From 6ed4fa50ddfa4c7957aaa909aaf72f0d7d317712 Mon Sep 17 00:00:00 2001 From: YurenHao0426 Date: Mon, 23 Mar 2026 18:21:26 -0500 Subject: Initial implementation: all models, methods, toy and CIFAR experiments Debug phase. Toy LQ experiments (3 seeds) complete with terminal gradient matching. Credit bridge matches state bridge on linear system (~0.94 cosine). CIFAR experiments in progress. --- results/toy_lq/sweep_results.json | 1070 +++++++++++++++++++++++++++++++++++++ 1 file changed, 1070 insertions(+) create mode 100644 results/toy_lq/sweep_results.json (limited to 'results/toy_lq/sweep_results.json') diff --git a/results/toy_lq/sweep_results.json b/results/toy_lq/sweep_results.json new file mode 100644 index 0000000..30c3f49 --- /dev/null +++ b/results/toy_lq/sweep_results.json @@ -0,0 +1,1070 @@ +{ + "base": { + "best_cos": 0.28987203165888786, + "best_step": 500, + "final_cos": -0.0006980087685709199, + "final_rho": 0.00831739305673788, + "final_nudge": 0.0027815375166634717, + "history": [ + { + "step": 500, + "avg_cos": 0.28987203165888786, + "avg_rho": 0.30755721777677536, + "avg_nudge": -0.10750458451608817, + "loss_term": 0.4910352826118469, + "loss_bridge": 0.21142145991325378 + }, + { + "step": 1000, + "avg_cos": 0.17731603110829988, + "avg_rho": 0.17737891773382822, + "avg_nudge": -0.06381315520654122, + "loss_term": 0.1414657086133957, + "loss_bridge": 0.16202926635742188 + }, + { + "step": 1500, + "avg_cos": 0.07300025007377069, + "avg_rho": 0.07345063022027414, + "avg_nudge": -0.02641519942941765, + "loss_term": 0.07946588099002838, + "loss_bridge": 0.08628662675619125 + }, + { + "step": 2000, + "avg_cos": 0.05074742032835881, + "avg_rho": 0.048574333622430764, + "avg_nudge": -0.016719758200148743, + "loss_term": 0.06007867306470871, + "loss_bridge": 0.016275471076369286 + }, + { + "step": 2500, + "avg_cos": 0.045629044994711876, + "avg_rho": 0.0567939051737388, + "avg_nudge": -0.016934571011612814, + "loss_term": 0.04579576849937439, + "loss_bridge": 0.008496936410665512 + }, + { + "step": 3000, + "avg_cos": 0.06145744491368532, + "avg_rho": 0.05695042138298353, + "avg_nudge": -0.020737762407710154, + "loss_term": 0.04881729558110237, + "loss_bridge": 0.013217507861554623 + }, + { + "step": 3500, + "avg_cos": 0.04631757119204849, + "avg_rho": 0.0402289762472113, + "avg_nudge": -0.015406294725835323, + "loss_term": 0.04255056008696556, + "loss_bridge": 0.01808304153382778 + }, + { + "step": 4000, + "avg_cos": 0.07397006265819073, + "avg_rho": 0.07162392682706316, + "avg_nudge": -0.025235373992472887, + "loss_term": 0.05486675351858139, + "loss_bridge": 0.02182137593626976 + }, + { + "step": 4500, + "avg_cos": 0.07251314166933298, + "avg_rho": 0.07301904633641243, + "avg_nudge": -0.027378834628810484, + "loss_term": 0.027670202776789665, + "loss_bridge": 0.013123266398906708 + }, + { + "step": 5000, + "avg_cos": -0.0006980087685709199, + "avg_rho": 0.00831739305673788, + "avg_nudge": 0.0027815375166634717, + "loss_term": 0.021854262799024582, + "loss_bridge": 0.0122066093608737 + } + ] + }, + "noise_0.1": { + "best_cos": 0.28885648772120476, + "best_step": 500, + "final_cos": 0.02330129671220978, + "final_rho": 0.03084004654859503, + "final_nudge": -0.005932313855737448, + "history": [ + { + "step": 500, + "avg_cos": 0.28885648772120476, + "avg_rho": 0.306567445397377, + "avg_nudge": -0.1071426725635926, + "loss_term": 0.4925232529640198, + "loss_bridge": 0.20981840789318085 + }, + { + "step": 1000, + "avg_cos": 