diff options
Diffstat (limited to 'results/nuis_ctrl_v1_rho0p2_residual_s4.json')
| -rw-r--r-- | results/nuis_ctrl_v1_rho0p2_residual_s4.json | 1 |
1 files changed, 1 insertions, 0 deletions
diff --git a/results/nuis_ctrl_v1_rho0p2_residual_s4.json b/results/nuis_ctrl_v1_rho0p2_residual_s4.json new file mode 100644 index 0000000..612a1c7 --- /dev/null +++ b/results/nuis_ctrl_v1_rho0p2_residual_s4.json @@ -0,0 +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": 0.2, "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_residual_s4", "n_in": 784}, "provenance": {"git_commit": "c8a372cae2612839bdbdafcf9c114f52cb4a6de3", "git_dirty": false}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.4633917808532715, "cos_r_negg": [-0.012779776938259602, 0.029937468469142914, 0.009443523362278938], "cos_innovation_negg": [-0.012779776938259602, 0.029937468469142914, 0.009443523362278938], "cos_apical_negg": [-0.014390179887413979, 0.02703464776277542, 0.006248843390494585], "cos_Ac_negg": [-0.012780384160578251, 0.029937095940113068, 0.009443294256925583], "r_norm": [4.848398208618164, 4.8553009033203125, 4.948000907897949], "ldr": {"dL_sdil": -0.003349781036376953, "dL_bp": -0.26721620559692383, "ratio": 0.012535845379938721}}, {"step": 100, "epoch": 0, "train_loss": 0.19523030519485474, "cos_r_negg": [0.45847442746162415, 0.37912869453430176, 0.3189619779586792], "cos_innovation_negg": [0.45847442746162415, 0.37912869453430176, 0.3189619779586792], "cos_apical_negg": [0.11247933655977249, 0.13046348094940186, 0.1326301544904709], "cos_Ac_negg": [0.4592515826225281, 0.3796048164367676, 0.31860604882240295], "r_norm": [0.9187071919441223, 0.9284980297088623, 0.943510890007019]}, {"step": 200, "epoch": 0, "train_loss": 0.24339696764945984, "cos_r_negg": [0.5118288993835449, 0.44126784801483154, 0.3616797924041748], "cos_innovation_negg": [0.5118288993835449, 0.44126784801483154, 0.3616797924041748], "cos_apical_negg": [0.12309147417545319, 0.13048416376113892, 0.14339151978492737], "cos_Ac_negg": [0.5134156346321106, 0.4426364302635193, 0.360500693321228], "r_norm": [0.7761707901954651, 0.7813184857368469, 0.7995097041130066]}, {"step": 300, "epoch": 0, "train_loss": 0.25752902030944824, "cos_r_negg": [0.528708815574646, 0.4626166522502899, 0.37011367082595825], "cos_innovation_negg": [0.528708815574646, 0.4626166522502899, 0.37011367082595825], "cos_apical_negg": [0.10177868604660034, 0.12452653795480728, 0.126957505941391], "cos_Ac_negg": [0.5307761430740356, 0.46267470717430115, 0.36981648206710815], "r_norm": [0.5919914245605469, 0.5966252684593201, 0.6120467185974121]}, {"step": 400, "epoch": 0, "train_loss": 0.13028958439826965, "cos_r_negg": [0.5409702062606812, 0.4840485453605652, 0.42399299144744873], "cos_innovation_negg": [0.5409702062606812, 0.4840485453605652, 0.42399299144744873], "cos_apical_negg": [0.11041324585676193, 0.11936397850513458, 0.12784609198570251], "cos_Ac_negg": [0.5410827994346619, 0.4839455187320709, 0.4234982132911682], "r_norm": [0.5470607876777649, 0.5520876049995422, 0.5642469525337219]}, {"epoch_end": 0, "step": 469, "test_acc": 0.947, "test_loss": 0.17841850128173828}, {"step": 500, "epoch": 1, "train_loss": 0.1807648241519928, "cos_r_negg": [0.5348552465438843, 0.5115889310836792, 0.4514876902103424], "cos_innovation_negg": [0.5348552465438843, 0.5115889310836792, 0.4514876902103424], "cos_apical_negg": [0.0851651281118393, 0.11966394633054733, 0.14762669801712036], "cos_Ac_negg": [0.5365169644355774, 0.5125844478607178, 0.4526808261871338], "r_norm": [0.5111985206604004, 0.5184361934661865, 0.5298259258270264], "ldr": {"dL_sdil": -0.01582382619380951, "dL_bp": -0.013664066791534424, "ratio": 1.1580612445200549}}, {"step": 600, "epoch": 1, "train_loss": 0.07896888256072998, "cos_r_negg": [0.5404101610183716, 0.5252118706703186, 0.42411017417907715], "cos_innovation_negg": [0.5404101610183716, 0.5252118706703186, 0.42411017417907715], "cos_apical_negg": [0.07069253921508789, 0.10805739462375641, 0.1227489709854126], "cos_Ac_negg": [0.5426425933837891, 0.5250234007835388, 0.42436668276786804], "r_norm": [0.4426986277103424, 0.4445212185382843, 0.45536065101623535]}, {"step": 700, "epoch": 1, "train_loss": 0.1936294138431549, "cos_r_negg": [0.5523302555084229, 0.5313393473625183, 0.4363704025745392], "cos_innovation_negg": [0.5523302555084229, 0.5313393473625183, 0.4363704025745392], "cos_apical_negg": [0.07309302687644958, 0.12037917226552963, 0.12745285034179688], "cos_Ac_negg": [0.5567302703857422, 0.5333513617515564, 0.439350962638855], "r_norm": [0.47314751148223877, 