1
|
{"args": {"mode": "sdil", "dataset": "mnist", "depth": 3, "width": 256, "act": "tanh", "residual": 0, "epochs": 15, "batch_size": 128, "eta": 0.05, "eta_A": 0.02, "eta_P": 0.002, "momentum": 0.9, "w_scale": 1.0, "a_scale": 1.0, "pert_sigma": 0.01, "pert_every": 4, "pert_ndirs": 16, "pert_mode": "simultaneous", "use_residual": 1, "raw_scale_control": "none", "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.5, "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": "nuis_ctrl_v1_rho0p5_residual_s0", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3603768348693848, "cos_r_negg": [-0.019524920731782913, -0.025532349944114685, 0.033747125416994095], "cos_innovation_negg": [-0.019524920731782913, -0.025532349944114685, 0.033747125416994095], "cos_apical_negg": [-0.01679173856973648, -0.02367609366774559, 0.02439393848180771], "cos_Ac_negg": [-0.019524451345205307, -0.02553233876824379, 0.03374704346060753], "r_norm": [4.89827823638916, 4.826774597167969, 4.823909759521484], "ldr": {"dL_sdil": -0.00920414924621582, "dL_bp": -0.24071121215820312, "ratio": 0.03823731002678246}}, {"step": 100, "epoch": 0, "train_loss": 0.20806199312210083, "cos_r_negg": [0.45580482482910156, 0.37684938311576843, 0.3593536615371704], "cos_innovation_negg": [0.45580482482910156, 0.37684938311576843, 0.3593536615371704], "cos_apical_negg": [0.07743188738822937, 0.08167853951454163, 0.13730955123901367], "cos_Ac_negg": [0.45889896154403687, 0.38018500804901123, 0.361653596162796], "r_norm": [0.8243230581283569, 0.8080554604530334, 0.8057764768600464]}, {"step": 200, "epoch": 0, "train_loss": 0.32909178733825684, "cos_r_negg": [0.4774864912033081, 0.40109360218048096, 0.38381052017211914], "cos_innovation_negg": [0.4774864912033081, 0.40109360218048096, 0.38381052017211914], "cos_apical_negg": [0.06533575057983398, 0.08173279464244843, 0.12964117527008057], "cos_Ac_negg": [0.48442214727401733, 0.40823638439178467, 0.39088404178619385], "r_norm": [0.743273138999939, 0.7259811162948608, 0.7249500155448914]}, {"step": 300, "epoch": 0, "train_loss": 0.17777316272258759, "cos_r_negg": [0.48456960916519165, 0.4558779001235962, 0.42146602272987366], "cos_innovation_negg": [0.48456960916519165, 0.4558779001235962, 0.42146602272987366], "cos_apical_negg": [0.05611689016222954, 0.06970221549272537, 0.11532211303710938], "cos_Ac_negg": [0.4899787902832031, 0.45894038677215576, 0.42386552691459656], "r_norm": [0.5654057264328003, 0.5556600093841553, 0.5548664331436157]}, {"step": 400, "epoch": 0, "train_loss": 0.16122844815254211, "cos_r_negg": [0.5220173597335815, 0.48386672139167786, 0.439599871635437], "cos_innovation_negg": [0.5220173597335815, 0.48386672139167786, 0.439599871635437], "cos_apical_negg": [0.060485437512397766, 0.07159489393234253, 0.11227110773324966], "cos_Ac_negg": [0.5238868594169617, 0.48481935262680054, 0.43951553106307983], "r_norm": [0.5609845519065857, 0.5467572212219238, 0.5457912683486938]}, {"epoch_end": 0, "step": 469, "test_acc": 0.947, "test_loss": 0.18223495864868164}, {"step": 500, "epoch": 1, "train_loss": 0.24727237224578857, "cos_r_negg": [0.5016535520553589, 0.48832446336746216, 0.46044179797172546], "cos_innovation_negg": [0.5016535520553589, 0.48832446336746216, 0.46044179797172546], "cos_apical_negg": [0.05814739316701889, 0.0640498548746109, 0.1180124431848526], "cos_Ac_negg": [0.5116339325904846, 0.4933936893939972, 0.4625237286090851], "r_norm": [0.5076783895492554, 0.4894869327545166, 0.49166637659072876], "ldr": {"dL_sdil": -0.01459231972694397, "dL_bp": -0.01246914267539978, "ratio": 1.1702745013683242}}, {"step": 600, "epoch": 1, "train_loss": 0.09304194897413254, "cos_r_negg": [0.4987279176712036, 0.4869357645511627, 0.45599937438964844], "cos_innovation_negg": [0.4987279176712036, 0.4869357645511627, 0.45599937438964844], "cos_apical_negg": [0.047639235854148865, 0.054375506937503815, 0.11665451526641846], "cos_Ac_negg": [0.5135130882263184, 0.493655800819397, 0.4611470699310303], "r_norm": [0.42564013600349426, 0.4166065454483032, 0.41495829820632935]}, {"step": 700, "epoch": 1, "train_loss": 0.19576694071292877, "cos_r_negg": [0.5116797089576721, 0.5149182081222534, 0.4854634404182434], "cos_innovation_negg": [0.5116797089576721, 0.5149182081222534, 0.4854634404182434], "cos_apical_negg": [0.04140525683760643, 0.06269320100545883, 0.11289717257022858], "cos_Ac_negg": [0.5239114761352539, 0.5192609429359436, 0.49155494570732117], "r_norm": [0.4615369439125061, 0.4475287199020386, 0.4478917717933655]}, {"step": 800, "epoch": 1, "train_loss": 0.13500498235225677, "cos_r_negg": [0.5412785410881042, 