1
|
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "match_innovation_norm", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "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_matched_s1", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.327335834503174, "cos_r_negg": [0.01368984766304493, -0.012001646682620049, -0.008926067501306534], "cos_innovation_negg": [0.026290064677596092, -0.012671238742768764, -0.006505622528493404], "cos_apical_negg": [0.01368984766304493, -0.012001646682620049, -0.008926068432629108], "cos_Ac_negg": [0.02628779411315918, -0.01267300546169281, -0.006506332661956549], "r_norm": [4.743828773498535, 4.803375244140625, 4.794331073760986], "ldr": {"dL_sdil": 0.021651029586791992, "dL_bp": -0.177276611328125, "ratio": -0.12213133714925116}}, {"step": 100, "epoch": 0, "train_loss": 8.629837036132812, "cos_r_negg": [0.037567175924777985, 0.008364958688616753, -0.5287129878997803], "cos_innovation_negg": [0.04334171861410141, 0.22261875867843628, 0.14754386246204376], "cos_apical_negg": [0.037567175924777985, 0.008364951238036156, -0.5287129282951355], "cos_Ac_negg": [0.04345659911632538, 0.22264283895492554, 0.14739736914634705], "r_norm": [5.985307693481445, 6.0046281814575195, 5.903804779052734]}, {"step": 200, "epoch": 0, "train_loss": 8.29200553894043, "cos_r_negg": [-0.05966085195541382, -0.09444378316402435, -0.09045639634132385], "cos_innovation_negg": [-0.09845882654190063, -0.06108493357896805, 0.19033609330654144], "cos_apical_negg": [-0.05966084823012352, -0.09444377571344376, -0.09045639634132385], "cos_Ac_negg": [-0.09845120459794998, -0.06108956038951874, 0.190270334482193], "r_norm": [4.763657569885254, 4.891393661499023, 4.839957237243652]}, {"step": 300, "epoch": 0, "train_loss": 14.548294067382812, "cos_r_negg": [-0.12011945992708206, -0.17529357969760895, -0.24218785762786865], "cos_innovation_negg": [-0.13343249261379242, -0.026975946500897408, 0.3038310706615448], "cos_apical_negg": [-0.12011946737766266, -0.17529356479644775, -0.24218782782554626], "cos_Ac_negg": [-0.13343918323516846, -0.02698030322790146, 0.3037066161632538], "r_norm": [4.711991786956787, 4.9175238609313965, 4.900652885437012]}, {"step": 400, "epoch": 0, "train_loss": 10.571653366088867, "cos_r_negg": [-0.07457352429628372, -0.24640749394893646, -0.37246787548065186], "cos_innovation_negg": [-0.11332274973392487, -0.2235855758190155, -0.0984620451927185], "cos_apical_negg": [-0.07457352429628372, -0.24640749394893646, -0.37246787548065186], "cos_Ac_negg": [-0.07068943977355957, -0.210202157497406, 0.04788782447576523], "r_norm": [4.917359828948975, 5.2306671142578125, 5.053191184997559]}, {"epoch_end": 0, "step": 469, "test_acc": 0.0892, "test_loss": 11.37957666015625}, {"step": 500, "epoch": 1, "train_loss": 11.507259368896484, "cos_r_negg": [0.0044118864461779594, -0.029640208929777145, -0.5220323801040649], "cos_innovation_negg": [-0.008758346550166607, 0.03197118267416954, 0.22668370604515076], "cos_apical_negg": [0.004411884117871523, -0.02964020147919655, -0.5220324397087097], "cos_Ac_negg": [-0.008773845620453358, 0.03197310492396355, 0.22666186094284058], "r_norm": [4.150497913360596, 4.425154209136963, 4.40374231338501], "ldr": {"dL_sdil": -0.8359193801879883, "dL_bp": -0.8359193801879883, "ratio": 1.0}}, {"step": 600, "epoch": 1, "train_loss": 8.379337310791016, "cos_r_negg": [-0.04307126998901367, -0.21659450232982635, -0.1836141049861908], "cos_innovation_negg": [-0.08859698474407196, 0.11656951904296875, 0.28489020466804504], "cos_apical_negg": [-0.04307126998901367, -0.21659448742866516, -0.1836141049861908], "cos_Ac_negg": [-0.08861087262630463, 0.1165703535079956, 0.28487831354141235], "r_norm": [3.9001951217651367, 4.099247932434082, 4.126674175262451]}, {"step": 700, "epoch": 1, "train_loss": 10.07689380645752, "cos_r_negg": [0.014078537002205849, 0.016958516091108322, -0.04757288843393326], "cos_innovation_negg": [2.461625263094902e-05, 0.192184180021286, 0.4833545982837677], "cos_apical_negg": [0.014078537002205849, 0.016958512365818024, -0.047572895884513855], "cos_Ac_negg": [1.161079853773117e-05, 0.192181795835495, 0.4833124279975891], "r_norm": [3.9420077800750732, 4.221686363220215, 4.267495155334473]}, {"step": 800, "epoch": 1, "train_loss": 6.442304611206055, "cos_r_negg": [0.037889618426561356, -0.06307996809482574, -0.063703253865242], "cos_innovation_negg": [-0.04553075507283211, 0.10371538996696472, 0.2536274194717407], "cos_apical_negg": [0.037889622151851654, -0.06307997554540634, -0.06370328366756439], "cos_Ac_negg": [-0.045541077852249146, 