1
|
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 0, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.2, "predictor_mode": "diagonal", "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 4, "device": "cuda", "log_every": 100, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "nuis_ctrl_v1_rho0p2_raw_s4", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4633917808532715, "cos_r_negg": [-0.011956857517361641, 0.0204080268740654, 0.002555444138124585], "cos_innovation_negg": [-0.010422075167298317, 0.02403554506599903, 0.006981857120990753], "cos_apical_negg": [-0.011956857517361641, 0.0204080268740654, 0.002555444138124585], "cos_Ac_negg": [-0.010422931984066963, 0.024035122245550156, 0.006981394253671169], "r_norm": [5.565400123596191, 5.211086273193359, 5.183325290679932], "ldr": {"dL_sdil": 0.013530254364013672, "dL_bp": -0.2579684257507324, "ratio": -0.052449265155758144}}, {"step": 100, "epoch": 0, "train_loss": 0.4750537872314453, "cos_r_negg": [0.17289072275161743, 0.2665292024612427, 0.574877142906189], "cos_innovation_negg": [0.4410594701766968, 0.45542776584625244, 0.6422987580299377], "cos_apical_negg": [0.17289072275161743, 0.2665292024612427, 0.574877142906189], "cos_Ac_negg": [0.4392107129096985, 0.45346736907958984, 0.6418181657791138], "r_norm": [3.9450879096984863, 3.9541139602661133, 3.9377858638763428]}, {"step": 200, "epoch": 0, "train_loss": 0.42211389541625977, "cos_r_negg": [0.1335192322731018, 0.19010284543037415, 0.6028484106063843], "cos_innovation_negg": [0.4481812119483948, 0.4370300769805908, 0.6816461086273193], "cos_apical_negg": [0.1335192322731018, 0.19010284543037415, 0.6028484106063843], "cos_Ac_negg": [0.44762662053108215, 0.43662548065185547, 0.6815295219421387], "r_norm": [3.7121119499206543, 3.7272558212280273, 3.7217602729797363]}, {"step": 300, "epoch": 0, "train_loss": 0.4736926853656769, "cos_r_negg": [0.13582423329353333, 0.1760597825050354, 0.5747314691543579], "cos_innovation_negg": [0.4511204957962036, 0.4937560260295868, 0.7376217842102051], "cos_apical_negg": [0.13582423329353333, 0.1760597825050354, 0.5747314691543579], "cos_Ac_negg": [0.4509219527244568, 0.49370697140693665, 0.7374743223190308], "r_norm": [3.64347767829895, 3.666696071624756, 3.659895896911621]}, {"step": 400, "epoch": 0, "train_loss": 0.3019845187664032, "cos_r_negg": [0.15092137455940247, 0.2851549983024597, 0.622618556022644], "cos_innovation_negg": [0.49672600626945496, 0.5648658275604248, 0.8030319213867188], "cos_apical_negg": [0.15092137455940247, 0.2851549983024597, 0.622618556022644], "cos_Ac_negg": [0.4966853857040405, 0.5648783445358276, 0.8029477000236511], "r_norm": [3.649062395095825, 3.6652636528015137, 3.6506729125976562]}, {"epoch_end": 0, "step": 469, "test_acc": 0.9203, "test_loss": 0.3187363037109375}, {"step": 500, "epoch": 1, "train_loss": 0.35978561639785767, "cos_r_negg": [0.11847931146621704, 0.2364443838596344, 0.4773114323616028], "cos_innovation_negg": [0.48284822702407837, 0.5253810882568359, 0.6845443248748779], "cos_apical_negg": [0.11847931146621704, 0.2364443838596344, 0.4773114323616028], "cos_Ac_negg": [0.4824906587600708, 0.5247441530227661, 0.6843276023864746], "r_norm": [3.6327996253967285, 3.647944450378418, 3.641751766204834], "ldr": {"dL_sdil": -0.011537283658981323, "dL_bp": -0.012406736612319946, "ratio": 0.9299208985805943}}, {"step": 600, "epoch": 1, "train_loss": 0.19335171580314636, "cos_r_negg": [0.13319805264472961, 0.23985205590724945, 0.5514774322509766], "cos_innovation_negg": [0.4706363081932068, 0.5092085003852844, 0.7829578518867493], "cos_apical_negg": [0.13319805264472961, 0.23985205590724945, 0.5514774322509766], "cos_Ac_negg": [0.47049611806869507, 0.5090044736862183, 0.7824071049690247], "r_norm": [3.580817222595215, 3.606771469116211, 3.6017746925354004]}, {"step": 700, "epoch": 1, "train_loss": 0.27263277769088745, "cos_r_negg": [0.12088846415281296, 0.2413065880537033, 0.67475825548172], "cos_innovation_negg": [0.48571228981018066, 0.5477913022041321, 0.8230524063110352], "cos_apical_negg": [0.12088846415281296, 0.2413065880537033, 0.67475825548172], "cos_Ac_negg": [0.48568645119667053, 0.5474509000778198, 0.8227138519287109], "r_norm": [3.570188283920288, 3.5988240242004395, 3.579251766204834]}, {"step": 800, "epoch": 1, "train_loss": 0.1933661699295044, "cos_r_negg": [0.09102176874876022, 0.2825019359588623, 0.6548992991447449], "cos_innovation_negg": [0.4500444531440735, 0.5296932458877563, 0.7932925224304199], "cos_apical_negg": [0.09102176874876022, 