summaryrefslogtreecommitdiff
path: root/results/nuis_ctrl_v1_rho0p5_raw_s2.json
blob: d41638348df9417080c151e6064478abb2a1a1da (plain)
1
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 2, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p5_raw_s2", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4534778594970703, "cos_r_negg": [0.006069933995604515, 0.028213273733854294, 0.000850403041113168], "cos_innovation_negg": [0.006572579964995384, 0.03945208340883255, 0.0016749617643654346], "cos_apical_negg": [0.006069933995604515, 0.028213273733854294, 0.000850403041113168], "cos_Ac_negg": [0.006574587430804968, 0.03945256024599075, 0.0016756001859903336], "r_norm": [8.09968090057373, 6.984550476074219, 6.564377784729004], "ldr": {"dL_sdil": -0.040816545486450195, "dL_bp": -0.1888291835784912, "ratio": 0.2161559178138577}}, {"step": 100, "epoch": 0, "train_loss": 8.607678413391113, "cos_r_negg": [0.043286263942718506, 0.02338339574635029, 0.01075268816202879], "cos_innovation_negg": [0.07204825431108475, 0.2613758444786072, -0.5655913949012756], "cos_apical_negg": [0.043286263942718506, 0.02338339574635029, 0.01075268816202879], "cos_Ac_negg": [0.07190786302089691, 0.2612081468105316, -0.5655913949012756], "r_norm": [10.865097999572754, 10.794242858886719, 10.00767707824707]}, {"step": 200, "epoch": 0, "train_loss": 14.23398208618164, "cos_r_negg": [-0.05621775984764099, -0.06742869317531586, -0.2560291588306427], "cos_innovation_negg": [-0.037481822073459625, 0.14581120014190674, -0.5063558220863342], "cos_apical_negg": [-0.05621775984764099, -0.06742869317531586, -0.2560291588306427], "cos_Ac_negg": [-0.03706343472003937, 0.1459081619977951, -0.5063564777374268], "r_norm": [10.567609786987305, 9.59535026550293, 9.80006217956543]}, {"step": 300, "epoch": 0, "train_loss": 7.152601718902588, "cos_r_negg": [-0.004294459708034992, -0.036559201776981354, 0.39814814925193787], "cos_innovation_negg": [-0.025412902235984802, 0.18632517755031586, 0.375], "cos_apical_negg": [-0.004294459708034992, -0.036559201776981354, 0.39814814925193787], "cos_Ac_negg": [-0.025513848289847374, 0.18630953133106232, 0.375], "r_norm": [9.94304084777832, 9.509531021118164, 9.850784301757812]}, {"step": 400, "epoch": 0, "train_loss": 11.771880149841309, "cos_r_negg": [0.02575899101793766, 0.20267914235591888, -0.46919432282447815], "cos_innovation_negg": [-0.011691069230437279, 0.26164400577545166, -0.2132701575756073], "cos_apical_negg": [0.02575899101793766, 0.20267914235591888, -0.46919432282447815], "cos_Ac_negg": [-0.011695858091115952, 0.26161447167396545, -0.2132701575756073], "r_norm": [9.711320877075195, 9.536703109741211, 9.406965255737305]}, {"epoch_end": 0, "step": 469, "test_acc": 0.0892, "test_loss": 5.7297685546875}, {"step": 500, "epoch": 1, "train_loss": 7.890752792358398, "cos_r_negg": [-0.00917352270334959, 0.009967482648789883, -0.4682352840900421], "cos_innovation_negg": [-0.04460432752966881, 0.25391969084739685, -0.030588233843445778], "cos_apical_negg": [-0.00917352270334959, 0.009967482648789883, -0.4682352840900421], "cos_Ac_negg": [-0.04462495446205139, 0.2539190351963043, -0.030588233843445778], "r_norm": [9.352081298828125, 9.198244094848633, 9.062174797058105], "ldr": {"dL_sdil": -0.7840461730957031, "dL_bp": -0.7840461730957031, "ratio": 1.0}}, {"step": 600, "epoch": 1, "train_loss": 8.19209098815918, "cos_r_negg": [-0.025425070896744728, 0.09573765099048615, 0.6352822184562683], "cos_innovation_negg": [-0.02527477592229843, 0.260063111782074, -0.9988667964935303], "cos_apical_negg": [-0.025425070896744728, 0.09573765099048615, 0.6352822184562683], "cos_Ac_negg": [-0.02532678283751011, 0.2600637972354889, -0.9988669157028198], "r_norm": [9.43122673034668, 9.104082107543945, 9.184896469116211]}, {"step": 700, "epoch": 1, "train_loss": 10.612958908081055, "cos_r_negg": [0.011208701878786087, -0.020073596388101578, 0.18309859931468964], "cos_innovation_negg": [0.05172453820705414, 0.2757587432861328, -0.07511737197637558], "cos_apical_negg": [0.011208701878786087, -0.020073596388101578, 0.18309859931468964], "cos_Ac_negg": [0.05172889307141304, 0.2757216989994049, -0.07511737197637558], "r_norm": [9.012006759643555, 8.955952644348145, 