diff options
Diffstat (limited to 'results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json')
| -rw-r--r-- | results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json | 2734 |
1 files changed, 2734 insertions, 0 deletions
diff --git a/results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json b/results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json new file mode 100644 index 0000000..d29ad2d --- /dev/null +++ b/results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json @@ -0,0 +1,2734 @@ +{ + "fa_log": [ + { + "hidden_norms": [ + 8.718070030212402, + 8.780964851379395, + 8.812002182006836, + 8.828720092773438, + 8.893179893493652 + ], + "bp_grad_norms_per_sample_med": [ + 0.0010068593546748161, + 0.000997236929833889, + 0.0009934091940522194, + 0.0009870644425973296, + 0.0009829498594626784 + ], + "bp_grad_norms_F": [ + 0.03544403985142708, + 0.03507080301642418, + 0.03477022796869278, + 0.03443896397948265, + 0.03407429903745651 + ], + "gamma_dfa": NaN, + "acc_eval": 0.115234375, + "loss_eval": 2.383650302886963, + "epoch": 0 + }, + { + "hidden_norms": [ + 634.95556640625, + 4686.67578125, + 8548.919921875, + 9006.0517578125, + 8484.302734375 + ], + "bp_grad_norms_per_sample_med": [ + 2.975244342451333e-06, + 1.1454501418484142e-06, + 1.1543062328200904e-06, + 1.164633545158722e-06, + 1.1064564660046017e-06 + ], + "bp_grad_norms_F": [ + 0.0001923188247019425, + 9.00624945643358e-05, + 6.289214798016474e-05, + 5.813155075884424e-05, + 5.113427323522046e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.27734375, + "loss_eval": 1.9464800357818604, + "epoch": 1 + }, + { + "hidden_norms": [ + 838.77099609375, + 8185.21484375, + 20076.259765625, + 21714.076171875, + 19607.720703125 + ], + "bp_grad_norms_per_sample_med": [ + 2.675848463695729e-06, + 5.853078732798167e-07, + 5.769501285612932e-07, + 5.794665298708424e-07, + 5.573179464590794e-07 + ], + "bp_grad_norms_F": [ + 0.00015455791435670108, + 2.4559509256505407e-05, + 2.031458825513255e-05, + 2.0197476260364056e-05, + 1.898946902656462e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2958984375, + "loss_eval": 1.932114601135254, + "epoch": 2 + }, + { + "hidden_norms": [ + 966.05859375, + 10760.646484375, + 31344.212890625, + 35515.1015625, + 30314.041015625 + ], + "bp_grad_norms_per_sample_med": [ + 2.748855649770121e-06, + 4.3298211949149845e-07, + 4.17521391682385e-07, + 4.17111380102142e-07, + 3.9905967241793405e-07 + ], + "bp_grad_norms_F": [ + 0.00015430455096065998, + 1.7030582966981456e-05, + 1.4290850231191143e-05, + 1.4243430996430106e-05, + 1.3498008229362313e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3203125, + "loss_eval": 1.9043192863464355, + "epoch": 3 + }, + { + "hidden_norms": [ + 1071.293212890625, + 12608.09765625, + 40531.1171875, + 47819.078125, + 38897.390625 + ], + "bp_grad_norms_per_sample_med": [ + 2.733014980549342e-06, + 3.670443504688592e-07, + 3.5215543903177604e-07, + 3.5238662121628295e-07, + 3.3813631716839154e-07 + ], + "bp_grad_norms_F": [ + 0.0001624817814445123, + 1.4486134205071721e-05, + 1.1960014489886817e-05, + 1.1927315426873975e-05, + 1.1350254681019578e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3115234375, + "loss_eval": 1.898539662361145, + "epoch": 4 + }, + { + "hidden_norms": [ + 1170.5045166015625, + 14295.359375, + 49885.03515625, + 61276.62890625, + 47640.921875 + ], + "bp_grad_norms_per_sample_med": [ + 2.690757355594542e-06, + 3.267070951551432e-07, + 3.0663491656923725e-07, + 3.0629081493316335e-07, + 2.9683877755815047e-07 + ], + "bp_grad_norms_F": [ + 0.0001624024735065177, + 1.2482374586397782e-05, + 1.0314163773728069e-05, + 1.0298784218321089e-05, + 9.855078133114148e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.322265625, + "loss_eval": 1.877720832824707, + "epoch": 5 + }, + { + "hidden_norms": [ + 1288.9276123046875, + 15613.23046875, + 57501.26171875, + 73376.796875, + 54583.2109375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7759115255321376e-06, + 3.0544967444257054e-07, + 2.801086793624563e-07, + 2.7996006224384473e-07, + 2.7101373234472703e-07 + ], + "bp_grad_norms_F": [ + 0.00017124316946137697, + 1.1784563866967801e-05, + 9.484625479672104e-06, + 9.465281436860096e-06, + 9.092000254895538e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.32421875, + "loss_eval": 1.8702466487884521, + "epoch": 6 + }, + { + "hidden_norms": [ + 1437.197021484375, + 16777.841796875, + 65706.7109375, + 87232.4140625, + 61590.078125 + ], + "bp_grad_norms_per_sample_med": [ + 2.883781235141214e-06, + 2.9475407359313976e-07, + 2.6204861569567584e-07, + 2.6302603828298743e-07, + 2.5517954327369807e-07 + ], + "bp_grad_norms_F": [ + 0.00017113517969846725, + 1.1177045962540433e-05, + 8.825853001326323e-06, + 8.807965969026554e-06, + 8.485528269375209e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.33984375, + "loss_eval": 1.8411049842834473, + "epoch": 7 + }, + { + "hidden_norms": [ + 1575.2144775390625, + 17884.404296875, + 73215.0, + 101692.0703125, + 67757.890625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8855845357611543e-06, + 2.785824051443342e-07, + 2.509512739834463e-07, + 2.5104517931140435e-07, + 2.451099874178908e-07 + ], + "bp_grad_norms_F": [ + 0.0001778283913154155, + 1.0928491974482313e-05, + 8.524083568772767e-06, + 8.510252882842906e-06, + 8.22034417069517e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.8426380157470703, + "epoch": 8 + }, + { + "hidden_norms": [ + 1746.853515625, + 18578.95703125, + 80693.078125, + 117047.6171875, + 73706.9765625 + ], + "bp_grad_norms_per_sample_med": [ + 2.964599843835458e-06, + 2.738325974860345e-07, + 2.414699338260107e-07, + 2.418815370219818e-07, + 2.3542932581221976e-07 + ], + "bp_grad_norms_F": [ + 0.0001854781003203243, + 1.0656134691089392e-05, + 8.172302841558121e-06, + 8.16663850855548e-06, + 7.904830454208422e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3671875, + "loss_eval": 1.8274496793746948, + "epoch": 9 + }, + { + "hidden_norms": [ + 1938.662109375, + 19766.384765625, + 88785.7734375, + 134051.8125, + 79702.1796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.93986431643134e-06, + 2.6766733185468183e-07, + 2.3016899319827644e-07, + 2.3030897011722118e-07, + 2.2565886581560335e-07 + ], + "bp_grad_norms_F": [ + 0.0001912089210236445, + 1.0689387636375614e-05, + 7.921232281660195e-06, + 7.924772944534197e-06, + 7.683997864660341e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3505859375, + "loss_eval": 1.8229554891586304, + "epoch": 10 + }, + { + "hidden_norms": [ + 2088.147705078125, + 20979.9453125, + 96985.96875, + 150667.6875, + 85126.3359375 + ], + "bp_grad_norms_per_sample_med": [ + 2.9922123303549597e-06, + 2.649965722412162e-07, + 2.2647282094112597e-07, + 2.272310553053103e-07, + 