1
|
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.05, "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_rho0p05_residual_s1", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.327335834503174, "cos_r_negg": [0.011935683898627758, -0.013785665854811668, -0.005787058733403683], "cos_innovation_negg": [0.011935683898627758, -0.013785665854811668, -0.005787058733403683], "cos_apical_negg": [0.011578163132071495, -0.01364164613187313, -0.006125854793936014], "cos_Ac_negg": [0.0119356419891119, -0.01378566026687622, -0.005787072703242302], "r_norm": [4.739142417907715, 4.79836368560791, 4.788496017456055], "ldr": {"dL_sdil": -0.04240679740905762, "dL_bp": -0.2198801040649414, "ratio": 0.19286327696358013}}, {"step": 100, "epoch": 0, "train_loss": 0.19800172746181488, "cos_r_negg": [0.4595581889152527, 0.36300790309906006, 0.3212670087814331], "cos_innovation_negg": [0.4595581889152527, 0.36300790309906006, 0.3212670087814331], "cos_apical_negg": [0.20975911617279053, 0.16929280757904053, 0.19154170155525208], "cos_Ac_negg": [0.4597140848636627, 0.3632028102874756, 0.32125383615493774], "r_norm": [0.9384980201721191, 0.9418078064918518, 0.9590500593185425]}, {"step": 200, "epoch": 0, "train_loss": 0.3248291611671448, "cos_r_negg": [0.4977818429470062, 0.42747363448143005, 0.3676988482475281], "cos_innovation_negg": [0.4977818429470062, 0.42747363448143005, 0.3676988482475281], "cos_apical_negg": [0.20688921213150024, 0.18004465103149414, 0.19594287872314453], "cos_Ac_negg": [0.4978838562965393, 0.42765089869499207, 0.3677062690258026], "r_norm": [0.7502566576004028, 0.7576274871826172, 0.7664961814880371]}, {"step": 300, "epoch": 0, "train_loss": 0.17866581678390503, "cos_r_negg": [0.5077928304672241, 0.4503347873687744, 0.4130593538284302], "cos_innovation_negg": [0.5077928304672241, 0.4503347873687744, 0.4130593538284302], "cos_apical_negg": [0.166174978017807, 0.16370157897472382, 0.1931830197572708], "cos_Ac_negg": [0.5077493190765381, 0.45025303959846497, 0.4126030206680298], "r_norm": [0.582416296005249, 0.5849134922027588, 0.5949047803878784]}, {"step": 400, "epoch": 0, "train_loss": 0.10736608505249023, "cos_r_negg": [0.5154123306274414, 0.46785473823547363, 0.4282694458961487], "cos_innovation_negg": [0.5154123306274414, 0.46785473823547363, 0.4282694458961487], "cos_apical_negg": [0.1695859581232071, 0.16620805859565735, 0.1936672031879425], "cos_Ac_negg": [0.5154713988304138, 0.46782737970352173, 0.4281441271305084], "r_norm": [0.566364049911499, 0.5695450305938721, 0.5781142711639404]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9461, "test_loss": 0.1800264015197754}, {"step": 500, "epoch": 1, "train_loss": 0.16453297436237335, "cos_r_negg": [0.5244979858398438, 0.4845229387283325, 0.43385422229766846], "cos_innovation_negg": [0.5244979858398438, 0.4845229387283325, 0.43385422229766846], "cos_apical_negg": [0.13939255475997925, 0.14868509769439697, 0.19115450978279114], "cos_Ac_negg": [0.5251206159591675, 0.48533526062965393, 0.4335957467556], "r_norm": [0.48811590671539307, 0.4888215661048889, 0.4921705722808838], "ldr": {"dL_sdil": -0.011702954769134521, "dL_bp": -0.00984230637550354, "ratio": 1.1890459738442951}}, {"step": 600, "epoch": 1, "train_loss": 0.10629040747880936, "cos_r_negg": [0.5180792808532715, 0.48784932494163513, 0.4331909120082855], "cos_innovation_negg": [0.5180792808532715, 0.48784932494163513, 0.4331909120082855], "cos_apical_negg": [0.12916310131549835, 0.1492573320865631, 0.1820889413356781], "cos_Ac_negg": [0.5184558033943176, 0.48788216710090637, 0.43305861949920654], "r_norm": [0.44508522748947144, 0.4487663507461548, 0.45049262046813965]}, {"step": 700, "epoch": 1, "train_loss": 0.16264066100120544, "cos_r_negg": [0.527536928653717, 0.49619653820991516, 0.4591195583343506], "cos_innovation_negg": [0.527536928653717, 0.49619653820991516, 0.4591195583343506], "cos_apical_negg": [0.13695219159126282, 0.15048012137413025, 0.18260575830936432], "cos_Ac_negg": [0.5289961099624634, 0.4966246485710144, 0.4590521454811096], "r_norm": [0.4516822099685669, 0.4543275237083435, 0.4520670175552368]}, {"step": 800, "epoch": 1, "train_loss": 0.19131723046302795, "cos_r_negg": [0.5477697849273682, 0.5233035087585449, 0.46972203254699707], "cos_innovation_negg": [0.5477697849273682, 0.5233035087585449, 0.46972203254699707], "cos_apical_negg": [0.10759079456329346, 0.14349976181983948, 0.1710243821144104], "cos_Ac_negg": [0.5484983921051025, 0.5231080055236816, 0.46940040588378906], "r_norm": [0.35974693298339844, 