9.195959091186523]}, {"step": 800, "epoch": 1, "train_loss": 7.530488967895508, "cos_r_negg": [0.01302462350577116, -0.06212879344820976, -0.4751131534576416], "cos_innovation_negg": [0.02423812448978424, 0.19756846129894257, 0.0], "cos_apical_negg": [0.01302462350577116, -0.06212879344820976, -0.4751131534576416], "cos_Ac_negg": [0.024241534993052483, 0.19757811725139618, 0.0], "r_norm": [8.682754516601562, 8.557674407958984, 8.5389986038208]}, {"step": 900, "epoch": 1, "train_loss": 9.516929626464844, "cos_r_negg": [0.019926870241761208, -0.11089521646499634, -0.5698447823524475], "cos_innovation_negg": [0.04081986844539642, 0.33930930495262146, -0.6585365533828735], "cos_apical_negg": [0.019926870241761208, -0.11089521646499634, -0.5698447823524475], "cos_Ac_negg": [0.04129152372479439, 0.33924928307533264, -0.6585365533828735], "r_norm": [8.787179946899414, 8.46986198425293, 8.557413101196289]}, {"epoch_end": 1, "step": 938, "test_acc": 0.0958, "test_loss": 14.253339453125}, {"step": 1000, "epoch": 2, "train_loss": 11.903938293457031, "cos_r_negg": [-0.01657753251492977, -0.01202379260212183, 0.2232142984867096], "cos_innovation_negg": [-0.16749955713748932, 0.013943457044661045, 0.2008928656578064], "cos_apical_negg": [-0.01657753251492977, -0.01202379260212183, 0.2232142984867096], "cos_Ac_negg": [-0.1675061583518982, 0.01390395499765873, 0.2008928656578064], "r_norm": [8.835855484008789, 9.250875473022461, 8.899043083190918], "ldr": {"dL_sdil": -1.3018312454223633, "dL_bp": -1.3018312454223633, "ratio": 1.0}}, {"step": 1100, "epoch": 2, "train_loss": 7.095750331878662, "cos_r_negg": [0.028792718425393105, 0.016014961525797844, -0.20770877599716187], "cos_innovation_negg": [0.003021157579496503, 0.043102797120809555, -0.010706637986004353], "cos_apical_negg": [0.028792718425393105, 0.016014961525797844, -0.20770877599716187], "cos_Ac_negg": [0.0030201401095837355, 0.04310184344649315, -0.010706637986004353], "r_norm": [8.63536262512207, 8.829824447631836, 8.958860397338867]}, {"step": 1200, "epoch": 2, "train_loss": 17.46408462524414, "cos_r_negg": [-0.023260295391082764, 0.1216074749827385, -0.4345114231109619], "cos_innovation_negg": [-0.018750956282019615, -0.11288260668516159, 0.82536381483078], "cos_apical_negg": [-0.023260295391082764, 0.1216074749827385, -0.4345114231109619], "cos_Ac_negg": [-0.01872233673930168, -0.11265920847654343, 0.82536381483078], "r_norm": [8.729228973388672, 8.940746307373047, 9.051870346069336]}, {"step": 1300, "epoch": 2, "train_loss": 13.53596305847168, "cos_r_negg": [0.025372661650180817, -0.006017977371811867, -0.42578125], "cos_innovation_negg": [-0.07081383466720581, -0.15056994557380676, -0.2421875], "cos_apical_negg": [0.025372661650180817, -0.006017977371811867, -0.42578125], "cos_Ac_negg": [-0.07078513503074646, -0.15052571892738342, -0.2421875], "r_norm": [8.374276161193848, 8.17935562133789, 8.201964378356934]}, {"step": 1400, "epoch": 2, "train_loss": 10.240300178527832, "cos_r_negg": [0.0203335490077734, 0.27930134534835815, 0.5859375], "cos_innovation_negg": [0.11001616716384888, 0.19296517968177795, 0.36328125], "cos_apical_negg": [0.0203335490077734, 0.27930134534835815, 0.5859375], "cos_Ac_negg": [0.10946781188249588, 0.19257411360740662, 0.36328125], "r_norm": [9.051124572753906, 9.292688369750977, 9.421853065490723]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.1028, "test_loss": 12.33814736328125}, {"step": 1500, "epoch": 3, "train_loss": 16.444610595703125, "cos_r_negg": [-0.0047040837816894054, -0.2855634391307831, -0.265625], "cos_innovation_negg": [0.04879166930913925, 0.22736075520515442, 0.34765625], "cos_apical_negg": [-0.0047040837816894054, -0.2855634391307831, -0.265625], "cos_Ac_negg": [0.04879727214574814, 0.22735252976417542, 0.34765625], "r_norm": [8.130634307861328, 8.34067153930664, 8.304515838623047], "ldr": {"dL_sdil": -1.3420801162719727, "dL_bp": -1.3420801162719727, "ratio": 1.0}}, {"step": 1600, "epoch": 3, "train_loss": 9.035846710205078, "cos_r_negg": [-0.003250044770538807, -0.20443597435951233, -0.21875], "cos_innovation_negg": [-0.016233408823609352, 0.14438506960868835, -0.01171875], "cos_apical_negg": [-0.003250044770538807, -0.20443597435951233, -0.21875], "cos_Ac_negg": [-0.016230184584856033, 