0.1037130281329155, 0.2535904049873352], "r_norm": [3.5866873264312744, 3.9869673252105713, 4.057910442352295]}, {"step": 900, "epoch": 1, "train_loss": 15.13424301147461, "cos_r_negg": [0.030607381835579872, -0.2586507201194763, -0.4286853075027466], "cos_innovation_negg": [0.014614340849220753, -0.10848779231309891, 0.6467256546020508], "cos_apical_negg": [0.03060738369822502, -0.2586507201194763, -0.4286853075027466], "cos_Ac_negg": [0.014606429263949394, -0.1084863692522049, 0.6467254757881165], "r_norm": [3.4449856281280518, 3.8409712314605713, 3.8702330589294434]}, {"epoch_end": 1, "step": 938, "test_acc": 0.0892, "test_loss": 5.97907939453125}, {"step": 1000, "epoch": 2, "train_loss": 16.947219848632812, "cos_r_negg": [-0.05575894936919212, -0.058172352612018585, -0.10303526371717453], "cos_innovation_negg": [-0.10887196660041809, -0.09698377549648285, 0.41782981157302856], "cos_apical_negg": [-0.05575894936919212, -0.058172356337308884, -0.10303527116775513], "cos_Ac_negg": [-0.1088702529668808, -0.09699638187885284, 0.4178636074066162], "r_norm": [3.4725301265716553, 3.893908977508545, 3.9083449840545654], "ldr": {"dL_sdil": -1.2378950119018555, "dL_bp": -1.2378950119018555, "ratio": 1.0}}, {"step": 1100, "epoch": 2, "train_loss": 14.489168167114258, "cos_r_negg": [0.007330155000090599, -0.2059539258480072, -0.6385909914970398], "cos_innovation_negg": [-0.058647677302360535, 0.10369955003261566, 0.1521846204996109], "cos_apical_negg": [0.007330155000090599, -0.2059539258480072, -0.6385909914970398], "cos_Ac_negg": [-0.058645546436309814, 0.10369760543107986, 0.15215173363685608], "r_norm": [3.404529094696045, 3.8415870666503906, 3.890641212463379]}, {"step": 1200, "epoch": 2, "train_loss": 11.066431045532227, "cos_r_negg": [0.040565118193626404, -0.26042842864990234, -0.30299752950668335], "cos_innovation_negg": [0.061117686331272125, 0.14500866830348969, 0.5352002382278442], "cos_apical_negg": [0.040565118193626404, -0.26042845845222473, -0.30299752950668335], "cos_Ac_negg": [0.06110852211713791, 0.14500820636749268, 0.5352024435997009], "r_norm": [2.9809021949768066, 3.4198238849639893, 3.4656436443328857]}, {"step": 1300, "epoch": 2, "train_loss": 10.231103897094727, "cos_r_negg": [0.01986650936305523, -0.09248112887144089, -0.06143999099731445], "cos_innovation_negg": [0.03777393698692322, 0.3218173682689667, 0.3766866624355316], "cos_apical_negg": [0.01986650750041008, -0.09248113632202148, -0.061439987272024155], "cos_Ac_negg": [0.03776753321290016, 0.32181769609451294, 0.3766733407974243], "r_norm": [2.943350076675415, 3.4399678707122803, 3.53225040435791]}, {"step": 1400, "epoch": 2, "train_loss": 15.886367797851562, "cos_r_negg": [0.03741500899195671, -0.041028186678886414, -0.22061492502689362], "cos_innovation_negg": [0.04061534255743027, 0.5060548782348633, 0.49722927808761597], "cos_apical_negg": [0.03741501271724701, -0.04102819040417671, -0.220614954829216], "cos_Ac_negg": [0.040607623755931854, 0.5060560703277588, 0.4972347617149353], "r_norm": [2.889829635620117, 3.452561855316162, 3.5183520317077637]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.098, "test_loss": 8.59035224609375}, {"step": 1500, "epoch": 3, "train_loss": 15.093576431274414, "cos_r_negg": [0.02216712012887001, -0.21593934297561646, -0.1491767168045044], "cos_innovation_negg": [0.018002334982156754, -0.013630302622914314, 0.025494003668427467], "cos_apical_negg": [0.022167127579450607, -0.21593934297561646, -0.1491767317056656], "cos_Ac_negg": [0.017906315624713898, -0.013631798326969147, 0.025503316894173622], "r_norm": [3.0101892948150635, 3.587337017059326, 3.8490867614746094], "ldr": {"dL_sdil": -2.548330307006836, "dL_bp": -2.548330307006836, "ratio": 1.0}}, {"step": 1600, "epoch": 3, "train_loss": 15.974457740783691, "cos_r_negg": [0.05967242270708084, -0.1192544624209404, -0.6155177354812622], "cos_innovation_negg": [0.12467913329601288, 0.18524280190467834, 0.41068893671035767], "cos_apical_negg": [0.05967242270708084, -0.1192544624209404, -0.6155176758766174], "cos_Ac_negg": [0.1246727854013443, 0.18524479866027832, 0.4107542335987091], "r_norm": [2.892625093460083, 3.2755002975463867, 3.376058578491211]}, {"step": 1700, "epoch": 3, "train_loss": 11.454575538635254, "cos_r_negg": [0.05775633454322815, -0.1839609146118164, -0.4234389662742615], "cos_innovation_negg": [-0.02099188044667244, -0.01595008373260498, 0.4565528631210327], "cos_apical_negg": [0.05775633454322815, -0.1839609295129776, -0.4234389662742615], "cos_Ac_negg": [-0.021001555025577545, -0.015945378690958023, 