{"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": 2, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p05_residual_s2", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4534778594970703, "cos_r_negg": [0.017675090581178665, 0.0372890904545784, 0.0025144016835838556], "cos_innovation_negg": [0.017675090581178665, 0.0372890904545784, 0.0025144016835838556], "cos_apical_negg": [0.01725875586271286, 0.0366811528801918, 0.002047688001766801], "cos_Ac_negg": [0.017675412818789482, 0.03728915750980377, 0.0025144696701318026], "r_norm": [4.74981689453125, 4.774450302124023, 4.856009483337402], "ldr": {"dL_sdil": -0.041077613830566406, "dL_bp": -0.24913907051086426, "ratio": 0.16487824951074917}}, {"step": 100, "epoch": 0, "train_loss": 0.17586416006088257, "cos_r_negg": [0.4970792233943939, 0.40330982208251953, 0.36845076084136963], "cos_innovation_negg": [0.4970792233943939, 0.40330982208251953, 0.36845076084136963], "cos_apical_negg": [0.19017258286476135, 0.17915573716163635, 0.21979236602783203], "cos_Ac_negg": [0.4971027076244354, 0.4034501314163208, 0.3684612214565277], "r_norm": [0.8698434829711914, 0.8818600177764893, 0.8946853876113892]}, {"step": 200, "epoch": 0, "train_loss": 0.3208368420600891, "cos_r_negg": [0.4926418364048004, 0.4460785388946533, 0.40469175577163696], "cos_innovation_negg": [0.4926418364048004, 0.4460785388946533, 0.40469175577163696], "cos_apical_negg": [0.19864384829998016, 0.190121129155159, 0.2166076898574829], "cos_Ac_negg": [0.492310106754303, 0.44555914402008057, 0.4045509696006775], "r_norm": [0.7164028882980347, 0.7233113050460815, 0.7400282025337219]}, {"step": 300, "epoch": 0, "train_loss": 0.19444385170936584, "cos_r_negg": [0.5353859663009644, 0.46626943349838257, 0.4177759885787964], "cos_innovation_negg": [0.5353859663009644, 0.46626943349838257, 0.4177759885787964], "cos_apical_negg": [0.16973918676376343, 0.1737947165966034, 0.2042451798915863], "cos_Ac_negg": [0.5351248383522034, 0.4660660922527313, 0.4175689220428467], "r_norm": [0.503006637096405, 0.5082999467849731, 0.5159482359886169]}, {"step": 400, "epoch": 0, "train_loss": 0.14225107431411743, "cos_r_negg": [0.5365469455718994, 0.5072477459907532, 0.4304060935974121], "cos_innovation_negg": [0.5365469455718994, 0.5072477459907532, 0.4304060935974121], "cos_apical_negg": [0.15741007030010223, 0.18071222305297852, 0.1913570910692215], "cos_Ac_negg": [0.5363162755966187, 0.5070614814758301, 0.4302181899547577], "r_norm": [0.5037429928779602, 0.5096166133880615, 0.5165624618530273]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9481, "test_loss": 0.17114948806762695}, {"step": 500, "epoch": 1, "train_loss": 0.20640233159065247, "cos_r_negg": [0.5260323286056519, 0.508039116859436, 0.46667802333831787], "cos_innovation_negg": [0.5260323286056519, 0.508039116859436, 0.46667802333831787], "cos_apical_negg": [0.14044275879859924, 0.1553192436695099, 0.19427356123924255], "cos_Ac_negg": [0.525843620300293, 0.5080826282501221, 0.46643710136413574], "r_norm": [0.473062664270401, 0.4806978702545166, 0.48895666003227234], "ldr": {"dL_sdil": -0.012392014265060425, "dL_bp": -0.010644689202308655, "ratio": 1.1641499370759274}}, {"step": 600, "epoch": 1, "train_loss": 0.10871553421020508, "cos_r_negg": [0.5452223420143127, 0.523560643196106, 0.45618343353271484], "cos_innovation_negg": [0.5452223420143127, 0.523560643196106, 0.45618343353271484], "cos_apical_negg": [0.1570211797952652, 0.15463560819625854, 0.17270517349243164], "cos_Ac_negg": [0.5453288555145264, 0.5229302644729614, 0.4559338092803955], "r_norm": [0.3991668224334717, 0.40303510427474976, 0.4114168882369995]}, {"step": 700, "epoch": 1, "train_loss": 0.17709866166114807, "cos_r_negg": [0.5708881616592407, 0.5400104522705078, 0.47234269976615906], "cos_innovation_negg": [0.5708881616592407, 0.5400104522705078, 0.47234269976615906], "cos_apical_negg": [0.14205503463745117, 0.15635958313941956, 0.18742695450782776], "cos_Ac_negg": [0.570570707321167, 0.539499044418335, 0.4720923900604248], "r_norm": [0.43761348724365234, 0.44108182191848755, 0.45164451003074646]}, {"step": 800, "epoch": 1, "train_loss": 0.15103770792484283, "cos_r_negg": [0.5683736801147461, 0.5581629276275635, 0.4943467378616333], "cos_innovation_negg": [0.5683736801147461, 0.5581629276275635, 0.4943467378616333], "cos_apical_negg": [0.13331329822540283, 0.15281610190868378, 0.1842871606349945], "cos_Ac_negg": [0.568236231803894, 0.5576614141464233, 0.49418404698371887], "r_norm": [0.33557310700416565, 0.3389970064163208, 