0.5338720083236694, 0.49602705240249634], "cos_innovation_negg": [0.5412785410881042, 0.5338720083236694, 0.49602705240249634], "cos_apical_negg": [0.042619746178388596, 0.062221676111221313, 0.09834045171737671], "cos_Ac_negg": [0.5519948601722717, 0.5353431105613708, 0.49804800748825073], "r_norm": [0.36324626207351685, 0.3547770380973816, 0.35415929555892944]}, {"step": 900, "epoch": 1, "train_loss": 0.19931073486804962, "cos_r_negg": [0.5573374629020691, 0.5556252002716064, 0.49063265323638916], "cos_innovation_negg": [0.5573374629020691, 0.5556252002716064, 0.49063265323638916], "cos_apical_negg": [0.04584997147321701, 0.06994988024234772, 0.10097363591194153], "cos_Ac_negg": [0.560693621635437, 0.5552884936332703, 0.488823801279068], "r_norm": [0.43848469853401184, 0.4289107024669647, 0.4265971779823303]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9536, "test_loss": 0.1556456199645996}, {"step": 1000, "epoch": 2, "train_loss": 0.04337725415825844, "cos_r_negg": [0.5406967401504517, 0.5441626906394958, 0.5123414993286133], "cos_innovation_negg": [0.5406967401504517, 0.5441626906394958, 0.5123414993286133], "cos_apical_negg": [0.02406124398112297, 0.052928850054740906, 0.10016977787017822], "cos_Ac_negg": [0.5500516891479492, 0.5479145050048828, 0.513395369052887], "r_norm": [0.3558533787727356, 0.34642890095710754, 0.3454371988773346], "ldr": {"dL_sdil": -0.010897994041442871, "dL_bp": -0.00895872712135315, "ratio": 1.2164667919695282}}, {"step": 1100, "epoch": 2, "train_loss": 0.12945353984832764, "cos_r_negg": [0.540961503982544, 0.547196626663208, 0.5078578591346741], "cos_innovation_negg": [0.540961503982544, 0.547196626663208, 0.5078578591346741], "cos_apical_negg": [0.035796165466308594, 0.048418864607810974, 0.1101682037115097], "cos_Ac_negg": [0.5532693862915039, 0.5568880438804626, 0.5110524892807007], "r_norm": [0.332541286945343, 0.3214121162891388, 0.32319897413253784]}, {"step": 1200, "epoch": 2, "train_loss": 0.12910637259483337, "cos_r_negg": [0.5410082936286926, 0.5529738664627075, 0.5163585543632507], "cos_innovation_negg": [0.5410082936286926, 0.5529738664627075, 0.5163585543632507], "cos_apical_negg": [0.01959778368473053, 0.039587631821632385, 0.11256879568099976], "cos_Ac_negg": [0.5523759126663208, 0.5613055229187012, 0.5173860788345337], "r_norm": [0.3318207263946533, 0.3226848244667053, 0.3240436911582947]}, {"step": 1300, "epoch": 2, "train_loss": 0.10018610954284668, "cos_r_negg": [0.5347980856895447, 0.5633956789970398, 0.5126184225082397], "cos_innovation_negg": [0.5347980856895447, 0.5633956789970398, 0.5126184225082397], "cos_apical_negg": [0.02419588342308998, 0.056053582578897476, 0.10105574876070023], "cos_Ac_negg": [0.5439138412475586, 0.5699706077575684, 0.5123409032821655], "r_norm": [0.32683128118515015, 0.31890031695365906, 0.31658512353897095]}, {"step": 1400, "epoch": 2, "train_loss": 0.1301070749759674, "cos_r_negg": [0.5010162591934204, 0.5402288436889648, 0.5228582620620728], "cos_innovation_negg": [0.5010162591934204, 0.5402288436889648, 0.5228582620620728], "cos_apical_negg": [0.00686885230243206, 0.04561745002865791, 0.10342801362276077], "cos_Ac_negg": [0.5327618718147278, 0.5623195171356201, 0.5346428155899048], "r_norm": [0.2709745168685913, 0.2633535861968994, 0.26226699352264404]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9658, "test_loss": 0.11943423690795898}, {"step": 1500, "epoch": 3, "train_loss": 0.09488650411367416, "cos_r_negg": [0.5477433204650879, 0.5659247040748596, 0.5450236797332764], "cos_innovation_negg": [0.5477433204650879, 0.5659247040748596, 0.5450236797332764], "cos_apical_negg": [0.02507450059056282, 0.04850952327251434, 0.10936501622200012], "cos_Ac_negg": [0.5674236416816711, 0.58188796043396, 0.5530956983566284], "r_norm": [0.25747135281562805, 0.2512120306491852, 0.25033754110336304], "ldr": {"dL_sdil": -0.005412191152572632, "dL_bp": -0.005684912204742432, "ratio": 0.9520272183021063}}, {"step": 1600, "epoch": 3, "train_loss": 0.09669487923383713, "cos_r_negg": [0.5311207175254822, 0.5536269545555115, 0.5428143739700317], "cos_innovation_negg": [0.5311207175254822, 0.5536269545555115, 0.5428143739700317], "cos_apical_negg": [0.028560880571603775, 0.04665552079677582, 0.10565381497144699], "cos_Ac_negg": [0.5533983111381531, 0.5779848098754883, 0.5536564588546753], "r_norm": [0.290454238653183, 0.2804765999317169, 0.282884418964386]}, {"step": 1700, "epoch": 3, "train_loss": 0.046024542301893234, "cos_r_negg": [0.5218465328216553, 0.5565448999404907, 0.5266669988632202], "cos_innovation_negg": [0.5218465328216553, 