{"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": 8, "use_residual": 0, "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 0, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "diag_nuis_v2_rho0_res0_s0", "n_in": 784}, "provenance": {"git_commit": "54633b2e0f51a8f48d4ba627d778c87927f54ca1", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3978137969970703, "cos_r_negg": [-0.02239900827407837, -0.021294258534908295, 0.034575555473566055], "cos_apical_negg": [-0.02239900827407837, -0.021294258534908295, 0.034575555473566055], "cos_Ac_negg": [-0.02239900827407837, -0.021294258534908295, 0.034575555473566055], "r_norm": [4.891855239868164, 4.819446563720703, 4.813877105712891], "ldr": {"dL_sdil": -0.007513999938964844, "dL_bp": -0.23199224472045898, "ratio": 0.03238901346904463}}, {"step": 100, "epoch": 0, "train_loss": 0.3326342701911926, "cos_r_negg": [0.44732898473739624, 0.38416725397109985, 0.3667517900466919], "cos_apical_negg": [0.44732898473739624, 0.38416725397109985, 0.3667517900466919], "cos_Ac_negg": [0.44732898473739624, 0.38416725397109985, 0.3667517900466919], "r_norm": [0.8364949822425842, 0.8173195123672485, 0.8178015947341919]}, {"step": 200, "epoch": 0, "train_loss": 0.2806997001171112, "cos_r_negg": [0.47987207770347595, 0.4345510005950928, 0.43154287338256836], "cos_apical_negg": [0.47987207770347595, 0.4345510005950928, 0.43154287338256836], "cos_Ac_negg": [0.47987207770347595, 0.4345510005950928, 0.43154287338256836], "r_norm": [0.7367554306983948, 0.7248662710189819, 0.7212702035903931]}, {"step": 300, "epoch": 0, "train_loss": 0.2950687110424042, "cos_r_negg": [0.5101386308670044, 0.4740510582923889, 0.4546336531639099], "cos_apical_negg": [0.5101386308670044, 0.4740510582923889, 0.4546336531639099], "cos_Ac_negg": [0.5101386308670044, 0.4740510582923889, 0.4546336531639099], "r_norm": [0.5894073843955994, 0.580111563205719, 0.5781813859939575]}, {"step": 400, "epoch": 0, "train_loss": 0.11995674669742584, "cos_r_negg": [0.5342420339584351, 0.48684829473495483, 0.45452749729156494], "cos_apical_negg": [0.5342420339584351, 0.48684829473495483, 0.45452749729156494], "cos_Ac_negg": [0.5342420339584351, 0.48684829473495483, 0.45452749729156494], "r_norm": [0.5368587374687195, 0.524689793586731, 0.5227817893028259]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9455, "test_loss": 0.183567440032959}, {"step": 500, "epoch": 1, "train_loss": 0.266681432723999, "cos_r_negg": [0.5489188432693481, 0.508673906326294, 0.4944116771221161], "cos_apical_negg": [0.5489188432693481, 0.508673906326294, 0.4944116771221161], "cos_Ac_negg": [0.5489188432693481, 0.508673906326294, 0.4944116771221161], "r_norm": [0.5609025359153748, 0.5504555702209473, 0.5459269881248474], "ldr": {"dL_sdil": -0.01507239043712616, "dL_bp": -0.012114167213439941, "ratio": 1.2441953434821542}}, {"step": 600, "epoch": 1, "train_loss": 0.26136401295661926, "cos_r_negg": [0.5187579393386841, 0.5020226836204529, 0.4812536835670471], "cos_apical_negg": [0.5187579393386841, 0.5020226836204529, 0.4812536835670471], "cos_Ac_negg": [0.5187579393386841, 0.5020226836204529, 0.4812536835670471], "r_norm": [0.48265740275382996, 0.47054633498191833, 0.469404935836792]}, {"step": 700, "epoch": 1, "train_loss": 0.05521097034215927, "cos_r_negg": [0.529063880443573, 0.5166817903518677, 0.49424171447753906], "cos_apical_negg": [0.529063880443573, 0.5166817903518677, 0.49424171447753906], "cos_Ac_negg": [0.529063880443573, 0.5166817903518677, 0.49424171447753906], "r_norm": [0.38553091883659363, 0.3786587417125702, 0.3761131763458252]}, {"step": 800, "epoch": 1, "train_loss": 0.1167590469121933, "cos_r_negg": [0.5241111516952515, 0.5325952768325806, 0.48102712631225586], "cos_apical_negg": [0.5241111516952515, 0.5325952768325806, 