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": 0.2, "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_rho0p2_residual_s1", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.327335834503174, "cos_r_negg": [0.011931603774428368, -0.013785354793071747, -0.005786198191344738], "cos_innovation_negg": [0.011931603774428368, -0.013785354793071747, -0.005786198191344738], "cos_apical_negg": [0.010163621976971626, -0.012931429781019688, -0.007024887017905712], "cos_Ac_negg": [0.011931432411074638, -0.013785326853394508, -0.005786254070699215], "r_norm": [4.739147186279297, 4.7983622550964355, 4.7884955406188965], "ldr": {"dL_sdil": -0.04240226745605469, "dL_bp": -0.21988940238952637, "ratio": 0.19283452042377447}}, {"step": 100, "epoch": 0, "train_loss": 0.19775736331939697, "cos_r_negg": [0.45872023701667786, 0.3625425696372986, 0.3213750123977661], "cos_innovation_negg": [0.45872023701667786, 0.3625425696372986, 0.3213750123977661], "cos_apical_negg": [0.1356997787952423, 0.10975304245948792, 0.14097663760185242], "cos_Ac_negg": [0.45958253741264343, 0.36344701051712036, 0.32139092683792114], "r_norm": [0.9387720823287964, 0.9419670104980469, 0.9592081904411316]}, {"step": 200, "epoch": 0, "train_loss": 0.32657188177108765, "cos_r_negg": [0.4945112466812134, 0.42765265703201294, 0.3720117211341858], "cos_innovation_negg": [0.4945112466812134, 0.42765265703201294, 0.3720117211341858], "cos_apical_negg": [0.12734103202819824, 0.11078793555498123, 0.14468501508235931], "cos_Ac_negg": [0.49576351046562195, 0.42875152826309204, 0.3724040389060974], "r_norm": [0.7496249079704285, 0.7569105625152588, 0.7658607959747314]}, {"step": 300, "epoch": 0, "train_loss": 0.1660933941602707, "cos_r_negg": [0.5047013759613037, 0.45309486985206604, 0.41172993183135986], "cos_innovation_negg": [0.5047013759613037, 0.45309486985206604, 0.41172993183135986], "cos_apical_negg": [0.10227082669734955, 0.09924061596393585, 0.13187697529792786], "cos_Ac_negg": [0.5056310892105103, 0.4530573785305023, 0.410241961479187], "r_norm": [0.5657568573951721, 0.5680766105651855, 0.5780221819877625]}, {"step": 400, "epoch": 0, "train_loss": 0.1028156578540802, "cos_r_negg": [0.5188640356063843, 0.4748266935348511, 0.4355982542037964], "cos_innovation_negg": [0.5188640356063843, 0.4748266935348511, 0.4355982542037964], "cos_apical_negg": [0.09269370138645172, 0.10672474652528763, 0.14090155065059662], "cos_Ac_negg": [0.5188056826591492, 0.4748728275299072, 0.4347767233848572], "r_norm": [0.5791422128677368, 0.5850809216499329, 0.586473822593689]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9444, "test_loss": 0.18449363555908202}, {"step": 500, "epoch": 1, "train_loss": 0.18749767541885376, "cos_r_negg": [0.5367656946182251, 0.47632521390914917, 0.44773295521736145], "cos_innovation_negg": [0.5367656946182251, 0.47632521390914917, 0.44773295521736145], "cos_apical_negg": [0.08224304020404816, 0.0912894457578659, 0.15145498514175415], "cos_Ac_negg": [0.5402472615242004, 0.47959578037261963, 0.44738397002220154], "r_norm": [0.5153515338897705, 0.518422544002533, 0.5192044377326965], "ldr": {"dL_sdil": -0.013696134090423584, "dL_bp": -0.011990353465080261, "ratio": 1.1422627473252644}}, {"step": 600, "epoch": 1, "train_loss": 0.1040426716208458, "cos_r_negg": [0.530276894569397, 0.48525846004486084, 0.4493897557258606], "cos_innovation_negg": [0.530276894569397, 0.48525846004486084, 0.4493897557258606], "cos_apical_negg": [0.07756748050451279, 0.07605090737342834, 0.1394515335559845], "cos_Ac_negg": [0.5312595367431641, 0.48635751008987427, 0.44880661368370056], "r_norm": [0.4344339072704315, 0.4356266260147095, 0.44007688760757446]}, {"step": 700, "epoch": 1, "train_loss": 0.15639904141426086, "cos_r_negg": [0.5480979681015015, 0.5102795958518982, 0.48049014806747437], "cos_innovation_negg": [0.5480979681015015, 0.5102795958518982, 0.48049014806747437], "cos_apical_negg": [0.07059018313884735, 0.0943416953086853, 0.1362384855747223], "cos_Ac_negg": [0.5526449680328369, 0.5116695165634155, 0.4787350594997406], "r_norm": [0.41975533962249756, 0.424714058637619, 0.4228808283805847]}, {"step": 800, "epoch": 1, "train_loss": 0.16617156565189362, "cos_r_negg": [0.5512282848358154, 0.5170258283615112, 0.47346439957618713], "cos_innovation_negg": [0.5512282848358154, 0.5170258283615112, 0.47346439957618713], "cos_apical_negg": [0.07387018203735352, 0.0937204658985138, 0.12942790985107422], "cos_Ac_negg": [0.5569518804550171, 0.5212909579277039, 0.4727581739425659], "r_norm": [0.4107813537120819, 