0.45654499530792236], "r_norm": [2.5749480724334717, 2.916334629058838, 3.2268168926239014]}, {"step": 1800, "epoch": 3, "train_loss": 12.229915618896484, "cos_r_negg": [0.08983103930950165, -0.10578735172748566, -0.16379764676094055], "cos_innovation_negg": [0.10483919829130173, 0.28883856534957886, 0.44382697343826294], "cos_apical_negg": [0.08983103930950165, -0.10578733682632446, -0.16379764676094055], "cos_Ac_negg": [0.10483229160308838, 0.2888367474079132, 0.44381600618362427], "r_norm": [2.640359878540039, 2.885666608810425, 3.2105512619018555]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.098, "test_loss": 10.5088134765625}, {"step": 1900, "epoch": 4, "train_loss": 17.48843002319336, "cos_r_negg": [0.012519306503236294, -0.4328705668449402, -0.5602304935455322], "cos_innovation_negg": [-0.049148138612508774, 0.28105664253234863, 0.7195155620574951], "cos_apical_negg": [0.012519307434558868, -0.4328705668449402, -0.5602304339408875], "cos_Ac_negg": [-0.04915396124124527, 0.2810572683811188, 0.7195122241973877], "r_norm": [2.3489015102386475, 2.694956064224243, 2.9692583084106445]}, {"step": 2000, "epoch": 4, "train_loss": 9.76813793182373, "cos_r_negg": [-0.018307412043213844, 0.06446070969104767, -0.10387585312128067], "cos_innovation_negg": [-0.0366482250392437, 0.2162434458732605, 0.6224699020385742], "cos_apical_negg": [-0.018307410180568695, 0.06446070969104767, -0.10387587547302246], "cos_Ac_negg": [-0.03666269779205322, 0.21624931693077087, 0.622467041015625], "r_norm": [2.511305332183838, 2.9758949279785156, 3.3254880905151367], "ldr": {"dL_sdil": -1.9794464111328125, "dL_bp": -1.9794464111328125, "ratio": 1.0}}, {"step": 2100, "epoch": 4, "train_loss": 7.482757568359375, "cos_r_negg": [0.03618372604250908, -0.13149124383926392, -0.48567989468574524], "cos_innovation_negg": [0.07447533309459686, 0.06922244280576706, 0.45074909925460815], "cos_apical_negg": [0.03618372604250908, -0.13149124383926392, -0.4856799244880676], "cos_Ac_negg": [0.07447478175163269, 0.06922130286693573, 0.4507486820220947], "r_norm": [2.409442901611328, 2.776766777038574, 3.0350341796875]}, {"step": 2200, "epoch": 4, "train_loss": 13.395278930664062, "cos_r_negg": [0.030296243727207184, -0.2916486859321594, -0.4456755518913269], "cos_innovation_negg": [0.0532541424036026, -0.2333734631538391, 0.8114545345306396], "cos_apical_negg": [0.030296241864562035, -0.2916486859321594, -0.4456755518913269], "cos_Ac_negg": [0.053279198706150055, -0.23337127268314362, 0.81145179271698], "r_norm": [2.322329521179199, 2.8261942863464355, 3.2641398906707764]}, {"step": 2300, "epoch": 4, "train_loss": 9.56633186340332, "cos_r_negg": [-0.0007144114933907986, -0.08392734825611115, -0.31144028902053833], "cos_innovation_negg": [0.02680947631597519, -0.05834192782640457, 0.5731551647186279], "cos_apical_negg": [-0.000714409165084362, -0.08392734825611115, -0.31144025921821594], "cos_Ac_negg": [0.02680298313498497, -0.058338649570941925, 0.5731555223464966], "r_norm": [1.9384572505950928, 2.2706122398376465, 2.636899709701538]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.0892, "test_loss": 11.87872587890625}, {"step": 2400, "epoch": 5, "train_loss": 15.17947769165039, "cos_r_negg": [0.027683284133672714, -0.12447544932365417, -0.5601911544799805], "cos_innovation_negg": [-0.037985533475875854, -0.10906483232975006, 0.2888246178627014], "cos_apical_negg": [0.027683287858963013, -0.12447544187307358, -0.56019127368927], "cos_Ac_negg": [-0.0379890538752079, -0.1090637668967247, 0.28882139921188354], "r_norm": [1.9936455488204956, 2.347970485687256, 2.744025468826294]}, {"step": 2500, "epoch": 5, "train_loss": 12.193578720092773, "cos_r_negg": [-0.0026660682633519173, -0.18455472588539124, -0.259928822517395], "cos_innovation_negg": [-0.03150944411754608, 0.10283806920051575, 0.5210461616516113], "cos_apical_negg": [-0.0026660673320293427, -0.18455474078655243, -0.2599288523197174], "cos_Ac_negg": [-0.03151656687259674, 0.1028411015868187, 0.5210443735122681], "r_norm": [1.9233973026275635, 2.319603443145752, 2.8037376403808594], "ldr": {"dL_sdil": -2.6694507598876953, "dL_bp": -2.6694507598876953, "ratio": 1.0}}, {"step": 2600, "epoch": 5, "train_loss": 13.220319747924805, "cos_r_negg": [0.02900480292737484, -0.2758943438529968, -0.4539209306240082], "cos_innovation_negg": [-0.011245804838836193, -0.12144125998020172, 0.09843893349170685], "cos_apical_negg": [0.02900479920208454, -0.2758943438529968, -0.45392096042633057], "cos_Ac_negg": [-0.011244049295783043, -0.12144289910793304, 