2.2135814958801348e-07 + ], + "bp_grad_norms_F": [ + 0.00019602400425355881, + 1.0553159881965257e-05, + 7.715923857176676e-06, + 7.728580385446548e-06, + 7.498896593460813e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.375, + "loss_eval": 1.8172534704208374, + "epoch": 11 + }, + { + "hidden_norms": [ + 2276.7451171875, + 22148.2578125, + 105343.15625, + 167593.515625, + 90718.5 + ], + "bp_grad_norms_per_sample_med": [ + 3.033006578334607e-06, + 2.6026333443951444e-07, + 2.1839655062194652e-07, + 2.189913459460513e-07, + 2.1411695172446343e-07 + ], + "bp_grad_norms_F": [ + 0.00019336632976774126, + 1.0404637578176335e-05, + 7.483776698791189e-06, + 7.508507223974448e-06, + 7.2862567321863025e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.8100277185440063, + "epoch": 12 + }, + { + "hidden_norms": [ + 2470.05517578125, + 22943.138671875, + 111998.0078125, + 181164.46875, + 93911.8515625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0791732115176274e-06, + 2.678765156360896e-07, + 2.2002203081683547e-07, + 2.2065630389533908e-07, + 2.1573649178208143e-07 + ], + "bp_grad_norms_F": [ + 0.00020370102720335126, + 1.0737571756180841e-05, + 7.488396931876196e-06, + 7.527175512223039e-06, + 7.304198788915528e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3447265625, + "loss_eval": 1.8219082355499268, + "epoch": 13 + }, + { + "hidden_norms": [ + 2655.71875, + 23620.771484375, + 118446.4453125, + 194098.046875, + 97110.40625 + ], + "bp_grad_norms_per_sample_med": [ + 3.03382375932415e-06, + 2.674543395642104e-07, + 2.1672505567948974e-07, + 2.1837567487636989e-07, + 2.127215310565589e-07 + ], + "bp_grad_norms_F": [ + 0.00019606151909101754, + 1.0727502740337513e-05, + 7.367311354755657e-06, + 7.4189733823004644e-06, + 7.2012962846201845e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3447265625, + "loss_eval": 1.800294041633606, + "epoch": 14 + }, + { + "hidden_norms": [ + 2864.51904296875, + 23962.1328125, + 123932.375, + 204264.46875, + 99147.40625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0961696211306844e-06, + 2.757700485744863e-07, + 2.1832956065281905e-07, + 2.1977299979880627e-07, + 2.1560039442647394e-07 + ], + "bp_grad_norms_F": [ + 0.00020294770365580916, + 1.0936742910416797e-05, + 7.422260750900023e-06, + 7.487843049602816e-06, + 7.265869498951361e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.8283555507659912, + "epoch": 15 + }, + { + "hidden_norms": [ + 3058.654296875, + 24515.28515625, + 129439.953125, + 215162.921875, + 101313.0703125 + ], + "bp_grad_norms_per_sample_med": [ + 3.0670128126075724e-06, + 2.704452413127001e-07, + 2.1300424180026312e-07, + 2.1552395423896087e-07, + 2.103476077763844e-07 + ], + "bp_grad_norms_F": [ + 0.00020699974265880883, + 1.1274227290414274e-05, + 7.348373856075341e-06, + 7.4233907980669755e-06, + 7.201060725492425e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3603515625, + "loss_eval": 1.8033101558685303, + "epoch": 16 + }, + { + "hidden_norms": [ + 3205.80859375, + 24759.69921875, + 132512.875, + 221758.34375, + 102296.3515625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0818489449302433e-06, + 2.7902248689315456e-07, + 2.1470867750394973e-07, + 2.1584412479569437e-07, + 2.1120550286468642e-07 + ], + "bp_grad_norms_F": [ + 0.00019346778572071344, + 1.1271905350440647e-05, + 7.35169760446297e-06, + 7.441570687660715e-06, + 7.2173174885392655e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3623046875, + "loss_eval": 1.789717197418213, + "epoch": 17 + }, + { + "hidden_norms": [ + 3365.503173828125, + 24903.955078125, + 137239.453125, + 228336.28125, + 103774.4140625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0317510208988097e-06, + 2.810125181440526e-07, + 2.1507479175397748e-07, + 2.1818523521233146e-07, + 2.1249555004487775e-07 + ], + "bp_grad_norms_F": [ + 0.00019401832832954824, + 1.144540510722436e-05, + 7.354777153523173e-06, + 7.453731996065471e-06, + 7.234774784592446e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3515625, + "loss_eval": 1.7909660339355469, + "epoch": 18 + }, + { + "hidden_norms": [ + 3517.615478515625, + 25088.482421875, + 140183.171875, + 234581.5, + 105112.640625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0766200325160753e-06, + 2.8522768502625695e-07, + 2.1497518787327863e-07, + 2.1691003837531753e-07, + 2.1271530670219363e-07 + ], + "bp_grad_norms_F": [ + 0.00020355181186459959, + 1.1879135854542255e-05, + 7.4169261097267736e-06, + 7.52632467992953e-06, + 7.302520316443406e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.34765625, + "loss_eval": 1.8049894571304321, + "epoch": 19 + }, + { + "hidden_norms": [ + 3643.212890625, + 25572.47265625, + 143095.625, + 240103.59375, + 106815.9609375 + ], + "bp_grad_norms_per_sample_med": [ + 2.9240111416584114e-06, + 2.7889984721696237e-07, + 2.1295824126355e-07, + 2.145163620070889e-07, + 2.1029546815043432e-07 + ], + "bp_grad_norms_F": [ + 0.00019266517483629286, + 1.1654319678200409e-05, + 7.349848146986915e-06, + 7.464911050192313e-06, + 7.246624591061845e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3681640625, + "loss_eval": 1.7967158555984497, + "epoch": 20 + }, + { + "hidden_norms": [ + 3727.509521484375, + 26266.642578125, + 147179.796875, + 243896.734375, + 107961.3359375 + ], + "bp_grad_norms_per_sample_med": [ + 3.034621840924956e-06, + 2.95065575528497e-07, + 2.1733650612532074e-07, + 2.1865858457204013e-07, + 2.140783124104928e-07 + ], + "bp_grad_norms_F": [ + 0.00019002222688868642, + 1.179500304715475e-05, + 7.40604900784092e-06, + 7.5316506809031125e-06, + 7.309085049200803e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3603515625, + "loss_eval": 1.7879303693771362, + "epoch": 21 + }, + { + "hidden_norms": [ + 3814.343994140625, + 26307.63671875, + 149997.296875, + 248837.015625, + 109946.75 + ], + "bp_grad_norms_per_sample_med": [ + 3.0296112072392134e-06, + 2.891725046083593e-07, + 2.132598950765896e-07, + 2.164179448982395e-07, + 2.1202281175192184e-07 + ], + "bp_grad_norms_F": [ + 0.00019085934036411345, + 1.1828436981886625e-05, + 7.313527476071613e-06, + 7.4397207754373085e-06, + 7.220100542326691e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3798828125, + "loss_eval": 1.769087791442871, + "epoch": 22 + }, + { + "hidden_norms": [ + 3903.16357421875, + 26762.390625, + 151661.765625, + 252546.703125, + 112163.4375 + ], + "bp_grad_norms_per_sample_med": [ + 3.10251039081777e-06, + 2.9397492085081467e-07, + 2.119260216204566e-07, + 2.151843716546864e-07, + 2.0996778005155647e-07 + ], + "bp_grad_norms_F": [ + 0.0001914280146593228, + 1.1861766324727796e-05, + 7.28844088371261e-06, + 7.418064342346042e-06, + 7.190795258793514e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.7941722869873047, + "epoch": 23 + }, + { + "hidden_norms": [ + 