0.2825019359588623, 0.6548992991447449], "cos_Ac_negg": [0.449919730424881, 0.5296264290809631, 0.7931093573570251], "r_norm": [3.5590052604675293, 3.581848621368408, 3.5757079124450684]}, {"step": 900, "epoch": 1, "train_loss": 0.3662659227848053, "cos_r_negg": [0.08493069559335709, 0.209713414311409, 0.5847129225730896], "cos_innovation_negg": [0.4701869785785675, 0.5524070262908936, 0.8087437152862549], "cos_apical_negg": [0.08493069559335709, 0.209713414311409, 0.5847129225730896], "cos_Ac_negg": [0.4702374339103699, 0.5523287653923035, 0.8087341785430908], "r_norm": [3.5570943355560303, 3.563816547393799, 3.5509824752807617]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9387, "test_loss": 0.23808800735473634}, {"step": 1000, "epoch": 2, "train_loss": 0.12149813771247864, "cos_r_negg": [0.13127610087394714, 0.25622662901878357, 0.6465858817100525], "cos_innovation_negg": [0.4869597852230072, 0.5245695114135742, 0.8300520777702332], "cos_apical_negg": [0.13127610087394714, 0.25622662901878357, 0.6465858817100525], "cos_Ac_negg": [0.48692601919174194, 0.5246900916099548, 0.8300127983093262], "r_norm": [3.572998046875, 3.5996079444885254, 3.5870606899261475], "ldr": {"dL_sdil": -0.010308444499969482, "dL_bp": -0.011464014649391174, "ratio": 0.8992002204495559}}, {"step": 1100, "epoch": 2, "train_loss": 0.33518916368484497, "cos_r_negg": [0.08790959417819977, 0.2497953474521637, 0.5649596452713013], "cos_innovation_negg": [0.434362530708313, 0.5430561304092407, 0.7986395359039307], "cos_apical_negg": [0.08790959417819977, 0.2497953474521637, 0.5649596452713013], "cos_Ac_negg": [0.4341791272163391, 0.5426937937736511, 0.7985544204711914], "r_norm": [3.540452003479004, 3.5672266483306885, 3.5559024810791016]}, {"step": 1200, "epoch": 2, "train_loss": 0.37320321798324585, "cos_r_negg": [0.12525154650211334, 0.2557481527328491, 0.5923508405685425], "cos_innovation_negg": [0.4628221392631531, 0.5563222169876099, 0.8131669759750366], "cos_apical_negg": [0.12525154650211334, 0.2557481527328491, 0.5923508405685425], "cos_Ac_negg": [0.46274644136428833, 0.5564808249473572, 0.8130512833595276], "r_norm": [3.544528007507324, 3.546799659729004, 3.547952175140381]}, {"step": 1300, "epoch": 2, "train_loss": 0.2107192575931549, "cos_r_negg": [0.057930342853069305, 0.2530653476715088, 0.5829254388809204], "cos_innovation_negg": [0.450573205947876, 0.5354846715927124, 0.8413370847702026], "cos_apical_negg": [0.057930342853069305, 0.2530653476715088, 0.5829254388809204], "cos_Ac_negg": [0.45045751333236694, 0.5354177951812744, 0.8411997556686401], "r_norm": [3.5318074226379395, 3.535956859588623, 3.531754493713379]}, {"step": 1400, "epoch": 2, "train_loss": 0.26081418991088867, "cos_r_negg": [0.05684400349855423, 0.20282453298568726, 0.6184252500534058], "cos_innovation_negg": [0.4863371253013611, 0.5721842646598816, 0.8038393259048462], "cos_apical_negg": [0.05684400349855423, 0.20282453298568726, 0.6184252500534058], "cos_Ac_negg": [0.48632127046585083, 0.5720475912094116, 0.8036677837371826], "r_norm": [3.508273124694824, 3.528763771057129, 3.511690139770508]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9449, "test_loss": 0.21544832916259765}, {"step": 1500, "epoch": 3, "train_loss": 0.22312194108963013, "cos_r_negg": [0.05475308746099472, 0.18134787678718567, 0.5367257595062256], "cos_innovation_negg": [0.3670569062232971, 0.46349018812179565, 0.7474660277366638], "cos_apical_negg": [0.05475308746099472, 0.18134787678718567, 0.5367257595062256], "cos_Ac_negg": [0.3670521080493927, 0.4633214473724365, 0.7475631237030029], "r_norm": [3.5198721885681152, 3.535693645477295, 3.5233917236328125], "ldr": {"dL_sdil": -0.0068939924240112305, "dL_bp": -0.007081389427185059, "ratio": 0.9735366900661583}}, {"step": 1600, "epoch": 3, "train_loss": 0.21695445477962494, "cos_r_negg": [0.0818386971950531, 0.1643141508102417, 0.5312960147857666], "cos_innovation_negg": [0.43134045600891113, 0.5168430209159851, 0.7232449054718018], "cos_apical_negg": [0.0818386971950531, 0.1643141508102417, 0.5312960147857666], "cos_Ac_negg": [0.43129485845565796, 0.5169353485107422, 0.7230604290962219], "r_norm": [3.5030550956726074, 3.524172306060791, 3.5076117515563965]}, {"step": 1700, "epoch": 3, "train_loss": 0.14829111099243164, "cos_r_negg": [0.05950741469860077, 0.23633649945259094, 0.5927807092666626], "cos_innovation_negg": [0.47093746066093445, 0.5254899263381958, 0.7582101821899414], "cos_apical_negg": [0.05950741469860077, 