1
|
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.05, "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_rho0p05_raw_s0", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3603768348693848, "cos_r_negg": [-0.018889345228672028, -0.02601044997572899, 0.032917968928813934], "cos_innovation_negg": [-0.01886402815580368, -0.02586054801940918, 0.03356533125042915], "cos_apical_negg": [-0.018889345228672028, -0.02601044997572899, 0.032917968928813934], "cos_Ac_negg": [-0.018863998353481293, -0.02586054801940918, 0.03356531634926796], "r_norm": [4.982118606567383, 4.8570404052734375, 4.847076892852783], "ldr": {"dL_sdil": -0.008685588836669922, "dL_bp": -0.23511672019958496, "ratio": 0.03694160427764105}}, {"step": 100, "epoch": 0, "train_loss": 0.2316625416278839, "cos_r_negg": [0.23470458388328552, 0.23169171810150146, 0.19987548887729645], "cos_innovation_negg": [0.45221105217933655, 0.43953999876976013, 0.4016229808330536], "cos_apical_negg": [0.23470458388328552, 0.23169171810150146, 0.19987548887729645], "cos_Ac_negg": [0.45217204093933105, 0.43955740332603455, 0.40168216824531555], "r_norm": [1.4844417572021484, 1.4598970413208008, 1.4545302391052246]}, {"step": 200, "epoch": 0, "train_loss": 0.3199096918106079, "cos_r_negg": [0.19267380237579346, 0.24690712988376617, 0.24164065718650818], "cos_innovation_negg": [0.4661179780960083, 0.46744492650032043, 0.41031360626220703], "cos_apical_negg": [0.19267380237579346, 0.24690712988376617, 0.24164065718650818], "cos_Ac_negg": [0.4661034941673279, 0.4674678444862366, 0.41032806038856506], "r_norm": [1.3621758222579956, 1.3459157943725586, 1.341383934020996]}, {"step": 300, "epoch": 0, "train_loss": 0.22720038890838623, "cos_r_negg": [0.20497924089431763, 0.2550729513168335, 0.2500893175601959], "cos_innovation_negg": [0.4957374930381775, 0.5101429224014282, 0.4892195165157318], "cos_apical_negg": [0.20497924089431763, 0.2550729513168335, 0.2500893175601959], "cos_Ac_negg": [0.4957354664802551, 0.5101265907287598, 0.48918506503105164], "r_norm": [1.2987923622131348, 1.2880769968032837, 1.2851369380950928]}, {"step": 400, "epoch": 0, "train_loss": 0.16089418530464172, "cos_r_negg": [0.19641119241714478, 0.22100770473480225, 0.22859039902687073], "cos_innovation_negg": [0.532158374786377, 0.5136614441871643, 0.4884665608406067], "cos_apical_negg": [0.19641119241714478, 0.22100770473480225, 0.22859039902687073], "cos_Ac_negg": [0.5320866703987122, 0.5137494802474976, 0.488487184047699], "r_norm": [1.2419109344482422, 1.2315348386764526, 1.231013536453247]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9414, "test_loss": 0.21403483505249024}, {"step": 500, "epoch": 1, "train_loss": 0.22588635981082916, "cos_r_negg": [0.14134198427200317, 0.2137194722890854, 0.2176074981689453], "cos_innovation_negg": [0.4921151399612427, 0.48354625701904297, 0.46166861057281494], "cos_apical_negg": [0.14134198427200317, 0.2137194722890854, 0.2176074981689453], "cos_Ac_negg": [0.49202507734298706, 0.4835682809352875, 0.46145182847976685], "r_norm": [1.2109417915344238, 1.200516939163208, 1.1987390518188477], "ldr": {"dL_sdil": -0.008084222674369812, "dL_bp": -0.006889566779136658, "ratio": 1.1734007280183238}}, {"step": 600, "epoch": 1, "train_loss": 0.1248762458562851, "cos_r_negg": [0.16898371279239655, 0.2130759358406067, 0.29900920391082764], "cos_innovation_negg": [0.47226765751838684, 0.47323399782180786, 0.5102213621139526], "cos_apical_negg": [0.16898371279239655, 0.2130759358406067, 0.29900920391082764], "cos_Ac_negg": [0.47219038009643555, 0.47315865755081177, 0.5101075768470764], "r_norm": [1.2194931507110596, 1.2160041332244873, 1.207112193107605]}, {"step": 700, "epoch": 1, "train_loss": 0.19223998486995697, "cos_r_negg": [0.14750772714614868, 0.20457768440246582, 0.2899758815765381], "cos_innovation_negg": [0.4767991900444031, 0.46232789754867554, 0.49343907833099365], "cos_apical_negg": [0.14750772714614868, 0.20457768440246582, 0.2899758815765381], "cos_Ac_negg": [0.47682324051856995, 0.4623897969722748, 0.4933580756187439], "r_norm": [1.2216745615005493, 1.2105355262756348, 1.20802903175354]}, {"step": 800, "epoch": 1, "train_loss": 0.15692606568336487, "cos_r_negg": [0.13524428009986877, 0.21434614062309265, 0.2651273012161255], "cos_innovation_negg": [0.4759499728679657, 0.4723067879676819, 0.476607084274292], "cos_apical_negg": [0.13524428009986877, 0.21434614062309265, 0.2651273012161255], "cos_Ac_negg": [0.47590887546539307, 0.47228991985321045, 