0.4118606448173523, 0.41608816385269165]}, {"step": 900, "epoch": 1, "train_loss": 0.2058173269033432, "cos_r_negg": [0.5738842487335205, 0.5367188453674316, 0.47506433725357056], "cos_innovation_negg": [0.5738842487335205, 0.5367188453674316, 0.47506433725357056], "cos_apical_negg": [0.061650656163692474, 0.09630943089723587, 0.14210359752178192], "cos_Ac_negg": [0.5736643671989441, 0.5366127490997314, 0.47380971908569336], "r_norm": [0.4241654872894287, 0.4231199622154236, 0.42675256729125977]}, {"epoch_end": 1, "step": 938, "test_acc": 0.955, "test_loss": 0.15527927742004394}, {"step": 1000, "epoch": 2, "train_loss": 0.021358666941523552, "cos_r_negg": [0.5856679081916809, 0.5437238216400146, 0.4949186146259308], "cos_innovation_negg": [0.5856679081916809, 0.5437238216400146, 0.4949186146259308], "cos_apical_negg": [0.046389296650886536, 0.08993111550807953, 0.13293297588825226], "cos_Ac_negg": [0.5890286564826965, 0.5449188351631165, 0.49256759881973267], "r_norm": [0.36267194151878357, 0.36412501335144043, 0.369210809469223], "ldr": {"dL_sdil": -0.009381487965583801, "dL_bp": -0.008346065878868103, "ratio": 1.1240610967782252}}, {"step": 1100, "epoch": 2, "train_loss": 0.11250459402799606, "cos_r_negg": [0.5836842060089111, 0.5534021258354187, 0.5075383186340332], "cos_innovation_negg": [0.5836842060089111, 0.5534021258354187, 0.5075383186340332], "cos_apical_negg": [0.060446105897426605, 0.08326318860054016, 0.12937980890274048], "cos_Ac_negg": [0.5876331329345703, 0.5538390874862671, 0.5071291923522949], "r_norm": [0.37702304124832153, 0.3828517198562622, 0.38200998306274414]}, {"step": 1200, "epoch": 2, "train_loss": 0.13226768374443054, "cos_r_negg": [0.5769443511962891, 0.5558546781539917, 0.5066117644309998], "cos_innovation_negg": [0.5769443511962891, 0.5558546781539917, 0.5066117644309998], "cos_apical_negg": [0.05843479931354523, 0.08167964220046997, 0.11882665008306503], "cos_Ac_negg": [0.58112633228302, 0.5567046999931335, 0.5058470964431763], "r_norm": [0.337144136428833, 0.3376275897026062, 0.34184199571609497]}, {"step": 1300, "epoch": 2, "train_loss": 0.11763103306293488, "cos_r_negg": [0.5968750715255737, 0.5797828435897827, 0.5279954671859741], "cos_innovation_negg": [0.5968750715255737, 0.5797828435897827, 0.5279954671859741], "cos_apical_negg": [0.04198881983757019, 0.0771612599492073, 0.12961767613887787], "cos_Ac_negg": [0.602980375289917, 0.5828290581703186, 0.528378427028656], "r_norm": [0.318722128868103, 0.3170597553253174, 0.3181980848312378]}, {"step": 1400, "epoch": 2, "train_loss": 0.11081075668334961, "cos_r_negg": [0.5979578495025635, 0.5789686441421509, 0.5127284526824951], "cos_innovation_negg": [0.5979578495025635, 0.5789686441421509, 0.5127284526824951], "cos_apical_negg": [0.046362608671188354, 0.07994739711284637, 0.11826172471046448], "cos_Ac_negg": [0.6016546487808228, 0.5812645554542542, 0.5135836601257324], "r_norm": [0.28814074397087097, 0.2862517237663269, 0.29094311594963074]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9643, "test_loss": 0.11845275306701661}, {"step": 1500, "epoch": 3, "train_loss": 0.05337446555495262, "cos_r_negg": [0.5874097347259521, 0.5748370289802551, 0.5361409187316895], "cos_innovation_negg": [0.5874097347259521, 0.5748370289802551, 0.5361409187316895], "cos_apical_negg": [0.04201585799455643, 0.07050016522407532, 0.12474855035543442], "cos_Ac_negg": [0.5923759937286377, 0.5785771608352661, 0.5364388227462769], "r_norm": [0.261452317237854, 0.2613041400909424, 0.2655849754810333], "ldr": {"dL_sdil": -0.005416393280029297, "dL_bp": -0.00526069849729538, "ratio": 1.0295958384260118}}, {"step": 1600, "epoch": 3, "train_loss": 0.05822928994894028, "cos_r_negg": [0.5893609523773193, 0.5796955227851868, 0.5560258626937866], "cos_innovation_negg": [0.5893609523773193, 0.5796955227851868, 0.5560258626937866], "cos_apical_negg": [0.028988126665353775, 0.06752461194992065, 0.1227165162563324], "cos_Ac_negg": [0.5977765917778015, 0.586108922958374, 0.5571414828300476], "r_norm": [0.2562752366065979, 0.2568072974681854, 0.25675591826438904]}, {"step": 1700, "epoch": 3, "train_loss": 0.11532272398471832, "cos_r_negg": [0.5804225206375122, 0.5829477906227112, 0.5600645542144775], "cos_innovation_negg": [0.5804225206375122, 0.5829477906227112, 0.5600645542144775], "cos_apical_negg": [0.041842132806777954, 0.08018238842487335, 0.12040457874536514], "cos_Ac_negg": [0.5935515761375427, 0.5934044122695923, 0.5652104616165161], "r_norm": [0.2907516658306122, 0.28905370831489563, 