0.36176374554634094, 0.3669247627258301]}, {"step": 900, "epoch": 1, "train_loss": 0.23413461446762085, "cos_r_negg": [0.5469974279403687, 0.5189642310142517, 0.4694017767906189], "cos_innovation_negg": [0.5469974279403687, 0.5189642310142517, 0.4694017767906189], "cos_apical_negg": [0.10924465954303741, 0.13199782371520996, 0.1644289791584015], "cos_Ac_negg": [0.5469458103179932, 0.518856406211853, 0.46931537985801697], "r_norm": [0.39200443029403687, 0.39184892177581787, 0.3952348828315735]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9594, "test_loss": 0.14740810661315917}, {"step": 1000, "epoch": 2, "train_loss": 0.020659351721405983, "cos_r_negg": [0.5646075010299683, 0.5425055027008057, 0.4808516502380371], "cos_innovation_negg": [0.5646075010299683, 0.5425055027008057, 0.4808516502380371], "cos_apical_negg": [0.1025833785533905, 0.12162604928016663, 0.15607091784477234], "cos_Ac_negg": [0.564576268196106, 0.5423530340194702, 0.4805763363838196], "r_norm": [0.35054507851600647, 0.3523258864879608, 0.35283398628234863], "ldr": {"dL_sdil": -0.010484561324119568, "dL_bp": -0.009114019572734833, "ratio": 1.1503773105211226}}, {"step": 1100, "epoch": 2, "train_loss": 0.10964615643024445, "cos_r_negg": [0.5504506230354309, 0.5405658483505249, 0.4947567582130432], "cos_innovation_negg": [0.5504506230354309, 0.5405658483505249, 0.4947567582130432], "cos_apical_negg": [0.09591667354106903, 0.1244964525103569, 0.16299867630004883], "cos_Ac_negg": [0.5513126254081726, 0.5401743650436401, 0.49388736486434937], "r_norm": [0.3291342258453369, 0.33226123452186584, 0.3298074007034302]}, {"step": 1200, "epoch": 2, "train_loss": 0.15681979060173035, "cos_r_negg": [0.5681092143058777, 0.5566937923431396, 0.5013537406921387], "cos_innovation_negg": [0.5681092143058777, 0.5566937923431396, 0.5013537406921387], "cos_apical_negg": [0.10015411674976349, 0.11984756588935852, 0.15649977326393127], "cos_Ac_negg": [0.5682443380355835, 0.5563015937805176, 0.5008739233016968], "r_norm": [0.31270861625671387, 0.3139200806617737, 0.3170466125011444]}, {"step": 1300, "epoch": 2, "train_loss": 0.13473723828792572, "cos_r_negg": [0.5784183144569397, 0.5624803304672241, 0.5229552984237671], "cos_innovation_negg": [0.5784183144569397, 0.5624803304672241, 0.5229552984237671], "cos_apical_negg": [0.09875530004501343, 0.12516838312149048, 0.16183607280254364], "cos_Ac_negg": [0.5786624550819397, 0.5625358819961548, 0.5228804349899292], "r_norm": [0.328178346157074, 0.3266424536705017, 0.32696443796157837]}, {"step": 1400, "epoch": 2, "train_loss": 0.09945511817932129, "cos_r_negg": [0.5877261757850647, 0.5833468437194824, 0.5220316648483276], "cos_innovation_negg": [0.5877261757850647, 0.5833468437194824, 0.5220316648483276], "cos_apical_negg": [0.08485932648181915, 0.10183429718017578, 0.1582496017217636], "cos_Ac_negg": [0.5899650454521179, 0.5846056342124939, 0.5218909978866577], "r_norm": [0.26476091146469116, 0.26450884342193604, 0.2684342861175537]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9662, "test_loss": 0.11500913486480713}, {"step": 1500, "epoch": 3, "train_loss": 0.044799480587244034, "cos_r_negg": [0.5729881525039673, 0.5794516801834106, 0.5339221954345703], "cos_innovation_negg": [0.5729881525039673, 0.5794516801834106, 0.5339221954345703], "cos_apical_negg": [0.05709003284573555, 0.10583391040563583, 0.1609860360622406], "cos_Ac_negg": [0.5750377178192139, 0.5808219909667969, 0.5336863994598389], "r_norm": [0.27066415548324585, 0.2718654274940491, 0.2756950855255127], "ldr": {"dL_sdil": -0.0055206939578056335, "dL_bp": -0.005840882658958435, "ratio": 0.9451814528987819}}, {"step": 1600, "epoch": 3, "train_loss": 0.09553006291389465, "cos_r_negg": [0.5864070653915405, 0.591616690158844, 0.5493444204330444], "cos_innovation_negg": [0.5864070653915405, 0.591616690158844, 0.5493444204330444], "cos_apical_negg": [0.06386640667915344, 0.09832250326871872, 0.15193292498588562], "cos_Ac_negg": [0.5889490842819214, 0.5921638607978821, 0.5484168529510498], "r_norm": [0.26048943400382996, 0.2610347867012024, 0.2635071277618408]}, {"step": 1700, "epoch": 3, "train_loss": 0.10861323773860931, "cos_r_negg": [0.5862084031105042, 0.587340235710144, 0.5640803575515747], "cos_innovation_negg": [0.5862084031105042, 0.587340235710144, 0.5640803575515747], "cos_apical_negg": [0.08238239586353302, 0.11745046079158783, 0.16670380532741547], "cos_Ac_negg": [0.5915169715881348, 0.590603232383728, 0.5643587708473206], "r_norm": [0.32203879952430725, 0.31957849860191345, 