0.5565448999404907, 0.5266669988632202], "cos_apical_negg": [0.013521810993552208, 0.05554120987653732, 0.10070943832397461], "cos_Ac_negg": [0.5539103746414185, 0.5861827731132507, 0.5446239709854126], "r_norm": [0.2801855206489563, 0.27018672227859497, 0.2713976502418518]}, {"step": 1800, "epoch": 3, "train_loss": 0.011728792451322079, "cos_r_negg": [0.52141273021698, 0.540260910987854, 0.5321792960166931], "cos_innovation_negg": [0.52141273021698, 0.540260910987854, 0.5321792960166931], "cos_apical_negg": [0.020364152267575264, 0.054958440363407135, 0.11829151213169098], "cos_Ac_negg": [0.5664126873016357, 0.5849656462669373, 0.5539586544036865], "r_norm": [0.253146767616272, 0.24607858061790466, 0.24518734216690063]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.967, "test_loss": 0.11927742767333985}, {"step": 1900, "epoch": 4, "train_loss": 0.030803844332695007, "cos_r_negg": [0.5550085306167603, 0.5748291015625, 0.5515085458755493], "cos_innovation_negg": [0.5550085306167603, 0.5748291015625, 0.5515085458755493], "cos_apical_negg": [0.011234588921070099, 0.04739664867520332, 0.1076519712805748], "cos_Ac_negg": [0.5830277800559998, 0.5995380878448486, 0.563378095626831], "r_norm": [0.2503376603126526, 0.24449379742145538, 0.24348139762878418]}, {"step": 2000, "epoch": 4, "train_loss": 0.09435570985078812, "cos_r_negg": [0.5165828466415405, 0.5491183996200562, 0.5441352128982544], "cos_innovation_negg": [0.5165828466415405, 0.5491183996200562, 0.5441352128982544], "cos_apical_negg": [-0.005702102556824684, 0.04362977296113968, 0.10505573451519012], "cos_Ac_negg": [0.5735284090042114, 0.5990222692489624, 0.5746065974235535], "r_norm": [0.20219786465168, 0.19464869797229767, 0.19419066607952118], "ldr": {"dL_sdil": -0.004922345280647278, "dL_bp": -0.004920199513435364, "ratio": 1.0004361138620608}}, {"step": 2100, "epoch": 4, "train_loss": 0.008065913803875446, "cos_r_negg": [0.5316856503486633, 0.5577585697174072, 0.5528817772865295], "cos_innovation_negg": [0.5316856503486633, 0.5577585697174072, 0.5528817772865295], "cos_apical_negg": [0.018969878554344177, 0.044544756412506104, 0.11087940633296967], "cos_Ac_negg": [0.5732593536376953, 0.5963599681854248, 0.5710786581039429], "r_norm": [0.24292318522930145, 0.23432767391204834, 0.23457136750221252]}, {"step": 2200, "epoch": 4, "train_loss": 0.0065501658245921135, "cos_r_negg": [0.5224059820175171, 0.5490351319313049, 0.5424318313598633], "cos_innovation_negg": [0.5224059820175171, 0.5490351319313049, 0.5424318313598633], "cos_apical_negg": [0.014763617888092995, 0.03863360360264778, 0.09883178770542145], "cos_Ac_negg": [0.5712668299674988, 0.6000847220420837, 0.5719034671783447], "r_norm": [0.219838947057724, 0.21344280242919922, 0.21355140209197998]}, {"step": 2300, "epoch": 4, "train_loss": 0.14067436754703522, "cos_r_negg": [0.49908602237701416, 0.5320106744766235, 0.5351196527481079], "cos_innovation_negg": [0.49908602237701416, 0.5320106744766235, 0.5351196527481079], "cos_apical_negg": [0.008820908144116402, 0.048831529915332794, 0.10290086269378662], "cos_Ac_negg": [0.5642155408859253, 0.5992565155029297, 0.5796096324920654], "r_norm": [0.2458912581205368, 0.23886513710021973, 0.23990841209888458]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9661, "test_loss": 0.1243594123840332}, {"step": 2400, "epoch": 5, "train_loss": 0.0868406593799591, "cos_r_negg": [0.4981728494167328, 0.5238622426986694, 0.5275735259056091], "cos_innovation_negg": [0.4981728494167328, 0.5238622426986694, 0.5275735259056091], "cos_apical_negg": [0.008215553127229214, 0.04122934117913246, 0.10164561867713928], "cos_Ac_negg": [0.5746393203735352, 0.6053228378295898, 0.5804059505462646], "r_norm": [0.20620393753051758, 0.19982397556304932, 0.19935940206050873]}, {"step": 2500, "epoch": 5, "train_loss": 0.040987178683280945, "cos_r_negg": [0.4637001156806946, 0.486184298992157, 0.49550023674964905], "cos_innovation_negg": [0.4637001156806946, 0.486184298992157, 0.49550023674964905], "cos_apical_negg": [0.01503700576722622, 0.045721884816884995, 0.09604103118181229], "cos_Ac_negg": [0.5619102120399475, 0.6024391651153564, 0.579123854637146], "r_norm": [0.19776150584220886, 0.1923750638961792, 0.19342955946922302], "ldr": {"dL_sdil": -0.005911648273468018, "dL_bp": -0.00665283203125, "ratio": 0.8885912413990825}}, {"step": 2600, "epoch": 5, "train_loss": 0.03735675662755966, "cos_r_negg": [0.48341357707977295, 0.5146516561508179, 0.5274155139923096], "cos_innovation_negg": [0.48341357707977295, 0.5146516561508179, 0.5274155139923096], "cos_apical_negg": [0.00393025204539299, 