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": 1, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p5_raw_s1", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.327335834503174, "cos_r_negg": [0.019747942686080933, -0.012253424152731895, -0.010189061984419823], "cos_innovation_negg": [0.032370083034038544, -0.011818937957286835, -0.006588317919522524], "cos_apical_negg": [0.019747942686080933, -0.012253424152731895, -0.010189061984419823], "cos_Ac_negg": [0.03236745670437813, -0.011820998974144459, -0.0065889982506632805], "r_norm": [8.022297859191895, 6.936640739440918, 6.413236618041992], "ldr": {"dL_sdil": 0.03699207305908203, "dL_bp": -0.16228103637695312, "ratio": -0.22795068287064243}}, {"step": 100, "epoch": 0, "train_loss": 10.255548477172852, "cos_r_negg": [0.019994664937257767, 0.1572040170431137, -0.39453125], "cos_innovation_negg": [0.026966765522956848, 0.23288485407829285, 0.1953125], "cos_apical_negg": [0.019994664937257767, 0.1572040170431137, -0.39453125], "cos_Ac_negg": [0.026951849460601807, 0.23288394510746002, 0.1953125], "r_norm": [10.35006046295166, 9.845701217651367, 9.498064994812012]}, {"step": 200, "epoch": 0, "train_loss": 8.509317398071289, "cos_r_negg": [-0.028591888025403023, 0.01346166804432869, -0.8125], "cos_innovation_negg": [-0.04961170628666878, -0.08859357237815857, 0.625], "cos_apical_negg": [-0.028591888025403023, 0.01346166804432869, -0.8125], "cos_Ac_negg": [-0.04960859194397926, -0.08857496082782745, 0.625], "r_norm": [10.754764556884766, 10.037720680236816, 10.064632415771484]}, {"step": 300, "epoch": 0, "train_loss": 17.83721160888672, "cos_r_negg": [-0.057352762669324875, -0.0851992815732956, -0.6171875], "cos_innovation_negg": [-0.041011743247509, -0.05717262253165245, -0.02734375], "cos_apical_negg": [-0.057352762669324875, -0.0851992815732956, -0.6171875], "cos_Ac_negg": [-0.04092578589916229, -0.057143256068229675, -0.02734375], "r_norm": [9.938236236572266, 9.55727767944336, 9.466873168945312]}, {"step": 400, "epoch": 0, "train_loss": 7.2290825843811035, "cos_r_negg": [0.11731478571891785, 0.27728092670440674, -0.58984375], "cos_innovation_negg": [0.15003573894500732, 0.16362930834293365, 0.0], "cos_apical_negg": [0.11731478571891785, 0.27728092670440674, -0.58984375], "cos_Ac_negg": [0.1500292718410492, 0.1636141985654831, 0.0], "r_norm": [9.497688293457031, 9.26835823059082, 9.029476165771484]}, {"epoch_end": 0, "step": 469, "test_acc": 0.1028, "test_loss": 11.4579041015625}, {"step": 500, "epoch": 1, "train_loss": 17.226133346557617, "cos_r_negg": [0.040775030851364136, 0.10044033080339432, -0.8046875], "cos_innovation_negg": [0.041798170655965805, 0.018371954560279846, 0.02734375], "cos_apical_negg": [0.040775030851364136, 0.10044033080339432, -0.8046875], "cos_Ac_negg": [0.041797056794166565, 0.018360260874032974, 0.02734375], "r_norm": [9.502269744873047, 9.298660278320312, 9.243734359741211], "ldr": {"dL_sdil": -2.046304702758789, "dL_bp": -2.046304702758789, "ratio": 1.0}}, {"step": 600, "epoch": 1, "train_loss": 11.26502513885498, "cos_r_negg": [-0.012791586108505726, 0.040445100516080856, -0.8046875], "cos_innovation_negg": [0.017647143453359604, 0.04069416970014572, 0.4296875], "cos_apical_negg": [-0.012791586108505726, 0.040445100516080856, -0.8046875], "cos_Ac_negg": [0.017636168748140335, 0.04069536551833153, 0.4296875], "r_norm": [9.236584663391113, 8.872383117675781, 8.91727066040039]}, {"step": 700, "epoch": 1, "train_loss": 11.11388874053955, "cos_r_negg": [0.013096985407173634, -0.15340036153793335, -0.484375], "cos_innovation_negg": [-0.040903255343437195, -0.2035406231880188, 0.4296875], "cos_apical_negg": [0.013096985407173634, -0.15340036153793335, -0.484375], "cos_Ac_negg": [-0.04156612977385521, -0.20494085550308228, 0.4296875], "r_norm": [9.124551773071289, 9.2476806640625, 9.073410034179688]}, {"step": 800, "epoch": 1, "train_loss": 12.390909194946289, "cos_r_negg": [-0.015279165469110012, 0.05406123772263527, -0.640625], "cos_innovation_negg": [-0.03402719646692276, -0.05897021293640137, -0.6484375], "cos_apical_negg": [-0.015279165469110012, 0.05406123772263527, -0.640625], "cos_Ac_negg": [-0.03403740003705025, -0.058980755507946014, -0.6484375], "r_norm": [8.950582504272461, 