0.23633649945259094, 0.5927807092666626], "cos_Ac_negg": [0.47104009985923767, 0.5257368087768555, 0.758141279220581], "r_norm": [3.506467819213867, 3.5251879692077637, 3.5080466270446777]}, {"step": 1800, "epoch": 3, "train_loss": 0.08678125590085983, "cos_r_negg": [0.055664125829935074, 0.19516195356845856, 0.5946141481399536], "cos_innovation_negg": [0.44033345580101013, 0.5145342350006104, 0.7555736899375916], "cos_apical_negg": [0.055664125829935074, 0.19516195356845856, 0.5946141481399536], "cos_Ac_negg": [0.4403555393218994, 0.5140544176101685, 0.7554497718811035], "r_norm": [3.530879020690918, 3.5565037727355957, 3.546597480773926]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9448, "test_loss": 0.23055702514648438}, {"step": 1900, "epoch": 4, "train_loss": 0.17653653025627136, "cos_r_negg": [0.09985731542110443, 0.249653622508049, 0.6090008020401001], "cos_innovation_negg": [0.4281187355518341, 0.5433008670806885, 0.726780354976654], "cos_apical_negg": [0.09985731542110443, 0.249653622508049, 0.6090008020401001], "cos_Ac_negg": [0.42824316024780273, 0.5437741875648499, 0.7255607843399048], "r_norm": [3.5450282096862793, 3.549558639526367, 3.5364203453063965]}, {"step": 2000, "epoch": 4, "train_loss": 0.21220366656780243, "cos_r_negg": [0.07960894703865051, 0.22611437737941742, 0.568345308303833], "cos_innovation_negg": [0.487714022397995, 0.5632803440093994, 0.8075501918792725], "cos_apical_negg": [0.07960894703865051, 0.22611437737941742, 0.568345308303833], "cos_Ac_negg": [0.4876750707626343, 0.5631519556045532, 0.8075096607208252], "r_norm": [3.4925031661987305, 3.5150599479675293, 3.4987144470214844], "ldr": {"dL_sdil": -0.009077027440071106, "dL_bp": -0.010187923908233643, "ratio": 0.8909594851543075}}, {"step": 2100, "epoch": 4, "train_loss": 0.08921875059604645, "cos_r_negg": [0.0939435362815857, 0.22298038005828857, 0.5617860555648804], "cos_innovation_negg": [0.4474589228630066, 0.5350453853607178, 0.8209345936775208], "cos_apical_negg": [0.0939435362815857, 0.22298038005828857, 0.5617860555648804], "cos_Ac_negg": [0.44751042127609253, 0.5347714424133301, 0.8209210634231567], "r_norm": [3.4979195594787598, 3.5182509422302246, 3.497817277908325]}, {"step": 2200, "epoch": 4, "train_loss": 0.11395369470119476, "cos_r_negg": [0.0822942778468132, 0.15268519520759583, 0.6701452732086182], "cos_innovation_negg": [0.3979695737361908, 0.4915057420730591, 0.8444207906723022], "cos_apical_negg": [0.0822942778468132, 0.15268519520759583, 0.6701452732086182], "cos_Ac_negg": [0.39813756942749023, 0.4918029010295868, 0.8443400859832764], "r_norm": [3.507755756378174, 3.515976905822754, 3.5087151527404785]}, {"step": 2300, "epoch": 4, "train_loss": 0.3364529609680176, "cos_r_negg": [0.0770307183265686, 0.21694979071617126, 0.6202293634414673], "cos_innovation_negg": [0.45409858226776123, 0.5880554914474487, 0.8336028456687927], "cos_apical_negg": [0.0770307183265686, 0.21694979071617126, 0.6202293634414673], "cos_Ac_negg": [0.4541519582271576, 0.5879000425338745, 0.8335167169570923], "r_norm": [3.4856796264648438, 3.501509666442871, 3.4909958839416504]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9479, "test_loss": 0.21531006774902345}, {"step": 2400, "epoch": 5, "train_loss": 0.22268988192081451, "cos_r_negg": [0.05968717485666275, 0.22637633979320526, 0.5694169402122498], "cos_innovation_negg": [0.43514344096183777, 0.5774223804473877, 0.7462446689605713], "cos_apical_negg": [0.05968717485666275, 0.22637633979320526, 0.5694169402122498], "cos_Ac_negg": [0.4350615441799164, 0.5765446424484253, 0.7457256317138672], "r_norm": [3.4932656288146973, 3.514101028442383, 3.500938892364502]}, {"step": 2500, "epoch": 5, "train_loss": 0.09341999888420105, "cos_r_negg": [0.019569875672459602, 0.19606637954711914, 0.4754149317741394], "cos_innovation_negg": [0.44084155559539795, 0.5889208316802979, 0.7859771251678467], "cos_apical_negg": [0.019569875672459602, 0.19606637954711914, 0.4754149317741394], "cos_Ac_negg": [0.44089609384536743, 0.5888198614120483, 0.7858589887619019], "r_norm": [3.501614570617676, 3.513394832611084, 3.5061142444610596], "ldr": {"dL_sdil": -0.009760826826095581, "dL_bp": -0.012443989515304565, "ratio": 0.7843808301261402}}, {"step": 2600, "epoch": 5, "train_loss": 0.17552168667316437, "cos_r_negg": [0.047727566212415695, 0.2561088502407074, 0.5877599120140076], "cos_innovation_negg": [0.39532262086868286, 0.5556920766830444, 0.7730623483657837], "cos_apical_negg": [0.047727566212415695, 0.2561088502407074, 