0.04374466463923454, 0.10936721414327621], "cos_Ac_negg": [0.590373694896698, 0.6134263277053833, 0.5952221155166626], "r_norm": [0.2101920247077942, 0.20476499199867249, 0.20305541157722473]}, {"step": 2700, "epoch": 5, "train_loss": 0.03785250335931778, "cos_r_negg": [0.4948693811893463, 0.5172471404075623, 0.5347311496734619], "cos_innovation_negg": [0.4948693811893463, 0.5172471404075623, 0.5347311496734619], "cos_apical_negg": [0.004449144005775452, 0.04389144480228424, 0.11013566702604294], "cos_Ac_negg": [0.602059006690979, 0.6207597255706787, 0.6133108735084534], "r_norm": [0.21967127919197083, 0.21421346068382263, 0.2138521820306778]}, {"step": 2800, "epoch": 5, "train_loss": 0.07534697651863098, "cos_r_negg": [0.4680801033973694, 0.501079797744751, 0.5174782276153564], "cos_innovation_negg": [0.4680801033973694, 0.501079797744751, 0.5174782276153564], "cos_apical_negg": [0.008842382580041885, 0.03690087050199509, 0.10078456252813339], "cos_Ac_negg": [0.5848636627197266, 0.6153844594955444, 0.6031122207641602], "r_norm": [0.2104223370552063, 0.2037733793258667, 0.20278412103652954]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.9694, "test_loss": 0.11499397735595702}, {"step": 2900, "epoch": 6, "train_loss": 0.0019426055951043963, "cos_r_negg": [0.4577125906944275, 0.48318934440612793, 0.5011618733406067], "cos_innovation_negg": [0.4577125906944275, 0.48318934440612793, 0.5011618733406067], "cos_apical_negg": [0.020965497940778732, 0.04053390026092529, 0.10396619886159897], "cos_Ac_negg": [0.5896856188774109, 0.6151741147041321, 0.604279100894928], "r_norm": [0.20870232582092285, 0.20246008038520813, 0.20229654014110565]}, {"step": 3000, "epoch": 6, "train_loss": 0.06829206645488739, "cos_r_negg": [0.446952223777771, 0.4674347937107086, 0.4866243898868561], "cos_innovation_negg": [0.446952223777771, 0.4674347937107086, 0.4866243898868561], "cos_apical_negg": [-0.009396068751811981, 0.03873144090175629, 0.09478671848773956], "cos_Ac_negg": [0.6048896312713623, 0.6227243542671204, 0.6129453182220459], "r_norm": [0.1770733892917633, 0.17196708917617798, 0.17103160917758942], "ldr": {"dL_sdil": -0.004875198006629944, "dL_bp": -0.005990125238895416, "ratio": 0.8138724671354174}}, {"step": 3100, "epoch": 6, "train_loss": 0.012289129197597504, "cos_r_negg": [0.4534164071083069, 0.4699392318725586, 0.497898131608963], "cos_innovation_negg": [0.4534164071083069, 0.4699392318725586, 0.497898131608963], "cos_apical_negg": [0.01680128276348114, 0.034895673394203186, 0.10088913887739182], "cos_Ac_negg": [0.6157838106155396, 0.6316499710083008, 0.6197147369384766], "r_norm": [0.17025022208690643, 0.16576413810253143, 0.16465052962303162]}, {"step": 3200, "epoch": 6, "train_loss": 0.030991695821285248, "cos_r_negg": [0.4131392240524292, 0.44730043411254883, 0.4769807457923889], "cos_innovation_negg": [0.4131392240524292, 0.44730043411254883, 0.4769807457923889], "cos_apical_negg": [0.00890885479748249, 0.03517932444810867, 0.09522219002246857], "cos_Ac_negg": [0.5946764945983887, 0.6279523372650146, 0.6239750981330872], "r_norm": [0.16218790411949158, 0.15704484283924103, 0.15644490718841553]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9685, "test_loss": 0.130343825340271}, {"step": 3300, "epoch": 7, "train_loss": 0.004070380236953497, "cos_r_negg": [0.40077900886535645, 0.4304547607898712, 0.4595605731010437], "cos_innovation_negg": [0.40077900886535645, 0.4304547607898712, 0.4595605731010437], "cos_apical_negg": [0.005518389865756035, 0.034604042768478394, 0.09811795502901077], "cos_Ac_negg": [0.5955986976623535, 0.6299235820770264, 0.6221448183059692], "r_norm": [0.14800630509853363, 0.14408552646636963, 0.14225590229034424]}, {"step": 3400, "epoch": 7, "train_loss": 0.02358859032392502, "cos_r_negg": [0.3879971504211426, 0.411190927028656, 0.4450938105583191], "cos_innovation_negg": [0.3879971504211426, 0.411190927028656, 0.4450938105583191], "cos_apical_negg": [0.016526248306035995, 0.035195671021938324, 0.10495409369468689], "cos_Ac_negg": [0.6020306348800659, 0.6280429363250732, 0.6263999342918396], "r_norm": [0.17126505076885223, 0.1664196103811264, 0.1652672290802002]}, {"step": 3500, "epoch": 7, "train_loss": 0.001445534871891141, "cos_r_negg": [0.3632965087890625, 0.402198851108551, 0.4473251700401306], "cos_innovation_negg": [0.3632965087890625, 0.402198851108551, 0.4473251700401306], "cos_apical_negg": [0.009577827528119087, 0.028438780456781387, 0.10632304847240448], "cos_Ac_negg": [0.5836133360862732, 0.6256428956985474, 0.635414719581604], "r_norm": [0.16099384427070618, 