0.47643569111824036], "r_norm": [1.1303777694702148, 1.1217707395553589, 1.1227798461914062]}, {"step": 900, "epoch": 1, "train_loss": 0.17302459478378296, "cos_r_negg": [0.1526738405227661, 0.2253347635269165, 0.3220912218093872], "cos_innovation_negg": [0.5346554517745972, 0.5626517534255981, 0.566016674041748], "cos_apical_negg": [0.1526738405227661, 0.2253347635269165, 0.3220912218093872], "cos_Ac_negg": [0.534640908241272, 0.5626353025436401, 0.5658434629440308], "r_norm": [1.147991418838501, 1.1383893489837646, 1.1372339725494385]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9454, "test_loss": 0.20151613082885741}, {"step": 1000, "epoch": 2, "train_loss": 0.05331810563802719, "cos_r_negg": [0.11788767576217651, 0.17947617173194885, 0.35381388664245605], "cos_innovation_negg": [0.4849860966205597, 0.5148710012435913, 0.5581153035163879], "cos_apical_negg": [0.11788767576217651, 0.17947617173194885, 0.35381388664245605], "cos_Ac_negg": [0.48491957783699036, 0.5147826671600342, 0.5579615831375122], "r_norm": [1.1214679479599, 1.1133793592453003, 1.1112723350524902], "ldr": {"dL_sdil": -0.007101535797119141, "dL_bp": -0.0070992112159729, "ratio": 1.0003274421728727}}, {"step": 1100, "epoch": 2, "train_loss": 0.1732659488916397, "cos_r_negg": [0.11807993799448013, 0.17467465996742249, 0.34313422441482544], "cos_innovation_negg": [0.47264716029167175, 0.48618900775909424, 0.5216293334960938], "cos_apical_negg": [0.11807993799448013, 0.17467465996742249, 0.34313422441482544], "cos_Ac_negg": [0.4726192355155945, 0.48613595962524414, 0.5215154886245728], "r_norm": [1.1383291482925415, 1.1279253959655762, 1.1292471885681152]}, {"step": 1200, "epoch": 2, "train_loss": 0.21508166193962097, "cos_r_negg": [0.11980476975440979, 0.157465398311615, 0.2798289656639099], "cos_innovation_negg": [0.4708786606788635, 0.4826098084449768, 0.5160412788391113], "cos_apical_negg": [0.11980476975440979, 0.157465398311615, 0.2798289656639099], "cos_Ac_negg": [0.470876544713974, 0.4825723469257355, 0.5159133076667786], "r_norm": [1.1140010356903076, 1.1040006875991821, 1.1062731742858887]}, {"step": 1300, "epoch": 2, "train_loss": 0.14503014087677002, "cos_r_negg": [0.15452975034713745, 0.17584452033042908, 0.3179256021976471], "cos_innovation_negg": [0.4997187852859497, 0.5223895311355591, 0.47794729471206665], "cos_apical_negg": [0.15452975034713745, 0.17584452033042908, 0.3179256021976471], "cos_Ac_negg": [0.4996158182621002, 0.5223785042762756, 0.47774845361709595], "r_norm": [1.112052321434021, 1.1025787591934204, 1.1016534566879272]}, {"step": 1400, "epoch": 2, "train_loss": 0.17568084597587585, "cos_r_negg": [0.09762503951787949, 0.1793743073940277, 0.3626384437084198], "cos_innovation_negg": [0.5365784764289856, 0.561907172203064, 0.5673563480377197], "cos_apical_negg": [0.09762503951787949, 0.1793743073940277, 0.3626384437084198], "cos_Ac_negg": [0.5362609624862671, 0.5614597797393799, 0.566525936126709], "r_norm": [1.1025522947311401, 1.095109224319458, 1.0913634300231934]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9572, "test_loss": 0.16043873100280762}, {"step": 1500, "epoch": 3, "train_loss": 0.1210009902715683, "cos_r_negg": [0.11008252948522568, 0.22029989957809448, 0.3482934832572937], "cos_innovation_negg": [0.48705393075942993, 0.5561088919639587, 0.5838615894317627], "cos_apical_negg": [0.11008252948522568, 0.22029989957809448, 0.3482934832572937], "cos_Ac_negg": [0.4869464337825775, 0.5559383630752563, 0.5836822986602783], "r_norm": [1.069291353225708, 1.0639028549194336, 1.0641794204711914], "ldr": {"dL_sdil": -0.0044784098863601685, "dL_bp": -0.004492044448852539, "ratio": 0.9969647311713815}}, {"step": 1600, "epoch": 3, "train_loss": 0.09388969838619232, "cos_r_negg": [0.08775987476110458, 0.15973791480064392, 0.25329574942588806], "cos_innovation_negg": [0.4820828139781952, 0.5090473890304565, 0.53265780210495], "cos_apical_negg": [0.08775987476110458, 0.15973791480064392, 0.25329574942588806], "cos_Ac_negg": [0.4820255935192108, 0.5089107751846313, 0.5325554609298706], "r_norm": [1.07172429561615, 1.0621823072433472, 1.062644124031067]}, {"step": 1700, "epoch": 3, "train_loss": 0.07022413611412048, "cos_r_negg": [0.10848211497068405, 0.17679093778133392, 0.29205521941185], "cos_innovation_negg": [0.4823445677757263, 0.5424801111221313, 0.5775178074836731], "cos_apical_negg": [0.10848211497068405, 0.17679093778133392, 0.29205521941185], "cos_Ac_negg": [0.4822928309440613, 0.542522668838501, 0.5774283409118652], "r_norm": [1.089035153388977, 