0.48102712631225586], "cos_Ac_negg": [0.5241111516952515, 0.5325952768325806, 0.48102712631225586], "r_norm": [0.39521896839141846, 0.39017635583877563, 0.38718611001968384]}, {"step": 900, "epoch": 1, "train_loss": 0.12190888077020645, "cos_r_negg": [0.556609034538269, 0.5399625301361084, 0.5007081031799316], "cos_apical_negg": [0.556609034538269, 0.5399625301361084, 0.5007081031799316], "cos_Ac_negg": [0.556609034538269, 0.5399625301361084, 0.5007081031799316], "r_norm": [0.3485320806503296, 0.3397080898284912, 0.33907851576805115]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9563, "test_loss": 0.14598272552490235}, {"step": 1000, "epoch": 2, "train_loss": 0.21153098344802856, "cos_r_negg": [0.5525427460670471, 0.5603439211845398, 0.5186597108840942], "cos_apical_negg": [0.5525427460670471, 0.5603439211845398, 0.5186597108840942], "cos_Ac_negg": [0.5525427460670471, 0.5603439211845398, 0.5186597108840942], "r_norm": [0.370191752910614, 0.36559224128723145, 0.3626173436641693], "ldr": {"dL_sdil": -0.010546453297138214, "dL_bp": -0.00943966954946518, "ratio": 1.1172481453798073}}, {"step": 1100, "epoch": 2, "train_loss": 0.08847551047801971, "cos_r_negg": [0.5112615823745728, 0.5405528545379639, 0.491232693195343], "cos_apical_negg": [0.5112615823745728, 0.5405528545379639, 0.491232693195343], "cos_Ac_negg": [0.5112615823745728, 0.5405528545379639, 0.491232693195343], "r_norm": [0.2634318768978119, 0.25687265396118164, 0.2574353814125061]}, {"step": 1200, "epoch": 2, "train_loss": 0.0304334107786417, "cos_r_negg": [0.5318017601966858, 0.5454596877098083, 0.5005227327346802], "cos_apical_negg": [0.5318017601966858, 0.5454596877098083, 0.5005227327346802], "cos_Ac_negg": [0.5318017601966858, 0.5454596877098083, 0.5005227327346802], "r_norm": [0.3200984597206116, 0.3153305649757385, 0.310947060585022]}, {"step": 1300, "epoch": 2, "train_loss": 0.1379236876964569, "cos_r_negg": [0.5548704862594604, 0.5711873769760132, 0.5169695615768433], "cos_apical_negg": [0.5548704862594604, 0.5711873769760132, 0.5169695615768433], "cos_Ac_negg": [0.5548704862594604, 0.5711873769760132, 0.5169695615768433], "r_norm": [0.2996469736099243, 0.2942494750022888, 0.2918258309364319]}, {"step": 1400, "epoch": 2, "train_loss": 0.2675240635871887, "cos_r_negg": [0.5424305200576782, 0.5742589235305786, 0.49545636773109436], "cos_apical_negg": [0.5424305200576782, 0.5742589235305786, 0.49545636773109436], "cos_Ac_negg": [0.5424305200576782, 0.5742589235305786, 0.49545636773109436], "r_norm": [0.3374136984348297, 0.3291166424751282, 0.3261111378669739]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.961, "test_loss": 0.1328551145553589}, {"step": 1500, "epoch": 3, "train_loss": 0.02847268432378769, "cos_r_negg": [0.5073622465133667, 0.5457200407981873, 0.4972744882106781], "cos_apical_negg": [0.5073622465133667, 0.5457200407981873, 0.4972744882106781], "cos_Ac_negg": [0.5073622465133667, 0.5457200407981873, 0.4972744882106781], "r_norm": [0.2855607867240906, 0.2791121304035187, 0.2787642478942871], "ldr": {"dL_sdil": -0.006588086485862732, "dL_bp": -0.006262242794036865, "ratio": 1.0520330658747608}}, {"step": 1600, "epoch": 3, "train_loss": 0.0700979083776474, "cos_r_negg": [0.5162990093231201, 0.5577571988105774, 0.5150010585784912], "cos_apical_negg": [0.5162990093231201, 0.5577571988105774, 0.5150010585784912], "cos_Ac_negg": [0.5162990093231201, 0.5577571988105774, 0.5150010585784912], "r_norm": [0.2842082977294922, 0.2784254848957062, 0.27576351165771484]}, {"step": 1700, "epoch": 3, "train_loss": 0.08875302970409393, "cos_r_negg": [0.5350693464279175, 0.5757480263710022, 0.5141506791114807], "cos_apical_negg": [0.5350693464279175, 0.5757480263710022, 0.5141506791114807], "cos_Ac_negg": [0.5350693464279175, 0.5757480263710022, 