0.34779369831085205]}, {"step": 900, "epoch": 1, "train_loss": 0.2185644805431366, "cos_r_negg": [0.5811727046966553, 0.5617357492446899, 0.48729169368743896], "cos_innovation_negg": [0.5811727046966553, 0.5617357492446899, 0.48729169368743896], "cos_apical_negg": [0.13030628859996796, 0.1538248360157013, 0.16992904245853424], "cos_Ac_negg": [0.5809875726699829, 0.5614639520645142, 0.4871039390563965], "r_norm": [0.38768357038497925, 0.39016199111938477, 0.39808550477027893]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9576, "test_loss": 0.14079903106689454}, {"step": 1000, "epoch": 2, "train_loss": 0.027318187057971954, "cos_r_negg": [0.5814144015312195, 0.5783625841140747, 0.4958611726760864], "cos_innovation_negg": [0.5814144015312195, 0.5783625841140747, 0.4958611726760864], "cos_apical_negg": [0.1055048257112503, 0.1402195692062378, 0.164465069770813], "cos_Ac_negg": [0.5812327265739441, 0.5781751275062561, 0.4956243634223938], "r_norm": [0.312904417514801, 0.31333309412002563, 0.32295629382133484], "ldr": {"dL_sdil": -0.0076730698347091675, "dL_bp": -0.006511993706226349, "ratio": 1.1782981036011555}}, {"step": 1100, "epoch": 2, "train_loss": 0.12655237317085266, "cos_r_negg": [0.5872601270675659, 0.5777523517608643, 0.5042988657951355], "cos_innovation_negg": [0.5872601270675659, 0.5777523517608643, 0.5042988657951355], "cos_apical_negg": [0.10918585956096649, 0.15159432590007782, 0.17792680859565735], "cos_Ac_negg": [0.588329553604126, 0.5776081681251526, 0.503897488117218], "r_norm": [0.35866644978523254, 0.3616602420806885, 0.37040048837661743]}, {"step": 1200, "epoch": 2, "train_loss": 0.19169515371322632, "cos_r_negg": [0.5778206586837769, 0.5714322328567505, 0.5065737366676331], "cos_innovation_negg": [0.5778206586837769, 0.5714322328567505, 0.5065737366676331], "cos_apical_negg": [0.103522390127182, 0.13620984554290771, 0.17392313480377197], "cos_Ac_negg": [0.5777586698532104, 0.571283757686615, 0.5059672594070435], "r_norm": [0.3244379758834839, 0.3290201425552368, 0.33428263664245605]}, {"step": 1300, "epoch": 2, "train_loss": 0.135965034365654, "cos_r_negg": [0.5901323556900024, 0.592176079750061, 0.5073062181472778], "cos_innovation_negg": [0.5901323556900024, 0.592176079750061, 0.5073062181472778], "cos_apical_negg": [0.09399458765983582, 0.12740005552768707, 0.1656610667705536], "cos_Ac_negg": [0.5904358625411987, 0.5921123027801514, 0.5068682432174683], "r_norm": [0.3208772540092468, 0.3238424062728882, 0.32930833101272583]}, {"step": 1400, "epoch": 2, "train_loss": 0.1040671095252037, "cos_r_negg": [0.5833787322044373, 0.5839993953704834, 0.5164324045181274], "cos_innovation_negg": [0.5833787322044373, 0.5839993953704834, 0.5164324045181274], "cos_apical_negg": [0.08873105049133301, 0.10539083927869797, 0.15748608112335205], "cos_Ac_negg": [0.5840877294540405, 0.584343433380127, 0.5154862403869629], "r_norm": [0.26289546489715576, 0.2639598548412323, 0.2716895341873169]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9691, "test_loss": 0.10747614059448242}, {"step": 1500, "epoch": 3, "train_loss": 0.062176160514354706, "cos_r_negg": [0.5943048000335693, 0.5952531099319458, 0.5310481786727905], "cos_innovation_negg": [0.5943048000335693, 0.5952531099319458, 0.5310481786727905], "cos_apical_negg": [0.08299660682678223, 0.10385184735059738, 0.15107882022857666], "cos_Ac_negg": [0.5947880148887634, 0.5951738357543945, 0.5299400687217712], "r_norm": [0.27155426144599915, 0.2732366621494293, 0.27872419357299805], "ldr": {"dL_sdil": -0.007850125432014465, "dL_bp": -0.007701881229877472, "ratio": 1.0192477912489637}}, {"step": 1600, "epoch": 3, "train_loss": 0.04608044773340225, "cos_r_negg": [0.596003532409668, 0.6089153289794922, 0.5469405055046082], "cos_innovation_negg": [0.596003532409668, 0.6089153289794922, 0.5469405055046082], "cos_apical_negg": [0.08411060273647308, 0.10125107318162918, 0.15174731612205505], "cos_Ac_negg": [0.5981305241584778, 0.6094847917556763, 0.5464131832122803], "r_norm": [0.2744586169719696, 0.27718842029571533, 0.28355464339256287]}, {"step": 1700, "epoch": 3, "train_loss": 0.042360708117485046, "cos_r_negg": [0.599576473236084, 0.6094846129417419, 0.5494703054428101], "cos_innovation_negg": [0.599576473236084, 0.6094846129417419, 0.5494703054428101], "cos_apical_negg": [0.08860689401626587, 0.10659298300743103, 0.15702548623085022], "cos_Ac_negg": [0.6054372191429138, 0.6122726798057556, 0.5488888025283813], "r_norm": [0.29002389311790466, 0.2930294871330261, 0.30011770129203796]}, {"step": 