0.29030248522758484]}, {"step": 1800, "epoch": 3, "train_loss": 0.024315420538187027, "cos_r_negg": [0.596096932888031, 0.5926241278648376, 0.5527631640434265], "cos_innovation_negg": [0.596096932888031, 0.5926241278648376, 0.5527631640434265], "cos_apical_negg": [0.04096980020403862, 0.06974849849939346, 0.11268502473831177], "cos_Ac_negg": [0.6052919626235962, 0.5986686944961548, 0.554040253162384], "r_norm": [0.2584753632545471, 0.258015513420105, 0.2584863305091858]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9632, "test_loss": 0.11869043083190918}, {"step": 1900, "epoch": 4, "train_loss": 0.04801398888230324, "cos_r_negg": [0.5939580798149109, 0.5847367644309998, 0.5548296570777893], "cos_innovation_negg": [0.5939580798149109, 0.5847367644309998, 0.5548296570777893], "cos_apical_negg": [0.06321965157985687, 0.07881684601306915, 0.1235065758228302], "cos_Ac_negg": [0.6099642515182495, 0.5955401062965393, 0.5598971843719482], "r_norm": [0.29030996561050415, 0.2912764251232147, 0.2929227948188782]}, {"step": 2000, "epoch": 4, "train_loss": 0.07748020440340042, "cos_r_negg": [0.5760419368743896, 0.5681276321411133, 0.5477563142776489], "cos_innovation_negg": [0.5760419368743896, 0.5681276321411133, 0.5477563142776489], "cos_apical_negg": [0.02792409248650074, 0.059769026935100555, 0.11725227534770966], "cos_Ac_negg": [0.6076132655143738, 0.5944979190826416, 0.561225414276123], "r_norm": [0.22954419255256653, 0.22847869992256165, 0.2283090054988861], "ldr": {"dL_sdil": -0.0057287514209747314, "dL_bp": -0.006127595901489258, "ratio": 0.9349101202287848}}, {"step": 2100, "epoch": 4, "train_loss": 0.050592273473739624, "cos_r_negg": [0.5596483945846558, 0.574286937713623, 0.5611082315444946], "cos_innovation_negg": [0.5596483945846558, 0.574286937713623, 0.5611082315444946], "cos_apical_negg": [0.024154040962457657, 0.0528561994433403, 0.11123406141996384], "cos_Ac_negg": [0.5905953645706177, 0.600906491279602, 0.5775448679924011], "r_norm": [0.2186432033777237, 0.2179013192653656, 0.22010205686092377]}, {"step": 2200, "epoch": 4, "train_loss": 0.01839619129896164, "cos_r_negg": [0.5920798778533936, 0.5911052227020264, 0.5681278705596924], "cos_innovation_negg": [0.5920798778533936, 0.5911052227020264, 0.5681278705596924], "cos_apical_negg": [0.011098718270659447, 0.05819759517908096, 0.11481077969074249], "cos_Ac_negg": [0.6115607023239136, 0.6076661348342896, 0.575154721736908], "r_norm": [0.21184080839157104, 0.21109355986118317, 0.21288445591926575]}, {"step": 2300, "epoch": 4, "train_loss": 0.16251692175865173, "cos_r_negg": [0.5427846312522888, 0.5610638856887817, 0.5550764203071594], "cos_innovation_negg": [0.5427846312522888, 0.5610638856887817, 0.5550764203071594], "cos_apical_negg": [0.028781840577721596, 0.057624928653240204, 0.11066931486129761], "cos_Ac_negg": [0.5845282077789307, 0.602412223815918, 0.5805932283401489], "r_norm": [0.19856134057044983, 0.19890277087688446, 0.20000869035720825]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.969, "test_loss": 0.11277374114990234}, {"step": 2400, "epoch": 5, "train_loss": 0.07764077931642532, "cos_r_negg": [0.5798075199127197, 0.5902670621871948, 0.579801082611084], "cos_innovation_negg": [0.5798075199127197, 0.5902670621871948, 0.579801082611084], "cos_apical_negg": [0.027051949873566628, 0.04981415718793869, 0.11517209559679031], "cos_Ac_negg": [0.6182402968406677, 0.624508261680603, 0.5985470414161682], "r_norm": [0.20267830789089203, 0.2023141235113144, 0.2052423655986786]}, {"step": 2500, "epoch": 5, "train_loss": 0.03252333402633667, "cos_r_negg": [0.5317485332489014, 0.5525341033935547, 0.5637630820274353], "cos_innovation_negg": [0.5317485332489014, 0.5525341033935547, 0.5637630820274353], "cos_apical_negg": [0.027274269610643387, 0.05618946626782417, 0.12815535068511963], "cos_Ac_negg": [0.6042139530181885, 0.6129974126815796, 0.6021047830581665], "r_norm": [0.2474762499332428, 0.24764174222946167, 0.2503415644168854], "ldr": {"dL_sdil": -0.008180499076843262, "dL_bp": -0.009082622826099396, "ratio": 0.9006758547031334}}, {"step": 2600, "epoch": 5, "train_loss": 0.05673728883266449, "cos_r_negg": [0.5518810749053955, 0.5768418312072754, 0.5871708393096924], "cos_innovation_negg": [0.5518810749053955, 0.5768418312072754, 0.5871708393096924], "cos_apical_negg": [0.010864458978176117, 0.05714278295636177, 0.11639561504125595], "cos_Ac_negg": [0.6136483550071716, 0.6258563995361328, 0.6129207611083984], "r_norm": [0.22200214862823486, 0.2225976437330246, 0.22385293245315552]}, {"step": 