0.5141506791114807], "r_norm": [0.28283974528312683, 0.27689802646636963, 0.27501118183135986]}, {"step": 1800, "epoch": 3, "train_loss": 0.08456698805093765, "cos_r_negg": [0.5278818011283875, 0.5696221590042114, 0.5028855204582214], "cos_apical_negg": [0.5278818011283875, 0.5696221590042114, 0.5028855204582214], "cos_Ac_negg": [0.5278818011283875, 0.5696221590042114, 0.5028855204582214], "r_norm": [0.2775178849697113, 0.2699923515319824, 0.2691011428833008]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9654, "test_loss": 0.11376221809387208}, {"step": 1900, "epoch": 4, "train_loss": 0.06340675801038742, "cos_r_negg": [0.5159375071525574, 0.5532490611076355, 0.4990085959434509], "cos_apical_negg": [0.5159375071525574, 0.5532490611076355, 0.4990085959434509], "cos_Ac_negg": [0.5159375071525574, 0.5532490611076355, 0.4990085959434509], "r_norm": [0.2617102265357971, 0.25448280572891235, 0.25316107273101807]}, {"step": 2000, "epoch": 4, "train_loss": 0.09591986238956451, "cos_r_negg": [0.47856640815734863, 0.5316683053970337, 0.4704130291938782], "cos_apical_negg": [0.47856640815734863, 0.5316683053970337, 0.4704130291938782], "cos_Ac_negg": [0.47856640815734863, 0.5316683053970337, 0.4704130291938782], "r_norm": [0.25978973507881165, 0.25386324524879456, 0.25181859731674194], "ldr": {"dL_sdil": -0.0057485103607177734, "dL_bp": -0.0059058889746665955, "ratio": 0.9733522565994891}}, {"step": 2100, "epoch": 4, "train_loss": 0.06319307535886765, "cos_r_negg": [0.45975053310394287, 0.5218613147735596, 0.4650382995605469], "cos_apical_negg": [0.45975053310394287, 0.5218613147735596, 0.4650382995605469], "cos_Ac_negg": [0.45975053310394287, 0.5218613147735596, 0.4650382995605469], "r_norm": [0.23031632602214813, 0.22501714527606964, 0.22409947216510773]}, {"step": 2200, "epoch": 4, "train_loss": 0.057848647236824036, "cos_r_negg": [0.479114830493927, 0.5228580236434937, 0.4676823019981384], "cos_apical_negg": [0.479114830493927, 0.5228580236434937, 0.4676823019981384], "cos_Ac_negg": [0.479114830493927, 0.5228580236434937, 0.4676823019981384], "r_norm": [0.20740358531475067, 0.20326529443264008, 0.2015407234430313]}, {"step": 2300, "epoch": 4, "train_loss": 0.11177002638578415, "cos_r_negg": [0.4605717957019806, 0.510822057723999, 0.4517856240272522], "cos_apical_negg": [0.4605717957019806, 0.510822057723999, 0.4517856240272522], "cos_Ac_negg": [0.4605717957019806, 0.510822057723999, 0.4517856240272522], "r_norm": [0.2687836289405823, 0.2638903856277466, 0.2608131766319275]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9672, "test_loss": 0.11259154510498047}, {"step": 2400, "epoch": 5, "train_loss": 0.020450670272111893, "cos_r_negg": [0.4683169722557068, 0.5224604606628418, 0.4606319069862366], "cos_apical_negg": [0.4683169722557068, 0.5224604606628418, 0.4606319069862366], "cos_Ac_negg": [0.4683169722557068, 0.5224604606628418, 0.4606319069862366], "r_norm": [0.23890022933483124, 0.23350372910499573, 0.2330227941274643]}, {"step": 2500, "epoch": 5, "train_loss": 0.04949410259723663, "cos_r_negg": [0.4616062641143799, 0.5154315233230591, 0.45055413246154785], "cos_apical_negg": [0.4616062641143799, 0.5154315233230591, 0.45055413246154785], "cos_Ac_negg": [0.4616062641143799, 0.5154315233230591, 0.45055413246154785], "r_norm": [0.19714675843715668, 0.19294852018356323, 0.1915959119796753], "ldr": {"dL_sdil": -0.005470342934131622, "dL_bp": -0.0060305967926979065, "ratio": 0.9070981068996915}}, {"step": 2600, "epoch": 5, "train_loss": 0.02772347629070282, "cos_r_negg": [0.450645387172699, 0.5067758560180664, 0.45540016889572144], "cos_apical_negg": [0.450645387172699, 0.5067758560180664, 0.45540016889572144], "cos_Ac_negg": [0.450645387172699, 0.5067758560180664, 0.45540016889572144], "r_norm": [0.21218553185462952, 