.20145821571350098, 0.20519861578941345]}, {"step": 6800, "epoch": 14, "train_loss": 0.0009065538179129362, "cos_r_negg": [0.1594417840242386, 0.21674184501171112, 0.2647378146648407], "cos_innovation_negg": [0.1594417840242386, 0.21674184501171112, 0.2647378146648407], "cos_apical_negg": [-0.004249837249517441, 0.032414667308330536, 0.08174674957990646], "cos_Ac_negg": [0.5946006774902344, 0.6216984987258911, 0.6755374073982239], "r_norm": [0.20176371932029724, 0.19887161254882812, 0.2025604546070099]}, {"step": 6900, "epoch": 14, "train_loss": 0.0006018164567649364, "cos_r_negg": [0.15775761008262634, 0.2153831422328949, 0.26428309082984924], "cos_innovation_negg": [0.15775761008262634, 0.2153831422328949, 0.26428309082984924], "cos_apical_negg": [-0.005284152925014496, 0.03316710144281387, 0.08256210386753082], "cos_Ac_negg": [0.5963074564933777, 0.6216413378715515, 0.6754178404808044], "r_norm": [0.20030176639556885, 0.19743712246418, 0.20103485882282257]}, {"step": 7000, "epoch": 14, "train_loss": 0.0007463142392225564, "cos_r_negg": [0.16179631650447845, 0.21716907620429993, 0.26553523540496826], "cos_innovation_negg": [0.16179631650447845, 0.21716907620429993, 0.26553523540496826], "cos_apical_negg": [-0.0037459516897797585, 0.03344787657260895, 0.08418938517570496], "cos_Ac_negg": [0.5923255681991577, 0.6215612888336182, 0.6758936047554016], "r_norm": [0.20400476455688477, 0.20118750631809235, 0.20500923693180084], "ldr": {"dL_sdil": -0.012740954756736755, "dL_bp": -0.02715706080198288, "ratio": 0.4691580892953803}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9725, "test_loss": 0.14923210258483888}], "final": {"test_acc": 0.9725, "test_loss": 0.14923210258483888, "wall_s": 29.01899290084839, "cos_r_negg": [0.16006410121917725, 0.2159397006034851, 0.26292920112609863], "cos_innovation_negg": [0.16006410121917725, 0.2159397006034851, 0.26292920112609863], "cos_apical_negg": [-0.0034941332414746284, 0.03366870433092117, 0.08231043815612793], "cos_Ac_negg": [0.5925276875495911, 0.6217751502990723, 0.6762490272521973]}}
|