1800, "epoch": 3, "train_loss": 0.04638168588280678, "cos_r_negg": [0.5836105942726135, 0.6037336587905884, 0.5602104663848877], "cos_innovation_negg": [0.5836105942726135, 0.6037336587905884, 0.5602104663848877], "cos_apical_negg": [0.048589885234832764, 0.09653361141681671, 0.15400061011314392], "cos_Ac_negg": [0.5913264155387878, 0.6077687740325928, 0.5613003969192505], "r_norm": [0.23215599358081818, 0.23303121328353882, 0.2408786416053772]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9667, "test_loss": 0.11098653583526612}, {"step": 1900, "epoch": 4, "train_loss": 0.030510950833559036, "cos_r_negg": [0.5977516174316406, 0.6129684448242188, 0.5715459585189819], "cos_innovation_negg": [0.5977516174316406, 0.6129684448242188, 0.5715459585189819], "cos_apical_negg": [0.07869121432304382, 0.11038168519735336, 0.15888436138629913], "cos_Ac_negg": [0.6013244390487671, 0.6158010959625244, 0.5713731646537781], "r_norm": [0.2544582784175873, 0.2550382614135742, 0.2616143226623535]}, {"step": 2000, "epoch": 4, "train_loss": 0.06422892212867737, "cos_r_negg": [0.5751674175262451, 0.6050205230712891, 0.5632307529449463], "cos_innovation_negg": [0.5751674175262451, 0.6050205230712891, 0.5632307529449463], "cos_apical_negg": [0.039523810148239136, 0.08830837160348892, 0.13610044121742249], "cos_Ac_negg": [0.5866235494613647, 0.6144043207168579, 0.5659879446029663], "r_norm": [0.20273345708847046, 0.2024599313735962, 0.20888051390647888], "ldr": {"dL_sdil": -0.004440866410732269, "dL_bp": -0.004471570253372192, "ratio": 0.9931335434981106}}, {"step": 2100, "epoch": 4, "train_loss": 0.02168159931898117, "cos_r_negg": [0.5857771635055542, 0.6053357124328613, 0.5740090608596802], "cos_innovation_negg": [0.5857771635055542, 0.6053357124328613, 0.5740090608596802], "cos_apical_negg": [0.05207394063472748, 0.09032325446605682, 0.14031092822551727], "cos_Ac_negg": [0.5964084267616272, 0.6146030426025391, 0.5765020847320557], "r_norm": [0.2210220992565155, 0.2213449776172638, 0.22735489904880524]}, {"step": 2200, "epoch": 4, "train_loss": 0.009620489552617073, "cos_r_negg": [0.5972596406936646, 0.6263022422790527, 0.5853859186172485], "cos_innovation_negg": [0.5972596406936646, 0.6263022422790527, 0.5853859186172485], "cos_apical_negg": [0.061493098735809326, 0.10348585247993469, 0.1426779329776764], "cos_Ac_negg": [0.6023486852645874, 0.6299389600753784, 0.5839298963546753], "r_norm": [0.26552489399909973, 0.26564037799835205, 0.27185261249542236]}, {"step": 2300, "epoch": 4, "train_loss": 0.17019157111644745, "cos_r_negg": [0.5706629753112793, 0.6006418466567993, 0.5666520595550537], "cos_innovation_negg": [0.5706629753112793, 0.6006418466567993, 0.5666520595550537], "cos_apical_negg": [0.06424862891435623, 0.09687228500843048, 0.14904698729515076], "cos_Ac_negg": [0.5960767269134521, 0.6189320087432861, 0.580556333065033], "r_norm": [0.23334062099456787, 0.23265743255615234, 0.23751312494277954]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9695, "test_loss": 0.1124767749786377}, {"step": 2400, "epoch": 5, "train_loss": 0.06642969697713852, "cos_r_negg": [0.5909543633460999, 0.6104036569595337, 0.5869391560554504], "cos_innovation_negg": [0.5909543633460999, 0.6104036569595337, 0.5869391560554504], "cos_apical_negg": [0.04040750488638878, 0.06977728009223938, 0.13496091961860657], "cos_Ac_negg": [0.6188592910766602, 0.6329231262207031, 0.5975795984268188], "r_norm": [0.1879744976758957, 0.18929597735404968, 0.19415967166423798]}, {"step": 2500, "epoch": 5, "train_loss": 0.01560700498521328, "cos_r_negg": [0.5764260292053223, 0.6069477796554565, 0.578482985496521], "cos_innovation_negg": [0.5764260292053223, 0.6069477796554565, 0.578482985496521], "cos_apical_negg": [0.029892697930336, 0.08527258038520813, 0.14372044801712036], "cos_Ac_negg": [0.6116330027580261, 0.6322923898696899, 0.5927160978317261], "r_norm": [0.24426546692848206, 0.24442210793495178, 0.24917516112327576], "ldr": {"dL_sdil": -0.01148965209722519, "dL_bp": -0.012451238930225372, "ratio": 0.9227717949684564}}, {"step": 2600, "epoch": 5, "train_loss": 0.05091261863708496, "cos_r_negg": [0.5831518173217773, 0.6210675239562988, 0.5870150327682495], "cos_innovation_negg": [0.5831518173217773, 0.6210675239562988, 0.5870150327682495], "cos_apical_negg": [0.04560829699039459, 0.09175842255353928, 0.14557278156280518], "cos_Ac_negg": [0.6158068776130676, 0.6439459919929504, 0.5956686735153198], "r_norm": [0.23803922533988953, 0.23771339654922485, 0.24180249869823456]}, {"step": 