0.5877599120140076], "cos_Ac_negg": [0.39529043436050415, 0.5555726289749146, 0.7726271152496338], "r_norm": [3.540367603302002, 3.5505189895629883, 3.543520927429199]}, {"step": 2700, "epoch": 5, "train_loss": 0.16440460085868835, "cos_r_negg": [0.08612929284572601, 0.22055748105049133, 0.6348462104797363], "cos_innovation_negg": [0.4664488434791565, 0.5508244037628174, 0.8962677717208862], "cos_apical_negg": [0.08612929284572601, 0.22055748105049133, 0.6348462104797363], "cos_Ac_negg": [0.4664746820926666, 0.550666332244873, 0.8961828351020813], "r_norm": [3.489314079284668, 3.5017409324645996, 3.4896774291992188]}, {"step": 2800, "epoch": 5, "train_loss": 0.29489392042160034, "cos_r_negg": [0.02368943765759468, 0.20780526101589203, 0.690775990486145], "cos_innovation_negg": [0.4487180709838867, 0.5598125457763672, 0.8652390241622925], "cos_apical_negg": [0.02368943765759468, 0.20780526101589203, 0.690775990486145], "cos_Ac_negg": [0.44904375076293945, 0.559325635433197, 0.8650967478752136], "r_norm": [3.4854655265808105, 3.5011043548583984, 3.488288402557373]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9493, "test_loss": 0.2088368148803711}, {"step": 2900, "epoch": 6, "train_loss": 0.08167441934347153, "cos_r_negg": [0.037279363721609116, 0.17545337975025177, 0.6865703463554382], "cos_innovation_negg": [0.41345784068107605, 0.5443288087844849, 0.8742119669914246], "cos_apical_negg": [0.037279363721609116, 0.17545337975025177, 0.6865703463554382], "cos_Ac_negg": [0.41342607140541077, 0.5439401865005493, 0.8740032911300659], "r_norm": [3.5052995681762695, 3.5192766189575195, 3.509443759918213]}, {"step": 3000, "epoch": 6, "train_loss": 0.42287296056747437, "cos_r_negg": [0.06268949806690216, 0.19135241210460663, 0.583855390548706], "cos_innovation_negg": [0.4322707951068878, 0.5566923022270203, 0.7725169062614441], "cos_apical_negg": [0.06268949806690216, 0.19135241210460663, 0.583855390548706], "cos_Ac_negg": [0.43265223503112793, 0.5564652681350708, 0.7723747491836548], "r_norm": [3.4727225303649902, 3.491758346557617, 3.4828524589538574], "ldr": {"dL_sdil": -0.007038027048110962, "dL_bp": -0.009097367525100708, "ratio": 0.773633364673046}}, {"step": 3100, "epoch": 6, "train_loss": 0.2642550468444824, "cos_r_negg": [0.01877940446138382, 0.2143135517835617, 0.557964563369751], "cos_innovation_negg": [0.4420211911201477, 0.5023664236068726, 0.8279370069503784], "cos_apical_negg": [0.01877940446138382, 0.2143135517835617, 0.557964563369751], "cos_Ac_negg": [0.44218194484710693, 0.5018340945243835, 0.8276281356811523], "r_norm": [3.5031251907348633, 3.5174241065979004, 3.5114688873291016]}, {"step": 3200, "epoch": 6, "train_loss": 0.21465060114860535, "cos_r_negg": [0.017408858984708786, 0.23596268892288208, 0.59035325050354], "cos_innovation_negg": [0.46481838822364807, 0.5952562093734741, 0.8424605131149292], "cos_apical_negg": [0.017408858984708786, 0.23596268892288208, 0.59035325050354], "cos_Ac_negg": [0.46495985984802246, 0.5914040207862854, 0.8423736095428467], "r_norm": [3.467249631881714, 3.4835939407348633, 3.471576690673828]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9511, "test_loss": 0.18839529342651368}, {"step": 3300, "epoch": 7, "train_loss": 0.13581418991088867, "cos_r_negg": [0.04182920604944229, 0.20473960041999817, 0.6690241098403931], "cos_innovation_negg": [0.4543980360031128, 0.5877878665924072, 0.850616455078125], "cos_apical_negg": [0.04182920604944229, 0.20473960041999817, 0.6690241098403931], "cos_Ac_negg": [0.45425844192504883, 0.5874686241149902, 0.8505607843399048], "r_norm": [3.4948437213897705, 3.496196985244751, 3.4916493892669678]}, {"step": 3400, "epoch": 7, "train_loss": 0.30225008726119995, "cos_r_negg": [0.04843023419380188, 0.18645831942558289, 0.6739591360092163], "cos_innovation_negg": [0.43378737568855286, 0.5751627087593079, 0.7943737506866455], "cos_apical_negg": [0.04843023419380188, 0.18645831942558289, 0.6739591360092163], "cos_Ac_negg": [0.43377113342285156, 0.5749982595443726, 0.7943326234817505], "r_norm": [3.4892115592956543, 3.4975619316101074, 3.4930148124694824]}, {"step": 3500, "epoch": 7, "train_loss": 0.13207051157951355, "cos_r_negg": [0.062053464353084564, 0.18894898891448975, 0.6230188012123108], "cos_innovation_negg": [0.426882803440094, 0.5420725345611572, 0.7607810497283936], "cos_apical_negg": [0.062053464353084564, 0.18894898891448975, 0.6230188012123108], "cos_Ac_negg": [0.4268535375595093, 0.5410705804824829, 0.7601210474967957], "r_norm": [3.4732272624969482, 