0.15612384676933289, 0.15620404481887817], "ldr": {"dL_sdil": -0.005360133945941925, "dL_bp": -0.007305599749088287, "ratio": 0.7337021093457591}}, {"step": 3600, "epoch": 7, "train_loss": 0.0047991215251386166, "cos_r_negg": [0.37746262550354004, 0.40822017192840576, 0.45053255558013916], "cos_innovation_negg": [0.37746262550354004, 0.40822017192840576, 0.45053255558013916], "cos_apical_negg": [0.00926753506064415, 0.039603784680366516, 0.10520198196172714], "cos_Ac_negg": [0.5973981022834778, 0.6330288648605347, 0.6374830007553101], "r_norm": [0.17061659693717957, 0.1651216298341751, 0.1666044294834137]}, {"step": 3700, "epoch": 7, "train_loss": 0.0014059263048693538, "cos_r_negg": [0.36543941497802734, 0.39931243658065796, 0.4484252333641052], "cos_innovation_negg": [0.36543941497802734, 0.39931243658065796, 0.4484252333641052], "cos_apical_negg": [0.010484463535249233, 0.03964608535170555, 0.10806763917207718], "cos_Ac_negg": [0.5976572632789612, 0.6345747709274292, 0.6466951370239258], "r_norm": [0.1540808379650116, 0.14826953411102295, 0.14882683753967285]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9716, "test_loss": 0.11724653453826904}, {"step": 3800, "epoch": 8, "train_loss": 0.0008633130346424878, "cos_r_negg": [0.3445333242416382, 0.377445250749588, 0.4279426336288452], "cos_innovation_negg": [0.3445333242416382, 0.377445250749588, 0.4279426336288452], "cos_apical_negg": [0.0023265022318810225, 0.0419115386903286, 0.10405996441841125], "cos_Ac_negg": [0.5913555026054382, 0.6292985677719116, 0.6440975666046143], "r_norm": [0.15268784761428833, 0.14833927154541016, 0.14836238324642181]}, {"step": 3900, "epoch": 8, "train_loss": 0.016011923551559448, "cos_r_negg": [0.30991411209106445, 0.3465301990509033, 0.4040083587169647], "cos_innovation_negg": [0.30991411209106445, 0.3465301990509033, 0.4040083587169647], "cos_apical_negg": [-0.000366229098290205, 0.03983292356133461, 0.10650406777858734], "cos_Ac_negg": [0.5823796987533569, 0.6275235414505005, 0.6463069915771484], "r_norm": [0.1643437147140503, 0.1599186658859253, 0.15844371914863586]}, {"step": 4000, "epoch": 8, "train_loss": 0.005876308307051659, "cos_r_negg": [0.2810835540294647, 0.32485589385032654, 0.38159894943237305], "cos_innovation_negg": [0.2810835540294647, 0.32485589385032654, 0.38159894943237305], "cos_apical_negg": [-0.004385711625218391, 0.03626742213964462, 0.10575808584690094], "cos_Ac_negg": [0.5818820595741272, 0.6265275478363037, 0.6461528539657593], "r_norm": [0.15862736105918884, 0.15402765572071075, 0.15324923396110535], "ldr": {"dL_sdil": -0.0052199587225914, "dL_bp": -0.00768677145242691, "ratio": 0.679083377839122}}, {"step": 4100, "epoch": 8, "train_loss": 0.009995711036026478, "cos_r_negg": [0.2915729284286499, 0.32985246181488037, 0.3921487331390381], "cos_innovation_negg": [0.2915729284286499, 0.32985246181488037, 0.3921487331390381], "cos_apical_negg": [-0.004570425022393465, 0.03635071963071823, 0.10753794014453888], "cos_Ac_negg": [0.5940312147140503, 0.6342158913612366, 0.6535048484802246], "r_norm": [0.174652099609375, 0.16777926683425903, 0.16745731234550476]}, {"step": 4200, "epoch": 8, "train_loss": 0.01584916189312935, "cos_r_negg": [0.2785198986530304, 0.3130256235599518, 0.37501060962677], "cos_innovation_negg": [0.2785198986530304, 0.3130256235599518, 0.37501060962677], "cos_apical_negg": [0.0033586015924811363, 0.034764379262924194, 0.10399192571640015], "cos_Ac_negg": [0.5832823514938354, 0.627549409866333, 0.6530169248580933], "r_norm": [0.14530107378959656, 0.14131084084510803, 0.14038027822971344]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.972, "test_loss": 0.13099193515777588}, {"step": 4300, "epoch": 9, "train_loss": 0.0013420312898233533, "cos_r_negg": [0.27485042810440063, 0.3154665231704712, 0.37518662214279175], "cos_innovation_negg": [0.27485042810440063, 0.3154665231704712, 0.37518662214279175], "cos_apical_negg": [-0.0016189543530344963, 0.038209930062294006, 0.10528067499399185], "cos_Ac_negg": [0.5919089317321777, 0.6336437463760376, 0.657653272151947], "r_norm": [0.15844403207302094, 0.15488797426223755, 0.15405455231666565]}, {"step": 4400, "epoch": 9, "train_loss": 0.0013598615769296885, "cos_r_negg": [0.27434292435646057, 0.3129573464393616, 0.3681192398071289], "cos_innovation_negg": [0.27434292435646057, 0.3129573464393616, 0.3681192398071289], "cos_apical_negg": [-0.002546069212257862, 0.037157535552978516, 0.10349668562412262], "cos_Ac_negg": [0.5906094908714294, 0.6344246864318848, 0.6580358743667603], "r_norm": [0.1486673355102539, 0.14501874148845673, 