0.2091660499572754, 0.20675276219844818]}, {"step": 2700, "epoch": 5, "train_loss": 0.09122030436992645, "cos_r_negg": [0.46868619322776794, 0.5180381536483765, 0.450004905462265], "cos_apical_negg": [0.46868619322776794, 0.5180381536483765, 0.450004905462265], "cos_Ac_negg": [0.46868619322776794, 0.5180381536483765, 0.450004905462265], "r_norm": [0.20776993036270142, 0.20399165153503418, 0.20168794691562653]}, {"step": 2800, "epoch": 5, "train_loss": 0.011094631627202034, "cos_r_negg": [0.39923107624053955, 0.45298054814338684, 0.41231513023376465], "cos_apical_negg": [0.39923107624053955, 0.45298054814338684, 0.41231513023376465], "cos_Ac_negg": [0.39923107624053955, 0.45298054814338684, 0.41231513023376465], "r_norm": [0.1927286684513092, 0.18906784057617188, 0.18704429268836975]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9689, "test_loss": 0.11404035301208496}, {"step": 2900, "epoch": 6, "train_loss": 0.0365702323615551, "cos_r_negg": [0.36084771156311035, 0.4027143120765686, 0.3576774001121521], "cos_apical_negg": [0.36084771156311035, 0.4027143120765686, 0.3576774001121521], "cos_Ac_negg": [0.36084771156311035, 0.4027143120765686, 0.3576774001121521], "r_norm": [0.2347666174173355, 0.22903230786323547, 0.22849947214126587]}, {"step": 3000, "epoch": 6, "train_loss": 0.014387298375368118, "cos_r_negg": [0.38353437185287476, 0.43034154176712036, 0.3947371542453766], "cos_apical_negg": [0.38353437185287476, 0.43034154176712036, 0.3947371542453766], "cos_Ac_negg": [0.38353437185287476, 0.43034154176712036, 0.3947371542453766], "r_norm": [0.18399973213672638, 0.18022119998931885, 0.1777746081352234], "ldr": {"dL_sdil": -0.006564125418663025, "dL_bp": -0.007992811501026154, "ratio": 0.8212536249378948}}, {"step": 3100, "epoch": 6, "train_loss": 0.043194446712732315, "cos_r_negg": [0.3321278393268585, 0.38589146733283997, 0.3507818579673767], "cos_apical_negg": [0.3321278393268585, 0.38589146733283997, 0.3507818579673767], "cos_Ac_negg": [0.3321278393268585, 0.38589146733283997, 0.3507818579673767], "r_norm": [0.17794635891914368, 0.17384949326515198, 0.17293182015419006]}, {"step": 3200, "epoch": 6, "train_loss": 0.06848011910915375, "cos_r_negg": [0.3206913471221924, 0.37333035469055176, 0.33710089325904846], "cos_apical_negg": [0.3206913471221924, 0.37333035469055176, 0.33710089325904846], "cos_Ac_negg": [0.3206913471221924, 0.37333035469055176, 0.33710089325904846], "r_norm": [0.1944277584552765, 0.18918633460998535, 0.1876811534166336]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9699, "test_loss": 0.11801956939697265}, {"step": 3300, "epoch": 7, "train_loss": 0.01468692533671856, "cos_r_negg": [0.3207220435142517, 0.3707606792449951, 0.33233171701431274], "cos_apical_negg": [0.3207220435142517, 0.3707606792449951, 0.33233171701431274], "cos_Ac_negg": [0.3207220435142517, 0.3707606792449951, 0.33233171701431274], "r_norm": [0.15375901758670807, 0.151533305644989, 0.14959079027175903]}, {"step": 3400, "epoch": 7, "train_loss": 0.009123355150222778, "cos_r_negg": [0.28374552726745605, 0.3274969458580017, 0.2981197237968445], "cos_apical_negg": [0.28374552726745605, 0.3274969458580017, 0.2981197237968445], "cos_Ac_negg": [0.28374552726745605, 0.3274969458580017, 0.2981197237968445], "r_norm": [0.1616383194923401, 0.1594000905752182, 0.1570923626422882]}, {"step": 3500, "epoch": 7, "train_loss": 0.017021456733345985, "cos_r_negg": [0.30124372243881226, 0.3384850025177002, 0.3080917298793793], "cos_apical_negg": [0.30124372243881226, 0.3384850025177002, 0.3080917298793793], "cos_Ac_negg": [0.30124372243881226, 0.3384850025177002, 0.3080917298793793], "r_norm": [0.15971358120441437, 0.15718066692352295, 0.15513375401496887], "ldr": {"dL_sdil": -0.00407978892326355, "dL_bp": -0.0053077638149261475, "ratio": 