2700, "epoch": 5, "train_loss": 0.04318687692284584, "cos_r_negg": [0.5957183837890625, 0.6321381330490112, 0.5912017226219177], "cos_innovation_negg": [0.5957183837890625, 0.6321381330490112, 0.5912017226219177], "cos_apical_negg": [0.03164352849125862, 0.08044534176588058, 0.13389158248901367], "cos_Ac_negg": [0.6257493495941162, 0.6570411324501038, 0.602735161781311], "r_norm": [0.19935742020606995, 0.19892239570617676, 0.2033696472644806]}, {"step": 2800, "epoch": 5, "train_loss": 0.08792144060134888, "cos_r_negg": [0.5771921873092651, 0.6217223405838013, 0.5921776294708252], "cos_innovation_negg": [0.5771921873092651, 0.6217223405838013, 0.5921776294708252], "cos_apical_negg": [0.02564740926027298, 0.0745646134018898, 0.1367609202861786], "cos_Ac_negg": [0.6057387590408325, 0.6472271084785461, 0.6045480370521545], "r_norm": [0.18388976156711578, 0.1832829862833023, 0.18803295493125916]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9703, "test_loss": 0.10882798805236817}, {"step": 2900, "epoch": 6, "train_loss": 0.013013310730457306, "cos_r_negg": [0.5745726823806763, 0.6178932785987854, 0.5860469341278076], "cos_innovation_negg": [0.5745726823806763, 0.6178932785987854, 0.5860469341278076], "cos_apical_negg": [0.024404019117355347, 0.08036196231842041, 0.14160391688346863], "cos_Ac_negg": [0.6131798028945923, 0.6507341861724854, 0.6046215295791626], "r_norm": [0.20108121633529663, 0.20018455386161804, 0.20396503806114197]}, {"step": 3000, "epoch": 6, "train_loss": 0.033054664731025696, "cos_r_negg": [0.562021017074585, 0.6133217215538025, 0.5866560935974121], "cos_innovation_negg": [0.562021017074585, 0.6133217215538025, 0.5866560935974121], "cos_apical_negg": [0.013995595276355743, 0.06926731765270233, 0.12820187211036682], "cos_Ac_negg": [0.6141505241394043, 0.6563743352890015, 0.6088958978652954], "r_norm": [0.15822497010231018, 0.15869250893592834, 0.16342765092849731], "ldr": {"dL_sdil": -0.005045607686042786, "dL_bp": -0.006118357181549072, "ratio": 0.8246670693333593}}, {"step": 3100, "epoch": 6, "train_loss": 0.03830147534608841, "cos_r_negg": [0.5494112372398376, 0.6081170439720154, 0.5899116396903992], "cos_innovation_negg": [0.5494112372398376, 0.6081170439720154, 0.5899116396903992], "cos_apical_negg": [0.030363958328962326, 0.08567628264427185, 0.13646987080574036], "cos_Ac_negg": [0.6140666007995605, 0.6589256525039673, 0.6149048805236816], "r_norm": [0.1938188672065735, 0.1934228241443634, 0.1997278928756714]}, {"step": 3200, "epoch": 6, "train_loss": 0.04877293109893799, "cos_r_negg": [0.5403801202774048, 0.5943267941474915, 0.5810519456863403], "cos_innovation_negg": [0.5403801202774048, 0.5943267941474915, 0.5810519456863403], "cos_apical_negg": [0.015142959542572498, 0.07472680509090424, 0.1282399594783783], "cos_Ac_negg": [0.6186503171920776, 0.6629937887191772, 0.6203477382659912], "r_norm": [0.1796012818813324, 0.17909282445907593, 0.18256071209907532]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9721, "test_loss": 0.10792610893249512}, {"step": 3300, "epoch": 7, "train_loss": 0.009267272427678108, "cos_r_negg": [0.5438717007637024, 0.5952125787734985, 0.5815542936325073], "cos_innovation_negg": [0.5438717007637024, 0.5952125787734985, 0.5815542936325073], "cos_apical_negg": [0.026580266654491425, 0.07269361615180969, 0.1270410120487213], "cos_Ac_negg": [0.6219054460525513, 0.6613340377807617, 0.6265897154808044], "r_norm": [0.16144952178001404, 0.1611819714307785, 0.16384157538414001]}, {"step": 3400, "epoch": 7, "train_loss": 0.01776883751153946, "cos_r_negg": [0.5103996992111206, 0.5722621083259583, 0.5747686624526978], "cos_innovation_negg": [0.5103996992111206, 0.5722621083259583, 0.5747686624526978], "cos_apical_negg": [0.023606907576322556, 0.07380744814872742, 0.13563251495361328], "cos_Ac_negg": [0.6165734529495239, 0.6589145660400391, 0.6254906058311462], "r_norm": [0.18773292005062103, 0.1874973475933075, 0.19095928966999054]}, {"step": 3500, "epoch": 7, "train_loss": 0.00803975947201252, "cos_r_negg": [0.5079407691955566, 0.5599184036254883, 0.5622813701629639], "cos_innovation_negg": [0.5079407691955566, 0.5599184036254883, 0.5622813701629639], "cos_apical_negg": [0.009438520297408104, 0.0698479413986206, 0.12053480744361877], "cos_Ac_negg": [0.6317951679229736, 0.6660887002944946, 0.636048436164856], "r_norm": [0.16335417330265045, 0.1640072762966156, 0.16649848222732544], "ldr": {"dL_sdil": -0.0055382028222084045, "dL_bp": -0.007310442626476288, "ratio": 0.7575742133794542}}, {"step": 3600, "epoch": 