1.0836055278778076, 1.0805248022079468]}, {"step": 1800, "epoch": 3, "train_loss": 0.038585830479860306, "cos_r_negg": [0.10026787966489792, 0.1584472954273224, 0.2623743414878845], "cos_innovation_negg": [0.4616909325122833, 0.49445784091949463, 0.5743393898010254], "cos_apical_negg": [0.10026787966489792, 0.1584472954273224, 0.2623743414878845], "cos_Ac_negg": [0.4615846872329712, 0.4943014979362488, 0.5740854740142822], "r_norm": [1.0575604438781738, 1.051081657409668, 1.0498106479644775]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9567, "test_loss": 0.1669241989135742}, {"step": 1900, "epoch": 4, "train_loss": 0.05220799520611763, "cos_r_negg": [0.12107962369918823, 0.19493287801742554, 0.3175318241119385], "cos_innovation_negg": [0.49412333965301514, 0.5541813373565674, 0.576842725276947], "cos_apical_negg": [0.12107962369918823, 0.19493287801742554, 0.3175318241119385], "cos_Ac_negg": [0.49395525455474854, 0.5539348125457764, 0.576566219329834], "r_norm": [1.04732346534729, 1.0417029857635498, 1.0409610271453857]}, {"step": 2000, "epoch": 4, "train_loss": 0.1620032787322998, "cos_r_negg": [0.09274173527956009, 0.18109172582626343, 0.2818826138973236], "cos_innovation_negg": [0.4649173319339752, 0.48723775148391724, 0.5584830045700073], "cos_apical_negg": [0.09274173527956009, 0.18109172582626343, 0.2818826138973236], "cos_Ac_negg": [0.46477028727531433, 0.4870375394821167, 0.5584304928779602], "r_norm": [1.039423942565918, 1.035604476928711, 1.0339125394821167], "ldr": {"dL_sdil": -0.005337387323379517, "dL_bp": -0.007765054702758789, "ratio": 0.6873599128005158}}, {"step": 2100, "epoch": 4, "train_loss": 0.05869222432374954, "cos_r_negg": [0.13352806866168976, 0.21133017539978027, 0.2777782082557678], "cos_innovation_negg": [0.44825026392936707, 0.48020094633102417, 0.5685089230537415], "cos_apical_negg": [0.13352806866168976, 0.21133017539978027, 0.2777782082557678], "cos_Ac_negg": [0.44824129343032837, 0.4801056683063507, 0.568452000617981], "r_norm": [1.0415667295455933, 1.0398322343826294, 1.037940502166748]}, {"step": 2200, "epoch": 4, "train_loss": 0.05471344292163849, "cos_r_negg": [0.08597137033939362, 0.17008939385414124, 0.25793978571891785], "cos_innovation_negg": [0.4834924042224884, 0.5145854949951172, 0.6063051819801331], "cos_apical_negg": [0.08597137033939362, 0.17008939385414124, 0.25793978571891785], "cos_Ac_negg": [0.4834393262863159, 0.514431357383728, 0.6061898469924927], "r_norm": [1.0674641132354736, 1.0618911981582642, 1.0563780069351196]}, {"step": 2300, "epoch": 4, "train_loss": 0.29510194063186646, "cos_r_negg": [0.11807355284690857, 0.20011469721794128, 0.29283344745635986], "cos_innovation_negg": [0.4751397371292114, 0.5576819181442261, 0.6443016529083252], "cos_apical_negg": [0.11807355284690857, 0.20011469721794128, 0.29283344745635986], "cos_Ac_negg": [0.47499263286590576, 0.5572212934494019, 0.6441560387611389], "r_norm": [1.061295747756958, 1.0528438091278076, 1.0521645545959473]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9605, "test_loss": 0.15810680198669433}, {"step": 2400, "epoch": 5, "train_loss": 0.11830407381057739, "cos_r_negg": [0.06817568093538284, 0.1392599493265152, 0.30653971433639526], "cos_innovation_negg": [0.43076619505882263, 0.5064486861228943, 0.6709274053573608], "cos_apical_negg": [0.06817568093538284, 0.1392599493265152, 0.30653971433639526], "cos_Ac_negg": [0.4307430684566498, 0.5063877105712891, 0.6708958148956299], "r_norm": [1.0599803924560547, 1.0531288385391235, 1.0495966672897339]}, {"step": 2500, "epoch": 5, "train_loss": 0.03706478327512741, "cos_r_negg": [0.09750305861234665, 0.18643105030059814, 0.2588551640510559], "cos_innovation_negg": [0.45465922355651855, 0.5231276750564575, 0.6625532507896423], "cos_apical_negg": [0.09750305861234665, 0.18643105030059814, 0.2588551640510559], "cos_Ac_negg": [0.4545869827270508, 0.5228502750396729, 0.6626551747322083], "r_norm": [1.0685884952545166, 1.062732458114624, 1.0581591129302979], "ldr": {"dL_sdil": -0.008816704154014587, "dL_bp": -0.010646924376487732, "ratio": 0.8280986923777787}}, {"step": 2600, "epoch": 5, "train_loss": 0.0740327388048172, "cos_r_negg": [0.09968943893909454, 0.18072400987148285, 0.26391273736953735], "cos_innovation_negg": [0.4368077516555786, 0.5317527055740356, 0.7128058671951294], "cos_apical_negg": [0.09968943893909454, 0.18072400987148285, 0.26391273736953735], "cos_Ac_negg": [0.4366735517978668, 0.5316020250320435, 0.7130361795425415], "r_norm": [1.0688101053237915, 