0.17456108083327612, + "avg_rho": 0.17607956007122993, + "avg_nudge": -0.06276426805804174, + "loss_term": 0.1324017345905304, + "loss_bridge": 0.1835721731185913 + }, + { + "step": 1500, + "avg_cos": 0.08755840392162402, + "avg_rho": 0.08556619860852759, + "avg_nudge": -0.03154423305143913, + "loss_term": 0.07669232785701752, + "loss_bridge": 0.10240059345960617 + }, + { + "step": 2000, + "avg_cos": 0.05209386876473824, + "avg_rho": 0.04985952338514229, + "avg_nudge": -0.01653688432027896, + "loss_term": 0.06218753010034561, + "loss_bridge": 0.02190624177455902 + }, + { + "step": 2500, + "avg_cos": 0.038306045811623335, + "avg_rho": 0.047700356847296156, + "avg_nudge": -0.01320391776971519, + "loss_term": 0.06128765642642975, + "loss_bridge": 0.01958899199962616 + }, + { + "step": 3000, + "avg_cos": 0.059713449950019516, + "avg_rho": 0.054914615117013454, + "avg_nudge": -0.01939693869402011, + "loss_term": 0.04238733649253845, + "loss_bridge": 0.006958359386771917 + }, + { + "step": 3500, + "avg_cos": 0.04283027160757532, + "avg_rho": 0.028526597811530035, + "avg_nudge": -0.01381517636279265, + "loss_term": 0.024001002311706543, + "loss_bridge": 0.006152431480586529 + }, + { + "step": 4000, + "avg_cos": 0.0656419579560558, + "avg_rho": 0.06320031352030735, + "avg_nudge": -0.02216823499960204, + "loss_term": 0.04236245155334473, + "loss_bridge": 0.01628262549638748 + }, + { + "step": 4500, + "avg_cos": 0.05787398883452018, + "avg_rho": 0.0610881638713181, + "avg_nudge": -0.02213552314788103, + "loss_term": 0.024835661053657532, + "loss_bridge": 0.008629711344838142 + }, + { + "step": 5000, + "avg_cos": 0.02330129671220978, + "avg_rho": 0.03084004654859503, + "avg_nudge": -0.005932313855737448, + "loss_term": 0.035377949476242065, + "loss_bridge": 0.020307490602135658 + } + ] + }, + "noise_0.3": { + "best_cos": 0.28320933257540065, + "best_step": 500, + "final_cos": -0.007130145425132166, + "final_rho": 0.003303757131410142, + "final_nudge": 0.004990910800794761, + "history": [ + { + "step": 500, + "avg_cos": 0.28320933257540065, + "avg_rho": 0.30105741570393246, + "avg_nudge": -0.10515742873152097, + "loss_term": 0.44116002321243286, + "loss_bridge": 0.210750013589859 + }, + { + "step": 1000, + "avg_cos": 0.1680816188454628, + "avg_rho": 0.1721916707853476, + "avg_nudge": -0.060552582144737244, + "loss_term": 0.14207889139652252, + "loss_bridge": 0.16882070899009705 + }, + { + "step": 1500, + "avg_cos": 0.08877109829336405, + "avg_rho": 0.08808179199695587, + "avg_nudge": -0.03221472934819758, + "loss_term": 0.12282441556453705, + "loss_bridge": 0.07336755841970444 + }, + { + "step": 2000, + "avg_cos": 0.04903770827998718, + "avg_rho": 0.04795513402981063, + "avg_nudge": -0.016069856472313404, + "loss_term": 0.06662434339523315, + "loss_bridge": 0.02520526573061943 + }, + { + "step": 2500, + "avg_cos": 0.05001032492145896, + "avg_rho": 0.053442057532568775, + "avg_nudge": -0.01783018947268526, + "loss_term": 0.04614880681037903, + "loss_bridge": 0.015394347719848156 + }, + { + "step": 3000, + "avg_cos": 0.03979540191357955, + "avg_rho": 0.03453826089389622, + "avg_nudge": -0.012916556559503078, + "loss_term": 0.04155049845576286, + "loss_bridge": 0.007860729470849037 + }, + { + "step": 3500, + "avg_cos": 0.06703585013747215, + "avg_rho": 0.04949762811884284, + "avg_nudge": -0.02297227829694748, + "loss_term": 0.059078700840473175, + "loss_bridge": 0.030750930309295654 + }, + { + "step": 4000, + "avg_cos": 