0.09843729436397552], "r_norm": [1.8653290271759033, 2.214156150817871, 2.741147518157959]}, {"step": 2700, "epoch": 5, "train_loss": 10.434335708618164, "cos_r_negg": [0.05317149683833122, -0.11676039546728134, -0.3232342302799225], "cos_innovation_negg": [-0.011293860152363777, 0.3770521879196167, 0.7010574340820312], "cos_apical_negg": [0.05317150056362152, -0.11676039546728134, -0.3232342302799225], "cos_Ac_negg": [-0.011303890496492386, 0.37705329060554504, 0.7010572552680969], "r_norm": [1.9065091609954834, 2.2450571060180664, 2.7173304557800293]}, {"step": 2800, "epoch": 5, "train_loss": 8.608712196350098, "cos_r_negg": [0.021214831620454788, -0.22861573100090027, -0.27367472648620605], "cos_innovation_negg": [0.025810478255152702, 0.183097243309021, 0.43575942516326904], "cos_apical_negg": [0.02121482789516449, -0.22861573100090027, -0.27367472648620605], "cos_Ac_negg": [0.02580975368618965, 0.18310371041297913, 0.4357585310935974], "r_norm": [1.9713449478149414, 2.2843751907348633, 2.9697022438049316]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.101, "test_loss": 18.1558404296875}, {"step": 2900, "epoch": 6, "train_loss": 10.255294799804688, "cos_r_negg": [0.03914017975330353, -0.0356130450963974, -0.04506968706846237], "cos_innovation_negg": [-0.03473691642284393, 0.1018490195274353, 0.42403528094291687], "cos_apical_negg": [0.039140187203884125, -0.0356130450963974, -0.045069705694913864], "cos_Ac_negg": [-0.03474500775337219, 0.10183992981910706, 0.4240354597568512], "r_norm": [1.725934624671936, 2.0168943405151367, 2.6580262184143066]}, {"step": 3000, "epoch": 6, "train_loss": 12.970020294189453, "cos_r_negg": [-0.004842756316065788, -0.12890617549419403, -0.6363621950149536], "cos_innovation_negg": [-0.047544751316308975, -0.005340326577425003, 0.28954434394836426], "cos_apical_negg": [-0.004842751659452915, -0.12890617549419403, -0.6363622546195984], "cos_Ac_negg": [-0.0472957044839859, -0.005311688408255577, 0.28955477476119995], "r_norm": [2.0303399562835693, 2.3841023445129395, 3.0937414169311523], "ldr": {"dL_sdil": -4.037936210632324, "dL_bp": -4.037936210632324, "ratio": 1.0}}, {"step": 3100, "epoch": 6, "train_loss": 14.473289489746094, "cos_r_negg": [0.04678209498524666, -0.010735692456364632, -0.17836451530456543], "cos_innovation_negg": [0.042138442397117615, -0.028546959161758423, 0.29168370366096497], "cos_apical_negg": [0.04678209871053696, -0.010735689662396908, -0.17836448550224304], "cos_Ac_negg": [0.042134158313274384, -0.028544962406158447, 0.291683167219162], "r_norm": [1.7289745807647705, 2.0565102100372314, 2.7483131885528564]}, {"step": 3200, "epoch": 6, "train_loss": 9.455717086791992, "cos_r_negg": [0.006103653460741043, -0.08823632448911667, -0.36146461963653564], "cos_innovation_negg": [-0.025493118911981583, -0.25060561299324036, 0.5729857683181763], "cos_apical_negg": [0.0061036525294184685, -0.08823632448911667, -0.36146461963653564], "cos_Ac_negg": [-0.0254963506013155, -0.25061002373695374, 0.5729857683181763], "r_norm": [1.564295768737793, 1.8498930931091309, 2.5144753456115723]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.1135, "test_loss": 8.644243359375}, {"step": 3300, "epoch": 7, "train_loss": 10.686386108398438, "cos_r_negg": [0.007331964559853077, -0.1437041163444519, -0.4923170804977417], "cos_innovation_negg": [-0.017183762043714523, 0.2204362154006958, 0.8977514505386353], "cos_apical_negg": [0.007331966422498226, -0.1437041312456131, -0.4923170804977417], "cos_Ac_negg": [-0.017086660489439964, 0.2204228937625885, 0.8977519273757935], "r_norm": [2.1216511726379395, 2.4493775367736816, 3.102795362472534]}, {"step": 3400, "epoch": 7, "train_loss": 18.421329498291016, "cos_r_negg": [0.046245694160461426, -0.17511767148971558, -0.40243658423423767], "cos_innovation_negg": [0.05551280826330185, 0.4356822371482849, 0.42793047428131104], "cos_apical_negg": [0.04624570161104202, -0.17511765658855438, -0.4024365544319153], "cos_Ac_negg": [0.05310817435383797, 0.43637800216674805, 0.4279768466949463], "r_norm": [1.860713243484497, 2.2496790885925293, 3.145846366882324]}, {"step": 3500, "epoch": 7, "train_loss": 6.89385986328125, "cos_r_negg": [0.06364455074071884, -0.23486125469207764, -0.5209711790084839], "cos_innovation_negg": [0.032999224960803986, -0.09086863696575165, 0.5897722244262695], "cos_apical_negg": [0.06364455074071884, -0.23486125469207764, -0.5209711790084839], "cos_Ac_negg": [0.03298337385058403, -0.09088343381881714, 0.5897735357284546], "r_norm": [1.8219056129455566, 