{ "operator_check": { "inside_max_error": 0.0, "max_output_residual": 0.7299999999998885, "tau": 0.73, "minimum_direction_cosine": 0.9999999999999996 }, "bias_scaling": { "taus": [ 2.0, 4.0, 8.0, 16.0, 32.0, 64.0 ], "tail_bias": [ 2.1078901113465762, 1.485287712794427, 1.047741566357388, 0.7380726599802617, 0.5182458643749303, 0.36457934163643035 ], "loglog_slope": -0.5062864083363753, "expected_slope": -0.5 }, "runs": { "heavy_tail/sgd": { "final_loss_mean": 1.5904582091610717, "final_loss_std": 4.791403952086443, "p99_update_mean": 1.4383751036429422, "divergences": 0, "final_losses": [ 0.0909823080030608, 0.05312237551349112, 0.09223474792462609, 0.07947341840642966, 0.15228072810776913, 0.1532125708388954, 0.10080784091637439, 0.10629070802367911, 0.09703321835086269, 0.5679931664106993, 17.47586128718058, 0.11620614025639271 ] }, "heavy_tail/global_clip": { "final_loss_mean": 0.05626926714363623, "final_loss_std": 0.01365064201168186, "p99_update_mean": 0.14999999999999325, "divergences": 0, "final_losses": [ 0.05423517478316375, 0.02972028068154693, 0.041147602941888496, 0.05373200563878424, 0.06955790364444688, 0.03964291055194818, 0.07311448424448197, 0.06882789460159842, 0.0468672212466314, 0.06692207337140618, 0.06285721333444437, 0.06860644068329401 ] }, "heavy_tail/centered": { "final_loss_mean": 0.0012210588447355575, "final_loss_std": 0.0004985480793692171, "p99_update_mean": 1.3289220351289603, "divergences": 0, "final_losses": [ 0.0013646504546732253, 0.0006280569779644884, 0.001277164577435138, 0.0007120214502243642, 0.0009400869903349203, 0.001395007860398821, 0.0010673624658662617, 0.0010298607072653143, 0.0011537571390326617, 0.0015756784939763432, 0.0008945247718432248, 0.0026145342478119268 ] }, "gaussian/sgd": { "final_loss_mean": 0.0016830468711389003, "final_loss_std": 0.0007888298476053901, "p99_update_mean": 1.3292792381825755, "divergences": 0, "final_losses": [ 0.0007722236026109946, 0.001467706388583531, 0.0013038306410087111, 0.0008065105614733637, 0.0024601787884927727, 0.002035386811184109, 0.0016890198835242236, 0.0008171132569252109, 0.0014557284199099977, 0.0014400223096305545, 0.0035888134542687105, 0.0023600283360546244 ] }, "gaussian/global_clip": { "final_loss_mean": 0.0016830755245458389, "final_loss_std": 0.0007888650369945481, "p99_update_mean": 0.14999999999999317, "divergences": 0, "final_losses": [ 0.000772225305850193, 0.001467882370216722, 0.001303797848379775, 0.0008065040691951997, 0.0024603130507194304, 0.002035403423231237, 0.001688927591324899, 0.0008171641409295063, 0.0014557480466857592, 0.0014400554891856017, 0.0035890348798259565, 0.0023598500790057864 ] }, "gaussian/centered": { "final_loss_mean": 0.0007145656274612843, "final_loss_std": 0.00021586353232532467, "p99_update_mean": 1.3292060953351132, "divergences": 0, "final_losses": [ 0.0006701332955378414, 0.00046670808528075574, 0.0005240023180628656, 0.0006419451792648101, 0.0012294410212866519, 0.00036388347280563916, 0.0008182256003758723, 0.0006522662877620825, 0.0007491919267796089, 0.0007148583521299482, 0.0008539547149137506, 0.0008901772753355874 ] } } }