0.06641693723698457, + "avg_rho": 0.06286459214364488, + "avg_nudge": -0.021883966866880655, + "loss_term": 0.025344880297780037, + "loss_bridge": 0.008583566173911095 + }, + { + "step": 4500, + "avg_cos": 0.03687915485352278, + "avg_rho": 0.028895257118468482, + "avg_nudge": -0.01225966370354096, + "loss_term": 0.020479349419474602, + "loss_bridge": 0.00925756897777319 + }, + { + "step": 5000, + "avg_cos": -0.007130145425132166, + "avg_rho": 0.003303757131410142, + "avg_nudge": 0.004990910800794761, + "loss_term": 0.03882830590009689, + "loss_bridge": 0.022515198215842247 + } + ] + }, + "lam_1.0": { + "best_cos": 0.2899630069732666, + "best_step": 500, + "final_cos": 0.0024576462844076255, + "final_rho": 0.010072723651925722, + "final_nudge": 0.0017782459035515785, + "history": [ + { + "step": 500, + "avg_cos": 0.2899630069732666, + "avg_rho": 0.3076498980323474, + "avg_nudge": -0.10753695170084636, + "loss_term": 0.49082812666893005, + "loss_bridge": 0.21152979135513306 + }, + { + "step": 1000, + "avg_cos": 0.17747302974263826, + "avg_rho": 0.17881879458824793, + "avg_nudge": -0.06420790683478117, + "loss_term": 0.1407778263092041, + "loss_bridge": 0.15578678250312805 + }, + { + "step": 1500, + "avg_cos": 0.08065215001503627, + "avg_rho": 0.08029377926141024, + "avg_nudge": -0.029326035796354216, + "loss_term": 0.07120706140995026, + "loss_bridge": 0.09130121767520905 + }, + { + "step": 2000, + "avg_cos": 0.0563324602941672, + "avg_rho": 0.058026942308060825, + "avg_nudge": -0.01875296530003349, + "loss_term": 0.06783974170684814, + "loss_bridge": 0.01712188683450222 + }, + { + "step": 2500, + "avg_cos": 0.039659525423000254, + "avg_rho": 0.04613731553157171, + "avg_nudge": -0.013568413676694036, + "loss_term": 0.0634445995092392, + "loss_bridge": 0.02091756835579872 + }, + { + "step": 3000, + "avg_cos": 0.04386034506993989, + "avg_rho": 0.039356538482631244, + "avg_nudge": -0.01407914562150836, + "loss_term": 0.05031818151473999, + "loss_bridge": 0.01259439717978239 + }, + { + "step": 3500, + "avg_cos": 0.0481002099889641, + "avg_rho": 0.03689269566287597, + "avg_nudge": -0.01590493693947792, + "loss_term": 0.03421805799007416, + "loss_bridge": 0.014973677694797516 + }, + { + "step": 4000, + "avg_cos": 0.07999587121109168, + "avg_rho": 0.07683291751891375, + "avg_nudge": -0.02731190746029218, + "loss_term": 0.04726963862776756, + "loss_bridge": 0.017824556678533554 + }, + { + "step": 4500, + "avg_cos": 0.07246351769814889, + "avg_rho": 0.07343382605661948, + "avg_nudge": -0.027415843835721414, + "loss_term": 0.02768528461456299, + "loss_bridge": 0.013396943919360638 + }, + { + "step": 5000, + "avg_cos": 0.0024576462844076255, + "avg_rho": 0.010072723651925722, + "avg_nudge": 0.0017782459035515785, + "loss_term": 0.02114972099661827, + "loss_bridge": 0.011466547846794128 + } + ] + }, + "noise_lam": { + "best_cos": 0.28980905935168266, + "best_step": 500, + "final_cos": 0.000333610107190907, + "final_rho": 0.009229002442831794, + "final_nudge": 0.0020260093733668327, + "history": [ + { + "step": 500, + "avg_cos": 0.28980905935168266, + "avg_rho": 0.30755361666282016, + "avg_nudge": -0.10748158146937688, + "loss_term": 0.4891643524169922, + "loss_bridge": 0.21093463897705078 + }, + { + "step": 1000, + "avg_cos": 0.1692807301878929, + "avg_rho": 0.17357065031925836, + "avg_nudge": -0.060928904761870704, + "loss_term": 0.12851648032665253, + "loss_bridge": 0.1991802453994751 + }, + { + "step": 1500, + "avg_cos": 