0.1442796289920807]}, {"step": 4500, "epoch": 9, "train_loss": 0.03456545248627663, "cos_r_negg": [0.25133368372917175, 0.28834646940231323, 0.3497832417488098], "cos_innovation_negg": [0.25133368372917175, 0.28834646940231323, 0.3497832417488098], "cos_apical_negg": [0.0015513217076659203, 0.03367505967617035, 0.10306073725223541], "cos_Ac_negg": [0.5931183099746704, 0.6332410573959351, 0.6618243455886841], "r_norm": [0.15447081625461578, 0.1503368765115738, 0.14995679259300232], "ldr": {"dL_sdil": -0.0061156973242759705, "dL_bp": -0.009683042764663696, "ratio": 0.6315883832088369}}, {"step": 4600, "epoch": 9, "train_loss": 0.000631060975138098, "cos_r_negg": [0.25374943017959595, 0.2884773910045624, 0.35207968950271606], "cos_innovation_negg": [0.25374943017959595, 0.2884773910045624, 0.35207968950271606], "cos_apical_negg": [0.0015299245715141296, 0.031631648540496826, 0.10145542025566101], "cos_Ac_negg": [0.5881377458572388, 0.6348575353622437, 0.6593854427337646], "r_norm": [0.15295439958572388, 0.14834043383598328, 0.14849190413951874]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9716, "test_loss": 0.13195653724670411}, {"step": 4700, "epoch": 10, "train_loss": 0.0018224127124994993, "cos_r_negg": [0.24005019664764404, 0.27142199873924255, 0.33488965034484863], "cos_innovation_negg": [0.24005019664764404, 0.27142199873924255, 0.33488965034484863], "cos_apical_negg": [0.004340329673141241, 0.03361248970031738, 0.09669365733861923], "cos_Ac_negg": [0.592002272605896, 0.637415885925293, 0.6620798110961914], "r_norm": [0.1530528962612152, 0.14844045042991638, 0.14824894070625305]}, {"step": 4800, "epoch": 10, "train_loss": 0.00042262973147444427, "cos_r_negg": [0.23037979006767273, 0.26626986265182495, 0.3306382894515991], "cos_innovation_negg": [0.23037979006767273, 0.26626986265182495, 0.3306382894515991], "cos_apical_negg": [-0.00018684379756450653, 0.03444229066371918, 0.09747174382209778], "cos_Ac_negg": [0.5839577913284302, 0.6326618790626526, 0.6625460386276245], "r_norm": [0.15761873126029968, 0.1525515615940094, 0.15248043835163116]}, {"step": 4900, "epoch": 10, "train_loss": 0.002176593756303191, "cos_r_negg": [0.22571918368339539, 0.26211169362068176, 0.32274991273880005], "cos_innovation_negg": [0.22571918368339539, 0.26211169362068176, 0.32274991273880005], "cos_apical_negg": [-0.002524089068174362, 0.036017339676618576, 0.09383861720561981], "cos_Ac_negg": [0.5886404514312744, 0.6347569227218628, 0.6628813743591309], "r_norm": [0.1424628496170044, 0.1377774327993393, 0.13758273422718048]}, {"step": 5000, "epoch": 10, "train_loss": 0.0006744620623067021, "cos_r_negg": [0.2222062051296234, 0.25270354747772217, 0.30892008543014526], "cos_innovation_negg": [0.2222062051296234, 0.25270354747772217, 0.30892008543014526], "cos_apical_negg": [0.0009590340778231621, 0.03153621777892113, 0.08860278129577637], "cos_Ac_negg": [0.5725927352905273, 0.6265392303466797, 0.6623731851577759], "r_norm": [0.15401682257652283, 0.15014326572418213, 0.1496882438659668], "ldr": {"dL_sdil": -0.006028294563293457, "dL_bp": -0.009351417422294617, "ratio": 0.6446396616754015}}, {"step": 5100, "epoch": 10, "train_loss": 0.0014848911669105291, "cos_r_negg": [0.21961820125579834, 0.2580984830856323, 0.30639171600341797], "cos_innovation_negg": [0.21961820125579834, 0.2580984830856323, 0.30639171600341797], "cos_apical_negg": [-0.0019001984037458897, 0.03560680150985718, 0.08627481758594513], "cos_Ac_negg": [0.5766924619674683, 0.6369302868843079, 0.6644871234893799], "r_norm": [0.1511916220188141, 0.14784136414527893, 0.14697614312171936]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9721, "test_loss": 0.1368281675338745}, {"step": 5200, "epoch": 11, "train_loss": 0.0008225785568356514, "cos_r_negg": [0.21912850439548492, 0.25657331943511963, 0.30576181411743164], "cos_innovation_negg": [0.21912850439548492, 0.25657331943511963, 0.30576181411743164], "cos_apical_negg": [-0.0035089156590402126, 0.0334937609732151, 0.08607956022024155], "cos_Ac_negg": [0.5823262929916382, 0.6361823081970215, 0.664984941482544], "r_norm": [0.13863006234169006, 0.1350957751274109, 0.134585902094841]}, {"step": 5300, "epoch": 11, "train_loss": 0.002895822050049901, "cos_r_negg": [0.21578525006771088, 0.24931523203849792, 0.29838377237319946], "cos_innovation_negg": [0.21578525006771088, 0.24931523203849792, 0.29838377237319946], "cos_apical_negg": [-0.0028217416256666183, 0.03173669055104256, 0.08159370720386505], "cos_Ac_negg": [0.5741730332374573, 0.629033625125885, 0.6636229157447815], "r_norm": [0.1457689106464386, 0.1420048177242279, 