9.133920669555664, 9.43021011352539]}, {"step": 900, "epoch": 1, "train_loss": 11.088262557983398, "cos_r_negg": [-0.011971227824687958, -0.16905423998832703, -0.3125], "cos_innovation_negg": [-0.05211924761533737, -0.2989688515663147, 0.6015625], "cos_apical_negg": [-0.011971227824687958, -0.16905423998832703, -0.3125], "cos_Ac_negg": [-0.05179830640554428, -0.29844844341278076, 0.6015625], "r_norm": [8.590007781982422, 8.724785804748535, 8.576166152954102]}, {"epoch_end": 1, "step": 938, "test_acc": 0.0958, "test_loss": 10.70472607421875}, {"step": 1000, "epoch": 2, "train_loss": 7.3060622215271, "cos_r_negg": [0.010516166687011719, -0.11391425132751465, -0.8046875], "cos_innovation_negg": [-0.04206956923007965, 0.28496024012565613, 0.4296875], "cos_apical_negg": [0.010516166687011719, -0.11391425132751465, -0.8046875], "cos_Ac_negg": [-0.042079176753759384, 0.28496259450912476, 0.4296875], "r_norm": [8.405755043029785, 8.252345085144043, 8.325518608093262], "ldr": {"dL_sdil": -1.3452386856079102, "dL_bp": -1.3452386856079102, "ratio": 1.0}}, {"step": 1100, "epoch": 2, "train_loss": 14.472942352294922, "cos_r_negg": [-0.011405833065509796, -0.06569328904151917, -0.6171875], "cos_innovation_negg": [0.02072197198867798, 0.23664258420467377, 0.19140625], "cos_apical_negg": [-0.011405833065509796, -0.06569328904151917, -0.6171875], "cos_Ac_negg": [0.020735356956720352, 0.23664231598377228, 0.19140625], "r_norm": [8.573768615722656, 8.777698516845703, 8.76460075378418]}, {"step": 1200, "epoch": 2, "train_loss": 12.044473648071289, "cos_r_negg": [0.014382035471498966, -0.22828882932662964, -0.796875], "cos_innovation_negg": [-0.05162845551967621, 0.24645277857780457, 1.0], "cos_apical_negg": [0.014382035471498966, -0.22828882932662964, -0.796875], "cos_Ac_negg": [-0.05164097994565964, 0.2464318722486496, 1.0], "r_norm": [8.376583099365234, 8.457967758178711, 8.215885162353516]}, {"step": 1300, "epoch": 2, "train_loss": 6.5063629150390625, "cos_r_negg": [0.017980705946683884, -0.09746164083480835, -0.58984375], "cos_innovation_negg": [0.024920618161559105, 0.1227433979511261, -0.00390625], "cos_apical_negg": [0.017980705946683884, -0.09746164083480835, -0.58984375], "cos_Ac_negg": [0.024907678365707397, 0.12276691198348999, -0.00390625], "r_norm": [7.911869049072266, 7.881860733032227, 7.70210599899292]}, {"step": 1400, "epoch": 2, "train_loss": 9.817485809326172, "cos_r_negg": [0.030312037095427513, -0.30478209257125854, -0.8046875], "cos_innovation_negg": [0.054487232118844986, 0.20171639323234558, 0.4296875], "cos_apical_negg": [0.030312037095427513, -0.30478209257125854, -0.8046875], "cos_Ac_negg": [0.05447912961244583, 0.2017064243555069, 0.4296875], "r_norm": [8.463221549987793, 8.475896835327148, 8.521780014038086]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.1032, "test_loss": 11.7731794921875}, {"step": 1500, "epoch": 3, "train_loss": 11.53721809387207, "cos_r_negg": [0.05731525644659996, -0.038483504205942154, -0.8046875], "cos_innovation_negg": [0.15988948941230774, -0.16430561244487762, -0.484375], "cos_apical_negg": [0.05731525644659996, -0.038483504205942154, -0.8046875], "cos_Ac_negg": [0.15984931588172913, -0.16432219743728638, -0.484375], "r_norm": [8.08842945098877, 8.42395305633545, 8.705155372619629], "ldr": {"dL_sdil": -2.464749336242676, "dL_bp": -2.464749336242676, "ratio": 1.0}}, {"step": 1600, "epoch": 3, "train_loss": 8.154747009277344, "cos_r_negg": [-0.027677316218614578, -0.08400880545377731, -0.6171875], "cos_innovation_negg": [0.04914071038365364, -0.031338125467300415, 0.0234375], "cos_apical_negg": [-0.027677316218614578, -0.08400880545377731, -0.6171875], "cos_Ac_negg": [0.04915933310985565, -0.03130488842725754, 0.0234375], "r_norm": [8.72436237335205, 8.844265937805176, 8.89121150970459]}, {"step": 1700, "epoch": 3, "train_loss": 6.9549784660339355, "cos_r_negg": [-0.03542960062623024, 0.022684037685394287, -0.2421875], "cos_innovation_negg": [-0.06088436022400856, 0.08918435871601105, 0.5859375], "cos_apical_negg": [-0.03542960062623024, 0.022684037685394287, -0.2421875], "cos_Ac_negg": [-0.06088074669241905, 