1.0605977773666382, 1.0566633939743042]}, {"step": 2700, "epoch": 5, "train_loss": 0.0685981884598732, "cos_r_negg": [0.11481891572475433, 0.20972947776317596, 0.2915850877761841], "cos_innovation_negg": [0.43444502353668213, 0.5356224179267883, 0.6728551387786865], "cos_apical_negg": [0.11481891572475433, 0.20972947776317596, 0.2915850877761841], "cos_Ac_negg": [0.4344581663608551, 0.5355129837989807, 0.6728971004486084], "r_norm": [1.0446279048919678, 1.040521264076233, 1.038030743598938]}, {"step": 2800, "epoch": 5, "train_loss": 0.08406370878219604, "cos_r_negg": [0.11313486844301224, 0.15717515349388123, 0.3214162588119507], "cos_innovation_negg": [0.4879230856895447, 0.582575798034668, 0.6949892044067383], "cos_apical_negg": [0.11313486844301224, 0.15717515349388123, 0.3214162588119507], "cos_Ac_negg": [0.4878532290458679, 0.5824505090713501, 0.6949414014816284], "r_norm": [1.0404664278030396, 1.0348474979400635, 1.0308709144592285]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9608, "test_loss": 0.1572404094696045}, {"step": 2900, "epoch": 6, "train_loss": 0.023374833166599274, "cos_r_negg": [0.08860492706298828, 0.15452654659748077, 0.2529580593109131], "cos_innovation_negg": [0.4613765478134155, 0.5762410163879395, 0.6754223704338074], "cos_apical_negg": [0.08860492706298828, 0.15452654659748077, 0.2529580593109131], "cos_Ac_negg": [0.4612863063812256, 0.5760371685028076, 0.6753679513931274], "r_norm": [1.0332884788513184, 1.0295827388763428, 1.0252538919448853]}, {"step": 3000, "epoch": 6, "train_loss": 0.25347787141799927, "cos_r_negg": [0.08315202593803406, 0.1285639852285385, 0.20180664956569672], "cos_innovation_negg": [0.41670238971710205, 0.4907757043838501, 0.6787306666374207], "cos_apical_negg": [0.08315202593803406, 0.1285639852285385, 0.20180664956569672], "cos_Ac_negg": [0.4165520966053009, 0.4903995394706726, 0.6787961721420288], "r_norm": [1.0317394733428955, 1.0286509990692139, 1.022667646408081], "ldr": {"dL_sdil": -0.006295233964920044, "dL_bp": -0.006876260042190552, "ratio": 0.9155026026203902}}, {"step": 3100, "epoch": 6, "train_loss": 0.09500887989997864, "cos_r_negg": [0.060014963150024414, 0.15525293350219727, 0.2267547845840454], "cos_innovation_negg": [0.42980897426605225, 0.5274608135223389, 0.6904079914093018], "cos_apical_negg": [0.060014963150024414, 0.15525293350219727, 0.2267547845840454], "cos_Ac_negg": [0.4297827482223511, 0.5274139046669006, 0.6906270980834961], "r_norm": [1.0476242303848267, 1.0429370403289795, 1.0376466512680054]}, {"step": 3200, "epoch": 6, "train_loss": 0.10645434260368347, "cos_r_negg": [0.0909779965877533, 0.12043160200119019, 0.22367516160011292], "cos_innovation_negg": [0.45399796962738037, 0.5186017751693726, 0.7197713851928711], "cos_apical_negg": [0.0909779965877533, 0.12043160200119019, 0.22367516160011292], "cos_Ac_negg": [0.4539446234703064, 0.518485426902771, 0.7198947072029114], "r_norm": [1.0119233131408691, 1.007584571838379, 1.0031044483184814]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9634, "test_loss": 0.14686290378570557}, {"step": 3300, "epoch": 7, "train_loss": 0.054971843957901, "cos_r_negg": [0.0789458379149437, 0.1460288166999817, 0.25119268894195557], "cos_innovation_negg": [0.4534943997859955, 0.5249248743057251, 0.7382019758224487], "cos_apical_negg": [0.0789458379149437, 0.1460288166999817, 0.25119268894195557], "cos_Ac_negg": [0.45348697900772095, 0.5246008038520813, 0.7382972240447998], "r_norm": [1.0166678428649902, 1.0093374252319336, 1.007851004600525]}, {"step": 3400, "epoch": 7, "train_loss": 0.10781344026327133, "cos_r_negg": [0.07544979453086853, 0.12924110889434814, 0.18920454382896423], "cos_innovation_negg": [0.4369913637638092, 0.5438998937606812, 0.7515773773193359], "cos_apical_negg": [0.07544979453086853, 0.12924110889434814, 0.18920454382896423], "cos_Ac_negg": [0.43686461448669434, 0.5438346266746521, 0.7518844604492188], "r_norm": [1.0381009578704834, 1.0323768854141235, 1.0269943475723267]}, {"step": 3500, "epoch": 7, "train_loss": 0.0555838942527771, "cos_r_negg": [0.050943776965141296, 0.12314064055681229, 0.24524903297424316], "cos_innovation_negg": [0.424956738948822, 0.5067443251609802, 0.7101840972900391], "cos_apical_negg": [0.050943776965141296, 0.12314064055681229, 0.24524903297424316], "cos_Ac_negg": [0.42496949434280396, 0.5064550638198853, 0.7104209661483765], "r_norm": [1.0190849304199219, 1.013673186302185, 1.0090439319610596], "ldr": {"dL_sdil": -0.0051877498626708984, "dL_bp": -0.00829671323299408, "ratio": 