3989.95361328125, + 27301.33984375, + 155319.25, + 259232.390625, + 115495.6796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.962329745059833e-06, + 2.8580453204085643e-07, + 2.0789140364740888e-07, + 2.0986689719393326e-07, + 2.051628200661071e-07 + ], + "bp_grad_norms_F": [ + 0.00018467944755684584, + 1.1658386029012036e-05, + 7.1250747168960515e-06, + 7.254411684698425e-06, + 7.026835191936698e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3642578125, + "loss_eval": 1.772944450378418, + "epoch": 24 + }, + { + "hidden_norms": [ + 4044.4833984375, + 27958.3046875, + 158277.765625, + 265767.4375, + 119247.171875 + ], + "bp_grad_norms_per_sample_med": [ + 2.939599653473124e-06, + 2.9012642244197195e-07, + 2.0495993169333815e-07, + 2.0806970724152052e-07, + 2.0302766756685742e-07 + ], + "bp_grad_norms_F": [ + 0.00017772591672837734, + 1.151475134975044e-05, + 7.005810857663164e-06, + 7.14335192242288e-06, + 6.918045983184129e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.359375, + "loss_eval": 1.7681654691696167, + "epoch": 25 + }, + { + "hidden_norms": [ + 4102.31298828125, + 28819.13671875, + 162886.59375, + 274888.375, + 124251.71875 + ], + "bp_grad_norms_per_sample_med": [ + 3.00775354844518e-06, + 2.8842808319495816e-07, + 1.9920004490359133e-07, + 2.0228824837431603e-07, + 1.9777876048010512e-07 + ], + "bp_grad_norms_F": [ + 0.00017743176431395113, + 1.1497872947074939e-05, + 6.855066203570459e-06, + 6.992981980147306e-06, + 6.764100362488534e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.361328125, + "loss_eval": 1.7643167972564697, + "epoch": 26 + }, + { + "hidden_norms": [ + 4156.68994140625, + 29263.09765625, + 167567.5625, + 286674.25, + 130533.7578125 + ], + "bp_grad_norms_per_sample_med": [ + 2.9521122542064404e-06, + 2.7994269657938275e-07, + 1.9472663836950233e-07, + 1.9846331156259112e-07, + 1.9457104372122558e-07 + ], + "bp_grad_norms_F": [ + 0.0001732901146169752, + 1.1156066648254637e-05, + 6.697610388073372e-06, + 6.834617579443147e-06, + 6.61404874335858e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.357421875, + "loss_eval": 1.7718162536621094, + "epoch": 27 + }, + { + "hidden_norms": [ + 4236.7529296875, + 30445.798828125, + 173905.171875, + 301198.6875, + 138844.546875 + ], + "bp_grad_norms_per_sample_med": [ + 2.825209094226011e-06, + 2.710314390697022e-07, + 1.868730663545648e-07, + 1.8901035048202175e-07, + 1.8439881444010098e-07 + ], + "bp_grad_norms_F": [ + 0.00016928529657889158, + 1.0838041816896293e-05, + 6.40136795482249e-06, + 6.533743089676136e-06, + 6.317317456705496e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3740234375, + "loss_eval": 1.7575714588165283, + "epoch": 28 + }, + { + "hidden_norms": [ + 4307.15966796875, + 30990.375, + 179990.59375, + 314925.53125, + 146306.109375 + ], + "bp_grad_norms_per_sample_med": [ + 2.882675744331209e-06, + 2.69783697603998e-07, + 1.8120582012670639e-07, + 1.8295163783932367e-07, + 1.7808179109124467e-07 + ], + "bp_grad_norms_F": [ + 0.0001705507020233199, + 1.0706517059588805e-05, + 6.220788691280177e-06, + 6.352388027153211e-06, + 6.134147952252533e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.765124797821045, + "epoch": 29 + }, + { + "hidden_norms": [ + 4360.767578125, + 32488.16015625, + 189370.28125, + 333218.25, + 156773.703125 + ], + "bp_grad_norms_per_sample_med": [ + 2.8181013931316556e-06, + 2.653597448443179e-07, + 1.7396521911905438e-07, + 1.755686014348612e-07, + 1.7160699883334019e-07 + ], + "bp_grad_norms_F": [ + 0.00016797029820736498, + 1.0374783414590638e-05, + 5.9415992836875375e-06, + 6.068700713512953e-06, + 5.864551440026844e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3564453125, + "loss_eval": 1.755420207977295, + "epoch": 30 + }, + { + "hidden_norms": [ + 4420.99462890625, + 33248.23046875, + 197272.421875, + 350132.625, + 166295.0625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8225279038451845e-06, + 2.5861882591016183e-07, + 1.676798291327941e-07, + 1.704754311049328e-07, + 1.6635479482829396e-07 + ], + "bp_grad_norms_F": [ + 0.00016677004168741405, + 1.0174546332564205e-05, + 5.748868716182187e-06, + 5.872501787962392e-06, + 5.676161890733056e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.365234375, + "loss_eval": 1.749579906463623, + "epoch": 31 + }, + { + "hidden_norms": [ + 4467.9111328125, + 34465.12109375, + 206817.0625, + 370568.6875, + 177992.796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.797579327307176e-06, + 2.4461206749037956e-07, + 1.6084129583759932e-07, + 1.623946417339539e-07, + 1.5842591949422058e-07 + ], + "bp_grad_norms_F": [ + 0.0001667520118644461, + 9.997165761888027e-06, + 5.521564617083641e-06, + 5.637548838421935e-06, + 5.447109742817702e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3818359375, + "loss_eval": 1.7478370666503906, + "epoch": 32 + }, + { + "hidden_norms": [ + 4500.455078125, + 35232.3984375, + 215903.1875, + 387229.5, + 187304.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.744678795352229e-06, + 2.4473743565067707e-07, + 1.5567927391657577e-07, + 1.5809170861302846e-07, + 1.5448634371750813e-07 + ], + "bp_grad_norms_F": [ + 0.00016586907440796494, + 9.867669177765492e-06, + 5.356784186005825e-06, + 5.468943072628463e-06, + 5.285164206725312e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.380859375, + "loss_eval": 1.748937964439392, + "epoch": 33 + }, + { + "hidden_norms": [ + 4559.64990234375, + 36405.82421875, + 224989.71875, + 406526.59375, + 197279.796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.676637450349517e-06, + 2.350530081685065e-07, + 1.5119638874239172e-07, + 1.5273755593625538e-07, + 1.4942507675641536e-07 + ], + "bp_grad_norms_F": [ + 0.00016609806334599853, + 9.700875125417951e-06, + 5.207733465795172e-06, + 5.317946943250718e-06, + 5.140104349266039e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.747586965560913, + "epoch": 34 + }, + { + "hidden_norms": [ + 4600.470703125, + 37569.390625, + 234249.046875, + 424273.0625, + 207188.484375 + ], + "bp_grad_norms_per_sample_med": [ + 2.6595262170303613e-06, + 2.3298444773445226e-07, + 1.4830226291451254e-07, + 1.5017273824469157e-07, + 1.4676733428586886e-07 + ], + "bp_grad_norms_F": [ + 0.0001615065266378224, + 9.520201274426654e-06, + 5.091071670904057e-06, + 5.198567123443354e-06, + 5.027129645895911e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.375, + "loss_eval": 1.7455894947052002, + "epoch": 35 + }, + { + "hidden_norms": [ + 4655.232421875, + 38777.5, + 243977.203125, + 443307.78125, + 217682.