0.08920293301343918, 0.5859375], "r_norm": [8.389348030090332, 8.436502456665039, 8.428319931030273]}, {"step": 1800, "epoch": 3, "train_loss": 6.131024360656738, "cos_r_negg": [0.035309989005327225, -0.2508002817630768, -0.8046875], "cos_innovation_negg": [-0.0997670516371727, 0.292773962020874, 0.2109375], "cos_apical_negg": [0.035309989005327225, -0.2508002817630768, -0.8046875], "cos_Ac_negg": [-0.099770687520504, 0.2927319407463074, 0.2109375], "r_norm": [8.066953659057617, 7.979948043823242, 7.825244903564453]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.0958, "test_loss": 15.22708388671875}, {"step": 1900, "epoch": 4, "train_loss": 11.77390193939209, "cos_r_negg": [0.08595188707113266, -0.5049331188201904, -0.6171875], "cos_innovation_negg": [-0.07709004729986191, 0.2698267698287964, 0.21484375], "cos_apical_negg": [0.08595188707113266, -0.5049331188201904, -0.6171875], "cos_Ac_negg": [-0.07708196341991425, 0.2698175311088562, 0.21484375], "r_norm": [8.575183868408203, 8.621679306030273, 8.65350341796875]}, {"step": 2000, "epoch": 4, "train_loss": 11.215351104736328, "cos_r_negg": [-0.009597906842827797, -0.3489310145378113, -0.421875], "cos_innovation_negg": [-0.02166447415947914, 0.18506285548210144, -0.16015625], "cos_apical_negg": [-0.009597906842827797, -0.3489310145378113, -0.421875], "cos_Ac_negg": [-0.02163964882493019, 0.18505842983722687, -0.16015625], "r_norm": [8.188566207885742, 8.312118530273438, 8.16801643371582], "ldr": {"dL_sdil": -1.9362831115722656, "dL_bp": -1.9362821578979492, "ratio": 1.0000004925285877}}, {"step": 2100, "epoch": 4, "train_loss": 14.727140426635742, "cos_r_negg": [0.014038543216884136, -0.674130916595459, -0.8046875], "cos_innovation_negg": [-0.024987127631902695, -0.42598649859428406, 0.0], "cos_apical_negg": [0.014038543216884136, -0.674130916595459, -0.8046875], "cos_Ac_negg": [-0.024976588785648346, -0.42599087953567505, 0.0], "r_norm": [8.830476760864258, 8.85820198059082, 8.99573802947998]}, {"step": 2200, "epoch": 4, "train_loss": 10.289848327636719, "cos_r_negg": [0.029824919998645782, -0.3603938817977905, -0.4296875], "cos_innovation_negg": [0.0957798957824707, 0.0723714679479599, 0.375], "cos_apical_negg": [0.029824919998645782, -0.3603938817977905, -0.4296875], "cos_Ac_negg": [0.09578461199998856, 0.07236636430025101, 0.375], "r_norm": [8.682836532592773, 8.686776161193848, 8.668912887573242]}, {"step": 2300, "epoch": 4, "train_loss": 9.162446975708008, "cos_r_negg": [0.09142471849918365, -0.6800583004951477, -0.8046875], "cos_innovation_negg": [-0.1525205671787262, -0.09147711098194122, 0.0234375], "cos_apical_negg": [0.09142471849918365, -0.6800583004951477, -0.8046875], "cos_Ac_negg": [-0.15252111852169037, -0.09148325026035309, 0.0234375], "r_norm": [8.597091674804688, 8.609399795532227, 8.657405853271484]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.0958, "test_loss": 11.5237857421875}, {"step": 2400, "epoch": 5, "train_loss": 14.908965110778809, "cos_r_negg": [0.2208012491464615, -0.6925382614135742, -0.8125], "cos_innovation_negg": [-0.016925405710935593, 0.1201857477426529, -0.1953125], "cos_apical_negg": [0.2208012491464615, -0.6925382614135742, -0.8125], "cos_Ac_negg": [-0.010413708165287971, 0.12075091898441315, -0.1953125], "r_norm": [8.76136589050293, 8.790179252624512, 8.832708358764648]}, {"step": 2500, "epoch": 5, "train_loss": 9.113646507263184, "cos_r_negg": [0.05305565521121025, -0.6685764193534851, -0.78515625], "cos_innovation_negg": [-0.006513811647891998, -0.4660874009132385, -0.1953125], "cos_apical_negg": [0.05305565521121025, -0.6685764193534851, -0.78515625], "cos_Ac_negg": [-0.006514234934002161, -0.4660630524158478, -0.1953125], "r_norm": [8.754865646362305, 8.771283149719238, 8.807271957397461], "ldr": {"dL_sdil": -3.0341272354125977, "dL_bp": -3.0341272354125977, "ratio": 1.0}}, {"step": 2600, "epoch": 5, "train_loss": 11.880722045898438, "cos_r_negg": [-0.10047242045402527, -0.325664222240448, -0.3984375], "cos_innovation_negg": [-0.17108644545078278, 0.05257571488618851, 0.19140625], "cos_apical_negg": [-0.10047242045402527, -0.325664222240448, -0.3984375], "cos_Ac_negg": [-0.17110908031463623, 