7, "train_loss": 0.0013797478750348091, "cos_r_negg": [0.5053645968437195, 0.5602267384529114, 0.5643929839134216], "cos_innovation_negg": [0.5053645968437195, 0.5602267384529114, 0.5643929839134216], "cos_apical_negg": [0.010374944657087326, 0.07207032293081284, 0.12697625160217285], "cos_Ac_negg": [0.6321369409561157, 0.6641747951507568, 0.6335896253585815], "r_norm": [0.1691068708896637, 0.1682838499546051, 0.17226824164390564]}, {"step": 3700, "epoch": 7, "train_loss": 0.0032279654406011105, "cos_r_negg": [0.5042350888252258, 0.5589630603790283, 0.5636758208274841], "cos_innovation_negg": [0.5042350888252258, 0.5589630603790283, 0.5636758208274841], "cos_apical_negg": [0.019520625472068787, 0.07704445719718933, 0.1362655609846115], "cos_Ac_negg": [0.6336003541946411, 0.6712273359298706, 0.6419575214385986], "r_norm": [0.18173480033874512, 0.18056946992874146, 0.18452316522598267]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9733, "test_loss": 0.11650427446365356}, {"step": 3800, "epoch": 8, "train_loss": 0.00982544757425785, "cos_r_negg": [0.46905720233917236, 0.5218753218650818, 0.5423223972320557], "cos_innovation_negg": [0.46905720233917236, 0.5218753218650818, 0.5423223972320557], "cos_apical_negg": [0.019694751128554344, 0.06497429311275482, 0.13053977489471436], "cos_Ac_negg": [0.6314952373504639, 0.666953444480896, 0.6417737603187561], "r_norm": [0.16658152639865875, 0.16555774211883545, 0.16968370974063873]}, {"step": 3900, "epoch": 8, "train_loss": 0.02925984561443329, "cos_r_negg": [0.4413347542285919, 0.49921470880508423, 0.5250418186187744], "cos_innovation_negg": [0.4413347542285919, 0.49921470880508423, 0.5250418186187744], "cos_apical_negg": [0.011488229967653751, 0.06818787753582001, 0.12613055109977722], "cos_Ac_negg": [0.6245931386947632, 0.6577432155609131, 0.6453928351402283], "r_norm": [0.15324461460113525, 0.15281051397323608, 0.15677839517593384]}, {"step": 4000, "epoch": 8, "train_loss": 0.008045164868235588, "cos_r_negg": [0.42673054337501526, 0.49403172731399536, 0.5183005332946777], "cos_innovation_negg": [0.42673054337501526, 0.49403172731399536, 0.5183005332946777], "cos_apical_negg": [0.0051508452743291855, 0.06453575938940048, 0.12304888665676117], "cos_Ac_negg": [0.6234814524650574, 0.6618207693099976, 0.6484739780426025], "r_norm": [0.17489200830459595, 0.17475000023841858, 0.17803341150283813], "ldr": {"dL_sdil": -0.006972499191761017, "dL_bp": -0.010512404143810272, "ratio": 0.6632639971196732}}, {"step": 4100, "epoch": 8, "train_loss": 0.0034962778445333242, "cos_r_negg": [0.4441763758659363, 0.5131612420082092, 0.5346510410308838], "cos_innovation_negg": [0.4441763758659363, 0.5131612420082092, 0.5346510410308838], "cos_apical_negg": [0.009656371548771858, 0.06788153201341629, 0.12598329782485962], "cos_Ac_negg": [0.6313721537590027, 0.6735833883285522, 0.6593194007873535], "r_norm": [0.1783696860074997, 0.1777813732624054, 0.18104714155197144]}, {"step": 4200, "epoch": 8, "train_loss": 0.008683789521455765, "cos_r_negg": [0.44431424140930176, 0.5114341974258423, 0.5370839834213257], "cos_innovation_negg": [0.44431424140930176, 0.5114341974258423, 0.5370839834213257], "cos_apical_negg": [0.009143955074250698, 0.0695948675274849, 0.12227039039134979], "cos_Ac_negg": [0.6305049061775208, 0.6718047857284546, 0.6562168598175049], "r_norm": [0.16913482546806335, 0.16838935017585754, 0.17084577679634094]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9739, "test_loss": 0.1178732195854187}, {"step": 4300, "epoch": 9, "train_loss": 0.0016776140546426177, "cos_r_negg": [0.43030187487602234, 0.4954873323440552, 0.5225053429603577], "cos_innovation_negg": [0.43030187487602234, 0.4954873323440552, 0.5225053429603577], "cos_apical_negg": [0.011808861047029495, 0.06989718228578568, 0.11824727803468704], "cos_Ac_negg": [0.6243135333061218, 0.6735286116600037, 0.6625051498413086], "r_norm": [0.16774168610572815, 0.16727542877197266, 0.16974994540214539]}, {"step": 4400, "epoch": 9, "train_loss": 0.0011276163859292865, "cos_r_negg": [0.3988119959831238, 0.4685785174369812, 0.4951285719871521], "cos_innovation_negg": [0.3988119959831238, 0.4685785174369812, 0.4951285719871521], "cos_apical_negg": [0.006170487031340599, 0.06447477638721466, 0.11312360316514969], "cos_Ac_negg": [0.6207109689712524, 0.6658036112785339, 0.6566041111946106], "r_norm": [0.16017498075962067, 0.15935488045215607, 0.1622387021780014]}, {"step": 4500, "epoch": 9, "train_loss": 0.08726860582828522, "cos_r_negg": [0.38311997056007385, 