0.08976124723752339, + "avg_rho": 0.08664474201699097, + "avg_nudge": -0.032548267083863415, + "loss_term": 0.08289942145347595, + "loss_bridge": 0.08361449092626572 + }, + { + "step": 2000, + "avg_cos": 0.0449219069754084, + "avg_rho": 0.03701058775186539, + "avg_nudge": -0.013820620176071921, + "loss_term": 0.07352523505687714, + "loss_bridge": 0.031066572293639183 + }, + { + "step": 2500, + "avg_cos": 0.0341803894067804, + "avg_rho": 0.04086895197785149, + "avg_nudge": -0.011834259377792478, + "loss_term": 0.05221429467201233, + "loss_bridge": 0.013445280492305756 + }, + { + "step": 3000, + "avg_cos": 0.043432267770792045, + "avg_rho": 0.035848827101290226, + "avg_nudge": -0.0138394293996195, + "loss_term": 0.08377102017402649, + "loss_bridge": 0.02437596395611763 + }, + { + "step": 3500, + "avg_cos": 0.04308730812044814, + "avg_rho": 0.02904196917855491, + "avg_nudge": -0.013877601828426123, + "loss_term": 0.028318829834461212, + "loss_bridge": 0.0096125528216362 + }, + { + "step": 4000, + "avg_cos": 0.07217423028002183, + "avg_rho": 0.07531639956869185, + "avg_nudge": -0.024754961564516027, + "loss_term": 0.04526568949222565, + "loss_bridge": 0.01877186819911003 + }, + { + "step": 4500, + "avg_cos": 0.041415012208744884, + "avg_rho": 0.039630077119606234, + "avg_nudge": -0.013629865366965532, + "loss_term": 0.0305576603859663, + "loss_bridge": 0.011333338916301727 + }, + { + "step": 5000, + "avg_cos": 0.000333610107190907, + "avg_rho": 0.009229002442831794, + "avg_nudge": 0.0020260093733668327, + "loss_term": 0.020293015986680984, + "loss_bridge": 0.00792770553380251 + } + ] + }, + "no_ln": { + "best_cos": 0.2994285201032956, + "best_step": 500, + "final_cos": -0.027601251068214577, + "final_rho": -0.03011056105606258, + "final_nudge": 0.013365049380809069, + "history": [ + { + "step": 500, + "avg_cos": 0.2994285201032956, + "avg_rho": 0.3287110353509585, + "avg_nudge": -0.11226618165771167, + "loss_term": 0.5378487706184387, + "loss_bridge": 0.20209679007530212 + }, + { + "step": 1000, + "avg_cos": 0.23427631705999374, + "avg_rho": 0.24483238657315573, + "avg_nudge": -0.08577251620590687, + "loss_term": 0.14587438106536865, + "loss_bridge": 0.17536549270153046 + }, + { + "step": 1500, + "avg_cos": 0.10971186744670074, + "avg_rho": 0.1081712432205677, + "avg_nudge": -0.039541066934665046, + "loss_term": 0.09813931584358215, + "loss_bridge": 0.1144137978553772 + }, + { + "step": 2000, + "avg_cos": 0.17568999342620373, + "avg_rho": 0.18043102137744427, + "avg_nudge": -0.06037815675760309, + "loss_term": 0.12119113653898239, + "loss_bridge": 0.05170433223247528 + }, + { + "step": 2500, + "avg_cos": 0.1501085925847292, + "avg_rho": 0.1694059126699964, + "avg_nudge": -0.05651436994473139, + "loss_term": 0.07073387503623962, + "loss_bridge": 0.015867799520492554 + }, + { + "step": 3000, + "avg_cos": 0.10612630782028039, + "avg_rho": 0.09910948442605634, + "avg_nudge": -0.034841354160259165, + "loss_term": 0.05772688612341881, + "loss_bridge": 0.0220709890127182 + }, + { + "step": 3500, + "avg_cos": 0.12348712608218193, + "avg_rho": 0.12745930068194866, + "avg_nudge": -0.04433920063699285, + "loss_term": 0.04150111600756645, + "loss_bridge": 0.014408521354198456 + }, + { + "step": 4000, + "avg_cos": 0.012287488003494218, + "avg_rho": 0.038615713633286454, + "avg_nudge": -0.006448045140132308, + "loss_term": 0.04848453402519226, + "loss_bridge": 0.026737889274954796 + }, + { + "step": 4500, + "avg_cos": 0.0344570055603981, + "avg_rho": 