0.32221710681915283]}, {"step": 1800, "epoch": 3, "train_loss": 0.04907859116792679, "cos_r_negg": [0.5856608152389526, 0.5890840888023376, 0.5482282638549805], "cos_innovation_negg": [0.5856608152389526, 0.5890840888023376, 0.5482282638549805], "cos_apical_negg": [0.06749321520328522, 0.09056306630373001, 0.14532238245010376], "cos_Ac_negg": [0.5878607630729675, 0.5907537341117859, 0.5485868453979492], "r_norm": [0.2197471410036087, 0.2204505205154419, 0.221063032746315]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9672, "test_loss": 0.11288363456726075}, {"step": 1900, "epoch": 4, "train_loss": 0.02294781059026718, "cos_r_negg": [0.5973317623138428, 0.6043586730957031, 0.5593540668487549], "cos_innovation_negg": [0.5973317623138428, 0.6043586730957031, 0.5593540668487549], "cos_apical_negg": [0.0686958059668541, 0.1018339991569519, 0.14995601773262024], "cos_Ac_negg": [0.6019316911697388, 0.6075939536094666, 0.5601412653923035], "r_norm": [0.25126156210899353, 0.2498759925365448, 0.25224488973617554]}, {"step": 2000, "epoch": 4, "train_loss": 0.09122863411903381, "cos_r_negg": [0.5711127519607544, 0.5799358487129211, 0.5689223408699036], "cos_innovation_negg": [0.5711127519607544, 0.5799358487129211, 0.5689223408699036], "cos_apical_negg": [0.04912835359573364, 0.09740493446588516, 0.14483867585659027], "cos_Ac_negg": [0.5858960747718811, 0.5902670621871948, 0.5742981433868408], "r_norm": [0.23155272006988525, 0.22967955470085144, 0.23148378729820251], "ldr": {"dL_sdil": -0.005922161042690277, "dL_bp": -0.006407603621482849, "ratio": 0.9242396054017725}}, {"step": 2100, "epoch": 4, "train_loss": 0.0664985328912735, "cos_r_negg": [0.5840332508087158, 0.6021450757980347, 0.5844618082046509], "cos_innovation_negg": [0.5840332508087158, 0.6021450757980347, 0.5844618082046509], "cos_apical_negg": [0.07027436792850494, 0.10700127482414246, 0.14804022014141083], "cos_Ac_negg": [0.5876871943473816, 0.6038587093353271, 0.5843597650527954], "r_norm": [0.22726424038410187, 0.225623220205307, 0.22834229469299316]}, {"step": 2200, "epoch": 4, "train_loss": 0.023848218843340874, "cos_r_negg": [0.596025288105011, 0.6008343696594238, 0.5844902396202087], "cos_innovation_negg": [0.596025288105011, 0.6008343696594238, 0.5844902396202087], "cos_apical_negg": [0.04419754445552826, 0.08808250725269318, 0.13897162675857544], "cos_Ac_negg": [0.6045669317245483, 0.6077235341072083, 0.5862392783164978], "r_norm": [0.22590483725070953, 0.2255634367465973, 0.22662967443466187]}, {"step": 2300, "epoch": 4, "train_loss": 0.14877982437610626, "cos_r_negg": [0.5799214839935303, 0.5900747776031494, 0.5790226459503174], "cos_innovation_negg": [0.5799214839935303, 0.5900747776031494, 0.5790226459503174], "cos_apical_negg": [0.04342891275882721, 0.0754004716873169, 0.1416110098361969], "cos_Ac_negg": [0.5919167399406433, 0.6006958484649658, 0.5831726789474487], "r_norm": [0.22525566816329956, 0.22617968916893005, 0.22663938999176025]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9692, "test_loss": 0.11152303066253662}, {"step": 2400, "epoch": 5, "train_loss": 0.07470186054706573, "cos_r_negg": [0.5891973972320557, 0.6044924259185791, 0.587180495262146], "cos_innovation_negg": [0.5891973972320557, 0.6044924259185791, 0.587180495262146], "cos_apical_negg": [0.040368471294641495, 0.06973414123058319, 0.13383647799491882], "cos_Ac_negg": [0.6007899641990662, 0.6130261421203613, 0.5905767679214478], "r_norm": [0.2216203510761261, 0.22197623550891876, 0.222914919257164]}, {"step": 2500, "epoch": 5, "train_loss": 0.02789377048611641, "cos_r_negg": [0.5571001768112183, 0.5795426368713379, 0.5836772918701172], "cos_innovation_negg": [0.5571001768112183, 0.5795426368713379, 0.5836772918701172], "cos_apical_negg": [0.03303978219628334, 0.07389634102582932, 0.1359260380268097], "cos_Ac_negg": [0.5869741439819336, 0.6067754626274109, 0.5961843132972717], "r_norm": [0.18438225984573364, 0.1837213933467865, 0.18523038923740387], "ldr": {"dL_sdil": -0.005010098218917847, "dL_bp": -0.006476372480392456, "ratio": 0.7735963664977843}}, {"step": 2600, "epoch": 5, "train_loss": 0.06555721163749695, "cos_r_negg": [0.5787664651870728, 0.6029772758483887, 0.5920038223266602], "cos_innovation_negg": [0.5787664651870728, 0.6029772758483887, 0.5920038223266602], "cos_apical_negg": [0.054474882781505585, 0.09032820165157318, 0.13570192456245422], "cos_Ac_negg": [0.6060702800750732, 0.6261075735092163, 0.6057282090187073], "r_norm": [0.2206229269504547, 0.22138342261314392, 