2.113788604736328, 2.887299060821533], "ldr": {"dL_sdil": -2.7791242599487305, "dL_bp": -2.779125213623047, "ratio": 0.9999996568437033}}, {"step": 3600, "epoch": 7, "train_loss": 10.256458282470703, "cos_r_negg": [0.020578347146511078, -0.14359118044376373, -0.3098403215408325], "cos_innovation_negg": [0.06512169539928436, -0.17804598808288574, 0.6493154764175415], "cos_apical_negg": [0.020578349009156227, -0.14359118044376373, -0.3098403811454773], "cos_Ac_negg": [0.06512264907360077, -0.1780502200126648, 0.6493154764175415], "r_norm": [1.506436824798584, 1.770148754119873, 2.5789108276367188]}, {"step": 3700, "epoch": 7, "train_loss": 10.41223430633545, "cos_r_negg": [0.008857296779751778, -0.2126459926366806, -0.5968126058578491], "cos_innovation_negg": [0.04915352910757065, 0.2753782570362091, 0.2443518340587616], "cos_apical_negg": [0.008857302367687225, -0.2126459777355194, -0.5968126654624939], "cos_Ac_negg": [0.04870340973138809, 0.2754344344139099, 0.24440880119800568], "r_norm": [1.762682557106018, 2.0029869079589844, 3.0203115940093994]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.0982, "test_loss": 11.48839501953125}, {"step": 3800, "epoch": 8, "train_loss": 13.540163040161133, "cos_r_negg": [-0.02213909476995468, -0.07909062504768372, -0.6160959005355835], "cos_innovation_negg": [-0.02084731124341488, 0.12267974764108658, 0.3846050500869751], "cos_apical_negg": [-0.02213909476995468, -0.07909061759710312, -0.6160959005355835], "cos_Ac_negg": [-0.020849252119660378, 0.12268407642841339, 0.3846050500869751], "r_norm": [1.7494614124298096, 1.9818449020385742, 3.011733055114746]}, {"step": 3900, "epoch": 8, "train_loss": 9.196820259094238, "cos_r_negg": [-0.0031318701803684235, -0.15781190991401672, -0.21378548443317413], "cos_innovation_negg": [-0.06686866283416748, 0.43560338020324707, 0.5173091888427734], "cos_apical_negg": [-0.003131865058094263, -0.15781190991401672, -0.21378546953201294], "cos_Ac_negg": [-0.06688021868467331, 0.4355728030204773, 0.5173093676567078], "r_norm": [1.6178021430969238, 1.7688393592834473, 2.8990988731384277]}, {"step": 4000, "epoch": 8, "train_loss": 10.564874649047852, "cos_r_negg": [0.013686101883649826, -0.18070998787879944, -0.41389593482017517], "cos_innovation_negg": [0.0857563391327858, -0.07245780527591705, 0.38016122579574585], "cos_apical_negg": [0.013686100021004677, -0.18070998787879944, -0.41389593482017517], "cos_Ac_negg": [0.08574873954057693, -0.07244805991649628, 0.38016101717948914], "r_norm": [1.4692742824554443, 1.6845234632492065, 2.6116912364959717], "ldr": {"dL_sdil": -2.3447189331054688, "dL_bp": -2.344719409942627, "ratio": 0.9999997966335945}}, {"step": 4100, "epoch": 8, "train_loss": 7.10745096206665, "cos_r_negg": [0.037930723279714584, -0.21142877638339996, -0.036002419888973236], "cos_innovation_negg": [-0.021827008575201035, 0.1278681755065918, 0.7639685869216919], "cos_apical_negg": [0.03793072700500488, -0.21142877638339996, -0.03600241243839264], "cos_Ac_negg": [-0.021835312247276306, 0.12787224352359772, 0.7639685869216919], "r_norm": [1.507741928100586, 1.8093645572662354, 2.850492000579834]}, {"step": 4200, "epoch": 8, "train_loss": 14.092636108398438, "cos_r_negg": [0.06115858256816864, -0.2938695549964905, -0.5185893774032593], "cos_innovation_negg": [0.09283484518527985, 0.08690566569566727, 0.6516421437263489], "cos_apical_negg": [0.06115858256816864, -0.2938695549964905, -0.5185893774032593], "cos_Ac_negg": [0.0929117202758789, 0.08681941777467728, 0.6516423225402832], "r_norm": [1.6354327201843262, 1.876108169555664, 2.946805953979492]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.1028, "test_loss": 17.0794158203125}, {"step": 4300, "epoch": 9, "train_loss": 5.8485517501831055, "cos_r_negg": [0.02849375270307064, 0.0641171932220459, 0.05944167077541351], "cos_innovation_negg": [0.08906682580709457, 0.1299668848514557, 0.7520847916603088], "cos_apical_negg": [0.02849375270307064, 0.0641172006726265, 0.059441663324832916], "cos_Ac_negg": [0.08903417736291885, 0.12996794283390045, 0.7520849704742432], "r_norm": [1.6232593059539795, 1.9098446369171143, 3.160547971725464]}, {"step": 4400, "epoch": 9, "train_loss": 11.522689819335938, "cos_r_negg": [0.0014117984101176262, -0.05524075776338577, -0.034860242158174515], "cos_innovation_negg": [-0.03014400601387024, -0.0960920974612236, 0.7208172082901001], "cos_apical_negg": [0.0014117946848273277, -0.055240750312805176, -0.03486024588346481], "cos_Ac_negg": [-0.030146276578307152, -0.09609593451023102, 0.7208172082901001], "r_norm": [1.3885986804962158, 