0.6252777114243789}}, {"step": 3600, "epoch": 7, "train_loss": 0.016231929883360863, "cos_r_negg": [0.04102472960948944, 0.11987172067165375, 0.23962877690792084], "cos_innovation_negg": [0.39390552043914795, 0.4806842803955078, 0.708020806312561], "cos_apical_negg": [0.04102472960948944, 0.11987172067165375, 0.23962877690792084], "cos_Ac_negg": [0.3938722014427185, 0.4802889823913574, 0.7081338167190552], "r_norm": [1.033786654472351, 1.0289814472198486, 1.024149775505066]}, {"step": 3700, "epoch": 7, "train_loss": 0.07463322579860687, "cos_r_negg": [0.06463445723056793, 0.16267982125282288, 0.2788849472999573], "cos_innovation_negg": [0.4540909230709076, 0.5231131315231323, 0.7329192161560059], "cos_apical_negg": [0.06463445723056793, 0.16267982125282288, 0.2788849472999573], "cos_Ac_negg": [0.4540967345237732, 0.5229589939117432, 0.733177125453949], "r_norm": [1.0419820547103882, 1.0341936349868774, 1.0294487476348877]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.964, "test_loss": 0.15205982093811035}, {"step": 3800, "epoch": 8, "train_loss": 0.02382911555469036, "cos_r_negg": [0.06859715282917023, 0.11703038960695267, 0.27915629744529724], "cos_innovation_negg": [0.4253789186477661, 0.5059475302696228, 0.7277424335479736], "cos_apical_negg": [0.06859715282917023, 0.11703038960695267, 0.27915629744529724], "cos_Ac_negg": [0.4253871440887451, 0.5056382417678833, 0.7274767756462097], "r_norm": [1.0048519372940063, 0.999269962310791, 0.9951889514923096]}, {"step": 3900, "epoch": 8, "train_loss": 0.1512923240661621, "cos_r_negg": [0.061481766402721405, 0.12657298147678375, 0.21981723606586456], "cos_innovation_negg": [0.41390371322631836, 0.5124120712280273, 0.7214871048927307], "cos_apical_negg": [0.061481766402721405, 0.12657298147678375, 0.21981723606586456], "cos_Ac_negg": [0.41383153200149536, 0.5121541023254395, 0.7217346429824829], "r_norm": [1.024421215057373, 1.0178453922271729, 1.014369249343872]}, {"step": 4000, "epoch": 8, "train_loss": 0.06986392289400101, "cos_r_negg": [0.06009560823440552, 0.13233833014965057, 0.17223654687404633], "cos_innovation_negg": [0.41721194982528687, 0.5196924209594727, 0.7315075397491455], "cos_apical_negg": [0.06009560823440552, 0.13233833014965057, 0.17223654687404633], "cos_Ac_negg": [0.4172259569168091, 0.5191183090209961, 0.7319979071617126], "r_norm": [1.0141053199768066, 1.0076427459716797, 1.003874659538269], "ldr": {"dL_sdil": -0.00461287796497345, "dL_bp": -0.007200658321380615, "ratio": 0.6406189210890098}}, {"step": 4100, "epoch": 8, "train_loss": 0.04138413816690445, "cos_r_negg": [0.0579494871199131, 0.11536955088376999, 0.24498414993286133], "cos_innovation_negg": [0.4185745418071747, 0.4919985830783844, 0.7416543960571289], "cos_apical_negg": [0.0579494871199131, 0.11536955088376999, 0.24498414993286133], "cos_Ac_negg": [0.41827186942100525, 0.49132803082466125, 0.7418818473815918], "r_norm": [1.024757742881775, 1.0176424980163574, 1.01226007938385]}, {"step": 4200, "epoch": 8, "train_loss": 0.08453892171382904, "cos_r_negg": [0.05851001292467117, 0.12678737938404083, 0.2562369108200073], "cos_innovation_negg": [0.432187020778656, 0.5198979377746582, 0.7205055952072144], "cos_apical_negg": [0.05851001292467117, 0.12678737938404083, 0.2562369108200073], "cos_Ac_negg": [0.4316670894622803, 0.5195059776306152, 0.7205809950828552], "r_norm": [1.0131231546401978, 1.0065301656723022, 1.0030956268310547]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9622, "test_loss": 0.1646196990966797}, {"step": 4300, "epoch": 9, "train_loss": 0.04316658526659012, "cos_r_negg": [0.054280661046504974, 0.10196350514888763, 0.2806333303451538], "cos_innovation_negg": [0.4090093970298767, 0.5053075551986694, 0.7531549334526062], "cos_apical_negg": [0.054280661046504974, 0.10196350514888763, 0.2806333303451538], "cos_Ac_negg": [0.40875399112701416, 0.5050163269042969, 0.7533770203590393], "r_norm": [1.0165133476257324, 1.0105763673782349, 1.005072832107544]}, {"step": 4400, "epoch": 9, "train_loss": 0.044569067656993866, "cos_r_negg": [0.04354625195264816, 0.10968311876058578, 0.2889139652252197], "cos_innovation_negg": [0.4083552956581116, 0.48645853996276855, 0.7443960905075073], "cos_apical_negg": [0.04354625195264816, 0.10968311876058578, 0.2889139652252197], "cos_Ac_negg": [0.4082323908805847, 0.4860506057739258, 0.7447205781936646], "r_norm": [1.006726861000061, 1.0011639595031738, 0.997625470161438]}, {"step": 4500, "epoch": 9, "train_loss": 