3.4815917015075684, 3.4748454093933105], "ldr": {"dL_sdil": -0.005243748426437378, "dL_bp": -0.008037254214286804, "ratio": 0.6524303308854202}}, {"step": 3600, "epoch": 7, "train_loss": 0.05919833481311798, "cos_r_negg": [0.07355920970439911, 0.22196289896965027, 0.5801942348480225], "cos_innovation_negg": [0.3847639858722687, 0.5737966299057007, 0.8547283411026001], "cos_apical_negg": [0.07355920970439911, 0.22196289896965027, 0.5801942348480225], "cos_Ac_negg": [0.3846800923347473, 0.5735874176025391, 0.8545604944229126], "r_norm": [3.475571393966675, 3.490208387374878, 3.476595878601074]}, {"step": 3700, "epoch": 7, "train_loss": 0.13432708382606506, "cos_r_negg": [0.03509465605020523, 0.2035141885280609, 0.4782654941082001], "cos_innovation_negg": [0.4349420368671417, 0.5155513882637024, 0.8384146094322205], "cos_apical_negg": [0.03509465605020523, 0.2035141885280609, 0.4782654941082001], "cos_Ac_negg": [0.43494024872779846, 0.5154975056648254, 0.839094340801239], "r_norm": [3.4590110778808594, 3.4725236892700195, 3.4619650840759277]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9533, "test_loss": 0.19848698043823243}, {"step": 3800, "epoch": 8, "train_loss": 0.12582886219024658, "cos_r_negg": [0.047408945858478546, 0.2342541217803955, 0.5734213590621948], "cos_innovation_negg": [0.43895184993743896, 0.5944777727127075, 0.7722045183181763], "cos_apical_negg": [0.047408945858478546, 0.2342541217803955, 0.5734213590621948], "cos_Ac_negg": [0.4389934539794922, 0.5940360426902771, 0.7718954086303711], "r_norm": [3.464224338531494, 3.4798431396484375, 3.4673376083374023]}, {"step": 3900, "epoch": 8, "train_loss": 0.2349289357662201, "cos_r_negg": [0.05087464675307274, 0.20161005854606628, 0.448138564825058], "cos_innovation_negg": [0.4174814224243164, 0.5588624477386475, 0.8307009935379028], "cos_apical_negg": [0.05087464675307274, 0.20161005854606628, 0.448138564825058], "cos_Ac_negg": [0.4174521863460541, 0.5583441257476807, 0.8310527205467224], "r_norm": [3.4680304527282715, 3.4799129962921143, 3.469026565551758]}, {"step": 4000, "epoch": 8, "train_loss": 0.20901921391487122, "cos_r_negg": [0.04730428010225296, 0.26220422983169556, 0.5256604552268982], "cos_innovation_negg": [0.4598146378993988, 0.6259366273880005, 0.8224747180938721], "cos_apical_negg": [0.04730428010225296, 0.26220422983169556, 0.5256604552268982], "cos_Ac_negg": [0.45933812856674194, 0.6254764795303345, 0.8220374584197998], "r_norm": [3.4767117500305176, 3.491227865219116, 3.4723434448242188], "ldr": {"dL_sdil": -0.008212298154830933, "dL_bp": -0.010396763682365417, "ratio": 0.789889854740116}}, {"step": 4100, "epoch": 8, "train_loss": 0.08099746704101562, "cos_r_negg": [0.07924340665340424, 0.205892875790596, 0.425891250371933], "cos_innovation_negg": [0.40887951850891113, 0.5891261100769043, 0.755929172039032], "cos_apical_negg": [0.07924340665340424, 0.205892875790596, 0.425891250371933], "cos_Ac_negg": [0.4087597727775574, 0.589033842086792, 0.7560677528381348], "r_norm": [3.4688241481781006, 3.4768028259277344, 3.4709818363189697]}, {"step": 4200, "epoch": 8, "train_loss": 0.3419545888900757, "cos_r_negg": [0.03113587200641632, 0.1751873791217804, 0.3959196209907532], "cos_innovation_negg": [0.3896183967590332, 0.5375776290893555, 0.6484522223472595], "cos_apical_negg": [0.03113587200641632, 0.1751873791217804, 0.3959196209907532], "cos_Ac_negg": [0.3896007239818573, 0.537344217300415, 0.6576241254806519], "r_norm": [3.488478899002075, 3.492347478866577, 3.489715814590454]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9551, "test_loss": 0.1832194664001465}, {"step": 4300, "epoch": 9, "train_loss": 0.09275281429290771, "cos_r_negg": [0.053618162870407104, 0.19253221154212952, 0.5308492183685303], "cos_innovation_negg": [0.417873740196228, 0.5843667984008789, 0.7965271472930908], "cos_apical_negg": [0.053618162870407104, 0.19253221154212952, 0.5308492183685303], "cos_Ac_negg": [0.41739410161972046, 0.5841286182403564, 0.7967294454574585], "r_norm": [3.476356029510498, 3.486741065979004, 3.473093032836914]}, {"step": 4400, "epoch": 9, "train_loss": 0.07180020958185196, "cos_r_negg": [0.043264977633953094, 0.21635767817497253, 0.490176796913147], "cos_innovation_negg": [0.44119763374328613, 0.571502149105072, 0.759745717048645], "cos_apical_negg": [0.043264977633953094, 0.21635767817497253, 0.490176796913147], "cos_Ac_negg": [0.4413753151893616, 0.5715869665145874, 0.7595043182373047], "r_norm": [3.4643805027008057, 3.4721014499664307, 