0.2254534661769867]}, {"step": 2700, "epoch": 5, "train_loss": 0.04593146592378616, "cos_r_negg": [0.6017680764198303, 0.6105853915214539, 0.5990234613418579], "cos_innovation_negg": [0.6017680764198303, 0.6105853915214539, 0.5990234613418579], "cos_apical_negg": [0.030534956604242325, 0.07209170609712601, 0.12952500581741333], "cos_Ac_negg": [0.6265010237693787, 0.6308919787406921, 0.6108779907226562], "r_norm": [0.17367111146450043, 0.17301763594150543, 0.1753917634487152]}, {"step": 2800, "epoch": 5, "train_loss": 0.11331872642040253, "cos_r_negg": [0.5852015018463135, 0.6050297021865845, 0.6058430671691895], "cos_innovation_negg": [0.5852015018463135, 0.6050297021865845, 0.6058430671691895], "cos_apical_negg": [0.037248946726322174, 0.08929359912872314, 0.1452261209487915], "cos_Ac_negg": [0.620559811592102, 0.6386481523513794, 0.6224511861801147], "r_norm": [0.1954781711101532, 0.19395135343074799, 0.19528909027576447]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9713, "test_loss": 0.10515918369293213}, {"step": 2900, "epoch": 6, "train_loss": 0.006533927284181118, "cos_r_negg": [0.5618231892585754, 0.5823687314987183, 0.5865188837051392], "cos_innovation_negg": [0.5618231892585754, 0.5823687314987183, 0.5865188837051392], "cos_apical_negg": [0.035457197576761246, 0.08563341200351715, 0.14092126488685608], "cos_Ac_negg": [0.6109321117401123, 0.6287125945091248, 0.6156300902366638], "r_norm": [0.1872854232788086, 0.18649417161941528, 0.18763405084609985]}, {"step": 3000, "epoch": 6, "train_loss": 0.08300692588090897, "cos_r_negg": [0.5744534730911255, 0.5794456005096436, 0.5861250162124634], "cos_innovation_negg": [0.5744534730911255, 0.5794456005096436, 0.5861250162124634], "cos_apical_negg": [0.04058567434549332, 0.0793231651186943, 0.13168630003929138], "cos_Ac_negg": [0.6225829720497131, 0.6238137483596802, 0.6147491931915283], "r_norm": [0.1884726881980896, 0.18706244230270386, 0.18900558352470398], "ldr": {"dL_sdil": -0.005829624831676483, "dL_bp": -0.00748685747385025, "ratio": 0.7786477640369043}}, {"step": 3100, "epoch": 6, "train_loss": 0.013627316802740097, "cos_r_negg": [0.5783639550209045, 0.584351658821106, 0.5909626483917236], "cos_innovation_negg": [0.5783639550209045, 0.584351658821106, 0.5909626483917236], "cos_apical_negg": [0.02789091132581234, 0.07112062722444534, 0.13366471230983734], "cos_Ac_negg": [0.6294291019439697, 0.6274665594100952, 0.6191610097885132], "r_norm": [0.14944246411323547, 0.15006566047668457, 0.15108169615268707]}, {"step": 3200, "epoch": 6, "train_loss": 0.04405490309000015, "cos_r_negg": [0.5540293455123901, 0.5780806541442871, 0.5798403024673462], "cos_innovation_negg": [0.5540293455123901, 0.5780806541442871, 0.5798403024673462], "cos_apical_negg": [0.024207837879657745, 0.06950204074382782, 0.1315123736858368], "cos_Ac_negg": [0.6184898018836975, 0.6359472274780273, 0.6188316345214844], "r_norm": [0.1712276190519333, 0.1713661104440689, 0.17324110865592957]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9721, "test_loss": 0.11621832408905029}, {"step": 3300, "epoch": 7, "train_loss": 0.013389818370342255, "cos_r_negg": [0.5641270279884338, 0.5790151953697205, 0.5924917459487915], "cos_innovation_negg": [0.5641270279884338, 0.5790151953697205, 0.5924917459487915], "cos_apical_negg": [0.034241147339344025, 0.07209589332342148, 0.1287578046321869], "cos_Ac_negg": [0.6335006952285767, 0.6423917412757874, 0.6323614716529846], "r_norm": [0.18159674108028412, 0.18028287589550018, 0.18132713437080383]}, {"step": 3400, "epoch": 7, "train_loss": 0.029773881658911705, "cos_r_negg": [0.5414331555366516, 0.555443286895752, 0.5770403742790222], "cos_innovation_negg": [0.5414331555366516, 0.555443286895752, 0.5770403742790222], "cos_apical_negg": [0.031215228140354156, 0.0754883885383606, 0.13154932856559753], "cos_Ac_negg": [0.6293191909790039, 0.6381579041481018, 0.6348534822463989], "r_norm": [0.19872957468032837, 0.1972137689590454, 0.19681642949581146]}, {"step": 3500, "epoch": 7, "train_loss": 0.0030391858890652657, "cos_r_negg": [0.5294433236122131, 0.5440030097961426, 0.572765588760376], "cos_innovation_negg": [0.5294433236122131, 0.5440030097961426, 0.572765588760376], "cos_apical_negg": [0.023815037682652473, 0.06500785052776337, 0.13215318322181702], "cos_Ac_negg": [0.6329994201660156, 0.6423251628875732, 0.6384730339050293], "r_norm": [0.15773600339889526, 0.15748322010040283, 0.15743637084960938], "ldr": {"dL_sdil": -0.005737520754337311, "dL_bp": -0.007938481867313385, "ratio": 0.7227478566099007}}, {"step": 