1
|
{"args": {"mode": "bp", "dataset": "fmnist", "depth": 7, "width": 256, "act": "tanh", "residual": 0, "epochs": 20, "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": 4, "use_residual": 1, "learn_A": 1, "learn_P": 1, "p_neutral": 1, "p_warmup_steps": 200, "p_warmup_eta": 0.05, "nuis_rho": 0.0, "normalize_delta": 0, "settle_steps": 0, "kappa": 0.0, "feedback": "error", "seed": 1, "device": "cuda", "log_every": 50, "max_steps": 0, "probe_bs": 512, "outdir": "results", "tag": "depth_fmnist_bp_d7_s1", "n_in": 784}, "steps": [{"step": 0, "epoch": 0, "train_loss": 2.3651061058044434}, {"step": 50, "epoch": 0, "train_loss": 0.6395633220672607}, {"step": 100, "epoch": 0, "train_loss": 0.6487054824829102}, {"step": 150, "epoch": 0, "train_loss": 0.4847731590270996}, {"step": 200, "epoch": 0, "train_loss": 0.4665953814983368}, {"step": 250, "epoch": 0, "train_loss": 0.4728131890296936}, {"step": 300, "epoch": 0, "train_loss": 0.6954860687255859}, {"step": 350, "epoch": 0, "train_loss": 0.4526720941066742}, {"step": 400, "epoch": 0, "train_loss": 0.5448930263519287}, {"step": 450, "epoch": 0, "train_loss": 0.49435555934906006}, {"epoch_end": 0, "step": 469, "test_acc": 0.842, "test_loss": 0.4577051788330078}, {"step": 500, "epoch": 1, "train_loss": 0.3384498059749603}, {"step": 550, "epoch": 1, "train_loss": 0.5434211492538452}, {"step": 600, "epoch": 1, "train_loss": 0.433641642332077}, {"step": 650, "epoch": 1, "train_loss": 0.5025926828384399}, {"step": 700, "epoch": 1, "train_loss": 0.36255010962486267}, {"step": 750, "epoch": 1, "train_loss": 0.3940623700618744}, {"step": 800, "epoch": 1, "train_loss": 0.4727742075920105}, {"step": 850, "epoch": 1, "train_loss": 0.23257189989089966}, {"step": 900, "epoch": 1, "train_loss": 0.4480585753917694}, {"epoch_end": 1, "step": 938, "test_acc": 0.8582, "test_loss": 0.38585543823242185}, {"step": 950, "epoch": 2, "train_loss": 0.35915422439575195}, {"step": 1000, "epoch": 2, "train_loss": 0.3213244676589966}, {"step": 1050, "epoch": 2, "train_loss": 0.29377466440200806}, {"step": 1100, "epoch": 2, "train_loss": 0.35081109404563904}, {"step": 1150, "epoch": 2, "train_loss": 0.35711967945098877}, {"step": 1200, "epoch": 2, "train_loss": 0.4872497618198395}, {"step": 1250, "epoch": 2, "train_loss": 0.21581903100013733}, {"step": 1300, "epoch": 2, "train_loss": 0.28140559792518616}, {"step": 1350, "epoch": 2, "train_loss": 0.18644744157791138}, {"step": 1400, "epoch": 2, "train_loss": 0.42379704117774963}, {"epoch_end": 2, "step": 1407, "test_acc": 0.8677, "test_loss": 0.3739507843017578}, {"step": 1450, "epoch": 3, "train_loss": 0.541132926940918}, {"step": 1500, "epoch": 3, "train_loss": 0.3661501407623291}, {"step": 1550, "epoch": 3, "train_loss": 0.24976524710655212}, {"step": 1600, "epoch": 3, "train_loss": 0.3311040699481964}, {"step": 1650, "epoch": 3, "train_loss": 0.37994784116744995}, {"step": 1700, "epoch": 3, "train_loss": 0.19890941679477692}, {"step": 1750, "epoch": 3, "train_loss": 0.4461594521999359}, {"step": 1800, "epoch": 3, "train_loss": 0.3191482126712799}, {"step": 1850, "epoch": 3, "train_loss": 0.3277038633823395}, {"epoch_end": 3, "step": 1876, "test_acc": 0.8698, "test_loss": 0.36183963928222657}, {"step": 1900, "epoch": 4, "train_loss": 0.23180453479290009}, {"step": 1950, "epoch": 4, "train_loss": 0.46234551072120667}, {"step": 2000, "epoch": 4, "train_loss": 0.30493828654289246}, {"step": 2050, "epoch": 4, "train_loss": 0.34819865226745605}, {"step": 2100, "epoch": 4, "train_loss": 0.26323258876800537}, {"step": 