0.14437103271484375, -0.01171875], "r_norm": [8.45271110534668, 8.598628997802734, 8.630088806152344]}, {"step": 1700, "epoch": 3, "train_loss": 13.682973861694336, "cos_r_negg": [0.008910275995731354, 0.5314902067184448, 0.5859375], "cos_innovation_negg": [-0.04581741616129875, 0.2546497583389282, 0.58203125], "cos_apical_negg": [0.008910275995731354, 0.5314902067184448, 0.5859375], "cos_Ac_negg": [-0.04582258313894272, 0.2546376585960388, 0.58203125], "r_norm": [8.633129119873047, 8.776876449584961, 8.845964431762695]}, {"step": 1800, "epoch": 3, "train_loss": 18.205738067626953, "cos_r_negg": [0.04461752995848656, -0.47234925627708435, -0.484375], "cos_innovation_negg": [0.0671183317899704, -0.16538923978805542, 0.78125], "cos_apical_negg": [0.04461752995848656, -0.47234925627708435, -0.484375], "cos_Ac_negg": [0.06705057621002197, -0.1654069423675537, 0.78125], "r_norm": [8.202865600585938, 8.286920547485352, 8.286673545837402]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.1032, "test_loss": 11.171415625}, {"step": 1900, "epoch": 4, "train_loss": 11.181070327758789, "cos_r_negg": [-0.06173565611243248, 0.09183380007743835, 0.11328125], "cos_innovation_negg": [0.0642334595322609, -0.040925033390522, 0.51171875], "cos_apical_negg": [-0.06173565611243248, 0.09183380007743835, 0.11328125], "cos_Ac_negg": [0.06420228630304337, -0.04095173254609108, 0.51171875], "r_norm": [8.587827682495117, 8.683780670166016, 8.845311164855957]}, {"step": 2000, "epoch": 4, "train_loss": 14.013265609741211, "cos_r_negg": [-0.03492019325494766, 0.10401894152164459, 0.11328125], "cos_innovation_negg": [-0.03786076605319977, -0.02154957689344883, 0.78515625], "cos_apical_negg": [-0.03492019325494766, 0.10401894152164459, 0.11328125], "cos_Ac_negg": [-0.03788123279809952, -0.021539095789194107, 0.78515625], "r_norm": [8.202961921691895, 8.270047187805176, 8.323112487792969], "ldr": {"dL_sdil": -1.8188533782958984, "dL_bp": -1.8188533782958984, "ratio": 1.0}}, {"step": 2100, "epoch": 4, "train_loss": 6.470243453979492, "cos_r_negg": [-0.0008236072608269751, -0.05182565748691559, -0.05078125], "cos_innovation_negg": [0.0933600664138794, -0.1702737808227539, 0.15234375], "cos_apical_negg": [-0.0008236072608269751, -0.05182565748691559, -0.05078125], "cos_Ac_negg": [0.09327870607376099, -0.17038270831108093, 0.15234375], "r_norm": [8.264612197875977, 8.322214126586914, 8.388392448425293]}, {"step": 2200, "epoch": 4, "train_loss": 10.790571212768555, "cos_r_negg": [-0.06804899871349335, -0.0514930784702301, -0.05078125], "cos_innovation_negg": [-0.14114263653755188, -0.04461637884378433, 0.56640625], "cos_apical_negg": [-0.06804899871349335, -0.0514930784702301, -0.05078125], "cos_Ac_negg": [-0.14115355908870697, -0.044624097645282745, 0.56640625], "r_norm": [8.39930534362793, 8.48637580871582, 8.370820999145508]}, {"step": 2300, "epoch": 4, "train_loss": 11.865091323852539, "cos_r_negg": [-0.07438310235738754, -0.05125802755355835, -0.05078125], "cos_innovation_negg": [-0.15275540947914124, 0.1405005156993866, 0.8359375], "cos_apical_negg": [-0.07438310235738754, -0.05125802755355835, -0.05078125], "cos_Ac_negg": [-0.15322351455688477, 0.13474678993225098, 0.8359375], "r_norm": [8.323041915893555, 8.41018295288086, 8.291877746582031]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.0958, "test_loss": 9.8487564453125}, {"step": 2400, "epoch": 5, "train_loss": 9.967241287231445, "cos_r_negg": [0.04543090984225273, -0.10339514911174774, -0.1015625], "cos_innovation_negg": [-0.04555587098002434, -0.23899617791175842, 0.78515625], "cos_apical_negg": [0.04543090984225273, -0.10339514911174774, -0.1015625], "cos_Ac_negg": [-0.04559876769781113, -0.2389882504940033, 0.78515625], "r_norm": [8.396261215209961, 8.439985275268555, 8.51211929321289]}, {"step": 2500, "epoch": 5, "train_loss": 14.84898567199707, "cos_r_negg": [-0.04485081508755684, -0.2646341025829315, -0.26953125], "cos_innovation_negg": [-0.25339028239250183, 0.06275664269924164, 0.12890625], "cos_apical_negg": [-0.04485081508755684, -0.2646341025829315, -0.26953125], "cos_Ac_negg": [-0.2533881366252899, 0.0627472847700119, 0.12890625], "r_norm": [8.10526180267334, 8.128353118896484, 