2700, "epoch": 5, "train_loss": 0.012749000452458858, "cos_r_negg": [0.5460054278373718, 0.5705655813217163, 0.5896921753883362], "cos_innovation_negg": [0.5460054278373718, 0.5705655813217163, 0.5896921753883362], "cos_apical_negg": [0.020181067287921906, 0.04799661040306091, 0.10917221009731293], "cos_Ac_negg": [0.604796826839447, 0.6227099299430847, 0.6224133968353271], "r_norm": [0.1813088357448578, 0.18050485849380493, 0.18305231630802155]}, {"step": 2800, "epoch": 5, "train_loss": 0.04988821595907211, "cos_r_negg": [0.5203732848167419, 0.555920422077179, 0.5674914717674255], "cos_innovation_negg": [0.5203732848167419, 0.555920422077179, 0.5674914717674255], "cos_apical_negg": [0.026939351111650467, 0.055087488144636154, 0.10572728514671326], "cos_Ac_negg": [0.6057394742965698, 0.6266579627990723, 0.6157872676849365], "r_norm": [0.18579533696174622, 0.18391703069210052, 0.18668165802955627]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9689, "test_loss": 0.11896831932067871}, {"step": 2900, "epoch": 6, "train_loss": 0.006228657439351082, "cos_r_negg": [0.5252091884613037, 0.5586499571800232, 0.5708284378051758], "cos_innovation_negg": [0.5252091884613037, 0.5586499571800232, 0.5708284378051758], "cos_apical_negg": [0.0154110724106431, 0.047098420560359955, 0.11612921953201294], "cos_Ac_negg": [0.6111032962799072, 0.6351197957992554, 0.6190087795257568], "r_norm": [0.182048499584198, 0.18028321862220764, 0.18298307061195374]}, {"step": 3000, "epoch": 6, "train_loss": 0.05967951565980911, "cos_r_negg": [0.503757119178772, 0.5327576398849487, 0.5590969324111938], "cos_innovation_negg": [0.503757119178772, 0.5327576398849487, 0.5590969324111938], "cos_apical_negg": [0.006705448497086763, 0.048671916127204895, 0.11587689816951752], "cos_Ac_negg": [0.6157150864601135, 0.6238573789596558, 0.6224365234375], "r_norm": [0.16073641180992126, 0.15966340899467468, 0.16142268478870392], "ldr": {"dL_sdil": -0.00339449942111969, "dL_bp": -0.004144594073295593, "ratio": 0.8190185482798169}}, {"step": 3100, "epoch": 6, "train_loss": 0.022607536986470222, "cos_r_negg": [0.49863696098327637, 0.51980060338974, 0.5562487244606018], "cos_innovation_negg": [0.49863696098327637, 0.51980060338974, 0.5562487244606018], "cos_apical_negg": [0.00047961785458028316, 0.0421375036239624, 0.10713469982147217], "cos_Ac_negg": [0.6123783588409424, 0.6199117302894592, 0.6267445087432861], "r_norm": [0.16553622484207153, 0.16404928267002106, 0.1662520319223404]}, {"step": 3200, "epoch": 6, "train_loss": 0.03901796042919159, "cos_r_negg": [0.5001592636108398, 0.5136443972587585, 0.5546040534973145], "cos_innovation_negg": [0.5001592636108398, 0.5136443972587585, 0.5546040534973145], "cos_apical_negg": [0.016849977895617485, 0.045684389770030975, 0.11689712107181549], "cos_Ac_negg": [0.6181316375732422, 0.6196607947349548, 0.6289545297622681], "r_norm": [0.17972540855407715, 0.1776731163263321, 0.179264634847641]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9685, "test_loss": 0.12659227161407471}, {"step": 3300, "epoch": 7, "train_loss": 0.0032453604508191347, "cos_r_negg": [0.49487438797950745, 0.5146627426147461, 0.5570589900016785], "cos_innovation_negg": [0.49487438797950745, 0.5146627426147461, 0.5570589900016785], "cos_apical_negg": [0.022871367633342743, 0.05089029669761658, 0.11626814305782318], "cos_Ac_negg": [0.621711015701294, 0.6350054144859314, 0.6384247541427612], "r_norm": [0.18347486853599548, 0.18063265085220337, 0.18405190110206604]}, {"step": 3400, "epoch": 7, "train_loss": 0.04436119645833969, "cos_r_negg": [0.4657537341117859, 0.4919852614402771, 0.5333935022354126], "cos_innovation_negg": [0.4657537341117859, 0.4919852614402771, 0.5333935022354126], "cos_apical_negg": [0.008652392774820328, 0.049738816916942596, 0.11082755029201508], "cos_Ac_negg": [0.6239668130874634, 0.6362565755844116, 0.6375622153282166], "r_norm": [0.18661317229270935, 0.1850159913301468, 0.1869591474533081]}, {"step": 3500, "epoch": 7, "train_loss": 0.006032860372215509, "cos_r_negg": [0.46196189522743225, 0.48990634083747864, 0.5319598317146301], "cos_innovation_negg": [0.46196189522743225, 0.48990634083747864, 0.5319598317146301], "cos_apical_negg": [0.005535840522497892, 0.053613126277923584, 0.1106598973274231], "cos_Ac_negg": [0.6271502375602722, 0.6420409679412842, 0.6415368914604187], "r_norm": [0.18070045113563538, 0.17858397960662842, 0.1823444664478302], "ldr": {"dL_sdil": -0.005282089114189148, "dL_bp": -0.0069892629981040955, "ratio": 0.75574336172812}}, {"step": 3600, "epoch": 