0.47849762439727783, 0.4916081428527832]}, {"step": 800, "epoch": 1, "train_loss": 0.17381304502487183, "cos_r_negg": [0.5510797500610352, 0.5477213859558105, 0.4589617848396301], "cos_innovation_negg": [0.5510797500610352, 0.5477213859558105, 0.4589617848396301], "cos_apical_negg": [0.0753069743514061, 0.10238130390644073, 0.13298794627189636], "cos_Ac_negg": [0.5540711283683777, 0.5492314100265503, 0.4595499634742737], "r_norm": [0.36570852994918823, 0.3675145208835602, 0.3761633038520813]}, {"step": 900, "epoch": 1, "train_loss": 0.2041294127702713, "cos_r_negg": [0.5688360929489136, 0.5709272623062134, 0.4859267473220825], "cos_innovation_negg": [0.5688360929489136, 0.5709272623062134, 0.4859267473220825], "cos_apical_negg": [0.07023810595273972, 0.12307806313037872, 0.14232924580574036], "cos_Ac_negg": [0.5693740844726562, 0.5711548328399658, 0.48617517948150635], "r_norm": [0.47735312581062317, 0.4816506505012512, 0.48987314105033875]}, {"epoch_end": 1, "step": 938, "test_acc": 0.9522, "test_loss": 0.164680269241333}, {"step": 1000, "epoch": 2, "train_loss": 0.03263162076473236, "cos_r_negg": [0.5729007720947266, 0.5703027248382568, 0.4814510941505432], "cos_innovation_negg": [0.5729007720947266, 0.5703027248382568, 0.4814510941505432], "cos_apical_negg": [0.05268578976392746, 0.09828369319438934, 0.12499398738145828], "cos_Ac_negg": [0.5749251842498779, 0.5714312791824341, 0.4817565381526947], "r_norm": [0.3504492938518524, 0.3510257601737976, 0.3581618666648865], "ldr": {"dL_sdil": -0.008551239967346191, "dL_bp": -0.0071846842765808105, "ratio": 1.1902040003650272}}, {"step": 1100, "epoch": 2, "train_loss": 0.13308821618556976, "cos_r_negg": [0.5845746994018555, 0.5814930200576782, 0.48799845576286316], "cos_innovation_negg": [0.5845746994018555, 0.5814930200576782, 0.48799845576286316], "cos_apical_negg": [0.057091135531663895, 0.10458144545555115, 0.14837689697742462], "cos_Ac_negg": [0.5915025472640991, 0.5857481956481934, 0.48922598361968994], "r_norm": [0.38480323553085327, 0.38755786418914795, 0.39720049500465393]}, {"step": 1200, "epoch": 2, "train_loss": 0.1610100418329239, "cos_r_negg": [0.5727114677429199, 0.5879212617874146, 0.49369505047798157], "cos_innovation_negg": [0.5727114677429199, 0.5879212617874146, 0.49369505047798157], "cos_apical_negg": [0.03205304965376854, 0.0988479033112526, 0.1303451955318451], "cos_Ac_negg": [0.5806068181991577, 0.5938493013381958, 0.49549737572669983], "r_norm": [0.316397100687027, 0.31673741340637207, 0.32411086559295654]}, {"step": 1300, "epoch": 2, "train_loss": 0.12316633760929108, "cos_r_negg": [0.5860399007797241, 0.5971480011940002, 0.48872262239456177], "cos_innovation_negg": [0.5860399007797241, 0.5971480011940002, 0.48872262239456177], "cos_apical_negg": [0.043977562338113785, 0.09919430315494537, 0.13200163841247559], "cos_Ac_negg": [0.5901158452033997, 0.5985755324363708, 0.48968470096588135], "r_norm": [0.31413739919662476, 0.3152046799659729, 0.32350340485572815]}, {"step": 1400, "epoch": 2, "train_loss": 0.13651013374328613, "cos_r_negg": [0.5844224691390991, 0.5855025053024292, 0.5037516951560974], "cos_innovation_negg": [0.5844224691390991, 0.5855025053024292, 0.5037516951560974], "cos_apical_negg": [0.05487751215696335, 0.09212378412485123, 0.12417959421873093], "cos_Ac_negg": [0.5932061672210693, 0.5950719118118286, 0.5087454319000244], "r_norm": [0.30646589398384094, 0.30732524394989014, 0.315845251083374]}, {"epoch_end": 2, "step": 1407, "test_acc": 0.9648, "test_loss": 0.1194626262664795}, {"step": 1500, "epoch": 3, "train_loss": 0.07648062705993652, "cos_r_negg": [0.582809329032898, 0.5990599393844604, 0.5306364297866821], "cos_innovation_negg": [0.582809329032898, 0.5990599393844604, 0.5306364297866821], "cos_apical_negg": [0.05203244462609291, 0.07996759563684464, 0.1280796378850937], "cos_Ac_negg": [0.5888310074806213, 0.6044595241546631, 0.5320906639099121], "r_norm": [0.26579931378364563, 0.2674559950828552, 0.2740623354911804], "ldr": {"dL_sdil": -0.006355859339237213, "dL_bp": -0.005833715200424194, "ratio": 1.089504564565485}}, {"step": 1600, "epoch": 3, "train_loss": 0.0905633419752121, "cos_r_negg": [0.5723593235015869, 0.5879327058792114, 0.5362997055053711], "cos_innovation_negg": [0.5723593235015869, 0.5879327058792114, 0.5362997055053711], "cos_apical_negg": [0.04090467840433121, 0.09057846665382385, 0.13063514232635498], "cos_Ac_negg": [0.582373857498169, 0.599800705909729, 0.5396597385406494], "r_norm": [0.3216865658760071, 0.32372450828552246, 0.32970255613327026]}, {"step": 1700, "epoch": 3, "train_loss": 