8.154784202575684], "ldr": {"dL_sdil": -2.2104549407958984, "dL_bp": -2.2104549407958984, "ratio": 1.0}}, {"step": 2600, "epoch": 5, "train_loss": 17.929655075073242, "cos_r_negg": [0.10165649652481079, -0.7821450233459473, -0.8125], "cos_innovation_negg": [-0.23728157579898834, -0.541267991065979, 0.23828125], "cos_apical_negg": [0.10165649652481079, -0.7821450233459473, -0.8125], "cos_Ac_negg": [-0.2373037338256836, -0.5412731170654297, 0.23828125], "r_norm": [8.668704986572266, 8.701017379760742, 8.750303268432617]}, {"step": 2700, "epoch": 5, "train_loss": 9.901445388793945, "cos_r_negg": [0.009795840829610825, 0.5564864873886108, 0.5859375], "cos_innovation_negg": [-0.030734334141016006, 0.02718561515212059, 0.796875], "cos_apical_negg": [0.009795840829610825, 0.5564864873886108, 0.5859375], "cos_Ac_negg": [-0.030703900381922722, 0.02718810923397541, 0.796875], "r_norm": [8.112945556640625, 8.154273986816406, 8.163640975952148]}, {"step": 2800, "epoch": 5, "train_loss": 15.594620704650879, "cos_r_negg": [0.12130018323659897, -0.18450616300106049, -0.20703125], "cos_innovation_negg": [0.09369021654129028, 0.09002826362848282, 0.19140625], "cos_apical_negg": [0.12130018323659897, -0.18450616300106049, -0.20703125], "cos_Ac_negg": [0.09368741512298584, 0.0744839608669281, 0.19140625], "r_norm": [8.264737129211426, 8.303056716918945, 8.341985702514648]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.1028, "test_loss": 20.358579296875}, {"step": 2900, "epoch": 6, "train_loss": 16.89208984375, "cos_r_negg": [-0.014575080946087837, -0.2592034637928009, -0.265625], "cos_innovation_negg": [0.0833456739783287, -0.1388159841299057, 0.8125], "cos_apical_negg": [-0.014575080946087837, -0.2592034637928009, -0.265625], "cos_Ac_negg": [0.0833434984087944, -0.13881127536296844, 0.8125], "r_norm": [8.071036338806152, 8.065722465515137, 8.09593677520752]}, {"step": 3000, "epoch": 6, "train_loss": 11.365478515625, "cos_r_negg": [0.047974660992622375, -0.6225115060806274, -0.640625], "cos_innovation_negg": [0.019792458042502403, -0.43774449825286865, 0.4375], "cos_apical_negg": [0.047974660992622375, -0.6225115060806274, -0.640625], "cos_Ac_negg": [0.01979689858853817, -0.43774959444999695, 0.4375], "r_norm": [8.169724464416504, 8.164944648742676, 8.221406936645508], "ldr": {"dL_sdil": -2.4173011779785156, "dL_bp": -2.4173011779785156, "ratio": 1.0}}, {"step": 3100, "epoch": 6, "train_loss": 14.204668045043945, "cos_r_negg": [0.02127210609614849, -0.26237258315086365, -0.26953125], "cos_innovation_negg": [-0.03792629763484001, -0.3400703966617584, 0.78125], "cos_apical_negg": [0.02127210609614849, -0.26237258315086365, -0.26953125], "cos_Ac_negg": [-0.03792750835418701, -0.34007057547569275, 0.78125], "r_norm": [8.430807113647461, 8.42910099029541, 8.496881484985352]}, {"step": 3200, "epoch": 6, "train_loss": 11.091636657714844, "cos_r_negg": [0.015585962682962418, -0.8003131747245789, -0.828125], "cos_innovation_negg": [0.12115953117609024, -0.7471803426742554, 0.78125], "cos_apical_negg": [0.015585962682962418, -0.8003131747245789, -0.828125], "cos_Ac_negg": [0.12116257101297379, -0.7471810579299927, 0.78125], "r_norm": [8.819147109985352, 8.829202651977539, 8.839988708496094]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.0974, "test_loss": 8.17626220703125}, {"step": 3300, "epoch": 7, "train_loss": 11.65943431854248, "cos_r_negg": [-0.07276209443807602, -0.5372828245162964, -0.56640625], "cos_innovation_negg": [0.06662250310182571, 0.558137834072113, 0.828125], "cos_apical_negg": [-0.07276209443807602, -0.5372828245162964, -0.56640625], "cos_Ac_negg": [0.06661158800125122, 0.5581486225128174, 0.828125], "r_norm": [7.814938545227051, 7.757964611053467, 7.753844261169434]}, {"step": 3400, "epoch": 7, "train_loss": 9.130935668945312, "cos_r_negg": [-0.06863047927618027, -0.8006556630134583, -0.828125], "cos_innovation_negg": [0.02796264924108982, -0.7023664712905884, 1.0], "cos_apical_negg": [-0.06863047927618027, -0.8006556630134583, -0.828125], "cos_Ac_negg": [0.027953851968050003, -0.702366828918457, 1.0], "r_norm": [8.880365371704102, 8.896379470825195, 8.904964447021484]}, {"step": 3500, "epoch": 7, "train_loss": 