0.043244189505154886, + "avg_nudge": -0.012965732564528784, + "loss_term": 0.04390523582696915, + "loss_bridge": 0.016968518495559692 + }, + { + "step": 5000, + "avg_cos": -0.027601251068214577, + "avg_rho": -0.03011056105606258, + "avg_nudge": 0.013365049380809069, + "loss_term": 0.05419892817735672, + "loss_bridge": 0.029718749225139618 + } + ] + }, + "big_vnet": { + "best_cos": 0.25947993124524754, + "best_step": 500, + "final_cos": 0.012725223108039549, + "final_rho": -0.0029713623225688934, + "final_nudge": -0.0007455882926781973, + "history": [ + { + "step": 500, + "avg_cos": 0.25947993124524754, + "avg_rho": 0.2872927797337373, + "avg_nudge": -0.09759780826667945, + "loss_term": 0.24058431386947632, + "loss_bridge": 0.20110812783241272 + }, + { + "step": 1000, + "avg_cos": 0.11903308952848117, + "avg_rho": 0.10757205138603847, + "avg_nudge": -0.04069022353117665, + "loss_term": 0.1535366326570511, + "loss_bridge": 0.10731971263885498 + }, + { + "step": 1500, + "avg_cos": 0.04738504672423005, + "avg_rho": 0.04187516961246729, + "avg_nudge": -0.01667174060518543, + "loss_term": 0.12195796519517899, + "loss_bridge": 0.03276967257261276 + }, + { + "step": 2000, + "avg_cos": 0.05584627948701382, + "avg_rho": 0.06464844088380535, + "avg_nudge": -0.019484267104417086, + "loss_term": 0.07039390504360199, + "loss_bridge": 0.023653611540794373 + }, + { + "step": 2500, + "avg_cos": 0.11334512817362945, + "avg_rho": 0.13342145457863808, + "avg_nudge": -0.04256325137491027, + "loss_term": 0.09321287274360657, + "loss_bridge": 0.03603611886501312 + }, + { + "step": 3000, + "avg_cos": 0.07876436489944656, + "avg_rho": 0.08206061793801685, + "avg_nudge": -0.02565166230003039, + "loss_term": 0.03433217480778694, + "loss_bridge": 0.014776414260268211 + }, + { + "step": 3500, + "avg_cos": 0.059695989514390625, + "avg_rho": 0.043808821588754654, + "avg_nudge": -0.01833860871071617, + "loss_term": 0.07867280393838882, + "loss_bridge": 0.04651641845703125 + }, + { + "step": 4000, + "avg_cos": 0.008286381295571724, + "avg_rho": 0.01973852072842419, + "avg_nudge": -0.0019318210737158854, + "loss_term": 0.03502834588289261, + "loss_bridge": 0.011813998222351074 + }, + { + "step": 4500, + "avg_cos": -0.006770744492920737, + "avg_rho": 2.3505534045398235e-05, + "avg_nudge": 0.002897862965861956, + "loss_term": 0.04147114232182503, + "loss_bridge": 0.026934277266263962 + }, + { + "step": 5000, + "avg_cos": 0.012725223108039549, + "avg_rho": -0.0029713623225688934, + "avg_nudge": -0.0007455882926781973, + "loss_term": 0.038749027997255325, + "loss_bridge": 0.01941092312335968 + } + ] + }, + "ema_0.999": { + "best_cos": 0.10180055970946948, + "best_step": 1000, + "final_cos": -0.01584776126158734, + "final_rho": -0.01703926082700491, + "final_nudge": 0.007290713644276063, + "history": [ + { + "step": 500, + "avg_cos": -0.005628384804974, + "avg_rho": 0.010925033983464042, + "avg_nudge": 0.0015428058492640655, + "loss_term": 0.5920301675796509, + "loss_bridge": 1.4890536069869995 + }, + { + "step": 1000, + "avg_cos": 0.10180055970946948, + "avg_rho": 0.10515290250380833, + "avg_nudge": -0.037424925404290356, + "loss_term": 0.5715007185935974, + "loss_bridge": 0.4665977954864502 + }, + { + "step": 1500, + "avg_cos": 0.021814276037427287, + "avg_rho": 0.003451728650058309, + "avg_nudge": -0.005690907438596089, + "loss_term": 0.2616257071495056, + "loss_bridge": 0.2758212387561798 + }, + { + "step": 2000, + "avg_cos": 0.03669632730695108, + "avg_rho": 