0.03675437346100807, "cos_r_negg": [0.5639013648033142, 0.5984659790992737, 0.5404385328292847], "cos_innovation_negg": [0.5639013648033142, 0.5984659790992737, 0.5404385328292847], "cos_apical_negg": [0.0356726348400116, 0.08682107925415039, 0.12229768186807632], "cos_Ac_negg": [0.5738296508789062, 0.6086991429328918, 0.5423985719680786], "r_norm": [0.3025714159011841, 0.302367240190506, 0.30920514464378357]}, {"step": 1800, "epoch": 3, "train_loss": 0.012730604037642479, "cos_r_negg": [0.5734977722167969, 0.5982744693756104, 0.5369141101837158], "cos_innovation_negg": [0.5734977722167969, 0.5982744693756104, 0.5369141101837158], "cos_apical_negg": [0.03080776333808899, 0.08617154508829117, 0.11317454278469086], "cos_Ac_negg": [0.586790144443512, 0.6150208711624146, 0.5431820154190063], "r_norm": [0.2756425440311432, 0.27534306049346924, 0.2815254330635071]}, {"epoch_end": 3, "step": 1876, "test_acc": 0.9663, "test_loss": 0.11451079864501953}, {"step": 1900, "epoch": 4, "train_loss": 0.032762207090854645, "cos_r_negg": [0.5686196088790894, 0.6121808290481567, 0.5366095900535583], "cos_innovation_negg": [0.5686196088790894, 0.6121808290481567, 0.5366095900535583], "cos_apical_negg": [0.019442813470959663, 0.07997725903987885, 0.11869443953037262], "cos_Ac_negg": [0.5839101672172546, 0.6255109310150146, 0.5402666926383972], "r_norm": [0.2687967121601105, 0.26841211318969727, 0.2760213613510132]}, {"step": 2000, "epoch": 4, "train_loss": 0.07402737438678741, "cos_r_negg": [0.550966203212738, 0.5759673714637756, 0.5259472131729126], "cos_innovation_negg": [0.550966203212738, 0.5759673714637756, 0.5259472131729126], "cos_apical_negg": [0.03513997420668602, 0.08047322928905487, 0.12142512202262878], "cos_Ac_negg": [0.5936437845230103, 0.6221067905426025, 0.5487080812454224], "r_norm": [0.2583365738391876, 0.2562132775783539, 0.2620803713798523], "ldr": {"dL_sdil": -0.008371703326702118, "dL_bp": -0.00857352465391159, "ratio": 0.976459935049304}}, {"step": 2100, "epoch": 4, "train_loss": 0.042866051197052, "cos_r_negg": [0.5705348253250122, 0.599862813949585, 0.5541979074478149], "cos_innovation_negg": [0.5705348253250122, 0.599862813949585, 0.5541979074478149], "cos_apical_negg": [0.05051989108324051, 0.08365020155906677, 0.13147062063217163], "cos_Ac_negg": [0.5924770832061768, 0.6245070695877075, 0.5610617399215698], "r_norm": [0.28235897421836853, 0.2823737859725952, 0.2904585003852844]}, {"step": 2200, "epoch": 4, "train_loss": 0.010305065661668777, "cos_r_negg": [0.5772885680198669, 0.607243537902832, 0.5564123392105103], "cos_innovation_negg": [0.5772885680198669, 0.607243537902832, 0.5564123392105103], "cos_apical_negg": [0.03302644193172455, 0.07795511931180954, 0.1127902939915657], "cos_Ac_negg": [0.6015622615814209, 0.632867693901062, 0.5630269646644592], "r_norm": [0.25257766246795654, 0.25193700194358826, 0.25840598344802856]}, {"step": 2300, "epoch": 4, "train_loss": 0.15979208052158356, "cos_r_negg": [0.5494014024734497, 0.5776619911193848, 0.5485407710075378], "cos_innovation_negg": [0.5494014024734497, 0.5776619911193848, 0.5485407710075378], "cos_apical_negg": [0.03557701036334038, 0.07048080116510391, 0.11928121745586395], "cos_Ac_negg": [0.5928793549537659, 0.6328431367874146, 0.5666439533233643], "r_norm": [0.25622111558914185, 0.25602421164512634, 0.26393648982048035]}, {"epoch_end": 4, "step": 2345, "test_acc": 0.9707, "test_loss": 0.11497614192962646}, {"step": 2400, "epoch": 5, "train_loss": 0.1078932136297226, "cos_r_negg": [0.5551625490188599, 0.5934861898422241, 0.5524869561195374], "cos_innovation_negg": [0.5551625490188599, 0.5934861898422241, 0.5524869561195374], "cos_apical_negg": [0.028840474784374237, 0.06992577016353607, 0.11109043657779694], "cos_Ac_negg": [0.6028305292129517, 0.6434247493743896, 0.5699541568756104], "r_norm": [0.21299418807029724, 0.21166028082370758, 0.21788239479064941]}, {"step": 2500, "epoch": 5, "train_loss": 0.02374871075153351, "cos_r_negg": [0.535438060760498, 0.5594796538352966, 0.5295423269271851], "cos_innovation_negg": [0.535438060760498, 0.5594796538352966, 0.5295423269271851], "cos_apical_negg": [0.023892957717180252, 0.06995686888694763, 0.12139956653118134], "cos_Ac_negg": [0.6035381555557251, 0.6296474933624268, 0.5638284087181091], "r_norm": [0.2603135108947754, 0.2593501806259155, 0.26533472537994385], "ldr": {"dL_sdil": -0.009871810674667358, "dL_bp": -0.011774621903896332, "ratio": 0.8383972543017015}}, {"step": 2600, "epoch": 5, "train_loss": 0.05304960906505585, "cos_r_negg": [0.52979576587677, 0.5670466423034668, 