0.768645528610492}}, {"step": 3600, "epoch": 7, "train_loss": 0.0063411155715584755, "cos_r_negg": [0.30416423082351685, 0.34374314546585083, 0.30913037061691284], "cos_apical_negg": [0.30416423082351685, 0.34374314546585083, 0.30913037061691284], "cos_Ac_negg": [0.30416423082351685, 0.34374314546585083, 0.30913037061691284], "r_norm": [0.15865351259708405, 0.15548795461654663, 0.15406224131584167]}, {"step": 3700, "epoch": 7, "train_loss": 0.005838679149746895, "cos_r_negg": [0.27810975909233093, 0.32046079635620117, 0.29577505588531494], "cos_apical_negg": [0.27810975909233093, 0.32046079635620117, 0.29577505588531494], "cos_Ac_negg": [0.27810975909233093, 0.32046079635620117, 0.29577505588531494], "r_norm": [0.17863257229328156, 0.17555025219917297, 0.1733328402042389]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.97, "test_loss": 0.12616020851135254}, {"step": 3800, "epoch": 8, "train_loss": 0.08425555378198624, "cos_r_negg": [0.2577763795852661, 0.2929215431213379, 0.26875755190849304], "cos_apical_negg": [0.2577763795852661, 0.2929215431213379, 0.26875755190849304], "cos_Ac_negg": [0.2577763795852661, 0.2929215431213379, 0.26875755190849304], "r_norm": [0.15988609194755554, 0.15718698501586914, 0.15486183762550354]}, {"step": 3900, "epoch": 8, "train_loss": 0.009738568216562271, "cos_r_negg": [0.26248815655708313, 0.3003271818161011, 0.27528297901153564], "cos_apical_negg": [0.26248815655708313, 0.3003271818161011, 0.27528297901153564], "cos_Ac_negg": [0.26248815655708313, 0.3003271818161011, 0.27528297901153564], "r_norm": [0.17696574330329895, 0.17444398999214172, 0.17152070999145508]}, {"step": 4000, "epoch": 8, "train_loss": 0.0004337254213169217, "cos_r_negg": [0.23113428056240082, 0.26362621784210205, 0.24363160133361816], "cos_apical_negg": [0.23113428056240082, 0.26362621784210205, 0.24363160133361816], "cos_Ac_negg": [0.23113428056240082, 0.26362621784210205, 0.24363160133361816], "r_norm": [0.173438161611557, 0.16961370408535004, 0.16862860321998596], "ldr": {"dL_sdil": -0.006462305784225464, "dL_bp": -0.009116381406784058, "ratio": 0.7088674218277513}}, {"step": 4100, "epoch": 8, "train_loss": 0.010536604560911655, "cos_r_negg": [0.23122937977313995, 0.26008719205856323, 0.24139222502708435], "cos_apical_negg": [0.23122937977313995, 0.26008719205856323, 0.24139222502708435], "cos_Ac_negg": [0.23122937977313995, 0.26008719205856323, 0.24139222502708435], "r_norm": [0.14982295036315918, 0.14678488671779633, 0.1456487774848938]}, {"step": 4200, "epoch": 8, "train_loss": 0.009496842510998249, "cos_r_negg": [0.2148902714252472, 0.24816416203975677, 0.22951406240463257], "cos_apical_negg": [0.2148902714252472, 0.24816416203975677, 0.22951406240463257], "cos_Ac_negg": [0.2148902714252472, 0.24816416203975677, 0.22951406240463257], "r_norm": [0.1480349898338318, 0.145409494638443, 0.14320231974124908]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9716, "test_loss": 0.12644243202209474}, {"step": 4300, "epoch": 9, "train_loss": 0.01001154538244009, "cos_r_negg": [0.21183210611343384, 0.24454396963119507, 0.227471262216568], "cos_apical_negg": [0.21183210611343384, 0.24454396963119507, 0.227471262216568], "cos_Ac_negg": [0.21183210611343384, 0.24454396963119507, 0.227471262216568], "r_norm": [0.16093958914279938, 0.15813371539115906, 0.15620499849319458]}, {"step": 4400, "epoch": 9, "train_loss": 0.0019738487899303436, "cos_r_negg": [0.19909998774528503, 0.229609876871109, 0.2141648679971695], "cos_apical_negg": [0.19909998774528503, 0.229609876871109, 0.2141648679971695], "cos_Ac_negg": [0.19909998774528503, 0.229609876871109, 0.2141648679971695], "r_norm": [0.15255683660507202, 0.1499708890914917, 0.1482204645872116]}, {"step": 4500, "epoch": 9, "train_loss": 