0.052568379789590836, 0.19140625], "r_norm": [8.301441192626953, 8.415926933288574, 8.310450553894043]}, {"step": 2700, "epoch": 5, "train_loss": 11.036749839782715, "cos_r_negg": [-0.20094235241413116, -0.6670376062393188, -0.8046875], "cos_innovation_negg": [-0.10651130229234695, -0.23025165498256683, 0.4296875], "cos_apical_negg": [-0.20094235241413116, -0.6670376062393188, -0.8046875], "cos_Ac_negg": [-0.10646158456802368, -0.23025259375572205, 0.4296875], "r_norm": [8.485367774963379, 8.47578239440918, 8.530881881713867]}, {"step": 2800, "epoch": 5, "train_loss": 23.670425415039062, "cos_r_negg": [-0.13499264419078827, -0.7049131393432617, -0.8359375], "cos_innovation_negg": [0.08350823819637299, -0.4116736650466919, -1.0], "cos_apical_negg": [-0.13499264419078827, -0.7049131393432617, -0.8359375], "cos_Ac_negg": [0.08351808786392212, -0.41167405247688293, -1.0], "r_norm": [8.97557258605957, 9.089457511901855, 9.230287551879883]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.0982, "test_loss": 11.89017373046875}, {"step": 2900, "epoch": 6, "train_loss": 10.981536865234375, "cos_r_negg": [-0.18273533880710602, -0.6601389646530151, -0.8046875], "cos_innovation_negg": [0.2024480402469635, -0.30576205253601074, 0.19140625], "cos_apical_negg": [-0.18273533880710602, -0.6601389646530151, -0.8046875], "cos_Ac_negg": [0.20248305797576904, -0.3057602047920227, 0.19140625], "r_norm": [8.712766647338867, 8.71328353881836, 8.782217979431152]}, {"step": 3000, "epoch": 6, "train_loss": 12.230125427246094, "cos_r_negg": [-0.1647941619157791, -0.49704235792160034, -0.6015625], "cos_innovation_negg": [-0.08325158804655075, 0.590111494064331, 0.8046875], "cos_apical_negg": [-0.1647941619157791, -0.49704235792160034, -0.6015625], "cos_Ac_negg": [-0.08324957638978958, 0.5901111364364624, 0.8046875], "r_norm": [7.905643463134766, 7.888843536376953, 7.838493347167969], "ldr": {"dL_sdil": -1.1818304061889648, "dL_bp": -1.1818304061889648, "ratio": 1.0}}, {"step": 3100, "epoch": 6, "train_loss": 12.053131103515625, "cos_r_negg": [-0.2135394811630249, -0.6514860391616821, -0.78515625], "cos_innovation_negg": [0.25970882177352905, -0.3043336272239685, -0.00390625], "cos_apical_negg": [-0.2135394811630249, -0.6514860391616821, -0.78515625], "cos_Ac_negg": [0.25969114899635315, -0.30433493852615356, -0.00390625], "r_norm": [8.471572875976562, 8.472350120544434, 8.493823051452637]}, {"step": 3200, "epoch": 6, "train_loss": 11.591326713562012, "cos_r_negg": [-0.2027834802865982, -0.6619418859481812, -0.8046875], "cos_innovation_negg": [0.13849501311779022, -0.3411552906036377, -0.37109375], "cos_apical_negg": [-0.2027834802865982, -0.6619418859481812, -0.8046875], "cos_Ac_negg": [0.13849510252475739, -0.34115615487098694, -0.37109375], "r_norm": [8.946978569030762, 9.021909713745117, 9.154298782348633]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.1135, "test_loss": 7.05854375}, {"step": 3300, "epoch": 7, "train_loss": 17.59749984741211, "cos_r_negg": [-0.08525792509317398, -0.35279908776283264, -0.42578125], "cos_innovation_negg": [0.08196399360895157, -0.3393060266971588, -0.58984375], "cos_apical_negg": [-0.08525792509317398, -0.35279908776283264, -0.42578125], "cos_Ac_negg": [0.08196558058261871, -0.33930644392967224, -0.58984375], "r_norm": [9.134340286254883, 9.198554039001465, 9.27498722076416]}, {"step": 3400, "epoch": 7, "train_loss": 14.165465354919434, "cos_r_negg": [-0.0696418508887291, -0.16134686768054962, -0.1953125], "cos_innovation_negg": [0.36436980962753296, 0.09845741093158722, 0.17578125], "cos_apical_negg": [-0.0696418508887291, -0.16134686768054962, -0.1953125], "cos_Ac_negg": [0.3667339086532593, 0.09845790266990662, 0.17578125], "r_norm": [8.55362606048584, 8.533271789550781, 8.539839744567871]}, {"step": 3500, "epoch": 7, "train_loss": 16.618972778320312, "cos_r_negg": [-0.21431967616081238, -0.6796319484710693, -0.828125], "cos_innovation_negg": [0.047853875905275345, 0.24010559916496277, 0.6015625], "cos_apical_negg": [-0.21431967616081238, -0.6796319484710693, -0.828125], "cos_Ac_negg": [0.04785699397325516, 0.24010488390922546, 0.6015625], "r_norm": [8.068330764770508, 8.11501407623291, 