0.5479004979133606], "cos_innovation_negg": [0.52979576587677, 0.5670466423034668, 0.5479004979133606], "cos_apical_negg": [0.04220541939139366, 0.08032345771789551, 0.1262071281671524], "cos_Ac_negg": [0.6043375730514526, 0.6363017559051514, 0.58698570728302], "r_norm": [0.22474563121795654, 0.22321724891662598, 0.23007771372795105]}, {"step": 2700, "epoch": 5, "train_loss": 0.034139081835746765, "cos_r_negg": [0.529230535030365, 0.5677273273468018, 0.539835512638092], "cos_innovation_negg": [0.529230535030365, 0.5677273273468018, 0.539835512638092], "cos_apical_negg": [0.024656590074300766, 0.07033709436655045, 0.11462977528572083], "cos_Ac_negg": [0.6117772459983826, 0.6435583829879761, 0.5806998014450073], "r_norm": [0.2135850340127945, 0.21193356812000275, 0.21844452619552612]}, {"step": 2800, "epoch": 5, "train_loss": 0.034261301159858704, "cos_r_negg": [0.5269429683685303, 0.5670950412750244, 0.5415679216384888], "cos_innovation_negg": [0.5269429683685303, 0.5670950412750244, 0.5415679216384888], "cos_apical_negg": [0.021020326763391495, 0.07434876263141632, 0.11197115480899811], "cos_Ac_negg": [0.6107158660888672, 0.642261266708374, 0.5806410312652588], "r_norm": [0.2211204618215561, 0.21918120980262756, 0.2257131189107895]}, {"epoch_end": 5, "step": 2814, "test_acc": 0.97, "test_loss": 0.11780296745300294}, {"step": 2900, "epoch": 6, "train_loss": 0.011397719383239746, "cos_r_negg": [0.5219845771789551, 0.5616589784622192, 0.542546808719635], "cos_innovation_negg": [0.5219845771789551, 0.5616589784622192, 0.542546808719635], "cos_apical_negg": [0.02675338089466095, 0.07400087267160416, 0.11950907111167908], "cos_Ac_negg": [0.6122092008590698, 0.6477441787719727, 0.5883749723434448], "r_norm": [0.23567363619804382, 0.2354898452758789, 0.24229687452316284]}, {"step": 3000, "epoch": 6, "train_loss": 0.06766141206026077, "cos_r_negg": [0.5165479183197021, 0.5659703612327576, 0.5425368547439575], "cos_innovation_negg": [0.5165479183197021, 0.5659703612327576, 0.5425368547439575], "cos_apical_negg": [0.018119461834430695, 0.07276603579521179, 0.11712069064378738], "cos_Ac_negg": [0.6148598194122314, 0.6590943932533264, 0.5922616720199585], "r_norm": [0.2285493016242981, 0.2258799523115158, 0.23218761384487152], "ldr": {"dL_sdil": -0.008624427020549774, "dL_bp": -0.011231310665607452, "ratio": 0.7678914133289462}}, {"step": 3100, "epoch": 6, "train_loss": 0.011940602213144302, "cos_r_negg": [0.49066847562789917, 0.5448637008666992, 0.5339353084564209], "cos_innovation_negg": [0.49066847562789917, 0.5448637008666992, 0.5339353084564209], "cos_apical_negg": [0.007285827770829201, 0.06124396622180939, 0.11916914582252502], "cos_Ac_negg": [0.6013911962509155, 0.6561592817306519, 0.5947411060333252], "r_norm": [0.175805002450943, 0.17356181144714355, 0.17831215262413025]}, {"step": 3200, "epoch": 6, "train_loss": 0.01863642781972885, "cos_r_negg": [0.4804835319519043, 0.5327173471450806, 0.5243409872055054], "cos_innovation_negg": [0.4804835319519043, 0.5327173471450806, 0.5243409872055054], "cos_apical_negg": [0.013016757555305958, 0.061587098985910416, 0.1157207190990448], "cos_Ac_negg": [0.6106147170066833, 0.6626662015914917, 0.5968565344810486], "r_norm": [0.1887008249759674, 0.18547572195529938, 0.19113068282604218]}, {"epoch_end": 6, "step": 3283, "test_acc": 0.9738, "test_loss": 0.11375721931457519}, {"step": 3300, "epoch": 7, "train_loss": 0.010094470344483852, "cos_r_negg": [0.4903387427330017, 0.5452684760093689, 0.5329725742340088], "cos_innovation_negg": [0.4903387427330017, 0.5452684760093689, 0.5329725742340088], "cos_apical_negg": [0.023183079436421394, 0.06250479817390442, 0.1161399632692337], "cos_Ac_negg": [0.609476625919342, 0.662117600440979, 0.600680947303772], "r_norm": [0.1918632984161377, 0.18966834247112274, 0.19530579447746277]}, {"step": 3400, "epoch": 7, "train_loss": 0.028262287378311157, "cos_r_negg": [0.4746207594871521, 0.5208358764648438, 0.5210437178611755], "cos_innovation_negg": [0.4746207594871521, 0.5208358764648438, 0.5210437178611755], "cos_apical_negg": [0.030959200114011765, 0.064633809030056, 0.11897359043359756], "cos_Ac_negg": [0.6199973821640015, 0.6660165786743164, 0.6130518913269043], "r_norm": [0.237336665391922, 0.23466970026493073, 0.2413531392812729]}, {"step": 3500, "epoch": 7, "train_loss": 0.012605790048837662, "cos_r_negg": [0.4489513337612152, 0.5001634359359741, 0.5048023462295532], "cos_innovation_negg": [0.4489513337612152, 0.5001634359359741, 0.5048023462295532], "cos_apical_negg": [0.011033246293663979, 0.06124911829829216, 0.11606670916080475], "cos_Ac_negg": [0.6154003739356995, 