0.036930523036668696, + "avg_nudge": -0.01006293793519338, + "loss_term": 0.11164076626300812, + "loss_bridge": 0.1407940685749054 + }, + { + "step": 2500, + "avg_cos": -0.020325756592986483, + "avg_rho": -0.027848235641916592, + "avg_nudge": 0.011153894321372112, + "loss_term": 0.15471391379833221, + "loss_bridge": 0.06181420385837555 + }, + { + "step": 3000, + "avg_cos": -0.0060501456415901584, + "avg_rho": -0.013405103546877703, + "avg_nudge": 0.005228333951284488, + "loss_term": 0.07506504654884338, + "loss_bridge": 0.08207326382398605 + }, + { + "step": 3500, + "avg_cos": -0.02149865326161186, + "avg_rho": -0.023167532014970977, + "avg_nudge": 0.010358475303898254, + "loss_term": 0.048137497156858444, + "loss_bridge": 0.03276998922228813 + }, + { + "step": 4000, + "avg_cos": -0.007104064881180723, + "avg_rho": -0.00720199760204802, + "avg_nudge": 0.005406570465614398, + "loss_term": 0.03773114085197449, + "loss_bridge": 0.037960827350616455 + }, + { + "step": 4500, + "avg_cos": -0.0034141440119128674, + "avg_rho": -0.003708663280121982, + "avg_nudge": 0.002461720102777084, + "loss_term": 0.0416095145046711, + "loss_bridge": 0.03200625628232956 + }, + { + "step": 5000, + "avg_cos": -0.01584776126158734, + "avg_rho": -0.01703926082700491, + "avg_nudge": 0.007290713644276063, + "loss_term": 0.023291591554880142, + "loss_bridge": 0.026996765285730362 + } + ] + }, + "K16": { + "best_cos": 0.3187306026617686, + "best_step": 500, + "final_cos": 0.012039402422184745, + "final_rho": -0.002023946028202772, + "final_nudge": -0.0010173787983755271, + "history": [ + { + "step": 500, + "avg_cos": 0.3187306026617686, + "avg_rho": 0.3298306291302045, + "avg_nudge": -0.12337777391076088, + "loss_term": 0.3945310413837433, + "loss_bridge": 0.27078020572662354 + }, + { + "step": 1000, + "avg_cos": 0.15608959334592024, + "avg_rho": 0.14268888781468073, + "avg_nudge": -0.05713697926451763, + "loss_term": 0.13724187016487122, + "loss_bridge": 0.13912135362625122 + }, + { + "step": 1500, + "avg_cos": 0.07919560000300407, + "avg_rho": 0.08190769484887521, + "avg_nudge": -0.029933936273058254, + "loss_term": 0.08082282543182373, + "loss_bridge": 0.07666538655757904 + }, + { + "step": 2000, + "avg_cos": 0.06641554242620866, + "avg_rho": 0.055935436549286045, + "avg_nudge": -0.020354578581949074, + "loss_term": 0.07134468108415604, + "loss_bridge": 0.011993557214736938 + }, + { + "step": 2500, + "avg_cos": 0.0844428426741312, + "avg_rho": 0.0917752521733443, + "avg_nudge": -0.028687965202455718, + "loss_term": 0.0323462039232254, + "loss_bridge": 0.00667245127260685 + }, + { + "step": 3000, + "avg_cos": 0.0888429010907809, + "avg_rho": 0.06338833862294753, + "avg_nudge": -0.026075587142258883, + "loss_term": 0.04170331731438637, + "loss_bridge": 0.010068882256746292 + }, + { + "step": 3500, + "avg_cos": 0.016165781184099615, + "avg_rho": 0.014142975055923065, + "avg_nudge": -0.003436643397435546, + "loss_term": 0.03528498113155365, + "loss_bridge": 0.012309195473790169 + }, + { + "step": 4000, + "avg_cos": 0.06803990031282107, + "avg_rho": 0.055018783935035266, + "avg_nudge": -0.022906929564972717, + "loss_term": 0.027342472225427628, + "loss_bridge": 0.007336604408919811 + }, + { + "step": 4500, + "avg_cos": 0.024216643689821165, + "avg_rho": 0.04094697698019445, + "avg_nudge": -0.008364889615525803, + "loss_term": 0.027580715715885162, + "loss_bridge": 0.016118880361318588 + }, + { + "step": 5000, + "avg_cos": 