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7372755084797973e-06, + 2.3094322898487007e-07, + 1.4448018248458538e-07, + 1.465641616960056e-07, + 1.4282748850291682e-07 + ], + "bp_grad_norms_F": [ + 0.00016169788432307541, + 9.384648365085013e-06, + 4.931174771627411e-06, + 5.035632966610137e-06, + 4.871027158515062e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3779296875, + "loss_eval": 1.739393949508667, + "epoch": 36 + }, + { + "hidden_norms": [ + 4674.50634765625, + 39366.46875, + 253330.671875, + 458747.40625, + 225757.234375 + ], + "bp_grad_norms_per_sample_med": [ + 2.720141083045746e-06, + 2.3003921967301721e-07, + 1.4151289917663235e-07, + 1.4314728957742773e-07, + 1.4079574839342968e-07 + ], + "bp_grad_norms_F": [ + 0.00016218192467931658, + 9.336513357993681e-06, + 4.852768142882269e-06, + 4.95553604196175e-06, + 4.79518030260806e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3818359375, + "loss_eval": 1.7406929731369019, + "epoch": 37 + }, + { + "hidden_norms": [ + 4730.9697265625, + 40562.7578125, + 261285.078125, + 476049.71875, + 235002.046875 + ], + "bp_grad_norms_per_sample_med": [ + 2.727818809944438e-06, + 2.3385531733310927e-07, + 1.4104001877512928e-07, + 1.4261416936278692e-07, + 1.39172499302731e-07 + ], + "bp_grad_norms_F": [ + 0.00016533929738216102, + 9.380160008731764e-06, + 4.782661108038155e-06, + 4.884728696197271e-06, + 4.725173766928492e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3759765625, + "loss_eval": 1.7494721412658691, + "epoch": 38 + }, + { + "hidden_norms": [ + 4769.4833984375, + 41614.734375, + 272033.25, + 494194.5, + 244904.296875 + ], + "bp_grad_norms_per_sample_med": [ + 2.728792651396361e-06, + 2.2869069482567284e-07, + 1.3784755026335915e-07, + 1.3816679711453617e-07, + 1.3538770815557655e-07 + ], + "bp_grad_norms_F": [ + 0.0001621924457140267, + 9.226439942722209e-06, + 4.664412244892446e-06, + 4.759738658322021e-06, + 4.6096511141513474e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.380859375, + "loss_eval": 1.744396686553955, + "epoch": 39 + }, + { + "hidden_norms": [ + 4807.61669921875, + 41796.7890625, + 279173.46875, + 509024.46875, + 251937.796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7552744086278835e-06, + 2.2436680069404247e-07, + 1.334142893938406e-07, + 1.3510819485418324e-07, + 1.3202044613080943e-07 + ], + "bp_grad_norms_F": [ + 0.00016327289631590247, + 9.199145097227301e-06, + 4.6049044613027945e-06, + 4.701471880252939e-06, + 4.55197186965961e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.37890625, + "loss_eval": 1.738051176071167, + "epoch": 40 + }, + { + "hidden_norms": [ + 4846.89013671875, + 43419.1953125, + 288358.6875, + 526415.1875, + 261093.328125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7159530873177573e-06, + 2.2739833127616293e-07, + 1.3493598771674442e-07, + 1.3701293255508062e-07, + 1.3374133800425625e-07 + ], + "bp_grad_norms_F": [ + 0.00016205597785301507, + 9.20344336918788e-06, + 4.573969818011392e-06, + 4.670639100368135e-06, + 4.519301455729874e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3759765625, + "loss_eval": 1.769773244857788, + "epoch": 41 + }, + { + "hidden_norms": [ + 4880.62109375, + 44368.0, + 297320.0, + 543147.8125, + 270529.90625 + ], + "bp_grad_norms_per_sample_med": [ + 2.704309054024634e-06, + 2.2085772855007235e-07, + 1.31169457517899e-07, + 1.3294274481268076e-07, + 1.296902354397389e-07 + ], + "bp_grad_norms_F": [ + 0.000160810406669043, + 9.048065294336993e-06, + 4.454658665054012e-06, + 4.546407581074163e-06, + 4.4059761421522126e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7465946674346924, + "epoch": 42 + }, + { + "hidden_norms": [ + 4899.1767578125, + 45389.57421875, + 307315.8125, + 560231.5625, + 278935.6875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7406351819081465e-06, + 2.2317071568522806e-07, + 1.2826720308112272e-07, + 1.3064222059711028e-07, + 1.270270928444006e-07 + ], + "bp_grad_norms_F": [ + 0.00016026092635001987, + 8.999614692584146e-06, + 4.385735792311607e-06, + 4.476342382986331e-06, + 4.342015927250031e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.73671293258667, + "epoch": 43 + }, + { + "hidden_norms": [ + 4931.53857421875, + 45391.578125, + 314715.1875, + 576663.1875, + 287751.21875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7501505428517703e-06, + 2.1621856660658523e-07, + 1.2502815138759615e-07, + 1.2642922797567735e-07, + 1.2348743894108338e-07 + ], + "bp_grad_norms_F": [ + 0.00015757110668346286, + 8.882019756129012e-06, + 4.295293820177903e-06, + 4.383605300972704e-06, + 4.2555047912173904e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.376953125, + "loss_eval": 1.7194021940231323, + "epoch": 44 + }, + { + "hidden_norms": [ + 4952.10205078125, + 46544.41015625, + 321747.78125, + 590124.0, + 293778.09375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7445450996310683e-06, + 2.20402526451835e-07, + 1.2519362257989997e-07, + 1.2609659449935862e-07, + 1.243117537796934e-07 + ], + "bp_grad_norms_F": [ + 0.00015927865752018988, + 8.919399988371879e-06, + 4.280226676200982e-06, + 4.3682503019226715e-06, + 4.240421731083188e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.376953125, + "loss_eval": 1.7289605140686035, + "epoch": 45 + }, + { + "hidden_norms": [ + 4985.6728515625, + 47313.03515625, + 330026.15625, + 605693.8125, + 301341.84375 + ], + "bp_grad_norms_per_sample_med": [ + 2.751742158579873e-06, + 2.1928519799985224e-07, + 1.2418719563811464e-07, + 1.259736563952174e-07, + 1.2327782883403415e-07 + ], + "bp_grad_norms_F": [ + 0.0001606208534212783, + 8.962500942288898e-06, + 4.251486188877607e-06, + 4.337850896263262e-06, + 4.2124443098146e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.7371184825897217, + "epoch": 46 + }, + { + "hidden_norms": [ + 4993.990234375, + 47521.25390625, + 335434.3125, + 618341.0625, + 307157.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7009368750441354e-06, + 2.1855973386664118e-07, + 1.2373016033961903e-07, + 1.255795467614007e-07, + 1.2245830305346317e-07 + ], + "bp_grad_norms_F": [ + 0.00015943347534630448, + 8.89301736606285e-06, + 4.223482847010018e-06, + 4.306960363464896e-06, + 4.1826651795418e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.384765625, + "loss_eval": 1.737732172012329, + "epoch": 47 + }, + { + "hidden_norms": [ + 5012.49951171875, + 48445.1953125, + 344918.46875, + 633703.8125, + 314618.65625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7400287763157394e-06, + 2.1858940613128652e-07, + 1.2219133793678338e-07, + 1.242535176970705e-07, + 1.2144892025389709e-07 + ], + "bp_grad_norms_F": [ + 0.0001590315659996122, + 8.84339078766061e-06, + 4.180086762062274e-06, + 4.265210009180009e-06, + 4.143982550885994e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.375, + "loss_eval": 1.7298643589019775, + "epoch": 48 + }, + { + "hidden_norms": [ + 5054.08984375, + 49201.015625, + 353929.5625, + 650649.9375, + 323234.34375 + ], + "bp_grad_norms_per_sample_med": [ + 2.710038643272128e-06, + 2.1170153274852055e-07, + 1.1893747853264358e-07, + 1.2112403169339814e-07, + 1.1869307314782418e-07 + ], + "bp_grad_norms_F": [ + 0.00015847652684897184, + 8.75549994816538e-06, + 4.11285736845457e-06, + 4.190015260974178e-06, + 4.073814125149511e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7279736995697021, + "epoch": 49 + }, + { + "hidden_norms": [ + 5079.68505859375, + 49818.85546875, + 363584.96875, + 665326.4375, + 330221.40625 + ], + "bp_grad_norms_per_sample_med": [ + 2.728055960687925e-06, + 2.1385960735642584e-07, + 1.1923917497824732e-07, + 1.2005260430214548e-07, + 1.1790963583280245e-07 + ], + "bp_grad_norms_F": [ + 0.00015702334349043667, + 8.663697371957824e-06, + 4.082327905052807e-06, + 4.160873686487321e-06, + 4.0494828681403305e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7221198081970215, + "epoch": 50 + }, + { + "hidden_norms": [ + 5103.2626953125, + 50236.62109375, + 366800.8125, + 675614.9375, + 334505.