0.13628464937210083, "cos_r_negg": [0.039948031306266785, 0.12244822829961777, 0.24630369246006012], "cos_innovation_negg": [0.41494235396385193, 0.4948727786540985, 0.740593433380127], "cos_apical_negg": [0.039948031306266785, 0.12244822829961777, 0.24630369246006012], "cos_Ac_negg": [0.41452139616012573, 0.4943540096282959, 0.7411041259765625], "r_norm": [1.0156197547912598, 1.009873390197754, 1.0053232908248901], "ldr": {"dL_sdil": -0.004762277007102966, "dL_bp": -0.007550597190856934, "ratio": 0.6307152781066957}}, {"step": 4600, "epoch": 9, "train_loss": 0.021538682281970978, "cos_r_negg": [0.0696280226111412, 0.13792680203914642, 0.3534112274646759], "cos_innovation_negg": [0.40972524881362915, 0.4834335744380951, 0.7794768810272217], "cos_apical_negg": [0.0696280226111412, 0.13792680203914642, 0.3534112274646759], "cos_Ac_negg": [0.40946805477142334, 0.48318007588386536, 0.7795393466949463], "r_norm": [1.04246187210083, 1.0343941450119019, 1.0295746326446533]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9619, "test_loss": 0.17044685344696045}, {"step": 4700, "epoch": 10, "train_loss": 0.08341549336910248, "cos_r_negg": [0.04571884498000145, 0.11891037225723267, 0.30167603492736816], "cos_innovation_negg": [0.38895994424819946, 0.47783076763153076, 0.7742821574211121], "cos_apical_negg": [0.04571884498000145, 0.11891037225723267, 0.30167603492736816], "cos_Ac_negg": [0.3886677026748657, 0.4773964285850525, 0.7738631367683411], "r_norm": [1.0328233242034912, 1.0247845649719238, 1.0186296701431274]}, {"step": 4800, "epoch": 10, "train_loss": 0.050148431211709976, "cos_r_negg": [0.02763083577156067, 0.10058742761611938, 0.3184281289577484], "cos_innovation_negg": [0.38614290952682495, 0.48786497116088867, 0.7830615043640137], "cos_apical_negg": [0.02763083577156067, 0.10058742761611938, 0.3184281289577484], "cos_Ac_negg": [0.3859814405441284, 0.48769450187683105, 0.7839152216911316], "r_norm": [1.0237292051315308, 1.0161826610565186, 1.0115747451782227]}, {"step": 4900, "epoch": 10, "train_loss": 0.055147189646959305, "cos_r_negg": [0.03308366984128952, 0.12296850979328156, 0.22171971201896667], "cos_innovation_negg": [0.37891247868537903, 0.4514099955558777, 0.7571605443954468], "cos_apical_negg": [0.03308366984128952, 0.12296850979328156, 0.22171971201896667], "cos_Ac_negg": [0.37868732213974, 0.451760470867157, 0.7571224570274353], "r_norm": [1.0504040718078613, 1.0420150756835938, 1.0372602939605713]}, {"step": 5000, "epoch": 10, "train_loss": 0.01705363765358925, "cos_r_negg": [0.054442863911390305, 0.11831051856279373, 0.326398104429245], "cos_innovation_negg": [0.4010009467601776, 0.4879360496997833, 0.7031821608543396], "cos_apical_negg": [0.054442863911390305, 0.11831051856279373, 0.326398104429245], "cos_Ac_negg": [0.40074652433395386, 0.48770245909690857, 0.7030981779098511], "r_norm": [1.0476109981536865, 1.0388975143432617, 1.0337738990783691], "ldr": {"dL_sdil": -0.007159620523452759, "dL_bp": -0.01317286491394043, "ratio": 0.5435127870988761}}, {"step": 5100, "epoch": 10, "train_loss": 0.09889398515224457, "cos_r_negg": [0.04644930362701416, 0.12823310494422913, 0.316743940114975], "cos_innovation_negg": [0.370052307844162, 0.4791232943534851, 0.779529333114624], "cos_apical_negg": [0.04644930362701416, 0.12823310494422913, 0.316743940114975], "cos_Ac_negg": [0.3699507713317871, 0.4789623022079468, 0.7796865105628967], "r_norm": [1.012260913848877, 1.0053958892822266, 1.0003650188446045]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.965, "test_loss": 0.15641935958862305}, {"step": 5200, "epoch": 11, "train_loss": 0.011618003249168396, "cos_r_negg": [0.029473181813955307, 0.13050757348537445, 0.2541808485984802], "cos_innovation_negg": [0.3835332691669464, 0.4815676808357239, 0.7470118403434753], "cos_apical_negg": [0.029473181813955307, 0.13050757348537445, 0.2541808485984802], "cos_Ac_negg": [0.3832237720489502, 0.48124536871910095, 0.7483774423599243], "r_norm": [1.0235294103622437, 1.0155537128448486, 1.0106277465820312]}, {"step": 5300, "epoch": 11, "train_loss": 0.06883976608514786, "cos_r_negg": [0.02978581003844738, 0.1577966809272766, 0.26428189873695374], "cos_innovation_negg": [0.3612690269947052, 0.44747382402420044, 0.7369123697280884], "cos_apical_negg": [0.02978581003844738, 0.1577966809272766, 0.26428189873695374], "cos_Ac_negg": [0.36127275228500366, 0.4472388029098511, 0.736887514591217], "r_norm": [1.0157907009124756, 1.0084893703460693, 1.003711462020874]}, {"step": 5400, "epoch": 