0.012039402422184745, + "avg_rho": -0.002023946028202772, + "avg_nudge": -0.0010173787983755271, + "loss_term": 0.027855150401592255, + "loss_bridge": 0.010500052943825722 + } + ] + }, + "best_combo": { + "best_cos": 0.30479515840609867, + "best_step": 500, + "final_cos": -0.025737087552746136, + "final_rho": -0.01576789258979261, + "final_nudge": 0.011260819776604572, + "history": [ + { + "step": 500, + "avg_cos": 0.30479515840609867, + "avg_rho": 0.33246727536122006, + "avg_nudge": -0.11396304952601592, + "loss_term": 0.5129790306091309, + "loss_bridge": 0.21324321627616882 + }, + { + "step": 1000, + "avg_cos": 0.24110793943206468, + "avg_rho": 0.24976263443628946, + "avg_nudge": -0.08793549550076325, + "loss_term": 0.14881110191345215, + "loss_bridge": 0.1860560178756714 + }, + { + "step": 1500, + "avg_cos": 0.12106851922969024, + "avg_rho": 0.11747027436892192, + "avg_nudge": -0.042971268917123474, + "loss_term": 0.10358402132987976, + "loss_bridge": 0.11228226125240326 + }, + { + "step": 2000, + "avg_cos": 0.19137668733795485, + "avg_rho": 0.2001398652791977, + "avg_nudge": -0.06597264126564066, + "loss_term": 0.0836295336484909, + "loss_bridge": 0.03368356451392174 + }, + { + "step": 2500, + "avg_cos": 0.1382010855789607, + "avg_rho": 0.15255235826286176, + "avg_nudge": -0.05140705577408274, + "loss_term": 0.058304790407419205, + "loss_bridge": 0.014804087579250336 + }, + { + "step": 3000, + "avg_cos": 0.0815443117171526, + "avg_rho": 0.07958398557578523, + "avg_nudge": -0.026181443439175684, + "loss_term": 0.059965550899505615, + "loss_bridge": 0.016236742958426476 + }, + { + "step": 3500, + "avg_cos": 0.09519872162491083, + "avg_rho": 0.09258671143713097, + "avg_nudge": -0.03418003007148703, + "loss_term": 0.03956954926252365, + "loss_bridge": 0.014235305599868298 + }, + { + "step": 4000, + "avg_cos": -0.011209671385586262, + "avg_rho": -0.007080828654579818, + "avg_nudge": 0.004587161975602309, + "loss_term": 0.03683673217892647, + "loss_bridge": 0.018107034265995026 + }, + { + "step": 4500, + "avg_cos": 0.04608155476550261, + "avg_rho": 0.05846519426753124, + "avg_nudge": -0.017059470837314922, + "loss_term": 0.043314699083566666, + "loss_bridge": 0.023334285244345665 + }, + { + "step": 5000, + "avg_cos": -0.025737087552746136, + "avg_rho": -0.01576789258979261, + "avg_nudge": 0.011260819776604572, + "loss_term": 0.03279898688197136, + "loss_bridge": 0.015516506507992744 + } + ] + }, + "noise_1.0": { + "best_cos": 0.2831856335202853, + "best_step": 500, + "final_cos": 0.010971122809375325, + "final_rho": 0.014127362189659229, + "final_nudge": -0.002180874968568484, + "history": [ + { + "step": 500, + "avg_cos": 0.2831856335202853, + "avg_rho": 0.30480503539244336, + "avg_nudge": -0.1052918794254462, + "loss_term": 0.4858350455760956, + "loss_bridge": 0.20565161108970642 + }, + { + "step": 1000, + "avg_cos": 0.14922113033632436, + "avg_rho": 0.14803783098856607, + "avg_nudge": -0.052950371988117695, + "loss_term": 0.1246427372097969, + "loss_bridge": 0.1951574981212616 + }, + { + "step": 1500, + "avg_cos": 0.032916761857147016, + "avg_rho": 0.030339293957998354, + "avg_nudge": -0.011420852970331907, + "loss_term": 0.059441905468702316, + "loss_bridge": 0.09553220868110657 + }, + { + "step": 2000, + "avg_cos": 0.02735897192421059, + "avg_rho": 0.019464978327353794, + "avg_nudge": -0.008191200438886881, + "loss_term": 0.07519456744194031, + "loss_bridge": 0.017568862065672874 + }, + { + "step": 2500, + "avg_cos": 0.02141591941472143, + "avg_rho": 