0.44503679871559143, 0.4763762950897217], "cos_innovation_negg": [0.38311997056007385, 0.44503679871559143, 0.4763762950897217], "cos_apical_negg": [0.00993735808879137, 0.06415434181690216, 0.11535833775997162], "cos_Ac_negg": [0.619747519493103, 0.6640850305557251, 0.6565393209457397], "r_norm": [0.1903897374868393, 0.18939487636089325, 0.19168758392333984], "ldr": {"dL_sdil": -0.012501165270805359, "dL_bp": -0.019839875400066376, "ratio": 0.6301030131854324}}, {"step": 4600, "epoch": 9, "train_loss": 0.0018958705477416515, "cos_r_negg": [0.3746110200881958, 0.44570279121398926, 0.4720151126384735], "cos_innovation_negg": [0.3746110200881958, 0.44570279121398926, 0.4720151126384735], "cos_apical_negg": [0.005030651111155748, 0.06676638126373291, 0.1082020178437233], "cos_Ac_negg": [0.6145462393760681, 0.6667003631591797, 0.6569769382476807], "r_norm": [0.1803533285856247, 0.17894554138183594, 0.18128520250320435]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.975, "test_loss": 0.11868392906188965}, {"step": 4700, "epoch": 10, "train_loss": 0.0014905092539265752, "cos_r_negg": [0.35386329889297485, 0.41931837797164917, 0.4501698613166809], "cos_innovation_negg": [0.35386329889297485, 0.41931837797164917, 0.4501698613166809], "cos_apical_negg": [0.005821586586534977, 0.061907459050416946, 0.10470347106456757], "cos_Ac_negg": [0.6195168495178223, 0.6647994518280029, 0.6555296182632446], "r_norm": [0.16766610741615295, 0.16651062667369843, 0.17018306255340576]}, {"step": 4800, "epoch": 10, "train_loss": 0.0012656109174713492, "cos_r_negg": [0.34898343682289124, 0.41529911756515503, 0.4481699466705322], "cos_innovation_negg": [0.34898343682289124, 0.41529911756515503, 0.4481699466705322], "cos_apical_negg": [0.005106424912810326, 0.06552715599536896, 0.10781456530094147], "cos_Ac_negg": [0.6147809028625488, 0.66548091173172, 0.6564418077468872], "r_norm": [0.1745816171169281, 0.17323486506938934, 0.1773829162120819]}, {"step": 4900, "epoch": 10, "train_loss": 0.0012779165990650654, "cos_r_negg": [0.3501048684120178, 0.4089609980583191, 0.4419053792953491], "cos_innovation_negg": [0.3501048684120178, 0.4089609980583191, 0.4419053792953491], "cos_apical_negg": [0.006679007317870855, 0.06383780390024185, 0.10522712767124176], "cos_Ac_negg": [0.6193428039550781, 0.6660074591636658, 0.6599264144897461], "r_norm": [0.17439880967140198, 0.17352086305618286, 0.17672832310199738]}, {"step": 5000, "epoch": 10, "train_loss": 0.000160726165631786, "cos_r_negg": [0.33845794200897217, 0.3988485336303711, 0.4336903691291809], "cos_innovation_negg": [0.33845794200897217, 0.3988485336303711, 0.4336903691291809], "cos_apical_negg": [0.0021523749455809593, 0.06200505793094635, 0.10135708749294281], "cos_Ac_negg": [0.614773154258728, 0.6642279028892517, 0.6590754985809326], "r_norm": [0.17232462763786316, 0.17092199623584747, 0.1742478609085083], "ldr": {"dL_sdil": -0.009563423693180084, "dL_bp": -0.01564066857099533, "ratio": 0.6114459653543757}}, {"step": 5100, "epoch": 10, "train_loss": 0.0022132722660899162, "cos_r_negg": [0.3401888608932495, 0.4020995497703552, 0.4394442141056061], "cos_innovation_negg": [0.3401888608932495, 0.4020995497703552, 0.4394442141056061], "cos_apical_negg": [0.0012293346226215363, 0.06330874562263489, 0.10881869494915009], "cos_Ac_negg": [0.6191054582595825, 0.668574869632721, 0.6650499105453491], "r_norm": [0.17797641456127167, 0.17698729038238525, 0.1799686849117279]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9754, "test_loss": 0.1257467162132263}, {"step": 5200, "epoch": 11, "train_loss": 0.0004062363877892494, "cos_r_negg": [0.3309895098209381, 0.3949207067489624, 0.4280523359775543], "cos_innovation_negg": [0.3309895098209381, 0.3949207067489624, 0.4280523359775543], "cos_apical_negg": [-0.0009264005348086357, 0.06143319979310036, 0.10022175312042236], "cos_Ac_negg": [0.6125749349594116, 0.6660000085830688, 0.6617434024810791], "r_norm": [0.16346126794815063, 0.16259120404720306, 0.1651577651500702]}, {"step": 5300, "epoch": 11, "train_loss": 0.0014113449724391103, "cos_r_negg": [0.3236422538757324, 0.38755232095718384, 0.41874659061431885], "cos_innovation_negg": [0.3236422538757324, 0.38755232095718384, 0.41874659061431885], "cos_apical_negg": [0.0011327746324241161, 0.061301615089178085, 0.097899429500103], "cos_Ac_negg": [0.6073694229125977, 0.6618682146072388, 0.661133885383606], "r_norm": [0.1701488494873047, 0.16920945048332214, 0.1716749519109726]}, {"step": 5400, "epoch": 11, "train_loss": 