3.4618239402770996]}, {"step": 4500, "epoch": 9, "train_loss": 0.19398769736289978, "cos_r_negg": [0.03966209292411804, 0.21499177813529968, 0.5362972021102905], "cos_innovation_negg": [0.4309678077697754, 0.6087491512298584, 0.7102959156036377], "cos_apical_negg": [0.03966209292411804, 0.21499177813529968, 0.5362972021102905], "cos_Ac_negg": [0.43100830912590027, 0.6087126731872559, 0.7100602388381958], "r_norm": [3.486661195755005, 3.4934773445129395, 3.477720022201538], "ldr": {"dL_sdil": -0.006987050175666809, "dL_bp": -0.009711220860481262, "ratio": 0.7194821615168733}}, {"step": 4600, "epoch": 9, "train_loss": 0.0560777485370636, "cos_r_negg": [0.06531397998332977, 0.1800817847251892, 0.5799179077148438], "cos_innovation_negg": [0.45342251658439636, 0.5431491136550903, 0.8237646818161011], "cos_apical_negg": [0.06531397998332977, 0.1800817847251892, 0.5799179077148438], "cos_Ac_negg": [0.45336800813674927, 0.5423874258995056, 0.824088454246521], "r_norm": [3.4609766006469727, 3.4702136516571045, 3.459541082382202]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9542, "test_loss": 0.19892877578735352}, {"step": 4700, "epoch": 10, "train_loss": 0.17763805389404297, "cos_r_negg": [0.04531599208712578, 0.21647030115127563, 0.514319896697998], "cos_innovation_negg": [0.4301847815513611, 0.6016221642494202, 0.7524961829185486], "cos_apical_negg": [0.04531599208712578, 0.21647030115127563, 0.514319896697998], "cos_Ac_negg": [0.43001091480255127, 0.6015416383743286, 0.7521877884864807], "r_norm": [3.4750311374664307, 3.488917827606201, 3.480609655380249]}, {"step": 4800, "epoch": 10, "train_loss": 0.141770139336586, "cos_r_negg": [-0.003987464122474194, 0.23780937492847443, 0.4706122875213623], "cos_innovation_negg": [0.386099636554718, 0.6022370457649231, 0.7465347051620483], "cos_apical_negg": [-0.003987464122474194, 0.23780937492847443, 0.4706122875213623], "cos_Ac_negg": [0.38721463084220886, 0.601956307888031, 0.7459104061126709], "r_norm": [3.459686756134033, 3.472586154937744, 3.4658868312835693]}, {"step": 4900, "epoch": 10, "train_loss": 0.07928487658500671, "cos_r_negg": [0.026156213134527206, 0.1351097822189331, 0.3670060634613037], "cos_innovation_negg": [0.4061506390571594, 0.5064048767089844, 0.6853375434875488], "cos_apical_negg": [0.026156213134527206, 0.1351097822189331, 0.3670060634613037], "cos_Ac_negg": [0.4061405062675476, 0.5060154795646667, 0.6849254369735718], "r_norm": [3.463350534439087, 3.473461389541626, 3.4622201919555664]}, {"step": 5000, "epoch": 10, "train_loss": 0.027447305619716644, "cos_r_negg": [0.0516609251499176, 0.12610000371932983, 0.5079575181007385], "cos_innovation_negg": [0.399018257856369, 0.5668156147003174, 0.7164534330368042], "cos_apical_negg": [0.0516609251499176, 0.12610000371932983, 0.5079575181007385], "cos_Ac_negg": [0.399252325296402, 0.5665684938430786, 0.716035783290863], "r_norm": [3.4591786861419678, 3.472064733505249, 3.4581494331359863], "ldr": {"dL_sdil": -0.004805013537406921, "dL_bp": -0.007359206676483154, "ratio": 0.6529254780629642}}, {"step": 5100, "epoch": 10, "train_loss": 0.21249383687973022, "cos_r_negg": [0.031447723507881165, 0.17383599281311035, 0.43652480840682983], "cos_innovation_negg": [0.4170268177986145, 0.5315840244293213, 0.8075850009918213], "cos_apical_negg": [0.031447723507881165, 0.17383599281311035, 0.43652480840682983], "cos_Ac_negg": [0.4170093536376953, 0.5316397547721863, 0.8076450824737549], "r_norm": [3.4762325286865234, 3.4863297939300537, 3.477323055267334]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9559, "test_loss": 0.1883988006591797}, {"step": 5200, "epoch": 11, "train_loss": 0.08720330148935318, "cos_r_negg": [0.039067670702934265, 0.19694335758686066, 0.48266273736953735], "cos_innovation_negg": [0.43498164415359497, 0.5526241064071655, 0.8344366550445557], "cos_apical_negg": [0.039067670702934265, 0.19694335758686066, 0.48266273736953735], "cos_Ac_negg": [0.4349234700202942, 0.5524110794067383, 0.8342981934547424], "r_norm": [3.466118097305298, 3.475189447402954, 3.4649057388305664]}, {"step": 5300, "epoch": 11, "train_loss": 0.1388130933046341, "cos_r_negg": [0.031038496643304825, 0.14290937781333923, 0.5434865951538086], "cos_innovation_negg": [0.4273046851158142, 0.507470965385437, 0.8129115104675293], "cos_apical_negg": [0.031038496643304825, 0.14290937781333923, 0.5434865951538086], "cos_Ac_negg": [0.4272947609424591, 0.507317066192627, 0.8124746680259705], "r_norm": [3.468296527862549, 3.478717088699341, 