11, "train_loss": 0.09158950299024582, "cos_r_negg": [0.048825714737176895, 0.1291065514087677, 0.19313271343708038], "cos_innovation_negg": [0.36108851432800293, 0.45001083612442017, 0.7426520586013794], "cos_apical_negg": [0.048825714737176895, 0.1291065514087677, 0.19313271343708038], "cos_Ac_negg": [0.3607645332813263, 0.4500146508216858, 0.7427655458450317], "r_norm": [0.9970957636833191, 0.9907493591308594, 0.9866862297058105]}, {"step": 5500, "epoch": 11, "train_loss": 0.09621293097734451, "cos_r_negg": [0.02780434675514698, 0.1293172538280487, 0.28517085313796997], "cos_innovation_negg": [0.3989107012748718, 0.4655280113220215, 0.7712547779083252], "cos_apical_negg": [0.02780434675514698, 0.1293172538280487, 0.28517085313796997], "cos_Ac_negg": [0.3988367021083832, 0.46450167894363403, 0.7711506485939026], "r_norm": [1.0169309377670288, 1.0087089538574219, 1.003431797027588], "ldr": {"dL_sdil": -0.00443209707736969, "dL_bp": -0.007188484072685242, "ratio": 0.6165551780535684}}, {"step": 5600, "epoch": 11, "train_loss": 0.009339344687759876, "cos_r_negg": [0.02955637313425541, 0.12707480788230896, 0.3031397759914398], "cos_innovation_negg": [0.3965030312538147, 0.4689267575740814, 0.7651573419570923], "cos_apical_negg": [0.02955637313425541, 0.12707480788230896, 0.3031397759914398], "cos_Ac_negg": [0.3964735269546509, 0.4691050052642822, 0.7656296491622925], "r_norm": [1.0046048164367676, 0.9975359439849854, 0.9935814142227173]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9646, "test_loss": 0.1676767059326172}, {"step": 5700, "epoch": 12, "train_loss": 0.039681173861026764, "cos_r_negg": [0.023674480617046356, 0.1559143364429474, 0.2809804081916809], "cos_innovation_negg": [0.3727792501449585, 0.48782408237457275, 0.7735583782196045], "cos_apical_negg": [0.023674480617046356, 0.1559143364429474, 0.2809804081916809], "cos_Ac_negg": [0.3725578486919403, 0.487379252910614, 0.7736832499504089], "r_norm": [1.0121532678604126, 1.004992961883545, 1.000563144683838]}, {"step": 5800, "epoch": 12, "train_loss": 0.009409253485500813, "cos_r_negg": [0.012268457561731339, 0.12750187516212463, 0.31009548902511597], "cos_innovation_negg": [0.3641427159309387, 0.45033109188079834, 0.7750030755996704], "cos_apical_negg": [0.012268457561731339, 0.12750187516212463, 0.31009548902511597], "cos_Ac_negg": [0.36390671133995056, 0.44966304302215576, 0.7762756943702698], "r_norm": [1.0208228826522827, 1.0131936073303223, 1.0075652599334717]}, {"step": 5900, "epoch": 12, "train_loss": 0.03900974616408348, "cos_r_negg": [0.037060368806123734, 0.11824866384267807, 0.3086228370666504], "cos_innovation_negg": [0.3732796907424927, 0.4792660176753998, 0.7758967876434326], "cos_apical_negg": [0.037060368806123734, 0.11824866384267807, 0.3086228370666504], "cos_Ac_negg": [0.3726726174354553, 0.4773159623146057, 0.7757237553596497], "r_norm": [1.0238292217254639, 1.015966773033142, 1.0104520320892334]}, {"step": 6000, "epoch": 12, "train_loss": 0.00767051987349987, "cos_r_negg": [0.026138661429286003, 0.11731883883476257, 0.30830878019332886], "cos_innovation_negg": [0.36157429218292236, 0.4710817337036133, 0.7574659585952759], "cos_apical_negg": [0.026138661429286003, 0.11731883883476257, 0.30830878019332886], "cos_Ac_negg": [0.36157506704330444, 0.47078239917755127, 0.7574666738510132], "r_norm": [1.0263004302978516, 1.0175118446350098, 1.012162446975708], "ldr": {"dL_sdil": -0.005875617265701294, "dL_bp": -0.01049591600894928, "ratio": 0.5598003319282933}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9656, "test_loss": 0.16640939407348632}, {"step": 6100, "epoch": 13, "train_loss": 0.013630226254463196, "cos_r_negg": [0.018729209899902344, 0.13422933220863342, 0.26772210001945496], "cos_innovation_negg": [0.3898214101791382, 0.4603390693664551, 0.7649310827255249], "cos_apical_negg": [0.018729209899902344, 0.13422933220863342, 0.26772210001945496], "cos_Ac_negg": [0.38953709602355957, 0.4590319097042084, 0.764623761177063], "r_norm": [1.0028607845306396, 0.9960696697235107, 0.9908932447433472]}, {"step": 6200, "epoch": 13, "train_loss": 0.014845678582787514, "cos_r_negg": [0.04792290925979614, 0.13098245859146118, 0.2532523572444916], "cos_innovation_negg": [0.3858796954154968, 0.4564515948295593, 0.7580298185348511], "cos_apical_negg": [0.04792290925979614, 0.13098245859146118, 0.2532523572444916], "cos_Ac_negg": [0.38527148962020874, 0.45281267166137695, 0.7577612996101379], "r_norm": [0.9996435046195984, 0.9916858673095703, 0.9880802631378174]}, {"step": 