7, "train_loss": 0.002231165301054716, "cos_r_negg": [0.44013679027557373, 0.4753730297088623, 0.5182855129241943], "cos_innovation_negg": [0.44013679027557373, 0.4753730297088623, 0.5182855129241943], "cos_apical_negg": [0.005495928227901459, 0.051755014806985855, 0.10805808752775192], "cos_Ac_negg": [0.619712769985199, 0.6374286413192749, 0.6382573843002319], "r_norm": [0.17825168371200562, 0.17635288834571838, 0.17897850275039673]}, {"step": 3700, "epoch": 7, "train_loss": 0.00967714935541153, "cos_r_negg": [0.452364444732666, 0.48681214451789856, 0.5281479358673096], "cos_innovation_negg": [0.452364444732666, 0.48681214451789856, 0.5281479358673096], "cos_apical_negg": [0.0055759381502866745, 0.04553765058517456, 0.11463244259357452], "cos_Ac_negg": [0.6161730289459229, 0.6399081945419312, 0.6419907212257385], "r_norm": [0.19381679594516754, 0.1908852458000183, 0.19416004419326782]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9717, "test_loss": 0.11634278068542481}, {"step": 3800, "epoch": 8, "train_loss": 0.007380823139101267, "cos_r_negg": [0.4137468934059143, 0.45212435722351074, 0.49357855319976807], "cos_innovation_negg": [0.4137468934059143, 0.45212435722351074, 0.49357855319976807], "cos_apical_negg": [0.00708778714761138, 0.05199868232011795, 0.10923279821872711], "cos_Ac_negg": [0.6250726580619812, 0.6390424966812134, 0.6447779536247253], "r_norm": [0.19211000204086304, 0.1904672384262085, 0.1922861933708191]}, {"step": 3900, "epoch": 8, "train_loss": 0.029728081077337265, "cos_r_negg": [0.4061935544013977, 0.43666213750839233, 0.4823722839355469], "cos_innovation_negg": [0.4061935544013977, 0.43666213750839233, 0.4823722839355469], "cos_apical_negg": [0.00817396305501461, 0.04910297691822052, 0.11961860209703445], "cos_Ac_negg": [0.6252659559249878, 0.6392565965652466, 0.6468268632888794], "r_norm": [0.1889357566833496, 0.18662671744823456, 0.18918143212795258]}, {"step": 4000, "epoch": 8, "train_loss": 0.016131335869431496, "cos_r_negg": [0.37239107489585876, 0.4060817062854767, 0.4575170874595642], "cos_innovation_negg": [0.37239107489585876, 0.4060817062854767, 0.4575170874595642], "cos_apical_negg": [9.378185495734215e-05, 0.046725645661354065, 0.11266472935676575], "cos_Ac_negg": [0.6202460527420044, 0.6352185010910034, 0.6429893970489502], "r_norm": [0.17835202813148499, 0.17747706174850464, 0.1794249564409256], "ldr": {"dL_sdil": -0.006239660084247589, "dL_bp": -0.009388342499732971, "ratio": 0.6646178582028789}}, {"step": 4100, "epoch": 8, "train_loss": 0.01696491427719593, "cos_r_negg": [0.40331804752349854, 0.43599748611450195, 0.4769810736179352], "cos_innovation_negg": [0.40331804752349854, 0.43599748611450195, 0.4769810736179352], "cos_apical_negg": [0.0021792910993099213, 0.048161644488573074, 0.10694894939661026], "cos_Ac_negg": [0.6172534823417664, 0.6348395347595215, 0.6427741050720215], "r_norm": [0.18196439743041992, 0.17999687790870667, 0.18242529034614563]}, {"step": 4200, "epoch": 8, "train_loss": 0.006257995031774044, "cos_r_negg": [0.37441813945770264, 0.41094475984573364, 0.45740368962287903], "cos_innovation_negg": [0.37441813945770264, 0.41094475984573364, 0.45740368962287903], "cos_apical_negg": [0.007493656128644943, 0.0479060634970665, 0.10380347073078156], "cos_Ac_negg": [0.610884964466095, 0.6348294019699097, 0.6460144519805908], "r_norm": [0.1809980869293213, 0.17913639545440674, 0.1820543110370636]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9712, "test_loss": 0.12717454872131348}, {"step": 4300, "epoch": 9, "train_loss": 0.0011023624101653695, "cos_r_negg": [0.3483300507068634, 0.382057249546051, 0.43449127674102783], "cos_innovation_negg": [0.3483300507068634, 0.382057249546051, 0.43449127674102783], "cos_apical_negg": [0.008317005820572376, 0.049297407269477844, 0.10036052763462067], "cos_Ac_negg": [0.6123583912849426, 0.6345103979110718, 0.6484857797622681], "r_norm": [0.18181782960891724, 0.17950648069381714, 0.1819377988576889]}, {"step": 4400, "epoch": 9, "train_loss": 0.0008195400005206466, "cos_r_negg": [0.35415226221084595, 0.3898577392101288, 0.4361424744129181], "cos_innovation_negg": [0.35415226221084595, 0.3898577392101288, 0.4361424744129181], "cos_apical_negg": [0.0060727475211024284, 0.04846321791410446, 0.10123865306377411], "cos_Ac_negg": [0.614793062210083, 0.6358656883239746, 0.648384690284729], "r_norm": [0.17908819019794464, 0.17724967002868652, 0.1796587109565735]}, {"step": 4500, "epoch": 9, "train_loss": 0.010645082220435143, "cos_r_negg": [0.33410248160362244, 