2150, "epoch": 4, "train_loss": 0.31695061922073364}, {"step": 2200, "epoch": 4, "train_loss": 0.34546929597854614}, {"step": 2250, "epoch": 4, "train_loss": 0.2306463122367859}, {"step": 2300, "epoch": 4, "train_loss": 0.1831660270690918}, {"epoch_end": 4, "step": 2345, "test_acc": 0.8676, "test_loss": 0.3606170928955078}, {"step": 2350, "epoch": 5, "train_loss": 0.21680787205696106}, {"step": 2400, "epoch": 5, "train_loss": 0.3507431149482727}, {"step": 2450, "epoch": 5, "train_loss": 0.35674044489860535}, {"step": 2500, "epoch": 5, "train_loss": 0.28412532806396484}, {"step": 2550, "epoch": 5, "train_loss": 0.23841333389282227}, {"step": 2600, "epoch": 5, "train_loss": 0.30879488587379456}, {"step": 2650, "epoch": 5, "train_loss": 0.174189031124115}, {"step": 2700, "epoch": 5, "train_loss": 0.19175726175308228}, {"step": 2750, "epoch": 5, "train_loss": 0.37364333868026733}, {"step": 2800, "epoch": 5, "train_loss": 0.2592906951904297}, {"epoch_end": 5, "step": 2814, "test_acc": 0.8737, "test_loss": 0.36097987670898435}, {"step": 2850, "epoch": 6, "train_loss": 0.15697525441646576}, {"step": 2900, "epoch": 6, "train_loss": 0.23301658034324646}, {"step": 2950, "epoch": 6, "train_loss": 0.12197292596101761}, {"step": 3000, "epoch": 6, "train_loss": 0.32307204604148865}, {"step": 3050, "epoch": 6, "train_loss": 0.373363733291626}, {"step": 3100, "epoch": 6, "train_loss": 0.24906697869300842}, {"step": 3150, "epoch": 6, "train_loss": 0.35010775923728943}, {"step": 3200, "epoch": 6, "train_loss": 0.2098419964313507}, {"step": 3250, "epoch": 6, "train_loss": 0.2019556760787964}, {"epoch_end": 6, "step": 3283, "test_acc": 0.868, "test_loss": 0.3677826171875}, {"step": 3300, "epoch": 7, "train_loss": 0.35043537616729736}, {"step": 3350, "epoch": 7, "train_loss": 0.2120496779680252}, {"step": 3400, "epoch": 7, "train_loss": 0.22139707207679749}, {"step": 3450, "epoch": 7, "train_loss": 0.25997334718704224}, {"step": 3500, "epoch": 7, "train_loss": 0.29301342368125916}, {"step": 3550, "epoch": 7, "train_loss": 0.2588314116001129}, {"step": 3600, "epoch": 7, "train_loss": 0.1799546480178833}, {"step": 3650, "epoch": 7, "train_loss": 0.26677218079566956}, {"step": 3700, "epoch": 7, "train_loss": 0.2200125753879547}, {"step": 3750, "epoch": 7, "train_loss": 0.3134715259075165}, {"epoch_end": 7, "step": 3752, "test_acc": 0.8782, "test_loss": 0.3489788330078125}, {"step": 3800, "epoch": 8, "train_loss": 0.21349477767944336}, {"step": 3850, "epoch": 8, "train_loss": 0.32520073652267456}, {"step": 3900, "epoch": 8, "train_loss": 0.21950021386146545}, {"step": 3950, "epoch": 8, "train_loss": 0.23741354048252106}, {"step": 4000, "epoch": 8, "train_loss": 0.2901303768157959}, {"step": 4050, "epoch": 8, "train_loss": 0.2241213023662567}, {"step": 4100, "epoch": 8, "train_loss": 0.26346465945243835}, {"step": 4150, "epoch": 8, "train_loss": 0.24012039601802826}, {"step": 4200, "epoch": 8, "train_loss": 0.23091445863246918}, {"epoch_end": 8, "step": 4221, "test_acc": 0.878, "test_loss": 0.3637749328613281}, {"step": 4250, "epoch": 9, "train_loss": 0.18444636464118958}, {"step": 4300, "epoch": 9, "train_loss": 0.1755490005016327}, {"step": 4350, "epoch": 9, "train_loss": 0.18485035002231598}, {"step": 4400, "epoch": 9, "train_loss": 0.26886507868766785}, {"step": 4450, "epoch": 9, "train_loss": 0.11517259478569031}, {"step": 4500, "epoch": 9, "train_loss": 0.2549457550048828}, {"step": 4550, "epoch": 9, "train_loss": 0.13501933217048645}, {"step": 4600, "epoch": 9, "train_loss": 0.2441372275352478}, {"step": 