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.6967995836457703e-06, + 2.110563741553051e-07, + 1.1874865890604269e-07, + 1.1944737821067974e-07, + 1.17826154166778e-07 + ], + "bp_grad_norms_F": [ + 0.00015892359078861773, + 8.714905561646447e-06, + 4.066732344654156e-06, + 4.143567366554635e-06, + 4.033038294437574e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3955078125, + "loss_eval": 1.7186274528503418, + "epoch": 51 + }, + { + "hidden_norms": [ + 5129.00830078125, + 51124.4453125, + 375390.84375, + 689126.5, + 340674.15625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7418275294621708e-06, + 2.1331975119665003e-07, + 1.1826992363239697e-07, + 1.1942577771151264e-07, + 1.1719654224862097e-07 + ], + "bp_grad_norms_F": [ + 0.00015799023094587028, + 8.683533451403491e-06, + 4.048787559440825e-06, + 4.127006377530051e-06, + 4.019248081021942e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7216696739196777, + "epoch": 52 + }, + { + "hidden_norms": [ + 5147.63671875, + 51383.28515625, + 382888.34375, + 701867.8125, + 346661.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.725489366639522e-06, + 2.1360351354360319e-07, + 1.1696960910967391e-07, + 1.1808428013182493e-07, + 1.1615074413384718e-07 + ], + "bp_grad_norms_F": [ + 0.0001584803540026769, + 8.698161764186807e-06, + 4.018610525235999e-06, + 4.092659764864948e-06, + 3.9867554733064026e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.39453125, + "loss_eval": 1.7209246158599854, + "epoch": 53 + }, + { + "hidden_norms": [ + 5179.68603515625, + 52093.16796875, + 388203.75, + 712999.25, + 350465.40625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7409823815105483e-06, + 2.16206004211017e-07, + 1.1817400746849671e-07, + 1.1920442943846865e-07, + 1.1731118831903586e-07 + ], + "bp_grad_norms_F": [ + 0.00015836946840863675, + 8.735608389542904e-06, + 4.036091922898777e-06, + 4.111263024242362e-06, + 4.005241407867288e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3837890625, + "loss_eval": 1.7180767059326172, + "epoch": 54 + }, + { + "hidden_norms": [ + 5197.8408203125, + 53041.44140625, + 398597.125, + 728487.4375, + 358157.9375 + ], + "bp_grad_norms_per_sample_med": [ + 2.763840029729181e-06, + 2.1291030805059563e-07, + 1.1598489635389342e-07, + 1.1754219286785883e-07, + 1.1593917292884726e-07 + ], + "bp_grad_norms_F": [ + 0.0001594589266460389, + 8.711985174159054e-06, + 4.000611625087913e-06, + 4.07353172704461e-06, + 3.971201294916682e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.7179795503616333, + "epoch": 55 + }, + { + "hidden_norms": [ + 5217.0146484375, + 53347.57421875, + 403986.625, + 738288.125, + 361666.125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7599005534284515e-06, + 2.135690380100641e-07, + 1.1644724651205252e-07, + 1.166295646726212e-07, + 1.1492483054098557e-07 + ], + "bp_grad_norms_F": [ + 0.0001592675835127011, + 8.73087174113607e-06, + 4.000759417976951e-06, + 4.0730465116212144e-06, + 3.972524154960411e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.719031810760498, + "epoch": 56 + }, + { + "hidden_norms": [ + 5245.46875, + 53932.046875, + 410963.8125, + 750011.5, + 367171.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.727937499003019e-06, + 2.1207910094744875e-07, + 1.1586875103830607e-07, + 1.1747369654813156e-07, + 1.1563617619003708e-07 + ], + "bp_grad_norms_F": [ + 0.0001595055073266849, + 8.686207365826704e-06, + 3.98215252062073e-06, + 4.0539648580306675e-06, + 3.954477506340481e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3837890625, + "loss_eval": 1.7162381410598755, + "epoch": 57 + }, + { + "hidden_norms": [ + 5273.81396484375, + 54359.40234375, + 415035.71875, + 757034.5, + 368386.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.8185475002828753e-06, + 2.183954137535693e-07, + 1.180805142553254e-07, + 1.196149383986267e-07, + 1.1741713734636505e-07 + ], + "bp_grad_norms_F": [ + 0.00016158133803401142, + 8.819352842692751e-06, + 4.027346676593879e-06, + 4.101018930668943e-06, + 3.999525233666645e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3828125, + "loss_eval": 1.7294056415557861, + "epoch": 58 + }, + { + "hidden_norms": [ + 5279.26513671875, + 54860.85546875, + 420308.09375, + 767223.875, + 372461.6875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7814132863568375e-06, + 2.1827422358455806e-07, + 1.1745117944883532e-07, + 1.1896155172053113e-07, + 1.1699857083158349e-07 + ], + "bp_grad_norms_F": [ + 0.0001601481344550848, + 8.751929271966219e-06, + 4.01089027946e-06, + 4.083126441400964e-06, + 3.984826435043942e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3740234375, + "loss_eval": 1.7233260869979858, + "epoch": 59 + }, + { + "hidden_norms": [ + 5281.876953125, + 55172.76171875, + 426239.0, + 777712.75, + 377365.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.729702600845485e-06, + 2.1274833272855176e-07, + 1.14989177291136e-07, + 1.1596039684036441e-07, + 1.1409560585207146e-07 + ], + "bp_grad_norms_F": [ + 0.00016002384654711932, + 8.679173333803192e-06, + 3.97576059185667e-06, + 4.045130026497645e-06, + 3.949881829612423e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.392578125, + "loss_eval": 1.7190814018249512, + "epoch": 60 + }, + { + "hidden_norms": [ + 5293.4951171875, + 55554.5703125, + 430637.53125, + 784743.25, + 379242.96875 + ], + "bp_grad_norms_per_sample_med": [ + 2.768365220617852e-06, + 2.1479348788489006e-07, + 1.1614937989179452e-07, + 1.1718410775074517e-07, + 1.1549084888429206e-07 + ], + "bp_grad_norms_F": [ + 0.00016024228534661233, + 8.705451364221517e-06, + 3.991565336036729e-06, + 4.061004801769741e-06, + 3.96601899410598e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3837890625, + "loss_eval": 1.7091575860977173, + "epoch": 61 + }, + { + "hidden_norms": [ + 5305.26953125, + 56114.8203125, + 436835.59375, + 