0.664509654045105, 0.619300365447998], "r_norm": [0.18742799758911133, 0.18515513837337494, 0.1903511881828308], "ldr": {"dL_sdil": -0.00667426735162735, "dL_bp": -0.009379982948303223, "ratio": 0.7115436550803839}}, {"step": 3600, "epoch": 7, "train_loss": 0.002501162001863122, "cos_r_negg": [0.4633944630622864, 0.5094193816184998, 0.5131481885910034], "cos_innovation_negg": [0.4633944630622864, 0.5094193816184998, 0.5131481885910034], "cos_apical_negg": [0.014765365980565548, 0.06386268138885498, 0.11611031740903854], "cos_Ac_negg": [0.6221897602081299, 0.6654284000396729, 0.6188482046127319], "r_norm": [0.19960951805114746, 0.19723404943943024, 0.20290613174438477]}, {"step": 3700, "epoch": 7, "train_loss": 0.047592177987098694, "cos_r_negg": [0.4487311840057373, 0.4822712242603302, 0.48509272933006287], "cos_innovation_negg": [0.4487311840057373, 0.4822712242603302, 0.48509272933006287], "cos_apical_negg": [0.016713546589016914, 0.06394006311893463, 0.10898618400096893], "cos_Ac_negg": [0.6234561204910278, 0.6643760204315186, 0.6124838590621948], "r_norm": [0.22732378542423248, 0.224981427192688, 0.23132050037384033]}, {"epoch_end": 7, "step": 3752, "test_acc": 0.9739, "test_loss": 0.1135302433013916}, {"step": 3800, "epoch": 8, "train_loss": 0.003513964358717203, "cos_r_negg": [0.42559632658958435, 0.46762120723724365, 0.46973276138305664], "cos_innovation_negg": [0.42559632658958435, 0.46762120723724365, 0.46973276138305664], "cos_apical_negg": [0.01379104983061552, 0.06488192081451416, 0.10912670195102692], "cos_Ac_negg": [0.6189025044441223, 0.6681501865386963, 0.6110430955886841], "r_norm": [0.19996729493141174, 0.1962798684835434, 0.20215703547000885]}, {"step": 3900, "epoch": 8, "train_loss": 0.017354629933834076, "cos_r_negg": [0.3996793031692505, 0.45104163885116577, 0.4575529396533966], "cos_innovation_negg": [0.3996793031692505, 0.45104163885116577, 0.4575529396533966], "cos_apical_negg": [0.01126004196703434, 0.0607152059674263, 0.1086505800485611], "cos_Ac_negg": [0.6127740740776062, 0.6692689061164856, 0.6164525747299194], "r_norm": [0.1811152696609497, 0.17744076251983643, 0.18241417407989502]}, {"step": 4000, "epoch": 8, "train_loss": 0.004856348503381014, "cos_r_negg": [0.37602561712265015, 0.41966673731803894, 0.43439748883247375], "cos_innovation_negg": [0.37602561712265015, 0.41966673731803894, 0.43439748883247375], "cos_apical_negg": [0.008020796813070774, 0.057990677654743195, 0.09969477355480194], "cos_Ac_negg": [0.6132957339286804, 0.6661556959152222, 0.6181684136390686], "r_norm": [0.16387838125228882, 0.16086477041244507, 0.16547046601772308], "ldr": {"dL_sdil": -0.007244512438774109, "dL_bp": -0.011032260954380035, "ratio": 0.6566661601580307}}, {"step": 4100, "epoch": 8, "train_loss": 0.013123110868036747, "cos_r_negg": [0.4056463837623596, 0.45457643270492554, 0.4590320885181427], "cos_innovation_negg": [0.4056463837623596, 0.45457643270492554, 0.4590320885181427], "cos_apical_negg": [0.011253751814365387, 0.06664575636386871, 0.10164335370063782], "cos_Ac_negg": [0.6166028380393982, 0.6658480167388916, 0.621646523475647], "r_norm": [0.20974087715148926, 0.2073712944984436, 0.21274694800376892]}, {"step": 4200, "epoch": 8, "train_loss": 0.020602025091648102, "cos_r_negg": [0.37572383880615234, 0.4216482937335968, 0.44565603137016296], "cos_innovation_negg": [0.37572383880615234, 0.4216482937335968, 0.44565603137016296], "cos_apical_negg": [0.003575623035430908, 0.05617605894804001, 0.10992729663848877], "cos_Ac_negg": [0.6122173070907593, 0.6653400659561157, 0.627249002456665], "r_norm": [0.2008437216281891, 0.1983548402786255, 0.2032732367515564]}, {"epoch_end": 8, "step": 4221, "test_acc": 0.9739, "test_loss": 0.12465989742279053}, {"step": 4300, "epoch": 9, "train_loss": 0.005610136780887842, "cos_r_negg": [0.3538643717765808, 0.3970605731010437, 0.42124608159065247], "cos_innovation_negg": [0.3538643717765808, 0.3970605731010437, 0.42124608159065247], "cos_apical_negg": [0.007544706575572491, 0.056428320705890656, 0.11097771674394608], "cos_Ac_negg": [0.619644045829773, 0.6742088794708252, 0.6325809955596924], "r_norm": [0.20761768519878387, 0.20515799522399902, 0.20977666974067688]}, {"step": 4400, "epoch": 9, "train_loss": 0.005376679822802544, "cos_r_negg": [0.3503841161727905, 0.3912612795829773, 0.40973806381225586], "cos_innovation_negg": [0.3503841161727905, 0.3912612795829773, 0.40973806381225586], "cos_apical_negg": [0.009758060798048973, 0.05871321260929108, 0.10685720294713974], "cos_Ac_negg": [0.6169899106025696, 0.6703310012817383, 