0.14134639501571655]}, {"step": 5400, "epoch": 11, "train_loss": 0.0014625702751800418, "cos_r_negg": [0.21218493580818176, 0.24303241074085236, 0.29275012016296387], "cos_innovation_negg": [0.21218493580818176, 0.24303241074085236, 0.29275012016296387], "cos_apical_negg": [-0.0010529994033277035, 0.031591467559337616, 0.08191727101802826], "cos_Ac_negg": [0.5728154182434082, 0.6315698623657227, 0.6656259894371033], "r_norm": [0.14087790250778198, 0.1373642235994339, 0.13662117719650269]}, {"step": 5500, "epoch": 11, "train_loss": 0.0016903358045965433, "cos_r_negg": [0.21171659231185913, 0.24361832439899445, 0.29311251640319824], "cos_innovation_negg": [0.21171659231185913, 0.24361832439899445, 0.29311251640319824], "cos_apical_negg": [-0.001549859531223774, 0.03345705568790436, 0.08360060304403305], "cos_Ac_negg": [0.5768427848815918, 0.6334128975868225, 0.6679232120513916], "r_norm": [0.13572141528129578, 0.1321655958890915, 0.13195744156837463], "ldr": {"dL_sdil": -0.005469560623168945, "dL_bp": -0.009053520858287811, "ratio": 0.6041363032992825}}, {"step": 5600, "epoch": 11, "train_loss": 0.0012983162887394428, "cos_r_negg": [0.20741267502307892, 0.23724700510501862, 0.28603723645210266], "cos_innovation_negg": [0.20741267502307892, 0.23724700510501862, 0.28603723645210266], "cos_apical_negg": [-0.002379343379288912, 0.03247354179620743, 0.08056950569152832], "cos_Ac_negg": [0.5771089196205139, 0.6324926614761353, 0.6671242713928223], "r_norm": [0.13510887324810028, 0.13125915825366974, 0.13129353523254395]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9727, "test_loss": 0.14068544807434083}, {"step": 5700, "epoch": 12, "train_loss": 0.0007156288484111428, "cos_r_negg": [0.19856968522071838, 0.23168379068374634, 0.2786247134208679], "cos_innovation_negg": [0.19856968522071838, 0.23168379068374634, 0.2786247134208679], "cos_apical_negg": [-0.0016147242859005928, 0.03143199533224106, 0.07900910079479218], "cos_Ac_negg": [0.5706033706665039, 0.6268273591995239, 0.6659523844718933], "r_norm": [0.1336199790239334, 0.12988582253456116, 0.12971554696559906]}, {"step": 5800, "epoch": 12, "train_loss": 0.00048588408390060067, "cos_r_negg": [0.1939568668603897, 0.23039600253105164, 0.27736300230026245], "cos_innovation_negg": [0.1939568668603897, 0.23039600253105164, 0.27736300230026245], "cos_apical_negg": [-0.0022492785938084126, 0.03173355013132095, 0.08023496717214584], "cos_Ac_negg": [0.5727983713150024, 0.6307904720306396, 0.6693071126937866], "r_norm": [0.1406402438879013, 0.13692748546600342, 0.13653405010700226]}, {"step": 5900, "epoch": 12, "train_loss": 0.0019222944974899292, "cos_r_negg": [0.19307759404182434, 0.23198121786117554, 0.2767632007598877], "cos_innovation_negg": [0.19307759404182434, 0.23198121786117554, 0.2767632007598877], "cos_apical_negg": [-0.004706338047981262, 0.03159979730844498, 0.07717008888721466], "cos_Ac_negg": [0.5650250911712646, 0.6260983943939209, 0.6689882278442383], "r_norm": [0.13981205224990845, 0.1357787847518921, 0.13550418615341187]}, {"step": 6000, "epoch": 12, "train_loss": 0.00045124097960069776, "cos_r_negg": [0.1851966679096222, 0.2210320681333542, 0.2665230631828308], "cos_innovation_negg": [0.1851966679096222, 0.2210320681333542, 0.2665230631828308], "cos_apical_negg": [-0.0034684371203184128, 0.031309306621551514, 0.07571516931056976], "cos_Ac_negg": [0.5698431730270386, 0.6257672309875488, 0.6678767204284668], "r_norm": [0.13368269801139832, 0.12989981472492218, 0.12956127524375916], "ldr": {"dL_sdil": -0.005572102963924408, "dL_bp": -0.00944441556930542, "ratio": 0.5899891764646483}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9721, "test_loss": 0.143893613243103}, {"step": 6100, "epoch": 13, "train_loss": 0.0005720598273910582, "cos_r_negg": [0.18247531354427338, 0.21841880679130554, 0.26117002964019775], "cos_innovation_negg": [0.18247531354427338, 0.21841880679130554, 0.26117002964019775], "cos_apical_negg": [-0.008093561977148056, 0.03186175972223282, 0.07311908900737762], "cos_Ac_negg": [0.5712630152702332, 0.6270120143890381, 0.6698102951049805], "r_norm": [0.13618148863315582, 0.13248328864574432, 0.13214844465255737]}, {"step": 6200, "epoch": 13, "train_loss": 0.0006833100924268365, "cos_r_negg": [0.1804307997226715, 0.21572285890579224, 0.25849825143814087], "cos_innovation_negg": [0.1804307997226715, 0.21572285890579224, 0.25849825143814087], "cos_apical_negg": [-0.004912731237709522, 0.031982652842998505, 0.07271470874547958], "cos_Ac_negg": [0.5690498352050781, 0.6263390779495239, 0.6701257824897766], "r_norm": [0.1367523968219757, 