794036.0, + 382300.59375 + ], + "bp_grad_norms_per_sample_med": [ + 2.793117346300278e-06, + 2.1690949836283835e-07, + 1.1630865515144251e-07, + 1.1756346651736749e-07, + 1.1506058683607989e-07 + ], + "bp_grad_norms_F": [ + 0.00016134958423208445, + 8.74852321430808e-06, + 3.9943656702234875e-06, + 4.064597760589095e-06, + 3.969457793573383e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3798828125, + "loss_eval": 1.7154879570007324, + "epoch": 62 + }, + { + "hidden_norms": [ + 5302.8837890625, + 56558.90625, + 441340.78125, + 802203.0625, + 385590.65625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8357644623611122e-06, + 2.1830236107689416e-07, + 1.174595283259805e-07, + 1.1849279246689548e-07, + 1.1699435020773308e-07 + ], + "bp_grad_norms_F": [ + 0.00016155940829776227, + 8.755636372370645e-06, + 4.0028130570135545e-06, + 4.07254128731438e-06, + 3.978685981564922e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7206263542175293, + "epoch": 63 + }, + { + "hidden_norms": [ + 5312.6650390625, + 56690.29296875, + 446444.65625, + 811108.375, + 389126.15625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7693422453012317e-06, + 2.0976162318220304e-07, + 1.1542074673798197e-07, + 1.161635623248003e-07, + 1.1368263841404769e-07 + ], + "bp_grad_norms_F": [ + 0.00016044928634073585, + 8.682382031111047e-06, + 3.968462351622293e-06, + 4.03698095396976e-06, + 3.944758191209985e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3935546875, + "loss_eval": 1.708216667175293, + "epoch": 64 + }, + { + "hidden_norms": [ + 5315.3154296875, + 57172.39453125, + 452320.78125, + 819345.4375, + 392431.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.7417515866545727e-06, + 2.1362264135404985e-07, + 1.1589729354000156e-07, + 1.165334211350455e-07, + 1.1489947127074629e-07 + ], + "bp_grad_norms_F": [ + 0.00016101017536129802, + 8.694333700987045e-06, + 3.9719238884572405e-06, + 4.039454324811231e-06, + 3.947835011786083e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7132065296173096, + "epoch": 65 + }, + { + "hidden_norms": [ + 5323.32421875, + 57353.53515625, + 455451.03125, + 825786.625, + 393529.90625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7779890388046624e-06, + 2.1516311221603246e-07, + 1.1656365472845209e-07, + 1.1796725374324524e-07, + 1.1570150348916286e-07 + ], + "bp_grad_norms_F": [ + 0.00016156666970346123, + 8.733955837669782e-06, + 3.986176579928724e-06, + 4.05381206292077e-06, + 3.9624515011382755e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3994140625, + "loss_eval": 1.715019702911377, + "epoch": 66 + }, + { + "hidden_norms": [ + 5342.94140625, + 57836.46875, + 461806.90625, + 835003.3125, + 397621.53125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7782559754996328e-06, + 2.1394832572241285e-07, + 1.1512971553884199e-07, + 1.1638152841442206e-07, + 1.1455212955979732e-07 + ], + "bp_grad_norms_F": [ + 0.00016023094940464944, + 8.671009709360078e-06, + 3.967585598729784e-06, + 4.035073743580142e-06, + 3.945795015170006e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7059491872787476, + "epoch": 67 + }, + { + "hidden_norms": [ + 5353.94189453125, + 58288.83984375, + 465855.71875, + 840827.8125, + 399727.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7943997338297777e-06, + 2.1217401524609159e-07, + 1.1551408363175142e-07, + 1.1611316352855283e-07, + 1.1411684397444333e-07 + ], + "bp_grad_norms_F": [ + 0.00015934662951622158, + 8.68945244292263e-06, + 3.9676901906204876e-06, + 4.034785888507031e-06, + 3.945680873584934e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.39453125, + "loss_eval": 1.703755259513855, + "epoch": 68 + }, + { + "hidden_norms": [ + 5350.173828125, + 58626.13671875, + 469388.5, + 847294.5, + 401525.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.773014784906991e-06, + 2.1550022211158648e-07, + 1.1623347972999909e-07, + 1.1767290430952926e-07, + 1.1548346634526752e-07 + ], + "bp_grad_norms_F": [ + 0.00016030117694754153, + 8.702772902324796e-06, + 3.981078407377936e-06, + 4.047733909828821e-06, + 3.958340130338911e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7165981531143188, + "epoch": 69 + }, + { + "hidden_norms": [ + 5357.58251953125, + 58903.7734375, + 474182.15625, + 855194.375, + 404878.59375 + ], + "bp_grad_norms_per_sample_med": [ + 2.75364641311171e-06, + 2.1325861609966523e-07, + 1.1510206121556621e-07, + 1.1569208169248668e-07, + 1.1389116849613856e-07 + ], + "bp_grad_norms_F": [ + 0.00015995267312973738, + 8.656907994009089e-06, + 3.959937203035224e-06, + 4.026182978122961e-06, + 3.939108864869922e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3955078125, + "loss_eval": 1.7067822217941284, + "epoch": 70 + }, + { + "hidden_norms": [ + 5361.0703125, + 59035.59375, + 477104.5, + 859434.875, + 405244.875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7923097150051035e-06, + 2.1804021343996283e-07, + 1.1658453757945608e-07, + 1.1743448169454496e-07, + 1.1549956013823248e-07 + ], + "bp_grad_norms_F": [ + 0.0001615561341168359, + 8.730263289180584e-06, + 3.989315700891893e-06, + 4.055812496517319e-06, + 3.967216798628215e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.392578125, + "loss_eval": 1.713646411895752, + "epoch": 71 + }, + { + "hidden_norms": [ + 5369.7529296875, + 59398.04296875, + 480994.0625, + 865578.875, + 407758.6875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7749495075113373e-06, + 2.139253751920478e-07, + 1.1538476485384308e-07, + 1.1588154080754975e-07, + 1.1468705451989081e-07 + ], + "bp_grad_norms_F": [ + 0.00016089060227386653, + 8.69345421961043e-06, + 3.971756996179465e-06, + 4.03762214773451e-06, + 3.950478458136786e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7055977582931519, + "epoch": 72 + }, + { + "hidden_norms": [ + 5376.67578125, + 59606.38671875, + 484413.75, + 871224.125, + 409605.40625 + ], + "bp_grad_norms_per_sample_med": [ + 2.767387286439771e-06, + 2.163243522090852e-07, + 1.1596397797575264e-07, + 1.1663173182796527e-07, + 1.1466888594213742e-07 + ], + "bp_grad_norms_F": [ + 0.0001625583681743592, + 8.740000339457765e-06, + 3.985041985288262e-06, + 4.0507725316274446e-06, + 3.964435109082842e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.717661738395691, + "epoch": 73 + }, + { + "hidden_norms": [ + 5388.92626953125, + 60038.8828125, + 487315.0, + 875672.0625, + 411209.