4650, "epoch": 9, "train_loss": 0.27016347646713257}, {"epoch_end": 9, "step": 4690, "test_acc": 0.8844, "test_loss": 0.34470483093261717}, {"step": 4700, "epoch": 10, "train_loss": 0.1851704716682434}, {"step": 4750, "epoch": 10, "train_loss": 0.2277248352766037}, {"step": 4800, "epoch": 10, "train_loss": 0.23483611643314362}, {"step": 4850, "epoch": 10, "train_loss": 0.20641453564167023}, {"step": 4900, "epoch": 10, "train_loss": 0.17134201526641846}, {"step": 4950, "epoch": 10, "train_loss": 0.18160204589366913}, {"step": 5000, "epoch": 10, "train_loss": 0.24632252752780914}, {"step": 5050, "epoch": 10, "train_loss": 0.35034987330436707}, {"step": 5100, "epoch": 10, "train_loss": 0.14342309534549713}, {"step": 5150, "epoch": 10, "train_loss": 0.2311842292547226}, {"epoch_end": 10, "step": 5159, "test_acc": 0.879, "test_loss": 0.36670565795898435}, {"step": 5200, "epoch": 11, "train_loss": 0.2527793049812317}, {"step": 5250, "epoch": 11, "train_loss": 0.11333931982517242}, {"step": 5300, "epoch": 11, "train_loss": 0.25337156653404236}, {"step": 5350, "epoch": 11, "train_loss": 0.2201198786497116}, {"step": 5400, "epoch": 11, "train_loss": 0.18673408031463623}, {"step": 5450, "epoch": 11, "train_loss": 0.2619165778160095}, {"step": 5500, "epoch": 11, "train_loss": 0.25838252902030945}, {"step": 5550, "epoch": 11, "train_loss": 0.29090118408203125}, {"step": 5600, "epoch": 11, "train_loss": 0.18005569279193878}, {"epoch_end": 11, "step": 5628, "test_acc": 0.8904, "test_loss": 0.3368552825927734}, {"step": 5650, "epoch": 12, "train_loss": 0.16821230947971344}, {"step": 5700, "epoch": 12, "train_loss": 0.18074707686901093}, {"step": 5750, "epoch": 12, "train_loss": 0.38870158791542053}, {"step": 5800, "epoch": 12, "train_loss": 0.18786704540252686}, {"step": 5850, "epoch": 12, "train_loss": 0.22646759450435638}, {"step": 5900, "epoch": 12, "train_loss": 0.2298225462436676}, {"step": 5950, "epoch": 12, "train_loss": 0.1624298393726349}, {"step": 6000, "epoch": 12, "train_loss": 0.14137963950634003}, {"step": 6050, "epoch": 12, "train_loss": 0.14458788931369781}, {"epoch_end": 12, "step": 6097, "test_acc": 0.8753, "test_loss": 0.37043748779296876}, {"step": 6100, "epoch": 13, "train_loss": 0.32490819692611694}, {"step": 6150, "epoch": 13, "train_loss": 0.12551774084568024}, {"step": 6200, "epoch": 13, "train_loss": 0.2624950408935547}, {"step": 6250, "epoch": 13, "train_loss": 0.19619466364383698}, {"step": 6300, "epoch": 13, "train_loss": 0.1979398876428604}, {"step": 6350, "epoch": 13, "train_loss": 0.20215047895908356}, {"step": 6400, "epoch": 13, "train_loss": 0.17269350588321686}, {"step": 6450, "epoch": 13, "train_loss": 0.16995848715305328}, {"step": 6500, "epoch": 13, "train_loss": 0.11076312512159348}, {"step": 6550, "epoch": 13, "train_loss": 0.17428122460842133}, {"epoch_end": 13, "step": 6566, "test_acc": 0.8856, "test_loss": 0.36011175537109374}, {"step": 6600, "epoch": 14, "train_loss": 0.19252288341522217}, {"step": 6650, "epoch": 14, "train_loss": 0.27086183428764343}, {"step": 6700, "epoch": 14, "train_loss": 0.1501627117395401}, {"step": 6750, "epoch": 14, "train_loss": 0.21651706099510193}, {"step": 6800, "epoch": 14, "train_loss": 0.12327341735363007}, {"step": 6850, "epoch": 14, "train_loss": 0.12542380392551422}, {"step": 6900, "epoch": 14, "train_loss": 0.2762380838394165}, {"step": 6950, "epoch": 14, "train_loss": 0.1922590285539627}, {"step": 7000, "epoch": 14, "train_loss": 0.27008992433547974}, {"epoch_end": 14, "step": 7035, "test_acc": 0.8839, "test_loss": 