8.141497611999512], "ldr": {"dL_sdil": -2.8917312622070312, "dL_bp": -2.8917312622070312, "ratio": 1.0}}, {"step": 3600, "epoch": 7, "train_loss": 17.49654769897461, "cos_r_negg": [-0.05472403019666672, -0.6517399549484253, -0.796875], "cos_innovation_negg": [0.2943275272846222, 0.1049397736787796, 0.41796875], "cos_apical_negg": [-0.05472403019666672, -0.6517399549484253, -0.796875], "cos_Ac_negg": [0.29432404041290283, 0.10493916273117065, 0.41796875], "r_norm": [8.525372505187988, 8.524673461914062, 8.515222549438477]}, {"step": 3700, "epoch": 7, "train_loss": 10.19696044921875, "cos_r_negg": [-0.22893820703029633, -0.6602572798728943, -0.8046875], "cos_innovation_negg": [0.2143981009721756, 0.029962781816720963, 0.19140625], "cos_apical_negg": [-0.22893820703029633, -0.6602572798728943, -0.8046875], "cos_Ac_negg": [0.21438570320606232, 0.02996237389743328, 0.19140625], "r_norm": [8.477952003479004, 8.456353187561035, 8.47607135772705]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.098, "test_loss": 9.13295830078125}, {"step": 3800, "epoch": 8, "train_loss": 9.953214645385742, "cos_r_negg": [-0.19311043620109558, -0.6523042321205139, -0.8046875], "cos_innovation_negg": [0.2931424677371979, 0.2158634513616562, 0.41015625], "cos_apical_negg": [-0.19311043620109558, -0.6523042321205139, -0.8046875], "cos_Ac_negg": [0.29322537779808044, 0.21586306393146515, 0.41015625], "r_norm": [8.249199867248535, 8.287460327148438, 8.300777435302734]}, {"step": 3900, "epoch": 8, "train_loss": 6.829085350036621, "cos_r_negg": [-0.20397111773490906, -0.6477262377738953, -0.8046875], "cos_innovation_negg": [0.17422093451023102, 0.44347500801086426, 0.625], "cos_apical_negg": [-0.20397111773490906, -0.6477262377738953, -0.8046875], "cos_Ac_negg": [0.1742263287305832, 0.44347482919692993, 0.625], "r_norm": [8.033866882324219, 8.039278030395508, 8.032636642456055]}, {"step": 4000, "epoch": 8, "train_loss": 10.8768310546875, "cos_r_negg": [-0.13105210661888123, -0.25949379801750183, -0.31640625], "cos_innovation_negg": [-0.10799551010131836, -0.32377392053604126, -0.421875], "cos_apical_negg": [-0.13105210661888123, -0.25949379801750183, -0.31640625], "cos_Ac_negg": [-0.10805001109838486, -0.3237740099430084, -0.421875], "r_norm": [8.77724838256836, 8.795995712280273, 8.798884391784668], "ldr": {"dL_sdil": -1.1976318359375, "dL_bp": -1.1976318359375, "ratio": 1.0}}, {"step": 4100, "epoch": 8, "train_loss": 13.615104675292969, "cos_r_negg": [0.04135633632540703, 0.04593779146671295, 0.05859375], "cos_innovation_negg": [0.2123039811849594, 0.08738474547863007, 0.359375], "cos_apical_negg": [0.04135633632540703, 0.04593779146671295, 0.05859375], "cos_Ac_negg": [0.2123110294342041, 0.08738459646701813, 0.359375], "r_norm": [8.383914947509766, 8.420074462890625, 8.391923904418945]}, {"step": 4200, "epoch": 8, "train_loss": 18.92833709716797, "cos_r_negg": [-0.13118402659893036, -0.4789707064628601, -0.58984375], "cos_innovation_negg": [0.2658495604991913, 0.11149245500564575, 0.46875], "cos_apical_negg": [-0.13118402659893036, -0.4789707064628601, -0.58984375], "cos_Ac_negg": [0.2658456563949585, 0.11149241775274277, 0.46875], "r_norm": [8.485389709472656, 8.5098237991333, 8.531970977783203]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.0982, "test_loss": 13.37454501953125}, {"step": 4300, "epoch": 9, "train_loss": 10.261911392211914, "cos_r_negg": [-0.09637166559696198, -0.5956960320472717, -0.73046875], "cos_innovation_negg": [-0.3425491750240326, -0.6807563900947571, -0.8359375], "cos_apical_negg": [-0.09637166559696198, -0.5956960320472717, -0.73046875], "cos_Ac_negg": [-0.3426416516304016, -0.6807563900947571, -0.8359375], "r_norm": [8.899179458618164, 8.92036247253418, 8.970678329467773]}, {"step": 4400, "epoch": 9, "train_loss": 10.811222076416016, "cos_r_negg": [-0.1516168862581253, -0.658687949180603, -0.8125], "cos_innovation_negg": [0.07316669821739197, -0.07895853370428085, -0.16015625], "cos_apical_negg": [-0.1516168862581253, -0.658687949180603, -0.8125], "cos_Ac_negg": [0.07483233511447906, -0.07895766943693161, -0.16015625], "r_norm": [8.702716827392578, 8.70215129852295, 8.762594223022461]}, {"step": 4500, "epoch": 