0.36796867847442627, 0.4165004789829254], "cos_innovation_negg": [0.33410248160362244, 0.36796867847442627, 0.4165004789829254], "cos_apical_negg": [0.007590829860419035, 0.048557184636592865, 0.10064555704593658], "cos_Ac_negg": [0.6158840656280518, 0.6369221210479736, 0.6506656408309937], "r_norm": [0.18416178226470947, 0.18281283974647522, 0.1849346160888672], "ldr": {"dL_sdil": -0.007865674793720245, "dL_bp": -0.013627573847770691, "ratio": 0.5771881981037275}}, {"step": 4600, "epoch": 9, "train_loss": 0.001619409304112196, "cos_r_negg": [0.33882132172584534, 0.37668660283088684, 0.4232252240180969], "cos_innovation_negg": [0.33882132172584534, 0.37668660283088684, 0.4232252240180969], "cos_apical_negg": [0.008896243758499622, 0.05055942013859749, 0.10552978515625], "cos_Ac_negg": [0.6186201572418213, 0.6367286443710327, 0.649141788482666], "r_norm": [0.1943180114030838, 0.1918736696243286, 0.19443336129188538]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9721, "test_loss": 0.12973360595703126}, {"step": 4700, "epoch": 10, "train_loss": 0.0021977010183036327, "cos_r_negg": [0.32870155572891235, 0.3676977753639221, 0.4149904251098633], "cos_innovation_negg": [0.32870155572891235, 0.3676977753639221, 0.4149904251098633], "cos_apical_negg": [0.0030493647791445255, 0.05364247411489487, 0.10174606740474701], "cos_Ac_negg": [0.6212400794029236, 0.6417281627655029, 0.6570532321929932], "r_norm": [0.1943536400794983, 0.19208161532878876, 0.19493183493614197]}, {"step": 4800, "epoch": 10, "train_loss": 0.0048472220078110695, "cos_r_negg": [0.3143385052680969, 0.3508572280406952, 0.40000349283218384], "cos_innovation_negg": [0.3143385052680969, 0.3508572280406952, 0.40000349283218384], "cos_apical_negg": [0.0022430382668972015, 0.050690457224845886, 0.10255388915538788], "cos_Ac_negg": [0.6173672676086426, 0.6412702798843384, 0.6560332179069519], "r_norm": [0.19109372794628143, 0.18872100114822388, 0.1914520412683487]}, {"step": 4900, "epoch": 10, "train_loss": 0.0014961820561438799, "cos_r_negg": [0.3018723726272583, 0.3383164405822754, 0.3870891332626343], "cos_innovation_negg": [0.3018723726272583, 0.3383164405822754, 0.3870891332626343], "cos_apical_negg": [-0.002491031074896455, 0.049107689410448074, 0.10321295261383057], "cos_Ac_negg": [0.6167256832122803, 0.6392117738723755, 0.655708372592926], "r_norm": [0.19168445467948914, 0.18950015306472778, 0.19210729002952576]}, {"step": 5000, "epoch": 10, "train_loss": 0.0008151744259521365, "cos_r_negg": [0.2916484475135803, 0.33086228370666504, 0.3762313425540924], "cos_innovation_negg": [0.2916484475135803, 0.33086228370666504, 0.3762313425540924], "cos_apical_negg": [-0.005651507060974836, 0.05090652406215668, 0.0991320013999939], "cos_Ac_negg": [0.6159945726394653, 0.6375445127487183, 0.6598638892173767], "r_norm": [0.18478374183177948, 0.1824667751789093, 0.18539118766784668], "ldr": {"dL_sdil": -0.008957244455814362, "dL_bp": -0.016509532928466797, "ratio": 0.5425498404239956}}, {"step": 5100, "epoch": 10, "train_loss": 0.006166377104818821, "cos_r_negg": [0.29230961203575134, 0.3324953317642212, 0.3814157247543335], "cos_innovation_negg": [0.29230961203575134, 0.3324953317642212, 0.3814157247543335], "cos_apical_negg": [-0.008434660732746124, 0.0491769015789032, 0.10000497102737427], "cos_Ac_negg": [0.6172416806221008, 0.6376235485076904, 0.6609718799591064], "r_norm": [0.17606347799301147, 0.1740131825208664, 0.17657995223999023]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9715, "test_loss": 0.12970616302490234}, {"step": 5200, "epoch": 11, "train_loss": 0.0005540874553844333, "cos_r_negg": [0.29465028643608093, 0.33385953307151794, 0.3777812421321869], "cos_innovation_negg": [0.29465028643608093, 0.33385953307151794, 0.3777812421321869], "cos_apical_negg": [-0.0082553057000041, 0.04947355017066002, 0.09809713065624237], "cos_Ac_negg": [0.6166052222251892, 0.6384375691413879, 0.6590942144393921], "r_norm": [0.17594361305236816, 0.17453598976135254, 0.1756635457277298]}, {"step": 5300, "epoch": 11, "train_loss": 0.0010943440720438957, "cos_r_negg": [0.28236421942710876, 0.3227180242538452, 0.36728060245513916], "cos_innovation_negg": [0.28236421942710876, 0.3227180242538452, 0.36728060245513916], "cos_apical_negg": [-0.008325804956257343, 0.048340387642383575, 0.09489801526069641], "cos_Ac_negg": [0.609477162361145, 0.6339924931526184, 0.6580001711845398], "r_norm": [0.17443129420280457, 0.1728535145521164, 0.17435342073440552]}, {"step": 5400, "epoch": 11, "train_loss": 