0.1327851414680481, 0.1326279640197754]}, {"step": 6300, "epoch": 13, "train_loss": 0.0008542236755602062, "cos_r_negg": [0.17706552147865295, 0.2127162516117096, 0.25362932682037354], "cos_innovation_negg": [0.17706552147865295, 0.2127162516117096, 0.25362932682037354], "cos_apical_negg": [-0.006028501316905022, 0.03235006332397461, 0.07116851955652237], "cos_Ac_negg": [0.566336989402771, 0.6259113550186157, 0.670574963092804], "r_norm": [0.1369779109954834, 0.1333196461200714, 0.13319005072116852]}, {"step": 6400, "epoch": 13, "train_loss": 0.0005410106969065964, "cos_r_negg": [0.18153151869773865, 0.21716755628585815, 0.25861719250679016], "cos_innovation_negg": [0.18153151869773865, 0.21716755628585815, 0.25861719250679016], "cos_apical_negg": [-0.005879516247659922, 0.031762123107910156, 0.07145000249147415], "cos_Ac_negg": [0.5664576292037964, 0.6267338395118713, 0.6720196008682251], "r_norm": [0.13243013620376587, 0.12902915477752686, 0.12883922457695007]}, {"step": 6500, "epoch": 13, "train_loss": 0.00020985145238228142, "cos_r_negg": [0.17772656679153442, 0.21242636442184448, 0.25414592027664185], "cos_innovation_negg": [0.17772656679153442, 0.21242636442184448, 0.25414592027664185], "cos_apical_negg": [-0.0075380546040833, 0.030885789543390274, 0.06898142397403717], "cos_Ac_negg": [0.5650582313537598, 0.6260440349578857, 0.6713216304779053], "r_norm": [0.1292339712381363, 0.12573862075805664, 0.12550702691078186], "ldr": {"dL_sdil": -0.0056751444935798645, "dL_bp": -0.009983845055103302, "ratio": 0.5684327493322806}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9727, "test_loss": 0.1467740493774414}, {"step": 6600, "epoch": 14, "train_loss": 0.00017112368368543684, "cos_r_negg": [0.17454290390014648, 0.2121538668870926, 0.2518329918384552], "cos_innovation_negg": [0.17454290390014648, 0.2121538668870926, 0.2518329918384552], "cos_apical_negg": [-0.007086961530148983, 0.03190644085407257, 0.0692686215043068], "cos_Ac_negg": [0.5636143684387207, 0.6269121766090393, 0.6718631982803345], "r_norm": [0.12923821806907654, 0.1260632574558258, 0.125556081533432]}, {"step": 6700, "epoch": 14, "train_loss": 0.0006778803071938455, "cos_r_negg": [0.17567873001098633, 0.2124112993478775, 0.2522124648094177], "cos_innovation_negg": [0.17567873001098633, 0.2124112993478775, 0.2522124648094177], "cos_apical_negg": [-0.006559518165886402, 0.030668459832668304, 0.06823772192001343], "cos_Ac_negg": [0.5630702972412109, 0.6271776556968689, 0.6719936728477478], "r_norm": [0.13151130080223083, 0.12808698415756226, 0.12786173820495605]}, {"step": 6800, "epoch": 14, "train_loss": 0.0005970008205622435, "cos_r_negg": [0.1692158281803131, 0.20591729879379272, 0.24470087885856628], "cos_innovation_negg": [0.1692158281803131, 0.20591729879379272, 0.24470087885856628], "cos_apical_negg": [-0.00712909922003746, 0.03100281022489071, 0.06647836416959763], "cos_Ac_negg": [0.5601401329040527, 0.6256040334701538, 0.6721820831298828], "r_norm": [0.12940508127212524, 0.1262277215719223, 0.12584228813648224]}, {"step": 6900, "epoch": 14, "train_loss": 0.0005270083784125745, "cos_r_negg": [0.16722643375396729, 0.20353162288665771, 0.24128787219524384], "cos_innovation_negg": [0.16722643375396729, 0.20353162288665771, 0.24128787219524384], "cos_apical_negg": [-0.0067224702797830105, 0.0305806752294302, 0.06410964578390121], "cos_Ac_negg": [0.5593805313110352, 0.625951886177063, 0.6732284426689148], "r_norm": [0.1306028664112091, 0.12735921144485474, 0.12685713171958923]}, {"step": 7000, "epoch": 14, "train_loss": 0.0004866862145718187, "cos_r_negg": [0.1670301854610443, 0.2021043747663498, 0.24019977450370789], "cos_innovation_negg": [0.1670301854610443, 0.2021043747663498, 0.24019977450370789], "cos_apical_negg": [-0.0063683222979307175, 0.03093770518898964, 0.06457505375146866], "cos_Ac_negg": [0.5627425909042358, 0.6281017065048218, 0.6729280352592468], "r_norm": [0.1315777450799942, 0.12827616930007935, 0.12790822982788086], "ldr": {"dL_sdil": -0.005945973098278046, "dL_bp": -0.010425247251987457, "ratio": 0.5703436047662575}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9721, "test_loss": 0.15021902580261232}], "final": {"test_acc": 0.9721, "test_loss": 0.15021902580261232, "wall_s": 23.304119110107422, "cos_r_negg": [0.16458654403686523, 0.2002539038658142, 0.2374999225139618], "cos_innovation_negg": [0.16458654403686523, 0.2002539038658142, 0.2374999225139618], "cos_apical_negg": [-0.008399410173296928, 0.030416259542107582, 0.06369166821241379], "cos_Ac_negg": [0.5622594356536865, 0.6273370981216431, 0.6730587482452393]}}
|