3600, "epoch": 7, "train_loss": 0.00181392056401819, "cos_r_negg": [0.5173041224479675, 0.5402019023895264, 0.5612258911132812], "cos_innovation_negg": [0.5173041224479675, 0.5402019023895264, 0.5612258911132812], "cos_apical_negg": [0.034563735127449036, 0.06581186503171921, 0.13177812099456787], "cos_Ac_negg": [0.623420000076294, 0.6413934230804443, 0.6396375894546509], "r_norm": [0.2031160295009613, 0.20182238519191742, 0.2009257823228836]}, {"step": 3700, "epoch": 7, "train_loss": 0.005001378245651722, "cos_r_negg": [0.5316832661628723, 0.5508448481559753, 0.5710439682006836], "cos_innovation_negg": [0.5316832661628723, 0.5508448481559753, 0.5710439682006836], "cos_apical_negg": [0.032789912074804306, 0.06641100347042084, 0.12860509753227234], "cos_Ac_negg": [0.6304998993873596, 0.6423559188842773, 0.6453312039375305], "r_norm": [0.16545230150222778, 0.16468030214309692, 0.16423511505126953]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9754, "test_loss": 0.11100868663787841}, {"step": 3800, "epoch": 8, "train_loss": 0.0019634724594652653, "cos_r_negg": [0.49285173416137695, 0.522550106048584, 0.5501976013183594], "cos_innovation_negg": [0.49285173416137695, 0.522550106048584, 0.5501976013183594], "cos_apical_negg": [0.018828824162483215, 0.06898509711027145, 0.12255098670721054], "cos_Ac_negg": [0.6295504570007324, 0.643524169921875, 0.6488214731216431], "r_norm": [0.16043861210346222, 0.15994001924991608, 0.16116008162498474]}, {"step": 3900, "epoch": 8, "train_loss": 0.009662585332989693, "cos_r_negg": [0.4753582775592804, 0.5029117465019226, 0.5357325673103333], "cos_innovation_negg": [0.4753582775592804, 0.5029117465019226, 0.5357325673103333], "cos_apical_negg": [0.017303794622421265, 0.061728790402412415, 0.12165973335504532], "cos_Ac_negg": [0.6283164024353027, 0.6399952173233032, 0.6502859592437744], "r_norm": [0.16895166039466858, 0.167774498462677, 0.16917681694030762]}, {"step": 4000, "epoch": 8, "train_loss": 0.012053707614541054, "cos_r_negg": [0.4598022401332855, 0.48842787742614746, 0.5198114514350891], "cos_innovation_negg": [0.4598022401332855, 0.48842787742614746, 0.5198114514350891], "cos_apical_negg": [0.022250913083553314, 0.06144747510552406, 0.12196500599384308], "cos_Ac_negg": [0.6305398941040039, 0.6407707333564758, 0.6521885991096497], "r_norm": [0.16535891592502594, 0.16453412175178528, 0.16724371910095215], "ldr": {"dL_sdil": -0.007221609354019165, "dL_bp": -0.011598825454711914, "ratio": 0.6226155727764189}}, {"step": 4100, "epoch": 8, "train_loss": 0.007018683943897486, "cos_r_negg": [0.46128901839256287, 0.4965739846229553, 0.5250366926193237], "cos_innovation_negg": [0.46128901839256287, 0.4965739846229553, 0.5250366926193237], "cos_apical_negg": [0.018589844927191734, 0.0649154856801033, 0.11730692535638809], "cos_Ac_negg": [0.632196307182312, 0.6469199657440186, 0.6585046648979187], "r_norm": [0.1577710211277008, 0.15650255978107452, 0.15849930047988892]}, {"step": 4200, "epoch": 8, "train_loss": 0.02269434742629528, "cos_r_negg": [0.47296077013015747, 0.5012197494506836, 0.5313794016838074], "cos_innovation_negg": [0.47296077013015747, 0.5012197494506836, 0.5313794016838074], "cos_apical_negg": [0.010709510184824467, 0.056064411997795105, 0.11694499850273132], "cos_Ac_negg": [0.6360347270965576, 0.6495529413223267, 0.6575896143913269], "r_norm": [0.1453334391117096, 0.1442568004131317, 0.14695820212364197]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.973, "test_loss": 0.12612433528900147}, {"step": 4300, "epoch": 9, "train_loss": 0.0055565061047673225, "cos_r_negg": [0.4508157968521118, 0.4840299189090729, 0.5116130709648132], "cos_innovation_negg": [0.4508157968521118, 0.4840299189090729, 0.5116130709648132], "cos_apical_negg": [0.018956102430820465, 0.06528937071561813, 0.11622902750968933], "cos_Ac_negg": [0.6355956196784973, 0.6468340158462524, 0.6545963287353516], "r_norm": [0.15752550959587097, 0.15622948110103607, 0.158013254404068]}, {"step": 4400, "epoch": 9, "train_loss": 0.0009345239959657192, "cos_r_negg": [0.43100717663764954, 0.4660802185535431, 0.4992997944355011], "cos_innovation_negg": [0.43100717663764954, 0.4660802185535431, 0.4992997944355011], "cos_apical_negg": [0.0012484766775742173, 0.05552934855222702, 0.11143714934587479], "cos_Ac_negg": [0.6282694339752197, 0.6399216651916504, 0.6582869291305542], "r_norm": [0.1593485325574875, 0.15823084115982056, 0.16016815602779388]}, {"step": 4500, "epoch": 9, "train_loss": 0.03349297121167183, "cos_r_negg": [0.4123862683773041, 