0.0020020240917801857, "cos_r_negg": [0.27028995752334595, 0.3119666278362274, 0.35435551404953003], "cos_innovation_negg": [0.27028995752334595, 0.3119666278362274, 0.35435551404953003], "cos_apical_negg": [-0.006102823652327061, 0.04947243630886078, 0.09166481345891953], "cos_Ac_negg": [0.6115628480911255, 0.6343180537223816, 0.6585925221443176], "r_norm": [0.18084609508514404, 0.17881640791893005, 0.18106326460838318]}, {"step": 5500, "epoch": 11, "train_loss": 0.0023182944860309362, "cos_r_negg": [0.26846668124198914, 0.31433433294296265, 0.3568243980407715], "cos_innovation_negg": [0.26846668124198914, 0.31433433294296265, 0.3568243980407715], "cos_apical_negg": [-0.009181167930364609, 0.050728313624858856, 0.09624015539884567], "cos_Ac_negg": [0.6113748550415039, 0.6360570192337036, 0.6586302518844604], "r_norm": [0.1936633437871933, 0.19143393635749817, 0.19355353713035583], "ldr": {"dL_sdil": -0.009663648903369904, "dL_bp": -0.018513984978199005, "ratio": 0.5219648236049266}}, {"step": 5600, "epoch": 11, "train_loss": 0.0003160181222483516, "cos_r_negg": [0.2535359859466553, 0.29683980345726013, 0.33718693256378174], "cos_innovation_negg": [0.2535359859466553, 0.29683980345726013, 0.33718693256378174], "cos_apical_negg": [-0.009216856211423874, 0.04909816384315491, 0.0910632386803627], "cos_Ac_negg": [0.6061104536056519, 0.6344951391220093, 0.6641950011253357], "r_norm": [0.18347963690757751, 0.18202190101146698, 0.1838439702987671]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9727, "test_loss": 0.13577726669311524}, {"step": 5700, "epoch": 12, "train_loss": 0.000947620952501893, "cos_r_negg": [0.25386619567871094, 0.2974931001663208, 0.3395717144012451], "cos_innovation_negg": [0.25386619567871094, 0.2974931001663208, 0.3395717144012451], "cos_apical_negg": [-0.01228018943220377, 0.04886765405535698, 0.09168899059295654], "cos_Ac_negg": [0.6076353788375854, 0.6335341930389404, 0.6627957820892334], "r_norm": [0.1760621815919876, 0.1745014786720276, 0.17572079598903656]}, {"step": 5800, "epoch": 12, "train_loss": 0.00036313594318926334, "cos_r_negg": [0.24559952318668365, 0.2888130843639374, 0.33031395077705383], "cos_innovation_negg": [0.24559952318668365, 0.2888130843639374, 0.33031395077705383], "cos_apical_negg": [-0.013709278777241707, 0.04781469330191612, 0.0893857330083847], "cos_Ac_negg": [0.6113327741622925, 0.6327759027481079, 0.6631534099578857], "r_norm": [0.18037748336791992, 0.17876042425632477, 0.18040570616722107]}, {"step": 5900, "epoch": 12, "train_loss": 0.003891424275934696, "cos_r_negg": [0.24846214056015015, 0.2900077998638153, 0.33367830514907837], "cos_innovation_negg": [0.24846214056015015, 0.2900077998638153, 0.33367830514907837], "cos_apical_negg": [-0.010402785614132881, 0.04911920055747032, 0.09120796620845795], "cos_Ac_negg": [0.6064143776893616, 0.6320850849151611, 0.6621131896972656], "r_norm": [0.1823580116033554, 0.1804198920726776, 0.1822509765625]}, {"step": 6000, "epoch": 12, "train_loss": 0.00027683001826517284, "cos_r_negg": [0.24804182350635529, 0.2855919599533081, 0.3256784677505493], "cos_innovation_negg": [0.24804182350635529, 0.2855919599533081, 0.3256784677505493], "cos_apical_negg": [-0.008538182824850082, 0.048475660383701324, 0.08708613365888596], "cos_Ac_negg": [0.6110955476760864, 0.6319037675857544, 0.6617473363876343], "r_norm": [0.17901301383972168, 0.17723117768764496, 0.17906996607780457], "ldr": {"dL_sdil": -0.009073600172996521, "dL_bp": -0.01850360631942749, "ratio": 0.4903692834985295}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9729, "test_loss": 0.13838680152893065}, {"step": 6100, "epoch": 13, "train_loss": 0.0006825184100307524, "cos_r_negg": [0.24415338039398193, 0.28162163496017456, 0.32339686155319214], "cos_innovation_negg": [0.24415338039398193, 0.28162163496017456, 0.32339686155319214], "cos_apical_negg": [-0.004759262781590223, 0.04672665894031525, 0.08735356479883194], "cos_Ac_negg": [0.6095070838928223, 0.630984902381897, 0.6619540452957153], "r_norm": [0.18518301844596863, 0.18320360779762268, 0.1852014660835266]}, {"step": 6200, "epoch": 13, "train_loss": 0.0008595471736043692, "cos_r_negg": [0.23484259843826294, 0.2732177972793579, 0.3125447928905487], "cos_innovation_negg": [0.23484259843826294, 0.2732177972793579, 0.3125447928905487], "cos_apical_negg": [-0.0061957696452736855, 0.04641520231962204, 0.08374373614788055], "cos_Ac_negg": [0.606260359287262, 0.6301473379135132, 0.6641076803207397], "r_norm": [0.18128171563148499, 0.17957058548927307, 0.1814786195755005]}, {"step": 