1.6415561437606812, 2.7161850929260254]}, {"step": 4500, "epoch": 9, "train_loss": 7.704092025756836, "cos_r_negg": [0.009524454362690449, 0.025368399918079376, -0.03347320854663849], "cos_innovation_negg": [0.0515335313975811, 0.05315358191728592, 0.4637925624847412], "cos_apical_negg": [0.00952446274459362, 0.025368396192789078, -0.03347321227192879], "cos_Ac_negg": [0.05153655260801315, 0.053153470158576965, 0.4637925326824188], "r_norm": [1.4529674053192139, 1.6545417308807373, 2.9059221744537354], "ldr": {"dL_sdil": -2.081070899963379, "dL_bp": -2.081070899963379, "ratio": 1.0}}, {"step": 4600, "epoch": 9, "train_loss": 6.669431686401367, "cos_r_negg": [0.01777573674917221, -0.09210150688886642, -0.10563072562217712], "cos_innovation_negg": [0.10005220770835876, 0.16834776103496552, 0.3849782347679138], "cos_apical_negg": [0.01777573861181736, -0.09210151433944702, -0.10563073307275772], "cos_Ac_negg": [0.1000543087720871, 0.16835026443004608, 0.3849782347679138], "r_norm": [1.3946471214294434, 1.578935146331787, 2.8077735900878906]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.1135, "test_loss": 9.015954296875}, {"step": 4700, "epoch": 10, "train_loss": 9.554258346557617, "cos_r_negg": [-0.01756708323955536, -0.04851745814085007, -0.002075500786304474], "cos_innovation_negg": [-0.08733142912387848, 0.07131969928741455, 0.6723060607910156], "cos_apical_negg": [-0.01756708323955536, -0.048517461866140366, -0.0020755035802721977], "cos_Ac_negg": [-0.08734530955553055, 0.07131726294755936, 0.6723060607910156], "r_norm": [1.3125298023223877, 1.5057177543640137, 2.6409571170806885]}, {"step": 4800, "epoch": 10, "train_loss": 5.9013471603393555, "cos_r_negg": [0.0754239559173584, -0.05168905481696129, 0.11058440804481506], "cos_innovation_negg": [0.007327036000788212, 0.2709810137748718, 0.7619322538375854], "cos_apical_negg": [0.0754239559173584, -0.05168905854225159, 0.11058438569307327], "cos_Ac_negg": [0.007336634211242199, 0.2709811329841614, 0.7619322538375854], "r_norm": [1.463352918624878, 1.5180680751800537, 2.9792442321777344]}, {"step": 4900, "epoch": 10, "train_loss": 10.935822486877441, "cos_r_negg": [0.0575990304350853, -0.33695322275161743, -0.34385931491851807], "cos_innovation_negg": [0.014982819557189941, 0.09094658493995667, 0.4001518189907074], "cos_apical_negg": [0.0575990229845047, -0.33695322275161743, -0.34385931491851807], "cos_Ac_negg": [0.015041286125779152, 0.09094700217247009, 0.40015166997909546], "r_norm": [1.4888331890106201, 1.616701602935791, 3.0466463565826416]}, {"step": 5000, "epoch": 10, "train_loss": 9.521804809570312, "cos_r_negg": [0.05206938087940216, -0.3930056393146515, -0.5807730555534363], "cos_innovation_negg": [-0.04468301683664322, 0.14414645731449127, 0.5752860903739929], "cos_apical_negg": [0.05206938832998276, -0.3930056393146515, -0.5807730555534363], "cos_Ac_negg": [-0.044713109731674194, 0.14414654672145844, 0.5752860903739929], "r_norm": [1.3562431335449219, 1.4976449012756348, 2.7944326400756836], "ldr": {"dL_sdil": -1.940413475036621, "dL_bp": -1.940413475036621, "ratio": 1.0}}, {"step": 5100, "epoch": 10, "train_loss": 7.874048233032227, "cos_r_negg": [-0.024124640971422195, 0.058301106095314026, 0.07119216024875641], "cos_innovation_negg": [-0.05157509446144104, 0.05782962590456009, 0.5174257159233093], "cos_apical_negg": [-0.024124637246131897, 0.058301106095314026, 0.0711921826004982], "cos_Ac_negg": [-0.05156651884317398, 0.057829566299915314, 0.5174257159233093], "r_norm": [1.3540897369384766, 1.5044630765914917, 2.9336659908294678]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.1009, "test_loss": 11.43567568359375}, {"step": 5200, "epoch": 11, "train_loss": 10.689697265625, "cos_r_negg": [0.019001197069883347, -0.2897270917892456, -0.2654634714126587], "cos_innovation_negg": [-0.05334196239709854, 0.08347544819116592, 0.39443832635879517], "cos_apical_negg": [0.019001193344593048, -0.2897270917892456, -0.2654634714126587], "cos_Ac_negg": [-0.0533544197678566, 0.08347558230161667, 0.39443832635879517], "r_norm": [1.2976155281066895, 1.4617466926574707, 2.853634834289551]}, {"step": 5300, "epoch": 11, "train_loss": 10.256033897399902, "cos_r_negg": [0.013968532904982567, -0.22812309861183167, -0.13791435956954956], "cos_innovation_negg": [-0.05052020028233528, 0.12567764520645142, 0.5747710466384888], "cos_apical_negg": [0.01396853569895029, -0.22812311351299286, -0.13791435956954956], "cos_Ac_negg": [-0.05052119493484497, 0.12567763030529022, 0.5747710466384888], "r_norm": [1.1869971752166748, 