9, "train_loss": 11.859151840209961, "cos_r_negg": [-0.09459742903709412, -0.47759270668029785, -0.58984375], "cos_innovation_negg": [0.2598411738872528, 0.034248627722263336, 0.46875], "cos_apical_negg": [-0.09459742903709412, -0.47759270668029785, -0.58984375], "cos_Ac_negg": [0.25983455777168274, 0.03424864262342453, 0.46875], "r_norm": [8.425926208496094, 8.45444107055664, 8.458234786987305], "ldr": {"dL_sdil": -1.6856966018676758, "dL_bp": -1.6856956481933594, "ratio": 1.0000005657452562}}, {"step": 4600, "epoch": 9, "train_loss": 13.60978889465332, "cos_r_negg": [-0.262395441532135, -0.6500300168991089, -0.8046875], "cos_innovation_negg": [0.1120535135269165, -0.09465036541223526, 0.625], "cos_apical_negg": [-0.262395441532135, -0.6500300168991089, -0.8046875], "cos_Ac_negg": [0.11534196883440018, -0.09464994072914124, 0.625], "r_norm": [8.468623161315918, 8.479522705078125, 8.519412994384766]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.1028, "test_loss": 19.0580978515625}, {"step": 4700, "epoch": 10, "train_loss": 13.967798233032227, "cos_r_negg": [-0.19663439691066742, -0.6369585990905762, -0.78515625], "cos_innovation_negg": [0.018648996949195862, -0.1480116993188858, -0.34765625], "cos_apical_negg": [-0.19663439691066742, -0.6369585990905762, -0.78515625], "cos_Ac_negg": [0.018578849732875824, -0.14801178872585297, -0.34765625], "r_norm": [8.440733909606934, 8.4483642578125, 8.463715553283691]}, {"step": 4800, "epoch": 10, "train_loss": 6.129626274108887, "cos_r_negg": [-0.10255283862352371, -0.6352357864379883, -0.78515625], "cos_innovation_negg": [-0.10318326950073242, -0.16379831731319427, -0.34765625], "cos_apical_negg": [-0.10255283862352371, -0.6352357864379883, -0.78515625], "cos_Ac_negg": [-0.10324013233184814, -0.1637982428073883, -0.34765625], "r_norm": [8.470785140991211, 8.488651275634766, 8.482808113098145]}, {"step": 4900, "epoch": 10, "train_loss": 12.707818984985352, "cos_r_negg": [0.040788255631923676, -0.004775119945406914, -0.00390625], "cos_innovation_negg": [0.289998322725296, 0.4808981120586395, 0.6484375], "cos_apical_negg": [0.040788255631923676, -0.004775119945406914, -0.00390625], "cos_Ac_negg": [0.28999438881874084, 0.4808982014656067, 0.6484375], "r_norm": [8.638195037841797, 8.648427963256836, 8.758851051330566]}, {"step": 5000, "epoch": 10, "train_loss": 7.799116134643555, "cos_r_negg": [-0.03289207071065903, 0.07600315660238266, 0.09375], "cos_innovation_negg": [-0.03519286960363388, 0.356702983379364, 0.41796875], "cos_apical_negg": [-0.03289207071065903, 0.07600315660238266, 0.09375], "cos_Ac_negg": [-0.035194121301174164, 0.356702983379364, 0.41796875], "r_norm": [8.377689361572266, 8.400264739990234, 8.440694808959961], "ldr": {"dL_sdil": -1.9702091217041016, "dL_bp": -1.9702091217041016, "ratio": 1.0}}, {"step": 5100, "epoch": 10, "train_loss": 10.47219467163086, "cos_r_negg": [-0.03785387799143791, -0.48272615671157837, -0.6015625], "cos_innovation_negg": [0.07259231805801392, 0.48528632521629333, 0.6328125], "cos_apical_negg": [-0.03785387799143791, -0.48272615671157837, -0.6015625], "cos_Ac_negg": [0.07262501865625381, 0.4852863848209381, 0.6328125], "r_norm": [7.70461368560791, 7.7034382820129395, 7.708110332489014]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.0892, "test_loss": 12.00000849609375}, {"step": 5200, "epoch": 11, "train_loss": 10.71782112121582, "cos_r_negg": [-0.18562965095043182, -0.5159285068511963, -0.640625], "cos_innovation_negg": [-0.03629561886191368, -0.4750930964946747, -0.31640625], "cos_apical_negg": [-0.18562965095043182, -0.5159285068511963, -0.640625], "cos_Ac_negg": [-0.03630782663822174, -0.47509312629699707, -0.31640625], "r_norm": [9.172325134277344, 9.21832275390625, 9.317131996154785]}, {"step": 5300, "epoch": 11, "train_loss": 5.943276405334473, "cos_r_negg": [0.061552613973617554, 0.041206296533346176, 0.05078125], "cos_innovation_negg": [0.08826838433742523, -0.12186316400766373, -0.1640625], "cos_apical_negg": [0.061552613973617554, 0.041206296533346176, 0.05078125], "cos_Ac_negg": [0.08823709189891815, -0.12186320126056671, -0.1640625], "r_norm": [8.823055267333984, 8.838237762451172, 8.881866455078125]}, {"step": 