9.979610443115234, "cos_r_negg": [5.192796015762724e-05, -0.7991563081741333, -0.828125], "cos_innovation_negg": [0.03197434917092323, -0.6567003726959229, 0.78125], "cos_apical_negg": [5.192796015762724e-05, -0.7991563081741333, -0.828125], "cos_Ac_negg": [0.031970880925655365, -0.6567010283470154, 0.78125], "r_norm": [8.684354782104492, 8.69792366027832, 8.674968719482422], "ldr": {"dL_sdil": -1.5388154983520508, "dL_bp": -1.5388145446777344, "ratio": 1.0000006197461024}}, {"step": 3600, "epoch": 7, "train_loss": 12.586816787719727, "cos_r_negg": [-0.012873931787908077, 0.2093500941991806, 0.21875], "cos_innovation_negg": [-0.11395280808210373, -0.3456953167915344, 0.8359375], "cos_apical_negg": [-0.012873931787908077, 0.2093500941991806, 0.21875], "cos_Ac_negg": [-0.11394085735082626, -0.34569495916366577, 0.8359375], "r_norm": [7.8183088302612305, 7.841979026794434, 7.790170669555664]}, {"step": 3700, "epoch": 7, "train_loss": 10.789178848266602, "cos_r_negg": [0.013845272362232208, 0.3656242787837982, 0.3828125], "cos_innovation_negg": [-0.21636852622032166, -0.49038785696029663, 0.796875], "cos_apical_negg": [0.013845272362232208, 0.3656242787837982, 0.3828125], "cos_Ac_negg": [-0.2163427770137787, -0.4903874397277832, 0.796875], "r_norm": [8.18099594116211, 8.203760147094727, 8.175332069396973]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.1028, "test_loss": 8.26958056640625}, {"step": 3800, "epoch": 8, "train_loss": 10.51145076751709, "cos_r_negg": [-0.05470547825098038, -0.3788420259952545, -0.39453125], "cos_innovation_negg": [-0.0007461022469215095, -0.27109163999557495, 0.4921875], "cos_apical_negg": [-0.05470547825098038, -0.3788420259952545, -0.39453125], "cos_Ac_negg": [-0.0008219259907491505, -0.27109235525131226, 0.4921875], "r_norm": [8.332502365112305, 8.355758666992188, 8.389439582824707]}, {"step": 3900, "epoch": 8, "train_loss": 8.812009811401367, "cos_r_negg": [-0.06844376027584076, -0.5706951022148132, -0.59765625], "cos_innovation_negg": [0.08148781210184097, -0.44096678495407104, 0.453125], "cos_apical_negg": [-0.06844376027584076, -0.5706951022148132, -0.59765625], "cos_Ac_negg": [0.0814904049038887, -0.4409674406051636, 0.453125], "r_norm": [8.454343795776367, 8.432502746582031, 8.447444915771484]}, {"step": 4000, "epoch": 8, "train_loss": 11.610506057739258, "cos_r_negg": [-0.0724836140871048, -0.7820213437080383, -0.8125], "cos_innovation_negg": [0.16154485940933228, -0.3440636694431305, 0.609375], "cos_apical_negg": [-0.0724836140871048, -0.7820213437080383, -0.8125], "cos_Ac_negg": [0.1615554839372635, -0.34406548738479614, 0.609375], "r_norm": [8.274002075195312, 8.270342826843262, 8.313014030456543], "ldr": {"dL_sdil": -4.135002136230469, "dL_bp": -4.135002136230469, "ratio": 1.0}}, {"step": 4100, "epoch": 8, "train_loss": 8.431884765625, "cos_r_negg": [0.02188153564929962, -0.09907101094722748, -0.1015625], "cos_innovation_negg": [0.11786270886659622, 0.002072244882583618, 0.78515625], "cos_apical_negg": [0.02188153564929962, -0.09907101094722748, -0.1015625], "cos_Ac_negg": [0.11759955435991287, 0.0020715761929750443, 0.78515625], "r_norm": [8.24980354309082, 8.273475646972656, 8.269441604614258]}, {"step": 4200, "epoch": 8, "train_loss": 9.99185848236084, "cos_r_negg": [-0.06288471817970276, -0.5857329368591309, -0.61328125], "cos_innovation_negg": [-0.002537989290431142, -0.5164992809295654, 0.4375], "cos_apical_negg": [-0.06288471817970276, -0.5857329368591309, -0.61328125], "cos_Ac_negg": [-0.002540719462558627, -0.5164992213249207, 0.4375], "r_norm": [8.435721397399902, 8.411832809448242, 8.466842651367188]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.0958, "test_loss": 15.21412236328125}, {"step": 4300, "epoch": 9, "train_loss": 8.622159004211426, "cos_r_negg": [-0.19204388558864594, -0.7476558685302734, -0.78125], "cos_innovation_negg": [0.2395847886800766, -0.043641019612550735, 0.828125], "cos_apical_negg": [-0.19204388558864594, -0.7476558685302734, -0.78125], "cos_Ac_negg": [0.23957504332065582, -0.04364091902971268, 0.828125], "r_norm": [8.409965515136719, 8.437745094299316, 8.472014427185059]}, {"step": 4400, "epoch": 9, "train_loss": 10.624998092651367, "cos_r_negg": [0.012787628918886185, -0.5864592790603638, -0.61328125], "cos_innovation_negg": [-0.002128981752321124, -0.16572080552577972, 