6300, "epoch": 13, "train_loss": 0.030698556452989578, "cos_r_negg": [0.024974389001727104, 0.1229492723941803, 0.23045440018177032], "cos_innovation_negg": [0.3792564272880554, 0.44726669788360596, 0.7672052979469299], "cos_apical_negg": [0.024974389001727104, 0.1229492723941803, 0.23045440018177032], "cos_Ac_negg": [0.37906143069267273, 0.4448671042919159, 0.7671007513999939], "r_norm": [0.9923502802848816, 0.9854509830474854, 0.9807218313217163]}, {"step": 6400, "epoch": 13, "train_loss": 0.04797133058309555, "cos_r_negg": [0.031597599387168884, 0.12947042286396027, 0.2632352113723755], "cos_innovation_negg": [0.3592355251312256, 0.45848363637924194, 0.755677342414856], "cos_apical_negg": [0.031597599387168884, 0.12947042286396027, 0.2632352113723755], "cos_Ac_negg": [0.3589644432067871, 0.45814529061317444, 0.7558613419532776], "r_norm": [0.9953324794769287, 0.9877307415008545, 0.9834027290344238]}, {"step": 6500, "epoch": 13, "train_loss": 0.003274256829172373, "cos_r_negg": [0.012578103691339493, 0.11179740726947784, 0.2535165250301361], "cos_innovation_negg": [0.3779268264770508, 0.4656335413455963, 0.7351794838905334], "cos_apical_negg": [0.012578103691339493, 0.11179740726947784, 0.2535165250301361], "cos_Ac_negg": [0.3777846097946167, 0.4647585153579712, 0.7358333468437195], "r_norm": [0.9992240071296692, 0.9908066391944885, 0.9862110614776611], "ldr": {"dL_sdil": -0.004559755325317383, "dL_bp": -0.008030295372009277, "ratio": 0.5678191291955524}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9639, "test_loss": 0.17209006233215332}, {"step": 6600, "epoch": 14, "train_loss": 0.007514375727623701, "cos_r_negg": [0.01508256234228611, 0.13311469554901123, 0.27300024032592773], "cos_innovation_negg": [0.38066548109054565, 0.46259987354278564, 0.7482706308364868], "cos_apical_negg": [0.01508256234228611, 0.13311469554901123, 0.27300024032592773], "cos_Ac_negg": [0.37996968626976013, 0.4604552388191223, 0.7488212585449219], "r_norm": [1.0007352828979492, 0.9943813681602478, 0.9901024103164673]}, {"step": 6700, "epoch": 14, "train_loss": 0.06227608770132065, "cos_r_negg": [0.00989779643714428, 0.14579452574253082, 0.3493391275405884], "cos_innovation_negg": [0.38116636872291565, 0.45595836639404297, 0.780990719795227], "cos_apical_negg": [0.00989779643714428, 0.14579452574253082, 0.3493391275405884], "cos_Ac_negg": [0.380575954914093, 0.4534801244735718, 0.7811557054519653], "r_norm": [1.0045349597930908, 0.9969815611839294, 0.9917994737625122]}, {"step": 6800, "epoch": 14, "train_loss": 0.06896437704563141, "cos_r_negg": [0.012832611799240112, 0.15619882941246033, 0.32386553287506104], "cos_innovation_negg": [0.3918267488479614, 0.4652990698814392, 0.7924189567565918], "cos_apical_negg": [0.012832611799240112, 0.15619882941246033, 0.32386553287506104], "cos_Ac_negg": [0.39119476079940796, 0.4643750488758087, 0.7919941544532776], "r_norm": [1.0126630067825317, 1.0044729709625244, 0.9982580542564392]}, {"step": 6900, "epoch": 14, "train_loss": 0.02285294607281685, "cos_r_negg": [-0.0004212576895952225, 0.13315272331237793, 0.2762680947780609], "cos_innovation_negg": [0.3808654546737671, 0.49239078164100647, 0.7823525667190552], "cos_apical_negg": [-0.0004212576895952225, 0.13315272331237793, 0.2762680947780609], "cos_Ac_negg": [0.38029754161834717, 0.49130237102508545, 0.7821752429008484], "r_norm": [0.9951581954956055, 0.9866114854812622, 0.9817646741867065]}, {"step": 7000, "epoch": 14, "train_loss": 0.052804600447416306, "cos_r_negg": [0.03642439842224121, 0.10317088663578033, 0.24736720323562622], "cos_innovation_negg": [0.36583250761032104, 0.4374212622642517, 0.7792471647262573], "cos_apical_negg": [0.03642439842224121, 0.10317088663578033, 0.24736720323562622], "cos_Ac_negg": [0.36550819873809814, 0.4369518756866455, 0.7794103622436523], "r_norm": [1.0036711692810059, 0.995650053024292, 0.9915581941604614], "ldr": {"dL_sdil": -0.003786906599998474, "dL_bp": -0.006322965025901794, "ratio": 0.5989131023950868}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.965, "test_loss": 0.17623918800354005}], "final": {"test_acc": 0.965, "test_loss": 0.17623918800354005, "wall_s": 27.3532133102417, "cos_r_negg": [0.02815653569996357, 0.10895447432994843, 0.2546684443950653], "cos_innovation_negg": [0.3487493693828583, 0.4369896352291107, 0.7721617221832275], "cos_apical_negg": [0.02815653569996357, 0.10895447432994843, 0.2546684443950653], "cos_Ac_negg": [0.3478301167488098, 0.4360348582267761, 0.77239590883255]}}
|