0.36318700561523437}, {"step": 7050, "epoch": 15, "train_loss": 0.15510906279087067}, {"step": 7100, "epoch": 15, "train_loss": 0.18050165474414825}, {"step": 7150, "epoch": 15, "train_loss": 0.24483339488506317}, {"step": 7200, "epoch": 15, "train_loss": 0.15059278905391693}, {"step": 7250, "epoch": 15, "train_loss": 0.24322853982448578}, {"step": 7300, "epoch": 15, "train_loss": 0.23706911504268646}, {"step": 7350, "epoch": 15, "train_loss": 0.1656387597322464}, {"step": 7400, "epoch": 15, "train_loss": 0.2189740389585495}, {"step": 7450, "epoch": 15, "train_loss": 0.1612611711025238}, {"step": 7500, "epoch": 15, "train_loss": 0.1748657077550888}, {"epoch_end": 15, "step": 7504, "test_acc": 0.8807, "test_loss": 0.37813473205566406}, {"step": 7550, "epoch": 16, "train_loss": 0.1814863234758377}, {"step": 7600, "epoch": 16, "train_loss": 0.2946085035800934}, {"step": 7650, "epoch": 16, "train_loss": 0.12428296357393265}, {"step": 7700, "epoch": 16, "train_loss": 0.1197979524731636}, {"step": 7750, "epoch": 16, "train_loss": 0.19343528151512146}, {"step": 7800, "epoch": 16, "train_loss": 0.12489385157823563}, {"step": 7850, "epoch": 16, "train_loss": 0.16141721606254578}, {"step": 7900, "epoch": 16, "train_loss": 0.11425430327653885}, {"step": 7950, "epoch": 16, "train_loss": 0.19923529028892517}, {"epoch_end": 16, "step": 7973, "test_acc": 0.8825, "test_loss": 0.36798621520996094}, {"step": 8000, "epoch": 17, "train_loss": 0.14948058128356934}, {"step": 8050, "epoch": 17, "train_loss": 0.12444821745157242}, {"step": 8100, "epoch": 17, "train_loss": 0.15714512765407562}, {"step": 8150, "epoch": 17, "train_loss": 0.15693198144435883}, {"step": 8200, "epoch": 17, "train_loss": 0.17939493060112}, {"step": 8250, "epoch": 17, "train_loss": 0.20760692656040192}, {"step": 8300, "epoch": 17, "train_loss": 0.15987138450145721}, {"step": 8350, "epoch": 17, "train_loss": 0.16027535498142242}, {"step": 8400, "epoch": 17, "train_loss": 0.3160420358181}, {"epoch_end": 17, "step": 8442, "test_acc": 0.8858, "test_loss": 0.3898866241455078}, {"step": 8450, "epoch": 18, "train_loss": 0.12129684537649155}, {"step": 8500, "epoch": 18, "train_loss": 0.15783561766147614}, {"step": 8550, "epoch": 18, "train_loss": 0.22940440475940704}, {"step": 8600, "epoch": 18, "train_loss": 0.11240377277135849}, {"step": 8650, "epoch": 18, "train_loss": 0.09250488877296448}, {"step": 8700, "epoch": 18, "train_loss": 0.18417538702487946}, {"step": 8750, "epoch": 18, "train_loss": 0.2293424904346466}, {"step": 8800, "epoch": 18, "train_loss": 0.08679106831550598}, {"step": 8850, "epoch": 18, "train_loss": 0.12526969611644745}, {"step": 8900, "epoch": 18, "train_loss": 0.16769009828567505}, {"epoch_end": 18, "step": 8911, "test_acc": 0.8869, "test_loss": 0.36922735900878906}, {"step": 8950, "epoch": 19, "train_loss": 0.1855911761522293}, {"step": 9000, "epoch": 19, "train_loss": 0.18688057363033295}, {"step": 9050, "epoch": 19, "train_loss": 0.16407130658626556}, {"step": 9100, "epoch": 19, "train_loss": 0.20362696051597595}, {"step": 9150, "epoch": 19, "train_loss": 0.2035364955663681}, {"step": 9200, "epoch": 19, "train_loss": 0.14054054021835327}, {"step": 9250, "epoch": 19, "train_loss": 0.172859326004982}, {"step": 9300, "epoch": 19, "train_loss": 0.1412552297115326}, {"step": 9350, "epoch": 19, "train_loss": 0.2628708779811859}, {"epoch_end": 19, "step": 9380, "test_acc": 0.887, "test_loss": 0.36895661010742187}], "final": {"test_acc": 0.887, "test_loss": 0.36895661010742187, "wall_s": 35.73411965370178}}
|