0.45048922300338745, 0.48813825845718384], "cos_innovation_negg": [0.4123862683773041, 0.45048922300338745, 0.48813825845718384], "cos_apical_negg": [0.0073779248632490635, 0.055708736181259155, 0.11663167178630829], "cos_Ac_negg": [0.6335850358009338, 0.6434621810913086, 0.66236412525177], "r_norm": [0.15378838777542114, 0.15259474515914917, 0.15409553050994873], "ldr": {"dL_sdil": -0.007371291518211365, "dL_bp": -0.013118952512741089, "ratio": 0.5618811037735206}}, {"step": 4600, "epoch": 9, "train_loss": 0.001479916274547577, "cos_r_negg": [0.4173874855041504, 0.4497365951538086, 0.4874991774559021], "cos_innovation_negg": [0.4173874855041504, 0.4497365951538086, 0.4874991774559021], "cos_apical_negg": [0.012430883944034576, 0.056131042540073395, 0.11374486982822418], "cos_Ac_negg": [0.6330933570861816, 0.6427613496780396, 0.6594250202178955], "r_norm": [0.1550893485546112, 0.1537916362285614, 0.1546764075756073]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9751, "test_loss": 0.12642392368316652}, {"step": 4700, "epoch": 10, "train_loss": 0.00215778942219913, "cos_r_negg": [0.39477741718292236, 0.43047553300857544, 0.47205987572669983], "cos_innovation_negg": [0.39477741718292236, 0.43047553300857544, 0.47205987572669983], "cos_apical_negg": [0.01019405759871006, 0.05123322829604149, 0.10903597623109818], "cos_Ac_negg": [0.6249011158943176, 0.6408618688583374, 0.6627616882324219], "r_norm": [0.15229037404060364, 0.15106777846813202, 0.15219895541667938]}, {"step": 4800, "epoch": 10, "train_loss": 0.002831760561093688, "cos_r_negg": [0.3821009397506714, 0.42313966155052185, 0.46219125390052795], "cos_innovation_negg": [0.3821009397506714, 0.42313966155052185, 0.46219125390052795], "cos_apical_negg": [0.003800086211413145, 0.052510738372802734, 0.10751186311244965], "cos_Ac_negg": [0.629231333732605, 0.6427340507507324, 0.6645975112915039], "r_norm": [0.14961805939674377, 0.14873483777046204, 0.14977099001407623]}, {"step": 4900, "epoch": 10, "train_loss": 0.001507880399003625, "cos_r_negg": [0.3845168352127075, 0.4220179319381714, 0.46184539794921875], "cos_innovation_negg": [0.3845168352127075, 0.4220179319381714, 0.46184539794921875], "cos_apical_negg": [0.006418270990252495, 0.04843633249402046, 0.10634401440620422], "cos_Ac_negg": [0.6274188756942749, 0.6420944929122925, 0.6620548963546753], "r_norm": [0.1594342142343521, 0.15761792659759521, 0.15847983956336975]}, {"step": 5000, "epoch": 10, "train_loss": 0.0006326170405372977, "cos_r_negg": [0.365106999874115, 0.40706682205200195, 0.4476078450679779], "cos_innovation_negg": [0.365106999874115, 0.40706682205200195, 0.4476078450679779], "cos_apical_negg": [0.0019759703427553177, 0.05384797602891922, 0.10979677736759186], "cos_Ac_negg": [0.6259748339653015, 0.6441224813461304, 0.6630863547325134], "r_norm": [0.15052375197410583, 0.149494469165802, 0.15060998499393463], "ldr": {"dL_sdil": -0.007718577980995178, "dL_bp": -0.014518767595291138, "ratio": 0.5316276281947332}}, {"step": 5100, "epoch": 10, "train_loss": 0.0012461576843634248, "cos_r_negg": [0.35911571979522705, 0.3983779549598694, 0.4413723349571228], "cos_innovation_negg": [0.35911571979522705, 0.3983779549598694, 0.4413723349571228], "cos_apical_negg": [0.002363624982535839, 0.04849231243133545, 0.10589759051799774], "cos_Ac_negg": [0.6248847842216492, 0.6437404751777649, 0.6656257510185242], "r_norm": [0.15122172236442566, 0.1501307189464569, 0.15147460997104645]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9759, "test_loss": 0.1276949785232544}, {"step": 5200, "epoch": 11, "train_loss": 0.0012402110733091831, "cos_r_negg": [0.35025155544281006, 0.3906233012676239, 0.4298197329044342], "cos_innovation_negg": [0.35025155544281006, 0.3906233012676239, 0.4298197329044342], "cos_apical_negg": [0.00504857674241066, 0.052778299897909164, 0.10604824125766754], "cos_Ac_negg": [0.62543785572052, 0.6403063535690308, 0.6640012264251709], "r_norm": [0.14704205095767975, 0.1458543986082077, 0.14721126854419708]}, {"step": 5300, "epoch": 11, "train_loss": 0.0012527708895504475, "cos_r_negg": [0.3447659909725189, 0.3828068673610687, 0.42176806926727295], "cos_innovation_negg": [0.3447659909725189, 0.3828068673610687, 0.42176806926727295], "cos_apical_negg": [0.0029204804450273514, 0.04953660070896149, 0.10046202689409256], "cos_Ac_negg": [0.6246902942657471, 0.6398322582244873, 0.663558840751648], "r_norm": [0.1457543671131134, 0.1447260081768036, 0.1462690830230713]}, {"step": 5400, "epoch": 11, "train_loss": 