0.2734375], "cos_apical_negg": [0.012787628918886185, -0.5864592790603638, -0.61328125], "cos_Ac_negg": [-0.0022580455988645554, -0.16572090983390808, 0.2734375], "r_norm": [8.358915328979492, 8.35004997253418, 8.399492263793945]}, {"step": 4500, "epoch": 9, "train_loss": 9.67203140258789, "cos_r_negg": [0.0016504075611010194, 0.1448075771331787, 0.15234375], "cos_innovation_negg": [0.2078697830438614, -0.5442118644714355, 0.8359375], "cos_apical_negg": [0.0016504075611010194, 0.1448075771331787, 0.15234375], "cos_Ac_negg": [0.20786729454994202, -0.5442118048667908, 0.8359375], "r_norm": [7.790171146392822, 7.81021785736084, 7.82426643371582], "ldr": {"dL_sdil": -2.327272415161133, "dL_bp": -2.327272415161133, "ratio": 1.0}}, {"step": 4600, "epoch": 9, "train_loss": 9.50709056854248, "cos_r_negg": [0.03133918344974518, -0.04891374707221985, -0.05078125], "cos_innovation_negg": [0.10768858343362808, -0.4323785901069641, 0.8359375], "cos_apical_negg": [0.03133918344974518, -0.04891374707221985, -0.05078125], "cos_Ac_negg": [0.10765126347541809, -0.43237853050231934, 0.8359375], "r_norm": [8.6129150390625, 8.625988006591797, 8.636297225952148]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.1032, "test_loss": 10.47324609375}, {"step": 4700, "epoch": 10, "train_loss": 17.704463958740234, "cos_r_negg": [-0.2113754153251648, -0.7924402952194214, -0.828125], "cos_innovation_negg": [-0.0420466884970665, -0.5955016613006592, 1.0], "cos_apical_negg": [-0.2113754153251648, -0.7924402952194214, -0.828125], "cos_Ac_negg": [-0.04216146096587181, -0.595502495765686, 1.0], "r_norm": [8.829102516174316, 8.858196258544922, 8.931103706359863]}, {"step": 4800, "epoch": 10, "train_loss": 13.77151107788086, "cos_r_negg": [-0.10815715789794922, -0.7763395309448242, -0.8125], "cos_innovation_negg": [-0.194504052400589, -0.6789554953575134, 0.828125], "cos_apical_negg": [-0.10815715789794922, -0.7763395309448242, -0.8125], "cos_Ac_negg": [-0.1944883018732071, -0.6789555549621582, 0.828125], "r_norm": [8.903640747070312, 8.928548812866211, 9.014246940612793]}, {"step": 4900, "epoch": 10, "train_loss": 12.760170936584473, "cos_r_negg": [0.093043252825737, -0.46575501561164856, -0.484375], "cos_innovation_negg": [-0.1991022378206253, -0.6289278268814087, 0.56640625], "cos_apical_negg": [0.093043252825737, -0.46575501561164856, -0.484375], "cos_Ac_negg": [-0.19910438358783722, -0.6289278864860535, 0.56640625], "r_norm": [8.481609344482422, 8.491607666015625, 8.50238037109375]}, {"step": 5000, "epoch": 10, "train_loss": 14.601521492004395, "cos_r_negg": [0.015325184911489487, -0.04902271926403046, -0.05078125], "cos_innovation_negg": [-0.008242804557085037, -0.9392389059066772, 0.6328125], "cos_apical_negg": [0.015325184911489487, -0.04902271926403046, -0.05078125], "cos_Ac_negg": [-0.008244137279689312, -0.9392389059066772, 0.6328125], "r_norm": [8.327488899230957, 8.346911430358887, 8.343986511230469], "ldr": {"dL_sdil": -1.96917724609375, "dL_bp": -1.96917724609375, "ratio": 1.0}}, {"step": 5100, "epoch": 10, "train_loss": 18.300006866455078, "cos_r_negg": [-0.09675931930541992, -0.06755118817090988, -0.0703125], "cos_innovation_negg": [0.2105976790189743, 0.13013581931591034, 0.1484375], "cos_apical_negg": [-0.09675931930541992, -0.06755118817090988, -0.0703125], "cos_Ac_negg": [0.21060597896575928, 0.1301369071006775, 0.1484375], "r_norm": [7.851202487945557, 7.871732234954834, 7.909472465515137]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.1032, "test_loss": 8.86439248046875}, {"step": 5200, "epoch": 11, "train_loss": 17.184865951538086, "cos_r_negg": [-0.03285696730017662, -0.6279933452606201, -0.65625], "cos_innovation_negg": [-0.10381213575601578, -0.7768880724906921, 0.23046875], "cos_apical_negg": [-0.03285696730017662, -0.6279933452606201, -0.65625], "cos_Ac_negg": [-0.10381168127059937, -0.7768881320953369, 0.23046875], "r_norm": [9.077016830444336, 9.087455749511719, 9.146000862121582]}, {"step": 5300, "epoch": 11, "train_loss": 15.585555076599121, "cos_r_negg": [-0.054122310131788254, -0.25450626015663147, -0.265625], "cos_innovation_negg": [-0.21315738558769226, -0.4217783212661743, 