3.466716766357422]}, {"step": 5400, "epoch": 11, "train_loss": 0.27817630767822266, "cos_r_negg": [0.017896460369229317, 0.1772950440645218, 0.48078250885009766], "cos_innovation_negg": [0.3642120063304901, 0.5362515449523926, 0.704262375831604], "cos_apical_negg": [0.017896460369229317, 0.1772950440645218, 0.48078250885009766], "cos_Ac_negg": [0.3674371838569641, 0.53584885597229, 0.7039588689804077], "r_norm": [3.456787586212158, 3.4661812782287598, 3.449216365814209]}, {"step": 5500, "epoch": 11, "train_loss": 0.20167073607444763, "cos_r_negg": [0.025979511439800262, 0.24400228261947632, 0.4427090287208557], "cos_innovation_negg": [0.3989015817642212, 0.5896512269973755, 0.702342689037323], "cos_apical_negg": [0.025979511439800262, 0.24400228261947632, 0.4427090287208557], "cos_Ac_negg": [0.3991009593009949, 0.5893043279647827, 0.7025435566902161], "r_norm": [3.4555277824401855, 3.4655354022979736, 3.454407215118408], "ldr": {"dL_sdil": -0.004147320985794067, "dL_bp": -0.006773233413696289, "ratio": 0.61231035939315}}, {"step": 5600, "epoch": 11, "train_loss": 0.11804436147212982, "cos_r_negg": [0.049389004707336426, 0.17493867874145508, 0.38279426097869873], "cos_innovation_negg": [0.4189903140068054, 0.5583648681640625, 0.8230645060539246], "cos_apical_negg": [0.049389004707336426, 0.17493867874145508, 0.38279426097869873], "cos_Ac_negg": [0.41872096061706543, 0.5583170652389526, 0.822784423828125], "r_norm": [3.485067367553711, 3.497248888015747, 3.486901044845581]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9546, "test_loss": 0.197890909576416}, {"step": 5700, "epoch": 12, "train_loss": 0.07822814583778381, "cos_r_negg": [0.042551275342702866, 0.2011890709400177, 0.44901496171951294], "cos_innovation_negg": [0.41427987813949585, 0.5747051239013672, 0.7529419660568237], "cos_apical_negg": [0.042551275342702866, 0.2011890709400177, 0.44901496171951294], "cos_Ac_negg": [0.40688616037368774, 0.573483943939209, 0.7563733458518982], "r_norm": [3.4518208503723145, 3.4615461826324463, 3.4539847373962402]}, {"step": 5800, "epoch": 12, "train_loss": 0.043576110154390335, "cos_r_negg": [0.04847339540719986, 0.15626055002212524, 0.4628142714500427], "cos_innovation_negg": [0.4235988259315491, 0.5542389154434204, 0.7477295398712158], "cos_apical_negg": [0.04847339540719986, 0.15626055002212524, 0.4628142714500427], "cos_Ac_negg": [0.4233926236629486, 0.5539628267288208, 0.7543405294418335], "r_norm": [3.4550275802612305, 3.4634969234466553, 3.453120231628418]}, {"step": 5900, "epoch": 12, "train_loss": 0.1297622174024582, "cos_r_negg": [0.018939029425382614, 0.1774609386920929, 0.49215948581695557], "cos_innovation_negg": [0.4190130829811096, 0.5573451519012451, 0.769503116607666], "cos_apical_negg": [0.018939029425382614, 0.1774609386920929, 0.49215948581695557], "cos_Ac_negg": [0.41877099871635437, 0.557030200958252, 0.7704774141311646], "r_norm": [3.4489479064941406, 3.457632541656494, 3.4471538066864014]}, {"step": 6000, "epoch": 12, "train_loss": 0.06887461245059967, "cos_r_negg": [0.022489607334136963, 0.18150678277015686, 0.5400077104568481], "cos_innovation_negg": [0.39491206407546997, 0.5437368154525757, 0.8377511501312256], "cos_apical_negg": [0.022489607334136963, 0.18150678277015686, 0.5400077104568481], "cos_Ac_negg": [0.39466768503189087, 0.5434008240699768, 0.8379849195480347], "r_norm": [3.4569315910339355, 3.4669337272644043, 3.4562478065490723], "ldr": {"dL_sdil": -0.004834681749343872, "dL_bp": -0.007802397012710571, "ratio": 0.6196405721815855}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9541, "test_loss": 0.19915148162841798}, {"step": 6100, "epoch": 13, "train_loss": 0.12441891431808472, "cos_r_negg": [0.037758029997348785, 0.21973922848701477, 0.5656598210334778], "cos_innovation_negg": [0.4033668041229248, 0.5921123027801514, 0.80461186170578], "cos_apical_negg": [0.037758029997348785, 0.21973922848701477, 0.5656598210334778], "cos_Ac_negg": [0.40334028005599976, 0.5919814705848694, 0.804319441318512], "r_norm": [3.448150634765625, 3.4538209438323975, 3.447100877761841]}, {"step": 6200, "epoch": 13, "train_loss": 0.03186078742146492, "cos_r_negg": [0.024448025971651077, 0.18667493760585785, 0.4622846245765686], "cos_innovation_negg": [0.35262954235076904, 0.5385372638702393, 0.7032085657119751], "cos_apical_negg": [0.024448025971651077, 0.18667493760585785, 0.4622846245765686], "cos_Ac_negg": [0.3525902032852173, 0.5382406115531921, 0.7020827531814575], "r_norm": [3.4546518325805664, 3.460237979888916, 