0.0026527205482125282, "cos_r_negg": [0.3370477557182312, 0.3748803436756134, 0.41634199023246765], "cos_innovation_negg": [0.3370477557182312, 0.3748803436756134, 0.41634199023246765], "cos_apical_negg": [0.0023171482607722282, 0.04821858927607536, 0.09994305670261383], "cos_Ac_negg": [0.622248649597168, 0.6417462825775146, 0.6651083827018738], "r_norm": [0.14654432237148285, 0.14567247033119202, 0.14684513211250305]}, {"step": 5500, "epoch": 11, "train_loss": 0.00147485570050776, "cos_r_negg": [0.3404698073863983, 0.3794620633125305, 0.4205111861228943], "cos_innovation_negg": [0.3404698073863983, 0.3794620633125305, 0.4205111861228943], "cos_apical_negg": [0.004639003425836563, 0.048866137862205505, 0.09987516701221466], "cos_Ac_negg": [0.6238670349121094, 0.6419881582260132, 0.6679314970970154], "r_norm": [0.15389621257781982, 0.15288250148296356, 0.1536381095647812], "ldr": {"dL_sdil": -0.008258819580078125, "dL_bp": -0.01560879498720169, "ratio": 0.5291132074481009}}, {"step": 5600, "epoch": 11, "train_loss": 0.0010807700455188751, "cos_r_negg": [0.32366716861724854, 0.3646085858345032, 0.40230053663253784], "cos_innovation_negg": [0.32366716861724854, 0.3646085858345032, 0.40230053663253784], "cos_apical_negg": [0.003534013871103525, 0.04884061589837074, 0.09961765259504318], "cos_Ac_negg": [0.615607500076294, 0.6380172967910767, 0.6647375226020813], "r_norm": [0.1503227800130844, 0.14935144782066345, 0.150526762008667]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9754, "test_loss": 0.13247075786590576}, {"step": 5700, "epoch": 12, "train_loss": 0.0005582757294178009, "cos_r_negg": [0.32820063829421997, 0.36593544483184814, 0.4054841995239258], "cos_innovation_negg": [0.32820063829421997, 0.36593544483184814, 0.4054841995239258], "cos_apical_negg": [0.0037302712444216013, 0.045040614902973175, 0.09691508114337921], "cos_Ac_negg": [0.6218763589859009, 0.6408955454826355, 0.6676142811775208], "r_norm": [0.1485031694173813, 0.147519052028656, 0.14880996942520142]}, {"step": 5800, "epoch": 12, "train_loss": 0.00036639836616814137, "cos_r_negg": [0.31826362013816833, 0.3544495701789856, 0.3930700719356537], "cos_innovation_negg": [0.31826362013816833, 0.3544495701789856, 0.3930700719356537], "cos_apical_negg": [0.003201207611709833, 0.04491714760661125, 0.09604141116142273], "cos_Ac_negg": [0.616691529750824, 0.6374433040618896, 0.6658985018730164], "r_norm": [0.14986775815486908, 0.14872559905052185, 0.14994212985038757]}, {"step": 5900, "epoch": 12, "train_loss": 0.0026673325337469578, "cos_r_negg": [0.3165438175201416, 0.3531757891178131, 0.3893282413482666], "cos_innovation_negg": [0.3165438175201416, 0.3531757891178131, 0.3893282413482666], "cos_apical_negg": [0.003569494467228651, 0.045752979815006256, 0.0928518995642662], "cos_Ac_negg": [0.6143794059753418, 0.6380530595779419, 0.6660316586494446], "r_norm": [0.14996939897537231, 0.1488775908946991, 0.15008890628814697]}, {"step": 6000, "epoch": 12, "train_loss": 0.00014879496302455664, "cos_r_negg": [0.30961406230926514, 0.3466123640537262, 0.3855319023132324], "cos_innovation_negg": [0.30961406230926514, 0.3466123640537262, 0.3855319023132324], "cos_apical_negg": [0.0015506725758314133, 0.044106483459472656, 0.09334377199411392], "cos_Ac_negg": [0.6153971552848816, 0.6390558481216431, 0.6667139530181885], "r_norm": [0.1464368999004364, 0.14536206424236298, 0.14683833718299866], "ldr": {"dL_sdil": -0.008332058787345886, "dL_bp": -0.016469426453113556, "ratio": 0.5059106831112935}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9753, "test_loss": 0.13651623115539552}, {"step": 6100, "epoch": 13, "train_loss": 0.00037324384902603924, "cos_r_negg": [0.30239158868789673, 0.3397728204727173, 0.37674129009246826], "cos_innovation_negg": [0.30239158868789673, 0.3397728204727173, 0.37674129009246826], "cos_apical_negg": [0.004823541268706322, 0.04640675708651543, 0.09096071124076843], "cos_Ac_negg": [0.6124904751777649, 0.6365949511528015, 0.6662158966064453], "r_norm": [0.14694678783416748, 0.14580979943275452, 0.1474391520023346]}, {"step": 6200, "epoch": 13, "train_loss": 0.0011788607807829976, "cos_r_negg": [0.2936674952507019, 0.3329145908355713, 0.37022751569747925], "cos_innovation_negg": [0.2936674952507019, 0.3329145908355713, 0.37022751569747925], "cos_apical_negg": [0.003545030951499939, 0.04557621479034424, 0.09047915041446686], "cos_Ac_negg": [0.6121279001235962, 0.6354484558105469, 0.6650893092155457], "r_norm": [0.14955121278762817, 0.14838676154613495, 0.15036849677562714]}, {"step": 