1.3122568130493164, 2.6022799015045166]}, {"step": 5400, "epoch": 11, "train_loss": 9.13038444519043, "cos_r_negg": [0.019583577290177345, -0.3641751706600189, -0.526611328125], "cos_innovation_negg": [-0.0005453093908727169, -0.12072335928678513, 0.5536854267120361], "cos_apical_negg": [0.019583577290177345, -0.36417514085769653, -0.5266112685203552], "cos_Ac_negg": [-0.0005325092934072018, -0.12072324752807617, 0.5536854267120361], "r_norm": [1.459100604057312, 1.5429048538208008, 3.1018805503845215]}, {"step": 5500, "epoch": 11, "train_loss": 13.559499740600586, "cos_r_negg": [0.06639349460601807, -0.5439437031745911, -0.5144651532173157], "cos_innovation_negg": [-0.06592432409524918, -0.18451941013336182, 0.861040472984314], "cos_apical_negg": [0.06639350205659866, -0.5439437627792358, -0.5144652128219604], "cos_Ac_negg": [-0.06593867391347885, -0.18451935052871704, 0.8610404133796692], "r_norm": [1.2953298091888428, 1.4854196310043335, 2.7311806678771973], "ldr": {"dL_sdil": -2.1915130615234375, "dL_bp": -2.1915130615234375, "ratio": 1.0}}, {"step": 5600, "epoch": 11, "train_loss": 11.20883560180664, "cos_r_negg": [-0.031799525022506714, -0.22811442613601685, -0.23479770123958588], "cos_innovation_negg": [-0.0690169632434845, -0.0075832391157746315, 0.4777218997478485], "cos_apical_negg": [-0.03179951757192612, -0.22811444103717804, -0.23479770123958588], "cos_Ac_negg": [-0.06903363764286041, -0.0075830332934856415, 0.4777218699455261], "r_norm": [1.340540885925293, 1.5263988971710205, 3.144881248474121]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.0974, "test_loss": 13.76685361328125}, {"step": 5700, "epoch": 12, "train_loss": 10.07309341430664, "cos_r_negg": [0.03777661919593811, -0.4412369728088379, -0.36332398653030396], "cos_innovation_negg": [0.03472915291786194, -0.17701226472854614, 0.8993306756019592], "cos_apical_negg": [0.03777661919593811, -0.4412369728088379, -0.36332398653030396], "cos_Ac_negg": [0.034731801599264145, -0.17701216042041779, 0.899330735206604], "r_norm": [1.3225698471069336, 1.5236485004425049, 2.700747013092041]}, {"step": 5800, "epoch": 12, "train_loss": 12.433907508850098, "cos_r_negg": [-0.003950262442231178, -0.547705352306366, -0.3867746889591217], "cos_innovation_negg": [-0.05912623554468155, -0.19530336558818817, 0.9709994792938232], "cos_apical_negg": [-0.003950268030166626, -0.547705352306366, -0.3867747187614441], "cos_Ac_negg": [-0.05908357352018356, -0.1953035295009613, 0.9709994792938232], "r_norm": [1.3607196807861328, 1.4850249290466309, 3.2782058715820312]}, {"step": 5900, "epoch": 12, "train_loss": 9.572389602661133, "cos_r_negg": [0.0695822685956955, -0.4782870411872864, -0.4824511706829071], "cos_innovation_negg": [-0.0036430712789297104, 0.21301300823688507, 0.6043546795845032], "cos_apical_negg": [0.0695822685956955, -0.478287011384964, -0.4824511408805847], "cos_Ac_negg": [-0.0036581167951226234, 0.21301306784152985, 0.604354739189148], "r_norm": [1.212158441543579, 1.313722848892212, 2.773188352584839]}, {"step": 6000, "epoch": 12, "train_loss": 8.03459644317627, "cos_r_negg": [0.05641988664865494, -0.30625995993614197, -0.32399484515190125], "cos_innovation_negg": [0.07821530103683472, -0.18439939618110657, 0.7435494661331177], "cos_apical_negg": [0.05641988664865494, -0.30625995993614197, -0.32399487495422363], "cos_Ac_negg": [0.07819831371307373, -0.18439938127994537, 0.7435494661331177], "r_norm": [1.2636423110961914, 1.4633649587631226, 2.6897456645965576], "ldr": {"dL_sdil": -1.2584352493286133, "dL_bp": -1.2584352493286133, "ratio": 1.0}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.0974, "test_loss": 6.334162353515625}, {"step": 6100, "epoch": 13, "train_loss": 5.32422399520874, "cos_r_negg": [-0.020511440932750702, -0.20042458176612854, -0.1201372966170311], "cos_innovation_negg": [-0.05921708047389984, 0.0025104722008109093, 0.7776116132736206], "cos_apical_negg": [-0.020511429756879807, -0.20042458176612854, -0.1201372966170311], "cos_Ac_negg": [-0.05921370908617973, 0.0025104996748268604, 0.7776116132736206], "r_norm": [1.1844608783721924, 1.339348554611206, 2.631443977355957]}, {"step": 6200, "epoch": 13, "train_loss": 15.338651657104492, "cos_r_negg": [0.006799392402172089, -0.4675729274749756, -0.5238707065582275], "cos_innovation_negg": [0.016532087698578835, 0.22660422325134277, 0.5020425319671631], "cos_apical_negg": [0.006799392402172089, -0.4675729274749756, -0.5238707065582275], "cos_Ac_negg": [0.016515430063009262, 0.22660425305366516, 0.5020426511764526], "r_norm": [1.2123239040374756, 