0.0038066008128225803, "cos_r_negg": [0.18125811219215393, 0.21069471538066864, 0.1968364417552948], "cos_apical_negg": [0.18125811219215393, 0.21069471538066864, 0.1968364417552948], "cos_Ac_negg": [0.18125811219215393, 0.21069471538066864, 0.1968364417552948], "r_norm": [0.1395764797925949, 0.1376519650220871, 0.13562631607055664], "ldr": {"dL_sdil": -0.005997225642204285, "dL_bp": -0.008848287165164948, "ratio": 0.6777837936606441}}, {"step": 4600, "epoch": 9, "train_loss": 0.011752025224268436, "cos_r_negg": [0.19314654171466827, 0.22306716442108154, 0.21022829413414001], "cos_apical_negg": [0.19314654171466827, 0.22306716442108154, 0.21022829413414001], "cos_Ac_negg": [0.19314654171466827, 0.22306716442108154, 0.21022829413414001], "r_norm": [0.147353857755661, 0.14492204785346985, 0.14291857182979584]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9727, "test_loss": 0.12896337757110596}, {"step": 4700, "epoch": 10, "train_loss": 0.0019247026648372412, "cos_r_negg": [0.20044180750846863, 0.2280893176794052, 0.2148800492286682], "cos_apical_negg": [0.20044180750846863, 0.2280893176794052, 0.2148800492286682], "cos_Ac_negg": [0.20044180750846863, 0.2280893176794052, 0.2148800492286682], "r_norm": [0.13999506831169128, 0.13813723623752594, 0.1367495357990265]}, {"step": 4800, "epoch": 10, "train_loss": 0.0010178547818213701, "cos_r_negg": [0.18150821328163147, 0.20863491296768188, 0.19851279258728027], "cos_apical_negg": [0.18150821328163147, 0.20863491296768188, 0.19851279258728027], "cos_Ac_negg": [0.18150821328163147, 0.20863491296768188, 0.19851279258728027], "r_norm": [0.15259207785129547, 0.150788351893425, 0.14868605136871338]}, {"step": 4900, "epoch": 10, "train_loss": 0.001019246643409133, "cos_r_negg": [0.1805490404367447, 0.2097999006509781, 0.19830119609832764], "cos_apical_negg": [0.1805490404367447, 0.2097999006509781, 0.19830119609832764], "cos_Ac_negg": [0.1805490404367447, 0.2097999006509781, 0.19830119609832764], "r_norm": [0.14551062881946564, 0.14358925819396973, 0.14128731191158295]}, {"step": 5000, "epoch": 10, "train_loss": 0.0010785356862470508, "cos_r_negg": [0.1676468402147293, 0.19102786481380463, 0.18200671672821045], "cos_apical_negg": [0.1676468402147293, 0.19102786481380463, 0.18200671672821045], "cos_Ac_negg": [0.1676468402147293, 0.19102786481380463, 0.18200671672821045], "r_norm": [0.13078957796096802, 0.12892042100429535, 0.12751910090446472], "ldr": {"dL_sdil": -0.004710696637630463, "dL_bp": -0.007292650640010834, "ratio": 0.6459512281838122}}, {"step": 5100, "epoch": 10, "train_loss": 0.0011707834200933576, "cos_r_negg": [0.16800034046173096, 0.1939515322446823, 0.18445813655853271], "cos_apical_negg": [0.16800034046173096, 0.1939515322446823, 0.18445813655853271], "cos_Ac_negg": [0.16800034046173096, 0.1939515322446823, 0.18445813655853271], "r_norm": [0.13529300689697266, 0.133498877286911, 0.1317964643239975]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9726, "test_loss": 0.13503065013885499}, {"step": 5200, "epoch": 11, "train_loss": 0.0011789256241172552, "cos_r_negg": [0.1572558730840683, 0.18001487851142883, 0.17088842391967773], "cos_apical_negg": [0.1572558730840683, 0.18001487851142883, 0.17088842391967773], "cos_Ac_negg": [0.1572558730840683, 0.18001487851142883, 0.17088842391967773], "r_norm": [0.13868428766727448, 0.13704298436641693, 0.13511855900287628]}, {"step": 5300, "epoch": 11, "train_loss": 0.0012689412105828524, "cos_r_negg": [0.16087812185287476, 0.18432626128196716, 0.17572149634361267], "cos_apical_negg": [0.16087812185287476, 0.18432626128196716, 0.17572149634361267], "cos_Ac_negg": [0.16087812185287476, 0.18432626128196716, 0.17572149634361267], "r_norm": [0.1293502151966095, 0.12786594033241272, 0.12634214758872986]}, {"step": 5400, "epoch": 