0.6331541538238525], "r_norm": [0.20245492458343506, 0.19962534308433533, 0.2046738564968109]}, {"step": 4500, "epoch": 9, "train_loss": 0.036763302981853485, "cos_r_negg": [0.33551520109176636, 0.37393367290496826, 0.3952620327472687], "cos_innovation_negg": [0.33551520109176636, 0.37393367290496826, 0.3952620327472687], "cos_apical_negg": [0.007033657282590866, 0.0598057396709919, 0.10435810685157776], "cos_Ac_negg": [0.619627058506012, 0.667587161064148, 0.6292416453361511], "r_norm": [0.20359715819358826, 0.201330304145813, 0.20609556138515472], "ldr": {"dL_sdil": -0.010631203651428223, "dL_bp": -0.01770325005054474, "ratio": 0.6005227074732018}}, {"step": 4600, "epoch": 9, "train_loss": 0.0037832045927643776, "cos_r_negg": [0.3121258616447449, 0.36032769083976746, 0.3776133954524994], "cos_innovation_negg": [0.3121258616447449, 0.36032769083976746, 0.3776133954524994], "cos_apical_negg": [0.015989728271961212, 0.06058131158351898, 0.10416124761104584], "cos_Ac_negg": [0.6033607721328735, 0.6596585512161255, 0.623542070388794], "r_norm": [0.2018270641565323, 0.19944502413272858, 0.2047208696603775]}, {"epoch_end": 9, "step": 4690, "test_acc": 0.9753, "test_loss": 0.12550905532836915}, {"step": 4700, "epoch": 10, "train_loss": 0.003072786144912243, "cos_r_negg": [0.2944188416004181, 0.32908353209495544, 0.35645240545272827], "cos_innovation_negg": [0.2944188416004181, 0.32908353209495544, 0.35645240545272827], "cos_apical_negg": [0.016930514946579933, 0.05618127062916756, 0.10497274249792099], "cos_Ac_negg": [0.6030189990997314, 0.655601441860199, 0.6225085854530334], "r_norm": [0.20414391160011292, 0.2015935778617859, 0.2070946842432022]}, {"step": 4800, "epoch": 10, "train_loss": 0.0008679613820277154, "cos_r_negg": [0.2837217450141907, 0.3221810460090637, 0.34900766611099243], "cos_innovation_negg": [0.2837217450141907, 0.3221810460090637, 0.34900766611099243], "cos_apical_negg": [0.015292130410671234, 0.05809130147099495, 0.10039294511079788], "cos_Ac_negg": [0.6085541844367981, 0.6584851145744324, 0.6256493330001831], "r_norm": [0.20482318103313446, 0.20216301083564758, 0.20747578144073486]}, {"step": 4900, "epoch": 10, "train_loss": 0.000689456588588655, "cos_r_negg": [0.29323333501815796, 0.33389848470687866, 0.3601685166358948], "cos_innovation_negg": [0.29323333501815796, 0.33389848470687866, 0.3601685166358948], "cos_apical_negg": [0.013220809400081635, 0.0575682707130909, 0.10602718591690063], "cos_Ac_negg": [0.6104445457458496, 0.6623284816741943, 0.6290529370307922], "r_norm": [0.20441797375679016, 0.2013956904411316, 0.20674476027488708]}, {"step": 5000, "epoch": 10, "train_loss": 0.002489793812856078, "cos_r_negg": [0.271761029958725, 0.3122059106826782, 0.33744847774505615], "cos_innovation_negg": [0.271761029958725, 0.3122059106826782, 0.33744847774505615], "cos_apical_negg": [0.004920518957078457, 0.052379921078681946, 0.09670887887477875], "cos_Ac_negg": [0.6035255789756775, 0.6595746278762817, 0.6313233971595764], "r_norm": [0.1856590211391449, 0.1828835904598236, 0.18777750432491302], "ldr": {"dL_sdil": -0.00991879403591156, "dL_bp": -0.017432942986488342, "ratio": 0.5689684205127766}}, {"step": 5100, "epoch": 10, "train_loss": 0.004342309199273586, "cos_r_negg": [0.2780485153198242, 0.3129868805408478, 0.3403642177581787], "cos_innovation_negg": [0.2780485153198242, 0.3129868805408478, 0.3403642177581787], "cos_apical_negg": [0.011281654238700867, 0.0573209710419178, 0.10063013434410095], "cos_Ac_negg": [0.606847882270813, 0.6650934219360352, 0.6345300078392029], "r_norm": [0.20424160361289978, 0.20170949399471283, 0.20710095763206482]}, {"epoch_end": 10, "step": 5159, "test_acc": 0.9749, "test_loss": 0.13234355907440185}, {"step": 5200, "epoch": 11, "train_loss": 0.0003710065211635083, "cos_r_negg": [0.26866614818573, 0.3031820058822632, 0.3304052948951721], "cos_innovation_negg": [0.26866614818573, 0.3031820058822632, 0.3304052948951721], "cos_apical_negg": [0.00830305926501751, 0.054413020610809326, 0.09989627450704575], "cos_Ac_negg": [0.6103471517562866, 0.6627001762390137, 0.6382420659065247], "r_norm": [0.19947873055934906, 0.19647422432899475, 0.20194187760353088]}, {"step": 5300, "epoch": 11, "train_loss": 0.001310388557612896, "cos_r_negg": [0.26046091318130493, 0.29670438170433044, 0.3230708837509155], "cos_innovation_negg": [0.26046091318130493, 0.29670438170433044, 0.3230708837509155], "cos_apical_negg": [0.003624703735113144, 0.0541350282728672, 0.09748238325119019], "cos_Ac_negg": [0.6059515476226807, 0.6595721244812012, 0.6382764577865601], "r_norm": [0.1994665414094925, 