0.027642763530214626, + "avg_nudge": -0.007318320373694102, + "loss_term": 0.05027623474597931, + "loss_bridge": 0.014622311107814312 + }, + { + "step": 3000, + "avg_cos": 0.048597725903770574, + "avg_rho": 0.03585781451935569, + "avg_nudge": -0.015588213689625263, + "loss_term": 0.06457968056201935, + "loss_bridge": 0.01680811122059822 + }, + { + "step": 3500, + "avg_cos": 0.06355043267831206, + "avg_rho": 0.04236283013597131, + "avg_nudge": -0.021118728754421074, + "loss_term": 0.035970453172922134, + "loss_bridge": 0.015534179285168648 + }, + { + "step": 4000, + "avg_cos": 0.04567302499587337, + "avg_rho": 0.05423017560193936, + "avg_nudge": -0.01528523334612449, + "loss_term": 0.034011729061603546, + "loss_bridge": 0.009947280399501324 + }, + { + "step": 4500, + "avg_cos": 0.024254882785802085, + "avg_rho": 0.023739464891453583, + "avg_nudge": -0.009648310175786415, + "loss_term": 0.04320281371474266, + "loss_bridge": 0.01757601648569107 + }, + { + "step": 5000, + "avg_cos": 0.010971122809375325, + "avg_rho": 0.014127362189659229, + "avg_nudge": -0.002180874968568484, + "loss_term": 0.024934478104114532, + "loss_bridge": 0.011717695742845535 + } + ] + }, + "lr_3e-4": { + "best_cos": 0.6101815849542618, + "best_step": 500, + "final_cos": -0.008048781737064322, + "final_rho": -0.026018392760306597, + "final_nudge": 0.007571722225596507, + "history": [ + { + "step": 500, + "avg_cos": 0.6101815849542618, + "avg_rho": 0.620405301451683, + "avg_nudge": -0.2231019102036953, + "loss_term": 3.107595205307007, + "loss_bridge": 2.7781143188476562 + }, + { + "step": 1000, + "avg_cos": 0.39656415830055874, + "avg_rho": 0.39799897621075314, + "avg_nudge": -0.14455867062012354, + "loss_term": 0.29117679595947266, + "loss_bridge": 0.13909414410591125 + }, + { + "step": 1500, + "avg_cos": 0.26989879210789997, + "avg_rho": 0.2640038679043452, + "avg_nudge": -0.09908402090271314, + "loss_term": 0.1437155306339264, + "loss_bridge": 0.06845193356275558 + }, + { + "step": 2000, + "avg_cos": 0.15282577524582544, + "avg_rho": 0.1327554533878962, + "avg_nudge": -0.05070468131452799, + "loss_term": 0.10841675102710724, + "loss_bridge": 0.044324424117803574 + }, + { + "step": 2500, + "avg_cos": 0.054395756063361965, + "avg_rho": 0.04450509278103709, + "avg_nudge": -0.01604464929550886, + "loss_term": 0.09302366524934769, + "loss_bridge": 0.016296718269586563 + }, + { + "step": 3000, + "avg_cos": 0.041961303912103176, + "avg_rho": 0.032989607813457646, + "avg_nudge": -0.01202150775740544, + "loss_term": 0.06656567752361298, + "loss_bridge": 0.0089980224147439 + }, + { + "step": 3500, + "avg_cos": 0.010103868204168975, + "avg_rho": -0.01982816867530346, + "avg_nudge": 0.0005857348442077637, + "loss_term": 0.0479322224855423, + "loss_bridge": 0.004413220100104809 + }, + { + "step": 4000, + "avg_cos": 0.028387469239532948, + "avg_rho": 0.012580555553237597, + "avg_nudge": -0.006568876715997855, + "loss_term": 0.06536682695150375, + "loss_bridge": 0.005300351418554783 + }, + { + "step": 4500, + "avg_cos": 0.015142000513151288, + "avg_rho": 0.019022303090120356, + "avg_nudge": -0.004821705476691325, + "loss_term": 0.037237197160720825, + "loss_bridge": 0.005878066644072533 + }, + { + "step": 5000, + "avg_cos": -0.008048781737064322, + "avg_rho": -0.026018392760306597, + "avg_nudge": 0.007571722225596507, + "loss_term": 0.02460472844541073, + "loss_bridge": 0.005403056740760803 + } + ] + } +} \ No newline at end of file -- cgit v1.2.3