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.811367039612378e-06, + 2.1376251879701158e-07, + 1.150508310843179e-07, + 1.1646184105984503e-07, + 1.1464238269809357e-07 + ], + "bp_grad_norms_F": [ + 0.0001601450494490564, + 8.67134895088384e-06, + 3.971680598624516e-06, + 4.03764806833351e-06, + 3.951014605263481e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.7057151794433594, + "epoch": 74 + }, + { + "hidden_norms": [ + 5394.13818359375, + 60238.7265625, + 490408.5625, + 880235.0625, + 412592.1875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7481701181386597e-06, + 2.1436494535009842e-07, + 1.1508839747875754e-07, + 1.1683697209718957e-07, + 1.148769968040142e-07 + ], + "bp_grad_norms_F": [ + 0.00016096947365440428, + 8.67969538376201e-06, + 3.976166681241011e-06, + 4.041281044919742e-06, + 3.95562074118061e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3974609375, + "loss_eval": 1.711783528327942, + "epoch": 75 + }, + { + "hidden_norms": [ + 5398.15234375, + 60386.79296875, + 492903.0625, + 883714.5, + 413466.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7468756798043614e-06, + 2.142960227047297e-07, + 1.155566522470508e-07, + 1.1617837003541354e-07, + 1.1447124137475839e-07 + ], + "bp_grad_norms_F": [ + 0.00016124852118082345, + 8.686849469086155e-06, + 3.9795741031412035e-06, + 4.044950401294045e-06, + 3.958904471801361e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3955078125, + "loss_eval": 1.7077192068099976, + "epoch": 76 + }, + { + "hidden_norms": [ + 5399.38623046875, + 60673.9140625, + 495671.34375, + 887343.3125, + 414498.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.782978071991238e-06, + 2.1490254198397452e-07, + 1.1598574900517633e-07, + 1.167711403127214e-07, + 1.1515145814655625e-07 + ], + "bp_grad_norms_F": [ + 0.00016076312749646604, + 8.698333658685442e-06, + 3.980411747761536e-06, + 4.046264621138107e-06, + 3.960503818234429e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7069586515426636, + "epoch": 77 + }, + { + "hidden_norms": [ + 5401.99755859375, + 60780.171875, + 498870.40625, + 892185.3125, + 416521.0625 + ], + "bp_grad_norms_per_sample_med": [ + 2.731095946728601e-06, + 2.1313489639851468e-07, + 1.1489591145164013e-07, + 1.1583144754467867e-07, + 1.1378151754115606e-07 + ], + "bp_grad_norms_F": [ + 0.00016048294492065907, + 8.671476280142087e-06, + 3.968764303863281e-06, + 4.033439381601056e-06, + 3.948562607547501e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7046585083007812, + "epoch": 78 + }, + { + "hidden_norms": [ + 5396.2353515625, + 60906.8125, + 499556.1875, + 893389.6875, + 416091.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8349218155199196e-06, + 2.165559180866694e-07, + 1.1532758037446911e-07, + 1.1579567171793315e-07, + 1.1431712465537203e-07 + ], + "bp_grad_norms_F": [ + 0.00016091841098386794, + 8.715818694327027e-06, + 3.983707301813411e-06, + 4.049661583849229e-06, + 3.963775725424057e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.39453125, + "loss_eval": 1.7074486017227173, + "epoch": 79 + }, + { + "hidden_norms": [ + 5399.12646484375, + 60870.21484375, + 502091.5625, + 896890.0625, + 417516.5625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7684259293891955e-06, + 2.148178452898719e-07, + 1.1549953882195041e-07, + 1.1610666916794798e-07, + 1.1443260916621512e-07 + ], + "bp_grad_norms_F": [ + 0.00016129721188917756, + 8.707292181497905e-06, + 3.980841484008124e-06, + 4.045855348522309e-06, + 3.9603528421139345e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.706810474395752, + "epoch": 80 + }, + { + "hidden_norms": [ + 5394.28564453125, + 61078.3671875, + 504476.65625, + 900378.125, + 418670.8125 + ], + "bp_grad_norms_per_sample_med": [ + 2.764037617453141e-06, + 2.1426544094538258e-07, + 1.156425071258127e-07, + 1.1587813730784546e-07, + 1.145299108884501e-07 + ], + "bp_grad_norms_F": [ + 0.00016113216406665742, + 8.712745511729736e-06, + 3.979288521804847e-06, + 4.0438499127049e-06, + 3.9589867810718715e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7053430080413818, + "epoch": 81 + }, + { + "hidden_norms": [ + 5396.7236328125, + 61213.38671875, + 505729.4375, + 902413.125, + 419251.0625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7847922865475994e-06, + 2.1385373827342846e-07, + 1.1590607584821555e-07, + 1.167506624710768e-07, + 1.1480997130774995e-07 + ], + "bp_grad_norms_F": [ + 0.0001612855412531644, + 8.721244739717804e-06, + 3.984357590525178e-06, + 4.048785740451422e-06, + 3.963945800933288e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.708754062652588, + "epoch": 82 + }, + { + "hidden_norms": [ + 5401.74169921875, + 61391.25, + 507562.5, + 904882.1875, + 420061.125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7846790544572286e-06, + 2.140303649866837e-07, + 1.1582537240428792e-07, + 1.1669516197798657e-07, + 1.1512172903849205e-07 + ], + "bp_grad_norms_F": [ + 0.00016096761100925505, + 8.707896085979883e-06, + 3.9824740269978065e-06, + 4.0467593862558715e-06, + 3.962396021961467e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7056825160980225, + "epoch": 83 + }, + { + "hidden_norms": [ + 5407.85205078125, + 61435.578125, + 508690.6875, + 906681.875, + 420568.9375 + ], + "bp_grad_norms_per_sample_med": [ + 2.766970737866359e-06, + 2.144422381888944e-07, + 1.157119697836606e-07, + 1.1730713112001467e-07, + 1.1475398054017205e-07 + ], + "bp_grad_norms_F": [ + 0.00016140809748321772, + 8.720307960174978e-06, + 3.985259354521986e-06, + 4.050050392834237e-06, + 3.965534688177286e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.7068425416946411, + "epoch": 84 + }, + { + "hidden_norms": [ + 5412.962890625, + 61492.68359375, + 510179.25, + 908609.25, + 421252.96875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7769999633164844e-06, + 2.1622201984428102e-07, + 1.1603597727116721e-07, + 1.1669756361243344e-07, + 1.1481289163839392e-07 + ], + "bp_grad_norms_F": [ + 0.00016168547153938562, + 8.720241567061748e-06, + 3.984777777077397e-06, + 4.049422386742663e-06, + 3.9649389691476244e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.7068109512329102, + "epoch": 85 + }, + { + "hidden_norms": [ + 5414.86083984375, + 61660.16015625, + 511083.3125, + 910132.875, + 421357.09375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7625901566352695e-06, + 2.1562310337230883e-07, + 1.1591524184950686e-07, + 1.168542453910959e-07, + 1.1494233120856734e-07 + ], + "bp_grad_norms_F": [ + 0.00016218819655478, + 8.743301805225201e-06, + 3.990852292190539e-06, + 4.055550562043209e-06, + 3.9708543226879556e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.384765625, + "loss_eval": 1.7097513675689697, + "epoch": 86 + }, + { + "hidden_norms": [ + 5414.11181640625, + 61679.67578125, + 512268.53125, + 911841.3125, + 422135.65625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7821745334222214e-06, + 2.1437830355353071e-07, + 1.1559337309563489e-07, + 1.1630524454631086e-07, + 1.143072410059176e-07 + ], + "bp_grad_norms_F": [ + 0.00016158135258592665, + 8.71670272317715e-06, + 3.983236638305243e-06, + 4.047654329042416e-06, + 3.9634205677430145e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7057619094848633, + "epoch": 