0.19642436504364014, 0.20184189081192017]}, {"step": 5400, "epoch": 11, "train_loss": 0.002470809267833829, "cos_r_negg": [0.25448858737945557, 0.2878149151802063, 0.3164098560810089], "cos_innovation_negg": [0.25448858737945557, 0.2878149151802063, 0.3164098560810089], "cos_apical_negg": [0.00452026491984725, 0.05260743573307991, 0.09517088532447815], "cos_Ac_negg": [0.6073060035705566, 0.6594626903533936, 0.6362428665161133], "r_norm": [0.1941535770893097, 0.19129878282546997, 0.1962924301624298]}, {"step": 5500, "epoch": 11, "train_loss": 0.002368900226429105, "cos_r_negg": [0.24936242401599884, 0.28135228157043457, 0.3080633878707886], "cos_innovation_negg": [0.24936242401599884, 0.28135228157043457, 0.3080633878707886], "cos_apical_negg": [0.00716947577893734, 0.051660820841789246, 0.091831274330616], "cos_Ac_negg": [0.6069409847259521, 0.6583317518234253, 0.6341017484664917], "r_norm": [0.1937277764081955, 0.1909867823123932, 0.19606642425060272], "ldr": {"dL_sdil": -0.01121484488248825, "dL_bp": -0.02140107750892639, "ratio": 0.5240317866149747}}, {"step": 5600, "epoch": 11, "train_loss": 0.0006278679938986897, "cos_r_negg": [0.24199867248535156, 0.2709590792655945, 0.2983156740665436], "cos_innovation_negg": [0.24199867248535156, 0.2709590792655945, 0.2983156740665436], "cos_apical_negg": [0.00971236452460289, 0.05224290490150452, 0.09043112397193909], "cos_Ac_negg": [0.6023350954055786, 0.6577886939048767, 0.6342024207115173], "r_norm": [0.19779030978679657, 0.1949300467967987, 0.20000602304935455]}, {"epoch_end": 11, "step": 5628, "test_acc": 0.9746, "test_loss": 0.13619753665924073}, {"step": 5700, "epoch": 12, "train_loss": 0.0006151688285171986, "cos_r_negg": [0.23634593188762665, 0.2623288035392761, 0.2931305766105652], "cos_innovation_negg": [0.23634593188762665, 0.2623288035392761, 0.2931305766105652], "cos_apical_negg": [0.006812026724219322, 0.04711467772722244, 0.08773602545261383], "cos_Ac_negg": [0.6036950945854187, 0.6565465927124023, 0.6355303525924683], "r_norm": [0.19501236081123352, 0.19225510954856873, 0.19715900719165802]}, {"step": 5800, "epoch": 12, "train_loss": 0.0005252254195511341, "cos_r_negg": [0.22883166372776031, 0.2535156011581421, 0.28534024953842163], "cos_innovation_negg": [0.22883166372776031, 0.2535156011581421, 0.28534024953842163], "cos_apical_negg": [0.00547691248357296, 0.04718494415283203, 0.08723905682563782], "cos_Ac_negg": [0.6058955192565918, 0.657025933265686, 0.6364992260932922], "r_norm": [0.1940699964761734, 0.19129586219787598, 0.1961495280265808]}, {"step": 5900, "epoch": 12, "train_loss": 0.001078218687325716, "cos_r_negg": [0.2313290238380432, 0.2567785084247589, 0.2869266867637634], "cos_innovation_negg": [0.2313290238380432, 0.2567785084247589, 0.2869266867637634], "cos_apical_negg": [0.005998366512358189, 0.04966701194643974, 0.0869307741522789], "cos_Ac_negg": [0.6020839214324951, 0.6555049419403076, 0.6366939544677734], "r_norm": [0.1965775191783905, 0.1937912553548813, 0.19868679344654083]}, {"step": 6000, "epoch": 12, "train_loss": 0.0009056039853021502, "cos_r_negg": [0.22886089980602264, 0.2550346851348877, 0.2855684757232666], "cos_innovation_negg": [0.22886089980602264, 0.2550346851348877, 0.2855684757232666], "cos_apical_negg": [0.0063621834851801395, 0.048350535333156586, 0.08751896023750305], "cos_Ac_negg": [0.6048064231872559, 0.6563602685928345, 0.637134313583374], "r_norm": [0.1951221227645874, 0.19225259125232697, 0.19705694913864136], "ldr": {"dL_sdil": -0.011640243232250214, "dL_bp": -0.023864418268203735, "ratio": 0.4877656392638466}}, {"epoch_end": 12, "step": 6097, "test_acc": 0.9752, "test_loss": 0.13830832386016845}, {"step": 6100, "epoch": 13, "train_loss": 0.00048515570233576, "cos_r_negg": [0.21801647543907166, 0.2452058345079422, 0.27572694420814514], "cos_innovation_negg": [0.21801647543907166, 0.2452058345079422, 0.27572694420814514], "cos_apical_negg": [0.004781823605298996, 0.047014400362968445, 0.0871681421995163], "cos_Ac_negg": [0.6030107140541077, 0.6560391187667847, 0.639000654220581], "r_norm": [0.19681480526924133, 0.19402167201042175, 0.199020117521286]}, {"step": 6200, "epoch": 13, "train_loss": 0.001061578281223774, "cos_r_negg": [0.21340787410736084, 0.23987334966659546, 0.2705043852329254], "cos_innovation_negg": [0.21340787410736084, 0.23987334966659546, 0.2705043852329254], "cos_apical_negg": [0.004939937964081764, 0.04700884968042374, 0.0865381509065628], "cos_Ac_negg": [0.603725790977478, 0.6578191518783569, 0.6400411128997803], "r_norm": [0.19550463557243347, 0.19258254766464233, 