3.4550435543060303]}, {"step": 6300, "epoch": 13, "train_loss": 0.03825074061751366, "cos_r_negg": [0.020515820011496544, 0.21622180938720703, 0.44693082571029663], "cos_innovation_negg": [0.4155089259147644, 0.5973267555236816, 0.6766473054885864], "cos_apical_negg": [0.020515820011496544, 0.21622180938720703, 0.44693082571029663], "cos_Ac_negg": [0.41552019119262695, 0.5969445705413818, 0.6782469749450684], "r_norm": [3.448761463165283, 3.455890655517578, 3.4471638202667236]}, {"step": 6400, "epoch": 13, "train_loss": 0.03932960703969002, "cos_r_negg": [0.026593996211886406, 0.2256653904914856, 0.39923524856567383], "cos_innovation_negg": [0.41374319791793823, 0.5692796111106873, 0.80968177318573], "cos_apical_negg": [0.026593996211886406, 0.2256653904914856, 0.39923524856567383], "cos_Ac_negg": [0.4138638973236084, 0.5679604411125183, 0.8093268871307373], "r_norm": [3.4697105884552, 3.482572555541992, 3.4712765216827393]}, {"step": 6500, "epoch": 13, "train_loss": 0.011997969821095467, "cos_r_negg": [0.041799116879701614, 0.12720957398414612, 0.3751963973045349], "cos_innovation_negg": [0.40569421648979187, 0.5459054708480835, 0.8217697143554688], "cos_apical_negg": [0.041799116879701614, 0.12720957398414612, 0.3751963973045349], "cos_Ac_negg": [0.4057433307170868, 0.5467766523361206, 0.8217895030975342], "r_norm": [3.4552578926086426, 3.4635722637176514, 3.454599380493164], "ldr": {"dL_sdil": -0.004309207201004028, "dL_bp": -0.00781385600566864, "ratio": 0.551482801561466}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9544, "test_loss": 0.1913643699645996}, {"step": 6600, "epoch": 14, "train_loss": 0.02708890289068222, "cos_r_negg": [9.342096745967865e-05, 0.13607776165008545, 0.5366811752319336], "cos_innovation_negg": [0.36983031034469604, 0.5505750179290771, 0.818997859954834], "cos_apical_negg": [9.342096745967865e-05, 0.13607776165008545, 0.5366811752319336], "cos_Ac_negg": [0.370279997587204, 0.5495230555534363, 0.8190932273864746], "r_norm": [3.4489996433258057, 3.452692747116089, 3.4436049461364746]}, {"step": 6700, "epoch": 14, "train_loss": 0.08523418009281158, "cos_r_negg": [0.02637433633208275, 0.1678479015827179, 0.5084254741668701], "cos_innovation_negg": [0.4090483784675598, 0.534423828125, 0.7576724290847778], "cos_apical_negg": [0.02637433633208275, 0.1678479015827179, 0.5084254741668701], "cos_Ac_negg": [0.4092547595500946, 0.5366582274436951, 0.7602380514144897], "r_norm": [3.457058906555176, 3.466118097305298, 3.4575705528259277]}, {"step": 6800, "epoch": 14, "train_loss": 0.09282908588647842, "cos_r_negg": [0.020635008811950684, 0.18371716141700745, 0.39045971632003784], "cos_innovation_negg": [0.40185490250587463, 0.5615551471710205, 0.7576956152915955], "cos_apical_negg": [0.020635008811950684, 0.18371716141700745, 0.39045971632003784], "cos_Ac_negg": [0.4020015597343445, 0.5608091354370117, 0.7574525475502014], "r_norm": [3.459503650665283, 3.4631452560424805, 3.455723285675049]}, {"step": 6900, "epoch": 14, "train_loss": 0.07403425872325897, "cos_r_negg": [0.004958726465702057, 0.21029961109161377, 0.4955659508705139], "cos_innovation_negg": [0.4068952202796936, 0.6397120356559753, 0.7893516421318054], "cos_apical_negg": [0.004958726465702057, 0.21029961109161377, 0.4955659508705139], "cos_Ac_negg": [0.40709972381591797, 0.6391990780830383, 0.7891943454742432], "r_norm": [3.4468507766723633, 3.451991319656372, 3.4429473876953125]}, {"step": 7000, "epoch": 14, "train_loss": 0.1459781378507614, "cos_r_negg": [0.006495658308267593, 0.20425459742546082, 0.4501078724861145], "cos_innovation_negg": [0.42327141761779785, 0.6309816837310791, 0.8001081943511963], "cos_apical_negg": [0.006495658308267593, 0.20425459742546082, 0.4501078724861145], "cos_Ac_negg": [0.4234600067138672, 0.6317270994186401, 0.8001537919044495], "r_norm": [3.457538604736328, 3.4639830589294434, 3.45184326171875], "ldr": {"dL_sdil": -0.003236740827560425, "dL_bp": -0.0056077539920806885, "ratio": 0.577190231977254}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9532, "test_loss": 0.2099518783569336}], "final": {"test_acc": 0.9532, "test_loss": 0.2099518783569336, "wall_s": 21.86707639694214, "cos_r_negg": [-0.013040256686508656, 0.19956542551517487, 0.3413127064704895], "cos_innovation_negg": [0.35866662859916687, 0.5476642847061157, 0.7576143145561218], "cos_apical_negg": [-0.013040256686508656, 0.19956542551517487, 0.3413127064704895], "cos_Ac_negg": [0.3587781488895416, 0.5472832918167114, 0.760552167892456]}}
|