11, "train_loss": 0.0007522239466197789, "cos_r_negg": [0.15411823987960815, 0.17842479050159454, 0.16960319876670837], "cos_apical_negg": [0.15411823987960815, 0.17842479050159454, 0.16960319876670837], "cos_Ac_negg": [0.15411823987960815, 0.17842479050159454, 0.16960319876670837], "r_norm": [0.13768011331558228, 0.13615527749061584, 0.1340952068567276]}, {"step": 5500, "epoch": 11, "train_loss": 0.002217905130237341, "cos_r_negg": [0.15328368544578552, 0.17673790454864502, 0.16894632577896118], "cos_apical_negg": [0.15328368544578552, 0.17673790454864502, 0.16894632577896118], "cos_Ac_negg": [0.15328368544578552, 0.17673790454864502, 0.16894632577896118], "r_norm": [0.12865038216114044, 0.12704630196094513, 0.12513038516044617], "ldr": {"dL_sdil": -0.005985930562019348, "dL_bp": -0.009874477982521057, "ratio": 0.6062022288788452}}, {"step": 5600, "epoch": 11, "train_loss": 0.0008059403044171631, "cos_r_negg": [0.14735251665115356, 0.1706986427307129, 0.16230052709579468], "cos_apical_negg": [0.14735251665115356, 0.1706986427307129, 0.16230052709579468], "cos_Ac_negg": [0.14735251665115356, 0.1706986427307129, 0.16230052709579468], "r_norm": [0.1310221403837204, 0.12954476475715637, 0.12770068645477295]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9729, "test_loss": 0.14027503452301027}, {"step": 5700, "epoch": 12, "train_loss": 0.00024155571009032428, "cos_r_negg": [0.14655159413814545, 0.17031948268413544, 0.16324445605278015], "cos_apical_negg": [0.14655159413814545, 0.17031948268413544, 0.16324445605278015], "cos_Ac_negg": [0.14655159413814545, 0.17031948268413544, 0.16324445605278015], "r_norm": [0.13277700543403625, 0.13118377327919006, 0.12941528856754303]}, {"step": 5800, "epoch": 12, "train_loss": 0.00038036395562812686, "cos_r_negg": [0.1454523503780365, 0.16926605999469757, 0.16230905055999756], "cos_apical_negg": [0.1454523503780365, 0.16926605999469757, 0.16230905055999756], "cos_Ac_negg": [0.1454523503780365, 0.16926605999469757, 0.16230905055999756], "r_norm": [0.1338832974433899, 0.13205818831920624, 0.12990295886993408]}, {"step": 5900, "epoch": 12, "train_loss": 0.001817264361307025, "cos_r_negg": [0.14318090677261353, 0.165847510099411, 0.15835271775722504], "cos_apical_negg": [0.14318090677261353, 0.165847510099411, 0.15835271775722504], "cos_Ac_negg": [0.14318090677261353, 0.165847510099411, 0.15835271775722504], "r_norm": [0.13398057222366333, 0.13230293989181519, 0.13060159981250763]}, {"step": 6000, "epoch": 12, "train_loss": 0.0019937404431402683, "cos_r_negg": [0.14001017808914185, 0.16120645403862, 0.1555197834968567], "cos_apical_negg": [0.14001017808914185, 0.16120645403862, 0.1555197834968567], "cos_Ac_negg": [0.14001017808914185, 0.16120645403862, 0.1555197834968567], "r_norm": [0.1423790156841278, 0.14043937623500824, 0.1384449005126953], "ldr": {"dL_sdil": -0.0069742351770401, "dL_bp": -0.011977218091487885, "ratio": 0.582291741184594}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9732, "test_loss": 0.14210308151245118}, {"step": 6100, "epoch": 13, "train_loss": 0.0017001167871057987, "cos_r_negg": [0.1384393870830536, 0.15989845991134644, 0.15438809990882874], "cos_apical_negg": [0.1384393870830536, 0.15989845991134644, 0.15438809990882874], "cos_Ac_negg": [0.1384393870830536, 0.15989845991134644, 0.15438809990882874], "r_norm": [0.1407199203968048, 0.13882631063461304, 0.13691718876361847]}, {"step": 6200, "epoch": 13, "train_loss": 0.00037529002293013036, "cos_r_negg": [0.13541299104690552, 0.1572113186120987, 0.1518983691930771], "cos_apical_negg": [0.13541299104690552, 0.1572113186120987, 0.1518983691930771], "cos_Ac_negg": [0.13541299104690552, 0.1572113186120987, 0.1518983691930771], "r_norm": [0.13515688478946686, 0.13340508937835693, 0.13153940439224243]}, {"step": 