87 + }, + { + "hidden_norms": [ + 5415.7666015625, + 61667.71875, + 513312.21875, + 913202.4375, + 422493.96875 + ], + "bp_grad_norms_per_sample_med": [ + 2.759925337159075e-06, + 2.1446679454584228e-07, + 1.1589692405777896e-07, + 1.1621393980476569e-07, + 1.1463799864941393e-07 + ], + "bp_grad_norms_F": [ + 0.000161813193699345, + 8.719586730876472e-06, + 3.9842125261202455e-06, + 4.0484619603375904e-06, + 3.964251391153084e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.7056890726089478, + "epoch": 88 + }, + { + "hidden_norms": [ + 5416.599609375, + 61688.796875, + 513735.6875, + 914162.125, + 422800.5625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7784999474533834e-06, + 2.1510695091819798e-07, + 1.1558491763707934e-07, + 1.1671390609535592e-07, + 1.1512633335541977e-07 + ], + "bp_grad_norms_F": [ + 0.0001624061696929857, + 8.740545126784127e-06, + 3.989640845247777e-06, + 4.053715656482382e-06, + 3.969642875745194e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7097570896148682, + "epoch": 89 + }, + { + "hidden_norms": [ + 5416.76220703125, + 61798.24609375, + 514785.84375, + 915175.75, + 423199.09375 + ], + "bp_grad_norms_per_sample_med": [ + 2.777203462756006e-06, + 2.1360766311318002e-07, + 1.1558247337006833e-07, + 1.1610516992277553e-07, + 1.1466264027149009e-07 + ], + "bp_grad_norms_F": [ + 0.0001616168738109991, + 8.719413926883135e-06, + 3.9822580220061354e-06, + 4.046389676659601e-06, + 3.9626197576581035e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7034153938293457, + "epoch": 90 + }, + { + "hidden_norms": [ + 5416.92138671875, + 61928.12109375, + 515030.59375, + 915798.25, + 423164.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.810946853060159e-06, + 2.1559939966664388e-07, + 1.1640541686119832e-07, + 1.1669366983824148e-07, + 1.1499378160806373e-07 + ], + "bp_grad_norms_F": [ + 0.00016196168144233525, + 8.733930371818133e-06, + 3.986880528827896e-06, + 4.051325959153473e-06, + 3.9673618630331475e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3857421875, + "loss_eval": 1.7049140930175781, + "epoch": 91 + }, + { + "hidden_norms": [ + 5416.93212890625, + 61895.34765625, + 515480.125, + 916358.0, + 423425.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.8099400424252963e-06, + 2.1475410960647423e-07, + 1.1618925555012538e-07, + 1.1643356145896178e-07, + 1.1480153716547647e-07 + ], + "bp_grad_norms_F": [ + 0.00016178778605535626, + 8.72767577675404e-06, + 3.985679995821556e-06, + 4.050090865348466e-06, + 3.966247277276125e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7039738893508911, + "epoch": 92 + }, + { + "hidden_norms": [ + 5417.095703125, + 61845.4921875, + 515914.6875, + 916804.125, + 423511.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.805000349326292e-06, + 2.15853873442029e-07, + 1.1573133917863743e-07, + 1.1623403395333298e-07, + 1.146071184621178e-07 + ], + "bp_grad_norms_F": [ + 0.00016200476966332644, + 8.730704394110944e-06, + 3.986714546044823e-06, + 4.050946245115483e-06, + 3.967099928559037e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7049956321716309, + "epoch": 93 + }, + { + "hidden_norms": [ + 5417.3349609375, + 61879.95703125, + 516017.53125, + 916914.5625, + 423434.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.7787407361756777e-06, + 2.167610944070475e-07, + 1.1583406234194626e-07, + 1.1677065714366108e-07, + 1.1465835569879346e-07 + ], + "bp_grad_norms_F": [ + 0.0001623090065550059, + 8.742457794141956e-06, + 3.991137873526895e-06, + 4.055534191138577e-06, + 3.971431851823581e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7075328826904297, + "epoch": 94 + }, + { + "hidden_norms": [ + 5417.50390625, + 61889.82421875, + 516312.4375, + 917429.0, + 423711.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7708026664186036e-06, + 2.1539200645293022e-07, + 1.1580210923511913e-07, + 1.16502334890356e-07, + 1.1404062405517834e-07 + ], + "bp_grad_norms_F": [ + 0.00016205437714233994, + 8.73019962455146e-06, + 3.986566753155785e-06, + 4.050732059113216e-06, + 3.966889835282927e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7053842544555664, + "epoch": 95 + }, + { + "hidden_norms": [ + 5417.48828125, + 61898.47265625, + 516375.0, + 917465.6875, + 423663.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7983871859760256e-06, + 2.1556444096404448e-07, + 1.160928064791733e-07, + 1.1642453046079027e-07, + 1.1440795333328424e-07 + ], + "bp_grad_norms_F": [ + 0.00016201405378524214, + 8.732341484574135e-06, + 3.987765467172721e-06, + 4.052003077958943e-06, + 3.9681326597929e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7055914402008057, + "epoch": 96 + }, + { + "hidden_norms": [ + 5417.54931640625, + 61916.09375, + 516515.78125, + 917698.75, + 423749.71875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7831354145746445e-06, + 2.1518340531656577e-07, + 1.1582258707676374e-07, + 1.1653285980628425e-07, + 1.1448688042037247e-07 + ], + "bp_grad_norms_F": [ + 0.00016204272105824202, + 8.731909474590793e-06, + 3.987126092397375e-06, + 4.051275027450174e-06, + 3.967459178966237e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7053790092468262, + "epoch": 97 + }, + { + "hidden_norms": [ + 5417.61865234375, + 61911.2734375, + 516535.4375, + 917715.875, + 423733.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.784623802654096e-06, + 2.1553401552409923e-07, + 1.157067259782707e-07, + 1.1653578724235558e-07, + 1.1444880243516309e-07 + ], + "bp_grad_norms_F": [ + 0.0001621013943804428, + 8.73380577104399e-06, + 3.987797754234634e-06, + 4.05199125452782e-06, + 3.968138116761111e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7057509422302246, + "epoch": 98 + }, + { + "hidden_norms": [ + 5417.64111328125, + 61917.359375, + 516576.09375, + 917756.125, + 423741.78125 + ], + "bp_grad_norms_per_sample_med": [ + 2.785410288197454e-06, + 2.1554964746428595e-07, + 1.1583354364574916e-07, + 1.1650187303757775e-07, + 1.145084596032575e-07 + ], + "bp_grad_norms_F": [ + 0.00016210223839152604, + 8.733919457881711e-06, + 3.987840045738267e-06, + 4.052033091284102e-06, + 3.968175860791234e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7057793140411377, + "epoch": 99 + }, + { + "hidden_norms": [ + 5417.64697265625, + 61917.6640625, + 516580.8125, + 917765.6875, + 423747.84375 + ], + "bp_grad_norms_per_sample_med": [ + 2.786067398119485e-06, + 2.1562047436418652e-07, + 1.1590604742650612e-07, + 1.1655521348075126e-07, + 1.1442210023915322e-07 + ], + "bp_grad_norms_F": [ + 0.00016209435125347227, + 8.73375483934069e-06, + 3.987807303929003e-06, + 4.0519980757380836e-06, + 3.968141299992567e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.705754280090332, + "epoch": 100 + } + ], + "seed": 42, + "depth": 4, + "d_hidden": 256 +}
\ No newline at end of file |