0.46484375], "cos_apical_negg": [-0.054122310131788254, -0.25450626015663147, -0.265625], "cos_Ac_negg": [-0.21315713226795197, -0.42177826166152954, 0.46484375], "r_norm": [9.133200645446777, 9.143876075744629, 9.220969200134277]}, {"step": 5400, "epoch": 11, "train_loss": 11.582035064697266, "cos_r_negg": [-0.02124510146677494, -0.258146733045578, -0.26953125], "cos_innovation_negg": [-0.2423168271780014, -0.5489672422409058, 0.625], "cos_apical_negg": [-0.02124510146677494, -0.258146733045578, -0.26953125], "cos_Ac_negg": [-0.24231775104999542, -0.5489672422409058, 0.625], "r_norm": [8.897274017333984, 8.908099174499512, 8.955784797668457]}, {"step": 5500, "epoch": 11, "train_loss": 10.815536499023438, "cos_r_negg": [0.19488923251628876, 0.36498355865478516, 0.3828125], "cos_innovation_negg": [-0.11542116105556488, -0.5704564452171326, 0.796875], "cos_apical_negg": [0.19488923251628876, 0.36498355865478516, 0.3828125], "cos_Ac_negg": [-0.11541452258825302, -0.5704564452171326, 0.796875], "r_norm": [7.822964668273926, 7.8375749588012695, 7.815549850463867], "ldr": {"dL_sdil": -2.248777389526367, "dL_bp": -2.248777389526367, "ratio": 1.0}}, {"step": 5600, "epoch": 11, "train_loss": 9.354345321655273, "cos_r_negg": [-0.19870108366012573, -0.4179230332374573, -0.4375], "cos_innovation_negg": [-0.19157151877880096, -0.5659834146499634, 0.29296875], "cos_apical_negg": [-0.19870108366012573, -0.4179230332374573, -0.4375], "cos_Ac_negg": [-0.1915706843137741, -0.5659831166267395, 0.29296875], "r_norm": [8.99805736541748, 9.00892448425293, 9.099859237670898]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.0892, "test_loss": 15.49762734375}, {"step": 5700, "epoch": 12, "train_loss": 12.49483871459961, "cos_r_negg": [0.032525885850191116, 0.10754494369029999, 0.11328125], "cos_innovation_negg": [-0.10293645411729813, -0.254874587059021, 0.52734375], "cos_apical_negg": [0.032525885850191116, 0.10754494369029999, 0.11328125], "cos_Ac_negg": [-0.10293140262365341, -0.25487449765205383, 0.52734375], "r_norm": [7.9246907234191895, 7.935434341430664, 7.929038047790527]}, {"step": 5800, "epoch": 12, "train_loss": 9.011554718017578, "cos_r_negg": [-0.1321522444486618, -0.40238243341445923, -0.421875], "cos_innovation_negg": [-0.12911565601825714, -0.4526625871658325, 1.0], "cos_apical_negg": [-0.1321522444486618, -0.40238243341445923, -0.421875], "cos_Ac_negg": [-0.13017389178276062, -0.4526626765727997, 1.0], "r_norm": [8.323644638061523, 8.33032512664795, 8.334144592285156]}, {"step": 5900, "epoch": 12, "train_loss": 7.745987415313721, "cos_r_negg": [-0.10914679616689682, -0.0486505962908268, -0.05078125], "cos_innovation_negg": [-0.19060179591178894, -0.37550675868988037, 0.6328125], "cos_apical_negg": [-0.10914679616689682, -0.0486505962908268, -0.05078125], "cos_Ac_negg": [-0.19060474634170532, -0.37550675868988037, 0.6328125], "r_norm": [8.682197570800781, 8.69565200805664, 8.712076187133789]}, {"step": 6000, "epoch": 12, "train_loss": 5.899158477783203, "cos_r_negg": [-0.056063584983348846, 0.3010885715484619, 0.31640625], "cos_innovation_negg": [-0.174213707447052, 0.010614042170345783, 0.73046875], "cos_apical_negg": [-0.056063584983348846, 0.3010885715484619, 0.31640625], "cos_Ac_negg": [-0.1742156744003296, 0.010614079423248768, 0.73046875], "r_norm": [8.285435676574707, 8.301414489746094, 8.360922813415527], "ldr": {"dL_sdil": -1.9572467803955078, "dL_bp": -1.9572467803955078, "ratio": 1.0}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.1135, "test_loss": 15.75487265625}, {"step": 6100, "epoch": 13, "train_loss": 13.229394912719727, "cos_r_negg": [-0.16768373548984528, -0.7887568473815918, -0.828125], "cos_innovation_negg": [-0.10204503685235977, -0.193100705742836, 1.0], "cos_apical_negg": [-0.16768373548984528, -0.7887568473815918, -0.828125], "cos_Ac_negg": [-0.10205547511577606, -0.1931007206439972, 1.0], "r_norm": [8.477532386779785, 8.494281768798828, 8.524208068847656]}, {"step": 6200, "epoch": 13, "train_loss": 9.279717445373535, "cos_r_negg": [-0.016502268612384796, -0.2530028223991394, -0.265625], "cos_innovation_negg": [-0.1391972303390503, -0.009444005787372589, 0.78515625], "cos_apical_negg": [-0.016502268612384796, -0.2530028223991394, -0.265625], "cos_Ac_negg": [-0.139200359582901, -0.009443961083889008, 