6300, "epoch": 13, "train_loss": 0.00033519911812618375, "cos_r_negg": [0.13752731680870056, 0.15933454036712646, 0.1538372039794922], "cos_apical_negg": [0.13752731680870056, 0.15933454036712646, 0.1538372039794922], "cos_Ac_negg": [0.13752731680870056, 0.15933454036712646, 0.1538372039794922], "r_norm": [0.13734742999076843, 0.13553787767887115, 0.13386204838752747]}, {"step": 6400, "epoch": 13, "train_loss": 0.0003998060419689864, "cos_r_negg": [0.13656961917877197, 0.15909065306186676, 0.1524784117937088], "cos_apical_negg": [0.13656961917877197, 0.15909065306186676, 0.1524784117937088], "cos_Ac_negg": [0.13656961917877197, 0.15909065306186676, 0.1524784117937088], "r_norm": [0.13284258544445038, 0.1312873661518097, 0.12953701615333557]}, {"step": 6500, "epoch": 13, "train_loss": 0.00031235706410370767, "cos_r_negg": [0.13381841778755188, 0.15428520739078522, 0.14882761240005493], "cos_apical_negg": [0.13381841778755188, 0.15428520739078522, 0.14882761240005493], "cos_Ac_negg": [0.13381841778755188, 0.15428520739078522, 0.14882761240005493], "r_norm": [0.13349714875221252, 0.1318046897649765, 0.13015931844711304], "ldr": {"dL_sdil": -0.006666526198387146, "dL_bp": -0.011768236756324768, "ratio": 0.5664847110425665}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9736, "test_loss": 0.1460468635559082}, {"step": 6600, "epoch": 14, "train_loss": 0.0015140974428504705, "cos_r_negg": [0.13230784237384796, 0.15244686603546143, 0.14781415462493896], "cos_apical_negg": [0.13230784237384796, 0.15244686603546143, 0.14781415462493896], "cos_Ac_negg": [0.13230784237384796, 0.15244686603546143, 0.14781415462493896], "r_norm": [0.1308879405260086, 0.12910108268260956, 0.12746359407901764]}, {"step": 6700, "epoch": 14, "train_loss": 0.0013997508212924004, "cos_r_negg": [0.12985539436340332, 0.14975348114967346, 0.14552007615566254], "cos_apical_negg": [0.12985539436340332, 0.14975348114967346, 0.14552007615566254], "cos_Ac_negg": [0.12985539436340332, 0.14975348114967346, 0.14552007615566254], "r_norm": [0.13147181272506714, 0.12963689863681793, 0.1280984878540039]}, {"step": 6800, "epoch": 14, "train_loss": 0.0015433531953021884, "cos_r_negg": [0.12913398444652557, 0.14907878637313843, 0.14442747831344604], "cos_apical_negg": [0.12913398444652557, 0.14907878637313843, 0.14442747831344604], "cos_Ac_negg": [0.12913398444652557, 0.14907878637313843, 0.14442747831344604], "r_norm": [0.13564874231815338, 0.13395941257476807, 0.13222013413906097]}, {"step": 6900, "epoch": 14, "train_loss": 0.0010377462022006512, "cos_r_negg": [0.12653705477714539, 0.14591452479362488, 0.14195674657821655], "cos_apical_negg": [0.12653705477714539, 0.14591452479362488, 0.14195674657821655], "cos_Ac_negg": [0.12653705477714539, 0.14591452479362488, 0.14195674657821655], "r_norm": [0.13570146262645721, 0.13370797038078308, 0.1319391280412674]}, {"step": 7000, "epoch": 14, "train_loss": 0.00036846072180196643, "cos_r_negg": [0.12514552474021912, 0.14422184228897095, 0.13999950885772705], "cos_apical_negg": [0.12514552474021912, 0.14422184228897095, 0.13999950885772705], "cos_Ac_negg": [0.12514552474021912, 0.14422184228897095, 0.13999950885772705], "r_norm": [0.1375933289527893, 0.13558220863342285, 0.1339113861322403], "ldr": {"dL_sdil": -0.006832651793956757, "dL_bp": -0.012336961925029755, "ratio": 0.5538358499829995}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9731, "test_loss": 0.14842316989898682}], "final": {"test_acc": 0.9731, "test_loss": 0.14842316989898682, "wall_s": 37.175920248031616, "cos_r_negg": [0.12640921771526337, 0.1457367241382599, 0.141432985663414], "cos_apical_negg": [0.12640921771526337, 0.1457367241382599, 0.141432985663414], "cos_Ac_negg": [0.12640921771526337, 0.1457367241382599, 0.141432985663414]}}