5400, "epoch": 11, "train_loss": 10.271692276000977, "cos_r_negg": [-0.15554723143577576, -0.4915660619735718, -0.6171875], "cos_innovation_negg": [0.0002655188145581633, 0.017148923128843307, 0.19140625], "cos_apical_negg": [-0.15554723143577576, -0.4915660619735718, -0.6171875], "cos_Ac_negg": [0.00026533339405432343, 0.017149074003100395, 0.19140625], "r_norm": [7.972602844238281, 7.987299919128418, 7.990858554840088]}, {"step": 5500, "epoch": 11, "train_loss": 12.208606719970703, "cos_r_negg": [-0.19998127222061157, -0.6461871862411499, -0.8046875], "cos_innovation_negg": [-0.08988816291093826, 0.005642070434987545, 0.40625], "cos_apical_negg": [-0.19998127222061157, -0.6461871862411499, -0.8046875], "cos_Ac_negg": [-0.08987265080213547, 0.005642108619213104, 0.40625], "r_norm": [8.203372955322266, 8.207939147949219, 8.234451293945312], "ldr": {"dL_sdil": -1.4868803024291992, "dL_bp": -1.4868803024291992, "ratio": 1.0}}, {"step": 5600, "epoch": 11, "train_loss": 8.259819030761719, "cos_r_negg": [0.05633822828531265, -0.33073586225509644, -0.4140625], "cos_innovation_negg": [0.11186914145946503, 0.27397483587265015, 0.39453125], "cos_apical_negg": [0.05633822828531265, -0.33073586225509644, -0.4140625], "cos_Ac_negg": [0.11187991499900818, 0.2739748954772949, 0.39453125], "r_norm": [8.12523365020752, 8.145190238952637, 8.183408737182617]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.098, "test_loss": 13.60080322265625}, {"step": 5700, "epoch": 12, "train_loss": 7.852395057678223, "cos_r_negg": [-0.28524982929229736, -0.6453471183776855, -0.8046875], "cos_innovation_negg": [0.05096225440502167, -0.10975302010774612, 0.78125], "cos_apical_negg": [-0.28524982929229736, -0.6453471183776855, -0.8046875], "cos_Ac_negg": [0.051130689680576324, -0.1097530722618103, 0.78125], "r_norm": [8.443368911743164, 8.461749076843262, 8.447426795959473]}, {"step": 5800, "epoch": 12, "train_loss": 12.04050064086914, "cos_r_negg": [-0.1931564211845398, -0.3389608561992645, -0.421875], "cos_innovation_negg": [0.3688060939311981, 0.4867904782295227, 0.640625], "cos_apical_negg": [-0.1931564211845398, -0.3389608561992645, -0.421875], "cos_Ac_negg": [0.3688081204891205, 0.4867904484272003, 0.640625], "r_norm": [7.668402671813965, 7.678630352020264, 7.639505386352539]}, {"step": 5900, "epoch": 12, "train_loss": 9.473217010498047, "cos_r_negg": [-0.20913226902484894, -0.6424360871315002, -0.8046875], "cos_innovation_negg": [-0.008895385079085827, -0.022748205810785294, 0.21875], "cos_apical_negg": [-0.20913226902484894, -0.6424360871315002, -0.8046875], "cos_Ac_negg": [-0.008909438736736774, -0.022748293355107307, 0.21875], "r_norm": [8.045585632324219, 8.056964874267578, 8.067422866821289]}, {"step": 6000, "epoch": 12, "train_loss": 4.6162109375, "cos_r_negg": [-0.048817843198776245, -0.3077565133571625, -0.38671875], "cos_innovation_negg": [0.03845749422907829, -0.09262891858816147, 0.265625], "cos_apical_negg": [-0.048817843198776245, -0.3077565133571625, -0.38671875], "cos_Ac_negg": [0.03863661363720894, -0.09262891113758087, 0.265625], "r_norm": [8.418392181396484, 8.431730270385742, 8.438087463378906], "ldr": {"dL_sdil": -1.201819896697998, "dL_bp": -1.201819896697998, "ratio": 1.0}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.1028, "test_loss": 8.93193388671875}, {"step": 6100, "epoch": 13, "train_loss": 6.096339225769043, "cos_r_negg": [0.12168270349502563, -0.1704285591840744, -0.21484375], "cos_innovation_negg": [-0.10440283268690109, 0.012353229336440563, 0.37109375], "cos_apical_negg": [0.12168270349502563, -0.1704285591840744, -0.21484375], "cos_Ac_negg": [-0.10440690815448761, 0.012353229336440563, 0.37109375], "r_norm": [7.819033622741699, 7.832849979400635, 7.832224369049072]}, {"step": 6200, "epoch": 13, "train_loss": 7.847175121307373, "cos_r_negg": [-0.048842836171388626, -0.5008080005645752, -0.6328125], "cos_innovation_negg": [0.06866491585969925, 0.4265682101249695, 0.8046875], "cos_apical_negg": [-0.048842836171388626, -0.5008080005645752, -0.6328125], "cos_Ac_negg": [0.06857999414205551, 0.42656785249710083, 0.8046875], "r_norm": [7.716916561126709, 7.72566032409668, 7.794162273406982]}, {"step": 