6300, "epoch": 13, "train_loss": 0.0005334530142135918, "cos_r_negg": [0.2936256527900696, 0.332280695438385, 0.36860334873199463], "cos_innovation_negg": [0.2936256527900696, 0.332280695438385, 0.36860334873199463], "cos_apical_negg": [0.002949201036244631, 0.04521677643060684, 0.08821894973516464], "cos_Ac_negg": [0.6121575236320496, 0.635464072227478, 0.666549563407898], "r_norm": [0.14792223274707794, 0.14679907262325287, 0.14844593405723572]}, {"step": 6400, "epoch": 13, "train_loss": 0.0006172251305542886, "cos_r_negg": [0.28886568546295166, 0.32823634147644043, 0.3629181385040283], "cos_innovation_negg": [0.28886568546295166, 0.32823634147644043, 0.3629181385040283], "cos_apical_negg": [0.001749833463691175, 0.043821848928928375, 0.0855003371834755], "cos_Ac_negg": [0.6091131567955017, 0.6342813372612, 0.6659475564956665], "r_norm": [0.15152990818023682, 0.15035876631736755, 0.1519782394170761]}, {"step": 6500, "epoch": 13, "train_loss": 0.00011808508133981377, "cos_r_negg": [0.284700870513916, 0.3244197964668274, 0.36029553413391113], "cos_innovation_negg": [0.284700870513916, 0.3244197964668274, 0.36029553413391113], "cos_apical_negg": [0.002069351729005575, 0.04354419559240341, 0.08564615249633789], "cos_Ac_negg": [0.6112993359565735, 0.6342726349830627, 0.6662527322769165], "r_norm": [0.14418646693229675, 0.143080472946167, 0.14480093121528625], "ldr": {"dL_sdil": -0.00850454717874527, "dL_bp": -0.01731761544942856, "ratio": 0.49109227558381313}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9752, "test_loss": 0.13994757270812988}, {"step": 6600, "epoch": 14, "train_loss": 0.0002561457804404199, "cos_r_negg": [0.28083300590515137, 0.3208739459514618, 0.3552173972129822], "cos_innovation_negg": [0.28083300590515137, 0.3208739459514618, 0.3552173972129822], "cos_apical_negg": [0.0004630244802683592, 0.04371214285492897, 0.08422216773033142], "cos_Ac_negg": [0.6097625494003296, 0.6335102319717407, 0.6659466028213501], "r_norm": [0.14630337059497833, 0.14519445598125458, 0.14684632420539856]}, {"step": 6700, "epoch": 14, "train_loss": 0.0006219938513822854, "cos_r_negg": [0.2804514765739441, 0.3197934031486511, 0.354887455701828], "cos_innovation_negg": [0.2804514765739441, 0.3197934031486511, 0.354887455701828], "cos_apical_negg": [0.0014164652675390244, 0.04353693500161171, 0.08363394439220428], "cos_Ac_negg": [0.6096209287643433, 0.6323015689849854, 0.6662194132804871], "r_norm": [0.1471884846687317, 0.1460249125957489, 0.1477433294057846]}, {"step": 6800, "epoch": 14, "train_loss": 0.001066845841705799, "cos_r_negg": [0.26744145154953003, 0.30664917826652527, 0.3402431309223175], "cos_innovation_negg": [0.26744145154953003, 0.30664917826652527, 0.3402431309223175], "cos_apical_negg": [1.106318086385727e-05, 0.04225609451532364, 0.08050717413425446], "cos_Ac_negg": [0.6058058738708496, 0.6291998028755188, 0.663305401802063], "r_norm": [0.14706626534461975, 0.14594686031341553, 0.14794312417507172]}, {"step": 6900, "epoch": 14, "train_loss": 0.0008502452983520925, "cos_r_negg": [0.26708921790122986, 0.30680036544799805, 0.34202855825424194], "cos_innovation_negg": [0.26708921790122986, 0.30680036544799805, 0.34202855825424194], "cos_apical_negg": [0.00013900105841457844, 0.04170065373182297, 0.08219294995069504], "cos_Ac_negg": [0.6062815189361572, 0.629963755607605, 0.6647016406059265], "r_norm": [0.14870262145996094, 0.147597074508667, 0.149154931306839]}, {"step": 7000, "epoch": 14, "train_loss": 0.0005537670804187655, "cos_r_negg": [0.26967987418174744, 0.30989396572113037, 0.3444839119911194], "cos_innovation_negg": [0.26967987418174744, 0.30989396572113037, 0.3444839119911194], "cos_apical_negg": [-0.0012382407439872622, 0.04133274778723717, 0.08160696923732758], "cos_Ac_negg": [0.6068159341812134, 0.6314824819564819, 0.6662921905517578], "r_norm": [0.1491071879863739, 0.14798906445503235, 0.14975789189338684], "ldr": {"dL_sdil": -0.009437493979930878, "dL_bp": -0.019233211874961853, "ratio": 0.4906873610754936}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9748, "test_loss": 0.14247735996246338}], "final": {"test_acc": 0.9748, "test_loss": 0.14247735996246338, "wall_s": 25.2902090549469, "cos_r_negg": [0.26692265272140503, 0.30770188570022583, 0.3412211537361145], "cos_innovation_negg": [0.26692265272140503, 0.30770188570022583, 0.3412211537361145], "cos_apical_negg": [-0.0005659900489263237, 0.04179733619093895, 0.08092036843299866], "cos_Ac_negg": [0.6041395664215088, 0.6299995183944702, 0.6659080386161804]}}
|