0.78515625], "r_norm": [7.61728572845459, 7.632712364196777, 7.58719539642334]}, {"step": 6300, "epoch": 13, "train_loss": 15.58309555053711, "cos_r_negg": [-0.05130664259195328, -0.6103756427764893, -0.640625], "cos_innovation_negg": [-0.2856462895870209, 0.17206421494483948, 1.0], "cos_apical_negg": [-0.05130664259195328, -0.6103756427764893, -0.640625], "cos_Ac_negg": [-0.28565138578414917, 0.17206427454948425, 1.0], "r_norm": [7.944525241851807, 7.95833683013916, 7.917252540588379]}, {"step": 6400, "epoch": 13, "train_loss": 8.362821578979492, "cos_r_negg": [-0.12808667123317719, 0.10740907490253448, 0.11328125], "cos_innovation_negg": [-0.20652149617671967, -0.21156379580497742, 1.0], "cos_apical_negg": [-0.12808667123317719, 0.10740907490253448, 0.11328125], "cos_Ac_negg": [-0.20650528371334076, -0.21156372129917145, 1.0], "r_norm": [7.8779730796813965, 7.89633321762085, 7.933520793914795]}, {"step": 6500, "epoch": 13, "train_loss": 14.536314964294434, "cos_r_negg": [0.2417798936367035, 0.5563158988952637, 0.5859375], "cos_innovation_negg": [0.0997333899140358, 0.3189339339733124, 1.0], "cos_apical_negg": [0.2417798936367035, 0.5563158988952637, 0.5859375], "cos_Ac_negg": [0.09973182529211044, 0.31893399357795715, 1.0], "r_norm": [8.752015113830566, 8.779094696044922, 8.890295028686523], "ldr": {"dL_sdil": -2.398242950439453, "dL_bp": -2.398242950439453, "ratio": 1.0}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.0974, "test_loss": 14.60253017578125}, {"step": 6600, "epoch": 14, "train_loss": 10.260499000549316, "cos_r_negg": [0.1832551658153534, 0.5563303232192993, 0.5859375], "cos_innovation_negg": [-0.05300292372703552, 0.023668130859732628, 1.0], "cos_apical_negg": [0.1832551658153534, 0.5563303232192993, 0.5859375], "cos_Ac_negg": [-0.05303143709897995, 0.02366822212934494, 1.0], "r_norm": [8.428667068481445, 8.447489738464355, 8.54556941986084]}, {"step": 6700, "epoch": 14, "train_loss": 6.828721046447754, "cos_r_negg": [-0.05906118452548981, -0.4595968723297119, -0.484375], "cos_innovation_negg": [-0.27511993050575256, -0.14427368342876434, 0.40234375], "cos_apical_negg": [-0.05906118452548981, -0.4595968723297119, -0.484375], "cos_Ac_negg": [-0.27510368824005127, -0.14427348971366882, 0.40234375], "r_norm": [8.577927589416504, 8.599018096923828, 8.619671821594238]}, {"step": 6800, "epoch": 14, "train_loss": 7.994805812835693, "cos_r_negg": [-0.06416022032499313, -0.19613522291183472, -0.20703125], "cos_innovation_negg": [-0.10924974083900452, -0.42175769805908203, 0.84375], "cos_apical_negg": [-0.06416022032499313, -0.19613522291183472, -0.20703125], "cos_Ac_negg": [-0.10924537479877472, -0.42175793647766113, 0.84375], "r_norm": [8.493719100952148, 8.510171890258789, 8.519445419311523]}, {"step": 6900, "epoch": 14, "train_loss": 4.80057430267334, "cos_r_negg": [-0.08321009576320648, -0.3997562527656555, -0.421875], "cos_innovation_negg": [-0.04034553840756416, -0.05306457355618477, 1.0], "cos_apical_negg": [-0.08321009576320648, -0.3997562527656555, -0.421875], "cos_Ac_negg": [-0.04039899259805679, -0.05306461080908775, 1.0], "r_norm": [8.382085800170898, 8.395641326904297, 8.394950866699219]}, {"step": 7000, "epoch": 14, "train_loss": 12.29056167602539, "cos_r_negg": [-0.13885396718978882, 0.10711666941642761, 0.11328125], "cos_innovation_negg": [-0.17264874279499054, -0.26289743185043335, 0.73046875], "cos_apical_negg": [-0.13885396718978882, 0.10711666941642761, 0.11328125], "cos_Ac_negg": [-0.17285296320915222, -0.2628975212574005, 0.73046875], "r_norm": [8.34173583984375, 8.353206634521484, 8.408172607421875], "ldr": {"dL_sdil": -2.191899299621582, "dL_bp": -2.191899299621582, "ratio": 1.0}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.1009, "test_loss": 13.5468400390625}], "final": {"test_acc": 0.1009, "test_loss": 13.5468400390625, "wall_s": 22.49405598640442, "cos_r_negg": [-0.15764665603637695, -0.40100422501564026, -0.421875], "cos_innovation_negg": [-0.2420637011528015, -0.1236443817615509, 0.84375], "cos_apical_negg": [-0.15764665603637695, -0.40100422501564026, -0.421875], "cos_Ac_negg": [-0.2420753389596939, -0.12364459037780762, 0.84375]}}