6300, "epoch": 13, "train_loss": 7.977858066558838, "cos_r_negg": [-0.09518624097108841, -0.46747344732284546, -0.5859375], "cos_innovation_negg": [0.11895081400871277, 0.5588080883026123, 0.8046875], "cos_apical_negg": [-0.09518624097108841, -0.46747344732284546, -0.5859375], "cos_Ac_negg": [0.11893948167562485, 0.5588085651397705, 0.8046875], "r_norm": [7.622480869293213, 7.630342483520508, 7.614863395690918]}, {"step": 6400, "epoch": 13, "train_loss": 9.694928169250488, "cos_r_negg": [0.07838189601898193, -0.34082505106925964, -0.4296875], "cos_innovation_negg": [-0.19109903275966644, -0.07497815042734146, 0.19140625], "cos_apical_negg": [0.07838189601898193, -0.34082505106925964, -0.4296875], "cos_Ac_negg": [-0.19108383357524872, -0.07497777789831161, 0.19140625], "r_norm": [8.213350296020508, 8.22939682006836, 8.26217269897461]}, {"step": 6500, "epoch": 13, "train_loss": 9.161271095275879, "cos_r_negg": [-0.09720490127801895, -0.17380550503730774, -0.21875], "cos_innovation_negg": [-0.24518650770187378, -0.010658462531864643, 0.40234375], "cos_apical_negg": [-0.09720490127801895, -0.17380550503730774, -0.21875], "cos_Ac_negg": [-0.24520382285118103, -0.010658405721187592, 0.40234375], "r_norm": [8.145859718322754, 8.16114330291748, 8.184585571289062], "ldr": {"dL_sdil": -1.7441749572753906, "dL_bp": -1.7441740036010742, "ratio": 1.00000054677705}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.101, "test_loss": 11.3324763671875}, {"step": 6600, "epoch": 14, "train_loss": 10.590892791748047, "cos_r_negg": [-0.023335391655564308, 0.014607099816203117, 0.01953125], "cos_innovation_negg": [-0.2865261137485504, -0.3393670618534088, -0.1640625], "cos_apical_negg": [-0.023335391655564308, 0.014607099816203117, 0.01953125], "cos_Ac_negg": [-0.28653809428215027, -0.3393670916557312, -0.1640625], "r_norm": [8.343790054321289, 8.360283851623535, 8.364086151123047]}, {"step": 6700, "epoch": 14, "train_loss": 7.228377342224121, "cos_r_negg": [0.11366799473762512, -0.210951030254364, -0.265625], "cos_innovation_negg": [-0.3458082675933838, -0.3371458947658539, 0.16796875], "cos_apical_negg": [0.11366799473762512, -0.210951030254364, -0.265625], "cos_Ac_negg": [-0.3459179401397705, -0.337146133184433, 0.16796875], "r_norm": [8.159287452697754, 8.17740249633789, 8.223154067993164]}, {"step": 6800, "epoch": 14, "train_loss": 10.455390930175781, "cos_r_negg": [-0.12113658338785172, -0.635749340057373, -0.796875], "cos_innovation_negg": [-0.01943564973771572, 0.00035860203206539154, 0.40625], "cos_apical_negg": [-0.12113658338785172, -0.635749340057373, -0.796875], "cos_Ac_negg": [-0.019462957978248596, 0.00035974010825157166, 0.40625], "r_norm": [8.254014015197754, 8.264790534973145, 8.324295043945312]}, {"step": 6900, "epoch": 14, "train_loss": 11.511268615722656, "cos_r_negg": [-0.2255789190530777, -0.6680182218551636, -0.8359375], "cos_innovation_negg": [-0.24502550065517426, -0.37816905975341797, -0.56640625], "cos_apical_negg": [-0.2255789190530777, -0.6680182218551636, -0.8359375], "cos_Ac_negg": [-0.24504362046718597, -0.37816905975341797, -0.56640625], "r_norm": [8.650300025939941, 8.67970085144043, 8.728560447692871]}, {"step": 7000, "epoch": 14, "train_loss": 13.167366027832031, "cos_r_negg": [-0.19883595407009125, -0.6421783566474915, -0.8046875], "cos_innovation_negg": [-0.2787978947162628, -0.12518706917762756, 0.16015625], "cos_apical_negg": [-0.19883595407009125, -0.6421783566474915, -0.8046875], "cos_Ac_negg": [-0.2788105010986328, -0.1251869797706604, 0.16015625], "r_norm": [8.168951988220215, 8.182437896728516, 8.196771621704102], "ldr": {"dL_sdil": -1.5362911224365234, "dL_bp": -1.5362911224365234, "ratio": 1.0}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.1009, "test_loss": 8.2447443359375}], "final": {"test_acc": 0.1009, "test_loss": 8.2447443359375, "wall_s": 22.726298093795776, "cos_r_negg": [-0.19490432739257812, -0.6292647123336792, -0.78515625], "cos_innovation_negg": [-0.1356714516878128, -0.28259608149528503, -0.078125], "cos_apical_negg": [-0.19490432739257812, -0.6292647123336792, -0.78515625], "cos_Ac_negg": [-0.13538646697998047, -0.28259608149528503, -0.078125]}}
|