0.0017204724717885256, "cos_r_negg": [0.32189029455184937, 0.3826256990432739, 0.4150451719760895], "cos_innovation_negg": [0.32189029455184937, 0.3826256990432739, 0.4150451719760895], "cos_apical_negg": [-0.001213527750223875, 0.060469381511211395, 0.09835895895957947], "cos_Ac_negg": [0.6155924201011658, 0.6649582386016846, 0.6622479557991028], "r_norm": [0.16895024478435516, 0.16803152859210968, 0.17112287878990173]}, {"step": 5500, "epoch": 11, "train_loss": 0.000993720255792141, "cos_r_negg": [0.32139259576797485, 0.38371407985687256, 0.41844815015792847], "cos_innovation_negg": [0.32139259576797485, 0.38371407985687256, 0.41844815015792847], "cos_apical_negg": [-0.001972053200006485, 0.061453819274902344, 0.09971621632575989], "cos_Ac_negg": [0.6149802207946777, 0.6682905554771423, 0.665856122970581], "r_norm": [0.17345094680786133, 0.17252591252326965, 0.17564040422439575], "ldr": {"dL_sdil": -0.009622596204280853, "dL_bp": -0.016050435602664948, "ratio": 0.5995224330661255}}, {"step": 5600, "epoch": 11, "train_loss": 0.0016657065134495497, "cos_r_negg": [0.30397194623947144, 0.36559394001960754, 0.39393600821495056], "cos_innovation_negg": [0.30397194623947144, 0.36559394001960754, 0.39393600821495056], "cos_apical_negg": [-0.003711450845003128, 0.05918329954147339, 0.08927440643310547], "cos_Ac_negg": [0.6115577220916748, 0.6611321568489075, 0.6605039834976196], "r_norm": [0.1635160744190216, 0.16261613368988037, 0.16581760346889496]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9757, "test_loss": 0.12609906377792357}, {"step": 5700, "epoch": 12, "train_loss": 0.0013611136237159371, "cos_r_negg": [0.3041272759437561, 0.3620166778564453, 0.39632081985473633], "cos_innovation_negg": [0.3041272759437561, 0.3620166778564453, 0.39632081985473633], "cos_apical_negg": [-0.0015211999416351318, 0.05659618228673935, 0.09308314323425293], "cos_Ac_negg": [0.6126233339309692, 0.665428102016449, 0.6645920276641846], "r_norm": [0.1696382761001587, 0.16894446313381195, 0.17202845215797424]}, {"step": 5800, "epoch": 12, "train_loss": 0.0020447673741728067, "cos_r_negg": [0.2991286516189575, 0.35727763175964355, 0.3913206160068512], "cos_innovation_negg": [0.2991286516189575, 0.35727763175964355, 0.3913206160068512], "cos_apical_negg": [-0.003953907638788223, 0.057026591151952744, 0.0929592177271843], "cos_Ac_negg": [0.6113331913948059, 0.6634926795959473, 0.663231372833252], "r_norm": [0.1672661155462265, 0.16649901866912842, 0.16933535039424896]}, {"step": 5900, "epoch": 12, "train_loss": 0.0014066663570702076, "cos_r_negg": [0.30047860741615295, 0.36044222116470337, 0.39695295691490173], "cos_innovation_negg": [0.30047860741615295, 0.36044222116470337, 0.39695295691490173], "cos_apical_negg": [0.0013358844444155693, 0.05819650739431381, 0.09572765231132507], "cos_Ac_negg": [0.6117909550666809, 0.6672633290290833, 0.6678286194801331], "r_norm": [0.1781907081604004, 0.17731031775474548, 0.17990857362747192]}, {"step": 6000, "epoch": 12, "train_loss": 0.0002284040383528918, "cos_r_negg": [0.2873755693435669, 0.34967178106307983, 0.38316577672958374], "cos_innovation_negg": [0.2873755693435669, 0.34967178106307983, 0.38316577672958374], "cos_apical_negg": [-0.0016045423690229654, 0.05784754827618599, 0.09123990684747696], "cos_Ac_negg": [0.6095637083053589, 0.6628272533416748, 0.6649200916290283], "r_norm": [0.17103305459022522, 0.17006808519363403, 0.1730501502752304], "ldr": {"dL_sdil": -0.009525716304779053, "dL_bp": -0.01639055460691452, "ratio": 0.5811710789066609}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9758, "test_loss": 0.12974705410003662}, {"step": 6100, "epoch": 13, "train_loss": 0.0007630693726241589, "cos_r_negg": [0.2865099608898163, 0.35144031047821045, 0.3844558000564575], "cos_innovation_negg": [0.2865099608898163, 0.35144031047821045, 0.3844558000564575], "cos_apical_negg": [-0.004450838081538677, 0.05715659260749817, 0.09215658903121948], "cos_Ac_negg": [0.6112799644470215, 0.6659401655197144, 0.6665726900100708], "r_norm": [0.1627330332994461, 0.161863774061203, 0.1650581657886505]}, {"step": 6200, "epoch": 13, "train_loss": 0.0005757695762440562, "cos_r_negg": [0.2834436893463135, 0.3464440703392029, 0.37994325160980225], "cos_innovation_negg": [0.2834436893463135, 0.3464440703392029, 0.37994325160980225], "cos_apical_negg": [-0.005105349700897932, 0.05714293196797371, 0.09181632101535797], "cos_Ac_negg": [0.6104350090026855, 0.663049042224884, 0.6663385629653931], "r_norm": [0.1573629379272461, 0.15636897087097168, 0.15965387225151062]}, {"step": 