1.321899175643921, 2.875114917755127]}, {"step": 6300, "epoch": 13, "train_loss": 10.865857124328613, "cos_r_negg": [-0.04579303413629532, -0.3763803243637085, -0.2255067229270935], "cos_innovation_negg": [-0.031234746798872948, -0.18446262180805206, 0.718535304069519], "cos_apical_negg": [-0.04579302668571472, -0.3763803243637085, -0.2255067080259323], "cos_Ac_negg": [-0.0312519297003746, -0.18446236848831177, 0.718535304069519], "r_norm": [1.1536757946014404, 1.2023568153381348, 2.787022829055786]}, {"step": 6400, "epoch": 13, "train_loss": 10.130316734313965, "cos_r_negg": [0.014571597799658775, -0.3073262870311737, -0.2725748121738434], "cos_innovation_negg": [-0.0355323888361454, -0.0392596572637558, 0.7970172166824341], "cos_apical_negg": [0.014571597799658775, -0.3073262870311737, -0.2725748121738434], "cos_Ac_negg": [-0.03557465225458145, -0.03925967589020729, 0.7970172166824341], "r_norm": [1.270211100578308, 1.2800111770629883, 3.166908025741577]}, {"step": 6500, "epoch": 13, "train_loss": 10.813982009887695, "cos_r_negg": [0.011883115395903587, -0.5175911784172058, -0.3062574565410614], "cos_innovation_negg": [-0.005955313332378864, -0.013851936906576157, 0.91746985912323], "cos_apical_negg": [0.011883115395903587, -0.5175911784172058, -0.306257426738739], "cos_Ac_negg": [-0.005974724888801575, -0.013851940631866455, 0.91746985912323], "r_norm": [1.1609448194503784, 1.2118589878082275, 2.9055025577545166], "ldr": {"dL_sdil": -1.5699901580810547, "dL_bp": -1.5699901580810547, "ratio": 1.0}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.1032, "test_loss": 11.97496240234375}, {"step": 6600, "epoch": 14, "train_loss": 10.166065216064453, "cos_r_negg": [0.0740952342748642, -0.42438608407974243, -0.5446914434432983], "cos_innovation_negg": [0.007209917530417442, -0.049284063279628754, 0.5896706581115723], "cos_apical_negg": [0.0740952342748642, -0.42438608407974243, -0.5446915030479431], "cos_Ac_negg": [0.007204357534646988, -0.04928404465317726, 0.589670717716217], "r_norm": [1.1379551887512207, 1.2085778713226318, 2.7993202209472656]}, {"step": 6700, "epoch": 14, "train_loss": 12.796271324157715, "cos_r_negg": [0.04596796631813049, -0.4727451801300049, -0.2784838080406189], "cos_innovation_negg": [-0.15685194730758667, -0.2215430736541748, 0.6244687438011169], "cos_apical_negg": [0.04596797004342079, -0.4727451801300049, -0.2784837782382965], "cos_Ac_negg": [-0.15674909949302673, -0.22154322266578674, 0.6244687438011169], "r_norm": [1.2866442203521729, 1.2906875610351562, 3.3445470333099365]}, {"step": 6800, "epoch": 14, "train_loss": 10.457527160644531, "cos_r_negg": [0.0034635011106729507, -0.33966535329818726, -0.3185226321220398], "cos_innovation_negg": [-0.015793822705745697, 0.06659657508134842, 0.4613923132419586], "cos_apical_negg": [0.003463505767285824, -0.33966535329818726, -0.3185226619243622], "cos_Ac_negg": [-0.01584102213382721, 0.06659658253192902, 0.4613923132419586], "r_norm": [1.164564847946167, 1.3001453876495361, 3.0376219749450684]}, {"step": 6900, "epoch": 14, "train_loss": 5.674578666687012, "cos_r_negg": [0.06117851287126541, -0.4614144265651703, -0.3935377895832062], "cos_innovation_negg": [0.08230698853731155, 0.23889562487602234, 0.44478482007980347], "cos_apical_negg": [0.06117851659655571, -0.4614144265651703, -0.3935377597808838], "cos_Ac_negg": [0.08229748904705048, 0.23889559507369995, 0.4447847902774811], "r_norm": [1.207074522972107, 1.301612138748169, 3.0882766246795654]}, {"step": 7000, "epoch": 14, "train_loss": 10.910627365112305, "cos_r_negg": [0.03731546550989151, -0.37948036193847656, -0.3181638717651367], "cos_innovation_negg": [-0.02078831009566784, 0.196212038397789, 0.4585264325141907], "cos_apical_negg": [0.03731546178460121, -0.37948036193847656, -0.3181639015674591], "cos_Ac_negg": [-0.02077524922788143, 0.1962120234966278, 0.45852649211883545], "r_norm": [1.1883717775344849, 1.262464165687561, 3.0732603073120117], "ldr": {"dL_sdil": -1.7473487854003906, "dL_bp": -1.7473487854003906, "ratio": 1.0}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.1028, "test_loss": 17.5008279296875}], "final": {"test_acc": 0.1028, "test_loss": 17.5008279296875, "wall_s": 27.01093864440918, "cos_r_negg": [0.11227206885814667, -0.525943398475647, -0.6144851446151733], "cos_innovation_negg": [-0.17259691655635834, -0.044658537954092026, 0.6417496800422668], "cos_apical_negg": [0.11227206885814667, -0.5259433388710022, -0.6144851446151733], "cos_Ac_negg": [-0.1726006418466568, -0.04465854912996292, 0.6417496800422668]}}
|