6300, "epoch": 13, "train_loss": 0.0006934639532119036, "cos_r_negg": [0.2264966070652008, 0.26610812544822693, 0.30659109354019165], "cos_innovation_negg": [0.2264966070652008, 0.26610812544822693, 0.30659109354019165], "cos_apical_negg": [-0.008863432332873344, 0.04641268402338028, 0.08453816175460815], "cos_Ac_negg": [0.6030253171920776, 0.626876950263977, 0.6629367470741272], "r_norm": [0.17757649719715118, 0.17598430812358856, 0.17771857976913452]}, {"step": 6400, "epoch": 13, "train_loss": 0.0002877316437661648, "cos_r_negg": [0.23091086745262146, 0.2681494355201721, 0.3060353994369507], "cos_innovation_negg": [0.23091086745262146, 0.2681494355201721, 0.3060353994369507], "cos_apical_negg": [-0.006315913982689381, 0.04519297927618027, 0.08115345239639282], "cos_Ac_negg": [0.6015437841415405, 0.6266802549362183, 0.6626092195510864], "r_norm": [0.17999890446662903, 0.1781638264656067, 0.17986194789409637]}, {"step": 6500, "epoch": 13, "train_loss": 0.0003943365009035915, "cos_r_negg": [0.2314969003200531, 0.26934683322906494, 0.30806347727775574], "cos_innovation_negg": [0.2314969003200531, 0.26934683322906494, 0.30806347727775574], "cos_apical_negg": [-0.006277681328356266, 0.04623973369598389, 0.08112269639968872], "cos_Ac_negg": [0.6035950183868408, 0.6283788681030273, 0.663296103477478], "r_norm": [0.18169091641902924, 0.1796780228614807, 0.18181093037128448], "ldr": {"dL_sdil": -0.009350590407848358, "dL_bp": -0.019655853509902954, "ratio": 0.4757153080703095}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9724, "test_loss": 0.14104432067871095}, {"step": 6600, "epoch": 14, "train_loss": 0.00024284770188387483, "cos_r_negg": [0.22123859822750092, 0.2598249316215515, 0.29724153876304626], "cos_innovation_negg": [0.22123859822750092, 0.2598249316215515, 0.29724153876304626], "cos_apical_negg": [-0.008594653569161892, 0.045161910355091095, 0.07877635955810547], "cos_Ac_negg": [0.5981976985931396, 0.6258737444877625, 0.662596583366394], "r_norm": [0.18268665671348572, 0.1807129681110382, 0.1825793981552124]}, {"step": 6700, "epoch": 14, "train_loss": 0.000819356762804091, "cos_r_negg": [0.21892967820167542, 0.25876832008361816, 0.2971835136413574], "cos_innovation_negg": [0.21892967820167542, 0.25876832008361816, 0.2971835136413574], "cos_apical_negg": [-0.008528337813913822, 0.04554855823516846, 0.07917456328868866], "cos_Ac_negg": [0.5992168188095093, 0.6250213384628296, 0.6616549491882324], "r_norm": [0.18037615716457367, 0.17839638888835907, 0.1804060935974121]}, {"step": 6800, "epoch": 14, "train_loss": 0.0006911448435857892, "cos_r_negg": [0.21183283627033234, 0.2506535053253174, 0.2896501421928406], "cos_innovation_negg": [0.21183283627033234, 0.2506535053253174, 0.2896501421928406], "cos_apical_negg": [-0.008370094001293182, 0.04392202943563461, 0.07889318466186523], "cos_Ac_negg": [0.5985738039016724, 0.6242616176605225, 0.6628355979919434], "r_norm": [0.17608055472373962, 0.17446398735046387, 0.17615440487861633]}, {"step": 6900, "epoch": 14, "train_loss": 0.0008076176745817065, "cos_r_negg": [0.21059472858905792, 0.24934405088424683, 0.28680405020713806], "cos_innovation_negg": [0.21059472858905792, 0.24934405088424683, 0.28680405020713806], "cos_apical_negg": [-0.007636239752173424, 0.043772052973508835, 0.07701065391302109], "cos_Ac_negg": [0.5949746966362, 0.6231884956359863, 0.6632137298583984], "r_norm": [0.1791018843650818, 0.1774611920118332, 0.17879526317119598]}, {"step": 7000, "epoch": 14, "train_loss": 0.0006948065129108727, "cos_r_negg": [0.2116452157497406, 0.250020295381546, 0.28750133514404297], "cos_innovation_negg": [0.2116452157497406, 0.250020295381546, 0.28750133514404297], "cos_apical_negg": [-0.007414512801915407, 0.04454676806926727, 0.07724614441394806], "cos_Ac_negg": [0.5964549779891968, 0.6238608360290527, 0.662742555141449], "r_norm": [0.18482258915901184, 0.18285025656223297, 0.1847810596227646], "ldr": {"dL_sdil": -0.009862080216407776, "dL_bp": -0.021312817931175232, "ratio": 0.46272999883239563}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9728, "test_loss": 0.1435254264831543}], "final": {"test_acc": 0.9728, "test_loss": 0.1435254264831543, "wall_s": 23.386398792266846, "cos_r_negg": [0.21129368245601654, 0.24853135645389557, 0.28675156831741333], "cos_innovation_negg": [0.21129368245601654, 0.24853135645389557, 0.28675156831741333], "cos_apical_negg": [-0.006689662579447031, 0.044270001351833344, 0.077663853764534], "cos_Ac_negg": [0.5961201190948486, 0.6233915090560913, 0.6630139350891113]}}
|