6300, "epoch": 13, "train_loss": 0.0006871781661175191, "cos_r_negg": [0.27674785256385803, 0.33835023641586304, 0.3720698058605194], "cos_innovation_negg": [0.27674785256385803, 0.33835023641586304, 0.3720698058605194], "cos_apical_negg": [-0.006839130073785782, 0.05593195557594299, 0.08954693377017975], "cos_Ac_negg": [0.6090211868286133, 0.660690426826477, 0.6654565930366516], "r_norm": [0.15734589099884033, 0.15652035176753998, 0.159844771027565]}, {"step": 6400, "epoch": 13, "train_loss": 0.0004738799179904163, "cos_r_negg": [0.2787921130657196, 0.3402276933193207, 0.3789054751396179], "cos_innovation_negg": [0.2787921130657196, 0.3402276933193207, 0.3789054751396179], "cos_apical_negg": [-0.005164789501577616, 0.05688251182436943, 0.09416162222623825], "cos_Ac_negg": [0.6102557182312012, 0.6640920639038086, 0.668114960193634], "r_norm": [0.16442690789699554, 0.1634214222431183, 0.16653040051460266]}, {"step": 6500, "epoch": 13, "train_loss": 0.000710058375261724, "cos_r_negg": [0.2747942805290222, 0.33808666467666626, 0.37204474210739136], "cos_innovation_negg": [0.2747942805290222, 0.33808666467666626, 0.37204474210739136], "cos_apical_negg": [-0.00583961745724082, 0.0555395670235157, 0.08962330967187881], "cos_Ac_negg": [0.6093389987945557, 0.6635269522666931, 0.6665266156196594], "r_norm": [0.16985037922859192, 0.16904151439666748, 0.1723356395959854], "ldr": {"dL_sdil": -0.009799778461456299, "dL_bp": -0.017330914735794067, "ratio": 0.5654507341852255}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9762, "test_loss": 0.13262854080200195}, {"step": 6600, "epoch": 14, "train_loss": 0.00024427290190942585, "cos_r_negg": [0.27022114396095276, 0.33344656229019165, 0.36703813076019287], "cos_innovation_negg": [0.27022114396095276, 0.33344656229019165, 0.36703813076019287], "cos_apical_negg": [-0.0057905614376068115, 0.056866422295570374, 0.08863642811775208], "cos_Ac_negg": [0.6120237112045288, 0.6635233163833618, 0.6657634973526001], "r_norm": [0.16316649317741394, 0.16228581964969635, 0.16553249955177307]}, {"step": 6700, "epoch": 14, "train_loss": 0.0004823197377845645, "cos_r_negg": [0.26939457654953003, 0.3327721655368805, 0.36749890446662903], "cos_innovation_negg": [0.26939457654953003, 0.3327721655368805, 0.36749890446662903], "cos_apical_negg": [-0.006343171000480652, 0.05564163625240326, 0.0884370505809784], "cos_Ac_negg": [0.6096881628036499, 0.6633120775222778, 0.6669067144393921], "r_norm": [0.16940899193286896, 0.16868823766708374, 0.1714802384376526]}, {"step": 6800, "epoch": 14, "train_loss": 0.0006483035394921899, "cos_r_negg": [0.2663818597793579, 0.3270002603530884, 0.360596239566803], "cos_innovation_negg": [0.2663818597793579, 0.3270002603530884, 0.360596239566803], "cos_apical_negg": [-0.005050738342106342, 0.05709920823574066, 0.08860805630683899], "cos_Ac_negg": [0.6111637949943542, 0.6632071733474731, 0.6660500764846802], "r_norm": [0.17243969440460205, 0.1715322583913803, 0.17472121119499207]}, {"step": 6900, "epoch": 14, "train_loss": 0.000526372401509434, "cos_r_negg": [0.2578943967819214, 0.3195595145225525, 0.3518974483013153], "cos_innovation_negg": [0.2578943967819214, 0.3195595145225525, 0.3518974483013153], "cos_apical_negg": [-0.006891245488077402, 0.056059278547763824, 0.08534833788871765], "cos_Ac_negg": [0.6097977161407471, 0.6608104705810547, 0.6642349362373352], "r_norm": [0.15979497134685516, 0.15897135436534882, 0.16206049919128418]}, {"step": 7000, "epoch": 14, "train_loss": 0.0003984446229878813, "cos_r_negg": [0.25704672932624817, 0.3179672658443451, 0.3501430153846741], "cos_innovation_negg": [0.25704672932624817, 0.3179672658443451, 0.3501430153846741], "cos_apical_negg": [-0.00590121652930975, 0.05605883151292801, 0.0864536315202713], "cos_Ac_negg": [0.6106324195861816, 0.6610426306724548, 0.6641703844070435], "r_norm": [0.15977753698825836, 0.15890425443649292, 0.16240063309669495], "ldr": {"dL_sdil": -0.008981242775917053, "dL_bp": -0.016131021082401276, "ratio": 0.55676839860531}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9759, "test_loss": 0.13539633264541626}], "final": {"test_acc": 0.9759, "test_loss": 0.13539633264541626, "wall_s": 23.59708571434021, "cos_r_negg": [0.2558465003967285, 0.3163191080093384, 0.34814485907554626], "cos_innovation_negg": [0.2558465003967285, 0.3163191080093384, 0.34814485907554626], "cos_apical_negg": [-0.0059834434650838375, 0.05518651381134987, 0.08496376872062683], "cos_Ac_negg": [0.6108644008636475, 0.6611005663871765, 0.6636170148849487]}}