0.19753095507621765]}, {"step": 6300, "epoch": 13, "train_loss": 0.0013569598086178303, "cos_r_negg": [0.21028536558151245, 0.23214983940124512, 0.2632784843444824], "cos_innovation_negg": [0.21028536558151245, 0.23214983940124512, 0.2632784843444824], "cos_apical_negg": [0.006486337631940842, 0.04463715851306915, 0.08417961001396179], "cos_Ac_negg": [0.602252721786499, 0.6572246551513672, 0.6402424573898315], "r_norm": [0.19806444644927979, 0.19515500962734222, 0.2000543177127838]}, {"step": 6400, "epoch": 13, "train_loss": 0.00040100750629790127, "cos_r_negg": [0.21405652165412903, 0.23466511070728302, 0.26365357637405396], "cos_innovation_negg": [0.21405652165412903, 0.23466511070728302, 0.26365357637405396], "cos_apical_negg": [0.008446392603218555, 0.046153582632541656, 0.085267573595047], "cos_Ac_negg": [0.6010825634002686, 0.6574031114578247, 0.6403003931045532], "r_norm": [0.19844256341457367, 0.1955283284187317, 0.2004726529121399]}, {"step": 6500, "epoch": 13, "train_loss": 0.00013707700418308377, "cos_r_negg": [0.21101930737495422, 0.2340274453163147, 0.26236242055892944], "cos_innovation_negg": [0.21101930737495422, 0.2340274453163147, 0.26236242055892944], "cos_apical_negg": [0.0074137235060334206, 0.04547731950879097, 0.08196429908275604], "cos_Ac_negg": [0.5996854901313782, 0.6558645963668823, 0.6396642923355103], "r_norm": [0.19560542702674866, 0.19276125729084015, 0.19759288430213928], "ldr": {"dL_sdil": -0.012395992875099182, "dL_bp": -0.02577637881040573, "ratio": 0.480905132806902}}, {"epoch_end": 13, "step": 6566, "test_acc": 0.9745, "test_loss": 0.1421777862548828}, {"step": 6600, "epoch": 14, "train_loss": 0.0003673648461699486, "cos_r_negg": [0.20384685695171356, 0.22617018222808838, 0.25491297245025635], "cos_innovation_negg": [0.20384685695171356, 0.22617018222808838, 0.25491297245025635], "cos_apical_negg": [0.007694783620536327, 0.04521997645497322, 0.08155819773674011], "cos_Ac_negg": [0.6003413200378418, 0.6558160781860352, 0.6403923034667969], "r_norm": [0.19539327919483185, 0.19239622354507446, 0.19721531867980957]}, {"step": 6700, "epoch": 14, "train_loss": 0.0005099893314763904, "cos_r_negg": [0.2055099457502365, 0.2272024303674698, 0.25549423694610596], "cos_innovation_negg": [0.2055099457502365, 0.2272024303674698, 0.25549423694610596], "cos_apical_negg": [0.008829940110445023, 0.045250680297613144, 0.08019733428955078], "cos_Ac_negg": [0.5977998971939087, 0.6545171141624451, 0.6388810873031616], "r_norm": [0.19672875106334686, 0.1939796656370163, 0.19889411330223083]}, {"step": 6800, "epoch": 14, "train_loss": 0.0005156504339538515, "cos_r_negg": [0.19749385118484497, 0.2178269326686859, 0.24622316658496857], "cos_innovation_negg": [0.19749385118484497, 0.2178269326686859, 0.24622316658496857], "cos_apical_negg": [0.00926391500979662, 0.043578363955020905, 0.0787520557641983], "cos_Ac_negg": [0.5952610969543457, 0.6540374159812927, 0.6387388706207275], "r_norm": [0.19368970394134521, 0.19087189435958862, 0.19562256336212158]}, {"step": 6900, "epoch": 14, "train_loss": 0.0004047742404509336, "cos_r_negg": [0.19575990736484528, 0.21598784625530243, 0.2438393384218216], "cos_innovation_negg": [0.19575990736484528, 0.21598784625530243, 0.2438393384218216], "cos_apical_negg": [0.008007603697478771, 0.043969012796878815, 0.07669824361801147], "cos_Ac_negg": [0.5954526662826538, 0.6536595225334167, 0.63812255859375], "r_norm": [0.19185319542884827, 0.18893596529960632, 0.1936325877904892]}, {"step": 7000, "epoch": 14, "train_loss": 0.0006776193622499704, "cos_r_negg": [0.1946800947189331, 0.21558788418769836, 0.2416897714138031], "cos_innovation_negg": [0.1946800947189331, 0.21558788418769836, 0.2416897714138031], "cos_apical_negg": [0.005928122438490391, 0.04235561937093735, 0.07497568428516388], "cos_Ac_negg": [0.5977321863174438, 0.6545623540878296, 0.6404196619987488], "r_norm": [0.19190290570259094, 0.18885603547096252, 0.19356811046600342], "ldr": {"dL_sdil": -0.012736454606056213, "dL_bp": -0.027106821537017822, "ratio": 0.46986160249968667}}, {"epoch_end": 14, "step": 7035, "test_acc": 0.9743, "test_loss": 0.14520231761932373}], "final": {"test_acc": 0.9743, "test_loss": 0.14520231761932373, "wall_s": 22.254133224487305, "cos_r_negg": [0.1939845085144043, 0.21528545022010803, 0.24152889847755432], "cos_innovation_negg": [0.1939845085144043, 0.21528545022010803, 0.24152889847755432], "cos_apical_negg": [0.006461089476943016, 0.042778365314006805, 0.07549141347408295], "cos_Ac_negg": [0.596068799495697, 0.6536844372749329, 0.6399118900299072]}}
\ No newline at end of file |
